Navigating in a three-dimensional world...three-dimensional environments into those that are locally planar surfaces, which allow movement only in a direction tangential to the surface,

Navigating in a three-dimensional world

Kathryn J JefferyDepartment of Cognitive Perceptual and Brain Sciences Division ofPsychology amp Language Sciences University College London London WC1H

0AP United Kingdomkjefferyuclacukwwwuclacukjefferylab

Aleksandar Jovalekic1Institute of Neuroinformatics University of Zurich CH-8057 Zurich Switzerlandajovalekiciniphysethzch

Madeleine Verriotis1Department of Neuroscience Physiology and Pharmacology University

College London London WC1E 6BT United Kingdommadeleineverriotisuclacuk

Robin HaymanInstitute of Cognitive Neuroscience Alexandra House London WC1N 3AR

United Kingdomrhaymanuclacuk

Abstract The study of spatial cognition has provided considerable insight into how animals (including humans) navigate on the horizontalplane However the real world is three-dimensional having a complex topography including both horizontal and vertical features whichpresents additional challenges for representation and navigation The present article reviews the emerging behavioral and neurobiologicalliterature on spatial cognition in non-horizontal environments We suggest that three-dimensional spaces are represented in a quasi-planar fashion with space in the plane of locomotion being computed separately and represented differently from space in theorthogonal axis ndash a representational structure we have termed ldquobicodedrdquo We argue that the mammalian spatial representation insurface-travelling animals comprises a mosaic of these locally planar fragments rather than a fully integrated volumetric map Moregenerally this may be true even for species that can move freely in all three dimensions such as birds and fish We outline theevidence supporting this view together with the adaptive advantages of such a scheme

Keywords ethology grid cells head direction cells hippocampus navigation neural encoding place cells spatial cognition three-dimensional

1 Introduction

Behavioural and neurobiological studies of spatial cognitionhave provided considerable insight into how animals(including humans) navigate through the world establish-ing that a variety of strategies subserved by differentneural systems can be used in order to plan and executetrajectories through large-scale navigable space Thiswork which has mostly been done in simplified labora-tory-based environments has defined a number of sub-pro-cesses such as landmark recognition headingdetermination odometry (distance-measuring) andcontext recognition all of which interact in the constructionand use of spatial representations The neural circuitryunderlying these processes has in many cases been ident-ified and it now appears that the basics of navigation arereasonably well understood

The real world however is neither simple nor two-dimensional and the addition of vertical space to the twohorizontal dimensions adds a number of new problemsfor a navigator to solve For one thing moving againstgravity imposes additional energy costs Also moving in a

volumetric space is computationally complicated becauseof the increased size of the representation needed andbecause rotations in three dimensions interact It remainsan open question how the brain has solved these problemsand whether the same principles of spatial encodingoperate in all three dimensions or whether the verticaldimension is treated differentlyStudies of how animals and humans navigate in environ-

ments with a vertical component are slowly proliferatingand it is timely to review the gathering evidence andoffer a theoretical interpretation of the findings to dateIn the first part of the article we review the currentliterature on animal and human orientation and locomotionin three-dimensional spaces highlighting key factors thatcontribute to and influence navigation in such spacesThe second part summarizes neurobiological studies ofthe encoding of three-dimensional space The third andfinal part integrates the experimental evidence to putforward a hypothesis concerning three-dimensionalspatial encoding ndash the bicoded model ndash in which wepropose that in encoding 3D spaces the mammalianbrain constructs mosaics of connected surface-referenced

BEHAVIORAL AND BRAIN SCIENCES (2013) 36 523ndash587doi101017S0140525X12002476

copy Cambridge University Press 2013 0140-525X13 $4000 523

maps The advantages and limitations of such a quasi-planarencoding scheme are explored

2 Theoretical considerations

Before reviewing the behavioural and neurobiological aspectsof three-dimensional navigation it is useful to establish atheoretical framework for the discussion which will be devel-oped in more detail in the last part of this article The study ofnavigation needs to be distinguished from the study of spatialperception per se in which a subject may or may not bemoving through the space Additionally it is also importantto distinguish between space that is encoded egocentrically(ie relative to the subject) versus space that is encoded allo-centrically (independently of the subject ndash ie relative to theworld) The focus of the present article is on the allocentricencoding of the space that is being moved through whichprobably relies on different (albeit interconnected) neuralsystems from the encoding of egocentric spaceAn animal that is trying to navigate in 3D space needs to

know three things ndash how it is positioned how it is orientedand in what direction it is moving Further all these thingsrequire a reference frame a world-anchored space-defin-ing framework with respect to which position orientationand movement can be specified There also needs to besome kind of metric coordinate system ndash that is a signalthat specifies distances and directions with respect to thisreference frame The core question for animal navigation

therefore concerns what the reference frame might beand how the coordinate system encodes distances anddirections within the frameThe majority of studies of navigation at least within neu-

robiology have up until now taken place in laboratory set-tings using restricted environments such as mazes whichare characterized both by being planar (ie two-dimen-sional) and by being horizontal The real world differs inboth of these respects First it is often not planar ndash air andwater for example allow movement in any direction andso are volumetric In this article therefore we will dividethree-dimensional environments into those that are locallyplanar surfaces which allow movement only in a directiontangential to the surface and those that are volumetricspaces (air water space and virtual reality) in which move-ment (or virtual movement) in any direction is uncon-strained Second even if real environments might be(locally) planar the planes may not necessarily be horizontalThese factors are important in thinking about how subjectsencode and navigate through complex spacesWhat does the vertical dimension add to the problem of

navigation At first glance not much The vertical dimen-sion provides a set of directions one can move in like anyother dimension and it could be that the principles thathave been elucidated for navigation in two dimensionsextend straightforwardly to three In fact however the ver-tical dimension makes the problem of navigation consider-ably more complex for a number of reasons First thespace to be represented is much larger since a volume islarger than a plane by a power of 32 Second the direc-tional information to be encoded is more complexbecause there are three planes of orientation instead ofjust one Furthermore rotations in orthogonal planesinteract such that sequences of rotations have different out-comes depending on the order in which they are executed(ie they are non-commutative) which adds to processingcomplexity Third the vertical dimension is characterizedby gravity and by consequences of gravity such as hydro-static or atmospheric pressure which add both informationand effort to the computations involved And finally thereare navigationally relevant cues available for the horizontalplane that are absent in the vertical for example the pos-ition of the sun and stars or the geomagnetic fieldThe internal spatial representation that the brain con-

structs for use in large-scale navigation is often referredto as a ldquocognitive maprdquo (OrsquoKeefe amp Nadel 1978) The cog-nitive map is a supramodal representation that arises fromthe collective activity of neurons in a network of brain areasthat process incoming sensory information and abstracthigher-order spatial characteristics such as distance anddirection In thinking about whether cognitive maps arethree-dimensional it is useful to consider the various 3Dmapping schemes that are possibleFigure 1 shows three different hypothetical ways of

mapping spaces which differ in the degree to which theyincorporate a representation of the vertical dimension InFigure 1a the metric fabric of the map ndash the grid ndashfollows the topography but does not itself contain any topo-graphical information the map is locally flat Such a surfaceis called in mathematics amanifold which is a space that islocally Euclidean ndash that is for any point in the space itsimmediate neighbours will be related to it by the usualrules of geometry even if more-distant points are relatedin a non-Euclidean way For example the surface of a

KATHRYN (KATE) JEFFERY is Professor of BehaviouralNeuroscience at University College London (UCL)United Kingdom After graduating in 1985 in medicinefrom the University of Otago New Zealand sheembarked on a research career to study neural rep-resentation training with Cliff Abraham (University ofOtago) Richard Morris (University of Edinburgh) andJohn OrsquoKeefe (UCL) In 1999 she joined the PsychologyDepartment (now Division of Psychology amp LanguageSciences) at UCL where she continues research intothe neural basis of spatial cognition

ALEKSANDAR JOVALEKIC currently is a PostdoctoralFellow at the Institute of Neuroinformatics in ZurichHe obtained his PhD at University College London in2012 where his doctoral work focused on how rodentsnavigate in three-dimensional environments

MADELEINE VERRIOTIS is a Postdoctoral ResearchAssociate at University College London (UCL) Shereceived her BSc in Psychology and her PhD in Behav-ioural Neuroscience from UCL during which sheexplored the representation of three-dimensionalspace in the rat and human brains Her postdoctoralresearch focuses on the mechanisms underlying thedevelopment of pain processing in the human infantbrain

ROBIN HAYMAN is a Postdoctoral Research Associate inNeuroscience at University College London (UCL) Hegained his BSc in Psychology with Literature at the Uni-versity of Staffordshire MSc in Neuroscience at UCLand PhD in Psychology also at UCL

Jeffery et al Navigating in a three-dimensional world

524 BEHAVIORAL AND BRAIN SCIENCES (2013) 365

sphere is a manifold because a very small triangle on itssurface will be Euclidean (its interior angles add up to180 degrees) but a large triangle will be Riemannian (thesum of its interior angles exceeds 180 degrees) Given asmall enough navigator and large enough surface undula-tions the ground is therefore a manifold For the purposesof the discussion that follows we will call a map that followsthe topography and is locally Euclidean a surface map Asurface map in its simplest form does not for a givenlocation encode information about whether the locationis on a hill or in a valley2 Although such a map accuratelyrepresents the surface distance a navigator would need totravel over undulating terrain because it does not containinformation about height (or elevation) above a referenceplane it would fail to allow calculation of the extra effortinvolved in moving over the undulations and nor wouldthe calculation of shortcuts be accurate

A variant of the surface map is the extracted flat mapshown in Figure 1b Here distances and directions areencoded relative to the horizontal plane rather than tothe terrain surface For a flat environment a surface mapequates to an extracted flat map For hilly terrain inorder to determine its position in x-y coordinates the navi-gator needs to process the slope to determine trigonome-trically the equivalent horizontal distance To generatesuch a map it is therefore necessary to have some proces-sing of information from the vertical dimension (ieslope with respect to position and direction) and so thisis a more complex process than for a purely surface mapThere is evidence discussed later that some insects can infact extract horizontal distance from hilly terrain althoughwhether this can be used to generate a map remains moot

In Figure 1c the map now does include a specific rep-resentation of elevation but the encoding method is differ-ent from that of the horizontal distances Hence we havetermed this a bicoded map In a bicoded map vertical

distance is represented by some non-spatial variable ndash inthis case colour Provided with information about x-y coor-dinates together with the non-spatial variable navigatorscould work out where they were in three-dimensionalspace but would not have metric information (ie quanti-tative information about distance) about the vertical com-ponent of this position unless they also had a key (someway of mapping the non-spatial variable to distances)Lacking a key navigators could still know whether theywere on a hill or in a valley but could not calculate efficientshortcuts because they could not make the trigonometriccalculations that included the vertical component Such amap could be used for long-range navigation over undulat-ing terrain but determining the most energy-efficientroute would be complicated because this would requirechaining together segments having different gradients(and therefore energetic costs) as opposed to simply com-puting a shortest-distance route Although the represen-tational scheme described here is an artificial one used byhuman geographers later on we will make the point thatan analogous scheme ndash use of a non-metric variable toencode one of the dimensions in a three-dimensionalmap ndashmay operate in the neural systems that supportspatial representationFinally the volumetric map shown in Figure 1d could be

used for navigation in a volumetric space Here all threedimensions use the same metric encoding scheme andthe shortest distance between any two points can be calcu-lated straightforwardly using trigonometry (or some neuralequivalent) with ndash on Earth ndash elevation being simply thedistance in the dimension perpendicular to gravityHowever calculating energy costs over an undulatingsurface would still be complicated for the same reasonsas above ndash that is energy expenditure would have to becontinuously integrated over the course of the plannedjourney in order to allow comparisons of alternate routes

Figure 1 Different kinds of three-dimensional encoding (a) A surface map with no information about elevation (b) An extracted flatmap in which the horizontal coordinates have been generated by trigonometric inference from hilly terrain in (a) (c) A bicoded mapwhich is metric in the horizontal dimension and uses a non-metric scale (ie shading) in the vertical (upper panel = birdrsquos eye view of theterrain lower panel = cross-section at the level of the dotted line) (d) A volumetric map which is metric in all three dimensions like thismap of dark matter in the universe

BEHAVIORAL AND BRAIN SCIENCES (2013) 365 525

The questions concerning the brainrsquos map of space thenare What encoding scheme does it use for navigation eitherover undulating surfaces or in volumetric spaces Is thevertical dimension encoded in the same way as the twohorizontal ones as in the volumetric map Is it ignoredentirely as in a surface map or is it encoded but in a differ-ent way from the horizontal dimensions as in the bicodedmap These questions motivate the following analysis ofexperimental data and will be returned to in the finalsection where we will suggest that a bicoded map is themost likely encoding scheme at least in mammals andperhaps even including those that can swim or fly

3 Behavioural studies in three dimensions

A key question to be answered concerning the represen-tation of three-dimensional space is how informationfrom the vertical dimension is incorporated into thespatial representation We will therefore now review exper-imental studies that have explored the use of informationfrom the vertical dimension in humans and other animalsbeginning with an examination of how cues from the verti-cal dimension may be processed before moving on to howthese cues may be used in self-localization and navigationThe discussion is organized as follows based on a logicalprogression through complexity of the representation ofthe third (usually vertical) dimension

Processing of verticality cuesNavigation on a sloping surfaceNavigation in multilayer environmentsNavigation in a volumetric space

31 Processing of verticality cues

Processing of cues arising from the vertical dimension is afaculty possessed by members of most phyla and it takesmany forms One of the simplest is detection of the verticalaxis itself using a variety of mechanisms that mostly rely ongravity or on cues related to gravity such as hydrostaticpressure or light gradients This information provides apotential orienting cue which is useful not only in staticsituations but also for navigation in three-dimensionalspace Additionally within the vertical dimension itselfgravity differentiates down from up and thus polarizesthe vertical axis Consequently animals are always orientedrelative to this axis even though they can become disor-iented relative to the horizontal axes The importance ofthe gravity signal as an orienting cue is evident inreduced or zero-gravity environments which appear tobe highly disorienting for humans Astronauts report diffi-culty in orienting themselves in three-dimensional spacewhen floating in an unfamiliar orientation (Oman 2007)The vertical axis once determined can then be used as a

reference frame for the encoding of metric information inthe domains of both direction and distance The technicalterms for dynamic rotations relative to the vertical axisare pitch roll and yaw corresponding to rotations in thesagittal coronal and horizontal planes respectively(Fig 2A) Note that there is potential for confusion asthese rotational terms tend to be used in an egocentric(body-centered) frame of reference in which the planes

of rotation move with the subject The static terms usedto describe the resulting orientation (or attitude) of thesubject are elevation bank and azimuth (Fig 2B) whichrespectively describe (1) the angle that the front-back(antero-posterior) axis of the subject makes with earth-hori-zontal (elevation) (2) the angle that the transverse axis ofthe subject makes with earth-horizontal (bank) and (3)the angle that the antero-posterior axis of the subjectmakes with some reference direction usually geographicalNorth (azimuth) Tilt is the angle between the dorso-ventral cranial axis and the gravitational verticalThe other metric category in addition to direction is dis-

tance which in the vertical dimension is called height ndash or(more precisely) elevation when it refers to distance froma reference plane (eg ground or water level) in anupward direction and depth when it refers to distance ina downward direction Determination of elevationdepthis a special case of the general problem of odometry (dis-tance-measuring)There is scant information about the extent to which

animals encode information about elevation A potentialsource of such information could be studies of aquaticanimals which are exposed to particularly salient cuesfrom the vertical dimension due to the hydrostatic pressuregradient Many fish are able to detect this gradient via theirswim bladders which are air-filled sacs that can be inflatedor deflated to allow the fish to maintain neutral buoyancyHolbrook and Burt de Perera (2011a) have argued thatchanges in vertical position would also for a constant gasvolume change the pressure in the bladder and thusprovide a potential signal for relative change in heightWhether fish can actually use this information for verticalself-localization remains to be determined howeverThe energy costs of moving vertically in water also add

navigationally relevant information These costs can varyacross species (Webber et al 2000) hence one mightexpect that the vertical dimension would be more salientin some aquatic animals than in others Thermal gradientswhich are larger in the vertical dimension could alsoprovide extra information concerning depth as well asimposing differential energetic costs in maintainingthermal homeostasis at different depths (Brill et al 2012Carey 1992)Interestingly although fish can process information

about depth Holbrook and Burt de Perera (2009) foundthat they appear to separate this from the processing ofhorizontal information Whereas banded tetras learnedthe vertical and horizontal components of a trajectoryequally quickly the fish tended to use the two componentsindependently suggesting a separation either during learn-ing storage or recall or at the time of use of the infor-mation When the two dimensions were in conflict thefish preferred the vertical dimension (Fig 3) possiblydue to the hydrostatic pressure gradientSeveral behavioural studies support the use of elevation

information in judging the vertical position of a goal Forinstance rufous hummingbirds can distinguish flowers onthe basis of their relative heights above ground (Hendersonet al 2001 2006) and stingless bees can communicate theelevation as well as the distance and direction to a foodsource (Nieh amp Roubik 1998 Nieh et al 2003) Interest-ingly honeybees do not appear to communicate elevation(Dacke amp Srinivasan 2007 Esch amp Burns 1995) Rats arealso able to use elevation information in goal localization

and ndash like the fish in the Holbrook and Burt de Perera(2009) study ndash appear to separate horizontal and verticalcomponents when they do so For example Grobeacutety ampSchenk (1992a) found that in a lattice maze (Fig 4A)rats learned to locate the elevation of a food item morequickly than they learned its horizontal coordinatessuggesting that elevation may provide an additional moresalient spatial cue perhaps because of its added energycost or because of arousalanxiety

Jovalekic et al (2011) also found a verticalhorizontalprocessing separation in rats but in the opposite directionfrom Grobeacutety and Schenk suggesting a task dependenceon how processing is undertaken Rats were trained on avertical climbing wall (studded with protruding pegs toprovide footholds Fig 4B) to shuttle back and forthbetween a start and goal location When a barrier wasinserted into the maze to force a detour the rats overwhel-mingly preferred in both directions the route in which thehorizontal leg occurred first and the vertical leg secondLikewise on a three-dimensional lattice similar to theone used by Grobeacutety and Schenk the rats preferred theshallow-steep route to the steep-shallow route around abarrier However the task in Jovalekic et alrsquos study differedfrom that faced by Grobeacutety and Schenkrsquos rats in that in thelatter both the starting location and the goal were welllearned by the time the obstacle was introduced It maybe that elevation potentiates initial learning but that in awell-learned situation the animals prefer to solve the hori-zontal component first perhaps due to an inclination todelay the more costly part of a journey

Humans can estimate their own elevation but with onlya modest degree of accuracy Garling et al (1990) foundthat subjects were able to judge from memory whichone of pairs of familiar landmarks in their hometown has

a higher elevation indicating that information aboutelevation is both observed and stored in the brain It isnot clear how this information was acquired althoughthere were suggestions that the process may be basedpartly on encoding heuristics such as encoding the steepestand therefore most salient portions of routes first asopposed to encoding precise metric information Notwith-standing the salience of the vertical dimension processingof the vertical dimension in humans appears poor (Mon-tello amp Pick 1993 Pick amp Rieser 1982 Tlauka et al 2007)As with the fish and rodent studies above a dissociation

between vertical and horizontal processing was seen in ahuman study in which landmarks arranged in a 3-by-3grid within a building were grouped by most participantsinto (vertical) column representations and by fewer partici-pants into (horizontal) row representations (Buumlchner et al2007) Most participantsrsquo navigation strategies were consist-ent with their mental representations as a result whenlocating the given target landmark a greater proportionof participants chose a horizontal-vertical route Theauthors suggested that these results might have beenreversed in a building that was taller than it was wideunlike the building in their study However given theresults from the rats of Jovalekic et al (2011) in whichthe climbing wall dimensions were equal it is also possiblethat this behaviour reflects an innate preference forhorizontal-first routesIn conclusion then studies of both human and nonhu-

man animals find that there is clear evidence of the useof elevation information in spatial computations and alsofor differential processing of information from verticalversus horizontal dimensions The processing differencesbetween the two dimensions may result from differentialexperience differential costs associated with the respective

Figure 2 (a) Terminology describing rotations in the three cardinal planes (b) Terminology describing static orientation (or ldquoattituderdquo)in each of the three planes

dimensions or differences in how these dimensions areencoded neurallyNext we turn to the issue of how the change in height

with respect to change in distance ndash that is the slope ofthe terrain ndashmay be used in spatial cognition and naviga-tion This is a more complex computation than simpleelevation but its value lies in the potential for addingenergy and effort to the utility calculations of a journey

32 Navigation on a sloping surface

Surfaces extending into the vertical dimension locally formeither slopes or vertical surfaces and globally form undulat-ing or crevassed terrain Slopes have the potential to provideseveral kinds of information to a navigating animal (Restatet al 2004) acting for instance as a landmark (eg a hillvisible from a distance) or as a local organizing feature ofthe environment providing directional informationAdditionally as described earlier the added vertical com-ponent afforded by slopes is both salient and costly asupward locomotion requires greater effort than horizontallocomotion and an extremely steep slope could be hazar-dous to traverse As a result it is reasonable to assume thatslopes might influence navigation in several waysA number of invertebrates including desert ants and

fiddler crabs appear able to detect slopes sometimeswith a high degree of accuracy (at least 12 degrees forthe desert ant Wintergerst amp Ronacher 2012) It is stillunknown how these animals perform their slope detectionwhich in the ant does not appear to rely on energy con-sumption (Schaumlfer amp Wehner 1993) sky polarization cues(Hess et al 2009) or proprioceptive cues from jointangles (Seidl amp Wehner 2008) The degree to whichslope is used in ant navigation has been under considerableinvestigation in recent years making use of the ability ofthese animals to ldquohomerdquo (find their way back to the nest)using a process known as path integration (see Etienne ampJeffery 2004 Walls amp Layne 2009 Wehner 2003) inwhich position is constantly updated during an excursionaway from a starting point such as a nestForaging ants will home after finding food and the direc-

tion and distance of their return journey provide a con-venient readout of the homing vector operative at thestart of this journey Wohlgemuth et al (2001) investigatedhoming in desert ants and found that they could compen-sate for undulating terrain traversed on an outboundjourney by homing across flat terrain the horizontallyequivalent distance back to the nest indicating that theyhad processed the undulations and could extract the corre-sponding ground distance Likewise fiddler crabs searchedfor their home burrows at the appropriate ground distanceas opposed to the actual travelled distance (Walls amp Layne2009) Whether such encoding of slope is relative to earth-horizontal (ie the plane defined by gravity) or whether itsimply relates the two surfaces (the undulating outboundone and the flat return one) remains to be determinedHowever despite the ability of Cataglyphis to extractground distance Grah et al (2007) found that the antsdid not appear to encode specific aspects of the undulatingtrajectory in their representations of the journey as antstrained on outbound sloping surfaces will select homingroutes that have slopes even when these are inappropriate(eg if the two outbound slopes had cancelled out leaving

Figure 3 The experiment by Holbrook and Burt de Perera(2009) (a) Line diagram of the choice maze (taken from theoriginal article with permission) in which fish were releasedinto a tunnel and had to choose one of the exit arms to obtain areward (b) The conflict experiment of that study in which thefish were trained with the arms at an angle so that one armpointed up and to the left and the other arm pointed down andto the right During testing the arms were rotated aboutthe axis of the entrance tunnel so that one arm had the samevertical orientation but a different left-right position and onehad the same horizontal (left-right) position but a differentvertical level (c) Analysis of the choices showed that the fishgreatly preferred the arm having the same vertical levelRedrawn from the original

the ant at the same elevation as the nest with no need toascend or descend) Thus ants do not appear to incorpor-ate elevation into their route memory their internal mapsseem essentially flat

The ability of humans to estimate the slope of surfaces hasbeen the subject of considerable investigation Gibson andCornsweet (1952) noted that geographical slant which isthe angle a sloping surface makes with the horizontal planeis to be distinguished from optical slant which is the anglethe surface makes with the gaze direction This is importantbecause a surface appears to an observer to possess a rela-tively constant slope regardless of viewing distance eventhough the angle the slope makes with the eye changes(Fig 5A) The observer must somehow be able to compen-sate for the varying gaze angle in order tomaintain perceptualslope constancy presumably by using information about theangle of the head and eyes with respect to the bodyHowever despite perceptual slope constancy perceptualslope accuracy is highly impaired ndash people tend to greatlyoverestimate the steepness of hills (Kammann 1967) typi-cally judging a 5-degree hill to be 20 degrees and a 30-degree hill to be almost 60 degrees (Proffitt et al 1995)

This mis-estimation of slope has been linked to the well-established tendency of people to underestimate depth inthe direction of gaze (Loomis et al 1992) If subjects per-ceive distances along the line of sight to be foreshortenedthen they would perceive the horizontal component of theslope to be less than it really is for a given vertical distanceand thus the gradient to be steeper than it is However thisexplanation fails to account for why subjects looking down-hill also perceive the hill as too steep (Proffitt et al 1995)Furthermore if observers step back from the edge of adownward slope there is a failure of slope constancy aswell as accuracy they see the slope as steeper than theydo if they are positioned right at the edge (Li amp Durgin2009) Durgin and Li have suggested that all these percep-tual distortions can collectively be explained by so-calledscale expansion (Durgin amp Li 2011 Li amp Durgin 2010) ndashthe tendency to overestimate angles (both gaze declinationand optical slant) in central vision by a factor of about 15(Fig 5B) While this expansion may be beneficial for pro-cessing efficiency it raises questions about how such distor-tions may affect navigational accuracy

Other nonvisual distortions of slope perception mayoccur as well Proffitt and colleagues have suggested thatthere is an effect on slope perception of action-relatedfactors such as whether people are wearing a heavy back-pack or are otherwise fatigued or are elderly (Bhalla ampProffitt 1999) whether the subject is at the top or bottomof the hill (Stefanucci et al 2005) and their level ofanxiety (Stefanucci et al 2008) or mood (Riener et al2011) or even level of social support (Schnall et al 2012)However the interpretation of many of these observationshas been challenged (Durgin et al 2012)Creem and Proffitt (1998 2001) have argued that visuo-

motor perceptions of slope are dissociable from explicitawareness By this account slope perception comprisestwo components an accurate and stable visuomotorrepresentation and an inaccurate and malleable consciousrepresentation This may be why subjects who appear tohave misperceived slopes according to their verbalreports nevertheless tailor their actions to them appropri-ately Durgin et al (2011) have challenged this viewsuggesting that experimental probes of visuomotor per-ception are themselves inaccurate The question ofwhether there are two parallel systems for slope perceptiontherefore remains openHow does slope perception affect navigation One

mechanism appears to be by potentiating spatial learningFor example tilting some of the arms in radial armmazes improves performance in working memory tasks inrats (Brown amp Lesniak-Karpiak 1993 Grobeacutety amp Schenk1992b Figure 6a) Similarly experiments in rodents invol-ving conical hills emerging from a flat arena that rats navi-gated in darkness found that the presence of the slopesfacilitated learning and accuracy during navigation(Moghaddam et al 1996 Fig 6B) In these experimentssteeper slopes enhanced navigation to a greater extentthan did shallower slopes Pigeons walking in a tiltedarena were able to use the 20-degree slope to locate agoal corner furthermore they preferred to use slope infor-mation when slope and geometry cues were placed in con-flict (Nardi amp Bingman 2009b Nardi et al 2010) In theseexamples it is likely that slope provided compass-like direc-tional information as well as acting as a local landmark withwhich to distinguish between locations in an environment

Figure 4 Rodent studies of vertical processing (a) Schematic of the lattice maze of Grobeacutety and Schenk (1992a) together with a close-up photograph of a rat in the maze (inset) (b) Schematic and photograph of the pegboard maze used in the experiments of Jovalekic et al(2011) and as described later by Hayman et al (2011) In the Jovalekic et al experiment rats learned to shuttle directly back and forthbetween start and goal locations as indicated in the diagram on the right On test trials a barrier (black bar) was inserted across the usualroute forcing the rats to detour The great majority of rats preferred the routes indicated by the arrows which had the horizontal leg firstand the vertical leg second

In humans similar results were reported in a virtualreality task in which participants using a bicycle simulatornavigated within a virtual town with eight locations(Restat et al 2004 Steck et al 2003 Fig 6C) Participantsexperiencing the sloped version of the virtual town(4-degree slant) made fewer pointing and navigationerrors than did those experiencing the flat version This

indicates that humans can also use slope to orient withgreater accuracy ndash although interestingly this appears tobe more true for males than for females (Nardi et al2011) When slope and geometric cues were placed in con-flict geometric information was evaluated as more relevant(Kelly 2011) This is the opposite of the finding reported inpigeons discussed above There are of course a number of

Figure 5 Visual slant estimation (a) Slope constancy A subject viewing a horizontal surface with a shallow gaze (β1) versus a steep gaze(β2) perceives the slope of the ground to be the same (ie zero) in both cases despite the optical slant (β) being different in the two casesIn order to maintain constant slope perception the nervous systemmust therefore have somemechanism of compensating for gaze angle(b) The scale expansion hypothesis An observer misperceives both the declination of gaze and changes in optical slant with a gain of 15so that all angles are exaggerated but constancy of (exaggerated) slant is locally maintained Increasing distance compression along the lineof sight additionally contributes to slant exaggeration (Li amp Durgin 2010)

reasons why pigeons and humans may differ in their naviga-tional strategies but one possibility is that different strat-egies are applied by animals that habitually travel throughvolumetric space rather than travel over a surface inwhich geometric cues are arguably more salient

The presence of sloping terrain adds complexity to thenavigational calculations required in order to minimizeenergy expenditure A study investigating the ability ofhuman participants to find the shortest routes linking mul-tiple destinations (the so-called travelling salesmanproblem or ldquoTSPrdquo) in natural settings that were eitherflat or undulating found that total distance travelled wasshorter in the flat condition (Phillips amp Layton 2009)This was perhaps because in the undulating terrain partici-pants avoided straight-line trajectories in order to minimizehill climbing and in so doing increased travel distanceHowever it is not known to what extent they minimizedenergy expenditure Interestingly Grobeacutety amp Schenk(1992a) found that rats on the vertical plane made a fargreater number of horizontal movements partly becausethey made vertical translations by moving in a stair patternrather than directly up or down thus minimizing the addedeffort of direct vertical translations This reflects the findingsof the TSP in the undulating environments mentioned above(Phillips amp Layton 2009) and is in line with the reportedhorizontal preferences in rats (Jovalekic et al 2011)

In summary then the presence of sloping terrain notonly adds effort and distance to the animalrsquos trajectorybut also adds orienting information However the extentto which slope is explicitly incorporated into the metric

fabric of the cognitive map as opposed to merely embel-lishing the map is as yet unknown

33 Navigation in multilayer environments

Multilayer environments ndash such as trees burrow systemsor buildings ndash are those in which earth-horizontal x-y coor-dinates recur as the animal explores due to overlappingsurfaces that are stacked along the vertical dimensionThey are thus conceptually intermediate between planarand volumetric environments The theoretical importanceof multilayer environments lies in the potential for con-fusion if ndash as we will argue later ndash the brain prefers to usea plane rather than a volume as its basic metric referenceMost behavioural studies of spatial processing in multi-

layer environments have involved humans and mosthuman studies of three-dimensional spatial processinghave involved multilayer environments (with the exceptionof studies in microgravity discussed below) The core ques-tion concerning multilayer environments is the degree towhich the layers are represented independently versusbeing treated as parts within an integrated whole Oneway to explore this issue is to test whether subjects areable to path integrate across the layers Path integrationintroduced earlier consists of self-motion-tracking usingvisual motor and sensory-flow cues in order to continu-ously update representations of current location andor ahoming vector (for a review see Etienne amp Jeffery 2004)An early study of path integration in three dimensions

explored whether mice could track their position across

Figure 6 Behavioural studies of the use of slope cues in navigation in rats (a and b) and in humans (c) (a) Schematic of the tilted-armradial maze of Grobeacutety amp Schenk (1992b) (b) Schematic of the conical-hill arena of Moghaddam et al (1996) (c) The virtual-realitysloping-town study of Restat et al (2004) and Steck et al (2003) taken from Steck et al (2003) with permission

independent layers of an environment (Bardunias amp Jander2000) The mice were trained to shuttle between their nest-box on one horizontal plane and a goal location on a parallelplane directly above the first by climbing on a vertical wiremesh cylinder joining the two planes When the entireapparatus was rotated by 90 degrees in darkness whilethe animals were climbing the cylinder all mice compen-sated for this rotation by changing direction on the upperplane thus reaching the correct goal location This suggeststhat they had perceived and used the angle of passiverotation and mapped it onto the overhead space Asecond control experiment confirmed that the mice werenot using distal cues to accomplish this thus suggestingthat they were able to rely on internal path integration mech-anisms in order to remain oriented in three dimensionsHowever it is not necessarily the case that the mice cal-

culated a 3D vector to the goal because in this environ-ment the task of orienting in three dimensions could besimplified by maintaining a sense of direction in the hori-zontal dimension while navigating vertically This is gener-ally true of multilayer environments and so to show trueintegration it is necessary to show that subjects incorporatevertical angular information into their encoding In thisvein Montello and Pick (1993) demonstrated that humansubjects who learned two separate and overlapping routesin a multilevel building could subsequently integrate therelationship of these two routes as evidenced by theirability to point reasonably accurately from a location onone route directly to a location on the other However itwas also established that pointing between these verticallyaligned spaces was less accurate and slower than were per-formances within a floor Further humans locating a goal ina multi-storey building preferred to solve the vertical com-ponent before the horizontal one a strategy that led toshorter navigation routes and times (Houmllscher et al2006) In essence rats and humans appear in such tasksto be reducing a three-dimensional task to a one-dimen-sional (vertical) followed by a two-dimensional (horizontal)task However their ability to compute a direct shortestroute was obviously constrained by the availability of con-necting ports (stairways) between the floors and so thiswas not a test of true integrationWilson et al (2004) investigated the ability of humans to

integrate across vertical levels in a virtual reality experimentin which subjects were required to learn the location ofpairs of objects on three levels of a virtual multilevel build-ing Participants then indicated by pointing the directionof objects from various vantage points within the virtualenvironment Distance judgements between floors weredistorted with relative downward errors in upward judge-ments and relative upward errors in downward judgementsThis effect is interesting as the sense of vertical spaceseems to be biased towards the horizontal dimensionHowever in this study there was also a (slightly weaker)tendency to make rightward errors to objects to the leftand leftward errors to objects to the right The resultsmay therefore reflect a general tendency for makingerrors in the direction towards the center point of thespatial range regardless of dimension (although this wasmore pronounced for the vertical dimension) Anotherinteresting finding from this study is that there appears tobe a vertical asymmetry in spatial memories with a biasin favour of memories for locations that are on a lowerrather than higher plane

Tlauka et al (2007) confirmed the findings of Wilsonet al (2004) and suggested that there might be a ldquocontrac-tion biasrdquo due to uncertainty reflecting a compressed ver-tical memory They further speculated that such biasesmight be experience-dependent as humans pay moreattention to horizontal space directly ahead than toregions above or below it This view is interesting andmight explain the experimental results seen with rats inJovalekic et al (2011) Finally young children appear tohave more problems in such three-dimensional pointingtasks (Pick amp Rieser 1982) once again suggesting thatexperience may modulate performance in vertical tasksIn the third part of this article we will argue that the

mammalian spatial representation may be fundamentallyplanar with position in the vertical dimension (orthogonalto the plane of locomotion) encoded separately and in adifferent (non-metric) way from position in the horizontal(locomotor) plane If spatial encoding is indeed planar thishas implications for the encoding of multilevel structuresin which the same horizontal coordinates recur at differentvertical levels On the one hand this may cause confusionin the spatial mapping system That is if the represen-tations of horizontal and vertical dimensions are not inte-grated then the navigational calculations that usehorizontal coordinates may confuse the levels As suggestedabove this may be why for example humans are confusedby multilevel buildings On the other hand if there is if notfull integration at least an interaction between the verticaland horizontal representations then it is possible that sep-arate horizontal codes are able to be formed for each verti-cal level with a corresponding disambiguation of the levelsFurther studies at both behavioural and neurobiologicallevels will be needed to determine how multiple overlap-ping levels are represented and used in navigation

34 Navigation in a volumetric space

Volumetric spaces such as air and water allow free move-ment in all dimensions Relatively little navigationalresearch has been conducted in such environments so farOne reason is that it is difficult to track animal behaviourin volumetric spaces The advent of long-range trackingmethods that use global positioning system (GPS) tech-niques or ultrasonic tagging is beginning to enable researchinto foraging and homing behaviours over larger distances(Tsoar et al 2011) However many of these studies stilladdress navigation only in the earth-horizontal plane ignor-ing possible variations in flight or swim height Laboratorystudies have started to explore navigation in three-dimen-sional space though the evidence is still limitedOne of the earliest studies of 3D navigation in rats the

Grobeacutety and Schenk (1992a) experiment mentionedearlier involved a cubic lattice which allowed theanimals to move in any direction at any point within thecube (see Fig 4A) As noted earlier it was found thatrats learned the correct vertical location before the horizon-tal location suggesting that these elements may be pro-cessed separately Grobeacutety and Schenk suggested that therats initially paid greater attention to the vertical dimensionboth because it is salient and because greater effort isrequired to make vertical as opposed to horizontal trans-lations therefore minimizing vertical error reduced theiroverall effort

By contrast Jovalekic et al (2011) found that rats shut-tling back and forth between diagonally opposite low andhigh points within a lattice maze or on a vertical subset ofit the pegboard (see Fig 4B) did not tend to take thedirect route (at least on the upward journey) but preferredto execute the horizontal component of the journey firstHowever the separation of vertical from horizontal mayhave occurred because of the constraints on locomotioninherent in the maze structure the rats may have foundit easier to climb vertically than diagonally Another exper-iment however also found verticalhorizontal differencesin behavior Jovalekic et al (2011) examined foraging be-haviour in the Grobeacutety and Schenk lattice maze andfound that rats tended to retrieve as much of the food aspossible on one horizontal layer before moving to the nextThese authors observed similar behaviour on a simplifiedversion of this apparatus the pegboard where again fora-ging rats depleted one layer completely before moving tothe next The implication is that rather than using a trulyvolumetric representation of the layout the animals tendto segment the environment into horizontal bands

A similar propensity to treat three-dimensional environ-ments as mainly horizontal has been reported in humans byVidal et al (2004) who studied participants in a virtual 3Dmaze and found that they performed better if theyremained upright and had therefore aligned their ego-centric and allocentric frames of reference in one of thedimensions Similar results were also reported by Aokiet al [2005] Likewise astronauts in a weightless environ-ment tended to construct a vertical reference using visualrather than vestibular (ie gravitational) cues and remainoriented relative to this reference (Lackner amp Graybiel1983 Tafforin amp Campan 1994 Young et al 1984)

Finally one intriguing study has shown that humans maybe able to process not only three- but also four-dimensionalinformation (Aflalo amp Graziano 2008) In this virtual-realitystudy in addition to the usual three spatial dimensions afourth dimension was specified by the ldquohotrdquo and ldquocoldrdquodirections such that turns in the maze could be forwardback left right updown or hotcold and the usual trigono-metric rules specified how ldquomovementsrdquo in the fourth (hot-cold) dimension related to movements in the other threeParticipants were required to path integrate by completinga multi-segment trajectory through the maze and thenpointing back to the start They eventually (after extensivepractice) reached a level of performance exceeding thatwhich they could have achieved using three-dimensionalreasoning alone This study is interesting because it ishighly unlikely that any animals including humans haveevolved the capacity to form an integrated four-dimen-sional cognitive map and so the subjectsrsquo performancesuggests that it is possible to navigate reasonably wellusing a metrically lower-dimensional representation thanthe space itself The question then is whether one can infact navigate quite well in three dimensions using only atwo-dimensional cognitive map We return to this pointin the next section

4 Neurobiological studies in three dimensions

So far we have reviewed behavioural research exploringthree-dimensional navigation and seen clear evidence that3D information is used by animals and humans although

the exact nature of this use remains unclear A parallelline of work has involved recordings from neurons thatare involved in spatial representation This approach hasthe advantage that it is possible to look at the sensoryencoding directly and make fine-grained inferences abouthow spatial information is integrated While most of thiswork has hitherto focused on flat two-dimensional environ-ments studies of three-dimensional encoding are begin-ning to increase in number Here we shall first brieflydescribe the basis of the brainrsquos spatial representation oftwo-dimensional space and then consider based on pre-liminary findings how information from a third dimensionis (or might be) integratedThe core component of the mammalian place represen-

tation comprises the hippocampal place cells first ident-ified by OrsquoKeefe and colleagues (OrsquoKeefe amp Dostrovsky1971) which fire in focal regions of the environment Ina typical small enclosure in a laboratory each place cellhas one or two or sometimes several unique locations inwhich it will tend to fire producing patches of activityknown as place fields (Fig 7A) Place cells have thereforelong been considered to represent some kind of markerfor location Several decades of work since place cellswere first discovered have shown that the cells respondnot to raw sensory inputs from the environment such asfocal odours or visual snapshots but rather to higher-order sensory stimuli that have been extracted from thelower-order sensory data ndash examples include landmarksboundaries and contextual cues such as the colour orodour of the environment (for a review see Jeffery 2007Moser et al 2008)For any agent (including a place cell) to determine its

location it needs to be provided with information aboutdirection and distance in the environment Directionalinformation reaches place cells via the head directionsystem (Taube 2007 Taube et al 1990a) which is a setof structures in areas surrounding the hippocampuswhose neurons are selectively sensitive to the direction inwhich the animalrsquos head is pointing (Fig 7B) Head direc-tion (HD) cells do not encode direction in absolute geo-centric coordinates rather they seem to use localreference frames determined by visual (or possibly alsotactile) cues in the immediate surround This is shown bythe now well-established finding that rotation of a singlepolarizing landmark in an otherwise unpolarized environ-ment (such as a circular arena) will cause the HD cells torotate their firing directions by almost the same amount(Goodridge et al 1998 Taube et al 1990b) Interestinglyhowever there is almost always a degree of under-rotationin response to a landmark rotation This is thought to bedue to the influence of the prevailing ldquointernalrdquo directionsense sustained by processing of self-motion cues such asvestibular proprioceptive and motor signals to motionThe influence of these cues can be revealed by removing asingle polarizing landmark altogether ndash the HD cells willmaintain their previous firing directions for several minutesalthough they will eventually drift (Goodridge et al 1998)The internal (sometimes called idiothetic) self-motion

cues provide a means of stitching together the directionalorientations of adjacent regions of an environment so thatthey are concordant This was first shown by Taube andBurton (1995) and replicated by Dudchenko and Zinyuk(2005) However although there is this tendency for thecells to adopt similar firing directions in adjacent

environments especially if the animal self-locomotedbetween them this is not absolute and cells may have dis-cordant directions flipping from one direction to the otheras the animal transitions from one environment to the next(Dudchenko amp Zinyuk 2005) This tendency for HD cellsto treat complex environments as multiple local fragmentsis one that we will return to laterThe source of distance information to place cells is

thought to lie at least partly in the grid cells which areneurons in the neighbouring entorhinal cortex thatproduce multiple place fields that are evenly spaced in agrid-like array spread across the surface of the environ-ment (Hafting et al 2005 for review see Moser ampMoser 2008) (Fig 7c) The pattern formed by the fieldsis of the type known as hexagonal close-packed which isthe most compact way of tiling a plane with circlesGrid cell grids always maintain a constant orientation for

a given environment (Hafting et al 2005) suggesting adirectional influence that probably comes from the head-direction system as suggested by the finding that thesame manipulations that cause HD firing directions torotate also cause grids to rotate Indeed many grid cellsare themselves also HD cells producing their spikes onlywhen the rat is facing in a particular direction (Sargoliniet al 2006) More interesting however is the influenceof odometric (distance-related) cues on grid cellsBecause the distance between each firing field and itsimmediate neighbours is constant for a given cell thegrid cell signal can theoretically act as a distance measurefor place cells and it is assumed that this is what they arefor (Jeffery amp Burgess 2006) though this has yet to beproven The source of distance information to the gridcells themselves remains unknown It is likely that someof the signals are self-motion related arising from themotor and sensory systems involved in commandingexecuting and then detecting movement through spaceThey may also be static landmark-related distance signals ndashthis is shown by the finding that subtly rescaling an

environment can cause a partial rescaling of the gridarray in the rescaled dimension (Barry et al 2007) indicat-ing environmental influences on the grid metric It seemsthat when an animal enters a new environment an arbitrarygrid pattern is laid down oriented by the (also arbitrary)HD signal This pattern is then ldquoattachedrdquo to that environ-ment by learning processes so that when the animal re-enters the (now familiar) environment the same HD cellorientation and same grid orientation and location canbe reinstatedGiven these basic building blocks of the mammalian

spatial representation we turn now to the question ofhow these systems may cope with movement in the verticaldomain using the same categories as previously slopemultilayer environments and volumetric spaces

41 Neural processing of verticality cues

The simplest form of three-dimensional processing isdetection of the vertical axis Verticality cues are thosethat signal the direction of gravity or equivalently the hori-zontal plane which is defined by gravity The brain canprocess both static and dynamic verticality cues Staticinformation comprises detection of the gravitational axisor detection of slope of the terrain underfoot whiledynamic information relates to linear and angular move-ments through space and includes factors such as effortBoth static and dynamic processes depend largely onthe vestibular apparatus in the inner ear (Angelaki ampCullen 2008) and also on a second set of gravity detectors(graviceptors) located in the trunk (Mittelstaedt 1998)Detection of the vertical depends on integration of vestib-ular cues together with those from the visual world andalso from proprioceptive cues to head and body alignment(Angelaki et al 1999 Merfeld et al 1993 Merfeld amp Zupan2002)The vestibular apparatus is critical for the processing not

only of gravity but also of movement-related inertial three-

Figure 7 Firing patterns of the three principal spatial cell types (a) Firing of a place cell as seen from an overhead camera as a ratforages in a 1 meter square environment The action potentials (ldquospikesrdquo) of the cell are shown as spots and the cumulative path ofthe rat over the course of the trial is shown as a wavy line Note that the spikes are clustered towards the SouthndashEast region of thebox forming a ldquoplace fieldrdquo of about 40 cm across (b) Firing of a head direction cell recorded as a rat explored an environmentHere the heading direction of the cell is shown on the x-axis and the firing rate on the y-axis Note that the firing intensifiesdramatically when the animalrsquos head faces in a particular direction The cell is mostly silent otherwise (c) Firing of a grid cell (fromHafting et al 2005) depicted as for the place cell in (a) Observe that the grid cell produces multiple firing fields in a regularlyspaced array Adapted from Jeffery and Burgess (2006)

dimensional spatial cues The relationship of the com-ponents of the vestibular apparatus is shown in Figure 8awhile the close-up in Figure 8b shows these core com-ponents the otolith organs of which there are two ndash theutricle and the saccule ndash and the semicircular canals Theotolith organs detect linear acceleration (includinggravity which is a form of acceleration) by means of special-ized sensory epithelium In the utricle the epithelial layer isoriented approximately horizontally and primarily detectsearth-horizontal acceleration and in the saccule it isoriented approximately vertically and primarily detects ver-tical accelerationgravity The two otolith organs worktogether to detect head tilt However because gravityis itself a vertical acceleration there is an ambiguityconcerning how much of the signal is generated bymovement and how much by gravity ndash this ldquotilt-translationambiguityrdquo appears to be resolved by means of the semi-circular canals (Angelaki amp Cullen 2008) These arethree orthogonal fluid-filled canals oriented in the threecardinal planes one horizontal and two vertical whichcollectively can detect angular head movement in anyrotational plane

Neural encoding of direction in the vertical plane (tiltpitch roll etc) is not yet well understood Head directioncells have been examined in rats locomoting on a verticalplane and at various degrees of head pitch to seewhether the cells encode vertical as well as horizontal direc-tion Stackman and Taube (1998) found pitch-sensitivecells in the lateral mammillary nuclei (LMN) Howeverthese were not true volumetric head direction cellsbecause their activity was not at all related to horizontalheading direction Nor did they correspond to the verti-cal-plane equivalent of HD cells Most of the observedcells had pitch preferences clustered around the almost--vertical suggesting that these cells were doing somethingdifferent from simply providing a vertical counterpart ofthe horizontal direction signal By contrast the HD cells inLMN were insensitive to pitch and their firing was uni-formly distributed in all directions around the horizontalplane Thus vertical and horizontal directions appear to beseparately and differently represented in this structure

In a more formal test of vertical directional encodingStackman et al (2000) recorded HD cells as rats climbedmoveable wire mesh ladders placed vertically on the sidesof a cylinder When the ladder was placed at an angularposition corresponding to the cellrsquos preferred firing direc-tion the cell continued to fire as the animal climbed theladder but did not fire as the animal climbed downagain Conversely when the ladder was placed on theopposite side of the cylinder the reverse occurred nowthe cell remained silent as the animal climbed up butstarted firing as the animal climbed down This suggeststhat perhaps the cells were treating the plane of the wallin the same way that they usually treat the plane of thefloor Subsequently Calton and Taube (2005) showedthat HD cells fired on the walls of an enclosure in amanner concordant with the firing on the floor and alsoinformally observed that the firing rate seemed to decreasein the usual manner when the ratrsquos head deviated from thevertical preferred firing direction This observation wasconfirmed in a follow-up experiment in which rats navi-gated in a spiral trajectory on a surface that was either hori-zontal or else vertically aligned in each of the four cardinalorientations (Taube et al 2013) In the vertical conditionsthe cells showed preferred firing directions on the verticalsurface in the way that they usually do on a horizontalsurface When the vertical spiral was rotated the cellsswitched to a local reference frame and maintained theirconstant preferred firing directions with respect to thesurfaceWhat are we to make of these findings In the Calton

and Taube (2005) experiment it appears that the HDcells seemed to be acting as if the walls were an extensionof the floor ndash in other words as if the pitch transformationwhen the rat transitioned from horizontal to vertical hadnever happened A possible conclusion is that HD cellsare insensitive to pitch which accords with the LMN find-ings of Stackman and Taube (1998) This has implicationsfor how three-dimensional space might be encoded andthe limitations thereof which we return to in the finalsection In the spiral experiment described by Taube(2005) it further appears that the reference frame provided

Figure 8 The vestibular apparatus in the human brain (a) Diagram showing the relationship of the vestibular apparatus to the externalear and skull (b) Close-up of the vestibular organ showing the detectors for linear acceleration (the otolith organs ndash comprising the utricleand saccule) and the detectors for angular acceleration (the semicircular canals one in each plane) (Taken from httpwwwnasagovaudienceforstudents9-12featuresF_Human_Vestibular_System_in_Spacehtml) Source NASA

by the locomotor surface became entirely disconnectedfrom the room reference frame ndashwhen the rat was clingingto the vertical surface the cells became insensitive tothe rotations in the azimuthal plane that would normallymodulate their firing on a horizontal plane The idea oflocal reference frames is one that we will also come backto laterThe Calton and Taube (2005) experiment additionally

replicated an observation made in microgravity which isthat head direction cells lose their directional specificityor are greatly degraded during inverted locomotion onthe ceiling (Taube et al 2004) In another study headdirection cells were monitored while animals were heldby the experimenter and rotated by 360 degrees (relativeto the horizontal plane) in the upright and in the invertedpositions as well as by 180 degrees in the vertical plane(Shinder amp Taube 2010) Although the animal wasrestrained head direction cells displayed clear directional-ity in the upright and vertical positions and no directional-ity in the inverted position Additionally directionality inthe vertical position was apparent until the animal wasalmost completely inverted These findings confirm thoseof the previous studies (Calton amp Taube 2005 Taubeet al 2004) that loss of HD cell directionality is a featureof the inverted position at least in ratsThis breakdown in signalling may have arisen from an

inability of the HD cells to reconcile visual evidence forthe 180 degree heading-reversal with the absence of a180 degree rotation in the plane to which they are sensitive(the yaw plane) This finding is consistent with a recentbehavioural study by Valerio et al (2010) Here ratswere trained to locomote while clinging upside-down tothe underside of a circular arena in order to find anescape hole Animals were able to learn fixed routes tothe hole but could not learn to navigate flexibly in amapping-like way suggesting that they had failed to forma spatial representation of the layout of the arena Takingthese findings together then it appears that the rodentcognitive mapping system is not able to function equallywell at all possible orientations of the animalWhat about place and grid cells in vertical space Proces-

sing of the vertical in these neurons has been explored in arecent study by Hayman et al (2011) They used twoenvironments a vertical climbing wall (the pegboard)and a helix (Fig 9) In both environments place and gridcells produced vertically elongated fields with grid fieldsbeing more elongated than place fields thus appearingstripe-like (Fig 9E) Hence although elevation wasencoded its representation was not as fine-grained as thehorizontal representation Furthermore the absence ofperiodic grid firing in the vertical dimension suggests thatgrid cell odometry was not operating in the same way asit does on the horizontal planeOne possible explanation is that grid cells do encode the

vertical axis but at a coarser scale such that the periodicityis not evident in these spatially restricted environments Ifso then the scale or accuracy of the cognitive map at leastin the rat may be different in horizontal versus verticaldimensions possibly reflecting the differential encodingrequirements for animals that are essentially surface-dwell-ing An alternative possibility explored below is that whilehorizontal space is encoded metrically vertical space isperhaps encoded by a different mechanism possibly evena non-grid-cell-dependent one Note that because the

rats remained horizontally oriented during climbing it isalso possible that it is not the horizontal plane so much asthe current plane of locomotion that is represented metri-cally while the dimension normal to this plane (the dorso-ventral dimension with respect to the animal) is rep-resented by some other means Indeed given the HDcell data discussed above this alternative seems not onlypossible but likely Thus if the rats had been verticallyoriented in these experiments perhaps fields would havehad a more typical hexagonal patternIt is clear from the foregoing that much remains to be

determined about vertical processing in the navigationsystem ndash if there is a vertical HDndashcell compass system ana-logous to the one that has been characterized for the hori-zontal plane then it has yet to be found and if elevation ismetrically encoded the site of this encoding is alsounknown The cells that might have been expected toperform these functions ndash head direction place and gridcells ndash do not seem to treat elevation in the same way asthey treat horizontal space which argues against the likeli-hood that the mammalian brain at least constructs a trulyintegrated volumetric mapWe turn now to the question of what is known about the

use of three-dimensional cues in navigationally relevantcomputations ndash processing of slopes processing of multi-layer environments and processing of movement in volu-metric space

42 Neural encoding of a sloping surface

Investigation of neural encoding of non-horizontal surfacesis only in the early stages but preliminary studies have beenundertaken in both the presence and the absence ofgravity In normal gravity conditions a slope is character-ized by its steepness with respect to earth-horizontalwhich provides important constraints on path integrationIf the distance between a start location and a goal includesa hill additional surface ground has to be travelled toachieve the same straight-line distance (ie that of flatground) Furthermore routes containing slopes requiremore energy to traverse than do routes without slopesrequiring a navigator to trade off distance against effort inundulating terrain Therefore one might imagine thatslope should be incorporated into the cognitive map andneural studies allow for an investigation of this issue thatis not feasible with behavioural studies aloneThe effect of terrain slope has been explored for place

cells but not yet for grid cells The earliest place cellstudy by Knierim and McNaughton (2001) investigatedwhether place fields would be modulated by the tilt of arectangular track A priori one possible outcome of thisexperiment was that place fields would expand and contractas the tilted environment moved through intrinsicallythree-dimensional ovoid place fields However fieldsthat were preserved following the manipulation did notchange their size and many place cells altered their firingaltogether on the track evidently treating the whole tiltedtrack as different from the flat track One interpretationof these findings is that the track remained the predomi-nant frame of reference (the ldquohorizontalrdquo from the perspec-tive of the place cells) and that the tilt was signalled by theswitching on and off of fields rather than by parametricchanges in place field morphology

This study also showed that slope could be used as anorienting cue In rotations of the tilted track cells weremore likely to be oriented by the track (as opposed to theplethora of distal cues) than in rotations of the flat trackThis observation was replicated in a subsequent study byJeffery et al (2006) who found that place cells used the30-degree slope of a square open arena as an orienting cueafter rotations of the arena and preferred the slope to localolfactory cues Presumably because the cells all reorientedtogether the effect was mediated via an effect on the headdirection system Place cells have also been recorded frombats crawling on a near-vertical (70-degree tilt) open-field

arena (Ulanovsky amp Moss 2007) Firing fields were similarto rodent place fields on a horizontal surface a pattern thatis consistent with the fields using the surface of the environ-ment as their reference plane (because if earth-horizontalwere the reference then fields should have been elongatedalong the direction of the slope as in the Hayman et al[2011] experiment) (see Fig 9e)Grid cells have not yet been recorded on a slope The

pattern they produce will be informative because theirodometric properties will allow us to answer the questionof whether the metric properties of place fields inhorizontal environments arise from distances computed

Figure 9 Adapted from Hayman et al (2011) (a) Photograph of the helical maze Rats climbed up and down either five or six coils ofthe maze collecting food reward at the top and bottom while either place cells or grid cells were recorded (b) Left ndash The firing pattern ofa place cell (top spikes in blue) and a grid cell (bottom spikes in red) as seen from the overhead camera The path of the rat is shown ingrey The place cell has one main field with a few spikes in a second region and the grid cell has three fields Right - the same data shownas a heat plot for clarity (red = maximum firing blue = zero firing) (c) The same data as in (b) but shown as a firing rate histogram as ifviewed from the side with the coils unwound into a linear strip The single place field in (b) can be seen here to repeat on all of the coils asif the cell is not discriminating elevation but only horizontal coordinates The grid cell similarly repeats its firing fields on all the coils (d)Photograph of the pegboard studded with wooden pegs that allowed the rat to forage over a large vertical plane (e) Left ndash The firingpattern of a place cell (top spikes in blue) and a grid cell (bottom spikes in red) as seen from a horizontally aimed camera facing thepegboard The place cell produced a single firing field but this was elongated in the vertical dimension The grid cell producedvertically aligned stripes quite different from the usual grid-like pattern seen in Fig 8c Right ndash the same data as a heat plot

in earth-horizontal or with respect to environment surfaceThe results discussed above lead us to predict the latterOnly one place cell study has been undertaken in the

absence of gravity in an intriguing experiment byKnierim and colleagues conducted on the NeurolabSpace Shuttle mission of 1998 (Knierim et al 20002003) Rats were implanted with microelectrodes forplace cell recording and familiarized with running on a rec-tangular track on Earth before being carried in the shuttleinto space Under microgravity conditions in space theanimals were then allowed to run on a 3D track theldquoEscher staircaserdquo which bends through three-dimensionalspace such that three consecutive right-angle turns in theyaw plane of the animal interspersed with three pitchrotations leads the animal back to its starting point(Fig 10a) The question was whether the cells could laydown stable place fields on the track given that the visualcues might have conflicted with the path integrationsignal if the path integrator could not process the 3Dturns On the ratsrsquo first exposure to the environment onflight day 4 the cells showed somewhat inconsistent pat-terns with those from one rat showing degraded spatialfiring those from a second rat showing a tendency torepeat their (also degraded) firing fields on each segmentof the track and those from a third looking like normalplace fields By the second recording session on flightday 9 place fields from the first two rats had gainednormal-looking properties with stable place fields(Fig 10b) This suggests either that the path integratorcould adapt to the unusual conditions and integrate theyaw and pitch rotations or else that the visual cuesallowed the construction of multiple independent planarrepresentations of each segment of the trackIt will be important to answer the question of which of

these explanations is correct If the cells could integrateturns in three dimensions this would imply a highly soph-isticated ability of the head direction system to co-ordinateturns in the three axes of rotation which ndash given that they

respond mainly to yaw ndash suggests an integrative processoutside the HD cells themselves

43 Neural encoding of multilayer environments

Very little work has examined neural encoding of space inmultilayer environments which is unfortunate given thetheoretical importance of this issue discussed earlierStackman et al (2000) compared the firing of HD cellsbetween the floor of a cylinder and an annular mezzanine(ldquoannulusrdquo) located 74 cm above the floor and found a30 increase in firing rate on the annulus However thewire-mesh studies of Stackman et al (2000) and Caltonand Taube (2005) did not report increases in firing rateas animals climbed the walls so the height effect mayperhaps have been due to arousal or anxiety on theannulus Directional firing preferences were shifted slightlyon the upper level with respect to the lower which mayhave resulted from transient slight disorientation duringthe climbThere is also little work on place and grid cells in multi-

layer environments Fenton et al (2008) recorded placecells in a large environment the walls of which supportedstairs leading to an upper ledge housing reward drinkingports They found that frequently fields that were presenton the floor were also present on the corresponding partof the stairsledge This could mean that either the fieldswere three-dimensional and extended from the floor to thestaircase or else that the cells were only responding to hori-zontal coordinates and ignoring the vertical Results from theHayman et al (2011) experiment on the helical track supportthe latter interpretation in that the data showed that firingtended to repeat on each successive coil (Fig 9 b and c)There was modulation of firing rate across the coils whichis not inconsistent with the notion that the fields may havebeen intrinsically three-dimensional (albeit very elongatedin the vertical dimension) However it could also simplybe rate modulation of the same field on each coil Rate

Figure 10 (a) The ldquoEscher staircaserdquo track on which rats ran repeated laps while place fields were recorded (b) The firing field of aplace cell showing a consistent position on the track indicating either the use of visual cues or of three-dimensional path integration(or both) to position the field (From Knierim et al [2000] with permission)

modulation of place fields has been seen in other situations(Leutgeb et al 2004) and suggests an influence of someother cues (in this case elevation) on field productionGrid cells in the Hayman et al experiment showed asimilar pattern with field location repeating from one coilto the next but also (as with the place cells) a slightdegree of rate modulation ndash something interestingly thathas not been reported in other experiments on grid cells

The implication of these findings is that the system wastreating the environment as a stack of horizontal planesrather than as an integrated three-dimensional spaceHowever the helical track experiment is not a true testof multilayered environment representation becausethere was no discrete boundary between layers as there isin say a multi-storey building Also the stereotypicalnature of the route the animals ran might have precludedthe formation of independent maps for each layer Atruly multi-storey apparatus will be required to resolvethis issue

44 Neural encoding of a volumetric space

The final and most important question concerning three-dimensional spatial encoding is whether there is metricencoding of volumetric space As with multilayer environ-ments there have been no studies done to date on neuralresponses to free movement through true volumetricspaces and so the answer to this question remains openHowever it is interesting to speculate about what kindsof evidence we should look for

A volumetric map requires direction and distance to becalculated for both the horizontal and vertical dimensionsWe have already seen that there is no evidence as yet forthree-dimensional head direction cells at least in rodentsso either there are three-dimensional HD cells in someother region of the brain or else the cognitive map has away of combining the planar signal from the classic HDareas together with pitch or elevation informationperhaps from the lateral mammillary nucleiWhat about distance in three dimensions Because grid

cells seem to provide the odometric representation forhorizontal distance by virtue of their regularly spaced gridfields we could predict that the periodicity observed inthe horizontal dimension might be part of a three-dimen-sional lattice-like arrangement of spherical fields Thereare two ways of close-packing spheres in a volumetermed hexagonal and face-centered cubic respectively(Fig 11a and b) Since we already know that the gridarray on the horizontal plane is hexagonal close-packed ifeither of these volumetric arrangements were to pertainthen we would predict it should be the hexagonal formAlthough no recordings have yet been made of grids in a

volumetric space the experiment by Hayman et al (2011)of grids recorded on a vertical plane might have beenexpected to reveal evidence of periodicity in the verticaldimension akin to what one would find by transecting aclose-packed volume (Fig 11c) However that experimentfound that the grid fields formed stripes on the verticalsurface implying instead that the true underlying patternof grids in a volumetric space might be in the form of

Figure 11 The two ways of efficiently filling a volumetric space with spheres (a)Hexagonal close-packing in which the spheres on eachlayer are planar hexagonal close-packed and offset between one layer and the next and (b) Face-centered cubic in which the spheres oneach layer are in a square arrangement with each sphere in a given layer nestling in the dip formed by four of those in the layers on eitherside (c) The result of transecting a three-dimensional hexagonal close-packed array is a regular set of circles (d) The result of transectinga set of horizontally hexagonal close-packed columns is stripes resembling the grid cell pattern shown in Figure 9e

hexagonal close-packed columns (Fig 11c) Whether thiswould also hold true for animals moving in an uncon-strained way through the space remains to be seen It isalso possible that there is periodicity in the vertical dimen-sion but at a much greater scale than could be seen in asmall laboratory setting

5 Is the cognitive map bicoded

At the beginning of this article we reviewed the encodingschemes that might exist to map three-dimensional spacehighlighting three different mapping schemes that vary intheir use of elevation information (see Fig 1) Based onthe foregoing review of behavioural and neurobiologicalstudies we return now to the question of map typesBelow we argue that the data collected to date favour ametrically planar map of the form we have termedldquobicodedrdquo ndash that is using a different encoding scheme forhorizontal than for vertical space Indeed we go a stepfurther and suggest that the schemes are referenced notto horizontal and vertical but more generally to theplane of locomotion (horizontal in the animalrsquos canonicalorientation) versus the axis orthogonal to thisFirst however let us consider the alternative hypotheses

for the possible structure of the cognitive map These are(1) that the map is a surface map lacking informationabout vertical travel or (2) that it is a fully integrated volu-metric map It is unlikely that the cognitive map could be asurface map because the evidence reviewed suggests thatalmost all animals investigated show some processingof vertical space A volumetric map by contrast is metric(ie distances and directions are explicitly encoded) in allthree dimensions No studies have yet been undertakenin animals that can move freely in all three dimensions(such as those that fly or swim) and it may be thatthrough evolution andor through developmental experi-ence a fully integrated volumetric map may have formedin such animals In particular animals that have to solvethree-dimensional spatial problems for which there is noplanar approximation such as those that have to path inte-grate across a volumetric space like the ocean may beexpected to possess a volumetric cognitive map if such athing exists Studies in birds fish or swimming and flyingmammals such as cetaceans or bats have the potential toreveal whether there are spatial metric computations fortravel in all three dimensions and also whether these areintegratedIf the cognitive map in animals that have a truly three-

dimensional ecology is volumetric we might expect to dis-cover in them a new class of head direction cells thatencode the vertical plane in the same way that the rodentHD cells discovered to date encode the horizontal planeAlternatively these animals might even possess HD cellsthat point uniquely to positions in three-dimensionalspace (ie that have tuning curves restricted in all threedimensions) We might also expect to see spherical gridcell grids in a 3D hexagonal close-packed array like thoseshown in Figure 11a and b

51 Arguments for a bicoded cognitive map

The above possibilities concerning 3D maps notwithstand-ing we suggest here that even ecologically three-

dimensional animals may turn out to form planar orquasi-planar cognitive maps Such animals often live inenvironments that are relatively more extended horizon-tally than vertically making a planar map useful (particu-larly as almost all animals maintain a canonical bodyorientation with respect to gravity) Also as discussedbelow encoding three dimensions poses considerable tech-nical challenges for a compass system Indeed there is littleevidence for such integrated encoding in the species thathave been studied so far Head direction cells evidentlyoperate in a plane rather than a volume and grid cellsare not periodic in the vertical dimension (or perhaps notin the dimension perpendicular to the animalrsquos body plane)In a bicoded map there is a spatially metric represen-

tation of a reference plane ndash usually the horizontalplane ndash and a non-metric or at least differently metricizedrepresentation for the axis perpendicular to this plane Inthe geographical example given earlier (Fig 1c) the non-spatial indicator of distance was colour which varies as afunction of elevation but from which elevation cannot beextracted without a key Loosely speaking it is as if thenon-metric variable conveys a coarse signal along thelines of ldquohighrdquo or ldquovery highrdquo but does not contain explicitinformation about distance that could enter into a trigono-metric calculationObviously the cognitive map does not use colour as its

non-metric variable but it could plausibly use an analogoussignal which we call here for want of a better word ldquocon-textualrdquo ndash that is information that pertains to a space butneed not itself contain distance or direction informationA real-world example of a contextual elevation cue mightbe hydrostatic pressure which Burt de Perera and col-leagues put forward as a possible mechanism for fish tolocalize themselves in the vertical axis (Burt de Pereraet al 2005) For terrestrial animals it could be visual orkinaesthetic cues to ascentdescent or aspects of thevisual panorama for flying animals it could be depthcues from stereopsis or motion parallax or perhaps olfac-tory gradients in the atmosphereOur hypothesis then is that the cognitive map is such a

bicoded map ndash it is metric in a given plane (the plane oflocomotion) and contextually modulated orthogonal tothis It is possible that contextual information concerningelevation could be processed by brain areas that arelocated outside the currently established navigation circui-try although it is likely that these signals would interact atsome pointWhat evidence supports this bicoding view Existing be-

havioural data from a number of studies reviewed earliersuggest a potential separation between horizontal and ver-tical dimensions However this does not prove that theencoding itself is different merely that execution of naviga-tional plans is different Theoretically this could bebecause the energetic costs of moving vertically are takeninto account when planning even if the planning is basedon a fundamentally volumetric three-dimensional mapThe neural data however tell a more revealing story Theobservation that head direction cells appear to use theplane of locomotion as a local reference (Stackman et al2000) implies that the directional signal for the verticalplane (or rather the plane normal to locomotion) must lieelsewhere in an as yet undiscovered region Furthermorethe finding that grid and place cells have different metricproperties in the vertical dimension at least for horizontally

oriented body posture implies that odometry does not takeplace equally in all three dimensions ndash the map is anisotropic(Hayman et al 2011)

Note that it remains to be determined experimentallywhether there is something special about the horizontalplane in the cognitive map or whether it is simply thatgravity primarily orients the locomotor plane and it isthis plane that is metrically encoded Recordings ofgrid cells on steeply sloping surfaces may answer thisquestion

A bicoded scheme seems suboptimal but in fact asurface-travelling animal could do quite well with such amap because it could use the metrically encoded loco-motor plane to calculate the directions needed for traveland the topography of the landscape would constrain theanimal to be at the correct elevation for that positionAlthough a bicoded map could not be used for path inte-gration in volumetric space an animal could divide thenavigational task into two parts It could first calculate thehorizontal trajectory and then ascend to the correctelevation (as indicated by the contextual gradient) or viceversa ndash or both simultaneously but in separate compu-tational circuits Indeed the findings of Grobeacutety andSchenk (1992a) seem to suggest that this is indeed whathappens at least in rats However in the case where ldquohori-zontalrdquo is not true earth-horizontal but lies on a slope thenit would be advantageous for the map to encode in someform information about that slope as this would affectthe energy costs of navigational routes involving thatsurface Hence there may be some mechanism for encod-ing the angle of a given planar fragment with respect toearth-horizontal

A bicoded map has a planar metric reference but asnoted earlier the real world is far from planar Howcould a bicoded map be used on undulating topologyOne possibility is that an undulating surface could beencoded as a manifold ndash not a single unified map but a

mosaic of contiguous map fragments In such a schemeeach fragment would comprise a local reference framedefined by a reference plane (which would be horizontalon a horizontal surface but need not be) and an axis orthog-onal to this (which would correspond to the gravitationallydefined vertical for an animal locomoting in its canonicalorientation on a horizontal surface) Figure 12 shows anillustrative example of such a mosaicized representationIn the example a rat in a fairly short excursion movesacross the ground from a rock to a tree up the trunk ofthe tree and out onto a branch ndash each of these planesbeing orthogonal to the one before To maintain orien-tation the animal must switch from a planar map refer-enced to the ground to a new one referenced to the treetrunk to a third referenced to the branchIf the planar fragments of the cognitive map need not be

oriented horizontally then how does the animal process thevertical axis which remains a special and important refer-ence due to the effects of gravity To form an integratedmap of large-scale space using mosaic fragments therealso should be some means of associating the local refer-ence frames both to one another (to enable navigationbetween fragments) and to earth-horizontal (to enableappropriate energy and other gravity-related calculationsto be made) A plausible means of achieving this could bethe vestibular system which tracks changes in orientationvia the semi-circular canals and could therefore providesome means of linking the map fragments together at soto speak their ldquoedgesrdquoWith such a scheme it would be poss-ible to use heuristic methods to compute for exampleapproximate shortcuts between map fragments For moresophisticated navigation however it would be preferableto weld fragments together into a larger unitary framethat would allow for precise metric computations acrossthe space It may be that such a reference frame adaptationcould occur with experience In fact the head direction cellstudies of Stackman et al (2000) in which reference frames

Figure 12 Hypothetical structure of the cognitive map in a dimensionally complex environment The map is assumed to be metric inthe plane of locomotion which is the ground for the space between the rock and the tree the vertical trunk as the animal climbs and thehorizontal branch The shading gradient represents the nonmetric encoding which conveys coarse information about distance in theplane orthogonal to the locomotor plane which is encoded differently from the precise metric of the locomotor plane

could be local to the surface or more globally referenced tothe surrounding room support the possibility of such experi-ence-dependent plasticityThe possibility that the cognitive map is formed of a

mosaic of locally planar fragments is consistent with thesudden reorientations experienced by humans and otheranimals as they transition from one region of the environ-ment to the next (Wang amp Spelke 2002) It is a common-place experience of urban navigators when taking anunfamiliar route between two locally familiar places forexample that a sudden almost dizzying re-alignment ofreference frames occurs across the interface between thetwo regions This subjective experience may reflect thesudden reorientation of head direction cells that occurswhen animals cross between local connected regionshaving different head direction cell orientations (Dud-chenko amp Zinyuk 2005 Taube amp Burton 1995 Zugaroet al 2003) The reorientation episodes experienced byastronauts in conditions of microgravity may have asimilar underlying cause (Oman 2007 See Fig 12)What about subjects that habitually navigate in volu-

metric spaces including animals such as fish birds andbats and also human divers aviators and astronauts Dothey use a surface map too In these environments aplane-referenced map would seem much less usefulHowever evidence from microgravity environmentstudies shows that even in a weightless environment astro-nauts tend to find a reference plane to serve as a localldquofloorrdquo (Oman 2007) This is usually the surface beneaththeir feet but if they drift too far away from this ldquofloorrdquoand too far toward the ldquoceilingrdquo or switch visual attentionthey frequently report visual reorientation illusions (VRIs)where surfaces abruptly and unexpectedly exchangeidentity (Oman et al 1986) testifying to the salience ofthe reference plane Volume-travelling animals tend tomaintain a constant body posture with respect to earth-horizontal and so they too may use a planar map for navi-gating even though their movement in space isunconstrained

52 Why would animals have a metrically planar map

Why would animal cognitive maps be (quasi-)planar Onthe face of it this seems maladaptive because a planarmap even a bicoded one has inherently less informationthan a fully integrated volumetric oneThe first possibility is an ontogenetic one that formation

of the map during development is constrained by experi-ence during infancy and the animals that have been inves-tigated to date have been mainly for technical reasonsthose species that are primarily surface-dwelling An excep-tion is fish in which we also saw a processing separationbetween horizontal and vertical space (Holbrook amp Burtde Perera 2009) but this could have been a behaviouraladaptation superimposed on an underlying volumetricmap It may be that if rats and mice were to be raisedfrom birth in a volumetric space ndash for example in micro-gravity or at least in a lattice system in which they couldmove in all directions ndash then we would see true three-dimensional HD cells and path integration behaviourthat seamlessly encompassed all three dimensionsSecond it may be that phylogenetic development of the

map was so constrained and therefore surface-dwellinganimals such as rats mice and humans would never

develop a 3D map even if raised in conditions that wouldallow it In other words surface-dwellers have either lostor else never evolved the neurological capacity to fully rep-resent three-dimensional space However if we were tostudy HD cells in animals that can move freely in all dimen-sions then we may find three-dimensional HD cells inthese species Emerging studies in bats may soon answerthis question we hope It should be noted however thatrats and mice naturally inhabit dimensionally complexenvironments and therefore might be expected to showintegrated three-dimensional encoding if this ever didevolve in vertebratesThe third and final possibility ndash and one for which we

argue here ndash is that a fully 3D map has never evolved inany species because of the difficulties inherent in stablyencoding an allocentric three-dimensional space using ego-centric sensory receptors A volumetric map requires threecoordinates for position and three for direction Monitoringposition and travel in such a space is complicated becausethe vestibular inputs to the head direction system originatein the semicircular canals which are themselves planar(one aligned in each plane) In order to extract three-dimensional heading the system would need to modulatethe rotational signal from each canal with each of theother two and do so dynamically and instantaneously Itis possible that the cost-benefit ratio of the required extraprocessing power is not sufficiently great even foranimals that have a three-dimensional ecological nicheAs we saw in the four-dimension experiment of Aflaloand Graziano (2008) it is possible to navigate quite wellin a space using a lower-dimensional representationtogether with some heuristics and what applies in fourdimensions may equally well apply in three

6 Conclusion

To conclude then we have reviewed emerging experimen-tal evidence concerning the encoding of three-dimensionalspace in animals and humans We find clear evidence ofability to process information about travel in the verticaldimension particularly in the domain of slope intensityand slope direction There is also some evidence of the pro-cessing of elevation information Evidence for true volu-metric coding is however weak and both behaviouraland neural studies suggest the alternative hypothesis thatthe neural representation of space in a wide variety ofspecies is metrically flat (referenced to a plane ndash usually theplane of locomotion) and is modulated in the vertical dimen-sion in a non-metric way We have termed such a mapldquobicodedrdquo to reflect its essentially anisotropic nature and wesuggest that a bicoded representation is a universal featureof vertebrate (and possibly invertebrate) cognitive maps

ACKNOWLEDGMENTSThe work was supported by grants to Kate Jeffery from theWellcome Trust (GR083540AIA) the European CommissionFramework 7 (ldquoSpacebrainrdquo) the Medical Research Council(G1100669) and the Biotechnology and Biological SciencesResearch Council (BBJ0097921) and by a grant to Axona Ltdfrom the European Commission Framework 7 (ldquoSpacebrainrdquo)

NOTES1 Madeleine Verriotis and Aleksandar Jovalekic contributed

equally to this target article

2 Note that some other part of the animalrsquos brain may be ableto deduce via postural and kinaesthetic mechanisms the slopeof the terrain The point is that this information is not integratedinto the metric fabric of the map itself

Open Peer Commentary

Semantic sides of three-dimensional spacerepresentation

doi101017S0140525X13000307

Arnaud BadetsCentre de Recherches sur la Cognition et lrsquoApprentissage Centre National dela Recherche Scientifique (CNRS) UMR-7295 Maison des Sciences delrsquoHomme et de la Socieacuteteacute 86000 Poitiers Francearnaudbadetsuniv-poitiersfrhttpcercalabouniv-poitiersfr

Abstract In this commentary I propose that horizontal and verticaldimensions of space are represented together inside a common metricsmechanism located in the parietal cortex Importantly this network isalso involved in the processing of number magnitudes and environment-directed actions Altogether the evidence suggests that differentmagnitude dimensions could be intertwined with the horizontality andverticality of our world representation

In their very attractive theory on navigation in a three-dimensionalworld Jeffery et al propose that our representations of theenvironment are based on cognitive maps which separately inte-grate horizontal and vertical dimensions of space Howeverknowledge of our environment is also built through our onlinemotor behaviours which improve the processing of other associ-ated dimensions to flexibly adapt future responses From this per-spective it has been suggested that a commonmetrics mechanismassociated with sensorimotor experience is in charge of processingdifferent magnitude dimensions such as space time andnumbers (Walsh 2003) My goal in this commentary is tosuggest that a cognitive map based only on three-dimensionalspaces might be an incomplete picture if abstract semantic dimen-sions are not fully considered Accordingly the representation ofnumber magnitudes can also interact with the representation ofthe horizontal and vertical dimensions of space during performedactions and contribute to a global map of the three-dimensionalworld representation

Evidence for a link among physical space actions and numberscomes from studies that show the association between smallnumbers with left and lower parts of space and large numberswith right and upper parts (Hubbard et al 2005 Umiltagrave et al2009) For instance Dehaene et al (1993) found that smallnumber processing can prime hand movement to the left partof space and large numbers can prime movement to the rightpart Schwarz and Keus (2004) revealed an association betweenthe vertical dimension and eye movements ndash that is downwardand upward saccades were initiated more quickly in response tosmall and large numbers respectively Interestingly Loetscheret al (2010) found that during a random number generationtask the leftward and downward adjustment of eye locations pre-dicted that the to-be-spoken number would be smaller than thelast one Conversely a prediction of large numbers was madethrough the right and upward location of eyes Altogether thesenumber-space associations occur in an automatic way and magni-tude representation could be represented as a number map thatintegrates spatial dimensions and actions (Schwarz amp Keus

2004 Walsh 2003) At a neurophysiological level number magni-tudes and three-dimensional Cartesian coordinates of externalspace are processed in a common parietal area (Hubbard et al2005) It is probable that daily life experience such as addingmore objects to a pile (for the vertical dimension) or culturalfactors such as the direction of writing (for the horizontal dimen-sion) could partially be in charge of the link between numbersand spatially directed actions (Gevers et al 2006) Anotherfactor could also come from the fact that during infancy countingstrategies often involve finger movements which in turn reinforcethe number-space association through sensorimotor experience(Butterworth 1999 Michaux et al 2010)

Based on sensorimotor accounts Walshrsquos ATOM (ldquoA Theory ofMagnituderdquo Walsh 2003) proposes that all magnitude dimensions(eg space time numbers and lengths) are processed inside acommon metrics mechanism located in the parietal cortex (Buetiamp Walsh 2009) Note that neurons in the parietal cortex of differ-ent animal species such as cats and macaque monkeys can alsorespond to number processing (Nieder amp Miller 2004 Thompsonet al 1970) Importantly the core assumption of ATOM is that werepresent space and time through environment-directed actionsAs stated by Walsh (2003) ldquothe inferior parietal cortex reflectsthe common need for space time and quantity information to beused in the sensorimotor transformations that are the main goalof these areas of cortexrdquo (p 483) It is worth noting that the parietalregion is also strongly involved in route-based navigation in pri-mates especially in the integration of self-movement information(Sato et al 2006) From an evolutionary viewpoint one maywonder why such common metrics mechanisms exist The mostprobable straightforward answer is that the brains of human andnonhuman animals are mainly shaped for anticipating upcomingevents in the environment (Corballis 2013 Hommel et al 2001Suddendorf amp Corballis 2007) On this view we have recentlyrevealed that the mere processing of number magnitudes is auto-matically associated with a general sensorimotor and anticipativemechanism (Badets et al 2013) In this theory distal referencesin the environment provide information of different magnitudesthat is simulated in an anticipative way to flexibly adapt futurenumerical- and motor-based responses (see Badets amp Pesenti[2011] for this ldquoanticipated-magnitude coderdquo)

Such anticipative mechanisms for sensorimotor adaptations arewell documented in the literature on motor control (Hommelet al 2001) For example envisage a person who wishes tomove a bottle of wine from the table to the upper part of thekitchen shelf During this motor sequence there is (1) the pro-cessing of a horizontal dimension for the reach-to-grasp move-ment of the hand towards the bottle on the table andsubsequently (2) the processing of a vertical dimension for theplacement of the bottle on the upper part of the shelf Accordingto Jeffery and colleagues both dimensions (ie horizontal andvertical) should be processed and represented separatelyHowever data on similar paradigms revealed that when thefinal placement was high on the shelf (vertical goal) the reach-to-grasp position of the hand (horizontal goal) was situated inthe lower part of the object (Cohen amp Rosenbaum 2004) Thisfinding indicates that the metric encoding the vertical goal is con-comitantly anticipated and accurately represented during enact-ment of the horizontal goal

In summary based on several lines of evidence that a commonmetrics mechanism located in the parietal region of the brain is incharge of the processing of space numbers and actions I proposethat different magnitude dimensions are most likely intertwinedwith the horizontality and verticality of space during environ-ment-directed actions Semantic knowledge such as themeaning of numbers is represented inside this common scaleand can refine the representation of physical space In otherwords semantic sides of three-dimensional space representationcould be anticipatorily activated in a sensorimotor mechanismwhich could give us the capacity to adapt different behavioursfor potential environmental constraints

CommentaryJeffery et al Navigating in a three-dimensional world

Human path navigation in a three-dimensionalworld

doi101017S0140525X13000319

Michael Barnett-Cowana and Heinrich H Buumllthoff bcaThe Brain and Mind Institute The University of Western Ontario LondonOntario N6A 5B7 Canada bDepartment of Human Perception Cognition andAction Max Planck Institute for Biological Cybernetics 72076 TuumlbingenGermany cDepartment of Brain and Cognitive Engineering Korea UniversitySeoul 136-713 Koreambarnettcowangmailcom hhbtuebingenmpgdewwwsitesgooglecomsitembarnettcowan wwwkybmpgde~hhb

Abstract Jeffery et al propose a non-uniform representation of three-dimensional space during navigation Fittingly we recently revealedasymmetries between horizontal and vertical path integration in humansWe agree that representing navigation in more than two dimensionsincreases computational load and suggest that tendencies to maintainupright head posture may help constrain computational processingwhile distorting neural representation of three-dimensional navigation

The target article is a well-written and timely paper that summar-izes much of the spatial cognition literature which has only recentlymade headway into understanding how animals navigate in a three-dimensional world Most of what is known about how the brains ofdifferent animals manage navigation has been constrained to thehorizontal plane Here Jeffery et al review experiments thatshow how insects fish rodents and other animals (includinghumans) navigate space which includes a vertical componentFrom these studies including some elegant work of their own onthe encoding of three-dimensional space by place and grid cells(Hayman et al 2011) the authors propose that 3D space is not uni-formly represented by the brain Rather they suggest that 3D spaceis represented in a quasi-planar fashion where spaces are con-strained to separate planes that are stitched into a non-Euclidianbut integrated map There is we think merit in this analysiswhich presents a reasonable and testable hypothesis one thataddresses the need to reduce the computational load required tofully represent navigation through 3D space while also requiringa mechanism to stitch representational spaces together

Given that the real world is three-dimensional it is somewhat sur-prising that investigations into three-dimensional navigation haveonly recently emerged One reason for this latency may be thatrepresenting movement in a volumetric space is not only computa-tionally complicated for the brain it poses additional constraints onexperimental design equipment and analysis than are required forhorizontal plane navigation After reviewing the literature onnonhu-man vertical navigation we became interested in human navigationwith a vertical component In a recent paper (Barnett-Cowanet al 2012) we showed for the first time that human path integrationoperates differently in all three dimensionsWehandled experimen-tal constraints by comparing performance in an angle completionpointing task after passive translational motion in the horizontalsagittal and frontal (coronal) planes To move participants in threedimensions we took advantage of the unique Max Planck Institute(MPI) CyberMotion Simulator (Teufel et al 2007) which is basedon an anthropomorphic robot arm and which offers a large motionrange to assess whether human path integration is similar betweenhorizontal and vertical planes We found that while humans tendto underestimate the angle throughwhich theymove in the horizon-tal plane (see also Loomis et al 1993) either no bias or an overesti-mate of movement angle is found in the sagittal and frontal planesrespectively Our results are generally in agreement with the theoryproposed by Jeffery et al for the representation of space being fixedto the plane of movement and our approach lends itself well totesting predictions that follow from this theory For example ifthe representation of space is fixed to the plane of movement andadditional processing is required to stitch planes together thenone would expect delays in response times to cognitive tasks as thenumber of planes moved through increases

As Jeffery et al point out the constant force of gravity pro-vides an allocentric reference for navigation However wewould like to clarify that gravity provides an allocentric referencedirection and not a reference frame A reference direction is fun-damental to perception and action and gravity is ideally suited asa reference direction because it is universally available to allorganisms on Earth for navigation and orientation Knowingonersquos orientation and the orientation of surrounding objects inrelation to gravity affects the ability to identify (Dyde et al2006) predict the behaviour of (Barnett-Cowan et al 2011)and interact with surrounding objects (McIntyre et al 2001)as well as to maintain postural stability (Kluzik et al 2005Wade amp Jones 1997) Tendencies to maintain upright headposture relative to gravity during self-motion and when interact-ing with objects have the functional advantage of constraining thecomputational processing that would otherwise be required tomaintain a coherent representation of self and object orientationwhen the eyes head and body are differently orientated relativeto one another and relative to an allocentric reference directionsuch as gravity Such righting behaviour is expressed in Figure 1where maintaining an upright head posture benefits gaze controlfor both fly and humanIn addition to studies of righting behaviour there are now

numerous studies which indicate that the brain constructs aninternal representation of the body with a prior assumption thatthe head is upright (Barnett-Cowan et al 2005 Dyde et al2006 MacNeilage et al 2007 Mittelstaedt 1983 Schwabe ampBlanke 2008) We speculate that the combination of rightingreflexes ndash to maintain an upright head posture during self-motion and optimize object recognition ndash along with priorassumptions of the head being upright may interfere with thebrainrsquos ability to represent three-dimensional navigation throughvolumetric space Further as vestibular and visual directional sen-sitivity are best with the head and body upright relative to gravity(MacNeilage et al 2010) future experiments aimed at disentan-gling how the brain represents three-dimensional navigationmust consider how the senses are tuned to work best whenupright

Figure 1 (Barnett-Cowan amp Buumllthoff) Optimization of gazecontrol in the blowfly Calliphora (left see Hengstenberg 1991)and in humans (right see Brodsky et al 2006) where uprighthead posture is maintained during self-motion Republishedfrom Hengstenberg (1991) with original photo copyright to MPIfor Biological Cybernetics (left) and Bike Magazine (August2001 p 71 right) with permissionNote The photo shown on the left above was taken at the MaxPlanck Institute for Biological Cybernetics which retains thecopyright Hengstenberg (1991) published part of thisphotograph as well In addition to holding the copyright for theoriginal photo we also have permission from the publisher ofHengstenbergrsquos article

Understanding how humans represent motion through volu-metric space is particularly important for assessing human aviationwhere loss of control in flight is todayrsquos major single cause of fatalaccidents in commercial aviation (see the EASA Annual SafetyReview European Aviation Safety Agency 2006) We suggest thatour understanding of three-dimensional navigation in humans canbe improved using advanced motion simulators ndash such as our MaxPlanck Institute CyberMotion Simulator which is capable ofmoving human observers through complex motion paths in avolume of space (An open access video is found in the originalarticle Barnett-Cowan et al 2012) Future experiments withfewer constraints including trajectories using additional degreesof freedom longer paths multiple planes and with the body differ-ently orientated relative to gravity will be able to more fully assessthe quasi-planar representation of space proposed by Jeffery et alas well as how vertical movement may be encoded differently

Learning landmarks and routes in multi-floored buildings

doi101017S0140525X13000320

Alain Berthoza and Guillaume ThibaultbaLaboratoire de Physiologie de la Perception et de lrsquoAction Collegravege deFrance 75231 Paris France bElectriciteacute de France Recherche ampDeacuteveloppement 92141 Clamart cedex Francealainberthozcolllege-de-francefrguillaumethibaultedffrhttpwwwlppacollege-de-francefrENhttpresearchedfcomresearch-and-innovation-44204html

Abstract Existing studies have proposed that humans preferentiallymemorize buildings as a collection of floors Yet this might stem fromthe fact that environments were also explored by floors We havestudied this potential bias with a learning and recognition experimentWe have detected a positive influence of the learning route ndash by floorsand also crucially by columns ndash on spatial memory performances

This commentary addresses the target articlersquos discussion ofldquonavigation in multilayer environmentsrdquo pace Jeffery et alrsquosreview of Montello and Pickrsquos (1993) study which ldquodemon-strated that human subjects who learned two separate andoverlapping routes in a multilevel building could sub-sequently integrate the relationship of these two routesHowever it was also established that pointing betweenthese vertically aligned spaces was less accurate andslower than were performances within a floorrdquo (sect 33para 5) By questioning the underlying methodology wepoint out how horizontal learning and also vertical learningof multifloored environments influence spatial memorywhich extends the findings reported by Jeffery et al

Existing studies on spatial memory in complex buildings(Buumlchner et al 2007 Houmllscher et al 2006 Montello amp Pick1993) have suggested that humans have a tendency tomemorize such environments as a collection of floors Yetthis might stem from the fact that environments werealso explored by floors of greater length than the heightof vertical junctions (staircases) In Thibault et al (2013)we have studied this potential bias with a learning and rec-ognition experiment for an object-in-place task undercarefully controlled conditions

First a virtual building was designed for the experimentas a repetition of rooms having the same height andlength connected by equivalent horizontal and verticalpassages Second a group of 28 participants ndash the floor lear-ners ndash learned the simulated space by watching a computer

movie depicting floor-by-floor travel similarly 28 other par-ticipants ndash the column learners ndash observed the buildingaccording to a column-by-column route (ie travelling byelevators across floors) One landmark was visible in eachvisited room of the building Third we equated the sequenceof landmarks encountered by floor-learning and column-learning participants (we changed the positions of the land-marks in the virtual building for each group to achieve this)We then observed the performances of the two groups in

a recognition phase where the participants viewed a cameramovement from one room to an adjacent one either fol-lowing a segment of the learned path (a familiar segment)or a shortcut (a novel segment) After being moved theyhad to indicate the correct object (among three distractors)that was there during the learning phaseUnder these controlled conditions we detected a posi-

tive influence of the learning route by floors and also bycolumns on spatial memory performances (in terms ofaccuracy and response times) We found that participantsprocessed familiar segments of the learned path more accu-rately than novel ones not only after floor learning (80correct answers in horizontal [familiar] trials vs 635 invertical [novel] trials) ndashwhich is in line with Montello andPickrsquos (1993) finding cited in section 33 of the targetarticle (see the first two sentences of sect 33 para 5) ndashbut crucially also after column learning Participantswho learned the multilevel building by a column-by-column route performed better in the recognition of land-marks within a column compared to the recognition oflandmarks across columns that is within a floor (66correct answers in vertical [familiar] trials vs 48 in hori-zontal [novel] trials)1 This finding suggests a key role ofthe learning mode on the exploitation of spatial memorywhich extends the findings reported by Jeffery et alAs a complement we here point out our own methodo-

logical limits1 Our work does not use a navigational task but an object-

in-place onewhichdiffers fromthe task consideredby Jefferyet al In our experiment the participants viewed a cameramovement ndash that is they were not free to change their direc-tion of movement and to update their navigation strategy2 Our experiment was conducted in a rectilinear

environment The participants cannot explore the (x y)plane but only the x-axis This can be a restricted casefor the bicoded map model proposed by the authors

NOTE1 The chance level was equal to 25 correct answers

Anisotropy and polarization of spaceEvidence from naiumlve optics andphenomenological psychophysics

doi101017S0140525X13000332

Ivana Bianchia and Marco BertaminibaDepartment of Humanities University of Macerata 62100 Macerata ItalybSchool of Psychology University of Liverpool Liverpool L69 7ZA UnitedKingdom

ivanabianchiunimcitmbertaminilivacukhttpdocentiunimcitdocentiivana-bianchihttpwwwlivacukvpmarcohtml

Abstract Additional evidence is presented concerning the anisotropybetween vertical and horizontal encoding which emerges from studiesof human perception and cognition of space in plane mirror reflectionsMoreover it is suggested that the non-metric characteristic ofpolarization ndash that Jeffery et al discuss with respect to gravity ndash is notlimited to the vertical dimension

As discussed by Jeffery et al evidence for true volumetric codingis weak and a bicoded map is the most likely encoding spatialscheme (at least in mammals) Moreover allocentric encodingof the horizontal dimension is metric whereas encoding of thevertical dimension is non-metric Among the features of the ver-tical dimension is its polarization In this commentary we focuson these two aspects (anisotropy and polarization) Jeffery et alexplicitly say that they confine their discussion to studies of navi-gation rather than spatial cognition However some parallels arestrong We draw attention to behavioral findings from naiumlveoptics and phenomenological psychophysics These data respect-ively (a) provide further evidence concerning the anisotropybetween vertical and horizontal encoding of space in humansand (b) suggest that polarization is a general characteristic ofnon-metric encoding of ecological space not exclusive to the ver-tical dimension

Many adults hold mistaken beliefs concerning the spatial behav-ior of reflections For example when adults imagine a personentering a room and walking parallel to a wall they predict thatthe person will see his or her reflection in a mirror on the wallbefore it is possible (ie before being aligned with the mirrorrsquosedge ndash this has been called the ldquoearly errorrdquo) A minority alsopredict that the image will appear at the mirrorrsquos farther edgerather than at the nearer edge The same mistake applies whenmoving in a real room not just when predicting an event in apaper-and-pencil task (Croucher et al 2002)

The early error is compatiblewith overestimation ofwhat is visiblein a mirror from left and right In other words people expect a coneof visibility for a givenmirror largely independent of the viewpoint ofthe observer and of distance (Bertamini et al 2010 Bianchi ampSavardi 2012b) The second error (the location of the image) is com-patible with the idea that the virtual world is spatially rotated arounda vertical axis (Bertamini et al 2003) However an alternative expla-nation was found in later studies The virtual space is thought of asallocentrically organized in an opposite way with respect to thereal environment (Savardi et al 2010)

These mistaken beliefs which are similar in males and femalesand in psychology and physics students are confined to the hori-zontal plane They disappear when vertical movements are con-sidered For instance if the imagined character approaches themirror by moving vertically ndash say climbing up a rope parallel tothe wall (Croucher et al 2002) or moving in a glass lift (Bertaminiet al 2003) ndash people neither expect to see the reflection beforereaching the near edge of the mirror nor to see it appearing atthe farther edge (ie opposite relative to the direction of theapproach) Experience does not lead to better understanding ofmirror reflections on the contrary adults make larger errors com-pared to children (Bertamini amp Wynne 2009)

Further evidence of a difference between transformationsaround horizontal and vertical axes comes from studies of percep-tion of mirror reflections of the human body Mirror-reflecting apicture or a real body around the horizontal axis ndash upside-downreversal ndash resulted in larger differences than did reflecting itaround the vertical axis ndash left-right reversal (Bianchi amp Savardi2008 Cornelis et al 2009)

With respect to polarization Jeffery et al argue that this is afeature typical of the dimension orthogonal to the plane of loco-motion (ie usually the vertical) They do not explicitly claimthat it is confined only to this dimension but they do notspecify that polarization can also characterize the representationof the horizontal dimension (at least in humans) Howeverspatial polarization emerges as a key aspect in both dimensionsfrom studies on the perception and cognition of space andobjects in mirrors (Bianchi amp Savardi 2008 2012b Savardi et al

2010) and it is a non-metric characteristic of direct experienceof space that goes beyond mirror perception This aspect ofspace has been emphasized in various anthropological (egBrown amp Levinson 1993) and linguistic (eg Cruse 1986 Hill1982) analyses of human cognition and in studies of memoryerrors in object recognition (Gregory amp McCloskey 2010) Onlyrecently has its importance been recognized within what hasbeen defined ldquophenomenological psychophysicsrdquo (Kubovy 2002Kubovy amp Gepshtein 2003) The basic structure of phenomenolo-gical experiences of space and its geometry invariably manifestsproperties of polarization (Bianchi et al 2011a 2011b Bianchiamp Savardi 2012a Savardi amp Bianchi 2009) In addition Stimu-lus-Response spatial compatibility can be thought of as a directeffect of an allocentrically polarized encoding of space inhumans (Boyer et al 2012 Chan amp Chan 2005 Roswarski ampProctor 1996 Umiltagrave amp Nicoletti 1990)Emerging from phenomenological psychophysics regarding

polarization is the operationalization in metric and topologicalterms of spatial dimensions and the finding that these dimensionsdiffer in the extent of the polarized area This in turn depends onthe extension of a central set of experiences perceived as neitherone pole nor the other For instance the perception of an objectas positioned ldquoon toprdquo of something (eg a flag on top of amountain)is confined to one specific position the same holds for ldquoat thebottomrdquo In between these two extremes there is a range of positionsthat are ldquoneither on top nor at the bottomrdquo In contrast in thedimension ldquoin front ofndashbehindrdquo (eg two runners) there is onlyone state that is perceived as ldquoneither in front nor behindrdquo (iewhen the two runners are alongside) Similarly in the dimensionldquoascending-descendingrdquo there is only one state that is perceived asldquoneither ascending nor descendingrdquo (ie being level) all the otherslants are recognized as gradations of ascending or of descending(Bianchi et al 2011b Bianchi et al 2013) Furthermore perhapscounter-intuitively polarization is not to be thought of necessarilyas a continuum Ratings of highlow largesmall widenarrow andlongshort (dimensions relevant for navigation) applied to ecologicalobjects or to spatial extensions do not behave as inverse measure-ments of the same continuum (Bianchi et al 2011a)

Navigating in a volumetric world Metricencoding in the vertical axis of space

doi101017S0140525X13000344

Theresa Burt de Perera Robert Holbrook Victoria DavisAlex Kacelnik and Tim GuilfordDepartment of Zoology University of Oxford Oxford OX1 3PS UnitedKingdom

TheresaBurtzoooxacuk RobertHolbrookzoooxacukVictoriaDaviszoooxacuk AlexKacelnikzoooxacukTimGuilfordzoooxacukhttpoxnavzoooxacuk

Abstract Animals navigate through three-dimensional environments butwe argue that the way they encode three-dimensional spatial information isshaped by how they use the vertical component of space We agree withJeffery et al that the representation of three-dimensional space invertebrates is probably bicoded (with separation of the plane oflocomotion and its orthogonal axis) but we believe that their suggestionthat the vertical axis is stored ldquocontextuallyrdquo (that is not containingdistance or direction metrics usable for novel computations) is unlikelyand as yet unsupported We describe potential experimental protocolsthat could clarify these differences in opinion empirically

Following behavioural and neurophysiological research on howthree-dimensional space is encoded by animals Jeffery et alhypothesize that vertebrates represent space in a bicodedfashion with spatial information about the plane of locomotioncomputed and represented separately from space in the

orthogonal axis We agree with this Indeed experimental evi-dence shows that fish that move freely through a volume representspace in two separable vertical and horizontal components andnot as an integrated three-dimensional unitary representation(Holbrook amp Burt de Perera 2009 2011b) This is supported bystudies on rats (Grobeacutety amp Schenk 1992a Hayman et al 2011Jovalekic et al 2011) However this still leaves open the funda-mental question of whether the vertical axis is represented ldquocon-textuallyrdquo (defined as not available for novel quantitativeprocessing) or metrically

Jeffery et al are ambiguous on this point as their bicodinghypothesis can accommodate the vertical axis being representedmetrically but they favour the idea that it is likely to beencoded contextually even for animals that move freely throughthree dimensions They reasonably argue that contextual codingmay free the animalrsquos brain from complex trigonometric calcu-lations in three dimensions but this comes at the cost of constrain-ing spatial computations For instance metric encoding allows theanimal to know when it is getting close is roughly where it shouldbe or has gone too far and it allows the animal to use the relationbetween multiple landmarks correcting for distance and perspec-tive when computing novel trajectories It has been argued that inthe horizontal plane ants integrate cross-modally metric infor-mation obtained from path integration with visual snapshots oflandmarks (Lent et al 2010 Mueller amp Wehner 2010) Foranimals that move freely in a volume the same computationalpossibilities are important in the vertical axis

Because metric information contains spatial relationships thepresence of metric representations can be expressed in howthe pattern of error or generalization is distributed along the ver-tical axis We offer two suggestions First we expect generaliz-ation along a metric vertical axis either side of significantlocations and for this ldquoerrorrdquo to obey Weberrsquos law displaying aroughly proportional relation between magnitude of thepercept and accuracy (Foley amp Matlin 2010) Such generalizationwould not apply to contextual representations This could betested by measuring error in recall of a rewarded vertical positionfor an animal that can move freely in the vertical axis (and henceexpress that error) Second generalization curves around a posi-tive and a negative instance should interfere if the instances areclose in a metric vertical axis and this interference should takethe behavioural form of a computable peak shift (Domjan2009) So if an animal is rewarded for visiting a place at acertain elevation the presence of a punishing instance marginallyabove this should result in a predictable downwards displace-ment of response

Whether an animal encodes the vertical axis metrically or not maybe related to the animalrsquos movement and whether it is surface boundwith three degrees of freedom (forward-backward left-right androtational (yaw) or moves with six degrees of freedom by theaddition of up-down roll and pitch In flying or swimminganimals the equality in the freedom of movement in the horizontaland vertical components of space is likely to favour metrical scalesfor space in both the horizontal and vertical dimensions

Navigating through a volumetric world doesnot imply needing a full three-dimensionalrepresentation

doi101017S0140525X13000356

Claus-Christian Carbon and Vera M HesslingerDepartment of General Psychology and Methodology and Graduate School ofAffective and Cognitive Sciences University of Bamberg D-96047 BambergBavaria Germany

cccexperimental-psychologycom verahesslingeruni-bambergdewwwexperimental-psychologycom

Abstract Jeffery et al extensively and thoroughly describe how differentspecies navigate through a three-dimensional environment Undeniablythe world offers numerous three-dimensional opportunities Howeverwe argue that for most navigation tasks a two-dimensional representationis nevertheless sufficient as physical conditions and limitations such asgravity thermoclines or layers of earth encountered in a specificsituation provide the very elevation data the navigating individual needs

As Jeffery et al correctly note most scientific efforts on large-scale spatial relations have focused on two-dimensional settingswhile neglecting further potentially important dimensions suchas elevation slant and distortion In theoretical terms generatinga more complete three-dimensional representation of theenvironment by integrating such information presumablyenhances navigation accuracy However it is rather debatablewhether this also leads to a significant improvement in actual navi-gation and localization performance in everyday tasks (Montello ampPick 1993)

As a series of empirical works confronting (human) participantswith navigation tasks has documented specific deficits in theassessing of the azimuth angle for instance arise in multi-flooredthree-dimensional versus single-flooredtwo-dimensionalsettings (Houmllscher et al 2006 Thibault et al 2013) In ecologicalcontexts offering a variety of orientation cues humans are never-theless able to actively navigate through three-dimensionalenvironments without any problems This might again indicatehere that it is not obligatory to have access to a perfect cognitivethree-dimensional representation of the environment Further-more the mismatch of evidence provided by empirical studiesand everyday experience might point to a lack of ecological val-idity in the paradigms commonly used to investigate the premisesof actual navigation This is partly due to laboratory and real-lifenavigation tasks requiring completely different and sometimeseven converse strategies or behaviors In a study by Carbon(2007) for example participants were asked to estimate nationallarge-scale distances as the crow flies ndash but what did they do inthe end Although they obviously used a consistent and steadystrategy as indicated by estimates being highly reliable as wellas strongly correlated with the factual physical distances (cfMontello 1991) they applied a strategy which differed entirelyfrom the one specified in the instructions Instead of linear dis-tances they used German Autobahn distances as the basis fortheir estimations thus replacing the requested but unfamiliarmode (no human was ever found to behave like a bird) withone derived from everyday behavior ndash that is from actually tra-velling these distances by car (instead of aircraft) Thus everydayknowledge was found to be preferred over (artificial) task affor-dances which is indeed reasonable since it is easier and moreeconomical to do something on the basis of knowledge and fam-iliar routines

Letrsquos get back to a point mentioned already and elaborate onwhy a complete three-dimensional representation of the environ-ment is not obligatory for the majority of typical real-life naviga-tion tasks Surface-travelling species for example are limitedto the surface they are travelling on they might hop and digfrom time to time but they mainly orient themselves to thesurface thus inherently to the current elevation of the given struc-ture of this surface Basically they navigate on an idealized planeWhen directions of places are to be assessed within a single planeazimuthal errors are relatively small (Montello amp Pick 1993) sonavigation will be quite accurate If sensory (eg visual) cuesare additionally taken into account it can be further tuned andoptimized (Foo et al 2005) Concerning navigation throughthree-dimensional environments the ability of extracting and uti-lizing supplemental information provided by external or sensorycues turns out to be quite economic Even a subject climbing amountain can still locate its target destination (the peak) on anidealized two-dimensional map provided some supplementalinformation on the elevation of this target is available This infor-mation can be ldquogleanedrdquo for instance from the required expendi-ture of energy while climbing Considering that most parts of the

three-dimensional space can thus be reduced to a surface-maprepresentation with sparse data requirements a cognitivesystem encoding topographies in full three-dimensional coordi-nates seems rather unnecessary as it would be too cost-intensive

Regarding species such as flying insects birds or fish that movemore freely through the third dimension very similar navigationroutines can be found (eg for honeybees Lehrer 1994)Research has indeed revealed specific skills in communicatingelevation (eg for fish Holbrook amp Burt de Perera 2009 egfor stingless bees Nieh amp Roubik 1998) and that elevation infor-mation can be highly relevant in some tasks (eg finding a birdrsquosnest) Still it is improbable that this information is fully integratedwithin a complete cognitive three-dimensional representationFrom an information theory perspective most parts of volumetricrepresentations of real-world contexts would comprise a greatnumber of ldquoempty cellsrdquo Furthermore reliable locations canhardly be imagined without any physical connection to thesurface A birdrsquos nest for example may be situated in a treetopthat is part of a tree that is itself solidly enrooted in the ground(ie the surface) Navigation requirements in the water wherespatial constraints are also obvious are similar Most relevantand reliable locations for hiding or for finding prey are near thebottom or the surface of the sea For navigating through thesea elevation information might be needed but not necessarilyin the form of a complete three-dimensional representationGibsonrsquos (1979) ecological approach offers a solid basis for redu-cing data for navigating tasks quite efficiently Moving through athree-dimensional world itself provides important directly visualacoustic and proprioceptive cues (cf Allen 1999) which help usto assess distances elevations and drifts of our movement trajec-tories both easily and accurately

Think local act global How do fragmentedrepresentations of space allow seamlessnavigation

doi101017S0140525X13000368

Paul A Dudchenkoab Emma R Woodb and RoderickM GrievesabaSchool of Natural Sciences University of Stirling Stirling FK9 4LA UnitedKingdom bCentre for Cognitive and Neural Systems University of EdinburghEdinburgh EH8 9JZ United Kingdompadudchenkostiracuk emmawoodedacuk rmg3stiracukwwwmemoryspacemvmedacuk

Abstract In this commentary we highlight a difficulty for metricnavigation arising from recent data with grid and place cells theintegration of piecemeal representations of space in environments withrepeated boundaries Put simply it is unclear how place and grid cellsmight provide a global representation of distance when their fieldsappear to represent repeated boundaries within an environmentOne implication of this is that the capacity for spatial inferences may belimited

The Appalachian Trail is a 2184-mile-long footpath through themountains of the eastern United States Each year approximately2000 ldquothru-hikersrdquo attempt to backpack the trailrsquos entire lengthfrom Georgia to Maine or vice versa There are maps for eachsection of the trail and hikers quickly learn that the most salientfeature of these is the elevation profile Miles are secondary toelevation change When carrying a backpack the amount ofelevation gain is of paramount interest in planning the dayrsquos pro-gress Elevation has also been formalised in William Naismithrsquosrule-of-thumb for estimating pedestrian distance On flatground walkers usually cover 3 miles per hour with an additionalhalf hour added for each 1000-foot elevation gain

The consideration of three-dimensional spatial mapping byJeffery et al rightly brings attention to what thru-hikers and

hill-walkers know firsthand ndash that elevation matters But are dis-tance and elevation represented in the same way or in differentways Jeffery et al based on a review of behavioural and neuro-physiological data argue that three-dimensional space is rep-resented in a bicoded (anistropic) internal map Distances arerepresented metrically ndash likely by grid cells ndashwhereas elevation isrepresented qualitatively perhaps as a contextual cue As ananimal moves up down and across a complex environment itmay represent each surface as a map fragment and these arelinked in some wayThere is however a difficulty with map fragmentation and it is

not just in identifying how maps of different planes are linkedRather it is with the representations of space provided by gridplace and to a lesser extent perhaps head direction cells them-selves Put simply while in an open space grid cells may encodedistances in more complex environments with similar repeatedgeometric features grid and place cells exhibit a fragmentedencoding of space Hence the metric representation of spaceitself appears to be piecemeal As we consider below this mayexplain why demonstrations of novel spatial inference in the ratare a challengeFor hippocampal place cells accumulating evidence suggests

that place fields are controlled by the local features of the environ-ment For example Skaggs and McNaughton (1998) found thatsome place cells showed similar fields in each of two identicalboxes This result was replicated by Fuhs et al (2005) in theirsame box orientation condition and repeated fields have beenfound in up to four adjacent boxes (Spiers et al 2009) These find-ings and the demonstration that some place fields duplicate whena barrier is inserted into a square environment (Barry et al 2006)provide clear support for the hypothesis that place fields aredriven by the boundaries of the immediate environment(Hartley et al 2000 OrsquoKeefe amp Burgess 1996) Thus an animaltraversing an environment with repeated geometric featuresmay experience a repeated fragmented place cell representationIndeed the repeated place cell firing in the spiral maze observedby Hayman et al (2011) may reflect this in the third dimensionFragmentary firing also occurs in medial entorhinal cortex grid

cells In an elegant experiment Derdikman et al (2009) showedthat grid fields fragmented in a hairpin maze in which theanimals ran in alternate directions through a series of alleywaysThe firing of these cells appeared anchored to the start of eachturn in a given direction for much of the distance in an alleywayalthough firing near the end of the alley was anchored to theupcoming turn In this same study hippocampal place cells like-wise exhibited repeated fields in different alleyways when therat was running in the same direction Thus both grid cells andplace cells exhibited a local not global representation of theenvironment Presumably this occurred even though the animaltreated the environment as a wholeWhat then of head direction cells Here the situation is differ-

ent In the same hairpin maze head direction cells wereunchanged (ie they tended to fire in the same direction) com-pared to recording sessions in an open field (Whitlock amp Derdik-man 2012) Remarkably in cells that showed grid fields anddirectional tuning (conjunctive cells) grid fields fragmented inthe hairpin maze while directional firing was unaffected The con-sistency of directional firing across maze compartments is inagreement with previous demonstrations that the preferredfiring of head direction cells is maintained with some erroracross environments when the animal walks between them (Dud-chenko amp Zinyuk 2005 Stackman et al 2003 Taube amp Burton1995)On the face of the existing place and grid cell evidence it is

unclear how rodents solve distance problems in environmentswith repeated boundaries This raises two possibilities First thedistance map may become more global with repeated experiencethough there is not strong evidence for this yet (Barry et al 2006Spiers et al 2009) Second it may be that rodents are not all thatgood at navigating multiple enclosure environments For example

we have found that rats are surprisingly limited in their ability totake a novel shortcut between familiar environments (Grieves ampDudchenko 2013) In short their representations of the worldmay be local and only incidentally global

To return to Jeffery et alrsquos main argument The firing of gridand place cells in three-dimensional mazes is certainly consistentwith a bicoded representation of space Challenges that emergefrom this view and from the findings reviewed above includeidentifying how global distance is represented when there is achange in elevation and how it is represented when the environ-ment contains repeated geometric features

Perceptual experience as a bridge betweenthe retina and a bicoded cognitive map

doi101017S0140525X1300037X

Frank H Durgin and Zhi LiDepartment of Psychology Swarthmore College Swarthmore PA 19081

fdurgin1swarthmoreedu zhilishgmailcomhttpwwwswarthmoreeduSocScifdurgin1publicationshtml

Abstract The bicoded cognitive maps described by Jeffery et al arecompared to metric perceptual representations Systematic biases inperceptual experience of egocentric distance height and surfaceorientation may reflect information processing choices to retaininformation critical for immediate action (Durgin et al 2010a)Different information processing goals (route planning vs immediateaction) require different metric information

Jeffery et al propose that the cognitive maps used by vertebratesare bicoded rather than fully volumetric insofar as they representa two-dimensional metric map in combination with a non-metricrepresentation of the third dimension In some cases the two-dimensional metric surface might be a vertical surface Addition-ally the maps are piecemeal This is an important hypothesis thatwill help to focus future research in the field but it raises ques-tions concerning the relationship between metric representationsin perception and those in cognition Our commentary willpropose possible links between metric perceptual experienceand bicoded cognitive maps

The visual world is normally understood as being sensedthrough a two-dimensional medium (eg the retina paceGibson 1979) that provides direct (angular) access to the verticaldimension and to the portion of the horizontal that is frontal tothe observer In contrast the information most relevant forroute planning might be said to be roughly along the line ofsight itself ndash the horizontal depth axis or the ground planewhich is highly compressed on the retina A large number ofvisual (and non-visual) sources of information may be used tolocate things in depth but foremost among these for locomotionis information about angle of regard with respect to the groundsurface (whether this is horizontal slanted or vertical) and eye-height

Relating the two-dimensional frontal retinal image to a cog-nitive metric surface map is an important achievement Theinformation evident to perceptual experience about surfacelayout in locomotor space includes information about surfaceinclination (ie slant relative to the gravitationally specifiedhorizontal plane) as well as information about surface extentand direction Over the past three decades several competingclaims about surface layout perception have been advancedincluding affine models of depth-axis recovery (Wagner1985) well-calibrated performance at egocentric distancetasks (Loomis et al 1992) energy-based models of slant per-ception (Proffitt et al 1995) intrinsic bias models of distance(Ooi amp He 2007) and our own angular scale expansionmodel of slant and distance (Durgin amp Li 2011 Hajnalet al 2011 Li amp Durgin 2012)

The common observations that many of these models seek toaddress include perceptual facts that are superficially inconsistentwith accurate metric spatial representations Distances are under-estimated (Foley et al 2004 Kelly et al 2004) surface slant rela-tive to horizontal is overestimated (Li amp Durgin 2010 Proffittet al 1995) and object height is consistently overestimated rela-tive to egocentric distance (Higashiyama amp Ueyama 1988 Liet al 2011) Alongside these facts are the observations thatangular variables such as angular declination (or gaze declination)relative to visible or implied horizons seem to control both per-ceptual experience and motor action (Loomis amp Philbeck 1999Messing amp Durgin 2005 Ooi et al 2001) emphasizing the impor-tance of angular variables in the perceptual computation ofegocentric distance

Because the perceptual input to vision is specified in a polarcoordinate system (the retinal ldquoimagerdquo and its polar transform-ations afforded by eye head and body movements) a basicissue for relating the bicoded maps proposed by Jeffery et al tothe perceptual input is to understand the transformation fromspherical coordinates to metric space One argument that wehave made in various forms is that metric spatial coding of dis-tance might be incorrectly scaled in perception without producinga cost for action Because our actions occur in the same perceptualspace as our other perceptions (ie we see our actions Powers1973) they can be calibrated to whatever scaling we perceiveMetric representation is important to successful action (Loomiset al 1992) but perceived egocentric distance can be mis-scaledas long as the scaling is stable and consistent

We have observed that angular variables (including slant) seemto be coded in a way that exaggerates deviations from horizontalthis scaling could have the felicitous informational consequencesof retaining greater coding precision while producing perceptualunderestimation of egocentric ground distance (Durgin amp Li2011) to which locomotor action can nonetheless be calibrated(Rieser et al 1995) Believing that the resulting cognitive mapsare metric does not require that perception also be metric inthe same way (there is substantial evidence that egocentric hori-zontal extents appear shorter than frontal horizontal extentssee eg Li et al 2013) though it suggests that anisotropies in per-ceptual experience are overcome in cognitive maps However thenotion that the vertical dimension is coded separately is intriguingand likely reflects a later transformation

The perception of surface slant appears to require integrationof vertical and horizontal metrics and the systematic distortionsevident in surface slant cannot be explained in detail merely byassuming that vertical scaling is expanded relative to horizontalscaling Rather the most elegant quantitative model of slant per-ception available suggests that perceived slant may represent theproportion of surface extent that is vertical (ie the sine of actualslant see Durgin amp Li 2012) Importantly the misperception ofsurface orientation shows constancy across changes in viewingdirection (eg Durgin et al 2010b) that implies that thecoding of slant in perceptual experience is with respect to a grav-itational (rather than an egocentric) reference frame even whenno horizontal ground surface is visible in the scene It wouldtherefore be of interest based on the suggestion that the two-dimensional cognitive map may sometimes be vertical to deter-mine how a vertical planar reference frame (eg a vertical navi-gation surface) affects the human interpretation of relativesurface orientation

The transformations between retinal input and cognitive mapsprobably involve an intermediate stage of perceptual represen-tations of space that resembles neither This perceptual stageclearly overcomes the superficial limitations of retinal imagesyet it integrates vertical and horizontal (in the form of perceivedsurface orientation) in a way that cognitive maps may not Theselection of a two-dimensional metric representation is probablyhighly determined by information processing constraints inspatial cognition Such ideas are also at the heart of scaleexpansion theory

Learning to navigate in a three-dimensionalworld From bees to primates

doi101017S0140525X13000381

Adrian G Dyerab and Marcello G P RosabaSchool of Media and Communication RMIT University Melbourne VIC 3001Australia bDepartment of Physiology Monash University Clayton VIC 3800Australiaadriandyerrmiteduau marcellorosamonasheduhttpwwwrmiteduaustaffadrian_dyerhttpwwwmedmonasheduauphysiologystaffrosahtml

Abstract We discuss the idea that environmental factors influence theneural mechanisms that evolved to enable navigation and propose that acapacity to learn different spatial relationship rules through experiencemay contribute to bicoded processing Recent experiments show thatfree-flying bees can learn abstract spatial relationships and we proposethat this could be combined with optic flow processing to enable three-dimensional navigation

Experiments that have investigated navigation in insects add con-siderably to the generality of the arguments presented by Jefferyet al in revealing additional evidence for potential bicoded pro-cessing in animal brains A contemporary point of interest inneuroscience is whether solving complex cognitive problems actu-ally requires large mammalian brains (Chittka amp Niven 2009) Inthis regard free-flying bees are an interesting model for under-standing how visual systems deal with navigation in a three-dimen-sional world

It is known that the mechanism by which both stingless bees(Melipona panamica) and honeybees (Apis mellifera) judge dis-tance in the horizontal plane is visually driven by optical flowwhere the velocity of angular image motion is integrated overtime to enable a bee to estimate the flown distance with accuracy(Dacke amp Srinivasan 2007 Eckles et al 2012) Interestingly whenhoneybees are presented with the equivalent type of visualproblem requiring judgment of distance using optic flow in thevertical plane they solve this task with a much lower level of pre-cision (Dacke amp Srinivasan 2007) We argue that the relative accu-racy of navigation in the horizontal and vertical dimensions makessense in terms of normal ecological imperatives and that brainplasticity can be recruited to facilitate robust bicoded processingif required

Studies on honeybees suggest that visual processing of opticflow is dominated by information impinging on the ventralvisual field (Dacke amp Srinivasan 2007) which makes ecologicalsense for an animal that predominantly flies in the horizontalplane when foraging for nutrition In contrast the stingless beeoperates in dense tropical environments where flower resourcesare typically scattered horizontally and throughout a range of ver-tical heights of up to 40 metres in the forest canopy These sting-less bees have been shown to be equally proficient at gaugingdistance in both the horizontal and vertical planes potentiallyusing optic flow mechanisms in both ventral and lateral visualfields (Eckles et al 2012) Nonetheless honeybees do sometimesalso have to operate in complex three-dimensional environmentssuch as tropical forests How might they deal with the complexityof reliable orientation in the vertical plane We suggest below thathoneybees are able to use information other than optic flow andthat their brains can learn to combine different visual perceptionsto solve novel problems in a manner consistent with the bicodedhypothesis

Whilst honeybees do communicate horizontal direction and dis-tance to hive mates through a symbolic dance communicationlanguage this does not reliably communicate elevation (Dackeamp Srinivasan 2007) It is therefore likely that individual foragersmust learn through their own experience to determine verticalcomponents of their three-dimensional world One possible sol-ution to this problem is relationship learning ndash estimating verticalposition by understanding the relative relationships such as

abovebelow between different elemental features in the environ-ment The capacity to process such relationship rules is not innatefor a brain Three-months-old human babies do not show evi-dence of relationship processing although by 6 months of agethis capacity has developed (Quinn et al 2002) Other adult pri-mates such as capuchin monkeys (Cebus apella) also show acapacity for solving problems requiring the learning of abovebelow rules (Spinozzi et al 2004) Recent work in honeybeesshows that while their brains do not innately code spatial relation-ship rules individual free-flying bees can learn such relationshipsthrough visual experience including the reliable processing ofabovebelow (Avarguegraves-Weber et al 2012)The relationship rule processing mechanism observed in hon-

eybees would thus give the potential for experienced individualsto forage in complex environments with acquired knowledgeabout relative vertical positions between biologically plausibleobjects such as flowers and tree limbs (Chittka amp Jensen 2011Dyer et al 2008) Could ldquovertical knowledgerdquo then be combinedwith the perception of optic flow in the horizontal plane to give atrue bicoded perception of three-dimensional space We believeso The work on how honeybees process complex visual relation-ship rules including abovebelow and samedifferent suggeststhat there is also a capacity to learn simultaneously two separaterules or types of visual information and then combine thisacquired knowledge to solve novel problems (Avarguegraves-Weberet al 2012) Thus although the honeybee appears to process hori-zontal and vertical optic flows as separate signals (Dacke amp Srini-vasan 2007) it does appear that their brains have the capacity tocombine multiple sources of sensory perception and otherspatial cues to make novel decisions in complex environmentsIn summary learning to build concepts to process multiple

dimensions of three-dimensional navigation using a bicodedsystem as hypothesised by Jeffery et al may represent a moregeneral biological capacity which likely extends to relativelysimple brains such as those of insects Further we suggest thatthe opportunity for a brain of an individual to learn complextasks through experience is critical to revealing the true behav-ioural capabilities in animals irrespective of the relative numberof neurons and synapses involved

Spatial language as a window onrepresentations of three-dimensional space

doi101017S0140525X13000393

Kevin J Holmesa and Phillip Wolff baDepartment of Linguistics University of CaliforniandashBerkeley Berkeley CA94720 bDepartment of Psychology Emory University Atlanta GA 30322

kjholmesberkeleyedu pwolffemoryeduhttpuserwwwserviceemoryedu~kholme2httpuserwwwserviceemoryedu~pwolffCLSLabhtm

Abstract Recent research investigating the languagendashthought interface inthe spatial domain points to representations of the horizontal and verticaldimensions that closely resemble those posited by Jeffery et al Howeverthe findings suggest that such representations rather than being tied tonavigation may instead reflect more general properties of theperception of space

Jeffery et al propose that bicoded representations may supportthe encoding of three-dimensional space in a wide range ofspecies including nonndashsurface-travelling animals Only humanshowever have the ability to draw on their representations ofspace to talk about their spatial experience Here we highlightthe potential of spatial language ndash not traditionally consideredin the study of navigation through space ndash to provide insight intothe nature of nonlinguistic spatial representation In particularwe suggest that recent research on spatial language spatial

cognition and the relationship between the two offers an unex-pected source of evidence albeit indirect for the kinds of rep-resentations posited by Jeffery et al Such evidence raises thepossibility that bicoded representations may support spatial cogni-tion even beyond navigational contexts

There are several striking parallels between Jeffery et alrsquosaccount of spatial representation and the semantics of spatiallanguage Jeffery et al argue that animals represent the verticaldimension in a qualitatively different manner than they do thetwo horizontal dimensions in large part due to differences in loco-motive experience This distinction between vertical and horizon-tal is also evident in spatial language Clark (1973) noted thatEnglish spatial terms rely on three primary planes of referenceone defined by ground level (dividing above from below) andthe other two defined by canonical body orientation (dividingfront from back and left from right) In a similar vein Landauand Jackendoff (1993) pointed to axial structure (eg the verticaland horizontal axes) as a key property encoded by spatial preposi-tions which otherwise tend to omit much perceptual detail (seealso Holmes amp Wolff 2013a) More recently Holmes (2012Holmes amp Wolff 2013b) examined the semantic organization ofthe spatial domain by asking native English speakers to sort a com-prehensive inventory of spatial prepositions into groups based onthe similarity of their meanings Using several dimensionalityreduction methods to analyze the sorting data including multidi-mensional scaling and hierarchical clustering Holmes found thatthe first major cut of the domain was between vertical terms (egabove below on top of under) and all other prepositions termsreferring to the left-right and front-back axes (eg to the left ofto the right of in front of behind) tended to cluster togetherThese findings suggest that the vertical-horizontal distinctionmay be semantically and perhaps conceptually privileged

Holmesrsquos (2012) findings are also consistent with Jeffery et alrsquosclaims about the properties that distinguish horizontal from verti-cal representations Jeffery et al propose that horizontal represen-tations are relatively fine-grained whereas vertical representationsare coarser and nonmetric in nature In Holmesrsquos study preposi-tions encoding distance information (eg near far from) clus-tered exclusively with horizontal terms implying that metricproperties are more associated with the horizontal dimensionsthan the vertical Further vertical terms divided into discrete sub-categories of ldquoaboverdquo and ldquobelowrdquo relations but horizontal termsdid not English speakers regarded to the left of and to the right ofas essentially equivalent in meaning Perhaps most intriguinglyHolmes found that the semantic differences among the dimen-sions were mirrored by corresponding differences in how spatialrelations are processed in nonlinguistic contexts (see also Franklinamp Tversky 1990) When presented with visual stimuli depictingspatial relations between objects (eg a bird above below tothe left of or to the right of an airplane) participants werefaster to discriminate an ldquoaboverdquo relation from a ldquobelowrdquo relationthan two different exemplars of an ldquoaboverdquo relation ndash but only inthe right visual field consistent with the view that the left hemi-sphere is specialized for categorical processing (Kosslyn et al1989) Though observed for vertical relations this effect of later-alized categorical perception demonstrated previously for color(Gilbert et al 2006) and objects (Holmes amp Wolff 2012) wasentirely absent in the case of horizontal relations Participantswere just as fast to discriminate two different exemplars of ldquoleftrdquoas they were to discriminate ldquoleftrdquo from ldquorightrdquo That the verticaldimension was perceived categorically but the horizontal dimen-sion was not suggests differences in how the mind carves upspatial information along different axes In characterizing thenature of the bicoded cognitive map Jeffery et al use colormerely as a way of illustrating the nonmetric property of verticalrepresentations but such an analogy seems particularly fittingWhereas the vertical axis may be represented in the same waythat we see a rainbow as forming discrete units the horizontalaxis may be represented more like color actually presents itselfnamely as a continuous gradient

Together the findings reviewed above tell a story about spatialrepresentation that is in many respects similar to that proposedby Jeffery et al in the target article However such findingssuggest an alternative explanation for the many differencesobserved between the horizontal and vertical dimensions Earlyin the target article Jeffery et al briefly distinguish betweenspatial navigation and spatial perception more generally implyingthat the representations supporting navigation may not extend toother spatial contexts But given the parallels between spatiallanguage and the representations implicated by Jeffery et alrsquosaccount and the fact that spatial terms often refer to staticspatial configurations rather than motion through space bicodedrepresentations may constitute a more general property ofspatial perception rather than being specifically tied to navigationThis possibility could be examined in future research on the rep-resentation of three-dimensional space More broadly our obser-vations suggest a role for research on the languagendashthoughtinterface in informing accounts of cognitive abilities ostensiblyunrelated to language lending support to the enduring maximthat language is a window into the mind (Pinker 2007)

Multi-floor buildings and human wayfindingcognition

doi101017S0140525X1300040X

Christoph Houmllscher Simon Buumlchner and Gerhard StrubeCenter for Cognitive Science amp SFBTR8 Spatial Cognition University ofFreiburg 79098 Freiburg Germanyhoelschcognitionuni-freiburgde simonbuechnercognitionuni-freiburgde strubecognitionuni-freiburgdehttpportaluni-freiburgdecognitionmembershoelscherhttpportaluni-freiburgdecognitionmembersbuechnerhttpportaluni-freiburgdecognitionmembersstrube

Abstract Multilevel wayfinding research in environmental psychologyand architecture exhibits a strong compatibility with Jeffery et alrsquosldquobicodedrdquo representation of space We identify a need for capturingverticality in spatial analysis techniques such as space syntax and arguefor investigating inter-individual differences in the ability to mentallyintegrate the cognitive maps of separate floors in buildings

Our commentary focuses on navigating multilayer environmentsand extends Jeffery et alrsquos view to an architectural and environ-mental psychology perspective

The functions of buildings are generally organized horizontallyprobably reflecting the constraints that humans encounter A hori-zontal plane is neutral to the axis of gravity and allows for stablewalking sitting and storing of objects Humans and buildingsldquoinhabitrdquo the same ldquotwo-dimensionalrdquo ecological niche and build-ings stack floors on top of one another As a consequence thestructure of typical buildings is highly compatible with theldquobicodedrdquo representation Whereas the horizontal plane is con-tinuous (albeit subdivided by corridors and partitions) and inline with the floors the vertical axis is discontinuous and discre-tized that is floors are on top of one another with only local con-nections via stairs or elevators Unless one has a view along amulti-storey atrium the vertical dimension is visually limited tothe current or directly adjacent floor Verticality is presented asldquocontextualrdquo at ordinal rather than metric scale and perceivedindirectly or derived by inference processes

Tlauka et al (2007) describe a systematic bias in vertical point-ing between floors Across several studies in our group we haveobserved a pattern in the process of pointing that links to thebicoded representation Pointing appears to be based on categori-cal and discretized rather than continuous information visible insmooth horizontal pointing and stepwise vertical pointing Partici-pants report counting floors rather than making spontaneous

judgments (Houmllscher et al 2009) This is further illustrated bylonger response times for the vertical component of the pointinggesture

Environmental researchers such as Weisman (1981) andCarlson et al (2010) differentiate between four types of environ-mental characteristics that impact navigation in buildings (1)visual access to decision points (2) architectural differentiation(the likelihood of confusing visual scenes) (3) layout geometryand (4) signage These features can be seen in the light of thebicoded representation with vertical integration as the centralchallenge Several studies (eg Soeda et al 1997) illustrate thatmultilevel public buildings are inherently difficult to navigateand wayfinding complexity can be traced to mentally linking indi-vidual floors of the building Participants expect a consistent cor-ridor layout across floors and are disoriented by layout differencesbetween floors Buildings in which the floors are partially horizon-tally offset from one another such as the Seattle Public Library(by architect Rem Koolhaas Carlson et al 2010) provide anadditional challenge for the mental alignment of correspondinglocations across floors

Vertical connections form salient elements in the mental rep-resentation of the building and are usually limited in numberLevel changes are a key source of disorientation about onersquosheading and position in a building (Houmllscher et al 2006)especially when a staircase is offset from the main corridor axisenclosed by walls or requires many turns An atrium in thecenter of a building can provide visual access between floorsthat helps people correctly align floors in their memory represen-tations (Peponis et al 1990) especially when the staircaseescala-torelevator is visually linked to this vertical core Views to theoutside with a distinctive global landmark can further supportintegration of cognitive maps between floors

The architectural community has developed spatial analysismethods to predict human movement patterns and cognitive rep-resentations namely ldquospace syntaxrdquo (Conroy Dalton 2005 Hillieramp Hanson 1984) and ldquoisovistrdquo approaches (Benedikt 1979 Wieneret al 2012) capturing visual access and layout complexity alike Itis noteworthy that these approaches concentrate on single-floorenvironments and that current space syntax analysis treats multi-level connections only in a simplified fashion (Houmllscher et al2012) While this reflects a correspondence with the bicodedview more fine-grained analysis of vertical connectors in buildinganalysis is called for

Human indoor navigation is strongly driven by the semantics offloor layout and expectations about the positioning of meaningfultypical destinations and landmark objects (Frankenstein et al2010) as well as by expectations of regularity good form andclosure (Montello 2005) Human navigation is further guided bysymbolic input like signage and maps Houmllscher et al (2007)have shown that signage and cross-sectional maps explicitlydesigned to highlight multilevel relations and connectionsbetween floors help to overcome vertical disorientation makingfirst-time visitors almost as efficient in wayfinding as a benchmarkgroup of frequent visitors

Based on their experience with typical buildings humansdevelop specific strategies for handling vertical layouts (egcentral-point or floor strategies Houmllscher et al 2006) Jefferyet al refer to Buumlchner et alrsquos (2007) finding that more people pre-ferred a horizontal-first route This could be partly due to the factthat the horizontal distances were longer than the vertical here (asin many buildings) and that people tend to follow the longest lineof sight (Conroy Dalton 2003 Wiener et al 2012) Furthermorepeople adapt their strategies to properties of the building theaforementioned floor strategy can then be replaced by a ldquoline-of-sightrdquo strategy depending on the connections in a complexbuilding (Houmllscher et al 2009)

Finally we wish to point to the need for understanding the roleof inter-individual differences in handling verticality along thedistinction of route versus survey knowledge and the degree towhich humans are able to coherently integrate cognitive maps

of stacked floors Studies cited by Jeffery et al as well as ourown are compatible with separate planar representations ofspatial information However a subset of participants in ourexperiments consistently reports imagining the environmentfrom an external perspective with walls and floors like glass (ldquoaglass doll houserdquo) This appears to require a consistent represen-tation of the vertical dimension and these people tend to reportallocentric rather than egocentric orientation strategies as wellas high levels of spatial abilities Jeffery et al state that investi-gating multi-floor environments alone is insufficient to identifyto what degree surface-dwelling animals can build true volumetricmental maps Virtual-reality simulations of such environments ndashfor example tilting a building layout by 90 degrees or comparingmultilevel navigation to ldquovertical floatingrdquo along tunnels ndash provideuseful extensions in this direction In studies currently under waywe focus on the role of individual differences in how the verticalcomponent is integrated

Applying the bicoded spatial model tononhuman primates in an arboreal multilayerenvironment

doi101017S0140525X13000411

Allison M Howard and Dorothy M FragaszyDepartment of Psychology University of Georgia Athens GA 30602ndash3013allisonmariehowardgmailcom doreeugaeduhttppsychologyugaeduprimate

Abstract Applying the framework proposed by Jeffery et al to nonhumanprimates moving in multilayer arboreal and terrestrial environments wesee that these animals must generate a mosaic of many bicoded spacesin order to move efficiently and safely through their habitat Terrestriallight detection and ranging (LiDAR) technology and three-dimensionalmodelling of canopy movement may permit testing of Jeffery et alrsquosframework in natural environments

Jeffery et al propose that a bicoded representation of three-dimensional space in which horizontal and vertical dimensionsare encoded in fundamentally different ways (ie metric andnon-metric) is common to all vertebrate animals Althoughbicoded spatial representation has implications for animalsmoving in all substrates this commentary focuses on how thetheoretical framework outlined by Jeffery et al might beapplied to nonhuman primates moving in natural multilayerenvironments and what techniques might be applied to thisproblemThe neural evidence upon which Jeffery et al base their con-

clusions comes largely from rats habitually moving within singlelayers of space or in multilayered compartmentalized space(eg tunnels) The authors also describe animals that move involumetric space (eg fish birds) and the need for data regardinghow these animals represent space in three dimensions ApplyingJeffery et alrsquos framework to arborealterrestrial nonhuman pri-mates we find that the space in which these animals move pre-sents significant challenges for generating spatial representationmosaics which Jeffery et alrsquos examples do not habitually encoun-ter A nonhuman primate that moves both on the ground and intrees (eg chimpanzees capuchin monkeys macaques) is pre-sented with the option of travel on substrates that may occur ata great variety of distances angles and heights from the horizontalplane of the animalrsquos location at any given moment Vertical hori-zontal and intermediately angled arboreal substrates comingleand contact substrates from neighboring trees The animal mustaccurately estimate distance to cross open spaces by leaping orbrachiation These locomotor patterns are associated not onlywith horizontal but also vertical displacements of the animal(eg Channon et al 2010) Considering Jeffery et alrsquos

Figure 12 in the context of a primate travelling in the trees wemight visualize similar arboreal bicoded map fragments thatoccur along numerous branches in many directions and extendinginto various distances from the central tree trunk These bicodedmap fragments meet and join fragments from neighboring treesforming a web of metrically mapped spaces that with increasingdensity of branches and variation in the branch angles willapproach metrically mapping both horizontal and vertical space(in the canonical orientation sense) assuming that the bicodedfragments are accurately joined in a mosaic

Landscapes within which animals move consist of both verticaland horizontal components that animals act upon Landscapecomponents such as topography have an impact on the mannerin which animals move through their environments (egMandel et al 2008) Animals moving primarily through a singlelayer of their environment (eg large grazing animals) areimpacted by elevation the one vertical component of the land-scape upon which they move (Bennett amp Tang 2006) Becausemoving uphill requires greater energy expenditure than movingon level ground (Taylor amp Caldwell 1972) terrestrial animalsmay make changes in the horizontal nature of their movementsin order to avoid a given area due to its vertical character (iedetouring around steep slopes) In contrast animals that movein multilayer environments contend with elevation as well asother additional vertical components to their movement decisions(ie substrate height) For example nonhuman primates andother vertebrates locomote in multilayer environments with acombination of arboreal and terrestrial substrates The varyingslopes of the multitude of potential substrates (ie branchestree trunks and terrestrial surfaces) and the locations to whichthese substrates convey present these animals with numerousoptions for movement These options also allow the animal toexert greater control over the vertical component of its movementdecisions For example continuous canopy forest would allowarboreal primates and other quadrupedal animals that movethrough arboreal habitats to travel at a constant height minimiz-ing vertical displacement while the elevation of the forest floorrises and falls We maintain that the use of non-compartmenta-lized multilayer environments requires a representation of spacethat is sufficiently accurate to allow for movement decisions inthe vertical and horizontal dimensions as well as precise aerialswingingleaping to distal substrates Logistically this spatialmodel may be hypothesized to include at a minimum preciseheuristics regarding where when and how to swing or leap orperhaps even a metric component of the vertical dimension

In studies of the movement of nonhuman primates in multi-layer environments movement observations are frequently sim-plified for their analysis Actual three-dimensional animalmovements are reduced to two-dimensional displacementsacross a planar surface (Janson 1998 2007 Normand amp Boesch2009 Noser amp Byrne 2007b Sueur 2011 Valero amp Byrne2007) Some prior studies have incorporated the vertical dimen-sion of movement in discussions of nonhuman primate movementpatterns by linking elevation to visibility of resource sites (eg DiFiore amp Suarez 2007 Noser amp Byrne 2007a) Draping the move-ments of nonhuman primates onto a digital elevation model oftheir habitat allows us to consider the energetic and viewpointeffects resulting from the vertical component of movement on aterrestrial substrate (eg Howard et al 2012) However thegreater vertical complexity of moving through multilayer environ-ments (eg arboreal and terrestrial) and its effects on adaptivemovement choice are not considered using this technique

One technique that may accurately represent the arboreal sub-strates upon which nonhuman primates move is Light Detectionand Ranging (LiDAR) technology Terrestrial LiDAR is three-dimensional laser scanning that generates a hemispherical pointcloud representing the returns of laser pulses from a ground-based vantage point (Beraldin et al 2010) This high-densitypoint cloud can be used in forest inventories and in measuringthe detailed geometric characteristics of trees (Maas 2010)

Point clouds can be used to generate three-dimensional modelsof the canopy through which animals move The use of thistechnique would allow researchers to catalog all possible sub-strates on a tree or group of trees estimate the energetic costsof movement on those substrates and even model animalsrsquomove-ment through a multilayer canopy based on heuristics or metricknowledge of vertical and horizontal spatial components of theirenvironment In this way the framework proposed by Jefferyand colleagues may be tested against movements of animals innatural environments

The complex interplay between three-dimensional egocentric and allocentric spatialrepresentation

doi101017S0140525X13000423

David M KaplanDepartment of Anatomy and Neurobiology Washington University School ofMedicine St Louis MO 63110kaplaneye-handwustledu

Abstract Jeffery et al characterize the egocentricallocentric distinctionas discrete But paradoxically much of the neural and behavioralevidence they adduce undermines a discrete distinction More strikinglytheir positive proposal ndash the bicoded map hypothesis ndash reflects a morecomplex interplay between egocentric and allocentric coding than theyacknowledge Properly interpreted their proposal about three-dimensional spatial representation contributes to recent work onembodied cognition

The efforts of Jeffery et al to synthesize and provide an overarch-ing theoretical interpretation for the literature on two-dimen-sional and three-dimensional spatial representation willundoubtedly prove useful to the field Their focus on 3D spacealso highlights the importance of investigating navigation per-formance in more complex ecologically valid task environmentsDespite these strengths the authors fall prey to a common con-fusion involving the central egocentricallocentric distinction

At the outset the authors flag this familiar distinction to empha-size how the ldquofocus of the present article is on the allocentricencoding of the space that is being moved throughrdquo (targetarticle sect 2 para 1) Although a discrete distinction is widelyassumed in the literature and the authors imply it is incidentalto the main goals of the article and hence can be glossed overtaking it at face value is problematic for two reasons Firstmuch of the neural and behavioral evidence the authors surveyundermines a clean division Second and more strikingly theirpositive proposal ndash the bicoded map hypothesis in which differentencoding schemes are employed for horizontal versus verticalspace yet both are referenced to the organismrsquos plane of loco-motion ndash clearly reflects a more complex interplay between ego-centric and allocentric spatial representation than the authorsexplicitly acknowledge

Egocentric and allocentric representations are primarily distin-guished by the different reference frames they employ Referenceframes specify locations in terms of distances along two (or three)perpendicular axes spanning out from an intersection point at theorigin Egocentric representations specify locations relative to areference frame centered on a body axis such as the midline oron a body part of the organism (eg ldquo20 cm to the right of myright handrdquo) Allocentric representations encode locations relativeto a reference frame centered on some environmental feature orobject (or set of these) and independently of the organismrsquos ownorientation or possibly even position (eg ldquo33degS 151degErdquo) Framedin this way the key difference underlying the egocentricallo-centric distinction concerns the origin or center of the referenceframe for the spatial representation a point which is clearly

reflected etymologically in the terms ldquoegocentricrdquo and ldquoallo-centricrdquo Nevertheless characterizing a spatial reference frameinvolves more than just specifying the origin Axis orientationand a metric for assessing distances along these axes must alsobe determined This gives rise to another facet of the ego-centricallocentric distinction that has been largely ignored inthe literature (for an exception see Grush 2007)

There are at least two senses in which a spatial representationmay be egocentric (or allocentric) As canvassed above a rep-resentation can be egocentric if the origin of the referenceframe is anchored to the observerrsquos location But a spatial rep-resentation can also be egocentric if the alignment or orientationof the reference frame axes is itself dependent on properties ofthe organism (independently of whether the origin is fixed tothe organism or not) Consider how bodily features such as thedorsalventral axis of the head might serve to determine the updown axis for certain egocentric spatial representations used inearly vision (Pouget et al 1993) or how the body midline mightserve to orient the leftright axis for spatial representationsemployed in reach planning (Pouget amp Sejnowski 1997) Theserepresentations are plausible candidates for being egocentric inboth senses Although these two kinds of egocentricity can gohand in hand this is not necessary Importantly allocentricspatial representations are subject to an equivalent analysis suchthat a representation may be allocentric with respect to itsorigin axes or both

A more nuanced egocentricallocentric distinction facilitatesthe identification of a richer representational taxonomy includinghybrid cases incorporating both egocentric and allocentricelements as building blocks For example consider a hybrid rep-resentation in which locations are specified relative to an ego-centric origin but where the horizontal and vertical axes arealigned with respect to some environmental feature such as theEarthrsquos magnetic field The axes defined by the cardinal directionsare allocentric as they bear no direct relationship to any egocen-trically defined axes and are invariant across changes in orientationand position of the subject

This is important because the authorsrsquo bicoded map hypothesisposits a hybrid spatial representation incorporating an allocentri-cally defined origin and egocentrically defined axes According tothe hypothesis different encoding schemes (metric vs non-metric) are used for the representation of horizontal and verticalspace in the hippocampus and medial entorhinal cortex (MEC)Importantly these schemes are not referenced to horizontal andvertical per se (ie gravitationally defined earth-horizontal andearth-vertical) as might be expected based on previous navigationstudies employing two-dimensional arenas Instead studies con-ducted by the authors using three-dimensional environments(Hayman et al 2011) and by other groups using microgravity con-ditions (Knierim et al 2000) in which egocentric and allocentricaxes can be explicitly dissociated indicate that the ldquohorizontalrdquoaxis is defined by the organismrsquos canonical orientation duringnormal locomotor behavior and the axis orthogonal to thisdefines ldquoverticalrdquo If correct this organism-dependent axial align-ment makes the spatial representation under consideration ego-centric in the second sense outlined above Nevertheless themetrically coded (horizontal) portion of this map implementedby assemblies of hippocampal place cells and entorhinal gridcells also possesses characteristic features of an allocentric rep-resentation as its reference frame is anchored to the externalenvironment and does not depend on the organismrsquos own locationThe bicoded map is a hybrid spatial representation

When the bicoded map hypothesis is interpreted as has beensuggested here its link to recent work in the field of embodiedcognition becomes evident Embodied cognition researchershave long sought to reveal how physical embodiment and motorbehavioral capacities shape and constrain our ability to representspace Until now findings in support of this link were restricted tohow organisms represent their local workspace If Jeffery et al areright evidence is emerging for how an organismrsquos motor

behavioral repertoire may also influence its representation oflarge-scale navigable space

The planar mosaic fails to account for spatiallydirected action

doi101017S0140525X13000435

Roberta L Klatzkya and Nicholas A GiudicebaDepartment of Psychology Carnegie Mellon University Pittsburgh PA15213 bSpatial Informatics Program School of Computing and InformationScience University of Maine Orono ME 04469-5711klatzkycmuedu nicholasgiudicemaineeduhttpwwwpsycmuedupeopleklatzkyhtmlhttpspatialumaineedufacultygiudice

AbstractHumansrsquo spatial representations enable navigation and reachingto targets above the ground plane even without direct perceptual supportSuch abilities are inconsistent with an impoverished representation of thethird dimension Features that differentiate humans from most terrestrialanimals including raised eye height and arms dedicated to manipulationrather than locomotion have led to robust metric representations ofvolumetric space

Consider some human capabilities for actions directed at spatialtargets at varying distances above the ground plane at an extremesnagging a fly ball on the run or pole vaulting at the mundanelevel turning on the wall switch or stepping over an obstacle onthe ground These actions are accurate and precise yet they gener-ally are not performed under closed-loop control that would free usfrommetric demands It seems unlikely that the planar mosaic rep-resentation proposed by Jeffery et al ndashwhere the third dimension isnot only non-metric but unstable ndash could support their executionHow do we resolve the disparity between what would be poss-

ible under a non-metric representation of space and what peoplecan actually do One avenue toward resolution is to say ldquoOh butJeffery et al are not referring to those types of behaviorsrdquo Butwhat then differentiates the behaviors ostensibly governed bythe planar mosaic from human spatially directed actions such aspointing reaching over-stepping and making contactFor one thing actions such as catching balls and reaching for

targets on a wall result from characteristics of human perceptionand action that most other terrestrial mammals do not shareHumans are ldquoecologically three-dimensionalrdquo to a high degreeOur raised eyes provide a perspective view of the terrain wherewe might travel within a volumetric context so vast it has beencalled ldquovista spacerdquo (Cutting amp Vishton 1995) Although notwithout error our representation of height variation acrossenvironmental undulations is simply not possible for animalswhose sense organs remain close to the ground during navigationHumans differ as well from rodents and ants by having armslimbs that are used not for locomotion (beyond infancy) butrather to reach and manipulate objects above the groundSpatially directed actions also potentially differ from terrestrial

navigation in that the corresponding motor commands mustspecify the disposition of the entire organism in volumetricspace ndash not only its location in point coordinates but limb posturesand joint angles Perhaps this provides an avenue of reconciliationwith the planar mosaic representation Actions directed towardtargets with particular metric relationships to the body may bedesignated as egocentric Hence they would lie specificallyoutside the scope of the current theory which restricts itself toallocentric (environmentally referred) representations This exclu-sion of metrically constrained behaviors from consideration isundermined however by the intrinsic ambiguity of frames ofreference (Klatzky 1998) Humans flexibly compute transform-ations between self-referred and environmentally referredframes even within a single task (Avraamides et al 2004)

Lacking reliable behavioral or neural signatures that would allowus to designate actions as egocentric on the basis of their metricdemands it seems inappropriate simply to exclude them from atheory of volumetric spatial representation

But wait ndash there is another feature of reaching jumping catch-ing and so on that might render such behaviors distinct fromthose guided by a volumetric mosaic Arguably these behaviorsare supported by perception rather than representational abstrac-tions This argument might work if perceptually guided actions aresomehow different from those guided by something called ldquorep-resentationrdquo presumably including navigation However a pos-ition to the contrary has been put forward by Jack Loomis andthe present authors along with collaborators (for review seeLoomis et al 2013) Our proposal stems from the fundamentalidea that the perceptual system functions to create represen-tations and it is these representations that guide action

Wehave used the term ldquospatial imagerdquo to refer to a particular typeof representation that can support behaviors such as navigation andreaching even when sensory systems no longer provide data aboutthe physical world For example when a sound is emitted andthen ceases a spatial image of the soundrsquos location still remains tosupport navigation (Loomis et al 2002) The spatial image makespossible not only orienting and direct locomotion toward thetarget but also spatial updating by means of path integration

Importantly for present concerns we have recently shown thatthe spatial image is three-dimensional and can be formed byactions with the arm as well as by vision (Giudice et al 2013)Subjects in our experiments formed representations of locationsin volumetric space that they viewed touched directly orprobed with a stick They then walked without vision by an indir-ect path requiring spatial updating and touched the target at theappropriate height Localization was not only accurate but mini-mally affected by how the target representation was formed Themean height error (signed distance between target and responseheight) was only 1 9 7 and 3 cm for targets originally exploredwith the hand long pole short pole and vision respectively Pre-cision as measured by variability around the response locationwas also little affected by mode of exploration

These findings demonstrate several pertinent points First fullythree-dimensional spatial representations were formed with highaccuracy Second those representations conveyed metric data suf-ficient to support navigation even in the absence of vision ndash thatis with perceptually guided homing precluded and spatial updat-ing as a requirement Third the spatial image afforded actiondirected toward the object with the arms as well as locomotion

We have posited that human action capabilities require a metricrepresentation of volumetric space that seems incompatiblewith theclaim that terrestrial mammals are restricted to an impoverishedrepresentation of the third dimension in the form of a looselybound mosaic of planar maps Although we have focused onhumans we see no reason not to extend these arguments to pri-mates more generally and we note that other mammals such assquirrels and cats routinely leap and jump to vertical targetlocations We concede however the potential for constraints onhuman spatial representation In particular when our circumstancesbecome more like those of animals confined to the ground plane(eg when travelling in a car) our cognitive maps too mayprovide only a local orientation relative to the direction of travel

Monkeys in space Primate neural datasuggest volumetric representations

doi101017S0140525X13000447

Sidney R Lehkya Anne B Serenob and Margaret E SerenocaComputational Neurobiology Laboratory The Salk Institute La Jolla CA92037 bDepartment of Neurobiology and Anatomy University of Texas Health

Science Center Houston TX 77030 cDepartment of Psychology University ofOregon Eugene OR 94703sidneysalkedu annebserenouthtmcedumserenouoregoneduhttpwwwsnlsalkedu~sidneyhttpnbauthtmceduhomepageserenoserenoindexhtmhttppsychwebuoregonedu~serenolab

Abstract The target article does not consider neural data on primatespatial representations which we suggest provide grounds for believingthat navigational space may be three-dimensional rather than quasindashtwo-dimensional Furthermore we question the authorsrsquo interpretation ofrat neurophysiological data as indicating that the vertical dimension maybe encoded in a neural structure separate from the two horizontaldimensions

The neurophysiological data presented by Jeffery et al in thetarget article in support of the view that spatial representationsfor navigation by surface-travelling mammals are quasi-planarcome from just one animal the rat There is no consideration ofneural data from primates We suggest here that when primateneural data are examined it raises the possibility that primateshave three-dimensional volumetric spatial representations fornavigation not quasi-planar ones Furthermore we question theauthorsrsquo interpretation of rat physiological data as suggestingthat the encoding of the vertical dimension occurs in a differentneural structure from the horizontal dimensions ldquoin an as yetundiscovered regionrdquo (sect 51 para 5)

One indication of possible differences in spatial representationsbetween rats and primates comes from comparing allocentricresponses in the hippocampus Rat hippocampal place cells fireonly when a rat is located at a particular place On the otherhand macaque monkeys also have hippocampal cells thatrespond when a monkey is merely looking from a distance at a par-ticular spot whose location has both horizontal and vertical com-ponents (Georges-Franccedilois et al 1999 Rolls 1999)

Although neurophysiological studies of navigation in ratshave focused on the hippocampus neurophysiological and func-tional magnetic resonance imaging (fMRI) studies in macaquemonkeys and humans have highlighted other structures as alsoimportant during navigation Among them are the posterior parie-tal cortex posterior parahippocampus and retrosplenial cortex(RSC) together with the nearby posterior cingulate cortex(PCC) (Epstein 2008 Kravitz et al 2011) RSC and PCCappear to be part of a network that transmits spatial informationfor navigation from the posterior parietal cortex to medial tem-poral lobe structures in particular the parahippocampus andhippocampus

In addition to being the ultimate source of spatial informationfor navigation the posterior parietal cortex is also involved inspatial representations for the control of action (which may bedistinct from spatial representations for object recognition andmemory [Sereno amp Lehky 2011] see also earlier work byGoodale amp Milner 1992) Control of action includes control of3D eye movements and 3D visually guided reaching and grasp-ing by the arm and hand (Blohm et al 2009 Breveglieri et al2012 Hadjidimitrakis et al 2012) Spatial representations forthe control of action in primates operate in a 3D volumetricspace and not a quasindash2D multiplanar space Furthermorerecent physiological studies in monkeys of populations of pos-terior parietal cells (Sereno amp Lehky 2011) show evidence fora 3D representation of space in primates even when simplyfixating

As the posterior parietal cortex in primates appears to be asource of spatial information both for control of action and fornavigation it seems a reasonable conjecture that known parietal3D spatial representations for control of action could also beused for navigation While the dimensionality of space represen-tation for navigation in primates is an important topic that hasnot been well studied there are physiological reasons to believethat it may be three-dimensional

It is also not clear that the representation of space in the rat isquasindashtwo-dimensional as Jeffery et al claim in the sense thatinformation about the vertical dimension is processed by differentneural circuitry The authors cite physiological evidence of neuralactivity in navigation circuits that is modulated by the verticaldimension for example elongated vertical (relative to horizontal)place-cell and grid-cell receptive fields for vertically arrangedspatial layouts It doesnrsquot follow from those response anisotropiesthat medial temporal lobe structures are not encoding the verticaldimension Computational studies for example might establishthat observed properties of rat grid and place cells are sufficientto account for behavioral abilities within the vertical dimensionwithout postulating other unknown neural centers for verticalspatial representations Indeed there is a debate within the theor-etical literature about whether narrowly tuned or coarsely tunedcells provide better representations within a population (Lehkyamp Sereno 2011 Pouget et al 1999 Zhang amp Sejnowski 1999)As the authors themselves state ldquomuch remains to be determinedabout vertical processing in the navigation systemrdquo (sect 41 para10) Therefore the conclusion of the target article that space fornavigation is quasindashtwo-dimensional for all surface-travellingmammals may be premature

Development of human spatial cognition in athree-dimensional world

doi101017S0140525X13000459

Kate A Longstaffe Bruce M Hood and Iain D GilchristSchool of Experimental Psychology University of Bristol Bristol BS8 1TUUnited KingdomKatelongstaffeBristolacuk BruceHoodBristolacukIaingilchristBristolacukhttpwwwbristolacukexpsychpeoplekate-a-longstaffeindexhtmlhttpeisbrisacuk~psidghomepagehtmlhttpwwwbristolacukexpsychpeoplebruce-m-hoodindexhtml

Abstract Jeffery et al accurately identify the importance of developing anunderstanding of spatial reference frames in a three-dimensional worldWe examine human spatial cognition via a unique paradigm thatinvestigates the role of saliency and adjusting reference frames Thisincludes work with adults typically developing children and childrenwho develop non-typically (eg those with autism)

Jeffery et alrsquos target article explores behavioural and neurobiolo-gical studies from many species and draws attention to the gaps inthe literature on human three-dimensional navigation Our ownwork addresses this gap by using a novel paradigm to examinehuman searching behaviour in large-scale space We have usedthis setup in a series of adult and developmental studies (Pellicanoet al 2011 Smith et al 2005 2008 2010)

Our large-scale search laboratory contains an array of greenlight-emitting diode (LED) light switches embedded in thefloor Participants press illuminated switches to discover targetsthat change colour when pressed Potential targets can bemanipulated by colour saliency and probability of distributionOur paradigm preserves the experimental rigor of classic visualsearch by recording button presses to millisecond accuracywhile scaling up to large-scale search in which the human partici-pant has to move through the search array This provides a firststep towards investigating the processes involved in human navi-gation which addresses a concern that has been raised regardingthe validity of modelling human search solely via two-dimensionalcomputer monitorndashbased search (Dalrymple amp Kingstone 2010Foulsham et al 2011 Foulsham amp Kingstone 2012 Ingram ampWolpert 2011)

Work comparing visually guided versus hidden target con-ditions revealed increased search latencies when the target is

not visually available (Smith et al 2008) In both visually andnon-visually guided conditions there were fewer revisits to pre-viously explored locations than are present in classic computerscreen visual search studies emphasizing a key distinctionbetween visual search and large-scale search (Smith et al 2008)Jeffery et al also highlight a gap in human developmental

work on navigation and spatial cognition which is crucial to build-ing an understanding of how humans develop complex spatialrepresentations Research in our paradigm has investigated therole of short-term memory in preventing revisits to previouslyinspected locations (Smith et al 2005) Children made signifi-cantly more revisits to previously examined search locationswhen performing a more effortful search with the non-dominanthand suggesting that when the task is more effortful individualsare less able to optimize search In addition although childrenrsquosgeneral fluid intelligence was unrelated to search time therewas a significant relationship between search latency and visuo-spatial short-term memory (measured via Corsi blocks) Thesedata highlight the role of spatial working memory in the deve-lopment of efficient exploration of large-scale space (Smithet al 2005)Further work focused on the learning of likely target locations

(Smith et al 2010) Here targets were located on one side of theroom in 80 of trials Participants were able to learn likely targetswhen both the start location and search locations were fixedthroughout the task However when room-based and referenceframes were disrupted there was no evidence of learningsuggesting that encoding of likely locations depends on a combi-nation of egocentric and allocentric cues (Smith et al 2010)This same paradigm has been used in individuals with autism

following the suggestion by Baron-Cohen (eg Baron-Cohen2008) that the exceptional visual search skills of children withautism are due to an increased aptitude at systemizing spatial cog-nition However children with autism were less efficient less sys-tematic less optimal and overall less likely to follow probabilisticcues than ability-matched typical children This provides furtherevidence that visual search and large-scale search do not alwaysshare the same cognitive mechanisms (Pellicano et al 2011)Recent work has investigated the role of saliency in building

spatial representations In work with adult participants the per-ceptual salience of search locations was manipulated by havingsome locations flashing and some static a paradigm adaptedfrom established visual search literature (Theeuwes 1994 Yantisamp Jonides 1984) Adults were more likely to search at flashinglocations even when explicitly informed that the target wasequally likely to be at any location (Longstaffe et al 2012) Wepropose that attention was captured by perceptual salienceleading to an automatic bias to explore these targets When thiswork was extended to a population of typically developing chil-dren they were also more likely to search at flashing locationsand the magnitude of this effect did not vary with ageHowever there was a strong developmental trend in thenumber of times children revisited previously examined locationsThe developmental trajectory for ability to remember previouslyvisited locations and limit revisits shows a development inspatial working memory occurring separately from perceptualinhibition This suggests individual executive sub-processes mayplay different roles during search with different developmentaltrajectories Further work in this paradigm examines the effectof mediating working memory load on search behaviour Thisinvestigation of working memory during search is crucial for atransition into real-world tasks such as searching for lost keys ora lost car in a parking lotThe review by Jeffery et al commendably addresses the scope

of research across many species and paradigms here we presentwork in a novel paradigm that expands our understanding ofhuman adult and child populations in the three-dimensionalworld Understanding how individuals incorporate referenceframes and employ cognitive mechanisms such as memory andattention is key to modelling human navigation

Has a fully three-dimensional space map neverevolved in any species A comparativeimperative for studies of spatial cognition

doi101017S0140525X13000460

Cynthia F MossDepartment of Psychology Institute for Systems Research University ofMaryland College Park MD 20742

cynthiamossgmailcomhttpwwwisrumdedufacultygatewaysmosshtm

Abstract I propose that it is premature to assert that a fully three-dimensional map has never evolved in any species as data are lacking toshow that space coding in all animals is the same Instead I hypothesizethat three-dimensional representation is tied to an animalrsquos mode oflocomotion through space Testing this hypothesis requires a large bodyof comparative data

The target article by Jeffery et al reviews the literature on three-dimensional spatial navigation and highlights important consider-ations for interpreting both behavioral and neurobiological dataon this topic In their review they lay the foundation for thehypothesis that three-dimensional spatial navigation is ldquobicodedrdquo ndashthat is based on different encoding mechanisms for horizontaland vertical space Furthermore they argue that a fully three-dimensional map has never evolved in any species

The ldquobicodedrdquo model is based on studies of animals that movealong surfaces to navigate which constrains the design of exper-iments and the data used to build theory For a complete theoryof three-dimensional spatial navigation a wider range of animalsmust be studied In particular data from animals that freely navi-gate three-dimensional volumetric space will help determinewhether a single model can account for three-dimensionalspatial navigation in terrestrial aerial and aquatic speciesCounter to the model proposed by the authors of the targetarticle I hypothesize that a fully three-dimensional volumetricmap has evolved in species that are not limited to navigatingalong surfaces More broadly I hypothesize that three-dimen-sional spatial representation depends upon an animalrsquos mode oflocomotion and the navigational tasks it encounters in thenatural environment

Another point I would like to make is that models of three-dimensional spatial navigation must be grounded in behavioraland neurobiological studies of animals engaged in biologically rel-evant and natural tasks For example rodents have convenientlyserved as animal models for a wide range of studies includingspatial navigation Many rodent maze studies are howeverhighly artificial requiring that animals perform tasks they are unli-kely to encounter in a natural setting and in spaces far morerestricted than they would navigate in the wild The study ofinbred rodents that have not navigated the natural environmentfor many generations raises additional concerns

Research traditions have limited the study of spatial navigationand advances in the field require a comparative approach Theimportance of choosing the right animals for the questionsunder study first articulated by Nobel Laureate August Krogh(1929) is widely recognized in the field of Neuroethology but isless influential in the broad field of Systems Neuroscience It isimportant that the spatial navigation research community inter-ested in problems of three-dimensional spatial navigation turnto the study of animals that have evolved to solve this problem

Echolocating bats present an excellent model system to pursuequestions about three-dimensional spatial navigation Bats belongto the order Chiroptera many species of which use biologicalsonar systems to represent the spatial location of targets andobstacles In turn this spatial information is used to build a cogni-tive map that can guide navigation in the absence of sensory cuesAnecdotal reports and laboratory studies of echolocating batsprovide evidence that bats rely strongly on spatial memory(Griffin 1958 Jensen et al 2005) and field studies show that

bats use memory on many different spatial scales (Schnitzleret al 2003 Tsoar et al 2011) Importantly bats use an absol-ute-spacendashbased (allocentric) navigational strategy that is hippo-campus-dependent and place cells have been identified andcharacterized in the bat hippocampus (Ulanovsky amp Moss 2007Yartsev amp Ulanovsky 2013) Furthermore grid cells have beenrecently described in the medial entorhinal cortex (MEC) of theEgyptian fruit bat (Yartsev et al 2011) Therefore echolocatingbats are particularly well suited for researchers to use in behavior-al neurobiological and computational studies of three-dimen-sional spatial navigation

Neural recording data from the MEC and hippocampus of batshave raised questions about the generality of the rodent model inspatial representation Although neural recordings from awakebehaving rats and bats show that neurons in the hippocampusand MEC represent two-dimensional space in a similar way (Ula-novsky amp Moss 2007 Yartsev et al 2011) there is one noteworthydifference namely an absence of continuous theta oscillations inhippocampal recordings from the big brown bat (Ulanovsky ampMoss 2007) and the MEC and hippocampus in the Egyptianfruit bat (Yartsev et al 2011 Yartsev amp Ulanovsky 2013) Wholecell patch clamp studies of the MEC of the big brown also demon-strate differences in subthreshold membrane potential resonancebetween bats and rats (Heys et al 2013) These reported speciesdifferences challenge a fundamental assumption of the oscillatoryinterference model of spatial coding (eg Burgess et al 2007Hasselmo et al 2007) By extension comparative data couldraise questions about the generality of three-dimensional spatialcoding mechanisms in the brains of animals that have evolved tonavigate along surfaces compared with those that move freelythrough three-dimensional volumetric spaces such as air andwater

In summary I propose that it is premature to assert that a fullythree-dimensional map has never evolved in any species asempirical data are lacking to support the notion that three-dimen-sional space coding in all animals is the same Recent findingsfrom Yartsev amp Ulanovsky (2013) demonstrating three-dimen-sional space representation in the hippocampus of the free-flying Egyptian Fruit bat are consistent with the hypothesis thatspace representation is tied to an animalrsquos mode of locomotionTo fully test this hypothesis requires a large body of comparativedata

ACKNOWLEDGMENTSupported by ONR Grant (N00014-12-1-0339) Three-dimen-sional Spatial Navigation to Cynthia F Moss and TimothyHoriuchi

Making a stronger case for comparativeresearch to investigate the behavioral andneurological bases of three-dimensionalnavigation

doi101017S0140525X13000472

Daniele Nardia and Verner P BingmanbaDepartment of Psychology Sapienza Universitagrave di Roma 00185 Rome ItalybDepartment of Psychology and JP Scott Center for Neuroscience Mind andBehavior Bowling Green State University Bowling Green OH 43403

dnardiuniroma1it vbingmabgsuedu

Abstract The rich diversity of avian natural history provides excitingpossibilities for comparative research aimed at understanding three-dimensional navigation We propose some hypotheses relatingdifferences in natural history to potential behavioral and neurologicaladaptations possessed by contrasting bird species This comparativeapproach may offer unique insights into some of the importantquestions raised by Jeffery et al

Jeffery et alrsquos target article is a fascinating synthesis of the rela-tively scarce but expanding research on three-dimensional naviga-tion in addition to proposing an engaging hypothesis on theneural representation of vertical space As recognized by Jefferyet al birds offer intriguing possibilities as model species for inves-tigating the behavioral and neurological bases of navigation in 3Dspace Studies on homing pigeons (Columba livia) suggest that thetopography of the landscape is an important source of informationfor bird navigation As mentioned by Jeffery et al sloped terrainprovides a global feature and ndash at least locally ndash a directional refer-ence frame The extent to which pigeons encode and use thisreference frame when walking on a slanted floor is impressive(Nardi amp Bingman 2009a Nardi et al 2010) A crucial attributeof slope ndash and more generally the vertical dimension ndash is theinvolvement of effort This has been highlighted in rats whichshow a strategy preference for horizontal movements relative tovertical movements (Grobeacutety amp Schenk 1992a Jovalekic et al2011) In pigeons evidence suggests that effort modulated bydifferent energetically demanding upward and downward move-ments affects how a sloped environment is represented or at leastused (Nardi et al 2012) This strengthens the hypothesis thateffort could be used as a ldquocontextualrdquo cue in the sense proposedby Jeffery et al derived from kinestheticproprioceptive infor-mation during locomotion on a slope Importantly hippocampallesions in pigeons do not impair the slope-based goal represen-tation (Nardi amp Bingman 2009b) despite having disruptiveeffects on memory for other spatial cues (eg Gagliardo et al1999 Vargas et al 2004) Possibly slope does not engage the hip-pocampus because it is used more as a source of directionalcompass-like information than for determining distances andmap-like information More studies are needed to ascertain therole of effort in the representation of sloped ndash and volumetric ndashspace and its neurological underpinnings

Beyond pigeons the class Aves with its rich diversity in naturalhistory offers exciting opportunities to apply comparativemethods for better understanding the representation of 3Dspace Although Jeffery et al acknowledge that evolutionary adap-tation (and constraints) reflected in species natural history maybe important in how vertical and horizontal space may be inte-grated they offer few explicit hypotheses that can be answeredby comparative-based research As a well-known example of thepower of comparative research the larger hippocampus andsuperior spatial cognitive abilities of food-storing birds (eg Pra-vosudov amp Clayton 2002) indicate that differential evolutionarypressure can lead to variation in natural history which co-occurswith variation in brain-behavior organization We are similarlyconvinced that interesting differences can also be found withrespect to navigating 3D space at both the relatively easilystudied behavioral level and the more difficult to study neurobio-logical level Almost all species of birds fly and necessarily inhabit athree-dimensional world but often differ with respect to theirldquo3D occupancy profilerdquo For example ground birds such assome quail species (family Odontophoridae) inhabit a robustlyhorizontal space whereas tree creepers (family Certhididae)search for food vertically but move horizontally among differenttrees and wallcreepers (family Tichodromidae) live in the sub-stantially vertical world of cliff faces1 How might natural differ-ences in horizontal vertical and volumetric occupancy influencethe representation of space and its neural encoding in the hippo-campal formation or elsewhere in the brain

In our opinion the most exciting and novel hypothesisadvanced by Jeffery et al is the notion that horizontal space ismore richly represented and is dissociated at least partiallyfrom the relatively impoverished contextual representation ofvertical space And this may indeed be true for species such asrats and even humans We wonder however how generally appli-cable this idea may be The black-capped chickadee (Poecile atri-capillus) is a remarkable species of bird that routinely lives in thecomplex volumetric space of deciduous trees and will seasonallystore and later recover seeds within the chaotic world of

intersecting tree branches ndash a behavior dependent on the integrityof the hippocampal formation (Sherry amp Vaccarino 1989) Chick-adees apply their robust spatial memory ability in a 3D spacewhich may appear multilayered but because deciduous treebranches can intersect at almost any angle would often bebetter characterized as true volumetric space (Coniferous treeshave a more multilayered quality) The survival of individuals isvery much dependent on ldquovolumetric encodingrdquo of stored foodthat would likely suffer if only a contextual heuristic would beused to encode vertical location From a purely adaptationist per-spective if any species ndash and more generally any natural historyprofile ndashwould support the evolution of a unified representationof volumetric space it would be the black-capped chickadeeand other species that store seeds in deciduous trees As anatural comparative experiment chickadees could be contrastedwith the Clarkrsquos nutcracker (Nucifraga columbiana) which alsostores seeds and has an extraordinary ability to recall the locationsof those seeds (eg Kamil et al 1994) However whereas chick-adees store seeds within the volumetric space of tree limbsnutcrackers store them in the ground in a rolling but two-dimen-sional horizontal space Would chickadees outperform nutcrack-ers in a spatial memory task in a sloping multilayered orvolumetric space Would hippocampal neurons in chickadees bemore likely to encode aspects of volumetric spaceWe are not advocating that comparative approaches would

answer all the important questions raised by Jeffery et al butwe hope we have convinced the reader that comparativeresearch ndash not necessarily limited to birds ndash is a powerful exper-imental strategy that can reveal much about the diversity ofthree-dimensional spatial representations

ACKNOWLEDGMENTWe would like to thank David Sherry for providing some helpfuladvice

NOTE1 We want to acknowledge here that comparative studies have inter-

pretative challenges associated with them and that the best comparisonsare among taxonomically closely related species (Harvey amp Pagel 1991)As such quails are probably not a good comparison group and we usethem only as a convenience to make our point

Which animal model for understanding humannavigation in a three-dimensional world

doi101017S0140525X13000484

Guy A OrbanDepartment of Neuroscience Division of Neurophysiology Parma University43100 Parma Italy

guyorbanmedkuleuvenbe

Abstract Single-cell studies of monkey posterior parietal cortex (PPC)have revealed the extensive neuronal representations of three-dimensional subject motion and three-dimensional layout of theenvironment I propose that navigational planning integrates this PPCinformation including gravity signals with horizontal-plane basedinformation provided by the hippocampal formation modified inprimates by expansion of the ventral stream

Jeffery et al ignore half of navigation namely navigation in ego-centric frameworks Many of the mechanisms of this type of navi-gation ndash as important as allocentric navigation as walking along acliff demonstrates ndash have been investigated in primates but areconspicuously absent from the present review Yet from an evol-utionary and genetic standpoint we humans are obviously muchcloser to apes and monkeys than to rodents Monkeys typicallylive and climb in trees and not surprisingly possess a three-dimensional (3D) space representation in their posterior parietal

cortex (PPC) (Andersen amp Buneo 2002) The PPC is dramaticallyexpanded in primates compared to rodents in which primary cor-tices V1 and S1 lie very close leaving little room for the extrastri-ate or posterior parietal cortex (Whitlock et al 2008) Single-cellstudies in monkeys have revealed neurons selective for opticflow components in several parietal regions middle superior tem-poral dorsal (MSTd Lagae et al 1994 Tanaka et al 1986) ventralintraparietal (VIP Schaafsma amp Duysens 1996) and medial PPC(Squatrito et al 2001) Human imaging studies suggest thatadditional regions such as V6 might be involved (Wall amp Smith2008) DeAngelis Angelaki and coworkers have comparedvisual and vestibular selectivities of neurons in MSTd VIP andvisual posterior sylvian (VPS) for rotation and translation inthree dimensions (see Chen et al 2010 2011a 2011b Takahashiet al 2007) Whereas MSTd and VIP neurons can support the per-ception of heading neurons in VIP and VPS support those of self-rotation In these areas all directions in space are representedwith no clustering of preferences close to the horizontal planeas would be expected if the vertical direction were poorly rep-resented This is unsurprising as in primates equipped with bin-ocular vision the z-axis (depth) is unique not the vertical axiswhich unlike depth is represented on the retina Indeed particu-larly in VIP and to a lesser degree in MSTd the depth axis wasunder-represented among preferred directions Preferred direc-tions and tuning widths of MSTd neurons however were non-uniformly distributed along the horizontal meridian explainingthe decrease in behavioral sensitivity to self-motion directionwith increasing eccentricity in humans and monkeys (Gu et al2010)

In their quest to demonstrate that humans mainly use a hori-zontal reference frame for navigation Jeffery et al quote psycho-physical studies of 3D slope perception and the misperception ofvertical slopes they reveal Contrary to the authorsrsquo claims thismisperception is part of a larger problem depth reconstruction(see above) which is complicated by the dependence of distancescaling on the derivation order of disparity Indeed human sub-jects have no precise representation of metric 3D structurewhether using stereo or monocular cues (Todd amp Norman2003) In no way can these misperceptions be taken as evidencethat neuronal mechanisms for estimating 3D slopes do not existIn fact such neurons have been documented in caudal intraparie-tal (CIP) part of the PPC (Taira et al 2000) and also in the infer-otemporal cortex (Liu et al 2004) ndash thus in both the dorsal andventral visual systems Similarly selectivity for second-order dis-parity gradients has also been documented (see Orban 2011 forreview) Some of the cortical areas in which these higher-orderneurons have been recorded may not even exist in rodentsagain underscoring the importance of using adequate modelsfor the human brain Together these results indicate that theprimate PPC includes representations in three dimensions ofsubject motion and of the layout of the environment two criticalpieces of information for navigating through 3D space

The starting point of the target article is the importance of thecognitive map in the hippocampus for spatial navigation Even ifgrid cells have recently been identified in monkey entorhinalcortex (Killian et al 2012) the hippocampal system of primatesbears only a vague resemblance to that of rodents Relativelysmall numbers of human hippocampal neurons are place cellsand these occur in several brain regions (Jacobs et al 2010)This is not surprising as the hippocampus receives major projec-tions from the ventral stream (Kravitz et al 2013) which in pri-mates processes information about objects at least as much asscenes and even information concerning conspecifics Indeedconsiderable evidence indicates that the human hippocampal for-mation processes not only spatial but also non-spatial and evennon-visual information (Liang et al 2013) This species differencemay be even further underestimated as the human brain evolvedto accommodate bipedalism and thus navigation in open spacesbut such navigation has so far been little investigated for technicalreasons On the other hand navigation supposes a transformation

of coordinate information from spatial maps into body-centeredrepresentations a function likely to be subserved by the PPCwhich is connected to the hippocampal formation via the retrospe-nial cortex in rodents (Whitlock et al 2008) A similar route linksthe primate precuneus and adjacent dorsal superior parietallobule with the hippocampal system (Kravitz et al 2011) Theinvolvement of these medio-dorsal PPC regions in the controland planning of navigation is supported by their activationduring observation of climbing and locomotion in humans (Abdol-lahi et al 2012) Hence I propose a much more balanced viewwhereby spatial navigation is both allocentric and egocentric andis controlled by both the hippocampal formation (fed by the quali-tative scene analysis in parahippocampal gyrus where gravity is notincorporated) and the metric space representation in the PPC (inwhich gravity is incorporated) These two streams of informationare integrated in the dorso-medial PPC devoted to planning loco-motor body movements A recent imaging study indicates that thisscheme clearly different from the bicoded scheme of Jefferyet al applies to humans (Indovina et al 2013) To what extentit may also apply to rodents is unclear but intriguingly one ofthe subregions of the rodent PPC (the anteromedial visual areaTorrealba amp Valdes 2008) may be a precursor of primate V6and supply optic flow information

In conclusion including nonhuman primates among the animalmodels is essential for understanding the human brain even fornavigation and spatial memory avoiding errors induced byjumping from rodents to humans as these authors have done

The study of blindness and technology canreveal the mechanisms of three-dimensionalnavigation

doi101017S0140525X13000496

Achille Pasqualottoa and Michael J ProulxbaBiological and Experimental Psychology Group School of Biological andChemical Sciences Queen Mary University of London London E1 4NSUnited Kingdom bDepartment of Psychology University of Bath Bath BA27AY United Kingdompasqualottoqmulacuk mjproulxbathacukhttppeoplebathacukmjp51

Abstract Jeffery et al suggest that three-dimensional environments arenot represented according to their volumetric properties but in a quasi-planar fashion Here we take into consideration the role of visualexperience and the use of technology for spatial learning to betterunderstand the nature of the preference of horizontal over verticalspatial representation

The article by Jeffery et al reviews the literature on spatial cogni-tion relative to three-dimensional environments which is an areavastly less investigated than are two-dimensional (or planar)environments The authors have concluded that three-dimen-sional spaces are represented in a quasi-planar fashion Here weadd to the discussion by examining the role played by develop-mental vision on the structure and functioning of brain areasinvolved in spatial cognition and the results of studies investi-gating the role of spatial learning on three-dimensional spacerepresentation

The role of developmental vision in spatial representation hasbeen reported by numerous studies on humans (Pasqualottoet al 2013 Postma et al 2007 Roumlder et al 2007 Vecchi et al2004 see Pasqualotto amp Proulx [2012] for a review) and animals(Buhot et al 2001 Hyvaumlrinen et al 1981 Paz-Villagraacuten et al2002) Yet the role played by vision on the establishment of thequasi-planar representation of three-dimensional spaces has notbeen investigated This represents a broad avenue for futurestudies Due to the crucial role played by vision in spatial cognition

(eg Frankenstein et al 2012) and vertical spatial perception (Ooiet al 2001 Sinai et al 1998) it is conceivable that visual experi-ence has an effect on both horizontal and vertical spatial represen-tation and therefore the results discussed by Jeffery et al may bedriven by experience with the visual world In other words blindindividuals especially if congenitally blind may not exhibit quasi-planar spatial representation of three-dimensional environmentsThis hypothesis may be well supported by studies reporting therole of blindness in the functioning of hippocampal cells inrodents (Hill amp Best 1981 Larsen et al 2009 Paz-Villagraacutenet al 2002 Poucet et al 2000) and on hippocampal structureand activation in humans (Chebat et al 2007 Deutschlaumlnderet al 2009 Kupers et al 2010 Leporeacute et al 2009)

The study by Passini and Proulx (1988) represents a rare inves-tigation on the spatial representation of a three-dimensionalenvironment by congenitally blind and sighted participantsThey found that overall blind participants took more decisionsat crucial points of the environment than did sighted onesThese points included intersections stairs open spaces etc Cru-cially vertical structures of the environment did not receive moreattention than horizontal ones

In a recent article Thibault et al (2013) investigated the effectof spatial learning on memory of a virtual three-dimensional build-ing The same building was learned either through ldquoplanarrdquo orldquocolumnarrdquo exploration It was followed by a spatial memorytask where participants re-experienced a segment of the explora-tory route The results showed that the group of participants whounderwent planar exploration had superior spatial memory per-formance Yet Thibault et al also found that for both groupsthe spatial memory task was better performed when participantsre-experienced the same segments as during the exploratoryphase (ie congruency effect) This suggests that according tothe learning type participants could store the three-dimensionalbuilding in quasi-planar and quasi-columnar fashions

These results suggest that quasi-planar spatial representationmay be the result of quasi-planar spatial exploration In otherwords by using virtual reality (Lahav amp Mioduser 2008 Peacuteruchet al 1995) or sensory substitution devices (Kupers et al 2010Proulx et al 2008 2012) three-dimensional environments couldbe explored and stored according to any fashion (ie planar orcolumnar) These devices would overcome the energetic costassociated with vertical exploration (Butler et al 1993 Houmllscheret al 2006) Finally the ability of the brain to learn and carryout spatial memory tasks in non-planar fashion is supported bystudies in ldquogravity-lessrdquo three-dimensional virtual mazes (Vidalet al 2003 2004) Therefore we advocate that future researchshould address the role played by visual experience and spatiallearning on spatial representation of three-dimensionalenvironments

Augmented topological maps for three-dimensional navigation

doi101017S0140525X13000502

Herbert Peremans and Dieter VanderelstActive Perception Lab FTEW-MTT Universiteit Antwerpen 2000 AntwerpenBelgiumherbertperemansuaacbe DieterVanderelstuaacbehttpwwwuaacbemainaspxc=herbertperemanshttpbitsofbatsweeblycom

AbstractWe describe an augmented topological map as an alternative forthe proposed bicoded map Inverting causality the special nature of thevertical dimension is then no longer fixed a priori and the cause ofspecific navigation behavior but a consequence of the combination ofthe specific geometry of the experimental environment and the motorcapabilities of the experimental animals

Based on a review of neurophysiological and behavioural evidenceregarding the encoding of three-dimensional space in biologicalagents Jeffery et al argue for a bicoded map in which heightinformation is available and used but not fully integrated withinformation from the horizontal dimensions in the cognitivemap We propose an alternative hypothesis namely an augmentedtopological map with ldquosemi-metricrdquo properties that under certaincircumstances will give rise to a representation which appears tobe bicoded but is based on the same principles of spatial encodingoperating in all three dimensionsAs an alternative to a metric representation roboticists

(Kuipers et al 2000 Wyeth amp Milford 2009) have found that atopological representation suffices for robots to map and navigateplanar environments In these experiments a graph represen-tation with nodes corresponding to distinct places and edges tospecific motor programs enables the robot to follow a pathfrom one distinct place to another To also allow the capabilityof executing shortcuts this graph representation is augmentedwith metric information Sensor-based estimates of travelled dis-tance and travel direction associated with the edges allow assign-ment of metric coordinates to the nodes in the graph Theseestimates will initially be unreliable However a mechanism asdescribed by Wyeth and Milford (2009) corrects the metric coor-dinate estimates for the graph nodes when the agent returns to apreviously visited distinct place As a consequence this represen-tation converges to a ldquosemi-metricrdquo map This differs from a realmetric map in that the reliability of the metric information associ-ated with the nodes varies across the graph For sparsely con-nected subgraphs for example less-travelled parts of theenvironment the metric information can be unreliable Forhighly interconnected subgraphs ndash for example well-exploredparts of the environment ndash the metric information can be usedfor optimizing paths for instance in calculating shortcutsWe propose that such a scheme can be extended to three

dimensions the nodes represent three-dimensional positionsand the edges the three-dimensional motor programs that movethe agent from one position (ie sensorially distinct place) toanother If the environment allows multiple three-dimensionalpaths from one place to another a ldquosemi-metricrdquo three-dimen-sional map would result Elsewhere we have described a mechan-ical analogue (Veelaert amp Peremans 1999) to model this processin which we map the nodes in the graph onto spherical joints andthe edges onto elastic rods To represent both the uncertaintyabout the actual distance travelled and the direction of travelbetween the distinct places represented by the nodes both thelengths of the rods and their orientations can vary (see Fig 1)The rigidity of the resulting mechanical construction tells uswhether all nodes are at well-defined positions relative to oneanother Whenever particular subparts of the mechanical con-struction can still move relative to one another this indicatesthat the corresponding parts of space are not well defined withrespect to each other However if the subparts of the mechanicalconstruction are rigid relative positions within those parts ofspace can be fully definedMultilayer environments modeled with this representation

result in a discrete set of highly interconnected subgraphs onefor each layer connected together by a small number of isolatedlinks The isolated links correspond to the few paths that can befollowed to go from one layer to another for example a staircasein a building As the subgraphs are connected by a few links onlythey can be metrically related to each other but the uncertainty onthis relationship is higher than that between distinct places at thesame layer The metric representations of the different layers areonly partially registered In terms of our mechanical analogue theconstructions representing the individual layers are rigid butthese rigid substructures are linked to one another through afew links only allowing them to move relative to each other asillustrated in Figure 1From the mechanical analogue it is clear that the extent to

which the augmented topological representation of space could

act as a metric representation ndash that is give rise to a rigid mechan-ical construction ndash depends on the interconnectivity of the nodesin the graph One factor that determines this interconnectivity iswhether the environment provides many opportunities for follow-ing different closed-loop paths that return to already-visited dis-tinct places Furthermore given an environment providingmany such opportunities whether a highly interconnectedgraph would actually be constructed also depends on the move-ment patterns of the mapping agent For example the rats inthe vertical plane maze seem to prefer movement patterns thatextensively explore horizontal layers only sparsely interspersedwith short movements up or down the vertical dimension Sucha movement pattern gives rise to a set of highly interconnectedsubgraphs linked together by a few isolated links similar to thatof a multilayer environment Hence to really test whether theaugmented topological representation (and the mechanism bywhich a three-dimensional ldquosemi-metricrdquo representation arisesfrom it) corresponds with the actual cognitive map asimplemented in mammalian brains the experimental agentsshould be able to treat all spatial dimensions equivalently

We propose that bats are a promising animal model for such astudy as they live in a true three-dimensional environment andhave the motor capabilities to follow arbitrary three-dimensionalpaths through this environment In particular the spatialmemory and orientation of nectar-feeding bats using both visualand echo-acoustic environment sensing have already been investi-gated in a series of studies (Thiele amp Winter 2005 Winter et al2004) Interestingly Winter and Stich (2005) note that the hippo-campus of nectar-feeding glossophagine bats is 50ndash100 larger insize than that of insectivorous members of the same family (Phyl-lostomidae) Although the reported results do not show evidencethat these animals treat vertical and horizontal dimensions differ-ently more specific experiments are necessary as the onesdescribed were not aimed at finding out the nature of the batsrsquocognitive map

Navigation bicoded as functions of x-y andtime

doi101017S0140525X13000514

James G Phillips and Rowan P OgeilSchool of Psychology and Psychiatry Monash University Clayton VIC 3800Australia

JimPhillipsmonashedu RowanOgeilmonashedu

Abstract Evidence from egocentric space is cited to support bicoding ofnavigation in three-dimensional space Horizontal distances and space areprocessed differently from the vertical Indeed effector systems arecompatible in horizontal space but potentially incompatible (or chaotic)during transitions to vertical motion Navigation involves changes incoordinates and animal models of navigation indicate that time has animportant role

Jeffery et al provide evidence that horizontal and vertical coordi-nates are processed by different systems to the effect that naviga-tion in a three-dimensional environment requires two codes Theauthors summarise evidence from allocentric navigation andsingle cell recording to support their position Although we aremore familiar with egocentric coordinate systems we agree withJeffery et al that bicoding is likely and note evidence from per-ceptual illusions neurological conditions studies of stimulusndashresponse compatibility and biomechanical constraints furthersupporting thisPerceptual illusions Distinctly different codes responsible for

horizontal and vertical distances are indicated by perceptual illu-sions In the horizontalvertical illusion two lines of equallength form an upside-down T Although of equal length the ver-tical line nevertheless appears longer perceptually (Avery amp Day1969) Such observations imply that horizontal and vertical dis-tances are processed differently and this is reinforced by thedifferential vulnerabilities of horizontal and vertical coordinatesystems to neurological disordersNeurological disorders Cerebro-vascular accidents (strokes)

affecting the right hemisphere particularly the parietal lobelead to hemineglect (Mattingley et al 1992) Hemineglect is a ten-dency to ignore the left side of space that occurs independently ofany deficits in the visual field These strokes affect the perceptionof space and the willingness to search and explore the left side ofspace Although hemineglect is a relatively common neurologicalsyndrome a comparable tendency to ignore the lower or upperzones of space is quite rare and the underlying pathophysiologyis not well understood (Shelton et al 1990) suggesting that differ-ent systems are responsible for horizontal and vertical coordinatesand this is reinforced by a consideration of effector systemsStimulusndashresponse compatibility The transition from a hori-

zontal plane to a three-dimensional coordinate system as occursin three-dimensional interfaces (Phillips et al 2005) poses poten-tial sources of incompatibility between stimulus and response(Worringham amp Beringer 1989) For instance in humanndashcompu-ter interfaces leftwards joystick or leftwards mouse motions bothindicate left cursor motion and remain so in three dimensionsNevertheless the use of motions to code for up can be arbitrarywith forward mouse movements coding for upwards cursor

Figure 1 (Peremans amp Vanderelst) (a) Two-layered environment with marked distinct places (b) three hypothetical representations of(a) where green = accurate travelled distance and orientation estimates blue = inaccurate travelled distance estimate and red = inaccuratetravelled distance and orientation estimates

motion on computer screens and forward joystick motion codingfor downwards cursor movements Indeed this arbitrariness isalso seen within animals with locomotor responses operating inopposition to trunkbody responses such that downward propul-sive movements code for up (flying swimming) or head-downmovements code for down (flying swimming)Biomechanical constraints Indeed navigation is necessarily a

corollary of activity and motion and there are biomechanical con-straints that argue for motion to be primarily coded in the horizon-tal plane Motion occurs within the context of a constantgravitational field that has implications for movement (Phillips ampOgeil 2010) For most mammals locomotion is roughly analogousto the oscillations of pendulums suspended vertically from hip orshoulder joints Such motions are mathematically predictableHowever there are potential sources of problems if motionsincorporate additional oscillators to create motion in the thirddimension If motion in the third dimension requires horizontallyoriented oscillators these physical systems have the potential tobe chaotic and unpredictable because the combination of hori-zontal and vertical oscillators approximates a pendulum (x oscil-lator) suspended on a rotating driving arm (y oscillator) Suchsystems are potentially chaotic (Baker amp Gollub 1996) in their be-haviour with system characteristics dependent upon speed andthe size of the y oscillator Given a potential unpredictabilityinherent in motion in three dimensions it should not be surprisingthat motion systems are primarily oriented in accordance withgravity and that coordinate systems are oriented perpendicularlyto gravity with other sources of information being more importantfor navigation than is altitudeCoordinate systems Navigation requires motion with coordi-

nates updated as time and location change For examplestudies of Monarch butterfly navigation during seasonal migrationhave demonstrated at the cellular level that the circadian systemintegrates information about navigating space based on time inaddition to an insectrsquos position relative to the sun (Reppert2006) Central to this is a time-compensated sun compasswhich likely receives direct input from the circadian system(Zhu et al 2008) to allow correction of flight direction and coor-dinates relative to changing light levels across the course of theday This time-compensated sun compass appears to be geneti-cally determined rather than learned (Reppert 2006) Hencewe feel the second code is more likely to involve a time com-ponent than an altitude component

Vertical and veridical ndash 25-dimensional visualand vestibular navigation

doi101017S0140525X13000526

David M W PowersCSEM Centre for Knowledge and Interaction Technology Flinders UniversityAdelaide SA 5001 AustraliaDavidPowersflinderseduauhttpflinderseduaupeopleDavidPowers

Abstract Does the psychological and neurological evidence concerningthree-dimensional localization and navigation fly in the face ofoptimality This commentary brings a computational and roboticengineering perspective to the question of ldquooptimalityrdquo and argues that amulticoding manifold model is more efficient in several senses and isalso likely to extend to ldquovolume-travellingrdquo animals including birds or fish

We think we live in a three-dimensional world so it is natural tothink that our representation of our environment would reflecta three-dimensional structure and be represented as such inour three-dimensional brain In fact we operate in at least afour-dimensional world with time being a dimension that alsoneeds representation in order to process events deal with

causality and so forth The traditional computing modelwhether the one-dimensional Turing Machine or the sequentialaddress space of a central processing unit (CPU) is to embed mul-tidimensional structures by arithmetic or referential mappings tolinear arrays Computer circuits and living brains also have astrong locally two-dimensional structure In the cortex thelayers of the brain represent both layers of complexity and amapping of time to depth with opportunity for increasinglydiverse associations to be formed The cochlear is essentially aone-dimensional representation physically coiled and temporallycoded for time and frequency The retina is essentially a two-dimensional representation with quaternary coding for frequencyand again an analog encoding of time and amplitude The vestib-ular system is a collection of essentially one-dimensional sensorsas described by Jeffery et al who note that there would be con-siderable complexity in producing a three-dimensional analog rep-resentation from thisAlthough the three-dimensional world nominally has volume

ldquolarger than a plane by a power of 32rdquo (sect 2 para 4) the verticaldimension is in general many orders of magnitude smaller thanthe distances navigated in the horizontal plane there is no needto encode it this way and even ldquovolume-travellingrdquo creaturestend to maintain a preferred orientation in relation to gravity asJeffery et al point out (sect 51 para 11) Animals also tend tooperate in strata being far from ldquounconstrainedrdquo In fact thereis arguably less constraint for a rat jumping around a three-dimen-sional maze or a squirrel jumping around an even more complexforest ndash for example they can move vertically in a way a bird ora fish cannot Similarly the ldquothree planesrdquo of directional infor-mation when viewed egocentrically combined with the verticaldimension ldquocharacterized by gravityrdquo (sect 2 para 4) provide anatural two-dimensional manifold which can be topologicallyexpanded locally when there is more information available as iswell known from self-organizing maps (von der Malsburg 1973)The constraints of gravity do not ldquoaddrdquo complexity but rathergive rise to both constraints and heuristics that reduce complexityThe visual system is also constrained by the locus of eyegaze and

the characteristic poses with respect to gravity Here we have two-dimensional sensors from which three-dimensional informationcould be reconstructed but again the evidence points to a 25-dimensional model (Marr 1982) But the two-dimensional planeof the retina tends to be near vertical whereas the two-dimen-sional plane of locomotion tends to be near horizontal and itwould seem natural that both should track egocentrically whenthese assumptions are invalid with discrepancies between thecanonical gravitational and egocentric navigational modelleading to errors that increase with the deviationJeffery et al have little to say about navigation focusing almost

exclusively on representation without much comment on aspectsof the reviewed experiments that relate to their actual titleWhen we consider vision and navigation in robotics whetherunderwater aerial or surface vehicles are used we tend to usea two-dimensional model coded with far more than just oneadditional piece of dimensional information We tend to trackand correct individual sensor readings to keep track of ourlocation based on simple models that predict based on locationvelocity and acceleration and we also attach information aboutthe terrain the temperature the sights and sounds and poten-tially also information about how hard or dangerous or howmuch energy is being expended These factors become inputsfor both algorithmic and heuristic components of the actualpath planning that is the key task in navigation One additionalfactor that is important here is the ability to deal with the worldat multiple scales with approximate paths being worked out fora coarse scale before zooming in recursively to plan the coarsestages in detail Even fully detailed two-dimensional maps leadto impossibly complex calculations of an optimal path so the sol-ution is to select a reasonable one at a coarse level based on high-level detail and then burrow down into the detail when neededHumans also have difficulty navigating more than two dimensions

and even 25 dimensions is not in general a good computer inter-face feature (Cockburn amp McKenzie 2001 2002)

In terms of Shannon Information the most frequent events orlandmarks or percepts are the least interesting and useful andKohonen Self-Organizing Maps (SOM) will tend to allocate areato its inputs in a monotonic increasing but sublinear way (p23

rather than -log p) for a gain in efficiency of representationwhilst the Zipfian Principle of Least Effort will ensure encodingso that the frequent data can be dealt with faster and sparse rep-resentation principles will lead us to store events that occur ratherthan placeholders for events that never occur or voxels that areseldom occupied as would be the case for a four-dimensionalmodel with a cell for every time and place In fact the anti-Shannon allocation of memory in a SOM means there is morereal estate available for the subtle details that correspond to thenext level of detail in our robotic model For the squirrel thismight be brain space for the vertical dimensions and horizontallayers of a forest and more importantly still the nutritional andterritorial dimensions The latter is an important fifth dimensionfor the squirrel marked by its own pheromones and those of itspotential mates and potential predators

Jeffery et al argue for a 25-dimensional model but allow thatthis could be a manifold with an egocentric normal rather than agravimetric plane They seem to prefer a mosaic model which isreminiscent of many robotic and vision models where extra detailsare linked in The manifold model would in many ways seem to fitbetter with current ideas of brain processing and self-organizationalthough a distributed mosaic of contextual associations with mul-tiple sensory-motor areas would be a highly plausible model forsurvival navigation

Foreshortening affects both uphill anddownhill slope perception at far distances

doi101017S0140525X13000538

Helen E RossDepartment of Psychology University of Stirling Stirling FK9 4LA ScotlandUnited Kingdom

herossstiracukwwwpsychologystiracukstaffstaff-profileshonorary-staffhelen-ross

Abstract Perceived slope varies with the viewing distance and isconsistent with the effects of foreshortening Distant viewing makesuphill slopes appear steeper and downhill slopes flatter than does nearviewing These effects are obvious to skiers and climbers inmountainous country They have also been measured in outdoorexperiments with controlled viewing distances There are many othersources of slope illusions

Jeffery et al discuss the human ability to estimate the slope of sur-faces and leave open the question of whether there is a dis-sociation between inaccurate conscious judgements of slope andaccurate motor actions on the slope They say that perceptualslope constancy exists ndash that is a surface appears to have a rela-tively constant slope regardless of viewing distance even thoughthe angle the slope makes with the eye (the optical angle)changes They also state that observers always overestimateslope and that this cannot be explained by foreshortening of dis-tance because that would make downhill slopes appear flatterSome of these statements can be challenged

To perceive a geographical scene correctly the two-dimen-sional image presented to the eye has to be turned into a three-dimensional image and related to gravity and the angle ofregard Gravity information is obtained through the vestibularsystem and postural cues and contributes to knowledge of theangle of regard Translation to a three-dimensional imagedepends on distance information which can be obtained fromvarious binocular and monocular cues at short distances and at

all distances from the probabilistic relationship between imagesand their physical sources (Yang amp Purves 2003) Local relation-ships between different parts of the terrain can produce largeslope illusions or ldquoanti-gravity hillsrdquo (Bressan et al 2003)perhaps by changing the probabilities

Outdoor observations and experiments using long viewingdistances show that apparent slope does vary with the viewingdistance and viewing height Distant uphill slopes appearsteeper than near uphill slopes Distant downhill slopes appearflatter than near downhill slopes and very distant downhillslopes appear to slope uphill Skiers at the top of a hill mayobserve skiers in the valley below apparently skiing uphillThese effects can all be explained by increased foreshorteningof distance at further distances (Fig 1) However when viewingdifferent parts of the terrain observers also vary their angle ofregard and the optical angle will change These effects are par-ticularly marked at short viewing distances and may contributeto different perceptual effects with close viewing

It is well known that uphill slopes appear steeper from a far thana near distance Alhazen used foreshortened distance in the 11thcentury to explain the vertical appearance of distant hills (Sabra1989) Ross (1974 p 73) found that skiers gave steeper verbalestimates of an uphill slope from a far viewpoint than when stand-ing on the slope Creem-Regehr et al (2004) varied the viewingdistance from the base of a hill to 70 meters away and found anincrease in slope estimates from the far distance Bridgemanand Hoover (2008) had observers stand on a paved hill with an11 degree slope and judge the slope when looking up to a coneplaced at 1 2 4 8 and 16 meters uphill from them their slopeestimates increased with the viewing distance Thus quite shortincreases in viewing distance are sufficient to cause an increasein apparent uphill slope at least when accompanied by a raisedangle of regard In this case the optical angle decreases with theviewing distance and cannot be used as an explanation of theincrease in apparent slope

The appearance of downhill slopes is more controversialUnlike uphill slopes downhill slopes can only be viewed with alowered angle of regard For evidence that downhill slopesappear too steep the authors cite Proffitt et al (1995) and Liand Durgin (2009) However Li and Durgin conducted a labora-tory experiment using very short viewing distances and found thatdownhill slopes appeared steeper when viewed further back fromthe edge of a drop The increased viewing distance wasaccompanied by a decreased optical angle and a less depressedangle of regard For downhill views a smaller optical angle isassociated with a steeper slope and this may explain the effectThe experiment does not compare uphill and downhill views ordeal with large distances Proffitt et al used real outdoor hillsviewed from the base or the top The hills had slopes of from 2to 34 degrees but their lengths were not specified The hillswere selected to be near footpaths on a university campus andto be sufficiently long that the top of the hill was ldquowell abovethe horizonrdquo when viewed from the base For an average viewer

Figure 1 (Ross) Distance foreshortening makes uphill slopesappear too steep (EF to EprimeFprime) and downhill slopes too shallow(AB to AprimeBprime) [Reprinted from Ross (1994) with permission]

with an eye-height of 15 meters the 2 degree hill would have tobe more than 43 meters in length but the steeper hills could beshorter The authors also conducted a virtual reality experimentin which they simulated the measurements for the real 5 degreehill and stated that it was 100 meters in length The outdoor hilllengths must have varied but are short in comparison with themountainous terrain experienced by skiers and hill walkers Theauthors found that slopes were overestimated both uphill anddownhill but very slightly more for the steeper downhill viewson visual and haptic measures (but not on verbal angular esti-mates) Similarly Ross (2006) found no difference betweenuphill and downhill views on verbal angular estimates foroutdoor slopes between 2 and 23 degrees with maximumviewing distances of 15 meters for the 23 degrees uphill viewand views of around 50 meters or more for the downhill viewsRoss (2010) varied the viewing distance to a bench on a 7degree downhill slope She had observers stand on the slope 20meters above the bench or view it from a window in a buildinggiving a viewing distance of 36 meters The verbal slope estimateswere significantly less steep from the higher and further view-point This shows that when viewing distances are well controlledthe flattening effect of a more distant viewpoint can be measuredover fairly short distances Ross (2010) also had observers makeangular estimates from the Wallace Monument 100 metersabove a flat plain The mean estimate for the flat plain was+645 degrees significantly steeper than the estimate of +093degrees given by observers standing on a horizontal footpathand demonstrating the apparent uphill slope of flat terrain whenviewed from a height

These errors have practical implications for skiers and mountai-neers who are trying to decide whether a distant slope is navigableand what effort will be required Large slope illusions are usuallynot detected as illusions (Bridgeman amp Hoover 2008) ndash unless evi-dence is available such as skiers gliding uphill or water runninguphill One can learn from experience to guess what is navigablebut experience does not seem to change the false appearance ofdistant slopes These observations cannot help the debate onwhether there is a dissociation between perception and actionbecause by the time the traveller reaches the distant slope his per-ception usually corresponds to reality

The problem of conflicting reference frameswhen investigating three-dimensional spacein surface-dwelling animals

doi101017S0140525X1300054X

Francesco Savelli and James J KnierimKrieger MindBrain Institute Johns Hopkins University Baltimore MD 21218fsavellijhuedu jknierimjhueduhttpkriegerjhuedumbiknierimlab

Abstract In a surface-dwelling animal like the rat experimental strategiesfor investigating the hippocampal correlates of three-dimensional spaceappear inevitably complicated by the interplay of global versus localreference frames We discuss the impact of the resulting confounds onpresent and future empirical analysis of the ldquobicoded maprdquo hypothesisby Jeffery and colleagues

At first glance the question of how the brain represents three-dimensional space appears to be a straightforward problemThis review by Jeffery and colleagues highlights instead thethorny nature of the question The main problem is that mostmammals are surface-dwelling animals and it becomes tricky todevise proper experiments to determine whether these organismsencode space as an isotropic volumetric representation when theymove along surfaces that extend in three dimensions Theldquobicoded maprdquo proposal of Jeffery et al suggests that three-

dimensional space is represented not as a true volumetric mapbut as a patchwork of diversely oriented local quasi-planarmaps Neurophysiological data informing this issue are themselvesfragmentary and limited to a few studies that employed comp-lementary strategies (eg with respect to the orientation of theanimalrsquos body) while addressing different functional correlatesof space (eg head-direction cells vs grid cells) Much exper-imental work is therefore needed to empirically evaluate thedebated theoretical scenariosThe implicit assumption in many experiments is that spatial

locations are represented in a global reference frame (eg the lab-oratory room) and that the local apparatus is merely a substratefor the animal to move (actively or passively) in and out of differ-ent locations in the global reference frame However it is increas-ingly clear from both behavioral and neurophysiological studiesthat hippocampal spatial representations are tied more stronglyto the local apparatus frame of reference than the global frameof reference when the two frames are translated relative to eachother (Knierim amp Hamilton 2011) Hence all experiments onthree-dimensional representations of space that employ surface-dwelling animals must address the problem of interacting refer-ence frames when the 3D volume of the global reference frameis sampled serially via surface-bound navigational epochs Forexample in an experiment not discussed in the target articleplace fields were bound to the behavioral track rather than thex y or z coordinates of the room when the track was moved todifferent locations in both horizontal and vertical axes of theroom (Knierim amp Rao 2003) These results suggest that thespatial representation was tied to the local reference frame ofthe track rather than to the room coordinates precluding anystrong interpretation regarding whether the cells representedtwo-dimensional or three-dimensional space in the global refer-ence frameOne key experiment considered by the authors involved rats

running up and down a helical path (Hayman et al 2011)Firing fields from grid cells and place cells repeated themselvesin corresponding positions of all or nearly all laps AlthoughJeffery et al interpret this result as evidence of a diminished rep-resentation of space in the z-axis this result could be also inter-preted as the laps sharing an identical representation under thecontrol of the frame of reference established by the recurringstereotypical lap This interpretation is strongly supported by astudy using planar circular and square spiral paths (Nitz 2011)Place fields recurred in corresponding positions of each spiralloop similar to the repeated firing on the different levels of thehelical track The sizes of the fields grew in lockstep with the con-centric loops except for those fields that recurred at the corners ofsquare spirals These findings suggest a form of pattern detectionoperating on the geometrical features of the lapThe second experiment with neural recordings considered by

Jeffery et al concerned rats climbing on a vertical pegboard(Hayman et al 2011) Grid cells fired in patterns of verticalstripes suggesting that the vertical wall was ldquoslicing intordquo a colum-nar extension of a classic two-dimensional grid pattern existing onthe contiguous floor (which was not sampled in this experiment)This is perhaps the strongest piece of evidence supporting theauthorsrsquo theoretical proposal but how would one test this modelrigorously One simple idea is to record the grids on the floorand then have the rat do the pegboard task as the pegboard ismoved to different locationsorientations relative to the floor Asthe pegboard slices into different columns the pattern of stripesshould be predictable based on the floor-based grid verticesbeing intersected by the pegboard However once again theinterplay of reference frames may prevent a simple answerPlace cells (Knierim amp Rao 2003 OrsquoKeefe amp Burgess 1996Siegel et al 2007) and grid cells (Derdikman et al 2009 Savelliamp Knierim 2012 Savelli et al 2008) are strongly influenced bylocal boundary manipulations and boundary-responsive neuronsare found in multiple hippocampus-related regions (Boccaraet al 2010 Lever et al 2009 Savelli et al 2008 Solstad et al

2008) If the floor-based grid becomes bound to the pegboard inthe first recording this grid may reset identically at the boundaryformed by the base of the pegboard regardless of where the peg-board is moved The vertical fields predicted to project over thepegboard from this grid would then remain unchanged tooThis is just one example of a general concern applicable tosimilar experimental designs in which a surface reference frameis moved relative to another

The authors make the provocative argument that even animalsthat truly navigate in three-dimensional volumes ndash birds batsinsects fish ndash are likely to use a bicoded strategy rather than avolumetric representation that is isotropic and metric in allthree axes The question can be addressed in these animalswithout the need of a surface substrate that may be in conflictwith a global reference frame A recent study (Yartsev amp Ula-novsky 2013) did not find elongation along a single common axisin three-dimensional place fields recorded from the hippocampusof freely flying bats seemingly in contrast with the bicoded maphypothesis Future experiments in the same species can test ifthis result applies to grid cells as well Regardless of theoutcome it will not be clear whether surface-dwelling creatureswill show the same results Understanding the proper referenceframe of the ldquocognitive maprdquo is one of the crucial problems thatmust be resolved in order to interpret the data from animalsrunning on surfaces in three-dimensional space and the referenceframes may differ across terrestrial arboreal and flying animals

Just the tip of the iceberg The bicoded mapis but one instantiation of scalable spatialrepresentation structures

doi101017S0140525X13000551

Holger Schultheis and Thomas BarkowskySFBTR 8 Spatial Cognition Universitaumlt Bremen 28359 Bremen Germanyschulthsfbtr8uni-bremende barkowskysfbtr8uni-bremendehttpcosyinformatikuni-bremendestaffholger-schultheishttpcosyinformatikuni-bremendestaffthomas-barkowsky

Abstract Although the bicoded map constitutes an interesting candidaterepresentation proposing it as the predominant representation for three-dimensional space is too restrictive We present and argue for scalablespatial representation structures as a more comprehensive alternativeaccount that includes the bicoded map as a special case

The question of how space including the third dimension is rep-resented in the human mind is doubtless of major importance tothe behavioral and brain sciences The bicoded map proposed byJeffery et al constitutes an interesting candidate representationHowever it seems unlikely that it captures more than one veryspecific type of the multitude of spatial representations availableto humans To assume that the bicoded map is more than aspecial case is to misjudge the nature of spatial representations

Even in one or two dimensions mental representations ofspatial information are usually neither as metric nor as consistentas assumed in the bicoded map More often humans use so-calledqualitative representations Such representations abstract from acontinuous property (eg distance) by partitioning it and dis-tinguishing only between classes that are defined by these par-titions (eg near and far in the case of distance see Forbus2011) Empirical behavioral (Knauff et al 1997) and neuroscien-tific (Sereno et al 2001) as well as computational (Krumnack et al2011) studies have shown that humans are prone to employingqualitative non-metric spatial representations Furthermorehuman spatial representations suffer from systematic distortionsthat lead to misrepresentations of for example distance (McNa-mara amp Diwadkar 1997) and orientation (Moar amp Bower 1983)According to these properties mental spatial representations are

often better viewed not as cognitivemaps but as cognitive collages(Tversky 1993) combinations and overlays of qualitative non-metric representations of parts and aspects of the overall rep-resented space In addition organization of human memory hasbeen found to favor representation structures that are economicin the sense of minimizing storage space and processing require-ments for representing a given body of knowledge both forsemantic knowledge generally (Collins amp Quillian 1969) and forspatial knowledge in particular (McNamara 1986 Stevens ampCoupe 1978)

Given these properties of human spatial representations itseems unlikely that the bicoded map constitutes more than oneout of many representations employed by humans A more com-prehensive view on human spatial representations that includesthe bicoded map as a special case is provided by our frameworkof scalable spatial representation structures (Schultheis amp Bar-kowsky 2011 Schultheis et al 2007) According to this con-ception representations are constructed such that they remainas simple as possible given the current task demands If the taskdemand changes the representations are adapted accordinglyand thus the representations scale to and with the current taskrequirements This on-demand scaling can occur with respect toseveral aspects of the representations

First scaling can occur with respect to the types of spatialknowledge that are represented (eg distance topology orien-tation) Evidence from psychological and artificial intelligenceresearch (reviewed in Schultheis et al 2007) indicates thatmental representations of spatial information are best viewed asbeing composed of several distinct knowledge-typendashspecific rep-resentation structures Such knowledge-typendashspecific represen-tation structures contribute to the scalability of spatialrepresentations If a certain situation provides or requires onlyknowledge about orientations between objects only a represen-tation structure specialized for representing orientation knowl-edge will be employed If the consideration of further types ofknowledge such as topology becomes necessary further represen-tation structures that are specialized for representing topologicalknowledge will be added to the overall spatial representation

Second each knowledge-typendashspecific representation is subjectto scaling such that (a) only task-relevant entities are included inthe representation and (b) the granularity of the representationthat is its ability to distinguish between different relationschanges with task demands Since the granularity of a represen-tation is directly related to how coarsely or finely the representedspatial continuum is partitioned spatial representations are alsoscalable in terms of how qualitative or metric they are

Finally spatial representations are scalable with respect to thenumber of dimensions that are encoded Although a two-dimen-sional representation may be considered the most commonform of spatial representation one-dimensional representationscan be sufficient and even superior to two-dimensional represen-tations in certain contexts (eg route representations MacEach-ren 1986)

When viewed in the framework of scalable representationstructures the conception of the bicoded map as the predominantmental representation appears too restrictive It seems implausi-ble to assume that the spatial plane is always represented metri-cally while the vertical dimension is always exclusivelyrepresented qualitatively which dimensions are represented andhow fine-grained the representation is depend on the demandsof the current spatial task For example when planning a biketrip the plane may not be represented but the vertical dimensionmay be instantiated by a fine-grained (perhaps metric) represen-tation of slope and elevation because this may constitute themost important information for planning the bike route On theother hand if required by the task humans may construct a fullthree-dimensional volumetric mental representation Researchon air traffic controllers for instance suggests that experiencedcontrollers form a continuous ldquofunctional picture of the momen-tary traffic situationrdquo that allows them ldquoto interpret very

complex dynamic constellations within a few secondsrdquo (Eyferthet al 2003 p 415)

That there is little empirical evidence for metric represen-tations of the vertical dimension or a three-dimensional volu-metric representation is therefore more indicative of thenecessity than of the ability to maintain and employ such rep-resentation structures For many spatial tasks it is sufficient toexclude construction of a representation of the vertical becausethe vertical information is (a) irrelevant to the task or (b) partlyredundant with horizontal information (eg in slanted terrainmoving in certain directions implies certain elevation changes)Due to this lack of a necessity to represent vertical informationin much detail (if at all) in many situations the data reviewed inthe target article do not allow unequivocally ruling out alternativerepresentations to the bicoded map Given the above consider-ations the framework of scalable representation structures pro-vides a much more convincing account of human mental spatialrepresentations than does the overly specific bicoded map

What is optimized in an optimal path

doi101017S0140525X13000563

Fraser T Sparksa Kally C OrsquoReillyb and John L KubieaaDepartment of Cell Biology The Robert F Furchgott Center for Neural andBehavioral Science State University of New York ndashDownstate Medical CenterBrooklyn NY 11203 bCenter for Neural Science New York UniversityNew York NY 10003

neurosparksgmailcom kallycogmailcom jkubiedownstateeduwwwneuraldynamicsca httpcoronaradiatanet

Abstract An animal confronts numerous challenges when constructing anoptimal navigational route Spatial representations used for pathoptimization are likely constrained by critical environmental factors thatdictate which neural systems control navigation Multiple codingschemes depend upon their ecological relevance for a particular speciesparticularly when dealing with the third or vertical dimension of space

Modeling the neural systems used for navigating between twolocations is ecologically valid when the specific constraints of anenvironment on a species are accounted for Navigational routesmay be optimized by certain factors in one species and by differ-ent factors in others Therefore models of navigational systemswithin terrestrial arboreal aquatic or aerial species shouldaccount for specific factors that dictate route optimizationJeffery et al have presented an argument for bicoding wherespace in the plane of locomotion is represented independentlyfrom that in the orthogonal axis Although we agree that the evi-dence presented points toward this coding scheme concept rep-resentation of space may be encompassed by ldquomulti-codingrdquowhere multiple systems (not limited to two) compute ecologicallyvalid representations to optimize navigation

What is the most relevant factor in optimizing a three-dimen-sional route Ignoring ecological validity and all factors beingequal the simplest answer is distance The optimal route is theshortest distance between two points which is a straight line Intwo-dimensional surface navigation a number of neural compu-tational models propose mechanisms by which the hippocampalformation could compute a straight path (Blum amp Abbott 1996Burgess amp OrsquoKeefe 1996 Muller amp Stead 1996) Extrapolatingfrom these models to account for the vertical dimension andmaintaining shortest distance as the most optimized factor calcu-lating the ldquostraight linerdquo is performed by treating the z-axis as ana-logous to the x- and y-axes

Treating the vertical dimension as equivalent to the two hori-zontal dimensions is not always possible for terrestrial animalswhen the straight line would require ecologically invalid oper-ations such as flight or tunneling To avoid this issue Jefferyet al argue that in a three-dimensional environment the

optimal shortest path (straight line) is calculated by using the tan-gents-to-ground contours However other factors that may comeinto play when optimizing a route should be taken into consider-ation for example metabolic expenditure and journey timeOn flat unobstructed terrain the straight-line distance between

two locations will yield the path requiring minimal time and energy(metabolic optimization) However given a terrain with a verticaldimension the energy expenditure for climbing or descending iscomplicated and relates to instantaneous slope Metabolic rate inhumans walking on an adjustable incline is a function of speedand slope For instance locomotion on a 10 degree slope willapproximately double metabolic rate compared with movementat the same speed on a level surface (Silder et al 2012) Similarlythe metabolic rate of horses on a 10 degree slope increases byapproximately 25 times (Eaton et al 1995) Additionallybecause climbing is slower than walking on the horizontal shortestpath distance is not equivalent to shortest path time For speciesthat optimize metabolic cost models solely optimizing shortestpath distance are not ecologically valid when metabolic cost is aprimary factor in route navigation For a terrestrial animal linearor tangential distance is a poor optimizer for routes in three-dimensional space Though we strongly concur with the overallconclusion of Jeffery et al our conclusions are based on the inap-propriate simplification that the shortest path is the optimal pathHowever the shortest path in three dimensions may be opti-

mized in some species A number of studies done with spatial rep-resentations in the barn owl (Tyto alba) offer insight into how thevertical dimension is processed and add support to the concept ofmulti-coding of three-dimensional space (Euston amp Takahashi2002 Takahashi et al 2003) Barn owls are known for theirability to hunt in total darkness (Konishi 1973 Payne 1971) andtheir navigation is guided by a topographical representation ofauditory space located in the external nucleus of the inferior col-liculus (Knudsen amp Konishi 1978a 1978b) Neurons within thisauditory space map have discrete spatial receptive fields notunlike place cells in the hippocampus that result from the compu-tation of inter-aural differences in the level and time-of-arrival ofsounds The most optimal route for the barn owl while hunting isthe route that minimizes distance and time to get to the preyThe auditory spatial map of the barn owl is used as a model of

mammalian auditory localization and navigation and it may be anexample of how information about the vertical dimension can beused in conjunction with two-dimensional representations ofspace Though terrestrial species may not use auditory infor-mation as primary spatial information arboreal mammals andprimate species likely do use this information in forming and uti-lizing three-dimensional representationsIn conclusion avoiding a steep direct path by taking a flat cir-

cuitous route is supported by optimizations other than straight-line distance In fact given the arguments presented here foradding metabolism and time into the optimization of a particularroute we suggest that the shortest path in three-dimensional aswell as two-dimensional terrains may or may not be a factorused for route optimization A multi-coding and multi-systemscheme for spatial representations in three dimensions is anattractive concept to explain how ecologically relevant factorscan contribute to spatial processing

Grid maps for spaceflight anyone They arefor free

doi101017S0140525X13000575

Federico Stellaa Bailu Sib Emilio Kropffc and AlessandroTrevesadaCognitive Neuroscience SISSA 34136 Trieste Italy bDepartment ofNeurobiology Weizmann Institute of Science Rehovot Israel cLaboratory ofNeuronal Plasticity Leloir Institute CABA C1405BWE Buenos Aires

Argentina dCentre for Neural Computation Norwegian University of Scienceand Technology NO-7491 Trondheim Norwayfstellasissait bailusiweizmannacilkropffgmailcomalesissaithttppeoplesissait~alelimbohtml

Abstract We show that given extensive exploration of a three-dimensional volume grid units can form with the approximateperiodicity of a face-centered cubic crystal as the spontaneous productof a self-organizing process at the single unit level driven solely by firingrate adaptation

In the target article Jeffery et al suggest that a fully three-dimen-sional representation of space may have not been found in anumber of studies because of the intrinsic computational com-plexity of 3D representations which would make the cost ofsetting them up too high and not because of anything to dowith the spatial behaviour of the species such as rats and primatesused in those studies If so genuine representations of 3Dvolumes would be unlikely to be revealed even when movingbeyond surface-dwelling species towards flying or marineanimals who appear capable of experiencing 3D space morethoroughly through three-dimensional navigation

Clearly managing directional information in three-dimensionalspace offers more challenges than on the plane (Finkelstein et al2012) Nevertheless how this may affect the development of agrid representation in three dimensions depends on the mechan-isms at work to generate it If such mechanisms were to relyheavily on head direction inputs then Jeffery et al would havea point If not there might be space for surprises

Since 2005 we have been analysing a model for grid cell for-mation (Kropff amp Treves 2008 Treves et al 2005) based on aself-organisation process driven solely by firing rate adaptation(head direction information and recurrent connections areneeded only to align the grids along common axes Si et al2012) The model produces grids in two dimensions spon-taneously Individual grid cells average in time the content ofspatially modulated inputs which can be very generic and neednot require any specific computation The emergence of a gridin the firing rate map is induced by the animalrsquos exploration ofspace and the final appearance of the grid its structuredepends on the way the environment is explored (Si et al 2012)and on the topology of the space itself (Stella et al 2013)

What would this model predict in three dimensions if the unitsreceive broad spatial inputs modulated also in the third dimen-sion Individual grids are expressed by the feedforward inputs

in the basic model and the same inputs could ldquocarryrdquo both two-dimensional crawling grids and three-dimensional flying gridsWe have looked at how the very same model behaves whenexpanding the original square environment into a cubic one andallowing the simulated animal to fly around for the time usuallyrequired for the two-dimensional exploration to generate goodgrid units

As in two dimensions the model produces a regular tiling ofspace Its units develop grid fields positioned at the vertices of alattice uniformly filling the available volume By computing the3D autocorrelogram of the spatial activity of these units (Fig 1)it is apparent that the configuration reached by the fields is theso-called face centered cubic (fcc) In this configuration eachfield is surrounded by 12 other fields and all pairs of neighbouringfields have roughly the same distance

Our model shows how grids in two and three dimensions (or inany number of dimensions) may be produced starting from thevery same principles of auto-organization without increasingcosts in higher dimensions Adaptation provides the means toshape regular forms out of rough and unpolished spatial inputsand it does so regardless of the topology of the external environ-ment Without a spatial behaviour engaging the full three-dimen-sional environment however no 3D grid units would appear asthere is no hard-wired or ad hoc designed structure in ourmodel to support them

In the words of Jeffery et al our model seems to indicate thatthe absence of three-dimensional grids in rats has an ontogeneticcause (target article sect 52 para 2) Rats do not possess three-dimensional grids because alas they have never learned to fly

What counts as the evidence for three-dimensional and four-dimensional spatialrepresentations

doi101017S0140525X13000587

Ranxiao Frances Wang and Whitney N StreetDepartment of Psychology University of Illinois at UrbanandashChampaignChampaign IL 61820franceswcyruspsychillinoisedu street1illinoiseduhttppublishillinoisedufranceswang

Abstract The dimension of spatial representations can be assessed byabove-chance performance in novel shortcut or spatial reasoning tasksindependent of accuracy levels systematic biases mosaicsegmentationacross space separate coding of individual dimensions and referenceframes Based on this criterion humans and some other animalsexhibited sufficient evidence for the existence of three-dimensional andor four-dimensional spatial representations

How can one prove that an animal has three-dimensional spatialrepresentations For navigation in two-dimensional space (eghorizontal plane) the classical paradigm is the novel shortcuttest If an animal traverses a route and then returns home by anovel shortcut this is taken as sufficient (but not necessary) evi-dence of a 2D spatial representation A variant of the novel short-cut test is the detour test where a familiar route is blocked and theanimal must navigate around the obstacle to reach the destinationAlthough novel shortcutting is usually attributed to path inte-gration rather than cognitive mapping both mechanisms require2D spatial representations and therefore the conclusion regard-ing dimensionality holds in both cases

Based on this logic novel shortcutting in three-dimensionalspace is sufficient evidence for 3D spatial representations If ananimal traverses a route in three-dimensional space and returnshome through a novel shortcut or navigates around an obstacleto reach the goal then that animal possesses a 3D spatial rep-resentation Among the four candidate models only the

Figure 1 (Stella et al) The three-dimensional autocorrelogramof a sample unit developing an approximate face-centered cubic(fcc) 3D firing rate map

volumetric map supports such behavior Animals using the surfacemap and the extracted flat map may take shortcuts within theencoded surface but they cannot return to the appropriateelevation once they are displaced along that dimension Abicoded map with non-spatial (eg contextual) coding of thethird dimension allows shortcuts within the surface(s) of the twospatial dimensions but the animal has to perform a randomsearch along the third dimension until it happens to reach thedesired context because by definition the contextual codingwithout spatial mapping provides no information about the direc-tion and distance an animal should travel to reach the goal contextfrom a novel context

The ability to perform novel shortcuts does not necessarilymean high accuracy or efficiency Performance in navigationtasks varies depending on the species the perceptual informationavailable and properties of the trajectory (Loomis et al 1993Wang amp Spelke 2003) Differences in accuracy may be due tothe quality of the spatial representation and spatial processingHowever lower accuracy cannot be used as evidence againstthe presence of 3D spatial representations Instead the criterionfor possessing a 3D representation should be above-chance per-formance in the third dimension Similarly biases occur inspatial judgments of lower dimensions (eg Huttenlocher et al1991 Sampaio amp Wang 2009) and veridical information is notrequired for spatial representations Hence systematic biasesare not evidence against 3D representations

Moreover spatial representations do not require integrationacross dimensions and none of the four models makes thisassumption Multiple dimensions can be encoded separately forexample as coordinates for each dimension as long as the infor-mation can be used together for spatial processing Thereforeseparation of one or more dimensions is not evidence against3D spatial representations Integration across space is also notrequired and fragmentation has been shown in 2D spatial rep-resentations (Wang amp Brockmole 2003) All four models areopen about whether the entire space is represented as a wholeor is divided into piecessegments and all can accommodateboth mosaic and integrated representations Thus segregationof space is not informative on the dimensionality of the spatialrepresentations

Finally the dimensionality of spatial representations for naviga-tion needs to be considered in a broader theoretical framework interms of reference frames and sensory domains Flexible efficientnavigation does not require an allocentric map For example anegocentric updating system can provide the same navigationalcapabilities as an allocentric cognitive map (Wang 2012 Wangamp Spelke 2000) It is also important to consider visual propriocep-tive and motor command information for self-motion estimationwhich may provide simpler and more efficient computation thanthe vestibular signals (Lappe et al 1999 Wang amp Cutting 1999)

As Jeffery et al discuss in the target article the dimensionalityquestion has been extended to four dimensions for humansAlthough humans probably did not evolve for navigation in four-dimensional space there are no known inherent constraints inimplementing 4D spatial representations with neurons andneural connections and the existing brain structure may be uti-lized to accommodate higher-dimensional spatial representationsTherefore 4D spatial representations cannot be dismissed a prioribased on evolution alone and should be treated as an empiricalquestion

A few studies have examined human 4D spatial representationsusing variations of the shortcutting task and spatial judgmenttasks Aflalo and Graziano (2008) showed that humans canperform path integration in four-dimensional space with extendedtraining and feedback Because the movements were orthogonal(90deg rotations) and determined by the observer updating ateach translation and rotation involved only one and two dimen-sions respectively and was relatively easy to compute algebrai-cally within an egocentric reference frame Neverthelessrepresentation of a 4D vector was required hence the study

provided some evidence of 4D representations Other studiesused virtual reality techniques to create visual simulations of 4Dgeometric objects (hyper-tetrahedron) and showed that observerscould judge the distance between two points and the anglebetween two lines embedded in the 4D virtual space (Ambinderet al 2009) Observers could also estimate novel spatial propertiesunique to 4D space such as the size (ie hyper-volume) of virtual4D objects (Wang in press) providing further evidence of 4Dspatial representationsIn summary an animal possesses three-dimensional spatial rep-

resentations if it can exhibit above-chance performance along thethird dimension in a novel shortcut or detour test in 3D spaceBased on this criterion humans (eg Wilson et al 2004) rats(eg Jovalekic et al 2011) and possibly ants (Grah et al 2007)all encode spatial information about the vertical dimension thatis beyond a surface map extracted flat map or bicoded mapwith contextual coding for the third dimension Thus thesespecies possess some form of 3D spatial representations for navi-gation which may vary in quality and format across species andthe individual dimensions may be represented separately and dif-ferently Empirical studies of four-dimensional path integrationand spatial judgments suggest that human 4D spatial represen-tations are possible although the condition and properties ofsuch representations require further investigation

Are all types of vertical information createdequal

doi101017S0140525X13000599

Steven M Weisberg and Nora S NewcombeDepartment of Psychology Temple University Philadelphia PA 19122smweistempleedu newcombetempleeduwwwspatiallearningorg

Abstract The vertical component of space occurs in two distinct fashionsin natural environments One kind of verticality is orthogonal-to-horizontal(as in climbing trees operating in volumetric spaces such as water or air ortaking elevators in multilevel buildings) Another kind of verticality whichmight be functionally distinct comes from navigating on sloped terrain (asin traversing hills or ramps)

Jeffery et al propose that vertical space is coded using a differentset of mechanisms from the head direction place and grid cellsthat code horizontal space In particular the authors present evi-dence that grid cells do not fire periodically as an organismmoves along the vertical axis of space as they do along the horizon-tal axis and that the head direction system uses the plane of loco-motion as its reference frame But a challenge for the bicodedmodel is whether it can account similarly for the different typesof information that can be specified in the vertical dimension Weargue that there are two markedly different components of verticalspace that provide entirely different spatial information and there-fore are represented differently orthogonal-to-horizontal (as inmultilevel buildings and volumetric space) and sloped terrainThe vertical dimension when orthogonal-to-horizontal can

specify contextual information that can be used to reference theappropriate planar map (ldquowhat floor am I onrdquo) as well as body-orientation information (ldquowhich way is the axis of gravityrdquo) Theformer is likely coded categorically in terms of low highhigher and so forth as described by Jeffery et al in the targetarticle Theoretically this information functions no differentlyfrom the contextual information provided by rooms with differ-ently colored walls or by the floor numbers outside elevatorsThe cue (wall color or floor number) specifies the planar cognitivemap to which the organism should refer The vertical dimensioncan also specify body-orientation information The organismconstantly refers to the plane of its body with respect to

earth-horizontal as a source of input for the direction of gravityThese cues combine for an organism to determine its elevationwith respect to the ground

In contrast none of this information is available in slopedenvironments Terrain slope can provide a directional cue by spe-cifying compass-like directions (ie up =North) which couldthen orient or augment the planar cognitive map Note that thisprocessing is distinct from orientation with respect to the direc-tion of gravity which provides no directional information for thehorizontal plane Recent work conducted in our lab shows thatslope is very useful for locating the goal in an otherwise ambiguousenvironment and that the directional information it provides iscoded with respect to onersquos own body (Weisberg et al 2012)Our data suggest that unlike North which is an invariant direc-tional cue that does not change based on the direction onefaces terrain slope is coded preferentially as uphill or downhilldepending on onersquos current facing direction It is unclear howthe vertical dimension with respect to onersquos own body couldprovide similar directional information

Data from our lab also suggest that the way slope informationaugments the planar map is computationally different from theway the orthogonal-to-horizontal dimension provides context Ina replication of Restat et al (2004) we found that participantswere able to use unidirectional-sloped terrain to make more accu-rate pointing judgments and sketch maps compared to the sameenvironment on a flat terrain However this was only the case for asimple environment In a complex environment only participantswith a high self-reported sense of direction used the slope infor-mation while low self-reporters did no better than in the flat con-dition (Weisberg et al 2012)

Neither of these experimental effects arises from informationprovided by the vertical dimension as orthogonal-to-horizontalThat is the participants in these experiments need not encodetheir elevation with respect to the ground but only the directionspecified by the gradient of the sloped terrain The goals of theorganism modulate which type of vertical information itencodes As further evidence of their dissociability considerhow the two possible reference frames might interact (see Fig 1)

For example an organism navigating along a sloped terraincould have a categorical representation of the vertical dimensioneither as it refers to the organismrsquos own body (the gradient atleft in the figure) or as it refers to earth-horizontal (the gradientat right) The rat on the left is unconcerned with the increase inelevation and is similar to the rat on the vertical trunk of thetree in Figure 12 of the target article The rat on the righthowever is encoding its elevation with respect to some low orhigh point Presumably Jeffery et al would predict that a ratwould encode the vertical dimension with respect to its ownbody as in the figure on the left But in doing so they may be dis-counting the information provided by the terrain gradient

By considering the possible representations of the vertical com-ponent of space one can characterize the qualities of each

representation separately For example are each of the resultantrepresentations derived from sloped terrain and volumetric spacenon-metric or do some of them contain metric properties If theyare non-metric are they functionally different susceptible to thesame distortions Whether these representations become inte-grated and can be used together is an empirical question butexploring these possibilities creates a richer picture of three-dimensional spatial cognition

ACKNOWLEDGMENTSWork on this commentary was supported by grant no SBE-1041707 from the NSF to the Spatial Intelligence and LearningCenter

Map fragmentation in two- and three-dimensional environments

doi101017S0140525X13000605

Homare Yamahachi May-Britt Moser and Edvard I MoserKavli Institute for Systems Neuroscience Norwegian University of Science andTechnology NO-7489 Trondheim Norway

homareyamahachintnuno maybmntnunoedvardmoserntnunohttpwwwntnunocbm

Abstract The suggestion that three-dimensional space is represented by amosaic of neural map fragments each covering a small area of space in theplane of locomotion receives support from studies in complex two-dimensional environments How map fragments are linked which braincircuits are involved and whether metric is preserved across fragmentsare questions that remain to be determined

Jeffery et al present an inspiring overview of behavioural and neu-robiological studies of neural map formation and navigation inthree-dimensional environments Based on their review of the lit-erature they suggest that three-dimensional space is ldquobicodedrdquomeaning that different neural mechanisms are used to mapenvironments in the horizontal and vertical planes Only the hori-zontal plane has a metric for distance and direction The suggesteddual coding scheme raises several interesting questions

The authors propose that three-dimensional space is encodedas a mosaic of contiguous map fragments These fragmentsneed not be horizontal but can in principle have any orientationrelative to the Earthrsquos surface The proposed fragmentation isreminiscent of the decomposition of grid cell maps that takesplace in complex environments in two-dimensional space A fewyears ago Dori Derdikman and colleagues measured the effectof environmental compartmentalization on the spatial periodicityin grid cells (Derdikman et al 2009) Rats were trained to run in azigzag fashion through 10 consecutive hairpin-shaped flat corri-dors in a large square enclosure The compartmentalization ofthe recording environment disrupted the global grid patternobserved in the same cells in the absence of the corridorsWhen internal walls were inserted separate maps were formedfor each corridor Each time the rat turned from one alley tothe next the grid map was reset and a new sequence of gridfields unfolded The internal walls broke the grid map intosmaller maps one for each compartment A similar pattern ofsub-maps is visible in entorhinal neurons during navigation inother multi-alley environments (Frank et al 2000) The sharptransitions at the turning points in these mazes appear to takeplace simultaneously in a coherent manner in grid cells andplace cells (Derdikman et al 2009) Taken together the studiessuggest that complex real-world environments are not rep-resented by a single universal map Instead the brain has multiplecontiguous grid cell and place cell maps each covering a smalluninterrupted space within the larger environment (Derdikmanamp Moser 2010) As argued by Jeffery and colleagues this

Figure 1 (Weisberg amp Newcombe) Two possible representationsof the vertical dimension ndash one specified with respect to the plane ofthe organismrsquos body the other specified by the organism travellingalong a sloped surface and therefore changing its elevation Whichof these representations would be predicted by the bicodedmodel Are they both categorical If so are they functionallysimilar (After Jeffery et al)

fragmentation of the cognitive map is probably not limited to thehorizontal plane Fragmentation may be an efficient mechanismfor mapping complex environments both within and acrossplanes

The fragmentation of the spatial map introduces new issueshowever How are fragments ldquogluedrdquo together at transitionpoints How does the brain adjust for changes in the slope ofthe terrain between segments Is metric information carriedover from one fragment to the next Can animals path-integrateacross segments It is possible as proposed by the authors thatthe integration of the fragments is solved by neural systems andcomputational algorithms not involving the currently knownspatial cell types in the hippocampus or entorhinal cortex Inorder to understand how animals navigate in real-world environ-ments we need to identify the factors that elicit fragmentationand the mechanisms that link one fragment to the next

We agree that some observations speak in favour of partlyspecialized mechanisms for encoding space in horizontal and ver-tical planes at least in terrestrial animals This includes the pres-ence of separate cell populations for encoding of pitch and yaw inthe lateral mammillary nucleus of the rat (Stackman amp Taube1998) as well as the finding that head direction cells in ratsrespond exclusively to variations in the horizontal (yaw) planeor the plane of locomotion if the surface is tilted (Stackmanet al 2000) The proposed segregation receives further supportfrom a study demonstrating less sensitivity to the vertical dimen-sion in place cells and grid cells in rats (Hayman et al 2011) Asindicated by Jeffery et al it remains possible that vertical modu-lation was not expressed in that study because the animals main-tained a horizontal orientation during the vertical movementMoreover with a predominance of horizontal movements wecannot exclude that horizontal and vertical dimensions differedwith respect to availability of metric cues and that precise fieldswould emerge also in the vertical direction if sufficient infor-mation were available Jeffery et al also acknowledge thatprecise representation of the vertical dimension may require aminimum of prior experience with active movement in three-dimensional environments Most laboratory rats lack such experi-ence Collectively the limited existing data raise the possibilitythat horizontal and vertical dimensions rely on somewhat segre-gated mechanisms but the findings do not rule out that experi-ence contributes to such segregation or that the brain as awhole has an integrated representation of volumetric spaceMore experimental work is clearly needed to settle these issuesThe target article of Jeffery et al takes important steps towardsidentifying critical questions that must be addressed

Does evidence from ethology support bicodedcognitive maps

doi101017S0140525X13000617

Shane Zappettini and Colin AllenProgram in Cognitive Science and Department of History amp Philosophy ofScience Indiana University Bloomington IN 47405szappettindianaedu colallenindianaeduhttpmypageiuedu~szappetthttpmypageiuedu~colallen

Abstract The presumption that navigation requires a cognitive map leadsto its conception as an abstract computational problem Instead of loadingthe question in favor of an inquiry into the metric structure andevolutionary origin of cognitive maps the task should first be toestablish that a map-like representation actually is operative in realanimals navigating real environments

ldquoThe core question for animal navigationrdquo according to Jefferyet al ldquoconcerns what the reference frame might be and how the

coordinate system encodes distances and directions within theframerdquo (sect 2 para 2) To answer this question they hypothesizethat a bicoded cognitive map underlies the navigational capacitiesof animals Their presumption that navigation requires a cognitivemap leads them to conceive of navigation as an abstract compu-tational problem that requires a representational schema powerfulenough to encode geometrical features of the familiar environ-ment of the navigating animal We argue that this presumptionleads them to ignore important ethological evidence against cog-nitive mapsThe work of ethologists is especially relevant to the problem of

navigating in a three-dimensional world Ethologists have pro-duced concrete methods for testing the cognitive map hypothesisduring navigational tasks Despite their passing references toethology Jeffery et al fail to note that most ethologists believethat the experimental evidence runs against the hypothesis thatcognitive maps underlie animal navigation (eg Bennett 1996Dyer 1991 Wray et al 2008 Wystrach et al 2011) In one suchexperimental paradigm the ldquodisplacement experimentrdquo ananimal is moved prior to navigating to a familiar goal in a familiarrange (eg see Wehner amp Menzel 1990) If the animal were totake a novel path from the displacement point to the goal thatwould count as evidence for map-like navigation However typi-cally the animal assumes a heading based on some other kind ofinformation providing evidence that the animal is using naviga-tional systems (eg landmark learning path integration systema-tic search behavior or a combination thereof) whoserepresentational structure does not preserve the geometricalproperties assumed by the cognitive map hypothesis Thesestudies unlike those to which Jeffery et al refer propose alternateexplanations that account for the navigational abilities of animalswithout presupposing the kinds of complex computation impliedby a cognitive map Simpler navigational strategies (eg visualsnapshot matching) have been proposed to explain navigation toa goal (Judd amp Collett 1998) with landmarks serving to providecontext-dependent procedural information rather than geometricpositional information (Wehner et al 2006) This work shows thatan animal may not need to abstract higher-order spatial character-istics of its environment in order to successfully navigate in itHowever without first determining what kind of information ananimal is actually using while navigating it is premature to beinvestigating the nature of the metric properties of the presumedcognitive mapWe suggest that reconceiving the problem of navigation in a

three-dimensional world in ethological terms would shift thefocus from describing the computational complexity of the geo-metric properties of the navigable environment to one whereinthe information used by the animal to determine its location rela-tive to some goal is investigated under ecologically valid con-ditions By failing to consider the alternative explanations ofhow animals navigate in real environments Jeffery et alrsquosbicoded map hypothesis puts the hypothetical cart before the evi-dential horseHaving ignored evidence against cognitive maps in animals

Jeffery et al appeal to human behavioral evidence suggestingthat the representations utilized in navigation preserve theplanar properties of the navigation environment For examplethey summarize several studies indicating that humans are rela-tively poor at estimating the height of landmarks in a familiarenvironment (sect 31 para 7) and other studies suggestingthat people overestimate the steepness of hills (sect 32 para3) They mention yet another study suggesting that people dopoorly when attempting to estimate the direction of objects in amultilayer virtual environment (sect 33 para 4) What is notclear is how these kinds of judgments are related to the infor-mation a navigating animal may use in an online navigationaltask Jeffery et al fail to acknowledge the possibility that thesekinds of distance judgments (and their failures) may not be evi-dence for a bicoded cognitive map precisely because the infor-mation contained in such a representation may not be that used

by a navigator in a real navigational task Differential performanceon vertical versus horizontal distance of objects though perhapshaving import for spatial perception in general may simply beirrelevant to navigation

We applaud Jeffery et al for acknowledging that in general theproblem of navigation is an evolutionary one However we findthe possibility they endorse to be problematic They argue thata fully three-dimensional map has never evolved on thegrounds that ldquo[it] is possible that the cost-benefit ratio of therequired extra processing power is not sufficiently greatrdquo (sect52 para 4) and they suggest that a lower dimensional map-likerepresentation supplemented with unspecified encoding heuris-tics could achieve successful navigational behavior at lower costParallel reasoning suggests that an even lower dimensional rep-resentation (not even a map at all) augmented with appropriateheuristics would have evolved under similar cost constraintsThat is to say given their assumption that evolution works to mini-mize the cost of a trait (ie the cost of a more complex compu-tational system) it becomes plausible that selection would act toproduce ever less costly representational systems for navigationas long as those systems are capable of approximating the sameresult This would be consistent with the computationallysimpler navigational systems found by ethologists to be sufficientfor animals to navigate In our view Jeffery et alrsquos hypothesis as tothe evolutionary origins of the bicoded cognitive map suffers froma deficiency common in attempts to connect the presence of a traitwith a story about selection history ndash that is that mere consistencybetween the presence of some trait and some hypothetical selec-tion pressure is the sole measure for plausibility They have pro-vided a list of how-possibly stories (two of which they rejectone they accept) when what is needed is a how-actually expla-nation that is properly grounded in the ethological evidenceFinally we agree that a fully volumetric map may not haveevolved but perhaps because no cognitive map at all is neededfor an animal to navigate in a three-dimensional world

Authorsrsquo Response

A framework for three-dimensional navigationresearch

doi101017S0140525X13001556

Kathryn J Jefferya Aleksandar Jovalekicb

Madeleine Verriotisc and Robin HaymandaDepartment of Cognitive Perceptual and Brain Sciences Division ofPsychology amp Language Sciences University College London London WC1H0AP United Kingdom bInstitute of Neuroinformatics University of Zurich CH-8057 Zurich Switzerland cDepartment of Neuroscience Physiology andPharmacology University College London London WC1E 6BT UnitedKingdom dInstitute of Cognitive Neuroscience Alexandra House LondonWC1N 3AR United Kingdom

kjefferyuclacuk ajovalekiciniphysethzchmadeleineverriotisuclacuk rhaymanuclacukwwwuclacukjefferylab

AbstractWe have argued that the neurocognitive representationof large-scale navigable three-dimensional space is anisotropichaving different properties in vertical versus horizontaldimensions Three broad categories organize the experimentaland theoretical issues raised by the commentators (1) frames ofreference (2) comparative cognition and (3) the role ofexperience These categories contain the core of a researchprogram to show how three-dimensional space is representedand used by humans and other animals

R1 Introduction

Scientists typically approach the problem of how space isrepresented in the brain using the vehicle of their own pre-ferred model species whether pigeon rat human or fishTherefore it is impressive to see in the commentaries howmany issues cut across species environments descriptivelevels and methodologies Several recurring themes haveemerged in the commentatorsrsquo responses to our targetarticle which we address in turn These are (1) Framesof reference and how these may be specified and related(2) comparative biology and how representations may beconstrained by the ecological characteristics and constraintsof a given species and (3) the role of developmental andlife experiences in shaping how representations are struc-tured and used In the following sections we explore eachof these themes in light of the commentatorsrsquo views andoutline research possibilities that speak to each of these

R2 Frames of reference

A fundamental issue in spatial representation is the frameof reference against which positions and movements arecompared The concept of frame is a mathematical con-struct rather than a real cognitive entity but it provides auseful way of organizing important concepts in spatial rep-resentation However the crisp formalism that we havedeveloped in order to treat space mathematically maybear scant resemblance to the brainrsquos actual representationwhich has been cobbled together using patchy multisen-sory variable probabilistic and distorted informationNevertheless it is useful to review the ingredients of aperfect map before trying to understand how the imper-fect and real-world brain may (or may not as Zappettiniamp Allen argue) have approximated this functionalityMaps can be either metric or non-metric (Fig R1) the

former incorporating distance and direction and the latter(topological maps ndash see commentary by Peremans amp Van-derelst) using only neighbourhood relationships withoutreference to precise distances or directions That saidthese parameters are necessarily loosely incorporated intothe map or else the relationship with the real worldbecomes too disconnected for the map to be useful Wefocused on metric maps in our analysis because of theclear evidence of metric processing provided by therecent discovery of grid cells However Peremans amp Van-derelst describe how an initially topological map couldpossibly be slowly metricized through experienceA metric spatial reference frame to be useful needs two

features an origin or notional fixed point within the framefrom which positions are measured and some way of spe-cifying directions in each of the cardinal planes Humanspatial maps generally fall into two classes Cartesian andpolar (Fig R1) which treat these two features slightly dif-ferently A classic Cartesian map has a single plane onwhich an origin is defined while directionality is imposedon the plane via the two orthogonal x and y axes whichintersect at this origin Positions are specified by projectingthem onto the axes yielding two distance parameters perposition Directions are not explicitly represented but canbe inferred by reference to the axes In a polar map by con-trast direction is explicitly represented Positions are speci-fied by the distance from the origin and the direction with

ResponseJeffery et al Navigating in a three-dimensional world

respect to a reference direction again yielding two par-ameters one specifying distance and one directionIf a map is a map of the real world then the frame of

reference of the map needs to correspond somehow (beisomorphic) with the world For example in a map of theArctic the origin on the map corresponds to the NorthPole and the reference direction on the map correspondsto the Greenwich Meridian Spatial inferences madeusing the map can enable valid conclusions to be drawnabout real-world spatial relationships However there aremany different levels of description of the real world If alookout on a ship navigating the polar waters sees aniceberg then she wonrsquot shout ldquoIceberg 3 km South 20degrees Eastrdquo but rather ldquoIceberg 100 m starboardbowrdquo ndash that is she will reference the object not to theEarthrsquos global reference frame but rather to the more rel-evant local reference frame of the ship Thus maps canalign with the world in many different waysIn the cognitive science of spatial representation the

global world-referenced frame has come to be called ldquoallo-centricrdquo whereas a local self-referenced frame is calledldquoegocentricrdquo As the icebreaker example shows howeverthis dichotomy is an oversimplification If the lookoutshould then bang her elbow on the gunwale she will com-plain not of her starboard elbow but of her right elbowwhich of course would become her port elbow if sheturned around ndash ldquostarboardrdquo and ldquoportrdquo are ship-referenceddirections and can therefore be egocentric or allocentricdepending on the observerrsquos current perception of eitherbeing the ship or being on the ship We will return to theegocentricallocentric distinction further onThe final important feature of a reference frame is its

dimensionality which was the primary topic of the targetarticle Cartesian and polar maps are usually two-dimen-sional but can be extended to three dimensions Forboth frameworks this requires the addition of anotherreference axis passing through the origin and aligned ver-tically Now the position of a point in this space requiresthree parameters distance in each of x y and z for the Car-tesian map and beeline distance plus two directions for the

polar map In the polar map the first direction is the anglewith respect to eitherboth of the orthogonal directionalreferences in the plane of those axes and the second direc-tion is the angle with respect to both axes in a plane orthog-onal to thatWith these concepts articulated the question now is how

the brain treats the core features of reference frames Weaddress these in the following order1 How are egocentric and allocentric reference frames

distinguished2 What is the dimensionality of these reference frames3 Where is the ldquooriginrdquo in the brainrsquos reference framesWe turn now to each of these issues as they were dealt

with by the commentators beginning with the egocentricallocentric distinction

R21 Egocentric and allocentric reference frames

When you reach out to pick up a coffee cup your brainneeds to encode the location of the cup relative to yourbody the location of the cup relative to the outside worldis irrelevant The frame of reference of the action is there-fore centred on your body and is egocentric Conversely ifyou are on the other side of the room then planning a pathto the cup requires a room-anchored allocentric referenceframe The fundamental difference between these frameslies in the type of odometry (distance-measuring) requiredIf the brainrsquos ldquoodometerrdquo is measuring distance betweenbody parts or movement of one part with respect toanother then the frame is egocentric and if the measure-ment is relative to the earth then it is allocentric Thereason that grid cells have proved so theoretically importantis that they perform allocentric odometry ndash their firingfields are equidistant in earth-centred coordinates ndash thusproving once and for all that the brain indeed has an allo-centric spatial systemActions in space typically require more than one refer-

ence frame and these frames therefore need to berelated to each other For example in reaching for thecup the movements the hand needs to make are

Figure R1 Three types of artificial map representation two metric and one nonmetric (a) Cartesian map showing orthogonal axes (xand y) an origin (black circle) and a grid with which other places on the map can be metrically specified (b) polar map with a specifieddirection (GreenwichMeridian GM) an origin (black circle) and a radial grid with which distances in a particular direction relative to theGM can be specified (c) topological (nonmetric) map ndash in this case of the London Underground ndash in which distances and directions arenot explicitly or veridically represented and in which positions are specified relative to other positions (eg ldquoRussell Square is betweenHolborn and Kingrsquos Crossrdquo) Note that the topological map being nonmetric does not have an origin or a metric grid

constrained by the position of the arm If reaching for thecup while also walking around the table then the handand arm frames need to be updated to account for the rela-tive motion of the cup and egocentric and allocentricframes must interact It seems that large parts of the pos-terior cortex are devoted to these complicated compu-tations with the parietal cortex being more involved withegocentric space (Andersen amp Buneo 2002) and the hippo-campus and its afferent cortices with allocentric spaceWhere and how these systems interact is not yet knownbut as Orban notes it may be in the parietal cortex

In our target article we restricted our analysis to allo-centric reference frames on the grounds that only allo-centric encoding is fully relevant to navigation Theimportance to the bicoded model of a possible egocentricallocentric unity is that there is better evidence for volu-metric spatial encoding in the realm of egocentric actionIf egocentric and allocentric space turn out to be encodedusing the same representational substrate the implicationis that allocentric space too should be volumetric ratherthan bicoded Indeed several commentators question theegocentricallocentric distinction and contest the exclusionof egocentric spatial processing from a theory of navigationOrban argues that navigation uses egocentric processesbecause allocentric and egocentric processing converge ona common neural substrate in the parietal cortex ndashwhichis true but it does not mean they are therefore indistin-guishable Klatzky amp Giudice referring to studies ofhumans ask ldquowhat then differentiates the behaviors osten-sibly governed by the planar mosaic from human spatiallydirected actions such as pointing reaching over-steppingand making contactrdquo ndash to which we would answer that ego-centric actions such as reaching do not require allocentri-cally referenced odometry They then say ldquoThis exclusionof metrically constrained behaviors from consideration isundermined however by the intrinsic ambiguity offrames of referencerdquo by which they presumably meanthat humans at least can flexibly move between frames ofreference However movement between referenceframes does not imply that the frames are therefore indistin-guishable Klatzky amp Giudice conclude that ldquolacking reliablebehavioral or neural signatures that would allow us to desig-nate actions as egocentric on the basis of their metricdemands it seems inappropriate simply to exclude themfrom a theory of volumetric spatial representationrdquo

We of course would not want to fully exclude egocentricfactors from a complete description of navigation but itneeds to be said that a distinction can be made As notedearlier the existence of allocentric odometry in the formof grid cell activity shows without a doubt that the brainpossesses at least one allocentric spatial system Giventhat there are actions that engage grid cells and othersthat do not there must be (at least) two forms of spatialrepresentation occurring in the brain a primarily ego-centric one in the parietal cortex that does not requiregrid cell odometry and a primarily allocentric one in theentorhinalhippocampal system that does The question iswhether these two systems use the same encodingscheme andor the same neural substrate In our viewboth are unlikely although in terms of neural substratethere are clearly close interactions Therefore arguing forallocentric isotropy (equivalent encoding in all directions)on the grounds of (possible) egocentric isotropy is unwar-ranted ndash they are separate domains

Of course there may be more than just these two refer-ence frame systems in the brain ndash there may even be manyAnd we accept that the distinction may not be as clear-cutand binary as we made outKaplan argues for the possibilityof hybrid representations that combine egocentric and allo-centric components and points out that the locomotor-plane-referenced form of the bicoded model is in fact sucha hybridWehave some sympathywith this view as it pertainsto the bicoded model but are not fully persuaded by itAlthough it is true that this plane is egocentrically referencedbecause it is always under the animalrsquos feet it is also the casethat the locomotor plane is constrained by the environmentSo one could argue that the egocentric frame of reference isforced to be where it is because of allocentric constraintsThus the mosaic form of the bicoded model in which eachfragment is oriented according to the local surface is argu-ably an allocentrically based rather than hybrid modelThe distinction may be mostly a semantic one howeverThe egocentricallocentric distinction can be blurred

even in discussions of explicitly allocentric neural represen-tation In fact the place cells themselves long regarded asthe prime index of allocentric encoding are hybrid inas-much as a place cell fires only when the ratrsquos egocentricreference frame occupies a particular place in the allo-centric world Indeed it is difficult to think of anyexample in the existing literature of fully allocentric encod-ing in which the position of the organism is truly irrelevantand neurons only encode the relative positions of objectswith respect to each other There may be such neuronsat least in humans but if they exist they have not beenfound yet Most encoding probably mixes allocentric andegocentric components to a certain degree Such mixingcan cause problems for the spatial machinery Forexample Barnett-Cowan amp Buumllthoff speculate that thecombination of righting reflexes to maintain an uprighthead posture during self-motion and object recognitioncombined with prior assumptions of the head beingupright may interfere with the brainrsquos ability to representthree-dimensional navigation through volumetric spaceEgocentric and allocentric distinctions aside within allo-

centric space one can also distinguish between locallydefined and globally defined space and several commenta-tors have addressed this issue It is relevant to whether alarge-scale representation is a mosaic a proposal forwhich Yamahachi Moser amp Moser (Yamahachi et al)advance experimental support noting that place and gridcell studies show fragmentation of large complex environ-ments based on internal structure (in their case walls)Howard amp Fragaszy argue that for a sufficiently denseand complex space such as in a forest the fragments ofthe mosaic could in principle become so small and numer-ous that the map as a whole starts to approximate a fullymetric one We would argue however that in order tobe fully metric in all dimensions such a map would needa global 3D directional signal for which evidence islacking at present However future studies may revealone at least in species with the appropriate ecologyIf complex space is represented in a mosaic fashion then

what defines the local reference frames for each fragmentHere it may be that distal and proximal environmental fea-tures serve different roles with distal features serving toorient the larger space and proximal ones perhaps defininglocal fragments Savelli amp Knierim observe that localversus global reference frames sometimes conflict They

suggest that apparent impoverishment of three-dimensionalencoding may result from capture of the activity by the localcues ndash the vertical odometer is in essence overridden bythe more salient surfaces of the local environment whichcontinually act to reset it For example consider a helicalmaze in which an animal circles around on a slowly ascend-ing track such as the one in the Hayman et al (2011) studyIt may be that the surface ndash the track ndash is highly salientwhereas the height cues are much less so leading tounder-representation of the ascending component of theanimalrsquos journey Studies in animals that are not constrainedto moving on a substrate such as those that fly will beneeded to answer the question more generally Thetheme of capture by local cues is also taken up by Dud-chenko Wood amp Grieves (Dudchenko et al) whonote that grid and place cell representations seem to belocal while head direction cells frequently maintain consist-ency across connected spaces The implication is that in atruly volumetric open space odometry has the capacity tooperate in all three dimensions leading to the creation ofan isotropic volumetric map Clearly experiments withother species with other modes of travel and in otherenvironments will be needed to answer the question ofwhether three-dimensional maps can be isotropic

R22 Dimensionality of reference frames

Having distinguished between egocentric and allocentricreference frames we turn now to the issue of dimensional-ity of these frames which was the theme of our targetarticle Although the earlier analysis of reference framesdrew a clear distinction between dimensions the braindoes not necessarily respect such distinctions as pointedout by some of the commentators Part of the reason forthis blurring of the boundaries is that the brain workswith information provided by the bodyrsquos limited senseorgans Thus it has to cope with the dimension reductionthat occurs in the transformation among the real three-dimensional world the two-dimensional sensory surfaces(such as the retina) that collect this information and thefully elaborated multidimensional cognitive representationthat the brain constructsSince information is lost at the point of reduction to two

dimensions it needs to be reconstructed again An exampleof such reconstruction is the detection of slope using visionalone in which slope is inferred by the brain from visualcues such as depth Orban explores how the primatevisual system constructs a three-dimensional represen-tation of objects in space and of slope based on depthcues in the parietal cortex However the reconstructionprocess introduces distortions such as slope overestimationOrban and Ross both observe that slope illusions varydepending on viewing distance meaning that the brainhas a complicated problem to solve when trying to tailoractions related to the slope Durgin amp Li suggestthat the action system compensates for such perceptual dis-tortion by constructing actions within the same referenceframe such that the distortions cancelPerception of three-dimensional spaces from a fixed

viewpoint is one problem but another quite different oneconcerns how to orchestrate actions within 3D space Forthis it is necessary to be able to represent the space andonersquos position within it ndash problems that are different foregocentric versus allocentric space

Taking egocentric space first How completely three-dimensional is the 3D reconstruction that the brain com-putes for near space (ie space within immediate reach)Several commentators argue in favour of a representationhaving full three-dimensionality Orban outlines how theparietal cortex generally believed to be the site of ego-centric spatial encoding (Galati et al 2010) is well special-ized for representing space in all three dimensions whileBadets suggests that the spatial dimensions may bemapped to a common area in the parietal cortex that inte-grates according to a magnitude-based coding scheme(along with other magnitudes such as number) LehkySereno amp Sereno (Lehky et al) agree that primatestudies of the posterior cortex in egocentric spatial tasksshow clear evidence of three-dimensional encoding Theysay ldquoWhile the dimensionality of space representation fornavigation [our emphasis] in primates is an importanttopic that has not been well studied there are physiologicalreasons to believe that it may be three-dimensionalrdquoalthough they do not outline what those reasons are By con-trast Phillips amp Ogeil argue that even egocentric space isbicoded First they appeal to evidence from perceptual illu-sions and neglect syndromes to show that vertical and hori-zontal spaces are affected differently Then they turn to atheoretical analysis of the constraints on integration of ver-tical and horizontal space and problems such as chaoticdynamics that can result from attempts at such integrationIt therefore seems that more research is needed to under-stand the representation of near space and whether ornot it is different in different dimensions (anisotropic)Turning to allocentric space the issue of the dimension-

ality was explored extensively in the target article and againin the commentaries The core questions concerningencoding of the third allocentric dimension have to dowith whether it is encoded and if so how and how itmight (or might not) be integrated with the other twoThe question of whether it is encoded is addressed by

several commentators Weisberg amp Newcombe makethe important point that ldquoverticalrdquo comes in (at least) twoforms orthogonal to horizontal or orthogonal to theanimal They suggest that the different types of verticalmay have different contributions to make to the overallencoding of the situation For example terrain slope maybe useful in for example helping orient the local environ-ment but this information would be lost if vertical encod-ing were earth-horizontalndashrelated They suggest that bothtypes of vertical may be encoded and may or may not beintegratedOther commentators focus on the encoding of volu-

metric (rather than slanting planar) three-dimensionalspace and speculate about how such encoding may beachieved Powers argues on the basis of studies in roboticsand engineering for dimension reduction as a means ofachieving coding efficiency by reducing redundancy (iewhat Carbon amp Hesslinger call ldquoempty cellsrdquo thatcontain no information but take up representationalspace) In this light the bicoded model would seem tooffer such efficiency Carbon amp Hesslinger agree arguingthat a two-dimensional map with alternative informationfor the vertical dimension should suffice for most thingsand they draw on comparative studies to support thisThis seems a very reasonable proposition to us For asurface-dwelling animal traversing say hilly terrain thereis only one z-location for each x-y position and so the

animal could in principle navigate perfectly well by com-puting only the x-y components of its journey The onlydrawback would be in say comparing alternative routesthat differed in how much ascent and descent theyrequired However this could be encoded using someother metric than distance ndash for example effort ndashmuchlike encoding that one route goes through a boggy swampand the other over smooth grassland The navigationsystem does not necessarily have to perform trigonometryin all three dimensions to compute an efficient routeacross undulating terrain For volumetric space theproblem becomes slightly less constrained because thereare unlimited z-locations for every x-y point Howeveragain it might be possible to navigate reasonably effectivelyby simply computing the trigonometric parameters in thex-y (horizontal) plane and then factoring in an approximateamount of ascent or descent

An interesting related proposal is put forward bySchultheis amp Barkowsky that representational complex-ity could be expanded or contracted according to currentneed According to their concept of scalable spatial rep-resentations the detail and complexity of the activated rep-resentation varies with task demands They suggest that thevertical dimension might be metrically encoded in situ-ations that demand it (eg air traffic control) but not in situ-ations that do not (eg taxi driving) Dimensionality ofencoding might also be affected not just by task demandsbut by cognitive load By analogy with the effects of loadon attentional processing (Lavie 2005) one mightsuppose that in conditions of high cognitive load (such asa pilot flying on instruments) the degree to which alldimensions are processed might be restricted Such proces-sing restrictions could have important implications for train-ing and instrument design for air- and spacecraft

Wang amp Street move in the opposite direction fromdimension reduction They argue not only that the cogni-tive representation of allocentric space is a fully fledgedthree-dimensional one but also that with the same neuralmachinery it is possible to encode four spatial dimensionsThey support their assertion with data showing that sub-jects can perform at above-chance levels on path-com-pletion tasks that cross three or four dimensionssomething they say should be possible only with a fully inte-grated volumetric (or perhaps ldquohypervolumetricrdquo) mapOur view on this suggestion is that it is a priori unlikelythat an integrated 4D map could be implemented by thebrain because this would require a 4D compass the selec-tion pressure for which does not exist in a 3D world Ourinterpretation of the 4D experiment is that if subjectscould perform at above-chance levels then they would doso using heuristics rather than full volumetric processingand by extension the above-chance performance on 3Dtasks may not require an integrated 3D map either Theexperimental solution to this issue would be to identifyneural activity that would correspond to a 4D compassbut computational modelling will be needed to make pre-dictions about what such encoding would look like

Given general agreement that there is some vertical infor-mation contained in the cognitive map (ie the map has atleast 25 dimensions) the next question concerns the natureof this information ndash is it metric and if not then what is itA fully volumetric map would have of course completemetric encoding of this dimension as well as of the othertwo and Wang amp Street argue for this Others such as

Savelli amp Knierim and Yamahachi et al argue that wewill not know for sure until the relevant experiments aredone with a variety of species and in a variety of environ-mental conditions A number of commentators agree withour suggestion however that the vertical dimension (or inthe mosaic version of the bicoded model the dimensionorthogonal to the locomotor plane) is probably not encodedmetrically There are a variety of opinions as to what theencoded information is however Phillips amp Ogeil suggestthat the other dimension is ldquotimerdquo rather than altitudealthough it is not clear to us exactly how this would workNardi amp Bingman support the proposal that ldquoeffortrdquo couldbe a height cue Sparks OrsquoReilly amp Kubie (Sparkset al) develop the notion of alternative factors further ndashthey discuss the issue of optimization duringnavigation point-ing out that the shortest path is not necessarily optimal andthat factors such as energy expenditure are also importantThey suggest a ldquomulti-codingrdquo (as opposed tomerely a ldquobicod-ingrdquo) scheme to take into account these other factors And asmentioned earlier Weisberg amp Newcombe suggest thatthere are two types of vertical information a true aligned-with-gravity vertical and another that is related to terrainslope raising the issue of how these two forms may berelated if at all Thus it appears that a variety of stimulustypes might serve to input into the encoding of vertical spaceThe final question regarding the vertical dimension is

whether ndash assuming it is represented at all ndash it is combinedwith the horizontal ones to make a fully integrated volu-metric map Burt de Perera Holbrook Davis Kacel-nik amp Guilford (Burt de Perera et al) refer to studiesof fish to argue that vertical space although in their viewcoded metrically is processed separately from horizontalspace They suggest some experiments to explorewhether this is the case by manipulating animalsrsquo responsesin the vertical dimension and showing that these show evi-dence of quantitative encoding (eg by obeying Weberrsquoslaw) However similar results could also be found if theanimal were using some kind of loose approximation to ametric computation (eg climbing to a certain level ofexhaustion) and so the experiments would need to be care-fully designed so as to show operation of true odometryBefore leaving the issue of reference frames it is worth

briefly examining the little-discussed issue of how the brainmay define an ldquooriginrdquo within these frames

R23 The origin of reference frames

As noted in the earlier analysis of reference frames ametric map needs an origin or fixed point against whichthe parameters of the space (distance and direction) canbe measured The question of whether the brain explicitlyrepresents origins in any of its reference frames is comple-tely unanswered at present although a notional origin cansometimes be inferred To take a couple of examplesamong many neurons in the visual cortex that respond pro-portionally to distance from the eye (Pigarev amp Levichkina2011) could be said to be using the retina as the ldquooriginrdquo inthe antero-posterior dimension while neurons that are gainmodulated by head angle (Snyder et al 1998) are using theneck as the ldquooriginrdquo (and straight ahead as the ldquoreferencedirectionrdquo) Whether there exist origins for the other ego-centric reference frame however is less clearFor allocentric space the origin becomes a more nebu-

lous concept although in central place forager species

such as rats the central place (the nest) may constitute anldquooriginrdquo of sorts However it may be that the brain uses nota fixed point as its reference so much as a fixed boundaryPlace and grid cells for example anchor their firing to theboundaries of the local environment moving their firingfields accordingly when the boundaries are moved enbloc Furthermore at least for grid cells deformation (asopposed to rigid translation) of these boundaries in somecases leads to a uniform deformation of the whole gridwith no apparent focus (Barry et al 2007) suggestingthat the whole boundary contributes to anchoring thegrid Hence it is not obvious that there is a single point-like origin in the entorhinalhippocampal representationof allocentric space However it could be that the placecells themselves in essence provide multiple origins ndashtheir function may be to anchor spatially localized objectsand events to the reference frame defined by the bound-aries and grids It may be that the simple singular notionof origin that has been so useful in mathematics will needto be adapted to accommodate the brainrsquos far more nebu-lous and multifarious ways of working

R3 Comparative studies

We turn now to the ways in which studies from differentspecies with different evolutionary histories and ecologicalconstraints have contributed to our understanding ofspatial encoding Principles derived for one species maynot apply to a different species in a different environmentusing a different form of locomotionOrban points out forexample that rodent and primate brains are very differentand that there are cortical regions in primates that do noteven have homologues in rodents Hence extrapolationfrom rodents to primates must be done with care a pointwith which we agreeThe importance of comparative studies in cognitive

science lies in understanding how variations in body typeecological niche and life experience correlate with vari-ations in representational capacity This information tellsus something about how these representations areformed ndashwhat aspects of them are critical and whataspects are just ldquoadd-onsrdquo that evolved to support a particu-lar species in a particular ecological niche While compara-tive studies of three-dimensional spatial cognition are rarethey promise to be particularly revealing in this regard

R31 Studies of humans

Humans are naturally the species that interests us themost but there are important constraints on how muchwe can discover about the fine-grained architecture of ourspatial representation until non-invasive single neuronrecording becomes possible There are however othermeans of probing encoding schemes such as behaviouralstudies Longstaffe Hood amp Gilchrist (Longstaffeet al) and Berthoz amp Thibault describe the methodsthat they use to understand spatial encoding in humansover larger (non-reachable) spaces However Carbon ampHesslinger point out that experimental studies need totake into account the ecological validity of the experimentalsituation something which is sometimes difficult to achievein laboratory studies in which natural large-scale three-dimensional spaces are rare Virtual reality (VR) will likely

help here and future studies will be able to expand on pre-vious research in two-dimensional spaces to exploit thecapacity of VR subjects to ldquofloatrdquo or ldquoflyrdquo through 3Dspace This could be used not just during behaviouralstudies but also in neuroimaging tasks in which the partici-pating brain structures can be identifiedSome commentators describe the human studies in real

(non-virtual) 3D spaces that have recently begunHoumllscher Buumlchner amp Strube (Houmllscher et al) extendthe anistropy analysis of our original article to architectureand environmental psychology and note that there are indi-vidual differences in the propensity to form map-basedversus route-based representations Pasqualotto ampProulx describe prospects for the study of blind peoplewhose condition comprises a natural experiment that hasthe potential to help in understanding the role of visualexperience in shaping spatial representation in the humanbrain Klatzky amp Giudice argue that humans are ecologi-cally three-dimensional inasmuch as their eyes are raisedabove the ground and their long arms are able to reachinto 3D space meaning that it is adaptive to representlocal space in a fully three-dimensional manner Indeedevidence supports that local (egocentric) space is rep-resented isotropically in primates a point made also byOrban though contested by Phillips amp OgeilA final important facet of human study is that humans

have language which provides insights unavailable fromanimal models In this vein Holmes amp Wolff note thatspatial language differentiates horizontal axes from verticalaxes more than it differentiates the horizontal ones fromeach other thereby supporting the notion of anisotropyin how space is encoded However as discussed in thenext section the degree to which such anisotropy couldbe due to experience rather than innate cognitive architec-ture is still open for debate

R32 Studies of nonhumans

Moving on to nonhuman animals Carbon amp Hesslingerreview the various ways in which the spatial problems facedby animals of a variety of different ecologies reduce to acommon set of mainly surface-referenced problemsHowever not all commentators agree that a planar orquasi-planar map is supported by comparative studiesOrban observes that the nonhuman primate brain is muchcloser to the human than to the rodent brain on whichmuch neurobiological work has been done and notes thatseveral parietal areas in primates are specialized for spatialrepresentation Howard amp Fragaszy discuss the issue ofnonhuman primates moving in a dense arboreal latticeThey suggest the use of laser scanning technology (LiDAR)to map out the movements of primates to determine howtheir movements are constrained (or at least informed) bythe structure of the complex spaces throughwhich theymoveNonhuman primates are relatively close to humans in

evolutionary terms and commonalities in spatial behaviourmight reflect this Useful insights can therefore be gleanedby moving phylogenetically further away to try and discernwhich features are ancient and foundational and which aremore species-specificMoss advocates bats for the study ofthree-dimensional spatial encoding Not only can bats flyand thus move through space in an unconstrained waybut they also have a sense ndash echolocation ndash unavailable tomost mammals This provides a tool for understanding

how a supramodal cognitive representation can be formedfrom sensory inputs that show local (species-specific) vari-ation Electrolocation and magnetic sense are additionalsensory systems in other species that may also be informa-tive here

Moving even further away from humans non-mamma-lian vertebrates such as birds and invertebrates havebeen valuable in providing comparative data that revealboth commonalities and differences in how different organ-isms represent space Zappettini amp Allen take issue withthe notion that there is a map-like representation operatingin nonhuman animals suggesting that this is putting ldquothehypothetical cart before the evidential horserdquo The debateabout the operation of cognitive maps in animals is a long-standing one that is too complex to fully revisit in thisforum However we would argue in defence of our prop-osition that the existence of grid cells is evidence of at leastsome kind of map-like process That there are neurons inthe brain that respond to distances and directions oftravel is incontrovertible proof that rodent brains ndash andtherefore likely the brains of other mammals at least ndashcompute these parameters Regardless of whether one con-siders a representation incorporating distance and directionto necessarily be a map it is nevertheless valid to askwhether encoding of these parameters extends into allthree dimensions To this extent we hope that commonground can be found with the map-skeptics in devising aprogram of study to explore real-world spatial processingZappettini amp Allen also object to our use of evolutionaryselection arguments to defend the bicoded hypothesisover the volumetric one We agree that such argumentsare weak and serve only to suggest hypotheses and not totest them However suggesting hypotheses was the mainaim of our target article given the paucity of hard dataavailable at present

Other comparative biologists have been more relaxedabout the notion of a map-like representation underlyingnavigational capabilities in nonhumans Citing studies ofavian spatial encodingNardi amp Bingman suggest that tax-onomically close but ecologically different species would beuseful comparators ndash for example chickadees (inhabitingan arboreal volumetric space) versus nutcrackers (storingfood on the ground) Burt de Perera et al use studiesof fish to argue that vertical space although coded separ-ately from horizontal space is nevertheless encoded metri-cally This finding is interesting because fish can ndash unlikehumans ndashmove in an unconstrained way through 3Dspace and would in theory benefit from an integrated 3Dmap if such had evolved That the fish map is evidentlynot integrated in this way suggests that maybe such inte-grated maps never did evolve Dyer amp Rosa suggest thatstudy of simple organisms such as bees can help revealthe ways in which complex behaviours can arise fromsimple building blocks particularly when plasticity isadded to the mix They agree however that evidencesuggests that in bees too space is represented in abicoded fashion

In summary then studies in comparative cognition maybe tremendously informative in the unravelling of the neu-rocognitive representation of three-dimensional spaceboth by revealing which features are central and whichare peripheral ldquoadd-onsrdquo and by showing how differentecological constraints and life experiences contribute toshaping the adult form of the representation

R4 Experience

The final major theme in the commentaries concerned thepossible role of experience in shaping spatial represen-tations Note that just because experience correlates withencoding format (bicoded volumetric etc) it does notnecessarily follow that it caused that format For examplean animal that can fly will have (perhaps) an ability toencode 3D space and also a lifetimersquos experience ofmoving through 3D space but the encoding may havebeen hard-wired by evolution rather than arising from theindividualrsquos own life experience Yamahachi et al stressthat more research is needed to determine the role ofexperience in shaping the structure of the map Thiscould be done by for example raising normally 3D-explor-ing animals in a restricted 2D space to see whether theencoding type changesExperience if it affects encoding at all can act in two

broad ways It can operate during development to organizethe wiring of the underlying neural circuits or it canoperate on the fully developed adult brain via learningto enable the subject to acquire and store new informationThese two areas will be examined in turn

R41 Developmental experience

It is now clear that infants are not blank slates on which alltheir adult capacities will be written by experience but areborn with many of their adult capacities having alreadybeen hard-wired under genetic control That said it isalso clear that developmental experience can shape theadult brain How experience and hard-wiring interact is amatter of considerable interestExperience during ontogeny can shape development of a

neural structure by affecting neural migration and axonalprojections or else subsequently by affecting dendriticconnections and synapse formation Migration and axondevelopment are generally complete by the time a develop-ing animal has the ability to experience three-dimensionalspace and so are unlikely to affect the final structure ofthe spatial representation On the other hand dendriticand synaptic proliferation and pruning processes are plen-tiful during infant development ndash however they also takeplace to an extent during adulthood when they comeunder the rubric of ldquolearningrdquo It is evident thereforethat development and learning show a considerableoverlapSome sensory systems show critical periods in infancy

during which experience is necessary for normal braindevelopment and interruption of which can cause lifelongimpairment Vision in mammals is a prime example of suchexperience-dependent plasticity Monocular deprivation orstrabismus (squint) occurring during the critical period butnot during adulthood can permanently affect connectionformation in the thalamus and visual cortex resulting inadult visual impairment (amblyopia Morishita amp Hensch2008) A natural question arising from studies of three-dimensional encoding in animals then is whether infantexperience can affect the formation of the adult cognitiverepresentationRecent studies have suggested that head direction cells

place cells and grid cells start to operate in adult-lookingways very early in infancy in rat pups In two studies pub-lished together in a 2010 issue of Science (Langston et al

2010 Wills et al 2010) cells were recorded from the firstday the pups started exploring 16 days after birth Headdirection cells already showed adult-like firing and stabilityeven though the pups had had very little experience ofmovement and only a few days of visual experience Placecells were also apparent from the earliest days of explora-tion although their location-specific firing improved incoherence and stability with age Grid cells were the lastcell type to mature appearing at about 3ndash4 weeks of agebut when they did appear they showed adult-like patternsof firing with no apparent requirement for experience-dependent tuning Thus the spatial system of the ratseems to come on-stream with most of its adult capabilitiesalready present suggesting a considerable degree of hard-wiring of the spatial representationIs this true for vertical space as well The experiment of

Hayman et al (2011) showed a defect in vertical grid cellodometry and perhaps this defect is due to the limited3D experience that rats in standard laboratory cages havehad However if extensive experience of 2D environmentsis not needed for the spatial cells to operate in the way thatthey do as is manifestly the case then experience of 3Denvironments may not be needed either Thus the ques-tion of whether infant experience is necessary for adult cog-nitive function in the spatial domain is one that awaitsfurther data for its resolutionAlthough the experiments have yet to be done some

theoretical analysis of this issue is beginning Stella SiKropff amp Treves (Stella et al) describe a model ofgrid formation in two dimensions that arises from explora-tion combined with simple learning rules and suggest thatthis would extend to three dimensions producing a face-centered cubic array of grid fields We would note herethat such a structure would nevertheless be anisotropicbecause the fields do not form the same pattern whentransected in the horizontal plane as in the vertical Whilesuch a grid map would be metric Peremans amp Vander-elst discuss a theoretical proposal whereby a non-metrictopological map could be built up from experience byrecording approximate distances travelled between nodesand using these to build a net that approximates (topologi-cally) the real space They note that this would look aniso-tropic because of differential experience of travel in thedifferent dimensions even though the same rules operatein all three dimensions This is an interesting proposalthat predicts that animals with equal experience in allthree dimensions should produce isotropic spatial rep-resentations something that may be tested during record-ings on flying animals such as bats

R42 Adult experience

As noted above there are considerable overlaps in the bio-logical processes supporting infant development and adultlearning because the processes of dendritic arborizationand synaptic pruning that support experience-dependentplasticity operate in both domains (Tavosanis 2012) Thequestion arises then as to whether adult experience canshape the structure of a complex cognitive representationsuch as that of spaceDyer amp Rosa suggest that in bees although movement

in the vertical dimension (determined using optic flow) isusually represented with a lower degree of precisionbrain plasticity could allow the experience-dependent

tuning of responses such that precision in this dimensioncould be increased In this way the organisms can learnto use navigationally relevant information particular to agiven environment This does not mean however thatthey learn a different mechanism for combining infor-mation Moreover Dyer amp Rosa agree that the evidencesupports the notion of a bicoded structure to spatial proces-sing in bees even with experienceExperiments involving human subjects have provided

insights into how experience might constrain spatial rep-resentation Berthoz amp Thibault discuss how the learningexperience during exploration of a multilayer environmentcan shape the way in which subjects subsequently tacklenavigation problems in that environment (Thibault et al2013) Pasqualotto amp Proulx explore the effects thatvisual deprivation (through blindness) can have on theadult spatial representation that forms They suggest thatcongenitally blind individuals may not exhibit quasi-planarspatial representation of three-dimensional environmentspositing that the absence of visual experience of the 3Dworld may shape how spatial encoding occurs This is amatter that awaits further study as 3D spatial encodinghas not yet been explored in subjects who are blindThese authors suggest several ways in which such studiesmight be conducted using sensory substitution devices(which convert visual information into tactile or auditory)in order to create 3D virtual reality environments Thedifference between subjects who are congenitally blindand those who had had visual experience in early life willbe particularly interesting hereExperience may play a role in how normal-sighted sub-

jects perceive and interpret the spatial world too Bianchiamp Bertamini discuss the interesting errors that human sub-jects make when predicting what will be visible in a mirror asthey approach it and show that the errors are different forhorizontal versus vertical relationships Tellingly adultsmake more errors than children These findings suggestthat experience may play a part in generation of errors ndashadults have more experience of crossing mirrors from oneside to the other than from top to bottom and more experi-ence of passing in front of mirrors generally than childrendo This may perhaps be a case where experience diminishesthe capacity to accurately represent a space because adultshave enhanced ability to make inferences which they do(in this case) erroneouslyWhile it is certainly the case that experience is likely to

inform the adult cognitive representation of space ourhypothesis is that experience is not necessary to constructit This is because the rodent studies described earliersuggest a high degree of representational complexity evenin animals that have had impoverished spatial life experi-ence Future research in animals reared in complex 3Denvironments will be needed to determine the extent towhich experience is needed for formation and refinementof the brainrsquos map of 3D space

R5 Conclusion

It is apparent from the commentaries on our target articlethat there are many questions to be answered concerningthe relatively nascent field of three-dimensional spatialencoding The most salient questions fall into the categoriesoutlined in this response but no doubt more will emerge as

spatial cognition researchers start to tackle the problem ofrepresenting large-scale complex spaces The work willrequire an interplay between behavioural and neurobiolo-gical studies in animals and humans and the efforts of com-putational cognitive scientists will be needed to place theresults into a theoretical framework The end result willwe hope be a body of work that is informative to many dis-ciplines involved both in understanding natural cognitivesystems and also in building artificial spaces and artificialdevices

ACKNOWLEDGMENTSThe work was supported by grants to Kate Jeffery from theWellcome Trust (GR083540AIA) the European Commis-sionrsquos Seventh Framework Programme (SPACEBRAIN)the Medical Research Council (G1100669) and the Bio-technology and Biological Sciences Research Council(BBJ0097921) and by a grant to Axona Ltd from theEuropean Commissionrsquos Seventh Framework Programme(SPACEBRAIN)

References

[The letters ldquoardquo and ldquorrdquo before authorrsquos initials stand for target article andresponse references respectively]

Abdollahi R O Jastorff J amp Orban G A (2012) Common and segregated pro-cessing of observed actions in human SPL Cerebral Cortex E-pub ahead ofprint August 23 2012 [GAO]

Aflalo T N amp Graziano M S (2008) Four-dimensional spatial reasoning inhumans Journal of Experimental Psychology Human Perception and Per-formance 34(5)1066ndash77 [aKJJ RFW]

Allen G L (1999) Spatial abilities cognitive maps and wayfinding Bases for indi-vidual differences in spatial cognition and behavior In Wayfinding behaviorCognitive mapping and other spatial processes ed R G Golledge pp 46ndash80Johns Hopkins Press [CCC]

Ambinder M S Wang R F Crowell J A Francis G K amp Brinkmann P (2009)Human four-dimensional spatial intuition in virtual reality Psychonomic Bul-letin and Review 16(5)818ndash23 [RFW]

Andersen R A amp Buneo C A (2002) Intentional maps in posterior parietal cortexAnnual Review of Neuroscience 25189ndash220 [GAO rKJJ]

Angelaki D E amp Cullen K E (2008) Vestibular system The many facets of amultimodal sense Annual Review of Neuroscience 31125ndash50 [aKJJ]

Angelaki D E McHenry M Q Dickman J D Newlands S D amp Hess B J(1999) Computation of inertial motion Neural strategies to resolve ambiguousotolith information Journal of Neuroscience 19(1)316ndash27 [aKJJ]

Aoki H Ohno R amp Yamaguchi T (2005) The effect of the configuration and theinterior design of a virtual weightless space station on human spatial orientationActa Astronautica 561005ndash16 [aKJJ]

Avarguegraves-Weber A Dyer A G Combe M amp Giurfa M (2012) Simultaneousmastering of two abstract concepts by the miniature brain of bees Proceedingsof the National Academy of Sciences USA 1097481ndash86 [AGD]

Avery G C amp Day R H (1969) Basis of the horizontalndashvertical illusion Journal ofExperimental Psychology 81376ndash80 [JGP]

Avraamides M Klatzky R L Loomis J M amp Golledge R G (2004) Use ofcognitive vs perceptual heading during imagined locomotion depends on theresponse mode Psychological Science 15403ndash408 [RLK]

Badets A Koch I amp Toussaint L (2013) Role of an ideomotor mechanism innumber processing Experimental Psychology 6034ndash43 [ABa]

Badets A amp Pesenti M (2011) Finger-number interaction An ideomotor accountExperimental Psychology 58287ndash92 [ABa]

Baker G C amp Gollub J G (1996) Chaotic dynamics An introduction CambridgeUniversity Press [JGP]

Bardunias P M amp Jander R (2000) Three dimensional path integration in thehouse mouse (Mus domestica) Naturwissenschaften 87(12)532ndash34 [aKJJ]

Barnett-Cowan M Dyde R T amp Harris L R (2005) Is an internal model of headorientation necessary for oculomotor control Annals of the New York Academyof Sciences 1039314ndash24 [MB-C]

Barnett-Cowan M Fleming R W Singh M amp Buumllthoff H H (2011) Perceivedobject stability depends on multisensory estimates of gravity PLoS ONE 6(4)e19289 [MB-C]

Barnett-Cowan M Meilinger T Vidal M Teufel H amp Buumllthoff H H (2012)MPI CyberMotion Simulator Implementation of a novel motion simulator toinvestigate multisensory path integration in three dimensions Journal of Visu-alized Experiments (63)e3436 [MB-C]

Baron-Cohen S (2008) Autism hypersystemizing and truth The Quarterly Journalof Experimental Psychology 61(1)64ndash75 [KAL]

Barry C Hayman R Burgess N amp Jeffery K J (2007) Experience-dependentrescaling of entorhinal grids Nature Neuroscience 10(6)682ndash84 [arKJJ]

Barry C Lever C Hayman R Hartley T Burton S OrsquoKeefe J Jeffery K J ampBurgess N (2006) The boundary vector cell model of place cell firing andspatial memory Reviews in the Neurosciences 17(1ndash2)71ndash79 [PAD]

Benedikt M L (1979) To take hold of space Isovists and isovist fields Environmentand Planning B Planning and Design 647ndash65 [CH]

Bennett A T (1996) Do animals have cognitive maps Journal of ExperimentalBiology 199219ndash24 [SZ]

Bennett D A amp Tang W (2006) Modelling adaptive spatially aware and mobileagents Elk migration in Yellowstone International Journal of GeographicalInformation Science 20(9)1039ndash66 doi 10108013658810600830806[AMH]

Beraldin J A Blais F amp Lohr U (2010) Laser scanning technology In Airborneand terrestrial laser scanning ed G Vosselman ampHMaas pp 1ndash42Whittles[AMH]

Bertamini M Lawson R Jones L amp Winters M (2010) The Venus effect in reallife and in photographs Attention Perception and Psychophysics 72(7)1948ndash64 DOI 103758APP7271948 [IB]

Bertamini M Spooner A amp Hecht H (2003) Naiumlve optics Predicting and per-ceiving reflections in mirrors Journal of Experimental Psychology HumanPerception and Performance 29(5)982ndash1002 doi 1010370096-1523295982 [IB]

Bertamini M amp Wynne L (2009) The tendency to overestimate what is visible in aplanar mirror amongst adults and children European Journal of CognitivePsychology 22516ndash28 doi 10108009541440902934087 [IB]

Bhalla M amp Proffitt D R (1999) Visual-motor recalibration in geographical slantperception Journal of Experimental Psychology Human Perception and Per-formance 25(4)1076ndash96 [aKJJ]

Bianchi I Burro R Torquati S amp Savardi U (2013) The middle of the roadPerceiving intermediates Acta Psychologica 144(1)121ndash35 DOI 101016jactpsy201305005 [IB]

Bianchi I amp Savardi U (2008) The relationship perceived between the real bodyand the mirror image Perception 5666ndash87 [IB]

Bianchi I amp Savardi U (2012a) The cognitive dimensions of contrariety In Thesquare of opposition A general framework for cognition ed J-Y Bezieau amp GPayette pp 443ndash70 Peter Lang [IB]

Bianchi I amp Savardi U (2012b) What fits into a mirror Naiumlve beliefs on the field ofview of mirrors Journal of Experimental Psychology Human Perception andPerformance 38(5)1144ndash58 doi 101037a0027035 [IB]

Bianchi I Savardi U amp Burro R (2011a) Perceptual ratings of opposite spatialproperties Do they lie on the same dimension Acta Psychologica 138(3)405ndash18 doi 101016jactpsy201108003 [IB]

Bianchi I Savardi U amp Kubovy M (2011b) Dimensions and their poles A metricand topological theory of opposites Language and Cognitive Processes 26(8)1232ndash65 doi 101080016909652010520943 [IB]

Blohm G Keith G P amp Crawford J D (2009) Decoding the cortical trans-formations for visually guided reaching in 3D space Cerebral Cortex 191372ndash93 [SRL]

Blum K I amp Abbott L F (1996) A model of spatial map formation in the hippo-campus of the rat Neural Computation 885ndash93 [FTS]

Boccara C N Sargolini F Thoresen V H Solstad T Witter M P Moser E Iamp Moser M-B (2010) Grid cells in pre- and parasubiculum Nature Neuro-science 13987ndash94 [FSa]

Boyer W Longo M R amp Bertenthal B I (2012) Is automatic imitation aspecialized form of stimulus-response compatibility Dissociating imitative andspatial compatibilities Acta Psychologica 139(3)440ndash48 doi101016jactpsy20120100 [IB]

Bressan P Garlaschelli L amp Barracano M (2003) Anti-gravity hills are visualillusions Psychological Science 14441ndash49 [HER]

Breveglieri R Hadjidimitrakis K Bosco A Sabatini S P Galletti C amp FattoriP (2012) Eye position encoding in three-dimensional space Integration ofversion and vergence signals in the medial posterior parietal cortex Journal ofNeuroscience 32159ndash69 [SRL]

Bridgeman B amp Hoover M (2008) Processing spatial layout by perception andsensorimotor interaction The Quarterly Journal of Experimental Psychology 61(6)851ndash59 [HER]

Brill R Lutcavage M Metzger G Stallings J Bushnell P Arendt M Lucy JWatson C amp Foley D (2012) Horizontal and vertical movements of juvenile

ReferencesJeffery et al Navigating in a three-dimensional world

North Atlantic Bluefin Tuna (Thunnus thynnus) in the western North Atlanticdetermined using ultrasonic telemetry with reference to population assessmentby aerial surveys Fortschritte der Zoologie 100155ndash67 [aKJJ]

Brodsky M C Donahue S P Vaphiades M amp Brandt T (2006) Skew deviationrevisited Survey of Ophthalmology 51105ndash28 [MB-C]

Brown M F amp Lesniak-Karpiak K B (1993) Choice criterion effects in the radial-arm maze ndashMaze-arm incline and brightness Learning and Motivation 24(1)23ndash39 [aKJJ]

Brown P amp Levinson S C (1993) ldquoUphillrdquo and ldquodownhillrdquo in Tzeltal Journal ofLinguistic Anthropology 3(1)46ndash74 [IB]

Buumlchner S Houmllscher C amp Strube G (2007) Path choice heuristics for navigationrelated to mental representations of a building In Proceedings of the 2ndEuropean Cognitive Science Conference Delphi Greece 23ndash27 May 2007 edS Vosniadou D Kayser amp A Protopapas pp 504ndash509 ErlbaumTaylor ampFrancis [aKJJ ABe CH]

Bueti D amp Walsh V (2009) The parietal cortex and the representation of timespace number and other magnitudes Philosophical Transactions of the RoyalSociety B 3641831ndash40 [ABa]

Buhot M-C Dubayle C Malleret G Javerzat S amp Segu L (2001) Explorationanxiety and spatial memory in transgenic anophthalmic mice BehavioralNeuroscience 115445ndash67 [AP]

Burgess N Barry C amp OrsquoKeefe J (2007) An oscillatory interference model of gridcell firing Hippocampus 17801ndash12 [CFM]

Burgess N amp OrsquoKeefe J (1996) Neuronal computations underlying the firing ofplace cells and their role in navigation Hippocampus 6749ndash62 [FTS]

Burt de Perera T de Vos A amp Guilford T (2005) The vertical component of afishrsquos spatial map Animal Behaviour 70405ndash409 [aKJJ]

Butler D L Acquino A L Hissong A A amp Scott P A (1993) Wayfinding bynewcomers in a complex building Journal of the Human Factors and Ergo-nomics Society 35159ndash73 [AP]

Butterworth B (1999) A head for figures Science 284928ndash29 [ABa]Calton J L amp Taube J S (2005) Degradation of head direction cell activity during

inverted locomotion Journal of Neuroscience 25(9)2420ndash28 [aKJJ]Carbon C C (2007) Autobahn people Distance estimations between German cities

biased by social factors and the autobahn Lecture Notes in Artificial Science4387489ndash500 [CCC]

Carey F G (1992) Through the thermocline and back again Oceanus 3579ndash85[aKJJ]

Carlson L Houmllscher C Shipley T amp Conroy Dalton R (2010) Getting lost inbuildings Current Directions in Psychological Science 19(5)284ndash89 [CH]

Chan A H S amp Chan K W L (2005) Spatial S-R compatibility of visual andauditory signals Implications for humanndashmachine interface design Displays 26(3)109ndash19 [IB]

Channon A J Crompton R H Guumlnther M M DrsquoAoucirct K amp Vereecke E E(2010) The biomechanics of leaping in gibbons American Journal of PhysicalAnthropology 143(3)403ndash16 doi 101002ajpa21329 [AMH]

Chebat D R Chen J K Schneider F Ptito A Kupers R amp Ptito M (2007)Alterations in right posterior hippocampus in early blind individualsNeuroReport 18329ndash33 [AP]

Chen A DeAngelis G C amp Angelaki D E (2010) Macaque parieto-insular ves-tibular cortex Responses to self-motion and optic flow Journal of Neuroscience30(8)3022ndash42 [GAO]

Chen A DeAngelis G C amp Angelaki D E (2011a) Convergence of vestibular andvisual self-motion signals in an area of the posterior sylvian fissure Journal ofNeuroscience 31(32)11617ndash27 [GAO]

Chen A DeAngelis G C amp Angelaki D E (2011b) Representation of vestibularand visual cues to self-motion in ventral intraparietal cortex Journal of Neuro-science 31(33)12036ndash252 [GAO]

Chittka L amp Jensen K (2011) Animal cognition Concepts from apes to beesCurrent Biology 21R116ndash19 [AGD]

Chittka L amp Niven J (2009) Are bigger brains better Current Biology 19R995ndash1008 [AGD]

Clark H H (1973) Space time semantics and the child In Cognitive developmentand the acquisition of language ed T E Moore pp 27ndash63 Academic Press[KJH]

Cockburn A amp McKenzie B (2001) 3D or not 3D Evaluating the effect of thethird dimension in a document management system In ed M Beaudouin-Lafon amp R J K Jacob Proceedings of CHIrsquo2001 Conference on HumanFactors in Computing Systems pp 434ndash41 ACM Press [DMWP]

Cockburn A amp McKenzie B (2002) Evaluating the effectiveness of spatial memoryin 2D and 3D physical and virtual environments In ed L Terveen Proceed-ings of CHIrsquo2002 Conference on Human Factors in Computing Systemspp 203ndash10 ACM Press [DMWP]

Cohen R G amp Rosenbaum D A (2004) Where objects are grasped reveals howgrasps are planned Generation and recall of motor plans Experimental BrainResearch 157486ndash95 [ABa]

Collins A M amp Quillian M R (1969) Retrieval time from semantic memoryJournal of Verbal Learning and Verbal Behavior 8240ndash47 [HS]

Conroy Dalton R (2003) The secret is to follow your nose Route path selection anangularity Environment and Behavior 35(1)107ndash31 [CH]

Conroy Dalton R (2005) Space syntax and spatial cognition World Architecture18541ndash47 107ndash111 [CH]

Corballis M C (2013) Mental time travel A case for evolutionary continuity Trendsin Cognitive Sciences 175ndash6 [ABa]

Cornelis E V van Doorn A J amp Wagemans J (2009) The effects of mirrorreflections and planar rotations of pictures on the shape percept of the depictedobject Perception 38(10)1439ndash66 [IB]

Creem S H amp Proffitt D R (1998) Two memories for geographical slant Sep-aration and interdependence of action and awareness Psychonomic Bulletinand Review 5(1)22ndash36 [aKJJ]

Creem S H amp Proffitt D R (2001) Defining the cortical visual systems ldquoWhatrdquoldquowhererdquo and ldquohowrdquo Acta Psychologica (Amsterdam) 107(1ndash3)43ndash68 [aKJJ]

Creem-Regehr S H Mohler B J amp Thompson W B (2004) Perceived slant isgreater from far versus near distances (Abstract) Journal of Vision 4(8)374a doi10116748374 Available at httpjournalofvisionorg48374 [HER]

Croucher C J Bertamini M amp Hecht H (2002) Naiumlve optics Understanding thegeometry of mirror reflections Journal of Experimental Psychology HumanPerception and Performance 28546ndash62 doi 1010370096-1523283546[IB]

Cruse D A (1986) Lexical semantics (Cambridge Textbooks in Linguistics)Cambridge University Press [IB]

Cutting J M amp Vishton P M (1995) Perceiving layout and knowing distances Theintegration relative potency and contextual use of different information aboutdepth InHandbook of perception and cognition vol 5 Perception of space andmotion ed W Epstein amp S Rogers pp 69ndash117 Academic Press [RLK]

Dacke M amp Srinivasan M V (2007) Honeybee navigation Distance estimation inthe third dimension Journal of Experimental Biology 210(Pt 5)845ndash53[aKJJ AGD]

Dalrymple K A amp Kingstone A (2010) Time to act and attend to the real mech-anisms of action and attention British Journal of Psychology 101213ndash16[KAL]

Dehaene S Bossini S amp Giraux P (1993) The mental representation of parity andnumber magnitude Journal of Experimental Psychology General 122371ndash96 [ABa]

Derdikman D ampMoser E I (2010) A manifold of spatial maps in the brain Trendsin Cognitive Sciences 14(12)561ndash69 [HY]

Derdikman D Whitlock J R Tsao A Fyhn M Hafting T Moser M ampMoserE I (2009) Fragmentation of grid cell maps in a multicompartment environ-ment Nature Neuroscience 12(10)1325ndash32 [PAD FSa HY]

Deutschlaumlnder A Stephan T Huumlfner K Wagner J Wiesmann M Strupp MBrandt T amp Jahn K (2009) Imagined locomotion in the blind An fMRI studyNeuroImage 45122ndash28 [AP]

Di Fiore A amp Suarez S A (2007) Route-based travel and shared routes in sym-patric spider and woolly monkeys Cognitive and evolutionary implicationsAnimal Cognition 10(3)317ndash29 doi 101007s10071-006-0067-y [AMH]

Domjan M (2009) The principles of learning and behavior 6th edition WadsworthCengage Learning [TBdP]

Dudchenko P A amp Zinyuk L E (2005) The formation of cognitive maps ofadjacent environments Evidence from the head direction cell system Behav-ioral Neuroscience 119(6)1511ndash23 [aKJJ PAD]

Durgin F H Hajnal A Li Z Tonge N amp Stigliani A (2010a) Palm boards arenot action measures An alternative to the two-systems theory of geographicalslant perception Acta Psychologica 134182ndash97 [FHD]

Durgin F H Hajnal A Li Z Tonge N amp Stigliani A (2011) An imputed dis-sociation might be an artifact Further evidence for the generalizability of theobservations of Durgin et al 2010 Acta Psychologica (Amsterdam) 138(2)281ndash84 [aKJJ]

Durgin F H Klein B Spiegel A Strawser C J ampWilliams M (2012) The socialpsychology of perception experiments Hills backpacks glucose and theproblem of generalizability Journal of Experimental Psychology Human Per-ception and Performance 201238(6)1582ndash95 [aKJJ]

Durgin F H amp Li Z (2011) Perceptual scale expansion An efficient angular codingstrategy for locomotor space Attention Perception and Psychophysics 73(6)1856ndash70 [aKJJ FHD]

Durgin F H amp Li Z (2012) Spatial biases and the haptic experience of surfaceorientation InHaptics rendering and applications ed A El Saddik pp 75ndash94Intech [FHD]

Durgin F H Li Z amp Hajnal A (2010b) Slant perception in near space is categ-orically biased Evidence for a vertical tendency Attention Perception andPsychophysics 721875ndash89 [FHD]

Dyde R T Jenkin M R amp Harris L R (2006) The subjective visual vertical andthe perceptual upright Experimental Brain Research 173612ndash22 [MB-C]

Dyer A G Rosa M G P amp Reser D H (2008) Honeybees can recognise imagesof complex natural scenes for use as potential landmarks The Journal ofExperimental Biology 2111180ndash86 [AGD]

Dyer F C (1991) Bees acquire route-based memories but not cognitive maps in afamiliar landscape Animal Behaviour 41239ndash46 [SZ]

Eaton M D Evans D L Hodgson D R amp Rose R J (1995) Effect of treadmillincline and speed on metabolic rate during exercise in thoroughbred horsesJournal of Applied Physiology 79(3)951ndash57 [FTS]

Eckles M A Roubik D W amp Nieh J C (2012) A stingless bee can use visualodometry to estimate both height and distance The Journal of ExperimentalBiology 2153155ndash60 [AGD]

Epstein R A (2008) Parahippocampal and retrosplenial contributions to humanspatial navigation Trends in Cognitive Sciences 12388ndash96 [SRL]

Esch H E amp Burns J E (1995) Honeybees use optic flow to measure the distanceof a food source Naturwissenschaften 8238ndash40 [aKJJ]

Etienne A S amp Jeffery K J (2004) Path integration in mammals Hippocampus14180ndash92 [aKJJ]

European Aviation Safety Agency (2006) EASA Annual Safety Review 2006 EASA(European Aviation Safety Agency Cologne Germany) [MB-C]

Euston D R amp Takahashi T T (2002) From spectrum to space The contributionof level difference cues to spatial receptive fields in the barn owl inferior col-liculus Journal of Neuroscience 22(1)284ndash93 [FTS]

Eyferth K Niessen C amp Spaeth O (2003) A model of air traffic controllersrsquoconflict detection and conflict resolution Aerospace Science and Technology7409ndash16 [HS]

Fenton A A Kao H Y Neymotin S A Olypher A Vayntrub Y Lytton W Wamp Ludvig N (2008) Unmasking the CA1 ensemble place code by exposures tosmall and large environments More place cells and multiple irregularlyarranged and expanded place fields in the larger space Journal of Neuroscience28(44)11250ndash62 [aKJJ]

Finkelstein A Derdikman D Foerster J Las L amp Ulanovsky N (2012) 3-Dhead-direction cells in the bat presubiculum Society for NeuroscienceAbstracts Abstract No 3820319 [FSt]

Foley H J amp Matlin M W (2010) Sensation and perception 5th edition Allyn andBacon [TBdP]

Foley J M Ribeiro N P amp Da Silva J (2004) Visual perception of extent and thegeometry of visual space Vision Research 44147ndash56 [FHD]

Foo P Warren W H Duchon A amp Tarr M J (2005) Do humans integrateroutes into a cognitive map Map- versus landmark-based navigation of novelshortcuts Journal of Experimental Psychology Learning Memory and Cog-nition 31(2)195ndash215 [CCC]

Forbus K D (2011) Qualitative modeling WIREs Cognitive Science 2374ndash91[HS]

Foulsham T amp Kingstone A (2012) Goal-driven and bottom-up gaze in an activereal-world search task In Proceedings of the Symposium on Eye-TrackingResearch and Applications (ETRA rsquo12) Santa Barbara CA March 28-30 2012ed C H Morimoto H O Istance J B Mulligan P Qvarfordt amp S NSpencer pp 189ndash92 ACM Digital Library [KAL]

Foulsham T Walker E amp Kingstone A (2011) The where what and when of gazeallocation in the lab and the natural environment Vision Research 51(17)1920ndash31 [KAL]

Frank L M Brown E N amp Wilson M (2000) Trajectory encoding in the hip-pocampus and entorhinal cortex Neuron 27169ndash78 [HY]

Frankenstein J Buumlchner S J Tenbrink T amp Houmllscher C (2010) Influence ofgeometry and objects on local route choices during wayfinding In SpatialCognition VII Proceedings of the Seventh International Conference on SpatialCognition Mt HoodPortland OR USA August 15-19 2010 ed C HoumllscherT F Shipley M O Belardinelli J A Bateman amp N S Newcombe pp 41ndash53Springer [CH]

Frankenstein J Mohler B J Buumllthoff H H amp Meilinger T (2012) Is the map inour head oriented north Psychological Science 23120ndash25 [AP]

Franklin N amp Tversky B (1990) Searching imagined environments Journal ofExperimental Psychology General 11963ndash76 [KJH]

Fuhs M C VanRhoads S R Casale A E McNaughton B L amp Touretzky D S(2005) Influence of path integration versus environmental orientation on placecell remapping between visually identical environments Journal of Neurophy-siology 942603ndash16 [PAD]

Gagliaro A Ioalegrave P amp Bingman V P (1999) Homing in pigeons The role of thehippocampal formation in the representation of landmarks used for navigationJournal of Neuroscience 19311ndash15 [DN]

Galati G Pelle G Berthoz A amp Committeri G (2010) Multiple reference framesused by the human brain for spatial perception and memory ExperimentalBrain Research 206(2)109ndash20 [rKJJ]

Garling T Anders B Lindberg E amp Arce C (1990) Is elevation encoded incognitive maps Journal of Environmental Psychology 10341ndash51 [aKJJ]

Georges-Franccedilois P Rolls E T amp Robertson R G (1999) Spatial view cells in theprimate hippocampus Allocentric view not head direction or eye position orplace Cerebral Cortex 9197ndash212 [SRL]

Gevers W Lammertyn J Notebaert W Verguts T amp Fias W (2006) Automaticresponse activation of implicit spatial information Evidence from the SNARCeffect Acta Psychologica 122221ndash33 [ABa]

Gibson J J (1979) The ecological approach to visual perception HoughtonMifflin [CCC FHD]

Gibson J J amp Cornsweet J (1952) The perceived slant of visual surfaces ndashOpticaland geographical Journal of Experimental Psychology 44(1)11ndash15 [aKJJ]

Gilbert A L Regier T Kay P amp Ivry R B (2006) Whorf hypothesis is supportedin the right visual field but not the left Proceedings of the National Academy ofSciences USA 103489ndash94 [KJH]

Giudice N A Klatzky R L Bennett C amp Loomis J M (2013) Perception of 3-Dlocation based on vision touch and extended touch Experimental BrainResearch 224141ndash53 [RLK]

Goodale M A amp Milner A D (1992) Separate visual pathways for perception andaction Trends in Neurosciences 1520ndash25 [SRL]

Goodridge J P Dudchenko P A Worboys K A Golob E J amp Taube J S(1998) Cue control and head direction cells Behavioral Neuroscience 112(4)749ndash61 [aKJJ]

Grah G Wehner R amp Ronacher B (2007) Desert ants do not acquire and use athree-dimensional global vector Frontiers in Zoology 412 [aKJJ RFW]

Gregory E amp McCloskey M (2010) Mirror-image confusions Implications forrepresentation and processing of object orientation Cognition 116(1)110ndash29doi 101016jcognition201004005 [IB]

Grieves R M amp Dudchenko P A (2013) Cognitive maps and spatial inference inanimals Rats fail to take a novel shortcut but can take a previously experiencedone Learning and Motivation 4481ndash92 [PAD]

Griffin D R (1958) Listening in the dark Yale University Press [CFM]Grobeacutety M C amp Schenk F (1992a) Spatial learning in a three-dimensional maze

Animal Behaviour 43(6)1011ndash20 [aKJJ TBdP DN]Grobeacutety M-C amp Schenk F (1992b) The influence of spatial irregularity upon

radial-maze performance in the rat Animal Learning and Behavior 20(4)393ndash400 [aKJJ]

Grush R (2007) Skill theory v20 Dispositions emulation and spatial perceptionSynthese 159(3)389ndash416 [DMK]

Gu Y Fetsch C R Adeyemo B DeAngelis G C amp Angelaki D E (2010)Decoding of MSTd population activity accounts for variations in the precision ofheading perception Neuron 66(4)596ndash609 [GAO]

Hadjidimitrakis K Breveglieri R Bosco A amp Fattori P (2012) Three-dimen-sional eye position signals shape both peripersonal space and arm movementactivity in the medial posterior parietal cortex Frontiers in Integratve Neuro-science 637 doi 103389fnint201200037 [SRL]

Hafting T Fyhn M Molden S Moser M B amp Moser E I (2005) Micro-structure of a spatial map in the entorhinal cortexNature 436801ndash806 [aKJJ]

Hajnal A Abdul-Malak D T amp Durgin F H (2011) The perceptual experience ofslope by foot and by finger Journal of Experimental Psychology Human Per-ception and Performance 37709ndash19 [FHD]

Hartley T Burgess N Lever C Cacucci F amp OrsquoKeefe J (2000) Modelling placefields in terms of the cortical inputs to the hippocampus Hippocampus 10369ndash79 [PAD]

Harvey P H amp Pagel M D (1991) The comparative method in evolutionarybiology Oxford University Press [DN]

Hasselmo M E Giocomo L M amp Zilli E A (2007) Grid cell firing may arise frominterference of theta frequency membrane potential oscillations in singleneurons Hippocampus 17(12)1252ndash71 [CFM]

Hayman R Verriotis M A Jovalekic A Fenton A A amp Jeffery K J (2011) Ani-sotropic encoding of three-dimensional space by place cells and grid cellsNatureNeuroscience 14(9)1182ndash88 [arKJJ MB-C TBdP PAD DMK FSa HY]

Henderson J Hurly T A amp Healy S D (2001) Rufous hummingbirdsrsquo memoryfor flower location Animal Behaviour 61(5)981ndash86 [aKJJ]

Henderson J Hurly T A amp Healy S D (2006) Spatial relational learningin rufous hummingbirds (Selasphorus rufus) Animal Cognition 9201ndash205[aKJJ]

Hengstenberg R (1991) Gaze control in the blowfly Calliphora A multisensorytwo-stage integration process Seminars in Neuroscience 3(1)19ndash29 Availableat httpwwwsciencedirectcomsciencearticlepii104457659190063T[MB-C]

Hess D Koch J amp Ronacher B (2009) Desert ants do not rely on sky compassinformation for the perception of inclined path segments Journal of Exper-imental Biology 212(10)1528ndash34 [aKJJ]

Heys JG MacLeod K Moss CF and Hasselmo M (2013) Bat and rat neuronsdiffer in theta frequency resonance despite similar spatial coding Science340363ndash67 [CFM]

Higashiyama A amp Ueyama E (1988) The perception of vertical and horizontaldistances in outdoor settings Perception and Psychophysics 44151ndash56 doi103758BF03208707 [FHD]

Hill A J amp Best P J (1981) Effects of deafness and blindness on the spatial cor-relates of hippocampal unit activity in the rat Experimental Neurology 74204ndash17 [AP]

Hill C (1982) Updown frontback leftright A contrastive study of Hausa andEnglish InHere and there Crosslinguistic studies on deixis and demonstrationed J Weissenborn amp W Klein pp 11ndash42 John Benjamins [IB]

Hillier B amp Hanson J eds (1984) The social logic of space Cambridge UniversityPress [CH]

Holbrook R I amp Burt de Perera T B (2009) Separate encoding of vertical andhorizontal components of space during orientation in fish Animal Behaviour 78(2)241ndash45 [TBdP CCC aKJJ]

Holbrook R I amp Burt de Perera T (2011a) Fish navigation in the verticaldimension Can fish use hydrostatic pressure to determine depth Fish andFisheries 12(4)370ndash379 [aKJJ]

Holbrook R I amp Burt de Perera T (2011b) Three-dimensional spatial cognitionInformation in the vertical dimension overrides information from the horizon-tal Animal Cognition 14613ndash19 [TBdP]

Holmes K J (2012) Language as a window into the mind The case of spaceDoctoral dissertation Department of Psychology Emory University [KJH]

Holmes K J amp Wolff P (2012) Does categorical perception in the left hemispheredepend on language Journal of Experimental Psychology General 141439ndash43 [KJH]

Holmes K J amp Wolff P (2013a) Spatial language and the psychological reality ofschematization Cognitive Processing 14205ndash208 [KJH]

Holmes K J amp Wolff P (2013b) When is language a window into the mindLooking beyond words to infer conceptual categories In Proceedings of the35th annual conference of the Cognitive Science Society eds M KnauffM Pauen N Sebanz amp I Wachsmuth pp 597ndash602 Cognitive ScienceSociety [KJH]

Houmllscher C Buumlchner S Broumlsamle M Meilinger T amp Strube G (2007) Signs andmaps ndashCognitive economy in the use of external aids for indoor navigation InProceedings of the 29th Annual Cognitive Science Society ed D S McNamaraamp J G Trafton pp 377ndash82 Cognitive Science Society [CH]

Houmllscher C Buumlchner S Meilinger T amp Strube G (2009) Adaptivity of way-finding strategies in a multi-building ensemble The effects of spatial structuretask requirements and metric information Journal of Environmental Psychol-ogy 29(2)208ndash19 [CH]

Houmllscher C Broumlsamle M amp Vrachliotis G (2012) Challenges in multi-level way-finding A case-study with space syntax technique Environment and PlanningB Planning and Design 3963ndash82 [CH]

Houmllscher C Meilinger T Vrachliotis G Broumlsamle M amp Knauff M (2006) Upthe down staircase Wayfinding strategies and multi-level buildings Journal ofEnvironmental Psychology 26(4)284ndash99 [aKJJ ABe CCC CH AP]

Hommel B Muumlsseler J Aschersleben G amp Prinz W (2001) The theory of eventcoding (TEC) A framework for perception and action planning Behavioral andBrain Sciences 24849ndash78 [ABa]

Howard A Bernardes S Nibbelink N Biondi L Presotto A Fragaszy D ampMadden M (2012) A maximum entropy model of the bearded capuchinmonkey habitat incorporating topography and spectral unmixing analysisAnnals of the International Society of Photogrammetry and Remote Sensing 1(2)7ndash11 [AMH]

Hubbard E M Piazza M Pinel P amp Dehaene S (2005) Interactions betweennumber and space in parietal cortex Nature Reviews Neuroscience 6435ndash48[ABa]

Huttenlocher J Hedges L V amp Duncan S (1991) Categories and particularsPrototype effects in estimating spatial location Psychological Review 98352ndash76 [RFW]

Hyvaumlrinen J Hyvaumlrinen L amp Linnankoski I (1981) Modification of parietalassociation cortex and functional blindness after binocular deprivation in youngmonkeys Experimental Brain Research 421ndash8 [AP]

Indovina I Maffei V Pauwels K Macaluso E Orban G A amp Lacquaniti F(2013 Simulated self-motion in a visual gravity field Brain sensitivity to verticaland horizontal heading in the human brain NeuroImage 71C114ndash24 [GAO]

Ingram J N amp Wolpert D M (2011) Naturalistic approaches to sensorimotorcontrol Progress in Brain Research 1913ndash29 [KAL]

Jacobs J Kahana M J Ekstrom A D Mollison M V amp Fried I (2010) A senseof direction in human entorhinal cortex Proceedings of the National Academy ofSciences USA 107(14)6487ndash92 [GAO]

Janson C H (1998) Experimental evidence for spatial memory in foraging wildcapuchin monkeys Cebus apella Animal Behaviour 55(5)1229ndash43 doi101006anbe19970688 [AMH]

Janson C H (2007) Experimental evidence for route integration and strategicplanning in wild capuchin monkeys Animal Cognition 10(3)341ndash56doi101007s10071-007-0079-2 [AMH]

Jeffery K J (2007) Integration of the sensory inputs to place cells What wherewhy and how Hippocampus 17(9)775ndash85 [aKJJ]

Jeffery K J Anand R L amp Anderson M I (2006) A role for terrain slope inorienting hippocampal place fields Experimental Brain Research 169(2)218ndash25 [aKJJ]

Jeffery K J amp Burgess N (2006) A metric for the cognitive map ndash Found at lastTrends in Cognitive Sciences 10(1)1ndash3 [aKJJ]

Jensen M E Moss C F amp Surlykke A (2005) Echolocating bats and their use oflandmarks and spatial memory Journal of Experimental Biology 2084399ndash410 [CFM]

Jovalekic A Hayman R Becares N Reid H Thomas G Wilson J amp Jeffery K(2011) Horizontal biases in ratsrsquo use of three-dimensional space BehaviouralBrain Research 222(2)279ndash88 [aKJJ TBdP DN RFW]

Judd S amp Collett T (1998) Multiple stored views and landmark guidance in antsNature 392710ndash14 [SZ]

Kamil A C Balda R P amp Olson D J (1994) Performance of four seed-cachingcorvid species in the radial arm-maze analog Journal of Comparative Psychol-ogy 108385ndash93 [DN]

Kammann R (1967) The overestimation of vertical distance and its role in the moonillusion Attention Perception and Psychophysics 2(12)585ndash89 [aKJJ]

Kelly J (2011) Head for the hills The influence of environmental slant on spatialmemory organization Psychonomic Bulletin and Review 18(4)774ndash80 [aKJJ]

Kelly J W Loomis J M amp Beall A C (2004) Judgments of exocentric direction inlarge-scale space Perception 33443ndash54 [FHD]

Killian N J Jutras M J amp Buffalo E A (2012) A map of visual space in theprimate entorhinal cortex Nature 491(7426)761ndash64 [GAO]

Klatzky R L (1998) Allocentric and egocentric spatial representations Definitionsdistinctions and interconnections In Spatial cognition ndash An interdisciplinaryapproach to representation and processing of spatial knowledge (Lecture Notesin Artificial Intelligence vol 1404) ed C Freksa C Habel amp K F Wenderpp 1ndash17 Springer-Verlag [RLK]

Kluzik J Horak F B amp Peterka R J (2005) Differences in preferred referenceframes for postural orientation shown by after-effects of stance on an inclinedsurface Experimental Brain Research 162474ndash89 Available at httplinkspringercomarticle1010072Fs00221-004-2124-6 [MB-C]

Knauff M Rauh R amp Renz J (1997) A cognitive assessment of topological spatialrelations Results from an empirical investigation In Proceedings of the 3rdInternational Conference on Spatial Information Theory (COSITrsquo97) ed S CHirtle amp A U Frank pp 193ndash206 Springer [HS]

Knierim J J amp Hamilton D A (2011) Framing spatial cognition Neural rep-resentations of proximal and distal frames of reference and their roles in navi-gation Physiological Reviews 911245ndash79 [FSa]

Knierim J J amp McNaughton B L (2001) Hippocampal place-cell firing duringmovement in three-dimensional space Journal of Neurophysiology 85(1)105ndash16 [aKJJ]

Knierim J J McNaughton B L amp Poe G R (2000) Three-dimensional spatialselectivity of hippocampal neurons during space flight Nature Neuroscience 3(3)209ndash10 [aKJJ DMK]

Knierim J J Poe G R amp McNaughton B L (2003) Ensemble neural coding ofplace in zero-g In The Neurolab Spacelab Mission Neuroscience research inspace Results from the STS-90 Neurolab Spacelab Mission NASA Report SP-2003ndash535 ed J C Buckey Jr amp J L Homick pp 63ndash68 NASA [aKJJ]

Knierim J J amp Rao G (2003) Distal landmarks and hippocampal place cells Effectsof relative translation versus rotation Hippocampus 13604ndash17 [FSa]

Knudsen E I amp Konishi M (1978a) A neural map of auditory space in the owlScience 200795ndash97 [FTS]

Knudsen E I amp Konishi M (1978b) Space and frequency are represented separ-ately in auditory midbrain of the owl Journal of Neurophysiology 41870ndash84[FTS]

Konishi M (1973) How the owl tracks its preyAmerican Science 61414ndash24 [FTS]Kosslyn S M Koenig O Barrett A Cave C B Tang J amp Gabrieli J D E

(1989) Evidence for two types of spatial representations Hemispheric special-ization for categorical and coordinate relations Journal of Experimental Psy-chology Human Perception and Performance 15723ndash35 [KJH]

Kravitz D J Saleem K S Baker C I amp Mishkin M (2011) A new neural fra-mework for visuospatial processing Nature Reviews Neuroscience 12(4)217ndash30 [SRL GAO]

Kravitz D J Saleem K S Baker C I Ungerleider L G amp Mishkin M (2013)The ventral visual pathway An expanded neural framework for the processingof object quality Trends in Cognitive Sciences 17(1)26ndash49 [GAO]

Krogh A (1929) The progress of physiology American Journal of Physiology90243ndash51 [CFM]

Kropff E amp Treves A (2008) The emergence of grid cells Intelligent design or justadaptation Hippocampus 18(12)1256ndash69 doi 101002hipo20520 [FSt]

Krumnack A Bucher L Nejasmic J Nebel B amp Knauff M (2011) A model forrelational reasoning as verbal reasoning Cognitive Systems Research 11377ndash92 [HS]

Kubovy M (2002) Phenomenology psychological In Encyclopedia of cognitivescience ed L Nadel pp 579ndash86 Nature Publishing Group [IB]

Kubovy M amp Gepshtein S (2003) Grouping in space and in space-time An exercisein phenomenological psychophysics In Perceptual organization in vision Be-havioral and neural perspectives ed R Kimchi M Behrmann amp C R Olsonpp 45ndash86 Erlbaum [IB]

Kuipers B Browning R Gribble B Hewett M amp Remolina E (2000) Thespatial semantic hierarchy Artificial Intelligence 119191ndash233 [HP]

Kupers R Chebat D R Madsen K H Paulson O B amp Ptito M (2010) Neuralcorrelates of virtual route recognition in congenital blindness Proceedings of theNational Academy of Sciences USA 10712716ndash21 [AP]

Lackner J R amp Graybiel A (1983) Perceived orientation in free-fall depends onvisual postural and architectural factors Aviation Space and EnvironmentalMedicine 54(1)47ndash51 [aKJJ]

Lagae L Maes H Raiguel S Xiao D K amp Orban G A (1994) Responses ofmonkey STS neurons to optic flow components A comparison of areas MT andMST Journal of Neurophysiology 71(5)1597ndash626 [GAO]

Lahav O amp Mioduser D (2008) Haptic-feedback support for cognitive mapping ofunknown spaces by people who are blind International Journal of Human-Computer Studies 6623ndash35 [AP]

Landau B amp Jackendoff R (1993) ldquoWhatrdquo and ldquowhererdquo in spatial language andspatial cognition Behavioral and Brain Sciences 16217ndash65 [KJH]

Langston R F Ainge J A Couey J J Canto C B Bjerknes T L Witter M PMoser E I amp Moser M-B (2010) Development of the spatial representationsystem in the rat Science 328(5985)1576ndash80 [rKJJ]

Lappe M Bremmer F amp van den Berg A V (1999) Perception of self-motionfrom visual flow Trends in Cognitive Sciences 3(9)329ndash36 [RFW]

Larsen D D Luu J D Burns M E amp Krubitzer L (2009) What are the effectsof severe visual impairment on the cortical organization and connectivity ofprimary visual cortex Frontiers in Neuroanatomy 330 [AP]

Lavie N (2005) Distracted and confused Selective attention under load Trends inCognitive Sciences 9(2)75ndash82 [rKJJ]

Lehky S R amp Sereno A B (2011) Population coding of visual space ModelingFrontiers in Computational Neuroscience 4155 doi1103389fncom201000155 [SRL]

Lehrer M (1994) Spatial vision in the honeybee The use of different cues indifferent tasks Vision Research 34(18)2363ndash85 [CCC]

Lent D D Graham P amp Collett T S (2010) Image-matching during ant navi-gation occurs through saccade-like body turns controlled by learned visualfeatures Proceedings of the National Academy of Sciences USA 10716348ndash53 [TBdP]

Leporeacute N Shi Y Lepore F Fortin M Voss P Chou Y Y Lord C LassondeM Dinov I D Toga AW amp Thompson PM (2009) Patterns of hippocampalshape and volume differences in blind subjects NeuroImage 46949 [AP]

Leutgeb S Leutgeb J K Treves A Moser M B amp Moser E I (2004) Distinctensemble codes in hippocampal areas CA3 and CA1 Science 305(5688)1295ndash98 [aKJJ]

Lever C Burton S Jeewajee A OrsquoKeefe J amp Burgess N (2009) Boundaryvector cells in the subiculum of the hippocampal formation Journal of Neuro-science 299771ndash77 [FSa]

Li Z amp Durgin F H (2009) Downhill slopes look shallower from the edge Journalof Vision 9(11)6ndash15 [aKJJ HER]

Li Z amp Durgin F H (2010) Perceived slant of binocularly viewed large-scalesurfaces A common model from explicit and implicit measures Journal ofVision 10(14)article 131ndash16 [aKJJ FHD]

Li Z amp Durgin F H (2012) A comparison of two theories of perceived distance onthe ground plane The angular expansion hypothesis and the intrinsic biashypothesis i-Perception 3368ndash383 [FHD]

Li Z Phillips J amp Durgin F H (2011) The underestimation of egocentric distanceEvidence from frontal matching tasks Attention Perception and Psychophysics732205ndash2217 [FHD]

Li Z Sun E Strawser C J Spiegel A Klein B amp Durgin F H (2013) On theanisotropy of perceived ground extents and the interpretation of walked dis-tance as a measure of perception Journal of Experimental Psychology HumanPerception and Performance 39477ndash493 [FHD]

Liang J C Wagner A D amp Preston A R (2013) Content representation in thehuman medial temporal lobe Cerebral Cortex 2380ndash96 [GAO]

Liu Y Vogels R amp Orban G A (2004) Convergence of depth from texture anddepth from disparity in macaque inferior temporal cortex Journal of Neuro-science 24(15)3795ndash800 [GAO]

Loetscher T Bockisch C J Nicholls M E R amp Brugger P (2010) Eyeposition predicts what number you have in mind Current Biology 20264ndash65[ABa]

Longstaffe K A Hood B M amp Gilchrist I D (2012) Inhibition attention andworking memory in large scale search Poster presented at the 5th InternationalConference on Spatial Cognition ldquoSpace and Embodied Cognitionrdquo RomeItaly September 4ndash8 2012 [KAL]

Loomis J L Klatzky R L amp Giudice N A (2013) Representing 3D space inworking memory Spatial images from vision hearing touch and language InMultisensory imagery Theory and applications ed S Lacey amp R Lawson pp131ndash55 Springer [RLK]

Loomis J M Da Silva J A Fujita N amp Fukusima S S (1992) Visual spaceperception and visually directed action Journal of Experimental PsychologyHuman Perception and Performance 18(4)906ndash921 [aKJJ FHD]

Loomis J M Klatzky R L Golledge R G Cicinelli J G Pellegrino J W amp FryP A (1993) Nonvisual navigation by blind and sighted Assessment of pathintegration ability Journal of Experimental Psychology General 12273ndash91[MB-C RFW]

Loomis J M Lippa Y Klatzky R L amp Golledge R G (2002) Spatial updatingof locations specified by 3-D sound and spatial language Journal ofExperimental Psychology Human Learning Memory and Cognition28335ndash45 [RLK]

Loomis J M amp Philbeck J W (1999) Is the anisotropy of perceived 3-D shapeinvariant across scale Perception and Psychophysics 61397ndash402 [FHD]

Maas H (2010) Forestry applications In Airborne and terrestrial laser scanninged G Vosselman amp H Maas pp 213ndash35 Whittles [AMH]

MacEachren A M (1986) A linear view of the world Strip maps as a unique form ofcartographic representation American Cartographer 137ndash25 [HS]

MacNeilage P R Banks M S Berger D R amp Buumllthoff H H (2007) A Bayesianmodel of the disambiguation of gravitoinertial force by visual cues ExperimentalBrain Research 179263ndash90 [MB-C]

MacNeilage P R Banks M S DeAngelis G C amp Angelaki D E (2010) Ves-tibular heading discrimination and sensitivity to linear acceleration in head andworld coordinates The Journal of Neuroscience 309084ndash94 [MB-C]

Mandel J T Bildstein K L Bohrer G amp Winkler D W (2008) Movementecology of migration in turkey vultures Proceedings of the National Academy ofSciences USA 105(49)19102ndash107 doi101073pnas0801789105 [AMH]

Marr D (1982) Vision A computational investigation into the human representationand processing of visual information WH Freeman [DMWP]

Mattingley J B Bradshaw J L amp Phillips J G (1992) Reappraising unilateralneglect Australian Journal of Psychology 44163ndash69 [JGP]

McIntyre J Zago M Berthoz A amp Lacquaniti F (2001) Does the brain modelNewtonrsquos laws Nature Neuroscience 4693ndash4 [MB-C]

McNamara T P (1986) Mental representations of spatial relations Cognitive Psy-chology 1887ndash121 [HS]

McNamara T P amp Diwadkar V (1997) Symmetry and asymmetry of human spatialmemory Cognitive Psychology 34160ndash90 [HS]

Merfeld D M Young L R Oman C M amp Shelhamer M J (1993) A multidi-mensional model of the effect of gravity on the spatial orientation of themonkey Journal of Vestibular Research 3(2)141ndash61 [aKJJ]

Merfeld D M amp Zupan L H (2002) Neural processing of gravitoinertial cues inhumans III Modeling tilt and translation responses Journal of Neurophysiol-ogy 87(2)819ndash33 [aKJJ]

Messing R M ampDurgin F H (2005) Distance perception and the visual horizon inhead-mounted displays Transactions on Applied Perception 2234ndash50[FHD]

Michaux N Pesenti M Badets A Di Luca S amp Andres M (2010) Let usredeploy attention to sensorimotor experience Behavioral and Brain Sciences33283ndash84 [ABa]

Mittelstaedt H (1983) A new solution to the problem of the subjective verticalNaturwissenschaften 70272ndash81 [MB-C]

Mittelstaedt H (1998) Origin and processing of postural information Neuroscienceand Biobehavioral Reviews 22(4)473ndash78 [aKJJ]

Moar I amp Bower G (1983) Inconsistency in spatial knowledge Memory andCognition 11107ndash13 [HS]

Moghaddam M Kaminsky Y L Zahalka A amp Bures J (1996) Vestibular navi-gation directed by the slope of terrain Proceedings of the National Academy ofSciences of the United States of America 93(8)3439ndash43 [aKJJ]

Montello D R (1991) The measurement of cognitive distance Methods and con-struct-validity Journal of Environmental Psychology 11(2)101ndash22 [CCC]

Montello D R (2005) Navigation In The Cambridge handbook of visuospatialthinking ed P Shah amp A Miyake pp 257ndash94 Cambridge University Press[CH]

Montello D R amp Pick H L (1993) Integrating knowledge of vertically aligned large-scale spaces Environment and Behavior 25(3)457ndash84 [aKJJ ABe CCC]

Morishita H amp Hensch T K (2008) Critical period revisited Impact on visionCurrent Opinion in Neurobiology 18(1)101ndash107 [rKJJ]

Moser E I Kropff E amp Moser M B (2008) Place cells grid cells and the brainrsquosspatial representation system Annual Review of Neuroscience 3169ndash89[aKJJ]

Moser E I amp Moser M B (2008) A metric for space Hippocampus 18(12)1142ndash56 [aKJJ]

Mueller M amp Wehner R (2010) Path integration provides a scaffold for landmarklearning in desert ants Current Biology 201368ndash71 [TBdP]

Muller R U amp Stead M (1996) Hippocampal place cells connected by Hebbiansynapses can solve spatial problems Hippocampus 6(6)709ndash19 [FTS]

Nardi D amp Bingman V P (2009a) Pigeon (Columba livia) encoding of a goallocation The relative importance of shape geometry and slope informationJournal of Comparative Psychology 123204ndash16 [DN]

Nardi D amp Bingman V P (2009b) Slope-based encoding of a goal location isunaffected by hippocampal lesions in homing pigeons (Columba livia) Behav-ioural Brain Research 205(1)322ndash26 [aKJJ DN]

Nardi D Mauch R J Klimas D B amp Bingman V P (2012) Use of slope andfeature cues in pigeon (Columba livia) goal-searching behavior Journal ofComparative Psychology 126288ndash93 [DN]

Nardi D Newcombe N S amp Shipley T F (2011) The world is not flat Can peoplereorient using slope Journal of Experimental Psychology Learning Memoryand Cognition 37(2)354ndash67 [aKJJ]

Nardi D Nitsch K P amp Bingman V P (2010) Slope-driven goal location behaviorin pigeons Journal of Experimental Psychology Animal Behavior Processes 36(4)430ndash42 [aKJJ DN]

Nieder A amp Miller E K (2004) A parieto-frontal network for visual numericalinformation in the monkey Proceedings of the National Academy of SciencesUSA 1017457ndash62 [ABa]

Nieh J C Contrera F A amp Nogueira-Neto P (2003) Pulsed mass recruitment bya stingless bee Trigona hyalinata Proceedings of the Royal Society of LondonB Biological Sciences 270(1529)2191ndash96 [aKJJ]

Nieh J C amp Roubik D W (1998) Potential mechanisms for the communication ofheight and distance by a stingless bee Melipona panamica Behavioral Ecologyand Sociobiology 43(6)387ndash99 [aKJJ CCC]

Nitz D A (2011) Path shape impacts the extent of CA1 pattern recurrence both withinand across environments Journal of Neurophysiology 1051815ndash24 [FSa]

Normand E amp Boesch C (2009) Sophisticated Euclidean maps in forest chim-panzees Animal Behaviour 77(5)1195ndash201 doi101016janbehav200901025 [AMH]

Noser R amp Byrne R (2007a) Mental maps in chacma baboons (Papio ursinus)Using inter-group encounters as a natural experiment Animal Cognition 10(3)331ndash40 doi 101007s10071-006-0068-x [AMH]

Noser R amp Byrne R (2007b) Travel routes and planning of visits to out-of-sightresources in wild chacma baboons Papio ursinus Animal Behaviour 73(2)257ndash66 doi 101016janbehav200604012 [AMH]

OrsquoKeefe J amp Burgess N (1996) Geometric determinants of the place fields ofhippocampal neurons Nature 381425ndash28 [PAD FSa]

OrsquoKeefe J amp Dostrovsky J (1971) The hippocampus as a spatial map Preliminaryevidence from unit activity in the freely-moving rat Brain Research 34(1)171ndash75 [aKJJ]

OrsquoKeefe J amp Nadel L (1978) The hippocampus as a cognitive map ClarendonPress [aKJJ]

Oman C (2007) Spatial orientation and navigation in microgravity In Spatialprocessing in navigation imagery and perception ed F W Mast amp L JanckeSpringer [aKJJ]

Oman C M Lichtenberg B K Money K E amp McCoy R K (1986) MITCanadian vestibular experiments on the Spacelab-1 mission 4 Space motionsickness Symptoms stimuli and predictability Experimental Brain Research64(2)316ndash34 [aKJJ]

Ooi T L amp He Z J (2007) A distance judgment function based on space per-ception mechanisms Revisiting Gilinskyrsquos (1951) equation PsychologicalReview 114441ndash54 [FHD]

Ooi T L Wu B amp He Z J (2001) Distance determined by the angular declinationbelow the horizon Nature 414197ndash99 [FHD AP]

Orban G A (2011) The extraction of 3D shape in the visual system of human andnonhuman primates Annual Review of Neuroscience 34361ndash88 [GAO]

OrsquoShea R P amp Ross H E (2007) Judgments of visually perceived eye level(VPEL) in outdoor scenes Effects of slope and height Perception361168ndash78 [HER]

Pasqualotto A amp Proulx M J (2012) The role of visual experience for theneural basis of spatial cognition Neuroscience and Biobehavioral Reviews361179ndash87 [AP]

Pasqualotto A Spiller M J Jansari A S amp Proulx M J (2013) Visual experienceis necessary for the representation of spatial patterns Behavioural BrainResearch 236175ndash79 [AP]

Passini R amp Proulx G (1988) Wayfinding without vision An experiment withcongenitally totally blind people Environment and Behavior 20227ndash52 [AP]

Payne R S (1971) Acoustic location of prey by barn owls (Tyto alba) Journal ofExperimental Biology 54535ndash73 [FTS]

Paz-Villagraacuten V Lenck-Santini P-P Save E amp Poucet B (2002) Properties ofplace cell firing after damage to the visual cortex European Journal of Neuro-science 16771ndash76 [AP]

Pellicano E Smith A D Cristino F Hood B M Briscoe J amp Gilchrist I D(2011) Children with autism are neither systematic nor optimal foragers Pro-ceedings of the National Academy of Sciences USA 108(1)421ndash26 [KAL]

Peponis J Zimring C amp Choi Y K (1990) Finding the building in wayfindingEnvironment and Behavior 22(5)555ndash90 [CH]

Peacuteruch P Vercher J L amp Gauthier G M (1995) Acquisition of spatial knowledgethrough visual exploration of simulated environments Ecological Psychology71ndash20 [AP]

Phillips F amp Layton O (2009) The traveling salesman problem in the naturalenvironment Journal of Vision 9(8)1145 [aKJJ]

Phillips J G amp Ogeil R P (2010) Curved motions in horizontal and verticalorientations Human Movement Science 29737ndash50 [JGP]

Phillips J G Triggs T J amp Meehan J W (2005) Forwardup directional incom-patibilities during cursor placement within graphical user interfaces Ergo-nomics 48722ndash35 [JGP]

Pick H L amp Rieser J J (1982) Childrenrsquos cognitive mapping In Spatial orien-tation Development and physiological bases ed M Potegal pp 107ndash28Academic Press [aKJJ]

Pigarev I N amp Levichkina E V (2011) Distance modulated neuronal activity in thecortical visual areas of cats Experimental Brain Research 214(1)105ndash11 [rKJJ]

Pinker S (2007) The stuff of thought Language as a window into human naturePenguin Books [KJH]

Postma A Zuidhoek S Noordzij M L amp Kappers A M L (2007) Differencesbetween early-blind late-blind and blindfolded-sighted people in hapticspatial configuration learning and resulting memory traces Perception 361253ndash65 [AP]

Poucet B Save E amp Lenck-Santini P-P (2000) Sensory and memory properties ofplace cells firing Reviews in the Neurosciences 1195ndash111 [AP]

Pouget A Deneve S Ducom J-C amp Latham P E (1999) Narrow versus widetuning curves Whatrsquos best for a population code Neural Computation 1185ndash90 [SRL]

Pouget A Fisher S amp Sejnowski T J (1993) Egocentric representation in earlyvision Journal of Cognitive Neuroscience 5(2)150ndash61 [DMK]

Pouget A amp Sejnowski T J (1997) Spatial transformations in the parietal cortexusing basis functions Journal of Cognitive Neuroscience 9(2)222ndash37 [DMK]

Powers W T (1973) Behavior The control of perception Aldine [FHD]Pravosudov V V amp Clayton N S (2002) A test of the adaptive specialization

hypothesis Population differences in caching memory and the hippocampus inblack-capped chickadees (Poecile atricapilla) Behavioral Neuroscience116515ndash22 [DN]

Proffitt D R Bhalla M Gossweiler R amp Midgett J (1995) Perceiving geo-graphical slant Psychonomic Bulletin and Review 2(4)409ndash28 [aKJJ FHDHER]

Proulx M J Brown D J Pasqualotto ANeuroscience and Biobehavioral Reviews[httpdxdoiorg101016jneubiorev201211017] Epub ahead of print[AP]

Proulx M J Brown D J Pasqualotto A amp Meijer P (2012) Multisensory per-ceptual learning and sensory substitution Neuroscience and BiobehavioralReviews 2012 Dec 7 pii S0149-7634(12)00207-2 [AP]

Proulx M J Stoerig P Ludowig E amp Knoll I (2008) Seeing ldquowhererdquo through theears Effects of learning-by-doing and long-term sensory deprivation on local-ization based on image-to-sound substitution PLoS ONE 3e1840 [AP]

Quinn P C Polly J L Furer M J Dobson V amp Narter D B (2002) Younginfantsrsquo performance in the object-variation version of the above-below categ-orization task A result of perceptual distraction or conceptual limitationInfancy 3323ndash47 [AGD]

Reppert S M (2006) A colorful model of the circadian clock Cell 124233ndash36[JGP]

Restat J D Steck S D Mochnatzki H F amp Mallot H A (2004) Geographicalslant facilitates navigation and orientation in virtual environments Perception 33(6)667ndash87 [aKJJ SMW]

Riener C R Stefanucci J K Proffitt D R amp Clore G (2011) An effect of moodon the perception of geographical slant Cognition and Emotion 25(1)174ndash82 [aKJJ]

Rieser J J Pick H L Jr Ashmead D amp Garing A (1995) Calibration of humanlocomotion and models of perceptual-motor organization Journal of Exper-imental Psychology Human Perception and Performance 21480ndash97 [FHD]

Roumlder B Kusmierek A Spence C amp Schicke T (2007) Developmental visiondetermines the reference frame for the multisensory control of action Pro-ceedings of the National Academy of Sciences USA 1044753ndash58 [AP]

Rolls E T (1999) Spatial view cells and the representation of place in the primatehippocampus Hippocampus 9467ndash80 [SRL]

Ross H E (1974) Behaviour and perception in strange environments Allen ampUnwin [HER]

Ross H E (1994) Scaling the heights Scottish Mountaineering Club Journal 255(185)402ndash10 [HER]

Ross H E (2006) Judgements of frontal slope in nearby outdoor scenes In FechnerDay 2006 ed D E Kornbrot R M Msetfi amp A W MacRae pp 257ndash62International Society for Psychophysics [HER]

Ross H E (2010) Perspective effects in frontal slope perception In Fechner Day2010 ed A Bastianelli amp G Vidotto pp 87ndash92 International Society forPsychophysics [HER]

Roswarski T E amp Proctor R W (1996) Multiple spatial codes and temporal overlapin choice reaction tasks Psychological Research 59196ndash211 [IB]

Sabra A I (1989) The Optics of Ibn al-Haytham Books I-III On direct vision Vol1 trans A I Sabra pp 163ndash66 The Warburg Institute University ofLondon [HER]

Sampaio C amp Wang R F (2009) Category-based errors and the accessibility ofunbiased spatial memories A retrieval model Journal of Experimental Psy-chology Learning Memory and Cognition 35(5)1331ndash37 [RFW]

Sargolini F Fyhn M Hafting T McNaughton B L Witter M P Moser M Bamp Moser E I (2006) Conjunctive representation of position direction andvelocity in entorhinal cortex Science 312(5774)758ndash62 [aKJJ]

Sato N Sakata H Tanaka Y L amp Taira M (2006) Navigation-associated medialparietal neurons in monkeys Proceedings of the National Academy of SciencesUSA 10317001ndash7006 [ABa]

Savardi U amp Bianchi I (2009) The spatial path to contrariety In The perceptionand cognition of contraries ed U Savardi pp 63ndash92 McGraw-Hill [IB]

Savardi U Bianchi I amp Bertamini M (2010) Naive prediction of orientation andmotion in mirrors From what we see to what we expect reflections to do ActaPsychologica 134(1)1ndash15 doi 101016JACTPSY200911008 [IB]

Savelli F amp Knierim J J (2012) Flexible framing of spatial representations in thehippocampal formation Poster presented at the Society for NeuroscienceConference New Orleans LA October 13ndash17 2012 Poster No 81213[FSa]

Savelli F Yoganarasimha D amp Knierim J J (2008) Influence of boundary removalon the spatial representations of the medial entorhinal cortex Hippocampus181270ndash82 [FSa]

Schaafsma S J amp Duysens J (1996) Neurons in the ventral intraparietal area ofawake macaque monkey closely resemble neurons in the dorsal part of themedial superior temporal area in their responses to optic flow patterns Journalof Neurophysiology 76(6)4056ndash68 [GAO]

Schaumlfer M amp Wehner R (1993) Loading does not affect measurement of walkingdistance in desert ants Cataglyphis fortis Verhandlungen der Deutschen Zool-ogischen Gesellschaft 86270 [aKJJ]

Schnall S Harber K D Stefanucci J K amp Proffitt D R (2012) Social supportand the perception of geographical slant Journal of Experimental Social Psy-chology 441246ndash55 [aKJJ]

Schnitzler H-U Moss C F amp Denzinger A (2003) From spatial orientation tofood acquisition in echolocating bats Trends in Ecology and Evolution 18(8)386ndash94 [CFM]

Schultheis H amp Barkowsky T (2011) Casimir An architecture for mental spatialknowledge processing Topics in Cognitive Science 3778ndash95 [HS]

Schultheis H Bertel S Barkowsky T amp Seifert I (2007) The spatial and thevisual in mental spatial reasoning An ill-posed distinction In Spatial cognitionV Reasoning action interaction (Lecture Notes in Computer Science vol4387) ed T Barkowsky M Knauff G Ligozat amp R D Montello pp 191ndash209Springer [HS]

Schwabe L amp Blanke O (2008) The vestibular component in out-of-body experi-ences A computational approach Frontiers in Human Neuroscience 217Available at httpwwwfrontiersinorghuman_neuroscience103389neuro090172008abstract [MB-C]

Schwarz W amp Keus I M (2004) Moving the eyes along the mental number lineComparing SNARC effects with saccadic and manual responses Perception andPsychophysics 66651ndash64 [ABa]

Seidl T ampWehner R (2008) Walking on inclines How do desert ants monitor slopeand step length Frontiers in Zoology 58 [aKJJ]

Sereno A B amp Lehky S R (2011) Population coding of visual space Comparison ofspatial representations in dorsal and ventral pathways Frontiers in Compu-tational Neuroscience 4159 doi1103389fncom201000159 [SRL]

Sereno I M Pitzalis S amp Martinez A (2001) Mapping of contralateral space inretinotopic coordinates by a parietal cortical area in humans Science 2941350ndash54 [HS]

Shelton P A Bowers D amp Heilman K M (1990) Peripersonal and verticalneglect Brain 113191ndash205 [JGP]

Sherry D F amp Vaccarino A L (1989) Hippocampus and memory for food cachesin black-capped chickadees Behavioral Neuroscience 103308ndash18 [DN]

Shinder M E amp Taube J S (2010) Responses of head direction cells in theanterodorsal thalamus during inversion Society for Neuroscience Abstracts No8951FFF6 [Society of Neuroscience Meeting Planner 2010 San Diego CAProgram No 8951] (Online) [aKJJ]

Si B Kropff E amp Treves A (2012) Grid alignment in entorhinal cortexBiological Cybernetics 106(8ndash9)483ndash506 doi 101007s00422-012-0513-7[FSt]

Siegel J J Neunuebel J P amp Knierim J J (2007) Dominance of the proximalcoordinate frame in determining the locations of hippocampal place cell activityduring navigation Journal of Neurophysiology 9960ndash76 [FSa]

Silder A Besier T amp Delp S L (2012) Predicting the metabolic cost of inclinewalking from muscle activity and walking mechanics Journal of Biomechanics45(10)1842ndash49 [FTS]

Sinai M J Ooi T L amp He Z J (1998) Terrain influences the accurate judgementof distance Nature 395497ndash500 [AP]

Skaggs W E amp McNaughton B L (1998) Spatial firing properties of hippocampalCA1 populations in an environment containing two visually identical regionsJournal of Neuroscience 188455ndash66 [PAD]

Smith A D Gilchrist I D amp Hood B M (2005) Childrenrsquos search behaviour inlarge-scale space Developmental components of exploration Perception 34(10)1221ndash29 [KAL]

Smith A D Hood B M amp Gilchrist I D (2008) Visual search and foraging com-pared in a large-scale search task Cognitive Processing 9(2)121ndash26 [KAL]

Smith A D Hood B M amp Gilchrist I D (2010) Probabilistic cuing in large-scaleenvironmental search Journal of Experimental Psychology Learning Memoryand Cognition 36(3)605ndash18 [KAL]

Snyder L H Grieve K L Brotchie P amp Andersen R A (1998) Separate body-and world-referenced representations of visual space in parietal cortex Nature394(6696)887ndash91 [rKJJ]

Soeda M Kushiyama N amp Ohno R (1997) Wayfinding in cases with verticalmotion In Proceedings of MERA 97 International Conference onEnvironment-Behavior Studies ed T Takahashi amp Y Nagasawa pp 559ndash64 University of Tokyo and Man-Environment Relations Association[CH]

Solstad T Boccara C N Kropff E Moser M-B amp Moser E I (2008) Rep-resentation of geometric borders in the entorhinal cortex Science 3221865ndash68 [FSa]

Spiers H J Jovalekic A amp Jeffery K J (2009) The hippocampal place code ndashMultiple fragments or one big map Society of Neuroscience Abstracts No10128 [Society of Neuroscience Meeting Planner 2009 Chicago IL ProgramNo 10128] (Online) [PAD]

Spinozzi G Lubrano G amp Truppa V (2004) Categorization of above and belowspatial relations by tufted capuchin monkeys (Cebus apella) Journal of Com-parative Psychology 118403ndash12 [AGD]

Squatrito S Raffi M Maioli M G Battaglia-Mayer A (2001) Visual motionresponses of neurons in the caudal area PE of macaque monkeys Journal ofNeuroscience 21(4)RC130 [GAO]

Stackman R W amp Taube J S (1998) Firing properties of rat lateral mammillarysingle units Head direction head pitch and angular head velocity Journal ofNeuroscience 18(21)9020ndash37 [aKJJ HY]

Stackman R W Golob E J Bassett J P amp Taube J S (2003) Passive transportdisrupts directional path integration by rat head direction cells Journal ofNeurophysiology 902862ndash74 [PAD]

Stackman R W Tullman M L amp Taube J S (2000) Maintenance of rat headdirection cell firing during locomotion in the vertical plane Journal of Neuro-physiology 83(1)393ndash405 [aKJJ HY]

Steck S D Mochnatzki H F ampMallot H A (2003) The role of geographical slantin virtual environment navigation In Spatial Cognition III Routes and navi-gation human memory and learning spatial representation and spatial learn-ing Lecture Notes in Artificial Intelligence No 2685 ed C Freksa W BrauerC Habel amp K F Wender pp 62ndash76 Springer [aKJJ]

Stefanucci J K Proffitt D R Banton T amp Epstein W (2005) Distances appeardifferent on hills Perception and Psychophysics 67(6)1052ndash60 [aKJJ]

Stefanucci J K Proffitt D R Clore G L amp Parekh N (2008) Skating down asteeper slope Fear influences the perception of geographical slant Perception37(2)321ndash23 [aKJJ]

Stella F Si B Kropff E amp Treves A (2013) Grid cells on the ball Journal ofStatistical Mechanics Theory and Experiment P03013 doi1010881742-5468201303P03013 [FSt]

Stevens A amp Coupe P (1978) Distortions in judged spatial relations CognitivePsychology 10422ndash37 [HS]

Suddendorf T amp Corballis M C (2007) The evolution of foresight What is mentaltime travel and is it unique to humans Behavioral and Brain Sciences 30299ndash351 [ABa]

Sueur C (2011) A non-Leacutevy random walk in chacma baboons What does it meanPloS ONE 6(1)e16131 doi101371journalpone0016131 [AMH]

Tafforin C amp Campan R (1994) Ethological experiments on human orientationbehavior within a three-dimensional space ndash in microgravity Advances in SpaceResearch 14(8)415ndash18 [aKJJ]

Taira M Tsutsui K I Jiang M Yara K amp Sakata H (2000) Parietal neuronsrepresent surface orientation from the gradient of binocular disparity Journal ofNeurophysiology 83(5)3140ndash46 [GAO]

Takahashi K Gu Y May P J Newlands S D DeAngelis G C amp Angelaki DE (2007) Multimodal coding of three-dimensional rotation and translation inarea MSTd Comparison of visual and vestibular selectivity Journal of Neuro-science 27(36)9742ndash56 [GAO]

Takahashi T T Bala A D Spitzer M W Euston D R Spezio M L amp KellerC H (2003) The synthesis and use of the owlrsquos auditory space map BiologicalCybernetics 89(5)378ndash87 [FTS]

Tanaka K Hikosaka K Saito H Yukie M Fukada Y amp Iwai E (1986) Analysisof local and wide-field movements in the superior temporal visual areas of themacaque monkey Journal of Neuroscience 6(1)134ndash44 [GAO]

Taube J S Wang S S Kim S Y amp Frohardt R J (2013) Updating of the spatialreference frame of head direction cells in response to locomotion in the verticalplane Journal of Neurophysiology 109(3)873ndash88 [aKJJ]

Taube J S (2007) The head direction signal Origins and sensory-motor integrationAnnual Review of Neuroscience 30181ndash207 [aKJJ]

Taube J S amp Burton H L (1995) Head direction cell activity monitored in a novelenvironment and during a cue conflict situation Journal of Neurophysiology 74(5)1953ndash71 [aKJJ PAD]

Taube J S Muller R U amp Ranck J B (1990a) Head-direction cells recorded fromthe postsubiculum in freely moving rats I Description and quantitative analysisJournal of Neuroscience 10(2)420ndash35 [aKJJ]

Taube J S Muller R U amp Ranck J B (1990b) Head-direction cells recordedfrom the postsubiculum in freely moving rats II Effects of environmentalmanipulations Journal of Neuroscience 10(2)436ndash47 [aKJJ]

Taube J S Stackman R W Calton J L amp Oman C M (2004) Rat head directioncell responses in zero-gravity parabolic flight Journal of Neurophysiology 92(5)2887ndash997 [aKJJ]

Tavosanis G (2012) Dendritic structural plasticity Developmental Neurobiology 72(1)73ndash86 [rKJJ]

Taylor C amp Caldwell S (1972) Running up and down hills Some consequences ofsize Science 178(4065)1096 [AMH]

Teufel H J Nusseck H-G Beykirch K A Butler J S Kerger M amp BuumllthoffH H (2007) MPI motion simulator Development and analysis of a novelmotion simulator In Proceedings of the AIAA Modeling and SimulationTechnologies Conference and Exhibit Hilton Head South Carolina AIAA2007-6476 Available at httpkybmpgdefileadminuser_uploadfilespubli-cationsattachmentsTeufel2007_45125B05Dpdf [MB-C]

Theeuwes J (1994) Stimulus driven capture and attentional set ndash Selective searchfor colour and visual abrupt onsets Journal of Experimental Psychology HumanPerception and Performance 20(4)799ndash806 [KAL]

Thibault G Pasqualotto A Vidal M Droulez J amp Berthoz A (2013) How doeshorizontal and vertical navigation influence spatial memory of multi-flooredenvironments Attention Perception and Psychophysics 75(1)10ndash15 doi103758s13414-012-0405-x [ABe CCC AP rKJJ]

Thiele H amp Winter Y (2005) Hierarchical strategy for relocation of food targets inflower bats Spatial memory versus cue-directed search Animal Behaviour69315ndash27 [HP]

Thompson R F Mayers K S Robertson R T amp Patterson C J (1970) Numbercoding in association cortex of the cat Science 168271ndash73 [ABa]

Tlauka M Wilson P N Adams M Souter C amp Young A H (2007) An inves-tigation into vertical bias effects Spatial Cognition and Computation 7(4)365ndash91 [aKJJ CH]

Todd J T amp Norman J F (2003) The visual perception of 3-D shape from multiplecues Are observers capable of perceiving metric structure Perception andPsychophysics 65(1)31ndash47 [GAO]

Torrealba F amp Valdes J L (2008) The parietal association cortex of the rat Bio-logical Research 41(4)369ndash77 [GAO]

Treves A Kropff E amp Biswas A (2005) On the triangular grid of entorhinal placefields Society for Neuroscience Abstracts Abstract No 3119811 [FSt]

Tsoar A Nathan R Bartan Y Vyssotski A DellrsquoOmo G amp Ulanovsky N (2011)Large-scale navigational map in a mammal Proceedings of the NationalAcademy of Sciences USA 108(37)E718ndash24 doi 101073pnas1107365108[aKJJ CFM]

Tversky B (1993) Cognitive maps cognitive collages and spatial mental models InSpatial information theory A theoretical basis for GIS ed A U Frank amp ICampari pp 14ndash24 Springer Verlag [HS]

UlanovskyNampMossC F (2007)Hippocampal cellular andnetwork activity in freelymoving echolocating bats Nature Neuroscience 10(2)224ndash33 [aKJJ CFM]

Umiltagrave C amp Nicoletti R (1990) Spatial stimulus-response compatibility InStimulus-response compatibility An integrated perspective ed R W Proctor ampT G Reeve pp 89ndash116 North Holland (Elsevier Science Publishing) [IB]

Umiltagrave C Priftis K amp Zorzi M (2009) The spatial representation of numbersEvidence from neglect and pseudoneglect Experimental Brain Research192561ndash69 [ABa]

Valerio S Clark B J Chan J H Frost C P Harris M J amp Taube J S (2010)Directional learning but no spatial mapping by rats performing a navigationaltask in an inverted orientation Neurobiology of Learning and Memory 93(4)495ndash505 [aKJJ]

Valero A amp Byrne R W (2007) Spider monkey ranging patterns in Mexican sub-tropical forest Do travel routes reflect planning Animal Cognition 10(3)305ndash15 doi101007s10071-006-0066-z [AMH]

Vargas J P Petruso E J amp Bingman V P (2004) Hippocampal formation isrequired for geometric navigation in pigeons European Journal of Neuroscience201937ndash44 [DN]

Vecchi T Tinti C amp Cornoldi C (2004) Spatial memory and integration processesin congenital blindness NeuroReport 152787ndash90 [AP]

Veelaert P amp Peremans H (1999) Flexibility maps A formalisation of navigationbehaviours Robotics and Autonomous Systems 27151ndash69 [HP]

Vidal M Amorim M A amp Berthoz A (2004) Navigating in a virtual three-dimensional maze How do egocentric and allocentric reference frames inter-act Brain Research Cognitive Brain Research 19(3)244ndash58 [aKJJ AP]

Vidal M Lipshits M McIntyre J amp Berthoz A (2003) Gravity and spatialorientation in virtual 3D-mazes Journal of Vestibular Research 13273ndash86[AP]

von der Malsburg C (1973) Self-organization of orientation sensitive cells in thestriate cortex Kybernetik 1485ndash100 [DMWP]

Wade M G amp Jones G (1997) The role of vision and spatial orientation in themaintenance of posture Physical Therapy 77619ndash28 [MBC]

Wagner M (1985) The metric of visual space Perception and Psychophysics38483ndash95 [FHD]

Wall M B amp Smith A T (2008) The representation of egomotion in the humanbrain Current Biology 18(3)191ndash94 [GAO]

Walls M L amp Layne J E (2009) Fiddler crabs accurately measure two-dimensionaldistance over three-dimensional terrain The Journal of Experimental Biology212(Pt 20)3236ndash40 [aKJJ]

Walsh V (2003) A theory of magnitude Common cortical metrics of time spaceand quantity Trends in Cognitive Sciences 7483ndash88 [ABa]

Wang R F (2012) Theories of spatial representations and reference framesWhat can configuration errors tell us Psychonomic Bulletin and Review19(4)575ndash87 [RFW]

Wang R F (in press) Human four-dimensional spatial judgments of hyper-volume Spatial Cognition amp Computation An Interdisciplinary Journal[RFW]

Wang R F amp Brockmole J R (2003) Human navigation in nested environmentsJournal of Experimental Psychology Learning Memory and Cognition 29398ndash404 [RFW]

Wang R F amp Cutting J E (1999) A probabilistic model for recovering cameratranslation Computer Vision and Image Understanding 76205ndash12 [RFW]

Wang R F amp Spelke E S (2000) Updating egocentric representations in humannavigation Cognition 77215ndash50 [RFW]

Wang R F amp Spelke E S (2002) Human spatial representation Insights fromanimals Trends in Cognitive Sciences 6(9)376 [aKJJ]

Wang R F amp Spelke E S (2003) Comparative approaches to human navigationIn The neurobiology of spatial behaviour ed K Jeffery pp 119ndash43 OxfordUniversity Press [RFW]

Webber D M Aitken J P amp OrsquoDor R K (2000) Costs of locomotion and verticdynamics of cephalopods and fish Physiological and Biochemical Zoology 73(6)651ndash62 [aKJJ]

Wehner R (2003) Desert ant navigation How miniature brains solve complex tasksJournal of Comparative Physiology A Neuroethology Sensory Neural andBehavioral Physiology 189579ndash88 [aKJJ]

Wehner R Boyer M Loertscher F Sommer S amp Menzl U (2006) Ant navi-gation One-way routes rather than maps Current Biology 1675ndash79 [SZ]

Wehner R amp Menzel R (1990) Do insects have cognitive maps Annual Review ofNeuroscience 13403ndash14 [SZ]

Weisberg S M Brakoniecki E amp Newcombe N S (2012) The other side of themountain Slope as a cue in navigation Paper presented at the 5th InternationalConference on Spatial Cognition Rome Italy September 4ndash8 2012 [SMW]

Weisberg S M Nardi D N Newcombe N S amp Shipley T F (2012) Sensing theslopes Sensory modality effects on a goal location task Poster presented at the53rd Annual Meeting of the Psychonomics Society Minneapolis MNNovember 15ndash18 2012 [SMW]

Weisman J (1981) Evaluating architectural legibility Way-finding in the builtenvironment Environment and Behavior 13(2)189ndash204 [CH]

Whitlock J R amp Derdikman D (2012) Head direction maps remain stable despitegrid map fragmentation Frontiers in Neural Circuits 6 article 9 doi 103389fncir201200009 [PAD]

Whitlock J R Sutherland R J Witter M P Moser M-B amp Moser E I (2008)Navigating from hippocampus to parietal cortex Proceedings of the NationalAcademy of Sciences USA 105(39)14755ndash62 [GAO]

Wiener J M Houmllscher C Buumlchner S amp Konieczny L (2012) Gaze behaviourduring space perception and spatial decision making Psychological Research 76(6)713ndash29 [CH]

Wills T J Cacucci F Burgess N amp OrsquoKeefe J (2010) Development of the hip-pocampal cognitive map in preweanling rats Science 328(5985)1573ndash76 [rKJJ]

Wilson P Foreman N Stanton D amp Duffy H (2004) Memory for targets in amultilevel simulated environment Evidence for vertical asymmetry in spatialmemory Memory and Cognition 32(2)283ndash97 [aKJJ RFW]

Winter Y amp Stich K P (2005) Foraging in a complex naturalistic environmentCapacity of spatial working memory in flower bats Journal of ExperimentalBiology 208539ndash48 [HP]

Winter Y von Merten S amp Kleindienst H-U (2004) Visual landmark orientationby flying bats at a large-scale touch and walk screen for bats birds and rodentsJournal of Neuroscience Methods 141283ndash90 [HP]

Wintergerst S amp Ronacher B (2012) Discrimination of inclined path segments by thedesert ant Cataglyphis fortis Journal of Comparative Physiology A Neu-roethology Sensory Neural and Behavioral Physiology 198(5)363ndash73 [aKJJ]

Wohlgemuth S Ronacher B amp Wehner R (2001) Ant odometry in the thirddimension Nature 411(6839)795ndash98 [aKJJ]

Worringham C J amp Beringer D B (1989) Operator orientation and compatibilityin visual-motor task performance Ergonomics 32387ndash99 [JGP]

Wray M K Klein B A Mattila H R amp Seeley T D (2008) Honeybees do notreject dances for ldquoimplausiblerdquo locations Reconsidering the evidence for cog-nitive maps in insects Animal Behaviour 76261ndash69 [SZ]

Wyeth G amp Milford M J (2009) Spatial cognition for robots Robot navigationfrom biological inspiration IEEE Robotics and Automation Magazine 16(3)24ndash32 [HP]

Wystrach A Schwarz S Schultheiss P Beugnon G amp Cheng K (2011) Viewslandmarks and routes How do desert ants negotiate an obstacle courseJournal of Comparative Physiology A 197167ndash79 [SZ]

Yang Z amp Purves D (2003) A statistical explanation of visual space NatureNeuroscience 6(6)632ndash40 [HER]

Yantis S amp Jonides J (1984) Abrupt visual onsets and selective attention Evidencefrom visual search Journal of Experimental Psychology Human Perception andPerformance 10(5)601ndash21 [KAL]

Yartsev M M amp Ulanovsky N (2013) Representation of three-dimensional space in the hippocampus of flying bats Science 340367ndash72[CFM FSa]

Yartsev M M Witter M P amp Ulanovsky N (2011) Grid cells without thetaoscillations in the entorhinal cortex of bats Nature 479(7371)103ndash107 doi101038nature10583 [CFM]

Young L R Oman C M Watt D G Money K E amp Lichtenberg B K (1984)Spatial orientation in weightlessness and readaptation to earthrsquos gravity Science225(4658)205ndash208 [aKJJ]

Zhang K amp Sejnowski T J (1999) Neural tuning To broaden or to sharpenNeuralComputation 1175ndash84 [SRL]

Zhu H Sauman I Yuan Q Casselman A Emery-Le M Emery P amp ReppertS M (2008) Cryptochromes define a novel circadian clock mechanism inMonarch butterflies that may underlie sun compass navigation PLoS Biology6138ndash55 [JGP]

ZugaroMB Arleo A Berthoz AampWiener S I (2003)Rapid spatial reorientationand head direction cells Journal of Neuroscience 23(8)3478ndash82 [aKJJ]

Introduction

Theoretical considerations

Behavioural studies in three dimensions

31Processing of verticality cues

32Navigation on a sloping surface

33Navigation in multilayer environments

34Navigation in a volumetric space

Neurobiological studies in three dimensions

41Neural processing of verticality cues

42Neural encoding of a sloping surface

43Neural encoding of multilayer environments

44Neural encoding of a volumetric space

Is the cognitive map bicoded

51Arguments for a bicoded cognitive map

52Why would animals have a metrically planar map

Conclusion


head19

head20

head21

head22

head23

head24

head25

head26

head27

head28

head29

head30

head31

head32

head33

head34

head35

head36

head37

head38

head39

head40

head41

head42

head43

head44

head45

head46

head47

head48

head49

head50

Introduction

Frames of reference

R21Egocentric and allocentric reference frames

R22Dimensionality of reference frames

R23The origin of reference frames

Comparative studies

R31Studies of humans

R32Studies of nonhumans

Experience

R41Developmental experience

R42Adult experience

Conclusion