The cognitive significance of resonating neurons in the cerebral cortex

Consciousness and Cognition 22 (2013) 1523–1550

Contents lists available at ScienceDirect

Consciousness and Cognition

journal homepage: www.elsevier .com/locate /concog

Review

The cognitive significance of resonating neurons in the cerebralcortex

1053-8100/$ - see front matter � 2013 Elsevier Inc. All rights reserved.http://dx.doi.org/10.1016/j.concog.2013.10.004

⇑ Corresponding author.E-mail addresses: [email protected] (D. LaBerge), [email protected] (R. Kasevich).

David LaBerge a,⇑, Ray Kasevich b

a Department of Cognitive Sciences, University of California, Irvine, USAb Stanley Laboratory of Electrical Physics, Great Barrington, MA, USA

a r t i c l e i n f o a b s t r a c t

Article history:Received 5 April 2013Available online 6 November 2013

Keywords:Resonating neuronApical dendriteCorticothalamic circuitConscious impressionsAspects of attentionFeelingsInsular cortex

Most neural fibers of the cerebral cortex engage in electric signaling, but one particularfiber, the apical dendrite of the pyramidal neuron, specializes in electric resonating. Thisdendrite extends upward from somas of pyramidal neurons, the most numerous neuronsof the cortex. The apical dendrite is embedded in a recurrent corticothalamic circuit thatinduces surges of electric current to move repeatedly down the dendrite. Narrow band-widths of surge frequency (resonating) enable cortical circuits to use specific carrier fre-quencies, which isolate the processing of those circuits from other circuits. Resonatinggreatly enhances the intensity and duration of electrical activity of a neuron over a narrowfrequency range, which underlies attention in its various modes. Within the minicolumn,separation of the central resonating circuit from the surrounding signal processing networkseparates ‘‘having’’ subjective impressions from ‘‘thinking about’’ them. Resonating neu-rons in the insular cortex apparently underlie cognitive impressions of feelings.

� 2013 Elsevier Inc. All rights reserved.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15242. Anatomy of the apical dendrite of the pyramidal neuron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15263. Short summary of the Dendritic Resonance Model (Kasevich & LaBerge, 2010) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15274. Pyramidal neurons within the minicolumn structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15305. Column clusters of minicolumns in the cortical fabric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15306. Pyramidal neurons of Layers 5 and 6 connect with the thalamus in recurrent circuits (loops) to produce resonant activity . 15307. Electric resonating of the apical dendrite of the pyramidal neuron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1530

7.1. Electric surges in the apical dendrite . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15317.2. Sharpening of the current input profile of frequencies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15317.3. Local control of peak resonance frequency by membrane channels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15317.4. Resonance curves produced from the simulation of the resonance model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15327.5. Column level control of the peak resonance frequency by layer 6 pyramidal neurons. . . . . . . . . . . . . . . . . . . . . . . . . . . 1532

8. The center-surround structure of the minicolumn: a loop in the center and a network in the surround . . . . . . . . . . . . . . . . . 15339. Connecting two corticothalamic loops across the cortex. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153410. The case of apical dendrites with compartments exhibiting different resonances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153711. Cognitive implications of the center loop and surrounding network of the cortical minicolumn. . . . . . . . . . . . . . . . . . . . . . . 1539

11.1. ‘‘Aboutness’’ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1539

http://crossmark.crossref.org/dialog/?doi=10.1016/j.concog.2013.10.004&domain=pdf

http://dx.doi.org/10.1016/j.concog.2013.10.004

mailto:[email protected]

mailto:[email protected]

http://dx.doi.org/10.1016/j.concog.2013.10.004

http://www.sciencedirect.com/science/journal/10538100

http://www.elsevier.com/locate/concog

FL

1524 D. LaBerge, R. Kasevich / Consciousness and Cognition 22 (2013) 1523–1550

12. Mechanisms of attention . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1541

ig. 1.aBerge

12.1. Selective attention. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1542
12.1.1. A role of inhibition in selective attention . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1545
12.2. Preparatory attention . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154612.3. Sustained attention . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1546

13. Describing feelings as resonant activity in the insular cortex. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1546
13.1. Anatomy and connections of the insular cortex . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154713.2. Attention to specific insular functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1548 Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1548References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1549
1. Introduction

From the early beginnings of central nervous system research with spinal cord experiments (Bell, 1811; Magendie, 1822),our understanding of brain activity has rested strongly on the assumption that the basic function of neural fibers is to com-municate all-or-none pulse signals from one location to another. For example, this kind of communication occurs in the re-flex circuit that connects a receptor in the knee tendon to a muscle cell in the thigh of the leg. In complex neural activities of

A camera lucida drawing of a human pyramidal neuron whose soma lies in Layer 5 of the visual primary cortex. From DeFilipe and Jones (1988) and(2005). The insert of spines is from Nagerl et al., 2008.

D. LaBerge, R. Kasevich / Consciousness and Cognition 22 (2013) 1523–1550 1525

the cerebral cortex, the concurrent development of information theory and computer design gives strong support to theall-or-none, digital view of neural functioning for describing cognitive activities. The all-or-none nature of axon pulses is eas-ily mapped onto the binary states of a computer wire, and circuits containing filtering and storage devices provide models forattention and memory activities in cognition. This confluence of research on the conductive properties of a nerve with re-search on the information-carrying properties of a wire gave rise to the now widely held view that the basic function of neu-ral fibers in cognitive activities of the brain can be regarded as signaling, the moving of binary pulse signals from one brainlocation to another. Included in the term signaling are the integrative processing of signals in the basal dendrites and soma.

Here we describe a neural fiber whose main function is apparently not signaling but resonating. This fiber, the apical den-drite, does not function mainly to move pulse information from one location to another, nor to integrate or process signalsthat contact its many synapses, but instead this fiber is part of a recurrent, closed circuit that propagates current pulsesrepeatedly in a closed loop. In the apical dendrite segment of the looping circuit, where these special properties of resonancetake place, the voltage of the injected current is below the threshold for producing action potentials. As the surges of currentmove down the apical dendrite their push-and-pause motion produces an oscillation, and this oscillation takes on a partic-ular frequency. When the neural membrane responds preferentially to inputs within a narrow range of frequencies, the

Fig. 2. Thicknesses of layers in parietal area 7 for the mouse, rat, cat, monkey, and human (Rockel, Horns, & Powell, 1980).


membrane is said to be exhibiting resonance (Higgs & Spain, 2009; Hutcheon & Yarom, 2000). In general the term resonancedenotes the ability of a system to oscillate most strongly at a particular frequency.

A recent study (Kasevich & LaBerge, 2010) conducted a circuit analysis of membrane function and determined that thepeak resonance frequency could be adjusted to 20, 40, 60 and 80 Hz by setting the rate of outward flow of potassium ions.Furthermore, it was determined that the width of the distribution of frequencies around the peak frequency could be re-duced to near zero by repeatedly moving electric current surges through the apical dendrites while pulses cycled aroundthe corticothalamic circuit.

The resonance properties of the apical dendrite serve several important functions in the operations in the cerebral cortex.They include: (a) setting the particular (peak) oscillation frequency of neural circuits, thereby enhancing a circuit’s responseto inputs of a particular frequency in the manner of a band-pass filter, and (b) setting the intensity level of attention, and (c)providing the neural substrates for cognitive states of sensations and feelings. A central theme of this article is that the majorfunctions served by the apical dendrite are other than signaling, either in transmission or integration.

2. Anatomy of the apical dendrite of the pyramidal neuron

Fig. 1 shows an example of a pyramidal neuron of a human brain whose soma lies in layer 5 of the primary visual corticalarea. The apical dendrite ascends vertically from the top of the soma, and is approximately 1600 lm in length; and the axondescends from the bottom of the soma. The apical dendrite typically is dotted with thousands of spines, which are small

Fig. 3. Resonance sharpening of current input. A longitudinal internal view of the apical dendrite beginning near the region of its principal input synapsesshows the general membrane structure and the movement of a current surge along the inside the membrane. The membrane shows the high density ofchannels, including the inward sodium (Na++) channels carried by the spines and the outward potassium (K+) channels, which operate on the passingcurrent surges to shape the profile of surge frequencies. Other membrane channels are omitted for clarity. The equation describing the change in shape isgiven in Kasevich & LaBerge, 2010.


protuberances that vary from 0.2 to 2 lm in length (Harris & Kater, 1994). A 15-lm length sample of spines taken from themiddle range of a monkey apical dendrite (Nagerl, Willig, Hein, Hell, & Bonhoeffera, 2008) is shown in Fig. 1.

Some of the longest apical dendrites in the human brain lie in the parietal area, where their estimated average length isapproximately 1800 lm. The comparative lengths of layer 5 apical dendrites in parietal area 7 of 5 species of mammals canbe observed using Fig. 2 by estimating the distance between layer 5 and layer 1 for each species.

The pyramidal neurons are marked by the close packing of spines on all of its dendrites (Yuste, 2010, 2011), somewhatlike new branches of some evergreen trees show small individual needles protruding everywhere along the shaft of itsbranches. The high density of spines that dot the dendrites enhances local electrical activity within and along their mem-branes. Spines come in many shapes and sizes (Garcia-Lopez, Garcia-Marin, & Freire, 2006; Yuste, 2010), and the diversityin spine morphology may be related to the local function of spinal synapses and to the source of the axons that contact thespines as axons pass across the many dendritic fibers in a section of cortex.

In general, pyramidal neurons vary somewhat across cortical areas and species in their shapes (morphology) and in theirconnectivity patterns to other neurons, both cortical and subcortical (Elston, 2013). Here we concentrate our description onthe shape and connectivity of the human layer 5 pyramidal neuron, which is typically of a large size, frequently shows burstfiring, and whose axon is usually (but not always) connected to a thalamic neuron. The characteristics of apical dendriteactivity in layer 5 pyramidal neurons generally apply to the pyramidal neurons whose somas lie in layers 6 and layers 2/3.

3. Short summary of the Dendritic Resonance Model (Kasevich & LaBerge, 2010)

The purpose of the present spine cable model is to describe how the apical dendrites of pyramidal neurons of the neo-cortex generate oscillations within a preferred narrow range of frequency. Dendritic oscillations take place within the den-dritic shaft, whose diameter is approximately 0.30 to 8.5 lm. The walls of the shaft are made of a thin, two-layered sheet oflipid cells into which is inserted a wide variety of protein channels (as shown diagrammatically in Fig. 3). The channels carrycharged ions (e.g., sodium, potassium, chlorine, calcium) into or out of the interior of the dendrite. The operation of the pres-ent spine cable model is based mainly upon the very rapid outward transport of potassium ions, inward transport of sodiumions, and the transport of ions along the outer surface of the membrane from compartment to compartment as longitudinalcurrent flow which is both capacitive and conductive in nature. The compartment electric circuit model therefore involvesboth radial and surface capacitances with conductance that together describe a band pass electric filter. The model producesa unique and possibly different resonance frequency for each compartment. The spine geometries and extracellular dielectricproperties are a measure of the surface capacitance. There is no limitation on the range of resonant frequencies possible withthe spine model as is the case for the models based on slow-activating H-type conductances and similar mechanisms de-scribed in Section 10.

Fig. 4. Longitudinal and radial capacitances in the membrane of the neocortical apical dendrite. (Top) Longitudinal surface capacitance, acting across themembrane segments between spine necks and holes where charges flow in and out of the dendrite channel openings. (Bottom) Radial capacitance, actingacross the dielectric segments between the intracellular and extracellular charges.

Fig. 5. A minicolumn of the cerebral cortex. Schematic diagram from Peters and Sethares (1991) of excitatory neurons in a minicolumn of the primaryvisual area of a monkey, showing pyramidal neuron apical dendrites, whose somas are located in layers 2,3,5, and 6. Neurons lacking apical dendrites arelayer 4 stellate neurons.


The source of surface capacitance can be described in more detail as follows. Along the interface between the outer mem-brane surface and extracellular fluid are moving charges, which encounter the shafts of spines and holes of channels. Thesetwo structures provide instantaneous capacitive stored energy for these charges (see Fig. 4). The direction of the moving po-sitive charges is opposite to the current in the dendritic core.


The electric force that determines this reversal of direction is seen in Fig. 8 D, where the arrow in the ellipse of the electricfield that intersects the surface current points in the opposite direction of the field force at the core current. The surfacecapacitance is modeled using theory of leaky cables as given by Delogne (1982).

Conventional neuron cable theory involves only radial membrane capacitance, so inductance must be introduced to pos-tulate such band pass behavior or resonance response as reported in the literature (e.g., Narayanan & Johnston, 2008). In thepresent model, the operation of the longitudinal surface capacitance contributes to the high-pass filtering of surge frequen-cies above the resonant frequency, and the operation of the radial capacitance contributes to the low-pass filtering of surgefrequencies below the resonant frequency. Together these operations produce one maximum or resonance condition in thefrequency distribution of the compartment transfer impedance. The electric coupling of compartments by radial and longi-tudinal current flow and the repeated current through the dendrite (via the corticothalamic loop) results in an overall trans-fer impedance with different possible resonant frequencies each with a bandwidth centered on the resonant frequency thatbecomes more narrow as the number of loops increases (see Fig. 10). The resonance frequency is set by the background acti-vation, which sets the level of current flow through the radial K+ channels.

In the apical dendrite, initial surge of charge is produced by the arrival of action potential pulses at the synapses at thesedistal dendritic locations. The volume of a single EPSP (excitatory postsynaptic potential) surge (i.e., the quantity of sodiumions) may vary according to the number of momentarily contributing synapses (spatial summation) and/or by the number ofpulses delivered to a synapse in a very short time period (temporal summation), as from a 10 ms burst of action potentials.Therefore, the current surge of a pulse moving down the apical dendrite may have varying levels of voltage amplitude, whichcontrasts with the discrete, all-or-none voltage amplitude feature of an action potential pulse propagating along an axon ofthe neuron.

Fig. 6. Locations of the pyramidal neurons in the cerebral cortex. A. Diagram of the lateral view of the cortex showing a highly convoluted sheet. B. Theconvoluted sheet was converted into a flat surface by Van Essen, Drury, Joshi, and Miller (1998). The area of this flat surface is approximately 3 times that ofthe lateral view, and has an average thickness of approximately 3 mm. C. The sheet is constructed of neurons organized in columns, which are clusters ofminicolumns. D. Within each minicolumn are shown the major pyramidal neurons with their somas and vertically aligned apical dendrites. The star-shapedneurons in the middle of the minicolumns (in layer 4) are stellate neurons (the relatively small size of layer 4 indicates that this column is not taken from aprimary sensory area). To show the apical dendrites with clarity, inhibitory neurons, which make up 15–20% of the total number of minicolumn neurons,are omitted. Axons, which exit at the bottom of the somas, are also omitted. From LaBerge and Kasevich (2007).


4. Pyramidal neurons within the minicolumn structure

The pyramidal neurons of the cerebral cortex are organized within small, cylindrical shaped structures, called minicol-umns (see Fig. 5), which are structural and functional units of the neocortex (Mountcastle, 1998), based on the observationthat neurons in a given minicolumn share many receptive field properties. Examples are the minimum receptive fields forwhisker movements in primary somatosensory areas and orientations of edges or lines in visual cortical areas. Within theminicolumn three classes of pyramidal neurons may be specified according to the location of their somas and their func-tions: Layer 5 pyramids have apical dendrites that typically extend to layer 1; layer 2/3 pyramids have apical dendrites thatalso typically extend to layer 1; layer 6 pyramids are mostly either short (extending upward to layer 3) or tall (extendingupwards to layer 1). In the sections that follow, we will describe the functions of each of these three classes of pyramids.

5. Column clusters of minicolumns in the cortical fabric

The minicolumns are arranged together in clusters, called columns, which are closely packed to make a convoluted sheetof gray matter that constitutes the cerebral cortex of the brain (see Fig. 6). As shown in Fig. 6 each column of neurons consistsof clusters of many minicolumns, and the structure of the minicolumn appears to be organized around a central corecontaining layer 5 pyramidal neurons, which are the pyramids that exhibit most of the longer apical dendrites in the cortex(Peters & Sethares, 1991). Layer 6 pyramids are located outside of the minicolumn and are organized within a whole column,which is a cluster of approximately 100–300 minicolumns (Mountcastle, 1998). The centrality in the structure of the mini-column of the long apical dendrites of layer 5 pyramidal neurons is apparent in the two diagrams in the lower part of Fig. 6.Later sections of the present paper will address the centrality of layer 5 pyramidal neurons in the function of theminicolumn.

The total number of neurons in the human neocortex is approximately 26–29 billion (Mountcastle, 1998). The percentageof excitatory neurons is 80–85%, and the percentage of inhibitory neurons is 15–20%. Of the excitatory neurons, 67–80% arepyramidal neurons and 5–8% are stellate neurons (Feldman, 1984; Mountcastle, 1998).

To visualize the packing density of pyramidal neurons in the neocortex, one can place a dime almost anywhere on theregions of the scalp that overlay the cortical fabric and estimate the number of pyramidal neurons that lie in the 6 corticallayers directly underneath the dime. Using data from Mountcastle (1998), we estimate the following quantities under thedime: 30,000,000 neurons; 20,000,000 pyramidal neurons; 2,000,000 layer 5 pyramidal neurons; 200,000 minicolumns,and 1000 columns. In the primary visual area of the primate, the packing density is approximately two and a half times thatof the other cortical areas; therefore for this area these quantities must be appropriately adjusted upward.

Given that the proportion of neurons in the cerebral cortex that contain an apical dendrite is about .60 (range .53–.68) ofthe total number of neurons, considerations of their comparatively large size indicates that they dominate the volume of thecerebral cortex. These anatomical proportions suggest that the apical dendrite serves a major role in the functioning of thecerebral cortex.

6. Pyramidal neurons of Layers 5 and 6 connect with the thalamus in recurrent circuits (loops) to produce resonantactivity

Although both layer 5 and layer 6 pyramids connect with thalamic neurons in corticothalamic loops, we select the layer 5pyramidal neurons (pyramids) to describe how resonance is produced and controlled in the apical dendrite (see Fig. 7). Layer5 pyramids tend to be more homogeneous in length than layer 6 pyramids, which simplifies modeling of apical dendrite res-onating activity.

The axon of a layer 5 pyramid typically activates a thalamic neuron, and this activation is returned to the pyramid of ori-gin forming a recurrent circuit referred to as the corticothalamic loop. When the pyramids are located in primary sensoryareas the thalamic axons that return to the pyramid make most of their synaptic contacts with the pyramid as its apical den-drite passes through layer 4 (see Fig. 5), and fewer contacts at layers 1/2. When the pyramids are located in higher-orderareas of the cortex (e.g., the parietal area of Fig. 2), the thalamic axons make their most numerous contacts with the apicaldendrite at layers 1/2 (Jones, 2002, 2007).

7. Electric resonating of the apical dendrite of the pyramidal neuron

The electric nature of the apical dendrite has been most frequently studied from data based on EEG measurements. Theapical dendrite is usually characterized as an electric dipole, which oscillates at frequencies between a fraction of a cycle (asin deep sleep) to over 80 Hz (as in meditation by expert meditators: e.g., Lutz, Greischar, Rawlings, Ricard, & Davidson, 2004).When thousands of apical dendrites (in groups of columns) are oscillating at a common frequency, then the sum of the indi-vidual electric dipole fields can produce measurable voltages at the scalp electrode as EEGs. The detailed nature of electricalactivity within the apical dendrite that generates the electric fields has thus far remained elusive.

Here, we describe a model of electric resonance in the apical dendrite that addresses the way that a specific peak reso-nance frequency can be made very narrow (Kasevich & LaBerge, 2010). Numerical simulations of a leaky spine cable model of


the neural fiber demonstrate that the apical dendrite fiber can tune its peak frequency in the 10–100 Hz range, graduallynarrowing the transfer impedance resonance profile to an arbitrarily small bandwidth (resonance condition).

7.1. Electric surges in the apical dendrite

Fig. 8 shows a simplification of the basic electrical events in the apical dendrite when it is an active part of the cortico-thalamic circuit shown in Fig. 7. Axon pulses from the thalamic relay neuron enter at the top of the apical dendrite and pro-duce a brief surge of current in the apical dendrite shaft. Each surge of current is made up of a movement of positive ions,which results in an electromagnetic field in the manner indicated by nested ellipses in Fig. 8. The electric field is shown sep-arately because dendritic activity is described here mainly in terms of electrical fields from charge flow. The circular mag-netic field of each dipole does not affect the current flow. While trains of current surges travel continuously down the apicaldendrite the surge voltage decreases as positive ions leak through the membrane (Stuart & Spruston, 1998). We assume thatif sufficient voltage remains by the time the surge reaches the soma the voltage will influence the production of action poten-tials, by enhancing the EPSPs from input axons contacting the basal dendrites. But if the voltage decays to near zero levels, orif the apical dendrite is not activated, then the production of action potentials will not take place, unless the EPSPs from inputaxons themselves are at sufficiently high levels to produce automatic information processing. The effects of voltage level atthe soma on production of action potentials in layers 2/3 are described later in the Selective Attention Section 12.1.

7.2. Sharpening of the current input profile of frequencies

The input of thalamic axon pulses entering at the top of the apical dendrite will typically exhibit a range of frequencies,which can be described as a profile or distribution of frequencies around a dominant peak frequency (the measure of thefrequency profile here is impedance magnitude, which indicates how much electrical opposition is being encountered byeach input frequency-dependent component of current across the range of frequencies). As the series of current surgesmoves down the apical dendrite the profile of pulse frequency narrows, as shown in Fig. 3. At the input synapses the fre-quency spectrum is wide and unfocused. As the current surges move down the apical dendrite the membrane channels re-move potassium ions in response to voltage increases of injected sodium ion current from spines along the dendrite. Thisprocess changes the shape of the frequency profile by increasing the occurrences of specific frequencies near the peak fre-quency region of the profile and decreasing the occurrences of specific frequencies remote from the peak frequency. Theequations which describe the way channel activity shapes the frequency profile are given in Kasevich & LaBerge (2010).

The apical dendrite is a sector of the corticothalamic loop. When this loop is active, current surges repeatedly pass alongthe membrane environment of the apical dendrite sector, as shown in Fig. 3. After each pass, the frequency profile narrowsslightly, so that many cycles of surges are needed to sharpen the profile to the point where the frequency of surges and con-sequent axon spikes are tuned closely to one particular frequency. The number of cycles needed to achieve sharp tuning ofthe apical dendrite depends upon the length of the apical dendrite and the noise contributed by each synapse in the corti-cothalamic loop. When pulses encounter synapses, noise produces a widening or flattening of the frequency spectrum. Butthe longer the apical dendrite, the more the narrowing can progress before encountering a synapse. Hence, cortical areas thatcontain longer apical dendrites are predicted to require fewer loop cycles to achieve the same degree of sharpness in thefrequency profile. Longer layer 5 apical dendrites are located in thicker cortical areas (see Fig. 2), so that it would seem likelythat species and individuals who have a thicker cortex in specific cortical areas because of longer layer 5 apical dendrites areable to achieve a sharper level of tuning of the apical dendrites, and are able to achieve a specific level of sharpness morequickly.

It would seem that the length of the apical dendrite imposes a limiting factor on the ability to achieve a sharp tuning,since very short apical dendrites may not compensate for the level of noise in the synapses of the corticothalamic loop.For example, the very short apical dendrites in layer 5 of the mouse may limit its ability to sustain attention to the locationof a recent appearance of a predator, while the longer apical dendrites of a cat may allow attention to be sustained for manyminutes at the location where the mouse was last seen (the mechanisms underlying the sustaining of attention will be dis-cussed later in Sections 12–12.3).

7.3. Local control of peak resonance frequency by membrane channels

The peak frequency of the resonance profile within the apical dendrite is determined by the rate of outflow of potassiumions. The particular rate of outward flow through a potassium channel is regulated by a particular level of voltage on theinside surface of the membrane. This level of internal voltage is maintained by the synaptic activity in local spines, whichin turn, is produced by pulse trains from axons of the local cortical circuitry. The local circuitry surrounding apical dendritesis presumed to be noisy (Ho & Destexhe, 2000) but we will attempt later to attribute much of this noise to synaptic activitywhose source is the extensive branching of layer 6 axons along the apical dendrite.

Fig. 9 illustrates the way that the internal voltage produced by the synapses at the spines influences the peak frequency ofthe resonance profile, according to the present model. Three levels of spine density produce three levels of average local volt-age. Low, medium, and high local voltages are shown at the left of the figure, and in the next column are shown the corre-sponding three density levels of open potassium channels. These three levels of outward potassium ion flow are represented


in the model by the parameter, gr, which denotes the value of the potassium channel radial conductance. In Fig. 9 low spinevoltage produces low peak frequency values, and high spine voltage produces high peak frequency values. Thus, the outwardpotassium flow through the apical dendrite may serve as a major factor in determining the peak frequency of local electricfield potentials and scalp EEGs in the 10–100 Hz range in the cerebral cortex.

The internal voltage level that regulates the outward flow of potassium ions could conceivably be influenced by the volt-age of the current surges that pass along the dendritic shaft as well as by local spine activity. The level of intensity of theresonant frequencies of layer 5 and layer 2/3 apical dendrites will vary from near zero when inactive, to high values of ampli-tude when the minicolumn is producing concentrated attention to a particular cognitive content. These large changes ofamplitude of current surges are produced in layer 5 pyramidal neurons whose somas produce ‘‘bursts’’ containing severalspikes, at rates of 100–300 Hz (e.g., Brumberg, Nowak, & McCormick, 2000) instead of producing a single spike. When theburst of spikes traverses the apical dendrite, it will raise the momentary voltage of each current surge inside the apical den-drite, and if this increased voltage were to affect the outward flow of potassium ions, then, according to the present model,bursting will raise the peak frequency value of the resonance profile. However, the durations of the individual current surgesare presumed to be brief, and may not greatly perturb the ongoing level of local voltage at the potassium channels. Therefore,we assume that the dominant influence on the local voltage is produced by the tonic inflow of positive sodium ions at theinside opening of the spine channels, which spreads along the inner surface of the membrane to the inside openings of near-by potassium channels.

7.4. Resonance curves produced from the simulation of the resonance model

A simulation of electric resonance in the apical dendrite (Kasevich & LaBerge, 2010) demonstrates narrowing of the profileas the current pulse surges through the corticothalamic loop. The resonating apical dendrite of Fig. 10 (e.g., as shown inFig. 1) was partitioned into 7 sections, or compartments, starting at the top of the dendrite. The top 6 compartments wereeach 200 lm in length, and assumed to be homogeneous with respect to the elements of the membrane. The 7th compart-ment, located between the 6th compartment and the soma, was approximately 240 lm in length and was not included in themodel because it’s diameter gradually increases to approach the larger diameter of the soma, and because it’s membraneelements are different from the other compartments (e.g., the 40 lm part of the apical dendrite adjacent to the soma con-tains no excitatory spines, and the number of inhibitory synapses decreases over the 240 lm length of this transitionalcompartment).

The resonance profile curves of Fig. 10 are numbered by the 6 compartments that the current surge first traverses, andthen by the successive loops through the entire corticothalamic circuit. One can immediately notice that the amount of nar-rowing produced on the moving current surge by the membrane over the initial 6 compartments is very small. However,with successive loops through the thalamus and the 6 compartments, the narrowing of the profile by the apical dendritemembrane becomes increasingly effective, and by the 7th loop the width of the profile has become very small, even forthe curve with a peak frequency value of 80 Hz.

Thus the resonance profile curves shown in Fig. 10 describe, through numerical simulation, the operations of the presentmodel leading to electric resonance in the apical dendrite and how a particular peak frequency is selected by the level ofoutward potassium conductance in the active apical dendrite. They also show how the resonance profile is sharpened byrepeated cycling of electric surges through the apical dendrite within the corticothalamic loop.

7.5. Column level control of the peak resonance frequency by layer 6 pyramidal neurons.

The peak frequency of the resonance profile is assumed to be regulated by the average voltage maintained at the insideentrance to the potassium channels. This voltage depends upon the average frequency of nearby spine activation by axonscontacting the outside of the spines. What is the origin of the axons that stimulate dendritic spines? Layer 5 pyramids wouldnot seem to be suppliers of axons that contact spines because they would deliver stronger synaptic activity at the outside endof the spines when their voltage amplitudes are increased by attention. More suitable modulators of average voltage levels atthe synapses of dendritic spines on layer 5 dendrites are the axons that arise from the layer 6 pyramidal neuron (Briggs,2010; Wiser & Callaway, 1996). Layer 6 axons are predominantly of the regular spiking type, and bursting is relatively rare(Brumberg, Hamzei-Sichani, & Yuste, 2003; Llano & Sherman, 2009; Mercer et al., 2005), unlike the activity of layer 5 pyra-midal neurons. Hence, layer 6 axons should provide a steady level of average voltage to target spines on layer 5 apical den-drites regardless of the changes in level of intensity of layer 5 pyramid resonating (owing to attentional demands of acognitive task, or to other influences such as the intensity of sensory input). Also, layer 6 axons can maintain a constant levelof average voltage delivered to spines of layer 2/3 apical dendrites as well as to the spines of apical dendrites of other layer 6axons. The observation that the anatomical locations of layer 6 pyramidal neurons lie within a column but outside the indi-vidual minicolumns that make up the column (see Fig. 6) is consistent with the hypothesis that layer 6 pyramidal neuronscan set the peak membrane oscillation frequency for all spiny dendrites within that column, including dendrites of stellateneurons. Thus, layer 6 pyramids identify each separate minicolumn as a member of that column by producing in their pyra-midal neurons a specific peak resonance frequency.

The foregoing descriptions of the targets of layer 6 axons are presented graphically in Fig. 11. Shown in Fig. 11 on the leftside are the two main types of layer 6 pyramidal neurons, short and tall. Other kinds of layer 6 pyramids exist (Briggs, 2010),


but the relevance of their functions to the present theory is not presently known. The axon, which exits the soma at the bot-tom, projects to the core type of neuron of the thalamus (Jones, 2002), and the pulse activity continues back to the top of thelayer 6 apical dendrite, completing the corticothalamic loop. This thalamic neuron also sends axons to dendritic spines (basalas well as apical) of all three kinds of pyramidal neuron, and probably also to spiny stellate neurons. The firing mode of theselayer 6 axons is regular, and therefore the activation of spine synapses is presumably uniform and stable over periods of time.The two pyramidal neurons in the middle of Fig. 11 are layer 5 neurons, which send their axon pulses to the matrix type ofthalamic neuron (Jones, 2002), before the pulses continue back to the top of the apical dendrite, completing the corticotha-lamic circuit. In primary sensory areas of the cortex (visual, auditory, tactual), the major proportion of returning axons fromthe thalamus contact the apical dendrite through intervening stellate neurons located in layer 4. One can compare the largearea of contacts of the thalamic axon at layer 4 for primary sensory minicolumns with that of higher order cortical areas byobserving that in Fig. 5 the thickness of layer 4 (from area V1) is much greater than the thickness of layer 4 (from the parietalarea) in Fig. 2. One could speculate that, for the primary sensory area, the large amount of thalamic input and the location ofthat input at the mid region of the apical dendrite (instead of at the top) allows the sensory receptors to exert (via the thal-amus) a more direct and powerful control over apical dendrite activity because of the density of thalamic projections and theproximity to the soma of layer 4. Perhaps this strength and directness of sensory influence on the apical dendrite of primarysensory areas produces the dominance of sensory impressions in everyday consciousness.

The significance of the layer 4 location of sensory inputs (via the matrix thalamic neuron) to the layer 5 apical dendrite issupported by the observation that layer 6 axons (via the core thalamic neuron) project strongly to the same layer 4 locationsof the sensory inputs. In this manner the layer 6 axons can set the peak resonant frequency value for the apical dendrite atthe point when the sensory ‘‘content’’ first appears at the cortical level. It has been conjectured that these layer 6 inputs servea modulatory function while the routine sensory thalamic inputs to layer 5 serve a ‘‘driving’’ function (Lee & Sherman, 2009).

On the right side of Fig. 11 are shown the layer 2 and layer 3 pyramidal neurons, which allow the resonating layer 5 apicaldendrites to influence the input–output information processing between minicolumns and columns. It is conjectured thatthe pair of longer layer 3 pyramidal neurons shown here exert local control of the resonating apical dendrites of layer 2/3, because their axons terminate on spines of these other layer 2/3 apical dendrites and basal dendrites. This local controlmay supplement the more global, column-wide control exerted by the layer 6 pyramidal neurons shown on the left sideof Fig. 11. The pair of resonating layer 3 neurons may belong to the class of ‘‘chattering’’ cells, identified by Grey andMcCormick (1996) in layers 2 and 3.

An overall view of the main cortical excitatory neurons shown in Fig. 11 suggests that the long apical dendrites on the lefthalf of the diagram represent the resonating operations of cortical function, while the shorter apical dendrites on the righthalf of the diagram represent the ways that the resonating operations influence signaling operations of cortical function.The thicker lines originating in the lower part of the right half of the diagram represent the route of signals arriving fromother minicolumns, the processing of those signals by a network of layer 2/3 neurons, and the transmission of those signalsto other minicolumns.

To maintain the same peak oscillation frequency in membranes of all the excitatory neurons within a column cluster ofapproximately 80 minicolumns requires that an enormous number of synaptic spines be distributed densely along the apicaland basal dendrites. Fig. 11 gives only a hint of the extensive degree to which spines dot the membranes of the pyramidalneurons in the cortex. A glance inside the apical dendrite (see Fig. 3) indicates that the moment-to-moment activity in themembranes of these dendrites contrasts dramatically with the almost quiescent ongoing activity of the inside of the smooth,virtually channel-free membrane of an axon surrounded by myelin.

8. The center-surround structure of the minicolumn: a loop in the center and a network in the surround

Fig. 6 D and C show many minicolumns of the neocortex in highly schematic form, with their two prominent structuralparts, the long cylindrical center made up of layer 5 pyramidal neurons and the short cylindrical surround made up of layer2/3 pyramidal neurons. To provide a more detailed representation of the minicolumn, we present in Fig. 12 an enlargementof the component neurons and add the representations of the current surges that move continuously down the apicaldendrites.

The diagrams in Fig. 12 show two views of the structure of the minicolumn of the neocortex: on the left side a view withthe surrounding cylinder positioned at the upper part of the center cylinder (here, the relative lengths of the center and sur-round cylinders correspond more closely to those found in minicolumns of the primary sensory area than to those found inthe parietal area, shown in Fig. 2 and most other cortical areas.); and on the right side of Fig. 12 is shown a view of the twocylinders separated so that details of the structures are seen more clearly. The center loop circuit actually extends below thecortex to connect with a thalamic neuron, but the resonating part of the loop takes place in the apical dendrite of the pyra-midal neuron, which is located in the cortex. Axons that leave the layer 5 pyramidal neurons typically project to the thala-mus in the corticothalamic circuit, but some axons project to subcortical structures, including the superior and inferiorcolliculi and the basal ganglia. Other axons project to adjacent columns where they may contact inhibitory neurons that sup-press activity in neighboring layer 5 pyramidal neurons, which forms a center-surround excitatory–inhibitory circuit struc-ture at the column level. The larger number of fibers returning to the thalamus shown in Fig. 12 is consistent with the finding

Fig. 7. The layer 5 pyramidal neuron, connected recurrently with a thalamic neuron. This cortico-thalamic loop contains the apical dendrite of thepyramidal neuron shown with its apical dendrite in the resonating state. The inhibitory neurons of the thalamic reticular nucleus have been omitted herefor simplicity.


that, in general, the number of cortex-to-thalamus fibers is about 10 times the number of thalamus-to-cortex fibers (Jones,1985).

In summary, the center and surrounding cylinders could be functionally labeled as the resonating and the signaling partsof the minicolumn, respectively. For convenience, the center cylinder of the minicolumn may be regarded as containing theloop circuit, even though the anatomy of the loop circuit actually extends into the subcortical region of the thalamus.

When the structure of the center loop circuit is considered in isolation, its looping circuitry strongly suggests that whenthis circuit oscillates it creates a cognitive event that has the property of variable duration. When a cognitive event has aduration longer than a fraction of a second, it is often referred to as a state. This event also appears to have the propertyof variable intensity, because some of the layer 5 pyramidal neurons that make up the center cylinder are capable of firingin bursts of several spikes. Thus, within the present theory, the duration and intensity of mental events or mental states canbe more directly related to the oscillatory activity in the center loop part of the minicolumn than to signaling activity in con-ventional neural circuit theories.

9. Connecting two corticothalamic loops across the cortex

The oscillatory activity in one minicolumn can influence the oscillatory activity in a minicolumn of a different column. Forexample, the frequency profile of apical dendrite oscillations in visual area V1 presumably induces a similar or even an iden-tical frequency profile in visual area V2. In this manner the circuits that link these areas can share a common peak frequency,which enables effective signaling to take place within the combined circuitry.

Fig. 8. A. Electromagnetic fields of current surges moving down the apical dendrite. B. An electromagnetic field is generated from each surge of current. C.Higher magnification of the electromagnetic field. D. The electric field component of the electromagnetic field is shown separately.


The route by which minicolumns in two different cortical areas can affect each other’s oscillatory activity is the connec-tion of the axon of the layer 5 pyramid of one minicolumn to the thalamic neuron that serves the layer 5 pyramid of the otherminicolumn, as shown in Fig. 13. This figure shows the pathways that connect center loop circuits of each participating mini-column, which carry neural signals into the loop circuits to generate oscillations in the central apical dendrites.

Fig. 14 adds direct minicolumn-to-minicolumn pathways to those shown in Fig. 13, along with layer 2/3 neurons. Thesepathways connect surrounding networks of each participating minicolumn, and carry neural signals between networks ofsignal processing. Together, the three types of pathways linking one minicolumn with another constitute a triangular circuit,whose structure and function has been explored in an earlier study (LaBerge, 1997).

The connection of a pair of minicolumns through the thalamus shown here in Fig. 14 has been generalized to a chain ofminicolumn links that reaches from primary visual areas to the prefrontal areas of executive control, as well as to a chain ofminicolumn links that reaches from prefrontal areas of executive control to primary visual areas (LaBerge & Kasevich, 2007,Figs. 7 and 8).

Evidence supporting the triangular connectivity between a pair of minicolumns just described is given in a recent study,which showed the participation of a thalamic pulvinar neuron in the synchronous oscillations of neurons in two adjacentvisual areas of monkey cortex: areas TEO (corresponds to the human inferotemporal area) and V4 (Saalmann, Pinsk, Wang,Li, & Kastner, 2012). Spike trains and local field potentials (LFPs) of individual thalamic neurons and cortical neurons in V4

and TEO were recorded from waking monkeys. The results showed greater thalamic pulvinar activity during the periodswhen the monkey attended to (and when it responded to) the expected receptive field of a target visual stimulus, comparedto a condition in which the monkey attended away from the target’s receptive field. The involvement of the pulvinar in sus-taining attention to a visual location shown here is consistent with earlier studies of pulvinar activity during visual attentionin monkeys (Peterson, Robinson, & Keys, 1985) and in humans (Buchsbaum et al., 2006; LaBerge & Buchsbaum, 1990).

The oscillations within visual areas V4 and TEO in the Saalamann et al. study (2012) showed increased synchronous activ-ity during visual attention, and the increased coherence exhibited itself in two frequency bands: bands in the 8–15-Hz alpharange and in the 30–60-Hz gamma range. The authors described the influence of the alpha peak frequency within the tha-lamic neuron as a source of modulation of the higher gamma oscillations. When the descending arm of the corticothalamicloop encounters the thalamus, the alpha frequency, native to thalamic activity (Lopes da Silva, 1991), mixes with the incom-ing frequency profile from the pyramidal neuron in such a way that it amplifies the frequency profile in the axon that leaves

Fig. 9. Resonance peak frequency and activity of local spines. A schematic diagram of the hypothesized relationship between the peak frequency of aresonance curve and the internal voltage level produced by subthreshold activity of an increasing number of local dendritic spines. From Kasevich &LaBerge, (2010).


the thalamus and returns to the apical dendrite of the pyramidal neuron. Details of the interactions between alpha and high-er-order frequencies have yet to be worked out in the literature (e.g., Busch, Dubois, & VanRullen, 2009; Palva & Palva, 2011).

In the active corticothalamic loop, it appears necessary to increase the voltage amplitude of the pulse activity upward as itpasses through the thalamus, so that the input voltage amplitude at the top of the apical dendrite is greater than the voltageamplitude at the soma. This adjustment seems to be required because the voltage decays considerably from the top of theapical dendrite to the soma (Stuart & Spruston, 1998). Axons leaving the layer 5 pyramidal neuron represent voltage ampli-tude by a rate code (including variations within bursts) superimposed on the carrier frequency, and when an axon contactsthe thalamus in the corticothalamic loop, the thalamus may increase spike trains to bursts (Lesica et al., 2006) and increasethe number of spikes within bursts, which increases the voltage amplitude in the rate code in the axon that leaves the thal-amus and projects back to the apical dendrite. In this way, the voltage amplitude of pulse surges delivered to the apical den-drite can be kept relatively constant from one loop to the next.

The connectivity dynamics between minicolumns in two cortical areas separated by a considerable distance are demon-strated by a study by Gregoriou, Gotts, Zhou, and Desimone (2011). These authors describe synchronous oscillations betweencortical areas V4 and area 8 (the frontal eye fields, FEF). Gregoriou et al. (2011) trained two monkeys to perform a visualattention task that required the release of a bar when a grating stimulus changed color. Recordings of local field potentials(LFPs) and multiunit spikes were taken from FEF and V4 while the monkey attended to the receptive field (RF) of the stim-ulus, prior to the onset of the color change. We make the assumption that the LFP measurements include the activity of theminicolumn center loop circuit, while spike measures could arise from both the center loop circuit and the surround networkcircuitry. Hence, for our purposes, it is the data from the LFP measurements that are most relevant here.

The measured results showed an increase of 63% in coherence of LFPs across the two areas, and an increase of 13% incoherence of spike trains. Both increases were statistically significant. The increase in LFP coherence between the FEF andV4 areas is consistent with the hypothesis that there is a relatively tight connection between the center loop circuits inthe minicolumns of these two areas. Our view is that the connection that mediates LFPs is a minicolumn-to-minicolumn linkthrough the thalamus, as shown in Fig. 14, or possibly a chain of these links.

Fig. 10. The narrowing of resonance curves. Resonance curves with peaks at 20, 40, 60, and 80 Hz are narrowed by repeated cascades through 6compartments of the apical dendrite via circuit loops through the thalamus. The resonance curves for the first 6 compartments are shown at the top of eachof the 4 sets of curves (for clarity, the resonance curves for the 3rd, 4th, and 5th compartments are omitted for the 60 and 80 Hz conditions). The remainingcurves in each of the 4 conditions show the effects of subsequent loops through the 6 compartments via the thalamus. Each curve representing a loop isactually the output of the 6th compartment of each successive loop. Radial conductances (for potassium ions) = 105, 310, 627, and 1056 nS for the curveswith peaks at 20, 40, 60, and 80 Hz, respectively. From Kasevich & LaBerge, 2010.


Additional findings of this study illustrate the top-down feature of the FEF-to- V4 connectivity. When the monkey at-tended to the receptive field of the target stimulus, the effect of attention began more strongly in FEF than in area V4, andlater in the delay period of sustained attention the effect of attention was greater in area V4.

Therefore, when two (or more) minicolumns lying in separate cortical area are connected through the thalamus (seeFigs. 13 and 14), the oscillations of neurons in one minicolumn influence the oscillations of the neurons in the other mini-column. In cases of strongly connected pairs of minicolumns, it is expected that the peak frequencies of the oscillation pro-files will be the same or very similar, as suggested by the studies of Saalmann et al. (2012) and Gregoriou et al. (2011)described earlier. When the peak frequencies of the two minicolumns match and the profile of frequency probability hasa variance near zero (see Fig. 10), then direct ‘‘horizontal’’ connections between the surrounding cylinders of networks willresonate at approximately the same frequency.

When fibers or wires resonate at a constant frequency, they can serve as ‘‘carriers’’ for digitally coded messages. A majorbenefit of precisely tuning a wire or neural fiber to a specific carrier frequency suggests that many neighboring axons arisingfrom different columns can transmit information at the same time without appreciable interference from cross talk. An anal-ogy is the separation of communications between many pairs of pilot cockpits in flight when the radio channel between eachpair of pilot cockpits is tuned to a different frequency. However, circuits carrying the same peak frequency may sometimesintersect, as in the case of synesthesia, when particular musical pitches or audio frequencies evoke particular colors.

10. The case of apical dendrites with compartments exhibiting different resonances

In the simulation of the present model, the radial conductance values remain constant for all compartments of the apicaldendrite, assuring that the resonant frequency of the outputs of each compartment remains the same. However, some recentarticles have reported evidence from data and simulations of resonance in apical dendrites of the hippocampus that indicatedifferent resonance frequencies in the compartments of a single apical dendrite. Investigation of how these resonance dif-


ferences come about in the hippocampus raise the question of whether similar compartmental differences might occur dur-ing typical operations of the neocortical apical dendrite.

Hu, Vervaeke, Graham, and Storm (2009) recorded from slices of the rat hippocampus and identified two separate thetaresonator mechanisms: (1) a M-type, slow potassium current channel located near the apical dendrite soma that operatedonly when the membrane was depolarized, and (2) a H-type, slow depolarizing potassium current channel located in thedistal dendrite region (often termed a ‘‘pacemaker channel,’’ owing to their help in producing rhythmic neural activity innetworks of the cortex and of the heart) which operates when the membrane is at subthreshold levels. Both mechanismsperform bandpass filtering of inputs at theta frequency (3–12 Hz) levels, which implies resonance.

Zhuchkova and Schreiber (2011) showed that the slow-activating H-type conductances, which act as high-pass filters, aredistributed along the apical dendrite with different values, resulting in changes of resonance frequencies along the apicaldendrite. The authors suggest that dendritic branches may increase the range of resonant frequencies in a dendrite. Theseresults are based on the apical dendrites in the hippocampus, which exhibits resonance frequencies mainly below 20 Hz,and the slow opening and closing of the H-channel is not sufficiently fast to account for filtering frequencies above 20 Hz,where most of the neocortical frequencies lie (Narayanan & Johnston, 2008).

Narayanan and Johnston (2008) determined that the subthreshold resonance of hippocampus pyramidal neuronsdepended on the location along the dendrite. This suggested that the frequency response of these neurons depends onthe dendritic location of their inputs, thereby regulating their oscillations. The mechanism that contributes to resonanceis the H-channel, which decreases in density along the hippocampal apical dendrite in the soma-to-distal direction. Asexpected, a compartment model showed a correnspondingly consistent change in resonant frequency (in the range below20 Hz) along the shaft of the apical dendrite.

In a neocortex layer 5 pyramidal neuron, the H-channel density was found to increase with distance from the soma(Williams & Stuart, 2000). This gradient distribution of the H-channel mechanism supports the claim that resonance can varyfrom one compartment to the next along the neocortical apical dendrite. However, the range of resonances subject to thisvariation lies in the range below 20 Hz. The present model describes the generation of resonance frequencies in the range0 to 100+ Hz, which includes the resonance frequencies of EEGs that accompany a wide variety of cognitive tasks (Kahana,2006).

Common to all of the foregoing examples of location-dependent resonances along the apical dendrite is the lack of amechanism that narrows their bandwidth, whereas the present model contains an effective mechanism for narrowing band-widths to less than 1 Hz.

Given the difficulty of identifying the type of potassium-leak related spine among the considerable diversity of spinetypes that dot the surface of the layer 5 apical dendrite, no data exists to our knowledge that identifies the distributionof densities of the potassium-leak related spine. However, another way to estimate the density of these particular spinesis to measure the amplitude of potassium channels in patches along the layer 5 apical dendrite, which is actually more clo-sely related to the outward potassium conductance parameter of the present model than the underlying density of its volt-age-producing spines.

Experiments by Bekkers (2000) and Korngreen and Sakmann (2000) positioned patches along the large layer 5 apical den-drite of the somatosensory area of the rat that contained two main types of single-channel, voltage-gated, outward potas-sium currents (only the first 450 lm of the total of 700 lm distance yielded measurements). Plots of the amplitude of IAand IK currents versus distance from the soma gave the following slopes: for the Bekkers (2000) experiment, the slopesfor IA and IK were 2.3 pA/100 lm and �.0.4 pA/100 lm, respectively, yielding an average of 0.95 pA/100 lm; for the Korn-green and Sakmann (2000) experiment the average slope (the individual IA and IF slopes were not given) was �0.9 pA/100 lm. The overall average of the two experiments yields a slope value of 0.025 pA/100 lm, which is close to zero.

Extrapolating these results to the entire length of the apical dendrite, one could conclude that a relatively constant cur-rent amplitude of potassium channels across the major extent of the layer 5 apical dendrite is functionally consistent withthe anatomically- based hypothesis that the axons from layer 6 pyramidal neurons distribute a constant rate of pulses to thesynapses of spines along the outside of the long apical layer 5 apical dendrites within a column of minicolumns. Taken to-gether, the Bekkers (2000) and Korngreen and Sakmann (2000) slope data give tentative support to the hypothesis that com-partments of the apical dendrites located in dendrites of neurons in an entire column share a common outward potassiumconductance parameter value.

The present spine model describes compartment resonant frequencies proportional to the radial leakage conductance in alinear relationship between leakage conductance and the compartment resonant frequency. Our published results (Kasevich& LaBerge, 2010) show the overall resonance narrowing process based on relatively constant leakage from compartment tocompartment (see Fig. 10). We conjecture that the case of large differences (e.g., more than 5 Hz) in outward conductancesfrom compartment to compartment are expected to produce unforeseen interactions with the narrowing processes, which islikely to result in soma resonances substantially different from the soma resonances calculated from the present constantleakage conductance case.


11. Cognitive implications of the center loop and surrounding network of the cortical minicolumn.

11.1. ‘‘Aboutness’’

The central proposal of the present paper is that apical dendrites do not process information but instead resonate as theyregister our cognitive experiences, especially our impressions of sensations and feelings. The circuits that serve the functionsof processing information and resonating are shown in Fig. 12. The center loop circuits contain layer 5 pyramidal neuronsand a thalamic neuron, and repeated cycles of pulses in this loop (spike pulses in the axons, current surges in the apical den-drites) induce electric oscillations in the membranes of apical dendrites of the pyramidal neurons. When thousands of apicaldendrites oscillate in synchrony, their voltages sum to a magnitude that reaches the scalp as a waveform that is measured asan EEG.

A second type of circuit shown in Fig. 12 is the surrounding network circuit, which, in the diagram, is displaced away fromthe center loop circuit, so that the features of the surrounding network can be viewed more clearly. The surrounding networkcontains approximately 39 pyramids on average, based on the monkey (Mountcastle, 1998). Because the pyramidal neuronsinterconnect in complex patterns, they are referred to here as a network. Inputs to and outputs from this network are madeup of spike signals from surrounding networks in other minicolumns located near and/or far from this network. Hence,according to the present theory, the ubiquitous information processing within the cerebral cortex takes place in the net-works within and between these surrounding networks of the minicolumn. The diagrams in Fig. 12 clearly show that axonconnections between the center looping circuits and the surrounding network circuits are assumed to be one-way and nottwo-way connections. Moreover, the one-way connection between the center looping circuits and the surrounding networkcircuit first passes through the thalamus, and, owing to the oscillatory nature of apical dendrite activity, spike trains withinthe loop circuits initiate electric field oscillations in the apical dendrites of layer 2/3 pyramidal neurons of the surroundingnetwork. However, apparently these spike trains do not exhibit (or need to exhibit) the complex coding of information that isbelieved to be typical of spike trains in axons of the surrounding network circuits. They need only express frequency andintensity of activity, presumably by variations in spike rate and burst size. Therefore, the center looping circuits do not sendcomplex information-bearing signals to the surrounding network circuitry. Instead, the center circuits send oscillatory activ-ity to the surrounding circuit pyramids via their apical dendrites, which vary in frequency and intensity. The activity receivedfrom the layer 2/3 apical dendrites induces an ongoing background oscillation in the soma and basal dendrites of these neu-rons, which sets the preferred frequency for receiving axon inputs at the basal dendrites. The activity received at the somafrom the layer 2/3 apical dendrites presumably modulates the ongoing base rate of voltage there, which influences the firingsensitivity of the pyramid. As a result, input from spike trains can induce spike outputs more readily if the sensitivity level ofsoma output is higher. If there were no activity in the apical dendrite of the surrounding network, then conceivably inputspike trains could not produce output spikes, with the possible exception of cases where the input–output connections weresufficiently strong to produce automatic processing. In sum, the ongoing activity in center corticothalamic loops enables nor-mal input–output information processing in the surrounding networks.

In addition to modulating input–output activity in the surrounding network part of the minicolumn, the oscillatory fre-quency of the center loop circuitry gives systemic meaning to the messages sent along the axons of a network by identifyingthe minicolumn that sent the messages. The addition of meaning to pulse trains can be clarified by noting the effect of itsabsence in the classical model of Shannon’s communication theory (Shannon & Weaver, 1949). This model consists of threemain features: the source of the information, the channel along which information is transmitted, and the destination of theinformation. A simple example is the telegraph, in which one key taps electric pulses into a wire that is connected to anotherkey that responds to pulses with an audible tap. Ironically, one of the strengths of the Shannon model is that the informationin the pulse train itself does not contain meaning, or semantic content. Therefore, in its abstract, meaning-free form, infor-mation can be quantified and used to compare many different applications of communication systems economically andobjectively.

Pure information in a telegraphic pulse train gains more meaningful content when knowledge about both the source anddestination of the message are given. An SOS pulse train intercepted in one of 5 wires strung between poles along a highwayis not meaningful unless it is known who is sending this message to whom and what code they are using. The center loop ofthe minicolumn confers systemic content to the pulse trains processed through the surrounding network by providing a car-rier frequency for pulses that are sent from a source neuron to a destination neuron. It is assumed that the membranes of thesource and destination neurons are in an ongoing state of oscillation independent of the generation and transmission ofpulses between them. For narrowly tuned oscillations at both source and destination neurons, axon pulses arriving at syn-apses on destination dendrites will activate the dendrites only if the axon carrier frequency matches the ongoing resonancefrequency of the destination dendrite. If the frequencies of the source and destination neurons are not narrowly tuned, someactivation may still occur, but the level of activation for increasing distance of a resonance miss-match will decrease accord-ing to the spread of the resonance curves (for resonance curves of varying amounts of spread see Fig. 10). If one intercepts apulse train message in a randomly selected axon, the peak frequency that carries that message is a marker for the minicol-umns whose center loop circuitries are generating that frequency at the source and usually also of the destination of theinformation flow. Therefore, it is the location of the minicolumns in the system that confers meaningful content to the


message. For a related discussion of the role of frequency matching in axonal connections see Izhikevich, Desai1, Walcott,and Hoppenstead (2003).

When information is exchanged between minicolumn network structures, the exchange can occupy extended durationsof time. For example, inspecting an object for flaws, and describing the features of a sunset both involve prolonged process-ing of information within and between networks of minicolumns. The center loops of these minicolumns have the propertyof duration: the oscillations that take place in the corticothalamic loops can continue as long as the appropriate input to theirthalamic neurons is sustained. The controlling source of thalamic input is presumably the minicolumn center loops that arelocated in frontal cortical areas of executive attentional control.

What is the cognitive manifestation of center loop activity? Information signals cannot be sent into it nor accessed from itand sent into the surrounding network, but the loop activity appears to be necessary for modulating information processingof signals in the surrounding network. Is the loop activity also silent cognitive-wise as it is apparently silent information-wise?

We propose that the looping activity at the center of the minicolumn is what the activity of that particular minicolumn is‘‘about.’’ The looping activity constitutes the identity of the minicolumn, e.g., a particular color, orientation of a line, a featureof a particular face, or an aspect of the feeling of joy. The identity of the minicolumn activity is not expressed in terms ofsignal codes of information, and consequently the identity is not expressible in linguistic terms. Therefore, the identity ofthe minicolumn activity is a non-linguistic entity. One response to this problem might follow Wittgenstein’s aphorism:‘‘Whereof one cannot speak, thereof one must be silent.’’ But one can still take an action of identifying something withoutspeaking, and that action is one of ‘‘pointing’’ to the thing. Instead of saying what something ‘‘is,’’ one keeps silent and pointsto it. Korzybski (1994) used this epistemological technique as the first stage of sensing the world, prior to verbal labeling of‘‘what the object is,’’ in order to avoid the philosophical pitfalls of the ‘‘is of identity.’’ Stated simply: the word is not the thing.

Fortunately, the minicolumn, by its structure, contains a non-informational means of pointing to its ‘‘identity’’, what thething is about. Instead of using neural fibers to probe the interior of the center loop, it uses neural fibers to surround the cen-ter loop (we commonly identify an object by drawing a circle around it with our finger as well as by pointing at it or touchingit with a finger). Thus, the brain indicates what a minicolumn’s particular activity is ‘‘about’’ by circling it with the networksurrounding structure that processes signals of information that are ‘‘about’’ the resonating activity it receives from the mini-column center. The features that distinguish one minicolumn from another are not only its peak resonant frequency (or pos-sibly, frequencies) and the shape and amplitude of the frequency profile around the peak frequency, but particularly theconnections to other minicolumns in the cortex. Minicolumns underlying the shape of an apple or banana have specific con-nections to minicolumns underlying the color red or yellow. In general, the cognitive content of minicolumns of higher levelcortical areas apparently depend upon their connections, direct or indirect, to the primary sensory areas.

By coincidence, the origin of the word ‘‘about’’ in the English language can be traced to the word ‘‘a-boutan’’ (circa 12thcentury England), which means ‘‘around the outside of.’’ An illustrative example is referring to the things outside of a smallhouse. This meaning is relatively concrete, compared to the abstract dominant modern dictionary meanings of ‘‘concerning’’or ‘‘associated with.’’ The anatomical separation of the network circuit from the loop circuit in the minicolumn correspondsto the more concrete meaning. Even the center circuit-to-surrounding network connection is not a direct one, because of theintervening route through the thalamus which carries a modulated frequency into layer 2/3 pyramidal neurons. This physicalseparation, instantiated in the anatomy of the minicolumn, perhaps should remind us, when we use the modern meanings of‘‘about,’’ that there is a ‘‘gap’’ of intentionality between the raw impression of a ‘‘thing’’ and its processed meaning.

Therefore it seems that, in its earliest etymological sense, the word ‘‘about’’ does not mean penetrating to or probing thecore of the object being considered, but instead it means a close circumscribing of the object, e.g., by describing it in terms ofthe informational content of its context.

However, our proposal concerning the cognitive significance of the center looping of the minicolumn goes further thanshowing how the neural structure of the minicolumn identifies itself by its context and its peak resonant frequency andhow it influences in an indirect manner the input–output signaling in the surrounding network of the minicolumn. We pro-pose that the resonant activity of the center loop circuit of the minicolumn also evokes in the brain a conscious impression ofthe minicolumn’s identity, including its intensity and its duration.

The consciousness proposal may be regarded as optional in the sense that adding conscious experience to the resonanceof the looping circuit does not necessarily imply that consciousness has a causal role in the functioning of the looping circuit.To our knowledge, there is no evidence that consciousness is a ‘‘something more’’ domain in which some unknown kind ofprocessing takes place and whose outcome subsequently contacts and influences the activity of the looping circuitry. There-fore, while consciousness can be viewed as being produced by the resonant activity of the minicolumn, its apparent lack ofan independent ‘‘downward’’ causal property would seem to give it the status of an epiphenomenon. However, if there exists‘‘something- more’’ than the present ‘‘nothing-but’’ neuroscience conception of brain-based consciousness, then the resonat-ing of particular minicolumns could conceivably serve as ‘‘receivers’’ (or resonators) by which the ‘‘something more’’ influ-ences the conscious events of mental life.

If one wishes to avoid referring to consciousness while considering the activities of the cortical minicolumn, one couldsubstitute for the terms conscious activity and conscious impressions the terms resonating activity and resonating impressions,even though many cases of resonating activity may be of such low intensity as to be undetectable.

Therefore, as a first approximation to the conditions in which a pyramidal neuron produces resonating impressions of theminicolumn’s identity or content, we suggest that the intensity of the resonant apical dendrite activity must be appropriately


high. Our field of vision contains many objects that differ from moment to moment in the intensity of their visual impres-sions, and those objects with the highest intensity dominate our field of subjective impressions, while most resonatingimpressions are of low intensity and ignored.

Variations in cognitive intensity may be traced to layer 5 pyramidal neurons, which are known to engage in the kind ofbursting activity that can produce high levels of intensity in the center looping circuits. Also, the close clustering of layer 5neurons within the center loop structure of the minicolumn may promote momentarily higher levels of amplitude throughhigher levels of synchrony, owing to their common source of axon input at the top of the relatively long apical dendrites.Lower levels of intensity are exhibited by layer 6 neurons, even of the tall type, because they are believed to engage in regularfiring but not bursting (e.g., Llano & Sherman, 2009; Mercer et al., 2005), A subset of Layer 2/3 pyramidal neurons in area V1

engage in ‘‘chattering’’ responses (Gray & McCormick, 1996), which could contribute to the overall intensity of minicolumnimpressions, but the apical dendrites of these neurons are relatively short, so that they do not produce as long an electricdipole and therefore do not produce electric waves with amplitudes as high as those produced by layer 5 pyramids. There-fore, given the available data, we tentatively conclude that it is the layer 5 pyramids that contribute the bulk of activity thatunderlies conscious impressions of sensations and feelings.

Impressions lie on a continuum bounded by being very sharp to being very vague. The sharpness–vagueness variation isassumed to correspond to the narrow-broad variation in the spread of the resonance profile around the peak frequency (seeFig. 10).

12. Mechanisms of attention

The layer 5 pyramidal gain control on input–output processing of layer 2/3 pyramidal neurons enables central loops ofminicolumns to determine which cognitive circuits shall momentarily dominate the activity of the cortex considered as awhole. The issue of circuit dominance is considered by many cognitive scientists to be the central problem of attention. Manyresponse outputs are available to the organism but physical limitations on the number of responses that can be emitted atone time compels the system to find ways to determine which particular minicolumn central loops will exhibit the highestlevel of amplitude at a particular moment of time. In the following sections we describe the way resonating loops in mini-columns participate in three different modes of attention: selective attention, preparatory attention, and sustained attention.

Fig. 11. The three major types of pyramidal neurons and the targets of their axons. A. Layer 6 tall and short pyramidal neurons. B. Layer 5 pyramidalneurons. Layer 5 and layer 6 pyramidal neurons connect with thalamic neurons in loop circuits. C. Layer 2/3 neurons connect with layer 2/3 neurons inother minicolumns located near and far across the cortex. (Adapted from Nieuwenhuys, Voogd, & van Huizen, 2008, Section II, p 564.).

Fig. 12. Center loop circuits and surrounding network neurons of the minicolumn. A. Diagram showing the center and surrounding parts of the minicolumnas they appear in the cortical fabric. B. The surrounding network of the minicolumn. C. The center loop circuit of the minicolumn. The dashed line indicatesthat the thalamocortical axon contacts the surrounding network as well as the center part of the minicolumn. The arrows entering and leaving thesurrounding network indicate inputs and outputs of the surrounding networks that contact minicolumns located at distances near and far. Current surgesmoving down the apical dendrites are represented by pairs of small ellipses, which represent electric fields produced by the current surges (see Fig. 8). Forclarity, the number of layer 5 pyramidal neurons in the center of the human minicolumn is 7 in this diagram. Actually the number of layer 5 pyramids in thecenter of the minicolumn may be closer to 20 in the human, given that the cat and monkey minicolumns (in the primary visual area) contain approximately13 and 17 layer 5 pyramids, respectively (Peters & Vilmez, 1993). The two axons that leave the cluster at the lower right represent projections to thesuperior colliculus, striatum, the contralateral cortex, or to neighboring columns.


12.1. Selective attention

Perhaps the most influential views of attention are based on the concept of a filter. In the electric physics that served as aframework for the development of information theory, a filter is regarded as a device that passes frequencies within a certainrange and blocks or attenuates frequencies outside that range. In its simple form, the filter performs the selection of frequen-cies by passively blocking unwanted frequencies.

The present view of attention is based on resonance phenomena, by which center looping circuits of minicolumns activelyselect input–output channels in the surrounding networks of minicolumns. This selection of input–output channels can bevery brief, as in visual search (e.g., Treisman & Gelade, 1980), or the selection can be prolonged, as in the extended durationsof tasks, such as driving a car, listening to music, or carrying on a conversation. In these complex activities, attention selects amulti-column network whose neural elements are all tuned to a common peak frequency.

Attentional selection by resonance is based on the columns that momentarily exhibit the highest voltage amplitude ofloop circuit activity. The voltage amplitude of looping activity in a given minicolumn center circuit is the sum of the voltageamplitudes of each participating individual layer 5 pyramidal neuron. The amplitude of a column is the combined voltageamplitudes of the minicolumns that make up the column. Thus, for example assuming an average of 10 layer 5 pyramidalneurons in a minicolumn, and 100 minicolumns in a column, and each layer 5 pyramidal neuron at maximum voltage ampli-tude, the total voltage amplitude of the column would involve 1000 pyramidal apical dendrites whose dipoles would con-tribute to a relatively large electromagnetic field. Furthermore, attention to a particular object may involve more than onecolumn.

Fig. 13. Layer 5 pyramidal neurons in separate minicolumns connecting to each other via thalamic neurons.


Minicolumn resonant amplitudes may be raised and sustained by projections to the thalamic neurons that serve thoseminicolumns from active frontal columns of the executive system, as in the Gregoriou et al. (2011) study described in Sec-tion 9, paragraph 8, or it may be raised briefly by projections to that thalamic neuron from the superior colliculus respondingto the abrupt onset of a sensory stimulus (e.g., Jonides & Yantis, 1988), as in orienting studies (e.g., Posner, 1980) and ERPstudies (e.g., Hillyard & Anllo-Vento, 1998).

Given that the driving source of attentional selection is the resonant center of the minicolumn, this looping activity se-lects a particular input–output signaling circuit by the intermediate step of transferring resonant activity to the apical den-drites of layer 2/3 pyramids. Examinations of the diagrams of Figs. 11 and 12 indicate that the axons that return to the tops ofthe apical dendrites of layer 5 pyramids also make contact with the tops of the apical dendrites of layer 2/3 pyramids. Hence,ongoing resonant activity in the layer 5 apical dendrites is copied to the layer 2/3 apical dendrites.

The resonating activity in the layer 2/3 apical dendrites is normally sustained over extended periods of time, correspond-ing to the sustained time of the connected layer 5 apical dendrites. For example, in a task containing a cue followed by atarget stimulus after a delay in time (e.g., a green traffic light following the onset of a red traffic light, or a knock on yourhotel room door following the sound of approaching foot steps), the cue initiates resonant activity in the apical dendritesof the layer 5 apical dendrite, which produces resonant activity in the apical dendrites of connected layer 2/3 apical den-drites. After a time delay, the target stimulus appears (the green light or the knock on the door) and delivers a train of inputpulses to the basal dendrites of particular layer 2/3 pyramids. The input voltages (EPSPs, excitatory post-synaptic potentials)from the basal dendrites sums with the ongoing level of voltage at the soma maintained by the apical dendrite, and a train ofoutput pulses is evoked in the soma’s output axon.

Since the level of subthreshold voltage at the soma can be varied, the number of input pulses at the basal dendrite neededto evoke an output axon pulse will vary. Higher levels of soma voltage require fewer input pulses to initiate pulses in theaxon, so that the latency of output pulses will be decreased. Also, when significant noise is present in the stimulus, moreinput pulses are normally needed to produce an axon output corresponding to a confident detection of the stimulus. Butthe additional input pulses can be offset by increasing the ongoing level of voltage at the soma produced by higher (atten-tional) activity in the apical dendrite. Thus, when the signal-to-noise level of a stimulus is low, attention to the stimulus canraise the probability of its detection response as well as shorten the response latency.

However, in cases when the attention level in the Layer 2/3 pyramid, represented by the amplitude of apical dendriteresonance, is inappropriately high, then the output pulses will often be evoked before the stimulus appears, which is an

Fig. 15. A family of exponential functions that describe the decay in a profile of input voltages from the top of an apical dendrite of a layer 2/3 pyramidalneuron. The voltage amplitude decays with distance, d, as the current surge travels toward the soma. The decay constant of the exponential function is g,which represents the average outward conductance of the combined active and passive membrane conductances. All voltage values are given as incrementsfrom baseline level of voltage. The Gaussian-like distributions describe the voltage variability across successive EPSP inputs, which decreases as the currentsurge travels toward the soma. The dark central decay curve represents the condition in which the width of the input frequency profile approaches zero.From LaBerge, 2005.

Fig. 14. Two minicolumns in separate cortical areas are connected in a triangular circuit: both directly between the surround networks and indirectlybetween the center loop circuits via thalamic neurons.



anticipatory error that is labeled a ‘‘false alarm.’’ Therefore in situations where quick responses are called for, the problem forthe person is one of adjusting upward the level of attention to the upcoming stimulus (and to the motor response) just shortof the level that produces an unacceptable rate of false alarms.

We consider now the neural mechanisms that may be involved in adjusting of the level of attention at the somas of layer2/3 pyramids. Fig. 15 shows a family of exponential curves that represent the decay of input voltage (ESPSs) along the shaftof an apical dendrite. Shapes of decay curves for layer 2/3 pyramids given by Larkum, Waters, Sakmann, and Helmchen(2007) are consistent with the exponential-like shape of the decay curves for layer 2/3 pyramidal apical dendrites reportedby others (e.g., Stuart & Spruston, 1998) and for the neural models that apply to the layer 5 apical dendrite (Destexhe,Rudolph, & Pare, 2003; Rall, 1977, 1989). At the input synapses of the apical dendrite a range of voltages (or currents) isshown, with both Gaussian-like and Rayleigh-like distributions that resemble the profiles of the more well-developedresonance curves in Fig. 10.

The input voltage (EPSP) in the apical dendrite decays as the current surge travels toward the soma, and for the layer 2/3pyramid reaches a low asymptotic level that is typically located below the threshold for firing of the axon exiting the soma.Hence, unlike the layer 5 pyramid, the voltage received by the layer 2/3 pyramidal soma from the apical dendrite will not, byitself, evoke a pulse output in the axon. To produce axon firings in the layer 2/3 pyramid voltage must be added from pulseinputs received in the basal dendrites. Cellular recordings in the cat visual cortex (Volgushev, Chistiankova, & Singer, 1998;Volgushev, Pernberg, & Eysel, 2002) offer support for the hypothesis that subthreshold activity at the soma influences input–output processing of the cortical neuron. These studies also showed that the spike count produced in response to the sametest stimulus was higher when the gamma (20–70 Hz) power of subthreshold membrane oscillations was higher.

The substantial convergence of the voltage decay curves in Fig. 15 implies that an initially high variance of voltage inputsat the top of the apical dendrite can be effectively reduced by the time the current surges reach the soma. The reduction involtage variance is described by the shrinking distributions drawn across the curves shown in Fig. 15. To achieve the highestlevels of stability of information processing in layer 2/3 pyramids, the variance in input voltage (current) to the soma by theapical dendrite should decrease toward zero, represented in Fig. 15 by the darkest curve at the center of the group of decaycurves. This reduction in variance translates into higher stability in voltage level at the soma, so that the variability of spikeoutputs, in turn, will be reduced. Therefore the voltage amplitude of the input at the top of the layer 2/3 apical dendrite canbe varied over a wider range without producing false alarms at the high end of the range or negating processing at the lowend. Since the length of the apical dendrite influences the degree of convergence of the decay curves, it would seem thatmammals with longer layer 2/3 apical dendrites potentially can achieve a higher level of stability in their responding to aconstant stimulus.

Even though the resonance of the layer 5 apical dendrite is narrowed by many cycles through the thalamus, the projectionfrom the layer 5 pyramid via the thalamus to the top of the layer 2/3 apical dendrites includes synapses whose noise willincrease the variance of the frequency distribution of pulses delivered to the layer 2/3 pyramid. The length of the layer 2/3 apical dendrite can partially compensate for this added variance in frequency. Thus, the length of the layer 2/3 apical den-drite contributes also to the stability of the peak resonant frequency among the many pyramids of the surrounding networkpart of the minicolumn. Taken together, the reduction in voltage amplitude variance and the added narrowing of the fre-quency distribution within the layer 2/3 apical dendrite enables more precise information processing in mental operationswithin the surrounding networks of minicolumns.

12.1.1. A role of inhibition in selective attentionCenter-surround networks involving layer 5 and 6 pyramids involve axons that project beyond the borders of columns. As

a result, activation of these pyramids can send signals to inhibitory neurons in surrounding columns that suppress activity inthe layer 5 and 6 pyramids of those columns. Given that inhibitory axon terminals virtually cover the pyramid somas, theaction of inhibition of apical dendrite activity can occur quickly and effectively.

Inhibition in search (e.g., Andrews, Watson, Humphreys, & Braithwaite, 2011) provides an illustrative example of theways that resonating minicolumn loops can indirectly select a particular object defined by a conjunction of features in highervisual areas such as V4 and IT. It is hypothesized that, when massed trials of the typical conjunctive search task involve fre-quent focal attention, all features of recently displayed objects, especially their location, will produce very high levels of res-onating in their corresponding minicolumns (of higher sensory areas). The resulting resonance activity in the minicolumnswill not shut down in these higher sensory areas when the display vanishes, but will continue oscillating with only slightdecay before the next display appears. Therefore, when the next display appears or is cued, all of the minicolumns represent-ing the features and locations will be active at levels near their upper limit. In this case, the only way to produce a relativelyhigher activity level in layer 5 pyramids for the target object relative to the surrounding objects is for the target column toinhibit columns representing surrounding locations. As a result, when the target display appears, the activation to the loca-tion and features will be increased only slightly, while the inhibition to the surrounding locations will receive sufficient addi-tional activation to suppress the activity in the corresponding minicolumns. In this way the activity in the targetminicolumns will now exceed the activities in the surrounding columns, which provides the needed condition for attentionalselection.

To briefly summarize this section on attention, the theory describes the selective aspect of attention as an adjustment ofvoltage base levels in layer 2/3 pyramids in the higher levels of sensory areas. This adjustment takes place by transmission ofoscillations from layer 5 pyramids to the layer 2/3 apical dendrites. The major neural benefit of these operations is that fewer


input pulses are needed to produce a specified pulse output from the layer 2/3 pyramids. The consequent cognitive benefit ofthis attentional condition is that processing through a given neuron occurs more quickly, and the negative effects of noise arereduced.

12.2. Preparatory attention

While the typical duration of selective attention to a location during search is relatively brief, the typical duration of pre-paratory attention over a period of time in expectation of a particular upcoming event is almost always much longer. Forexample, waiting for a red light to change to green at a traffic intersection usually involves attention to anticipating the greencolor that is about to appear below the red color, and involves holding that anticipation over an extended period of time.When the green light appears, the driver responds more quickly than if he or she did not prepare to see the green light. Alaboratory example is given by the preparatory attention investigated in the Gregoriou et al. (2011) experiment describedearlier.

The application of the present theory to preparatory attention was extensively described in an earlier article (LaBerge,2005) and the main points of that description are included in the present paper. A simple experimental procedure involvingthe appearance of dots in three boxes was developed to measure preparatory attention in normal humans (LaBerge, Auclair,& Sieroff, 2000) and patients (e.g., Sieroff et al., 2004).

12.3. Sustained attention

A third mode of attention is the prolonging of attention without anticipation of some upcoming event, also termed main-tenance attention (LaBerge, 1995). The simple act of sustaining attention to something has sometimes been interpreted asattending to something ‘‘for its own sake,’’ i.e., for the experiencing of the content of the attention. Examples are: gazing atpaintings, listening to music, pausing at the end of a line of verse to prolong an image, meditating, imagining the face of aspecial person, and imagining scenes from last year’s vacation. In all of these cases, the dominant activity of the minicolumnis the resonating center loop; the surround signaling networks are presumed to be relatively quiescent, except for pulse pro-cessing that is highly automatic, for example, the stream of words that some people report make up the background of mostof their waking life and even dreaming life. One may convert sustained attention into preparatory attention simply by addingthe image of an upcoming event or goal, so that one attends less to ‘‘what is’’ and more to ‘‘what will be.’’

Pure sustained attention has received much less study in the cognitive sciences than other aspects of attention, perhapsbecause open-ended extending of attention precludes an overt response which could measure the accuracy and responsetime resulting from neural activities taking place during the sustained attention interval. Measurements which are moreappropriate for inferring the nature of neural processing during prolonged attention are the scalp EEG (electroencephalo-gram) and the cellular LFP (local field potential). The three aspects of attention, selective attention, preparatory attention,and sustained attention, described in this paper, were more extensively treated in an earlier publication (LaBerge, 1995).

A particularly persistent target of prolonged attention is the feeling of our bodies, feelings experienced at special timesbut also at random occasions during daily life. We will take up this topic next, and give descriptions and analyses of bodilyfeeling using the neural model of cortical activity developed here, with special emphasis on the resonating center of theminicolumn.

13. Describing feelings as resonant activity in the insular cortex

We describe the neural activity underlying feelings, or underlying ‘‘having a feeling’’ as the resonating of the looping cir-cuit in the center of minicolumns in particular areas of the cerebral cortex. Current research on feeling has been directed

Fig. 16. The insular cortex revealed by pulling back the cortical tissue (the opercula, or ‘‘lids’’) at the border of the temporal and parietal lobes.


mainly to the insular cortex, but some investigators have suggested that the anterior cingulate cortex, which is closely con-nected to the insula, may be also be involved in the activities of feeling, in particular the regulation of feelings (Craig, 2009).

The key idea by which the present theory of neural resonance is applicable to the functions of the insular cortex is that inthis area of the cortex the resonant activities in the circuit loops at the center of insular minicolumns are ‘‘about’’ the wholebody. The peak frequencies generated by the several sectors of the insula establish the basis for specific circuit activities thatspread outward to contact neural structures that represent the different aspects of the body perceived as-a-whole, e.g., itstemperature and balance. But the shared referent of activity in the several sectors of the insula appears to be the particularimpression of the whole body at a given moment, which is typically verbalized as ‘‘how we are feeling now.’’

While the center circuit of the minicolumn produces the subjective experience of feeling it does not itself produce infor-mation–bearing signals that are integrated into circuit networks that process characteristics of particular kinds of body feel-ings. As pointed out earlier, the information-processing part of the minicolumn activity takes place in the network thatsurrounds the center loop; the center loop activity enables and modulates the processing within the surrounding networkthrough the intermediate structure of the apical dendrites of layer 2/3 neurons. These resonant dendrites modulate the in-put–output processing that involves the basal dendrites, soma, and axon of the layer 2/3 neurons, as we describe and analyzethe specific feelings that are currently happening to us. Therefore, as we explore the different parts of insular cortex, we willbe especially interested in the different participations of the resonating layer 5 pyramids and the information-processinglayer 2/3 pyramids in producing whole-body feelings.

13.1. Anatomy and connections of the insular cortex

To support the proposal that the whole-body impression is provided by the insula, we briefly describe the anatomy of theinsula, along with the densities of the layer 5 pyramids and their connections with the thalamus, both of which are necessarycomponents of the corticothalamic loop in which the resonating activity of the whole-body impression takes place. For arecent review of the anatomy and functions of the insular cortex see Nieuwenhuys (2012, chap. 7).

The insula, sometimes called the fifth lobe of the cortex, is a small triangular-shaped part of the cerebral cortex (less than2% of the surface area), and it is hidden under the lid-like boundaries of the temporal and parietal cortices, as shown inFig. 16. According to a study by Morel, Gallay, Baechler, Wyss, and Gallay (2013), the human insula exhibits seven subdivi-sions identified by the characteristics of the layers of neurons. Within each subdivision are many sectors serving a variety offunctions. A meta-analysis of 1768 experiments by Kurth, Zilles, Fox, Laird, and Eickhoff (2010) combined 13 different func-tional categories of the insula into four domains: sensorimotor, cognitive, chemical sensory, and social emotional, corre-sponding roughly with the posterior to anterior orientation of the insula within the brain. These four categories areconsistent with the categories suggested by Cerliani et al. (2012), who used probabilistic fiber tracking with 10 in vivohumans.

At the posterior end of the insula (located at the right portion in Fig. 16), the thickness pattern of the 6 layers of cortexresembles that of the cortical layering in the adjacent parietal and temporal lobes (see Fig. 2 for the parietal pattern of layerthicknesses). At the anterior end of the insula, layers 2 and 4 (containing small granule neurons) have vanished, while thepyramidal neurons of layer 5 have increased in number and size (Morel et al., 2013). Thus a gradient of layer thickness existsbetween the posterior and anterior sectors of the insula: as one moves from the posterior to the anterior parts of the insulathe thicknesses of layers 2 and 4 decrease, and the thickness of layer 5 increases, owing to the increase in number and size ofthe layer 5 pyramidal neurons.

One is tempted to conjecture that the increase in size and number of layer 5 pyramids in the anterior sector of the insulaindicates that stronger impressions of body feeling of the emotional kind are more available than the sensorimotor kind. Forexample, the bodily mood during a good dinner generally may be more vivid than the bodily mood during movements of thelimbs and torso in simple calisthenics or yoga. The decrease in layer 2 and layer 4 thickness indicates more sparse minicol-umn network circuitry and therefore less within and between minicolumn information processing in emotional whole-bodyfeelings than in sensorimotor whole-body feelings.

We turn now to the thalamic inputs to the many sectors of the insula, which form part of the corticothalamic loops thatunderlie the resonant activity treated in the present paper. Relayed through these thalamic inputs are the signals from thebody’s internal milieu and external gestures that communicate the current status of bodily functions. The thalamic fiberscontact layer 5 and layer 6 pyramids at their apical dendrites, and then the current surges flow down these apical dendritesto the soma; subsequently the pyramidal axons return activity to the thalamic neuron of origin. This forms the circuit loopthat generates resonant electrical activity in the apical dendrites of the several sectors of the insula.

The group of thalamic nuclei which project to the anterior insular cortex relay signals from visceral, gustatory, and auto-nomic structures, with the ventral posterior medial (VPM) thalamic nucleus having a prominent role (Mufson & Mesulam,1984). The anterior insula is concerned with moods and emotional states, such as disgust and feelings of unease as well ashappiness, maternal and romantic love (Craig, 2009). The anterior insula is also involved with taste (Rolls, 2006), and the leftanterior insular responds to all odors, while the right insula is activated specifically by disgusting odors (Heining et al., 2003).A review of studies of socially relevant functions of the insula (Lamm & Singer, 2010), indicate involvement of the anteriorinsula in empathy, compassion, fairness, and cooperation.

Of the thalamic nuclei that send fibers to the posterior insular cortex, the ventral posterior inferior (VPI) thalamic nucleusappears to be prominent, and the oral and medial pulvinar nuclei of the thalamus are reciprocally connected with both the


anterior and posterior pulvinar (Mufson & Mesulam, 1984). The posterior insula is involved in somatosensory functions, suchas feeling one’s sense of being, and the awareness of the functioning of one’s own body parts (Karnath, Baier, & Nägele, 2005).

Each of the foregoing thalamic nuclei sends signals arising from bodily changes to the sectors of the insula that combineto evoke a medley of whole-body feelings. A description of these different feelings in terms of mental content depends on theway the minicolumn surrounding network is connected to the variety of limbic and sensorimotor structures. But the feelingsthemselves resist description because they are not connected by information-bearing signals outside of their center minicol-umn loops. Therefore feelings themselves are more accurately treated as subjective impressions occurring in the minicolumncenter loop that are pointed to by the information-processing in the surrounding network of that minicolumn.

13.2. Attention to specific insular functions

While feelings are generally produced by external circumstances and mental images, can we modulate them by directingour attention to them, for example, can we intensify the feeling of happiness or sadness by focusing on our body in certainways? If it is possible to modulate the feeling by attention, then neural connections must exist that control the attentionbeing expressed by enhanced levels of resonance in minicolumn center loops of the insula. Earlier in this paper we describedan experiment by Gregoriou et al. (2011) which investigated the control of attention to the color of a circle so as to detect achange in the color of the circular pattern. The color display is represented in area V4 and the expression of attention in V4

was controlled by the frontal eye field (FEF) area, which is reciprocally connected to area V4. A diagrammatic account by thepresent theory of those connections is shown in Fig. 14. By analogy we seek a frontal cortical area that is reciprocally con-nected to a sector of the insular cortex through thalamic nuclei.

A promising candidate for the frontal area that controls attention to the insula is the orbital prefrontal cortex (OPFC), ow-ing to its involvement in the motivational effects of decision-making (e.g., Damasio, 1994). We use the diagram of intercon-nections shown in Fig. 13 as a guide for locating the connections required to support the transfer of resonant activity in aminicolumn of the orbital prefrontal area to a minicolumn of one of the sectors of the insula.

It appears that the OPFC does not send fibers to the all the thalamic nuclei that relay signals from bodily activity to theinsula. According to a fiber labeling study by Cavada, Company, Tejedor, Cruz-Rizzolo, and Reinoso-Suarez (2000), the OPFCis connected with the mediodorsal (MD) nucleus and the medial pulvinar (Pul M), nucleus but no connections were shownwith the ventral posterior medial (VPM) nucleus, which provides major inputs to the anterior insula, nor with the ventralposterior inferior (VPI) nucleus, which provides major inputs to the posterior insula. Therefore, according the present anal-ysis, the circuitry that controls attention to the anterior and posterior parts of the insula would appear to involve the medi-odorsal (MD) thalamic nuclei for the OPFC minicolumns, and the medial pulvinar (Pul M) thalamic nuclei for both theanterior and posterior sectors of the insula (in Fig. 13 the left thalamus represents the MD and the right thalamus representsthe Pul M).

According to the present theory, the stronger attentional effects are predicted to take place in the anterior insula, owing tothe finding (e.g., Morel et al., 2013) that the layer 5 pyramidal neurons are larger and more numerous in the anterior insulathan in the posterior insula (with a gradient of these features extending across the insular sections between these two end-points of the insula). Groups of more numerous and larger layer 5 pyramidal neurons produce a higher level of total oscil-lation amplitude in minicolumns of the anterior insula, so that prolonged attention to anterior insular functions, such asemotion-based feelings of sadness or happiness, presumably can be intensified more than attention to posterior insular func-tions such somatosensory-based feelings of bodily motion of owning the movement of one’s limbs. Evidence from medita-tion studies that compare expert meditators with novices indicates that when they attend to emotional states of compassionthe anterior insula of experts is more active than that of controls (Lutz, Brefczynski-Lewis, Johnstone, & Davidson, 2008), andthat the cortical thickness of their right anterior insula is greater than that of controls (Lazar et al., 2005).

The present concept of the resonating neuron extends the range of descriptions of the contents of cognition by providing amore direct substrate for the cognitive features of duration and intensity than is provided by the concept of signaling. Dura-tion and intensity feature prominently in human activity concerned with the arts, e.g., the sounds of music and the images ofpoetry. But perhaps more universal are the durations and intensities of feelings, which are whole-body responses to eventsand situations encountered in everyday life. While future research will undoubtedly illuminate the details of the neuroanat-omy and function of the sectors of the insula, anterior, posterior, and the other sectors located between them, the concept ofthe resonating apical dendrite could help to more closely relate the neuroanatomical findings with their counterparts in cog-nitive feelings.

Acknowledgments

The authors express their appreciation to the editor and the reviewers for their penetrating comments and efforts to im-prove the manuscript. We also thank three people who read an earlier draft of the manuscript and made helpful comments:We thank John Compton for clarifying philosophical issues and for his continual encouragement, Anne La Berge for her mu-sical ideas about the resonant dendrite and her demonstrations of resonance ideas in musical performance, and Eric Sierofffor his collaborative work in the area of preparatory attention and his general neuropsychological expertise.


References

Andrews, L. S., Watson, D. G., Humphreys, G. W., & Braithwaite, J. J. (2011). Flexible feature-based inhibition in visual search mediates magnifiedimpairments of selection: Evidence from carry-over effects under dynamic preview-search conditions. Journal of Experimental Psychology: HumanPerception and Performance, 37, 1007–1016.

Bekkers, J. M. (2000). Distribution and activation of voltage-gated potassium channels in cell-attached and outside-out patches from large layer 5 corticalpyramid neurons of the rat. Journal of Physiology, 525(3), 611–629.

Bell, C. (1811). An idea of a new anatomy of the brain; submitted for the observations of his friends. A privately printed pamphlet. London: Strahan & Preston.Briggs, F. (2010). Organizing principles of cortical layer 6. Frontiers in Neural Circuits, 4, 3. http://dx.doi.org/10.3389/neuro.04.003.2010.Brumberg, J. C., Nowak, L. G., & McCormick, D. A. (2000). Ionic mechanisms underlying repetitive high-frequency burst firing in supragranular cortical

neurons. Journal of Neuroscience, 20, 4829–4843.Brumberg, J. C., Hamzei-Sichani, F., & Yuste, R. (2003). Morphological and physiological characterisation of layer VI corticofugal neurons of mouse primary

visual cortex. Journal of Neurophysiology, 89, 2854–2867.Buchsbaum, M. S., Buchsbaum, B. R., Chokron, S., Tang, C., Wei, T.-C., & Bynea, W. (2006). Thalamocortical circuits: fMRI assessment of the pulvinar and

medial dorsal nucleus in normal volunteers. Neuroscience Letters, 404, 282–287.Busch, N. A., Dubois, J., & VanRullen, R. (2009). The phase of ongoing EEG oscillations predicts visual perception. The Journal of Neuroscience, 29, 7869–7876.Cavada, C., Company, T., Tejedor, J., Cruz-Rizzolo, R. J., & Reinoso-Suarez, F. (2000). The anatomical connections of the macaque monkey orbitofrontal cortex.

A review. Cerebral Cortex, 10, 222–242.Cerliani, L., Thomas, R. M., Jbabdi, S., Siero, J. C. W., Nanetti, L., Crippa, A., et al (2012). Probabilistic tractography recovers a rostrocaudal trajectory of

connectivity variability in the human insular cortex. Human Brain Mapping, 33, 2005–2034.Craig, A. D. (2009). How do you feel—now? The anterior insula and human awareness. Nature Reviews Neuroscience, 10, 59–70.Damasio, A. R. (1994). Descartes’ error: Emotion, reason, and the human brain. New York: Putnam.DeFilipe, J., & Jones, E. G. (1988). Cajal on the cerebral cortex. New York: Oxford University Press.Delogne, P. (1982). Leaky feeders and subsurface radio communication. New York: Peter Peregrines Ltd..Destexhe, A., Rudolph, M., & Pare, D. (2003). The high-conductance state of neocortical neurons in vivo. Nature Reviews of Neuroscience, 4, 739–751.Elston, G. N. (2013). Cortex, cognition and the cell: New insights into the pyramidal neuron and prefrontal function. Cerebral Cortex, 13, 1124–1138.Feldman, M. L (1984). Morphology of the neocortical pyramidal neuron. In A. Peters & E. G. Jones (Eds.). Cerebral cortex. cellular components of the cerebral

cortex (Vol. 1, pp. 123–200). New York: Plenum Press.Garcia-Lopez, P., Garcia-Marin, V., & Freire, M. (2006). Three dimensional reconstruction and quantitative study of a pyramidal cell of a Cajal histological

preparation. Journal of Neuroscience, 26, 11249–112152.Gray, C., & McCormick, D. A. (1996). Superficial pyramidal neurons contributing to the generation of synchronous oscillations in the visual cortex. Science,

274, 109–113.Gregoriou, G. G., Gotts, S. J., Zhou, H., & Desimone, R. (2011). High-frequency, long-range coupling between prefrontal and visual cortex during attention.

Science, 324, 1207–1210.Harris, K. M., & Kater, S. B. (1994). Dendritic spines: Cellular specializations impartingboth stability and flexibility to synaptic function. Annual Reviews of

Neuroscience, 17, 341–371.Heining, M., Young, A. W., Ioannou, G., Andrew, C. M., Brammer, M. J., Gray, J. A., et al (2003). Disgusting smells activate human anterior insula and ventral

striatum. Annals of the New York Academy of Sciences, 1000, 380–384.Higgs, M. H., & Spain, W. J. (2009). Conditional bursting enhances resonant firing in neocortical Layer 2–3 pyramidal neurons. Journal of Neuroscience, 29,

1285–1299.Hillyard, S. A., & Anllo-Vento, L. (1998). Event-related brain potentials in the study of visual selective attention. Proceedings of the National Academy of

Sciences USA, 95, 781–787.Ho, N., & Destexhe, A. (2000). Synaptic background activity enhances the responsiveness of neocortical pyramidal neurons. Journal of Neurophysiology, 84,

1488–1496.Hu, H., Vervaeke, K., Graham, L. J., & Storm, J. F. (2009). Complementary theta resonance filtering by two spatially segregated mechanisms in CA1

hippocampal pyramidal neurons. Journal of Neuroscience, 29, 14472–14483.Hutcheon, F., & Yarom, Y. (2000). Resonance, oscillation and the intrinsic frequency preferences of neurons. Trends in Neurosciences, 23, 216–222.Izhikevich, E. M., Desai1, N. S., Walcott, E. C., & Hoppenstead, F. C. (2003). Bursts as a unit of neural information: Selective communication via resonance.

Trends in Neurosciences, 3, 161–167.Jonides, J., & Yantis, S. (1988). Uniqueness of abrupt visual onset in capturing attention. Perception and Psychophysics, 1988(43), 346–354.Jones, E. G. (1985). The thalamus. New York: Plenum Press.Jones, E. G. (2002). Thalamic circuitry and thalamocortical synchrony. Philosophical transactions of the Royal Society of London, B, 357, 1659–1673.Jones, E. G. (2007). The thalamus (2nd ed.). Cambridge, UK: Cambridge University Press.Kahana, M. J. (2006). The cognitive correlates of human brain oscillations. Journal of Neuroscience, 26, 1669–1672. http://dx.doi.org/10.1523/jneurosci.3737-

05c.2006.Karnath, H.-O., Baier, B., & Nägele, T. (2005). Awareness of the functioning of one’s own limbs mediated by the insular cortex? Journal of Neuroscience, 25,

7134–7138.Kasevich, R. S., & LaBerge, D. (2010). Theory of electric resonance in the neocortical apical dendrite. PLoS One, 6(8), e23412.Korngreen, A., & Sakmann, F. (2000). Voltage-gated K+ channels in layer 5 neocortical pyramidal neurons from young rats: Subtypes and gradients. Journal of

Physiology, 525(3), 621–639.Korzybski, A. (1994). Science and sanity: An introduction to non-aristotelian systems and general semantics (5th ed.). Institute of General Semantics.Kurth, F., Zilles, K., Fox, P. T., Laird, A. R., & Eickhoff, S. B. (2010). A link between the systems: Functional differentiation and integration within the human

insula revealed by meta-analysis. Brain Structure and Function, 214, 519–534.LaBerge, D. (1995). Attentional processing: The brain’s art of mindfulness. Harvard University Press.LaBerge, D. (1997). Attention, awareness, and the triangular circuit. Consciousness and Cognition, 6, 149–181.LaBerge, D. (2005). Sustained attention and apical dendrite activity in recurrent circuits. Brain Research Reviews, 50, 86–99.LaBerge, D., Auclair, L., & Sieroff, E. (2000). Preparatory attention: Experiment and theory. Consciousness and Cognition, 9, 396–434.LaBerge, D., & Buchsbaum, M. S. (1990). Positron emission tomographic measurements of pulvinar activity during an attention task. Journal of Neuroscience,

10, 613–619.LaBerge, D., & Kasevich, R. S. (2007). The apical dendrite theory of consciousness. Neural Networks, 20, 1004–1020.Lamm, C., & Singer, T. (2010). The role of anterior insular cortex in social emotions. Brain Structure and Function, 214, 579–591.Lazar, S. W., Kerr, C. E., Wasserman, R. H., Gray, J. R., Greve, D. N., Treadway, M. T., et al (2005). Meditation experience is associated with increased cortical

thickness. Neuroreport, 16, 1893–1897.Larkum, M. E., Waters, J., Sakmann, B., & Helmchen, F. (2007). Dendritic spikes in apical dendrites of neocortical layer 2/3 pyramidal neurons. Journal of

Neuroscience, 27, 8999–9008.Lee, C. C., & Sherman, S. M. (2009). Modulator property of the intrinsic cortical projection from Layer 6 to Layer 4. Frontiers in Systems Neuroscience, 3, 3.

http://dx.doi.org/10.3389/neuro.06.003.2009.Lesica, N. A., Weng, C., Jin, J., Yeh, C.-I., Alonso, J.-M., & Stanley, G. B. (2006). Dynamic encoding of natural luminance sequences by LGN bursts. PLoS Biology,

4(7), 3209. http://dx.doi.org/10.1371/journalpbio.0040209.

http://refhub.elsevier.com/S1053-8100(13)00143-8/h0005






http://dx.doi.org/10.3389/neuro.04.003.2010













































http://dx.doi.org/10.1523/jneurosci.3737-05c.2006

http://dx.doi.org/10.1523/jneurosci.3737-05c.2006





















http://dx.doi.org/10.3389/neuro.06.003.2009

http://dx.doi.org/10.1371/journalpbio.0040209


Llano, D., & Sherman, S. M. (2009). Differences in intrinsic properties and local network connectivity of identified layer 5 and layer 6 adult mouse auditorycorticothalamic neurons support a dual corticothalamic projection hypothesis. Cerebral Cortex, 19, 2810–2816.

Lopes da Silva, F. (1991). Neural mechanisms underlying brain waves: From neural membranes to networks. Electroencephalography and ClinicalNeurophysiology, 79, 81–93.

Lutz, A., Brefczynski-Lewis, J., Johnstone, T., & Davidson, R. J. (2008). Regulation of the neural circuitry of emotion by compassion meditation: Effects ofmeditative expertise. PLoS One, 3(3), e1897.

Lutz, A., Greischar, L. L., Rawlings, N. B., Ricard, M., & Davidson, R. J. (2004). Long-term meditators self-induce high-amplitude gamma synchrony duringmental practice. Proceedings of the National Academy of Sciences USA, 10, 16369–16373.

Magendie, F. (1822). Expériences sur ls fonctions des racines des nerfs rachidiens. Journal de Physiologie Expérimentale et de Pathologie, 276–279.Mercer, A., West, D. C., Morris, O. T., Kirchbecker, S., Kerhoff, J. E., & Thomson, A. M. (2005). Excitatory connections made by presynaptic cortico-cortical

pyramidal cells in layer 6 of the neocortex. Cerebral Cortex, 15, 1485–1496.Morel, A., Gallay, M. N., Baechler, A., Wyss, M., & Gallay, D. S. (2013). The human insula: Architectonic organization and postmortem MRI registration.

Neuroscience, 236, 117–135.Mountcastle, V. B. (1998). The cerebral cortex. Cambridge, MA: Harvard University Press.Mufson, E. J., & Mesulam, M. M. (1984). Thalamic connections of the insula in the rhesus monkey and comments on the paralimbic connectivity of the

medial pulvinar nucleus. Journal of Comparative Neurology, 227, 109–120.Nagerl, U. V., Willig, K., Hein, B., Hell, S. W., & Bonhoeffera, W. (2008). Live-cell imaging of dendritic spines by STED microscopy. Proceedings of the National

Academy of Sciences USA, 105, 18982–18987.Narayanan, R., & Johnston, D. (2008). The h-channel mediates location dependence and plasticity of intrinsic phase responses in rat hippocampal neurons.

Journal of Neuroscience, 28(22), 5846–5860. http://dx.doi.org/10.1523/JNEUROSCI.0835-08.2008.Nieuwenhuys, R., Voogd, J. K., & van Huizen, C. (2008). The human central nervous system (4th ed.). Berlin: Springer (Section II).Nieuwenhuys, R. (2012). The insular cortex: A review. In M. A. Hofman & D. Falk (Eds.) Evolution of the primate brain: From neuron to behavior. Progress in

brain research, Vol. 195, pp. 123–163.Palva, S., & Palva, J. M. (2011). Functional roles of alpha-band phase synchronization in local and large-scale cortical networks. Frontiers in Psychology. http://

dx.doi.org/10.3389/fpsyg.2011.00204.Peters, A., & Sethares, C. (1991). Organization of pyramidal neurons in area 17 of monkey visual cortex. Journal of Comparative Neurology, 306, 1–23.Peters, A., & Vilmez, E. (1993). Neuronal organization in area 17 of cat visual cortex. Cerebral Cortex, 3, 49–68.Peterson, S. E., Robinson, D. L., & Keys, W. (1985). Pulvinar nuclei of the behaving rhesus monkey: Visual responses and their modulation. Journal of

Neurophysiology, 54, 867–886.Posner, M. I. (1980). The orienting of attention. Quarterly Journal of Experimental Psychology, 32, 3–25.Rall, W. (1977). Core conductor theory and cable properties of neurons. In Handbook of Physiology. In T. M. Brookhardt, E. R. Kandel, & V. B. Mountcastle

(Eds.). Handbook of physiology, the system: Cellular biology of neurons (Vol. 1, pp. 139–198). Bethesda, MD: Williams & Wilkins.Rall, W. (1989). Cable theory for dendritic neurons. In: C. Koch & I. Segev (Eds.), Methods in neuronal modeling. Cambridge, Massachusetts: MIT Press.Rockel, A. J., Horns, R. W., & Powell, T. P. S. (1980). The basic uniformity in the structure of the neocortex. Brain, 103, 221–244.Rolls, E. T. (2006). Brain mechanisms underlying flavour and appetite. Philosophical Transactions of the Royal Society of London B Biological Sciences, 361,

1123–1136.Saalmann, Y. B., Pinsk, M. A., Wang, L., Li, X., & Kastner, S. (2012). The pulvinar regulates information transmission between cortical areas based on attention

demands. Science, 337, 753–756.Shannon, C. E., & Weaver, W. (1949). The mathematical theory of communication. Urbana, Illinois: University of Illinois Press.Sieroff, E., Piquard, A., Auclair, L., Lacomblez, L., Derouesne, C., & LaBerge, D. (2004). Deficit of preparatory attention in frontal-temporal dementia. Brain and

Cognition, 55, 444–451.Stuart, G., & Spruston, N. (1998). Determinants of voltage attenuation in neocortical pyramidal neuron dendrites. Journal of Neuroscience, 18, 3501–3510.Treisman, A., & Gelade, G. (1980). A feature-integration theory of attention. Cognitive Psychology, 12, 97–136.Van Essen, D. C., Drury, H. A., Joshi, S., & Miller, M. I. (1998). Functional and structural mapping of human cerebral cortex: Solutions are in the surfaces.

Proceedings of the National Academy of Sciences, USA, 95, 788–795.Volgushev, M., Chistiankova, M., & Singer, W. (1998). Modification of discharge patterns of neocortical neurons by induced oscillations of the membrane

potential. Neuroscience, 83, 15–25.Volgushev, M., Pernberg, J., & Eysel, U. T. (2002). A novel mechanism of response selectivity of neurons in cat visual cortex. Journal of Physiology (London),

540, 307–320.Williams, S. R., & Stuart, G. J. (2000). Site independence of EPSP time course is mediated by dendritic I(h) in neocortical pyramidal neurons. Journal of

Neurophysiology, 83, 3177–3182.Wiser, A. K., & Callaway, E. M. (1996). Contributions of individual layer 6 pyramidal neurons to local circuitry in macaque primary visual cortex. Journal of

Neuroscience, 16, 2724–2739.Yuste, R. (2010). Dendritic Spines. Cambridge, MA: MIT Press.Yuste, R. (2011). Dendritic spines and distributed circuits. Neuron, 71, 772–781.Zhuchkova, E., & Schreiber, S. (2011). Dendritic versus somatic resonance. BMC Neuroscience, 12(Supplement 1), p289. http://dx.doi.org/10.1186/1471-2202-

12-S1-P289.






















http://dx.doi.org/10.3389/fpsyg.2011.00204

http://dx.doi.org/10.3389/fpsyg.2011.00204






























http://dx.doi.org/10.1186/1471-2202-12-S1-P289

http://dx.doi.org/10.1186/1471-2202-12-S1-P289

Documents

The cognitive significance of resonating neurons in the cerebral cortex