POSNER, M. I.; BOIES, S. J., Componentes da Atenção. 1971

Psychological Review1971, Vol. 78, No. S, 391-408

COMPONENTS OF ATTENTION1

MICHAEL I. POSNER 2 AND STEPHEN J. BOIES

University of Oregon

The study of human attention may be divided into three components. Theseare alertness, selectivity, and processing capacity. This paper outlines ex-perimental techniques designed to separate these components and examinetheir interrelations within comparable tasks. It is shown that a stimulusmay be used to increase alertness for processing all external information, toimprove selection of particular stimuli, or to do both simultaneously. Devel-opment of alertness and selectivity are separable, but they may go on to-gether without interference. Moreover, encoding a stimulus may proceedwithout producing interference with other signals. Thus, the contact be-tween an external stimulus and its representation in memory does not appearto require processing capacity. Limited capacity results are obtained whenmental operations such as response selection or rehearsal must be per-formed on the encoded information.

The subject of attention is a broad one.There are many definitions of it and subcate-gories included under it. In a recent book,Moray (1970) has proposed at least six dif-ferent meanings of the term attention in cur-rent psychological research. It would bepremature to propose any single taxonomyor exhaustive set of categories for the term.However, both in looking at the various defi-nitions that Moray proposes and in review-ing the literature, there do appear to be threemajor topics under which studies of atten-tion might be grouped.

First is the notion of alertness. Maintain-ing attention in the sense of alertnessa ispresumably involved in human ability toperform in long, boring tasks like thosethat psychologists design to study vigilance(Mackworth, 1970). The ability to develop

1 Portions of this paper were presented in an in-vited address to the Western Psychological Associ-ation, Los Angeles, April 1970. The research in-cluded here was sponsored by National ScienceFoundation Grants GB 5960 and 21020 and by theAdvanced Research Projects Agency of the De-partment of Defense monitored by the Air ForceOffice of Scientific Research under Contract No.F44620-67-C-0099 all to the University of Oregon.The authors are grateful to Barbara Lewis andLarry Moore for help in running the subjects.

2 Requests for reprints should be sent to MichaelI, Posner, Department of Psychology, Universityof Oregon, Eugene, Oregon 97403.

3 The term alertness is used here rather thanarousal to avoid the emotional connotation of manyuses of the latter term. The degree of separationof the terms is not clear at this point.

and maintain an optimal sensitivity to ex-ternal stimulation is also studied when sub-jects (Ss) receive a warning to preparethemselves to take in information. The fore-period of a reaction time task may be con-sidered as a miniature vigilance situationwhere alertness must be developed rapidlyand maintained over a relatively brief inter-val. Recent studies have investigated thetime course (Bertelson, 1967) and the brainactivity (Karlin, 1970; Naatanen, 1970;Posner & Wilkinson, 1969) during this in-terval. It appears that there is some simi-larity between the brain processes which takeplace during the foreperiod and those in-volved in vigilance tasks (Wilkinson &Haines, 1970). These studies provide somerationale for including alertness as a generalcomponent of attention.

A second sense of attention is the abilityto select information from one source or ofone kind rather than another (Egeth, 1967;Triesman, 1969). Studies of selective atten-tion may require that ^s report informationfrom a particular sensory modality, spatiallocation, or content (letters rather than dig-its). Broadbent (1958) has argued that atleast some aspects of selection are performedby filtering mechanisms which block out orattenuate input. Other theories have stressedadditional operations which are performedon information of particular importance(Deutsch & Deutsch, 1963; Norman, 1968).These views bear a relationship to physio-logical models of attention that have stressed

391

392 MICHAEL I. POSNER AND STEPHEN J. BOIES

either a tuning process which blocks inputfrom unselected sources (Hernandez-Peon,1966) or a general alerting reaction whichenhances input from whatever source (Naa-tanen, 1970). The former view is more inline with filtering mechanisms, while the lat-ter fits better with the idea of additionaloperations performed on highly processed in-put. One way of viewing this issue is to askwhether 5s can prepare for signals in a waywhich is not selective. If they prepare for avisual signal, is the capacity to process audi-tory signals decreased or enhanced over anunprepared state? Thus, the separability ofalertness and selection as internal processeshas important consequences for a theory ofattention.

A third sense of attention relates to theidea of a limited central processing capacity(Broadbent, 1958). It is well known that.S"s have difficulty in handling two taskssimultaneously, even when they wish to doso. Signals which arrive during the reactiontime (RT) period tend to be delayed (Wei-ford, 1968). Yet it is not clear what aspectsof signal processing are subject to the single-channel limitation. Some have argued thatthe single channel extends to almost all pro-cessing (Welford, 1968), while other haveseen it as applying mainly to response ex-ecution (Reynolds, 1964). A limited capac-ity mechanism suggests that any two opera-tions requiring it will interfere with eachother. Thus, a study of the ability of 5"s totime share mental operations can lead to anunderstanding of what aspects of processingrequire access to the limited capacity system.

Most recent articles related to attention in-volve one of these three components (seee.g., Koster, 1969; Sanders, 1967, 1970).Unfortunately, the experiments used to studyeach of the components are different andthere has been little attempt to develop taskswhich allow each of them to be separated andinterrelated. Moray (1970) suggests theimportance of doing so as follows:

It might well be, for example, that the relation be-tween selective listening and arousal is such thatarousal level acts as a parameter which will alterthe over-all efficiency of selection and rejection asit varies. But, no systematic investigation has sofar been carried out [p. 85].

This paper is an attempt to develop means

for studying the separation and interrelation-ship of these components of attention at thebehavioral level.

Consider first a task which involves allthree components mixed in a fairly complexway. The 6"s see a single letter for a briefperiod of time, followed after a variable in-terval by a second letter. The task is to re-spond as rapidly as possible "same" if thetwo letters have the same name; otherwise,"different" (Posner, 1969; Posner, Boies,Eichelman, & Taylor, 1969). The first let-ter has two functions. It serves as a warningsignal and also tells the 5" what letter he islooking for. Thus, one might expect it toincrease alertness and to introduce some biasof the perceptual system toward tests relatedto the configuration of the first letter. Thebias might involve, for example, activation oftrace systems related to the physical con-figuration of the letter (Posner, 1969).What happens during the interval betweenthe two letters depends greatly on how theattention (processing capacity) of 5 is di-rected. For example, if he knows that allmatches will be physically identical, he main-tains his speed on physical matches muchbetter than if some matches must be based onthe name (Posner, 1969). If the first letterdisappears, there is a rapid increase in timesfor a physical match, while name matches re-main relatively constant, suggesting that 5"skeep the letter name in mind. However, ifthe physical form remains present during theinterval, physical matching remains constantwhile name matches increase in efficiencyuntil they resemble the physical matches.This suggests that 5"s generate a visual codeof the opposite case (Boies, 1969; Posner,1969). These results indicate that 5s arevery flexible in the coding they use for let-ters and similar visual material (Cohen,1969; Tver sky, 1969), but their choices areconstrained by the rivalry between physicaland name codes for SV limited processingcapacity.

This brief description of the letter-match-ing sequence suggests that it may be a usefultask for separating the three components ofattention described above. The experimentsreviewed in this paper are an effort to do so.Moreover, they attempt to connect the opera-tions of this experimental paradigm to the

COMPONENTS OF ATTENTION 393

three theoretical components of attentionoutlined above. Two major questions areconsidered. First, Can the warning functionand the selective function of the first letter beseparated? If so, how do the two functionsrelate? Second, What is the relationshipbetween these two components of attentionand central processing capacity?

PREPARATION

Many investigators have studied the periodof time between a warning signal and a sig-nal related to the required response (Bertel-son, 1967). Of primary interest are studieswhere 51 has little uncertainty about when theresponse-related signal will occur. This isachieved by using blocks of trials with a fixedwarning interval. Whether the responsetask involves reaction time (Bertelson,1967), signal detection (Egan, Greenberg,& Schulman, 1961), or other types of pro-cessing (Leavitt, 1969), the results are simi-lar. Performance is worse with no warningand improves as the warning interval in-creases to some optimal value. This valueis most frequently about .2-.5 second. Aninterval longer than optimal tends to producea decrease in performance, but this is usually

not as pronounced, particularly when feed-back is used to keep motivation high.

Figure 1 indicates functions obtained usingtwo-letter matching tasks (Posner & Wilkin-son, 1969). Both tasks involved a warningsignal followed after a foreperiod by a pairof letters. The foreperiod was constant fora block of trials but varied between blocks.The 5"s had a single key which they pressedwhen the letter pair was defined as "same"for that task. The solid line is for a taskwhich required 51 to press a key if the letterswere physically identical, while the dottedline required him to press the key if bothletters were either vowels or consonants.Although RT differs greatly in the two tasks,the time course of preparation is very similarand like that described in the studies citedabove.

ENCODING

If 5 is provided with part of the informa-tion he needs to perform a speeded task, hisRT when he receives the remainder will bereduced (Cohen, 1969; La Berge, VanGelder, & Yellott, 1970; Leonard, 1958).If the time between the two signals is varied,it is possible to discover how long it takes toencode the first signal in a form which is op-

550

S2450

aCO

350

O--OV-C MATCH

I PI MATCH

750

650 >

i

550

10 100 IOOO

WARNING INTERVAL (MILLISEC)

6000

FIG. 1. Preparation functions for physical (PI) match task (solid line,left ordinate) and vowel-consonant (V-C) match task (dotted line, rightordinate). (Data are from constant block foreperiods of each warninginterval collapsed across four days; after Posner & Wilkinson, 1969.)


timal for processing the second. If the firstsignal is a letter that must be matched againsta second letter, the improvement should rep-resent the encoding of the information fromthe first letter needed for the match. Inorder to reduce the involvement of generalpreparation in such functions, a warning sig-nal was given .5 second before the first letter.It was supposed that preparation could bemaintained during the time between presen-tation of the first letter and the presentationof the second letter.

MethodExperiments were run on a PDP-9 computer.

Displays were presented on a Tektronix 503 scopeusing a P 11 phosphor which decays to .01 ofits energy in 20 milliseconds (msec.). The warn-ing signal was a cross presented .5 second be-fore the first letter. The first letter was presenteddirectly below the warning signal and remainedon until S's response. After a variable interval(interstimulus interval [ISI]), the second letterwas presented adjacent to the first. At a distanceof 16 inches the two letters were foveal (3 degreesof visual angle). The time between the two letterswas varied between blocks of trials.

Two separate experiments were run. In the firstexperiment, S responded "same" if the letters werephysically identical; in the second, "same" was de-fined as both vowels or both consonants. Experi-ment I used only capital letters, while ExperimentII used mixed uppercase and lowercase letters.The computer chose the letters at random from analphabet of 15 letters with the restriction that halfof the trials in each block be "same." In these ex-periments, 5s had two keys and were instructed topress the left one if the letters were defined as"same" and the right one if "different."

In Experiment I, each of eight vS"s ran in six 40-trial blocks for each of four days. Each block con-tained one of six intervals between the two letters.The intervals were assigned randomly so that each 6"had all six blocks on each day. Experiment II wassubstantially the same except that each block con-tained four physical matches (e.g., AA), eightname matches (e.g., Aa), 12 vowel consonant"same" responses, and 24 "different" responses.

Results

First, it is clear from Figures 2 and 3,as found previously (Posner & Mitchell,1967), that the level of match has sys-tematic effects on the reaction time. More-over, the encoding function for all three

430

400

_Js

350

300

O---O DIFFERENTSAME

PHYSICAL MATCH

200 400 600

INTERVAL (MILLISEC)

800 1000

FIG. 2. Encoding function for physical matches. (Constant warninginterval of SOO msec., with constant blocks of various ISIs.)


TOO

600

<0_l

Ife

500

400

V-C SAME

200 400 6OO

INTERVAL (MILL!SEC)

800 1000

FIG. 3. Encoding functions for vowel-consonant (V-C) match task.(Constant warning interval of 500 msec, with constant blocks of variousI Sis. NI=name identity, PI = physical identity.)

levels is strikingly similar in both experi-ments. The optimal RT is at 500 msec, inall functions. This finding appears to berobust over the distribution of intervals,since a control study using intervals of 0, 1,SO, ISO, 250, and 500 showed almost thesame function as that obtained for the firstfive points of Figure 2. The same result isalso obtained in encoding functions discussedin the next section (see Figures 4 and 5).There is some tendency for sharpest im-provement in RT to occur later for morecomplex matches. Most of the improvementfor physical matches appears to come in thefirst 150 msec. This is much less true ofvowel-consonant matches, which show asharp improvement between 250 and 500msec. The extent of improvement appearsto be slightly greater for vowel-consonantmatches than for other types, but name

matches do not show much more improve-ment than physical matches.

In general, encoding and preparation func-tions appear to be rather similar. Both showan optimal point at about SOO msec. Bothtend to give greater improvement for vowel-consonant than for physical matches. Thenext experiment is designed to compare themmore directly.

PREPARATION AND ENCODING

If S is given 500 msec, to prepare, his RTwill be about optimal. If he is given infor-mation at the end of that 500 msec, aboutwhat he is to select, he shows a further im-provement which also goes on for about 500msec. Are these two improvements due totwo separate processes, or are they merelytwo different ways of studying the sameoverall process? A paradigm was designed


TABLE 1CONDITIONS OF WARNING AND INTERSTIMULUS

INTERVALS FOR EXPERIMENTS ON ENCODINGAND PREPARATION

Inter-

interval1*

0

150500

Warning interval"

0

PreparationBothBothBoth

150

Preparation

500

PreparationEncodingEncodingEncoding

a In milliseconds.

to compare them directly. The method usedwas to obtain preparation functions, encod-ing functions, and functions where both pro-cesses might go on simultaneously.

MethodThree experiments were run involving a com-

mon method. The apparatus and displays werethe same as discussed previously. Three con-ditions varied only the warning interval (WI) andwere used to produce a preparation function. Threeconditions involved an optimal WI (.5 second) and

varied only ISI. These produced an encodingfunction. Three conditions used zero foreperiodand varied ISI. In this case, the first letter servedas both a warning signal and as a source of spe-cific information. Two conditions enter into twofunctions, making a total of seven different combi-nations of WI and ISI as shown in Table 1.Each condition involved a block of 40 trials; 20"same" and 20 "different." Each trial was sepa-rated by a long variable intertrial interval of8-15 seconds. This was followed by a warningsignal, the first letter and the second letter. Thewarning signal remained on only during the WI,while the first and second letters remained presentuntil ^s responded.

Each experiment involved 10 6"s run in one blockof each of the seven conditions in a random orderon each of two days. The first day was consideredpractice. After each trial, -S^ were given feed-back as to the correctness of their responses andthe RTs. The three experiments differ only in thelevel of instruction. In the first experiment, "same"was defined as physical identity and all letters wereuppercase; in the second experiment, "same" re-sponses were defined as letters with the same name;while in the final experiment, "same" was definedby the rule both vowels or both consonants. In allexperiments, the "same" response was made withthe left index finger and the "different" responsewith the right index finger.

600

500

.J2

400

300

0/0

500/150

PREPARATION

5̂00/0

/ENCODING500/500

100 200 300

INTERVAL (MILLISEC)400 600

FIG. 4. Preparation, encoding, and "both" functions for physical matchingtask. (Values of WI/ISI are shown for each point.)


760

700

650

600

I

K 600

500

450

500/150

PREPARATION

500/0

BOTH

ENCODING500/500

100 200 300 400

INTERVAL (MILLISEC)600

FIG. 5. Preparation, encoding, and "both" functions for name matchingtask. (Values of WI/ISI are shown for each point.)

Results

Three functions can be drawn from theseven conditions. Each condition involvinga combination of one WI and one ISI. Apreparation function involves conditions 0/0,150/0, and 500/0 (WI/ISI). This functionis shown in the upper curve of Figures 4,5, and 6.

The encoding function is obtained fromconditions where S has always had a full 500-msec. WI and either 0-, 150-, or 500-msec.ISI. This is shown in the lower curve ofeach figure. When there is no warning sig-nal, the first letter serves both as a warningsignal and to provide selective information.The function obtained from conditions 0/0,0/150, and 0/500 is called the "both" func-tion and is shown as the middle line of eachfigure. The figures show the mean of themedian RTs for each condition.

The main effects of WI and ISI on RTare highly significant in all experiments. Inall three experiments there is an interactionbetween WI and ISI. An interaction be-

tween two independent variables has been in-terpreted (Sternberg, 1969) as indicatingthat they affect the same stage of processing.In the present case, WI and ISI may bethought to interact because they both af-fect alertness. At zero WI, alertness mustbe varying during the ISI, while with anoptimal WI, 5" may already be alert and thusISI would mainly involve encoding of in-formation from the first letter.

Table 2 shows the amount of improvement(from the 0/0 condition for preparation and"both" and from the 500/0 for encoding)due to preparation, encoding, and both to-gether. The results clearly indicate that inthe case of physical and name instructions,the improvement obtained in the "both" con-dition is almost exactly the sum of that ob-tained separately from the preparation andencoding functions. In the vowel-consonantcase, there appears to be a more substantialdeparture from additivity, but these are notsignificant due to high variability among 6"sin this condition.


900

850

800

760

700

650

600

550

500/0••

PREPARATION

500/150

ENCODING500/500

100 200 300

INTERVAL (MILLISEC)

400 500

FIG. 6. Preparation, encoding, and "both" functions for vowel-consonantmatching task. (Values of WI/ISI are shown for each point.)

The additivity of improvement due topreparation and encoding suggests that thetwo processes can proceed in parallel with-out interference. This is true even at 150msec, when the functions are changingrapidly. Moreover, the additivity indicatesthat the amount of improvement due to en-coding is the same regardless of the initiallevel of alertness.

It remains to be shown that the encodingfunction is really related to specific informa-tion from the letter presented. A strikingfinding in the physical match encoding func-tion is a significant interaction between in-terval and type of response (same vs. differ-ent) F = 3.9, df = 2/9, p < .05. Recently anumber of investigators have pointed out thatthe response "different" does not appear tobe the mirror image of the response "same"(Bamber, 1969; Nickerson, 1965; Smith &

Nielsen, 1970; Tversky, 1969). In manyexperiments where there is an interval be-tween the two items to be matched, "same"responses are faster than "different" re-sponses (Bamber, 1969; Nickerson, 1965).Figures 7 and 8 show the data from the prep-aration, encoding, and "both" functions forthe physical match. It is clear that the rela-tive efficiency in processing "same" as com-pared to "different" stimuli increases as afunction of interval, but only when 5 hasbeen given the first letter. The relative in-crease in efficiency of the "same" responseas compared to "different" shown in theRTs is also reflected in the error rates. Nosuch change is found when the interval in-volves the period between a warning signaland the letter pair. Nor does it occur forname or vowel-consonant level matches.

The finding that "same" and "different"

COMPONENTS OF ATTENTION 3Q9

responses are identical in simultaneous pre-sentation has not been universal (Krueger,1970; Posner & Mitchell, 1967). This ap-pears to depend on many factors. It wassuspected that one such factor was assign-ment of hands. To check on this, a controlstudy was run which assigned six ,9s "same"on the left and six 5"s "same" on the right.The experiment involved ISIs of 0, SO, 150,and 500 msec, with a 500-msec. WI. Thephysical match instruction was used. Thedata for left-hand "same" 5s replicated Fig-ure 7 perfectly. For Ss assigned "same"on the right, however, "same" was fastereven with simultaneous presentation. How-ever, regardless of hand assignment, the sig-nificant divergence over time between "same"and "different" responses was found (seealso Figure 2 and Corballis, Lieberman, &Bindra, 1968).

The finding that presentation of the firstletter improves the efficiency of handling anidentical letter suggests that the first letterchanged 5's sensitivity to a letter of iden-

TABLE 2AMOUNT OF IMPROVEMENT IN REACTION TIME

FROM PREPARATION, ENCODING, AND BOTH

.. , .JVL&tcn

PhysicalPreparationEncodingBoth

NamePreparationEncodingBoth

Vowel-ConsonantPreparationEncodingBoth

Interval0

ISO

5853

101

4258

111

71124157

500

12273

195

127H98201

170150296

Note.—Improvement is measured by subtracting RT atspecified interval from RT at 0/0 for preparation and "both"function and 500/0 for encoding.a In milliseconds.

tical form. One way of conceptualizing thisis in terms of Sokolov's neural model idea(Sokolov, 1963). The first letter serves as

600

800

11

I-K

400

300

PREPARATION

ENCODING

O O DIFFERENT

•—• SAME

100 200 300INTERVAL (MILLISEC)

400 900

FIG. 7. "Same" versus "different" match RTs as a function of intervalfor preparation and encoding functions of physical match task.


600

500

2w

3i-tc

400

300

O—--O DIFFERENT

• • SAME

100 200 300INTERVAL (MILLISEC)

400 SCO

FIG. 8. "Same" versus "different" match RTs as a function ofinterval for "both" functions of physical match task.

a model of what 5" is looking for. When thesecond letter matches it, processing proceedsrapidly; if it does not, further tests are made.A similar view would be that the first letteractivates internal units representing thestored information concerning that letter(Morton, 1969; Posner, 1969), which maythen be more sensitive to identical stimula-tion for some period of time. The latter ideawould presumably allow for activation of re-lated items necessary to handle improve-ments at the name and vowel-consonantlevels, since access to past experience isclearly needed for these matches. Both ofthese ideas are consistent with recent ideasabout "same" and "different" responses tosuccessive stimuli (Bamber, 1969; Tversky,1969). What is new is that the divergenceover time indicates that at least part of theadvantage of "same" is due to the encodingof the first letter rather than in the criteriongoverning the match itself.

The results of this section suggest thatencoding and preparation are separable func-tions with different effects on behavior.However, it appears possible to carry onboth simultaneously as efficiently as eitherone alone. In some ways this provides ananswer to Moray's question concerningwhether selectivity is more efficient for analerted organism than for an unalerted one.According to these data, whether S has hada previous warning signal does not changethe efficiency with which he encodes a letter,at least up to the level of a name match.Nor does the relative bias in physical match-ing for "same" as opposed to "different"differ for alerted and unalerted conditions.These data strongly suggest that a warningsignal does not improve the process of en-coding information from the first letter.

So far, we have defined two separate func-tions having to do with attention and ob-served something of their relationship. It


now remains to relate each of them to thethird component of attention.

PROCESSING CAPACITY

One sense of attention involves competi-tion between signals for some limited capac-ity system (James, 1890). The idea is thatmany tasks place demands on a centrallimited capacity system and thus tend to in-terfere. This basic idea has led to a numberof experiments which have attempted to mea-sure the attention demands of one task by itsinterference with a secondary task (Welch,1898; Welford, 1968). One way of doingthis is to use a probe as the standard second-ary task. For example, Naatanen (1970)measured the evoked potential to auditoryclicks, which were probes occurring duringa visual RT task. Posner and Keele (1967)and Ells (1969) used a simple or choice RTtask as a probe during a movement task.Foss (1969) used a probe RT to studyprocessing of sentences. The advantage ofa probe RT task is that the experimentercan measure both the RT to primary andsecondary tasks and thus be certain whateffects each is having on the other. Instudying movements, it was found that theprobe had little effect on the primary move-ment task, but that the probe RT reflectedthe central processing demands of the pri-mary task in a very sensitive way. Forexample, prior to the movement, probe timewas a function of a number of alternativemovements and was unrelated to targetwidth. During the movement, probe timewas a function of target width and unrelatedto number of alternatives (Ells, 1969).These findings suggested very strongly that

probe RTs could provide a dynamic pictureof the central processing demands of primarytasks.

The probe technique applied to the letter-matching sequence might indicate how prep-aration and encoding are related to centralprocessing capacity. Of particular interestis whether preparation for a visual signalserved to improve or retard RT to signalscoming from another modality (Hernandez-Peon, 1966; Naatanen, 1970). Anotherquestion is whether encoding itself requiresprocessing capacity in the sense of interfer-ence with a probe RT. Finally, there is thequestion of what aspects of letter matchingseem to demand the greatest involvement ofprocessing capacity.

MethodThe same apparatus was used as described pre-

viously. A trial consisted of a warning signalfollowed after a WI with the first letter, whichremained on until the presentation, after an I SIof a second letter adjacent to it. After eachtrial, feedback was provided on the correctness andRT for the letter match. The values of the WI,ISI, and intertrial interval (ITI) are given foreach experiment in Table 3. On half of the trials,SO decibels of white noise was turned on in 6"s leftear at one of eight positions in the trial. The probeoccurred equally often in each of the eight positions.The SB were instructed that the letter task wasprimary and that they were to press their rightindex finger if the letter had the same name andtheir right middle finger if the names were differ-ent. They were told to press their left index fingerwhenever they heard the white noise burst. Feed-back on the letter-matching task did not occur untilboth responses were complete. No feedback wasprovided on the noise RTs.

In Experiment I, 5"s ran six 48-trial blocks perday. Each block was identical in form. All thefirst letters were uppercase and the second letters

TABLE 3

INTERTRIAL, INTERSTIMULUS, AND WARNING INTERVALS AND FIRST-LETTEREXPOSURE DURATIONS (IN SECONDS) FOR EACH EXPERIMENT

Experi-

iii

in

ITI

8-15 variable4-6 variable4-6 variable

WI

1.5.5

Exposure duration

.51

.05 "I

son r variab'e

l.OOOJ

ISI

.511

Probe positions(milliseconds following warning signal)

1

-3000,-500,-500,

2

-2000,150,150,

3

150,550,550,

4

600,650,650,

5

1100,800,800,

6 7 8

1300, 1600, 19001000, 1650, 18001000, 1650, 1800


lowercase, so all matches were at the name level.In Experiment II, half the blocks were all upper-case (physical match) and half were identical toExperiment I (name match).

All 6"s were run individually for two days. Datafrom the second day only were used for those 5swhose median probe RTs were under 1 second forall positions. Twelve ,5s were included in theanalysis of each experiment (a total of three wereexcluded for not meeting the criterion).

Results

The basic results of each experimentare the mean of the median probe RTsat each position and the mean of the medianletter match RTs at each probe position.These are shown in Figures 9,10,11, and 12.Differences in letter match times betweenphysical and name matches are consistentwith the instructions and suggest 6"s used avisual code for the former and the letternames for the latter (Posner, 1969).

The first point of interest is that the RTto the probe is longer when it is in the ITIthan when it occurs after the warning signal.Some evidence on this point appears in Fig-ure 9 where the two probes after the warningare slightly faster than those during the ITI.However, this improvement does not reachsignificance. In Experiment II, RTs atPosition 2 are faster than at Position 1, butagain this is not significant. However, atPositions 3 and 4, each 5* shows a definiteimprovement in RTs over the times obtainedduring the ITI (p < .01, by sign test).Thus, a prepared -S" gives faster RTs to thesecondary task than an unprepared S, eventhough the 6" is instructed to prepare for thevisual signal. An analysis of the distribu-tions at each probe position revealed no evi-dence of anticipation errors (e.g., RTs to thenoise of less than 200 msec.). Every effort

600

500

i400

300

WARNINGFIRST SECOND

LETTER LETTER

-4 -2 .5TIME (SEC)

I 1.5

FIG. 9. Probe RTs as a function of probe position for Experiment I.


BOO

400

e<0ai

300

WARNING FIRST SECONDI I

NONE -4 -2 0 .9TIME (SEC)

1.5

FIG. 10. Letter match RTs as a function of probe position for Experiment I.

600

900

Het

400

300

P.I.

FIRSTWARNING LETTER

SECONDLETTER

-I I 1.50 .5

TIME (SEC)

FIG. 11. Probe RTs as a function of probe position for Experiment II.

2


400

COJ

1300

O--

WARNINGI

FIRST LETTERSECOND LETTER

I

NONE -I 0 .5

TIME (SEC)1.5

FIG. 12. Letter match RTs as a function of probe position forphysical (PI) and name (NI) matches of Experiment II.

was made to insure that the preparation of.S"s was specific to the visual letter-matchingtask. The auditory probe occurred on only.5 of the trials and were distributed overeight different positions. All feedback in-volved the times for the letter task and nonereferred to times on the probe task.

A second point of interest is that probetimes do not necessarily increase as a con-sequence of presenting the first letter. Whenthe first letter was on for only .5 second(Figure 9), RT to probes after the first let-ter (Probe Positions 5 and 6) was increasedsharply. However, when the first letter waspresent for 1 second (Figure 11), RT to theprobe did not begin to increase until 300-500msec, after the first letter was presented(Probe Positions 5 and 6). From previousresults (see Figures 2 and 3), it would seemlikely that the encoding of the first letter waspretty much complete by the time the probebegan to show interference.

Third, it is clear that probe times aresystematically increased at certain points.Probes which occur shortly after the secondletter are always long (Figure 9, Probe Posi-tion 7, and Figure 11, Probe Positions 7 and8). These times decrease quite consistentlyfrom 100-400 msec, after the second letter.

This result suggests that 5"s have troubleprocessing the probe during the responsephase of the letter-match task. If probes aredelayed until close to the letter-match RT,they drop back to nearly the same values asobtained during the ITI (Figure 9, ProbePosition 8). Probes which occur betweenthe two letters are increased if they fall closeto the occurrence of the second letter (seeFigure 11, Probe Position 6). This increasedoes not appear to be due to the actual ap-pearance of the second letter. When probedistributions were computed for probes 500msec, before the second letter in Experiment11 (Figure 11, Probe Position 6), it wasfound that few of the RTs were long enoughto occur after the second letter. Even withthese excluded from the analysis, RTs toprobes in this position still remained longerthan those during the ITI and WI for all12 5s.

Do the probe RTs really reflect the pro-cessing requirements of the letter-matchingtask? With the long I SI it might be possi-ble for 5"s to switch from processing the let-ter to listening for the probes. In order tocheck on this, an experiment was run inwhich the first letter was turned off ran-domly after 50, 150, 500, or 1000 msec. The


500

g400

<2

.1

300

•—•50

X—X 500

WARNING FIRSTI

SECONDI

.5

TIME (SEC)1.5

FIG. 13. Probe RTs compared for exposure durations of thefirst letters of SO and 500 msec.

experiment was otherwise identical to thename letter-matching condition of Experi-ment II. Figure 13 shows the results forseven "̂s who ran in this experiment for fivedays. The data represent probe times onDays 4 and 5 for trials which had only 50-msec. exposure of the first letter and thosefor which the first letter was on for 500msec. The RT results together with gen-ally very low error rate indicates that 5"s arenot switching away from the letter task. Ifthey were, they would miss many of the50-msec. exposures. Moreover, the probecurves from this study are quite similar tothose illustrated in the previous studies.Even if the first letter was turned off after50 msec., 5" does not show substantial inter-ference with the probe until 500 msec.

Another possibility is that ,9s are learningthe distribution of the probes and respondmost rapidly to those in the middle of thedistribution. This possibility seems ratherremote—in part because of the low frequencyof probes at any position (only on 1/16 ofthe trials), but also because of the complexshape of the RTs as a function of probe posi-tion. Probe Positions 4 and 5 are the two

mid positions and should be fastest based onan expectancy idea. These show very differ-ent results in both Figure 9 and in the namematch condition of Figure 11. Moreover,the probes at Position 8 are quite fast al-though they are at the end of the distribu-tion. In addition, the results shown in Fig-ure 13 illustrate that the point at which thefirst letter is turned off has small but sys-tematic effects on the probe RTs in the di-rection which would be expected if 5"s wereshowing interference from letters which wereremoved somewhat more rapidly than thosewhich remained present. Thus, it appearsthat probe RT reflects the attention demandsof the primary task.

CONCLUSIONSThe introduction outlined three component

processes involved in the study of attention.These are alertness, selectivity, and process-ing capacity. In this paper each of thesecomponents was identified with a particu-lar experimental operation. Alertness wasstudied by varying the time between a warn-ing signal and a pair of letters which 5" wasrequired to match. Since the warning signal


told -S nothing about what letters he was toreceive, it provided only rather minimal se-lective information. Selectivity was studiedby providing one of the two letters so that 6"knew what he was to look for. Improve-ments in RT to the second letter provide afunction related to encoding of the first let-ter. Both alertness and selectivity werefound to improve performance in the letter-matching task. In both cases, the time re-quired to reach optimal performance wasabout 500 msec. This value appeared to besurprisingly consistent both between the twoprocesses and over various levels of process-ing (physical, name, vowel). The encodingfunction appeared to be somewhat steeperover the first 150 msec, than the preparationfunction. As the level of encoding requiredby the match increased, the steepness of thefunction appeared to be somewhat reduced.

The main consistent difference betweenthe preparation and encoding functions wasin the rate of responding to identical versusdifferent letter pairs. The preparation in-terval improved physical matches and differ-ent responses equally. The encoding of oneletter appeared to improve processing ofphysical matches more than of mismatches.This divergence between "same" and "differ-ent" responses was particularly sharp in thefirst 150 msec. For higher levels of instruc-tions, no such divergence was shown.

The preparation and encoding functionsare additive at least for physical and namematches. This is true even after only a 150-msec. interval when both are improvingrapidly. The additivity between the twofunctions suggest that encoding may proceedwith nearly equal efficiency regardless of theinitial level of alertness. The bias intro-duced from the first letter appears to developjust as rapidly whether £ has been preparedby an earlier warning signal or not. Thisprovides one kind of answer to Moray'squestion about the relationship of selectionand alertness. The additivity also impliesthat the two operations can be carried onsimultaneously without interference. If, forexample, 5" must both extract the name of aletter for later matching and use it as awarning signal, he may perform both func-

tions with as great efficiency as either onealone.

This conclusion is similar to that reachedby Posner and Wilkinson (1969), in whichpreparation, as measured by a negative shiftin the electroencephalogram (contingent neg-ative variation), was unaffected by makingthe warning signal a letter which had to beclassified. The correspondence between thefindings of the present study and previouswork suggests a relationship between alert-ness, as used in this paper, and the contingentnegative variation. If this relationship iscorrect, then alertness might be thought ofas a general and widespread increase in cor-tical activation. Results of studies with thecontingent negative variation support thatview and suggest that this improvement is notconfined merely to tasks in which there is anovert motor response (Karlin, 1970). Thus,the improvement does not appear to be con-fined to the motor system. How does alert-ness improve task efficiency ? The failure ofthe initial level of alertness to vary the ef-ficiency of encoding indicates that alertnessdoes not influence the earliest stages of pro-cessing. However, it is possible that withour method, the parallel development ofalertness and encoding obscures the impor-tance of the initial state of alertness so thatit does not affect the encoding function.Perhaps a more interesting possibility is thatalertness varies the time when the resultsof the encoding process begin to occupy thelimited capacity system. This would suggestthe early stages of processing are unaffectedby level of alertness. Studies of the rela-tion of early components of the evoked po-tential to alertness (Karlin, 1970; Naatanen,1970) are not conclusive on this point.

If preparation is viewed as a general pro-cess which is not selective, there must beother mechanisms which account for the ob-vious ability of 5"s to select information.The first letter appears to have a very spe-cific effect on selection of a new letter whichresembles it. If the first letter is viewed asactivating a trace system related to the pro-cess of recognition (Morton, 1969; Posner,1969), this activation may serve to makeselection of already active areas of memory


more efficient. Thus, in the "both" condi-tion, a single letter would lead to a generalalertness and a specific activation, which con-tribute additively to the overall reduction inRT. Of course, the activation of such areasof memory could come either from internalor external sources.

The nearly perfect time sharing betweenpreparation and encoding suggests that atleast one of these mental operations does notrequire central processing capacity. If theyboth did, they would be expected to interfere.In order to further test whether alertnessand encoding were free of demands on pro-cessing capacity, the attention demands ofthe letter-matching tasks were measured byintroduction of a simple probe RT taskto white noise. The results suggest thata warning signal serves to reduce RTboth to the main task and to the probetask, even when probes were introduced ononly half the trials. Moreover, the encodingof the first letter did not seem to interferewith the probe task, provided 5s had onesecond before the second letter was intro-duced. This was the case even when thefirst letter was turned off so that 5" had toencode quickly. This suggests that encodingof a letter does not require processing ca-pacity as we have defined it here. Of coursethe reasonableness of the conclusion dependson the assumption that probe RT will besensitive to processing demands of the pri-mary task.

Interference with the probe was significantwhen it was introduced 500 msec, prior tothe second letter. The interference was notdue to the physical occurrence of the secondletter, since the effect was still present whenall RTs above 500 msec, were eliminated.What aspect of the letter matching causesinterference with the late probes ? One pos-sible explanation is that such mental opera-tions as rehearsal or generation of distinctivefeatures for testing, which follow encodingof the first letter, require processing capacity,while encoding itself does not.

These results suggest that attention in thesense of central processing capacity is relatedto mental operations of which we are con-scious, such as rehearsing or choosing a re-

sponse, but is not related to the contact be-tween the input and long-term memory thatleads to the letter name. Several lines ofevidence have tended to relate single-channeleffects to response selection rather than allforms of stimulus processing (Herman &Kantowitz, 1970; Keele, 1970). Moreover,other research has indicated that consciousawareness is itself rather late in the sequenceof mental processing (Fehrer & Raab, 1962;Libet, 1966). If single-channel effects weretied to conscious processing, one would notexpect to see them closely correlated with en-coding of the first letter (Deutsch & Deutsch,1962; Norman, 1968), but rather with opera-tions on the retrieved product of such encod-ing. The data obtained in this study appearto be consistent with such a view.

REFERENCES

BAMBER, D. Reaction times and error rates for"same" and "different" judgments of multi-dimensional stimuli. Perception and Psycho-physics, 1969, 6, 169-174.

BERTELSON, P. The time course of preparation.Quarterly Journal of Experimental Psychology,1967, 19, 272-279.

BOIES, S. Retention of visual information from asingle letter. Unpublished master's thesis, Uni-versity of Oregon, 1969.

BROADBENT, D. E. Perception and communication.London: Pergamon Press, 19S8.

COHEN, G. Pattern recognition: Differences be-tween matching patterns and matching descrip-tions to patterns. Journal of Experimental Psy-chology, 1969, 82, 427-434.

CORBALLIS, M. C., LlEBERMAN, W., & BlNDRA, D.Discriminability and central intermittency insame-different judgments. Quarterly Journal ofExperimental Psychology, 1968, 20, 51-61.

DEUTSCH, J. A., & DEUTSCH, D. Attention: Sometheoretical considerations. Psychological Review,1963, 70, 80-90.

EGAN, J. P., GREENBERG, G. E., & SCHULMAN, A. J.Interval of time uncertainty in auditory detection.Journal of the Acoustical Society of America,1961, 35, 771-728.

EGETH, H. E. Selective attention. PsychologicalBulletin, 1967, 67, 41-56.

ELLS, J. Attention requirements of movement con-trol. Unpublished doctoral dissertation, Uni-versity of Oregon, 1969.

FEHHER, E., & RAAB, D. Reaction time to stimulimasked by meta contrast. Journal of Experi-mental Psychology, 1962, 63, 143-147.

Foss, D. J. Decision processes during sentencecomprehension: Effects of lexical item difficulty


and position upon decision time. Journal ofVerbal Learning and Verbal Behavior, 1969, 8,457-462.

HERMAN, L. M., & KANTOWITZ, B. H. The psy-chological refractory period: Only half the doublestimulation story. Psychological Bulletin, 1970,73, 74-88.

HERNANDEZ-PEON, R. Physiological mechanismsin attention. In R. W. Russell (Ed.), Frontiersin physiological psychology. New York: Aca-demic Press, 1966.

JAMES, W. Principles of psychology. Vol. 1.New York: Holt, 1890.

KARLIN, L. Cognition, preparation, and sensory-evoked potentials. Psychological Bulletin, 1970,73, 122-136.

KEELE, S. W. The effects of input and outputmodes on decision time. Journal of ExperimentalPsychology, 1970, 85, 157-164.

KOSTER, W. G. (Ed.) Attention and performanceII. Amsterdam: North-Holland Press, 1969.

KRUEGER, L. E. Effect of bracketing lines on thespeed of "same"—"different" judgment of two ad-jacent letters. Journal of Experimental Psy-chology, 1970, 84, 324-330.

LA BERGE, D., VAN GELDER, P., & YELLOT, J. Acueing technique in choice reaction time. Percep-tion and Psychophysics, 1970, 7, 57-62.

LEAVITT, F. Accuracy of report and central readi-ness. Journal of Experimental Psychology, 1969,81, 542-546.

LEONARD, J. A. Partial advance information in achoice reaction task. British Journal of Psy-chology, 1958, 49, 89-96.

LIBET, B. Brain stimulation and the threshold ofconscious experience. In J. T. Eccles (Ed.),Brain and conscious experience. New York:Springer-Verlag, 1966.

MACKWORTH, J. A. Vigilance and habitation.London: Penguin Press, 1970.

MORAY, N. Listening and attention. London:Penguin Press, 1970.

MORTON, J. Interaction of information in wordrecognition. Psychological Review, 1969, 76,165-178.

NAATANEN, R. Evoked potential, EEG, and slowpotential correlates of selective attention. InA. F. Sanders (Ed.), Attention and performanceIII. Amsterdam: North-Holland Press, 1970.

NICKERSON, R. S. Response times for same-differ-ent judgments. Perceptual and Motor Skills,1965, 20, 15-18.

NORMAN, D, A. Toward a theory of memory and

attention. Psychological Review, 1968, 75, 522-536.

POSNER, M. I. Abstraction and the process ofrecognition. In G. Bower & J. T. Spence (Eds.),Psychology of learning and motivation. Vol. 3.New York: Academic Press, 1969.

POSNER, M. L, & KEELE, S. W. Attention demandsof movement. In, Proceedings of the 16th Inter-national Congress of Applied Psychology. Am-sterdam : Swets & Zeitlinger, 1967.

POSNER, M. I., BOIES, S. J., EICHELMAN, W. H., &TAYLOR, R. L. Retention of visual and namecodes of single letters. Journal of ExperimentalPsychology, 1969, 79(1, Pt. 2).

POSNER, M. I., & MITCHELL, R. F. Chronometricanalysis of classification. Psychological Review,1967, 74, 392-409.

POSNER, M. I., & WILKINSON, R. T. On the pro-cess of preparation. Paper presented at the meet-ing of the Psychonomics Society, November1969.

REYNOLDS, D. Effects of double stimulation:Temporal inhibition of response. PsychologicalBulletin, 1964, 62, 333-347.

SANDERS, A. F. (Ed.) Attention and performanceI. Amsterdam: North-Holland Press, 1967.

SANDERS, A. F. (Ed.) Attention and performanceIII. Amsterdam: North-Holland Press, 1970.

SMITH, E. E., & NIELSEN, G. D. Representationsand retrieval processes in short-term memory:Recognition and recall of faces. Journal of Ex-perimental Psychology, 1970, 85, 397-405.

SOKOLOV, Y. N. Perception and the conditioned re-flex. London: Pergamon Press, 1963.

STERNBERG, S. The discovery of processing stages:Extension of Bonders method. In W. G. Koster(Ed.), Attention and performance II. Amster-dam: North-Holland Press, 1969.

TRIESMAN, A. Strategies and models of selectiveattention. Psychological Review, 1969, 76, 282-299.

TVERSKY, B. Pictorial and verbal encoding in ashort-term memory task. Perception and Psy-chophysics, 1969, 6, 225-233.

WELCH, J. On the measurement of mental activitythrough muscular activity and the determinationof a constant of attention. American Journal ofPhysiology, 1898, 1, 288-306.

WELFORD, A. T. Fundamentals of skill. London:Methuen, 1968.

WILKINSON, R. T., & HAINES, E. Evoked responsecorrelates of expectancy during vigilance. ActaPsychologica, 1970, 33, 402-413.

(Received November 4, 1970)

Documents

POSNER, M. I.; BOIES, S. J., Componentes da Atenção. 1971