0L A

1EPORT DOCUMENTATION PAGE

A - A 197i 088 1 b. RESTRICTIVE MARKINGS

3 DISTRIBUTION/ AVAILABILITY OF REPORT_______________________________ APPROME FOR PUJBLIC RELEASE:

Zb EC -ASSF'CA71ON. DOWNGRADING SCyIEDULE DISTRIBUTION UNLRIIED.

1 ;ER;ORMING3 ORGANIZATION REPORT NUMBER(S) S. MONITORING ORGANIZATION REPORT NUMBER(S

ONR TECHNICAL REPORT # 88-26a. XAME OF PERFORMING ORGANIZATION 6b OFFICE SYMBOL 7a. NAME OF MONITORING ORGANIZATIONI ifippicabilo)

W(ASINGTON UNIVERSITYl OFFICE OF NAVAL RESEARCHi (Q)DE 1142PT)

,~ , c- ADDRESS Cjry, State, and ZIP Code) 7b. ADDRESS (City, State. and ZIP Code)660 S. Euclid, Box 8111, NEUROLOGY DET., PERSONNEL & TRAINING RESEARCH PROGRAMSST. LXXIS, MD) 63110 800 NOKITH QUINCY STREET

* _______________________ ___________ ARLINGTON. VA 22217-5000Sa. 'lAME OF ;uNDiNG. SPONSORING Sb. OFFICE SYMBOL 9 PROCUREMENT INS7RUMENr IDENTIFICATION NUMBER

ORGANIZATiON j (if appicable) N00014-86-0289

Sc. ADDRESS(Ciy, Stare, and ZIP Code) 10. SOURCE OF FUNDING NUMBERS

PROGRAM PROJECT T ASK -vo RK UNITELEMENT NO NO.20 NO ~ ACCESSION NO

I TILE Indfuce Security Classification)

Relating Sensitivity and Criterion Effects to the Internal Mechanisms of VisualS tial Attention

12 -DERSONAL AuTisOR(S)

Gordon I.- Shiilmon and MWu-hata T -PQ-nplr

3a, TYPE OF REPORT 1i3b. rIME COVERED I14 DATE OF REPORT (Year, MfonthI Day) 115 PAGE* COUNT

N;TECHNICAL IFRomOIMAY88 TCOIMY89 APRIL 30, 1988& 146- SwPPLEMENTARY NO0TA71ON

COSA"I CODES I8 SUBjECT TERMS (Continue on reverse it necessary and identify by block number)E!D CROUP I SUB-C*ROUP

TO 10

.d 9 ABSTRACT Cantinuo an~ reverse if necessary and crentify by block number)

SJUL 221988

- P.3. C . E 71 F A S RAC-2 BS. SC Q..

-,NC -ASFEO'.NL,MrI-;: a SAME AS RP- r- D71C UIERS UNCLASSIFIED3;'EC ESC~i'. MDV:DUAL 22b 7EL.-'HONE (include Area Coc) '-c . C- S f-M30L

MIHELI OSE 314) 362-3317 ONR 1142PT0 0 FO RM 74 73 .3z v.IA ea *,On "'ay :e 4sed ,,rn emira..sea SECURi7Y C'.AS-F CA71 OF 'I-S PAGEJ* .'-,All other ecit,ons are obsolete

0

5...~ % %,.~ ..

, . .RELATING SENSITIVITY AND CRITERION EFFECTS TO THE- ' INTERNAL MECHANISMS OF VISUAL SPATIAL ATTENTION

~Gordon L. Shulman and Michael I. Posner

'Ni

-. 0ONR Technical Report 88-2

" Research sponsored by: Personnel and Training Research Program

Psychological Science Division,Office of Naval Research

Under Control Number: N00014-86-K-0289Contract Authority Number: NR-442a554

: Reproduction in whole-.or part is permitted for any' purpose of the United States Government

N-',

°

% %%

NJZ

N.

4lk

iON

Relating Sensitivity and Criterion Effects to theInternal Mechanisms of Visual Spatial Attention

1

Gordon L. Shulman and Michael I. Posner

.A recent paper by Muller and Findlay (1987) raises the important issueof how to relate the parameters d' and beta to the iniernal mechanisms thatprocess visual stimuli. In'this commentary'weconsider-'the widely heldview that d' changes reflect a variety of mechanisms leading to perception,but that beta changes reflect a single high level decision mechanism thatis postperceptual and under conscious control.' We will argue that-'in acomplex highly parallel, multi-level system, both sensitivity and criterionshifts may influence perception in lawful ways - neither being necessarilymore basic and important. °Later iIthe paper, we will alsoraisesomemethodological considerations that qualify Muller and Findlay's results.

We do not argue that Muller and Findlay's conclusion thato probabilitymanipulations produce beta shifts in detection tasks and d' shifts inidentification tasks is necessarily wrong.- We do question, however, theimplication Muller and Findlay along with others often draw from this kindof result - that detection tasks involve 'radically' different selectionmechanism., tnan identification tasks. In the following discussion, theterms d' and beta will refer to the quantities one computes from data

-. . collected in an experiment; the term criterion or signal-noise ratio willrefer to the theoretical variables that may underly changes in thosemeasured quantities.

The Standard Interpretation of Bela Shifts

What mechanisms produce a beta shift? The usual answer is that shiftsin beta reflect the operation of a conscious high level decision mechanismunder the observer's control. It is a mechanism that operates fairly latein processing after stimuli have been encoded. The same mechanism is

. assumed to operate in all detection tasks, whether one is detecting tones,points of light or tumors. The mechanism uses in a rough sense the rulesof statistical decision theory to set criteria, which are thereforeinfluenced by probability manipulations and changes in payoffs.Experiments have confirmed that changes in signal probability oL payoffsproduce changes in beta.

According to this view, beta shifts in location cueing experiments* result from something like the following process. If subjects are cued

that a stimulus at a particular location is unlikely, then a percept atthat location is disregarded, consciously rejected. Although signaldetection theory only speaks of signal and noise levels and criterionvalues (it is not necessarily inteded as a process model) thisinterpretation of the mechanism producing beta shifts is widespread. It

* doubtless results from the fact that instructing subjects to change theirllingness to report a signal produces measured shifts in bota. Thp

-. assumption seEms to be that since beta shifts in this case are caused by a

- 2

9,, L,-

particular mechanism, any shift in beta must be caused by the samemechanism.

Empirical Problems for the Standard Interpretation of Beta Shifts

Psychologists often ascribe d' shifts produced by different variablesto different mechanisms (Kahneman, 1973). For example, luminance masksinterfere with target identification only when target and mask arepresented to the same eye, while pattern masks are effective underdichoptic presentation (Turvey, 1973). These results have suggested adistinction between central and peripheral masking. Changes in stimulusduration or other temporal variables produce different effects on detection

C depending upon the spatial frequev,:y of the target stimulus (Tolhurst,1975a, 1975b; Legge, 1978). These latter effects are attributed to thedifferent spatial and temporal properties of sustained and transientmechanisms.

The recognition that changes in d' may be caused by differentmechanisms is fortunate. Given that there are two basic parameters insignal detection theory, assigning each parameter to a particular mechanismwould produce limited two box theories of a system which is a much morecomplicated multi-level affair. However, our interpretations of beta aremore restricted (but see Kahneman, 1973). This may partly result from

* parsimony. If only a single mechanism is required to account for betachanges, there is no reason to postulate multiple mechanisms. We willargue below that the literature on spatial attention cannot be explained interms of the decision mechanism commonly assumed to underlie beta shifts;i.e. the mechanism that produces beta shifts when subjects are instructed

- -to change their willingness to report a signal.

detection, but as Shaw (1984) and Duncan (1980) have noted, the samedecision framework is applicable. When the decision criterion is loweredfor a location, it takes less time for a stimulus to exceed that criterion,resulting in a faster response.

A number of observations from location cueing experiments indicateconstraints on performance that are inconsistent with the operation of thestandard decision mechanism.

1. The performance deficit for noncued locations is greater if those• locations cross either the vertical or horizontal meridians relative to ,

the cued location (Hughes & Zimba, 1987; Rizzolatti, Riggio, Descola & *, ..Umilta, 1987). Further, deficits for uncued locations are a function

4. of the distance/number of intervening positions of the uncued locationfrom the cued location (Downinr & Pinker, 1985; Shulman, Wilson &Sheehy, 1985; Shulman, Sheehy & Wilson, 1986; Rizzolatti, et. al.,1987; Hughes & Zimba, 1987, argue that this effect depends upon the useof an articulated visual field). Distance effects have also beenreported for the effects of distractors in identification tasks

S3%

(Eriksen & St. James, 1986) andl for probe tasks in which the targetoccurs at different distances from a second stimulus that the subjectis processing or has just processed (LaBerge, 1983; Sagi & Julesz,1986)). Moreover, the slope of the distance function decreases withthe eccentricity of the cue. This latter result has also been foundusing both probability manipulations (Downing & Pinker, 1985; Shulman,Wilson & Sheehy, 1985; Shulman, Sheehy & Wilson, 1986) and probemethods (Sagi & Julesz, 1986). Why should a decision maker raisecriteria for noncued locations to a degree dependent on the distancefrom the cued location and why should these changes depend oneccentricity? Why should a decision maker care whether the uncuedstimulus crosses the vertical oz horizontal meridians?

2. In some situations, probability effects are weaker if the samelocation is made highly probable over a long block of trials, ratherthan being cued on each trial 'Posner, et. al., 1980). Blockedconditions would seem ideal fof the standard decision criterionmechanism.

3. Without any manipulation of probability similar changes in

performance can be obtained in both detection and identification tasks

by presenting a peripheral stimulus near the target (Jonides, 1976,1980; Maylor, 1985), or by having a target occur near a task beingperformed by the subject (Laberge, 1983; Hoffman, Nelson & Houck, 1983;Sagi and Julesz, 1986). If a peripheral stimulus is a target for aneye movement, facilitation is also found at the target location priorto the actual movement (Posner, 1980; Remington, 1980; Shepard, Findlay& Hockey, 1985).

It is not known if these uon-probabilistic methods produce beta ord' shifts during detection tas's. But suppose they produce betashifts. Why would the decision maker be forced to attend to thelocation of a projected eye movement even when a foveal target is givena higher probability? This kind of constraint is clearly outside ofthe characteristics usually given to an ideal observer. However, it isconsistent with the operation of a spatial selection mechanism.

Nor does it seem reasonable to take the view that probabilitymanipulations involve the standard decision process while peripheralcues and eye movements affect performance via some other mechanism. Ifa peripheral cue is presented or an eye movement is prepared to alocation with a low target probability, that location will initially befacilitated in comparison with higher probability locations.Probabilistic and non-probabi] istic manipulations direct theorientation of a single select on mechanism (Posner, 1980; Posner &Rafal, 1982; Shepard, Findlay , Hockey, 1985; For a possibledissociation of these two manipulations, see Briand & Klein, 1987).

44. Patients with lesions of the parietal lobe show characteristicperformance shifts in reporting targets in the field contralesional to

4

4

4

.~ %. ,

the cued location (Baynes, Holtzman & Volpe, 1986; Ladavas, 1987;Morrow & Ratcliff, 1987, Posn-i, et. al., 1984). These impairments arenot due to a failure to understand the task or to a ploblem in sensoryinput or motor output (Posner, et al, 1984). Subjects show they canrespond to probability manipulations when targets are in theipsilesional field. To the extent that these cueing deficits representa malfunction of a decision mechanism, that mechanism must be fieldspecific. Moreover, the effects of lesions can be specific tomodalities. De Renzi, Gentilini and Pattaciri (1984) have shownstatistical independence between visual and auditory deficits (see also

Sieroff & Michel, in press), indicating that the decision mechanism ismodality specific.

These effects occur in luminance detection (Posner, et. al, 1984),visual search (Friedrich, Walker & Posner, 1980) and identificationtasks (Riddoch & Humphreys, 1987). There is no evidence that patientsshow a basic difference between detection and identification.

In (1) and (2) criteria shifts resulting from cueing manipulations arebased on rules that are not derived from the theory of the ideal detector.The non-probability manipulations of (3) also cannot be explained through

standard decision rules, and as noted, they probably affect the samemechanism as probability cues. One could say that the mechanism of thestandard theory incorporates rules;in addition to those originallyenvisioned, but we are then left without a theory explaining why the systemfollows these extra rules. Moreover, the field and modality specificity ofthe lesion results (4) suggest a much different mechanism than that thoughtto be responsible for beta shifts.

Alternative Mechanisms for Producing Beta Shifts

The constraints outlined abov seem more consistent with a selectiveattention system than the standard decision mechanism. While we do notintend to develop a theory of selective attention, consider a simpleselection mechanism that governs whether information is transmitted fromone process to another; for example in detection tasks, from sensorypathways to those involved in making decisions and responses. Selectionmight be accomplished by only passing activity that exceeds a certaincriterion value. When one cues a -ertain region that value is lowered,increasing the likelihood that noise generated activity and signalgenerated activity will be passed to other systems. If these later systems

* do not change signal/noise ratios, a beta shift will result. Alternately,a colleague has suggested that overall activity is boosted at cuedlocations. Since this increase will occur for both signal and noise, the

N effect is again equivalent to shifting a criterion.

Kahneman (1973) has noted that a criterion at one level of the system* essentially controls what categorizations are made at that level and

therefore, what information is pas ,ed to the next. In the standard signaldetection model, the criterion controls the categories 'yes' and 'no' and

5

0

It-

0

therefore the information sent to an overt response stage. In the present-model, a criterion controls a selection process that determines whether

S.. information from different spatial regions is input to other processingstages. This criterion shift may have powerful perceptual consequencessince a categorization at one stage can affect processing at a subsequentstage. For example, Kahneman notes that the rectangular perception of theAmes room results from a criterion bias since a trapezoidal interpretationis equally consistent with the sensory input. Yet the categorization ordescription of the Ames room as rectangular produces a very powerful effecton size perception. Effects of expectations or top-down processing onperception can be treated similarly (Broadbent, 1971). Since a criterioncontrols the flow of information from one process to another, the functionor effect of the criterion shift vill depend on the nature of theseprocesses.

Spatial selection, for example, may serve a number of functions orproduce a variety of effects depending on the task. Selection maydetermine which stimuli engage an eye movement control system. Severalpapers have shown that a preparatory shift of attention precedes an eyemovement to a peripheral stimulus (Remington, 1980; Shepard, Findlay &Hockey, 1986). If the task involves identification or conjunctive search,spatial selection may enable analy:!ers limited in capacity to process therelevant stimulus without suffering interference. Other authors haveproposed additional reasons for spatial selection (Ullman, 1984; Navon,1985).

Reconsider Muller and Findlay' argument that a radically differentmechanism operates in detection anJ identification tasks. The abovediscussion suggests the alternative possibility that the selectionmechanism is the same in all of these paradigms. The task dependency wouldreflect how the stimuli are processed once they are selected. To theextent that selection controls access to mechanism that change signal/noiseratios, selection will produce d' Thanges. In the case of detection,signal/noise ratios are determined early in the system, prior to selection.Selection therefore produces a beta shift. In the case of identification,many processes will determine the final signal/noise ratio; some of therelevant processes may well occur subsequent to selection. Selectiontherefore produces a d' shift.

When subjects are cued that a stimulus will occur at a location,introspectively, one has the strong impression of attending to that

- location. This impression does no, depend on what stimuli are subsequentlypresented, i.e. whether it is a siigle stimulus in an empty field(detection), or four stimuli (identification). Although completelydifferent mechanisms may underlie -he performance consequences of using thecue, it seems likely that the same mechanism is involved.

6%%.N.

Methodological issues in Muller & Findlay's Experiment

Aside from our concern over th? interpretation of Muller and Findlay'sresults, we believe there may be problems with their methodology. Mullerand Findlay make two specific clains in their article.

1. Probability manipulations produce beta shifts in detection tasksand d' shifts in identification tasks. Muller and Findlay suggest that theselection process in detection tasks is radically different than inidentification tasks. In identification tasks, the probabilitymanipulation directs a selection process that determines which inputs areanalyzed by limited capacity form inalyzers. In detection tasks, theprobability manipulation does not 'iffect this selection mechanism, or if itdoes, the selection mechanism has no effect on performance (either d' orbeta). Observers use the location probabilities to affect a quitedifferent mechanism, a decision process that differentially weightsevidence from different regions.

Muller and Findlay's conclusion concerning the effect of locationprobability manipulations on d' and l beta during detection tasks rests on aparticular procedure for assigning false alarms to the different validityconditions. When a subject makes a false alarm it is not possible toassign that observation to a particular location since no signal was in

0 fact presented. Muller and Findlay adopt Bashinski and Bachrach's methodof asking the subject to make a loriation response and using that responseto assign false alarms to valid and invalid conditions. Subjects in Mullerand Findlay's experiment make three responses on each trial. First, theygive a yes-no response, then a conidence rating of the response, andfinally, a location response. Mul'.er and Findlay suggest that theseresponses are determined as folloxrq. On any trial, strength values for thefour possible locations are sampled. These samples are then weightedaccording to the apriori probabilities assigned to their location. If thesum of the weighted samples exceed, a criterion, a detection response isgiven. Different degrees of confijence essentially correspond to differentdegrees of (weighted) strength with respect to the yes-no criterion(equivalently, criterion are set up for each confidence level). Thelocation response is determined b, the location of the weighted samplegiving the largest strength response.

This account makes a number of assumptions. During the detectionprocess, the observer must know th, location of each sample so that its

0 strength can be appropriately wei.iited. This knowledge must then bereflected in the location response so that the latter is an accuratemeasure of the use of location in )rmation and location probabilities in

" the detection process.

Suppose there was some uncertainty during the detection stage• concerning the location of the sam:)le being weighted. For weak signals,

subjects may know that something was presented, but may not know where, adissociation that could depend upoi signal strength. Uncertainty

.4."

'Vo-

w.

0

% 1concerning sample location will de':rease the effect of the probabilitymanipulation on detection; valid samples will receive their strongerweighting on fewer trials. The same uncertainty, however, will increase

the effect of the probability manipulation on the location response. Withmore uncertainty concerning the actual location of the strongest sample,subjects will rely more on apriori probabilities. Imperfect knowledge oflocation will therefore increase the estimated false alarm rate for thevalid condition.

Although imperfect knowledge of location could arise during the initialperceptual stage, later processes night also contribute. Since thelocation response is the third rendonse the subject makes, the subject mayforget where the stimulus was presented, particularly when the initialsignal is weak and ambiguous. Any loss of location information between thedetection and location response will again increase the effect of the

% apriori location probabilities or the location response.

This analysis suggests that an estimate of the false alarm rate basedon the location response will inflate the 'true' false alarm rate for validtrials. In the luminance detection condition of Experiment 2 of Muller andFindlay, the false alarm rate was 7.2% for valid trials, 4.8% for invalid.This small increase might well be accounted for by the factors suggestedabove.

Given the complexities and assumptions involved in the method of Mullerand Findlay, the issue of d' shift7 and spatial attention is perhaps bestaddressed using 'criterion free' methods such as two interval forcedchoice.

-. We therefore do not believe that Muller and Findlay have convincinglydemonstrated that cueing produces d' shifts in identification tasks andbeta shifts in detection tasks. However, even if this result wereestablished, as noted above, we question their interpretation of theunderlying mechanisms.

2. Spatial attention can be divided between different locationscontrary to the conclusion of Posner, Snyder & Davidson (1980).

We do not see how Findlay and Muller's experiment contradicts theresults of Posner et. al. The latter authors were concerned with whetherattention could be split to disparite locations. They used a linear

* display of four locations, and sho,,ed that when a location was cued withhigh validity, a secondary cue with lower validity affected cue performanceif it was directed to an adjacent location but not if it was directed to alocation that was separated by an intervening location from the primary cue

p. location.

0 The data of Posner, et. al. show that attention can be spread tomultiple adjacent locations but not to locations separated by noncuedpositions. Kiefer and Siple (1987, in press) has recently replicated both

.'Vo

8

6 % %

results using a trial by trial technique in which two equally probable% locations were used. and other authors using different methods have%confirmed that attention can be spread over regions of variable size

(LaBerge, 1983; Eriksen & St. Jamep, 1986).

Muller and Findlay used a display in which targets could appear at oneof four points on an imaginary square. Subjects were never required toattend to regions separated by non-_ued regions. Trials in which twolocations were cued can be grouped into three categories 1) both cuedpositions in the left and right field, 2) both in the upper or lower field,and 3) both along the diagonal of the square. In the first two cases,subjects could use the cue by spreiding attention to adjacent regions. Thethird case is somewhat ambiguous but does not require the spatialdistribution of attention that Po;ner et. al. studied. Muller andFindlay's experiment is therefore :onsistent with Posner et. al.'s claimthat attention can be spread to adjacent regions; it does not pertain tothe claim that it cannot be split (at least in the sense Posner et. al.studied).

Currently, the study of attention is entering an exciting phase inwhich the operations involved in internal mechanisms of attention are beingrelated to neural systems (see Berlucchi & Rizzolatti, 1987; Posner &Marin, 1985 for reviews). These efforts require combining carefulperformance :tudies of the type done by Muller & Findlay with efforts tounderstand the neural systems involved. For these efforts to succeed, itis necessary to relate performance parameters (e.g. d' and beta) to themany mechanisms that may influencE them.

I Writing of this critique was supported by a contract No. N00014-86-K-0289 from the Office of Naval Rese-irch. The authors appreciate an analysis

*, by Dr. Harolo Hawkins that aided 1he development of this paper.

'.-

0

.9

-. % V..

%

References

Baynes, K., Holtzman, J.D. & Volpe, B.T. (1986) Components of visualattention: alterations in response pattern to visual stimuli followingparietal lobe infarction. Brain, 109, 99-114.

Berlucchi, G. & Rizzolatti, G. (eds.) (1987) Selective visual attention.Neuropsychologia, 25, 1A.

Briand, K. & Klein, R. (1987) Is Posner's "beam" the same as Treisman's"glue": On the relation betwE(n visual orienting and feature

% integration theory. JEP:HPP, .13, 228-241.

Broadbent, D. (1971) Decision and S;tress. London:Academic Press.

De Renzi, E., Gentilini, M. & Pattacini, F. Auditory extinction following

hemisphere damage. Neuropsychiogia, 1984, 22, 733-744.

Downing, C. and Pinker, S. (1985) The spatial structure of visualattention. In Mechanisms of Attention: Attention and Performance XI,M. I. Posner and O.S.M. Matin (eds.). Erlbaum and Associates,

* Hillsdale, New Jersey.

Duncan, J. (1980) The demonstration of capacity limitation. CognitivePsycnology, 12, 75-96.

Eriksen, C. and St. James, J. (1966) Visual attention with and around thefield of focal attention: a ziom lens model. Perception &Psychophysics, 40, 225-240.

Friedrich, F.J., Walker, J. & Posn-r, M.I. (1985) Effects of parietallesions on visual matching: ipiplications for reading errors.Cognitive Neuropsycholoy, 2, 253-264.

Hoffman, J., Nelson, B. & Houck, M. (1983) The role of attentionalresources in automatic detectj,)n. Cognitive Psychology, 15, 379-410.

Hughes, H. & Zimba, L. (1987) Nat,ral boundaries for the spatial spread ofdirected visual attention. Neuropsychologia, 25, 5-18.

O Jonides, J. (1976) Voluntary versus reflexive control of the mind's eyemovement. Paper presented at -he Psychonomics Society, St. Louis, MO.

Kahneman, D. (1973) Attention and ffort. Englewood, NJ:Prentice-Hall.

Kiefer, R.J. & Siple, P. (1987) SIatial constraints on the voluntary* ontrol of attention across vi.;ual space. Canadian Journal of

Psychology, 41, 474-489.

10

0-0

i&A % *~ .*~~. . , NIR ** .*'

0

LaBerge, D. (1983) Spatial extent of attention to letters and words.

Journal of Experimental PsychoLogy: Human Perception and Performance,9, 371-380.

Ladavas, E. (1987) Is the hemispatial deficit produced by right parietallesions associated with retinal', or gravitational coordinates? Brain,

167-180.

Legge, G. (1978) Sustained and triansient mechanisms in human vision:

temporal and spatial propertie;. Vision Research, 18, 69-81.

Maylor, E. (1985) Facilitatory and inhibitory components of orienting in

visual space. In M. Posner anI 0. Matin (eds.) Attention and

Performance XI. Hillsdale, NJ:Erlbaum.

Morrow, L.A. & Ratcliff, G. (1987) Attentional mechanisms in visualneglect. J. of Clinical and E;;:perimental Neuropsychology, 9, 1, 74-75.

Muller, H. and Findlay, J. (1987) Sensitivity and criterion effects in thespatial cueing of visual attention. Perception & Psychophysics, 42,

383-399.

* Navon, D. (1985) Attention division or attention sharing? In M. Posnerand 0. Matin (eds). Attention and Performance XI.

Hillsdale,NJ:Erlbaum.

Nissen, M., Posner, M. & Snyder, S. (1978) Relationship between attention

shifts and saccadic eye movemeits. Paper presented at the PsychonomicsSociety, November, 1978.

Posner, M.I. (1980) Orienting of ittention. Quarterly J. of ExperimentalPsychology, 32, 3-25.

Posner, M.I. & Marin, O.S.M. (eds) (1985) Attention and Performance XI.-Hillsdale, N.J.:Erlbaum.

Posner, M.I., Snyder, C. & Davidson, B. (1980) Attention and the detection

of signals. Journal of Experimental Psychology: General, 109, 160-174.

Posner, M.I., Cohen, Y., and Rafa!, R.D. Neural systems control of spatial* orienting. Proceedings of the Royal Society of London B 298, 1982,

187-198.

Posner, M.1., Walker, J., Friedti,'h. F. & Rafal, R. (1984) Effects of

parietal injury on covert onii,,ting of attention. Journal ofNeuroscience, 4, 1863-1874.

Remington, R. (1980) Attention and saccadic eye movements. Journal of

E.perimental Psychology:Human Perception and Performance, 6, 726-744.

¢- 1

0%

- I

S

Riddoch, J. & Humphreys, G.W. (1987) Perceptual and action systems inunilateral visual neglect. In M. Jeannerod ( d Neurophysiologicaand Neuropsychological Aspects of Spatial Neglect, Elsevier:Amsterdam,151-181.

Rizzolatti, G., Riggio, L., Dascola, I., & Umilta, C. (1987) Reorientinge. ~:attention across the horizontal and vertical meridians: evidence in

favor of a premotor theory of attention. Neuropsychologia, 25, 31-40.

Sagi, D. & Julesz, B. (1986) Enhanced detection in the aperture of focalattention during simple discri.nination tasks. Nature, 321, 693-695.

Shaw, M. (1984) Division of attenlion among spatial locations: afundamental difference between detection of letters and detection ofluminance increments. In H. Bouma and D. Bowhuis (eds.) Attention andPerformance X, Hillsdale,NJ:ErLbaum.

Shepard, M., Findlay, J. & Hockey, R. (1986) The relationship between eye

movements and spatial attentioa. Quarterly Journal of ExperimentalPsychology, 38a, 475-491.

Shulman, G., Wilson, J. & Sheehy, 1. (1985) Spatial determinants of thedistribution of attention. P*.tception & Psychophysics, 37, 59-66.

Shulman, G., Sheehy, J. & Wilson, J. (1986) Gradients of spatialattention. Acta Psychologia, 51, 167-181.

Sieroff, E. & Michel, F. (1987) V,!rbal visual extinction in right/lefthemisphere patients and the problems of lexical access.Neuropsychologia, in press.

Tolhurst, D. (1975a) Reaction times in the detection of gratings by humanobservers: a probabilistic me~hanism. Vision Research, 15, 1143-1150.

Tolhurst, D. (1975b) Sustained and transient channels in human vision.Vision Research, 15, 1151-1155,.

*i. Turvey, M. (1973) On peripheral and central processes in vision:inferences from an information -processing analysis of masking withpatterned stimuli. Psychologj.:al Review, 80, 1-52.

*. Ullman, S. (1984) Visual Routines. Cognition, 18, 97-159.

N.d

12

%

Wz7NfImOfnI I

24 La Q a

tz La 06.

tu

Q Q Itac0-a 24~ CLK ~~J~A ~CIZ1 OZ

L'Q

6. 4 1.Z :

Ck ;ac cqc z~b

~~c I.. Q ~.r4

ka cc.~ A

Zl cc -cnv0

- ~ ~ A.

Ln0

2Q C

-4 C4 ac Ac11 -L

Z Q ac2 c L.a C C1 c c,-c Q4 -c

~ ~ . ac

uJJ-

0~~-s6 L,'. tA~ W0~ 0I-o k ~U~ i

14 U1

cc uU4~

*~C Z0 U i u A0 .~ac L -

4Lcz k.

-t k.. xQ .: ~ ZZ

.4 ac 0l* ,.-,.~ ne

U4 ,4z,

~ -~ ~ QC

cc L5' ,j k QZ

ac CCO aC a L. a

to z, " u

La.

La ;2

4-1

044

t ia. go

-. 0~ Q Q **, .4

k. Q k3 IQ

*0 A6

e~Ic 8 a

p IS

Ia..2 Co~

.EQ~.-

a", ~ I CC C9

- .. U~. ~ , ~ I J0i

~c IC~ 06 hJ~~

0c

.. & '. aica , a ,,1

C4 L'J

. 6 -2~ .~ ~ AG

'-L. CA

*~ L- zJ-,Q -J -J Z

Ul. nI

Cc k

Z m'

cc

S~Zk 41.~~ -

~Z

25 ~ o -- 0 Z ~ Q'~~~ ~

co :?

Q0 UCc -3.

1w '

*553 ..3. 3'.- '0~ C

WQ Z- 3

Im vi'W ~

lw'-0IC LQ ~3

-IW Qia

m -S cc

00~ 00

*Q tw5 43

ag A C -

04

-. 5.~~ ~ ~~~ E! *5 *,*,S * z- .*-S -''--'L'YQ ~ ~ j

1 .. ~ ' C 19 * kh~ .A.-, lW

WU r --

~.0

lVkj*1-2

~~Z514 0- _4 4

~ ~ ~ cc

C2 CC

~1 2

06*

Cc. . N. . 4N

~~4J

Wo U2

L.Z uI-, u

2,- E - .t

co N

*z.

UUg14

't N. i ,

46- w I

21 CcN InCa

*Q Q

Q cc

'N4z,

w tz

~uw 0!

Q CQ

Co '

CKhi ~L

'o CL

ti

06 zih

hi hi U 'C~ ~ ..

0

-4,

-J

~

-4-.,

'~.. -'p4..',* ~1

-4

4,-4.

J4 4,,

4-.

-. 4

'<'h

0

-- 4.

0

Sb I* U U S S S S S

S V

I4, ,' *4~~ ~

f7 J---. 4-.-..

4 - - .4 .'.~*.~ 4,~.~ -, 'p

.4 b. - 'b~ S