eU resulti indicated that only the time intercept parameter of the timeaccuracy func' tion is affected by retinal location of sHmulation. This finding suggests that it ...
Perception & Pq,choPltr t;rt 1 9 8 1 , 3 0( 2 ) ,1 1 4 - t I 8
Retinal location and visual processingrate JOHN J. STERLING and TIMOTHY A. SALTHOUSE University of Missouri-Columbia, Columbia, Missouri 6520I Although previous studies have shown that the time required to process visual stimuli,in' ".".r"r fo"rpresentations away from the fovea, the evidence concerning the exact nature of this increaseis inconclusive. Three experiments were conducted using both reaction time and tachis' toscopic masking tasks to gene"ale timeaccuracy functions for stimuli at different retinal loca' tions.-eU resulti indicated that only the time intercept parameter of the timeaccuracy func' tion is affected by retinal location of sHmulation. This finding suggests that it takes longer for information to becomeavailable to some decision mechanism with stimuli displaced away{rom the fovea. but that the actual rate of extracting information is not influenced by retinal locus of stimulation.
The purpose of the present experimentswas to At least since Poffenberger(1912), it has been parameter,the slope known that reaction time increasesas stimuli are determinewhich time-accuracy located farther from the foveal center of the retina or the intercept, is primarily affected by the retinal (e.g., Eriksen & Schultz, 1977;Lefton & Haber, location of stimulation. Some relevantevidenceis 1974;Payne,1966;Rains,1963).The mechanismfor available, but it is presentlycontradictory. Estes experimentin which this slowerperformancein peripherallocationsis not (1978)describeda tachistoscopic to limit processing was used poststimulus mask yet clear, although it is known that the density of a peripheral locationswere that he concluded and from the time, with distance retinal receptorsdecreases function reof the slopes shallower with is needed associated presumed more time that might be fovea. It when fewer receptorsare available,but the greater lating processingtime to identification accuracy. time might be required either for initial sensoryin- Eriksen,Becker,and Hoffman (1970),however,retegration or for actual information extraction.These porteda similar studyin which parallel(equal-slope) two possibilitiescan be investigatedby examiningthe functions were obtained for stimuli at different retfunctions relating time (in either reaction time or ta- inal locations. maskingtasks)to stimulusclassification chistoscopic accuracy. EXPERIMENT 1 Salthouse(Note l) hasdemonstratedthat the timeaccuracyfunctions can be characterizedin terms of tradeThe first experimentutilized speed-accuracy the intercept(the point in time at which accuracybe- off procedures (e.g., Salthouse,1979,l98l; Wood ginsto exceedchance),the slope(the rate of increase & Jennings,1976)to generatetime-accuracyfuncin accuracyper unit time), and the accuracyasymp- tionsin a reactiontime task.An eyemovementmonitote (the final level of accuracy achievedwith un- tor was used to ensurethat subjectsdid not move limited time) parameters,and that theseparameters their eyesto fixate on the peripheralstimuli. are differentially sensitiveto various experimental manipulations.For example,increasedstimulusin- Method Subjcct!. Four right-handcdfemaleswith normal or corrcctedtensity reducesthe time-axisintercept, while increased assubjectsfor 16 l-h scssions. stimulus discriminability increasesthe slope of both to-normalvisualacuity served Apprntur. Stimuli were presentedon a Mini-Bce CRT contimemasking reaction time and tachistoscopic trolled by a PDP lll34 computer. Eye movementswere monaccuracyfunctions.(Seealso Lappin & Disch, 1972; itored by a Narco BiosystemsModcl 2fl) Eye MovementRecorder Link & Tindall, l97l; Pachella& Fisher, 1969; interfacedwith the computer.Head position was held constantby that the inter- a chin cup and a foreheadrestraintdevice. Swensson,l9?2). Salthousesuggested all processes Proccdurr. Target stimuli consistedof the .3-deglettersX and of duration parameter the reflects cept O prescnted3.3, 5.6, or 7.6 deg to the left or right of center exceptactual information extraction, while the slope fixation. Subjectsrespondedusing either thc secondor third finrepresentsthe rate of extractionof information from Berof the designatedhand to pressone of two spccifiedkcys on the stimulus. The asymptote parameter, unlike the the keyboard. Stimulus-responsepairings were balancedacross intercept and slope parameters,is independent of subjectsand sessions. The first four scssionswereconsideredpracticeto stabilizep€rtime and is therefore likely to reflect state, rather formance and establishthe time boundariesfor the specd-accuracy than process,limitations.For this reason'it is not of manipulations.The rcmaining 12 sessions,consistingof 4 sessions primary interest in this investigationand will not be with eachlocation condition in a balanccdorder for eachsubject, each involved 7 blocks of trials. The first block consistedof further. discussed Inc. Society, Copyrightl98l Psychonomic
tt4
l r7 181|o6orl4-0ss00.7sl0 0031-s
LOCATIONAND PROCESSING RATE 20 practice trials during which the subjectsrespondedat their slowesttime boundaries.The responsesfrom this block were discardedand not utilized in any analyscs.The remaining6 blocks of40 trials eachwereperformedat one of the threcspeed-accuracy emphases.A goal region was defined in terms of a set of minimum and maximum reaction times within which the subject was instructed to attempt to produce all of her responsesat whateverlevel of accuracywas appropriateto attain the dcsiredspeed. The minimum and maximum times of the three speed-accuracy regionswere determinedindividually for eachsubject to produce approximatcaccuraciesof 6090,75o/0,and 900/0. A trial beganwith the subject fixating on a centerasteriskfor |.5 sec.The termination of the fixation point was followed by the presentationof either an X or an O in one of the visual field locationson either side of the fixation point. Subjectsresponded by pressingone of two keys as rapidly as was consistentwith the speed-accuracy region. The feedbackafter eachtrial consistedof the word CORRECT or WRONG printcd on the screen, followed by a time line with vertical bars representingthe minimum and maximum times acceptablein that goal region and an arrow indicating how fast the subject actually respondedon that trial. Subjectswerc instructed to respond so that the arrow appearcd betweenthe verticalbars. The speed-accuracy regions for cach subject wcre counterbalancedwithin each scssionbeginning with the slowestregion (i.e.,ABCCBA order). Any deviationof the eyesfrom the centerfixation point during the target presentstionwould result in unknown visual field location. Conscquently,I deviation of more than 2 dcg from thc central fixation point detectedby thc cye movementmonitor causcd the computcr to: (l) discard the trial; (2) instruct the subject to stop moving her eyesfrom the centralfixation point; and (3) begin a newtrial.
115
F
U q c
o z q c
Frgu,r r. ,."" ;:'::-r* rocrtronr *,.,;:';:':: derlvcdby prcdlctlngrcrcdon llme rl rccurrcla of659o rnd tt9o (Expcrlmcnt 1).
ducted on these values revealedsignificant effects of accuracy[F(1,3)=24.25, p < .021and location [F(2,6)=7.81,p< .05], but not the interactionof locationx accuracy[F(2,6)( 1.0,n.s.l. The meanregressionlines for the threelocations, computedby averagingthe predictedvaluesfor each subject, are displayedin Figure l. Notice that the Results time-accuracyfunctions are nearly parallel for the Trials with reaction times greater than 900 msec variouslocationconditions. were excludedfrom analyses.Lessthan l9o of the trials weredeletedfrom eachlocation condition. The Dlscusslon remaining trials were subjectedto linear regression The major result of Experiment I is that the timeanalysesto determinethe correlation, slope, and accuracyparametersensitiveto retinal location apinterceptparametersof the function relating reaction pearsto be the intercept and not the slope. The Actime to classificationaccuracy.'For each location curacyby Location interactionwas not significant, condition, there were 24 reaction-time,/accuracyindicatingthat the reactiontime changebetween6590 pairs,corresponding to the numberof blocksof trials and 8590 accuracy was roughly equivalent for all during the experimentalsessions.Pairs with ac- locationconditions.Sincethe time requiredto achieve curaciesless than 5590 or greaterthan 9590 were a comparableincreasein accuracywas not different, omitted from the computationsto avoid inclusion of it can be inferred that the rate of accumulatingindata from the chanceaccuracyand perfect accuracy formation from different retinal locations also was segmentsof the time-accuracyfunction. The correla- not different. tions betweenreaction time and percentcorrect for eachsubjectrangedfrom +.49 to +.81, with a mean EXPERIMENT 2 of +.67. Thesevaluesare not particularlyhigh, but they do indicatethat subjectswereableto trade speed Experiment2 was designedto demonstratethat for accuracy. retinal location affects the interceptbut not the slope Becausethe slope and intercept of the speed- parameterin time-accuracyfunctions derived from accuracytradeoff functions are possibly correlated tachistoscopicmaskingtasks, as well as those gener(i.e., Salthouse,1979;Wood & Jennings,1976),it ated from speed-accuracyreaction time tasks. The may be unreasonableto expectthat separateanalyses previousresultswill also be reexaminedwith a modof eachmeasurewill provide meaningfulresults.We ified speed-accuracy procedurethat allows complete thereforeusedthe slope and interceptparametersto tradeoff functions to be derived within a single sesgeneratepredicted reaction times at accuraciescor- sion. Eye movementswere not monitored because respondingto 65q0and 85vocorrectjudgments.A 3 there were relatively few contaminating movements (location) x 2 (accuracy)analysisof variance con- detectedin Experiment l.
116
AND SALTHOUSE STERLING
Method
Sublectr. Forty-eight undergraduatepsychologystudentswith normal or corrected-to-normalvision servedas subjectsin a l-h scssion.None of thc subjectshad participatedin Expcriment l. Apprrrtrs. A PDP I l/03 computer,interfacedwith a Hewletr PackardModel l3llA DisplayMonitor and two lGkey telephone keyboards,wasusedto presentstimuli and recordresponses. Procrdurc. All subjcctscompletedsix blocks of trials-three in the reaction time task and threc in the tachistoscopictask. The first two blocks were consideredpractice (one for each task), and the remainingfour blocks (two for eachtask) were presented in counterbalanccdorder. The target stimulus for both the reaction time and tachistoscopictasks was an arrow subtcnding 1.5 deg of visual anglc, pointing at I 45-degsngle to either the left or the right. For a given subjcct, the target stimulus appearedonly at fixation or at 7,5 or 15.0 deg to the left or right of fixation. That is, the location variable was manipulated acrosssubjects,with 16 subjects recciving targets at each location. The subjectswere instructed to respondby pressingthe key on the left of a keyboard if the arrow was pointing to the left, and to respondby prcssingthe key on the right of a keyboardif the arrow was pointing to the right. Each trial in both taskswas initiated by thc appearanceof four centerfixation dots placedat the cornersof a squaresubtending 2.0 dcg of visual angle. Onc secondafter the presentationof the fixation square,the arrow was presentedin the appropriatelocation. The stimulusremainedon until thc subjectrespondedin the reaction time task but was prescntcdfor only 5 msec with a 5to 130-msccblank interval before a poststimulusmask (a square with diagonallinescovcringthe shaftsof the arrows)in thc tachistoscopictask. In the reaction time task, a timc line with two vertical markers and an arrow pointer appcaredon the scre€naftcr the subject responded.The subjectswere instructed to respond within the designatedtimc region regardlessof the accuracythat might result. A trial block beganwith the desiredtime region between375 and 475 mscc,and the time region was reducedin 50-msecsteps until it reached 125 to 225 mscc. That is. in the first l0 trials, the subject was to respond between 375 and 475 msec, in the second l0 trials, hc or she was to respond between 325 and 425 msec, in the third l0 trials, bctween275 and 375 msec, and so on. The time region remainedat 125to 225 msecuntil the subjcct produccd l0 responseswithin that interval, at which point the time region incrcasedto progressivclyhigher valuesafter l0 responsesat each interval or until a total of 250 responseshad beenproduced.This is a procedurcfor generatingreaction-time/ spced-accuracyfunctions different from those used in Experiment l, but it has thc advantageof being much more efficient in the collcctionof data. A trial block in the tachistoscopictask consistcdof 2O trials at eachinterstimulusintcrval from 5 to 130 msec.Ten trials were presentedat cach interval, starting from the largestinterval and thcn decreasingto the smallestinterval bcfore increasingagain.
Results Data from the two experimentalblocks of the reaction time task were combined,and accuracieswere determinedfor each 50-msecinterval ranging from 175 to 475 msec. Regressionparameterswere then derived from thesevaluesafter deleting points with low or high accuraciesto obtain the bestexcessively fitting regressionline with a minimum of threepairs of values.The slopeand interceptparameterswere then usedto predictreactiontimesat 6590and 8590 accuracyfor each subject. The correlationsbetween reactiontime and percentcorrect for thesefunctions for eachsubjectrangd from .43to .99,with a median of .98. Note that the presentprocedurefor generat-
functions is not ing reaction-time/speed-accuracy only more efficient, but also appearsto better reflect the negativerelationshipbetweenspeedand accuracy than doesthe procedureusedin Experimentl. A 3 (location) x 2 (accuracy)analysisof variance conducted on the predicted reaction times revealed significant effects of accuracy[F(1,45)=624.97, p < .00011 and location[F(2,45):21.70,p < .0001]' but not of accuracyx location[F(2,45)( 1.1,n.s.]. (Similar results were obtained in an analysis conducted on the mean accuraciesat each s0-msecreaction time interval. The location x time interaction was significant in this analysis,but the data illustrated in Figure 2 indicate that it is probably due to convergenceof accuraciesat the shortest reaction times.) The mean accuraciesat each time interval for the three locations are illustrated in the right side of Figure 2. Both statisticallyand graphically, the presentresultsconfirm those of Experiment l. The data of Figure 2 suggestthat therewas no differencebetweenthe reactiontimeswith stimuli at 7.5 and 15.0degfrom fixation. Subjectsin both of these conditionsoften complainedthat it was difficult to respondto an arrow on one sideof the screenwhen the correct responsewas to hit the key on the oppositeside of the keyboard.It is thereforepossible incompatibility effect was that a stimulus-response masking the effects of visual field location with the extremepositions. The data from the two tachistoscopictrial blocks werecombined,and the accuracywasdeterminedfor each interstimulusinterval. Becauseaccuracyhad reachedan asymptotic level by 105 msecin all conditions, only the intervals between 5 and 80 msec were examinedin a 3 (location) x 4 (interstimulus too t s 9 0 U c
3 8 0 L,, U 7 0 { z u 6 0 c {e.
5 30 55 80tO5 ,NIER ST'A4UIUS ,NIERYAT
400 REACT'ON f 'ME
fN l,tsEcs Flgurc 2. Lcft sldc: Mcrn rccuncy rs r functlon of Intcntlmulus lntcrvrl for thrcc rtlmulus locrdons (trchlstoscoplcmrsklng terl, Expcrlncnt 2). Rlght cldc: Mcrn lccurlcy rs r functlon of rcrcllon tlne for three stlmulus locrtlons (rcrctlon tlmc t$k' Expcrlment 2).
r
LOCATIONAND PROCESSING RATE interval) analysisof variance.Location [F(2,45)= 7.87, p < .0021and interstimulusinterval [F(3,135) :202.32, p ( .00011,but not their interaction [ F ( 6 , 1 3 5 ) =1 . 8 4 , n . s . ] , f a c t o r sw e r e s i g n i f i c a n t . The mean accuraciesat each interval for the three locationsare illustratedin the left side of Figure2. EXPERIMENT 3
II7
Proccdurc.The task was identical to thc tachistoscopictask in Experiment2, exceptthat the interstimulusintervalswerechanged to 5, 15, 25, 35, and 45 msec.Subjectsfirst completeda practice block of trials with thc target arrow presentedat 7.5 deg to the right or left of fixation with the interstimulusintervals from Experiment2. Two blocks of trials were then administeredat each of the three visual field locations. in an order balanccd across conditionsand subjects.
Results Data for the two blocks completedat each visual field location were combined, and accuracy was determined at each interstimulus interval. These valueswere then subjectedto a 3 (location) x 5 (interstimulus interval) analysis of variance. Main effects of location lF(2,225):39.42, p ( .00011and interstimulusinterval IF(4,225):91.14, p < .00011 were significant, but not their interaction lF(8,225) =1.04, n.s.l. The mean accuraciesat each interstimulusintervalareillustratedin Figure3.
Although the InterstimulusInterval by Location interaction for the tachistoscopictask in Experiment 2 was not significant,inspectionof Figure2 suggeststhat the slope parametermight be affected by visual field location. One possibleinterpretation of this result is that the interval increments(25 msec) were too large and thus the data were not precise enoughto accuratelydepict the true parallel nature of the functions.Another possibilityis that the apparentdifferenceis real but that the between-subjects GENERAL DISCUSSION designused in Experiment2 was not powerful enoughto detectthis difference.Experiment3 was The major finding of theseexperimentsis that only designedto investigatethese alternative interpretations by usinga smallervalue(10 msec)of interstim- the time interceptof the time-accuracyfunction is afulus interval incrementand having each subject fected by the retinal location of stimulation. A difreceiveall stimuluslocationsto providea more sen- ferencein the time-accuracyslopewould be reflected sitive test of the InterstimulusInterval by Location in an interaction of either location X accuracy or location x time. The interactions were not signifinteraction. icant, but the main effects of location were significant, in all analyses.Moreover, the figures illusMethod Sublcctr. Sixtcenundergraduatepsychologystudentswith nor- trated that shifting the locus of stimulation away mal or corrected-to-normalvision participatcd in a l-h session. from the fovea delaysthe time at which accuracybeNone of the subjectshad servcdin the previousexpcriments. gins to improve abovethe chancelevel. Theseresults Apperrtur. The apparatuswas the sameas that usedin Experiwere obtainedwhen processingtime was limited by ment2. the occurrenceof the responsein a speed-accuracy reaction time task (ExperimentsI and 2) and when processingtime was limited by the occurrenceof a poststimulusmask in a tachistoscopic maskingtask o @xperiments2 and 3). The conclusion that retinal locus influences the b e o 7.5 time at which accuracybegins to improve above a l! q level, but not the actual rate of improvement chance c onceit begins,is consistentwith a finding by Eriksen, 3ro Becker, and Hoffman (1970).Estes(1978)argued E that retinal location affects the rate of accuracyim(9 provement, but his inferencewas basedon unpubi, 70 lished data without reported statisticalsupport, and u thus it is difficult to evaluate. q q The analysisof time-accuracyparametersis based t60 on the assumptionthat the total time to achievea given level of accuracycan be divided into two distinct componentscorresponding to (l) the amountof increasein stimulus information per unit time, and (2) the duration of all other processesexcept infor5 t 5 2 5 3 5 4 5 mation extraction. The presentresultsindicatethat a , N T E R S I ' A 4 U I U 5 ' N I E R YI .A reduction in the number of availablereceptorsas the ,N /t,tsEc locus of stimulation is moved from fovea to periphFlgure 3. Mern rccurrcy rs r functlon of Intentlmulus Intervrl ery doesnot alter the rate at which accuracy(stimulus for threestlmuluslocrllons (Expcrincnt 3). information) is accrued.Instead,a reductionin numq
118
STERLING AND SALTHOUSE
ber of receptorsincreasesthe amount of time re- functions. Whether the present interpretation of quired before accuracy begins to improve above a theseparametersis correctmust await more extensive analyticalinvestigationof the type reportedhere. chancelevel. An interpretationconsistentwith theseresultsis REFERENCE NOTE that the time intercept correspondsto, among other things, the duration necessaryto establishan initial l. Salthouse,T. A. Parometercol the time'accttracyrelotionrepresentationof the visual stimulus, and that the ship. Paper presentedat the 2fth Annual Mccting of the Psychoslope representsthe rate of discriminating between nomic Society,Phoenix,Arizona, November1979' alternative representations.A fewer number of receptorstherefore delaysthe establishmentof the iniREFERENCESi tial representationbut does not impair the speedof distinguishingbetweendifferent possiblerepresenta- Enrxssr.r,C. W., Bncrnn, B. 8., & Horru,lx, J. E. Safari to masking ltnd. Perception & Psychophys,'ct,1970' t, U5-250. tions. This interpretationis alsosupportedby the reC. W., & Scsulrz, D. W. Retinallocusand acuityin Enrxsrr, sults cited earlier with respectto manipulations of visual information proccssing.Bulletin oJ the Psychonomic stimulus intensity and stimulus discriminability. ReSociety,1977,9, El-84. ducedintensity would be expectedto hamperthe for- Esrrs, W. K. Perceptual processingin letter recognition and reading. In E. C. Carterette& M. P. Friedman (Eds.), Hand' mation of a representation,and reduceddiscrimbook oJ perception (Yol.9l Perceptuolprocessing.New York: inability or distinctivenesswould be expectedto slow AcademicPress,1978. precisely has what process. is This the discrimination Lepprr,r,J. S., & Dtscg, K. The latencyopcratingcharacteristic: been found in severalstudiesin which stimulus inII. Effects of visual stimulus intensity on choicercaction time. Journal of Experimentol Psychology, 1972,93, 267-272. tensity hasbeenreportedto affect the time intercept, whereasstimulusdiscriminabilityhasbeenreported Lpprox, L. A., & Herrn, R. N' lnformation extraction from different retinal locations. Journol of Experimental Psychology' to affect the slope(e.g., Lappin & Disch, 19'72;Link r974,102,975-9E0. & Tindall, l97l; Pachella& Fisher,1969;Swensson' Lrnr. S. W., & TInorr,r,, A. D. Spccdand accuracyin compara1972;Salthouse,Note l). tive judgnrentsof line length.Perception& hychophysis' l97l' 9,28,1-2EE. It is important to stressthat there is no intrinsic Frennn, D. F. Effcct of stimulusdegradation temporal relationshipimplied betweenthe two com- Prcnnlr.r, R. G.,on& thc tradcoff betweenspeedand accuracyin and similarity ponentsof the presentmodel of the time-accuracy absolute judgments. Journol of Experimentol Psychology, contributingto the functions. That is, the processes 1969,tl,7-9. time intercept may precede,follow, or occur simul- Pevxn, W. H. Reaction time as a function of rctinal location. VisionResearch,1966,6,729-732. taneouslywith the processesresponsiblefor the slope. PorrpNsrnonn, A. T. Rcaction time to rctinal stimulation with This is particularly apparent in the comparison of specialreferenceto the time lost in conduction through nerve the reaction time and tachistoscopictime-accuracy centers.A rch ivesof Psychol ogy, 1912,23, l -73. functions. The time interceptsof the former are Rerxs, J. D. Signal luminancc and position effects in human reactiontime. Yision Reseorch,1963,3, 239-251. much larger than thoseof the latter, presumablybetradeoff. T. A. Adult age and the speed-accuracy cause responsepreparation and executionare re- Srr-rnousn, ics,1979,22,8I l -821. Ergonom quired in the measuredinterval for the reaction time S,rr,nrousn,T. A. Convergingevidcncefor information processing tisk. In the reactiontime functions, therefore,it may stages:A comparative-influencestage-analysismethod.,4cfa Psychologica,lgEl, 47, 39-61. be speculatedthat the time intercept is due to the vs. accuracy in establishmentof the stimulus representation,which Swnr,rssoN.R. G. The elusive tradeoff: &Spced 1972, tasks.Perception Psychophysics, visual discrimination process, at leastpartially precedesthe discrimination t2, 16-12. response' the and to the preparationand executionof tradeoff funcWooo, C. C., & Jnr.rrvIxcs,J. R. Speed-accuracy tions in choicc reaction time: Expcrimcntal designsand computawhich follows the discriminationprocess.However, tional procedura. Perception& Psychophysis,1976,19,92-l0l. the similar finding of retinal location affecting the interceptand not the slopein both reactiontime and NOTE tachistoscopicfunctions suggeststhat the current finding is not merelya reflection of responsefactors. l. The percentcorrectmeasureof accuracywasutilized because We suggestthat this two-componentmodel of Salthouse(1979, l98l) had reported that this measureyielded as those obtime-accuracyparametersmay prove extremelyuse- reaction-time/accuracycorrelationsat least as largesuch as informore theoreticallydependcntmeasurcs from tained perin simple processess involved ful in analyzingthe d', (d')2, and log odds correct to incorrect transmittcd, mation ceptualdecisions.The slopeand interceptparameters responses. have already been demonstratedto be differentially (ManuscriPtreceivedMarch 9, l98l; sensitiveto a variety of experimentalmanipulations revisionacceptedfor publicationMay 19' 1981.) under severalproceduresfor generatingtime-accuracy
