Mrmnrv & Cognition1994 . 22 10 145-156

Procedural differences in processing intactand degraded stimuli

SANDER A. LOSVrtje Unauersiteit, De Boelelaan, The Netherlands

In three experiments, the extent to which the processing of a visual stimulus profits from equalprocessing demandsofa preceding stemulus was examined. Subjects identified two subsequentlypresented digits (S1 and 52) that were either intact or degraded by noise, yielding four combina-tions of stimulus quality . In Experiments 1 and 2, Sl and S2 differed with respect to the valuesof the digits, so that stimulus quality was the only dimension of possible agreement . The resultsrevealed a faster response to SZ when the stimulus pairs were homogeneous (both intact or bothdegraded stimuli) than when they were not homogeneous (degraded-intact pairs and intact-degraded pairs, respectively ). The occurrence ofequal values of Sl and S2 (Experiment 3) tendedto magnify this homogeneous-stimuTrcs effect, but was not a prerequisite for its occurrence . Rela-tive to conditions considered to be neutral, the homogeneous-stimulus effect proved to he dueto deviant behavior following the processing of a degraded Sl . The suggestion that this reflectsthe involvement of controlled processing is discussed .

A common distinction in the description ofhuman co 'g-nition is between representation and procedure. It fre-quently occurs in broad theories in which a representa-tion is conceived aN a substrate that results from proceduralaceivity, or on which procedure,, operate (e .g . . Ander-son, 1983) . On a more basic level, the distinction mayset the stake for discussions regarding the nature of vari-ous experimental results, namely, whether these originatefrom the operation of procedures or from the presenceor activation of certain representationsA specific example concerns the controversy regard-

ing the nature of the fast-same effect in sequential match-ing tasks, in which subjects judge whether two sequen-tially presented stimuli (S1 and 52) are "same" or"different" (Farell, 1985 ; Krueger, 1983 ; Krueger &Shapiro, 14$1b ; Proctor 1981 ; Proctor & Rao . 1983) . Thematching response time (T2) it defined as the time elaps-ing from the presentation of S2 until completion of theresponse . The fast-yarrre effect refers to the common ob-servation that T2 is shorter on same trials than on differ-ent trims (e .a . . Bamber, 1969 ; Krueger, 1978 ; Proctor .1981) . Contrary to the common belief that this effectoccurs at the level of matching, for instance, due to dif-ferences m time to compare or recheck (Krueger, 1978)representations of some and different,stimuli, Proctor(1981) proposed that the process ofencoding is completedfaster ifa stimulus is preceded by a stimulus that has anidentical form . The empirical basis for this encoding facili-

The author i5 grateful to [Jatiid E Irwm, Andric; F Sanders, Gor-don D Loan, and cw n anonvmou5 reviewers fir invaluable commentsnn earlier draft; of this paper Correspondence should be addrcesed toS Log, Department of Psychology, Vnkle Um~ersitert, De Boeiclaan11 11 . 1081 HV Amsterdam, The tietherlan&

-Rteepred b1 pre,, pniic ednnr tilarearer Jean. huorrs Prre"nn

tation derives from the observation that the fast-same ef-fect has about the same size in a naming task, in whichS2 must be named instead of matched with 51 (Proctor,1981) . Fareli (1985) questioned Proctor's implicit claimthat matching did not occur in his naming task, but ad-mitted that encoding facilitation is an entirely reasonableconstruct . Krueger and Shapiro (1481b) argued furtherthat stimulus repetition may entail a shift in criterion ratherthan increased sensitivity, but Krueger (1983) discon-firmed this view when he failed to find differences be-tween error distributions in a simultaneous and a sequen-tial matching, task, thuti rendering support for the encodingfacilitation hypothesis .

It should be noted that an explanation ofthe fast-sameeffect in terms of either matching or encoding facilita-tion assumes the physical identity of S1 and 52 . Nicker-son (1975) confirmed the relevance of stimulus identityto a letter-matching task, in which the letters were de-graded by various noise patterns . The noise was irrele-vant to the task in that the matching response was madewith respect to the value ofthe letters and not with respectto the noise patterns . However, Nickerson found that T2was shorter when the noise patterns corresponded thanwhen they did not, This result, though interesting in it-self. failed to cnticallv contribute to the discussion on thenature of the fast-sane effect, because it stands open toseveral interpretations . Nickerson concluded from his datathat all features, whether relevant or not, are representedin memory, and that the comparison of representationsevolves faster if S1 and S2 have identical features . Proc-tor (I98I), by contrast, preferred an encoding locus forthis effect, arguing that the encoding pcoce~,, is facilitatedin the case of stimulus identity .From a theoretical point of view, it seems more fertile

to focus on departures from stimulus identity as a neces-

145 Copyright 1994 Psychonomic Society, Inc

146 LOS

sarv condition for performance enhancement . In this re-spect, the empirical data obtained by Hansen and Sanders(1988, Experiment 1) are of interest . They had ;ub:lcct%match two digits that were sequentially presented in thevisual field . Either digit was intact or degraded by noise,constituting four conditions of 51-S2 quality : intact-intact,intact-degraded, degraded-intact, and degraded-degraded .The data showed that T2 was not only affected by the qual-ity of S2, but also by the relation of the qualities of Stand S2 . Specifically, if S2 was intact, T2 was shorter whenS1 was also intact than when 51 was degraded . whereasthe opposite was true if S2 was degraded (i e . T2 waslonger when S1 was intact than when Sl was degraded) .Because this effect reflects an advantage of processing ho-mogeneous stimulus pairs, it will be denoted here as thehomogeneous-ctimuluc effect (HSE) . In line with 'Vicker-son's (1975) account, Hanscn and Sanders assumed thattheir HSE reflected a tore efficient matching process mthe case of identical stimulus representations On closeinspection, the data do not fully support this znterpreta-tion, however . First, HanSen aid sanders used three dif-ferent noise patterns for each degraded digit, so it is doubt-ful w hether, in a physical sense, the majority of degraded.some pairs was any more similar than same pairs in whichone of the stimuli was degraded and the other intact . Sec-ond, the EiSE was, although attenuated relative to sometrials, still present on different trials, in which, apart fromthe digits, the more patterns were always different . Neitherresult is predicted by a matching account, which heavilyrelies on the identity of the stimuli .At first sight, it seems that an account of Hansen and

Sanders's (1988) data m terns ofa recapitulation of encod-ing processes encounters the same problems as does theaccount based on representations . However, encodingrecapitulation makes it easier to byPasS the problem ofstimulus identity by postulating a process that operateson the noise in a more abstract way, independent of itsspecific physical structure . Thus, Van Duren and Sanders(1988) suggested that a degraded stimulus needs a pro-cess that separates relevant from irrelevant features forits encoding, whereas a fast route, bypassing this process,may be used for encoding an intact stimulus . Conse-quently, by processing stimulus pairs of the same visualquality, a processing route is opened by S 1, which mayfacilitate the processing of S2 along the same route . Bycontrast, processing stimuli of different visual qualitiesrequires a less efficient, and possibly time-consumingswitch of processing route . The important feature of thisexplanation is that it does not exclusively rely on the spe-cific shape of the stimuli, but rather on their molar struc-ture, demanding a specific perceptual analysis .The proposed account for the HSE in terms of recapit-

ulation of processes relates well to the procedural viewof the human mind, which has an important status m re-search on implicit memory . The basic claim ofthis viewis that our knowledge can be better described in tenorof actions than in terms of static representations . such aspropositions (Kniere & RocdiQer, 1984) Knowledge, itis argued . is not a structure of mere results that is stored

in memory, but is inseparable from the way we obtainedit (Kolers, 1976 : Kolen & Roedicrer, 1984) . A mainsource ofevidence for this view relies on the dependenceof performance on specific stimulus conditions that arevery similar to the HSE described here . For instance,Roediper and Slaxton (1987, Experiment 3) had subjectsstudy words that were presented either clearly focusedor moderately blurred . In a test episode, the subjects' taskwas so complete words from which same letters wereomitted (e .g ., -YS-E-Y for misten'), again presentedeither blurred or clearly focused . Half of the test wordswere studied previously and, ofthese, half were presentedin the same stimulus quality a% in the study episode, whilethe other half was presented in the alternate stimulus qual-ity . The data showed a clear advantage of studied wordsover nonstudied words and, for the studied words, aninteraction between the effects of stimulus quality m thestudy and test episodes- Words studied m clear focus werebetter completed when they recurred in the test episodein clear focus rather than blurred, whereas the oppositewas true for the blurred words . This HSE was regardedas evidence for a closer recapitulation of encoding pro-cesses in the case of identical stimulus qualities in the studyand test episodes .

Research Questions and Experimental ApproachThree questions are addressed in the present study . The

first and foremost question is whether evidence can befound for a procedural account for the HSE reported byHansen and Sanders (1988) . The experimental approachto this question basically relies on depriving alternativerepresentation accounts from any possible basis, so as toarrive at a procedural account by exclusion . For this pur-pose, Hansen and Sanders's four stimulus conditions wereused, but the experimental design was adapted so that cueoverlap of actions with respect to S l and 52 was restricted,as much as possible, to encoding processes To achievethis, three measures were taken . First, following Hansenand Sanders, the functional visual field paradigm (San-ders, 1963) was employed . This experimental paradigmrequires different responses to 51 and S2-a saccade anda manual keypress, respectively . This precludes a prioriany overlap of response or response-selection processes,and thus any tendency of the subject to repeat a previousresponse (see also Krueger & Shapiro, 1981a) . Second,instead of a matching task, a task that requires indepen-dent responses to S1 and 52 was used, reducing the pos-sibality that matching would be the focus of the HSE .Third . the possibility of matching was further reduced inthe first two experiments, in which S 1 and 52 representedboth different digits and noise patterns, rendering stimu-lus quality as the only dimension of possible agreement.Only m Experiment 3 were conditions added in which 5Iand S2 represented identical digits and noise patterns, inorder to establish whether stimulus idenGtv would furtheradd to the }ASE .The second question is whether the HSE as a procedural

phenomenon relics on a data-driven or a strategic mech-anism The proposed procedural account relics on a data-

PROCESSING INTACT AND DEGRADED STIMULI 147

driven mechanism, in that :t assumes that the HSE on T2is caused by the act of processing S1 . That is, subjectsare fast in pra:c5timg 52 of homogeneous stimulus pairs(or, alternatively, slow in processing S2 of heterogene-ous stimulus pairs) because they engaged in similar(respectively . dissimilar) processing activity for the ben-efit of S I . In this respect, the mechanism differs from theoriginal proposal of Van Duren and Sander (1988),which is strategic m that it assumes a preparatory activityon the part of the subject . The preparation implies the ad-vance selection ofa processing route for the "worst case"expected on a trial . According to this principle . the slowroute is selected in advance whenever a degraded stimu-lus on a trial is expected, whereas the fast route is onlyselected when both S1 and S2 arc expected to be intact .Thus, the processing of intact stimuli of heterogeneouspairs evolves relatively slowly, because it takes place (inpart) along the suboptimal5law route . In Hansen and San-derti's (1988) Study, such a preparatory mechanism mighthave played a role, because the specific 51-52 pairs werepresented in pure blocks of trials, which allowed their sub-jects to select the fast processing route in advance of atrial in the intact-intact condition . In order to establishthe contribution of this strategic mechanism to the H5E,a "blocked" and a "mixed" presentation of intact anddegraded Sts were compared in Experiment I . The ab-sence of advance knowledge on the forthcoming S2 in themixed condition should, according to the worst case prin-ciple, provoke the selection of the slow route in and con-dition . Consequently, the HSE should at least be reducedin the mixed condition relative to the blocked conditionif it relics on the strategic principle proposed by VanDuren and Sanders, but not if it relies ors a data-drivenprinciple .The third and final question is whether the HSE is

facilitating or inhibiting in nature . HanSCn and Sanders(1988} tacitly assumed facilitation to he the driving forcebehind their effect, but inhibition constitutes a perfectlyequivalent alternative . That is, instead of being advanta-geous to repeat a same route, it could very well be disad-vantageous to have to switch from one route to another .The problem of deeding between facilitation and inhibi-tion is that it depends on a control condition, whose neu-trality is not easy to defend (e .g ., Jonides & Mack . 19&4) .In the present study, a control condition was used m whichan auditory S1 was combined whit an intact or dcQradedS2 . The virtues of this control condition are threefold .First, it allows the employment of essentially the tiametask as in the experimental conditions (which is not thecase if 52 is presented m isolation) . Second . the infor-mation value of the tone can be easily equated to that ofa visual S1 by varying the pitch of the uanat . Third, itavoids the enoaaement into visual processing before S2i5 presented, which is obvioush desirable given the pur-pose of the present study . There arc, howcver, at leastt+;o possible objections to this control condition, First .contrary to ~ ~, itiual S I , an auditnrv S I may tend to alertsuhlectti (Niemt, 1974, Sanders . 1977) . thus shortening

the subsequent T2 . Second, it is unclear whether costsare involved in changing from auditory to visual process-mg, thus prolonging tine subsequent T2 . In the GeneralDiscussion, the seriousness of these objections will becommented upon, and an evaluation of the neutrality ofthe control condition will take place in the light of thedata obtained . For the present, however, the control con-dition will be referred to as auditory, rather than neutral .

EXPERIMENT]

Experiment I was an attempt to acquire insight into thethree research questions introduced above . First, it wasa test of whether an HSE would occur if stimulus qualityis the onlv common factor between S1 and S2 . Second,it was an attempt to establish the strategic contributionto the HSE by comparing conditions in which intact anddegraded Sts were presented in either pure or mixedblocks of trials . Third, it was an attempt to establishwhether the HSE is due to facilitation or inhibition, bycomparing it with the auditory conditions .

:MethodSubjects . The subjects were 24 students, 12 is other of two

groups Six subjects of either group were female All the sub-jects had normal or corrected-to-normal vision They were paid fortheir servicesApparatus The experiment took place ma sound-attenuated, air

conditioned, dimly illuminated cubicle . The subject was seated mfront of a table, nn which a semicircular, 50-em-high black screenwas mounted, which was enclosed by a black ceiling and bottom .The subject had hislher head on a chinrest, positioned 10 cm abovethe bottom and at a distance of 70 cm from the screen . The screenhad two windows at eye level and at either side of the visual mend-tan, subtending a bmocuLar visua[ field of45° The windows werecovered with translucent paper, on which the visual stimuli were pro-jected from behind by means ofa Kodak carousel prolutor Auditorystimuli were presented via padded earphones, which the subject woreduring an experimental session . Horizontal saccadic eve movementswere registered by AgCI electrodes, yielding a standard electro-oculogram (EOG), which was sampled with a frequency of250 Hz .On the table, 20 cm below the bottom of the screen, d responsepanel was mounted m front of the subject . The panel consisted oftwo keys, on top ofwhich the subject rested the left and right indexfinger, An Olivetti M280 microcomputer controlled the timing ofevents and stored the sampled EUG data and reaction timesStimuli Audrton and visual stimuli were used . The auditory stim-

uh were binaural 70-dBA tones of ether low (500 Hz) or high(10(10 Hz) frequency The visual stimuli were very similar to theones used by Hansen and Sander-, (1988), and consisted of intactand degraded qualities of the alerts 2, 3, 4, and 5 (see examplesmFigure 1) . The digits (5 0 x 3~5 cm) were surrounded by a rect-angular frame measuring 7 5 x 6 5 cm, and constituting a visualangle of ';.3' wide Both the digit end the frame convsteci of adja-cent w bite dots (7 mm diameter, ? cd'm2 luminance) against a blackbackground (0 05 cd'm2 luminance) 1n the intact form . the digitconsisted of 14 dots and the frame of 36 dots A degraded formwas a deformation of the intact loon, m which 10 dots from tieframe were p~eudorandomk distributed around the digit within [heframe, Ihuti securing, a constant illumination of the stimulus thatwas independent of its form The degraded digit remained unam-bi,uousl,, identifiable in that no conspicuous alternative algaemerged from a combination of the digit at issue and arc dote of

148 LOS

1

00" 38 *" fte

00 0"n 00801so" 30~ I V : :21 ~ 0: L 0:

r em n r *"

6,

S2

I.I L .4na0

Figure 1 . An overview ofthe visual stimulus conditions. The stim-uli are depicted as negatives of the authentic stimuli . The stimulipresented first (S I) are depicted in the upper row, the subsequentpresented stimuli (S2) in the bottom row. The stimulus conditionsare constituted by the stimulus qualities of Sl and S2 . The condi-tion name is below each S1-S2 pair; abbreviations in parenthesesare used throughout the text . Note that the digits and noise patternsrepresented by S1 and S2 are always different .

the noise pattern . There were four versions of each degraded digit,so as to prevent pattern-specific encoding . Noise patterns were alsodifferent between the digits .

Task . A trial of the experimental task started with the onset ofa light-emitting diode (LED) underneath the left window for 1 sec,summoning the subject to fixate the center of that window . An in-terstimulus interval of 1.5 sec separated the offset ofthe LEDfromthe presentation of a visual stimulus (S 1) on the left window . S1was either an intact or a degraded digit, whose value indicatedwhether fixation on S1 should be continued ("no go") or a quicksaccade should be made to the right window ("go") . In case ofa go response, the presentation of S2 was triggered by the onsetof the saccadic eye movement . As has been shown by Houtmansand Sanders (1983), subjects are unable to do any processing onS2 during the saccadic eye movements. S2 was also either an in-tact or a degraded digit, whose value determined the response keyto bepressed. The offset of both stimuli took place 4 sec after theonset of S1 . The next trial started 4 sec after the offset of the stim-uli. A trial of the auditory control task was identical to a trial ofthe experimental task, except for one point. The stimulus that waspresented while the subjects fixated the left window was not a digit,but a tone presented for 100 msec, whose pitch determined the go/no-go saccadic response .The assignment of digits and pitch to the responses requested by

S1 and S2 was as follows . The total set of digits was subdividedinto two sets : {2,3} and {4,5}. For half ofthe subjects the set {2,3}was used as S1 and {4,5} as S2, whereas for the other half of thesubjects this was reversed . Again, for halfof the subjects of eachgroup, the lower digitat S1 indicated the go response and the higherone the no-go response, whereas this was reversed for the otherhalf of the subjects . At S2, the lower digit was always assignedto the left response key and the higher digit to the right one, sincethis proved to be the compatible stimulus-response setting . The as-signment of pitch in the control task to the go and no-go responsewas compatible with the subject's digit condition in the experimentaltask, with a go response to the lower digit yoked to a go responseto the low-pitched tone, and vice versa.Design and Procedure . The independent combination of the

visual qualities of 5l and S2 yielded four experimental conditions :intact-Intact, intact-degraded, degraded-intact, and degraded-degraded . An example of each condition is depicted in Figure 1 .In the control conditions, S2 was preceded by a tone, yielding tone-intact and tone-degraded pairs. One group of subjects, the "blocked"group, received the total of six conditions in six separate blocks,

while the other "mixed" group received intact and degraded Stsmixed within blocks . This left three separate blocks, determinedby the quality of S1 (being intact, degraded, or a tone). Blocks werepresented in a Latin-square arrangement in both groups . There werethree dependent variables : (I) T1, the time between the presenta-tion of S1 and the initiation of the saccade, (2) Tmov, the time be-tween the initiation and the completion ofthe saccade, and (3) T2,the time between the completion of the saccade and pressing theresponse key .The subjects were tested for 3 h in either in the morning or the

afternoon. They were instructed not to move their heads during theexecution of the task . Speed and accuracy ofresponses were equallystressed, both with respect to responding to S1 and S2 . All prac-tice and experimental blocks contained 34 trials . Each trial had anequal probability of a go or no-go response at S1 and of a rightor left response at S2 and, for the mixed group, an equal probabil-ity of an intact or a degraded S2 . Furthermore, for the degradeddigits, each noise pattern had an equal chance of occurrence, ashad the specific combinations of noise patterns of Sl-S2 pairs inthe degraded-degraded condition . The subjects received one prac-tice block in all experimental and control conditions, after whichthese conditions were presented for measurement The blockedgroup received two series of six blocks in each S1-S2 condition,and the mixed group received three series of three blocks in eachS I condition.' After a block, the subjects were given feedback onerror performance and information about the stimulus condition ofthe next block. Breaks of about 5 min were inserted after trainingand after each series of blocks .Data analysis . The on-line calculation of the initiation and comple-

tion times of saccades was carried out as follows . First, the point intime was established with the maximal tangent to the EOG samplevalues . Then, starting from this point, the first points back and forthin time of which the tangent approached zero were determined andregarded as the saccade initiation and completion times, respectively .Incorrect scorings due to eventual artifacts in the EOGsignal werecorrected afterward by the experimenter . This procedure yieldedthe points in time necessary for the calculation of TI, Tmov, and T2 .The resulting T1, Tmov, and T2 were pooled across series, after

which means were calculated for each specific SI-S2 condition.The first two trials of each block and trials in which T1, Tmov,or T2 deviated more than two standard deviations from the meanwere excluded from analysis . Error trials with respect to S1 (sac-cade initiation on a no-go trial) or S2 (pressing the wrong responsekey) were separately analyzed .Separate univariate analyses of variance (ANOVAs) were car-

ried out on mean TI, Tmov, and T2, as well as on mean error per-formance with respect to S 1 and S2 . These analyses were carriedout on the data resulting from the experimental conditions only .S1 quality and S2 quality were within-subjects factors, whereas S2presentation (blocked vs . mixed) and digit assignment ({2,3}/{4,5}vs . {4,5}/{2,3}) were between-subjects factors . An additional anal-ysis was carried out with respect to T2, in which the results fromthe main experimental conditions were contrasted with those fromthe auditory conditions .

ResultsIn this section, the results regarding T2 are presented

first, since they are ofmain interest for the present study .Then the results regarding T1 and Tmov are presented,followed by the data on error performance . Furthermore,to allow an easy reference to specific stimulus conditions,the following abbreviations are used throughout this sec-tion : I and D for intact and degraded digits, respectively,and A for the auditory signal . Pairs of these abbrevia-tions denote the specific stimulus conditions . For instance,DI indicates that S1 was a degraded digit and S2 was an

PROCESSING (N~IACTA:VD DEGRADED STIMULI 149

'BLOCKED MIXED'440

Ayap

g420-

400-

---

- ~ ~

420-

53S0y

, ~ ~

400-

X380y

F360- .f

F3B0-

340- ,. . .

-

"" 340

320- ° - ' - ~ , 32P-'-

~II ~ ~ q ~ 1p OD fl ~ q ~ ~ ID ~ DD

STIMULUS QUALITY STIMULUS QUALITY

dso C - ----490 p

ago- .470-

450- 450,

W430- T 430

F610 , F410^

390-~-" 39p- .

370-

i /

nrA

Al ~ ro DID f w °~n of Al io 00 noSTIMULUS OVqLFTY STIMULUS OUALITV

Figure 2. Mean response times in Experiment 1 asa function of Sl quality, 52 qual-ity, and 52 presentation (blocked, panels AandC: mixed, panels B and D) . TI is theresponse time to Sl, and '1'2 is the response tame to 52 . I = intact stimuli; D - degradedstimuli; A = auditory stimuli .

intact digit, AD indicates that 51 was a tone and S2 wasa degraded digit, and so on .

Figures 2C and 21) show the mean T2 ati a function ofSI and S2 quality for the blocked and mixed groups,respectively The A'.VOVA repealed a main effect of S2quaGty[F(1 .20) -- 242.24, MS, = 338.58,p C .00E],indicating that responding to an intact 52 was faster thanto a degraded S2 . Thiti effect interacted with S1 qualityEF(E .20) = 10.87 . MS, = 304.81 . p c .{)I], reflectingthe HSE . There was no indication that the HSE dependedon the presentation ofintact and degraded Sts m pure ormixed blocks of trials, as testified by an absence of theS1 qaalit_y x S2 quality X 52 presentation interactionjF(1 .20) -- 0 .33, MSS = 3(14 .8I, p = .57] Planned com-parisons revealed that the response to an intact S2 wasfaster when it followed an intact S 1 than when it followeda dcuradcd 51 ; that is . T2(H) C 'P,01) [F(L ,20) = I2 .64,MS, = 367.65 . p C O11, The corresponding tendencyto respond taster to a degraded S2 when preceded by adegraded S I than when preceded b~ an intact S I way notsignificant [F(I .2O) -- 2 .18, MS, = L,012.25, p = .16]Finally . some significant effects related to the grouping

factor "digit assignment" were observed, but since theseeffects did not seem to affect the conclusions with respectto the HSE, :hey are ignored here and in the followingexperiments .An additional analysis contrasted the 14SE on T2 with

the auditory control conditions, in which a tone precededthe intact or degraded 52 . This analysts revealed that theAD-.4I difference did not significantly differ from the ID-3f difference IF(1 .20) = 0.95 . MS, = 1,191, p = .34] .whereas the AD-Al difference was significantly largerthan the DD-BI difference IF(1,20) = 14 .6 . MS, --1,515_26, p c .0011 . Thus, compared with the auditoryconditions. the pretient H5E i5 a manifestation ofdeviantbehavior following a degraded 5I This was specified inTreater detail by simple contrasts, which rep Baled that theTZs of the S I -audrtorN and the S ?-intact conditions werealso absolutely comparable ; that is, nether T2(AI) andT2(IT) [F(1,20) = 1 .37 . MS, -- 589 .64, p = .2(i] norT2(AD) and T2(ID) [F( 1,20) = 0 02, M5~ = 1,24'. .60,p = .881 differed cionificantly . BN contrast_ T2(DI) wassignificantly longer than T2(Al) [F(1,20) = 337 .5b, .WS, -248.18, p < ()()l], whereas T2(DD) tended to he some-

150 LOS

what shorter than ~I'2(AD) [F{ 1,20) = 1 .82, MS, -- 1,501 .14]

Figures 2A and 2$ show the mean 1'] from the ex-perimentdl conditions as a function ofS 1 and 52 quality .The A10VA repealed a significant main effect of S1 qual-itv [F(1,20) -- 65.91, MS, = 1,677 54,p C 001 J . in-dicating a saccade initiation that was faster lollowing anintact S1 than following a degraded S1 . Furthermore, aninteraction between the effects of 52 quality and S2 pre-sentation [F(1,20) -- 5 12, MS, = 309 .20, p C .OS] in-dicates that, m the blocked group, the saccade was initi-ated faster when an intact 52 followed Sl than when adegraded S2 followed S1, whereas this difference was ab-sent in the mixed group .The average Tmov amounted to 338 msec . Tmo,~ was

significantly affected by S2 quality [F(1 .20) - 7.01,MSe = 3.15 . p C .05] . althouah the saccade precedingan intact S2 was only 1 msec faster than the saccade pre-ceding a degraded S2 . This effect occurred only m theblocked group {a 2-rxtsec effect}, and not m the mixedgroup, as indicated by an Interaction between the effectsofS2 quality and S2 presentation [F(1,20) -- 7 .63, MS, =3.15 . p c OS] . Generally . differences on Tmov are read-ily significant due to a very high consistency among sub-jects . but they typically do not exceed z msec and thusare not considered to be of interest for the present study .Table 1 shows the error percentages that occurred in

the experimental and control conditions of the blocked andmixed groups . With respect to S1, the ANOVA did notreveal any significant differences between conditions .With respect to S2, the ANOVA revealed a main effectof 52 quality [F(1,20)- 8.26 . MS, = 12 .09, p C .O1],indicating that fewer errors were made when 52 was intactthan when S2 was degraded (I 75 % vs 3 78 % , respec-tively) . This effect depended on the quality ofS 1 (F(1,20)= 758, MS, = 4.35, p C .051, constituting an HSE onerror performance, which m turn depended on S2 presen-tation IF(l,20) = 4.58, MSe = 4.35, p e .OS] . The latterthree-way interaction indicates that the HSE on error per-formance was prctient in the mixed group, m that fewererrors occurred in the homogeneous conditions than in

Table i:14ean error Percentages in Responding to Sl and S2 in theExperimental and Control Conditions of Experiments 1 and 2

Expenmental ControlII D1 FD DD 41 AD

Experimrrt I BlockedS1 3.39 365 3 13 495 6 77 7 29S2 2 08 2 08 3 65 3 13 2 08 3.91

Expenment I Mixed51 3 3U 2 60 ? 30 260 295 2 95S2 0 35 2 43 5 21 3 13 2 48 ? 1?

Fxpcnmcnt 2

51 ? ?9 1 56 1 56 3 65 469 5 99S2 2.09 1 04 2 8r 3 39 ? 13 469

\otc-I = intact stimuli Lf = degraded sumult . A - audito+\ sumuh

the heterogeneous conditions (1 .74% vs . 3 .82%, re5pcc-tively), but not in the blocked group (2 .61 % vs 2.87%,respectively) Finally, simple contrasts ofthe experimentalcondctions with auditory conditions did not yield any sig-nificant difference,,. [n `um, the data on error performanceshow- that the HSE, as found on T2, was not an artifactdue to differences in speed-accuracy tradeoff betweenconditions. because the "faster" conditions tended to pro-duce fewer errors .

DiscussionThe morn result of Experiment 1 was the replication

of Hansen and 5anders's (1988) HSE . This effect indi-cates that T2 was tihorter when 52 had the same qualityas the preceding 51 than when 52 had a different qualityfrom the preceding S 1 . In the present study, matching wasexcluded as an explanation for this effect, because S1 andS2 required independent responses and always representeddifferent digits and noise patterns . Thus, an explanationshould proceed from the specific processing demands ofintact and degraded stimuli .Apart from this main result . two other results are of

importance First, by mixing intact and degraded Sts, theHSE was divided across T2 and error distributions ratherthan reduced . A reduction was expected if the HSE re-lied on a preparatory mechanism to deal with the worstcase expected on a teal, because absence of knowledgeon the forthcoming S2 provokes the selection of the slowproccti5ing route in all stimulus conditions . Effects thatare possibly related to this worst case principle occurred,m fact, in the blocked group, where saccadic eye move-ments were initiated and executed faster when in antici-pation of an intact 52 than when in anticipation of a de-graded S2 . However, the data of the mixed group showthat these effects can be dissociated from the HSE on T2 .Thus, although advance knowledge on the forthcomingS2 hack several strategic effects on prior events, it did notcause the H5E on T2_ The second important result is that,compared with the auditory conditions, the present HSEmanifested itself neither as a facilitating effect nor as aninhibiting effect. Instead, the auditory conditions mimickedthe results of the S I -intact conditions with respect to T2 .The implications ofthis finding will be further elaboratedin the General Discussion .

EXPERIMENT 2

The HSE m Experiment I was quite small . Because ofthis, it was difficult to acquire insight into the effect sizeofits various components relative to each other and rela-tive to the auditory conditions . Experiment 2, then, wasan attempt to magnify the HSE by increasing the process-ing demands of degraded stimuli . For this purpose, de-graded stimuli were presented 180° rotated, so as to in-crease the difference in processing demands between intactand degraded stimuli . This manipulation added someSO msec to the reaction time to degraded, upright stimuliin a study by Sanders and Rath (1991), who used the digits

PROCESSING INTACT AND DEGRADED STIMULI 151

2-7 for S1 and S2 in a matching task . An interesting re-suit oftheir study was that T2 to an intact S2 was sigmfi-cantlv shorter when S1 wati degraded than when S1 wasboth de,raded and rotated . Referring to Hansen andSanders (1488), Sanders and Rath attributed this effectto differences m matching times, but a procedural accountagain poses a viable alternative .

MethodThe method was the same as that used for the blocked condition

of Experiment 1, except for two points- First, there was a changem Stimulus material with respect to the degraded stimuli . whichwere presented ISO` rotated . viefdin ; degraded and rotated siim-uli intact digits . by contrast, were presented in their normal, up-right position Second, because problems were expected to distin-guishing "2" and "3" in the degraded category, the digit sets werechanged to {2,4} far S I . and {3.5} for S?, fit vice versa S1 and52 quality were the within-subjects factors m the A\OVAs, anddigit assignment ({2,4}!{3 .5}vs. {3,5}1{2,4}) and session (thefirst or second series of blocks) were the between-subjects Furors .Subjects Six male and 6 female students, none of whom had

engaged in the previous experiment, served as paid subjects . Twoof the subjects were unable tea recognize several of the degradeddigit,, after practice, and therefore were replaced by 2 others

ResultsFigure 313 shows the mean T2 as a function of S1 and

S2 quality . The effect of S2 quality was highly signifi-cant [F(1,10) = 372.04, MSe = 847.41, p < .001] andinteracted with the effect of S1 quality [F(! ,10) -- 24.40,MS, -- 803.77 . p < .{)()l] . This interaction reflects, theH5E that was also found in Experiment l . Planned com-parisons revealed that T2 was shorter when the stimuliwere homogeneous than when they were not . Thai i5, ashorter T2 was found at II than at DI JF'(1,10) = 27 .54,MS, = l,SR1 .95, p C .001], and a shorter TZ at DDthan at ID [F(1,14) = 6.65, MS, = 5,277.48, p e .45] .Planned comparisons to the auditory conditions replicatedthe main result of Experiment 1, m that the effects of theauditory conditions and the 51-intact conditions were addi-tive J,4D-AI -- ID-II. F(1, 10) = 0.60 . MS, = 11.248 .95,p = .46], whereas the effects of the auditory conditions

and the S 1-degraded conditions interacted [AD-AI >DD-DI, F'(1 .10) = 26.49, MS, = 3,716 88.p < .001] .Furthermore, simple contrasts between the componentso£ the HSE and auditory conditions also yielded resultsthat were very similar to those obtained in Experiment 1 .That is, T2 at AI was not different from T2 at II [F(I, IO) _0.39, MS, _ 1,302 47, p = .55], and T2 at AD was notdifferent from T2 at ID [F(1,10) = 0.47, MS, -- 7.571 .62,p - 51] By, contrast . T2 at DI was longer than at AI[F( l , lfl) = 17.84, MS, = 1,943-02, p < .01), whereasT2 at DD was shorter than at AD [F(1,10) = 5 .39, MS,= 3,023 .27 . p < .OS] .Figure 3A Shows T1 resulting from the experimental

conditions as a function of 51 and S2 quality A main ef-fect of SI quality [F(1,10) = 43 .39, MS, -- 3,500.04,p C .0011 reflected faster responding to intact stimulithan to degraded stimuli . Mast importantly however, incontrast with the blocked condition of Experiment 1, therewas neither a main effect of 52 quality [F(1,10) = 0.0(),MS, = 1,I99.94], nor an interaction between the effectsof S 1 and 52 quality [F(1,10) = 0.23, MS, = 499.75]-Also . there were some significant effects involving thefactors of session and digit assignment . However, thesevariables did not affect the findings that are ofimportanceto the present study, so their effects will be ignored . Fi-nally, Tmov (M = 739 cosec) did not vary significantlyamong conditions .

Error percentages (Table 1) remained well below 5% .Slightly mare errors occurred in the homogeneous thanm the heterogeneous stimulus conditions, but the ANOVAsnn error performance with respect to S 1 and S2 did notyield any significant difference between these means, andthe simple contrast between the experimental and audi-tory conditions did not result in significance .

DiscussionExperiment 2 proved to be successful as an attempt to

magnify the ASE by increasing tie differences in process-ing demands between intact and degraded stimuli . Thesizable HSE on T2 again indicated that the response to

440- -- -

A4zo-

~, . .. .._ ; _ ;

aoo-

W 3B0-

f- 360

""_"""'" `` o~\~a

L~ \ ~'`~\1 `

320- t ~ "

ip ._

H o~ ID 00STIMULUS QUALITY

550- g -

soo-

~\\\

'r\

u o1 a, ro DD ADSTIMULUS QUALITY

Figure 3. Mean response limes in Experiment 2 a5 a function of S1 and 52 quality .'fl is the response time to Cl, and T2 is the response time to 52 . I = intact %tim-uli; D = degraded stimuli, A -- auditor stimuli .

152 LOS

S2 was faster when stimulus pairs were homogeneous thanwhen they were not . This result replicated the main re-sult of Experiment 1 . Another excellent replication con-cerned the observation that a tone and an intact S 1 yieldedidentical results with respect to T2 . Thus, relative to theauditory conditions, it was again a degraded S 1 that some-how caused deviant behavior on T2 . Furthermore, thedegraded manipulation proved to be strong enough todemonstrate significant differences between all separatehomogeneous and heterogeneous components, as well asbetween the components of the auditory and S 1-degradedconditions . One result deviating from Experiment 1 wasthat, although the stimulus conditions were presented inpure blocks of trials and the subjects had advance knowl-edge of the qualities of the forthcoming S1 and S2, noeffects of S2 quality were observed on either T1 or Tmov.Thus, in line with the data of the mixed condition of Ex-periment 1, Experiment 2 provided another case in whichdissociation occurred between strategic effects on T1 andTmov on the one hand and the HSE on T2 on the other,although it is not clear what exactly caused this deviantresult from the blocked condition of Experiment 1 . In all,Experiment 2 provided a useful replication of the data ofExperiment 1 in that it confirmed the main findings, whileit shed more light on the relations of the separate compo-nents of the HSE relative to each other and relative tothe auditory conditions .

EXPERIMENT 3

The primary aim of the experiments conducted so farhas been to demonstrate that the HSE is not dependenton the presence or activation of representations . This wasachieved by restricting the overlap between S 1 and S2 tothe level of stimulus quality, implying that the digits andthe specific noise patterns of S1 and S2 were always dif-ferent . Experiment 3 was an examination ofwhether same-ness of digits and noise patterns would further add to theHSE, as was the case in Hansen and Sanders's (1988)study . For this purpose, nominally sauce and differentdigits were used for the S1-S2 pairs, constituting twolevels of the variable digit correspondence . The differentcondition was exclusively used in the previous experi-ments and served in Experiment 3 as a baseline for theevaluation of the effects resulting from thesame condition .As far as the stimulus conditions are concerned, Ex-

periment 3 comes very close to Hansen and Sanders's(1988) experiment described in the introduction, but thepresent task is different in that matching is not required,because S1 and S2 require independent responses. Thus,following Proctor's (1981) argument, if the main find-ings of Hansen and Sanders can be replicated, a proceduralaccount should be favored over a representational one .

MethodRelative to she preceding experiments, the following adjustments

were made . Regarding S1, the digit set {2,3,5,6,8,9} was used,with the subset {2,3,5,6} requiring a go response and the subset{8,9} requiring a no-go response, for all the subjects . Regarding

S2, the digit set was identical to the set requiring a go responseat S1, that is, the set {2,3,5,6} . Of this set, the subset {2,3} re-quired a left manual response and the subset {5,6} required a rightmanual response . Note that by leaving out the digits 4 and 7, iso-lated groups of digits were created, so that the decision processesthat they give rise to would be facilitated .In addition to the variables of stimulus quality employed in the

previous experiments, digit correspondence was added as an in-dependent variable, comprising a same and a different condition.In the same condition, the digits of S 1 and S2 were identical, whereasthe digits were different in the different condition. Thus, the samecondition contained the S1-S2 pairs 2-2, 3-3, 5-5, and 6-6 in allconditions of stimulus quality, whereas the different condition con-tained all the other possible pairs.

Each experimental block contained SO trials, of which the first2 were discarded. The48 remaining trials were composed such thatall individual stimuli had an equal probability of presentation, withone exception. Given that in the degraded-degraded condition S1and S2 were same (an event with a probability of .25), the proba-bility that their noise patterns were identical was .5 instead of .25,which would have been the expected probability because four dif-ferent noise patterns were used for each digit. The reason for thisdeviation was to obtain equal numbers of trials in which same stimulihad identical noise patterns and in which they had different noisepatterns, so as to enable a fair comparison between these specificsame conditions . Thus, each block contained a random presenta-tion of (1/s x 48) 16 no-go trials and (% x 48) 32 go trials . Thego trials comprised (3/a x 32) 24 different trials and ('/4 x 32) 8same trials . In the condition with two degraded stimuli ('/z x 8),four same digits had identical noise patterns, whereas the other ('/zx 8) four same stimuli had different noise patterns .Auditory conditions were omitted, because the previous experi-

ments demonstrated that conditions with an intact S I yielded iden-tical results, as far as T2 is concerned. Thus, there were eight ex-perimental conditions, constituted by an independent combinationof Sl quality (2), S2 quality (2), and digit correspondence (2). Thevisual qualities of Si and S2 were varied between blocks, as in theblocked condition of Experiment 1, whereas digit correspondencenaturally occurred within blocks . Four separate blocks, constitutedby the quality of S1 and S2, were presented four times each in aLatin-square arrangement. Means of the pooled data ofeach stimuluscondition were calculated and subsequently entered into ANOVAswith S1 quality, S2 quality, and digit correspondence (same/different)as independent factors .Subjects . Twelve new subjects (6 females/6 males) were tested

during about 4 h and paid for their services . One subject was re-placed by another, because of unacceptable error rates.

ResultsAlthough in the DD condition same pairs with identi-

cal noise patterns had a shorter T2 than same pairs withdifferent noise patterns (496 vs . 512 msec), a preliminaryt test showed that this difference was not significant[t(11) = 1 .41, p = 28], due to a strong variability be-tween subjects . For this reason, the data of these condi-tions were collapsed and treated as a single same factor .Figures 4B and 4C show the mean T2 as a function of

S1 and S2 quality for different and same stimuli, respec-tively . The ANOVA revealed a significant main effect ofS2 quality [F(1,11) = 65 .89, MS, = 1,216 .5l, p < .001],which interacted with the effect of S 1 qual ity [F(1,11) =12 .93, MS, = 450.66, p < .O1], reflecting the HSE. TheHSE depended, in turn, on digit correspondence [F(1,11)= 7.29, MS, -- 208 .44, p < .OS] . This effect reflectsa more robust HSE in the case of same stimuli, although

PROCESSING INTACT AND DEGRADED STIMULI 153

asoA 540- B

430- 520

Ii .-500

V410- ~~jy480

3 I

,7390 j. 460. .

440_

.

370- ~~ -420-

.

II DI

lo. .

.Do

STIMULUS QUALIT Y

different same

540- (,'

s2o-

500-W

NS80-

ZF460-~_

44p- .

420-

. . . __ . . 9 . . .~.~._ - . .DI ID DO II p1 10 Up

STIMULUS QUALITY STIMULUS QUALITY

Figure 4, Mean response times in Experiment 3 as a function a£ SL quality, 52 quality, and digit correspondence(same or different) . TI is the response time to 51, and T2 is the response time to 52 . I = intact stimuli, Q -- degradedstimuli .

a planned comparison showed that the HSE was stillpresent in the case of different stimuli [F(1,1 I) = 4.87,MS, = 573 .11, p = .OS], corroborating the results oftheprevious two experiments . In farther agreement with Ex-periment i , comparisons between the components of theHSE in the different condition yielded a longer T2 at DIthan at II [F(1,11) = 6.85, MS, - 172.27, p C .OS],while T2 at DD was not significantly shorter than at ID[F(1,11) = 1 .27, MSS = 268 .42, p - .2$] . This patternwas reversed on same trials, in which responding at DIwas not significantly slower than at II [F(t,l I) = 1 .68,MS, = 938.44, p = .23], but responding at DD was fasterthan at ID [F(1,11) = 13 .71, MS, = 1,113.24,p C .O1] .The ANOVA showed, furthermore, that the faster responseto same stimuli than to different stimuli was no more than atrend [F(3 , 11) = 3.48, MS, = 1 ,692 .33, p = .{?9] . Plannedcontrasts showed that this was mainly due to the absenceof any effect of digit correspondence at ID [F(1,11) _0.01 . MS, = 2,094.36], which was . interestingly, alsothe case in the study by Hansen and Sanders ([988) . Inall other conditions of stimulus quality, there was at leasta trend toward a faster response to same stimuli than todifferent stimuli at II [F(1 . 11) = 3.97, MS, = 478.36,p = .071, DI [F(1,11) = 4 .00, MSr = 815 91, p = .07],and DD I F(3,11) = 9 .67, MS, = L062.21, p C .O1] .Figure 4A shows the mean T1 as a function of SI and

52 quality . The only significant effect revealed by theANOVA was the main effect of 51 quality [F(1,1 I) -115.47 . MS, = 409 68, p e .QOI], reflecting lamer re-tipcmding to intact than to degraded digits . The effect sug-gested by Figure 4, that 5 Z of homogeneous stimulus pairs(II and DD) was responded to faster than S1 of heterogene-ous stimulus pairs (ID and DI . respectively), failed to reachsignificance [F( I , 1 1) = 3 .77 . MSS _ 19 1 .11, p = .081 .The mean Tmov amounted to 141 msec and again showed

maximal deviatiom of 2 msec among conditions As inthe blacked condition of Experiment 1, the saccade du-ration was slightly but sianificantlr shorter (1 mcec) when

the subjects anticipated an intact 52 rather than a degraded52 [F(1,11) = 5.69, MS, = 5 .64, p c .05] .

Errorpercentages are presented in Table 2_ The AIVOVAdid not reveal any significant differences in error perfor-mance among conditions for either S 1 or S2 .

DiscussionThree main results of Experiment 3 are better explained

in procedural terms than in terms of matching represen-tations . First, in agreement with the previous two exper-iments, it was found that a small but significant ASE onT2 was brought about by digits of a different value . Asargued previously, the physical dissimilarity of these stim-uli and the absence of matching requirements exclude arepresentational account for this effect . Second, there wasa tendency for T2 to be shorter for same stimuli than fordifferent stimuli, which :s in agreement with Proctor's(1981) original finding that the matching of stimuli is nota requirement to observe the fast-same effect . Althoughsame representation may have a stronger physical resem-blance than different representations, a procedural accountin terms of encoding facilitation is still more obvious thana matching account, because of the inadequacy of match-mg, to arrive at the correct response in the present task .Finally, m agreement with the result found by Hansen andSanders (3988) in a matching task, same stimuli yieldeda more robust HSE than different stimuli . This interactionsuggests that the fast-same effect and the H5E have a com-

'fable 2Mean Error Percentages in Responding to S1 and S2- in Experiment 3

Stimulus Conditionll D! ID

^ ~DD

S1 599 573 4R2 599S. different -'()4 3 56 ~ 47 2 955. same 3.34 3 65 3.65 4 17

Note-f = intact aumuh, D - degraded <rimulr

154 LOS

moa underlying source, which completes a coherent pic-ture outlined by the first two main results . In sum . then .because of the absence of matching requirements in thepresent task, the HS£ and the fast- .came effect, as wellas their interaction, are probably not caused by any mech-am5m based on repre,,cntations, but rather by a mecha-nism based on encoding procedures .A puzzling result of Experiment 3 concerns the absence

of any facilitation an T2 of same stimuli relative to differcnt Stimuli in the condition with an intact S1 and a dc,oruded 52, while this facilitation was, at least as a trend,present in other conditions . This seems to be more thana coincidence, because the sane result was found by Han-sen and Sanders (1988) in their matching task . It is im-portant to note that this absence of a fast-same effectcannot he explained as a lack of correspondence betweensame representations of intact and degraded stimuli, be-cause a strong trend toward a fast-carne effect was ob-served both in a degraded S1 and an intact S2, and to twodegraded digits with different noise patterns . However,a harmonious procedural account for this result is alsonut readily apparent .

GENFRAL DISCUSSION

A31 three experiments showed the HSE, as previouslyfound by Hanserr and Sanders (1988) : The response fvi-lowtng the second stimulus of a pair is faster when thestimuli share the same structural properties than when theydo not The rationale of the design of the present studywas to restrict the overlap of actions with respect to S1and 52 to encoding processes, while excluding the iden-tz of digits and noise patterns . Thus, it is concluded thatthe occurrence of the HSE is not dependent on a processof matching 51 and SZ (Experiments 1-3), or on anymechanism batted on the identity ofthe representations of5I and 52 (Experiments 1-2 : Experiment 3, the differentcondition) . Rather, the mere requirement of encoding 51and S2 is a sufficient condition for the observation of theeffect .

It could be argued that the present HSE was not sym-metrical, in that the fester response to 52 at intact-intactpairs than to S2 at degraded-intact pairs did not alwayshave a significant counterpart at degraded-degraded pairsrelative to intact-degraded pairs (Experiments 1 and 3) .It is, however, not clear how much attention this asym-metry deserves . The observed differences an T2 betweendegraded-degraded pugs and intact-degraded pairs in Ex-periments I and 3 seem too sizable to accept the null hy-pothesis of no effect . Furthermore, it could be that an ad-ditionat hidden factor that broke the symmetry of the HSEwas present . Suppose, for instance, that the task loadvanes with the number of degraded digits presented ona trial and, furthermore, that an increasing task load slowsdown the response to S2 . It follows that `f2 in the de-gradcd-dcaraded condition is affected in opposite direc-tions by two variables-task load and repetition of prmce~,tiin, route-and so the advantage of repeating the slow

processing route should be assessed as stronger than thepresent data suggest,' Whatever the precise mechanism,it is not detrimental to the main conclusion that the inter-action of the effects of S 1 and S2 quality establishes theHSE as a procedural phenomenon .

In the remainder of this section, three more issues areaddressed . the implications of the present findings formatching phenomena, the strategic contribution to theHSE. and the neutrality of the auditory conditions and itsimplications for the nature of the effect .To obtain access to the first issue, a comparison is made

between the present data and those of Hanscn and Sanders(1988), who used stimulus conditions identical to thoseof the present study in a matching task . The comparisonyields a clear analogy of the HSEs of both studies . Innether study was nominal sameness of the digits repre-sented by S1 and 52 a prerequisite for the occurrence ofthe HSE, though it increased the magnitude of the HSEas compared with different stimuli . Furthermore, in bothstudies, the greater robustness of the HSE for nominalsame stimuli was especially die to the absence of a fast-same effect in the intact-degraded condition . It should headmitted, though, that the present HSE remained yuan-titative[v less than the HSE of Hansen and Sanders Al-though the importance of this difference i5 not clear, be-cause it did not proceed from a singe experiment, cautionm interpreting the similarities of both HSEs is m order-It is therefore not warranted to conclude that Hansen andSanders's 145E relics entirely on the same proceduralmechanism proposed for the present HSE . However, itis tempting to suppose that any larger effect is a match-ing task should be additive to the effect observed atpresent If it is assumed that the matching process subse-quently follows the identification of 52, variables foundto affect T2 m an identification task should necessarilyaffect T2 in a matching task in an identical way, althoughthe latter taste allows additional effects to occur duringa possible process of matching representations of S1 andS2 . Several models on matching phenomena, usually re-ferred to as single-process models (Farell . 19&5), in factassume the seriality of identification and matching pro-cesses (e .g ., Krueger, 1478 ; Proctor, 1481), and the ob-served analogy of the present H5E and that of Hansenand Sanders lends further support for this assumption .The second issue readdresses the question of whether

the HSE relies on a data-driven or a strategic mechanism .The first mechanism explains the HSE as the mere con-sequence of engaging in similar or dissimilar processingbehavior before an actual stimulus is processed . By con-trast, strategic accounts assume that advance preparationis the key to the occurrence of the effect . One possiblestrategic mechanism, based an the preparation for theworst case occurring an a trial, could oat be confirmedby the data presently obtained, as already explained inthe discussions to Experiments I and 2 . Preparing for the%tiorst case i5, however, only one out of several possiblestrategic mechanisms that could explain the HSE, and soits discQnfirmaTion does not necessarily imply the correct-

PROCESSING INTACT AND DEGRADED STIMULI 155

ness ofthe alternative data-driven account . In particular,it could be that subjects always prepare for the qualityof the first signal to come, and that it is this preparationthat causes the subsequent HSE. To test this mechanism,subjects should be deprived of their advance knowledgeon the quality of S1, by mixing its intact and degradedforms . In that case, stimulus-specific preparation regard-ing S1 would fail, and so the HSE should disappear ifit relies on the strategic mechanism proposed here, butnot if it relies on a data-driven mechanism .The third issue to be discussed is whether the auditory

conditions could be considered as neutral, and if so, whattheir neutrality implies for the nature of the HSE. As ar-gued in the introduction, there are two challenges to theneutrality of the auditory conditions . First, an auditoryS1 could provide a T2 that is shorter than it should bein an ideal neutral condition, due to its alerting property .Second, an auditory S1 could provide a T2 that is longerthan it should be in an ideal neutral condition, due torefractoriness to switch from an auditory mode of pro-cessing to a visual one . From a theoretical point of view,only the second of these challenges should be seriouslyconsidered, because a great deal of evidence argues againstthe first one . An alerting effect of an auditory stimuluson the response to a visual command stimulus (i .e ., S2in the present study) is only observed under specific con-ditions, none of which are met in the present study . Spe-cifically, (1) the auditory stimulus should have an inten-sity beyond 70 dBA, (2) the preparedness for processingthe visual stimulus should be low, and (3) the relation ofthe visual stimulus to the response should be either sim-ple or highly compatible (Sanders, 1980 ; Van der Molen& Keuss, 1981) . One finding argues, however, againstboth challenges to the neutrality of the auditory condi-tions, namely, the robust finding of Experiments 1 and2 that, for T2, the auditory conditions yield results thatare identical to the S 1-intact conditions . Rejecting the neu-trality of the auditory conditions implies the acceptancethat the equality of T2 in these conditions, relative to theS1-intact conditions, relies on a coincidence . That is,facilitation or inhibition on T2 due to an auditory S1, asimplied by the first and second challenge, respectively,is precisely offset by an equal amount of facilitation orinhibition in switching from processing an intact S1 toa visual S2 . Although the possibility that this is what hap-pens cannot be excluded, the more likely possibility is thatthe auditory conditions are indeed neutral .The acceptance of the auditory conditions as neutral has

an important implication for the nature of the HSE,namely, that this effect is caused by deviant behavior fol-lowing the processing of a degraded S1 . Processing adegraded SI delays the response to an intact SZ and tendsto speed up the response to a degraded S2, whereas pro-cessing an intact S1 does not affect the responding to S2 .In terms of the "alternative routes" model of Van Durenand Sanders (1988), this implies that switching from theslow route to the fast route is offset by costs, while a repe-tition of this route may yield some benefits . By contrast,

using the fast route leaves subsequent processing un-affected, in that there are no costs involved in switchingto the slow route or benefits in repeating the fast route .It seems, then, that using the slow route has the propertyof imposing an inflexibility on the processing system : itstimulates its reuse while blocking alternative routes . Asuggestion as to why the assumed inflexibility is strictlytied to the use of the slow route, but does not occur afterusing the fast route, is that these routes could differ intheir reliance on controlled processing . The fast routecould represent an automatic processing route, which isplausible given that ample evidence exists on the automa-ticity of processing intact familiar symbols (e .g ., LaBerge& Samuels, 1974 ; Logan, 1978, 1988 ; Posner & Snyder,1975a) . By contrast, the additional process that charac-terizes the slow route could be a controlled process, pos-sibly engaged in separating relevant from irrelevant fea-tures, as suggested by Van Duren and Sanders (1988) or,alternatively, in searching for a familiar form in a noisybackground . Whatever the activity of this additional pro-cess, the consequence of its invocation could be a tem-porary fixation on the slow route, which yields the subse-quent cost-benefit characteristic . Similarly, the concurrenceof costs with benefits has often been interpreted in theliterature as an indicant of the involvement of attention .For instance, Posner and Snyder (1975b ; see also Posner,1978) inferred the attentional load of a mental pathwayfrom the observation that benefits of a valid prime on sub-sequent processing concur with costs ofan invalid prime .Research that explores the conjecture that the HSE reflectsswitching problems from controlled to automatic process-ing is in progress .

REFERENCES

ANDERSON, J . R. (1983) . The architecture ofcognition . Cambridge,MA : Harvard University Press.

BnMaee, D. (1969) . Reaction times and error rates for "same"-"dif-ferent" judgments of multidimensional stimuli . Perception & Psycho-physics, 6, 169-174.

FneELL, B. (1985) . "Same"-"different" judgements : A review of cur-rent controversies in perceptual comparisons. Psychological Review,98, 419-456 .

HnNSEx, W., & SANDERS, A. F . (1988) . On the output of encoding dur-ing stimulus fixation . Acta Psychologica, 69, 95-107 .

HouTMnxs, M. J. M ., & SANDERS, A. F . (1983) . Is information ac-quisition during large saccades possible? Bulletin of the PsychonomicSociety, 21, 127-130.

JoNiDes, J., & MACK, R. (1984) . On the cost and benefit of cost andbenefit . Psychological Bulletin, 96, 29-44.

Kot,exs, P. A. (1976) . Reading a year later . Journal ofExperimentalPsychology: Human Learning & Memory, Z, 554-565.

Koreas, P. A ., & RoeDIGea, H . L., III (1984) . Procedures of mind .Journal of Verbal Learning & Verbal Behavior, 23, 425-449.

KRUEGER, L. E. (1978) . Atheory of perceptual matching . Psychologi-cal Review, 85, 278-304.

KRUEGER, L . E . (1983) . Probing Proctor's priming principle : The ef-fect of simultaneous and sequential presentation on "same"-"differ-ent" judgements . Journal of Experimental Psychology : Learning,Memory, X Cognition, 9, 511-523.

KRUEGER, L. E., & SenPiao, R. G . (1981a). Intertrial effects of same-different judgements . Quarterly Journal ofExperimentalPsychology,33A, 241-265.

KRUEGER, L. E., & SxAriao, R. G. (198!6)_ A reformuladon ofProc-

156 LOS

tor's unified theory for matching-task phenomena. Ps.crhologiculReview, 88, 573-581 .

LnBrRCe, D., &S4MUEGS, S. J . (1974) . Toward a theory of automaticinformation processing in reading . Cognitive Psvchology, 6, 293-323.

LocnN, G. D. (1978) . Attention in character-classification tasks: Evi-dence for the automaticity of component stages . Journal of Experimen-tnl Pslchology: General, 107, 32-63.

Locnv, G. D. (1988) . Toward an instance theory of automatizauon.Ps_vchologicn! Review, 95, 492-527.

NicKeasoN, R. S . (1975) . Effects of correlated and uncorrelated noisenn visual pattern matching . In P. M. A. Rabbit & S. Dornic (Eds .),Attention and performance V (pp. 655-668) . New York : AcademicPress.

NieMi, P. (1979) . Stimulus intensity effects on auditory and visual reac-tion processes . Acm Psycho/ogica, 43, 299-312 .

PosNea, M . L (1978) . Chronometric explorations of mind. Hillsdale,NJ : Erlbaum.

PosNea, M. I ., & StvvDee, C. R. R. (1975a) . Attention and cognitivecontrol In R. L. Solso (Ed.), Information processing and cognition :The Loyola symposium (pp . 55-SS) . Potomac, MD : Erlbaum.

PosNFa, M . I ., & SNVnea, C . R . R . (1975b) . Facilitation and inhibi-tion in the processing of signals . In P . M . A . Rabbit & S . Dornic(Eds .), Attention and performance V (pp . 669-682) . New York : Aca-demic Press .

PROCTOR, R. W. (198 1) . A unified theory for matching-task phenom-ena . Psychological Review, 88, 291-326.

PROCTOR, R. W ., & Rno, K. V . (1983) . Reinstating the original prin-ciples of Proctor's unified theory of matching-task phenomena: Anevaluation of Krueger and Shapirds reformulatioo . PsychologicalReview, 90, 21-37 .

RoeoIcE2, H . L., III, & BLnxro :v, T. A. (1987) . Effects ofvarying mo-

dality, surface features, and retention interval on priming in word-fragment completion . Memory & Cognition, 15, 379-388.

SANDERS, A. F. (1963) . The selective process in the functional visualfield . Assen : van Gorcum .

SANDERS, A. F. (1977) . Structural and functional aspects of the reac-tion process . In S. Dornic (Ed.), Attention and performance VI (pp. 3-25) . Hillsdale, NJ : Erlbaum.

SANDERS, A. F . (1980) . Stage analysis of reaction processes . In G. E.Stelmach &J. Requin (Eds.), Tutorials in motor behavior (pp. 331-354) . Amsterdam: North-Holland.

SANDERS, A. F., & RnTe, A . (1991) . Perceptual processing and speed-accuracy trade-off. Acta Psychologica, 77, 275-291 .

VAN pea MoLeN, M.W ., & Keuss, P. J . G. (1981) . Response selec-tion and the processing of auditory intensity . Quarterly Journal ofExperimental Psychology, 33A, 177-184.

VAN DuxEN, L., & SANDERS, A . F . (1988) . On the robustness of theadditive factors stage structure in blocked and mixed choice reactiondesigns . Acta Psychologica, 69, 83-94.

NOTES

1 . The number of measurements per condition was 25% less in themixed group than in the blocked group. This was due to the fact thatthe blocked and mixed conditions had originally been intended for twoseparate experiments instead of one.2. The author thanks an anonymous reviewer for suggesting this

possibility .

(Manuscript received October 8, 1992 ;revision accepted for publication July 9, 1993 .)