313-328 overshadowing and summation in compound stimulus conditioning of the rabbit's nictitating...

7/30/2019 313-328 Overshadowing and Summation in Compound Stimulus Conditioning of the Rabbit's Nictitating Membrane Response.

1/16

Journal of Experimental Psychology:Animal Behavior Processes1982, Vol. 8, No. 4, 313-328Copyright 1982 by the American Psychological Association, Inc.0097-7403/82/0804-03i3$00.75

Overshadowing and Summationin Compound Stimulus Conditioningof the Rabbit's Nictitating Membrane Response

E. James KehoeUniversity of New South Wales, Kensington, New South Wales, AustraliaThe present experiments examined acquisition of the rabbit's nictitating mem-brane response to a light + tone simultaneous compound stimulus and its com-ponents as a function of the intensity of the tone. In Experiment 1, the toneintensity was varied across the values of 85, 89, and 93 dB, and the CS-US(conditioned stimulus-unconditioned stimulus) interval'was 400 msec. In Ex-periment 2, the tone intensities were 73, 85, and 93 dB, and the CS-US intervalwas 800 msec. Experiments 3 and 4 further examined the effects of the 73-dBtone at CS-US intervals of 400 and 800 msec, respectively. All experimentsincluded control groups, which were trained with either a light or a tone CS. Inbrief, the experiments revealed repeated instances of overshadowing, i.e., theimpairment of conditioned response (CR) acquisition to one or both of the com-ponents of a compound. Moreover, two types of summation were obtained:within-subjects summation, in which rabbits trained with a compound showeda higher level of responding to the compound than to either of its componentCSs (Experiments2, 3, and 4), and between-groupssummation, in whichagrouptrained with a compound showed faster CR acquisition than either of its cor-responding control groups trained with a single CS (Experiments 1 and 2). Theresults are discussed in terms of perceptual and distributive processing modelsof compound stimulus conditioning.

Any organism in even the most sterile en-vironment is faced with a continual andmultifaceted stream of stimulus events. Todiscover the laws governing behavioral ad-justments to the exigencies of an environ-ment, Pavlov (1927, pp. 110-113) and sub-sequent investigators used compoundsof twoseparable stimuli (e.g., tone + light) as a lab-oratory model for the array of innocuousevents that antedate a biologically significantstimulus (e.g., Hull, 1943; Kehoe & Gor-mezano, 1980; Razran, 1965, 1971; Wick-ens, 1954, 1959, 1965). In any compoundstimulus conditioning procedure, it is logi-This research was supported by the Australian Re-search Grants Committee. The author expresses his grat-itude to Nancy Amodei and Ann Topple for collectingthe data. The author also thanks Linda A.Kehoe for hercritical reading of the manuscript and J. C. Clarke,C. R. Gustavson, and R. F. Westbrook for their sug-gestions concerning interpretation of the data.Requests for reprints should be sent to E. James Ke-hoe, School of'Psychology, University of New SouthWales, P.O. Box 1,Kensington, New South Wales 2033,Australia.

cally possible to observe both the interactionand integration of the component stimuli.The delineation of an interaction concernsthe extent to which conditioned response(CR) acquisition to a given conditioned stim-ulus (CS) is influenced by the other com-ponent in a compound, A prototypical ex-ample of an interaction is "overshadowing."In the overshadowing phenomenon, re-sponse acquisition to one CS is reducedthrough compound training with anotherCS,which often is highly "salient" in that it pro-duces rapid response acquisition (Kamin,1969; Mackintosh, 1976; Pavlov, 1927, p.14L). While the study of interaction concernsthe relation between the components of acompound, the studyofintegration concernsthe relation of the components to the com-pound as a whole. The study of integrationhasbeen conducted both on a within-subjectsbasis, by periodically testing the subject withindividual presentations of the componentsduring compound training, and on a be-tween-groups basis, by comparing the re-sponding of a group trained with a com-

313


2/16

314 E. JAMES KEHOEpound to control groups trained with eachof the components. The most common ex-perimental outcome has been "summation,"in which the level of responding to the com-pound exceeds that to the components (Ke-hoe & Gormezano, 1980). Although the labelsummation carries the connotation of anatomistic process of combination, the termis intended strictly as a descriptive one with-out prejudice to alternative theories.Many conditioning theories address over-shadowing and summation. These can beroughly grouped into two classes, which willbe denoted "perceptual" and "distributive."The most extreme perceptual hypotheses as-sume that the separate componentsofa com-pound fuse into a single functional stimulus,i.e., a "configuration" (Razran, 1965, 1971).Less extreme perceptual hypotheses assumethat the component stimuli of a compoundeither modify each other's sensory encoding(Hull, 1943)'or contribute independently toa joint signal (Heinemann & Chase, 1975;Kehoe & Gormezano, 1980, pp. 370-373).For perceptual hypotheses, the results of testswith components outside the context of thecompound (i.e., overshadowing and within-subjects summation) presumably reflect thedegree of generalization (or lack thereof)from the compound to the functional stimuliof the separate components. In contrast tothe perceptual hypotheses, the distributivehypotheses assume that the components ofa compound remain functionally the samein all contexts but are processed in such away that there is a trade-off jn the amountof associative strength ultimately gained byeach of the contending stimuli. Distributivehypotheses postulated a fixed processing ca-pacity (Sutherland & Mackintosh, 1971) ora fixed associative strength (Rescorla &Wag-ner, 1972;cf. James & Wagner, 1980). How-ever, Mackintosh (1975) argued that trade-offs can be achieved by a process in whichthe stimulus with the greatest relative asso-ciative strength on a trial gainsa proportionalincrement in its learning rate parameterwhile all other stimuli suffer a decrement intheir respective learning rate parameters (cf.Moore & Stickney, 1980). To account forsummation results, the distributive accountsall assume a relatively simple addition of as-sociative strengths (Mackintosh, 1975; Res-corla & Wagner, 1972).

The variety of alternative formulationsconcerning compound conditioning betraysthe fragmented nature of the available data.Although there is a growing body of researchconcerning stimulus interaction (e.g., James& Wagner, 1980; Mackintosh, 1976; Mack-intosh & Reese, 1979) and stimulus integra-tion (e.g., Bellingham & Gillette, 1981; Gil-lette & Bellingham, 1982; Kehoe & Gor-mezano, 1980), no single study has soughtto establish the relation between overshadow-ing, within-subjects summation, and be-tween-groups summation. Moreover, thereare major unresolved empirical questionsconcerning each of these phenomena.

Reciprocal OvershadowingReciprocal overshadowing is said to occurwhen acquisition to both components of acompound show some impairment relativeto corresponding single-stimulus controls (cf.Mackintosh, 1976). The issue of reciprocalovershadowing has been raised in connectionwith distributive hypotheses which postulatea fixedprocessing capacity. According to thispostulate, which has been designated theinverse hypothesis, acquisition to onestimulus occurs always at the expense ofother concurrent stimuli (Mackintosh, 1975,1976). When two CSs are of equal salience,reciprocal negative interactions would be ex-pected, because the two CSs would evenlydivide the limited resource. On an empiricallevel, overshadowing effects have been largelyunidirectional in nature (e.g., Kamin, 1969;Mackintosh, 1976). However; Mackintosh(1976) found that reciprocal overshadowingeffects were detectable but only when equis-alient stimuli were weak or moderate in ab-solute intensity (cf. James & Wagner, 1980).

The Constancy of Within-Subjects SummationAlthough within-subjects summation hasbeen demonstrated repeatedly, there have

been relatively few attempts to determinewhether a fixed empirical rule can be foundto describe the quantitative relation betweenresponding to a compound and to its com-ponents (Kehoe & Gormezano, 1980).Among the available theories, the configuralhypotheses generally argue against any stable


3/16

OVERSHADOWING AND SUMMATION 315relation (Razran, 1965, 1971). Other percep-tual hypotheses postulate a stable relation butcontend that the quantitative relation be-tween responding to the compound and re-sponding to its components will vary withfactors affecting the degree of generalizationbetween the compound and its nominalcomponents (Bellingham & Gillette, 1981;Heinemann & Chase, 1975; Hull, 1943). Incontrast, the distributive accounts appear topredict a relativelyfixed quantitative relationbetween responding to a compound and re-sponding to its components (Mackintosh,1975; Rescorla &Wagner, 1972). While theydo not specify an exact rule for mapping as-sociative strength into responding, they doassume that there is a fixed monotonic rela-tion between the associative strengths of cur-rent stimuli and overt performance.

Between-Groups Summationand OvershadowingBetween-groups summation occurs when-ever CR acquisition in a group trained witha compound is faster than that in corre-sponding groups trained with only one of thecomponents. The only formulation that pre-dicts that between-groups summation willoccur under all circumstances is that ofHull(1943). Hull assumed that the associativestrengths of the component stimuli func-tional inside a compound will summate toyield betweeq-groups summation even if oneor both of the nominal stimuli suffer sub-stantial generalization decrements whentested outside the context of the compound(Kamin, 1969, p. 55).Allother formulations,both perceptual (e.g., Heinemann & Chase,1975) and distributive (Mackintosh, 1975;Rescorla & Wagner, 1972), generally predictbetween-groups summation, but they alsoexpect it to disappear when profound over-shadowing occurs. Under the distributivehypotheses, overshadowing represents adef-icit in acquisition of associative strength. Ifone component of a compound can be en-

tirely prevented from acquiring associativestrength, then between-groups summationdisappears. The available evidence is ambig-uous: In conditioned suppression paradigms,Kamin (1969, pp. 53-55) found that be-tween-groups summation and overshadow-ingwere mutually exclusive, but Mackintosh

.(1976) observed between-groups summationin conjunction with substantial overshadow-ing of one component.In order to provide additional evidencerelating to these issues, the present experi-ments were designed to examine acquisitionof the rabbit's nictitating membrane responseto a light + tone compound and its compo-nents as a function of tone intensity. Testtrials for the individual components were in-terspersed among the compound trainingtrials to permit a determination of within-subjects summation throughout acquisition.Moreover, a complete set of single-stimuluscontrols were included in order to assess bothovershadowing and between-groups sum-mation effects.

Experiment 1- The design of Experiment 1 was modeledon that ofMackintosh (1976, Experiment 1).In order to detect overshadowing, a fixedvisual CS (flashing house light) was com-pounded with a auditory CS (1000-Hz tone)which was varied in intensity across groups.

The values of the tone were selected on thebasis of previous research in the Universityof New South Wales laboratory. An 85-dB(SPL) tone had been found to produce a CRacquisition rate comparable with that of thelight CS. According to the convention thatdefines salience in terms of demonstratedconditionability, the 85-dB tone CS and thelight CS would be regarded as having equalsalience. If an inverse law is correct, mutualnegative interactions would be expected foranimals trained with a compound of the lightand 85-dB tone. In order to determinewhether unidirectional overshadowingof thelight could be obtained, two more intensevalues were used, namely, 89 and 93 dB.Method

Subjects. The subjects were 56 naive male and femalealbino rabbits (Oryctolagus cuniculus). On arrival, eachrabbit was 70-80 days old and weighed about 1.5 kg.Each rabbit had free access to food and water in its homecage.Apparatus. The apparatus and recording procedurefor the nictitating membrane response were patternedafter those of Gormezano (1966) and were described indetail by Kehoe, Schreurs, and Amodei (1981). Duringtraining sessions, each rabbit was held in a Perspex re-strainer and placed in one of eight sound-attenuating,ventilated chambers. On the wall ofthe chamberin front


4/16

316 E. JAMES KEHOEof the subject was located a stimulus panel. A speakerwas mounted at the midpoint of the stimulus panel, 8cm anterior to and 16 cm above the subject's head. Thespeaker provided both white noise and an auditory CS,which was a 1000-Hz tone of 85, 89, or 93 dB (SPL)superimposed on an 82-dB background provided by thewhite noise and exhaust fans. Mounted on the stimuluspanel 4 cm above the speaker was an 8-W frosted neonlight tube. To provide a visual CS, the light was flashedat a rate of 20 Hz. The unconditioned stimulus (US) wasa 50-msec, 3-mA, 50-Hz ACshock delivered via stainlesssteel Autoclipwound clips positioned 10 mm apart and10-15 mm posterior to the dorsal canthus of the righteye. The sequence and timing of stimulus events on eachtrial were controlled by solid-state circuitry.Each rabbit's right external eyelids were held open byNo. 3 tailor-hooks mounted on a Velcro strap whichfitted about the head. A muzzle-like head-set, fitted se-curely about the snout, supported aphotosensitive trans-ducer for monitoring movements of the nictitatingmembrane. A small hook was attached to a silk loopsutured in the nictitating membrane of the rabbit's righteye. The hook was connected by a thread to one end ofan L-shaped piano-wire lever, which mechanically trans-mitted the movement of the nictitating membrane to aphotoelectric transducer. The variations in electrical sig-nal from the transducer were amplified and recorded ona four-channel, ink-writing oscillograph operating at aspeed of 50 mm/sec.Procedure. All animals received 1 day of preparation,1 day of recovery, 1 day of adaptation, and 6 days ofacquisition training. On the preparation day, a smallloop of silk (000 Dynex) was sutured into the rabbit'sright nictitating membrane, the surrounding hair wasremoved, and the animals were returned to their homecages for 1 day of recovery. On the adaptation day, theanimals were placed in the conditioning apparatus fora period equal to the length of the subsequent trainingsessions, but neither the CSs nor the US was presented.Following preparation, the animals were assigned ran-domly to one of seven groups (n = 8). Three groups weretrained with a compound of light and tone. Across thethree groups, the light was fixed, but the tone was variedover the values of 85, 89, and 93 dB. These compoundgroups were designated as LT85, LT89, and LT93,which denotes light, tone, and intensity of the tone. Thecompound groups received 54 reinforced trials per day,which were interspersed with 3 unreinforced test trialseach to the light (L ) and tone (T). The test trials werepresented on Trials 10, 20, 30, 40, 50, and 60. On odd-numbered days, the sequence of test trials wasLTTLLT,and on even days, the sequence was TLLTTL. In ad-dition to the compound stimulus groups, there were foursingle-CS control groups: One was trained with the lightand the others with the 85-, 89-, and 93-dB tones, re-spectively. The control groups were designated L, T85,T89, T93. The light group received 54 reinforced trialsper day, 3 unreinforced light test trials, and 3 unrein-forced test trials with the 85-dB tone. Similarly, the tonecontrol groups received 54 reinforced trials per day, 3unreinforced tone test trials, and 3 unreinforced lighttest trials. For all groups, the CS duration and the CS-US interval were 400 msec. The intertrial intervals were50, 60, and 70 sec (M = 60 sec).

A response was defined as any extension of the nic-titating membrane exceeding .5 mm (1 mm of pen de-flection). A CR on reinforced trials was defined as anyresponse initiated after the onset of the CS(s) but beforeUS onset, i.e., a response with a latency of less than 400msec following CS onset. On the test trials, a CR wasdefined as any response occurring with a latency lessthan 1,000 msec after CSonset. For analysisof the data,a set of contrasts was written by using the Bonferronitechnique (Miller, 1966). The rejection level was set ac-cording to a Type I error rate (.05) for a family of con-trasts.Results

Figure 1 depicts the mean CR percentageshown by the seven groups across successive2-day blocks of training. Panels A, B, and Cshow the CR acquisition curves for thelight + tone compound and its componentsfor Groups LT85, LT89, and LT93, respec-tively. Panel D shows the CR acquisitioncurves as measured on reinforced trials forthe single-stimulus control groups, namely,T85, T89, T93, and L.Overshadowing. Inspection of Figure 1reveals clear evidence of overshadowing inthat CR acquisition to the fixed light CS wasimpaired by training in compound with themore intense tones. Across the three com-pound groups, the rate and level of acquisi-tion to the light decreased as tone intensityincreased. Overall, there was a significantdownward linear trend in the means for re-sponding to the light, which were 82%, 47%,and 34% for Groups LT85, LT89, and LT93,respectively, F(\, 28) = 19.54. More impor-tant, inspectionof panel D of Figure 1 revealsthat CR acquisition in Group L , which wasgiven reinforced training with only the light,rapidly reached an asymptote near 100% andshowed an overall mean of 74%. In compar-ison with Group L, Groups LT89 and LT93showed significantly lower levels of respond-ing to the light CS, Fs(\, 28) = 8.97 and17.74, but Group LT85's responding to thelight failed to differ from that of Group L(F


5/16

OVERSHADOWING AND SUMMATION 317responding onlight test trials in the tone con-trol groups was examined. The overall meanlevels of responding to the light in GroupsT85, T89, and T93 were 29%, 18%,and 9%,respectively. These levels were, respectively,53,29, and 25 percentage points below thoseobtained with light in the correspondingcompound groups, smallest f\l, 42) = = 6.89.Accordingly,crossmodal generalization fromtone accounts only for a portion of the re-sponding to the light in groups trained witha compound. Nevertheless, the differences inresponding to the light between the tone andcompound groups were larger at the lowerintensities, for the interaction between thelinear trend across tone intensity and the

10 0908070eo604030

1 2O 10

10090807060SO403020100

_C OM P OU N Doo LIGHT T E S T S4* T O N E T E S T S/

1- 2 3-4 6-6 1- 2 3-4 6-6TWO-DAY BLOCKS

(Figure 1. Mean percentage of conditioned responses(CRs) in Experiment 1 plotted as a function of 2-dayblocks. (Panels A, B, and C depict CR acquisition to thecompound and itscomponents for Groups LT85, LT89,and LT93, respectively. Panel D shows the CR acqui-sition curves for Groups T85, T89, T93, and L.)

compound versus tone training conditionswas significant, F(\, 42) = 4.14. Since thelevel of responding to light in both the com-pound and the tone control groups was in-versely related to tone intensity, different lev-els of crossmodal generalization may havecontributed to the differences among com-pound groups in their respective levels of re-sponding to the light component.Inspection of the acquisition curves fortone, as shown in Figure 1, reveals that thelevel of responding to tone in both the tonecontrol and the compound groups wasgen-erally high across all three intensities. Despitethe large effects of tone intensity on CR ac-quisition to light, there was little evidence ofpronounced differences in salience amongthe tone intensities as measured in the con-trol groups. In connection with the issue ofreciprocal overshadowing, there was littlesuggestion of impairment in CR acquisitionto tone in any of the compound groups. Inthe tone groups, the mean levels of respond-ing shown to the tone by Groups T85, T89,and T93 were 79%, 87%, and 75%, respec-tively. The means of the corresponding com-pound groups were 78%, 83%, and 85%, re-spectively. As may be apparent, there is onedisparity in the pattern ofmeans between thetone groups and the compound groups:Group T89 showed a higher level ofrespond-ing than groups T85 or T93, while in thecompound groups, the level ofresponding tothe tone appeared to be a positive functionof intensity. These differences were most pro-nounced early in training and were con-firmed by a significant three-way interactionbetween the tone versus compound condi-tions (Factor 1), linear trend across tone in-tensity (Factor 2), and linear trend acrossblocks of trials (Factor 3), F(l, 42) = 4.20.Befween-groups summation. Even whenovershadowing occurred, CR acquisition tothe compound appeared to proceed morerapidly than in the single-stimulus controlgroups. In Groups LT85, LT89, and LT93,the overall mean levels of responding to thecompound were 91%, 89%, and 91%, re-spectively. In Groups T85, T89, T93, and L,the overall means for reinforced trials were77%, 85%, 73%, and 74%, respectively. In-spection of Figure 1 reveals that differenceswere most apparent in the first block of train-


6/16

318 E. JAMES KEHOEing, after which the performance of all groupsconverged. Statistical analysis of the overallmeans confirmed that all three compoundgroups were individually superior to GroupL , smallest F(l, 49) = 12.24. Group LT85was superior to Group T85, f\l, 49) = 10.72,and Group LT93wassuperior to Group T93,F(l, 49) = 18.30. However, Group LT89 wasnot superior to Group T89 (F < 1). An anal-ysis of differences in linear trends acrosstraining blocks revealed an identical patternof results.Within-subjects summation. Inspectionof panels A, B, and C of Figure 1 revealssome evidence for within-subjects summa-tion. InGroup LT85, responding to the com-pound was initially higher than respondingto either component, but responding on allthree types of trials converged to a meanasymptotic level near 100%. The initial dif-ference in the acquisition curves for the com-pound and its components was confirmed inthe statistical comparison between the lineartrends for compound versus tone trials, F(l,21) = 16.19, but not for compound versuslight trials, F( 1,21) = 7.25, p < .10. In GroupsLT89 and LT93, the level of responding tocompound was only marginally higher thanresponding to the tone throughout acquisi-tion, but responding to the compound wassubstantially higher than responding to thelight. In both groups, there was no significantdifference between responding to the com-pound and responding to the tone, but therewas a significant overall difference betweenresponding to the compound and respondingto the light, Fs(l t 21) = 22.97 and 43.55 fo rGroups LT89 and LT93, respectively. In or-der to determine whether the group meancurves reflected the performance of the in-dividual subjects, the relative levels of re-sponding to the compound and its compo-nents over Days 1 and 2 were examined foreach subject. Out of eight subjects in eachgroup, Groups LT85, LT89, and LT93 con-tained, respectively, six, seven, and three sub-jects that showed a higher level of respondingto the compound than to either component.Discussion

The results of Experiment 1 revealed (a)unidirectional overshadowing of light as afunction of tone intensity, (b) between-groups

summation in Groups LT85 and LT93, and(c) weak evidence of within-subjects sum-mation, which was most apparent at thelower tone intensities. The pattern of resultslies between that of Mackintosh (1976) andthat of Kamin (1969): The observation thatovershadowing occurred in conjunction withbetween-groups summation agrees with thefindings of Mackintosh (1976, Experiment1). However, with respect to the question ofreciprocal overshadowing, the present resultsresemble those ofKamin, i.e., unidirectionalovershadowing of the light was observed withthe 89- and 93-dB tones, but unimpeded CRacquisition to both light and tone apparentlyoccurred with the 85-dB tone. Mackintosh(1976, Experiment 2) gave reinforced train-ing with equally salient light and tone stimulito three compound groups which differed inthe absolute intensities of both stimuli. Re-ciprocal overshadowing was found in thegroup trained with the least intense CSs,whereas the other two groups trained withmore intense CSs showed no evidence ofovershadowing. In the present experiment,the CSs used were highly salient, as dem-onstrated by the rapid CR acquisition seenin all the single-CS control groups. Accord-ingly, the unimpeded CR acquisition to thetone and light in Group LT85 can be con-strued as consistent with Mackintosh's (1976,Experiment 2) findings.In the assessment of both overshadowingand within-subjects summation, the intervalfor scoring of responses on light and tone testtrials, 1,000 msec, was longer than that ofreinforced compound trials, 400 msec. Con-sequently, the level of responding on testtrials may have provided a higher estimateof CR strength than reinforced compoundtrials, which would attenuate the apparentmagnitude of overshadowing and within-subjects summation effects. However, otherexperimentation with the rabbit nictitatingmembrane response (NMR) preparationfailed to reveal any differences in the per-centage CR measure for observation inter-vals of 400 msec on CS-US trials and 1,000msec on CS test trials (Tait, Kehoe, Gor-mezano, Note 1). Yet, in order to avoid anypotential biases in subsequent experiments,the interval of measurement on all trialswasequal to the CS-US interval.


7/16

OVERSHADOWING AND SUMMATION 319Experiment 2

The present experiment was conducted to(a) further test Mackintosh's (1976) hypoth-esis that reciprocal overshadowing occursonly when the salience of the two compo-nents is equal and low and (b) examine sum-mation effects when the overall rate of CRacquisition is slower than that observed inExperiment 1. In Experiment 1, CR acqui-sition on reinforced trials in both the com-pound and the control groups was relativelyrapid and reached asymptotes near 100%CRs. With the exception of the overshadow-ing results, differences between groups andwithin subjects were largely confined to thefirst 2 days of training. Accordingly, it couldbe argued that the failure to observe recip-rocal overshadowing in Group LT85 and theweakness of within-subjects summation couldbe attributed partly to the ceiling on mea-surement imposed by the rapid approach to100% CRs. In fact, Kamin (1969, pp. 54-56)found that even unidirectional overshadow-,ing varied with the overall rate of acquisitionand could be eliminated if the overall rate ofconditioning was raised by increases in theintensity of a shock US. Consequently, in the^present study, the overall rate of CR acqui-sition was reduced by lengthening the CS-US interval to 800 msec, a value that pro-duces a modest but reliable rate of CR ac-quisition (Kehoe, 1979; Smith, Coleman, &Gormezano, 1969).While the available theories (Hull, 1943;Mackintosh, 1975; Rescorla&Wagner, 1972)differ regardingquestions of reciprocal over-shadowing, they allagree that overshadowingis a function of relative CS salience, Thus,it should be possible to reduce tone intensityto a point at which the tone will be overshad-owed by the light used in the present exper-iments. Consequently, the present experi-ment extended the rangeof the tone intensitymanipulation by including a tone of 73 dB,which wasdemonstrably less salient than thelight CS.MethodSubjects. The subjects were 44 naive male and femalealbino rabbits, 70-90 days old and weighing about 2 kgon arrival.Apparatus andprocedure. Unless otherwise specified,the apparatus, recording technique, and procedure were

identical to those described for Experiment 1. The tim-ing of stimulus presentations wascontrolled by an AppleII computer with interfaces and software described byScandrett and Gormezano (1980). The only change inprocedure caused by the introduction of computer con-trol was in the intertrial interval (ITI) values. With thecontrol system used for Experiment 1, the ITIs wererandomized among the valuesof50,60, and 70 sec (M =60), whereas the computer system denominated the ITIvalues in 1-sec units over the range 50-70 sec (M = 60).For recording purposes, the signals from the movementtransducer were sent to the 8-bit analog-to-digital con-verter of the computer as well as to the oscillograph.During trials, the computer sampled the signal from eachtransducer every 5 msec.The rabbits were randomly assigned to one of sevengroups, all of which received 8 days of training. Groups(n = 8) labeled LT73, LT85, and LT93 received rein-forced training with a light + tone compound, in whichthe tone intensities were 73,85, and 93 dB, respectively.Groups (n = 4) labeled T73, T85, and T93 received re-inforced training with a tone CS,for whichthe intensitieswere 73, 85, and 93 dB, respectively. Finally, a group(n = 8) labeled L received reinforced training with thelight CS. During training, four animals died, one inGroup LT85, one in Group LT93, and two in GroupL. Their incomplete data were not included in the de-scription or analysis of the results. For all groups, boththe CSduration and the CS-US interval were 800 msec.Moreover, the response scoring interval for all trials was800 msec measured from CS onset.Comparison of computer versus human data. Prelim-inary statistical analyses compared CR frequency datacollected by the computer with those collected by thehuman observer. Separate analyses of the data from toneand from light test trials failed to reveal any significantdiscrepancies between the two methods of responsecounting. However, the analysis of data from reinforcedtrials revealed a small but significant difference of fourpercentage points between the overall mean CR fre-quencies collected by the computer (55%) and those col-lected by the human (59%), f\l, 29) = 5.48, p < .05.There were no significant interactions of data collectionmethod with the tone intensity variable, compound ver-sus single stimulus manipulation, or trends across train-ing. A trial-by-trial examination indicated that the hu-man observer counted as responses some movementsthat, according to the computer, were just under thenominal .5 mm (1 mm of pen movement) criterion fora response. The human observer's criterion appeared tobe equivalent to 3/8ths mm (6/8ths mm of pen move-ment). Since the two methods of data collection werein good agreement, the results reported for Experiments2, 3, and 4 are based on the computer-collected data.Results and Discussion

Figure 2 depicts the mean CR percentageshown by the seven groups across successive2-day blocks of training. Panels A, B, and Cshow the CR acquisition curvesto the light +tone compound and its components forGroupsLT73, LT85, and LT93, respectively.


8/16

320 E. JAMES KEHOEPanel D shows the CR acquisition curves onreinforced trials for the single-stimulus con-trol groups, T73, T85, T93, and L.Overshadowing. Inspection of Figure 2reveals that all three compound groups dis-played overshadowing of light. The level ofresponding to the light in the compoundgroups rose to terminal levels less than 50%CRs, which diverged significantly from thehigh terminal level shown by Group L (M =88% CRs). Statistical comparisons confirmedthat there were significant differences in thelinear trends across training for each com-pound group relative to Group L, smallestF(\, 24) = 12.23. Comparisons between the

D CI t !

100

90

80

70

60

50

40

30

20

10

010 0

90

80

70

60

SO

40

30

20

10.

0

LT73

ICOMPOUNDo U I Q H T TESTiTONE TEST

LT85

1-2 3-4 5-6 7-8 1-2 3-4 6-6 7-8TWO-DAY BLOCKS

Figure 2. Mean percentage of conditioned responses(CRs) in Experiment 2 plotted as a function of 2-dayblocks. (PanelsA, B, and C depict CR acquisition to thecompou nd and its componentsfor Groups LT73, LT85,and LT93, respectively. Panel D shows the CR acqui-sition curves for Groups T73, T85, T93, and L.)

compound groups failed to reveal any sig-nificant differences in their respective levelsof responding to light. To determine whetherthe observed responding to the light in thecompound groups could be attributed tocrossmodal generalization from the tonecomponent, I compared the levels of re-sponding on light test trials in the tone con-trol groups with those of the compoundgroups. Specifically, Groups T73, T85, andT93 showed overall mean levels of 5%, 11%,and 12% CRs, respectively. Collectively, theselevels of generalized responding to light weresignificantly lower than the overall meanlevels of responding to the light shownby Groups LT73, LT85, and LT95, whichwere 32%, 16%, and 26%, respectively,F(\, 33)= 5.29.Examination of responding to the tone CSrevealed some evidence of reciprocal over-shadowing in Groups LT85 and LT93. BothGroups LT85 and LT93 showed terminallevels of responding (45%and 64%) that weresubstantially lower than those of their re-spective tone control groups^ T85 and T93(90% and 90%). In comparisons between lin-ear trends across training, Group LT85 ver-sus Group T85 failed to attain the declaredlevel of significance, F( 1,28) = 6.18, p > .05,but Groups LT93 and T93 did differ signif-icantly, F(l, 28) = 8.14. As shown in panelD of Figure 2, Groups T93 and T85 bothshowed rapid CR acquisition which followedlargely the same course as that of Group L.Thus, the salience of the 93- and 85-dBtones appeared close to that of the light, cre-ating the conditions for reciprocal overshad-owing under the inverse hypothesis.While the reciprocal overshadowing evi-dent in Groups LT85 and LT93 was consis-tent with an inverse law, Group LT73's re-sults were perplexing both theoretically andempirically. In examining the control groups,it is clear that the light was more salient thanthe 73-dB tone. Specifically, Group L showedmore rapid CR acquisition and a higher over-all level of responding than Group T73, F(\,14) = 13.74. However, in Group LT73, CRacquisition to the light but not the tone wasimpeded. Although the rate of CR acquisi-tion to the tone in Group LT73 was slow, itscourse of acquisition failed to differ from theequally slow acquisition of Group T73 (F


9/16

OVERSHADOWING AND SUMMATION 3211). If salience is to be anchored to the rateand level of CR acquisition produced by astimulus when trained by itself, then GroupLT73 appears to be a case in which a lesssalient stimulus can unidirectionally over-shadow a more salient stimulus.Between-groups summation. WhereasExperiment 1 revealed consistent between-groups summation, only Group LT93 in thepresent experiment showed facilitation ofGRacquisition to the compound: Group LT93'soverall level of responding to the compound(79%) was significantly greater than that ofGroup L (52%), F(\, 33) = 8.78, and therewas a significant divergence between the lin-ear trends across training for Groups LT93and T93 (overallM = 55%), F(\, 33) = 8.49.Among the other groups, Group LT85'soverall mean level of performance (69%)failed to differ from that of either Group T85(52%) orGroup L, while Group LT73's over-all mean (59%) differed from that of GroupT73 (17%), F(\, 33) = 16.10, but not that ofGroup L. The inconsistent nature of be-tween-groups summation in the present ex-periment is particularly damaging to conten-tions that the associative strengths of stimulifunctional inside a compound should alwayssummate to facilitate CR acquisition to acompound (Hull, 1943).Within-subjects summation. Inspectionof panels A, B, and C of Figure 2 revealsstrong evidence for within-subjects summa-tion, especially in comparison with the weakeffects seen in Experiment 1. In all three com-pound groups, the level of responding to thecompound significantly exceeded the level ofresponding to either the tone or the lightcomponent throughout training, smallest F( 1,19) = 11,14. To determine whether the groupmean curves reflected the performanceof theindividual subjects, I examined the mean lev-els of responding to the compound and itscomponents over all days for each subject.In Groups LT73, LT85, and LT93, 7/8, 7/7,and 6/7 subjects, respectively, showed ahigher level of responding to the compoundthan to either component.Evaluation of a combination rule. One ofthe key issues dividing the available theoriesconcerns the constancy of within-subjectssummation: Configure! hypotheses expect anunstable relation between the level of overt

responding to the compound and that to itscomponents. The remaining perceptual hy-potheses expect a stable relation but wouldinclude a correction for generalization dec-rements from compound training to com-ponent testing. Likewise, the distributive hy-potheses would also expect a relatively stablequantitative relation between performanceto the compound and its components. Sincenonconfigural hypotheses, both perceptualand distributive, specify the combinationrule as a summation oftheoretical quantities(i.e., associative strength), the exact form ofthe quantitative relation for overt respondingcannot be specified. However, a fair test forthe stability of within-subjects summationcan be accomplished by comparing observedlevels of responding to a compound withpredicted levels derived from a fixed rule forcombining the observed levels of respondingon tone and light test trials. In particular, thepercentage CRs to a compound (PAE) canbe compared with predicted levels (PAS')based on the percentage CRs to the separatecomponents (PA, PB) combined accordingto the formula PAS' = PA +PB- (PA XPB). This formula is, of course, the generalformula for the probability of at least one"hit" when sampling simultaneously fromtwo independent sources. Informal exami-nation of compound conditioning data in avariety of procedures has indicated that pre-dicted levels from the independent combi-nation rule approximate the observed levelsof responding to the compound (Kehoe &Gormezano, 1980, pp. 360-361).

To assess the stability and accuracy of theindependent combination rule, I comparedthe obtained levels of responding to the com-pounds with the predicted levels in a set oforthogonal comparisons. In the analysis ofExperiment 1's results, the only comparisoninvolving a difference between the obtainedand the predicted performances was in aninteraction of the linear trend across dayswith the linear trend across tone intensity,F(l, 21) = 5.17. Examination of the meansused in testing the interaction revealed somemodest discrepancies between obtained andpredicted performances.Asthe most notable,in Group LT93, there was a 7% overpredic-tion for the first 2-day block and a 7% un-derprediction for the last 2-day block. For all


10/16

322 E. JAMES KEHOE1009080

70605040

302010

*.EXPERIMENT 1o-0 EXPERIMENT 2w f /

0 10 20 30 40 50 60 70 80 90 100OBTAINED CR PERCENTAGE

Figure 3. A scattergram of the obtained percentage ofconditioned responses (CRs) on reinforced compoundtrials (abscissa) versus the predicted percentage of CRsbased on the independent combinations of CR proba-bilities estimated from light and tone test trials (ordi-nate). (Each point represents the datum of a single sub-ject from the compound groups in Experiment 1 [Days1-2] and Experiment 2 [Days 3-4].)

other points, the largest discrepancy betweenthe obtained and the predicted means was2% . However, in the analysis of Experiment2's results, there was a significant main effectcomparison between the overall mean of thepredicted performance (48%) and the meanof the obtained performance (69%). More-over, there were no interactions involvingthecomparison between predicted and obtainedlevels. Thus, there was a consistent, 21 per-centage point underproduction by the com-bination rule under consideration.To depict the subject-by-subject corre-spondence between obtained and predictedlevels ofresponding, Figure 3 is a scattergramof the actual percentage CRs on reinforcedcompound trials and the predicted percent-age CRs based on the independent combi-nation of CR probabilities estimated fromtone and light test trials. For Experiment 1,the data shown are from Days 1-2 of train-ing, during which time the levels of perfor-mance across subjects varied most widely.After the first 2 days, the range of variationbecame constricted as the level of respondingto both the compound and the components

in many subjects reached 100% CRs. Forcomparable data from Experiment 2, thedata from Days 3-4 are shown. Figure 3 alsoshows the best fitting straight lines for theregression of the predicted performance onthe observed performance in Experiments 1and 2. Inspection of Figure 3 reveals thatthere was moderate correspondence betweenthe actual and the predicted levels of perfor-mance, which was statistically confirmed bysignificant product-moment correlation coef-ficients. For Experiment 1, r = .68, n = 24,and for Experiment 2, r = .79, n = 22. Asindicated by the statistical analyses, the re-sults of Experiment 1 showed some overpre-diction, particularly in subjects that showeda moderate level of responding to the com-pound, and the results of Experiment 2showed a consistent underprediction.Despite the discrepanciesbetween the pre-dicted and the obtained compound perfor-mance levels in both experiments, the out-come of the independent combination ruleindicates that there are stable relations be-tween responding to a compound and re-sponding to its components. Thus, the eval-uation of a combination rule supports non-configural hypotheses. If the underpredictionsare taken seriously, they particularly supportarguments by perceptual hypotheses thatgeneralization decrements in the level of re-sponding to components outside the com-pound underestimate the associative strengthof the stimuli functional inside the com-pound (Heinemann & Chase, 1975; Hull,1943). The underpredictions can be handledby distributive hypotheses but only at theexpense of adding a perceptual postulate.Specifically, the Rescorla-Wagner model hasbeen appended with a configural, "uniquestimulus" hypothesis. This hypothesis con-tends that the observed responding to a com-pound represents the summation of not onlythe associative strengths of the componentsbut also that of an additional, configuralstimulus that arises during each presentationof a compound stimulus (Rescorla, 1972,1973; Whitlow & Wagner, 1972). Accord-ingly, a prediction of responding to the com-pound based on only the levels of respondingto the separate components would not in-clude the associative strength of the uniquestimulus.


11/16

OVERSHADOWING AND SUMMATION 323Experiments 3 and 4

The most startling result of Experiment 2was the observation that a demonstrablyweak 73-dB tone was able to impair CR ac-quisition to a more salient light without thetone itself suffering any noticeable reciprocalimpairment. Experiment 3 wasconducted toconfirm the overshadowing of light by the73-dB tone under an 800-msec CS-US in-terval. Subsequently, Experiment 4 was con-ducted to determine whether the same resultwould be obtained when the overall rate ofCR acquisition was increased by shorteningthe CS-US interval to 400 msec.Method

The subjects were 32 albino rabbits, 70-90 days oldand weighing2 kg on arrival. In each experiment, threegroups of rabbits were trained, which consisted ofGroups LT73 (n = 8), T73 (n = 4), and L ( = 4). Thesethree groups were trained and tested in the same fashionasdescribed for the correspondinggroups in Experiment2, with the exception that Group L was tested with the73-dB tone three times in each block of 60 trials. Theprocedure was identical to that of Experiment 2, withthe following exceptions: In Experiment 3, training wasconducted in 3days, each composed of 180 trials, whichcorresponded in sequence to three successive sessions of60 trials used in Experiment 2. In Experiment 4, theCS-US interval was 400 msec, and training was con-ducted across 5 days, each composed of 120 trials cor-responding in sequence to two successive sessions inExperiment 2. The reduction in the number oftrials persession from Experiment 3 to Experiment 4 was neces-sitated by observations that the 180 trials per session inExperiment 3 produced within-session decrements inresponding by some animals, which increased the within-cell variance. The reduction in the number of trials persession from Experiment 3 to Experiment 4 was ex-pected to have a positive effect on the rate of CR ac-quisition along with the reduction in the CS-US interval(Kehoe & Gormezano, 1974).

Results and DiscussionExperiment 3. In Figure 4, the left panelshows the CR acquisition curves for GroupLT73's compound and its components, andthe right panel shows the CR acquisitioncurves for the single-stimulus control groups,T73 and L. Inspection of Figure 4 revealsthat there was clear overshadowing of thelight in that the mean level of responding tothe light in Group LT73 (27%)was signifi-cantly lower than in Group L (70%), F(l,10) = 31.12. With respect to the tone, the

EXPERIMENT 3, SOO-mnec CS-US Interval100

9080

70o 60 soO_ 40cK 30

2010

G R O U P LT 73 C O M P O U N D 0 0 LIGHT T E S T S** TON E T E S T S

-X/, /

^ / 'X/ x' xxvS *1

CONTROL GROUPS f A

y GROUP L'-' /. /

/ GROUP T73*-' / / /; / i i

0 1 2 3 1 2 3ONE-DAY BLOCKS

Figure 4. Mean percentage of conditioned responses(CRs) in Experiment 3 (800-msec CS-US interval) plot-ted as a function of days. (Left panel depicts CR ac-quisition to the compound and its components forGroup LT73. Right panel shows the CR acquisitioncurves for Groups T73 and L.)

control group, T73, showed a relatively lowoverall level of responding (17%), and thusthere was little room to observe whether CRacquisition to the tone in Group LT73 (M =10%) suffered any impairment indicative ofreciprocal overshadowing. Although bothgroups showed some acquisition to the toneacross days, F(\, 10) = 10.17, any tendencyof Group LT73's responding to the tone todiverge to a lower level than that of GroupT73 was not significant,F(\, 10) = 3.91, p >.05. Finally,a comparison between the levelsof responding in the control groups con-firmed that the light was more salient thanthe 73-dB tone, F(\, 13) = 9.79.To determine whether the low level of re-sponding to the tone in Group LT73 mayhave reflected generalization from the lightcomponent of the compound rather than thetone's own associative strength, I estimatedthe degree of crossmodal generalization fromlight to tone by examining responding on thetone test trials given to Group L. In fact, noneof the four subjects in Group L ever re-sponded to the tone over 27 test presenta-tions, whereas in Group LT73, six subjectsshowed at least one response to the tone.Since there was no within-cell variance forGroup L, I decided to use the nonparametric


12/16


13/16

OVERSHADOWING AND SUMMATION 325of responding to the compound in GroupLT73 (77%) failed to differ significantly differfrom that of either Group L (71%) or GroupT73 (56%). However, Group LT73's perfor-mance over days did diverge to a higher levelthan Group T73's, as indicated by a signifi-cant difference in their respective lineartrends, F(l, 13) = 7.56. Third, Group LT73did show within-subjects summation, in thatthe overall level of responding to the com-pound (77%)was significantly higher thanthe overall level of responding to either thetone (49%), F(l, 13) = 20.41, or the light(46%), F(l, 13) - 12.70. Individually, sevenof the eight subjects showed a higher level ofresponding to the compound than to eithercomponent. Finally, an assessment of the in-dependent combination rule revealed thatthe mean overall level of responding to thecompound (76% GRs) wasunderpredicted bythe combined levels of responding to thecomponents (64%), However, the apparentunderprediction failed to attain the declaredlevel of significance, F(l, 6) = 5.09, .05


14/16

326 E. JAMES KEHOEunder consideration, the tendency for thedescriptive rule to underpredict respondingto the compound tends to favor the gener-alization decrement postulates of perceptualformulations (cf.Bellingham &Gillette, 1981;Heinemann & Chase, 1975; Kehoe & Gor-mezano, 1980). Since the quantitative rela-tion between responding to the componentsand responding to the compound was rela-tively stable, the within-subjects summationresults can also be regarded as supporting thedistributive accounts if supplemented by aconfigural, unique-stimulus hypothesis to ac-count for underproductions of responding tothe compound (cf. Rescorla, 1972, 1973;Whitlow & Wagner, 1972).Reciprocal Overshadowing

The results of all four experiments pro-vided repeated and systematic demonstra-tions ofunidirectional overshadowing, whichpreviously has been shown only indirectly inthe NMR preparation (cf. Marchant &Moore, 1973). Moreover, reciprocal over-shadowing was found in Group LT93 of Ex-periment 2 and perhaps Group LT85 of Ex-periment 2 and Group LT73 of Experiment4. In terms of their theoretical implications,the present results contradicted the expectedone-to-one relations between overshadow-ing, physical CS intensity, and CS salienceas denned by the rate of CR acquisition.Under the 400-msec CS-US interval of Ex-periment 1, the level of responding to thelight component of a compound declined astone intensity increased, but, as measured insingle-CS groups, the tone stimuli appearedto have equivalent salience. In Experiment2, under the 800 msec CS-US interval, theopposite pattern of results appeared, namely,CR aquisition to the light was equally im-paired across variations in tone intensity,which, in the control groups, proved to bedifferentially salient. More important, in Ex-periments 2 and 3, it appeared that a de-monstrably weak, 73-dB, 800-msec tone im-paired CR acquisition to the light, withoutCR acquisition to the tone itself appearingto suffer a reciprocal impairment.Although the correspondence between thedegree of overshadowing and salience wasnot all what the available theories would pre-

dict, appeals to ceiling effects on measure-ment and other vagaries of comparing over-shadowing and salience across differentgroups might be used to defend the assump-tion of a one-to-one relation between thedegree of overshadowing and the relative CSsalience. However, in Experiments 2 and 3,the unidirectional overshadowing of the lightby the less salient 73-dB, 800-msec tone rep-resents an apparent reversal of the relationbetween overshadowing and relative CS sa-lience which defies any simple appeal tomeasurement difficulties. Since the overshad-owing of light by the 73-dB, 800-msec tonewas undeniable, the only way out of the theo-retical paradox posed by the anomalous over-shadowing lies in explaining why CR acqui-sition to the 73-dB, 800-msec tone failed toshow a reciprocal impairment. In examiningthe total pattern of results, there are twopieces of evidence that CR acquisition to the73-dB, 800-msec tone was subject to recip-rocal overshadowing by the light. First, underthe 800-msec CS-US interval in Experiment2, CR acquisition to the 85-dB and 93-dBtones did show some reciprocal impairment.Second, under the 400-msec CS-US intervalin Experiment 4, the 73-dB tone itself suf-fered impairment in the final days of acqui-sition training.On the basis of this converging evidence,it appears possible to infer that the associativestrength of the 73-dB, 800-msec tone wasreduced but that overt responding arose fromany of three sources: (a) a within-compoundassociation between the tone and light (Res-corla, 1981, 1982; Rescorla & Cunningham,1978), (b) stimulus generalization from thelight component of a compound, and (c)stimulus generalization from a configuralstimulus. In considering the possible sourcesof responding to the 73-dB, 800-msec tone,the contribution of a within-compound as-sociation between the light CS and tone CSis difficult to assess. On the basis of resultsfrom other rabbit NMR studies, the evidencefor associations between simultaneous stim-uli is nonexistent: Conditioning between si-multaneous CSs and USs has never beenfound (Kehoe, Feyer, & Moses, 1981, Ex-periment 3A; Smith et al., 1969). However,until it is determined whether second-orderconditioning can be obtained with simulta-


15/16

OVERSHADOWING AND SUMMATION 327neousCSs in the NMR preparation, a within-compound association remains a possiblecontributor. As a source of generalized re-sponding to the 73-dB tone, it is possible torule out direct generalization from the lightitself: In Experiment 3, examination of thecontrol groups revealed negligible levels ofgeneralization from light to tone. Finally, inthe compound groups, there is the additionalpossible source of generalized responsestrength, namely, from a configural stimulus.What is most notable is that the anomalousovershadowing of light but not tone occurredonly under the 800-msec CS-US interval forwhich the independent combination rule sig-nificantly underpredicted the likelihood ofresponding to the compound. This un-derprediction was construed as supportingthe general hypothesis that some perceptualintegration (e.g., Bellingham &Gillette, 1981;Heinemann & Chase, 1975; Kehoe & Gor-mezano, 1980) between the components oc-curred over the CS-US interval. Under aconfigural hypothesis, it would be expectedthat the level of responding to the compo-nents would be largely attributable to stim-ulus generalization from the associativestrength of the configural stimulus. In thecase of the 73-dB, 800-msec tone in Exper-iments 2 and 3, the level of responding wasgenerally low, and, thus, with even modeststimulus generalization from a configuralstimulus, it would be possible to account forresponding to the 73-dB, 800-msec tone.Despite the anomalies between overshad-owing and salience measurements, the over-shadowing results agree with Mackintosh's(1976) hypothesis that reciprocal overshad-owing will be found only when the CSs areof relatively low salience. Thus, these resultstend to challenge a strict inverse hypothesisof overshadowing (e.g., Rescorla & Wagner,1972) and favor the more open-ended Mack-intosh (1975, 1976) model or the perceptualformulations (Heinemann & Chase, 1975;Hull, 1943). Consequently, the theoreticalburden now shifts toward detailing the exactconditions under which reciprocal over-shadowing occurs. In connection with theMackintosh (1975, 1976) model, it will benecessary to specify more precisely the initiallearning rate parameters, the size of the dec-rements in them from trial to trial, and the

mapping of associative strength unto overtresponding (cf. Moore & Stickney, 1980). Forthe perceptual formulations, it will be nec-essary to obtain independent scales of thegeneralization decrements between the com-pound and its components (cf. Kehoe &Gormezano, 1980).

Reference Note1. Tail, R. W., Kehoe, E. J., & Gormezano, I. Effectsof US duration on classical conditioning of the rabbit'snictitating membrane response. Unpublished manu-script, Universityof Iowa, 1982.

ReferencesBellingham, W. P., & Gillette, K. Spontaneous config-uring to a tone-light compound using appetitive train-ing. Learning and Motivation, 1981, 12, 416-428.Blough, D. S. Recognition by the pigeonof stimuli vary-ing in two dimensions. Journal of the ExperimentalAnalysis of Behavior, 1972, 18, 345-367.Blough, D. S. Steady state data and a quantitative modelof operant generalization and discrimination. Journalof Experimental Psychology: Animal Behavior Pro-cesses, 1975, 1, 3-21.Gillette, K., & Bellingham, W. P. Loss of within-com-pound flavour associations: Configural precondition-ing. Experimental Animal Behaviour, 1982, 1, 1-17.Gormezano, I. Classical conditioning. In J. B. Sidowski(Ed.), Experimental methods and instrumentation inpsychology. New York: McGraw-Hill, 1966.Heinemann, E. G., & Chase, S. Stimulus generalization.In W. K. Estes (Ed.), Handbook of learning an d cog-nitive processes: Vol. 2. Conditioning and behaviortheory. Hillsdale, N.J.: Erlbaum, 1975.Hull, C. L. Principles of behavior. New York: Appleton-Century-Crofts, 1943.James, J. H., &Wagner,A. R. One-trial overshadowing:Evidence of distributive processing. Journal of Ex-

perimental Psychology: Animal Behavior Processes,1980, 6, 188-205.Kamin, L. J. Selective association and conditioning. InN. J. Mackintosh & F. W. K. Honig (Eds.), Funda-mental issues in associative learning. Halifax, NovaScotia: Dalhousie University Press, 1969.Kehoe, E. J. The role of CS-US contiguity in classicalconditioning of the rabbit's nictitating membrane re-sponse to serial stimuli. Learning and Motivation,1979, 10, 23-38.Kehoe, E. J., Feyer, A., & Moses, J. L. Second-orderconditioning of the rabbit's nictitating membrane re-sponse as a function of the CS2-CS1 and CS1-USintervals. Animal Learning &Behavior, 1981, 9, 304-315.Kehoe, E.J., &Gormezano, I. Effects oftrials per sessionon conditioning of the rabbit's nictitating membraneresponse. Bulletin of the Psychonomic Society, 1974,2, 434-436.Kehoe, E. J., &Gormezano, I. Configuration and com-bination lawsin conditioning with compound stimuli.Psychological Bulletin, 1980, 87, 351-378.


16/16

328 E. JAMES KEHOEKehoe, E. J., Schreurs, B. G,, & Amodei, N. Blockingacquisition of the rabbit's nictitating membrane re-sponse to serial conditioned stimuli. Learning andMotivation, 1981, 12, 92-108.Mackintosh, N. J. A theory of attention: Variation in

the associability of stimuli with reinforcement. Psy-chological Review, 1975, 82, 276-298.Mackintosh, N. J. Overshadowing and stimulus inten-sity. Animal Learning &Behavior, 1916,4,186-192.Mackintosh, N. J., & Reese, B. One-trial overshadowing.Quarterly Journal of Experimental Psychology, 1979,31, 519-526.Marchant, H. G., Ill, & Moore, J. W. Blocking of therabbit's nictitating membrane response in Kamin'stwo-stage paradigm. Journal of Experimental Psy-chology, 1973, 101, 155-158.Miller, R. G. Simultaneous statistical inference. NewYork: McGraw-Hill, 1966.Moore, J. W., & Stickney, K. J. Formation of atten-tional-associative networks in real time: Role of thehippocampus and implications for conditioning.Physiology and Behavior, 1980, 8, 207-217.Pavlov, I. P. Conditioned reflexes: An investigation ofthe physiological activity of the cerebral cortex (G. V.Anrep, trans.). London: Oxford University Press,1927.Razran, G. Empirical codification and specific theoret-ical implicationsofcompound-stimulusconditioning:Perception. In W. Prokasy (Ed.), Classical condition-ing. New York: Appleton-Century-Crofts, 1965.Razran, G. Mind in evolution.New York: Houghton-

Mifilin, 1971.Rescorla, R. A. "Configural" conditioning in discrete-trial bar pressing. Journal of Comparative and Phys-iological Psychology, 1972, 79, 307-317.Rescorla, R. A. Evidence for "unique stimulus" accountof configural conditioning. Journal of Comparativeand Physiological Psychology, 1973, 85, 331-338.Rescorla, R. A. Simultaneousassociations. In P. Harzem& M. D. Zeiler (Eds.), Advances in analysis of behav-ior: Vol. 2. Predictability, correlation, & contiguity.New York: Wiley-Interscience, 1981.Rescorla, R. A., Simultaneous second-order condition-

ingproduces S-S learning in conditioned suppression.Journal of Experimental Psychology: Animal BehaviorProcessess, 1982, 8, 23-32.Rescorla, R. A., & Cunningham, C. L. Within-com-pound flavor associations. Journal of ExperimentalPsychology: Animal Behav ior Processes, 1978,4,267-275.Rescorla, R. A., & Wagner, A. R. A theory of Pavlovianconditioning: Variations in the effectiveness of rein-forcement and nonreinforcement. In A. Black &W. F. Prokasy (Eds.), Classical conditioning II. NewYork: Appleton-Century-Crofts, 1972.Saavedra, M. A. Pavlovian compound conditioning inthe rabbit. Learning and Motivation, 1975, 6, 314-326.Scandrett, J., & Gormezano, I. Microprocessor controland A-D data acquisition in classical conditioning.Behavior Research Methods & Instrumentation, 1980,12 , 120-125.Smith, M. C., Coleman, S.R., &Gormezano, I. Classicalconditioning of the rabbit's nictitating membrane re-sponse at backward, simultaneous, and forward CS-US intervals. Journal of Comparative and Physiolog-ical Psychology, 1969, 69, 226-231.Sutherland, N. S., & Mackintosh, N. J. Mechanisms ofanimal discrimination learning. New York: AcademicPress, 1971.Whitlow,J. W., Jr., &Wagner, A. R. Negative patterningin classical conditioning: Summation of response ten-dencies to isolable and configural components. Psy-chonomic Science, 1972, 27, 299-301.

Wickens, D. D. Stimulus-response theory as applied toperception. In Learning theory, personality theory,an d clinical research: The K entucky symposium.NewYork: Wiley, 1954.Wickens, D. D. Conditioning to complex stimuli.Amer-ican Psycho logist, 1959,14, 180-188.Wickens, D. D. Compound conditioning in humans andcats. In W. F. Prokasy (Ed.), Classical conditioning.New York: Appleton-Century-Crofts, 1965.

Received September 18, 1981

Correction to Timberlake,Wahl, and KingIn the article "Stimulus and Response Contingencies in the MisbehaviorofRats" by William Timberlake, Glenda Wahl, and Deborah King (Journal ofExperimental Psychology: Animal Behavior Processes, 1982, Vol. 8, No. 1, pp.62-85), the abscissa of Figures 1, 2, 5, and 6 were incorrectly labeled. In eachcase, the word DAYS should replace the word TRIALS.

313-328 overshadowing and summation in compound stimulus conditioning of the rabbit's nictitating...

Documents