JOURNAL OF THE EXPERIMENTAL ANALYSIS OF BEHAVIOR

ACQUISITION OF MATCHING TO SAMPLEVIA MEDIATED TRANSFER1

MURRAY SIDMAN, OSBORNE CRESSON, JR.,AND MARTHA WILLSON-MORRIS

NORTHEASTERN UNIVERSITY AND THE EUNICE K. SHRIVER CENTERAT THE WALTER E. FERNALD STATE SCHOOL

Two severely retarded Down's-syndrome boys learned a matching-to-sample performancethrough mediated transfer. The transfer paradigm involved three sets of stimuli, one audi-tory set (A) and two visual sets (B and C). The subjects were taught directly to do B-Aand C-B matching, but experienced no direct association between C and A. They acquiredthe ability to do C-A matching without having been taught that performance directly.They also learned indirectly to name some of the visual stimuli, but naming was apparentlynot the mediator in the emergent C-A matching. The use of words and letters as stimulihighlighted the possible relevance of mediated associations in the indirect acquisition ofelementary reading comprehension and oral reading. The acquisition of matching viamediated transfer also raised some new considerations concerning the role of coding re-sponses in arbitrary matching to sample.

In his germinal paper on mediated associa-tion, or mediated transfer, Peters (1935) citedearly associationist philosophers and psycholo-gists in his description of this phenomenon:". . . the successive appearance of apparentlyunrelated ideas in consciousness can be ac-counted for only on the basis of a third ideawhich is definitely related to both of them.. . . the term 'mediate association' refers to anindirect associative connection between twoitems of experience, ideas, or mental acts, asopposed to a direct type of associative connec-tion. And this indirect associative connectionis one supposedly mediated by a third itemthat has been directly associated with both ofthem. When we find two ideas A and B ap-pearing in succession with no possibility oftheir previous, direct association, we will usu-ally find that both A and B have formerly beenassociated with an item C. Thus, the connec-tion A-B is the result of two other connections,A-C and B-C" (Peters, 1935, pp. 20-21).

Subsequent experimentation, usually usingpaired-associate techniques in the verbal learn-ing tradition, has amply confirmed the exis-

'This research was supported by Grants HD 05124and HD 04147 from the National Institute of ChildHealth and Human Development. We are indebted toF. Garth Fletcher for his technical assistance, and toNan Haar for performing the IQ evaluations. Reprintsmay be obtained from Murray Sidman, Department ofPsychology, 440UR, Northeastern University, Boston,Massachusetts 02115.

tence of mediated transfer, although the de-scriptive language has tended away from ideasand mental acts to stimuli and responses (Jen-kins, 1963).We have recently demonstrated mediated

transfer by using matching-to-sample proce-dures with retarded subjects (Sidman, 1971;Sidman and Cresson, 1973). In these earlierexperiments, the subjects first learned to matchsets of 20 pictures and 20 printed picturenames to corresponding dictated names. Theythen proved able for the first time to matchprinted words to the corresponding pictures,with word-picture matching never having beentaught directly.The relation of the matching-to-sample ex-

periments to the familiar mediated-transfer, ormediated-association paradigm, can be seen inFigure 1. A, B, and C are three sets of stimuli.Arrows point from comparison to sample stim-uli. Subjects first learned to match B to A, andC to A; they were then able to match B to C,and C to B.

In this example, matching printed words topictures and pictures to printed words may beregarded as simple reading comprehension;matching pictures to dictated words as simpleauditory comprehension; and the matching ofprinted to dictated words may be called "audi-tory receptive reading". The earlier studiesthus had some relevance both to the principlesand practicalities of teaching elementary read-ing.

261

1974, 22, 261-273 NUMBER 2 (SEPTEMBER)

MURRAY SIDMAN et al.

The present experiment, however, was con-cerned with the nature of the mediated-trans-fer process itself. One interpretation of theoriginal finding is that the subjects' emergentability to match printed words to pictures wasmediated by their previous training to matchboth of these sets of stimuli to the same audi-tory stimuli. For example, in the course oflearning the crossmodal matching, they hadlearned to match the picture of a car and theprinted word car to the dictated word "car";they were then able to match the picture tothe printed word through the mediation ofthe dictated word.A second interpretation is also possible. Be-

fore they learned the visual-auditory matchingtasks, the subjects had some proficiency innaming the pictures, but could not name theprinted words. After they had learned thecrossmodal matching tasks, they proved ablefor the first time to name the printed words-to read them aloud. The emergence of oralreading raised the possibility of a differentmediation path for the indirectly learnedword-picture matching. This path is illustratedby the triangle on the right of Figure 1. Forexample, subjects had learned to say "car" inresponse to pictures of cars (B-D) and to theprinted word car (C-D); they were then able tomatch picture with printed word (B-C and C-B)through the mediation of the spoken word.Was the subjects' emergent reading compre-

hension (matching printed words to pictures)mediated by the dictated words spoken to

'RCEPT IVE' EXPRESSIVE

Fig. 1. Stimuli, responses, and mediated-transferparadigms for Subject J.C. In the "receptive" triangle,arrows point from comparison to sample stimuli. Onthe "expressive" side, arrows point from visual stimulito oral naming responses.

them? Or did names spoken by the subjects,either aloud or implicitly, mediate their word-picture matching? The first possibility wouldimply stimulus mediation via receptive chan-nels. The second would require response medi-ation, and would be consistent with stimulus-response models of mediated transfer (Jenkins,1963; Schoenfeld and Cumming, 1963). Toemphasize the distinction, the mediation tri-angle on the left side of Figure 1 has been la-belled, "receptive", and the triangle on theright, "expressive".A major purpose of the present experiment

was to determine whether mediated transfer,demonstrated within the context of matchingto sample, requires a response as the mediatoror whether stimulus mediation is sufficient tocarry the explanatory burden. Are the pre-requisites for the emergence of B-C and C-B tobe specified as B-A and C-A, or as B-D and C-D(Figure 1)?

If naming mediated word-picture matchingin the earlier experiments, it becomes neces-sary to explain how word naming itselfemerged. A possibility was suggested (Sidmanand Cresson, 1973) when one subject's picturenaming improved as a consequence of learningauditory comprehension (matching pictures todictated words). Could he also have learnedoral reading (naming printed words) simplyas a consequence of learning auditory recep-tive reading (matching printed to dictatedwords)? If so, printed-word naming and pic-ture naming might then have mediated read-ing comprehension (picture-word matching).Although the conditions under which recep-

tive training facilitates expressive speech arestill unknown (Guess, 1969; Guess and Baer,1973), it seems reasonable to suppose that thesubjects learned to read the words aloud be-cause of their receptive training in matchingprinted words to dictated names. In the pres-ent experiment, we eliminated this avenue forthe learning of oral reading, in the hope thatwe could assess the likelihood of receptivelymediated transfer.By proceeding around the receptive triangle

(Figure 1) in anotlher direction, we preventedthe subject from experiencing any direct asso-ciation between printed words and their audi-tory counterparts. He was taught auditorycomprehension (B-A) and reading comprehen-sion (C-B) directly, and was then tested for theemergence of auditory receptive reading (C-A)

262

MEDIATED TRANSFER IN SAMPLE MATCHING

and oral reading (C-D). If the absence of directC-A training prevented the subject from learn-ing to name the printed words, but C-A itselfstill emerged from the B-A and C-B training,the possibility of receptive mediation wouldhave to be seriously considered.Attainment of this objective involved the

assumption that the outcomes of the media-tion paradigm used in the previous studies(B-A, C-A, test for B-C) and in the present ex-periment (B-A, C-B, test for C-A) would besimilar-that mastery of the first two stageswould result in mastery of the third. On thebasis of paired-associate experiments (Jenkins,1963), this assumption appeared reasonable,but its test in the contexts of matching to sam-ple and crossmodal associations was a secondobjective.A third objective was to use another set of

stimuli, upper- and lower-case letters, to testthe generality of mediated learning of match-ing to sample. Two experiments are described;each used different stimuli and differed in cer-tain procedural features.

METHOD

SubjectsThe subjects were two institutionalized

Down's-syndrome boys, one (J.C.) 14 yr and9 months and the other (P.A.) 18 yr old at thestart of the first complete test session. Bothwere participants in a behavior modificationproject (Mackay and Sidman, 1968; Sidman,1970), but received no reading instruction dur-ing the period covered by the experiment.Their experimental sessions were scheduled soas not to interfere with their project activities,and the pennies and tokens they earned werespent in accord with contingencies prevailingin the project. Approximately two monthsafter the experiment was completed, SubjectJ.C. achieved a mental age of 3-4 on the Pea-body Picture Vocabulary Test, 4-11 on theStanford-Binet L-M Intelligence Test, and 6-9on the Leiter Performance Scale. Subject P.A.did not undergo such formal testing, but allwho were acquainted with him during the 8 yrof the project judged him considerably lessadvanced than Subject J.C. in all areas offunctioning.

General Experimental DesignThe subjects were to be taught two legs of

the receptive mediation triangle, the cross-modal matching task B-A and the visualmatching task C-B. They were then to betested for the emergence of the third leg, thecrossmodal matching task C-A. Also to betested was the effect of the teaching on the twooral naming tasks, C-D and B-D. First, the sub-jects' pre-experimental capabilities were as-sessed by pretesting them on all these tasks.They were also pretested on, and if necessary,were taught identity matching (Subject J.C.:printed picture names to printed picturenames; Subject P.A.: printed upper-case toupper-case, and lower-case to lower-case let-ters) in order to ensure that they could dis-criminate the visual stimuli. The left columnsof Figures 3 and 4 list the types of stimuli andthe responses required of the subjects in eachkind of test. This test battery, administeredbefore, during, and after the teaching (see be-low), provided the experimental data.

After the pretests, teaching was accom-plished gradually, first one leg of the receptivetransfer triangle (B-A) and then the other(C-B). Tests interspersed at various stages ofthe teaching evaluated the requisite baselinebehavior, and evaluated nonspecific transfereffects arising from the teaching proceduresthemselves, apart from the associative transferwith which the experiment was concerned.Variations in this general procedure are de-scribed below.

Apparatus and General Test ProceduresThe subject sat before a panel of nine trans-

lucent windows onto which stimuli were pro-jected from the rear. The windows, each 7.32cm in diameter, were arranged in a circle ofeight with the ninth in the center (Figure 2).The display diameter, from outer edge toouter edge, was 32.72 cm; the center-to-centerdistance was 9.53 cm between adjacent win-dows on the perimeter, and 12.7 cm betweenthe center window and each of the others.Each trial began with a sample stimulus.

Visual samples (black letters, words, or linedrawings on a white background) appeared onthe center window. With auditory samples,words or letter names were dictated and re-peated through a speaker at 2.0- to 2.5-sec in-tervals by a continuous-loop tape system(Fletcher, Stoddard, and Sidman, 1971); thecenter window was illuminated but blank.Each sample, visual or auditory, remained

263

MURRAY SIDMAN et al.

Q)erha

a x

ao

nE iFig. 2. Schematic diagrams of the display matrix, illustrating one trial from the identity word-word matching

test, and one trial from the test of matching lower-case to upper-case letters. In other kinds of tests, the samplewindow (center) or the comparison windows (periphery) contained pictures or were blank.

present throughout the trial, and trial dura-tions had no time limit.In matching-to-sample tests, the subject

pressed the center window to bring compari-son stimuli, always visual, onto the outer win-dows. This "observing response" was alwaysrequired, even when the sample was dictated.On each trial, one comparison stimulus, thecorrect one, corresponded to the sample. Sub-ject J.C. had seven incorrect comparison stim-uli (words or pictures) on every trial. SubjectP.A. had five incorrect comparison stimuli (up-per- or lower-case letters) and two blank win-dows on every trial (Figure 2). Sample stimuliand window positions of correct and incorrectcomparison stimuli varied from trial to trial.

After the comparison stimuli appeared, thesubject pressed a comparison window. Correctchoices caused chimes to ring and a penny(Subject J.C.) or a token (Subject P.A.) to bedelivered. Neither chimes nor penny or tokenfollowed an incorrect choice. All stimuli dis-appeared after each choice, and 1.5 to 2.0 seclater a new sample began the next trial (non-correction procedure). Any window press dur-ing the intertrial interval postponed the nextsample for 2 sec. Solid-state logic scheduledthe procedures and recorded the responses au-tomatically. Test sessions lasted about 1 hr.In oral naming tests, the subject had simply

to name the sample picture, word, or letteraloud. He was asked, "Tell me what you see".

He pressed no windows, and no comparisonstimuli appeared. Oral responses of both sub-jects were clear, and there was no ambiguityin determining whether they were correct.After each response, the experimenter presseda handswitch to activate the reinforcement (ifcorrect), intertrial-interval, and slide-changeapparatus.

Visual and auditory samples never occurredtogether. With dictated samples, the centerwindow was always blank; with visual word,letter, or picture samples, no stimuli weredictated.

Subject J.C.'s tests each had 20 trials. Thestimuli, taken from a list of 20 pictures, or theprinted (lower case) or dictated names of thepictures, were: axe, bed, bee, box, boy, bug,car, cat, cow, dog, ear, hat, hen, hut, hoe, man,pie, pig, saw, zoo.

Subject P.A.'s tests each had 24 trials. Thestimuli were six letters, A, B, E, G, L, and N,presented in upper-case or lower-case form oras dictated letter names.

General Teaching ProceduresTeaching, which took place between each

administration of the test battery, always in-volved matching to sample, sometimes with acorrection and sometimes with a noncorrec-tion procedure (see below). On the correctionprocedure, an error produced the usual inter-trial interval, after which the same sample and

264

MEDIATED TRANSFER IN SAMPLE MATCHING

comparison display were presented again. Onboth the correction and noncorrection proce-dure, chimes and a penny or token followedany response to the correct comparison win-dow.Each test listed in Figures 3 and 4 had six

versions, differing in their sequence of samplepresentations and in the arrangements of stim-uli on the panel. Also, each test set used a dif-ferent variation of each picture. For example,the car in one set was a VW, and in anotherwas an MG; the cat in one set was standing,and in another was lying down; etc. One ver-sion of each test was used only in the test bat-tery. A systematic rotation through the otherfive sets during teaching prevented the subjectsfrom learning display configurations, irrele-vant display features, and sequences of samplesor window positions, all of which would havepermitted them to achieve high scores withoutobserving the samples. Varied numbers of setswere presented during teaching sessions, whichlasted 20 to 65 min and occurred one to fivetimes a week.

Order of Conditions and Detailed TeachingProcedures for Subject J.C.Table 1 summarizes the subject's testing and

teaching sequence. Dates correspond to thoseat the bottom of Figure 3.

Pretesting. Three and 2 yr before the experi-ment began, Subject J.C. received the tests ofauditory and reading comprehension. Duringthese years he also had considerable matching-to-sample and oral-naming experience withother kinds of visual, auditory, and tactilestimuli, including colors and color names,numbers and number names, nonsense tri-grams, and geometric forms. In preparationfor the present experiment he received thecomplete test battery twice. The early scoreshave been included in order to show that mereexposure to the tests did not appreciably im-prove his reading and auditory comprehensionor his oral reading.

Preteaching. Some of Subject J.C.'s low pre-test scores (see Results) might have beencaused simply by an inability to distinguishthe printed words from each other. He wastherefore taught word-word identity matching,selecting each of the 20 printed words fromdifferent sets of eight alternatives in responseto printed-word samples. A typical display isin the left side of Figure 2. The teaching in-

Table 1

Testing and Teaching Sequence for Subject J.C.

Procedures Dates

A. Pretesting1. Test: Matching printed

words to pictures (C-B) andpictures to dictated words(B-A)

2. Test: Samie3. Test: Complete battery4. Test: Complete battery

B. Preteaching1. Teach: Identity matching of

printed words to printedwords

2. Test: Complete battery

C. First leg of transfer triangle(B-A)1. Teach: Matching pictures to

dictated words2. Test: Comtiplete battery

D. Second leg of transfer triangle(C-B)1. Teach: Matching 9 printed

words to pictures2. Test: Complete battery3. Teach: Matching 14 printed

words to pictures4. Test: Complete battery5. Teach: Matching 20 printed

words to pictures6. Test: Complete battery7. Reteach: Matching 20

printed words to pictures8. Test: Complete battery

7/07/679/27/688/26/709/10/70

9/30/70 to 11/12/7011/15/70

11/17/70 to 12/04/7012/08/70

12/09/70 to 12/30/7012/30/70

1/05/71 to 1/24/711/24/71

2/03/71 to 2/15/712/15/71

2/18/71 to 2/24/712/26/71

volved sets of noncorrection trials with all 20samples, and correction trials with subsets ofthose samples he found troublesome. After1240 trials he achieved a criterion of 95% cor-rect in 120 trials of word-word identity match-ing, and was given the test battery again.

First leg of transfer triangle. He nextlearned auditory comprehension, matchingeach variation of the 20 pictures to its dictatedname. Again, a mixture of correction and non-correction procedures was used, the correctionprocedure being applied to special sets of sam-ple stimuli with which the subject had diffi-culty. He required 540 trials to reach a cri-terion of 95% correct in 120 trials, and thenreceived the test battery again.Second leg of transfer triangle. The critical

teaching involved the matching of printed-word comparisons to picture samples. Five

265

MURRAY SIDMAN et al.

variations of each picture were presented inmixed sequences. The correction procedurewas used throughout. Subject J.C. was firsttaught to match only nine of the printed wordsto their equivalent pictures. He was initiallygiven sets of 10 trials with only two samplepictures, axe and zoo. When he achieved ascore of 100% in a set of 10 trials, he was thengiven a second pair of pictures, hen and saw,to learn. After achieving 100% with the sec-ond pair, he was given variations of all fourpictures as samples until he reached a criterionof 19 correct trials in a set of 20. This systemof teaching the subject to match new wordsand then adding those to the previouslylearned set continued, with the following se-quence of pairs: axe and zoo, hen and saw,man and bug, bug and dog (only one newword at this point), hoe and cow. He required545 trials to meet the criterion for matchingthe first nine printed words to their corre-sponding pictures, after which he received thetest battery again.

Subject J.C. then learned to match five moreprinted words to their equivalent pictures.These were taught in the following pairs: hutand hat, bed and bee, bed and boy. Each pairwas presented until he achieved perfect per-formance in a set of 10 trials. He was thengiven a complete set of 20 sample pictures, in-cluding the six he had not yet been taught. Ifhe did not score perfectly on all 14 sampleshe had been taught, he was given remedialtrials with those he had matched incorrectly.When he achieved perfection on the remedialset, he was retested with the complete set of20 samples. This process of testing with thecomplete set and then giving remedial teach-ing on those he had been taught but founddifficult continued until, when given the com-plete set, he correctly matched all the wordshe had been taught. The subject met this cri-terion for the 14 samples in 640 trials and re-ceived the test battery.The next teaching phase, carried out in the

same manner as the preceding phase, addedthe last six pictures in the following sequenceof pairs: boy and box, cat and car, car and ear,pig and pie. The subject required 380 trialsto reach the learning criterion which, by thistime, included all 20 samples. He then re-ceived the battery of tests again.The final teaching stage involved special

remedial sets of samples the subject sometimes

still matched incorrectly. This teaching usedall six sets of word-picture matching trials, in-cluding the test set, and the subject achieved100% on all six sets after 380 trials. The finaltest battery followed this reteaching.

Order of Conditions and Detailed TeachingProcedures for Subject P.A.

For this subject, all six letters were compari-son stimuli on every trial. He was taught tomatch printed upper-case letters to six dictatedletter names, and printed lower-case lettersto the six corresponding upper-case letters(Table 2).

Table 2Testing and Teaching Sequence for Subject P.A.

Procedures Dates

A. Pretesting1. Test: Complete battery 10/30/722. Test: Same 1/23/73

B. Two legs of transfer triangle1. Teach: Alternate visual-audi-

tory matching (B-A) andvisual-visual matching (C-B)for each letter consecutively 1/26/73 to 5/11/73

2. Test: Complete battery 5/11/733. Teach: Review previously

taught matching, and teachthe other visual-visual match-ing task (B-C) 5/14/73 to 5/30/73

4. Test: Complete battery 5/30/73

Pretesting. Like Subject J.C., Subject P.A.had considerable matching-to-sample and oral-naming experience with a variety of stimuliother than those used here. He was pretestedtwice on the complete battery, and did nothave to be taught identity matching.Two legs of transfer triangle. First, the sub-

ject learned to match two upper-case compari-son stimuli, B and L, to their dictated names.When he reached 100% in a set of 24 trials, hethen learned to match the two lower-case let-ters to their upper-case equivalents. A typicaldisplay is in the right side of Figure 2. Whenhe reached 100% in the visual-visual task, hereturned to visual-auditory matching. This al-ternation continued until he went through 96consecutive trials, two sets of each task, per-fectly within a single session.A third sample, "A", was then added. When

all three upper-case letters were matched per-fectly to their dictated equivalents, the subject

266

MEDIATED TRANSFER IN SAMPLE MATCHING

was given the visual-visual task of matchingthe three lower-case comparisons to their ap-propriate upper-case samples. When hereached criterion on each task the other wasre-introduced, the alternation continuing untilhe achieved perfect scores in four consecutive24-trial sets, two sets of each type, within asingle session.The same procedure then continued, one

new letter being added in each teaching step.The final samples to be introduced were "G","N", and "E". After achieving a perfect per-formance with all six letters in both matchingtasks, the subject had to meet additional cri-teria before receiving the test battery. First, hehad to perform perfectly again on four alter-nating 24-trial sets, two of each type, withinone session. Then, on the day of the test bat-tery, he had to perform perfectly on one 24-trial set of each type. He required 5768 trialsto meet these criteria.

In the final teaching phase, Subject P.A. re-ceived additional experience with the twotasks he had been taught. He also learned tomatch upper-case comparison letters to lower-case samples. With all six letters and letternames as samples, the teaching proceeded asbefore, with the addition of the third match-ing task. The subject again had to proceederrorlessly through a 24-trial set before goingon to the next task. He also had to performerrorlessly on six consecutive 24-trial sets, twoof each type, within a single session. Then, inthe same session as the final test battery, hehad to perform three 24-trial sets perfectly, oneof each type. This teaching phase required1824 trials.

RESULTS

Subject J.C.Pretests. The sequence of teaching proce-

dures and test results can be followed by scan-ning the dated columns in Figure 3. The sub-ject's test scores before explicit teaching arein the four left columns of bars. In 1967 and1968 (the first two columns), he matched pic-tures to dictated and printed words poorly, al-though he did somewhat better in auditorythan in reading comprehension. His poorestperformances on the complete test battery(8/26/70 and 9/01/70) were in the four testsat the top, all involving printed words. He didbetter, although short of perfection, in picture

naming and auditory comprehension, whichdid not involve printed words. In identitymatching of printed word to printed word, hescored no higher than 70%.

Tests after teaching identity matching (pre-teaching). After being taught visual word-wordmatching (9/03/70 to 11/12/70), the subjectdisplayed his new skill by scoring 100% on theidentity matching test (11/15/70). The otherscores changed little; his poor reading did notreflect an inability to discriminate the printedwords.

Tests after teaching auditory comprehen-sion (first leg of transfer triangle). After he hadlearned auditory comprehension, Subject J.C.scored 100% on his newly taught skill (12/08/70). Learning to match pictures to dictatedwords also seemed to improve his picture nam-ing slightly, although he did not consistentlydo as well in later tests. The change in audi-tory comprehension was not accompanied byappreciable increases in any of the readingscores. Learning one leg of the mediation tri-angle (B-A) did not cause the other legs toemerge. In the three matching tests of reading,his scores were close to the chance level foreight comparison stimuli per trial, and oralreading remained at zero. These poor scoresapparently did not stem from an inability todistinguish the pictures from each other, orthe dictated or printed words from each other.Although some difficulty in oral reading mayhave stemmed from an inability to say thewords aloud, the subject's picture naming in-dicated that this problem could have beenonly minor and could not have accounted forhis complete lack of oral reading.

Tests after teaching reading comprehension(second leg of transfer triangle). Would theestablishment of this performance, matchingprinted words to pictures, cause the emergenceof the third leg of the triangle, matchingprinted words to dictated words?Double bars in the four reading tests (12/

30/70) distinguish the scores for the ninewords he was first taught to match to pictures(+) and the scores for the 11 words he had notyet been taught (-). The direct effect of theteaching was a large increase in the subject'saccuracy in matching the nine printed wordsto their equivalent pictures, an increase notreflected in his score for the 11 untaughtwords. The reason for his less-than-perfectscore on the nine critical words is unclear.

267

MURRAY SIDMAN et al.

SAMPLESTIMULUS RESPONSE T,EST SCORES (PERCENT CORRECT) S:J.C.100. LL LPRINTED NAME 50 L

25WORDo ORALLY

MATCH 1001751

AUDITORY 0PRINTED 1PRINTD |251

WORD WORtD J

MATCH 7 I

PRItNTED 5SWORD PICTURE

.~~~~* L .. c ec.. leec

MATCH 10 11 i L I

PRINTED II l lPICTURE WORD a1rds kdwords -words words

1001 1 1 1 1 n y 1

N4AME 50

PICTUR-E ORALLY 0Jm a -1001 1U UU

MATCH 7511 Lu . .AUDITORY ao ecWORD PICTURE US Iwords

MATCH10LI II

PRINTED 50 1ocPRINTED 25 U aOll5 U

WORD WORD wordslu lme~~0* 0e 00to. Wwu-

FPf lzac

Fig. 3. Subject J.C. The two left columns identify the sample stimuli and responses that comiprised each type oftest. The row of bars to the right of each test represents the subject's scores in successive administrations of thetest. The dates of each test, and of the interpolated teaching, are given below the columns of bars. In some testseries, each of the four reading tests has two scores: the bars marked + represent the subject's performance withwords he had been taught to match in the word-picture matching task; the bars marked - represent his perform-ance with words he had not yet been taught.

Accuracy also improved considerably in match-ing picture-comparison stimuli to printed-word samples, although he had not receivedexplicit training on this task.

In matching printed to dictated words, thesubject improved modestly on the nine criticalwords. This improvement did not extend tothe 11 untaught words, and the selectivity ofthe change provides evidence that the im-provement was a transfer effect from the pre-vious teaching.

In oral reading, the subject named one ofthe printed words aloud for the first time, buthis score did not show an improvement com-parable to that of the matching tasks.Although the subject's scores after he had

been taught 14 of the 20 words (1/24/71) weresimilar to those of the preceding tests, the per-centages represent a larger number of correctresponses. Furthermore, all of his correctchoices in response to dictated samples wereprinted words learned in word-picture match-

MEDIATED TRANSFER IN SAMPLE MATCHING

ing. In spite of his improvement in matchingprinted to dictated words, he still showed noability to read the printed words aloud.Having learned to match all 20 printed

words to picture samples, the subject correctlymatched 12 (60%) of the printed words todictated samples (2/15/71).

Also, for the first time, his oral reading im-proved distinctly. The way this came about isof interest. In the first seven trials of this test,he responded to the printed words by utteringnonsense syllables or letter names unrelated toletters within the words. On Trial 8, with beeas the sample word, he said, "bee-fire". OnTrial 9, with box as the sample word, he sud-denly sat up straight, clapped his hands,laughed, said, "I get it", and responded cor-rectly. A similar "aha" reaction occurred onTrial 10. After failing the first seven trials, hewent on to name nine of the next 13 wordscorrectly. The first seven trials were then re-peated, and he read three of those words aloudcorrectly. The correct responses upon retestwere included in the score shown in Figure 3.Although transfer to auditory receptive

reading and to oral reading did occur, thistransfer was far from perfect. Furthermore,the subject fell short of 100% even on theword-picture matching task he had beentaught. Therefore, he was retaught word-pic-ture matching and received the test series oncemore (2/26/71).

This time, Subject J.C.'s word-picturematching was almost perfect, as was picture-word matching, which had not been directlytaught. Along with this improvement was anincreased accuracy in matching printed to dic-tated words. Oral reading, however, remainedat its previous level.

Subject P.A.Pretests. The subject did visual-visual iden-

tity matching of upper- and lower-case lettersperfectly in pretests, and continued errorlesslyin every administration of the test battery.These consistently perfect performances weretherefore omitted from Figure 4.The two initial administrations of the test

battery, nearly three months apart (Figure 4;10/30/72 and 1/23/73), yielded consistent re-sults. His best scores, less than 50%, were onthe two tasks he was to be taught directly. Theother matching tasks were within the range ofchance performance for a display of six com-

parison stimuli. He showed a slight ability toname upper-case letters aloud, but was com-pletely unable to name the lower-case letters.

Tests after teaching two legs of the transfertriangle. Although the subject maintained analmost perfect performance on the visual-audi-tory task he had been taught, visual-visualmatching deteriorated from the relatively rig-orous learning criterion he had attained (5/11/73). On the reverse visual-visual task,lower-case samples and upper-case compari-sons, in which he had not received direct train-ing, his performance was still less satisfactory.Nevertheless, he did better on these tasks afterthe teaching than before, and the improve-ment was reflected in considerable transfer tothe critical visual-auditory task-matchinglower-case letters to their dictated names.

Oral naming of upper-case letters also im-proved considerably over the pretest perform-ances.

Oral naming of lower-case letters was testedtwice. For the first time, Subject P.A. named afew lower-case letters correctly, but his scorewas still quite low, as indicated by the first ofthe two bars in the uppermost row (5/11/73).This test was administered before the upper-case naming test, and because he did so muchbetter in the latter we suspected that he mighthave learned something about naming thatcould help him with lower-case letters. Forthis reason, the lower-case test was re-adminis-tered. His score did improve, as indicated bythe second of the two bars in the uppermostrow, but was still quite low.Teaching the subject two matching tasks

brought about a much improved performancein the third matching task, as well as in oralnaming of both sets of letters. The incomplete-ness of the transfer might have reflected hisless-than-perfect matching of lower- to upper-case, and upper- to lower-case letters. He wastherefore retaught the two tasks he had previ-ously learned. He was also taught explicitlyto match upper-case comparison letters tolower-case samples.The final column in Figure 4 (5/30/73)

shows Subject P.A. maintaining perfect per-formances on all three matching tasks he hadbeen taught. In matching lower-case compari-sons to dictated samples, he also scored almostperfectly. His oral-naming scores, however,were practically the same as in the precedingtests.

269

MURRAY SIDMAN et al.

Lo C.LETTER

U.C.LETTER

AUD.

LETTERNAME

L. CoLETTER

U.C.L ETTE R

AUD.

L E TT E R

NAME

NAME

ORALLY

NAME

ORALLY

MATCH

L. C.LETTER

MATCH

U.C.LETTER

MATCHL. C.LETTER

U.C.LETTER

EST SCORES

#tft0

M

0VW

Fig. 4. Subject P.A. The two left columns identify the Fample stimuli and responses that comprised each typeof test. The row of bars to the right of each test represents the subject's scores in successive administrations ofthe test. The dates of each test, and of the interpolated teaching, are given below the columns of bars.

HX

_ _o% o%in in

270

TEACH

100

50 TEAC TEACH

0

MEDIATED TRANSFER IN SAMPLE MATCHING

DISCUSSION

Matching to SampleIn previous experiments (Sidman, 1971; Sid-

man and Cresson, 1973), the ability to matchprinted words with pictures emerged after sub-jects had learned to match each set of visualstimuli to a set of auditory stimuli, dictatedwords. This learning sequence was consistentwith the commonly observed developmentalsequence in which auditory comprehensionprecedes reading comprehension. It was alsoconsistent with the suggestion that the abilityto perform crossmodal tasks is helpful in learn-ing reading comprehension (Birch, 1962; Birchand Belmont, 1964, 1965; Geschwind, 1965;Kahn and Birch, 1968; Wepman, 1962).The present and the earlier experiments to-

gether, however, have shown that the receptivemediation triangle (Figure 1) can be bidirec-tional. Facilitation can apparently work bothways; the normal developmental sequence isapparently not a necessary sequence.The present data suggested also that oral

naming need not mediate the emergence ofvisual-auditory matching. After Subject J.C.had learned to match nine and then 14 printedwords to pictures, he showed a substantialability to match printed to dictated words andalmost no ability to read words aloud. Oralreading could not at this stage have mediatedthe emergent matching of printed to dictatedwords.

After Subject J.C. had learned to match all20 printed words to pictures, he named a sub-stantial number of words for the first time. His"aha" reaction during the oral reading testsuggested that although he had failed to readthe words aloud up to that point, he had actu-ally been capable of doing so, and exercisedhis new capability only in the course of thistest. Nevertheless, even though he may havebeen capable of naming the printed words, heclearly had not been doing so.

Practically all of Subject P.A.'s ability toname upper- and lower-case letters can be at-tributed to the teaching procedures. Afterlearning the two matching-to-sample perform-ances, his performance improved considerablyin matching lower-case letters to their dictatednames. Correlated with this improvement incrossmodal matching was a corresponding im-provement in upper-case letter naming, but on

first testing, only a slight development of

lower-case naming. The emergence of cross-modal lower-case matching in the absence oflower-case naming indicates that the emergentmatching was not mediated by naming, butwas fostered rather by the receptive trainingin the other two matching-to-sample tasks.The improvement in lower-case letter nam-

ing upon retest was analogous to Subject J.C.'s"aha" experience. Subject P.A. was apparentlycapable of naming more lower-case letters thanhe actually did upon initial testing, but he hadnot been using those letter names to mediatethe crossmodal matching of lower-case to dic-tated letters.On the second posttest, Subject P.A.'s match-

ing of lower-case letters to dictated names hadbecome almost perfect, but the two namingperformances remained at their previous lev-els. This again indicated that the mediationwas independent of oral naming.

The Emergence of Printed-Word andLower-Case Letter Naming

It was possible for Subject P.A. to learn up-per-case names in the course of learning tomatch upper-case to dictated letters (B-A).None of the subjects' training, however, in-volved the association of printed lower-caseletters with dictated letter names, or of printedwith dictated words. The interesting possibil-ity here is that receptively mediated matching(C-A) might, in turn, have fostered oral nam-ing of the words and lower-case letters (C-D);oral naming, rather than mediating the emer-gent visual-auditory capability (C-A), mightitself have been a product of that capability.This possibility received some support from

the data: (1) Oral naming did emerge aftercrossmodal matching. (2) In his test on 2/15/71, Subject J.C. matched 12 dictated wordswith their appropriate printed words; he alsonamed nine of these 12 correctly. Of the eightwords matched incorrectly, five were alsonamed incorrectly. On 2/26/71, 10 of the 15words that he matched correctly were alsonamed correctly. Four of the five wordsmatched incorrectly were also named incor-rectly. These data are consistent with the pos-sibility that the receptively mediated matchingof printed to dictated words actually generatedoral reading.On the other hand, oral naming might be

explained by the expressive mediation para-digm. For instance, teaching Subject P.A. di-

271

MURRAY SIDMAN et al.

rectly to match upper-case letters to their dic-tated names (B-A) partially established B-D inthe expressive mediation triangle. Then, teach-ing him directly to match lower- to upper-caseletters established a second side of this trian-gle, C-B. At this point, the third side of theexpressive triangle, C-D, would have emerged,making the subject capable for the first time ofnaming some of the lower-case letters.This second possibility was suggested by two

of Subject J.C.'s naming errors. He gave theincorrect response, "hammer", both to the pic-ture of an axe and to the printed word axe;in response to the picture of a pig and to theword pig, he said, "cow". In these instances, hegave the same incorrect name to the pictureand to the corresponding printed word. Simi-larly, Subject P.A. consistently said, "Seh", inresponse to both upper- and lower-case G dur-ing his final test. It is difficult to attribute suchnaming errors to any process other than thatillustrated by the expressive mediation tri-angle.

Mediation by Stimuli or by Responses?The present study has suggested that new

matching-to-sample performances, not directlytaught, were generated by an apparently re-ceptive mediation process. The suggestion thatan untrained association can be mediated by astimulus is not consistent with stimulus-re-sponse theories, which require differential re-sponses to intervene between stimulus termsin the mediated-transfer paradigm (Jenkins,1963; Schoenfeld and Cumming, 1963). Evenstimulus-stimulus association theories postulateintervening representational processes (Estes,1969) whose functions are difficult to distin-guish from those attributed to response proc-esses.

In arbitrary matching to sample, however,no differential responses to the individualstimuli are required. The necessary overt re-sponse is pointing, or touching, which is thesame for all sample and comparison stimuli.Matching to sample has been accommodatedinto a stimulus-response framework by hypoth-esizing that each sample generates a differen-tial coding response; the stimulus conse-quences of these coding responses then controlappropriate comparison responses (Schoenfeldand Cumming, 1963). Without such coding,matching to sample would have to be viewedas a direct stimulus-stimulus association.

Although it is plausible for each sample togenerate its own coding response, the transfertests in our earlier experiments (Sidman, 1971;Sidman and Cresson, 1973) did not involve theoriginal sample stimuli. The stimuli in thetransfer tests, B and C, had training historiesonly as comparison stimuli. To explain themediated transfer of matching to sample, acoding hypothesis would therefore have to as-sume that each sample and its correct compari-son generated the same coding response. Forexample, the dictated sample word "boy", thecomparison picture of a boy, and the compari-son printed word boy would all have to becoded the same way. The common coding re-sponse could then mediate the matching of thecomparison stimuli to each other.

In the present experiments, the mediationof C-A matching via identical sample and com-parison coding would require an additionalstep. In B-A matching, each B-comparisonstimulus would have had to generate the samecoding response as its corresponding A-sample.In C-B matching, these same B-coding re-sponses would have had to persist and, in addi-tion, transfer to the C-comparison stimuli.Coding responses common to the A-samplesand the C-comparisons could then mediateC-A matching.Names can clearly serve as coding responses

and can meet the requirement of identicalsample and comparison coding. To maintainthe coding hypothesis in the present experi-ments, however, one must assume some classof coding responses other than names. Then,each comparison stimulus would have to gen-erate the same unknown coding response as itsappropriate sample, even though the matchingrelation was arbitrary.

Schoenfeld and Cumming (1963) presenteda strong argument for nonverbal mediatingprocesses, although such processes become diffi-cult to identify in matching-to-sample tasksthat involve large numbers of stimuli. The ad-ditional assumption of identical sample-com-parison coding, required if mediated transferof matching is to be interpreted as other thanstimulus-stimulus learning, does not seem tohave been considered, and is under investiga-tion in this laboratory.

Until these issues have been clarified, itwould seem desirable to restrict the term medi-ation to its procedural sense. If an associationbetween stimuli B and C is established by their

272

MEDIATED TRANSFER IN SAMPLE MATCHING 273

common association with stimulus A, and wecannot identify actual events or processes thatintervene between B and C, the utility of pos-tulating such events or processes will remaindebatable. In the procedural sense, the termmediation refers to the observation that theB-C association, for example, was broughtabout by some form of prior learning that in-volved elements other than B and C. The criti-cal aspects of that prior learning are mattersof primary concern.

REFERENCESBirch, H. G. Dyslexia and the maturation of visual

function. In J. Money (Ed.), Reading disability.Baltimore: Johns Hopkins, 1962. Pp. 161-169.

Birch, H. G. and Belmont, L. Auditory-visual integra-tion in nomial and retarded readers. AmericanJournal of Orthopsychiatry, 1964, 34, 852-861.

Birch, H. G. and Belmont, L. Auditory-visual inte-gration, intelligence, and reading ability in schoolchildren. Perceptual and Motor Skills, 1965, 20,295-305.

Estes, W. K. New perspectives on some old issues inassociation theory. In N. J. Mackintosh and W. K.Honig (Eds.), Fundamental issues in associativelearning. Halifax: Dalhousie Univ. Press, 1969.Pp. 162-189.

Fletcher, F. G., Stoddard, L. T., and Sidman, M. Asystem for unlimited repetitive presentation ofauditory stimuli. Journal of Experimental ChildPsychology, 1971, 11, 165-169.

Geschwind, N. Disconnexion syndromes in animalsand man: part I. Brain, 1965, 88, 237-293.

Guess, D. A functional analysis of receptive languageand productive speech: acquisition of the pluralmorpheme. Journal of Applied Behavior Analysis,1969, 2, 55-64.

Guess, D. and Baer, D. M. An analysis of individualdifferences in generalization between receptive andproductive language in retarded children. Journalof Applied Behavior Analysis, 1973, 6, 311-329.

Jenkins, J. J. Mediated association: paradignms andsituations. In C. N. Cofer and B. S. Musgrave (Eds.),Verbal behavior and verbal learning. New York:McGraw-Hill, 1963. Pp. 210-245.

Kahn, D. and Birch, H. G. Development of auditory-visual integration and reading achievement. Per-ceptual and Motor Skills, 1968, 27, 459-468.

Mackay, H. A. and Sidman, M. Instructing the men-tally retarded in an institutional environment. In G.A. Jervis (Ed.), Expanding concepts in mental re-tardation. Springfield: Charles C. Thomas, 1968.Pp. 164-169.

Peters, H. N. Mediate association. Journal of Experi-mental Psychology, 1935, 18, 20-48.

Schoenfeld, W. N. and Cumming, W. W. Behaviorand perception. In S. Koch (Ed.), Psychology: thestudy of a science, Vol. 5. New York: McGraw-Hill,1963. Pp. 213-252.

Sidman, M. Behavior shaping with the mentally re-tarded. In N. R. Bernstein (Ed.), Diminished people:the problems and care of the mentally retarded.Boston: Little, Brown and Co., 1970. Pp. 263-276.

Sidman, M. Reading and auditory-visual equivalences.Joutnal of Speech and Hearing Research, 1971, 14,5-13.

Sidman, M. and Cresson, 0. Reading and crossmodaltransfer of stimulus equivalences in severe retarda-tion. American Journal of Mental Deficiency, 1973,77, 515-523.

Wepman, J. M. Dyslexia: its relationship to languageacquisition and concept formation. In J. Money(Ed.), Reading disability. Baltimore: Johns Hopkins,1962. Pp. 179-186.

Received 31 October 1973.(Final Acceptance 15 April 1974.)