00 ‘.ll. r33. ”N‘I Arm”. 1 MESS ‘I. {'0‘ In ‘1!!! avidlw "I”; m {fob Ru. 1 «Ha {1H0 . n 1.. It. a - .. .\ a 6... 14”.! K M. LI hr.“ '0’ Cara bx a\ F... aortas i .. 3c 9d a L. A... no...“ Pu! «My ens: nan..— . I‘D ’ t. o in”. Q.“ .Hha Unfit. J o.) . ”P11”- “‘1‘ 5.0 7w .1 mp”... n43 y! ‘- AM u b I I! o I. .V ~2 r. gum” . A . .2“ .. f. . ‘2 no} i ”DJIQC :5!“ 1 r. ’- t .0». fix. . ?3 if. n f“. t ‘ 1r 5% '- PVO Eb. , .- I on |U.- A .\ . d Flirt F .‘ 3. “but; I x. E to ..:l 0’ (In. ‘ a .I .V Orrfiel A v «fit. up .. . . f‘ E C .L -” [1... IR” . a £0.va III: I “to e V.4 .1. ”I!" r!‘ M. £1; 0 v. n: v ., _. E... P ,.. rt F .c Vbu .3! t rljfiwc . l .lb ML, Mum art CC I .- r. [is 1 V vh. .' "[205 A § .1 n.“ It \ I. “ % it fl. {3 I!“ “f ‘ EVU oruu “(HI- ‘3... K r. ' nu . .I‘I .if:5.2323,::.:_:_:::_+._ m m m m . 'V L I B R A R Y thhigan 3 rate University TH ESIS TRANSFER IN THE A-B, A-D PARADIGM AS A FUNCTION OF RESPONSE MEANINGFULNESS BY . “ 1 Wayne H. Decker A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF ARTS Department of Psycholoqy 1968 ABSTRACT TRANSFER IN THE A-B, A-D PARADIGM AS A FUNCTION OF RESPONSE MEANINGFULNESS by Wayne H. Decker The present study of paired-associate learning was con— cerned with comparing transfer in the A-B, A~D paradigm (old stimuli, new responses on List 2) at two levels of response meaningfulness. Transfer was measured by comparing perform- ance of the A-B, A-D subjects with that of the A-B, C-D (new stimuli, new responses) control subjects on List 2. Previous studies had found less negative transfer (inferiority of the A-B, A-D subjects to the A-B, C-D subjects) with low meaning— fulness responsasthan with high meaningfulness responses. The two stage analysis of paired associate learning maintains that there are two overlapping stages in P-A learn-' ing. These are the response learning stage and the associa- tive stage. Since the A-B, A-D paradigm involves learning new responses to old stimuli, it was expected that the A-B, A-D subjects would be equal to the A—B, C—D subjects in re- sponse learning but would be inferior in the associative . stage of learning. Thus, in the A-B, A—D paradigm, negative transfer occurs in the associative stage alone and should not be affected by response meaningfulness. The previous findings of less negative transfer with low meaningfulness responses Wayne H. Decker were not predicted by the two stage analysis. Since subjects in previous studies were givena specified number of trials on each list, it was suggested that those in the low response meaningfulness condition had not been given sufficient opportunity to demonstrate the effects of associa— tive interference as had those with high meaningfulness responses. In order that all stages of learning could be studied in all gs, the S3 in the present study learned List 2 to the criterion of one perfect trial. Two purposes of the present study were to determine whether transfer differed as a function of re- sponse meaningfulness when the trials to criterion measure was used and to determine whether elimination of the response learn— ing stage would eliminate the difference in transfer. The latter hypothesis was tested by utilizing an associative matching pro« cedure for half of the §S, whereby the responses were selected from a card rather than learned. The results indicated that the three variables, meaning- fulness, paradigm, and procedure all influenced learning, but no interactions were significant. It was concluded that trans- fer does not differ as a function of response meaningfulness when a trials to criterion measure is used. Since transfer did not interact with meaningfulness, the hypothesis that as- sociative matching would eliminate the interaction was mean- ingless. The failure to find a difference in transfer as a function of response meaningfulness was seen as support for the adequacy of the two stage analysis. CL Approved fimé’pkawi , Date leifl‘wi-U' C, L Qfi if J 7 i I ACKNOWLEDGMENTS I wish to thank Dr. Gordon Wood, my adviser, for his help and guidance in this project. The assistance and criticism of Dr. James Phillips and Dr. David Raskin are also greatly appreciated. ii TABLE OF CONTENTS INTRODUCTION....................................o.... METHOD...o........................................... RESULTS..............o............................... DISCUSSION.................................o......... SUMMARY...................................o.o...o.... REFERENCESOOOOOOOOCOOCOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO iii Page 10 14 l7 l9 LIST OF TABLES Table Page 1. Mean number of trials to each criterion on List 2 for each condition............... 11 iv LIST OF APPENDICES Appendix Page A. PAIRED-ASSOCIATE LISTSooooooooooooooooooo0.... 20 B. TRIALS TO SUCCESSIVE CRITERIA FOR EACH SUBJECTOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO00...... 23 INTRODUCTION Various studies have utilized a two stage conception in the analysis of paired-associate learning (e.g. Underwood, Runquist, & Schulz, 1959: Horowitz, 1962). The two stages assumed are a response learning stage and an association learning stage. According to Underwood et a1. (1959) the response learning stage necessarily precedes the association learning stage, as responses must be learned before they can be available for association to stimuli. It also seems that the two stages overlap, since some responses in a list prob— ably are being associated to their respective stimuli as other responses are being learned. Predictions yielded by the two stage analysis have been partially successful in transfer studies involving response meaningfulness as a variable. Jung (1963) found negative transfer for the A—B, C-B paradigm (relative to an A-B, C-D control group) with responses of high meaningfulness and posi- tive transfer in the same paradigm when the responses were low in meaningfulness. Jung suggested that positive trans- fer occurred in the response learning stage since the respon- ses were the same in both lists and negative transfer occurred in the associative stage because these old responses had to be associated with new stimuli in List 2. With low response meaningfulness the response learning was so difficult that the l 2 positive transfer during the response learning stage was greater than the negative transfer of the associative stage. With high meaningfulness responses, however, the response learning was less difficult and negative transfer was the net result. The A-B, A—Br paradigm (same stimuli and responses in both lists, but they are repaired) has also been found to follow the predictions of the two stage analysis. Merikle and Battig (1963) obtained positive transfer with low re- sponse meaningfulness and negative transfer with high response meaningfulness. The above explanation of transfer in the A-B, C-B paradigm also applies to the A-B, A—Br paradigm. Thus, both the A-B, A-Br and the A—B, C—B paradigms have been found to interact with response meaningfulness since positive transfer has been found with low response meaning- fulness and negative transfer with high response meaningful— ness. Although the two stage analysis predicts a paradigm X response meaningfulness interaction for the A~B, C—B and A—B, A—Br paradigms (relative to A-B, C—D controls), the two stage analysis does not predict a Paradigm X Meaningfulness inter— action for the A—B, A~D paradigm. Negative transfer is to be expected in the associative stage, as new responses must be associated with old stimuli. Yet, no difference as a function of paradigm (zero transfer) would be predicted in the response learning stage since both the experimental (A-B, A—D) and control (A-B, C-D) groups must learn new re- 3 sponses. Since the amount of response learning is the same for both the experimental and control groups, there is little reason to expect an interaction between response meaningful- ness (low vs. high) and paradigm (A—B, A—D vs. A-B, C-D). However, several studies (Jung, 1963: Merikle & Battig, 1963: Goulet, 1965: Houston, 1965) have found an interaction between response meaningfulness and paradigm (A-B, A-D vs. A—B, C-D). Jung and Goulet found less negative transfer with low than with high response meaningfulness. That is, the performance of the subjects in the experimental group with low meaningfulness responses was less inferior to the low meaningfulness control group than the performance of the high meaningfulness experimental group was to the high. meaningfulness control group. Neither Merikle and Battig nor Houston found significant differences among three levels of response meaningfulness, although slightly less negative transfer occurred with low meaningfulness responses than with high meaningfulness responses. One purpose of the present study was to assess further whether the Paradigm X Meaningfulness interaction in the A-B, A-D paradigm is a re- liable finding. A second purpose, which depended upon ob- taining of the interaction, was to test a two stage analysis explanation of the Paradigm X Meaningfulness interaction. However, the reliability of the Paradigm X Meaningfulness in- teraction is placed in doubt by some aspects of previous Studies. In presenting data which showed the mean number of overt intrusion errors (responses that belong in the list but are 4 given to the wrong stimuli) across trials on List 2 with each of two response meaningfulness levels, Jung (1963) suggested that the portion of training in which the overt errors in- creased to a peak corresponded to the response learning stage, while after the peak, as the number of errors per trial de— creased, the associative stage was dominant. With the A-B, A-D and A-B, C-D paradigms the mean number of intrusions in- creased throughout the ten trials given on List 2 with low meaningfulness responses, but a peak had been reached at approximately the fourth trial with high meaningfulness responses. Possibly less negative transfer occurred with low response meaningfulness because fewer trials were spent by the gs in the associative stage as compared to the §s with high meaningfulness responses. Subjects in the low meaning- fulness groups may have been primarily in the response learn- ing stage during the ten trials and performance in this stage may be expected not to differ as a function of paradigm since both the A—D and C-D groups are in the process of learning new responses. Those in the high meaningfulness groups, however, learned the responses in fewer trials and, therefore, had suf- ficient trials to manifest the negative transfer of the asso- ciative stage. Further support for this view was suggested by the fact that Jung's (1963) low response meaningfulness List 1 required a mean of 23.45 trials to reach a criterion of a perfect trial. It seems to follow that ten trialscnzList 2 was not enough to adequately demonstrate the effects of the negative transfer of the associative stage when such difficult responses had to be 5 learned. Jung's high meaningfulness groups needed only 10.50 trials to learn List 1, therefore, ten trials on List 2 seem to have been sufficient in showing negative transfer which occurred in the associative stage. Thus, the Paradigm X Meaningfulness interaction may have been due to the differences in the degree of learning achieved in ten trials as a function of response meaningfulness. Similarly, Goulet (1965) presented fifteen List 2 trials, but the high meaningfulness §s had re- quired 13.82 trials to learn List 1, while 26.45 trials were needed by the low response meaningfulness subjects. The studies which have obtained the least differences in transfer among response meaningfulness levels have given the most trials on List 2 relative to the number of trials necessary to learn List 1 to the criterion of one perfect trial (Merikle and Battig, 1963; Houston, 1965). In order to allow the occurrence of the associative stage at all meaningfulness levels, gs in the present study were taken to a criterion of one perfect trial on List 2. Ekstrand (1966) points out some of the difficulties involved in determining the point at which response learning is completed. A major difficulty results from the fact that the response learning and associative stages overlap. To eliminate the uncertainty as to the stage involved and to view more clearly the effects of the interference in the as- sociative stage, half of the SS in the present study learned List 2 using an associative matching procedure so that response learning was not necessary (Horowitz, 1962). It was expected that if, a paradigm X meaningfulness interaction occurred 6 using a regular anticipation method, the interaction would be eliminated by the associative matching procedure since response meaningfulness should have little effect upon the associative stage. That is, a response meaningfulness X paradigm X learn- ing procedure interaction was expected. METHOD ' Design. A 2 X 2 X 2 factorial design was used with a separate group of §s utilized in each of the eight conditions. The variables investigated were response meaningfulness (high vs. low with the same level of meaningfulness used on both lists in every condition), paradigm (A-B, A-D vs. A-B, C-D), and learning procedure on List 2 (regular anticipation vs. associative matching). Materials. The lists utilized were identical to those of Jung (1963). These are presented in.Appendix A. Each list contained six paired associates which had two-syllable adjectives as stimuli and trigrams as responses. Interlist and intralist similarity were minimized as much as possible for both stimuli and responses. Response similarity was minimized by using any given letter only once in a given list. Similarity of List 1 and List 2 responses was mini- mized by not permitting any letter used in both lists to occupy the same position in the two trigrams in which it appeared. Trigrams with summed letter counts of 9—14 and 46-79 were used for the low and high meaningfulness lists respectively (Underwood and Schulz, 1960, Appendix F). Procedure. The §s were given instruction regarding P—A learning and the anticipation method. They were told that they were to learn six pairs of items consisting of an 7 8 adjective and a nonsense syllable and were shown examples similar to the pairs utilized in the task. They were told that the pairs would appear on the memory drum in front of them with the adjectives appearing initially alone and then, after a short time, with the appropriate nonsense syllable. The SS were instructed that after studying thepairs the first time through the list, they should attempt to anticipate the correct nonsense syllable for each adjective presented thereafter by spelling it aloud. They were informed that the appearance of a strip of masking tape in the memory drum window indicated the end of one trial and the beginning of the next trial. About one min. after learning List 1 to the criterion of one perfect anticipation trial, §s were given the instruc— tions for List 2. Half of the §s were instructed to learn List 2 in the same manner as they had learned the first list. The remaining half of the SS, the associative matching groups, did not learn the second list in the same manner as they had learned the first. These §s were given a 3 in.x 5 in. card 'which had the correct responses for List 2 randomly printed on it. The SS held the card and could refer to the list for the correct response upon being presented each stimulus. Thus, the necessity of the response learning stage in paired asso- ciate learning was eliminated. In this way the difficulties in studying the associative stage were lessened. All gs were taken to a criterion of two perfect trials on List 2 to in— sure that all §s had at least two trials. 9 The lists were presented on a Stowe memory drum at a 3:3 sec. rate with an intertrial interval of 6 sec. The lists were randomized with three different orders being utilized in order to minimize the possibility of serial learning. Subjects. The §s were 104 undergraduate students en- rolled in introductory psycholOgy classes at Michigan State University. They were randomly assigned to the eight groups with each group consisting of 13 students. Two persons had to be discarded and replaced due to failure to learn List 1 to a criterion of one perfect trial. RESULTS List 1 Learning. The mean number of trials to criterion for the S3 with high meaningfulness responses was 8.77 for the A—D, associative matching group, 7.62 for the C—D, associa- tive matching group, 9.15 for the A-D, regular anticipation group, and 9.08 for the C-D, anticipation group. The means of the trials to criterion for the low response meaningful- ness conditions were 15.00 for the A—D, associative matching group, 17.00 for the 0-D, associative matching group, 10.77 for the AnD, anticipation group, and 12.38 for the C—D anti- cipation group. An analysis of variance indicated that response meaningfulness was the only variable affecting group differences, ‘§(lw96):= 20.99, p<;001. Neither paradigm, §