CUE UTILIZATION IN SERIAL LEARNING Thesis for the Degree of Ph. D. MICHIGAN STATE UNIVERSITY Louis GT Lippman 196267 THESIS WITHHIM/WNW ' (x 3 1293 01056 2522 This is to certify that the thesis entitled Cue Utilization In Serial Learning presented by Louis G. Lippman has been accepted towards fulfillment of the requirements for Mo— degree in. Psychology (/71? / // l/‘Lj {/ :17 -Lg‘ll: %’{_,-"i “It” Major prd’fessor l 4. I. L7 3:) I ., 0.169 L____,_,___— V_A, , ,7 f Bi 7%” LIBRARY Mien“... ,t . at: University v- -. ., xi}! Lam... MHZ... ~.T 7 i“ 1095' ABSTRACT CUE UTILIZATION IN SERIAL LEARNING by Louis G. Lippman Two experiments investigated gs' use of cues in serial learning. Experiment I was based upon a theory of the serial position effect which asserts that the intertrial interval (ITI) makes the first item of a series (list) highly dis- criminable or an anchor for the primacy effect. This led to a three—pronged hypothesis. Early in learning--up to the point when the first item is learned both as a response and as a cue for primacy--§ is highly dependent upon the ITI as an order cue. Late in learning, though § is no longer dependent upon the ITI as a primacy or order cue, he has acquired a habit or set regarding the location of the ITI within the series. At an intermediate stage of learning, the first item has supplanted the ITI as a pri- macy or order cue, and the habit regarding the locus of the ITI has not acquired much strength. From this general hypothesis the following prediction is made: If the ITI is shifted from the end to the middle of a list, a performance decrement will occur if the shift Louis G. Lippman is made either early or late in learning but will not occur if the ITI is shifted at an intermediate stage. A list of 12 nonsense syllables alternating with 12 blank spaces was presented at a 1.5 second rate (a three-second rate from syllable to syllable). The ITI consisted of a 9 second presentation of a blank space. The ITI was shifted by decreasing the duration of the blank space between items 12 and 1 from 9 to 1.5 seconds and increasing the duration of the blank between items 6 and 7 from 1.5 to 9 seconds. Four experimental groups (N a 2M each), in which the ITI was shifted after § attained at least 1, h, 7, or 10 correct anticipations on a single trial (groups E1, Eu, E7, and E10, respectively) were compared to a control group (N - 2D) in which the ITI had a constant location. The results showed no significant impairment in per- formance for groups Eh and E7, a temporary decrement for E10, and a more extensive decrement for group E1’ in sup- port of the prediction. It was suggested that interference effects were minimized in Experiment I due to the visual identity of the ITI to the inter-item space. Experiment II was based upon a former study which ‘ indicated that serial position effects are eliminated when the ends of a serial list are disguised (continuous serial learning (no ITI) with extralist items preceding the first trial). For Experiment II, it was suggested that the re- sulting serial position curve represents the relative Louis G. Lippman distinctiveness or ease in learning of the individual items of the list, provided that item effects are not counter- balanced by list rotation. These findings and contentions were tested under conditions of extreme intralist similar- ity, requiring a minimum of response learning. Two groups of 22 §s each were presented an eight cluster list presented in the same order at a H second rate for 40 trials. All clusters (four nonsense syllables) were identical except for the position of the syllables within each cluster. The §s were instructed to anticipate clusters by pronouncing the syllables as they would appear, from left to right, in the aperture. Group C learned with a long (12 second) visually distinctive ITI; group B learned a contin- uous serial list and three extralist clusters (other permu- tations of the same four nonsense syllables) preceded the first trial only. The results for group C supported previous findings that the advantage in learning the first cluster (item) is present from the onset of acquisition. The results of group B supported previous findings that primacy effects are eliminated and that §s select an idiosyncratically dis- tinctive item for a primacy cue when the ends of a temporal series are disguised. Of particular interest was the ab- sence of a recency effect in group C. This was attributed to the high intralist similarity (interference) among the responses in §s' task. CUE UTILIZATION IN SERIAL LEARNING By Louis G. Lippman A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Psychology 1966 ACKNOWLEDGMENTS The author is most grateful for the personal guidance and interest provided by his major professor, Dr. M. Ray Denny. The author is likewise indebted to Drs. Charles Hanley, Glenn I. Hatton, Donald M. Johnson, and Bertram P. Karon for not only their instruction but also their patience and kindness. The author also wishes to thank Drs. Paul Bakan and David C. Raskin for their assistance and to thank Dr. William T. Stellwagen who provided counsel and procurred subjects for Experiment I. 11 TABLE OF CONTENTS Page ACKNOWLEDGMENTS . . . . . . . . . . . . . ii LIST OF TABLES . . . . . . . . . . . . . iv LIST OF FIGURES . . . . . . . . . . . . . v LIST OF APPENDICES . . . . . . . . . . . . vi GENERAL INTRODUCTION . . . . . . . . . . . 1 Single Factor Theories. . . . . . . . . 2 Dual Factor Theories . . . . . . . . . 3 EXPERIMENT I . . . . . . . . . . . . . . 6 Method . . . . . . . . . . . . . . 10 Procedure . . . . . . . . . . . . . 12 Results. . . . . . . . . . . . . . 13 Discussion. . . . . . . . . . . . . 18 Summary. . . . . . . . . . . . . . 22 EXPERIMENT II. . . . . . . . . . . . . . 2“ Method . . . . . . . . . . . . . . 27 Procedure . . . . . . . . . . . . . 29 Results. . . . . . . . . . . . . . 29 Discussion. . . . . . . . . . . . . “1 Summary. . . . . . . . . . . . . . 45 REFERENCES. . . . . . . . . . . . . . . H7 APPENDIX . . . . . . . . . . . . . . . “9 iii LIST OF TABLES Table Page 1. Summary of the analysis of variance on trials to criterion . . . . . . . . . . . 15 2. Trials to criterion. . . . . . . . . 15 3. Summary of four t-tests performed on the number of correct anticipations following partial criterion per trial . . . . . . 15 A. Summary of trtests performed on the total number of correct anticipations on: (a) The first two trials following partial criterion; (b) The third and fourth trials following partial criterion; (c) The fifth and sixth trials following partial criterion. . . . 16 5. Summary of the analysis of variance on the number of items recalled . . . . . . . l7 6. Number of items recalled . . . . . . . 17 7. Summary of the analysis of variance of the serial position effects in groups C and E . HO iv Figure 1. LIST OF FIGURES Mean number of clusters correctly anticipated per trial by groups C and E . . . . Mean of the total number of clusters correctly anticipated by groups C and E, in the eight serial positions. . . Mean number of correctly anticipated clusters, per block, in groups C and E for clusters land2,3andlI,5and6,and7and8. The maximum score is 4 clusters . . . . Mean number of clusters correctly anticipated, per trial, for the first cluster in groups C and E. . . . . . . . . . Mean number of syllables correctly anticipated, per block, for clusters 1, 5, and 8 in grups C and E . . . . . Mean percent of the total number of clusters correctly anticipated for each of the eight serial positions. The curves are plotted from the cluster actually presented first in groups C and E from Ss' "objective first" clusters in group E. . . . . . Page 3“ 35 36 37 38 39 Appendix A. LIST OF APPENDICES Instructions for Acquisition in Experiment I O O O O O O O O O O I O 0 Instructions for Recall in Experiment I Instructions and Worksheet for the Interval Between Acquisition and Recall List of Nonsense Syllables Used in Experi- ment I . . . . . . . . . . . . Supplementary Acquisition Data vi Page 50 51 52 53 54 GENERAL INTRODUCTION Since Hull's (1940) theoretical analysis of rote learning, several investigators have attempted to explain serial learning and the bow-shaped serial position effect using a limited number of explanatory concepts.l Some of the recent theories focus on an explanation of the serial position effect without explaining serial learning. Such approaches have been excluded from the present analysis. The theories with which we are concerned ask the question, "what is learned in serial learning." These theories are most generally applicable to serial learning of nonsense syllables by the anticipation methOd, and response learning is usually assumed to occur both before and during the association or ordering phase of acquisition. Thus, these theories represent attempts to describe what is learned in the association phase (response learning is assumed to be a necessary prerequisite to association). 1According to Hull, the bow-shaped function is the result of the number of forward and backward associations which are present at different points in the list, with the center of the list, which is most difficult, having the most associations. Single Factor Theories The simplest single factor theory is one that Young (1961) has named the "specificity hypothesis." According to this view, serial learning represents the acquisition of a chain or sequence of 8—H associations in which the stimulus for a particular item (response) in the list is its immediately preceding item. Thus, every item in the list (except the first and last item) has a dual function: As a response to the preceding item and as a stimulus for the following item.’ As a second approach, the specificity hypothesis has been elaborated to produce a "compound stimulus" or "cluster" hypothesis (Horowitz and Izawa, 1963). This hypothesis differs from the specificity hypothesis (or sequential as- sociation hypothesis) only in that two or more preceding items are assumed to serve as the functional stimulus for each successive item in the series. The more remote items in the "cluster" are assumed to have weaker stimulus function. A third hypothesis of serial learning is the "position" hypothesis (Jensen and Rohwer, 1965). In this case the functional stimulus is considered to be the ordinal position occupied by each item in the list. In other words, an as- sociation between each ordinal position and its correspond- ing list item is assumed to develop. The most recent suggestions about serial learning view the serial position effect as analogous to end- anchoring in Judgment (Simon, 1962). According to an "anchoring" hypothesis, the list is learned in a system- atic fashion, around the item occupying the first position on the list (i.e., forward and backward from this item). The anchoring hypothesis, therefore, is more a description of Ss' strategy in learning (rather than a description of specific S-R connections). Dual Factor Theories The two major dual-process theories are essentially combinations of the specificity or cluster hypothesis (sequential associations with either single or compound stimuli) and the position hypothesis. One version of a dual factor theory (Young, Patterson, and Benson, 1963) suggests that the ends of a serial list are learned by sequential association with ordinal position. The second version (Ebenholtz, 1963b) states the Opposite; that the ends of a serial list are learned by association with ordinal position and the middle items are learned by sequential associations (specificity or cluster). These two versions do not assert that the two types of learning (sequential and positional) are mutually exclusive for a particular item in the series; but that both sequential and positional associations (or cues) occur throughout the list. Thus these two hypotheses differ only in their predictions of the relative strengths of sequential and positional cues in different portions of the list. More recently, Lippman and Denny (196“) have offered an hypothesis which seems most similar to Ebenholtz's dual factor hypothesis. Lippman and Denny have suggested that the order cues (or positional cues) at the beginning and end of the list are of two different types (primacy and recency cues, respectively). The primacy cue for the first three or four items of the list might even be viewed as exploiting memory span for these few items. The recency cue is viewed as a "temporal" cue for the end of the list: S learns where, in time, the end of the list comes. It is important to note that Lippman and Denny have asserted that neither of these "positional" cues is item specific; that these cues are not precise "Position—Item" associations. Instead these cues provide non-specific information as to approximately where a particular item occurs in the series. The present two experiments are based upon the Lippman- Denny hypothesis and elaborations thereof. In Experiment I, the intertrial interval (ITI), as is customary, represented the starting point and the end point of each trial. In this experiment the ITI was shifted from the end to the middle of the list at differential stages of acquisition (for different groups of §S). According to the specificity hypothesis, a shift in the ITI should produce no effect upon performance since the sequence of items remains unchanged (thus the chain of S-R associations from item to item should not be influenced). Considering the position hypothesis, a shift in the locus of the ITI should produce a significant decrement in performance because the ordinal positions of each of the items has been changed (analogous to an A-B, A-Br transfer paradigm). These two predictions may be viewed as "pitted against" a prediction based upon an elaboration of the Lippman-Denny hypothesis (which asserts that associations with ordinal positions are non-specific). In Experiment II, one group of Ss learned a serial list with an ITI which clearly demarcated the ends of the list, and another group learned with the ends of the list disguised (no ITI and extralist items preceded the first item on the first trial only). Lippman and Denny's hy- pothesis allows for prediction of the performance and serial position effects for both of these groups. It appears difficult, however, to derive a definite prediction from any of the other hypotheses unless some modifications or elaborations are imposed. EXPERIMENT I Recent research on serial learning has lead some in- vestigators to designate the first item of a temporal series as either the anchor for the serial position effect (Jensen, 1962; Feigenbaum and Simon, 1961; Simon, 1962) or as the cue for primacy and recency effects (Lippman and Denny, 1964).1 In turn, Lippman and Denny have demonstrated that the first item can serve as such a cue or anchor only when it can be perceived as first in a temporal series (con- ventionally established by the intertrial interval (ITI)). Once the first item has been established as a cue for primacy, it may be asked whether the ITI must hold a constant position in the series in order to maintain its cue value. Two differing opinions may be considered: (1) It may be asserted that the most important attribute of the ITI is its constancy. Therefore any change in the ITI's location in a temporal series would be expected to impede learning. Such a conclusion is supported by the findings of Lippman and Denny (196“), Ebenholtz (1963a), lThe ITI makes the first item of a temporal series distinctive by virtue of its visual and temporal dis- tinctiveness and its constant location. The first item may be regarded as either an anchor about which the re— mainder of the list is learned; or, in conjunction with the ITI, as the source of both an order cue (primacy effect) and a temporal cue (recency effect). 6 Keppel (1964), Winnick and Dornbush (1963) and Bowman and Thurlow (1963) with the varied-position method (continual shift of ITI position). (2) On the other hand, the ITI not only makes the first item a highly discriminable cue but also facilitates the acquisition of the first item as a response. Once the first item becomes a well-learned re- sponse, the ITI could then become a redundant cue; i.e., the first item could supplant the ITI as a cue for primacy. Such an interpretation seems consistent with certain find- ings and speculations about information processing and en- coding (Miller, 1956). If this were the case, then changing the position of the ITI after the first item has been learned would produce no detrimental effect in learning the entire list. In fact, if the ITI were still able to serve as a facilitating cue, then placing it in a different posi- tion in the list might facilitate the learning of the whole list (the reasoning behind this expectation is elaborated below). Both of these alternative explanations and predictions must be considered conjointly with the degree to which S uses the ITI as a cue. Direct empirical research has shown that S learns a temporal series from both ends, working toward the middle items of the series (Jensen, 1962). Furthermore, Lippman and Denny (1964) have shown that pri- macy effects occur almost immediately while several repe- titions of the list are required before recency effects begin to accrue (before performance on end items exceeds middle items). Early in learning, S is highly dependent upon the ITI for a primacy cue. Thus if the position of the ITI were shifted before the first item had been learned (learned Sggg as a response SEQ as a primacy cue), then learning of the entire list would be impaired. Late in learning, S has already used the ITI as a cue for primacy and for the establishment of recency effects (recency is learning approximately where the end of the list comes). Changing the location of the ITI late in acquisition, there- fore, should probably produce only a slight impairment in performance as the result of the small change in stimulus conditions. At an intermediate stage of acquisition, however: 1. The first item has been learned as a response serving as a cue for primacy and the ITI has probably had sufficient opportunity to bring about the identification of the location of the end of the list (recency effect); 2. Thus the ITI has become a redundant cue upon which S is no longer highly dependent; 3. A habit (set) with regard to the locus of the ITI has gg£_acquired much strength (less generalization decrement from the switch in ITI placement than late in learning). If these assumptions hold true, then a change in location of the ITI, during an intermediate stage of acqui- sition, would produce at the most a small, temporary per- formance decrement and might even facilitate the learning of the entire list. Facilitation would most likely occur if the ITI were moved to the middle of the list when S was concentrating upon learning the middle of the series: The ITI could then provide S with an additional cue for learn- ing this portion of the list. Facilitation will probably occur only if all three conditions above can be satisfied (and only if the ITI has n93 been particularly distinctive, i.e., ITI highly similar to the interstimulus interval and no asterisks present; interference effects would most likely occur, as a result of switching the ITI to the middle of the list, when the association between the ITI (in its original location) and the first item of the list is strong). The purpose of the present study is to test for differ— ential effects in performance as the result of changing the location of the ITI from the end to the middle of the list after different levels of acquisition have been reached. After learning to criterion, recall, starting from the ITI's original location, will be tested. Since responses are con- stant throughout acquisition and recall (responses are identi— cal irrespective of the ITI location) and extraneous factors are constant, any interference effects produced by returning the ITI to its original location can not be explained by un- learning. Any interference effects would be due solely to an interfering set regarding the starting point (the first item following the ITI). The strength of the interfering set should be a direct function-of the relative number of trials the ITI is located in the middle of the list, after lO relocation. Differential performance decrement would there- fore be attributed to the relative difficulty in reinstating the set for the original ITI location. Method Subjects The Ss were 120 students enrolled in an introductory psychology course at Michigan State University who had no prior experience with either nonsense syllables or serial learning. Participation was on a volunteer basis. Each S was systematically assigned to one of five groups, ac- cording to his order of appearance at the laboratory. No five successively appearing Ss were assigned to the same group. Instructions The instructions (Appendix A) were the same as used formerly (Lippman and Denny, 1964), with minor modifications for added clarity. The Ss were handed copies of the in— structions which they followed while S read the instructions aloud. The instructions preceding recall (Appendix B) were spoken aloud by S; the instructions for S's task between acquisition and recall were read by S silently and appeared at the top of his worksheet (Appendix C). Apparatus The apparatus consisted of an MTA lOO Scholar (teaching machine), adapters for endless loop material, an MTA 11 external timer, and a Hunter decade interval timer. The apparatus was wired so that the teaching machine would close an independent delay circuit at a particular position in the series (selected by S); when the delay had timed out, the teaching machine and its external timer resumed function- ing. The aperture of the teaching machine was reduced to 7/16 by 1 inch. Materials Nonsense syllables (Appendix D) were identical to those used by Lippman and Denny, and were presented in the same order. The syllables were typed in upper-case elite and were presented alternately with blank spaces. Each syllable was presented for 1.5 seconds, each blank space for 1.5 seconds, and an ITI consisted of a 9 second presentation of a blank space. The syllables were always presented in the same order, but twelve different starting positions were em- ployed with two Ss in each group beginning their learning from each of the twelve starting positions. Other materials consisted of ten simple sentences which were selected from experimental materials used by Bulgarella (1965). These sentences (Appendix C) were used only to fill two minutes of time between acquisition and recall. 12 Procedure There were four experimental groups and one control group. The experimental groups continued in acquisition until attaining at least 1, 4, 7, or 10 correct antici— pations on a single trial (according to each S's indivi— dual rate of acquisition). Each of these groups was labeled El’ E4’ E7, or ElO respectively, with regard to the partial criterion. After an S reached partial criterion (according to the assigned experimental group) the location of the ITI was shifted from the end of the list to the middle of the list on the trial immediately following attainment of the partial criterion. Specifically, the ITI shifted from the blank space between positions 12 and l to the blank between positions 6 and 7. Following the shift in location of the ITI, all experi- mental groups continued learning until attaining one perfect recitation of the list. Since the ITI was moved to the middle of the list, the criterion for learning was defined lacoordingly--i.e., in reference to the changed ITI location. This definition of criterion prevented the stopping of S when he was half-way through a l2-item trial; it also en- sured one ITI for each list presentation for all groups. Finally, the control group (group C) continued learning to a criterion of one perfect recitation without a shift in ITI location. 13 Immediately following attainment of criterion, Ss were handed worksheets (Appendix C) from which they counted vowels in simple sentences. All Ss read the instructions on their worksheets and counted vowels for a total time of two minutes. (Since the instructions for recall required an average of 20 seconds to present, the total time between acquisition and recall was approximately 140 seconds). While S counted vowels, S reset the teaching machine to the original starting point in the series (the machine was set to run without intermittent stopping until reaching a parti- cular item in the series; thus S was not informed as to where the list would start). Finally, each S was given one trial of serial anticipation starting from the original ITI location. Since S was not informed of the starting position for recall, his anticipation to the first item was not scored; S was told to look at the first syllable in order to know the starting point. Results An analysis of variance on trials to criterion was performed to test for the predicted differential effects upon performance (Table l). The S was not significant and multiple comparisons by the Newman-Keuls method revealed no differences among the means (p > .05; Table 2). (Addi- tional acquisition data are presented in Appendix E.) Since the trials measure represents the effects of the combination of many factors, a more sensitive measure 14 was derived in order to test the present hypotheses directly. The mean number of correct anticipations following partial criterion per trial was calculated for each S in the E-groups and four corresponding measures were derived for Ss in group C. FourxS-tests between each E—group and its appropriate C measure (Table 3) revealed overall impairment in performance in groups E and E10 and no difference in groups E4 and E7. 1 The effect of shifting the location of the ITI upon subse- quent performance was directly examined in the first two trials following the shift, the third and fourth trials, and in the fifth and sixth trials following partial criterion. (For these analyses, trials were defined from the ITI's original location for all groups.) For each of these three analyses, the total number of correct anticipations was ob— tained; for each measure in each E—group, an appropriate corresponding measure was derived from group C. As above, if a score could not be obtained for an S (occasionally in group E and particularly in group C in which S is likely to 10 reach criterion soon after attaining partial criterion), the S was discarded rather than make the untenable assumption of continued perfect recitation after one perfect trial. The results of these analyses are presented in Table 4. It can be seen that the shift in ITI location did not significantly affect performance in groups E4 and E7. For group E1 the performance on the first two trials just missed significance (p < .10, two—tailed) but group El subsequently showed im- paired performance on the second and third pairs of trials H U": TABLE l.-—Summary of the analysis of variance on trials to criterion. Source d.f. Mean Square F p Between groups 4 131.11 1.44 n.s. Within groups 115 91.15 Total 119 TABLE 2.--Trials to criterion. C El E4 E7 E10 M 30.25 34.38 34.29 31.67 36.08 S.D. 8.70 10.47 11.62 8.44 8.01 TABLE 3.--Summary of four S-tests performed on the number of correct anticipations following partial criterion per trial. C El C4 E4 U7 E7 Clo El0 M 6.27 5.49 7.16 7.20 8.55 8.37 9.89 9.35 S.D 0.87 1.22 0 83 0.74 0.73 0 84 0.81 0 {2 N 24 24 24 24 24 24 21 24 p .05 .90 .50 .05 10 is from Ss reaching criterion. Reduced N in C All tests are two-tailed. NOTE: 16 TABLE 4.--Summary of Estests performed on the total number of correct anticipations on: (a) The first two trials following partial criterion; (b) The third and fourth trials following partial criterion; (0) The fifth and sixth trials following partial criterion. M 3.33 2.25 7.63 7.71 12.96 12.33 18.76 15.58 S.D. 1.99 1.73 2.16 2.29 3.24 2.90 2.51 3.01 M 5.67 3.54 8.83 8.96 14.33 14.04 18.58 17.17 S.D. 2.84 3.36 2.64 3.29 3.90 2.87 2.85 3.31 p .05 .90 .80 .20 M 7.00 4.92 10.29 9.79 15.52 15.33 18i33 18.29 S.D. 3.50 3.12 2.85 4.12 2.78 4.04 2.31 1.96 N 24 24 24 24 23 24 16 17 p .05 .70* .90* .70 *Heterogeneity of variance. NOTE: All tests are two—tailed. 17 TABLE 5.--Summary of the analysis of variance on the number of items recalled. Source d.f. Mean Square F p Between groups 4 2.56 1.19 n.s. Within groups 115 2.16 Total 119 TABLE 6.--Number of items recalled. M 9.17 8.58 8.71 9.38 8.88 S.D. 1.76 1.74 1.49 1.09 1.12 (p < .05, two-tailed). For group E10, performance was significantly impaired on the first two trials following partial criterion (p < .01, two-tailed), but did not differ from group C thereafter (p < .20; p < .70, respectively). Finally, an analysis of variance was performed on the number of items correctly recalled. The S was not significant and none of the means differed significantly (p 2 .05; Tables 5 and 6). Discussion It is of considerable interest to note that the groups did not differ on trials to criterion (although the means were in the predicted direction) and that two of the E- groups never differed from the control condition when more sensitive measures of performance were employed. On the basis of previous research using the varied-position method, a change in the location of the ITI may be expected to im- pair performance. The present procedure, however, employed an 1T1 which was visually identical to the interstimulus interval (blank space), differing only temporally (9 seconds vs. 1.5 seconds). Such a procedure should minimizg the interference effects of a shift in the location of the ITI. On this basis, the effects found for all E-groups are under— estimated; greater performance decrement might have occurred if the ITI had differed BREE temporally and visually from the interstimulus interval. This would be consistent with l9 Lippman and Denny's assertion that the cue value of the ITI is highly dependent upon its perceptual distinctiveness. When the more sensitive measure of mean number of correct anticipations following partial criterion per trial was employed, the predicted effects were found——i.e., no effect in groups E4 and E7 and a decrement in the performance of groups El and E10. Additional testing of performance on the first, second, and third pairs of trials following partial criterion further substantiated the results based on the mean number of correct anticipations following partial criterion. And these more specific tests pin—point where the effects occurred. It is of prime importance to note that during the first two trials following partial criterion, all Ss in the E-groups were exposed to novel or changed stimulus conditions. The fact that groups E4 and E7 showed no impair- ment in performance on these trials supports the contention that, at an intermediate stage of learning, the habit or set regarding the locus of the ITI has not acquired much strength. And the fact that these two groups failed to show a decrement in performance on the next two pairs of trials offers support for the contention that Ss, at these stages of learning, are no longer very dependent upon the ITI as a cue for primacy or recency. Group Elo showed a significant drop in performance on the first two trials following partial criterion (when ex- posed to changed stimulus conditions), but did not show a decrement on the four trials thereafter. This finding of a tempgrary performance decrement supports the prediction that when Ss are near criterion performance, they are not highly dependent upon the 1T1 as a cue, but instead have acquired a habit or set regarding the location of the ITI. Finally, group E showed a marginally significant drOp in 1 performance on the first two trials following partial criterion (p < .10, two—tailed). On both pairs of subse- quent trials, however, group E showed a significant decre— l ment in performance. This finding of an gxgggggg perfor- mance decrement is in keeping with the prediction that early in learning Ss have not acquired a strong habit or set con— cerning the location of the ITI, but instead Ss are highly dependent upon the ITI for a cue. When this cue is changed, a long—term decrement in performance is expected since the first item has not yet been learned both as a response and as a cue. It is of interest to compare the present results to predictions which may be derived from the specificity and position hypotheses of serial learning. The specificity hypothesis would predict that the shift in 1T1 location would produce no effect upon performance since the sequence of items remained unchanged (thus the chain of S-R associ- ations would not be disrupted). The results of the analysis on trial to criterion (no difference) is in keeping with this prediction. However, the specificity hypothesis fails to predict not only the fact that some of the groups showed a performance decrement (when more sensitive measures were 21 used) but it also fails to predict which groups showed decrement and the extent of the decrement (temporary or more extended). The position hypothesis would predict that the shift in locus of the ITI would produce a significant performance decrement because the ordinal positions of each of the items have changed (negative transfer from response competition). The results of the analysis of trials to criterion fail to support this prediction. And, as with the specificity hy- pothesis, the position hypothesis fails to predict absence of decrement in some groups, which groups would show a drop in performance, and the extent of the decrement. The recall data appear to be inconclusive. Although it is of interest to note that no decrement occurred in re- call (as would be expected with response identity through- out learning and recall), the single trial of serial antici— pation does not provide sufficient information or sensi- tivity to warrant elaborate explanation. And the fact that the ITI was perceptually constant throughout acquisition and recall (and perceptually identical to the interstimulus interval) undoubtedly minimized the strength of S's set re- garding the ITI's position in the series for recall. It may have been informative to have run a condition like group C, but in which the recall trial started in the middle of the list. Had this group shown a decrement in recall (in comparison to an apprOpriate control group), it may be 22 suggested that Ss in the present E-groups had learned to make changes in their set regarding the starting position. Summary One theory of the serial position effect asserts that the intertrial interval (ITI) makes the first item of a series (list) highly discriminable or an anchor for the primacy effect. In the present experiment, there was a three-pronged hypothesis. Early in learning--up to the point when the first item is learned both as a response and as a cue for primacy—-S is highly dependent upon the ITI as an order cue. Late in learning, though S is no longer de- pendent upon the ITI as a primacy or order cue, he has ac- quired a habit or set regarding the location of the ITI within the series. At an intermediate stage of learning, the first item has supplanted the ITI as a primacy or order cue, and the habit regarding the locus of the ITI has not acquired much strength. From this general hypothesis the following prediction is made: If the ITI is shifted from the end to the middle of a list, a performance decrement will occur if the shift is made either early or late in learning but will not oc— cur if the ITI is shifted at an intermediate stage. A list of 12 nonsense syllables alternating with 12 blank spaces was presented at a 1.5 second rate (a three second rate from syllable to syllable). The ITI consisted of a 9 second presentation of a blank space. The 1T1 was 23 shifted by decreasing the duration of the blank space be- tween items 12 and 1 from 9 to 1.5 seconds and increasing the duration of the blank between items 6 and 7 from 1.5 to 9 seconds. Four experimental groups (N = 24 each), in which the ITI was shifted after S attained at least 1, 4, 7, or 10 correct anticipations on a single trial (groups El’ E4’ respectively) were compared to a control group 10’ (N = 24) in which the ITI had a constant location. E7, and E The results showed no significant impairment in per- formance for groups E4 and E7, a temporary decrement for E and a more extensive decrement for group E1, in support 10’ of the prediction. It was suggested that interference effects were mini— mized in the present experiment due to the visual identity of the ITI to the inter—item space. EXPERIMENT II In a recent study, Lippman and Denny (1964) found that serial position effects are eliminated when the ends of a serial list are disguised, that is, when there was no intertrial interval (ITI) 222 when extralist items preceded the first item on the first trial only. When the data for each S of such a group were pooled and plotted from the item which each S had correctly anticipated most frequently (S's "objective first" item), then a curve re- sembling a serial position curve, though with minimal re- cency effect, was generated. In this case, of course, the "objective first" item and its respective list position differed from S to S. In other words, when S is unable to perceive an item as first (extralist items on the first trial and no ITI), he apparently EEISEEE a discriminable idio- syncratically meaningful item for establishing his own serial position curve. The nonsense syllables used by Lippman and Denny con- stituted a typical, homogeneous list of highly differing CVC trigrams which followed the typical rules for list con- struction. It is important to note that Lippman and Denny, for the purpose of their design, did 22: counterbalance for item difficulty (or differences) by rotating syllable 24 25 position across Ss. Item difficulty was purposely con- founded with list (serial) position. This confounding raises two questions about the results of their study: (1) The absence of a bow-shaped serial position curve in the list with the disguised ends may have been due, in part, to an adventitious distribution of item difficulty which could have effectively camouflaged a small bow-shaped effect. (2) The peaks and troughs of the fairly flat, but irregular, serial position curve in the list with the disguised ends may very well represent on one hand the easiest (most dis— tinctive) and on the other hand the most difficult (least distinctive) items of the list (given a context of no pri— macy and no recency effects). Although the first possibility can not be disregarded, it is informative to note that the peaks and troughs, excluding the primacy and recency effects in the control group, consistently occur at the same position (item) in all four groups (three experimental and one control group of the Lippman and Denny study). This observation, albeit inconclusive, is more in keeping with the second possibility and leads to the following hypotheses: (1) If disguising the ends of a continuous serial list eliminates primacy and recency effects, and if the syllable position is £93 rotated across Ss, then the resulting serial position curve represents the relative difficulty or distinctiveness of each of the items in the list. Thus if differences among the serial items are decreased (by an increase in intralist similarity), then the serial position curve 26 representing the relative distinctiveness of each of the items of the list should become more nearly flat and smooth. (2) If the ends of the same, unrotated serial list are not disguised and a highly distinctive ITI is present on all trials, then the resulting serial position curve should show primacy and recency effects and be smooth throughout the list. The purpose of the present study is to compare a group in which the ends of a serial list are disguised to a group in which the ends of a list are clearly demarcated on all trials by a visually and temporally distinctive III. Both groups are run under conditions of extremely high intra- list similarity. In the present study, all serial items are clusters composed of the same four nonsense syllables (identi— cal elements) which differ only in the order in which the four elements appear. Clusters (sequences of syllables) are used in preference to redundant sequences of letters; thus problems of meaningfulness, which Ss might discover in (or impart to) certain letter sequences, are reduced. The clusters are not rotated across Ss so that the above hy- potheses can be assessed. It is also of interest to test whether each S will select a particular cluster for estab- lishing his own serial position curve when the ends of the list are disguised, even with high similarity among items. Finally, it is important to note that since all elements of all clusters (items) are identical throughout the list, S's task requires a minimum of response learning. 27 Consequently, the present conditions permit focusing upon S's acquisition of an gaggg (sequence). The major diffi- culty in S's task is produced by the high similarity among the orders he must differentiate; that is, difficulty does not center upon response learning (or response availability). Method Subjects The Ss were 44 students enrolled in an introductory psychology course at Michigan State University; participation was on a voluntary basis. Instructions Instructions were a modification of Lippman and Denny's instructions. The Ss were told that four three- letter nonsense syllables would appear in the window of the apparatus; then the same set of four syllables would appear again but in a different order, again in a different order, and so on. The Ss were instructed to anticipate all four syllables by pronouncing each syllable, as it would appear in the window, from left to right. If unsure of any sylla- ble (response) in a set (cluster), S was told he could omit the syllable and instead say "blank" (to inform S of which syllable in a cluster S was anticipating). Guessing was encouraged but not required. The S3 were given a copy of the instructions which they followed while S read the in- structions aloud. Comments clarifying S's understanding of his task were made when required. 28 Apparatus Apparatus for this experiment consisted of an MTA 100 Scholar (teaching machine) and an MTA external timer. End— less loop adapters were used for group C; fanfolded paper (which provides no consistently repeating cues--e.g., flaws, folds, or joints in the paper) was used for group E. The window or aperture of the teaching machine was reduced to 11/16- by 2 1/4-inches. Materials The list for this experiment consisted of eight clusters of the same four nonsense syllables. For example: The first cluster was MEP KOJ XIG ZUR; the second cluster was ZUR XIG MEP KOJ; the third cluster was XIG ZUR KOJ MEP; and so on. The four syllables comprising a cluster were presented horizontally in the window, each typed in upper— case Elite and separated by a space of one letter width. Syllables were grouped into clusters according to the follow- ing restrictions: (1) Throughout the list a syllable ap- peared no more than twice in any position (within a cluster); (2) the sequence of syllables within a cluster did not correspond to a sequence of syllables between clusters; (3) the sequence of two syllables within a cluster did not repeat in the same positions in any other cluster; (4) the sequence of all four syllables within a cluster did not occur in reversed order in any other cluster. Three addi— tional permutations of the same four syllables, conforming 29 to the above restrictions, were constructed for presen- tation prior to trial one in group E. The clusters were presented in the same order on each trial. Procedure All clusters were presented at a four-second rate. For group C, the ITI was 12 seconds in duration, and con— sisted of two 4-second blank spaces and one 4-second pre- sentation of three asterisks (centered in the aperture). For group E, learning was continuous (no ITI) and three extralist clusters preceded the first cluster of the list on the first trial only. All Ss were given 40 trials, with trial one being Ss' first exposure to the material. For all Ss the task started when S removed a slip of paper (shield) from the aperture at the same time as the first cluster (which, in group E, would be the first of the three extralist clusters) appeared. Results The main results are presented graphically as the mean number of clusters correctly anticipated (all four syllables correctly anticipated as to respective position within each cluster). This is plotted in blocks of four trials in Figure l and over the eight serial positions in Figure 2. An analysis of variance was performed upon the data repre— sented in the serial position curves of Figure 2 and is summarized in Table 7. The significant groups' effect 30 indicates that group C correctly anticipated more clusters than group E, as is reflected in Figure 10 With regard to Figure 1, it is of interest to note that if all Ss antici- pated each cluster with random permutations of all four syllables on every trial, a mean of 0.33 clusters would be correctly anticipated per trial by chance, whereas perfect performance would be 8.00. Group C did not consistently exceed this level until trial 16; group E did not consis- tently exceed this level until trial 33. Learning was ex- tremely slow. The significant position effect (Table 7) indicates that the eight clusters were not learned equally well over- all (groups C and E combined) and the position by groups interaction indicates that the serial position curves for groups C and E differed significantly in shape. Multiple comparisons by the Newman-Keuls method indicates that cluster one in group C yielded significantly more correct anticipations than clusters 2 and 3 in group C (p < .01) and these three serial clusters yielded significantly more correct anticipations than the remaining clusters (p < .01). All other comparisons (involving clusters 4-8 in group C and 1-8 in group E) were not significant. In other words, group B does not yield the typical bow-shaped serial position curve and approximates a horizontal line, as can be seen in Figure 2. The same sort of curve plotted in terms of number of correctly positioned syllables in a 31 cluster instead of complete clusters yields even a smoother and more nearly flat function (no figure, however, is in— cluded). The acquisition of the first pair of clusters, the second pair, the third pair, and the fourth pair of clusters for both groups E and C is presented in separate curves in Figure 3. It can be seen that group C learned the first two clusters consistently sooner and at a higher rate than did group E (as can also be seen in Figure 5). In Figure 4, the data for the first cluster only are plotted for the first four trials. This shows that the C group had an immediate advantage over group E in learning the first cluster. Figure 3 indicates that the primacy effect was not limited to the first cluster of the list, but spread to clusters 3 and 4 as well. It is of particular interest to note that the acquisition of clusters 3 and 4 does not begin to differ appreciably from the remainder of the list (clusters 5—8) until after the fifth trial block, where the rate of acquisition of these clusters rather suddenly increases. Since clusters 5—8 in both groups are not significantly different (Table 7 and multiple comparisons), it appears that the primacy effect has a definite limit at cluster 3 or 4. Although not statistically reliable, clusters 7 and 8 tended to be correctly anticipated more frequently than clusters 5 and 6 in SEES group E and C. Since this tendency is present in Group E, nothing even resembling a recency effect is indicated. The number of syllables anticipated 32 in the correct position in a cluster, rather than the number of whole clusters correctly anticipated, are plotted over trial blocks for clusters 1, 5, and 8 (representing the beginning, middle, and end of the list, respectively) for groups C and E (Figure 5). In group E, with this mea- sure, the syllables, as such, appear equally difficult for clusters 1, 5, and 8. Group C showed a consistent advantage in learning the first cluster, as has been noted in Figures 3 and 4. Since the acquisition of cluster 8 appears to be superior to cluster 5 in group C, S-tests were performed to compare E with C on the learning of cluster 5 and on the learning of cluster 8. Also, S—tests were run to compare the learning of cluster 5 with cluster 8 for both groups C and E. The £3 for all four of these comparisons were nonsignificant, in keeping with the comparisons performed upon the data in Figure 2 (the number of entire clusters correctly anticipated). Consequently, there is no evidence for a significant recency effect in group C. Since Lippman and Denny (1964) found that each S generated his own primary effect even when the ends of a serial list were disguised, serial position curves were plotted from the cluster actually first in groups C and E and from Ss'"objective first" cluster in group E (Figure 6). This made it possible to explore a primacy effect for each S in group E. The "objective first" cluster was defined as the cluster correctly anticipated most frequently by an "1 S. In case of ties, the tied cluster correctly anticipated 33 earliest in learning was employed. Since learning in groups C and E was not equal, the mean percent correct anticipations (rather than number of correct anticipations) was used so that serial position effects in group E could be examined with greater comparability. Thus, Figure 6 shows how Ss distributed their correct anticipations, p33- portionally, among clusters. The curve for group C repeats the information presented in Figure 2, showing a strong primacy effect which extends approximately to the fourth cluster, plus the absence of a recency effect. Plotting from Ss' "objective first" cluster in group E also produces a strong primacy effect and no recency, closely approxi- mating, in relative magnitude, the curve for group C. Plotting group E's data from the cluster which was actually first in the list, however, discloses a recency effect which was not apparent in any of the analyses above. Since it may be suggested that a disproportionate number of Ss "selected" items 6, 7, or 8 for their "objective first" cluster, a chi-square test was performed. Of the 18 Ss in group E who correctly anticipated at least one cluster, 11 of the "objective first" clusters occurred among clusters 6—8. Comparing this proportion to the proportion expected upon the basis of chance (3/8 of the 18 Ss, assuming random selection) yielded a significant chi-square (X2 = 4.29, d.f. = l, p < .05). Consequently, the "objective first" clusters were not randomly or equally selected from the Mean Number Correct Anticipations per Trial 34 —---- Group E .6- ——-——-Group C . O I 1 l I r r I r i I 1 2 3 4 5 6 7 8 9 10 Trial Blocks [4 trials per block] Figure 1.-—Mean number of clusters correctly antiCipated per trial by groups C and E. Mean Number Correct Anticipations 35 ---- Group E Group C Serial Position Figure 2.——Mean of the total number of clusters correctly anticipated by groups C and E, in the eight serial positions 36 .mnmpwsao : ma whoom Essaxme one .m com 5 wow .0 ocm m .3 now m .m cam fl mpmpmSHo pom m now 0 masonw 2H .xooan poo .mpmpmzao ompmafiofipcm manomppoo mo popes: cmmZIu.m opsmHm mxooan poo mamfipp nu mxooam awake oammsmmzmma Smmsmmamma _ m a a 2398 |-l w a m 2338 |-i a w m 2338 In! N a H 2338 III Hootg Jed saaqsnIQ qoaaaoo JeqmnN ueew t Anticipations I" v Corre Mean Number 37 0.8 -— 0.6 ‘- O.4 ‘fi O°2 —* 0.0 ‘ Trial Figure 4.——Mean number of clusters correctly anticipated, per trial, for the first cluster in groups C and E. 38 .m wow 0 mozopm CH m ocm .m .H mumpmSHo pom .xooao poo .pmpmaaofipcm mapoopnoo moanmaamm mo Lopes: cwozll.m mpswfim flxOOHp pom mawfipp 3g mxooam Hanna Smmammzmmfi oammammzmma . _ _ _ _ F _ _ _ me _ _ I CDNKOLOQ'MNHO I O\ m pmpmSHo IIIII l l H O H H m mopmzao liuln H boomsao x0013 Jed suotqedtorquv qoeaaoo JaqwnN ueew Mean Percent of Total Correct Anticipations 39 50 a 45 i ‘ —-"- Group E("objective first") \ ---- Group E 40 -» ‘ ‘ Group C 35 r 25 ~ 10 ” l I T l I I l l l 2 3 4 5 6 7 8 Serial Position Figure 6.--Mean percent of the total number of clusters correctly anticipated for each of the eight serial posi— tions. The curves are plotted from the cluster actually presented first in groups C and E and from §s' "objective first" clusters in group E. 40 TABLE 7.--Summary of the analysis of variance of the serial position effects in groups C and E. df Mean Square F p Groups (G) 1 635.59 11.35 .01 error g. 42 56.04 Positions (P) 7 238.88 14.45 01 G x P 7 275.50 16.67 01 error p. 294 16.53 Total 351 41 eight clusters. Or, the relative proportions of Ss' correct responses were not equally distributed among the eight clusters (or list positions). Discussion The present procedure of using serial items which were clusters of syllables repeated in different sequences, permitted comparison between a group in which the end of the list was disguised (group E) and a group in which the ends of the list were clearly demarcated by the presence of a visually and temporally distinctive ITI (group C). This procedure, then, allowed for replication of two of Lippman and Denny's (1964) groups, but under conditions re- quiring S to learn only an order (minimal response learning). In both experiments, group C learned significantly faster than group E.- In the present experiment, however, the task was very difficult; even at the end of 40 trials group C performed with no more than 25 percent correct anticipations (complete clusters anticipated correctly). Examination of the acquisition of the different parts of the eight cluster list showed that the primacy effect for group C was not limited to the first cluster, but spread to the second and third clusters as well. The ad- vantage for group C in learning the first cluster began on the first trial and continued at an accelerated rate. For group E, the different portions of the list all appeared to be learned at about the same rate. Thus the results support 42 Lippman and Denny's (1964; 1966) contention that serial position effects (conventionally plotted) are eliminated when the ends of a serial list are disguised and support as well the present prediction that idiosyncratic item effects are reduced when intralist similarity is high. It is of considerable importance to note that group C provided no strong evidence for a recency effect. Since group C had a temporally and visually distinctive ITI this- finding is contradictory to the results of previous re- search. One possible explanation is that the learning was not continued for a sufficient number of trials. Lippman and Denny found that the recency effect required over 9 trials to develop in a typical list of CVC trigrams and the present task was much more difficult. Another, and perhaps more cogent, explanation can be formulated upon examination of S's task. In the present study, a minimum of response learning was required. For all clusters (except cluster one in group C which was preceded by asterisks), the clusters were available. That is, the cluster in the window contained all the necessary responses for anticipating the following cluster. The Ss could simply read the four sylla- bles in the window in a different sequence to make an antici- pation. In other words, the present task required S to learn eight different sequences in addition 29 the serial order of the entire list. Since recency is also a serial order effect that is learned as a temporal sequence (Lippman and Denny), it is easy to see that learning the eight different sequences 43 (within clusters) could interfere with learning the serial order of the list (recency effect). This interpretation is consistent with the results shown in Figure 5 where all clusters are well matched in group B. In group C, the syllables of cluster 8 tended to be learned better than those of cluster 5 (small recency effect). This superiority seemed to manifest itself early in learning but remained minimal throughout. Presumably the extreme amount of inter- ference prevented S from learning the temporal sequence well enough to mediate a good recency effect. This interpretation gains support from Figure 6 (group E plotted from_the cluster actually presented first in the list), if the graph may be interpreted as showing that the last three clusters of the list were relatively easier to learn than the other five clusters of the list. The serial position curves plotted in Figure 6 (which present the proportion of S's correct responses made to each of the 8 clusters) show that if group E's data are plotted from each S's "objective first“ item, then primacy effects for group C and E appear nearly equivalent. This finding implies that, although groups C and E learned at different rates, the method or style of learning was quite similar for both groups. In group C, Ss learned from an item which was made distinctive by a highly discriminable ITI; in group E, Ss learned from an item which they found to be idiosyncrati- cally discriminable. Thus in group C, the ITI provided an objective cue which led all S5 to "select" (perceive) the 44 same item as the starting point of the series. When learn— ing was equated for groups C and E (i.e., measured in terms of percent of correct anticipations at each serial position), then it can be seen that the distribution of the proportion of correct responses is that which was altered by the dis— tinct starting point for the series. This interpretation of the re-distribution of correct responses is in keeping with research on the isolation effect in serial learning which has shown that, although the isolated term is correctly anticipated more frequently than its control (non—isolated) term, the number of correct responses at the ends of the series is proportionately reduced. Although intralist similarity was extremely high in the present study, Ss in group E nevertheless reached a moderate consensus as to which serial items were more dis- tinctive, namely, clusters 6, 7, and 8. If it can be assumed that Ss' "selection" of an "objective first" cluster is based upon the distinctiveness or difficulty of the cluster (and not list position), then the present results for group E lend support to Murdock's (1960) contention that a list of items having identical distinctiveness is impossible to .construct: ". . . assuming no two stimuli are identical it is literally impossible to achieve an 'equal-distinctive- ness' scale; that is, a group of stimuli (larger than two) where all stimuli have the same SZ value." (p. 23). 45 Summary A former study indicated that serial position effects are eliminated when the ends of a serial list are disguised (continuous serial learning (no ITI) with extralist items preceding the first trial). For the present experiment, it was suggested that the resulting serial position curve repre- sents the relative distinctiveness or ease in learning of the individual items of the list, provided that item effects are not counterbalanced by list rotation. The present experiment tested these findings and contentions under conditions of extreme intralist similarity, requiring a minimum of response learning. Two groups of 22 S3 each were presented an eight cluster list presented in the same order at a 4 second rate for 40 trials. All clusters (four nonsense syllables) were identical except for the position of the syllables within each cluster. The Ss were instructed to anticipate clusters by pronouncing the syllables as they would appear, from left to right, in the aperture. Group C learned with a long (12 second) visu- ally distinctive ITI; Group E learned a continuous serial list and three extralist clusters (other permutations of the same four nonsense syllables) preceded the first trial only. The results for group C supported previous findings that the advantage in learning the first cluster (item) is present from the outset of acquisition. The results of 46 group E supported previous findings that primacy effects are eliminated and that Ss select an idiosyncratically dis- tinctive item for a primacy cue when the ends of a temporal series are disguised. Of particular interest was the ab- sence of a recency effect in group C. This was attributed to the high intralist similarity (interference) among the responses in Ss' task. REFERENCES Bowman, R. E., and Thurlow, W. R. Determinants of the effect of position in serial learning. Amer. S. Psychol., 1963, SS, 436-445. Bulgarella, Rosaria A. The role of awareness in verbal learning. Unpublished doctoral dissertation, Michigan State University, 1965. Ebenholtz, S. M. Serial learning: Position learning and sequential associations. S. e p. Psychol., 1963, SS, 353-362. (a) . Position mediated transfer between serial learn- ing and a spatial discrimination task. S. e p. Psychol., 1963, SS, 603-608. (b) Feigenbaum, E. A., and Simon, H. A. Comment: The dis— tinctiveness of stimuli. Psychol. Rev., 1961, SS, 285-288. Horowitz, L., and Izawa, C. Comparison of serial and paired-associate learning. S. exp. Psychol., 1963, 62. 352-361. Hull, C. L., Hovland, C. 1., Ross, R. T., Hall, M., Perkins, D. T., and Fitch, F. B. Mathematico-deductive theory of rote learning. New Haven, Conn.£ Yale University Press, 1940. Jensen, A. R. An empirical theory of the serial-position effect. S. Psychol., 1962, SS, 127-142. Jensen, A. R., and Rohwer, W. D., Jr. What is learned in serial learning? S. verbal Learn. verbal Behav., 1965, S, 62-72. Keppel, G. Retroactive inhibition of serial lists as a function of the presence or absence of positional cues. S. verbal Learn. verbal Behav., 1964, 3, 511— 517. 47 48 Lippman, L. G., and Denny, M. R. Serial position effect as a function of intertrial interval. S. verbal Learn. verbal Behav., 1964, 3, 496-501. Comment on the role of the intertrial interval in serial learning: A clarification. Psychon. Sci., 1966, S, 234. Miller, G. A. The magical number seven, plus or minus two: Some limits on our capacity for processing information. Psychol. Rev., 1956, SS, 81-97. Murdock, B. B., Jr. The distinctiveness of stimuli. Psychol. Rev., 1960, S1, 16-31. Simon, H. A. A theory of the serial position effect. Brit. S. Psychol., 1962, SS, 307-320. Winnick, Wilma A., and Dornbush, Rhea L. Supplementary reports: Role of position cues in serial rote learning. S. exp. Psychol., 1963, SS, 419-421. Young, R. K. The stimulus in serial verbal learning. Amer. S. Psychol., 1961, 13, 517-528. Young, R. K., Patterson, Judith, and Benson, W. M. Back- ward serial learing. S. verbal Learn. verbal Behav., 1963, 1. 335-338. APPENDIX 50 APPENDIX A INSTRUCTIONS FOR ACQUISITION IN EXPERIMENT I This is an experiment in learning a list of nonsense syllables, and not a psychological test. We are interested in certain complex relationships in the learning process which are common to all people, and are not concerned with your personal reactions, intelligence, or personality. Shortly after the apparatus starts you will see a three-letter syllable in the window. Then a blank space, then another syllable, another blank space, and so on and so on. The syllables will always appear in the same order. As you learn, you are to try to anticipate the syllables; in other words, as you see one syllable or the blank after it, you are to spell out the syllable that will follow BEFORE it appears. If you think you know what a syllable will be, but are not sure, guess, because it will not hurt your score any more than to say nothing. And if you get it right it will count as a success. Always try to spell the syllables as distinctly as possible. Please continue anticipating the syllables until I tell you to stop. Please do not ask questions about the purpose of the study until the experiment is over. Do you have any questions regarding your task? 51 APPENDIX B INSTRUCTIONS FOR RECALL IN EXPERIMENT I Okay, stOp (counting vowels). You now have an opportunity, on one trial, to recall the list of syllables you Just learned. You will know the starting point when you see the first syllable. Anticipate the syllables as you did before. Be sure to make a guess each time. 52 APPENDIX C INSTRUCTIONS AND WORKSHEET FOR THE INTERVAL BETWEEN ACQUISITION AND RECALL For your next task, we want you to count the vowels (S, S, S, S, and S) in the sentences below. Count each of the vowels in the first sentence before going to the second sentence, and so on. Try to work quickly and accurately. Continue counting until you are told to stop. The quiet librarian whispers. The poor farmer struggles. The thirsty carpenter drinks. The strict teacher scolds. The hopeful fisherman waits. The clumsy typist errs. The confused pilot crashes. The lucky gambler wins. A; The preoccupied professor forgets. A The careless barber nicks. 53 APPENDIX D LIST OF NONSENSE SYLLABLES USED IN EXPERIMENT I MEP GAW KOJ CIB ZUR TEY WOQ XIG NAH JEC QUT YOF 54 APPENDIX E SUPPLEMENTARY ACQUISITION DATA Group E4 E7 E10 C Trials to one perfect M 3.58 3.04 3 5H 2.96 S.D. 1.18 0.81 1.72 0.86 Trials to four perfect M 8.88 7 58 10 00 7.96 S.D. 3.60 2 69 3.80 3.43 Trials to seven perfect M 15.50 16.13 15.33 S.D. 5 59 4.98 4.71 Trials to ten perfect M 24.75 22.96 S.D. 6.94 6.19 "11111111111111.1111?