130 633 THS ASSUCIATWE RELAUONS EN VERBAL DISCRIMINAHON LEARMNG Thesis for the Degree of M. A. MICHIGAN S'f'ATE UNWERSITY CHERYL BOWSHIER WHALEN 1968 1m LIBRARY Michigan State University ‘ BINDING av ‘5 BUUK BINUERY INC. LIBRARY amozns SH‘IGPN". RICHIE“ IIIIAI; & sous“ ' I ABSTRACT ASSOCIATIVE RELATIONS IN VERBAL DISCRIMINATION LEARNING by Cheryl Bowshier Whalen According to the two-stage memory model, storage processes in verbal discrimination (VD) learning are cru- cial. Any existing relation which allows items of a pair to be more easily combined during storage, such as strong intrapair associative relations, aids learning because the pair may be stored as a unit, thus reducing the memory load. The present experiment is a test of the two-stage memory model assumption that storage processes are crucial to the learning of associative pairs. Four fourth-grade and four sixth-grade groups were given three study trials and three test trials alternately on each of two 24-pair VD lists. Half the pairs on each list were strong associates; the other half were unrelated pairs. Pairing on the study and test trials was manip- ulated. The pairing was either constant or varied from trial to trial resulting in four conditions: CC, CV, VC, and VV. The first letter designates whether the pairing was constant (C) or varied (V) on the study trials; the second, the test trial condition. Cheryl Bowshier Whalen Under constant study Conditions associative rela- tions facilitated performance. Under varied study condi- tions, performance on associative pairs was inferior or equivalent to that on unrelated pairs. All associative conditions were significantly different from each other suggesting that rearranging the pairing on test trials af- fected performance as well as rearranging the pairing on study trials. A corollary to the two-stage memory model was suggested to account for these differences. All unrelated conditions were statistically equiv- alent with the exception of U-VV which was significantly inferior to all other unrelated conditions. Association by contiguity was suggested as a possible explanation. Under constant study conditions, the grades were statistically equivalent. However, the grades significantly differed when the pairing on study trials was varied. One hypothesis presented for this difference was that sixth graders were more able to c0pe with an inconsistent sit- uation. A second hypothesis presented was that this dif- ference tapped a critical change in children within this Approved W Committee C airman Date 262 /MW /?63 age range. ASSOCIATIVE RELATIONS IN VERBAL DISCRIMINATION LEARNING BY Cheryl Bowshier Whalen A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF ARTS Department of Psychology 1968 5,5317/ /,;2 2'47 ACKNOWLEDGMENTS The author gratefully acknowledges the assistance of Doctor Abram M. Barch, without whose help this study would not have been possible. The author is also indebted to the other members of her guidance committee, Doctors David C. Raskin and Ellen Strommen, for their advice and encouragement. ii TABLE OF CONTENTS Page ACKNOWLEDGMENTS. . . . . . . . . . . . . . . . . . . . ii LIST OF TABLES . . . . . . . . . . . . . . . . . . . . iv LIST OF APPENDICES . o . . . . . . . . . . . . . . . . V INTRODUCTION . . . . . . . . . . . . . . . . . . . . . l Interpair Similarity . . . . . . . . . . . . . . . . Intrapair Similarity . . . . . . . . . . . . . . . . Purpose. 0 O O O O 9 O O 0 0 O O O O O O 0 O O O O O comm METHOD 0 O O O O 9 O O O O O O O O O O O O O O O O O O 10 Subjects . . . . . . . . 10 Materials 0 O O 0 O O O O O O 0 O O O O O O O O O O O 10 Procedure. 0 O O O O O O O O O O O O O O O O O O O 0 13 Experimental Conditions. . . . . . . . . . . . . . . 13 RESULTS 0 O O O O O O O O O 0 O O O O O O O O O I I O O 15 DISCUSSION 0 O O O O O O O O O O O O O O O O O O O O O 20 REFERENCES 0 O O G O O O O O O O O O O O O O O O O O O 28 APPENDICES O Q 0 0 O O O O O O I O O O O O O O O O O O 30 iii L IST OF TABLES Table 'Page 1. Summary of Grades K Study Conditions x Test Conditions X Item Type Analysis of Variance Comparing the Mean Total Correct Responses (L-l and L-2 Combined). . . . . . . . . . . . . . l6 2. Newman-Keuls Test for the Item Type X Study Conditions x Test Conditions Interaction (Mean Number Correct) . . . . . . . . . . . . . . l7 3. Newman-Keuls Test for the Grades x Study Conditions Interaction (Mean Number Correct). . . l7 4. Newman-Keuls Test for the Grade X Item Type X Test Conditions Interaction (Mean Number correct) I I O O O 0 O O O O 0 O O O O I O O O O O 18 iv LIST OF APPENDICES Appendix Page A. Lists 0 O O C O O O O O O O O O O I O O O O O O 30 B. Instructions . . . . . . . . . . . . . . . . . 33 INTRODUCTION The frequency theory of verbal discrimination (VD) learning proposed by Ekstrand, Wallace, and Underwood (1966) postulates that the cue for discrimination is the subjec- tive frequency difference between the correct (C) and in- correct (I) alternatives (items) of a VD pair. Each re- sponse to an item adds one frequency unit. Whenever a subject perceives, pronounces, or rehearses an item, frequency units are added. The perception of each alter- native is referred to as a representational response (RR). Pronouncing an item aloud during anticipation is called a pronounciation response (PR). The rehearsal response is referred to as the rehearsal-of-the-correct-alternative response (RCR). The RCR is a pronouncing of the C item implicitly or explicitly, and it may occur more than once after the subject is supplied with information about which .item is C. No rehearsal of the I item is assumed to occur. The frequency theory also proposes a fourth response mechanism capable of adding frequency units to VD items. The frequency of a given item may be increased if that item has occurred elsewhere in the list as an implicit associa- tive response (IAR). The subject may confuse an IAR with an RR, so that IARs may add frequency units in the same sense that an RR does. During a particular VD trial, all items receive at least one frequency unit from the RR. As VD learning progresses, the C item accumulates more frequency units than the I item as a result of the subject's rehearsal of the C item and the additional presentation of the C item under some feedback conditions, e.g., anticipation. Ignoring whether the PR occurs to the C or I item, a dif- ferential frequency of at least 2:1 in favor of the C item is built up due to RCR. This increase in favor of the C item makes discrimination on the basis of differ- ential frequency possible: the subject is able to choose either the most frequent item in a pair (Rule 1); or the least frequent item in a pair (Rule 2). Frequency theory specifies that the more similar two items are, whether within a pair or between pairs, the more likely an RR, PR, RCR, or IAR made to one item will be confused with an RR, PR, RCR, or IAR made to the other one. This confusion between items leads to a blur- ring of frequency differentials, thus producing poorer learning by the subject. Ekstrand gt_§l. (1966) specif- ically predicted greater interference from intrapair associative relations than from interpair associative relations due to the possible priming effect that might take place for the IARs when associates are presented simultaneously. Interpair Similarity Using nonsense syllables, Underwood and Archer (1955) obtained interference when formal interpair simi- larity was increased from a low to a high level of simi— larity. However, their rules for forming the high formal similarity list resulted in similarity not only between pairs but within pairs. When intrapair similarity was held to a minimum, Battig and Brackett (1963) did not find clear-cut differences with three levels of formal interpair similarity. Ekstrand et_al. (1966) tested aspects of the fre- quency theory by varying relative frequency within VD pairs through the use of similarity relations between pairs. They reasoned that when the same item or strong associates are presented as the C item in two pairs, additional fre- quency units are added to the C items. The result is a greater differential frequency in favor of the C item and a facilitation of learning. When the same item or strong associates are the I items in two pairs, the I items re— ceive additional frequency units, thereby decreasing the frequency difference between C and I items and producing interference. When the same item or strong associates are C in one of the two pairs and I in the other pair, maximum interference occurs because the frequency cue is very diffi- cult to use. Rule 1 is apprOpriate for some pairs and Rule 2 for others. VD learning was found to be facilitated to an equivalent degree when the same item or strong associates were presented as the C item in two pairs. Interference occurred as predicted when the same item was I in two pairs. However, the group receiving strong associates as I items in two pairs did not show measurable interference effects in comparison to a control group receiving unre- lated VD pairs; nor did it differ from the group receiving the same words as I items in two pairs. That is, there was no indication that strong associates as I items in two pairs measurably influenced the subjective frequency difference between C and I items. The condition in which the same item was C in one pair and I in the other pro- duced twice as many errors as strong associates under the same condition and both showed interference in relation to the control group. In summary, contrary to the frequency theory pre- diction, increasing formal interpair similarity does not seem to produce interference when intrapair similarity is low. In the extreme case of formal similarity, where one of the items in each of two pairs is identical, the fre- quency theory predicts the results exactly. These gener— alizations are dangerous, for the two studies on which they are based used different experimental materials. Words with their concomitant meanings may not necessarily behave as nonsense syllables behave. However, the frequency theory does not differentiate between the two: presumably they should behave similarly. Frequency theory does not fully explicate inter- pair associative similarity. The fact that strong asso- ciates are facilitative when they are the C items in two pairs but not interfering when they are the I items in two pairs is unexplained: the first result implies that IARs are functioning as assumed; the second, that IARs are not functioning as assumed. Intrapair Similarity Underwood and Viterna (1951) compared VD lists of high and medium intrapair meaningful similarity. Inter- pair similarity was low. They found a small but signifi- cant difference in learning in favor of the medium simi- larity group. Edwards (1966) used a four-choice VD task to study separately the effects of within-display and between-display similarity. Within-display similarity did not affect learning of nonsense syllables, but did interfere with the learning of forms. Between-display similarity retarded performance with both nonsense syl- lables and forms. Two studies have compared the performance of normal and deaf subjects on VD tasks in which intrapair similarity was varied. Putnam, Iscoe, and Young (1962) found that highly meaningful similar word pairs were easier to learn than unrelated pairs for both normal and deaf adolescents. Youniss, Feil, and Furth (1965) reported that deaf adoles- cents find physically similar pairs more difficult than unrelated ones, but that this is not true for pairs with similar meaning. Normal adolescents and normal and deaf college students had greater difficulty with the meaning- fully similar pairs. Barch, Lippman, and Whalen (1967) found that the VD learning of grade school children depended upon the type of intrapair similarity and frequency of usage. For low frequency of usage words, learning was facilitated by both formal similarity (e.g., "leopard-leper") and meaning- ful similarity (e.g., "muck-filth"). For very frequently used words, formal similarity (e.g., "desk-deck") and meaningful-formal similarity combined (e.g., "began-begun") produced interference. However, meaningfully similar pairs did not differ from unrelated pairs. Strongly associated word pairs were significantly easier to learn than any other class of common words. In the above study, the only criterion for calling a pair meaningful was that the words have similar meaning. Thus, the meaningfully similar class of pairs included some strong associates with similar meaning (e.g., "pretty- beautiful"). Meaningfully similar pairs which were strongly associated were among the easiest in the list. Thus, what- ever made high associative pairs easier did not exclude pairs that were synonyms of each other. Subsequent unpub- lished research by Barch and associates has confirmed the facilitation found for associative pairs in both mixed and unmixed lists with grade school children. To explain the paradoxical finding that intrapair associative similarity facilitated learning, Barch gt_31. (1967) proposed a two-stage memory model which asserts that subjects attempt to reduce the memory load as much as possible in the VD task. To do this they "tag" or "code" one item of a pair as C, then collapse the pair and its tag together for storage. The number of memory units is the number of collapsed pairs. During the test trial the collapsed unit is searched for and retrieved, unfolded, and inspected for the tagged C item. The lin- guist Yngve (1960) has proposed a similar memory load reduction model to account for sentence production. In the Barch eE_al. (1967) model storage processes are most crucial. Any existing relation which allows items of a pair to be more easily combined aids learning, be- cause once the pair is collapsed the tag is safe from interference. Any relations conflicting with the intra- pair relation interfere with the combining process. As a result, the tag for the C item may be lost before storage, i.e., the C item may be forgotten; or the members of the pair may have to be stored separately, increasing memory load, and leaving the tag highly susceptible to interference. The ease of combination for any pair depends on the rela- tive strength of the intrapair relation as compared to alternative relations for the items. The more linguistically sophisticated the subject,’ the more likely it is that he has potential alternative relations. Thus, college students may not benefit as much by intrapair associative relations. Because the obtained associative effect was so strong with grade school children, Barch et_al. (1967) hypothesized that associative pairs are at least as easy to learn as unre— 1ated pairs for college students. Support for this pre- diction has come from a study by Eberlein and Raskin (1968) in which interpair association produced significant inter- ference in the VD learning of college students, whereas intrapair associations had no effect. In summary, whether or not intrapair similarity facilitates, interferes with, or makes no difference in learning is a complex interaction between the type of interpair similarity, the age of the subject, and the frequency of usage of the stimulus material. Purpose The purpose of the present experiment was to com- pare the predictions of the frequency theory and the two- stage memory model concerning the effects of intrapair and interpair associative similarity. Both the two-stage memory model and the frequency theory predict interference from having associates in the list but not paired together (interpair associations). The frequency theory predicts that rearranging the pair- ing of unrelated items is not detrimental to learning. For the frequency theory, the unrelated items will always retain their frequency units even when they are not paired consistently with the same other unrelated item. The two- stage memory model says nothing about unrelated items so cannot make a prediction. The present experiment is a direct test of the two- stage memory model's assumption that storage processes are crucial. If this is so, rearranging the pairing on the study trials, and thus upsetting the combining of items in pairs, will produce more interference than rearranging on the test trials. Rearranging the pairing should interact with item type according to the two-stage memory model, i.e., the learning of associative items is more interfered with if the rearranging occurs on study trials than is the learning of unrelated items. The frequency theory does not make specific differential predictions concerning rearranged pairing on study trials and test trials. Pre- sumably both cause equal interference for associative pairs but none for unrelated pairs. METHOD Subjects The SS were 105 fourth-grade and 113 sixth—grade children from two Lansing, Michigan elementary schools. The four classrooms in each grade were alphabetically halved, resulting in groups of 11 to 16 gs each. Each group was assigned to one of the four experimental con— ditions. Seven fourth-grade and three sixth-grade gs were omitted for response alternation or responding more than once per item. To assist statistical analysis, addi- tional gs were nonsystematically eliminated until the four experimental conditions contained 22 fourth-grade and 26 sixth-grade gs each. Materials All but six of the words used to form 48 verbal discrimination pairs were A or AA on the Thorndike-Lorge (1944) general count. The exceptions, which occur in associative pairs (A pairs) only, were: bug, hammer, insect, navy, subtract, and thirsty. Only bug, thirsty, and subtract occurred less than 100 times in the Thorndike- Lorge count of 120 juvenile books. 10 11 Half of the 48 verbal discrimination pairs were high A pairs; the other half were unrelated pairs (U pairs). Except for two continuation pairs ("sail-boat" and "cry- baby") and two contrast pairs ("bad-good" and "army-navy"), the A pairs were a stimulus item and its most frequent free associate as given by the Palermo and Jenkins (1964) and Entwisle (1966) word association norms for children. The four exceptions lacked children's norms, but previous data indicated they would act similarly to the other A pairs. The U pairs had no primary associative links within or between pairs. Two VD lists of 12 A pairs and 12 U pairs each were constructed. Formal and meaningful similarity be- tween pairs on a list was minimized. The correct words were chosen so that formal and meaningful similarity within or between lists could not be used as differential cues for learning the correct words. Each of the lists had an alternatively-keyed version in which all the cor- rect words on Key A became incorrect on Key B, and vice versa. For each list three study trials and three test trials using six different serial orders of the pairs were prepared. A pairs and U pairs were intermixed. No pair appeared in the same serial position, nor was it preceded or followed by a given pair more than once. The left-right sequence of C and I items within a pair varied 12 across the trials. Approximately half of the correct words on each trial occurred first in a pair and the other half second. Voice recordings of both study and test trials were recorded by the E’on Scotch magnetic low-print tape using a 602-2 Ampex recording system. A Stowe memory drum was used for timing. The clicking of the drum mechanism was minimally present on the tape. A 5 sec. presentation rate per pair was used for both study and test trials, i.e., 5 sec. occurred between the beginning of the first word of one pair and the beginning of the first word of a succeeding pair, with about 1 sec. re- quired for saying each word of the pair. The test trials consisted of the sequential number of each pair followed by the pair itself (e.g., "one, pretty-beautiful"). Test trials were recorded first on a master tape. The same version of these test trials served as a presentation for the two alternately-keyed conditions. Thus, possible accidental uses of stress or intonation could not in any systematic way indicate which word of a pair was correct. On the study trials, 2.5 sec. after the beginning of a pair, the correct word was repeated and identified as correct (e.g., "beautiful is correct"). After each study trial was recorded, the relevant test trial was recorded from the master tape onto the final presentation tape. The intertrial interval was 10 sec. 13 Procedure The experiment was given to groups of 11 to 16 SS each. The E read the instructions to the SS and clarifica- tion was given whenever necessary. The VD lists were pre- sented via a Wollensak tape recorder. The alternating study trial-test trial procedure was used to separate in time the hypothesized storage and retrievalpmocessesof the two-stage memory model and to study separately the effects of intrapair and interpair associative relations. The SS received three alternating study and test trials on each list. Lists were presented in the same order for all groups since counterbalancing the lists would have required an unwieldy number of groups. The list given first will be referred to as List 1 (L-l), the second as List 2 (L-2). Each group received both lists under one experimental condition. There was 5 to 8 min. between lists. gs responded on a special answer sheet on which, for each pair, they circled a "1" if the correct word occurred first in the pair and a "2" if the correct word occurred second. Experimental Conditions Pairing on the study and test trials was manipu- lated. The pairing was either constant or varied. On the l4 constant trials, each word was always paired with its mate, i.e., each word was paired with the word with which it had originally been paired when constructing the pairs. On the varied trials, the words were never paired with any other word more than once. Any correct word from a given A pair was always paired with an incorrect word from another A pair; U words were likewise paired with U words. All possible combinations of pairing were used, resulting in four conditions: CC, CV, VC, and VV. The first letter designates whether the pairing was constant (C) or varied (V) on the study trial; the second, the test trial condition. The study trials for Conditions CC and CV were identical; likewise, the study trials for VC and VV. The test trials for Conditions CC and VC were identi- cal; likewise, the test trials for CV and VV. RESULTS Since the two lists could not be counterbalanced, it was not possible to distinguish differences due to pre- sentation order of the lists from differences due to possible differential list difficulty. Therefore, the two VD lists were combined and list position was not used as a variable in the statistical analysis. A 2 X 2 X 2 X 2 analysis of variance (Table 1) showed all main effects to be significant except Grades. 49.31) were easier than U items (Y = 47.95). A items (2 Constant Study Conditions (Y = 51.64) were easier than varied Study Conditions (Y = 45.62). Constant Test Conditions (2 = 49.91) were easier than varied Test Conditions (it: 47.35). However, the interpretation of the main effects was modified by the significant interactions found. The Item Type X Study Conditions X Test Conditions interaction was significant, F (1,184) = 3.95, p<.05. The Newman-Keuls Test (Winer, 1962) comparing the means for this interaction is shown in Table 2. All comparisons between A and U items within a particular Condition were significant except A-VV and U-VV. A items were easier than U items under both Condition CC and Condition CV; U items were easier than A items under Condition VC; and, A and U 15 16 Table 1.--Summary of Grades X Study Conditions X Test Con- ditions X Item Type Analysis of Variance Comparing the Mean Total Correct Responses (L-l and L-2 Combined) Source df Mean Square E Between 88 191 Grades (G) 1 113.27 1.08 Study Conditions (SC) 1 3,468.01 33.02** Test Conditions (TC) 1 625.26 5.95* G X SC 1 932.03 8.87** G X TC 1 .03 <1 SC X TC 1 13.86 <1 G X SC X TC 1 367.54 3.50 Error (b) 184 105.03 Within 88 192 Item Type (I) 1 178.76 ll.37** I X G l .01 <1 I X SC 1 969.01 61.64** I X TC 1 8.76 <1 I X G X SC 1 7.88 <1 I X G X TC 1 82.40 5.24* I X SC X TC 1 62.16 3.95* I X G X SC X TC 1 23.36 1.49 Error (w) 184 15.72 Total 383 *p<.05 **p<.01 17 Table 2. Newman-Keuls Test for the Item Type X Study Con- ditions X Test Conditions Interaction (Mean Num- ber Correct) A-VV U-VV A-VC U-VC U-CV U-CC A-CV A-CC 43.43 44.74 45.86 48.17 48.76 49.82 52.19 55.45 Table 3. Newman-Keuls Test for the Grades X Study Conditions Interaction (Mean Number Correct) 43.34 47.56 50.70 52.74 Note: Any two means underlined by the same line are not significantly different. items did not differ significantly under Condition VV. For the A items, all Conditions were significantly different and were ordered from easiest to most difficult in the fol- lowing manner: A-CC, A-CV, A-VC, A-VV. Conditions U-CC, U-CV, and U-VC were not significantly different. However, Condition U-VV showed significantly poorer performance in comparison to all other U Conditions. 18 The only significant interactions involving the Grade variable were the Grade X Study Condition interaction, F(1,184) = 8.87, p<.Ol, and the Grade X Item Type X Test Conditions interaction, F(1,184) = 5.24, p<.05. The Newman- Keuls comparison between means for the Grade X Study Con- ditions interaction appears in Table 3. The performance of constant Study Conditions was better than the perfor- mance of varied Study Conditions for both grades. Under constant Study Conditions, the fourth grade and sixth-grade groups do not differ significantly. Under varied Study Conditions, the performance of the fourth graders was sig- nificantly inferior to the sixth graders. Table 4.--Newman-Keuls Test for the Grade X Item Type X Test Conditions Interaction (Mean Number Correct) 4-U-VT 4-A-VT 6-U-VT 4-U-CT 6-A-VT 6-U-CT 4-A-CT 6-A-CT 46.70 46.75 46.92 48.02 48.85 49.96 50.68 50.79 Note: Any two means underlined by the same line are not significantly different. The Newman-Keuls comparison between means for the Grade X Item Type X Test Conditions interaction is shown in Table 4. Generally, the groups under constant Study Conditions consistently performed better than those groups 19 under varied Test Conditions although not always signifi- cantly better. Irregardless of grade, the A items under constant Test Conditions were easiest. Under constant Test Conditions, the grades do not significantly differ on the A items. However, under constant Test Conditions the sixth graders made significantly more correct responses on the U items than the fourth graders. There were no significant differences between grades or item types under varied Test Conditions. DISCUSSION Barch, §E_21. (1967) found A pairs to be easier than U pairs for fourth, sixth, and eighth grade children. The present experiment replicated this finding for fourth and sixth graders under presentation conditions which elim- inated any possible accidental effects from voice cues. The facilitation produced by associative relations resulted in a significant Item Type main effect even though more errors occurred on A items than on U items under Conditions VC and VV. Varying the conditions under which gs study or are tested affected performance. It can be seen from Table 3 and Table 4 that constant study or constant test conditions facilitated performance irregardless of both grades and item types. As predicted by the two-stage memory model, there was a significant Item Type X Study Conditions X Test Con- ditions interaction. A comparison of A pairs and U pairs showed that under constant study conditions, associative relations facilitated performance. It is assumed that the strong intrapair associative relationship expedited tagging, collapsing, and combining during storage. When the pairing was varied during study trials, interpair associative 20 21 relations were much stronger than intrapair, thus hindering performance on A pairs. Significant differences were found between all associative conditions. The conditions for A items ranked from least to most difficult were A-CC, A-CV, A-VC, A-VV. Neither the frequency theory nor the two-stage memory model explain these results entirely. The frequency theory predicts that intrapair and interpair associative relations are interfering when one associate is C and the other I in two pairs. It does not make predictions as to the differential effects of study or test conditions. Therefore, it offers no explanation as to why the conditions ordered in this fashion. The two-stage memory model predicts that intrapair associative relations are facilitating under constant study conditions, i.e., interpair associative relations are rel- atively more interfering when they occur on study trials than when they occur on test trials. Intrapair associative relations were the most facilitating under constant study conditions. A comparison of the means shows that there is a large drOp from A-CV to A-VC. The A pairs under constant study conditions performed better than all other groups and the A pairs under varied study conditions performed worse than all other groups. However, there was some decrement in performance when A pairs were rearranged on test trials only (Condition 22 CV as compared to CC) and some gain from having the two items of an A pair presented together on test trials only (Conditions VC as compared to Condition VV). The two-stage memory model makes no specific predictions as to the effects of intrapair associative relations on test trials. Perhaps storage processes also occur during the test trials. If this corollary to the two-stage memory model is true, on the test trials the gs would be retrieving, unfolding, and inspecting the pairs for the tagged C item as well as saving the tag, recombining, and recollapsing. Some of these Operations may be incompleted, thus causing a decrement in performance. Applying this corollary to the results obtained for Conditions CV and VC makes it clearer. Under Condition CV, the items of a particular A pair are in two different serial positions in the list. For the item first to appear, the A pair is retrieved, un— folded, and inspected for the tagged C item. If that pair is not re-stored correctly, the tag may be lost before the second item of that pair appears. A testable consequence of this corollary is that the items appearing first for particular A pairs should show more correct responses than those items occurring second, i.e., performance should be better the first time either item of an A pair is presented than the second time an item from that A pair is presented. Another explanation may be made for these results. Suppose the SS prefer to know which item of a pair presented 23 on a CV test trial is correct. If the pair "water-beautiful" were presented and "beautiful" were correct, the gs may first retrieve the pair "water-thirsty," look for its tag, and see that "thirsty" is correct. However, this does not tell the §_which of the two words presented was correct. Therefore, if allowed enough time, the §_may then retrieve the pair "beautiful-pretty," see that "beautiful" is correct, and only after retrieving both these pairs respond. Indirect evidence has shown that younger children perform best when given time to search for the correct response (Suppes and Ginsberg, 1962). If grade school children perform in this fashion on a VD task, there should be a developmental difference between grade school and adult gs, because there is evidence that adults perform equally well under conditions in which the verbal reinforcement is right or wrong on a VD task (Spence, Lair, and Taylor, 1963). This assumes that a self-generated reinforcement functions similarly to reinforcement given by an outside source. The specific prediction would be that no difference would occur between Groups A-CC and A-CV for adult gs. An unpublished study by Whalen and Barch, using essentially the same design as the present experiment and college students as gs, showed no differences between A-CC and A-CV. This second explanation may be tested in other ways. One obvious way is to present items singly on the test trials. This technique would also allow a direct test of the two 24 explanations presented. The first explanation treats cor- rect and incorrect alternatives as the same. All that is important is which alternative occurs first in serial posi- tion. It predicts that when A items are presented singly, the first item of an A pair is more likely to be responded to correctly. The second explanation predicts performance is better, and the latency of responding shorter, when the correct alternative is presented singly than when the incor- rect alternative is presented. A comparison of Groups A-VC and A-VV showed some gain from having A pairs presented together on test trials. This gain is not enough to counteract the interference gen- erated from strong interpair A relations on the study trials under Condition VC, for Group A-VC performs significantly poorer than Group U-VC. The interpretation that storage processes also occur during the test trials may be extended to cover the results found under Condition VC. Because the A items are not presented together on study trials, collapsing and combining does not occur. Tagging, however, may occur. On the test trials when A pairs are presented simultaneously, the §_may spend most of his time trying to combine and collapse, and in so doing lose the tags. This would most likely affect Test Trial 2. Having the two items of an A pair presented together on Test Trial 1 for the first time sensitizes the S to the fact that these relationships are there. The §_may spend 25 Study Trial 2 trying to utilize these strong associative relationships and thus not be tagging the correct items. The trial analysis on L-l showed Test Trial 2 under Condi- tion VC to be significantly inferior to Test Trial 3. This effect is all the more striking because no other trials within experimental conditions were significantly different. This suggests that this sensitization to the associative relations is occurring, but that by the third trial, the S has, in effect, learned to work with it. When U items were rearranged on either study trials (Condition CV) or test trials (Condition VC), but not on both (Condition VV), no significant differences occurred when compared to constant study and test conditions (Con- dition CC). When the pairing of U items was rearranged on both study and test trials, performance significantly dropped in comparison to all other unrelated conditions. This decrement is unexplained if one assumes the necessary and sufficient condition for VD learning is the presence of a differential frequency between C and I items. The two-stage memory model, if it predicts anything for U items, would predict a difference in favor of U-CC over U-VV. It has no specific predictions concerning Groups U-CV and U-VC. This model was designed to account only for strong associative relations. However, Spear, Ekstrand, and Underwood (1964) reported that an association due to contiguity is develOped 26 between the C and I items in a pair. Let us assume asso- ciations by contiguity are facilitating in the same way that strong pre-existing associative relations are. Then, to the extent that associations by contiguity develOp, VD learning should be facilitated. The present results showed that learning was-facilitated whenever U items were pre- sented under conditions that would be expected to build up contiguous associative strength. in a.pair, i.e., whenever conditions were constant, either on study trials, test trials, or both. When conditions did not favor such con- tiguous associations, performance deteriorated. Contiguous associations thus seem to perform in a manner similar to strong associative relations. They do not perform exactly in the same way, either because they did not reach sufficient strength to cause differences between Groups U-CC, U-CV, and U-VC, or because interpair contiguous associations are not interfering. Under constant study conditions, the performance of school children in different grades was statistically equivalent. However, the grades significantly differed when the pairing on study trials was varied. Varying the study trials interfered with the performance of both grades, but significantly more interference occurred for the fourth grade. One hypothesis for this difference is that the sixth graders are more able to c0pe with a varied and inconsistent study situation because they have two years more experience 27 with using words and with learning situations in general. A second hypothesis is that this difference taps a critical change in children within this age range with respect to their ability to handle inconsistent situations when pairs of commonly used words are used.. In no previous research using commonly used words by Barch and associates were dif- ferences found between fourth and sixth graders. Only when §s were given low frequency words under constant conditions did fourth and sixth graders differ. REFERENCES Barch, A. M., Lippman, M., and Whalen, S. Intra-pair rela- tions in verbal discrimination learning: A para- doxical finding. Paper read at Midwestern Psychol— ogical Association, Chicago, 1967. Battig, W. F. and Brackett, H. R. Transfer from verbal- discrimination to paired-associate learning: II. Effects of intralist similarity, method, and per- centage occurrence of response members. J. exp. Psychol., 1963, 65, 507-514. Eberlein, E. and Raskin, D. C. Intrapair and interpair associations in verbal discrimination learning. Psychon. Sci., 1968, 11, 145-146. Edwards, D. C. Similarity in verbal discrimination. Percept. mot. Skills, 1966, 23, 815-820. Entwisle, D. R. Word Associationslof Young Children. Baltimore: Johns Hopkins Press, 1966. Ekstrand, B. R., Wallace, W. P., and Underwood, B. J. A frequency theory of verbal—discrimination learning. Psychol. Rev., 1966, 13, 566-578. Palermo, D. S. and Jenkins, J. J. Word Association Norms. Minneapolis: University of Minnesota Press, 1964. Putnam, V., Iscoe, I., and Young, R. K. Verbal learning in the deaf. J. comp. physiol. Psychol., 1968, 55, 843-846. I Spear, N. E., Ekstrand, B. R., and Underwood, B. J. Asso- ciation by contiguity. J. exp. Psychol., 1964, 61, 151-161. Spence, J. T., Lair, C. V., and Goodstein, L. D. Effects of different feedback conditions on verbal- discrimination learning in schizophrenic and non- psychiatric subjects. J. verb. Learn. and verb. Behav., 1963, 2, 339-345. 28 29 Suppes, P. and Ginsberg, R. Application of a stimulus sampling model to children's concept formation with and without overt correction responses. 'J. exp. Psychol., 1962, 63, 330-336. Thorndike, E. L. and Lorge, I. The Teacher's Word Book of 30,000 Words. New York: Columbia University,71944. Underwood, B. J. and Archer, E. J. Studies of distributed practice: XIV. Intralist similarity and presen- tation rate in verbal-discrimination learning of consonant syllables. J. exp. Psychol., 1955, 50, 120-124. - Underwood, B. J. and Viterna, R. D. Studies of distributed practice: IV. The effect of similarity and rate of presentation in verbal—discrimination learning. J. exp. Psychol., 1951, 43, 296-299. Winer, B. J. Statistical Principles in Experimental Design. New York: McGraw Book Company, 1962. Youniss, J., Feil, R. N., and Furth, H. G. Discrimination of verbal material as a function of intrapair sim- ilarity in deaf and hearing gs. J._educ. Psychol., 1965, 56, 184-190. Yngve, V. H. A. A model and an hypothesis for language structure. Proc. Amer. Phil. Soc., 1960, 104, 444-466. APPENDICES APPENDIX A LISTS Associative Pairs bad - good cold - hot high - low chair - table army - navy king - queen you - me always - never go - stop cry - baby thirsty - water sad - happy 31 List 1 Unrelated Pairs able - deep calm - huge tall - poor fruit - shoulder bird - desk back - curtain under - along then - quite lift - act best - noise smooth - train. important - yellow Associative Pairs black - white there - here baby - cry long - short beautiful - pretty bug - insect girl — boy nail - hammer boat - sail hard - soft begin - start subtract - add 32 List 2 Unrelated Pairs evil - fast heat - dollar they - of sweet - quiet better - large inch - ear note - game strong - deck milk - car miss - prevent bent - glad jump - receive APPENDIX B INSTRUCTIONS 34 Instructions I'm Mrs. Whalen from Michigan State University. We are going to play a word game. In this word game you will have to learn some words. You will have to listen very carefully. When I turn on the tape recorder, you will hear a list of pairs of words. One word in each pair will be cor- rect. You must learn the correct words. First, all the pairs of words will be presented in a study trial and you will be told which word is correct in each pair. For example, you might hear, "Book-Marble, Book is correct" or "Frog-Apple, Frog is correct." After you have heard all the pairs and have been told which word is correct in each pair, the pairs will be presented again. This will be a test to see how well you can remember the correct words. On the test trials, only the pairs will be presented and you must tell me which word in each pair is correct. This is how you are to tell me. If you will look at your answer sheet you will see that under Test 1, for example, there are the numbers 1 to 24. After each of these numbers is a l and a 2. If the first word of pair presented during the test trials is correct, you would circle 1; if the second word of a pair is correct you would circle 2. For example, you know that Book is correct. If Book occurred first in the pair you would circle 1; if Book occurred second 35 in the pair you would circle 2. What would you do if I said "Book - Marble?" (Answer from children.) What would you circle if I said "Apple - Frog?" (Answer from children.) The correct words are always the same, but whether they come first or second in the pair changes. Are there any questions? Now we're ready to start Study Trial 1. Remember on the study trials you don't write--just listen and try to learn the correct words. (St0p tape between Study 1 and Test 1.) Now we're ready for Test Trial 1. Please answer every item. If you're not sure which word of a pair is correct, guess. Remember if the first word of pair is correct, circle 1; if the second word of a pair is correct, circle 2. (At beginning of List 2.) On the second part of the experiment, everything is exactly the same, except that you will be learning a new list of words. Your job is still to learn the correct words. Ready? We'll start with Study Trial 1 again. AT ”'II'II'I‘IIII’IIII’I III II IIIII'IIII'I’IIIIIIIIII’“