h;t‘1!:.‘§lfi~?ag 1:..‘27; I, ' h f. V... :"‘I.li-"' ' l ,. “at“. In I» ”v" .ar- “mi-i} i‘f '3: U‘. I i. ASE; MSU LIBRARIES “ \— RETURNING MATERIALS: Place in book drop to remove this checkout from your record. FINES wi11 be charged if book is returned after the date stamped below. fl ABSTRACT THE EFFECT OF ENMILIARIZATION WITH ELEMENTS OF COMPOUND STIMULUS ITEMS UPON THE SUCCESSFUL EMPLOYMENT OF ASSOCIATIVE STRATEGIES BY EDUCABLE RETARDATES by Daniel B. Berch Previous research employing the paired-associates (PA) task with educable retardates has shown that experimenter-induced-strategies can facilitate learning. However, one study has shown that retardates _did not make optimal use of the strategy aids provided them. The purpose of the present study was to determine if retarded children, given prior familiarization with relevant stimulus cues, could make maximally effective use of strategy aids in a PA task. One group received a "relevant differentiation" treatment. Four other groups were formed, each receiving one of the following treat- ments: 1) familiarization with elements from irrelevant stimulus items (irrelevant differentiation); 2) familiarization with the entire relevant stimulus items (relevant familiarization); 3) familiariza- tion with the entire irrelevant stimulus items (irrelevant familiariza- tion); and 4) no familiarization of any kind (control). All groups were first given a practice task. Following this session, the experimental treatments were administered individually. Immediately after this task, the same RA task was presented to all 53. The list contained eight dissyllabic pairs of low-high meaningfulness (M). All subjects were provided with strategy aids on one-half of the pairs. Learning and test trials were presented alternately, and a recognition method was used on the test trials. After the learning task, S was asked to describe how he learned each pair. This session was tape recorded. Then a final task was administered in which each stimulus item from the PA list was divided into three elements and listed adjacent to the eight response items from the RA task. The subjects had to select the response term that was associated with each element. No time limit was imposed on this task. The results demonstrated that: l) aid was effective (as shown in previous studies); 2) differentiation training appeared to facili- tate overall learning, whereas familiarization training appeared to inhibit learning; and 3) the results were inconclusive regarding the nature of the facilitative effect of differentiation training upon the successful employment of associative strategies in paired-associate learning. THE EFFECT OF FAMILIARIZATION WITH ELEMENTS OF COMPOUND STIMULUS ITEMS UPON THE SUCCESSFUL EMPLOYMENT OF ASSOCIATIVE STRATEGIES BY EDUCABLE RETARDATES By Daniel B. Berch A THESIS Submitted to 'Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF ARTS College of Education 1966 ACKNOWLEDGEMENTS The writer wishes to express his appreciation to the many people who have been instrumental in the development and completion of this study. I am sincerely grateful to my major adviser, Dr. Clessen J. ‘Martin, for his support and guidance throughout all phases of the study and for the opportunity to make use of project facilities. I am indebted to Dr. Rosaria Bulgarella for her valuable assistance in determining appropriate statistical analyses for the experiment. I am especially grateful to Cornelius vanderVeen who assisted in con- structing the materials and conducting the experiment. I also wish to thank R” Lawrence Hohn, who offered some valuable suggestions and criticisms. ii TABLE OF AMOWLEDGMNTS..... LIST OF TABLES . . o . ; . LIST OF FIGURES O O O 3 0 LIST or APPENDICES . . . '. INTRODUCTION... . . . . METHOD 0 q o '0 o o o o o 0 Subjects . . . . . 9 Materials . . o . o 0 Procedure . . . . . 9 mm D O O O O O O O 0 DISCUSSION AND CONCLUSIONS BIBLIOGRAPHY . . . . . . . APPENDICES O 0 O O 3 9 .0 ‘3 iii CONTENTS Page ii iv vi 11 ll 11 13 18 32 36 LIST OF TABLES Table Page 1 The Irrelevant Stimuli and Their Embedded words 0 O O O O O O O O O O 0 o O I 9 O O O O O 15 2 The A and B Pairs of the Criterion Task and the Strategy Aid Given for Each Pair . z . a 17 3 Means and Standard Deviations of the TOtal Number of Correct Responses on the PraCtiCETaSkoooooooooooocoo: 18 4 Summary of Analysis of variance for Total Number of Correct Responses on the criterionTaSkoooooooooaooooo 19 5 Summary of Analysis of variance for Number of Correct Responses on Aided Pairs . o . . o 20 6 Summary of Analysis of variance for Number of Correct Responses on Unaided Pairs . . . . 21 7 Means and Standard Deviations of the Number of Correct Responses for the Five Groups on the Aided Pairs of the Criterion Task. . o 22 8 Means and Standard Deviations of the Number of Correct Responses for the Five Groups on the Unaided Pairs of the Criterion Taakoooooooooaoooooooooo 22 9 Haans and Standard Deviations of the Total Number of Correct Responses for the Five Groups on the Criterion Task . . . . . . o . 23 10 Summary of Lindquist Type 1 Analysis of variance for the Five Groups on Total Number of Correct Responses . . . . . a . a . 25 11 Summary of Lindquist Type 1 Analysis of Variance for the Five Groups on Number of Correct Responses for Aided and Unaided Pairs o . . . o . o . . . o . . . . . . . . . 27 iv LIST OF FIGURES Figure Page 1 Mean Number TOtal Correct Responses on Criterion Task . . . . . . . . . . . . . . . 24 2 Mean Number Correct Responses on Aided and Unaided Pairs on Criterion Task . . . o . o 26 3 Mean Number Correct Responses on Aided and Unaided Pairs for the Five Groups . . . . . 3O Appendix A Means and Standard LIST OF APPENDICES the Five Groups " Means and Standard the Five Groups . Means and Standard Achievement Grade Five Groups . . . Instructions Instructions Instructions Instructions Instructions for the for the for the for Strategy Collection Task for the Differentiation Task 9 O O O I O 0 Practice Task Familiarization Task Criterion Task vi 0 Deviations of CA for Deviations of IQ for Deviations of Reading Equivalents for the 3 Page 39 4O 41 42 44 45 46 47 i. .3 ti! 111' . INTRODUCTION Investigators have recently begun to utilize the paired-associates (PA) task as a tool for examining the learning process among the mentally retarded. The majority of these studies have employed common pictures for RA materials rather than printed words. At least three of these studies, with matching on CA, have found no significant difference between normals and retardates (Akutagwa & Benoit, 1959; Berkson & Cantor, 1960; Eisman, 1958). In one study using common pictures, however, normals learned more rapidly than retardates (Ring & Palermo, 1961). Jensen (1965) attempted to induce verbal mediation in a PA task in order to determine if learning could be facilitated. Using common pictures, the experimenter (E) presented a phrase containing the stimulus and response items of the pair, e.g., "I threw a SHOE at the CLOCK." ‘With subjects matched on MA, mediators were provided for half of the normals and half of the retardates. The phrase for each pair was given only on the first presentation of the pair. The results showed that normals were superior to retardates under all conditions, and that the mediators facilitated learning in both groups. Davidson (1964), working with normal children, also attempted to induce verbal mediation in a PA task. Five experimental conditions were administered, with pictures used for RA materials. Three of these treatments consisted of the following: 1) pairs named by E and joined by one of seven prepositions -- over, under, onto, etc., 2) 2 pairs embedded in sentences employing the same prepositions used in 1), and 3) pairs named in the sentences used in 2) and in addition, pictures of the pairs as described by the sentences. «The results showed that links formed by a single preposition were as effective in facilitating learning as links consisting of sentences and of pictures describing these sentences. ZMartin, Boersma, and Cox (1965) conducted a study with college students in order to examine in detail what kinds of mediators or associative cues are reported in PA learning. From verbal reports collected at the conclusion of the learning task, they developed a strategy classification scheme consisting of seven categories. These categories were rank ordered along an apparent underlying continuum of cue complexity. Analysis of the data revealed an increasing relation- ship between complexity of the associative strategy reported and performance on the learning task. In another study conducted by Martin, Boersma, and Bulgarella (1966), the types of associative strategies reported by mildly retarded children were compared to those reported by normal children matched on CA. The results showed that the normals’ performance on the RA task was significantly better than that of the retardates. Analysis of the Ss' verbal reports indicated that the normals reported using more high level strategies, whereas the retardates reported more low level ones. iMartin, Cox, and Bulgarella (1966) devised a third study to determine in part if high level strategies given to normal children by 3 the E would facilitate their learning of a PA task. Materials con- sisted of dissyllabic pairs of low-high meaningfulness OM). The findings clearly showed that strategy aids resulted in better perfor- mance. A further study by Martin (1966) was designed to examine whether the performance of educable retarded children on a PA task could be facilitated by providing them with high level strategy aids. The materials consisted of eight low-high M dissyllabic pairs. It was found that retarded Ss who received these aids learned significantly faster than unaided retardates and performed as well as unaided normals matched on CA. However, since the aided retardates did not perform significantly better than the unaided normals, Martin con- cluded that the retarded Ss did not make optimal use of the strategy aids. I In the Martin (1966) study, a stimulus differentiation task had been devised to distinguish aided Ss for whom aid was effective from those for whom it was not. The stimulus items were divided into three. segments which appeared with their respective response words. One segment of each stimulus was the element which had been combined with the response to form the strategy. The results indicated that effec- tively aided 38 (those whose acquisition scores fell within the top third of the range of all scores) were able to differentiate relevant cues significantly better than 83 not effectively aided. It was concluded that the ability to differentiate visually the relevant 4 stimulus cue was an important factor in effective utilization of strategy aids. The purpose of the present study was to determine if retarded Ss given prior familiarization training with relevant stimulus cues could make greater use of strategy aids than Ss not given such training. Studies which have examined the effects of familiarization of the stimulus members of paired-associates upon acquisition have revealed conflicting results (Goss, 1965). Cannon and Noble (1961) conducted a study to investigate the hypothesis that familiarization is the sole basis of meaningfulness. Subjects from one of the treatment groups were presented with 20 independent, 1 sec. exposures of low M dissyllables which were sub- sequently used as stimulus members in a paired-associates list. During the familiarization trials, Ss pronounced each item as it was exposed. The results showed that familiarization with the stimulus items facilitated acquisition of the RA list. In offering one interpretation of the results, the authors made reference to the concept of stimulus predifferentiation. They did not elaborate, however, on how it may have functioned in facilitating learning. 0n the basis of results from previous studies investigating stimulus predifferentiation, it is probably safe to assume that the authors believed familiarization resulted in an increase in the dis- tinctiveness of the stimulus items. This distinctiveness reduced the probability of these stimuli being associated with incorrect response items. 5 Underwood and Schulz (1960) designed an experiment in which they tested the hypothesis that increasing amounts of familiarization with the stimulus units should increase the rate of PA learning. The stimulus and response items consisted of syllables of low-low M. Subjects received 1, 10, 20, or 40 familiarization trials at an ex- posure rate of 2 sec. per syllable, and Ss spelled each item aloud as it was presented. The findings revealed that familiarization of stimulus members consisting of less than 40 trials had no effect upon acquisition of the PA list, while 40 trials appeared to inhibit perfor- mance. Underwood and Schulz concluded that inhibition may have resulted from a loss of differentiation between stimulus and response systems during the learning of the RA list. This loss could have occurred because during familiarization training, 8 became accustomed to giving ' Then as S learned the the stimulus items of the list as "responses.' response items of the PA list and the strength of these items approached that of the stimuli as "responses," he may have confused the actual stimulus and response syllables of the list. In order to test this loss-of—differentiation hypothesis, a second experiment was designed which employed essentially the same procedure used in the first experiment. TWO groups received equal amounts of familiarization training with the same low M syllables, and these syllables were used as stimulus items in the PA task. One group's PA list also used low M syllables for response items, while the other group’s list used low M dissyllables for response items. It was 6 reasoned that since the stimulus and response members of the syllable- dissyllable list belonged to different classes of verbal materials, the possibility for loss of differentiation would be minimized. Thus, the learning of this list had to be faster than the learning of the syllable-syllable list in order to support the hypothesis. ApprOpriate groups were employed to control for the different response terms. The results showed a nonsignificant difference between the experimental groups, thus leading to rejection of the hypothesis. There was neither facilitation nor inhibition following relevant stimulus familiarization. Kanungo and Lambert (1963) also reasoned that during stimulus familiarization training, 83 become accustomed to giving the stimulus items of a PA list as "responses." They stated that familiarization of stimulus members "should promote the formation of word-word habits, i.e. positive reaction tendencies to connect verbal elements with themselves." From conventional transfer paradigms they derived the hypothesis that this treatment would result in negative transfer as compared with a control group receiving familiarization of words which were not used in the PA task. Low M dissyllables were used for both stimulus and response items. During familiarization training the exposure rate was 1 sec. per item. The subject was asked to repeat each word aloud continuously for 15 sec. at a rate of three to four repetitions per sec. Immediately after the repetitions, 3 made his rating for that word on semantic differential scales. This procedure was repeated three times for each 7 of the stimulus words and once for each semantic scale. The S was then presented with the RA list. Results showed that the group receiving familiarization with actual stimulus items performed signif- icantly lower than the group familiarized with irrelevant words. The authors concluded that the explanation of this inhibitory effect could not be attributed to a decrement in meaning of the low M stimuli, because the familiarization treatment did not produce any significant decrement in meaning as measured by the semantic differential. This conclusion was supported in a study by Schulz and Thysell (1965). Employing both low and high M dissyllables, they found that neither 0, 5, nor 20 l-sec. familiarization trials had any effect upon the number of kinds of associations given during the production test. They concluded that presumed changes in the number or nature of the associations elicited by familiarized items cannot serve as the basis for an explanation of the effects of familiarization with these items upon subsequent performance in a PA task. By familiarizing S with an element of a stimulus, this element should actually become the functional stimulus (Underwood, Ham & Ekstrand, 1962). In discussing the concept of cue selection, Underwood et al. stated that there may be a discrepancy between the nominal stimulus (the stimulus actually presented to S) and the functional stimulus ( the component of the nominal stimulus which becomes the effective cue for response elicitation). Using word-color and tri- gram-color compounds for stimulus materials, they hypothesized that given two components of different classes as the nominal stimulus, the 8 more meaningful will become the functional stimulus. The results showed that for the trigram—color compound, color, the more meaningful component, became the effective (functional) stimulus on a transfer test. For the word-color compound, words became the functional stimuli on the transfer test. The authors concluded, however, that the latter finding may have been due to a bias which 38 have toward dealing with verbal material rather than the higher meaningfulness of the words. Cohen and Musgrave (1964) designed a PA study to investigate cue selection where the compound stimulus terms each consisted of two CVC (consonant, vowel, consonant) trigrams of known M value, and the response units were single letters. Four treatment groups were formed, each receiving one of the following four combinations of‘M values (H = high H, L = low M): HH, HL, LH, and LL. On the transfer task the component CVCs of the compound stimuli were presented singly. The findings showed significantly better performance to high M than to low M CVCs. Although the main effect of position was non- significant, performance to low M CVCs previously given in the left- hand position in the compound stimulus was significantly greater than to high M CVCs in that position for both the LL and LH groups. It was concluded that in PA learning, where the stimulus term consists of two verbal components, cue selection is a function of meaningfulness of the individual component. In addition, the position of the individual component is an important variable if the component is of low'M. 9 For the present study, it was assumed that in order to make optimal use of a syntactical strategy aid such as "map of the village" in learning a dissyllabic pair of low-high M (Zumap-Village), one must first be able to differentiate the more meaningful component "map." "Map" is the functional stimulus for optimal utilization of the associative strategy. It is hypothesized, therefore, that l) familiarization with the relevant meaningful stimulus element in a syntactical strategy will lead to maximally effective use of the strategy aid as measured by performance on the criterion task. This treatment (Relevant Cue Differentiation) will result in significantly better performance compared to a control condition in which strategy aids are given but not familiarization training. On the basis of Kanungo and Lambert's study, it is hypothesized that 2) familiarization of the entire stimulus item (Relevant Familiar- ization) will interfere with differentiation of the stimulus components, resulting in the ineffective use of strategy aids and hence signifi- cantly poorer performance than the control condition. A fourth group receiving strategy aids will be given familiariza- tion training with the meaningful elements of the stimuli which will not be used in the criterion task (Irrelevant Cue Differentiation). This group is included in the study in order to determine whether there is some general transfer involved in selecting high M components from irrelevant stimuli. If this treatment condition does produce 10 some general transfer, then performance on the criterion task ought to be facilitated when compared to the control group. A fifth group receiving strategy aids will also be given familiar- ization training with the irrelevant stimulus items. These Ss will be familiarized with the entire stimulus item (Irrelevant Familiarization). This group is included in the study to control for the effect of famil- iarization with relevant items. If irrelevant familiarization produces a general set to connect each stimulus item with itself, performance on the criterion task should be inhibited when compared to the control group. METHOD Subjects The 88 tested in this study were 80 educable retarded children selected from Junior Special B classes in the Detroit public schools. The criteria for initial assignment to these special classes were CA, 11-14, and IQ, 56—75. Mban CA of the Ss used in this study was 13-7 (range 11-4 to 15-5) and mean IQ was 71 (range 53-87). In order to ensure Ss' ability to read the items, only children who had received a grade equivalent of 2.5 or higher on the reading subtest of either the Metropolitan, Iowa, or Stanford achievement tests were selected for the experiment. (See appendices for further descrip- tion of 33.) Materials For all Ss on the practice task, four dissyllabic pairs of low- high leere utilized. These pairs were: Lemur-Kitchen, Bodkin-wagon, Holbut-Farmer, and Olpret-Balloon. The first two pairs consisted of items selected from Noble's (1952) list. Mean m values of the stimu- ulus and response terms were 1.84 and 8.87 respectively. Constructed specifically for this task, the stimulus items of the last two pairs were designed to approximate Noble's low E paralogs. The high M response items were taken from second-grade readers. For the criterion task, eight dissyllabic pairs of low-high M were constructed. These pairs were: Gokem-Uncle, Sagrole4Money, Tarop-Jelly, Zumap-Village, FlotsamaArmy, Meardon-Insect, Binest- 11 12 Outside, and Lenear-Garden. The first six pairs were selected from Noble's (1952) list, with mean 3 values of the stimulus and response items, 1.39 (range 1.05 to 2.19) and 7.89 (range 6.57 to 9.43) respectively. The last two pairs in this list were also devised specifically for this task. The stimulus items were designed to approximate Noble's low E dissyllables, and the response items were selected from second-grade readers. Each stimulus selected for the criterion list contained a familiar word, thus making the pairs easily amenable to the construction of syntactical strategies. A third list consisting of eight irrelevant stimuli was formed. These low M dissyllables were: Attar, Byssus, Delpin, Sumpage, Endore, Fardel, Standage, and Caratch. The first three items were selected from Noble's (1952) list. Mean 2 value was 1.48 (range 1.13 to 1.71). The last five items were selected from Cieutat's (1963) list and had a mean association value (a) of .64 (range .49 to .77). Each of these eight paralogs contains a familiar word, as do the stimulus items on the criterion list. All 16 stimulus items were selected because they possessed this attribute. Even though the stimuli were taken from two different sources, it was assumed that for retardates, differences in meaningfulness between the low M items from the two different lists would be negligible. Separate test booklets were constructed for the practice task and the criterion task. They contained all the response items of the task on each page in a randomized order. Separate 2 x 2 inch slides were 13 prepared for the S-R pairs and for the stimulus items. A Kodak 700 Carousel projector with a Lafayette TLZK.automatic timer was utilized for slide presentation. A tape recorder was used to collect Ss' verbal reports at the conclusion of the criterion task. Procedure PRACTICE TASKi All Ss were given the practice task in order to acquaint them with the paired-associate learning task and to enable assessment of the initial comparability of the treatment groups. The practice task was administered to groups of two to six Ss, depending upon the number of Ss that could be tested completely on the same day. The task was introduced to the 89 as a word game in which they were instructed to learn four associations. Three learning (L) and three test (T) trials were given. A test trial was presented after each learning trial (LTLTLT). During the learning trials, each pair was presented automatically at a six sec. exposure rate with an interval of approximately .75 sec. between pairs. The exposure rate on the test trials was controlled manually so that 33 had as much time as they needed to respond. On all learning and test trials, E pronounced the items as they appeared on the screen. Slides for each learning and test trial were presented randomly. The remain- ing portions of the experiment were administered individually. FAMILIARIZATION TASK. Each S was first given examples of all levels of associative strategies in order to facilitate collection of strategy infbrmation after completion of the criterion task. Following 14 discussion of associative strategies, each S received one of five familiarization treatments prior to the criterion task. The five experimental treatments were: 1) Relevant Cue Differentiation (RD) -- Each S was given an example of a paralog and was then shown that a familiar word was embedded in it. S was then given a relevant differentiation trial in which each of the stimulus items from the criterion list was projected on the screen. E pronounced each stimulus as it appeared on the screen and then pronounced the word embedded in it, e.g., Zumap-map. The embedded word of each stimulus was the first word of the syntactical strategy given for the pair (ZumaprVillage) on the criterion task, e.g., map- map of the village. After the relevant differentiation trial, S was given a pronunciation trial in which E again pronounced each stimulus but this time S had to pronounce the embedded words. A11 88 received two relevant differentiation trials and two pronounciation trials with a six sec. exposure rate for each item on all four trial. The same exposure rate was employed for all treatment conditions. 2) Irrelevant Cue Differentiation (ID) -- This group received the same instructions and the same treatment as the RD group. However, the list used in this treatment contained irrelevant stimuli. None of the embedded words in the irrelevant stimuli were contained in the syntac- tical strategies employed in the criterion task. Table 1 presents the irrelevant stimuli and the embedded words. 15 Table l The Irrelevant Stimuli and Their Embedded Words Irrelevant Stimuli Embedded Words attar at sumpage page delpin pin endore end standage stand byssus by caratch car fardel far :3) Relevant Familiarization (RF) -- These 38 were given two relevant faunliarization trials alternated with a pronunciation trial on the :stimmli from the criterion list. The embedded word was neither pro- ruounced nor pointed out to the S. 0n the relevant familiarization trrials, E simply pronounced each stimulus and on the pronunciation tirials S had to pronounce each stimulus after E. ‘4) Irrelevant Familiarization (IF) -- This group received the same 7 treatment as the RF group but with the irrelevant stimuli. '5) Control (C) -- This group received no familiarization of any type. 16 The 88 were randomly assigned to these five conditions. CRITERION TASK. Following the experimental treatment, the same task was administered to all Ss. A total of five learning and five test trials were presented. The learning and test trials were alter- nated. A six sec. exposure rate was used for each learning trial. No time limit was imposed on the test trials. All Ss were instructed that E would give them some associative strategies, and that they should try to use these strategies to help them learn which words went together. One-half of each treatment group received syntactical strategies on four of the eight pairs (A pairs), and the other half of each group received the same type of aids on the other four pairs (B pairs). Aids were given on the first three trials only. The A and B pairs and their respective strategy aids are presented in Table 2. ASSOCIATIVE STRATEGY TASK. Immediately following the criterion task, each S was again shown the eight criterion pairs separately and was asked to describe how he learned each pair. This session was tape recorded. DIFFERENTIATION TASK. After strategy collection, each S was given a sheet of paper containing two columns. The right-hand column con- sisted of all eight response items, and the left-hand column consisted of 24 elements contained in the stimulus items from the criterion list. Three elements from each stimulus were selected, one of which consisted of the embedded word used in the syntactical strategy, e.g., Sagrole- 17 Table 2 The A and B pairs of the Criterion Task and the Strategy Aid Given for Each Pair Word-Pairs Strategy Aids A Gokem-Uncle Go to uncle Pairs Sagrole-Money Roll of Money Binest-Outside Nest is outside Tarop-Jelly Tar is like jelly B Lenear-Garden Near the garden Pairs Zumap-Village Map of the village FlotsamaArmy Sam is in the army Meardon-Insect Don's insect sag, gro, role. The 24 elements were arranged so that no three elements of a stimulus appeared successively. S was instructed to select the word on the right which was associated with each element on the left. No time limit was imposed on this task. RESULTS The five treatment groups did not differ significantly at the .05 level on CA, IQ, or reading achievement. In order to determine the initial comparability of the five groups on a paired-associates task, the practice task data were subjected to a 1 x 5 analysis of variance. The analysis yielded a nonsignificant F ratio (.23, df = 4/75, p:>.25) indicating that the groups did not differ significantly prior to the introduction of the experimental treatments. The means and standard deviations of the total number of correct responses for the five groups on the practice task are presented in Table 3. Table 3 Means and Standard Deviations of the Total Number of Correct Responses on the Practice Task Groups RD ID RF IF C Mean 5.88 5.81 5.63 5.38 5.25 S.D. 2.63 1-60 2.28 2.33 2.27 The criterion task data of all groups except control were sub- jected to three 2 x 2 x 2 factorial analyses of variance. The factors were: 1) Stimulus -- Relevant, Irrelevant; 2) Type of familiariza- tion -- Differentiation, Familiarization of nominal stimulus; 3) Aided 18 19 pairs -- A, B. A separate analysis was carried out on the total number correct, number correct on aided pairs, and number correct on unaided pairs. The results from the analysis of total number correct are summarized in Table 4. The main effect of type of familiariza- tion was significant beyond the .05 level indicating that differentia— tion was superior to familiarization of the entire word. Table 4 Summary of Analysis of variance for Total Number of Correct Responses on the Criterion Task Source df MS F A: Stimulus 1 6.25 B: Type of Familiarization 1 451.57 6.62* C: Aided pairs 1 105.07 1.54 A x B: Stim x Type Famil. 1 56.25 .___. A x C: Stim. x Aided pairs 1 240.25 3.52 B x C: Type Famil. x Aided pairs 1 1.56 .____ A x B x C: Stim. x Type Famil. x Aided pairs 1 20.24 ____. Error: ‘Within treatments __6 68.21 _____ Total 63 * Significant beyond the .05 level The results from the analysis of number correct on aided pairs are summarized in Table 5. Familiarization of the nominal stimulus was again shown to be significantly inferior to differentiation. The nonsignificant main effect of aided pairs suggested that if strategy aids did facilitate learning, they were equally effective for both sets of aided pairs. 20 Table 5 Summary of Analysis of variance for Number of Correct Responses on Aided Pairs Source df . MS F A: Stimulus 1 .25 ___. B: Type of Familiarization 1 105.07 5.38* C: Aided pairs 1 56.25 2.88 A x B: Stim. x Type Famil. l 30.25 1.55 A x C: Stim. x Aided pairs 1 39.06 2.00 B x C: Type Famil. x Aided pairs 1 .77 ___. A x B x C: Stim. x Type Famil. x Aided pairs 1 2.54 ___. Error: Within treatments ,_6 19.52 .___ Total 0‘ L0 * Significant beyond the .05 level The results from the analysis of number of correct responses on the unaided pairs are summarized in Table 6. 0n the unaided pairs also, differentiation was significantly better than familiarization of the nominal stimulus. The nonsignificant main effect of aided pairs indicated that the two sets of unaided pairs were of equal difficulty. Further analysis of the difficulty level of the A and B pairs was performed for the control group with the use of a g-test. Half the Ss were aided on the A pairs and half on the B pairs. The resulting g value comparing the mean number of total correct responses did not approach significance at the .05 level. This finding permitted the pooling within each treatment group of Ss aided on A and B pairs. 21 Table 6 Summary of Analysis of variance for Number of Correct Responses on Unaided Pairs fi V V v v 7 fi —v— Source df MS F A: Stimulus l 9.00 .___ B: Type of Familiarization 1 121.00 5.66* C: Aided pairs 1 7.56 A x B: Stim. x Type Famil. 1 4.00 ____. A x C: Stim. x Aided pairs 1 85.56 4.00 B x C: Type Famil. x Aided pairs 1 3.06 .____ A x B x C: Stim. x Type Famil. x Aided pairs 1 7.57 -———— Error: Within treatments ,_§ 21.39 .____ Total 63 * Significant beyond the .05 level The criterion data were then subjected to three 1 x 5 analyses of variance. The means and standard deviations of the number of correct responses for the groups on the aided pairs are presented in Table 7. The analysis of these data revealed a nonsignificant F ratio (1.91, df = 4/75, p>.05). Table 8 presents the means and standard deviations of number correct on the unaided pairs. The analysis of these data yielded a nonsignificant F ratio (1.73, df = 4/75, p).05). The means and standard deviations of total number correct are presented in Table 9. The analysis of total number correct revealed a nonsignificant F (2.05, df = 4/75, p>.05); however, the means of the groups are in the expected direction. Compared to the performance 22 Table 7 Means and Standard Deviations of the Number of Correct Responses for the Five Groups on the Aided Pairs of the Criterion Task RD ID RF IF C Mean 16.44 14.94 12.50 13.75 15.31 S.D. 3.37 4.06 5.33 4.82 4.01 w—-v wv- v w —v W fi v—w Table 8 Means and Standard Deviations of the Number of Correct Responses for the Five Groups on the Unaided Pairs of the Criterion Task RD ID RF IF C Mean 13.56 13.81 10.31 11.56 11.69 S.D. 4.46 4.04 5.35 4.69 3.70 of the control group, RD is the highest group and RF the lowest. In order to determine if the groups differed in rate of learning, an analysis of variance, Lindquist Type 1 design, was computed 23 (Lindquist, 1953). The results are summarized in Table 10. "The main effect of trials was, of course, significant and is of little interest. However, the significant interaction term indicated that the perfor- mance of all groups did not increase at the same rate. Performance Table 9 Means and Standard Deviations of the Tbtal Number of Correct Responses for the Five Groups on the Criterion Task RD ID RF IF C ——fi fifi Mean 30.00 28.75 22.81 25.31 27.00 S.D. 6.65 7.75 10.00 8.65 5.91 curves, presented in Figure 1, show that the RD group was near asymp- tote by the third trial. This suggested a ceiling effect which may have minimized differences among groups. To investigate this hypothesis of a ceiling effect, the total number of correct responses on the first three trials were subjected to a simple 1 x 5 analysis of variance. This analysis yielded an F value of 2.46 which approached significance at the .05 level (critical value 8 2.49). Thus it appears that differences among groups were minimized by a ceiling effect. In order to determine whether aid was effective, the criterion 24 .vflmmu EOHHUUWHU 50 WGQCOQWMH UUQHHOU HNUOH .HUQESC 5mm: .H .th may; m a m N H _ a _ a e U ml I .II .0 .5 Q I I I IO \.0 mm “1 O \ QH ‘ 11111 ‘ ...\ II a ‘ - ..\_. “WV \. \\\\ \ \ \ X 1 m I \ Q. \ \ \x\\ \ \ \ \ \\ \ \ \ \ \\\ 1 IO \ \Q \ \ \ \ \O \ \ \ \ \\.. Q I. II .I IO\\\ \ \ .\.\ O y 11111 '01 ..\ |\ 1‘ ||\ ' \ \X. 1 J SHSNOJS‘HH JDSHHOO NVEH MICHIGAN suns unwmm INSTRUCTIONAL MATERIALS cm {Es marmnm 25 Table 10 Summary of Lindquist Type 1 Analysis of variance for the Five Groups on TOtal Number of Correct Responses Source df MS F A: Groups 4 25.76 2.05 Error (a) 75 12.56 ___. B: Trials 4 126.48 1149.82*** A x B: Groups x Trials 16 32.36 294.18*** Error (b) 300 .11 Total 399 ~—v fifi fi fi *** Significant beyond the .001 level task data were subjected to another Lindquist Type 1 analysis of variance. The factors were: 1) Groups -- RD, ID, RF, IF, C; 2) Pairs -- Aided, Unaided. The results are summarized in Table 11. The significant main effect of pairs indicated that aid was effective. The interaction term, however, was also singificant. A graphic rep- resentation of the interaction is presented in Figure 2. It can be observed from Figure 2 that although the ID group was the third highest on the aided pairs, this group performed better than the other groups on the unaided pairs. To investigate further this finding, learning curves were plotted for each group comparing the levels of performance on aided and un- aided pairs. These curves are presented in Figure 3. Of particular 26 SSSNOdSBH 1038803 NVBN £3 53 EIEIID Iii UNAIDED AIDED and unaided pairs on on aided 011868 Fig. 2. Mean number corre task. criterion 27 Table 11 Summary of Lindquist Type 1 Analysis of variance for the Five Groups on Number of Correct Responses for Aided and Unaided Pairs Source df MB F A: Groups 4 72.20 2r33 Error (a) 75 30.99 B: Pairs 1 230.39 44.74*** A x B: Groups x pairs 4 56.67 ll.00*** Error (b) _15 5.15 Total 159 *** Significant beyond the .001 level interest is the fact that only the ID group showed approximately the same level of performance on the aided and unaided pairs. This finding suggests that the irrelevant differentiation training may have resulted in general transfer to the stimuli of the unaided pairs, thus aiding the formation of high level strategies for use in learning these associations. To assess the reliability of two judges' independent ratings of the verbal reports, a Spearman rank order correlation coefficient was computed for each pair from 40 of the Ss. The coefficients ranged from .89 to .99 indicating high agreement between the judges. A total strategy score was then computed from each S's verbal report for the unaided pairs. This score was obtained for an S by summing the 28 strategy ratings assigned to each of the four unaided pairs. For example, if an S reported using four 7 level (syntactical) strategies, his total strategy score would be 28. In order to determine the relationship between performance on the unaided pairs and the complexity of strategies reported for these pairs, three Spearman rank order correlation coefficients were com— puted between number of correct responses on the unaided pairs and total strategy score for these pairs. Because the analysis of variance for number correct on the unaided pairs revealed a significant difference between the combined differentiation groups and the com- bined familiarization groups, a separate coefficient was computed for each of these combined groups and the third computed for the control group. The coefficient for the differentiation groups was .14; for the familiarization groups, .34; and for the control group, .53. The latter two were significant beyond the .05 level indicating a signif- icant positive relationship between complexity of reported strategy levels and performance on the unaided pairs. ‘When differentiation training was compared with familiarization training, the results based upon total number of correct responses, number correct on aided pairs, and number correct on unaided pairs all revealed that differentiation training resulted in significantly better performance than familiarization training. Because the control con- dition was excluded from these analyses, three separate 1 x 3 analy- ses of variance comparing the control group, the combined differentia- 29 tion groups, and the combined familiarization groups were performed. The F values for total number correct and number correct on unaided pairs were significant beyond the .05 level. The F value for the aided pairs approached significance at the .05 level. Duncan's multiple range test (Winer, 1962) was used to investi- gate further the significant findings obtained for two of the three dependent variables. The results of both analyses indicated that the differentiation and familiarization treatments differed significantly from each other (p<:.05), but that neither treatment differed signif- icantly from the control condition. To investigate further the significant difference between the combined differentiation and familiarization groups on total number correct and number correct on the unaided pairs, the data from the differentiation task were subjected to two factorial 2 x 2 x 2 analyses of variance. The factors were: 1) Stimulus -- Relevant, Irrelevant; 2) Type of familiarization -- Differentiation, Familiarization of nominal stimulus; 3) Aided pairs -- A, B. One analysis was computed for number of correct responses on the elements of all pairs, and the other was.computed for number correct on elements of the unaided pairs. It was hypothesized that if Ss from the differentiation groups success- fully selected out embedded words which were used as cues for syntacti- cal strategies, then they should have performed significantly better than the familiarization groups on the differentiation task. Both analyses revealed no significant difference between the 30 d .4 anHaw msu wow enema confine: non woven to noncommwu uoouuoo Hones: :wwz .m .mwm mA¢HMH QMQH