7ivioibg‘7 IllllllllVIIAl'iHllmIlllfi’lEIRH'IHYIIiIIMiHIIHH 3 1293 00548 0342 LIBRARY Michigan State University This is to certify that the dissertation entitled VERBAL ENCODING AND THE SYNTAGMATIC-PARADIGMATIC SHIFT presented by Kathleen Rose Johnson has been accepted towards fulfillment of the requirements for Doctor of degree m Audiology and Phllosophy Speech Selences /,< ////J // ’ Major professor/ l ' 'I‘ ‘ ‘ \, (g V’trfr *‘r / Dat€; )“L/t 2 3i ///’:,/(/ /\ MS U is an Affirmative Action/Equal Opportunity Institution 0- 12771 )V1ESI_J RETURNING MATERIALS: Piace in book drop to ngRARJES remove this checkout from w your record. FINES wiii r be charged if book is returned after the date stamped below. VERBAL ENCODING AND THE SYNTAGMATIC-PARADIGMATIC SHIFT BY Kathleen Rose Johnson A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Audiology and Speech Sciences 1989 J7 m r\ W‘) ABSTRACT VERBAL ENCODING AND THE SYNTAGMATIC-PARADIGMATIC SHIFT BY Kathleen Rose Johnson The purpose of this study was to determine whether first, second, and fifth grade children encode syntagmatic, paradigmatic, and evaluative attributes of orally presented word relationships and whether encoding differences exist on the basis of children's word association styles. Sixty-four children were given the Word Association Task. Subject groups were formed by grade and association style (i.e., syntagmatic or paradigmatic). A short-term memory paradigm (Wickens,1970) was used in three tasks (Syntamatic, Paradigmatic, Evaluative) as indices of encoding. Tasks consisted of a control and experimental condition consisting of four trials comprising three words which were remembered and recalled subsequent to a distractor task. Trial level data for each task (Syntagmatic, Paradigmatic, Evaluative) was used in ANOVA and post-hoc tests to determine differences in Groups, Tasks, Task Conditions, and Trials. Differences were not related to word association response preference of subjects. Fifth graders demonstrated better recall across trials. Proactive inhibition and release from proactive inhibition were derived for each subject on each task. Kathleen Rose Johnson Group means were analyzed by ANOVA and post-hoc tests to determine differences. Children did not consistently demonstrate proactive inhibition in tasks. Therefore, release from proactive inhibition was small and did not differ for groups. Large variances within trial level data and in performance across trials accounts for this outcome. AN ad-hoc analysis of the percent of subjects in each group demonstrating the proactive inhibition effect for each task revealed small percentages. This outcome provided support and insight into the lack of group differences. An ad-hoc analysis was performed on trial level error patterns (Substitution, Omission, Interference, Order) on the three tasks. ANOVA revealed differences in the frequency of error types as well as group level differences. The four-level analysis in this study provides new information on the efficacy of using this paradigm to determine encoding of attributes by children. Error pattern analysis of trial level data provided useful descriptive information as to the nature of children's short-term memory for stimuli. Copyright by Kathleen Rose Johnson 1989 ACKNOWLEDGEMENTS Appreciation is extended to the members of my dissertation guidance committee: Leo V. Deal, Michael R. Casby, Lois A. Bader, and Nickola w. Nelson. Special gratitude is owed to Leo V. Deal, who served as academic and dissertation advisor, and who continuously provided support and encouragement throughout my program. Special gratitude is owed to the first, second, and fifth graders who participated in this study. A sincere thank you is given to the school principles, Mr. James Hengstebeck of Elmhurst School, and Mr. Dennis Semrau of Lewton School and the first, second, and fifth grade teachers for allowing this work to be conducted in their schools. My thanks is extended to the Lansing School District Research Committee for its warmth, support, and selection of the schools in which this project was carried out. Thanks go to my colleagues at Michigan State University for their patience and understanding. iii The members of my family deserve the most thanks. My husband, Mark, provided me with love, support, encourage- ment, understanding, and patience throughout my entire program, and particularly during the completion of this project. My parents, Rosemary and Ernest Balchak, gave me the attributes of optimism, positive self-esteem, independence, and perseverance. They also gave unlimited love, support, and encouragement throughout my endeavors. iv TABLE OF CONTENTS CHAPTER PAGE LIST OF TABLES X LIST OF FIGURES XV I LITERATURE REVIEW 1 Introduction 1 The Syntagmatic-Paradigmatic Shift 1 Variables that affect word associa- tions 2 Age, Form Class, and Word Frequency 2 Gender, IQ, and SES 8 Theories of the Syntagmatic-Paradigmatic Shift 11 Associative Theories 11 Linguistic and Feature-Matching Theory 13 Theories of Associative Structure 15 Cognitive Development Theories 16 Models of Memory 25 Verbal Encoding 28 CHAPTER Semantic Encoding Dimensions Developmental Characteristics of Encoding Summary of Literature Review II METHOD Subjects Word Association Pretest Letter Identification Pretest Subject Age Subject Race and Gender Procedures Proactive Inhibition Procedure Proactive Inhibition Tasks Evaluative Encoding Task Syntagmatic Encoding Task Paradigmatic Encoding Task Reliability of Task Administration and Scoring III RESULTS Management of Encoding Data Percent-Correct Recall Derived Dependent Variables: PI AND PRI Encoding Task Outcomes Evaluative Encoding Task Percent-Correct Recall Derived Scores: PI and PRI vi PAGE 30 35 44 47 47 50 54 55 56 57 57 64 64 67 69 71 72 72 73 76 76 76 81 CHAPTER Statistical Analysis Syntagmatic Encoding Task Percent-Correct Recall Derived Scores: PI and PRI Statistical Analysis Paradigmatic Encoding Task Percent-Correct Recall Derived Scores: PI and PRI Statistical Analysis Item Error Analysis Within Tasks Description of Error Types Reliability of Error Coding Evaluative Encoding Error Patterns Statistical Analysis Syntagmatic Encoding Error Patterns Statistical Analysis Paradigmatic Encoding Error Patterns Statistical Analysis IV DISCUSSION Encoding Task Outcomes Evaluative Encoding Syntagmatic Encoding Paradigmatic Encoding Overall Encoding Task Findings and vii PAGE 81 9O 90 94 96 105 105 109 111 116 116 119 119 122 124 129 129 135 139 139 140 145 148 CHAPTER Implications V SUMMARY AND CONCLUSIONS Introduction Background Purpose Experimental Design Subjects Stimuli Evaluative Encoding stimuli Syntagmatic Encoding stimuli Paradigmatic Encoding stimuli Procedures Grade level performance Rapid Naming Word Association Task Dependent Variables Findings Conclusions APPENDIX A WORD LIST FOR WORD ASSOCIATION TASK B DECISION RULES FOR CODING WORD ASSOCIATION TASK RESPONSES C SEMANTIC DIFFERENTIAL RATINGS OF STIMULI D SETS OF TRIADS FOR EVALUATIVE ENCODING TASK E SETS OF TRIADS FOR SYNTAGMATIC ENCODING TASK viii PAGE 149 156 156 156 157 158 158 158 159 160 160 160 161 161 161 162 162 164 166 167 168 170 171 CHAPTER PAGE F SETS OF TRIADS FOR PARADIGMATIC ENCODING TASK 172 G SCORING OF ITEMS AND TRIALS FOR ENCODING TASKS 173 H SUBSTITUTION ERRORS MADE BY SUBJECT GROUPS ON THE EVALUATIVE ENCODING TASK 176 I SUBSTITUTION ERRORS MADE BY GROUPS ON THE SYNTAGMATIC AND PARADIGMATIC ENCODING TASKS 178 J STUDENT TASK SUMMARY FORM 180 K PERCENT OF SUBJECTS IN GROUPS DEMONSTRATING THE PROACTIVE INHIBITION EFFECT 182 L INFORMED CONSENT 186 REFERENCES 187 ix TABLE 10 LIST OF TABLES Mbdified summary of the percentage of paradigmatic and syntagmatic responses by subject grade and form class for stimulus words and associates (EnUwisle, 1966). Association responses by form class relative to grade level and frequency of occurrance (Entwisle, 1966). Summary of age, gender, and racial characteristics of 64 children classified in four groups. Each group contained 16 children. The group means and ranges of the percent of paradigmatic responses on the werd Association Task. Evaluative Encoding Task: Percent means and standard deviations of correct recall across Trials (1-4) within Conditions (Control and Experimental) for subject Groups (1-4). Evaluative Encoding Task: RAST means and standard deviations of correct recall across Trials (1-4) within Conditions (Control and Experimental) for subject Groups (1-4). Evaluative Encoding Task Derived Scores: RAST Proactive Inhibition (RAST PI), Percent Release from Inhibition (PRI), and RAST Release from Inhibition (RPI) means and standard deviations fer subject Groups (1-4). Results of a three-way analysis of variance as a function of Group (1-4), Condition (Control and Experimental) and Trial (1-4) for the Evaluative Encoding Task. Results of the simple main effects test for Group (1-4) on Condition (Control and Experimental) in the Evaluative Encoding Task. Results of the Neuman-Keuls specific comparison test on pairs of Trial means (1, 2, 3, 4) in the Evaluative Encoding Task. .A difference between any two means is significant when it exceeds the appropriate critical value (CV) for «=0.05. PAGE 49 53 77 78 82 84 85 86 TABLE PAGE 11 Results of a one-way analysis of variance as a 88 function of Group (1-4) for the RAST Release from Proactive Inhibition on the Evaluative Encoding Task. 12 Results of an ad-hoc analysis of the percent of 89 subjects in each group who demonstrated the Proactive Inhibition Task Effect (PIE): Proactive Inhibition for Control and Experimental Conditions (PI-C, PI-E), Release from Proactive Inhibition (RPI) and the Proactive Inhibition Effect (PIE) (If PI-C, PI-E, and RPI occur). 13 Syntagmatic Encoding Task: Percent means and standard 91 deviations of correct recall across Trials (1-4) within Conditions (Control and Experimental) for subjects in Groups (1-4). 14 Syntagmatic Encoding Task: RAST means and standard 92 deviations of correct recall across Trials (1-4) within Conditions (Control and Experimental) for subjects in Groups (1-4). 15 Syntagmatic Encoding Task Derived Scores: RAST 9S Proactive Inhibition (RAST PI), Percent Release from Inhibition (PRI) and RAST Release from Inhibition (RPI) means and standard deviations for subject Groups (1~4). 16 Results of a three-way analysis of variance as a 97 function of Groups (1- -4), Condition (Control and Experimental) and Trial (1- 4) for the Syntagmatic Encoding Task. 17 Results of the Neuman-Keuls specific comparison test 98 on pairs of mean SIOpes fer the Groups (1, 2, 3, 4) for the Syntagmatic Encoding Task. .A difference between any two means is significant when it exceeds the appropriate critical value (CV) for o<=0.05. 18 RAST means for Condition (Control and Experimental) 99 by Trial (1-4) for the Syntagmatic Encoding Task. 19 Results of the Neuman- Keuls specific comparison test 100 on pairs of means for Trials (1, 2, 3, 4,) for the Syntagmatic Encoding Task. .A difference between any two means is significant when it exceeds the appro- priate critical value (CV) for afi=0.05. 20 Results of the simple main effects test for Condition 102 (Control and Experimental) at Trial (1, 2, 3, 4) in the Syntagmatic Encoding Task. xi TABLE 30 31 32 33 34 35 36 37 PAGE Results of ad-hoc analysis of the percent of 117 subjects in each group who demonstrated the Proactive Inhibition Effect (PIE): Proactive Inhibition for Control and Experimental Con— ditions (PI-C, PI-E); Release from Proactive Inhibition (RPI); the Proactive Inhibition Effect (PIE) (If PI-C, PI-E, and RPI occur). Group means and standard deviations of Error 120 Type (Substitution, Onission, Interference, Order) within subject Groups (1-4) and Evalu- ative Encoding Conditions (Control and Ex- perimental) . Results of a three-way analysis of variance as 123 a function of Group (1-4) , Condition (Control and Experimental), and Type of Error (Sub- stitution, Omission, Interference, Order) for the Evaluative Encoding Task . Results of a simple main effects test for 125 Group (1-4) at Condition (Control and Experi- mental) for the Evaluative Encoding Error Analysis. Results of Neuman-Keuls specific comparison 126 test on pairs of mean slopes for Error Type (Substitution, Omission, Interference, Order) for the Evaluative Encoding Task. A differ- ence between any two means is significant when it exceeds the appropriate critical value (CV) for cc ==r0 .05. Group means and standard deviations of Error 127 Type (Substitution, Omission, Interference, Order) within subject Groups (1-4) and Syntag- matic Task Conditions (Control and Experimental). Results of a three-way analysis of variance as 130 a function of Group (1-4) , Condition (Control and Experimental) and Type of Error (Sub- stitution, Qnission, Interference, Order) for the Syntagmat ic Encoding Task . Results of Neuman-Keuls specific comparison 131 test on pairs of means for Group (1, 2, 3, 4) for the Syntagmatic Encoding Task Error Analysis . A difference between any two means is signifi- cant when it exceeds the appropriate critical value (CV) for (Dc-0.05. xiii TABLE I PAGE 38 Results of Neuman-Keuls specific comparison test 132 on pairs of mean slopes for Error Type (Substi- tution, Onission, Interference, Order) for the Syntagmatic Encoding Task. A difference between any two means is significant when it exceeds the appropriate critical value (CV) for cc =0.0S. 39 Group means and standard deviations of Error Type 133 (Substitution, Omission, Interference, Order) within subject Groups (1-4) and Paradigmatic Task Conditions (Control and Experimental). 40 Results of a three-way analysis of variance as a 136 function of Group (1-4), Condition (Control and Experimental), and Type of Error (Substitution, Onission, Interference, Order) for the Paradig- mat ic Encoding Task . 41 Results of Neuman-Keuls specific comparison test 137 on pairs of means for Group (1, 2, 3, 4) for the Paradigmatic Encoding Task Error Analysis . A difference between any two means is significant when it exceeds the appropriate critical value (CV) for oc=0.05. 42 Results of Neuman-Keuls specific comparison test 138 on pairs of mean $10pes for Error Type (Substi- tution, Onission, Interference, Order) for the Paradigmatic Encoding Task. A difference between two means is significant when it exceeds the critical value (CV) for oC=0.05. xiv FIGURE LIST OF FIGURES Schematic presentation of the method used in quantifying release from PI, the percentage release being given by X/Y x 100 (Wickens, 1970). Percent-correct recall means for Group (1-4) on the Evaluative Encoding Task Control and Experimental Conditions. Percent-correct recall means for Group (1-4) on the Syntagmatic Encoding Task Control and Experimental Conditions. Percent-correct recall means for Group (1-4) on the Paradigmatic Encoding Task Control and Experimental Conditions. Histogram.of mean number of errors by error Type (Substitution, Omission, Interference, Order) for the Evaluative Encoding Task. Evaluative Encoding Task Item Error Analysis: Convergent Interaction of Group (1-4) by Condition (Control and Experimental). Histogram of mean number of errors by error type (Substitution, Omission, Interference, Order) for the Syntagmatic Encoding Task. Histogram.of mean number of errors by error type (Substitution, Omission, Interference, Order) for the Paradigmatic Encoding Task. PAGE 63 79 93 108 121 123 128 134 CHAPTER I LITERATURE REVIEW Intredustien he 8 t 'c- ad' In word association tasks, individuals are presented with a word and are expected to respond with another. This task has been of interest to developmental and clinical psychologists for some time. It has been used to study thought, concept formation and manipulation, memory, organization of meaning, and linguistic structure (Brown and Berko, 1960: Clark, 1970: McNeill, 1963: Thysell and Schultz, 1964). Research has demonstrated that children manifest two distinct response patterns on word association tasks. Young children (kindergarten and first graders) produce word associations that are syntagmatic in nature. That is, stimulus words produce responses of a different grammatical form class, but which occur in syntactic and semantic sequence in the language. For example, for the stimulus word "cat", the association that might be produced by a young child is "meow" or "run." Somewhat older children (second through fifth graders) typically respond in a paradigmatic fashion. That is, stimulus words produce response words of the corresponding grammatical form class. 2 For example, "cat" might be followed by a coordinate associate such as "dog" or perhaps a superordinate term, such as "animal." Paradigmatic responses are often synonyms, antonyms, coordinates, contrasts, or super- ordinates (Palermo, 1971: Woodward and Lowell, 1916). This developmental change in word association patterns has been labeled the syntagmatic-paradigmatic shift. Variables that Affect Word Associations e ss nd W r e nc It is known that variations occur in the type of responses individuals exhibit during word association tasks. Several studies of word associations in children (Castaneda, Fahel, and Odom, 1961: Entwisle, 1966: Entwisle, Forsyth, and Muuss, 1964: Palermo, 1971: Palermo and Jenkins, 1964; 1966) demonstrate that subject age and the form class of the stimulus word influences that associate which is produced. Entwisle et a1. (1964) originally studied associations from 500 children. Entwisle later expanded the study (1966) and presented data on form class in word associations of 1040 children's and 200 adults. The data were sorted according to form class and the frequency of occurrence of the stimulus word. Additionally, subject variables including grade, socio-economic status (SES), high or low urban 3 population, and intelligence quotient (IQ) were related to word association data. Table 1 is a modified summary based on Entwisle (1966), which presents the percentage of paradigmatic and syntagmatic responses by subject grade and by form class of both the stimulus word and its associate. These data are consistent with results from other studies (Brown and Berko, 1960: Ervin, 1961; Palermo, 1971). That is, paradigmatic responses increase remarkably with age. However, the increase is at different rates and different asymptotes for various form classes. Table 2 displays the most-to-least common response types for each form class relative to grade level. For nouns, the most frequent response at any age level is another noun. These responses of young children, however, should not be construed as paradigmatic because prior to and during kindergarten and first grade, children are more likely to use a noun response to any stimulus including other nouns. During these years, there is also a selective increase in syntactic responding. For example, from kindergarten to third grade, the number of adjectival responses to nouns doubles. The total pattern of response to nouns observed over the kindergarten to fifth grade range includes a gradual increase in paradigmatic responses, a decrease in the noun-verb syntactic response, and an increase in the adjective-noun response. In contrast, verb Table 1. 4 Mbdified summary of the percentage of paradigmatic and syntagmatic responses by subject grade and form class for stimulus words and associates (Entwisle, 1966). Grade Level % Syntagmatic Responses % Paradigmatic Responses OU‘ILNi—‘Vx‘ OU'ICNI—‘x Omar—*7: Omar-47s Adjectives Nouns 57.1 44.1 Nouns 35.0 28.1 10.5 6.6 [—1 KOOWU‘I «ora.s~q Response to Nouns verbs Response to Adjectives verbs Adverbs 13.4 3.0 13.0 2.6 7.0 1.7 4.4 1.6 Response to Verbs Adverbs “Prepositions 4.8 4 5.2 7.3 4.2 8.6 2.5 7.6 1.4 Response to Adverbs Adjectives Verbs 8.2 31.9 12.1 27.8 19.0 15.0 19.2 9.7 Nouns 61.2 62.7 72.9 78.1 77.1 Adjectives 16.8 31.7 70.6 78.5 63.8 verbs 16.6 20.4 47.5 59.6 60.0 Adverbs 8.6 17.9 51.2 62.2 78.0 Table 2. and frequency of occurrance (Entwisle, 1966). 5 Association responses by form class relative to grade level Grade Stimulus-Associate Form Class UI Most to Least'Commonly Produced Associates noun-noun noun-verb noun-adjective adjective-noun adjective-adjective adjective-verb adjective-adverb verb-verb verb-noun verb-adverb verb-preposition adverb-noun adverb-verb adverb-adverb adverb-adjective adverb-adverb adverb-verb/adjective adverb-adjective/verb adverb-noun 6 and other form class responses to nouns decrease steadily from the kindergarten level. For adjectives, verbs, and adverbs, kindergarten and first graders respond predominately with nouns, but paradigmatic responses to adjectives and verbs do occur during these ages. Third and fifth graders respond to adjectives and verbs in a paradigmatic manner, however, syntactic linkage to nouns persists to the fifth grade level. A dramatic change toward paradigmatic responding occurs for adjectives between the first and second grades. This change has been attributed to the development of contrasts (Entwisle, 1966: Nelson, 1977). Syntagmatic responding is considered to be most prevalent at around the kindergarten or first grade level, but it is still present on some form classes at the fifth grade level (Entwisle, 1966). Large changes in responses to various form classes occur between kindergarten and first grade. Changes are particularly dramatic for adjectives, adverbs, and pronouns. Even larger changes occur between first and third grade. Adjectives, pronouns, verbs, and adverbs continue to develop paradigmatic associates. Adjectives and pronouns reach an asymptote, whereas the other form classes continue to develop at a slower pace beyond third grade. By fifth grade the majority of responses are paradigmatic except responses to nouns. 7 Entwisle (1966) found evidence in her data that certain combinations of stimulus and response words strongly suggested syntactic patterns that are frequent sequences in phrases and sentences. She found, for example, that first grade children produced a small percentage (10%) of adverbial responses to high-frequency verbs, and fifth graders produced a similar percentage for medium-frequency verbs. The proportion of adverbs was smaller for low- frequency verbs, but doubled between kindergarten and fifth grade. In contrast, the proportion of adverbial responses to other form classes (e.g., nouns and adjectives) remained very small (3%) for all ages. Adverbial responses to verbs appear to develop over a period of time and are seemingly related to the frequency of the verbs. Noun—verb sequences are common for high-and medium-frequency verbs even at the fifth grade level, whereas adjective-verb combinations were less likely to occur. Similarly, noun responses to high- frequency adjectives were more numerous than to nouns or verbs. Verb responses to high-frequency nouns and pronouns exceeded high-frequency adjectives. These responses patterns suggest that children respond according to frequently occurring syntactic linkages in the language. The college-aged subjects in the Entwisle study demonstrated a decrease in paradigmatic responding and an increase in syntagmatic responding. This phenomenon has been observed by others (Deese, 1962: Fillenbaum and Jones, 8 1965: McNeil, 1964). McNeil explained that the syntactic patterns of adults differ in kind from those of children. Children's syntagmatic responses reflect grammatical pairings, such as the noun-verb sequence in the associate "cat-jump". Adult syntagmatic responses reflect semantic refinements related to newly acquired meanings for previously known words, such as in the adjective-noun combination "high-noon" (Brown and Berko, 1960; Entwisle, 1966: Entwisle et a1., 1964; Ervin, 1961; Nelson, 1977). Gender, IQ, and SES The effect of gender differences has been studied in relation to the occurrence of the shift. Entwisle (1966) found that boys lagged behind girls in kindergarten but this difference was temporary. No difference attributable to gender was evident after first grade in the proportion of paradigmatic responding. Differences were seen, however, in the types of responses given by males and females, particularly of older ages (e.g., grade five). Females characteristically gave more common responses. That is, they responded to stimulus words with more predictable, popular words (e.g., coat-hat), whereas males gave more idiosyncratic responses (e.g., coat-belt). Overall, the small gender differences in responses of the youngest children suggested that form class development sequence are similar for boys and girls of similar intelligence quotients 9 (IQ). Other studies have verified that gender differences are minimal (Masters, 1969: Weingarten and Anisfeld, 1981). Entwisle (1966) reported on the effect of IQ on the percentage of paradigmatic responses that are characteristic of children. The rapid development of paradigmatic responses to adjectives and pronouns that occurs between kindergarten and first grade was highly correlated with the general intelligence of the child. However, by third grade IQ differences were no longer associated with systematic differences in paradigmatic responding to adjectives. The effects of SES on paradigmatic responding were also considered by Entwisle (1966). A comparison of two urban samples of children in three grade levels, who represented different socioeconomic classes (i.e., high and low) and who were of average intelligence, revealed no statistical difference in the percentage of paradigmatic responding. A difference was observed in the total number of paradigmatic responses according to form class favoring the high students. The differences associated with SES were small. Differences in form class (i.e., verbs, adverbs, and pronouns) that remain at the fifth grade level are due to the later development of these form classes in terms of paradigmatic responding. It has been speculated that experience with language may be related to the general socioeconomic status of the children within a particular school. Evidence suggests that the effect of education and 10 experiences with taxonomic categorization significantly enhance paradigmatic responses (Masters, 1969: Sharp and Cole, 1972: Sharp and Gollin, 1985). In summary, IQ measures for young children are related to their ability to respond in a paradigmatic pattern during a word association task. Though, beyond third grade, IQ is not a good predictor of paradigmatic responding. Perhaps this is because most children have developed the paradigmatic response pattern by the third grade regardless of within-normal-limit IQ differences. Socioeconomic status alone does not appear to influence the development of paradigmatic responding. A mid—to-high SES average of children in a school may provide the environment for increased experience. Experiences such as antonym and synonym tasks and word category games may facilitate the development of paradigmatic word relationships. In summary, the syntagmatic-paradigmatic shift has been described in relation to subject and form class variables. The subject factor which contributes most to the occurrence of the shift is age, with gender, IQ, and SES contributing in lesser degrees. Form class and word frequency are variables that strongly influence the types of associates to stimuli. Further, form classes develop at different rates relative to age. The data for verbs and adjectives show most clearly the displacement by age of the shift. Children demonstrate evidence of the shift between five and nine 11 years of age (Entwisle, 1966; Nelson, 1977), with asymptote of most form classes complete by fifth grade. In contrast, adult subjects (between fifth grade and college age) develop a later syntactic phase which conveys richer semantic associates (Entwisle, 1966: McNeil, 1964). h or' o he n t - ' at'c Shift Several theories have been proposed and tested in attempts to explain the occurrence and timing of the syntagmatic-paradigmatic shift. Nelson (1977) reviewed four major classes of these theories and identified their inadequacies in explaining the shift. The following presentation is based upon that work. ss at v h r s Ervin (1961) studied association patterns of children in kindergarten through sixth grade. She found that most responses of kindergarten children were syntagmatic and there was a significant increase in paradigmatic responses with increasing age. Ervin proposed that this pattern of word associations is explained by "contiguity", i.e., "the frequency of associative pairing in language use" (Nelson, 1977, p.100). According to this theory, words become associated when they share the same or similar sentence contexts. For example, in the sentence: "We c1imbed.", one 12 could use "the hill" or "the mountain" interchangeably (Nelson, 1977). Thus, these two words would be associated. As the number and variety of sentential contexts increases with language experience, paradigmatic associations increase. In other words, associations are formed through shared links in grammatical form, meaning, and experience. Similarly, Palermo (1971) observed an increase in the frequency of the most popular responses as a function of age. He attributed this increase to increasing "associative strength" of words as children increase in age. Thus, the low frequency of contrasts, common responses, and paradigmatic responses is due to low associative strength in young children. While the associative theory may be feasible for explaining some syntagmatic associations, Nelson (1977) argued that it failed to account for associations built from common but nonspecific sentence frames such as "See the ____." It did not account for the relatively infrequent anticipation errors seen in children's sentence completions and the infrequent co-occurrence of paradigmatic associates that share common linguistic contexts and form class in our language (e.g,. "shallow and deep," "light and dark," big and small"). In addition, it did not recognize that the timing of the shift occurs well after children have developed vocabulary and syntactic knowledge and use (Francis, 1972). 13 c nd eat re- tch o McNeill (1966) interpreted the shift from syntagmatic to paradigmatic responding as a change representing increased semantic ability. His theory of semantic feature acquisition, based on prevalent linguistic theories (Chomsky, 1965: Katz and Fodor, 1963: Brown and Berko, 1960), assumed that the lexicon can be specified by features or markers that determine meaning, and by syntactic features that govern and restrict lexical use. For example, the syntactic feature of a the word "flower" is a common noun, whereas semantic features include physical object, living, small and plant (Nelson, 1977). The development of semantic features is assumed to be a gradual process. Because feature entries for words are built up gradually, associations change over time as children develop recognition of words that share common features. In earlier stages, when few features have been acquired, associations are few and unconstrained. Clifton (1967) supported McNeil's theory and distinguished syntagmatic responding as a matching of the restricted sentence contexts in which words can appear with the semantic features of the words. Paradigmatic responses were achieved by the matching of the whole set of features. Associates were described as pairs of words whose sets of features differ minimally. 14 Another supporter of the feature matching theory was Clark (1970), who suggested that linguistic structure rules may be followed in generating word associations. Clark pointed to examples illustrating that paradigmatic associations can be generated from the minimal contrast rule (e.g., dog-cat), the marking rule (e.g., man-woman), feature deletion and addition rules (e.g., couch-chair), and the category preservation rule (e.g., pants-shirt). Syntagmatic rules include the selectional feature realization rule described by Clifton (1967), and the idiom completion rule described by Nelson (1977). Nelson (1977) offered several arguments against a feature-matching hypothesis to explaine the syntagmatic- paradigmatic shift. She argued that interclass differences found in response patterns, especially those that are neither paradigmatic nor syntagmatic (e.g., eat-fat, see- blind) are not accounted for by the theory. In addition, the theory does not account for differences in responses to the various form classes. If semantic features were the primary means by which words were associated, variability within individuals and within form class would be minimal. Finally, the theory attributes too little semantic knowledge to young children who are able to produce and interpret complex utterances. Other theories have been proposed to explain the syntagmatic-paradigmatic shift. These theories consider 15 mechanisms for associative processes that differ from linguistic structure explanations. Thegries 9f Associative Structure The word association task has led investigators to assume that it taps unconscious organization of meaning. Its structure has been considered equivalent to the structure of memory and thought (Mandler, 1967: Tulving, 1968). Specifically, according to the associative structure theory, associations represent the way in which semantic or cognitive information is stored in memory (Nelson, 1977). The assumption is that there is a psychological structure of semantic memory that is not necessarily isomorphic with linguistic structure. Deese (1965) stated that certain underlying structural principles or schemata govern word associations. A hierarchical coding system, whereby words are represented by semantic features, develops to replace the word dictionary which is stored by young children (Anderson and Beh, 1968). Models of semantic memory have included other aspects of coding that can be employed in assigning meaning. Propositional relationships are those that consist of predications that can be made about a word (Miller, 1969). Although this is similar to Clark's claim that associations are derived from language structure, the mechanism proposed here is that associations are derived from language 16 function. Functional relationships are governed by psychological motives based on the semantic and pragmatic intention of speakers such as the semantic notion of agent in identifying the subject of a sentence or topicality (Bates and Mac Whinney, 1982). o ' v ev m 5 Word association differences between children and adults have also been described as representing stages in the development of logical thought. Woodward and Lowell (1916), Goett (1911), and others (Stolz and Tiffany, 1972: Moran, 1966; Riegel, 1970) emphasized that word association responses shift from syntactic to logical ones including synonyms, coordinates, contrasts, and superordinates. Although logical responses increase over the same age period as the shift from preoperational to concrete logical operations (Piaget, 1970), experimental studies have not revealed a clear developmental hierarchy connecting the two types of shifts (Cramer, 1974: Moran, 1966; Moran and Huang, 1974; Penk, 1971: Sullivan and Moran, 1967). Nelson's (1977) summary of studies emphasizing the development of logical concrete operations as a basis of word association changes concluded with the observation that paradigmatic responses are symptomatic of more mature thinking. It has been suggested that syntagmatic and paradigmatic responders do not differ in the amount or kind 17 of information present in the lexicon, but that they do differ in the cognitive set, type, or style to respond (Nelson, 1977: Riegel, 1970; Winokur and Tweeney, 1974; Woodward and Lowell, 1916). The process that allows children to use information differently as they mature has been referred to as cognitive-linguistic reorganization (Nelson, 1977). Cognitive reorganization is defined and explained by Nelson (1982, 1985) in terms of the development of event representations for concepts. According to a "functional core model," children enter into routine scripts that build up event knowledge. The concept of a single object develops when viewed as the focal point in a dynamic situation. That is, the function of the object (along with its perceptual properties) is important to the child's formation of the concept. Children establish knowledge of the various syntagmatic relationships (i.e., agent, action, object) that are possible for an object and are thus able to decontextualize the object concept from particular scripts. It is at this point that the ability to substitute these elements helps the child to establish categories of objects. Categorization requires the abstraction that different things can and do occupy the same position in a given semantic and syntactic structure. Category knowledge becomes context-free as a more advanced cognitive operation allows the child the ability to recognize same or similar 18 procedures of functions in different scripts. That is, across contexts paradigmatic relations are extracted from syntagmatic ones. Nelson referred to the concept of "animal" to illustrate that the recognition of function aids the development of context-free categories. Nelson (1977) used the example of how the words "dog" and "cat" become associated with the superordinate category "animal." Originally, through episodic experiences children store possible functional relations and definitional perceptual attributes that enable the child to use the words "dog" and "cat" in sentences. This non-hierarchical contextual level is reflected in early syntagmatic associations (Schank, 1976; Halperin, 1971). Common categories build up in vertical relationships as the child recognizes functional similarities that contrast with perceptual characteristics, which may be quite diverse. In this was, the child comes to recognize that "dog" is an animal and that "cat" is an animal. The coordination of group members within horizontal relationships appears to be difficult for young children, which may explain why paradigmatic responses increase more readily during the school years. The child develops the ability of inclusion, for example, which allows him to group "dog" and "cat" together as instances of animal. Coordination of concepts not only develops later as children develop advanced cognitive abilities, but it is also 19 reinforced and becomes salient in the kinds of concept tasks that school-age children are obliged to do (Nelson, 1977). School-age children respond paradigmatically because they have developed the conceptual structure to do so, along with responding to the saliency of coordinated relations. Other influences of word association choices may include recency of acquisition, frequency of use, instructions to respond, personal interest, and cognitive style (Nelson, 1977). Petrey (1977) reanalyzed the nature of the word association responses obtained by Entwisle (1966) to test the adequacy of the syntagmatic-paradigmatic distinction as compared to an episodic-semantic one and to hypothesize about the mechanisms which underlie these changes. Petrey found that indeed kindergarten and first grade children produced associations of an episodic nature that were usually syntagmatic (predicate-like) in structure. Syntagmatic associations formed from episodic experiences were either situational ("examine-needle"), or were formed from syntactic juxtaposition ("examine-blood pressure"). Older children and adults produced associations of a semantic or paradigmatic nature. That is, "The stimulus effects immediate cognitive input into semantic memory, whose systematic organization leads to a semantically related response." (Petrey, 1977, p.62). Petrey described the developmental changes in word associations as reflecting competence versus performance implications. She 20 hypothesized that linguistic maturation (i.e., word association changes) "involves a major reorganization of the mental lexicon during which memories of parole (performance) decline in importance and sensitivity to langue (competence) develops" (Petrey, 1977, p. 70). What is not clear, however, is the process of change from one type of organization to the other. The change from predicative to attributive function appears to provide an example in support of Nelson's hypothesis of cognitive reorganization, or Petrey's (1977) episodic-semantic shift. As children enter the school-age years, further changes toward increased semantic development occur within the adjective form class as with other form classes. As stated earlier, the adjective form class reflects the greatest change between syntagmatic and paradigmatic responses of first and third graders. Differences in cognitive organization may explain this phenomenon. For example, Nelson (1976) found that modifiers constitute an early form of predication for children in the early linguistic stage. Adjectives are used by children in this stage to comment on a topic or to describe (e.g., "Elizabeth is beautiful"). With an increase in language development and cognitive development during the pre-school and early school years, adjectives are used more to describe properties for the purpose of classification. 21 Another change in adjective use is the development of contrasts. Contrast adjectives have been noted to be asymmetrical in acquisition and use (Bierswisch, 1967: Greenberg, 1966; Lyons, 1968; Sapir, 1944, Vendler, 1968). Polar adjectives (e.g., long-short, high-low), where one is positive and the other negative, are particularly asymmetrical in development. Several factors may influence asymmetrical development. First, the positive form rather than its negative form provides the name for dimensions (e.g., long-length). Positive forms also maintain neutrality in questions (e.g., Which one is bigger?). On the other hand, in questions using negative adjectives, (e.g., Which one is smaller), the negative end of the dimension is being referred to. Thirdly, measure phrases contain positive rather than negative adjectives (e.g., five feet wide). Positive adjectives have been referred to as "unmarked" because they are less complex, whereas negative, contrastive adjectives are "marked" (Greenberg, 1966: Klaskey, Clark, and Macken, 1973). A sequence of learning stages in the acquisition of contrasts was suggested by Clark (1970) whereby young children understand and produce positive, unmarked forms earlier than negative, marked forms (Donaldson and Wales, 1970; Clark, 1971). The positive term names the dimension initially, then comparative meaning is acquired whereby both terms are equated with the positive end of the dimension. 22 Finally, correct assignment of the polar terms to the extended ends of the dimension continuum develops. The asymmetric learning of adjectives has been attributed to children's asymmetry of conceptual knowledge of the extent of dimensions. Klatsky, Clark, and Macken (1973) tested the conceptual hypothesis with a concept attainment task to determine children's learning of polarity of dimensions. In this task, CVC nonsense words were used to replace English words for the positive and negative ends of four dimensions: size, height, length, and thickness. This was done to eliminate the possibility that asymmetries in adult usage toward positive terms influence children's acquisitions. Children grouped by age (3-7 to 3-11: 4-0 to 4-4: 4-7 to 4-11) chose items which were, for example, either long or short relative to a given standard. Significant differences occurred in polarity due to superior performance with positive labels. There were no differences between age groups. However, age differences were not expected, because positive and negative terms are learned gradually, with all distinctions being acquired after age five (Clark, 1972). Klatsky, Clark, and Macken (1973) attributed the asymmetries in the acquisition of both adjective and CVC labels for dimensional terms to asymmetric conception of extension and lack of extension. Other possibilities for explaining the data, such as the adult usage hypothesis, 23 simple translation of CVC's to their English equivalents, verbal mediation, and response-bias, were ruled out by examining the data relative to each hypothesis. Based on these results the conceptual hypothesis was favored, and the nature of the cognitive mechanism underlying asymmetric acquisition of dimensional extent was proposed. Clark (1973) explained that positive terms require the knowledge of the presence of extension and or that an item has greater extension relative to a standard (another item). These reference points, primary and secondary, are simple comparative measures. The negative term, however, requires consideration of the standard, then selection of an appropriate item with less extension than the standard. This task is more complex because the standard is the item with greater extension. The normal way of encoding for the positive, unmarked forms where the standard is the less extended object is referred by Klasky, Clark, and Macken (1973) as canonical encoding. Children's preference for items with more extension reflect the canonical encoding of this type of term. The choice of an item with the lack of extension (negative, marked term) does not match canonical encoding and therefore is more complex at the cognitive level. In summary, the syntagmatic-paradigmatic shift can be explained by and is consistent with the functional core model of concept development and with cognitive-linguistic 24 reorganization theory (Francis, 1972: Nelson, 1977: Nelson, 1982). The lexical system undergoes change as the context- bound conceptual representation, based in event structures, moves to a categorical system which is removed from concrete event representation. The syntagmatic-paradigmatic distinction in word associations is present in the associations of children and adults. As described previously, the type of associations depend on form class, word frequency, and development of semantic information and organization. In the previously discussed works regarding the development of contrast adjectives (Klatsky, Clark and Macken, 1973: Clark, 1970; Nelson, 1976), positive forms developed earlier than negative forms. Positive dimensions of adjectives conform to canonical encoding, whereas negative dimensions violate canonical encoding and therefore represent a more complex cognitive task. Petrey (1977) suggested that syntagmatic-paradigmatic (episodic-semantic) differences in word associations may also reflect encoding changes. Children who respond in a syntagmatic manner may encode the stimulus word as a verbal unit which co-occurs with other units in natural language (e.g., examine-doctor), whereas children and adults who respond in a paradigmatic manner react (encode) to the stimulus as a lexical vehicle for its abstract cognitive content (e.g., examine—look). 25 The cognitive-linguistic reorganization theory emphasizes that cognitive changes underlie word association changes described as the syntagmatic-paradigmatic shift. More specifically, development of verbal encoding strategies which allow for classification by episodic experience and by comparison and inclusion of semantic content may underlie the shift. If young children encode episodic verbal units, while older children and adults encode stimuli based on their abstract cognitive content, the syntagmatic- paradigmatic shift may be a linguistic manifestation of encoding differences. It is of interest then to consider the nature of verbal encoding and developmental changes in encoding processes. Models of Memory Memory tasks, such as digit span, have been used as a measure of cognitive functioning for over 50 years. The tasks reveal a reliable development sequence, and have been related to other measures of cognitive ability. In the early 1960s, several models of memory were developed within the framework of the information-processing approach. This approach has sought to specify the mental processes that determine input-output relationships and attempts to analyze the transmission and transformation of information within the cognitive system. The assumptions 26 that underlie this approach include the premise that the cognitive system is composed of a well organized pattern which allows for a series of operations to be performed on information flowing through the system. The approach further assumes that information is transformed by "mental programs" that allow for input and retrieval of events (Hagen et al., 1975). Information-processing models have influenced the development of models of memory. Waugh and Norman (1965) proposed that memory is composed of two separate storage systems: primary memory and secondary memory. Primary memory has limited capacity and information is lost rapidly as new information replaces the old. By employing the process of rehearsal, however, information can be maintained and also transferred to secondary memory where a more permanent storage of information can occur. The model of Waugh and Norman was expanded by Atkinson and Shiffrin (1968) whose multistore model distinguished between structural memory features and control processes. Structural components, likened to "hardware" in computer terminology, include the physical system which receives sensory information (sensory register), allows it to remain for a short time in working memory (short-term store), and transfers it to a permanent part of memory (long-term store). Control processes, analogous to software programs, facilitate the flow of information within and between the 27 structural components. Control processes are those directly under the control of the individual which may be established, selected, and modified to suit the demands of a situation. Craik and Lockhart (1972) developed a model that stressed levels of analysis of stimuli that result in a memory trace. The existence of a memory trace is a function of the level of processing that a stimulus has undergone. The analysis of physical or sensory features comprises an initial level of processing. Deeper levels of analysis are more cognitive and semantic, concerned with pattern recognition and the extraction of meaning. The deeper the analysis, the more permanent the trace. Processes which may strengthen memory traces include elaborative rehearsal, mnemonic devices, storage strategies, and retrieval processes, all of which require intervention of the cognitive system (Handler, 1974). The changes in the emphasis of models of memory from structural storage characteristics to functional, conscious control processes are supported by empirical evidence regarding adult and children's encoding of stimulus attributes. The review of verbal encoding which follows, supports the claim that developmental changes in memory can be attributed to changes in the type of processing done, rather than to changes in the characteristics of the structural components of memory (Hagen et al., 1975). 28 V E C The ability to extract symbolic information from words depends upon the number and kind of attributes that are contained in a memory trace. The process of encoding, according to Wickens (1972) involves one's ability to respond to the psychologically prominent stored aspects of stimuli. The meaning of a word results from encoding of the word along a number of these attributes. The view that elements, attributes, components, or dimensions contribute to a words's unique meaning is supported by several researchers (Bower, 1967; Norman and Rumelhar, 1970: Osgood, Suci, and Tannenbaum, 1957; Underwood, 1969: Wickens, 1970). Specific attributes of encoding have been defined from a large number of studies. One encoding task paradigm used extensively in these studies is an adaptation of a short- term memory task. It grew out of the Keppel and Underwood (1962) paper on "proactive inhibition" in short-term memory, which demonstrated that the Peterson and Peterson (1959) paradigm was subject to proactive interfering effects. The "release from proactive inhibition" (RPI) technique, described below, is considered by Wickens (1972) as an index of encoding. It has been used as a technique for identifying features of symbolic information. 29 The procedures for the RPI technique include the presentation of a triad of words that are considered members of a group representing a particular attribute (e.g., food, animals, feminine). Following the triad presentation, a rehearsal prevention activity occurs for 20 seconds and recall of the words is then requested. Another triad of the same attribute is presented, and so on. This procedure continues for four trials in the control condition. For the experimental condition, the procedure is similar except the fourth-trial triad consists of materials representing a different attribute. Typically, recall is excellent for Trial 1, but declines steadily across the first three trials for both the control and experimental conditions. In the experimental condition, when there is a shift to psychologically different materials on Trial 4, marked improvement in recall occurs. This shift has been labeled the release from proactive inhibition (Wickens, 1972). More than 20 experiments have been performed using the release from PI procedure to identify and quantify attributes of words encoded by children and adults. Studies have focused on semantics, grammatical class, connotation, physical characteristics of words, and manner of word presentation. Semantic attributes appear to be particularly salient to adults and children (i.e., taxonomic category, semantic differential, sense impression, masculine-feminine distinctions) as found by Goggin and Wickens (1971), Wickens 30 and Clark (1968), Rail and Levine (1975), and Rail and Schroll (1974). Manner of word presentation (e.g., mode oflanguage presented to bilingual speakers, word frequency, digits versus printed words for numbers) is also a salient attribute (Baldwin, 1969: Goggin & Wickens, 1971; Swanson & Wickens, 1970), whereas grammatical class (e.g., verb- adjective, verb-noun, tense) and physical characteristics (e.g., number of syllables or phonemes, articulation position, modality of presentation) are not (Wickens, Clark, Hill, & Wittlinger, 1968). Semantic Encoging Dimensions Evidence of semantic encoding by adults is provided from the research regarding RPI. Some discussion of these psychological dimensions will help to clarify the studies involving developmental trends in encoding that are important for this study. One important attribute of encoding is taxonomic class or categorical (conceptual) encoding. Categories are exemplified by items such as the names of various animals, foods, clothing, and body parts (Goggin & Wickens, 1971). A shift in the class on the fourth trial of the release from PI procedure produces a large increase in recall. Taxonomic class appears to be highly salient for adults. Developmental considerations of this attribute will be addressed in the following section. 31 A second semantic encoding attribute found to be salient for adults is the value of the word on the semantic differential (Osgood, Suci, and Tannenbaum, 1957). The semantic differential is a code for the connotative meaning of words described by a point in three dimensional space. That is, a word's semantic value is determined by its location in a sphere having three orthogonal axes identified as evaluation (i.e, good-bad, pleasant-unpleasant), activity (i.e., active-passive, lively-still), and potency (i.e., strong-weak, tough-tender). Heise (1965) used these bipolar adjective scales to assemble a dictionary of semantic profiles of 1,000 words. From this report, it is possible to group words together that are relatively homogeneous with respect to their connotative meaning. Wickens and Clark (1968) demonstrated that words at the end of the semantic differential scale are encoded similarly in short-term memory and represent a psychological class differing from words at the other end of the scale. They verified that release from PI occurs on the dimensions of evaluation, potency, and activity in college age subjects. This result supports the theory that adult subjects encode verbal materials by some meaning characteristic which is associated with the extremes on each of the Osgood scales, and that the dimensions of the scales are bipolar and represent different classes of connotative meaning. 32 Green and Goldfried (1965) challenged the validity of bipolarity of semantic space. These authors criticized the use of semantic differential scales to investigate meaning because of the forced bipolar structure of the scales. In order to test the assumption of bipolarity for semantic space, single-adjective scales were devised. Subjects rated words against a series of descriptive adjectives. The adjectives were divided into the "most obvious" scale and the "least obvious" scale. The results of correlation coefficients for these ratings showed definite negative relationships in 10 of the obvious adjective pairs: good-bad, clean-dirty, honest-dishonest, large-small, hard- soft, strong-weak, happy-sad, long-short, fast-slow, and alive-dead. Pairs which did not show negative correlations were more common among the least obvious scale. These words are less clearly opposed in meaning than the most obvious pairs. The results failed to yield support for a generalized, bipolar, and symmetrical model of semantic space. Green and Goldfried's results suggest that bipolar factors will be found if one selects adjective pairs which appear to be unambiguous in their opposed meaning. The more typical outcome is for unimodal factors to appear. Two other semantic encoding attributes that are less salient to adults than taxonomic class and the semantic differential are sense impression and masculine-feminine connotations. Sense impression describes physical 33 characteristics of stimuli (e.g., roundness, whiteness). Wagner (1970) demonstrated the release from PI in college students. The masculine-feminine connotation includes sex role knowledge and preference for either masculine or feminine items (Kail & Levine, 1974). This attribute is not particulary salient for adults, but does have implications for children. Further evidence in the literature supports the claim that adults encode verbal material of a semantic nature. For example, free recall tasks have been used to observe clustering of items from the presented list by taxonomic class (Bousfield, 1953), by associative relatedness (Russell, 1952; Cofer, 1965), and by sense impression Hudson, 1968). Cofer and Bruce (1965) did not obtain clustering by grammatical classes in free recall. These findings are similar to results of experiments of encoding using the release from PI procedure. Underwood (1965) and Kimble (1968) used a recognition memory task to determine encoding of single words. Subjects heard a large group of words, then judged whether particular words were heard previously in the list. Words were included in the later portion of the list which bore relationships as antonyms, associates, categorical names, sense impression to the earlier words. False recognition errors occurred for category names and associates. Kimble found that errors occurred based on sense impression. 34 Using a transfer of training design, Ory (1968) investigated the hypothesis that single words are categorically encoded. Paired lists were constructed containing several relationships. List A-B represented the stimulus (A) and the response (B) term. List A-C represented the same stimulus but a different response (C). The D-C list contained different stimulus and different response words. List A-B' indicated that the stimulus was the same and response (8') was similar in some way to (B). A'-B represented the reverse of this condition. Finally, A- Br indicated that the same stimulus and response are present, but the pairing differed in the two lists. Four groups of 20 subjects participated in the transfer of training experiment. The type of list design, given independently to one group, influenced their ability to learn a second list. The D-C group was the control group. Relative to the performance of this group, the A-B' group was superior to the A-C group. The A-B'r group performed at the lowest level of mean number correct words learned. The ability of A-B subjects to recognize relationships between the responses used the first-list responses as mediators. The A-B'r group evidenced interference effects from List 1. The results of this study suggested that categorical encoding for single words occurs at both an automatic as well as conscious level. (Wickens, 1970). 35 Various studies of semantic encoding in adults have been discussed. The experimental findings demonstrate that certain attributes are more salient than others. Taxonomic class, the semantic differential, and sense impression are the most prominent attributes that adults encode. It is of interest to this study to review the finding of semantic encoding in children from a developmental perspective. Deveiopmentai Characterisries of Egeoding The developmental research regarding encoding of semantic attributes using the RPI index has focused on the attributes discussed previously that have been used for adult encoding tasks. Kail and Schroll (1974) used visual presentation of stimuli (i.e., printed words) to investigate taxonomic encoding in 7 and 11 year old children. The categories, body parts and animals, were used in the release from PI procedure. The results confirmed that older children's recall was higher than the younger children's and that taxonomic encoding was accomplished by both young and older children. These findings have been replicated with first graders (Libby & Kroes, 1971) and second graders (Pender, 1969: Wagner, 1970) using the auditory presentation of stimuli. Research on children's ability to encode along the semantic differential has resulted in equivocal findings. Pender (1969) found that second and sixth grade subjects 36 demonstrated release from PI for the evaluative dimension when stimuli were presented auditorily. Cermak, Sagotsky, and Moshier (1972) and Kail and Schroll (1974) found that developmental differences in the ability to use the evaluative dimension as an encoding category occurred. Sixth graders demonstrated the accumulation and release from PI, whereas second graders did not demonstrate these effects. Kail and Schroll (1974) pointed out that the manner in which stimuli were chosen in the Cermak et al. study may have been problematic. Kail and Schroll used the ratings of Heise (1965) to maximize evaluative values and minimize scores on the activity and potency dimensions. No age differences in recall were found and developmental differences were evident in encoding of the evaluative dimension. The ability to use evaluative classes in memory appears to develop sometime after the ability to place words correctly in evaluative categories. Di Vesta (1966) demonstrated that seven year olds accurately classify words into positive and negative categories along the evaluative dimension, yet only older children used the categories to encode words. Kail and Schroll suggested that older children may have richer semantic networks in which to encode items. Improvement in recall may also reflect more sophisticated retrieval strategies. 37 The different outcomes between the evaluative dimension studies were explained by Kail and Schroll (1974) as the possible result of modality presentation differences. They suggested that while seven year olds can encode using taxonomic categories with both auditory and visual stimuli, they may only use the auditory modality for evaluative encoding. Older children are able to use both modalities to encode this dimension. Auditory presentation of stimuli may be encoded more completely than visual stimuli. A greater amount of interference may built up and thus allow for more sensitivity to a change in subsequent stimuli. This suggestion is supported by the Pender (1969) data discussed previously, as well as by Roberts (1968) who found that recall by seven-year-old children of auditorily presented stimuli was superior to recall of a visual nature. Another suggested difference between auditory and visual presentation of stimuli is the simultaneous versus sequential nature of the stimuli. Kail and Schroll (1974) hypothesized that sequential (auditory) presentation may result in increased interference and as a result he more responsive to changes in stimuli. The encoding of semantic differential attributes has been related to children's word associations based on the premise that associations reflect similarities in semantic coding. Semantic differential ratings (Di Vesta, 1966) were correlated with the Palermo and Jenkins (1963) norms of 38 children's word associations to determine whether the stimulus word and its primary associate have similar locations in semantic space. The representational mediational response paradigm (Osgood, Suci, and Tannenbaum, 1959) predicts this to be the case. Di Vesta restricted correlational data to fifth grade responses. The coefficients of each salient factor were .64 for Evaluation, .47 for Potency, and .62 for Activity. For opposites, the semantic differential ratings of the stimulus word and its opposite were inversely correlated. These findings support the representational mediation response paradigm and suggest that the primary associate is mediated by similarity in connotative meaning. As with adults, two additional semantic encoding categories have been investigated in children using the PI task. The sense impression category, based on shared physical properties, and the masculine-feminine dimension were found to have saliency for children. For the sense impression category, college age subjects demonstrated the release from PI effect. There was a slight effect for 11- year olds (Wagner, 1970), whereas there was no effect for younger children (Kroes, 1973: Wagner, 1970). The masculine-feminine dimension was studied in children using words that represented games that were judged as being appropriate for boys (e.g., hunting, airplane) or girls (e.g., hopscotch, dolls) (Kail & Levine, 1974). The 39 release effect was found for 7 and 10 year old boys as well as for 7 year old girls. These children were the most extreme in their preference for stimuli which matched their gender. The 10 year old girls, however, did not show this strong stimuli preference. Those girls who did not show sex-typed choices on the sex-role preference task did not show evidence of encoding in the release from PI task. False recognition procedures and the influence of facilitative instructions on generalization errors have been used to investigate developmental changes in encoding processes. The false recognition procedure consists of the presentation of word lists that contain the key stimuli. A recognition test is then presented that includes the key words and new words that are related either acoustically, semantically, or orthographically. The dependent variable is the number of words that a subject falsely identifies as being on the original key word list. Encoding attributes can be defined when a child consistently makes errors that bear a type of relationship with the key words. Several experiments using the false recognition procedure have attempted to determine changes in structural (sensory) versus semantic encoding (Bach and Underwood, 1970: Cramer, 1972: Felzen and Anisfeld, 1970: Freund and Johnson, 1972). The findings from these studies collectively indicate that semantic encoding is established in children as young as four years of age. The 40 developmental age of nonsemantic, orthographic, and acoustic encoding is less clear. A flexible system which favors the encoding of semantic information and possible nonsemantic characteristics under particular circumstances is suggested by Hagen et al. (1974). Cramer's (1972) study of the influence of facilitative instructions on phonetic and semantic generalization errors of first and fifth graders provides further support for developmental changes in children's responses to verbal material. Materials for this study included a 12 word training list for which each word could be associated with a synonym or a rhyming word. A list of 12 control words bearing no associations to the training list was con- structed. The test list consisted of 12 presentation words, 12 generalization words (six synonyms, six rhymes), and 12 control words. Subjects were randomly assigned to one of three instructional groups. The "neutral learning" group listened to the word lists and were told to remember them. Subjects in the "semantic-facilitation" group were told to think of other words that mean the same thing. The "sound facilitation" group was instructed to think of words that sound the same as the stimulus word. The measured variable was the number of generalization errors. Sound instruction decreased the generalization errors of first graders (for whom phonetic memory organization was assumed to be dominant), whereas semantic instructions had a greater 41 effect on decreasing fifth grader's generalization errors (for whom semantic organization was assumed to be dominant). Semantic instructions minimally facilitated semantic encoding in first graders only. Because sound was already established as a basis for memory organization in both age groups, phonetic encoding was not influenced by sound instruction. Cramer concluded that the sound facilitative instructions decrease generalization errors only when a new memory organization is emerging. First graders but not fifth graders were influenced, suggesting the emergence of new organization in the younger group. Felzen and Anisfeld (1970) found that third graders make more generalization errors to rhyming foils than to semantically related foils. The reverse was found to be true for sixth graders. These data support the assertion that memory organization shifts from a sound or phonetically based organizational structure to a semantically based one. Another example of developmental changes that occur as representational abilities mature includes Luria and Vinogradova's (1957) report that a direct relationship between mental age and the changeover from clang responses (rhyming or otherwise acoustically related words) to meaningful responses was found (Nelson, 1977). Children who are just entering the verbal mediational stage (five to seven years) show these clang responses, whereas older 42 children's responses are based on more meaningful similarities (e.g., synonyms, antonyms). Cramer (1973) later demonstrated that facilitative instructions interacted with grade level in producing generalization errors for synonyms and antonyms. Similarly, Reiss (1946) found generalization to homonyms and synonyms to change as a function of age. These studies support the claim that age effects underlie organizational of verbal memories. Clark (1972) investigated the acquisition of antonyms in terms of semantic features (semantic fields). Dimensional and spatio-temporal terms were studied in a task which required children to give a word opposite to adjectives that conformed to the positive-negative contrast. Clark hypothesized that children should be able to group words that share features of meaning into a field and substitute one for another on the basis of their shared features. The second hypothesis predicted that the child's responses would reflect the order of acquisition based upon semantic complexity and that substitutions also follow the order of acquisition. Clark's subjects were preschool children, ages 4-0 to 5-5, who were placed into three groups based on five month age difference intervals. The results of the investigation confirmed that semantic fields are set up by young children even before the full meanings for the words have been established. The older the group, the 43 larger the number of semantically related responses to the dimensional and spatio-temporal terms. There was no overall difference in acquisition between the members of a pair. It was concluded from this finding that the positive-negative poles are learned as pairs, not as single items. Between pair differences were sizable. The order in which the pairs in each field were acquired was closely related to the semantic complexity of the pair. That is, less complex pairs were learned first. The substitution errors reflected increasing semantic appropriateness with increases in age, and more fully specified terms were always substituted. Hagan et al. (1974) offerred several conclusions relative to the experimental findings of the encoding of attributes by adults and children. The first is that the pattern of results from studies of children is generally consistent with the pattern from adult studies. Specifically, the semantic attributes are the most important features in the encoding process, while the role of nonsemantic attributes such as acoustic features is ambiguous. The attributes that are dominant in the encoding of adults tend to be present in the encoding of very young children. Less salient attributes in the encoding of adults are not detected until later in a child's development. The memory trace for a stimulus appears to contain a number of semantically based attributes for children as young as 6 years of age. With increases in age, the representation of 44 stimuli are encoded into an increasing number of semantic categories. ummar u R V 6W The development of word associations from syntagmatic to paradigmatic relationships has been considered to be linguistic evidence of conceptual development. Association changes are gradual and depend on the form class and word frequency of the stimulus, as well as subject variables such as age, gender, IQ, and SES. Recent explanations of the syntagmatic—paradigmatic shift offered by Nelson (1977), Francis (1972) and Petrey (1977) favor the cognitive reorganization theory. This theory assumes that the kind of knowledge available to the child (perceptual, functional, relational, logical, episod- ic) remains unchanged, but the availability of hierarchical conceptual relations develops as vertical and horizontal connections between concepts increases. These conceptual relations increase with age, cognitive development and experience with concepts. As episodic knowledge develops into greater semantic knowledge, the linguistic system is mapped onto these changes. The syntagmatic-paradigmatic shift then, is thought to be an example of the change in linguistic production that reflects underlying conceptual development and change. 45 What is not clear in the empirical studies and the theoretical discussions is the mechanism that underpins the change from one type of lexical organization to the other. Petrey (1977) asked the question "...by what process can episodic memories of words in context lead to the abstract semantic organization of mature lexical storage" (p. 70). Petrey proposed that children "react to" a verbal stimulus either in terms of a verbal unit, which co-occurs with other units, or as a vehicle for the abstraction of semantic information. According to Petrey, this difference between syntagmatic (episodic) and paradigmatic (semantic) storage~ retrieval systems defines the change in word associations. The notion that children "react to" the associations between words appears to refer to a relationship between how children encode associations and how they produce them. The empirical evidence on encoding of attributes suggests that developmental differences are present in children between the ages of six and 12 years. Differences in evaluative encoding, sense impression, and the masculine- feminine connotation have been revealed through the release from PI procedure. Other paradigms such as false recognition and facilitative instruction have illustrated changes in memory organization toward semantically based attributes. It has not been demonstrated, however, that children "react to" the syntagmatic versus paradigmatic association styles in terms of encoding processes. 46 The purpose of this study was to determine whether children encode syntagmatic and paradigmatic attributes of word relationships and whether encoding differences exist on the basis of children's word association styles. It was of interest to determine whether children who respond primarily in a syntagmatic fashion encode syntagmatic associations versus paradigmatic ones. Similarly, do children who respond primarily in a paradigmatic fashion encode paradigmatic associations as well as syntagmatic ones? Group differences in encoding abilities were predicted based on the hypothesis that word association response differences reflect encoding differences. A second purpose of this study was to determine whether the evaluative dimension, in auditorily presented stimuli, is encoded by first, second, and fifth graders. It was also of interest to determine whether encoding differences are attributable to age or to differences in word association styles. CHAPTER II METHOD Subjecrs Sixty-four subjects were selected from a middle socio-economic status school district in mid-Michigan. The socio-economic status of individual subjects was not available: however, approximately 14% of the population of the school from which these children were drawn came from families who receive government income subsidy. The school district mean for this subsidy is 24% with a range from 2% to 49%. Children were selected from first, second, and fifth grade classrooms. These grade levels were chosen because the syntagmatic-paradigmatic shift occurs most dramatically between grades one and two, and it is complete by grade five (Entwisle, 1966: Palermo, 1971). Children were placed into four experimental groups based on age and performance on a word association pretest 47 48 described below. Table 3 provides a summary of the age, gender, and race of children in each group. School records were reviewed to insure that the children performed within the average range on grade level classroom work and on available standardized measures. First grade subjects had scores within the average range on the "Curriculum Monitoring System" in reading and mathematics or on the Deboron Kindergarten Screening instrument. Second and fifth grade subjects had percentile rankings of between 25 and 99 in reading and mathematics on the sranrord Aehievemenr Ies . All of the subjects attended full-time education classrooms and were within age-grade appropriate limits. They were reported by their classroom teacher as performing at grade level. None were classified as being speech, language, hearing, learning, or emotionally impaired. Subjects were seen for a brief (10 minutes) initial session to complete a word association pretest described below. They also completed a pretest of rapid naming. Table 3. 49 Summary of age, gender, and racial characteristics of 64 children classified in four groups. Each group contained 16 children. Group Age Gender Race (1-4) Years-months Male Female White Black Other Group 1 (first and second grade syntagmatic responders) mean 7.0 8 8 12 4 0 mode 7.1 S.D. 4.8 Range 6.4 to 7.1 Group 2 (first grade paradigmatic responders) mean 6.8 11 5 13 2 1 mode 7.1 8.0. 3.2 Range 6.2 to 7.1 Group 3 (second grade paradigmatic responders) mean 7.8 10 6 12 2 2 mode 8.2 S.D. 3.83 Range 7.2 to 8.2 Group 4 (fifth grade paradigmatic responders) mean 10.9 10 6 14 2 0 mode 11.1 S D. 3.04 Range 10.4 to 11.2 50 Worg Association Pretesr Each subject received a word list consisting of frequently occurring nouns, verbs, and adjectives (Thorndike & Lorge, 1944; Entwisle, 1966). A sample is in Appendix A. Seven nouns, eight verbs, and 15 adjectives were included in the lists. Half of the words were adjectives because this form class is sensitive to a high rate of paradigmatic responses, particularly constrastive in nature (e.g., hot-cold). Children who do not respond paradigmatically on adjectives are likely not to respond in this manner to other from classes. Thus, children who met the criteria for Group 1 did not demonstrate contrastive responding on the pretest. No more than two adjectives appeared consecutively on any list. These quasi-randomized lists were constructed to insure that subjects did not have the opportunity, through practice, to adopt a contrastive responding style. 51 Subjects were tested individually in a quiet room. The following instructions were given verbally to each subject: "Today I want to play a word game with you. You may not have played this game before, so let me explain it you. I'm going to say some words to you, one at a time. Each time I say a word, I want you to tell me the first word that you think of. I'll write it down and then say another word. Let's try a few practice words, I'll say the word and you tell me the first word you think of. CAT- (etudeg; respogse) That's fine. Let's try another one. GRASS-(erggegr response) O.K. Now we'll start the game." (If the subject does not understand the task as demonstrated by responses that deviate from the instructions, several examples will be given.) "For CAT you could say DOG or SCRATCH or maybe FURRY. For GRASS you could say GREEN, MOW, or TREE . " Responses were transcribed by the examiner, then classified according to their principle usage. Syntagmatic responses were those that differed in form class from the stimulus word (e.g., eat iggeh). Paradigmatic responses were those consistent with the form class of the stimulus word (e.g., long ehgrr) (Entwisle, 1966). Appendix B provides the specific decision rules used to classify each response. The percentage of syntagmatic and paradigmatic responses was calculated for each subject and group 52 membership was determined by grade level and percentage of paradigmatic associations on the word association pretest (See Table 4). The criteria for group eligibility (Palermo, 1971: Entwisle, 1966) is as follows: Group 1: 16 first and second grade children (ages 6.2 to 7.1 and 7.2 to 8.2)) who exhibit syntagmatic responses (40% and below of association responses in a quasi-randomized list are paradigmatic) Group 2: 16 first grade children (ages 6.2 to 7.1) who exhibit paradigmatic responses (60% or greater of association responses in a quasi-randomized list are paradigmatic) Group 3: 16 second grade children (ages 7.2 to 8.2) who exhibit paradigmatic responses (60% or greater of association responses in a quasi-randomized list are paradigmatic) Group 4: 16 fifth grade children (ages 10.0 to 11.6) who exhibit paradigmatic responses (60% or greater of association responses adjectives embedded in a quasi-randomized list are paradigmatic) During the second session, the word association task was randomly repeated on 28% of the subjects. This was done following the experimental tasks so that subjects were not influenced in any way by these stimuli. The test- retest consistency of word association style was calculated for individual subjects by (a) dividing the 53 Table 4. The group means and ranges of the percent of paradigmatic responses on the word Association Task. Paradigmatic Responses Group ‘Mean % Range % (1-4) Group 1 22.0 3.4- 43.3 (first and second graders) Group 2 85.9 67.9- 96.7 (first.graders) Group 3 76.1 56.7- 90.0 (second graders) Group 4 75.8 56.7-100.0 (fifth graders) 54 number of paradigmatic responses in the first trial (P1) by the mean number of stimuli in that trial (29); (b) dividing the number of paradigmatic responses in the second trial (P2) by the mean number of stimuli in that trial (29): and (c) finding the ratio of the two quotients. In other words, (P1/29)/(P2/29)=P3. Consistency information for each group is reported below: Grogp Tesr-Retest (3) Negber gr Subjects 1 93.6 7 2 81.6 3 3 91.6 5 4 90.7 3 Letrer Identification Prerest Following the word association pretest, subjects performed a naming task. Printed letters, presented individually on 3 x 5" cards were identified. The children then practiced naming the letters at the rate of one per second. All subjects were able to perform this task which was subsequently used as the rehearsal preventing distractor task in the experimental tasks to be described below. 55 Subject Age For Group 1, the mean age was 83.4 (7.0 years) with a standard deviation of 4.8, a mode of 85, and a range from 77 to 85 (6.4 to 7.1 years). For Group 2, the mean age in months was 81.4 (6.8 years) with a standard deviation of 3.2. The mode was 85 and range was 75 to 85 months (6.2 to 7.1 years). Note that most of the subjects in Group 1 and 2 were matched for age. The subjects in Group 1 were from first and second grade classrooms. They were grouped together because they performed in a similar manner on the word association pretest described later. Their syntagmatic response preference was the primary basis upon which they were classified. Group 2 was comprised of first graders only. They were of the same mean age as Group 1, but differed in the type of responses given on the word association pretest. Their paradigmatic response preference was the primary basis upon which they were classified. 56 Group 3 subjects were second grade paradigmatic responders. The mean age in months was 94 (7.8 years) with a standard deviation of 3.8 and a mode of 98. The range was from 87 to 98 months (7.2 to 8.2 years). The Group 4 mean age in months was 130.8 (10.9 years) with a standard deviation of 3.0 and a mode of 133. The range in months was from 125 to 135 (10.4 to 11.2 years). Subjeer Race end gender The total subject population was comprised of 39 males and 25 females. There were 10 black subjects, three in the category "other" (neither black nor white), and 51 white subjects. In Group I there were eight males and eight females. Seven males were white, one was black, whereas five females were white and three were black. Group 2 contained 11 males and five white females. Eight males were white, two were black, and one was other. There were 10 males and six females in Group 3. Nine males were white and one was other. 57 Three females were white, two were black, and one was other. Group 4 was comprised of 10 males and six females. Nine males were white and one was black. Five females were white and one was black. W Subjects were seen for a forty minute session between two and four weeks following their classification into groups. The experimental tasks described below were accomplished as well as the follow- up word association retest procedure. P o ct'v nh b't on Pr ce r Wickens et al. (1963) developed and extensively used an experimental short term memory paradigm to investigate attributes of words encoded by adults. It was a modification of the Peterson-Peterson (1959) short-term memory paradigm. In this Procedure, subjects are asked to remember two or three words that share a particular attribute (e.g., apple, orange, cherry (fruit)). A distractor task is used during a 20 second 58 delay before recall of the words is requested. Typically, subjects decline in their ability to recall presented words across several trials. The systematic decline in recall performance is attributed to "proactive inhibition" (PI). That is, because the words share an attribute, learning of new information that also contains this attribute is inhibited. On the last trial (generally the fourth), stimuli can either be presented which represents the attribute of previous trials or stimuli can represent a different attribute. When stimuli are changed on the fourth trial it is known as the "shift" trial. Subjects who encode the change in the attribute of the shift trial improve in their ability to recall those items. This improvement in recall is called "release from proactive inhibition" (RPI). The degree of recall improvement on a shift trial is indicative of the psychological importance of a different attribute being encoded in the short-term memory of the subject. In experiments using this procedure, the control condition is when all of the 59 trials contain one attribute. Therefore, the last trial is not a shift trial. In the experimental condition, the last trial of the task contains a different attribute than the attribute represented by previous trials. In this case, the shift trial represents a change in the stimuli. This procedure has been used in the investigation of attributes children encode (Cermak et al., 1972: Pender, 1969: Kail and Schroll, 1974; Kail and Levine, 1977). It is used in this study to investigate the encoding of various attributes by children who respond in a word association task with syntagmatic or paradigmatic responses. The following instructions were given verbally to each subject: "I am going to say three words that I want you to try to remember in the same order that they are given. After I say the words, then you will say them. Next, you will name the letters that you see on these cards. When you are finished naming the letters, tell me the three words that I told to you. Let us practice first. Then we will begin. Ready? happy, sun, cookie." (The subject then was gesturally prompted to name the letters). "O.K., now tell me the words." 60 A trial consisted of the following events. First, a word triad was presented by live voice at the rate of approximately one word per second (Pender, 1969). The subject was asked to repeat each triad aloud immediately after it was presented. Nesxt, a distractor task consisting of letter naming occurred for 20 seconds to prevent rehearsal of the triad (Kail and Schroll, 1974). Letters were presented on 3x5" index cards at an approximate rate of one per second and were randomized before each trial. Finally, following the distractor task, the subject was asked to verbally recall the word triad. Four trials, similar to that just described, were administered. Each trial employed a different set of words representing the selected attribute. Each subject received the control and experimental conditions for each of three encoding tasks to be described below. This accounts for six tasks which were presented in a randomized order. A one minute interval between tasks and conditions occurred. During this time the letter cards for the distraction task were shuffled 61 and general verbal reinforcement and general conversation were engaged in. This was done to minimize interference between the stimuli in the tasks and their conditions. Subjects received one point for each word recalled. An additional point was earned for each word recalled in correct order. Therefore, each word was worth a total of two points. Subjects could earn between 0 and 6 points on each trial. Raw scores were converted to proportional values (e.g., .00: .167: .333; .50: .667: .833: 1.0). Scores for Control condition Trial 1 served as the baseline control (CTl). The scores for Control Trial 2 (CT2), Control Trial 3 (CT3), and Control Trial 4 (CT4) represent measures of decline in recall. Scoring procedures for the experimental condition were identical to the control condition (i.e., ET1, ET2, ET3, ET4). Performance measures were defined as described by Wickens (1973). First, proactive inhibition (PI) was defined as any decrease in correct recall of stimulus 62 items occurring between Control Trial 1 and Control Trial 4: (CTl-CT4). An additional index of PI for experimental trials was defined as (ET1-ET3). Second, release from proactive inhibition (RPI) was defined as the difference between the Control Trial 4 and Experimental Trial 4 (ET4-CT4) divided by PI for the Control Condition (CTl-CT4). Thus, the equation for estimating RPI is [(ET4-CT4)/(CTl-CT4)]. Third, the percent of RPI (PRI) was derived by multipying the product of the RPI equation by 100. Figure 1 illustrates the relationship between these indices. An example of a subject's score profile for one task is as follows: Control 1 (C1) = 2 (Proportional value = .333) Control 4 (C4) = 0 (Proportional value = .000) Experimental 4 (E4) 1 (Proportional value = .167) [(Er4-cr4)/(cr1-cr4)j [(1-0)/(2-0)] = .50 x 100 = 50% Reliebiiiry. Ten subjects, who were randomly chosen, were audiotaped during administration of the tasks. The tapes were compared to the transcription of responses on the student's response sheet. The mean item to item agreement in transcription for the 72 Correct Recall (%) 63 '00 l l I l 80—- —' 60— — A 40 — '— — 4) " 1' Y X 20 — '— I—I Experimental "" O-O Control "‘ 0 | l l l l 2 3 4 Trial Figure l. Schanatic presentation of the method used in quantifying release from PI, the percentage release being given by X/Y x 100. (Wickens, 1970) 64 responses per subject was 99%. The assignment of scores to each item and the trial totals were examined for each of the 64 subjects to determine the consistency with which the three trained graduate clinicians scored responses. The mean inter-judge agreement was 99%. Preactive inhibition Iasks v u t've d s . The evaluative dimension of the semantic differential has been found to be encoded by adults and children (Hagen et al., 1975) depending upon the modality of stimulus presentation. When presented via the auditory modality, children as young as seven years of age have been reported to demonstrate release from PI (Pender, 1969). However, other studies using printed information (Cermak it al., 1972; Kail and Schroll, 1974) demonstrate that only older children (11 year olds) encode the evaluative attribute. This study sought to determine whether young children (six, seven, and eight year olds) encode the 65 children (six, seven, and eight year olds) encode the evaluative attribute auditorily. It was of further interest to determine whether word association development, which is considered to be related to conceptual development is predictive of the ability to encode the evaluative dimension. The criteria established by Kail and Schroll (1974) to select positive and negative words as rated by Heise (1965) was used as a basis of word selection for this task. In order for a word to be considered heavily loaded on the positive-negative evaluation continuum it required a (+,-) rating of 1.0 or greater. Ratings of less than (+,-) 1.0 were required on the activity and potency dimensions. The selected positive and negative words and their semantic differential ratings are listed in Appendix C. The 13 words that are comprised the positive word pool, and the 14 words that are comprised in the negative word pool are listed below. Four groups of triads were randomly selected and arranged into eight 66 triads were randomly selected and arranged into eight sets presented in Appendix D. Posirive wgrde: true, uncle, save, good, open, church, farm, health, milk, fix, fresh, offer, live Negarive gorgs; bad, cut, hate, burn, fall, kill, fear, quit, shoot, worry, trouble, problem, no, miss In each task condition (control and experimental), subjects received word triads for four trials. No word appeared more than once across the four trials for a particular subject. One-fourth of the subjects (four) in each group received the stimulus triads in the following manner for the control and experimental conditions: positive control-positive experimental negative control-negative experimental positive control-negative experimental negative control-positive experimental On the Control condition, the four trials consisted of the assigned attribute. On the Experiemental condition, however, the first three trials consisted of the assigned attribute with the fourth shift trial consisting of the opposite attribute. For example, a 67 subject assigned to the positive triads received them for four trials in the Control condition. If positive triads were assigned to the Experimental condition, the first three trials were positive, followed by a negative triad on the shift trial. WW Young children who associate words in a predominantly syntagmatic fashion are thought to have less or differently developed conceptual organization than do those who produce paradigmatic associations. A question of interest in this study is whether children encode the difference between syntagmatic versus paradigmatic verbal material. To determine whether encoding of these associative relationships is occurring, the release from PI procedure was used. Stimuli for this task were randomized combinations of phrase-like frames which included size words (adjectives), color words (adjectives), and shape words (nouns). An example of a triad is "big, red, star." 68 Six size words (small, short, fat, big, tall, thin), 10 color words (red, blue, pink, white, green, orange, black, brown, gray, tan), and six shape words (circle, square, tube, line, star, triangle) were in the pool of words used to construct eight sets of stimuli. These types of adjectives and nouns were chosen to constrain semantic information and to maximize the syntactic associations of the words within the triads. Word triads were similarly constructed to represent paradigmatic responses for the fourth (shift) trial in the experimental condition. Specifically, the triads included color words (adjectives). The 10 color words that were comprised in the pool of words were the same as for the syntactic color adjectives, however, care was taken to insure that no color was repeated in a given trial series for a particular subject. The eight sets of stimuli for the syntagmatic encoding task (control and experimental conditions) are in Appendix E. The task sequence was similar to that described in the Evaluative Encoding Task. The trial sequence 69 remained unchanged (i.e., word triad presentation, word repetition, letter naming, recall) and each subject received four trials. For the control condition, subjects received syntagmatic word triads across the four trials. For the experimental condition, subjects received syntagmatic triads across the first three trials. On the fourth trial a paradigmatic word triad was presented. The indices derived from the previous task were also derived for this task. That is, Proactive Inhibition (PI), Release from Inhibition (RPI), and Percentage Release from Inhibition (PRI) will be calculated. a a atic ncodin Task. An early form of paradigmatic responses to word associations is categorical or taxonomic in nature (e.g., shirt-pants). Categorical relationships are considered to be highly semantic, and categories are salient attributes encoded by children as young as six years of age, as well as 70 adults (Kail and Schroll, 1974: Libby 8 Kroes, 1971: Pender, 1969). In this study, it was of interest to determine whether children of different ages were able to encode the difference between taxonomic or paradigmatic stimuli and the phrase structure of the syntagmatic stimuli. Of particular interest was whether children's association preference influenced their recall of stimuli. The paradigmatic encoding task was similar to the two previously described tasks. The proactive inhibition procedure was used to determine whether experimental groups encoded a shift from paradigmatic to syntagmatic stimuli. The stimuli for this task consisted of triads comprising adjectives (colors). Twelve colors were in the pool of words from which triads were formed. The colors (pink, blue, white, tan, orange, green, red, gray, black, beige, rust, brown) did not appear more than once in any set of four triads. Eight sets were formed (Appendix F). For the control condition subjects received paradigmatic word triads for 71 four trials. For the experimental condition subjects received three trials of paradigmatic word triads. On the fourth experimental trial, a syntagmatic word triad was presented. The scoring procedures were identical to the other two tasks. Re abi it of Tas dm' '5 at'o and Scorin . Three graduate students were trained to administer and score the tasks described above. These judges assisted the author in data collection. Judges were trained in task administration via live demonstration and in scoring of audiotapes of six children, two each from first, second, and fifth grades. Training continued until each judge demonstrated 95% item-to-item agreement in scoring with the author. CHAPTER III RESULTS Percent-Correcr Receii The dependent variables of this experiment were children's responses to three encoding tasks: Evaluative, Syntagmatic, and Paradigmatic. Each task consisted of a control and an experimental condition. Each condition included four trials consisting of word triads. Responses to trials were scored in terms of the number of words correctly recalled as well as scored for the order of the recalled words. The maximum total raw score per trial for a subject was 6: 1 point for each correct word in a triad and 1 point for each word given in correct order. Trial level scores were converted to a proportion (expressed as percent-correct in figures and tables). An example of the stimuli and scoring procedures for each of the three tasks is provided in Appendix G. Because proportions tend to be skewed and because skewed data may violate the assumptions of ANOVA, all proportional data generated in this study were transformed prior to statistical analysis. Studebaker (1985) suggested a Rationalized Arcsine Transform as a way to normalize such 72 73 distributions and to produce scales for which the real effects are linear and additive. A spreadsheet was written in Microsoft Excel (Macintosh version 1.5, 1988) to accomplish this transformation. Inputs to the spreadsheet were proportions of correctly recalled responses for each subject and the number of test items equal to a proportion of 1.0. Proportions were converted to radians using Studebaker's equation 3: T = arcsinelJX/(N-l) + arcsine .¢X+1)/(N+1). Resulting values for T were multiplied by a scaling factor of 46.4732433: the value 23 was subtracted from this product. The outputs of the spreadsheet were Rationalized Arcsine Transforms (RAST), expressed in units of "raus." The percent-correct recall means and standard deviations for groups on the four trials within each task and condition are presented in separate tables for each task. A corresponding table expressed in RAST is also provided. Derived Dependenr Veriebies; P1, 231 RAST values were used to calculate two dependent variables from subject recall performance on the four trials within each task condition. The first derived score was Proactive Inhibition (PI). This was calculated by subtracting Control condition Trial 4 from Control condition Trial 1 expressed by the equation (CTl-CT4). Similarly, PI 74 was calculated for the Experimental condition by subtracting Trial 3 from Trial 1. PI is an index of the amount of memory decline that subjects demonstrate on successive trials. The second derived score was RAST Release from Proactive Inhibition (RPI). This was calculated by subtracting the RAST value of the Control condition Trial 4 from the Experimental condition Trial 4. The PI value was divided into the difference. The equation [(ET4-CT4)/(CT1- CT4)] x 100 expresses the RPI. The values were calculated for each subject, then means and standard deviations were calculated. The means and standard deviations of RPI for all of the subjects within groups are presented in separate tables for each task. The following analyses were performed on the data using the Brain Power StatView 512+ and Clear Lake Research ANOVA, both written for the Macintosh computer. First, experimental results were described by means, standard deviations, and ranges for each of the measures taken from each subject. Rationalized Arcsine Transforms (Studebaker, 1985) of proportional data were accomplished. A three-way repeated measures analysis of variance for group means for each task was performed to assess the significance of differences among groups, task conditions, and task trials. Newman-Keuls post hoc tests were used for significant main effects and Simple Effects tests were used 75 when interaction effects occur. The level of significance was set at .05 for ANOVA and post hoc tests. Task outcomes (Evaluative Encoding, Syntagmatic Encoding, Paradigmatic Encoding) are further expressed as Proactive Inhibition (PI) values and Percent Release from Memory Inhibition (PRI) and were analyzed with a one—way, randomized block design ANOVA to assess the significance of differences among groups. In the event of a significant F- ratio for the group main effect, a Newman-Keuls test was run to identify patterns of differences among groups. The level of significance for these tests was set at .05. An item error-type analysis was used to assess whether response differences were evident among groups. A three-way ANOVA determined differences in group, condition, and error type. When significant F-ratios were found, a Newman-Keuls test was performed. The level of significance for these tests was set at .05. Proportion correct scores for individual trials were analyzed to determine the number of children whose data exhibited the constituent phenomena necessary for the proactive inhibition effect (PIE). The analysis was done separately for each group within each of the three encoding tasks. The numbers of children whose data satisfied the following definitions were summed, then expressed as a percentage of the total number of youngsters in the group. Proactive inhibition-control (PI-C): a child was said to exhibit PI-C if and only if the RAST score for 76 Trial 1 exceeded, by any amount, the RAST score for Trial 4. Proactive inhibition-experimental (PI-E): a child was said to exhibit PI-E if and only if the RAST score for Trial 1 exceeded the RAST score for Trial 2, or if the Trial 1 score exceeded Trial 3. Release from proactive inhibition (RPI): a child was said to exhibit RPI if and only if the RAST score for experimental Trial 4 exceeded the RAST score for the Control Trial 4. Proactive inhibition Effect (PIE): a child was said to exhibit PIE if and only if the child exhibited both PI-C and RPI. Encoding Task Outcgmes Eva uat ve Encodi s e - o c ecal . Tables 5 and 6 present the proportional means and standard deviations (represented in percentage) for group recall performance and the RAST of this data by trial on the Control and Experimental conditions of the Evaluative Encoding Task. Figure 2 displays each group's percent-correct recall performance on the control and experimental conditions. In the Control condition, Group 1 subjects (first and second grade syntagmatic responders) correctly recalled a mean of 32.3% of stimuli in Trial 1, 26% in Trial 2, 34.4% in Trial 3, and 35.4% in Trial 4. The mean percent across trials was 32. The mean percent of stimuli recalled by Group 1 in the Experimental condition was 44.8% in Trial 1, 77 o.om w.mv e.wm_ A.mH mH.v H.m~ ~.mm m.c~ .Q.m v m.mo m.mm N.mm H.~m «.8w w.sm o.mv ¢.co m.~c v.mm AM m.mm H.Hv s.mm 0.5m N.mm m.mm m.cm c.mm .:.m m m.em H.mm m.mv N.vm o.mo n.0m m.sm H.u~ N.m~ H.mm .M m.cm m.m~ m.om H.cm m.~m m.mm m.m~ m.m~ .:.m N s.mm H.mm o.qm ~.cm o.Hm c.m~ w.o~ o.m~ N.m~ m.me .M m.cm m.Hm a.~m m.vm m.nm o.mm m.mm u.mH .a.m a c.Hm m.HN m.~m n.5a m.ev o.~m «.mm v.8m c.o~ m.~m .M m .N .H mdsosu Raise. 4 RE... m as: N E: H 32... again. a as: m RE. N RE. H St... 06835 cowpwpcou Hmucoewaoaxm :ofiudpcou Houucoo .Av-~v me30pmcuuon23m pom nfimucoswaocxm tam Hosucouv chMHMp:6n.cm:uwz Av-~v mammks mmOLOm Hfimuos uuoscou mo meowumm>op tsapcaum tam memos accuse; ”xmms mamvouzm o>wum3~m>m .m oases 78 N.wm 4.84 N.H4A 4.NN 4.44 8.24 4.8m 8.8N .a.m 4 8.8 a: as 4.4m 884 4.4m 4.4.4 NS 4.48 38 w o.N4 8.44 m.Nm H.s4 4.Nm 8.84 m.Nm 4.84 .a.m m N.4m H.4m m.m4 o.mm N.mo N.om 8.84 4.NN N.NN 4.4m .M m.Nm m.¢m 4.4m N.Hm N.mm o.mm 4.4m 8.4N .a.m N N.mm m.NN a.mN N.4N 4.Nm N.NN a.NH 4.NN N.mN N.m4 .m 4.Nm m.NN 4.4m m.om m.mm N.mm 8.NN 8.NN .a.m 4 N.om o.oN N.Nm N.44 o.m4 o.NN 4.4m m.4m H.mN N.4n .m .m .N .N masons 4H4NNH.M. 4 44449 n HmNNe N Haste N Haste mNmata.m_ 4 Neath m Haste N Nests 4 Haste Hoonasm :ofluwpcou Hmucosfluomxm cowumvcoo Houucou .973 330.6 36.33 sow 2355,4353 use 3558 25:36:81 553 :73 MHmmhL. mmOhum HHGUQH HUQMHOU MO mfiOwwaNrOfi Uhmfizmum Em mzmvfi g "xmme wcmtoocm o>flumsfim>m .o ofinmh — Mean Correct Recall (93) Mean Correct Recall (R) 79 H Experimental O-O Control 00 I I I I I I I I .. Group I ._ _. Group 2.. 80 - a - - 60 - - - '- I— .1 P - 40 - - *- '- 20 _ - — -l 01 I I I I I I I I l 2 3 4 l 2 3 4 Trial Trial 00 I I I I _ Group 3- _ 80 - - - 60 r - '- 40 - - - d 20 _ - I.- u- r - — - 0 I I I I I I I I i 2 3 4 l 2 3 4 Trial ‘ Trial Figure 2. Percent-correct recall means for Group (1-4) on the Evaluative Encoding Task Control and Experimental conditions . 80 17.7% in Trial 2, 32.3% in Trial 3, and 21.9% in Trial 4. The average percent across trials was 29. Group 2, which has a mean age similar to Group 1 (first grade paradigmatic responders), recalled a mean of 43.8% in Control Trial 1. Trials 2, 3, and 4 were recalled with 29.2%, 25%, and 20.5% accuracy respectively. The average percent of recall for the four trials was 29.6. In the Experimental condition, subjects recalled a mean of 51% in Trial 1 and 26.1, 24, and 28.1% on subsequent trials for an average of 32.3%. Recall means for Group 3 subjects (second grade paradigmatic responders) in the Control condition was 53.1 % for Trial 1, 29.2% for Trial 2, 27.1% for Trial 3, and 37.5% for Trial 4. The trial mean was 36.7. In the Experimental condition, percent scores per trial were 63.6, 54.2, 45.8, and 53.1 for an average of 54.2%. The mean percentage of recall for Group 4 subjects (fifth grade paradigmatic responders) was 59.4 in Trial 1, 62.5 in Trial 2, 60.4 in Trial 3, and 49 in Trial 4 in the Control condition. The average trial mean for the Control condition was 57.8%. The mean percent recall in the Experimental condition for Trial 1 was 84.4, 52.1 for Trial 2, 55.2 for Trial 3, and 73.9 for Trial 4. The trial mean was 66.4%. 81 e 've c es: . RAST values of correct recall were used to calculate PI which is the decline in recall from Trial 1 to Trial 4 in the Control condition and from Trial 1 to Trial 3 in the Experimental condition. Because RAST values are used in the inferential analysis they, rather than percentages, are reported here. Table 7 is a summary of the RAST group means by condition of PI. The mean RAST PI for Group 1 was -.11. For Group 2, the mean RAST PI was 24.82. Group 3 produced a mean RAST PI of 17.88. For Group 4, the mean RAST PI was 10.29. The Percent Release from Proactive Inhibition (PRI) was calculated from the trial data for each subject by the formula [(ET4-CT4)//(CT1-CT4)] x 100, as previously defined. Table 7 displays the PRI means and standard deviations as well as RAST values (RPI) for each group. The derived calculations were performed subsequent to per subject calculations. The mean RPI for Group 1 was 77.72 with a standard deviation of 147.07. For Group 2, the mean RPI was 58.86 with a standard deviation of 86.22. Group 3 had a mean RPI of 23.97 with a standard deviation of 114.41. Group 4 had a mean RPI of 83.53 with a standard deviation of 130.36. Seatierieel Analysis. A three-way repeated measures ANOVA was performed on the RAST-correct recall data 82 Table 7. Evaluative Encoding Task Derived Scores: RAST Proactive Inhibition (RAST PI), Percent Release from Inhibition (PRI) and RAST Release from Inhibition (RAST RPI) means and standard deviations fer subject Groups (1-4). Group I *RAST PI *PRI ‘ RAST RPI (1-4) S.D. (C1-C4) (E4‘C4)/(C1-C4)X100 1 . I -.11 69.0 77.7 5.0. 114.9 147.1 . I 24.8 50.5 58.9 - S.D. 75.2 86.2 3 I 17.88 36.1 24.0 S.D. 150.4 114.4 4 I 10.29 78.1 83.5 5.0. 111.2 130.4 *Individual cells were calculated by: (l) Calculating values designated by column headings for each subject individually, then; (2) Calculating summary statistics across subjects within the respective groups. 83 from the Evaluative Encoding Task (See Table 8). Main effects of this ANOVA were groups (1-4), conditions (control and experimental), and trials (1-4). Each of the main effects was significant, as was the interaction between groups and conditions, indicating that differences between control and experimental conditions were not the same for each of the four groups. The group by condition interaction was investigated through tests of simple main effects (See Table 9). The group effect produced significant differences for each condition (control and experimental), but the condition effect was significant only for Group 3. Thus, when RAST-correct recall data were collapsed across trials, only the second grade paradigmatic responders differed between control and experimental conditions. Specifically, these children produced higher RAST correct-recall scores in the Experimental condition than in the Control condition. The significant main effect of trial was probed through the Newman-Keuls test of mean differences (Table 10). Trial 4 differed from Trials 1, 2, and 3. In particular, Trial 4 produced significantly higher RAST-correct recall scores than the other trials. This outcome, plus the lack of a significant trial by condition interaction, suggests that although PI may have occurred, RPI was at best inconsistent. IFigure 2 shows the large variability within trials. 84 Table 8. Results of a three-way analysis of variance as a function of Group (1-4), Condition (Control and Experimental) and Trial (1~4) for the Evaluative Encoding Task. Source of Sum of Mean variation Squares df Square F P G 97431.762 3 32477.254 14.734 .000 Error 132254.961 6O 2204.249 C 5831.933 1 5831.933 6.652 .0124 GC 8115.003 3 2705.001 3.086 .0339 Error 52600.774 60 876.680 T 27604.187 3 9201.396 7.669 .0001 GT 6123.130 9 680.348 .567 .8229 Error 215974.251 180 1199.857 CT 3677.906 3 1225.969 .980 .4036 GCT 11132.841 9 1236.982 .989 .4511 Error 225235.559 180 1251.309 85 Table 9. Results of the simple main effects test for Group (1-4) on Condition (Control and Experimental) in the Evaluative Encoding Task. Effect MSn DFn DFe ‘MSe F P G at CONT 11804.985 3 6O 1371.295 8.609 .000 G at EXPER 23377.270 3 60 1709.633 13.674 .000 C at GROUP 1 476.556 1 60 876.680 .544 .464 C at GROUP 2 365.817 1 60 876.680 .417 .521 C at GROUP 3 10069.579 1 60 876.680 11.486 .001 C at GROUP 4 3034.984 1 60 876.680 3.462 .068 86 Table 10. Results of the Neuman-Keuls specific comparison test on pairs of Trial means (1, Z, 3, 4) in the Evaluative Encoding Task..A difference between any two means is significant when it exceeds the apprOpriate critical value (CV) for q=0.05. Trial 1 Trial 2 Trial 3 Trial 4 Means 55.088 36.871 38.132 39.962 Trial 1 * * Trial 2 * Trial 3 * * Denotes a significant difference between pairs of means. 87 A one—way repeated-measures ANOVA of RPI was performed to assess differences between groups (See Table 11). The main effect of group was not significant, indicating that groups did not differ in amount of RPI for the Evaluative Encoding Task. The ad-hoc analysis of the percent of subjects in each group who demonstrated the expected phenomena of PI-C (proactive inhibition-control condition), PI-E (PI experimental condition), RPI (release from PI), and PIE (proactive inhibition effect) is presented in Table 12. In Group 1, 43.75% of subjects demonstrated PI-C, as compared to 68.75% who exhibited PI-E. There were 37.50% of. subjects who evidenced RPI and 25% who demonstrated PIE. For Group 2, similar percentages of subjects were noted. PI-C occurred for 62.50% of subjects and PI-E occurred for 87.50%. RPI was found for 50% of subjects and 37.50% demonstrated PIE. Group 3 PI-C occurred for 56.25% of subjects and PI-E was found for 68.75% of subjects. RPI was produced in 43.75% of subjects and 18.75% demonstrated PIE. In Group 4, 62.50% of subjects demonstrated PI-C as compared to 87.50% for PI-E. RPI occurred for 50% of subjects and PIE was evident for 50% of subjects. 88 Table 11. Results of a one-way analysis of variance as a function of Group (1-4) for the RAST Release from Proactive Inhibition on the Evaluative Encoding Task. Source of Sum of Mban variation Squares df Square F P G 34607.944 3 11535.981 .780 .5097 Error 887198.035 60 14786.634 Table 12. 89 Results of an ad-hoc analysis of the percent of subjects in each group who demonstrated the Proactive Inhibition Task Effect (PIE): Proactive Inhibition for Control and Experimental Conditions (PI-C, PI-E), Release from Proactive Inhibition (RPI) and the Pro- active Inhibition (PIE) Effect (If PI-C, PI-E and RPI occur). Groups Evaluation Encoding Task: Percent Subjects Within (1-4) Groups PIeC PI-C RPI PIE Group 43.75 68.75 37.50 25.00 Group 62.50 87.50 50.00 37.50 Group ’ 56.25 68.75 43.75 18.75 Group 62.50 87.50 50.00 50.00 90 S ta a 'c ncodin T sk Outc es Percent-Qorreer Reeali. The means and standard deviations of the proportions of correct recall (indicated as percent) and the RASTs of these data by trial on control and experimental conditions of the syntagmatic encoding task are present in Table 13 and 14. Figure 3 provides a display of each group's percent-correct recall performance on the control versus the experimental conditions. For Group 1, subjects recalled 39.6% in Control Trial 1. Recall was 28.1% for Trial 2, 21.9% for Trial 3, and 41.7% for Trial 4. The mean percent-correct recall across trials was 32.15. In the Experimental condition, subjects in Group I recalled 51% for Trial 1, and 24%, 31.2%, and 45.8% on successive trials. The mean across trials was 38%. Group 2 subjects recalled 46.9% in Control Trial 1, 22.9% in Control Trial 2, 33.3% in Control Trial 3, and 15.6% in Control Trial 4. The average percent-correct recall was 29.68%. Experimental condition trials for subjects within Group 2 yielded percent-correct recall across trials as follows: 40.6%, 40.6%, 31.2%, and 49%. The average percent-correct recall across trials was 40.35%. The mean percent-correct recall for Group 3 subjects in Control Trial 1 was 46.9. Trial 2 yielded 55.2%, Trial 3, 35.4%, and Trial 4, 39.6%. The mean of trials was 44.28%. In the Experimental condition, Group 3 subjects produced 64.6% recall in Trial 1, 41.7% recall in Trial 2, 49% recall 91 s cm m.vm o.v~ m.¢m m.mN m.Nm a.HN .D.m v mn.ch w.on w.mo m.Nh m.nw Nw.cu w.mo m.Hw o.mo o.om .W h.cm v.Nv o.mm m.wN m.vm o.mv a.cm m.ov .O.m m nn.Hm o.mv c.mv N.Hv o.vo mm.vv c.mm e.mm N.mm m.ov .M c.mm m.mm m.hm h.vv “.mn w.~m ¢.om m.vm .Q.w N cv.nm o.mv N.Hm 0.00 o.ov wc.mm c.mH m.mm a.- m.ov .M _ m.mN v.0N m.mm v.Nv o.mm m.m~ m.Hm v.om .D.m a ev.mm m.mv N.Hm o.v~ o.~m mH.Nm B.Hv m.~N H.mN o.mm .M m .N .H mmsouo mammhh.k_ v Hmmhh m awash N Hawks H awash mfimwhb x v Hmfihh m awash N awash a Awash Hummnzm sewumpcou Hmucosmuoaxm cowuwpcou Monacou .av-Hv mdsoau_cw muuomnsm how Ammucoswaoaxm can Hosucouv mcofiuwpcoo casuwz Av-~v mumwuh mmocum Hamuoh uoosaoo mo mcowumw>op usepcwum use memos accused "xmme mcmtoucu emumsmmucxm .mH manna 92 54m mém 5mm 9mm 0.5m EON N.mm NHNN .D.w v omdw v.2. 9mm 9.05 N.Hm Ndn 0.65 m.mw ago hem M. 50.3“ adv Eon mdN m.om QKQ m.wm 06¢ .Q.m m m¢.~m .N..om 0.94 0.94 ~60 mi; vdm m.vm wém mdv 1% H6». min oém N69 ¢.NN 9mm Emm 5mm .Q.m N mm.\.m ode Ném ~44. N444 «flaw 04H 0.,mm mtnN o.w¢ M m.om fiww c.mm $.34 me m.mN mém m.Nm .940. H swim 5.34 4.4.3” 0.3.. N.mm N.mm m.Nv NHON HdN m.ov M. m .N .4 mnsoho mHmTfi. M v HGT; m Hmwb. N Hawhh H an“: mamwhp. x #4 Hawhp. m Hmwfi. N Haw: H Hm“: “~00ng :OMHNpcou Heucoswsooxm cowuwncoo Houuaou mmopom 44muos 4004400 mo meowumw>oc pumpcmum paw memos.Hm®© fihmfiflmum mun—m mCmOE HfiOUhOQ uxmme mchoucm oHomeHpmsem .mN oHnmb 107 0.0m 0.04 o.mN m.mN 0.0m .H.mm 0.mm 5.mN .m.m 4 m.N5 0.Nm 0.5m n.05 0.Nw 0.55 N.05 0.00 m.m5 0.00 .M H.H4 4.0m 5.4N 4.xm 0.mm m.mm m.Hm 4.Nm .Q.m m m.0m 0.0m 4.mm 0.Hm 0.40 4.Hm 5.m4 m.N4 m.mm N.o0 .M N.m4 m.4m N.NN n.4m w.mN 5.4m m.mm 0.mm .n.m N 0.5m 0.mv 5.0m N.mN w.om m.H4 m.mN H.m4 4.0m o.N4 .M 4.m4 0.4m N.mm m.mm 0.4m m.0m 4.0m m.mm .m.m H H.H4 N.me m.Nm m.04 0.00 N.m4 H.04 4.00 0.0m 5.0m ”M miN .4 HmH49.w_ 4 HmHyb m HmHhh N HmHsh H HmHsh mHmHse.M 4 HaHsh m HmHsh N HaHue H HmHye masosu :oHuncou HmucoEHsodxm :oHqucou Houucou uuomnsm .HQIHV masouu Hoomcsm sow HHmucoEHsocxm use Hosucouv mcoHuHccou :quHz HQIHV mHmHuh mmouom HHmuos Hoopsoo mo mcoHumH>00 tampcmum 0cm mcmos 5m44034m>m won 44-40 muncho poonnom c4044: 446040 .oocopomsoHGH :04mm4ec .:04u:u4pmnomv 0044 youum mo m:04um4>oc wsmwcmam 0cm mamas 0:040 .40 04004 121 _ ExperImental Control \\\\\\\\\\\\\\\\\‘6 0\\\\\\\\\\\\\\\\\\\\\\\\\V\\\\\\\\\\\\\\\\\. \\\\,\\\\\ INT 0RD Total . Q\V\\V\V_VV\.\\V.. \\\\\\\\V\V\\\b Group 2 SUB OMS \\\\\\\\\.\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\ .\\\.\,\\\\. INT 0RD Total ~\\\\\V\V\\V\\\. .\\V\\\.V_\V.\.\VVM SUB OMS Group I 28$ :8: 1m V\\\\\\\\\\\\,\\\\\\\\\\\\\. M T 0 .\\\\\\\~ m T. m I S \\\\\\\\\\\ M 4 0 p U B m 0 G _ _ — 4 _ _ _ _ _ _ _ _ I . I .m \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\. MW. .1 m \\ I \\\\\\\ 0 T N , S ‘\\\\\\\\\\\\\\\\\\\\ M 3 O m. : .4 B m V\\\\\\\\ w G 80:0 :8: Histogram of mean numlaer of errors by error type . (Substitution, mission, Interference, Order) for the Evaluative Encoding Task. Figure 5 . 122 8.50 errors in the Control condition and a mean of 6.4 errors in the Experimental condition. Group 4 (fifth grader paradigmatic responders) produced a mean of 5.80 in the Control condition as compared to a mean of 4.50 in the Experimental Condition. The most common error types were Omissions and Substitutions. This was true for all groups and both conditions. Order errors were the third most common made by all the groups with very little distinction among groups in mean errors. Interference errors were infrequent for all groups. The total mean Interference error was .50 for both control and experimental conditions. This indicates that stimuli in one encoding task probably did not contaminate responses to other encoding tasks. Statisticai Anniysis, A three-factor (Group x Condition x Error Type) repeated-measures ANOVA (See Table 32) produced significant results for the main effects of Group (1-4), Condition (Control and Experimental) and error Type (Substitution, Omission, Order, Interference). The Group by Condition interaction effect was significant. Figure 6 shows this effect to be convergent. For the Control Condition, Groups 1, 2, and 3 produced similar (and higher) mean errors; Group 4 yielded fewer error. For the Experimental Condition, Groups 1 and 2 demonstrated similar (and higher) mean errors; Group 3, an 123 Table 32. Results of a three-way analysis of variance as a function of Group (1-4) , Condition (Control and Experimental), and Type of Error (Substitution, Onission, Interference, Order) for the Evaluative En- coding Task. Source of Sum of Mean Variation Squares df Square F P G 90.662 3 30.221 14.852 .0000 Error 122.086 60 2.035 c 4.314 1 4.314 5.055 .0282 CC 7.600 3 2.533 2.968 .0390 Error 51.211 60 .854 T 467.037 3 155.679 20.686 .0000 GT 35.205 9 3.912 .520 .8590 Error 1354.633 180 7.526 CT 1.693 3 .564 .176 .9128 GCT 18.299 9 2.033 .633 .7681 Error 478.383 180 3.213 123 a u 2.5+ C C 2 2‘ 3 .- I g 1.5? \ 2 II I. U 0 Ir mum EXPEBIIEIITII. o—om: I—Imz +——+m3 o-——em4 Figure 6. Evaluative Encoding Task Item Error Analysis: Convergent Interaction of Group (1-4) by Condition (Control and Experimental) . 124 intermediate number of mean errors; and Group 4, the fewest mean errors. Simple main effects tests (Table 33) revealed that the Group effect was significant for the Control condition and for the Experimental condition considered independently; the condition effect was significant for Group 3, but not for the other groups. A Neuman-Keuls test of mean differences of error type was performed ( Table 34). Interference errors differed from Order, Substitution, and Omission errors. Omission differed from Substitution and from Order. There were significantly fewer Interference errors than any other type. The mean number of Order and Substitution error did not differ. Syntagnatic Encoging Error Enrterns Table 35 summarizes group means of errors by type on the Syntagmatic Encoding task. The means are displayed in histogram form in Figure 7. Subjects in Group 1 demonstrated a mean of 8.88 errors in the Control condition and a mean of 8.44 errors in the Experimental condition. Group 2 subjects produced a mean of 8.81 Control condition errors and 7.50 mean errors in the IExperimental condition. The mean errors for Group 3 subjects was 7.19 in the Control condition and 6.94 in the 125 Table 33.. Results of a simple main effects test for Group (1-4) at Condition (Control and Experimental) for the Evaluative Encoding Error Analysis. Effect MSn DFn DFe MSe F P ‘ G at CONTROL 10.542 3 103 1.444 7.300 .000 G at EXPERIMENTAL 22.212 3 103 1.444 15.381 .000 c at GROUP 1 .281 1 60 .854 .330 . .568 C at GROUP 2 .000 1 60 .854 .000 1.000 c at GROUP 3 8.508 1 60 .854 9.968 .002 c at GROUP 4 3.125 1 60 .854 3.661 .060 126 Table 34. Results of Neuman-Keuls specific comparison test on pairs of mean slopes for Error Type (Substitution, Onission, Interference, Order) for the Evaluative Encoding Task. A difference between any two means is significant when it exceeds the appropriate critical value (CV) for oc=0.05. Interference Order Substitution Omission Means .5000 1.7812 2.3672 3.1094 Interference * * * Order * * Substitution . * *Denotes a significant difference between pairs of means. 127 40.0 0N.0 40.4 0N. 04.N 04.N 00.0 04.4 04.4 40.N 44.N .:.m 00.04 40.0 04.4 N0. 04.N 4N.N 00.0 N4.4 N0 40.4 44.N :00: 4-4 0=o40 00.0 00.N N0.4 40. 00.4 00. 04.N . 44.4 00. 00.4 00. .0.0 . 00.0 40.0 00.4 04. N4.4 40. 00.0 40. 00. 00.4 00. can: 4 0:O4u 04.4 40.N 00.4 04. 04.N 40.N 00.N 00.4 00. 44.0 4N.N .0.0 04.44 40.0 40.N 0H. 04.4 00.N 04.4 00.4 0N. 00.0 04.4 :06: 0 03040 00.4 N4.0 00.4 40.N 0N.N 00.4 0N.N 44.4 4N.4 00.0 40.N .0.0 40.04 00.4 N4.4 04.4 44.N 04.N 40.0 N0.4 40. 44.0 40.N :06: N 03040 4N.0 NN.N 04.4 00.4 00.0 0N.N 00.4 44.4 00. 00.N 00.N .0.0 40.44 44.0 00.4 00. 44.0 40.N 00.0 04.4 00. N4.0 N4.0 cum: 4 asouo mmmosu 0464444460 s<404 040 424 020 000 04404 040 424 020 000 0684050 44464 :o4u4wcou Hmucos44oaxm :04u4wcou Hohucou .4Hmucos44ocxm 0cm 4044:000 0:0444vcoo xmmb u4umswmuc4m 0cm 44-40 mo:O40 4064030 c4044: 446040 0cm 00:040446404 .co4mm4so .co4usu4400300 06044 4044m mo m¢04404>60 0400cmum 0cm 0:065 0:040 .00 04004 128 _ Experimental Control \\\\\\\\\\\\\\\\\\\ V\\\\\\\\\\V.\.\.\\\\\V\\\V\\\\\V_\\..\\\V\\\\\ INT 0RD Total \\\\\\\\\\\\V\\\. SUB OMS o‘\\\\\\\\\\\. Group 2 Group’l V\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\u \\\\\.\\\ INT ORD Total \\\\.\\\\\\\\\.\. SUB OMS __—_P_ l0 8 6 4 2 0 280 cam: . .VVVV\\\\\\\\.\\. \\\\\\\\V\\\\\\\\\\\,\\\\.\\.\\\\\\\N \\\\\\\\§\\wx§\. . , . \\\\\\\\ Group 3 4F—_____P 0 8 6 .4 2 mcoccm cam—L INT 0RD Total SUB OMS INT 0RD Total OMS SUB Figure 7 . Histogram of mean number of errors by error type (Substitution, Onission, Interference, Order) for the Syntagmatic Encoding Tasks . 129 Experimental condition. Group 4 subjects produced a mean of 3.06 and 3.81 in the Control and Experimental conditions. Substitution and Omission errors were the most frequent for all groups, while Order and Interference errors printwera less frequent. Statisrical Anaiysis, A three-way (Group x Condition x Error Type) repeated-measures ANOVA was performed.- (See Table 36). Significant differences were found for the main effects of Group and Error Type. A post-hoc Neuman-Keuls test of mean differences revealed that Group 4 had significantly fewer errors on the Syntagmatic Encoding task than did Groups 3, 2, and 1. Further, Groups 3, 2 and 1 did not differ from each other (Table 37). A similar post-hoc test of Error Type means (Table 38) revealed fewer Interference errors than Order, Substitution, and Omission. Also significant was fewer Order errors than either Substitutions or Omissions. E 2' t' E 3' E E !! The mean number of errors for subject groups by Error Type and Condition is presented in Table 39. Figure 8 shows similar information in histogram form. Group 1 subjects produced a mean of 7.75 errors in both the Control and Experimental conditions. Subjects in Group 8.50 in the Experimental condition. The mean errors for 130 Table 36. Results of a three-way analysis of variance as a function of Group (1-4), Condition (Control and Experimental) and Type of Error (Substitution, Omissibn, Interference, Order) for the Syntagmatic En- coding Task. Source of Sum of Mean variation Squares df Square F P G 133.258 3 44.419 17.586 .000 Error 151.547 60 2.526 C .781 1 .781 .799 .3750 GC 4.297 3 1.432 1.465 .2332 Error 58.672 60 .978 T 306.148 3 102.049 16.548 .0000 GT 60.086 9 6.676 1.083 .3777 Error 1110.016 180 6.167 CT 18.781 3 6.260 2.562 .0563 GCT 38.703 9 4.300 1.760 .0787 Error 439.766 180 2.443 131 Table 37. Results of Neuman-Keuls specific comparison test on pairs of means for Group (1, 2, 3, 4) for the Syntagmatic Encoding Task Error Analysis. .A difference between any two means is significant when it exceeds the appropriate critical value (CV) for C(==0.05. Group 4 Group 3 Group 2 Group 1 Means .8594 1.7656 2.0391 2.1641 Group 1 * Group 2 * Group 3 * *Denotes a significant difference between pairs of means. 132 Table 38. Results of Neuman-Keuls specific comparison test on pairs of mean SIOpes for Error Type (Substitution, Omission, Interference, Order) for the Syntagmatic Encoding Task. A difference between any two means is significant when it exceeds the appropriate critical value (CV) for oc=0.05. Interference Order Substitution Omission Means .5156 1.5625 2.2031 2.5469 Interference * * * Order * * Substitution *Denotes a significant difference between a pair of means 133 04.0 NN.0 00.4 44.4 00.N 00.4 00.0 40.4 N4. 44.N 44.N .0.0 04.04 00.0 40.4 00. NH.N 00.N 00.0 00.N 0N. 00.4 0N.N :00: 4-4 00040 00.4 N0.N 0N.H 00. 00.N 44.4 04.N 00.N 00. N0.4 0N. .0.0 .40.4 04.0 40.4 00. 0N.4 04.4 00.4 00.N 00. 00.4 00. :00: 4 00040 N0.4 40.N 04.N 04. 40.0 00.4 40.N 00.N 00. 04.N 04.4 .0.0 00.04 04.0 44.N 40. 04.4 0N.N N4.4 00.N 00. NH.N 00.N :00: 0 00040 00.4 00.N 44.4 4N.N 44.N 40.4 00.N 00.4 00.4 N4.N 00.4 .0.0 04.04 00.0 0N.N 0N.4 40.N 04.N 0N0 N0.N 00. 00.N 00.0 :00: - N 00040 N4.4 00.N 44.N 00.4 N0.N 04.N 00.N 00.4 00. 00.0 04.N .0.0 00.04 04.4 04.4 N0. 00.N 00.N 04.4 00.4 00. 44.N 00.N :03; 4 00040 .2404 50 424 020 000 .2494 040 424 020 000 44-: 000040 40404 004440000 404005440040 004440000 4044000 4004000 .000040440404 .00400450 .0044044400000 00044 4044m mo 0004404>00 04000040 0:0 00005.0:040 .44040054400xm 0:0 40440000 0004440000 4004 044050400400 000 44-40 000040 4004000 040443 440040 .00 04004 Control Group 2 ammo. ,...w.,..wwwm..mwmnwan,”AHHuman”umma.”ufifimm0.00.0“,H.,0.43“?0w.m.,m...nww.,...n. A ,. \.\\\\ \ \\\\\\\\.\\.\\\.\\\\\\.\\\\.\\.\\\.\\ V\\\\\.\mvv® INT 0RD Total INT 0RD Total SUB OMS SUB OMS Group 4 134 — Experlmental \\\\\\\\\\\\\\\\\\\ Group I V\\\\\\\\\\\\\\.\\\\\\\\.\\\\\\\\\.\. \\\\\\\\\\\. INT 0RD Total INT 0RD Total \\\\.\\\\\s SUB OMS SUB OMS §VVNVV\\\\\ Group 3 _r_______ 0 . 8 6 4 2 08.5 c002 08.5 c002 Histogram of mean number of errors by error type (Substitution, Onission, Interference, Order) for the Paradig- matic Encoding Task . Figure 8 . 135 Group 3 was 7.12 in the Control condition and 6.75 in the Experimental condition. Group 4 subjects had a mean of 4.06 errors in the Control condition and 3.75 in the Experimental condition. Statistical Analysis, A three-way repeated-measures ANOVA (Groups x Condition x Error Type) revealed statistically significant differences for the main effects of Group and Error Type. Table 40 shows the ANOVA summary table. Condition was not significant; no interactions were significant. A post-hoc Neuman-Keuls test of mean differences (Table 41) demonstrated that Group 4 differed from Groups 3, 2, and 1. Group 3 differed from Group 4, but did not differ from Groups 2 or 1. Group 2 differed from Group 4, but did not differ from Groups 3 and 1. Group 1 differed from Group 4, but not from Groups 2 or 3. A post- hoc test of Error Type means (Table 42) revealed fewer Interference errors than other types. The error type analysis of all of the tasks demonstrates that subjects performed similarly with regard to the type of errors that occurred regardless of the type of stimuli to be remembered. Substitution and Omission errors were the most common, whereas Interference and Order errors did not occur with great frequency. 136 Table 40. Results of a three-way analysis of variance as a function of Group (1-4), Condition (Control and Experimental), and Type of Error (Substitution, Omission, Interference, Order) for the Paradigmatic En- coding Task. Source of Sum of Mean variation df Squares Square F P G ’ 3 94.100 31.367 13.574 .0000 Error 60 138.648 2.311 C 1 .096 .096 .077 .7819 GC 3 .506 .169 .136 .9380 Error 60 74.273 1.238 T 3 277.896 92.632 13.601 .0000 GT 9 47.346 5.261 .772 .6421 Error 180 1225.883 6.810 CT 3 10.990 3.663 1.875 .1354 GCT 9 23.002 2.556 1.308 .2353 Error 180 351.633 1.954 137 Table 41. Results of Neuman-Keuls specific comparison test on pairs of means for Group (1, 2, 3, 4) for the Paradigmatic Encoding Task Error Analysis. .A difference between any two means is significant when it exceeds the appropriate critical value (CV) for<0¢=0.05. - Group 4 Group 3 GrOup 1 Group 2 iMeans .9766 1.7344 1.9375 2.0937 Group 1 * Group 2 * Group 3 * *Denotes a significant difference between pairs of means. 138 Table 42. Results of Neuman-Keuls specific comparison test on pairs of mean slopes for Error Type (Substitution, Omission, Interference, Order) for the Paradigmatic Encoding Task. .A difference between any two means is significant when it exceeds the critical value (CV) for oc=0.05. Interference Omission Order Substitution Means .4141 2.0078 2.1562 2.1641 Interference * * ' * Omission Order Substitution *Denotes a significant difference between.pairs of means. CHAPTER IV DISCUSSION Encoding Task Outcgmeg The purpose of this study was to investigate verbal encoding abilities of children. One goal was to determine whether stimuli constructed as syntagmatic (phrase-like) or paradigmatic (form class consistency) are uniquely encoded by children. Another goal was to determine whether specific groups of children, who produce word associations according to syntagmatic or paradigmatic preference, encode these stimuli. Further, this study sought to determine whether first, second, or fifth grade children encode the evaluative dimension of words from auditory stimuli and whether the syntagmatic or paradigmatic groups were different in encoding of this attribute. The proactive inhibition procedure was used to investigate encoding abilities. Trial level RAST-correct recall was analyzed by subject group and task condition. Two derived measures from trial level data provided indices of encoding. First, a decline in memory performance from Trial 1 to Trial 4 (Control condition) and from Trial 1 to Trial 3 (Experimental condition) indicated the occurrence of proactive inhibition. This happens when subjects attempt to remember successive materials across trials that are 139 140 similar in an attribute. The similarity of stimuli inhibits memory for subsequent information. The second measure was a comparison between the fourth trial of the Control condition and the fourth trial of the Experimental condition.. Recall performance declines in the Control condition if PI occurs. In the Experimental condition, the stimulus on the fourth trial represents a different attribute. An improvement in recall for this trial indicates that encoding has occurred. The amount of recall improvement on this trial as compared to the Control condition was called the RAST release from proactive inhibition (RPI). Evaluative Encoding Analysis of the trial level RAST correct recall on the Evaluative Encoding Task revealed that a group by condition interaction occurred. Specifically, Group 3 (second grade paradigmatic responders) produced higher correct recall in the Experimental condition than in the Control condition. The main effect of group was not significant. This finding is consistent with other research. Kail and Schroll (1974) and Cermak et al. (1972) did not find an age effect for the amount of recall on evaluative encoding. The difference found in correct recall by condition for Group 3, in this study is puzzling because an equal number of subjects in each group received the attributes (positive-P, negative-N) 141 in the four possible arrays: PP, PN, NP, NN for the Experimental and Control conditions. A significant trial effect was found in the Evaluative Encoding Task, however, a Condition-by-Trial interaction did not occur. Trial differences were seen between Trial 4 and the first three trials, but this pattern did not differ for the Experimental versus the Control condition. Examination of group means revealed that each group demonstrated a decline in performance on Trial 2 and variable increases or decreases on Trials 3 and 4. Variability at the level of trials and conditions has been found in other investigations. Hoemann et al. (1974) studied categorical encoding in deaf and normal hearing children. Recall means were variable between trials. Kail and Levine (1976) investigated the encoding of sex-typed children's games. Trial level variability occurred for 7 and 11 year old children. An investigation of taxonomic features in 7 and 11 year olds (Kail and Schroll, 1974) attributed the significant difference between the experimental and control groups in recall to sampling error. Non-systematic memory decline has not been discussed or explained in these studies. Perhaps this is because significantly different results between the Control and Experimental Trial 4 occurred. It is interesting to note, however, that variability among trials is a common occurrence in studies using the PI experimental procedure 142 when control and experimental conditions are based on separate subject groups, as well as in repeated-measures designs where subjects serve as their own controls. Trial level variability affected the overall group means of correct recall, and the derived measures of PI and RPI. Several ad-hoc questions were raised relative to non- significant group outcomes on these measures. What percentage of the number of subjects in each group demonstrated the proactive inhibition phenomena? What percentage evidenced the release effect upon comparison of the Control versus the Experimental conditions? What percentage of subjects demonstrated both phenomena: PI on each condition and RPI? It was not surprising, in view of the trial level variability, that few subjects per group (from 18% to 50%) demonstrated PIE (i.e., PI on both conditions and RPI). It is interesting to note that PI-E was greater than PI-C for each of the subject groups. This may be explained by the variability of decline in Control trials with unexpected instances of increases in correct recall on Trial 4. For the Experimental condition, variability between only two trials (2 and 3) was considered. It may be that there were too few subjects in the sample to demonstrate the task phenomena, or that subjects of these ages are too inconsistent in producing PI and RPI on this sort of task. The findings from the analysis of group performance on 143 the Evaluative Encoding Task indicate that when presented in the auditory modality, this attribute was not encoded by children in grades one, two, and five. This finding supports the Cermak et al. (1972) and the Kail and Schroll (1974) studies which found that children aged 10 years and younger did not improve on the shift trial for the evaluative attribute presented visually. Pender (1969) did find a release effect for second graders, however, stimuli were presented orally. In the Cermak experiment, only the sixth graders demonstrated both an accumulation and release of proactive inhibition. Recall did not decline systematically over trials indicating a lack of buildup of inhibition. Similarly, systematic decline did not occur in this investigation. Variability occurred at the trial level, within the control and experimental conditions: the percentage of subjects who demonstrated RPI in all subject groups was small. The ad-hoc analysis of errors at the trial level was accomplished by classifying each error response by type. It was of interest to investigate whether children of different ages and association style preference differed in how they responded to the task stimuli. The convergent group-by- condition interaction indicated that Group 3 differed in the number of errors on the Control versus the Experimental condition. The group effect was significant for each condition, indicating that the younger children (Groups 1, 144 2, and 3) produced more errors than Group 4. This finding conflicts with the earlier conclusion that age effects did not underlie percent-correct recall. It appears that when errors are considered, age effects become apparent. The types of errors produced by each group were consistent. That is, Omissions and Substitution errors occurred most frequently, whereas Order and Interference errors occurred seldomly. Children tended to admit when they could not recall a stimulus word, rather than to guess. Substitution errors were mostly words that hadpreviously appeared in trials within the task, or words having semantic associations to the stimulus word (e.g., cut-knife). The prevalence of semantic substitutions is interesting. It may suggest that children do indeed encode the stimuli according to particular attributes. The substitution errors that each group of subjects made on the Evaluative Encoding Task is presented in Appendix H. It is interesting to note the large number of syntagmatic substitutions present in Groups 1 (15/23), Group 2 (29/37), and Group 3 (17/23), whereas Group 4 subjects produced fewer syntagmatic errors (8/15). Another interesting observation is the larger number of errors on words within positive triads (15) than negative triads (8) in Group 1. A similar pattern of errors existed for Group 2 with 22 in the positive triads and 15 in negative triads. Group 3 produced more errors on words within negative triads (16) than positive triads (7) and 145 Group 4 produced a similar number of errors within each triad type. The asymmetrical development of contrast adjectives may explain these observations. Positive unmarked forms are acquired earlier than negative marked forms (Clark, 1970; Donaldson and Wales,1970: Klatsky, Clark, and Macken, 1973). The asymmetry is thought to represent the conceptual asymmetry of extension. If the Group 1 and 2 children in this study were asymmetrical in the development of the positive-negative contrasts, then the asymmetrical substitution errors in recall favoring the positive dimension may be due to their ability to extract the attributes from the positive triads with greater ease. Errors then would be more likely to reflect a meaningful association. Perhaps by grade two (Group 3), the negative contrast is more conceptually represented, and by grade 5 (Group 4) it is developed as well as the its positive contrast terms. S nc The main effects of group (1-4), condition (control and experimental) and trial (1-4) were significantly different on the Syntagmatic Encoding Task. Group differences demonstrated that main effects occurred in correct recall. That is, Groups 1, 2, and 3, comprised of first and second graders, did not differ from 146 each other but did produce less correct recall than Group 4 fifth graders. The syntagmatic versus paradigmatic classification of groups did not contribute to differences in correct recall. An increase in recall performance on memory tasks as a function of age is supported by a large body of research confirming that various memory and organizational strategies develop over a period of years and facilitate a child's ability to remember and recall stimuli (Cole, Frankel and Sharp, 1971; Flavell, Beach, & Chinsky, 1966; Hagan & Kingsley, 1968; Reese, 1962; Vaughan, 1968). For the release from PI procedure, older children have recalled more than younger children (Kail & Levine, 1976: Rail & Schroll, 1974) for various attributes, whereas other attributes do not demonstrate an age effect for the amount of recall (Kail & Schroll, 1974). The condition effect, specifically that the Experimental condition produces more correct recall than the Control condition should be due to higher scores in Experimental (shift) Trial 4. Subjects evidenced recall decline in the characteristic fashion from Trial 1 to 2, but correct recall did not systematically decline across Trials 2, 3, and 4. This finding indicates that variability of correct recall occurred for the Syntagmatic Encoding Task as it did in the Evaluative Encoding Task. Trial-level variability again appears to account for 147 the the lack of group differences in PI-C, PI-E, RPI, and PIE. The ad-hoc analysis of the percent of subjects who demonstrated the desired encoding phenomenon was small. PI- C was evident for 37.5% to 50% of subjects across groups. PI-E occurred for 50% to 68.75% of subjects across groups, and RPI ranged from 37.50% to 75% of subjects. PIE ranged from 18.75% to 37.50% of subjects within groups. The Syntagmatic Encoding Task error analysis revealed that the main effects of group and error type were significant. The group differences confirm that age effects underlie the amount of errors in recall. The first and second grade children (Groups 1, 2, and 3) did not differ. Consequently, the syntagmatic-paradigmatic word association preferences of these children did not contribute to different recall performance on this task. The pattern of error types found in the Evaluative Encoding Task also was observed here. That is, Omission and Substitution errors were frequent, whereas Order and Interference errors seldom occurred. Groups 1, 2, and 3 produced a greater number of idiosyncratic responses than did Group 4. A list of the substitution errors for this task by group is in Appendix I. Some of the errors appear to be phonemically based (e.g., pink-sink, line-blind), whereas other errors reflect both syntactic associations (e.g., thin-person, big-fish) and semantic associations (e.g., small-little, short-skinny). 148 E 1' !' E 2' The analysis of trial-level RAST correct recall on the Paradigmatic Encoding Task revealed significant differences as a function of groups and trials. As in the Syntagmatic task, Groups 1, 2, and 3 did not differ from each other, but differed from Group 4. Again, age effects explain this difference. The syntagmatic versus paradigmatic word association preference did not distinguish correct-recall performance. Variability of correct recall across trials occurred where differences between Trial 1 and 2 and 1 and 3 were found, but not between Trials 1 and 4. Again, trial level variability influenced the derived measures PI and RPI. Groups did not differ in the RPI for paradigmatic stimuli and the percentage of subjects demonstrating the encoding phenomena was small. PI-C ranged from 31.25% to 50% of subjects, whereas PI-C occurred for 62,50% to 81.25% of subjects. RPI was demonstrated for 25 to 50% of subjects and PIE occurred for 18.75% to 25% of subjects in groups. Group and Error Type main effects were significant for the Paradigmatic Encoding Task. Groups 1, 2, and 3 each differed from Group 4, but not from each other. This finding confirms that age effects underlie the amount of error in memory recall. As in the other two encoding tasks, 149 Omission and Substitution errors were more frequent than were Order or Interference errors. The substitution errors on this task reflect some syntactic and/or phonemic associations (e.g., star-stand, pink-peaches, blue-book) as well as semantic associations (e.g., blue-yellow, orange- apple). Groups 1, 3, and 4 produced a greater number of substitution errors on the Syntagmatic Encoding Task than on the Paradigmatic Encoding Task. Group 2 produced a similar number of substitutions on these tasks. The Paradigmatic task produced fewer errors overall. Perhaps this is because young children are able to encode taxonomic stimuli which results in better recall. Indeed, most of the substitution errors on this task were color names. Ov l cod T F' ' s ' t'o In this study of verbal encoding of three attributes (Evaluative, Syntagmatic, Paradigmatic), four levels of analysis were accomplished. The first analysis performed on trial-level data revealed that age effects underlie group differences in correct-recall performance for the Syntagmatic and Paradigmatic Tasks but not for the Evaluative Task. The large body of research which supports the finding that recall improves with age makes one wonder why age effects did not occur for the Evaluative Task. The 150 error analysis indeed contradicted this finding. Groups 1, 2, and 3 did differ from Group 4 in the number of errors on the Evaluative Task. It appears then that age effects did occur for each of the three encoding tasks. The second finding that emerged from trial-level analysis for each of the three tasks is that trial variability is problematic in the PI procedure. Inter-trial variability in correct-recall occurred on all of tasks. This at least indicates that the syntagmatic and paradigmatic stimuli did not produce unusual correct recall patterns relative to the evaluative task. The problem created by this sort of variability is that it influences the derived measures. The derived PI measure, for example, is calculated by taking the difference between Control Trials 1 and Trial 4, essentially assuming systematic decline in recall. It fails to recognize the pattern of decline across all four trials. Decline may not occur systematically between Trials 1 and 3, but recall may decline dramatically on Trial 4. Another possibility is that of systematic decline across Trial 1, 2, and 3 followed by increase in recall in Trial 4. Finally, recall may be variable between Trials 1 and 4. Each of these pattern occurred with some frequency at the subject level. This trial level variability appears to be a persistent problem in the PI procedure with children in this study as well as in other studies. 151 The second level of analysis of derived data for each task revealed that groups did not differ in PI or RPI. In other words, the attributes included in this study are not encoded by first, second, or fifth grade children tested with the PI procedure. Further, because children did not demonstrate encoding, no conclusions can be made regarding encoding differences of children based upon word association style preference. The third level of analysis was sought to determine the percentage of subjects in each group (and each task) who evidenced the PI procedure phenomena noted in other studies. The small proportion of the total number of subjects in each group (16) who produced PI, RPI, and PIE may indicate that: (1) larger numbers of subjects are needed to observe the phenomena: (2) the PI procedure is not as sensitive to the encoding of attributes by children as it is with adults: (3) another unknown variable is present in some children which influences their performance. These issues cannot be explored from the data generated by this study. The fourth level of analysis (of error types) revealed that subjects within groups performed similarly across tasks. That is, first and second graders produced similar numbers of errors and all children produced errors which favored Omissions and Substitutions. Children readily admitted when they could not remember stimuli. On the other hand, Substitution errors reflected that children encoded 152 particular attributes of the stimuli. For the most part, Substitutions consisted of words previously heard within preceding trials, or which resembled the phonemic, syntactic, or semantic content of the stimulus word (e.g., line-blind, live-right, blue-yellow). These frequent error types support an argument that encoding of attributes does occur in children of the ages sampled. Anecdotal observations of children's comments indicate they were aware of the attributes of words, and also the structural (syntactic-semantic) aspects of phrase stimuli. Comments from the children indicated awareness of the nature and the attributes of the stimuli. For example, following the Syntagmatic task, some children volunteered that remembering the group of words was easy because they could visualize stimuli (e.g., big, blue, star). Similarly, some children commented after the Paradigmatic task that all of the words were colors (e.g., red, yellow, tan). This four-level analysis procedure of each task provided new and different insights into the PI procedure as an index of encoding abilities in children. The analysis of trial level data, derived dependent variables, the percentage of subjects demonstrating the PI procedure phenomena, and the error analysis more fully describe the performance of groups on the various tasks. Clearly, RPI is only one index of stimulus encoding. Additional analyses revealed (1) a high degree of consistency in performance 153 across tasks, (2) the PI was questionable and therefore problematic, and (3) consistency in the type and patterns of errors across tasks. Francis (1972) hypothesized that the syntagmatic- paradigmatic shift is caused by a reorganization of the lexical filing system whereby isolated words from sentences are compared across related constituents. It was hoped that the PI procedure would first, isolate words according to a specific relationship, either in a syntactic arrangement (i.e., size, color, shape), in a paradigmatic arrangement (i.e., colors), or in an evaluative arrangement (i.e., positive or negative words). Secondly, it was hypothesized that a child's knowledge of these arrangements would influence (decrease) recall across the four trials. Finally, when given the alternative arrangement, children would recognize the difference and recall would improve. Performance was hypothesized to be related to the organizational filing system of words. In this study, using the PI procedure, children did not show strong PI or RPI on each of the three tasks. Certainly one could question the construction of the triads in the syntagmatic task. Are phrase structures encoded on the PI task so that inhibition occurs on subsequent trials? Although the performance data per se do not answer this question, one could argue that children responded to this task similarly to the other tasks. In the evaluative task, the triads of positive or 154 negative words should facilitate PI. In the paradigmatic task, color word triads closely resemble other taxonomies that have been successfully encoded by children. Yet, the children in this study did not evidence PI or RPI with the consistency needed to determine that encoding occurred. No comparisons can be made relative to the groups of children based on age and word association style. It is possible, based upon the evidence from this study, that the PI procedure is not useful in determining encoding of attributes in children. Evidence for encoding exists in young children. Perhaps other paradigms, such as clustering in free recall of lists, or procedures which use transfer of training facilitative instruction, and generalization errors would be more fruitful in investigating cognitive reorganization as related to the syntagmatic-paradigmatic shift. There are studies that have used these paradigms to investigate these relationships in normal and language-learning disabled (LLD) children. It has been suggested that LLD children, as a heterogeneous population do not employ paradigmatic organization when stimuli are presented auditorily (Freston and Drew, 1974: Parker, Preston, and Drew, 1975), but may use visual stimuli more effectively in paradigmatic organization (Suiter and Potter, 1978). Paradigmatic organization has implication for learning in the academic areas including the development of reading. 155 Paradigmatic organization can be taught and appears to be able to be transfered to other tasks (Cartelli, 1978). While LLD children appear to make the syntagmatic- paradigmatic shift (Preston and Drew, 1974), as do their normal counterparts, they may have difficulty in transferring the organizational structure to other tasks. The syntagmatic-paradigmatic shift appears to tell us much more than information of word associations per se. The nature of the word associations may help us to determine the underlying conceptual organization that is present and operational for a particular child. The level of conceptual organization has implications in terms of the educational content and the teaching methods that teachers use in normal early elementary classrooms, as well as having implications for content and methods used in the education of language- learning impaired children. CHAPTER V SUMMARY AND CONCLUSIONS Intreduetien Bagkgrougd Syntagmatic and paradigmatic associative relationships occur in the process of language development as newly acquired knowledge of word use develops from event-based or episodic relationships to hierarchically organized ones. Children produce word associations of a syntagmatic nature with greater frequency in the early years. An increase in paradigmatic associations develops gradually with increases especially noted during the first and second grade period. This change in word association relationships toward increased semantic organization has been attributed to cognitive reorganization (Nelson, 1977). Petrey (1973) suggested that this reorganization occurs as children discover new ways to react to verbal stimuli. Verbal encoding differences may underlie the syntagmatic- paradigmatic shift. The short-term memory proactive inhibition task has been used extensively by Wickens (1974) and others. It is an encoding task that has been used in experiments with children and adults to define attributes of stimuli that are Psychologically real to specific age groups and populations. 156 157 The cognitive reorganization theory that has been proposed as an explanation for the syntagmatic-paradigmatic shift has not been tested empirically. Specifically, verbal encoding abilities of children who respond in a syntagmatic- associative style versus a paradigmatic-associative style have not been investigated. Purpose This study was designed to determine whether syntagmatic and paradigmatic associative relationships are attributes of stimuli that are encoded by children. The aim of the study was to determine if association response type was related to the encoding of the corresponding stimulus type. The primary goals of the study were to determine: 1. Whether children who produce syntagmatic associations also "encode" syntagmatic-associative relationshipsin a PI task; 2. Whether children who produce paradigmatic associations encode paradigmatic-associative relationships and syntagmatic ones: 3. Whether children perform on these tasks in a similar manner to performance on another attribute (evaluative). Secondary goals were established for analyzing trial data and group means from the proactive inhibition task in order to determine: 4. Whether means differed as a function of group, trial, task condition, and task; 158 5. Whether indices of memory inhibition and percent release from memory inhibition could be derived as measures of encoding; 6. Whether error patterns could be identified relative to groups, task conditions, and task types. Experimental Design W Four groups of subjects were tested. Group 1 consisted of 16 first and second grade children with a mean age of 83.4 months who earned a syntagmatic word association score of 60% or greater on the Word Association Task pretest. Group 2 comprised 16 first grade children, with a mean age of 81.4 months who earned a paradigmatic Word Association Task score of 60% or greater on the pretest. Group 3 and Group 4 consisted of 16 children who also produced 60% or greater paradigmatic word associations. The mean age for the Group 3 second graders was 94 months and the mean age for the Group 4 fifth graders was 130.8 months. Stimuli The proactive inhibition task was used as the structure for the three tasks in this experiment. Each task consisted of two conditions (control and experimental) and each condition consisted of four trials. A trial comprised 159 three words that were to be remembered and recalled by subjects. The subjects listened to each set of stimuli presented verbally. Then they were asked fisrst, to produce one immediate repetition, next to perform a letter naming distractor task for 20 seconds, and finally they were given a cue to recall the words. Evaluative Encoding stimuli. The semantic differential ratings from Heise (1965) were used to devise a set of positive words and a set of negative words. The criteria for inclusion in a set was a rating of 1.0 (-1.0) or more on the evaluative dimension and ratings of less than 1.0 on the activity and potency dimensions. Word triads were constructed from the 13 positive words or the 14 negative words. These words were arranged in sets of four to correspond to the four trials needed for the PI task. No word appeared more than once in each set. Subjects were given different sets for the control and experimental conditions. They received either positive-positive, negative-negative, positive-negative, or negative-positive stimuli. On the control condition, four trials were given with one attribute featured. On the experimental condition the first three trials were given as in the control condition. On the fourth "shift" trial the opposite attribute was given. 160 nta t cod 5 ' '. The stimuli for the syntagmatic task comprised sentence-like frames, which included triads consisting of a size adjective (e.g., big), a color adjective (e.g., blue), and a shape noun (e.g., star). Six size words, 10 color words, and six shape words were combined to make triads. Triads were arranged in sets of four for the PI task. No word within each triad was repeated in a set. On the control condition, four trials consisted of syntagmatic stimuli. On the experimental condition, the fourth trial consisted of paradigmatic stimuli. WW1; Color words were used as stimuli for this task. 12 color words were arranged in triads and in sets as described for the other tasks. Color names were not repeated in any triad within a set. On the control condition four trials consisted of paradigmatic stimuli. On the experimental condition the "shift" trial comprised the syntagmatic stimuli described above. Procedures All subjects were selected to meet the following criteria: (1) They were judged as performing at grade level by their teachers and as being free from educationally handicapping conditions, (2) They passed a screening test 161 for rapid letter naming, and (3) The were responsive to the Word Association Task pretest. Grade_lexel_eerf2rmanse- School records were reviewed for each prospective subject. For first graders the results from the DabgIgn_§inggrga;t§n_§greening instrument or the "Curriculum Monitoring System" were reviewed for average range performance. Teachers identified any potential subject who was not performing at grade level as well as those who were suspected of being or certified as a child in need of special education services. Bapig_flaming. During the initial session, potential subjects were asked to name 26 upper case letters of the alphabet from visually presented cards. Letter naming was chosen as the distractor task in the experimental paradigm. This pretest insured that all potential subjects were able to identify all of the letters and ruled out rapid naming difficulties. Wgrd Association Task. Each potential subject received a quasi-randomized word list comprising frequently occurring words. The list included seven nouns, eight verbs, and 15 adjectives (Entwisle, 1966; Thordike and Lorge, 1944), with no more than two adjectives appearing consecutively. Subjects were given each word verbally and 162 they responded to it with a verbal associate. Responses were transcribed and scored to produce a percent of syntagmatic and paradigmatic associates. Subjects were selected for inclusion of the study if they met the criterion defined. This classification was a derived independent variable. e nd Va ' s Two dependent variables were derived from each of the three encoding tasks. The PI task paradigm provided the basis for (1) an index of memory inhibition across successive trials and (2) percent release from memory inhibition when comparing the control versus experimental condition trials. 1111311090 Findings of the study include the following: 1. Children who produce syntagmatic associations do not encode syntagmatic-associative relationships in a proactive inhibition paradigm. 2. Children who produce paradigmatic associations do not evidence encoding of syntagmatic or paradigmatic- associative relationships in a proactive inhibition paradigm. 163 3. Children within the age range of this study do not encode the evaluative dimension. Performance measures were similar to the other tasks which supports the consistency of the findings. 4. Group effects in performance across the three encoding tasks confirmed that age effects were present in the amount of recall that was demonstrated. The younger children in Groups 1, 2, and 3 produced less recall across trials as compared to the older children in Group 4 whose recall was greater. 5. The derived indices of proactive inhibition and release from proactive inhibition were produced along with the overall task effect in less than 38% of subjects within groups. This indicates that the children in this study did not consistently evidence behavior that was to be used as the index of encoding. This experiemntal paradigm may not be practical to use with young children due to the inconsistent occurrence of the desired performance. 6. Error patterns during recall of words were consistent across subject groups and encoding tasks. This finding provided another means for evaluating and comparing children's performance on the PI tasks. The prevelance of substitution errors in all groups suggests that further analysis of these errors may provide information on the way in which the words are being encoding in memory. 164 Qenslueiene The results of this study provide the basis for the following conclusions. 1. Excellent temporal reliability of subject's associative response patterns on the Word Association Task suggests that word association pattern is a quantifiable independent variable. 2. The proactive inhibition task (PI) can be used in a repeated measures design without a significant amount of between-task interference occurring. The percentage of interference errors from stimuli heard in other tasks was very small. 3. The high degree of consistency of groups across PI tasks relative to trial variability, proactive inhibition and release from proactive inhibition, and subjects demonstrating the overall task effect, suggests that the construction of the Syntagmatic and Paradigmatic Tasks was similar to the Evaluation Task which has been used in other experiments. This consistency allowed for comparability between tasks. 4. The four-level analysis approach used in this study may provide important information about children's performance on PI encoding tasks. Analysis procedures that go beyond derived indices that are similar to the ones described in this study, such as the percentage of subjects 165 in groups who demonstrate the effect, and the item error- analysis, appear to be practical for use in further studies involving verbal encoding abilities in children. 5. The findings of this study appear to indicate that associative relationships may not be encoded by children in this age range at least as measured by the PI tasks. Further research on encoding of these or similar attributes might look toward other tasks that purport to measure cognitive organization, such as clustering in free recall, transfer of training, and the use of facilitative instruction. Nouns man flower book river table bird pencil 7 APPENDIX A WORD LIST FOR WORD ASSOCIATION TASK Verbs (8) run sit give begin eat add tell IOVO v s 5 dark sad short hard quiet clean loud fat cold black pretty pleasant long tall bright 166 APPENDIX DECISION RULES FOR CODING WORD ASSOCIATION TASK RESPONSES Paradigmatic responses: When a stimulus word is followed by a semantic associate of the same grammatical class. Stimulus Form Class nggntig_A§§ggiatg Examplg noun antonym man-woman synonym/related pencil-crayon class inclusion table-chair superordinate bird-animal subordinate flower-rose verb antonym run-walk synonym/related begin-start adjective antonym dark-light synonym/related loud-noisy class inclusion black-brown Syntagmatic responsgs: When a stimulus word is followed by a semantic associate of a different grammatical class. Stimulus-Associatie Form glass noun-verb noun-adjective verb-noun verb-preposition verb-adverb adjective-noun 167 Beagles. man-walks pencil-write table-eat man-black pencil-long table-wooden run-bear eat-food give-present run-around sit-down give-up run-slowly move-quickly eat-fast dark-room short-man pretty-flower APPENDIX C SEMANTIC DIFFERENTIAL RATINGS OP STIMULI osi ve Words 0 u ° mg 31211133120 395.1213}! M22003: church 2.40 -0.40 -o.77 1.53 farm 1.40 0.56 0.01 1.50 fix 1.26 0.30 0.97 1.53 fresh 1.60 -0.30 -o.60 1.70 good 1.57 0.36 -0.71 1.77 health 1.29 0.70 0.07 1.47 live 1.29 0.07 0.26 1.50 milk 1.30 -0.96 0.15 1.60 offer 1.00 -0.26 -0.43 1.12 open 1.15 -0.27 0.25 1.20 save 1.11 0.07 -0.19 1.42 uncle 1.36 0.29 -0.23 1.41 true 1.23 0.41 -0.35 2.23 Negative Worgs on in; Egaigativg Qimgnsion: Egrd Evaluation Activity Potengy Polarity bad -3.35 -0.79 0.50 3.40 burn -2.47 0.74 -0.52 2.63 cut -1.00 0.95 -0.67 2.21 fall -2.20 -0.29 -0.96 2.49 fear -1.02 0.07 -0.54 1.90 hate -3.11 0.11 -0.61 3.17 worg kill miss no problem quit shoot trouble worry -3029 -1082 -1.60 -1.83 -20“ -2075 -2.20 -00‘3 -0039 0.30 -0.27 -0086 0.37 SETS OP TRIADS APPENDIX D Unigge Sets of Positive Triads: 80; 1, milk true health good hope live open uncle farm offer save fix Set 4 uncle save church farm open fix good live true 83; 3 live church open health good fix true offer fresh milk save uncle §2§_§ offer church milk true open save health fix live offer health fresh fresh uncle farm 8e 7 uncle milk church fresh save good offer true fix live health Open save offer uncle fix fresh church farm good true live open milk Unigge Sets of Nggativg Triggs: Set 1 out bad burn trouble shoot quit no problem worry miss kill hate Set 4 no quit miss shoot out problem hate burn kill trouble fall bad QC; 2 worry fear miss kill shoot fall burn bad hate problem out quit Set 2 fear fall bad no quit shoot miss out trouble problem burn fear get 5 sheet fall bad problem fear worry trouble hate burn kill no quit fig; 0 fall no hate burn bad problem out trouble miss kill quit worry 170 FOR EVALUATIVE ENCODING TASK 5.; 3 health fix true good farm live church milk save offer fresh open get 6 good church offer save fix live true fresh milk farm uncle health £22.; fall kill hate worry fall cut burn fear worry trouble bad hate QB; 6 fall worry fear burn hate problem kill bad cut quit miss shoot Appendix 3 SETS OP TRIADS POR SYNTAGMATIC ENCODING TASK 0 e 8 ts 8 ts a T ° as; i get 3 set 3 small red circle tall gray star fat blue star short blue square thin brown triangle short green circle fat tan tube small pink tube thin pink square big orange line short white line big white line 8e; 4 §2£_§ 515.5 tall orange triangle short blue tube fat green line fat red square tall white triangle short red star small black line thin tan circle big blue square big green circle small orange square small brown tube get 7 £22.! thin gray triangle big green square small white circle fat pink star tall orange tube short blue circle short black line tall tan triangle 171 APPENDIX F SETS OF TRIADS FOR PARADIGNATIC ENCODING TASK Set i pink blue white tan orange green red gray black rust brown beige Sgt 4 blue red pink rust tan white beige black green gray brown orange set 7 red rust blue tan beige pink gray green white brown black orange t s m}. black green tan white pink brown orange rust red blue beige gray figs 5 tan white gray black red brown pink beige rust orange blue green SOL Q green tan black rust orange blue green white red beige gray brown 172 QC; 3 brown gray red pink beige blue green orange tan black white rust get 9 green red white rust pink gray black beige tan brown blue orange APPENDIX G SCORING OF ITEMS AND TRIALS FOR ENCODING TASKS EVALUATIVE ENCODING-POSITIVE TRIADS (TRIAL 1,2,3, AND 4) Subject ID4 Group I II III IV Task Order 1 2 3 4 5 6 CONTROL CONDITION- (Positive) gm; 120.3523: 11111111223 (2) (2) (2) =(6) 1. milk ___2___ true___2___ health;__2___ ___6___ 2. book____2___ hope (milk) live____2____ ___4___ 3. open (uncle-1) uncle (open-1) farm;__2_ ___4___ 4. offer___2____ save___o___ fix 0 2 EVALUATIVE ENCODING-POSITIVE TRIADS (TRIAL 1,2,3, AND 4) g§PERINBNTAL CONDITIO -(Positive with Negative 4th trial) SET 1 co e Trig; Score (2) (2) (2) =(6) 1. milk ___2___ true___2___ health___2___' ‘___6___ 2. book (live-1) hope__2____ live 0 ___3___ 3. open ___2___ uncle (milk) farm 2 ___4___ 4. cut ___2___ problem;_2__ kill___2____ 6 173 174 EVALUATIVE ENCODING-NEGATIVE TRIADS (TRIALS 1 , 2 , 3 , AND 4) Subject ID# Group I II III IV Task Order 1 2 3 CONTROL CONDITION- (Negative) ‘ 5 6 3314.1 mason moors (2) (2) (2) =(6) 1. cut 2 bad 2_____ burn___2____ 6 2. trouble (bad) shoot (quit-1) quit (shoot-1) 2 3. no 2 problem (trouble-1) worry_2___ ___ ____ 4. miss 0 kill (shoot) hate___o___ o EVALUATIVE ENCODING-NEGATIVE TRIADS (TRIALS 1 , 2 , 3 , AND 4) EXPERIMENTAL QONDITION-(Negative with Positive 4th trial) SET ii 1;gg_ggg;g Tgig; Score (2) (2) (2) =(6) 1. cut 2 bad 2_____ burn___2____ ___6_____ 2. troub1e___2___ shoot__2_____ quit___o____ 4 3 . no 2 problem_o_ worry_o_ _ _ 4. farm;___2_____ save____2____geed 2 6 SYNTAGMATIC ENCODING TRIADS (TRIALS 1,2,3, AND 4) §§LIICII unseen 1:01.020: (2) (2) (2) =(6) 1. small____2 red 2 circle____2____ __6____ 2. short (small) blue 2 square___2_____ __4____ 3. fat______2 tan 2 tube(circle) __4____ 4. big (small) orange 2 line a 2 175 SYNTAGMATIC ENCODING TRIADS (TRIALS 1,2,3,AND 4) gggggiggNT;L_gQ§giT;QN-(Color as 4th trial) £31.11 1&2!_§2210 a o e (2) (2) (2) =(6) 1. small 2 red 2 circle 2 __6____ 2. short 2 blue___2 square (circle) __4____ 3. fat (short) tan____o____ tube 2 __2____ 4. brown 2 pink 2 green 2 __6____ PARADIGMATIC ENCODING TRIADS (TRIALS 1,2,3, AND 4) CO 0 CO I £1.11 1M 11M 8 o (2) (2) (2) =(6) 1. pink 2 blue (white-1) white (blue-1) __2____ 2. tan 2____ orange 2 green 0 __4____ 3. red (black-1) gray _____o black___o____ __1____ 4. rust (red-1) brown_____o______ beige_____o 1 PARADIGNATIC ENCODING TRIADS-COLORS (TRIALS 1,2,3, AND 4) ggPEgINENTAL CONgIziog-(syntagmatic triad on 4th trial) §§LL1 More More (2) (2) (2) =(6) 1. pink;____2___ blue______2______ white___2____, ___ ____ 2. tan (green-1) orange____2______ green (ban) 3 3. red 2____ gray o______ black 2 __4____ 4. tall 2 brown 2 star 2 6 APPENDIX H SUBSTITUTION ERRORS MADE BY SUBJECT GROUPS ON THE EVALUATIVE ENCODING TASK Group Pogitiyg_gtimuli Negative Stimuli 1 farm-farmer kill-shout save-home hate-queen milk-have shoot-gun open-my bad-face true-blue burn-far church-place kill-fat good-break worry-scared hope-big fear-frightened farm-listen‘ uncle-care health-school church-house live-tell offer-made true-cute 2 offer-help bad-cookie health-grandpa burn-bad uncle-grandpa miss-she milk-eggs fall-boy fresh—new quit—stop good-family no-rope live—light worry-weigh fresh-wash kill-anger milk-man fall-angry fix-city fear-anger good-barn bad-mad live-right burn-run offer-helper shoot-gun true-nice fall-eight health-many bad-book health-need true-pay open-there uncle-grandpa offer-read church-wrist save-grandpa 3 health-help kill-son good-life fall-happy true-circle hate-naughty true-help problem-easy live-home fear-tall 176 177 fresh-ask fix-school milk-thin fix-buy open-field save-wide fix-clear true-life live-help save-receive quit-stop miss-stop burn-out problem-blood miss-scared bad-hurt worry-frighten cut-open quit-school worry-scared quit-hit cut-hit fall-hurt fear-afraid fear-fire quit-story trouble-hurry APPENDIX I SUBSTITUTION ERRORS MADE BY GROUPS ON THE SYNTAGMATIC AND PARADIGMATIC ENCODING TASKS Group Syntagmatic iParadiqmatig 1 thin-person pink-peaches tan-people star—stand circle-she orange-shook line-blind short-knife triangle-some gray-use big-bad beige-grass pink-men red-men star-people small-little tan-sun white-pan brown-silver pink-sink 2 circle-purple red-zero blue-sun blue-album line-little brown-bath black-big beige-bathe orange-big white-sun green-ball beige-new fat-bee pink-face pink—call blue-drink white-number orange-apple black-book brown-van red-happy 3 triangle-two white-peach tan-Rick gray-easy thin-easy red-see line-cute star-chair red-away white-read star-winter tan-crayon big-fun blue-yellow blue-seven tube-trunk tan-mom square-bad fat-sun pink-happy star-plaid tall-rectangle green-round short-skinny 178 179 line-slime star-lime big-fish line-gun blue-round square-box yellow-green APPENDIX J STUDENT TASK SUMMARY FORM Name: I.D.# Gender/Race B.D. C.A. Assgcigtigg Qgtg; Set # # Syntagmatic responses: N A NR Total # Paradigmatic responses:N Total Classification: Syntagmatic A NR Paradigmatic Evaluative Encodigg; Stimuli sets: PP NN PN NP Control Condition: Set # 1. Task Order 2. 3. 4. Experimental Condition: Set # 1. Task Order 2. 3. 4. gyntagmgtic Encoding; Control Condition: Set # 1. Task Order 2. 3. 4. Experimental Condition: Set# Task Order 180 181 Parndignatic Encoging: Control Condition: Set # Task Order 1. 2. 3. 4. Experimental Condition: Set # Task Order 1. 2. 3. 4. Response Analysis: Evaluative Encoding Control Experimental Totals substitutions omissions interference order error Syntagmatic Encoding: substitutions omissions interference order error Paradigmatic Encoding: substitutions omissions interference order error APPENDIX K PERCENT OF SUBJECTS IN GROUPS DEMONSTRATING THE PROACTIVE INHIBITION EFFECT An ad-hoc analysis of the percent of subjects in each group who demonstrated the expected phenomenon in the short- term memory procedure used the three encoding tasks is included in the following tables. Trial level RAST data is represented by task and condition in the eight colums on the left (i.e., ECl, EC2,...EE1, EE2...) for the 16 subjects in each group. The five columns on the right represent decision rules performed by Microsoft Excel for determining proactive inhibition by condition: Proactive Inhibition-Control (PI-C) was coded as 1 (true) when Control Trial 1 (C1) was greater than Control Trial 4 (C4). CN was a constant criteria equal to 1. Progctive Innipition-Expgrinentai (PI-E) was coded as 1 when Experimental Trial 1 (E1) was greater than Experimental Trial 2 (E2) or Trial 3 (E3). Pgrgent Bgigase fton innipition (PRI) was coded as 1 when E4 was greater than C4. Eroactivg Innibition Effgct (PIE) was coded as 1 when PI-C and PRI were coded as 1. A more stringent criteria for PIE, adopted for and reported in the manuscript, was coded as 1 when PI-C, PI-E, and PRI were equal to 1. 182 APPENDIX K 183 Evaluative Encoding Task Ad-hoc Analysis of Subjects in Groups (535) demonstrating the Proactive Inhibition Effect. -‘--‘-“““-C-‘ "20 “I. -....... .0...- banana-000000000 : " 8:83 0:20;! (“330333.03 (133%: “”9330“?! IP33" 88 t 8 I I 80 E 0...: V A gnu-OO—OO-DOCOO-OOO 3 g g ‘ .0.-dn‘-..d.....¢i .‘Od.¢¢¢-.d.¢nn..e .O‘...-...ooueanuei I: 95-: 33 " 00.7. :-“‘¢-‘-,n..--“‘ 07... :ue-eceoca..--.o.e one-e. 00.7. :‘.“..‘.‘.-¢“‘. 0'... .00-.40-000-40o—agunoao40.4-0.0-0.0:04004400—0—04-«ncogccooo—oo-ooooo-o—o gOGOOOOOO-DOOOO‘... . OOOC‘COOOOOOOOCOOPfl‘COOdOOGO-O.duo-0.. ‘0‘ 3 I .‘O-I.‘OOOOOOOOCOOEC‘...‘OO-fi...‘aa..£.fi...‘..d....d... as Q . .COCOOO‘OOOOOO-nn' 184 APPENDIX K Syntagmatic Encoding Task Ad-hoc Analysis of Subjects in Groups (%) demonstrating the Proactive Inhibition Effect. I I l I [IUFIII IIHUL IF! 'flfllflfl : 0- v = I v u C i ii : 3 00 e 8 3: : E'iéfiééii‘ z 5: mm 3 3: e 3 3 3 3 3 a: titre! 9: e :e a i i s g a 5 : O : 3 3 C O 3' i 3 3 3 3 i 833 333838;! 3 0 5 s I .§ : . 3333 3333333 0 O O 3 ii 3 3 3 i. 33 £3 33 3 i ' é i i i p a e s b 3 3 3 f 3 L s e 3 . i'. . 3 8 3 3 f 3 eeoe-e-4eoee-e-ee a S :--..¢----4400.--: ”9‘ 60.3. 4 D ; .0¢‘.a.-e..-e-.-OO:OOGOOO-OOOO-OOOO .- 00-40—0-—.o-oo0.03.04.00-4000000000 C -------~N 2 :3 i 3: i: tfiééib3égggeiéié i 333 33333 3 3 3 :3 3 :e 3 i: 83 :3 .. :rooa.‘¢..---o.¢ 3 a g 3 E . 3 5 3 s I L 3 UUUUUHUHUUUHUUHU I : ° ! 9 3 : .‘..-“‘..‘.‘.--.;;d‘fl.““‘d“.‘..§~.“..-II.‘..‘.... 50" ”‘30 ”3‘ 3 3 3 3 8 3 O 3 . . 3 3 3: ~ ‘.‘..o..-..o-o..-. i a i 3 i 3 3 i s i s 5 3 3 3 f s f a i 3 i 5 f 3 £33:- a I f s i s i 3 3 i s i i i 3 3 2 E {tilltiitlzsitéi aneooene-e—eeeeee :‘ddnno..-ueeeueuea..:..as..-e-¢-...-eceace.:.eeceono.on...e..... .‘O-C...‘....‘...: . e' econ-nee.nee...00.30—00-0440040—00-0 fl I-OC-OOOOOOOOOOOO;OO00.0-COOOOCOOOO e "eeeeeeeeeeeeeeee 3 . 3 3 3 3 '0' 3 O 7. 00.3. If... 185 APPENDIX K Paradigmatic Encoding Task Ad-hoc Analysis of Subjects in Groups ('3) demonstrating the Proactive Inhibition Effect. 9:4 PI-I "1 129.9" “129" l I 1 If C!) I! 9191911011 I! 99) I! 91099 d I! "04. (640) (I 999 (ls-G) (69W em rte-1pm 019. P99 '99 '09 '09 999 992 '99 '99 1 use 99 199 99 99.. 9.99 use 19999 91. 9 9 9 9 9 1 19999 199.99 991 99.. 99. 199.99 999 91.. 1 1 9 9 9 1 19999 use 19999 99.. use use 99 -9.99 1 9 9 9 9 1 99. use 9999 99 99 19 99.. 99.. 1 9 1 1 9 1 u. use use 99.. 99 -9.19 99.. us 9 1 9 9 9 1 199.99 199.99 99 91.. 99. use -9.99 -9.99 1 1 9 9 9 1 91. 99 9999 199.99 -9 99 -9 99 -9.99 9.99 9 9 9 9 9 1 99. 199.99 999 use use use 99.. 199.99 1 9 1 1 9 1 99 4.99 use 4.99 99 use 99.. 91. 1 1 1 1 1 1 99. use use use 199.99 use use 99. 1 1 1 1 1 1 u 99 use use 199.99 use 199 99 9 99 use 9 1 9 9 9 1 999 .999 use 99. use 999 99. e1. 9 1 9 9 9 1 999 999 999 99.. 199.99 999 use 19999 9 1 1 9 9 1 99. 999 -999 «.99 use 9999 99... 19999 1 9 1 1 9 1 19999 99.1 99.99 199.99 use use 999 19999 9 1 9 9 9 119999 99 99.. 99.. 199.99 199.99 999 19999 1 1 1 1 1 m 1 9 9 19 9 9 9 m 9 § 99.99 92.99 99.29 92.99 1 9.29 9 19999 use 99 9.99 use use 99 999 1 9 9 9 9 2 21.. 29.99 999 21.. use use 99.. 29. 9 9 9 9 9 2 99 991. 2999 99.. 999.99 use 99 999.99 9 1 1 1 1 9 19999 use 19999 199.99 4.99 use 999 99. 9 9 9 9 9 2 2129 99.. 29.. use use 999 99. 9 1 1 9 9 9 u... use 99 -9.99 use use 99 99.. 1 1 1 1 1 2 99.. ... 21.29 99... 29.29 99 99 999 9 9 9 9 9 9 999 use use -e.99 .. .999 91. 999 9 1 9 9 9 e 999 19999 use 99 use use 91. 19999 9 1 1 9 9 2 99. 99999 99999 99.. ... 99.. 99.. 99999 9 I 1 1 1 9 u. 99 99. -9.99 99 use 99.. -999 1 1 9 9 9 2 999 999 19999 99.. .. 0.99 19999 99 9 1 1 9 9 2 999 19999 use 91.. 4.99 -999 -999 -999 9 9 9 9 9 2 99.. 99 use 99 199.99 use 199.99 199.99 1 1 1 1 1 2 99.. 99. 99.29 99.. ... .999 4.99 99.. 9 1 9 9 9 9 199.99 use 99.. ...199.99 -999 99 ... 1 1 9 9 9 W99 9 11 9 9 9 “I 9. 99.99 99.29 99.29 99.99 99.99 9 99.. 99 99 4.99 -9.99 use 9.99 .999 1 9 9 9 9 9 99.. 998 99.. 99.. 99 ... 99 999.99 9 9 1 9 9 9 19999 199.99 499 99.. 199.99 91.99 99.. u. 1 1 9 9 9 9 99. use use 91.. .. 99 099 .991 1 1 9 . 9 9 9 19999 99.. .999 199.99 199.99 use 19999 99 9 1 9 9 9 9 91.. .999 199.99 99.. 99.. 199.99 99.. 199.99 9 1 1 9 9 9 99. 99 99.. 21.. use use 99 .. 9 9 1 9 9 9 99.. 99 99.. 99.. 199.99 99 199.99 ... 9 1 9 e 9 9 99 99.. 9999 999.99 999.99 ... 99 999.99 9 1 9 9 9 9 99.. use 99.. 99.. use 199.99 999.99 -999 9 9 9 9 9 9 999 99- 99. 99.. 199.99 99 99.. .. 9 1 1 9 9 9 99 use 999 999.99 ... use 99.. 499 9 1 9 9 9 9 19999 use 999 us use use 91. 99 1 1 1 1 1 9 99.. 19999 91.. 199.99 use 99.. 99. 1 1 1 1 1 9 99. 19999 99999 29.29 ... 99 199.. 99.. 9 1 1 1 1 9 199.99 99 use 99.. -9.99 ... 99.. 199.99 1 9 1 1 9 W99 9 99 9 9 9 ms. 99 99 99.99 99.99 99.99 19.99 9 199.99 199.99 199.99 199.99 199.99 199.99 99.. 99.. 9 1 9 9 9 9 199.99 O. 99.99 999.. ... 999.99 99.. 999.99 9 9 9 9 9 9 19999 99 19999 99.. ... 199.99 99. 999.99 1 1 1 1 1 9 19999 99.. 99. 199.99 .. 99 99.. -999 9 1 9 9 9 9 199.99 99.. use 99 199.99 199.99 99 .999 1 1 9 9 9 9 199.99 99.. use 91. 199.99 .. 99 199.99 1 1 1 1 1 9 19999 9.99 19999 199.99 99.. 99.. 499 999.99 9 1 e 9 9 9 19999 19999 19999 999 199.99 199.99 199.99 199.99 1 9 1 1 9 9 19999 99.. 99.99 19999 199.99 99.. 19999 .. 9 1 9 9 9 9 199.99 199.99 9999 199.99 99.. use 19999 19999 9 1 9 9 9 9 19999 199.99 19999 19999 199.99 199.99 19999 199.99 9 9 9 9 9 9 99.. 29.29 19999 99 99 99.. 29. 999.99 9 1 1 9 9 9 99 199.99 use 199.99 99 199.99 999 199.99 9 1 9 9 9 9 99.. 199.99 199.99 199.99 199.99 199.99 199.99 9 1 9 9 e 9 99 99.. ee 91.. ... use 499 99 9 1 1 1 1 4 99.. 999.99 use 999.99 999.99 999.99 99.. 999.99 9 1 9 9 9 en” 9 9 e 1 9 e 9 9 m 9 e 91 .99 91 .99 91 .99 99.99 1 9.99 APPENDIX L INFORMED CONSENT Michigan State University requires that all persons conducting studies obtain a signed consent from the parent or guardian that permits participation of their child in the study. Please read the statements below and sign this form. If you have any questions about this study prior to or after you have consented to allow the participation of your child, please call Kathleen Johnson (353-8780). 1. I, , on (Date) freely and voluntarily consent to allow my child to serve as a subject in a study of relationships between measures of word association and measures of verbal encoding conducted by Kathleen R. Johnson, M.A., working under the supervision of Michael W. Casby, Ph.D. and Leo V. Deal, Ph.D. 2. I understand that this study does not present any risk to my child's physical or psychological well being. 3. I agree to allow the experimenter to review my child's school/educational records. Information concerning his/her educational, intellectual, physical, and emotional development (e.g., achievement test results, report cards, special education testing results) and performance on tasks for this study will remain confidential. I understand that any publication of the results of this study will maintain my child's anonymity. 4. I understand that the procedures for this study are designed to extend our knowledge about learning. The results of this study will not be of direct personal benefit to me or my child. 5. I agree to allow my child to participate in this study freely, without payment to me or from me, and without implication of personal benefit. I understand that I may withdraw consent at any time by phoning or writing and that my child may cease to participate in the study at any time without penalty. 6. I have had an opportunity to question the nature and purpose of the study and I have been provided with a copy of this informed consent form. I understand that upon completion of the study (and at my request) I can obtain additional explanation about the results of the study. 186 REFERENCES Anglin, J.M. (1977). W e ve - ment. New York: Norton. Anderson, S. W., & Beh, W. (1968). The reorganization of verbal meaning in childhood. gogrnel of Verbal Learnigg and Vegbal flenevlor, 7, 1049-1053. Anderson, J. R., & Bower, G. H. (1973). Hu an ssoc' iv memory. Washington, D.C.: V. H. Winston. Anderson, R. C., Spiro, R. J., Montague, W. E., (Eds.) (1977). S o i s o o knowlegge, New Jersey: Lawrence Erlbaum Assoc., Publications. Atkinson, R. C. (1968). Human memory: A proposed system and its control process. In K. W. Spence & F. T. Spence (Eds. ), T e s c o o of ea in motlyetion: Advances in reeearch ang teeoryn (Vol. 2). New York: Academic Press. Atkinson, R. C., Hansen, D. N., & Bernback, H. A. (1964). Short-term memory with young children. Esychonomle eelenee, 1, 255-256. Atkinson, R. C., & Shiffrin, R. M. (1968). Human memory: A proposed system and its control processes. In K. W- Spence 5 J. T. Spence (Eds.), Ih§_EY§QanQQY_2§ learning agd motive§log, (Vol. 2). New York: Academic Press. Bach, M. J., & Underwood, B. J. (1970). Developmental changes in memory attributes. Journa of Edueatlonal Psychology, 61, 292-296. Barclay, J, R., & Reid, M. (1974). Semantic intefration in children's recall of discourse. QeyelepmeQEel Eeyehology, 10, 277-281. Bierwisch, M. (1967). Some semantic universals of German adjectivals. Eeundetlens e: Legggege, ;. l-ge. Bousfield, W. A., Cohen, B. H., & Whitmarsh, G. A. (1958). Associative clustering in the recall of words of 187 188 different taxonomic frequencies of occurrence. Esxshglsgisal_3spsrtsi 4. 38-44- Bower, G. H. (1967). A multi-component theory of the memory trace. In K. W. Spence & J. T. Spence (Eds.), The 2sxsb2l2gx_2f_1earning_and_mgtixatign. (Vol 1)- New York: Academic Press. Bransford, J. D., & Franks, J. J. (1971). Abstraction of linguistic ideas. Cognihive Esyehelogy, 2, 331- 350. Bnain Beyer §§e§ Viey 512+, Version 1.1, 1986. Brown, A. (1977). Development, schooling, and the acquisition of knowledge about knowledge: Comments on chapter 7 by Nelson. In R.C. Anderson, R. J. Spiro, W. E. Montague (Eds.). Sehoeling end Lhe eegnisinion e: hneyledge, New Jersey: Lawrence Erlbaum Assoc., Publications. Brown, R., & Berko, J. (1960). Word association and the acquisition of grammar. thlg_heyelennenh, 31, 1- l4. Bruner, J. (1964). The course of cognitive growth. Anerican Esyehologish, 19, 1-15. Cermak, L. S., Sagotshy, C., & Moshier, C. (1972). Development of the ability to encode within evaluative demensions. J o ent Child Esychology, 13, 210-219. Chomsky, N. (1965). Aspeege of the theezy ef syntex. Cambridge, Massachusetts: M.I.T. Press. Clark, E. V. (1972). On the child's acquisition of antonym pairs in two semantic fields. Jen;ne1_ej_yezhe1 Learning and Verhel Beheyien, 11, 750-758. Clark, H. H. (1970). Word associations and linguistic theorY- In J- Lyons (Ed-). Nsfl_ngrizgn§_in lingnienie_. Baltimore, Md.: Penguin Books, 271- 286. Qlear_Lake_Bs§sargh_AnQYA. Version 1.12. 1987. Clifton, C. (1967). The implications of grammar for word associations. In K. Salzinger & S. Salzinger (Eds.), Reseanch in verbal behevio; end eeme neur2nhxsiglssisal_implisatign§- New York: Academic Press. 189 Cofer, C. N., & Bruce, D. R. (1965). Form-class as the basis for clustering in the recall of nonassociated words. l2urnal_2f_2srbal_Lsarning and.!srbal.§snaxior. 4. 386-389. Cole, M., Frankel, F., 8 Sharp, D. (1971). Development of free recall learning in children. neyelennennel Esxshologx. 4, 109-123. Craik F. M., & Lockhart, R. S. (1972). Levels of processing: A framework for memroy research. Journal of Venbe; Leezning end Verhal Behevior, 11, 671-684. Cramer, P. (1972). A developmental study of errors in memory. Developmentel Esyehology 7, 204-208. Cramer, P. (1973). Evidence for developmental shift in the basis for memory organization. Jen;nel_efi Er2srimsn:al.§hild_gsxsnolosx. 16. 12-22- Deese, J. (1962). Form class and the determinants of word association. on n a r n nd V rba MI 1! 79'840 Deese. J- (1965)- Ihs_sLrusturg_of_assosiation§_in languags_and_tnought- Baltimore. Md-= Johns Hopkins University Press. DiVesta, F. J. (1966). A developmental study of the semantic structures of children. Jonrnel of Esrba1_Lsarning_and_2srbal_nsnaxior. 5. 249-259. DiVesta, F. J. (1966). A normative study of 220 concepts rated on the semantic differential by children in grades 2 through 7. e J a o ene Eélgnglgglp 109: 205-229- Entwisle, D. R. (1966). e d sso at'ons f oun ehilgzen. Baltimore, Md.: Johns Hopkins University Press. Entwisle, D. R., Forsyth, D. F. & Muuss, R. (1964). The syntactic-paradigmatic shift in children's word associations. l22rnal_of_Ysrbal_Learni_g_and Esrbal_Bebaxior. 3. 19-29- Ervin, S. M. (1961). Changes with age in the verbal determinants of word association. Amenieen logrnal_of_£§xshologx. 74. 361-372- 190 Estes, W. K. (1972). An associative basis for coding and organization in memory. In A. W. Melton & E. Martin (Ed8-) Q2ding_2rosesse§_in_fluman_nsmorx (p- 161-215). Washington, D. C.: V. H. Winston & Sons. Felzen, E., & Anisfeld, M. (1970). Semantic and phonetic relations in the false recognition of words by third- and sixth-grade children. DQXQLQDEQDLQL Esxshologx. 3. 163-168- Fillenbaum, S., & Jones, L.V. (1965). Grammatical contingencies in word associations. Jennnel_ef Yerbal_Learning_and_!erbal_flehaxior. 4. 248-255- taneous verbal rehearsal in a memory task as a function of age. Chlld_peyelennen§, 37, 283-299. Francis, H. (1972). Toward an explanation of the syntagmatic-paradigmatic shift. Child Dexelopmsnt. 43: 949-959- Freund, J. S., & Johnson, J. W. (1972). Changes in memroy attribute dominance as a function of age. dengnel of_Edusafional_£§xsholog¥. 63. 386-389. Green, R. F., & Goldfried, M. R. (1965). On the bipolarity of semantic Space. Esxshologisal_nonosrann§i General end Annlled, 79, (No. 6), Whole No. 599, 1-31. Greenberg, J. H. (1966). 'v s s. The Hague: Mouton. Hagan, J. W., Jongeward, R. H. Jr., & Kail, R. V., Jr., (1975). Cognitive perspectives on the development of memory. In Reese, H. W. (Eds.), Adveneee in h'l ev o t a v , (Vol 10). New York: Academic Press. Hagan, J. W., & Kingsley, P. R., (1968). Labeling effects in short-term memory. ghild_neyelennen_,39,113- 121. Halperin, M. (1971). Memory encoding in young children. Unpublished paper based on senior honors project, Northwestern University. Heise, D. R. (1965). Semantic differential profiles for 1000 most frequent English words. Esychologieal Meneggenhe, 79, (8, whole No. 601). 191 Hoemann, H. W., Andrews, C. E., & DeRosa, D. V., (1974). Categorical encoding in short-term memory by deaf and hearing children. den;nel_efi_§neeeh_end Hearing_8esearsn. 17. 426-431- Inhelder. 3.. & Piaget. J- (1964)- Tns_earlx_gr22th_of legie_in_;he_ehlld. London: Routledge & Kegan. Kail, R. V., & Levine, L. (1974). Encoding processes and sex-role preferences. Developmental Reprts Series, No. 48, University of Michigan. Kail, R. V., & Schroll, J. T. (1974). Evaluative and taxonomic encoding in children's memory, dennnel of Exneglnentel thld Esyeholegy, 18, 426-437. Katz, J. J., & Fodor, J. A. (1963). The structure of a semantic theory. Lengnege, 39, 170-210. Kendler, H. H., & Kendler, T. S. (1968). Mediation and Conceptual Behavior. In Spence, K. W., & Spence, J- T- (Eds.) Ine_2szonologz_of_Learning_and Meniyehien, (Vol. 2) New York: Academic Press. Keppel, G., & Underwood, B. J. (1962). Proactive inhibition in short-term retention of single items. dournel of_Yerbal_Lsarning_and_2erbal_hehaxior. 1. 153- 161. Kimble, G. A. (1968). Mediating associations. Jou n of Ernerimsnfal_£sxsholos¥. 76. 263-266. Kroes, W. H. (1973). Conceptual encoding by sense impression. Pencentuel end notor Shills, 37, 432. Libby, W. L., & Kroes, W. H. (1971). Conceptual encoding and concept recall-recovery in children. Child Development, 42, 2089-2093. Loess, H. (1967). Short-term memory, word class and sequence of items. deunnal of Ennerinentel Esyehology, 74, 556-561. Luria,A. R., & Vinogradova, 0. S. (1957). An objective investigation of the dynamics of semantic systems. BriLisn.12grnal__of_2§xsnologx. 50. 89-105- Lyons, J. (1968). Intrednetien he Theorenleal Llnguistics. Cambridge: Cambridge University Press. Mandler, G. (1967). Organization and memory. In K. W. Spence & J. T. Spence (Eds.), The psyehology ef 192 learning_and_motixation (Vol- 1)- New York: Academic Press. Masters, J.C. (1969). Word association and the functional definition of words. Qevelepnental Psychology, 1, 517-519. McNeill, D. (1963). The origin of associations within the same grammatical class. den;nel_efi_ye;hel Learning_and_Yerbal__Benaxior. 2. 250-262- McNeill, D. (1966). A study of word association. degenel 2f_Yer2al_Learning_and_Yerhal_Benaxior. 5. 548- 557. Melton, A. W., & Martin, E. (Eds.), (1972). Qeding Ereeessee in annan Menogy. Washington, D. C.: V. H. Winston & Sons. Mlcrosofh Excel, Version 1.5, 1988. Miller, G. A. (1969). The organization of lexical memory: Are word associations sufficient? In G. A. Talland & N. C. Waugh (Eds.), Ihe_ne§helegy_ef neneny. New York: Academic Press. Moran, L. J. (1966). Generality of word association response sets. Psychologleal Monegnenhs, 80, (4, Whole No.. 612). Nelson, K. (1973). Some evidence for the cognitive primacy of categorization and its functional basis. Merrill_221mer_92arterlx. 19 (1). 21-39- Nelson, K. (1974a.). Concept, word and sentence: Interrelations in acquisition and development. Esxshologisal_8exiew. 81. 267-285- Nelson, K. (1974b.). Variations in children's concepts by age and category- Qnild_nexelopment. 45. 577-584- Nelson, K. (1977). The syntagmatic-paradigmatic shift revisited: A review of research and theory. s o u t' , 84, 1, 93-116. Nelson, K. (1976). Some attributes of adjectives used by young children. Qegnlnlen, 4, 13-30. Nelson, K. (1982). The syntagmatic and paradigmatics of conceptual development. In S. Kucyaj (Ed.) Lengnage Qevelonneng: Ihonghn end Qulture, (Vol. 2), 335-364, New Jersey: Lawrence Erlbaum and Associates. 193 Ory, N. (1968). Transfer as a function of response and stimulus encoding in the learning of paired associates. Unpublished master's thesis, Ohio State University. Osgood, C. E., Suci, G. J., 8 Tannenbaum, P. H. (1957). he Measgremen§_of_ueaning- Urbano. 11: Univ. Illinois Press. Palermo, D. S. (1971). Characteristics of word association responses obtained from children in grades one through four. Dexelonmental_£sxsholosx. 5. 118- 123. Palermo, D. S., 8 Jenkins, J. J. (1964). Weed-Asseclation Normsi__Qrade_§shool_1hr229h_sollese- Minneapolis, Minn.: University of Minnesota Press. Pender, N. A. (1969). A developmental study of conceptual, semantic differential, and acoustical dimensions as encoding categories in short-term memory. Final Report of Project #9-E-070, U. S. Department of Health, Education, and Welfare, Northwestern University. Penk, W. E. (1971). Developmental changes in idiodynamic sets of children's word associations. DQYELQEEQDEQL_E§YQEQLQQYp 5: 55-63- Peterson, L. R., 8 Peterson, M. J. (1959). Short-term retention of individual verbal items. geurnel ef Ernerimental_£§xshologx. 58. 193-198- Piaget, J. (1962). ' o n c i hoo . NewYork: Norton. Reese, H. W. (1975). hdvences in th1d Deyelopnent and Beheyler, Vol. 10, New YOrk: Academic Press. Reutener, D. (1969). Background, symbolic and class shift in short-term memory. Unpublished doctoral dissertation, Ohio State University. Riegel, K. F. (1970). The language acquisition process: A reinterpretation of related research findings. In L. R. Goulet 8 P. B. Baltes (eds.), Lifeeenen QEYQLQEEQDLQL_ S O ' e C 80 - New York: Wiley. 194 Sapir, E. (1944). A study in semantics, os f , 11, 93-116. Sharp, D., 8 Cole, M. (1972). Patterns of responding in the word associations of West African children. ghlld Dexelonment. 43. 55-65- Sharp, M. J.,D., 8 Gollin, E. S. (1985). Memory and the syntagmatic-paradigmatic shift: A developmental study of priming effects. Bellegln_ef_§he Esxshsnomis_§22ietx. 23 (2). 95-97- Spence, K. W., 8 Spence, J. T. (Eds.) (1968). The Esxsho1o9Y_of_Learnins_and_notization (Vol- 2)- New York: Academic Press. Stolz, W. S., 8 Tiffany, J. (1972). The production of "child-like" word associations by adults to unfamiliar adjectives. o b i and_Yerhal_Behaxior. 11. 38-46- Studebaker, G. A. (1985). A "Rationalized" arcsine transform. Beseereh, 28, 455-462. Thorndike, E. L., 8 Lorge, I. (1944). e c e word hoeh of 39,990 words, New York: Columbia University, Teachers College. Tulving, E. (1968). Theoretical issues in free recall. In T. R. Dixon 8 D. L. Horton (Eds.), Venhel behevlor and_seneral_behaxior_theorxl_ Englewood Cliffs. New Jersey: Prentice-Hall. Turvey, M. T., 8 Egan, H. (1969). Contextual changes and release from proactive interference in short-term verbal memory. 0 1 n a Esxsholosx. 81. 396-397- Underwood, B. J. (1965). False recognition produced by implicit verbal responses, den;nel_ef Experimental_£§xsholos¥. 70. 122-129. Vendler, z. (1968). v s m' 2 ion. The Hague: Mouton. Weingarten, M., 8 Anisfeld, M. (1981). The syntagmatic- paradigmatic shift in dichotic listening. degenel of_Er2erimental.§hild_£§xshologx. 31. 271-278- White, H. (1984). The syntagmatic-paradigmatic shift and long-term memory activation. The donrnal ef c s o 4 6 -5 . 195 Wickens, D. D. (1970). Encoding categories of words: An empirical approach to meaning. Eeyehelegieel Review, 77, 1-15. Wickens, D. D. (1972). Characteristics of word encoding. In A. W. Melton 8 E. Martin (Eds.), Qodlng grocesses in heman Menony. Washington, D.C.: Winston 8 Sons. Wickens, D. D. Born, D. G., 8 Allen, C. K. (1963). Proactive inhibition and item similarity in short- term memory. ou o Ve a n and Ve bal Behavior, 2, 440-445. Wickens, D. D., 8 Clark, S. (1968). Osgood dimensions as an encoding class in short-term memory. Jounnal of Experlmental Psychology, 78, (vol. 4), 580-584. Winokur, S. C., 8 Tweney, R. D. (1974). The paradigmatic shift and development of the subjective lexicon. Unpublished manuscript, Bowling Green State University. Wittlinger, R. P. (1967). Phasic arousal in short-term memory. Unpublished doctoral dissertation, Ohio State University. Woodward, H., 8 Lowell, F. (1916). Children's association frequency tables. Esycheloglcal Monognaphs, 22, (5, Whole No. 97). "Tll'iiifljiifilflflilflfliiii