ABSTRACT IDENTIFICATION AND DESCRIPTION OF THE INTRINSIC SOURCES OF INDIVIDUAL DIFFERENCES IN CONCEPT LEARNING By William Leslie Logan The present study attempted to identify and describe individual differences in concept learning. It was hypothesized that if a signifi- cant portion of the reliable individual variation in a complex learning situation could be identified and described, it would be reasonable to eXpect that subsequent predictions could be made about the conditions necessary to maximize efficiency in learning for various categories of subjects. Scores were obtained for 39 subjects (19 male, 20 female) on 11 reference tests representing measures of 14 intrinsic individual difference variables. The scores were factored using the method of principle-axes solution, and rotated to a varimax solution. The factor analysis yielded five factors interpreted as l) strength of the initial registration of the stimulus trace, 2) immediate visual memory, 3) field independence, 4) susceptibility to response competition, and 5) neuroti- cism. Normalized factor scores for each subject were obtained from the varimax solution. In addition, each subject was tested every week for five weeks with six types of conceptual learning conditions. The de- pendent measures obtained from this testing series were 1) memory errors, 2) perceptual-inference errors, and 3) a measure of rule acquisition. Logan The second phase of the study proposed to utilize the factor scores as predictor variables in a series of multiple regression analyses using the dependent measures of the concept learning task as criterion vari- ables. Because of the lack of reliability manifested by the dependent measures the second phase of the study was not carried out. It was evident from the data that the within subject variance, respective to a problem condition, was approximating the between subject variance. It was suggested that this unexpected relationship was due to a differ- ential learning phenomenon that manifested itself as unstable performance characteristics on the part of the subjects. IDENTIFICATION AND DESCRIPTION OF THE INTRINSIC SOURCES OF INDIVIDUAL DIFFERENCES IN CONCEPT LEARNING by William Leslie Logan A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Counseling, Personal Services and Educational Psychology 1973 The research reported herein was performed pursuant to a grant awarded to the author by Region VIII, National Center for Education Research and Development, Office of Education, U. S. Department of Health, Education and Welfare. Project No. l-H-OOZ. ii ACKNOWLEDGMENTS This thesis is indebted to many people. My advisor, Dr. Joe L. Byers, was not only of inestimable help in the conception and planning of the work but continues to provide help and encouragement throughout. I am very grateful to Dr. William Connett of the University of Northern Colorado, and Drs. Lawrence Lezotte and John Schweitzer of Michigan State University for their suggestions and aid. Analysis of the data in its present form was made possible only through their kind assistance. iii TABLE OF CONTENTS LIST OF TABLES ....0....0.00000000000000000COOOOO0.0....0.00.0.0... Vii CHAPTER I STATEMENT OF PROBLmOOOOOOOOOOOOO ...... ......OOOOOOOOOOOOO 1 General Background of the Present Study ............... 1 BaSi-c PrOblem.....OOOOOOOOOOOOOI0.0.0.000......... 2 Categories of IDs in learning ................. 4 Definition of concept learning: classification scheme ..................... 5 Approach of the Present Study ......................... 7 Procedural Variables .. ...... . ........ . ........... . 7 Stage of Learning .......... ...... ...... ......... .. 8 Overview .......................................... 10 II REVIEW OF RELATED RESEARCH ................................ 11 Selection of Tests ... ......... ........................ 15 Group I .... ...... ......... ................ . ..... .. 15 Raven Progressive Matrices ..... .......... ..... 15 Stroop Test ....... ..... ....................... 15 Eysenck Personality Inventory (EPI) . ..... ..... l6 Witkins Test of Field Independence: Embedded Figures Test (EFT) ............... 16 Kagan's Matching Familiar Figures Test (MFF) .. 16 Group II .................... ...... ................ 17 Immediate Digit Span Memory (IDs) . ....... ..... l7 Delayed Digit Span Memory (DDS) ............... l7 Proactive Inhibition of Digit Span (PI) ....... 18 Retroactive Inhibition of Digit Span (RI) ..... 18 Immediate Visual Memory (VMI) ................. l9 Delayed Visual Memory (VMD) .. ....... .......... 19 iv CHAPTER Page III METHOD AND PROCEDURES . ............ . ...... . ........... ..... 22 Selection of Subjects .... ......... . . ......... ........ 22 Administration and Scoring of Reference Tests . ....... . 22 GroupI000000000000000 00000000000000 0.00 ........ 00 22 Raven Progressive Matrices (RPM) .. ............ 22 Stroop TeSt 0 0 0 O 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 O 0 0 0 0 0 0 23 Eysenck Personality Inventory (EPI) ........... 25 Witkins Test of Field Independence: Embedded Figures Test (EFT) ....... ....... 26 Kagan' 3 Matching Familiar Figures (MFF) ....... 27 Group II ......................................... . 28 Immediate Digit Span (IDS) and Delayed Digit Span (DDS) . ................. 28 Retroactive Inhibition (RI) and Proactive Inhibition (PI) . ..... ... ....... . 29 Immediate Visual Memory (VMI) . ..... ........... 31 Delayed Visual Memory (VMD) ................... 33 Laboratory Tasks . ............. . ................... .... 33 Procedural Variables . ........... ...... ......... ... 33 Stimulus Availability (SA) .................... 33 Concept Complexity (CC) ......... ......... ..... 34 Apparatus 0 0 0 0 0 0 0 0 O 0 0 0 0 0 0 0 0 ..... 0 0 0 0 O 0 0 0 0 0 0 0 0 0 0 0 0 0 0 34 Stimulus . .................................. ....... 34 concept PrOblemS 0 0 C 0 O C 0 O 0 0 0 0 O 0 0 0 0 0 0 O 0 0 C 0 O 0 0 0 0 0 0 0 O 0 0 0 0 0 35 Procedure 0 ..... 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 36 Procedural Modifications .......... ....... .-....... 37 Scoring .. .......... . ............ . ................ . 38 IV RESULTS 000.00 ..... 0000000000 ..... 000 00000000000000000 0.0.0 39 sanlple validity 0 000000000 O O 0 0 0 0 O 0 O 0 0 O 0 0 0 0 0000000 0 0 0 0 0 0 39 Eysenck Personality Inventory .. .................. . 39 Raven Progressive Matrices . .................... ... 41 Task Validity . ...................................... . 42 Reliability of Dependent Measures ..... ................ 43 Factor Analysis ....................................... 44 Rationale . ...................................... .. 44 Factor Analysis Procedure . ....................... . 45 V CHAPTER Page Factor Analysis of Reference Tests . ...... ......... 45 Varimax Rotation Factors ... ............ . ......... . 45 V DISCUSSION AND SUMMARY .......... .............. ............ 51 Discussion . ................. . ....... ........ ....... ... 51 Summary . ............................................. . 55 REFERENCES 0.0.0.0000. 00000 0.00.0.0... 0000000 0 00000000000 00 000000 0. 57 APPENDIX A: SUMMARY OF THE ANALYSIS OF VARIANCE OF THE DEPENDENT MEASURES, CONCEPT COMPLEXITY, STIMULUS AVAILABILITY AND RULE ACQUISITION ............ 60 vi Table LIST OF TABLES Page Means, SDs and Correlations Among 14 Reference Variables . .............................. ..... 40 Summarization of the Data Obtained by Jensen (1964) ....... 41 Reliability Coefficients of Dependent Measures Under All Conditions of the Conceptual Task . .......... .... 44 Principal Components and Communalities (hz) of the 14 Reference Variables ................................ 46 Varimax Rotation Solution ........ ....................... .. 47 vii CHAPTER I STATEMENT OF PROBLEM General Background of the Present Study An analysis of the experimental literature in the area of conceptual learning points up two (2) basic methodologies used in the investigation of learning processes as they relate to conceptual behavior. The first of these methodologies falls into what might be broadly termed "ecological conditions"; the second deals with the characteristics of the performing subject (S) as he solves a conceptual problem. Studies investigating the effects of various ecological conditions on concept learning have generally been concerned with the manipulation of such factors as: 1) the utilization of positive instances (Friebergs and Tulving, 1961); 2) the number of relevant or irrelevant dimensions presented to a subject in a concept identification task (Walker and Bourne, 1961); 3) the logical structure of concepts (Haygood and Bourne, 1965); and other functional relationships identified as task variables. The methodologies followed by experimental psychologists utiliz- ing the characteristics of the performing Spas experimental variables has produced a reasonable amount of useful information (Bourne, Goldstein, and Link, 1964; Dickstein, 1968). However, an examination of the research literature concerned with these "organismic" or "subject variables" suggests two (2) inadequacies. In the search for general behavioral laws, investigators in comparative studies have been 1 2 preoccupied with means of avoiding error variance and have tended by design to randomize out the effect of individual differences (IDs) rather than studying their influence on behavior. In addition, the research has been relatively unsystematic and therefore has tended not to be programmatic in design. Basic Problem Recently, a high degree of interest has been shown in the individualization of instruction. If the goal of education is to bring each student to a common level of mastery in cognitive or instructional tasks, then a major requirement would be the adaptation of the mode and method of instruction to individual variation. It would seem to follow that the success of an individualized instructional program would be dependent upon a complete understanding of the IDs contributing to the learning of complex tasks. The present state of knowledge is such that given sufficient data, experimental or instructional psychologists enjoy a reasonable degree of success in the prediction and control of behavior. For example, if a group of §§ are presented a series of digits and instructed to recall them in the exact serial order that they were presented, it is safe to predict that they will recall the first and last parts of the series better than the central portion of the series. However, in relation to conceptual abilities and the learning of complex tasks, (i.e., conceptual learning) the degree of predictive success that has been demonstrated is in advance of the understanding of the underlying IDs that produce the behavior. It would seem reasonable to assume, that if the knowledge of these IDs as they relate to conceptual learning could be acquired it would provide " . . . both 3 a source of hypotheses about the nature of learning processes and a means of testing certain deductions from theoretical formulations (Jensen, 1966, p. 142)." In a recent review, Glaser (1967), supporting this line of thought, has pointed out the importance of developing techniques which will allow the identification of individual learning functions. In addition he has de-emphasized the practice of averaging data as a method of demonstrating learning in performance terms. Given the identification of these individual learning functions, it would seem to follow that subject variables could then be defined as initial state measures. This concept of a behavioral baseline is prevalent in the area of physiological psychology; Skinner, in studies of individual behavior, defines this procedure as the description of an "Operant level." In either case, qualifying conditions are being placed on the general behavioral laws. In consideration of this position, the direction of future research should be towards the identification of differences in performance that are directly dependent upon IDs in learning processes. If these performance measures can be identified, then a "Taxonomy of Processes" (Melton, 1967) as they relate to cognitive tasks can be develOped. These individual performance functions, as initial state measures, must be considered as variables in the learning process. When the point is reached at Which these variables can be included as given in the experimental paradigm, it will, in effect, increase the level of information obtained from the data and consequently the generalizability of the results. 4 Categories of IDs in learning. Jensen (1964) in studies of IDs has introduced two useful concepts into the literature: 1) intrinsic and extrinsic sources of IDs, and 2) phenotypic and genotypic variables. In conceptualizing the importance of a) the production of individual functions and b) a taxonomy of processes, the preceding section has been an attempt to develOp the initial structure of a model in terms of the IDs which are inherent in the learning process. Jensen terms these IDs which cannot exist as functions independent of learning, as "intrinsic individual differences." The differences " . . . consist of intersubject variability in the learning process (Jensen, 1967, p. 122)." On the other hand, "Extrinsic individual differences" are variables which may influence performance on a learning task. These variables that may be identified as extrinsic sources are sex, IQ and personality. Phenotypic variables are defined in terms of the ecological conditions, i.e., task variables. Operationally a phenotypic source of variance is defined as any significant interaction between the gs and a task variable. A genotype is defined as the underlying process variable that is the causal factor for the pattern of relationships between the phenotypic variables. For example, the process of "retroactive inhibition" is by definition an intrinsic source of IDs. The construct that is posited as an explanation of this process is interference with the consolidation of the stimulus trace and is considered to be the underlying genotype. In the "retroactive inhibition" paradigm (i.e., learn A, learn B, test A) the observed behavioral measure taken during the testing of material A, is considered to be the phenotypic aspect of the "retroactive inhibition" process. An‘g who scores high on the testing of material A is said to have little susceptibility to retro- active inhibition and therefore is resistant to interference effecting 5 the consolidation of the stimulus trace. Though extrinsic IDs seem to contribute to the between-subjects variation in learning, within the constructs of the model as it is defined by Jensen (1964, 1967), the majority of the phenotypic variation of IDs in learning will ultimately be explained in terms of genotypic-intrinsic factors. Common to both concepts, as they are presented above, is the basic distinction between task and process. At the present, there seems to be a concensus among a number of experimental psychologists that a more fruitful approach to the understanding of IDs can be found in the study of process variables. Melton (1967) makes this point in saying, that "what is necessary is that we frame our hypotheses about individual differences variables in terms of the process constructs of contemporary theories of learning and performance Ep. 239]." In accordance with this the present study is designed to deter- mine some of the characteristics of the performing § (i.e., IDs) in terms of the relative contribution of intrinsic factors to the variation found in the type of performance error the § commits while performing the task, and the decision processes of the performing §§ as they relate to conceptual learning and relevant process variables. Definition of concept learning: classification scheme. Before investigating the effects of intrinsic IDs on conceptual behavior we must in some systematic fashion define the specific learning behaviors with which we will be concerned. Concept learning has been chosen as the task in this study for the following considerations. First, the task structures are similar in form to school learning and natural learning situations. Second, there is a large body of definitive literature that is well documented in 6 relation to the relevant task and subject variables. Third, the designs of concept learning tasks lend themselves more to eXperimental control than other complex learning behaviors. Lastly, concept learning is particularly well suited to the goal of identifying intrinsic individual differences as subjects can be presented with a series of many different concept learning tasks whose relationship to one another may be clearly specified. The area of conceptual learning manifests distinctions in phenomena that defines a number of specific classes and/or levels of behavior. However, a taxonomic analysis of conceptual learning has been fully explicated elsewhere (Kendler, 1964; Jensen, 1966; Haygood and Bourne, 1965) and the classification scheme will be concerned only with those classes of behavior that are immediately related to the problem under study. The first distinction that we will be concerned with relates to the type of conceptual task. Basic to all learning is the process of simple discrimination; within the taxonomy of conceptual behavior a type of concept learning is found which depends largely upon discrimina- tion learning. The laboratory learning task designed to investigate this behavior requires the § to divide a series of complex stimuli into two mutually exclusive sets labeled positives and negatives. Positives are classified as being exemplars of a concept, negatives as non- exemplars of the same concept; the relevant attributes and the relational rule are defined by the experimenter (E). The process of definition is fully explained in Procedures: section (IV C-3). The second distinction, related to the conceptual task, is made between "attribute identifi- cation" and "rule learning." Haygood and Bourne (1965) make this 7 distinction in relation to the task requirements of the learning condition. In the former the S is given the relational rule with the task instructions and must discover the relevant attributes. In rule identification the S is told what the relevant attributes of the concept are and must discover and verbalize the rule of relationship between the attributes. If sources of intrinsic IDs can be found in this simplistic form of concept learning, research should then lead to the discovery of their effects in more complex forms of learning behavior. Approach of the Present Study The present study is an attempt to identify and describe intrinsic sources of IDs in concept learning. As mentioned above there is a consensus that the most fruitful approach to the understanding of IDs will be found in the study of process variables. More specifically, it is highly probable that the greatest source of ID variance in learn- ing can be found in the interaction between the process variables and the procedural variables of the learning condition. Procedural Variables Procedural variables are a class of task variables dealing with the procedure of the learning condition, excluding the content and the sensory modality of the presentation. This particular class of variables would include such factors as CS-UCS interval, pacing, distribution of practice, type and amount of stimulus available to the S from previous events, task complexity, and stage of learning. The last three variables are of particular importance to this study. 8 The importance of the type and amount of past information as well as task complexity as procedural variables is well documented in the literature (Walker and Bourne, 1961; Bourne, et a1., 1964) and are used as independent variables in this study. Stage of Learning Previous research in concept learning has involved the S at most one to three hours in a laboratory task. It is well known that if a learner is presented with a series of related learning tasks, his performance, in addition to showing a greater stability in the final stages of the series, is more efficient than in the initial stages of learning. Therefore, it would seem that previous research is inadequate insofar as it has been investigating basically the initial stages of learning and not the more stable behavior found in the final stages of practice. In addition, Fleishman (1962, 1967) in developing a taxonomy of IDs as they relate to perceptual-motor skills, has found changes in the factorial composition of the IDs contributing to the performance, at different stages of learning. The changes are systematic and do stabilize in the later stages of practice. In consideration of these factors the study follows a program suggested by Jensen (1964, 1967). The first step in a systematic approach to the identification of these intrinsic ID sources is to limit the area of research to one type of learning. It is assumed that by limiting the focus of study to a single type of learning and manipulating relevant process variables within this narrow class of behavior, it should make the interpretation of any evolving structure a simple process. 9 Within the design of this study two additional methodological or procedural innovations have been added. The first departure from earlier procedures will be the use of long-term experimentation. The §§ will be tested on the laboratory tasks, three hours a week for a period of approximately five weeks. This extended period of testing allowed the investigator to collect data on the reliable performance measures found in the final stages of practice. The second procedural change will be in the methodology used in the selection or develOpment of the tests for the reference battery. Earlier studies have attempted to explain individual variation in complex tasks through the use of psychometric tests designed as indexes of general ability (e.g., general reasoning, induction, deduction, and verbal comprehension). The approach of this study is much simpler in structure, the emphasis being on the identification of intrinsic sources of IDs. As stated above, it is expected that the greatest source of ID variance will be found in the interaction between the process variables and the procedural variables. In consideration of this, two types of test instruments were utilized in constructing the reference battery. Wherever possible the reference tests were selected from the methodology of established studies in literature, the procedures of which were designed to assess process functions. The second type of test included in the reference battery was selected from standardized psychometric instruments. The criterion of selection for an instrument is to be its relationship to the relevant process variables, and its factorial simplicity. 10 Overview The intent of the study is the identification and description of intrinsic sources of IDs in concept learning. The general structure is atheoretical in concept but the design of the study is systematic in its approach to the problem. The reference battery is comprised of two types of test instruments. In addition to standardized tests selected for their factorial simplicity, the instruments were selected with consideration given to tested procedures found in the literature. The §§ were tested on various forms of concept learning materials, similar to those used by Bruner, Goodnow, and Austin (1956). The procedural variables that have been selected are the difficulty of the task (concept complexity), and the type and amount of stimuli available to the S from previous events (memory). CHAPTER II REVIEW OF RELATED RESEARCH At present, many psychologists have basically rejected the definition of intelligence as being a unitary learning ability. While not discounting the concept of a general learning factor (g), using the process of factor analysis, American psychologists by tradition tend toward the identification of specific or group factors. Woodrow (1946) began a trend in the research of IDs that has been labeled the psychometric approach. The goal of this method has been the identification of IDs in learning in terms of group factors or abilities as they are defined through the factor analysis of psycho- metric reference tests. Examples of such a battery would be the "Kit of Reference Test for Cognitive Factors" (French, Ekstrom, and Price, 1963) or Thurstone's tests of Primary Mental Abilities (PMA). Typical factors included in this type of reference battery are verbal fluency, perceptual speed, general reasoning, numerical ability, etc. A number of criticisms can be directed of this general approach. With few exceptions, studies of IDs though quite competent in design in relation to the psychometric method have encountered the same problems. The difficulties most consistently found were the following: 1. The reference test as a measure of an ability factor does not present in a simple form the initial state of the S. A great deal of transfer from prior learning is involved in this type of assessment. 11 12 2. In many instances in terms of the processes and procedures involved in the reference test, the assessment is more complex than the task. In terms of scientific explanation, it would seem more logical, in relation to the initial states of the learner, that Specific factors relating to performance are more basic and therefore necessary to the understanding of the learning process. It would seem more likely that the aptitude or ability could be defined by the interaction of the intrinsic sources of IDs. Unless the interaction of these specific factors can be explicated, using an aptitude measure to explain IDs in learning is using one incomplete construct to explain another. 3. When the reference battery and the learning tasks have been factor analyzed little or no common variance has been evidenced between the two measures. The usual result was two distinct factor types, one for the reference tests and one for the learning tasks. A number of studies have been conducted to investigate IDs in learning. The review will be concerned only with those correlational studies that directly deal with the cognitive factors of learning, as Opposed to psychomotor learning. Illustrative of such investigations are Stake (1958); Allison (1960); Duncanson (1964); Lemke, Klausmeier, and Harris (1967); and Dunham, Guilford, and Hoepfner (1966). Stake (1958) investigated the relationships between learning tasks, ability factors, and scholastic achievement. The learning tasks were categorized as to their verbal or non-verbal content, and as to whether rote or relational learning was required. The instruments used in the reference battery are subject to the first criticism in that the assessment involved a good deal of transfer from prior learning ex- perience. In fact the criterion for selection was that they parallel 13 some scholastic learning experience. In addition, the factor analysis yielded two factor groupings: (1) reference and achievement factors, and (2) learning factors. The intercorrelations between factors within these two groupings were negligible, as well as the intercorrelation of the learning factors. Lastly, the majority of what might be labeled as Specific factors (e.g., verbal reasoning) seem to be complex in them- selves and are in need of eXplication. Allison (1960) administered 13 learning tasks which were repre- sentative of three types of learning: rote, conceptual and motor learning. Thirty-nine reference measures of aptitude and achievement were used by Allison in an attempt to assess any relationships between the learning process and human abilities. As with Stake (1958) the instruments used by Allison in the reference battery involve a great deal of transfer from prior learning. The psychological processes and/or procedures involved in the reference assessment (e.g., deduction, verbal knowledge) are as complex as the learning tasks. The investigation carried out by Allison yielded factors that were common to both factor domains, the reference and learning. Nonetheless, like Stake (1958), the factors interpreted by Allison such as "Spatial Conceptual Learning," or an interbattery factor "Conceptual Process Factor," did not yield much information about the learning process. Duncanson (1964) investigated the interrelationships of ability and learning measures by administering a battery of ability tests in conjunction with learning tasks. The tasks included three types of learning: paired-associate, rote-memory, and concept-formation. Following the psychometric method the reference tests were taken from a battery of available instruments (French, et a1., 1963). The ability 14 measures and learning scores were then combined and the resulting correlation matrix factor analyzed. Seven factors were extracted and then rotated to an equimax solution. Three factors were common to both the learning and ability measures, verbal ability, reasoning ability and rote-memory ability; three factors were restricted to the learning measures, concept formation, verbal learning, and nonverbal learning; and one factor was restricted to the ability measures, speed. Though three factors are common to both domains, little of the variance in the learning tasks is explained by the reference tests. Lemke, Klausmeier, and Harris (1967), following the psychometric method encountered the same difficulties. The selected 16 psychometric tests represented eight ability factors. Scores obtained from each of the gs on these instruments were intercorrelated with 18 scores obtained from the same §§ on series information-processing (IP) and concept attainment (CA) tasks and the resulting matrix factor analyzed. Low correlations were found between the CA factors, IP factors and the set of cognitive abilities. The CA and IP tasks were seen as relatively distinct activities. As with the other investigations little of the variance in the learning tasks is accountable for by the ability measures. The investigation by Dunham, Guilford, and Hoepfner (1966) though similar to the other studies is quite different in procedure. The study was carried out within the structure-of-intellect model (SI), the selection of tests being made in relation to this systematic theory. We find common agreement between this study and the others in that factors were found that were common to the learning tasks but not to the reference tests, and others that were common to both domains. 15 In addition, the abilities identified as factors (e.g., cognition of figural classes) seem to be as complex as the learning task. Selection of Tests The reference battery is comprised of two types of instruments. The first group consists of standardized psychometric tests that have been selected for their ability to measure the organismic variables of interest as well as for their factorial simplicity. Group I Raven Progressive Matrices. The progressive matrices are purported to be a pure measure of the general factor "g," common to most intelligence tests. With the college sample used by Jensen (1964) very little spread in the scores was found among SE and therefore the matrices proved to have low discriminatory power. In an attempt to overcome this problem a 30-minute limit was placed on the time allowed to complete the test. This restriction should make the test suffic- iently difficult and therefore add some spread to the scores of the college sample used in the study. Stroop Test. This test yields a measure of response competi- tion. This measure provides an index of interference (reSponse competition) between two unequal habit strengths, in this case color naming and word reading, and is distinguished from other measures such as retroactive and proactive interference. Though it is not a formally standardized test a basic format has been developed and its extensive use has been well reviewed by Jensen and Rohwer (1966). The procedures of testing and the obtained measures used here will be in the same format as those used by Jensen (1964). l6 Eysenck Personality Inventory (EPI). Though the test carries the label of "Personality Inventory," it is only the hypothesized under- lying genotypic aspect and the intrinsic aspect of learning that are of importance to this study. The inventory measures two independent dimensions: extraversion-introversion (E), neuroticism-stability (N). The E factor is hypothesized as being closely related to the magnitude of excitation and/or inhibition found in the central nervous system (CNS), while the N factor is hypothesized as being closely related to the degree of lability of the autonomic nervous system (Eysenck, 1960). For example, §§ who score low on the E scale are postulated as having strong excitatory and weak inhibitory potentials, whereas §§ who scores high on the E scale are characterized as having weak excitatory and strong inhibitory potentials (Eysenck and Eysenck, 1968b). Witkins Test of Field Independence: Embedded Figures Test (EFT). The EFT gives us a measure of the trait or characteristic that has been labeled field independence or field articulation in some factor analytic studies (Gardner, et al., 1960). Field Independence defines the ability of an individual to differentiate the figure from the ground in a visual structure. Witkin (1962) characterizes the typical field-dependent person as one who takes a long time to locate a familiar figure hidden in a complex background. Whereas, the field-independent person is more analytical in his approach to his environment and tends to impose structure on a field which lacks it. The importance of this character- istic as a subject variable in concept learning has recently been demonstrated by Dickstein (1968). Kagan's Matching Familiar Figures Test (MFF). The MFF is a measure of the trait labeled reflection-impulsivity. The trait is l7 descriptive of two discrete cognitive styles, and in this fashion is somewhat less simplistic than some of the other factor measures included in the reference battery. Group II The second group of reference tests were devised using the methodologies and procedures found in the eXperimental literature as guidelines. Selection again is based upon the factorial simplicity of the measures and the judgment of the eXperimenter as to their relevance as organismic variables. Respective to this, all measures derived from the tests included within Group II meet the criterion of falling within the definition of "intrinsic sources of IDs" as it is stated above. The reference tests in Group II were designed to assess the following functions. Immediate Digit Span Memory (IDs). This is a measure of short- term memory (STM) where the S is required to recall a series of stimulus items immediately after their presentation. Basically, the S is presented a set of stimulus items in serial order, one at a time. Depending upon the experimental requirements, the §§ are required to reproduce the items in their exact serial order or reproduce as many items as they can in any order (i.e., free recall). The number of items that the §'is able to recall is considered to be a measure of the Sig ability to retain and recall material in their STM. This ability is hypothesized to be dependent upon the strength of the initial registra- tion of the stimulus trace(s). Delayed Digit Span Memory (DDS). This test is also a measure of STM and follows the same basic paradigm as IDs with the exception that an unrelated task is interpolated between the learning and recall phases 18 of the experimental trial. The interpolated task is of a specific time duration and therefore inserts a measured delay between the learning and recall phases. Primarily, in addition to causing a delay between learn- ing and recall, the interpolated task prevents covert rehearsal of the stimulus item presented in the learning phase. The measure of retention of the stimulus items in this paradigm is hypothesized to be dependent upon the decay of the stimulus trace that takes place during the time delay between learning and recall. Proactive Inhibition of Digit Span (PI). The basic paradigm for PI is: learn list A, learn list B, test retention of list B. PI takes place when interfering stimulus items occur before the acquisition of the criterion items. The interfering items are said to act forward or proactively in effecting the retention of the criterion items. The measure of retention of the criterion stimulus items in this paradigm is postulated to be dependent upon the weakening of the stimulus trace due to the persistance of the trace of the previous list. Retroactive Inhibition of Digit Span (RI). The basic paradigm for RI is: learn list A, learn list B, test retention of list A. In this paradigm a list of stimulus items is interpolated between the learning of the criterion items and a test for their retention. The interfering items (i.e., interpolated list) are said to act backward or retroactively on the remembering of the criterion items. In the RI paradigm, retention of the criterion items is dependent upon the amount of interference with the consolidation of the stimulus trace of the criterion list. Actually, the terms proactive and retroactive are somewhat misleading in suggesting that in the one sense list A works forward in 19 time, while list B works backwards in time. In effect, the acquisition of the lists are successive in time, and it is the interaction of the two traces that produces any decrement found in retention. Immediate Visual Memory (VMI). VMI is a measure of visual short- term memory. Sperling (1967) presented a model of STM which emphasized the acquisition and storage of visual stimulus materials. Basic to his model was a component given the label of Visual Information Storage (VIS), which is similar in concept to the sensory memory component of Atkinson and Shiffrin (1968). VIS is a very brief visual storage system that is capable of holding a great deal of information for a short duration. The decay time of the contents of VIS vary from a fraction of a second to several seconds (Sperling, 1963). Though the visual sensory data is transformed for storage in VIS, the information is then scanned and encoded in a verbal form in a component labeled Auditory Information Storage (AIS). AIS is similar in concept to the Primary Memory component of Waugh and Norman (1965). The relevance of a concept of VIS to the processing of visual information is obvious. In relation to conceptual learning, involving the processing of visual information, it would seem that the efficiency with which a §’is able to retain and recall sensory data in VIS could be considered to be an intrinsic source of variation. The procedures followed in measuring VMI are explicated in a later section (IV, B-2C). Delayed Visual Memory (VMD). VMD is also a measure of visual short-term memory and will follow the same basic procedural format as VMI with the exception that an unrelated task is interpolated between the presentation of the visual data and its recall. The interpolated task is of a specific time duration and therefore inserts a measured 20 delay between the presentation and recall phases. Primarily, as in the DDS paradigm, in addition to causing a delay between presentation and recall the interpolated task prevents covert rehearsal of the visual items presented in the acquisition phase. It was stated that storage in the VIS lasted at best for only a few seconds, after which time the material has decayed. Within the structure of the model it is assumed that the information is quickly recoded and stored in a somewhat more permanent form of memory, the AIS. Once in the AIS the information may be rehearsed, discarded or placed in long-term memory. The ability of a S’to scan his VIS and store information in the AIS is a process that is intrinsic to learning and therefore by definition a probable intrinsic source of variation. The effect of the delay on the recall phase raises a theoretical question. If the mode of presentation of the interpolated task was visual it would undoubtedly interfere with the retention of the stimulus materials and therefore cause a decrement in the recall measure. But, if the mode of presentation of the interpolated task was in a non- interfering auditory mode two outcomes are possible. The first probable result is that the delay may cause the visual information to decay without being transferred to AIS, or a partial loss and storage in AIS resulting in a decrement in recall. The second probable result is that since visual memory is not susceptible to auditory interference (Sperling, 1963) it is possible that a.§ is able to encode and transfer the visual information from VIS to the somewhat more permanent AIS while he is performing the interpolated delay task. In this case, the resulting recall measure would be dependent upon a respective S's 21 ability to encode and transfer visual information from his VIS to AIS. It would be expected the S's recall measure would be at least equal to or better than his performance on the VMI task. CHAPTER III METHOD AND PROCEDURES Selection of Subjects The subject sample to be used in this study is the group of students enrolled in the introductory general psychology course at the University of Northern Colorado. This course being a general education requirement for the undergraduate degree presents a fairly representative cross section of the college p0pulation. From the initial group of volunteers, 40 Sp were selected for participation in the study (20 males and 20 females). The basic criterion for selection was that the §§ demonstrate a willingness to participate in the study and to maintain a strict testing schedule for an extended period of time. One S was drOpped from the study after the second week of testing because of his frequent absences during his allotted testing time. Administration and Scoring of Reference Tests Group I Raven Progressive Matrices (RPM). The advanced progressive matrices Set II was administered to all 39 subjects in a single session. The Sp were given a 30-minute time limit within which to finish the test. All Sp required the full 30 minutes with six_§§ completing all 36 items in the set. The mean score of the group placed it at the 90th percentile according to the norms provided by Raven (1965). Though no test-retest 22 23 reliability measures were made, the re-test reliability for college students is reported by Raven (1965) to be rtt = .91. Since a limiting time factor was added to the testing procedure, rather than using the absolute number of correct items a‘S received as a score on the matrices, the original score for each §_was transformed into the percent correct of these items attempted. number of correct items number of items attempted % correct = A Pearson Product-Moment Correlation calculated between the original scores and the transformed score, yielded an rxy of .75. Str00p Test. Though the Str00p Test is not a formally standard- ized test a basic format has been developed and its extensive use has been well reviewed by Jensen and Rohwer (1966). The procedures of test- ing and the obtained measures used here will be in the same format as those used by Jensen (1964). There were three cards--the color card (C) on which there were 100 patches of five different colors, the word card (W) on Which the names of the colors were printed, and the color-word card (CW) on which were printed the names of colors but they were printed in a color conflicting with the printed name (e.g., the word yellow printed in red, green or blue ink). Each card has 100 items to be named. The SL3 task on card C is to verbally state the names of the color patches, reading from left to right as fast as he can. 0n card W the Slp task is to read aloud the color names as fast as he can. On card CW the Sip task is to name the color of the inks that the words are printed in, while ignoring the conflicting printed color names. Card C consisted of ten rows and ten columns of evenly spaced colored dots. The dots were all 5/8" in diameter and l-l/2" center to 24 center. The five colors used were red, orange, green, blue and yellow. The placement of all colors within the 10 x 10 matrix was random except for the following restrictions: 1. Adjacent dots (reading from left to right) were never of the same color. 2. All colors appeared at least once in a row of dots. 3. All colors appeared an equal number of times. Card W consisted of 20 rows of five columns. The words were printed in off-white on a flat gray background. All letters were in block capitals 7/8" high. Their line width was 1/8". All rows and columns were in exact line with the words being distinctly separated. The word names were in random order except for the same restriction applied to Card C (above), with the additional restriction that the color names were never in the same order as the color dots on Card C. Card CW consisted of the same word format as was used with Card W, but with the words colored with the five colors, the actual color conflicting with the color name. The order of the colors was the same as with the color dots. The cards were placed on an easel five feet from the S, with the cards being approximate with the SLp eye level. The order of administration was Card C, Card W and Card CW. The task was explained to the S by the experimenter (E); in addition the five colors to be used were named. When the S indicated that he understood the task Card C was presented, E said "Go," and simultaneously started a stOp- ‘watch. The procedure was similar on Cards W and CW. Prior to the I>resentation of each card, S was told what was expected of him. On Caird W he was told to read the color names; on Card CW he was told to némme the color and ignore the printed words. 25 Each S received a score as to how many seconds it took him to complete each task. Jensen and Rohwer (1966) in reviewing the literature on the Stroop Test found no less than 16 scores derived from the three basic time scores on Cards C, W, and CW. When factor analyzed (Jensen, 1964) only three factors emerged from all the Stroop scores. These factors were identified as: 1. Color difficulty factor (Cd). 2. Speed factor (Sp). 3. Interference factor (Intf). The scores which most clearly represented the factors were chosen for use in this study. The scores are as follows: C = C/(C+W) d Sp = W Intf = CW - C The test, re-test reliabilities of these scores as reported by Jensen (1964) are respectively: Cd: rtt = .97 Sp: rtt = .98 Intf: rtt = .93 Eysenck Personality Inventory (EPI). Since the EPI carries the label of personality inventory, permission to administer the test was solicited from all Sp before administration. The general form and purpose of the inventory was eXplained to all Sp and in addition they were told that if after taking the test they objected to its format, they may personally destroy the answer sheet. All Sp granted permission and did not object to the format and question. 26 Form A of the EPI was administered to all 39 Sp in a single group session. Whereas in standardization norms for American College Students (Eysenck and Eysenck, 1968a) the correlation between the E and N scales was zero (i.e., .00) the correlations found for Form A between the rEN = two scales in this study was rEN = .13. This indicates that at least in the college sample used in this study there is some relationship between the two scales. In respect to the stability of the scales the test-retest reliabilities are quite satisfactory, with the reported = .82; N-scale, r = .84. reliabilities on Form A being: E-scale, r tt tt Witkins Test of Field Independence: Embedded Figures Test (EFT). The EFT was administered by the E to individual Sp following the standardized format of Witkin (1971). The tests were conducted as part of a normal weeks testing schedule. The test-retest reliabilities reported by Witkin (1971) are as follows: males: rtt = .82 females: r .79 tt Three basic scores may be derived from the EFT, a measure indicating the average amount of time required by the S to complete an item (Xt)’ the number of errors a'S makes in performing the task (e), and number of times a‘S request that the simple forms be shown after their initial presentation (XS). The intercorrelations of the three scores are as follows: Xt with XS, r = .70; Xt with e, r = .72; XS with e, r = .69, all were significant at p < .01. It is felt that X8, in addition to being a measure of field independence also contains a visual memory component, and therefore for the purposes of this study would be the best measure of the three scores to use. The magnitude of the 27 reliability correlation coefficients indicate that the measure of field- independence is a stable construct. Kagan's Matching Familiar Figures (MFF). The MFF was administered by the S to individual Sp. The stimulus cards used were ones modified by Shulman, et a1. (1968) for use with adults. The set consisted of one pretest example card and 12 test cards. Each card contained one sample figure and eight test figures (two rows of four), with one of the eight test figures being a match of the sample. The SLp task was to choose the test figure that matched the sample figure. The format of test was explained to the S by S. The S was then given the pretest example card; if there were no questions on the part of the S_and S was assured that S understood the task, the S was then presented the other 12 cards in succession. On each card the S was given three minutes within which to correctly choose the matching figure. If the S did not make the correct choice within the three minutes, the trial was terminated and a new card presented. The obtained measure on the MFF is the amount of time required to correctly match the sample figure. For the purposes of analysis in this study, a Sip score was the average time it took him to correctly match the sample over the 12 test cards. Using the stimulus cards, modified by Shulman, et a1. (1968) for an adult population, Lezotte (1969) found an internal consistency reliability using an analysis of variance procedure of r = .71. Group II Immediate Digit Span (IDS) and Delayed Digit Span (DDS). In the first test (IDS), the S would hear a series of digits spoken by a female 28 voice at a one-second rate and would write down the series on his answer sheet immediately after the presentation. In the second test (DDS) the .§L§ recall was delayed by ten seconds. The delay interval was filled by the verbal presentation, of pluses (+) and minuses (-), spoken by the female voice at a one-second rate. There were eight delay items in all, and the S was required to write down on his answer sheet, in the spaces provided, the corresponding symbol as it was spoken. In almost all cases, with few exceptions, the Sp were conscientious in attending to the spoken (+) and (-) and writing them down. In all cases, 13 seconds were allowed for the S to write down the digit series. The IDS and DDS series were randomly interspersed within sets of eight; there were ten such sets in all. Each length of a series was replicated five times throughout the entire test. To summarize: 2 conditions (IDS and DDS) 8 series lengths (2, 3, 4, 5, 6, 7, 8, 9) 5 replications The test was administered to the Sp in groups of five. The Sp sat around a semi-circular table with a tape recorder containing the recorded digit series placed in the center, equi-distant from all Sp. The task was explained to the Sp and an example task was presented. When the S was sure that the task was understood the tape recorder was started. To summarize the sequence: Immediate Digit Span Events Time 1. "ready" command 1 second 2. pause 1 second 3. digits (2 to 9) 2-9 seconds 4. "write" command 1 second 5. blank for writing response 13 seconds 0\ etc. 29 Delayed Digit Span Events Time 1. "ready” command 1 second 2. pause 1 second 3. digits (2—9) 2-9 seconds 4. pause 1 second 5. + and - 8 seconds 6. "write" command 1 second 7. blank for writing response 13 seconds In both IDS and DDS, the serial order position method of scoring was used. This method consists of giving one point credit for every item recalled in the serial position it occupied in the order of presentation. For example, if a.S recalled the series 12345 as 23451, his score would be zero, whereas if he recalled it 13425, his score would be two. The test-retest reliabilities of IDS and DDS using the same format as was used in the present study, were found to be satisfactory (Jensen, 1962). This would indicate that these phenomenon as they are investigated in this paradigm are quite stable. To summarize these: IDS: r tt DDS: rtt 070 .79 Retroactive Inhibition (RI) and Proactive Inhibition (PI). This test grouping consisted of two conditions: 1. Digit span with Retroactive Inhibition (RI) for digit series lengths of from four to seven digits. In this paradigm the S heard one series of digits, then a second series, and then was asked to recall the first series. 2. Digit span with Proactive Inhibition (PI) for digit series lengths of from four to seven digits. This paradigm is formally the same as the one above, except that the S_is asked to recall the second series. 30 The administration followed the same basic format as used with the IDS and DDS paradigm. The Sp were presented with 16 RI and 16 PI conditions, which were randomly interspersed over the 32 conditions. In both conditions, ten seconds always intervened between the last digit of the series—to-be-recalled and the "write" signal. In both conditions the S did not know until time of recall whether he would have to write the first or second series. The sequence is summarized as follows: 3; Events Time 1. "ready" command 1 second 2. pause 1 second 3. first digit series (4-7) 4-7 seconds 4. pause 3 seconds 5. second digit series(4-7) 4-7 seconds 6. + and - 8 minus 2nd series 7. pause 1 second 8. "write" command 1 second 9. blank for writing response 13 seconds g; Events Time 1. "ready" command 1 second 2. pause 1 second 3. first digit series(4-7) 4-7 seconds 4. pause 3 seconds 5. second digit series(4-7) 4-7 seconds 6. + and - 9 seconds 7. pause 1 second 8. "write" command 1 second 9. blank for writing response 13 seconds As with the IDS and DDS paradigm, serial order position scoring was used with this test series. Using the same format and procedures as were used in this study, Jensen (1964) found the measures of RI and PI to be reasonably stable phenomenon. are summarized as follows: This is indicated by the test-retest reliabilities which 31 RI: rtt = .60 PI: rtt = .58 Immediate Visual Memory (VMI): The VMI was administered by S to Sp in groups of two as part of a normal weeks testing schedule. During the VMI condition a.S saw a stimulus pattern flashed on a screen for the duration of 250ms (1/25 seconds). The Slp task was to view the image while it was projected on the screen and immediately write down on his answer sheet what he saw during the brief exposure. The answer sheet contained a check list indicating the possible choices of stimulus items. This format was as follows: Border Figures one red one red large circle two blue two blue small ellipse This format enabled the S to quickly indicate his recall reSponses without any memory loss due to the time that would be needed to write them down in long hand. . The stimulus materials were twenty—four 35mm transparencies of geometric designs. The stimulus materials were of the same format as the stimulus patterns used in the concept learning phase of the study. A stimulus pattern could contain any one of the possible (64) designs generated by using all possible combinations of levels within the following six binary dimensions: one or two, red or blue, solid borders; one or two, large or small, red or blue, circular or elliptical, solid figures. The 24 transparencies were divided into three groups of eight, the group division being respective to the number of stimulus elements each transparency contained. The group divisions were: (1) two elements, (2) three elements, and (3) four elements. For example, if a 32 tranSparency contained two borders and one figure, it would belong to the same stimulus group as one which contained one border and two figures. The target to background contrast of the projected image, measured by a digital photometer (Gamma Scientific Instruments) was 80 percent. During the testing phase, the 24 transparencies were randomly interspersed in order to avoid any chance of a perceptual response set being develOped in the Sp. The duration of exposure was controlled tachistascopically using a model T-AP Tadhistascope, manufactured by Lafayette Instrument Company. The transparencies were projected onto the screen using a Viewlex Projector, with a five-inch Luxtar lens. The Sp were seated side by side, approximately three feet from each other and five feet from a flat gray screen upon which the stimulus image was projected. Each S was shown a card containing samples of the visual materials that he was to view. The dimensions of the stimulus patterns were pointed out and explained to the Sp by the S. The task was explained to the Sp by the S; if there were no questions a pretest example was presented. If there were no further questions after the example was presented and the S felt the Sp understood the task, the normal testing session was begun. In all trials, the Sp were allowed ten seconds to make their recall responses. The total number of recall errors committed by a S over the 24 trials was used as the VMI measure. Using a seven-day interval between sessions, the test-retest reliabilities (n = 20) computed for the VMI paradigm yielded an r of tt .69 (i.e., r = .69). The magnitude of the correlation coefficient tt ‘Would indicate that the VMI performance is a stable phenomenon. 33 Delayed Visual Memorx (VMD). The VMD followed the same basic format as was used in VMI with the exception that a ten—second delay interval was interpolated between presentation and recall. The delay interval was filled by the verbal presentation of pluses (+) and minuses (-), spoken by the Slat one-second rate. The S was required to write down on his answer, in the spaces provided, the corresponding symbol as it was spoken. In almost all cases with few exceptions, the Sp were conscientious in attending to the spoken (+) and (-) and writing them down. In all cases the Sp were allowed ten seconds to make their recall responses. The total number of recall errors committed by a S over the 24 trials was used as the measure of VMD. Because of scheduling problems test-retest reliabilities were not made on the VMD paradigm. However, computed Spearman-Brown Split-half reliabilities resulted in an rtt = .71. This would indicate that within one administration of the test a 8'3 performance was relatively consistent. Laboratory Tasks The format to be used in this series of tasks is a modification on a procedure suggested by Bourne, et a1. (1964). Procedural Variables Stimulus Availability (SA). SA is Operationally defined in terms of the number of previously presented stimuli to which the subject has access on any trial. The design matrix will include three levels of SA: two available stimulus cards--SA2, four available stimulus cards-- 8A4, and six available stimulus cards--SA6. 34 Concept Complexity (CC). CC is defined by the number of relevant attributes defining a particular concept. There are two conditions of complexity: two relevant attributes--CC2, and four relevant attributes-- CC4. Apparatus The apparatus consists of three "memory boards" constructed of clear plastic, one board is assigned to each SA level. On the front of the boards are pegs on which the S can hang the stimulus cards, the number of pegs available are equal in number to the assigned SA level. In addition, in order to aid the subject in remembering the identities of the instances on the board, each peg has a bi-colored disc (red-- positive, black--negative). For example, if the subject were to hang a positive instance on a particular peg, he would then rotate the disc so that the red portion of the disc was showing above the stimulus card. Stimulus The stimulus patterns to be used in the eXperiment are geometric designs printed on 2 x 2-1/2 inch white cardboards. Each card contains one of the possible (256) designs generated by using all possible combinations of levels within the following eight binary dimensions: one or two, red or blue, solid or broken borders; one or two, large or small, solid or spotted, red or blue, and circular or ellipitical figures. In a technical sense, those dimensions which are important to the definition of a concept, are labeled as "relevant," and those which are not as "irrelevant." The levels or different values of a dimension are referred to as "attributes," and therefore, in relation to the relevant dimension those attributes which specify a concept are termed 35 "relevant attributes." A stimulus event which contains all of the necessary relevant attributes in their pr0per relationship is referred to as a "positive instance" (PI); those events which do not as "negative instances" (NI). Concept Problems The variables SA and CC being crossed produce six independent problem conditions. The stimulus arrays generated for each were selected from the 256 possible designs. Each series begins with a positive focus card and contains an equal number of positive and negative instances. The number of trials presented to a subject in any one problem series is equal to the assigned SA level plus twenty. This will result in stimulus arrays of 22, 24 and 26 trials in length. IrreSpective of the CC in a problem condition, the amount of information presented up to and including trial (SA level + 1) will leave 64 possible hypotheses remain- ing until solution, with the final hypothesis being eliminated on the last trial. The resulting design matrix is summarized: CC Level SA Level 2 4 2 4 6 Ergo—«1232 The procedure was patterned after Bourne, et al. (1965). Preliminary instructions given to‘S concerning the concept learning task included: (a) a description of all the stimulus dimensions and 36 their levels, (b) an explanation of the information contained in positive and negative instances, (c) he was told that all of the concepts are to be conjunctive but was given no indication as to how many attributes were relevant to the concept he was to attain, and (d) respective to the SA level to which he has been assigned the S was told that he may retain as a maximum only the SA level number of cards, and after SA level + 1 trials he must discard at least one of the previously presented cards, and after each successive presentation the maximum number of cards that a'S may retain is controlled by the SA level. In addition to the initial instructions concerning the SA level and discard procedure, the S was instructed to arrange the cards in any order or fashion that he may choose on the memory board placed in front of him. After the instruc- tions the Sp were shown examples of what would be positive and negative instances of a given concept. During each problem the stimulus was placed before the S one at a time and he was allowed to arrange them in any order he chooses. The experimenter described the first trial presentation (focus card) as a positive instance of the concept. On every successive trial, within fifteen seconds of the stimulus presentation S was required to verbalize whether or not he thought the stimulus card was a positive or negative instance, his response was then confirmed or invalidated by the experi- menter. After the subject identified the instance, he was given 15 seconds in which to hang the card on the memory board and study it. Therefore, in each of the treatment cells, a number of previous- ly presented stimuli, vis., 2, 4, or 6, remain before S as he responds to each stimuli. In treatment 8A4, for example, stimuli which have been presented on the four preceding trials (except for the first four trials 37 of the problem), plus the particular stimulus instances just presented were available toIS for inspection. Each time a new stimulus is pre- sented, the S is required to identify it as a PI or N1 and discard one of the exposed cards within 30 seconds of the presentation of the stimulus card on that particular trial. An additional criterion require- ment such as the identification of the relevant attributes, was required of the subjects. A trial by trial record was kept of the identification error, the order in which S discards the stimulus cards relative to their order of presentation, and the cards which he retains. Each of the 39 Sp were tested on the concept attainment task for a period of five weeks. The testing sessions were designed such that a.S would be presented with all six problem conditions within any one week's testing program. Under these conditions, 90 independent stimulus arrays were developed and were presented to each S in a random order over the five-week period. Procedural Modifications It became evident after one week's testing and the Sp had become familiar with the task procedures, that the 30 seconds allowed for each trial was much too long. The Sp were making their identification responses, on the average, within five seconds and felt that they needed only an additional five seconds to study the cards. It was further evident that the procedure of having the S present the stimulus cards to the S was unnecessary. Therefore, the original procedure was modified as follows: 1. The stimulus cards were arranged in their prescribed order and placed face down on the table in front of the S. 38 2. The S was allowed to select a single card at a time, identify its class (i.e., PI or NI) and study it if he wanted to. 3. The time restriction that aIS must identify the card within 15 seconds and that only 15 seconds was allowed to study it after identification was still imposed. Scoring The dependent measures in the CA task were (a) memory errors, (b) judgment errors, and (c) identification of relevant attributes. A memory error is operationally defined as the type of error a S commits when he has had sufficient information to properly classify an instance, as a PI or NI, but does not. A judgment error is when a_S is presented with a stimulus card that contains new information and he makes an error in classifying it (i.e., an error in judgment). At the end of the CA task, the Sp were asked to identify the relevant attributes of the concept and they were given one point for a correct answer and zero for an incorrect answer. CHAPTER IV RESULTS Sample Validity In order to establish the comparability of the sample used in this study with the general experimental p0pu1ation, which is generally composed of volunteers from undergraduate courses in psychology, the mean scores obtained from the sample on the reference tests (Table l) were compared to the standardization norms of the respective tests. In cases where the reference tests have not been standardized the compari- sons were made to data available in the literature. 0f the 14 tests used, there were four on which there were no available data or norms which applied to the age range of the sample; these four were the EFT, MFF, VMI and VMD. Eysenck Personality Inventory The mean scores for the Extraversion (E) and Neuroticism (N) scales placed the sample at the 62nd and 4lst percentile, respectively, for American college students (Eysenck and Eysenck, 1968a). The mean scores and standard deviations computed for the standardization sample = 4.1 and R = 10.9, S on the E and N scales are: R = 13.1, S = 4.7, d d respectively. In comparing the sample scores found in Table l, with the scores obtained by Jensen (1964) little difference was found between the three sets of data. 39 40 No mHn HH: 5N HN- mNn oo: oo: H¢.HH m¢.om Hm .oH wN qN: om: ON 00- NH: NN: so- NH.oH w¢.nm Hm .MH on 0H: Hm: HN Ho 0N: Ho: NH nn.NN mN.wHH man .NH NH: wNu mH Ho: mNu mHu oH Hn.wH mn.HoH mQH .HH on so: «N: «0 mo Ho: mo.m om.NH 92> .oH «H- no oN «o. «N: oou NH mm.oH Nm.NN Hz> .m u NH MN- Ho mH 0N: mH N¢.m mN.o¢ am .w I OH no- «H mo ON- o¢.NH mq.o¢ mucH .5 HH- ma- so- oo mm.m oo.mm so "aoosum .o n Ho HHu mHn q¢.nH no.mn 2mm .m - no mu- mm.e~ mm.oo em "as: .e No- no 0H.m wm.m x "Ham .m u mHu mq.¢ o¢.m z .N n om.m mN.mH m "Hmm .H HH w n m o N H mm coo: oHanum> .mmHanum> monouomom «H waoa< mnOHumHouuoo was mmw..momoz .H oHAMH 41 Raven Progressive Matrices The sample mean scores, found in Table 1, place the sample at the 90th percentile (Raven, 1965). Though the ranking indicates that the experimental sample respective to performance on the RPM is superior to the general papulation, the ranking is similar to that found by Jensen (1964). Therefore, the assumption is that the college sample used in this study is comparable to other experimental samples drawn from the papulation of college students. The four remaining tests in the reference battery follow basic procedures found in the experimental literature. Presented in Table 2 are the data obtained by Jensen (1964) using the same testing format as was used in the present study. Table 2. Summarization of the Data Obtained by Jensen (1964). Variables X Sd Stroop: color difficulty .60 .04 interference 42.12 14.96 speed 38.09 5.84 IDS 176.68 21.96 DDS 152.52 26.74 RI 55.70 12.81 PI 52.38 13.25 Comparing the measures obtained in the present study with the Jensen data, it can be seen that though the sample profiles respective 42 to the Stroop measures are similar, the performance of the sample used in the present study is consistently poorer in relation to the digit Span memory and the interference-measures (i.e., IDS, DDS, RI and P1). In general, it appears that the test profile of the experimental sample is comparable to the general experimental pOpulation and may be considered to be a valid sample with one qualification. This qualifica- tion being, that any inferences made within this study in relation to immediate memory and interference variables as ID performance factors should be made respective to the differences noted between the data of the present study and that obtained by Jensen (1964). Task Validity In a study of this design, it is necessary to establish that the sample is not only a representative one but also that the performance of the sample on the experimental task is comparable to what might be expected from the literature. In this sense, the validity of the task is established. A two-way analysis of variance for repeated measures (Winer, 1962) was carried out on the dependent measures of memory errors, total errors and rule acquisition under the variables of CC and SA. The complete analysis is presented in Appendix A. In that the data in the literature has been obtained generally in one testing session, the task validity will be evaluated in relation to the data obtained during the first week testing in the present study. The literature in the area of concept learning would predict that those problems which have the greatest concept complexity (i.e., greatest number of relevant attributes) would be the most difficult (Walker and Bourne, 1961; Bourne, et al., 1964). In addition the earlier findings of Bourne, et al., indicate that concept learning is facilitated 43 by the availability of previous stimuli. The degree of this effect being a non-linear function. Data from the present study does not confirm these earlier conclusions. Though the analysis of the memory errors (Appendix A) yielded a significant effect due to CC, the data is contrary to the predicted direction. For example, the problems of greatest complexity (i.e., CC4) produced the fewest number of errors. SA as a variable in the present study did not yield a significant F-ratio, and the data is ..-—.- -‘ --— —- somewhat equivocal as it relates to the earlier finding of Bourne, et al. (1964). Respective to the variables of CC, SA and their interrelation- ships, the present study did not manifest the stability and generality of earlier findings. In light of this fact, the generalizability of any findings that might result from this study, on IDs as they relate to conceptual learning, will have to be strictly qualified. Reliabilityiof Dependent Measures Basic to the design of the present study was the use of long term experimentation. It was felt that the extended period of testing would allow the investigator to collect data on the reliable performance measures found in the final stages of practice. Following within this assumption reliability coefficients were computed for each of the six conditions across the five weeks of testing (Hoyt, 1941; Winer, 1962). Presented in Table 3 are the reliability coefficients for the dependent measures of M errors and rule acquisition. In relation to the stability of the concept learning performance, as can be seen from Table 3, the necessary assumption of stability respective to these dependent measures is not confirmed. In that, in theory, multiple regression analysis 44 Table 3. Reliability Coefficients of Dependent Measures Under All Conditions of the Conceptual Task. Condition Dependent Measure r coefficient SA2:CC2 Memory error -.35 SA2:CC4 .16 SA4:CC2 .14 SA4:CC4 .27 SA6:CCZ .ll SA6:CC4 .00 SA2:CC2 Rule acquisition .14 SA2:CC4 .25 SA4:CC2 .00 SA4:CC4 .02 SA6:CC2 .02 SA6:CC4 .28 requires that the criterion variable (i.e., dependent measure) be statistically reliable, the proposed analysis will not be carried out. Factor Analysis Rationale Previous to this investigation the 12 reference tests employed in this study had never been used concurrently in any one study. Therefore, a determination of their empirical relationship had not been made. ReSpective to this, though the reference measures are considered to be indices of individual process variables that are at least 45 phenotypically different, one could speculate that some of the measures share a common variance. Factor analysis, as a statistical procedure, supplies a sound method of determining the covariation or interrelation- ship among a number of variables and reducing them to a generally more fundamental and lesser number of variables. If, through the procedures of factor analysis, a number of phenotypically different kinds of varia- bles demonstrate an interdependence one could hypothesize that they represent a single common factor (i.e., the same intrinsic source of variance and/or genotype). Factor Analysis Procedure An intercorrelation.matrix was computed between the SLp scores on the 14 reference measures. The resulting correlation matrix was first subjected to a principal components analysis. The principal component solution was then rotated to a varimax solution, with unities placed in the diagonal of the correlation matrix and only factors having eigen values of one or greater being rotated. Factor Analysis of Reference Tests Presented in Table l are the means and standard deviations (SDS) of the measures of the 12 reference tests that represented the 14 variables which entered into the factor analysis. The correlation matrix (Table l) was first subjected to a principal components analysis (Table 4). Varimax Rotation Factors An orthogonal varimax rotation yielded five factors which accounted for 66 percent of the variance. Interpretations of the rotated factors are based on loadings equal to .40 or greater. The five 46 SSH SN SH- HS- SH- SS SS- NS- NN- SS HN- HN SH- NN- SN HS .SH SSH HN- HN- SH- SS SH NS NH- HN- SS SH- NH- SH- SH- NN HS .SH SSH SH- SN SN SH- SN NS SN HH- SS- NH- HN NS- SH SN SSS .NH SSH SS SS- SS SS NH- SH SS- NS SS- SN- SN SS NN SS SSH .HH SSH SS HH SH- SS SN NS SH- NN- NN- SH- SS SS- SS- NS- Sz> .SH SSH SS SH HN SN SS- SS- SH HH- NN SN SS- SH- SS- SN Hz> .S SSH NS- HS- NH- SH NH SS- SN HN- SS SS- SH- HS HS HS- .S SSH HH- HH- NH SS- SN- HH SN SS- SH- SS- SS HN- SS- SS- .N SSH SS SS- NH- HS SH HH- SH NS SH- SN- HH- NS- NS- SS "SoouSS .S SSH NS SS SH- HS- HS NS SS NS- SS SS- NS- NN- SS SS- SSS .S SSH SS SS- SH SH SH SH SS SN SS NN- SS SS SH SS- Sm "SS2 .S SSH HS- SH NN- NS- SS- SH SH SH SS SS SH SS SS- SS SS "SSS .S SSH SS- SH- SH- NH SS SS- HS HH- SS- NS NS NN- HH- SH- 2 .N SSH HS SN- SH SS- SH NN SH SH SH- SN SH- SN HN- NN S "HSS .H NS SH SH NH HH SH S S N S S S S N H SHSSHSS> moHuHHmSDEEoo ocm mucoaomfiou HomHocHum .S oHan SSHSSHNS> Seemumwmm SH SSS So HNSV 47 factors are presented below. The same format will be followed throughout. The variables are listed in descending order respective to the magnitude of their factor loadings. An asterisk following the factor loading indicates that the variable had its highest loading on this factor. Table 5 presents the factor analysis. Table 5. Varimax Rotation Solution. Variable A B C D E h2 1. EPI: E -14 -21 76 -25 -22 76 2 N -15 11 05 02 85 76 3. EFT: xS 30 -04 74 -03 -02 65 4. MFF: it -10 56 -09 23 01 39 5. RPM -16 -05 -59 -39 -32 64 6. Stroop: Cd 44 -37 -07 28 00 42 7. Intf -12 06 -15 81 -11 72 8. 3p -25 34 3o 15 -59 65 9. VMI 01 -85 09 14 02 75 10. VMD -26 -01 10 72 05 61 11. IDS 66 17 08 -33 -1o 60 12. DDS 77 oo 11 -33 04 72 13. R1 66 -39 -01 -12 -03 63 14. pl 84 -16 15 02 10 77 Percent of Variance Accounted For: 19 ll 12 13 O9 48 Factor A Variable Loading 14 Proactive Inhibition .84* 12 Delayed Digit Span .77* ll Immediate Digit Span .66* 13 Retroactive Inhibition .66* 6 Stroop Color .44* Interpretation: Strength of the initial registration of the stimulus trace. Discussion: Proactive inhibition (PI) has its largest loading on this factor, and in addition it also has the highest loading in the rank ordering of the factor loadings. The factor was not called a PI factor because of the substantial loadings of other variables on this factor. Factor A is interpreted as the strength of the initial registration of the stimulus trace. ReSpective to this, the stronger the initial registration, the more trace is left to be consolidated, and the less susceptible it is to decay because of delay. In the PI paradigm the persisting trace of list 1 presumably weakens the registration of list 2. Factor B Variable Loading 9 Visual Memory: immediate -.85* 4 Matching Familiar Figures .56* 6 Stroop-color difficulty -.37 Interpretation: Visual Memory: immediate. Discussion: The interpretation of this factor is quite clear. VMI has its largest loading on this factor as well as occupying the 49 highest ranking. Since the Matching Familiar Figures (MFF) test has its highest loading on this factor, it suggests that a major portion of variance on the MFF might be due to a visual memory factor. Factor C Variable Loading 1 EPI: E Scale .76* 3 Embedded Figures: XS .74* 5 RPM:-% Correct -.59* Interpretation: Field-independence. Discussion: The interpretation of this factor is relatively clear with the EFT having its highest loading on the factor. With the RPM having its highest loading on this factor, it suggests that a major portion of the variance found on the RPM is due to a field-independence factor. The loading of the EPI: E might have been expected. It has been found that people scoring low on the S scale do better on visual performance tasks (e.g., Raven Progressive Matrices and Embedded Figures Test) than those who score high on the S scale (Eysenck and Eysenck, 1968b). Factor E Variable Loading 2 EPI--N Scale .85* 8 Stroop--Speed -.60* Interpretation: Neuroticism. Discussion: The definition of Factor E is clearly Neuroticism. The loading of Stroop-~speed with the N scale might be used as the definition of a cognitive style. In relation to this we would expect people scoring high on the N scale being able to perform a simple task, such as word naming much better than complex ones. 50 A summarization of the identifying factor labels are presented below: Factor Interpretation A Strength of the initial registration of the stimulus. B Visual Memory: immediate. C Field independence. ; D Susceptibility to Response Competi- E tion. % E Neuroticism. E CHAPTER V DISCUSSION AND SUMMARY Discussion The proposed second phase of the study was to utilize multiple regression procedures to determine the relationship between the in- trinsic ID variables, defined by the factor analysis, and the dependent measures. In that the measures proved to be unreliable across the five weeks the proposed analyses were not carried out. Whereas the literature in the area of concept learning would predict those problems which have the greatest concept complexity would be the most difficult, and that learning is facilitated by the avail- ability of previous stimuli, the present study did not confirm these earlier conclusions. The lack of comparable findings between the present and earlier findings is a point of dilemma for the author. Though logically the comparisons between the data of the present study and earlier findings should be made on data obtained during the first week, it should be noted that the data obtained during the fifth week confirms earlier findings of Bourne, et al. (1964) and Logan (1969). Bourne, et al. (1964) investigated the effect of an availability continuum, ranging from zero (0) to ten (10) available stimuli. Their results manifested a facilitative effect with increasing availability under their more complex problem condition (i.e., three relevant attributes). An Optimum level of facilitation was reached at four 51 52 available stimuli after which performance decreased. The data obtained in the present study during the fifth and final week of testing mani- fested the same facilitative effect due to SA. In addition, though no significant main effect due to CC was found, the present study yielded a highly significant interaction between the variables of CC and SA. These findings confirm the conclusions of Bourne, et a1. (1964). They demonstrate in general that SA has a facilitative effect up to some Optimal level, the degree Of effect being greatest with the more complex problems. The lack of a main effect due to CC in the study confirms earlier findings of Logan (1969). The present study and that of Logan (1969) have four procedural modifications respective to the Bourne, et al. (1964) study. The first provides feedback to the S in relation tO the correctness or incorrectness Of his Classification response. The three remaining modifications concerned with increased control of the information as it is presented to the S are as follows: (a) the stimulus arrays were balanced with an equal number of positive and negative instances; (b) the amount of information being transmitted was equal between stimulus arrays; and (c) the information was selected and pre- sented in a systematic manner rather than in a random fashion. The assumption made in relation to the present findings is that if in con- junction with giving the S feedback respective to the correctness or incorrectness of their classification response, the information is presented in a controlled fashion, the S can maximize their information and consequently reduce the conceptual complexity of the problem. If the controls are met the prediction would be that no differences would be found between Sp on conceptual learning tasks Of different concept complexities. 53 As a note to future investigators, it should be pointed out that an additional control should be added to the experimental paradigm. This control would entail presenting each Of the stimulus arrays specific to a SA-CC condition to the Sp under a Sa 0 or no memory condition. This additional variable will allow the experimenter to investigate the effect of available memory at the lower end Of the continuum. There is no Obvious explanation for the lack Of reliability manifested by the dependent measures. The reliability coefficients were determined utilizing ANOVA procedures (Winer, 1962) based on methodology first suggested by Hoyt (1941). The computational formula is as follows: MSbetween - MSwithin r = M Sbetween In relation to the obtained reliability coefficients, it is evident that reSpective to an experimental condition, the Obtained variance within a subject's performance was approximately equal to or greater than (e.g., r = -.35) the variance between subjects' performances. There are two possible inferences that may be made in relation to this data. The first is that the measurement Of a subject's ability to solve a conceptual problem solving task as it was presented in this study is an unreliable one. The second, is that the measure Of the subject's ability to solve the experimental problem in a simple test- retest situation would prove to be a reliable phenomenon. However, because Of the design Of the present study the Sp were exposed to a number of different experimental conditions (i.e., six task conditions) resulting in a differential learning phenomenon which manifested itself as an increased within subject variance. 54 To state this more simply, the assumption is that if the Sp were presented with a single problem condition their performance would prove to be a more stable or reliable phenomenon. However, in the present study, since the Sp were presented with a number of different problem types in a random order, it is quite possible that: (a) the Sp attempted to develop a problem solving style for each of the task conditions, (b) since the problem types were presented in a random order the Sp could not determine exactly which problem condition they were working under and (c) were therefore unable to develop a problem solving style or skill that was efficient for any one or all problem conditions. Following this line Of thought the logical assumption is that because Of the lack of an efficient problem solving skill the Sp would evidence unstable performance characteristics. These characteristics would manifest themselves as an increased within subject variance re- spective to all conditions. The validity Of this assumption can be evaluated utilizing the following procedures. Procedure A: T-la, T-la' Procedure B: T-la, T-2a, . . . T-6a, T-la' Procedure C: T-la, T-lb. The conditions within the procedures are identified as: T-la--first administration Of condition one, form a. T-la'--second administration of condition one, form a. T-lb--first administration of condition one, form b. T-2a . . . T-6a--first administration Of condition two through condition six, form a. 55 Within procedure A, the correlation Of Eplp with I;lél would provide the test-retest reliability coefficient of problem condition one. Procedure S, with conditions ligp through ngp interpolated between the first and second administrations Of Elli: is similar in design to a single week's testing series in the present study. Any decrement in the correlation between 2:1p and Ezlp: that might be found between procedure A and procedure S can be considered to be due to the differential learning phenomenon mentioned above. In procedure S, a high correlation between zplp and 3112 would establish the format of a problem condition as a reliable one, and therefore each problem within a condition could be considered to be a parallel form of the others. If, in following the suggested procedures, the assumption of differential learning is not upheld, and the reliability Of the dependent measures cannot be verified, it would mean that earlier findings would have to be qualified respective to this deficiency. Summary The present study was undertaken for the purpose of identifying and describing individual difference variables that are intrinsic to the learning situation. In addition, an attempt was made to determine the relationship between the intrinsic ID variables and selected dependent measures taken while the Sp were performing a concept learning task. The first phase Of the study involved the factor analysis Of 14 variables selected as measures of intrinsic sources of IDs. The factor analytical procedures were used in order to determine the interrelation- ships Of the variables and reduce them to a lesser number Of more fundamental variables. The analysis yielded five factors the interpreta- tions of which are summarized below: 56 Factor Interpretation A Strength Of the initial registration of the stimulus trace. B Visual Memory: immediate. C Field Independence D Susceptibility to Response Competition. E Neuroticism. Because of the lack Of reliability manifested by the dependent measures the prOposed second phase of the study (i.e., multiple re- gression analysis) was not carried out. It was evident from the data that the within subject variance, respective to a problem condition, was approximating the between subject variance. It was suggested that this unexpected relationship was due to a differential learning phenomenon that manifested itself as unstable performance character- istics on the part of Sp. Though the lack Of reliability prevented the second phase Of the study from being carried out, the author believes strongly in the validity of the proposed procedures. It is again emphasized that the success Of an individualized instructional program is dependent upon a complete understanding Of the IDs contributing to learning. It is hoped that future researchers will take note Of the shortcomings of this study and continue towards the goal Of filling this void in the knowledge Of instructional methodology. BIBLIOGRAPHY BIBLIOGRAPHY Allison, R. B. Learning parameters and human abilities. Princeton, New Jersey: Princeton University and Educational Testing Service, 1960. (Technical Report, Office Of Naval Research Contract NONR 1858-15.) Atkinson, R. C. and Shiffrin, R. M. Human memory: a prOposed system and its control processes. In K. W. Spence and J. T. Spence (Eds.) Advances inppsyphology of learningpand motivation research and theory. Vol. II. New York: Academic Press, 1968. Bourne, L. E., Jr., Goldstein, S., and Link, W. E. Concept learning as a function Of availability of previously presented informa- tion. Journal Of Experimental Psychology, 1964, 67, 439-448. Bruner, J. S., Goodnow, Jacqueline J” and Austin, G. A. A study of thinkipg. New York: John Wiley and Sons, Inc., 1956. Dickstein, L. 8. Field independence in concept attainment. Perceptual and Motor Skills, 1968, 27(2), 635-642. Duncanson, J. P. Intelligence and the ability to learn. Research Bulletin (RB-64-29). Princeton, N.J.: Educational Testing Service, 1964. Dunham, J. L., Guilford, J. P., and Hoepfner, R. Abilities pertaining to classes and the learning of concepts. Reports from the Psychological Laboratory, NO. 39, Los Angeles: University of Southern California, 1966. Eysenck, H. J. The structure of human personality. (2nd ed.) London: Methuen, 1960. Eysenck, H. J. and Eysenck, S. B. G. Eysenck personality inventory. San Diego: Educational and Industrial Testing Service, 1968(a). Eysenck, H. J. and Eysenck, S. B. G. Personality structure and measurement. San Diego: Knapp, 1968(b). Fleishman, E. A. The description and prediction Of perceptual-motor skill learning. In R. Glaser (ed.), Training research and education. Pittsburg: University Of Pittsburg Press, 1962. 57 58 Fleishman, E. A. Individual differences and motor learning. In R. M. Gagne (ed.), Learning and individual differences. Columbus: Charles E. Merrill Books, Inc., 1967, pp. 165-191. Friebergs, Vaira and Tulving, E. The effect Of practice on utilization of information from positive and negative instances in concept identification. Canadian Journal Of Psychology, 1961, 15, 101- 106. French, J. W., Ekstrom, R. B., and Price, L. A. Manual for Kit of refer- ence tests for cognitive factors. Princeton, N.J.: Educational Testing Service, 1963. Gardner, R. W., et a1. Personality organization of cognitive controls and intellectual abilities. Psychological Issues, 1960, 2, 289-300. Glaser, R. Some implications of previous work on learning and individual differences. In R. A. Gagne (ed.), Learning and Individual Differences. Columbus: Charles E. Merrill Books, Inc., 1967, pp. 1-18. Haygood, R. C. and Bourne, L. E., Jr. Attribute and rule learning aspects of conceptual behavior. Psychological Review, 1965, 72, 175-195. Hoyt, C. J. Test reliability estimated by analysis of variance. Psychometrika, 1941, 6, 153-160. Jensen, A. R. Individual differences in concept learning. In H. J. Klausmeier and C. W. Harris (eds.), Analysis Of Concept Learn- ing. New York: Academic Press, Inc., 1966, pp. 139-154. Jensen, A. R. Varieties Of individual differences in learning. In R. M. Gagne (ed.), Learning and individual differences. Columbus: Charles E. Merrill Books, Inc., 1967, pp. 117-135. Jensen, A. R., and Rohwer, W. B., Jr. The Stroop color word test: a review. Acta Psychologia, 25, 1966, 36-93. Kendler, H. H. The concept of the concept. In A. W. Melton (ed.), Categpries Of human learning. New York: Academic Press Inc., 1964, pp. 211-236. Lemke, E. A., Klausueier, H. J., and Harriss, C. W. Relationship of selected cognitive abilities to concept attainment and informa- tion processing. Journal Of Educational Psychology, 1967, 58, 27-35. Lezotte, L. W. The relationship between cognitive ability and the learning of structured and unstructured materials. Doctoral dissertation, Michigan State University, East Lansing, Michigan, 1969. 59 Logan, W. L. Unpublished research, Michigan State University, East Lansing, Michigan, 1969. Melton, A. W. Individual differences and theoretical process variables. In R. M. Gagne (ed.), Learning and individual differences. Columbus: Charles E. Merrill Books, Inc., 1967, pp. 238-252. Raven, J. C. Advanced progressive matrices. London: H. K. Lewis, 1965. Shulman, L. S., et a1. Structure Of inquiry process. Final Report, U. S. Department Of Health, Education and Welfare, 1968. Sloboda, W. and Smith, E. E. Disruption effects in human short-term 1 memory: Some negative findings. Percgptual and Motor Skills, 1968, 27(2), 575-582. Sperling, G. A. A model for visual memory tasks. Human Factors, 1963, 5, 19-31. Sperling, G. A. Successive approximations to a model for short term memory. Acta Psychologia, 1967, 27, 285-292. Stake, R. E. Learning parameters, aptitudes, and achievements. Technical Report. Princeton, N.J.: Princeton University and Educational Testing Service, June 1958. Walker, C. M., and Bourne, L. E., Jr. Concept identification as a function of amounts of relevant and irrelevant information. American Journal Of Psychology, 1961, 74, 410-417. Witkin, H. A., et al. Psychological differentiation. New York: John Wiley and CO., Inc., 1962. Waugh, N. C. and Norman, D. A. Primary memory. Psychological Review, 1965, 72, 89-104. Winer, B. J. Principles in experimental design. New York: McGraw-Hill Book CO., 1962. Witkin, H. A., et al. A manual for the embedded figures test. PalO Alto: Consulting Psychologists Press Inc., 1971. Woodrow, H. A. The ability to learn. Psychological Review, 1946, 53, 147-158. APPENDIX 60 no. om.m oh.N x663 ousmmoz uaooooaon ZOHHHmHDOU< MHDM 024 NHHHHm¢4H¢>< mDHDZHHm .wHHxMHmZOU Emmozoo .mmm3m mo mHme