AN EXPLORATGRY STUDY OF THE DERIVATiON AND RETENWON 0F HEGHER-QRBER CGDENG SCHEMES DU-RiHG THE FORMATEON 0F VERBAL ASSGCIATIONS Thesis gov Hm Dogma of DE. D. MECBEGAN STATE UNIVERSITY Donald J. Freeman 1967 ; L: a" LIBRARY E 5 MiChigan QM"! University ‘ I Thu-.315 a, i ' . Q- This is to certify that the thesis entitled An Exploratory Study of the Derivation and Retention of Higher—Order Coding Schemes During the Formation of Verbal Associations presented by Donald J. Freeman has been accepted towards fulfillment of the requirements for Ph .D o degree in Educat ion ' . - fl at kw» Major rofessor Date June 19, 1967 0-1694 ABSTRACT AN EXPLORATORY STUDY OF THE DERIVATION AND RETENTION OF HIGHER-ORDER CODING SCHEMES DURING THE FORMATION OF VERBAL ASSOCIATIONS by Donald J. Freeman Problem Recent investigations have shown that the comparative ease with which a given S is able to form a set of associ- ations is dependent upon the nature and quantity of the mediational links (encoding schemes) which he establishes between the stimulus and response terms (Underwood and Schulz, 1960; and Martin, Boersma, and Cox, 1965a). But these studies have focused on the perception of relations between the stimulus and response terms in each pair (first—order coding schemes). They have not determined whether or not Ss attempt to further condense the units of retention by looking for relations among the first-order coding schemes, i.e., by deriving higher-order coding schemes. Based on the assump— tion that 83 will form higher-order coding schemes, this dissertation attempted to answer three questions. First, what is the impact of the derivation of higher- order coding schemes on short-term retention? Second, what is Donald J. Freeman the impact of higher—order coding schemes on long-term re- tention? Third, what role does perception play in the derivation of higher—order coding schemes? Procedure In the two experiments reported in this dissertation, 85 were presented with a list of nine triads, each consisting of three familiar terms (e.g., mammal-bare-dear). Their task was to form associations among the three terms in each triad in such a way that the presentation of any one of them dur- ing the test trial would elicit the other two terms as responses. Because each triad was constructed according to the same relationship, or principle, it was possible for a given S to derive a single relationship which would apply for all nine triads. Through use of this list, the problem of identifying 85 who had derived higher-order coding schemes was reduced to that of determining whether or not a given S had discovered the "built-in" principle. The first two ques— tions were thus resolved by comparing the performance of 85 who discovered the principle with that of 85 who did not dis- cover the principle. The final question was approached through the use of lists with varying perceptual cues. Eight lists were con- structed according to all possible combinations of the two conditions on each of the following variables: Donald J. Freeman (1) Degree of embeddedness of the underlying principle - The principle underlying each triad was constructed according to one of the following two rules: (a) two-homonym lists - "If two terms in each triad are transformed into their homonyms, they represent exemplars of the third, or categori- cal term." (b) one—homonym lists - "If one term in each triad is transformed into its homonym, it and one other term represent exemplars of the third, or categorical term.“ e.g., mammal-bare—dear vs. mammal-bear-dear (2) Size of print - The categorical term in each triad was either capi- talized with the other two terms in small print (caps) or all three terms were printed in small letters (no caps). e.g. MAMMAL-bare-dear vs. mammal-bare-dear (5) Order of presentation - The categorical term was either fixed in the first position across all nine triads (fixed), or it was varied in position from one triad to the next (random). e.g., mammal—bare-dear bare-mammal-dear metal-steal-lead steal-lead—metal In an attempt to answer the third question, the per cent of 83 who discovered the principle as well as other general measures of performance were determined for each condition along the three variables. In Experiment I, 218 college sophomores were presented with eight learning trials and eight test trials. A question— naire (L.P.F.Q.) was administered at the completion of each Donald J. Freeman experimental session to determine which 85 had derived higher- order coding schemes. Long-term retention was also deter- mined by administering a single test trial three weeks after initial acquisition. The conditions in Experiment II were highly similar except that the 244 Ss in this experiment were presented with only one learning and one test trial. Major Findings The results of these two experiments provided partial answers to the above three questions. First, the short-term retention of those 83 who derived higher-order coding schemes in this task was clearly superior to the corresponding per- formance of Ss who derived different relations for each triad, i.e., a set of first-order coding schemes. 85 who discovered the principle not only formed the required associ- ations more rapidly than those who did not, but they also made fewer intrusions in recall. Only the difference in number of intrusions made on the first test trial of Experi- ment II failed to reach a statistically significant level. These findings were explained in terms of a simple extension of Underwood's two stage model of associational learning (1962) . Second, the derivation of higher—order coding schemes during acquisition apparently has little or no impact on long-term retention. Differences between 55 who either did Donald J. Freeman or did not discover the principle failed to reach statistical significance on any measure of long-term retention, including number of words correctly recalled and various types of in- trusions. Although several explanations may be advanced for this unexpected finding, the author attributed these results to the fact that a sizable number of 85 who discovered the principle during acquisition were apparently unable to recall the principle on the retention trial. This loss in retention was, in turn, accounted for by a simple modification of inter- ference theory. Finally, the results of this study failed to yield any clear conclusions regarding the role of perception in the derivation of higher-order coding. Although differences be- tween Ss presented one- and two-homonym lists were highly significant across all measures of short-term retention, the corresponding differences along the other two variables failed to ever reach statistical significance (alpha e .05). However, there was some evidence to suggest that capitalizing the categorical terms may have facilitated performance on one- homonym lists, while registering little impact on two-homonym lists. AN EXPLORATORY STUDY OF THE DERIVATION AND RETENTION OF HIGHER-ORDER CODING SCHEMES DURING THE FORMATION OF VERBAL ASSOCIATIONS BY q 0 A Donald J: Freeman A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY College of Education 1967 ACKNOWLEDGEMENTS There are a number of people who have made a significant contribution to this dissertation. A special word of thanks must first be extended to Dr. Clessen Martin, the chairman of my academic committee, who served as the primary source of guidance throughout the development of this dissertation. I am also sincerely appreciative of the assistance provided by the other members of this committee--Dr. Abram Barch, Dr. Lee Shulman, and Dr. Ted Ward. Each of these individuals played a critical role in the design and development of this dissertation. I would also like to thank the following colleagues for their assistance during the data collection phase of this investigation: David Cox, Ronald Crowell, William Darnell, and Brian Bolton. Conversations with these individuals also led to many refinements in the theory and methodology on which this study is based. Above all, I am indebted to my wife Kathy for her abun- dant patience, encouragement, and secretarial assistance. ii TABLE OF CONTENTS CHAPTER I. STATEMENT OF THE PROBLEM . . . . . . . . . . I. Statement of the Problem . . . . . . . II. Statement of the Hypotheses. . . . . . I I C RELEVANT RE SEARCH 0 O O O O O O O O O I O O 0 I. General Overview of Chapter II . . . . II. Research Involving the Use of Encoding Schemes in Paired-Associate Learning . III. General Evidence Related to the Stated Problem of the Investigation . . . . . A. The Effect of Higher-Order Coding Schemes on Short-Term Retention . B. The Effect of Higher-Order Coding Schemes on Long-Term Retention. . C. The Role of Perception in the Formation of Higher-Order Coding Schemes . . . . . . . . . . . . . III. METHOD AND PROCEDURE . . . . . . . . . . . . I. General Overview of the Experimental Design . . . . . . . . . . . . . . . . II. Experiment I . . . . . . . . . . . . . Subjects. . . . . . . . . . . . . Materials . . . . . . . . . . . . Procedure . . . . . . . . . . . . III. Experiment II. . . . . . . . . . . . . Subjects. . . . . . . . . . . . . Materials . . . . . . . . . . . . Procedure . . . . . . . . . . . . IV. Level of Principle-Formation Question- naire. . . . . . . . . . . . . . . . . IV. RESULTS. . -.- . . . . . . . . . . . . . . . I. Definition of the Dependent Variables. II. Higher—Order Coding Schemes and Short— Term Retention . . . . . . . . . . . . iii Page 12 19 19 19 31 51 55 56 4O 4O 42 42 42 46 49 49 49 49 51 62 62 64 n“ w. TABLE OF CONTENTS - Continued CHAPTER III. Configural Perception and the Encoding Process. . . . . . . . . . . . . . . . IV. Higher-Order Coding Schemes and Long- Term Retention . . . . . . . . . . . . V. DISCUSSION AND CONCLUSIONS . . . . . . . . . I. Higher-Order Coding Schemes and Short- Term Retention . . . . . . . . . . . . II. Configural Perception and the Encoding Process. . . . . . . . . . . . . . . . III. Higher-Order Coding Schemes and Long- Term Retention . . . . . . . .y. . . . IV. Implications for Education . . . . . . V. Conclusions. . . . . . . . . . . . . . BIBLIOGRAPW. O O O O O O O O O O O O O O O O O O O APPENDICES. . . . . . . . . . . . . . . . . . . . . iv Page 75 100 108 108 117 128 142 147 150 165 LIST OF TABLES TABLE 1. List of triads which were presented during the study. . . . . . . . . . . . . . . . . . . . . . 2. The operational criteria used to assign indi- viduals to one of four levels of principle form- ation according to their responses to the Level of Principle Formation Questionnaire . . . . . . 5. The four criteria ultimately adopted to identify 85 who discovered the principle during acquisi- tion 0 O O I C C C O O I O O O C O O O O O O O O 4. Tests of differences between the mean perform- ance levels of 83 who either did, or did not, discover the principle at some point during Experiment I . . . . . . . . . . . . . . . . . . 5. Tests of differences in the first test trial performance of 85 who either did or did not dis- cover the principle during the first learning trial of Experiment II . . . . . . . . . . . . . 6. Per cent of 85 who discovered the principle during the first learning trial of Experiment II (column I) or at some point during the eight learning trials of Experiment II (column II) when presented lists with varying perceptual cues . . . . . . . . . . . . . . . . . . . . . . 7. A 2x2x2, fixed-effects, analysis of variance test of the number of words correctly recalled on the first test trial by Ss presented lists with varying perceptual cues . . . . . . . . . . 8. A 2x2x2 fixed effects analysis of variance test of the number of trials taken to reach criterion by Ss presented lists with varying perceptual cues (Experiment I). . . . . . . . . . . . . . . Page 44 52 56 69 72 76 81 82 LIST OF TABLES - Continued TABLE Page 9. A 2x2x2 fixed effects analysis of variance test of the total number of intrusions made by 85 presented lists with varying perceptual cues (Experiment I) . . . . . . . . . . . . . . . . . 82 10. Distribution of means in the interaction between number of homonyms and capitalization of the categorical term . . . . . . . . . . . . . . . . 85 11. Results of a 2x2x2x8 analysis of variance test of the total number of words recalled by Ss pre- sented lists with varying perceptual cues. . . . 91 12. Results of a 2x2x2x5 fixed—effects analysis of variance test of the number of words recalled over the first three test trials of Experiment I, with "type of response" as the fourth variable . 97 13. Results of a 2x2x2x3 fixed—effects analysis of variance test of the number of words recalled over the first three test trials of Experiment I, with "type of association" as the fourth inde- pendent variable . . . . . . . . . . . . . . . . 99 14. Mean performance levels on the test of long-term retention for those 83 who either did or did not discover the principle during acquisition. . . . 105 15. Results of an analysis of variance test of per- formance on the test for long—term retention of 83 presented lists with varying perceptual cues. 104 16. The mean number of words correctly recalled over the first three test trials (total possible = 18) for each condition in the interaction be- tween number of homonyms and type of associ- ation. . . . . . . . . . . . . . . . . . . . . . 121 17. Mean number of words correctly recalled over the first three test trials for each triad (total possible = 6.00) . . . . . . . . . . . . . . . . 182 18. Results of a one-way analysis of variance test of the differences among 83 classified at dif- ferent levels of principle formation . . . . . . 185 vi LIST OF TABLES - Continued TABLE 19. 20. 21. 22. 25. Results of a one-way analysis of variance test of differences among individuals classified according to the four criteria described in Table 5. . . . . . . . . . . . . . . . . . . . . Differences in the mean performance levels of 55 who either did or did not discover the principle according to classifications made by the Independent Judge for Experiment I . . . . . . . The most frequently occurring extra-list intru- sions for one- and two-homonym lists . . . . . . One-tailed t-tests of differences in short-term retention among 83 who either did, or did not, discover the principle, with only those 85 who ultimately reached criterion included in the sample . . . . . . . . . . . . . . . . . . . . . One-tailed t-tests of differences among 85 in the two-homonym group who discovered the princi- ple and who either did, or did not, make spelle ing errors on the test for long-term retention . vii Page 185 186 188 189 191 LIST OF FIGURES FIGURE 1. Mean number of words correctly recalled over eight test trials by 85 who either did, or did not, discover the principle (Experiment I). 2. Learning curves (number of words correctly recalled) for Ss presented lists with varying perceptual cues (Experiment II) . . . . . . . . 5. Cumulative proportion of Spelling distortions made by Ss who either did, or did not, discover the principle (Experiment I). . . . . . . . . . viii Page 67 88 116 LIST OF APPENDICES APPENDIX Page A. Instructions Given at Some Point During the Experimental Session. . . . . . . . . . . . . 166 I. General Instructions Given at the Begin- ning of Each Experimental Session. . . . 167 II. Instructions Given Between the First Learning and First Test Trial. . . . . . 169 III. Instructions Given at the End of the First Test Trial in Both Experiment I and Experiment II. . . . . . . . . . . . 169 IV. Instructions Given Immediately Prior to the Test for Long—Term Retention . . . . 170 V. Instructions Given Prior to the Comple— tion of the Level of Principle Formation Questionnaire in Experiment II . . . . . 170 B. The Level of Principle Formation Question- naire O O O O O O O O O O O O O C O O O O O O 171 C. Table 17 through 25--Additional Analyses of the Data. . . . . . . . . . . . . . . . . . . 181 ix H- 3“ ‘p In '1' ' ‘8‘] NM (I) (D “ VJ CHAPTER I STATEMENT OF THE PROBLEM I. Statement of the Problem Several recent investigations have shown that performance on associational tasks is improved when Ss interject some form of mediational link between the stimulus and response terms (Underwood and Schulz, 1965 a; Martin, Boersma, and Cox, 1965 a; and Cox, 1965). Several explanations might be advanced regarding the impact of this mediational activity. However, an approach which seems especially fruitful suggests that the critical function which these links serve is that of somehow uniting the two terms in each pair. Epstein, Rock, and Zuckerman (1960), for example, have shown that associations between pairs of nouns presented as meaningful units (e.g., CAKE near ROAD) are formed more rapidly than associations between the same two nouns presented in a non-unit form (e.g., CAKE and ROAD). This view of the function of mediational links repre- sents a direct application of Miller's (1956 a) analysis of hunuan learning. Miller suggests that during any learning task, individuals actively attempt to transform incoming information into a small number of compact units of retention. When presented with the task of retaining the number series 2, 5, 8, 11, 14, 17, and 20, for example, most individuals perceive the underlying relationship—~each number differs from the preceding number by three. By transforming the number series into a sentence which describes this relation- ship, the task of retaining the complete series is reduced from that of retaining seven independent numbers to the task of retaining a single sentence. In an analogous fashion, the formation of mediational links between each stimulus and response term in an associational task may reduce the number of units of information with which an individual must effec- tively deal. Miller further argues that an individual is forced to make these transformations because the human storage system can only absorb a limited number of units or "chunks of information" at any one point in time. It is as if each storage register could accept one of a tremendous variety of alternative symbols, but the number of registers available was quite limited. (Miller, 1956 a, p. 129.) Therefore, Miller (1956 a) depicts the process of transform- ing information into more efficient units as analogous to Carrying a purse which will hold only seven coins, irre- SPective of the monetary value of each. In the same way that 5“? Jindividual can carry more money if the coins are silver dCXIILars, instead of pennies, an individual can retain more information if he stores efficient units of recall, rather t115111 informationally poor units. Bruner (1959) appears to be in complete agreement with Miller. At one point he states . . . One of the most notable things about the human mind is its limited capacity for dealing at any one moment with diverse arrays of information. . . . The seven things we deal with must be worth their weight. (Bruner, 1959, p. 77.) The overall process of transforming incoming information into a smaller number of efficient units of retention shall be referred to as the encoding process throughout this dis- sertation. Other authors have used the terms "recoding" (Miller, 1956 a) and "decoding" (Osgood, 1955) to depict the same phenomenon. The particular form which a given trans— formation takes shall be referred to as either a coding scheme or an encoding scheme (e.g., naming and describing each figure as it is presented while attempting to retain a series of figures; a statement of the relationship in the above number series, etc.)1 These terms have been adopted in preference to the terms "mediation" and "verbalization" because the latter have a very general reference. Coding schemes, on the other hand, refer only to an attempt on the part of the individual to reduce the number of units of retention. The process of encoding or storing information represents only one phase of the learning process. Once information has been encoded or stored, the task of drawing this information out of storage still confronts the individual. This process lThe terms "coding scheme" and "encoding scheme" are used interchangeably throughout this dissertation. of regenerating the stored information will be referred to as the decoding process throughout this dissertation. Decod- ing in associational learning corresponds to the S's attempts to generate the appropriate response term as each stimulus is presented. Few would question the assertion that encoding and decod- ing are interrelated processes. It therefore seems reasonable to suggest that the nature of the coding schemes which are derived during encoding will have a significant impact on both the nature and extent of decoding. A S who discovers the relationship among the numbers in the above series, for example, should be able to recall more digits during decoding than a S who does not discover this relationship. And, as noted in the introductory statement, associations are formed more rapidly between pairs of terms which are linked by some form of coding scheme than between pairs for which no medi- ational link has been established (Underwood and Schulz, 1960; Martin §£__l,, 1965 a; and Cox, 1965). However, several questions regarding the nature of encod- ing and decoding during the formation of verbal associations have not been answered by empirical research. The following example should provide the necessary background for a dis- cussion of some of these issues. Imagine that three individuals are asked to serve as subjects in a paired-associates task involving the following three pairs of familiar terms: tree - wood mirror - reflection river - water Further imagine that the encoding schemes adopted by these three individuals take the following form: Individual no. 1: transforms the two words in each pair into a compact unit by forming a sentence which links the two words together; namely . . . l'WOOD comes from TREES." "I see my REFLECTION in a MIRROR." "The WATER in the RIVER is cold." Individual no. 2: initially derives sentences of this type, but then discovers that the sentences he has derived may, in turn, be organized into a short story or theme; namely . . . "When I look at my REFLECTION in the MIRROR, it reminds me of staring into the clear WATER of a RIVER. The experience is as re- freshing as chopping WOOD in a grove of TREES." Individual no. 5: discovers the functional relationship between the first pair of terms; i.e., "WOOD comes from TREES." He then examines each succeeding pair on the list to determine whether or not they fit this same relationship. Since REFLECTIONS do come from MIRRORS and WATER does come from RIVERS, this search results in the discovery of a single relationship which will hold for all three pairs of terms; namely . . . "(A given response term) comes from (The rele- vant stimulus term.)" It should be apparent from this example that even in comparatively simple associational tasks, diverse forms of encoding may occur. Individual one, for example, has derived three independent coding schemes, each of which unites a .- n“ ‘9‘" d d are unV $5: (I. .—h - fly?“ Ebb .. 0" 0.. .A -\u v ‘OIL Y“. (I) (b v.. single pair of stimulus and response terms. Coding schemes of this type which are based on relationships between only one set of stimulus and response terms will be referred to as first-order coding schemes throughout this dissertation. Individuals two and three, on the other hand, have re— duced the number of coding schemes to one-—a theme and a general relationship, respectively. Coding schemes of this type which are based on relationships between two or more first-order coding schemes, and which effectively reduce the number of units which the 8 must recall, will be referred to as higher-order coding schemes throughout this dissertation. This illustration gives rise to a basic assumption of this dissertation; namely, when individuals are presented with a learning task involving the formation of verbal associ- ations, a significant prOportion of these Ss will attempt to formulate higher-order coding schemes. The following set of questions, which are based on this assumption, therefore, serve as the focus of this investigation. Given that a group of individuals has been presented with a task which involves the formation of verbal associations . . . (1) What effect, if any, does the derivation of higher-order coding schemes have on short-term retention? (2) What effect, if any, does the derivation of higher-order coding schemes have on long-term retention? (5) Does an individual's overall perception of the stimulus and response terms affect the prob- ability that he will derive a higher-order coding scheme? Since available literature fails to provide an answer to the more fundamental question of whether or not 85 will attempt to formulate higher-order coding schemes during associational tasks, there is little or no direct basis for predicting an answer to any of these specific questions. This is not to say, however, that no other authors have re- ferred to the basic question of whether or not Ss will attempt to formulate higher—order coding schemes. The following statement by Underwood and Schulz (1960), for example, raises the question of whether or not 85 will attempt to utilize the same relationship for more than one pair of terms (similar to individual no. 5 in the above illustration). Commenting on the verbal reports of a group of Ss who had just completed a paired associate task involving pairs of nonsense trigrams and familiar three-letter words, these authors state . . . Another fact we have not tried to express numerically, appears quite universally in the subjects‘ reports. This is the fact that a subject will use several types of mediators in learning the list. It is possible that the subject tried to use a consistent mediation and failed; all we know is that at the end of learning the associations are quite varied in type.2 (Underwood and Schulz, 1960, p. 500.) Although Battig (1966) does not argue that Ss will attempt to use the same mediator for more than one pair of terms, he does argue that 85 do not learn each pair inde— pendently. Commenting on an experiment in which 85 learned 2Underlining has been added by this author. pairs of nonsense shapes and two-digit numbers, he states . . . . . we were both surprised and impressed by the frequency of reports indicating some sort of inter- relating or grouping of two or more pairs together, instead of each pair being learned as a separate indi- vidual entity. (Battig, 1966, p. 178.) As a result of these observations, Battig undertook a series of experiments which demonstrated that under experi- mental conditions which favored the formation of inter-pair groups according to similarities in degree of learning, fewer errors were made than under conditions which did not favor such grouping. For example, when pairs were presented in the same serial position until responded to correctly and then varied in position from trial to trial, fewer errors were made than when the position of every pair was changed on each trial. Battig attributed these findings to the 53' use of a form of "subjective organization" whereby the number of units of retention were reduced. At one point in the discussion he states . . . Since it is the rare paired-associate list in which the number of pairs does not exceed the 'magical number seven', it is highly unlikely that the S could simultaneously carry each individual pair in the form of a separate bit of information. Consequently, some form of grouping or recoding into higher-order multiple- pair chunks may be a necessary prerequisite to successful performance on paired-associate tasks. (Battig, 1966, p. 181.) Thus, Battig not only argues that S's form higher-order coding schemes, but he also maintains that the use of these schemes facilitates performance on paired-associate tasks. However, one may question whether the S's initial attempts to form interrelationships among more than one pair of terms will be based on similarities in item difficulty or degree of learning. It seems more reasonable to suggest that such bases will be adopted only under those conditions in which more efficient approaches have failed (e.g., deriving the same relationship among the terms in more than one pair). Furthermore, Battig's evidence for the facilitating effect of higher-order coding schemes is somewhat questionable. Other differences exist between the experimental and control con- ditions which may also favor the experimental group (e.g., presence or absence of position cues). In short, Battig's study points to the need for further research regarding the role of higher-order coding schemes in associational learning. Other authors have also suggested that 83 may attempt to form themes during associational tasks (similar to indi- vidual no. 2 in the preceding illustration). Miller, Gallanter, and Pribram (1960), for example, maintain that individuals faced with a paired-associates task begin their transformations by forming words from the dominant aspects of nonsense syllables. These words are then organized into sentences which, in turn, are organized into themes. Of major importance, however, is the fact that these authors rely on logical, rather than empirical, support for their asser- tions. ' In short, at least three other authors have made some reference to the basic question of whether or not 85 will 10 attempt to formulate higher—order coding schemes during an associational task. But, with the possible exception of Battig (1966), none of these authors have successfully treated this question as the subject of an empirical investi- gation. One problem which may have curtailed attempts to answer this question empirically is that of finding a suitable method for studying the derivation of higher-order coding schemes. An investigation which involves the collection of verbal reports, for example, must overcome several problems which stem from the idiosyncratic nature of the higher-order coding schemes which individuals may adopt. The technique of present- ing Ss with some form of higher-order coding scheme immediately prior to the first learning trial is also somewhat inappro— priate in that it never really answers the question of whether or not 83 would have developed such schemes on their own. In short, the two methods which have been adopted in other in— vestigations of encoding during associational learning do not lend themselves to an initial study of higher—order coding schemes. Fortunately, however, an answer to this methodological problem is suggested in the literature; namely, the technique of building some a priori structure into the list of stimulus materials. This technique has proved especially fruitful in studies of encoding involving the free recall of serial lists (examples include Miller and Selfridge, 1950; Bousfield, 1955; Epstein, 1961; Underwood and Keppel, 1965; and Lindley, 11 1965). But it has rarely, if ever, been applied to studies of associative learning. In this investigation, nine triads of familiar terms were constructed according to the same relationship or prin- ciple; namely, when two of the terms in each triad are trans— formed into their homonyms, they represent exemplars of the third or categorical term (e.g., bare-dear—mammal).3 During the learning, or encoding trials, each S attempted to form associations among all three terms. During the decoding, or test trials, one term from each triad was presented, and Ss attempted to write the other two terms. Because each triad was constructed according to the same principle, any S who attempted to look for common rela- tionships among the terms in more than one triad should ultimately have discovered the "built-in" principle. Identi— fication of 83 who have derived higher-order coding schemes was, therefore, reduced to the problem of determining whether or not a given S had discovered this consistent relationship. A questionnaire which was administered soon after the com- pletion of the task was used in making these identifications.4 3Lists were also constructed in which only one term was written as a homonym (e.g., bear-dear-mammal). These two lists, known as two-homonym and one-homonym lists, respectively, are presented in Table 1 in Chapter III. 4This questionnaire, known as the Level of Principle- Formation Questionnaire (L.P.F.Q.), was actually designed to identify 85 at four different levels ranging from Ss who formulated no first-order coding schemes to 85 who derived a Single higher-order coding scheme for all nine triads; i.e., 12 This resolution of the methodological problem cleared the way for an attempt to answer the three specific questions posed above. The following statement of the hypotheses sug- gests how these answers were derived. II. Statement of the Hypotheses The following hypotheses are based on the three questions which were stated in the preceding section of this chapter.5 A. The Derivation of higher-order coding schemes facilitates acquisition or short-term retention: Hypothesis 1: (When level of principle formation is determined soon after the S reaches criterion or after the eighth test trial if he fails to reach criterion.) The mean level of performance over all eight test trials will be greater for those individuals at the higher levels of principle formation than for those at the lower levels. ' Corollary 1a: The mean number of trials to criterion will be lower for $5 at the higher levels of principle formation than for 85 at the lower levels of principle formation. Corollary 1b: The mean number of words recalled over all eight test trials will be greater for 83 at the higher levels of principle formation than for Ss at the lower levels of principle formation. discovered the principle. Since these four levels might be said to correspond to the process of discovering a common relationship or principle, they will be referred to as levels or principle formation throughout this dissertation. 5Many of the hypotheses refer to "levels of principle formation." These levels refer to successive steps in the discovery of the "built-in" principle, ranging from the formulation of no first-order coding schemes to the dis- ?Overy and utilization of a single higher-order coding scheme; l-e., the principle. For a more complete description, see Table 2 in Chapter III. 15 Hypothesis 2:6 With one exception, Ss at the higher levels of principle formation will make fewer intrusions in recall over all eight test trials than $5 at the lower levels. The single exception is number of improper plurals where the opposite prediction is made. Corollary 2a: The mean number of spelling distor- tions made by 85 at the higher levels of principle formation will be lower than the corresponding mean for $5 at the lower levels of principle formation. Corollary72b: The mean number of extra-list intru- sions made by 85 at the higher levels of principle formation will be lower than the corresponding mean for $5 at the lower levels of principle formation. Corollary 2c: The mean number of intra-list intru- tions made by 85 at the higher levels of principle formation will be lower than the corresponding mean for $5 at the lower levels of principle formation. Corollary 2d: The mean number of improper plurals formed by $5 at the higher levels of principle for- mation will be greater than the corresponding mean for Ss at the lower levels of principle formation. Corollary 2e: The mean total number of intrusions made by 83 at the higher levels of principle for- mation will be lower than the corresponding mean for 88 at the lower levels of principle formation. Hypothesis 5: (When level of principle formation is determined soon after the first test trial.) The mean number of words recalled on the first test trial by $5 at the higher levels of principle formation will be greater than the corresponding mean for 85 at the lower levels of principle formation. Hypothesis 4: With one exception, 85 who derive higher levels of principle formation during the first learning trial will make fewer intrusions in recall on the first test trial than Ss at the lower levels. The single exception is the number of improper plurals where the Opposite prediction is made. (If scores on the first 6Hypotheses involving the number of intrusions made by Ss (Hypotheses 2, 4, 9, and 12) were not included in the proposal on which this dissertation is based. However, these hypotheses were formulated prior to the derivation of scores on these variables and are therefore not post hoc in nature. 14 test trial of Experiment II are interchanged with scores over all eight test trials of Experiment I, the five corollaries of hypothesis four are identical to those of hypothesis two. For purposes of convenience, therefore, these corollaries will not be repeated here.) B. A 85 overall perception of the stimulus and response terms will affect the likelihood that he will derive a higher- order coding scheme: Hypothesis 5: (When level of principle formation is determined soon after the first test trial.) Those 88 who are assigned lists with positive perceptual cues will have educed higher levels of principle formation during the first learning trial than those assigned lists which lack these cues. Corollary 5a: At the end of the first test trial, the median level of principle formation will be greater for 85 presented lists with one homonym than for 85 presented lists with two homonyms. Corollary 5b: At the end of the first test trial, the median level of principle formation will be greater for 85 presented lists in which the cate- gorical terms are capitalized than for 85 presented lists in which the categorical terms are not capi- talized. Corollary 5c: At the end of the first test trial, the median level of principle formation will be greater for 85 presented lists in which the cate- gorical term is always fixed in the first position than for Ss presented lists in which the categorical term varies in position from triad to triad. Hypothesis 6: (When level of principle formation is determined soon after the subject reaches criterion or at the end of the eighth test trial if he fails to reach criterion.) Those 53 who are assigned lists with posi- tive perceptual cues will have educed higher levels of principle formation during the eight learning trials than Ss assigned lists which lack these cues. Corollary 6a: At the completion of the task, the median level of principle formation will be greater for 55 presented lists with one homonym than for 85 presented lists with two homonyms. 15 Corollary 6b: At the completion of the task, the median level of principle formation will be greater for 55 presented lists in which the categorical terms are capitalized than for Ss presented lists in which the categorical terms are not capitalized. Corollary 6c: At the completion of the task, the median level of principle formation will be greater for individuals presented lists in which the cate- gorical term is always fixed in the first position than for Ss presented lists in which the categorical term varies in position from triad to triad. Hypothesis 7: Because of their effect on the derived level of principle formation, positive perceptual cues will aid in the formation of associations as early as the first learning trial. Corollapy 7a: The mean number of words correctly recalled on the first test trial by 55 presented lists with one homonym will be greater than the correspond- ing mean for 88 presented lists with two homonyms. Corollary 7b: The mean number of words correctly re- called on the first test trial by 55 presented lists in which the categorical terms are capitalized will be greater than the corresponding mean for Ss pre- sented lists in which the categorical terms are not capitalized. Corpllary 7c: The mean number of words correctly recalled on the first test trial by Ss presented lists in which the categorical terms are fixed in the first position will be greater than the correspond- ing mean for Ss presented lists in which the cate- gorical terms are not fixed in the first position. Hypothesis 8: Because of their effect on the derived level of principle formation, lists containing positive perceptual cues will be learned more rapidly than lists which lack these cues. Corollary 8a: The mean number of trials to criterion for 53 presented lists with one homonym will be less than the corresponding mean for 85 presented lists with two homonyms. Corollary 8b: The mean number of trials to criterion for 35 presented lists in which the categorical terms are capitalized will be less than the corresponding mean for Ss presented lists in which the categorical terms are not capitalized. 16 Corollary 8c: The mean number of trials to cri- terion for Ss presented lists in which the categor- ical terms are capitalized will be less than the corresponding mean for Ss presented lists in which the categorical terms are not capitalized. Hypothesis 9: Because of their effect on the derived level of principle formation, lists with positive per- ceptual cues will elicit fewer intrusions in recall than lists which lack these cues. Corollary 9a: The mean total number of intrusions made by 35 presented lists with one homonym will be less than the corresponding mean for 85 presented lists with two homonyms. Corollary 9b: The mean total number of intrusions made by 85 presented lists in which the categorical terms are capitalized will be less than the corres— ponding mean for 53 presented lists in which the categorical terms are not capitalized. Corollaryg9c: The mean total number of intrusions made by 88 presented lists in which the categorical term is fixed in the first position will be less than the corresponding mean for Ss presented lists in which the categorical term is not fixed in the first position. Hypothesis 10: Because of their effect on the derived level of principle formation, positive perceptual cues will enhance the formation of verbal associations through- out the learning task. Corollary 10a: The mean total number of words re— called over all eight test trials by 55 presented lists with one homonym will be greater than the corresponding mean for 85 presented lists with two homonyms. Corollary 10b: The mean total number of words re- called over all eight test trials by Ss presented lists in which the categorical terms are capitalized will be greater than the corresponding mean for Ss presented lists in which the categorical terms are not capitalized. Corollary 10c: The mean total number of words re- called over all eight test trials by Ss presented lists in which the categorical terms are fixed in the first position will be greater than the corres- ponding mean for 35 presented lists in which the categorical terms are not fixed in the first position. 17 C. The derivation of higher—order coding schemes during acguisition will result in higher levels ofpperformance on the test for long-term retention: Hypothegis 11: The mean number of words correctly re— called on the test for long—term retention by 85 at the higher levels of principle formation during acquisition will be greater than the corresponding mean for Ss at the lower levels of principle formation. Hypothesis 12: With one exception, the number of intru- sions made on the test of long-term retention will be lower for $8 at the higher levels of principle formation than for $8 at the lower levels of principle formation. The single exception is number of improper plurals where the opposite prediction is made. Corollary 12a: The mean number of spelling distor- tions made by 85 at the higher levels of principle formation will be less than the corresponding mean for $5 at the lower levels of principle formation. Corollary 12b: The mean number of extra-list intru- sions made by $5 at the higher levels of principle formation will be less than the corresponding mean for $5 at the lower levels of principle formation. Corollary 12c: The mean number of intra-list intru- sions made by 85 at the higher levels of principle formation will be less than the corresponding mean for $5 at the lower levels of principle formation. Corollary 12d: The mean number of improper plurals formed by 85 at the higher levels of principle forma— tion will be greater than the corresponding mean for Ss at the lower levels of principle formation. Corollary 12e: The mean total number of intrusions formed by 88 at the higher levels of principle forma- tion will be less than the corresponding mean for $5 at the lower levels of principle formation. Hypothesis 15: Because of their effect on the derived levels of principle formation, positive perceptual cues will aid in long-term retention. Corollary 15a: The mean number of words correctly recalled on the test of long-term retention by 85 presented lists with one homonym will be greater than the corresponding mean for 88 presented lists with two homonyms. 18 Corollary 15b: The mean number of words correctly recalled on the test of long-term retention by 85 presented lists in which the categorical terms are capitalized will be greater than the corresponding mean for 85 presented lists in which the categorical terms are not capitalized. Corollary 15c: The mean number of words correctly recalled on the test of long-term retention by 85 presented lists in which the categorical terms are fixed in the first position will be greater than the corresponding mean for 85 presented lists in which the categorical terms are not fixed in the first position. Hypothesis 14: Among Ss who discover the principle dur— ing acquisition, the mean number of misspelled exemplar terms on the test of long-term retention made by Ss pre- sented one-homonym lists will be greater than the corres— ponding mean for Ss presented two-homonym lists. CHAPTER I I RELEVANT RE SEARCH I. General Overview of Chapter II The function of this Chapter is to review that research which has some bearing on one or more of the three questions which were raised in the statement of the problem. Section II of this Chapter, for example, describes research which has been concerned with the role of encoding in paired associ- ate learning. The purpose of section II is to provide a general background of the methods and issues which have characterized research in this area. Section III then re- examines each of the three questions in terms of relevant re— search. The purpose of section III is to provide evidence for the general predictions which were made in Chapter I. II. The Use of Encoding Schemes in Paired-Associate Learning Three comparatively independent approaches have been adopted in an attempt to study the use of encoding schemes in the formation of verbal associations--experimenta1 establish- ment of a mediational chain via training on one or more lists; presentation of a single list, followed by verbal reports; 19 20 and, presentation of some form of encoding scheme as a part of the basic instructions. The paradigm which underlies the majority of studies based on the first approach is to require 85 to learn one list, (A—B); followed by a second list, (B—C); followed by still a third, or critical list, (A-C). Performance on the final list is believed to be based on the use of an estab- lished mediational chain; namely, A-B-C. In other words, the "B" terms are believed to form a simple mediational link between "A" and "C". Although most investigators have adopted this paradigm in an unmodified form, a few others have utilized minor vari- ations in design. The B-C relationship, for example, is sometimes assumed to exist, and therefore only one training list, A-B, is presented (Foley and Cofer, 1945; Russell and Storms, 1955; and, Barnes and Underwood, 1959). The direction of the relationships established in the training phase has also been varied (Horton and Kjeldergaard, 1961). Finally, different forms of learning, such as serial lists have been used to establish the mediational chain (Foley and Cofer, 1945; and,Ri¢hardson, 1962). The results of studies based on this approach are some- what contradictory. The majority of investigators report that the establishment of a mediational chain via previous training facilitates performance on the final list (Foley and Cofer, 1945; Murdock, 1952; Bugelski and Scharlock, 1952; 21 Russell and Storms, 1955; Horton and Kjeldergaard, 1961; Richardson, 1962, 1966; Schulz and Lovelace, 1964; Goulet, 1966; and Shanmugan and Miron, 1966). But other authors main- tain that little or no mediation occurs as a result of train- ing (Peters, 1955; Katona, 1960; Barclay, 1961; and Mandler and Earhard, 1964). Two of the latter authors have proposed alternative explanations for the positive effects which have been observed. Barclay (1961), for example, found that differential reinforce- ment in the establishment of the B-C associations had no effect on the comparative speed with which the A-C associations were formed. He therefore argued that the facilitating ef- fects which were reported in other studies could be attributed to general transfer of training such as "learning to learn" or to simple stimulus generalization without any reference to mediation. But the findings on which this conclusion is based have not gone unchallenged. Shanmugan and Miron (1966), for example, found that transfer did increase as the degree of learning on the second, (B-C), list increased. Mandler and Earhard (1964) have also proposed an alterna- tive. These authors demonstrated that a list of A-E associ- ations is learned more rapidly when preceded by lists A-B and B—C than when preceded by lists A-B and D—C. They there— fore argued that as learning progresses on the second, (B-C), list, the A-B associations undergo extinction in the experi- mental paradigm (A-B, B-C, A-C), while remaining intact in the p.- by I 7 A I. (1 pa I U r. u. .1 22 control paradigm (AFB, D-C, A-C). These intact A-B associ- ations then interfere with the formation of A-C associations for Ss in the control group, while no comparable interference is present for Ss in the experimental group. It is, there- fore, the presence or absence of interference conditions rather than the presence or absence of mediational chains which gives rise to the obtained differences in performance between these two groups. But these conclusions have also been challenged. Goulet (1966), for example, found that the retention of A-B associations following second list learning was as high, or higher, under experimental conditions as under control conditions. It therefore seems reasonable to conclude that the estab— lishment of mediational chains (encoding schemes) does facili- tate performance on the critical A—C lists. Further, the extent to which these chains facilitate performance has been shown to be directly related to the degree of second stage learning (Shanmugan and Miron, 1966); the length of the an- ticipation interval (Schulz and Lovelace, 1964; and Richardson, 1966); and the polarization of the mediating ("B") terms (Shanmugan and Miron, 1966). In short, the procedure of establishing a coding scheme via training on other paired associate lists has been reasonably fruitful in answering the general question of whether or not coding schemes have an effect on the formation of verbal associations. 25 The second technique--presentation of a single list, followed by verbal reports--has yielded an even more defini- tive answer to this basic question. This technique was used as early as 1918 when Reed (1918) presented 85 with a series of paired-associate lists and then asked them to report any aids which they had interjected between the stimulus and response terms. He found that whenever an associational aid was reported, the corresponding pair of terms was learned more rapidly and retained longer than pairs for which no association was reported. Reed also developed a crude classification sys- tem of the types of associations which were formed. He noted, for example, that logical associations have a greater impact on performance than associations based on some "likeness in sensory quality" such as the sound or sight of letters. Reed's study was apparently followed by a long interval in which psychologists failed to ask 85 how they formed associ- ations. Beginning about ten years ago, however, verbal reports began to find their way back into studies of paired- associate learning. Use of this procedure by contemporary psychologists has confirmed most of Reed's original findings. In 1957, for example, Rock (1957) collected verbal re- ports following a paired-associate task. Noting that most Ss derived at least some mnemonic devices in forming the associations, Rock stated . . . The theoretical significance of the widespread use of such devices in rote learning experiments has not been sufficiently emphasized in the past. The success— ful use of such devices may mean that an idea suddenly 24 occurs to a S which enables him to link two items then and there; it has, to some extent, the character of insightful learning. (Rock, 1957, p. 191.) It remained for Mattocks (as reported in Underwood and Schulz, 1960) to show that these devices aid in the formation of verbal associations. Mattocks collected verbal reports following a paired-associate task which involved low meaning- ful trigrams as stimulus terms and common three letter words as response terms. He found that pairs which were linked by some mediational scheme involving pre-existing associations were learned more rapidly than pairs for which the S reported no associations. Using a somewhat different approach, Clark, Lansford, and Dallenbach (1960) reported that under experi- mental conditions which favored the formation of mnemonic devices (e.g., long intertrial intervals and long exposure times), verbal associations were formed almost twice as rapidly as under conditions which did not favor the formation of these devices. Verbal reports were used to confirm the speculation that these differences in rate of learning were the result of differences in the extent to which Ss were able to form and utilize mnemonic devices. In 1961, Bruner (1961) developed a crude scheme for classifying associative aids. He found that the aids which children used in forming associations between familiar words almost always involved one of three general types of relation- ship--equivalence, thematic, or functional. Perhaps more important, Bruner also noted that those pairs which were 25 linked by the type of relationship which a given youngster used most frequently were learned more rapidly than pairs linked by one of the other two types of relationship. In a more recent study, Martin, Boersma, and Cox (1965a) developed an elaborate scheme for classifying associative aids used during associational tasks which involve nonsense syllables as stimulus and response terms. These authors found that each reported "associative strategy" could be reliably ordered along a dimension of cue complexity which ranged from no reported association to syntactical strategies. More im- portant, these authors also reported that the speed with which a given association is formed is directly related to the level of the associative strategy which the individual reports. In an extension of this research, Martin, Cox, and Boersma (1965b) demonstrated that the general level of strategy which is derived by a given S is a direct function of the meaning— fulness of the items which are presented. These authors also found that differences in meaningfulness of the stimulus terms have a greater effect on general strategy level than corres— ponding differences in response terms. All of the above investigators have been concerned with coding schemes which somehow unite a single pair of terms, i.e., first-order coding schemes. However, Battig (1966) ob- tained some evidence in the verbal reports of Ss which indi— cated that they frequently formed groups consisting of two or more pairs, instead of learning each pair independently. 26 He therefore undertook a series of experiments which demon— strated that under experimental conditions which facilitated the formation of "higher-order multiple pair chunks" on the basis of degree of learning, associations were formed more rapidly than under conditions which did not facilitate such grouping. This research suggests that the encoding process does not always terminate with the derivation of a set of first-order coding schemes, as the other investigations in this area have assumed. The third approach--presentation of some form of encod- ing scheme as a part of the general instructions--has also yielded a set of conclusive findings. Studies involving this approach present one group of $5 with a set of coding schemes prior to the initial presentation of the list; while a second, or control group is not presented with these aids. The encoding schemes which are most frequently presented are sentences which unite the two words or two nonsense syl— lables in each pair (e.g., Someone was playing a DRUM in a TENT). Sentences of this type have been shown to aid in the formation of associations between pictures of familiar objects which are presented to mentally retarded adults (Jensen and Rohwer, 1965a), as well as in the formation of associations between nonsense syllables presented to fourth through sixth grade children (Cox, 1965). Using a slight variation of this technique, Spiker (1960) has shown that the presentation of sentences during a practice 27 list aids fifth grade youngsters in forming associations on a second, or critical list. In a comparable study, Bruner (1961) contrasted the effectiveness of coding schemes derived by individual 88 with those presented by the experimenter. In this study, one group of twelve year old Ss was told to produce a word, or an idea which would tie each pair of words together. A second group was then presented with the medi- ators derived by Ss in the first group. The performance scores of Ss in the self-mediational group were clearly superior to that of Ss in the second group. These four studies, therefore, provide convincing evidence that imposed coding schemes facilitate the formation of associations. But, contrary to what one might expect, these aids do not appear to reduce the differences in speed of learning which occur among children when no aids are given (Davidson, 1964; and Cox, 1965). And, the effect of these aids on long- term retention is also somewhat questionable. Martin, Cox, and Bulgarella (1966), for example, found that the two-day retention of a group of children who were given a set of first-order coding schemes during acquisition was superior to that of a comparable group who were not given these aids. Jensen and Rohwer (1965a), on the other hand, found no dif- ferences in the seven day retention of a group of mentally re- tarded adults who were given a set of coding schemes during acquisition and a comparable group who were not given these aids. 28 But, as indicated earlier, there can be little doubt that imposed coding schemes do facilitate the formation of associations. The critical function which these aids serve in this regard, appears to be that of somehow uniting the two terms in each pair. Epstein, Rock, and Zuckerman (1960), for example, devised three methods of presenting the same six pairs of concrete nouns. In list one, the two nouns were connected in meaning- ful units by the addition of a single term (e.g., CAKE near ROAD). In list two, a grammatically equivalent term was interjected between the two nouns, but the addition of this term did not form a unit (e.g., CAKE and ROAD). Finally, a third list was devised which did not involve any extraneous terms (e.g., CAKE - ROAD). Following a single presentation of the list, the mean number of words correctly anticipated was computed. Scores on list one (meaningful units) were clearly superior to scores on the other two lists. In a comparable study with children, Davidson (1964) found that supplying a prepositional link between each pair (e.g., shoe on chair) was as effective as a picture showing this condi- tion, accompanied by a sentence describing the picture (e.g., large shoe resting on the arms of a chair). These two studies, therefore, clearly demonstrate that the effectiveness of various encoding schemes which are utilized during paired- associate learning Stems from the fact that these schemes somehow unite the two terms in a given pair. 29 The final procedure which has been used to determine the role of encoding schemes in associational learning has been to present lists which have some form of "built-in" coding scheme. In perhaps the only study which has utilized this technique, Underwood and Erlebacher (1965) presented lists of stimulus and response trigrams. These lists were constructed in such a way that the letters in each trigram could be rearranged to form familiar words. They found that most 83 formed words during encoding. But, the use of this coding scheme facilitated performance only under those con- ditions in which the letters in each trigram had been re— arranged according to a single rule. When 83 were required to use more than one decoding rule, use of the "built-in" coding schemes had little or no effect on performance. Unfortunately, however, the imposed coding scheme in this study was aimed at facilitating either response learning or the formation of discriminations among highly similar stimuli. It did not serve the function of somehow uniting the two terms in each pair. The combined results of studies reviewed in this section leave little doubt that the derivation of encoding schemes facilitates the formation of verbal associations. There is even evidence to suggest that some "strategies," or encoding schemes, which are used to link nonsense syllable pairs may be more effective than others (Martin §£_§l,, 1965a). Research in this area has also identified the critical function which 50 these coding schemes serve; namely, joining pairs of terms into some meaningful unit (Epstein, Rock, and Zuckerman, 1960; and Davidson, 1964). In short, a general appraisal of the research in this area tempts one to conclude that the major issues have been largely exhausted. But a closer review reveals at least two major short- comings of the existing research. First, the full range of procedures which might be used to study the encoding process during associational learning is far from exhausted. Underwood and Erlebacher (1965), for example, are perhaps the only authors to use the technique of deriving lists with some form of "built-in" structure. Yet, as noted in Chapter I, the use of this technique has proved its value in studies which involve the free recall of serial lists. Second, with the exception of Battig (1966), no one has traced the encoding process beyond the formation of relations between each pair of terms, i.e., first-order coding schemes. Yet, as Battig (1966) notes, general descriptions of the encoding process seem to suggest that Ss will attempt to form fewer units of retention than there are pairs on the list. Thus, studies which examine only intra-pair associational aids do not seem totally consistent with general descriptions of the encoding process. 51 III. General Evidence Related to the Stated Problem of this Investigation A. The Effect of Higher-Order Coding Schemes on Short-Term Retention AS noted above, Battig's (1966) research represents the only investigation of associational learning which has traced the encoding process beyond the derivation of a set of first— order coding schemes. Thus, Battig is the only investigator to provide direct evidence for the assertion that the deri- vation of higher-order coding schemes facilitates the forma- tion of verbal associations. And even this evidence may be questioned on several grounds. First, in terms of metho- dology, Battig has failed to control certain extraneous con- ditions such as the presence or absence of serial position cues which may have also contributed to his obtained results. Second, Battig‘s assertion that 85 group pairs on the basis of item difficulty or degree of learning is reasonable only if one assumes that the experimental conditions have effective- 1y prevented a S from adopting a more efficient strategy. Bruner's study (1961), for example, suggests that Ss are more likely to initially attempt to derive the same relationship among the terms in more than one pair. Bruner observed that children usually exhibit a strong preference in the type of relationship which they form between pairs of familiar terms. Furthermore, those pairs which are linked by the child's favorite type of relationship--equivalence, 52 functional, or thematic——tend to be learned more rapidly than pairs joined by one of the other two types of relationship. Thus, Bruner might have argued that Ss group pairs on the basis of the type of relationship which they have established between each. But Bruner's failure to provide any statistical evidence for his assertions severely limits the strength of any such argument. In short, investigations of associational learning have provided only equivocal evidence for the general hypothesis that the derivation of higher-order coding schemes will facilitate the formation of verbal associations. However, indirect evidence is provided by the results of investigations involving other types of learning. Studies involving the free recall of serial lists, for example, have shown that recall is improved when 85 transform letter trigrams into common words (Underwood and Keppel, 1965; and Lindley, 1965); when Ss alphabetically organize lists of unrelated nouns (Tulving, 1962; and Earhard, 1967); when 83 group words according to common categories (Reid, Brackett, and Johnson, 1965); and when 55 link words through the use of well-established grammatical rules (Miller and Selfridge, 1950; Marks and Jacks, 1952; Richardson and Voss, 1960; Tulving and Patkau, 1962; and Coleman, 1962). Collectively, these studies suggest that a 8'5 active attempts to reduce the number of units of retention ultimately facili— tate short—term retention. It is therefore at least logically 55 consistent to assert that similar attempts in associational learning will also facilitate acquisition. B. The Effect of Higher-Order Coding Schemes on Long-Term Retention Although available evidence leaves little doubt that the use of first-order coding schemes facilitates the formation of verbal associations, one can only speculate as to their effect on long-term retention. Martin pp _l., (1966) and Jensen and Rohwer (1965a) appear to be the only authors to ever examine the effect of coding schemes on the long-term retention of verbal associations. And, these two studies were not only limited to a concern with first-order coding schemes, but they also yielded contradictory results. Thus, studies of associational learning do not even provide indirect evidence for the general hypothesis that the derivation of higher-order coding schemes will aid in the long-term reten- tion of verbal associations. As noted earlier, however, the derivation of higher-order coding in this investigation might be said to correspond to the discovery of a principle. Therefore, studies which have been concerned with the retention of principles may also be relevant to this hypothesis. Unfortunately, however, most studies which have focused on the formation and retention of principles have been concerned with problem-solving or non- reproductive forms of learning rather than with the use of principles as aids in storing and retaining information. 54 i There is a sizable body of research, for example, which is concerned with the basic question of whether or not some form of aid in deriving the relevant principles will have any impact on an individual's ability to solve a given set of 7 When retention is determined in these studies, problems. it is usually expressed in terms of an individual's ability to apply the relevant principles to a new set of problems rather than in terms of retention of information which has been organized around some principle. Occasionally, however, these investigations have also examined the effect of various forms of training on the long- term retention of information; namely, retention of the solu- tions to a given set of problems. Results of studies in this area have consistently shown, for example, that when 7The results of studies in this area are loaded with contradictions. Several authors, for example, maintain that their results favor the conclusion that some form of guidance or aid in deriving the relevant principles is superior to training involving self-discovery of these principles. Guidance has been shown to aid in both the derivation of solutions to a set of problems (Ewert and Lambert, 1952; Duncker, 1945; and French, 1954) and in an individual's abil- ity to apply these principles to new problem situations (Judd, 1908; Waters, 1928; Katona, 1940; Hendrickson and Schroeder, 1941; Craig, 1955; Kittell, 1957; Corman, 1957; Gagné and Brown, 1961; and Ray, 1961). However, other authors have been forced to conclude that different levels of guidance do not yield differences in an individual's ability to solve a given set of problems (Olander, 1951; Stacey, 1949; Marks, 1951; and Hilgard, Edgren, and Irvin, 1959) nor in his ability to apply these principles to new problem situations (Thiele, 1958; Hendrix, 1947; Craig, 1956; and Haslerud and Meyers, 1958). Some of these authors have even maintained that self-discovery is superior to some form of guidance. 55 individuals are given some aid in discovering the principles which underlie the solution to a given set of problems ("guided discovery") they will be able to retain these solu— tions longer than individuals who are either given a direct statement of the relevant principles, together with the solutions (Kittell, 1957; and Ray, 1961) or those individuals who merely memorize the solutions with no understanding of the underlying principles (a series of studies by Katona, 1940; Swenson, 1945; and Kittell, 1957). Katona (1940), for‘ example, found that a group of 83 who were given somektraining in deriving the relevant principles retained more solutions to a set of matchstick problems than a group of 85 who originally memorized the solutions to these problems with no understanding of the relevant principle.8 There is also some evidence to suggest that individuals who are given direct statements of the underlying principles will retain solutions to a set of problems longer than indi- viduals who are merely instructed that there is a principle which underlies these solutions (Postman, 1954; and Craig, 1956). But Kersh (1958, 1962) maintains that his results support the opposite conclusion; namely, that self-discovery is superior to a statement of the relevant principles. 8It is interesting to note that these differences in long-term retention have occurred despite the fact that 85 are often unable to verbalize the relevant principles. Katona (1940), for example, reported that many 85 who had solved most or all of the retention tasks were nevertheless unable to state the principle. Kersh (1958) also found that only about 50 per cent of the Ss could remember the relevant rules four weeks after acquisition. 56 Collectively, these studies suggest that 85 who dis- cover the relevant principle during acquisition should retain the associations longer than 85 who initially memorize these associations, i.e., form no coding schemes whatsoever. In fact, these studies even suggest that these differences should increase over time (Katona, 1940; Postman, 1954; Kittell, 1957; and Ray, 1961). But, this research has little to say regarding the contrast in long-term retention between 55 who have derived a higher-order coding scheme and Ss who have derived a set of first-order coding schemes. Thus, these studies provide only partial support for the hypothesis that the derivation of high-order coding schemes will aid in the long-term retention of verbal associations. C. The Role of Perception in the Formation of Higher-Order Coding Schemes To the author's knowledge, no one has examined the effect of perceptual variables on the discovery of relations among pairs of stimulus and response terms. Consequently, there is no direct evidence to support the general hypothesis that a 8'8 overall perception of the stimulus and response terms will affect the likelihood that he will derive a higher- order coding scheme. The source of this hypothesis must, therefore, be traced to a group of studies which have shown that perceptual vari- ables affect the probability that Ss will discover implicit relations among a series of numbers (Katona, 1940; Duncker, 57 1945; Kersh, 1958; and Wertheimer, 1959). Consider the following example from Katona (1940). Suppose that three individuals are presented with the number series which was described in Chapter I and are asked to recite each digit in the order in which it is presented. Further imagine that this series is displayed to each individual in a different manner, such as . . . 2, 5, 8, 1, 1, 1, 4, 1, 7, 2, O 258, 111, 417; 20 2. 5; 8, 11, 14, 17, 20 individual no. 1 individual no. 2 individual no. 5 Because people tend to group on the basis of contiguity, it seems reasonable to predict that individual three is more likely to perceive the implicit relation among the numbers than either individual one or individual two. In other words, individual three is most apt to reduce the entire series to a single unit of retention, i.e., a statement of this rela- tionship. Derivation of a higher-order coding scheme in this in- vestigation involves an analogous discovery of an implicit categorical relation among the terms in each triad. It therefore seems reasonable to predict that any method of presenting each triad which somehow makes this categorical relation more obvious should ultimately affect the probability that a given S will derive a higher-Order coding scheme. These logical assertions give rise to a second basic assumption of this dissertation; namely, the categorical terms in each triad are more suggestive of the "built-in" principle than either of the exemplar terms. Based on this 58 assumption, two methods of drawing attention to the cate— gorical terms will be utilized in an attempt to determine whether or not perceptual variables have any effect on the likelihood that a given S will derive a higherrorder coding scheme. The first variable involves differences in the size of print—-the categorical terms will either be capitalized with exemplar terms written in small print, or all three terms will be written in small print. A number of studies support the assertion that isolating the categorical terms in this fashion should result in differential attention to these terms (Kohler and Von Restorf, as described in Katona, 1940; Siegel, 1945; Pillsbury and Raush, 1945; and Kimble and Dufort, 1955). The second variable involves differences in the order of presenting the three terms—-the categorical terms will either be presented in the first position in all nine triads, or their position will vary from one triad to the next. This variable is based on evidence from a series of studies which suggests that differential attention is usually given to the first term in a series (Cofer, 1965; Howes and Osgood, 1954; Asche, 1946; Astin and Ross, 1958; and Suppes, 1965). This study will also consider a third perceptual vari— able which does not involve the assumption that the cate- gorical terms are more suggestive of the principle than the exemplar terms. Either one or two terms in each triad will 59 be written as homonyms. Those 85 presented two-homonym lists must transform two terms in each triad before the underlying categorical relationship is apparent, while 33 presented one-homonym lists need only transform one word in each triad. Thus, the "degree of embeddedness" of the underlying relation— ship should vary for one-homonym and two-homonym lists. By determining what per cent of the Ss discover the "built—in" principle for each condition on these three vari- ables, it should be possible to determine whether or not a (33' overall perception of the three terms (configural per- ception) has any effect on the probability that he will derive a higher-order coding scheme. CHAPTER I I I METHOD AND PROCEDURE I. General Overview of the Experimental Design This investigation consisted of two independent experi— ments. In Experiment I, eight groups of 85 were presented with a list consisting of 27 familiar terms arranged as nine triads. During the learning trials, each triad was projected on a screen for four seconds and Ss attempted to form associ— Iations between all three terms. During the test trials, one term from each triad was presented, using a five second ex- posure, and Ss attempted to write the other two terms. Each experimental session consisted of eight learning trials and eight test trials. The interval between a given test trial and the following learning trial was fixed at two minutes. During this period, two trained graduate students scored the answer sheets. Those Ss who reached criterion—- two consecutive perfect trials--were identified by the scorers during this interval. As each S was identified, he quickly went into the adjoining hall where he completed the Level of Principle Formation Questionnaire. Those 85 who did not reach criterion within eight trials completed the question— naire immediately after the eighth test trial. 40 41 Each of the eight groups in Experiment I received a dif- ferent list of triads. Each list represented a unique combi— nation of the following three treatment variables: (1) type of underlying principle-—one vs. two homonyms (2) size of print--all terms in small print vs. cate- gorical term capitalized and other two terms in small print (5) order of presentation within the triad—-categorical term always in the first position vs. position of the categorical term varied from first through third pOSltlon. Thus, Experiment I utilized a 2x2x2 factorial design and in- volved a group testing procedure. The independent or treat- ment variables are presented above. The four dependent variables are: level of principle formation, total number of words correctly recalled, number of trials to criterion, and number of intrusions in recall. Exactly three weeks after the initial learning session, the Ss in Experiment I were presented with a single retention test trial. During this trial, one term from each triad was presented, and Ss attempted to write the other two terms. The independent variables on the test for retention were the three treatment variables listed above, as well as the level of principle formation utilized during initial acquisition. The dependent variables were number of words correctly re- called and number of intrusions. With one major exception, the design of Experiment II was identical to that of Experiment I. This exception was that the eight groups in Experiment II were given only one 42 learning and one test trial. Following the test trial, all 85 completed the Level of Principle Formation Questionnaire. If perceptual variables do affect relations between terms which 85 perceive and subsequently utilize during en- coding, then it is likely that this effect will be most prominent during the first learning trial. Beyond that point, the effect of perception is probably confounded by other vari— ables. Thus, Experiment II was conducted primarily as a test of the hypothesis that perceptual attributes of the stimulus materials affect the perceived relations between terms. Secondarily, Experiment II provided clarification of the general results of Experiment I. II. Experiment I Subjects Two-hundred and eighteen undergraduate students--67 males and 151 females—-served as subjects in this experiment. The majority of these students were college sophomores. Each was enrolled in one of eight sections of educational psychology at Michigan State University. During acquisition, one of the eight lists of triads was randomly selected for presentation to a given section. The sample size was considerably smaller on the test of long-term retention. This reduction resulted from two factors. First, only 157 of the 218 Ss reached criterion within the 45 eight acquisition trials. Second, twenty-one of these 157 85 were absent on the day retention was determined. Thus the sample was reduced from 218 85 during acquisition to 156 $5 on the test for long-term retention. Materials Nine triads were constructed according to a predetermined relation; namely, when two of the terms in each triad are transformed into their homonyms, they become exemplars of the third or categorical term. These triads formed the basis of two distinct lists. The first list, which shall hereafter be referred to as the two-homonym list, consisted of the nine triads in an unmodified form. The second list, hereafter referred to as the one-homonym list, was constructed by chang— ing one of the terms in each triad into its homonym. Thus, the one-homonym lists have an underlying principle which is different from that for the two-homonym lists; namely, when pp§_of the terms in each triad is transformed into its homonym, two terms are exemplars of the third or categorical term. Both of these lists are presented in Table 1 on the following page. The frequency of occurrence of each term, as determined by the Thorndike-Lorge word count (Thorndike and Lorge, 1944) is also depicted in Table 1. Due to the nature of the under- lying principles, it was impossible to equate the one and two- homonym lists on this variable. Therefore, the unique terms 44 Table 1. List of triads which were presented during the study. Numbers in parentheses indicate the number of times per million that the words occur in printed text. (Thorndike and Lorge, 1944.) Common Terms Uniqpe Terms misspelled two-homonym one-homonym categorical exemplar lists (or) lists mammal (6) dear (AA) bare (A) bear (AA) metal (A) steal (A) led (AA) lead (AA) receptor (*) knows (AA) ayes (16) eyes (AA) study (AA) reed (22) rite (10) write (AA) weather (AA) son (AA) rein (25) rain (AA) food (AA) meet (AA) foul (27) fowl (20) appendage (5) tale (A) feat (15) feet (AA) beverage (8) bier (4) whine (11) wine (A) group (AA) teem (6) heard (AA) herd (57) (*) not listed (A) 50-100 times/million words (AA) 100 or more times/million words on one—homonym lists occur somewhat more frequently than the corresponding terms on two-homonym lists. Type of underlying principle was only one of the stimulus variables examined in this study. Variations in the method of presenting the stimuli gave rise to two additional variables. The first of these variations involved differences in the size in which the terms were printed--either the categorical term was written in capital letters with exemplar terms in small letters (caps condition) or all three terms in each triad were printed in small letters (no-caps condition). The second variation involved differences in the order of presentation of the three terms within each triad--the 45 categorical term was either fixed in the first position on all nine triads (fixed condition), or its position varied such that it appeared in the first position on triads 1, 4, and 7; in the second position on triads 2, 5, and 8; and in the third position on triads 5, 6, and 9 (random condition). In short, all possible combinations of conditions along the three variables led to the development of eight distinct lists of triads. The final phase of the construction of materials involved the development of stimulus lists for presentation during the test trials. The selection of one term from each triad gave rise to three different stimulus lists for both the one- and two-homonym conditions. By assigning each term in a triad to a different stimulus list, a given term appeared as a stimu- lus on every third test trial. Due to potential differences in the recall of categorical terms as opposed to exemplar terms or in the recall of words in different positions within the triad, each stimulus list was counterbalanced for type of word and position within the triad. Thus, each stimulus list consisted of three terms of each type (categorical, exemplar, and misspelled exemplar) as well as three terms which appeared in each position within the triad (first, second, and third). Once this balance was insured, the order of presentation of the nine terms in each of the three stimulus lists was determined by a table of random numbers. In this fashion, order of presentation was 46 altered from test trial to test trial, thereby counteracting the potential effects of serial position cues. Examination of Table 17 of Appendix C, which depicts the mean number of words recalled for every triad over the first three test trials, reveals that this procedure did eliminate systematic serial effects. Construction of the three stimulus lists for the one- homonym condition preceded the construction of the three lists for the two-homonym condition. The latter lists were identi- cal to the one-homonym lists except that the unique terms were replaced by their corresponding terms on the two-homonym lists. (Terms in column four of Table 1 were replaced by the corres— ponding terms in the third column.) The eight lists of triads as well as the six stimulus lists were transferred to slides for display during the learn- ing and test trials. Each list of triads consisted of nine slides with three terms on each, while the stimulus lists consisted of nine slides with a single term on each. A carousel slide projector was used for the display of the fourteen lists. Procedure Each experimental session began with a general set of instructions which described the nature of the task (see Section I of Appendix A). The nine slides in the list of triads were then projected on a screen in the front of the 47 room. A four-second exposure time was used for each slide during this and later learning trials. The first learning trial was followed by a second set of instructions (see Section II in Appendix A). This set of instructions served two functions. First, it described the appropriate way to complete the answer sheets. Second, an attempt was made to maximize the number of intrusions in recall through presentation of the following paragraph: Other studies have shown that those people who write the most responses in situations like this usually learn the list fastest. So even if you aren't sure of the answers, write whatever comes into your mind as I show each slide. The nine slides in the first stimulus list were then pro- jected. A five-second exposure time was used for each slide during this and later test trials. The first test trial was followed by a third set of instructions (see Section III in Appendix A). These directions stressed that 85 were to re- main quiet during the intertrial interval. They also restated that the scorers would identify those 83 who reached criterion on a given test trial. When the two-minute interval elapsed, the second learning trial was presented, followed by the second test trial. The complete experimental session consisted of eight learning and eight test trials. During every intertrial interval, two trained graduate students collected and scored the answer sheets. Responses of those Ss who received a perfect score on the preceding trial 48 were scored first. In this manner, it was possible to rapidly identify any S who reached criterion.9 As each of these in- dividuals was identified, he quickly left the room and went into the adjoining hall where he completed the Level of Principle Formation Questionnaire. The time needed for identi- fying 55 who reached criterion occasionally extended beyond two minutes, but never exceeded two and one—half minutes.10 A test of long-term retention was administered three weeks after the initial acquisition session. 83 had not been in- formed that they would be tested for long-term retention. Each experimental session was introduced by a set of general instructions (see Section IV in Appendix A). The instructions were followed by a single test trial. There were two differences between the test of long-term retention and earlier tests of short-term retention. First, the list of stimulus terms consisted of all nine categorical terms, rather than three categorical terms and six exemplar terms. This change was introduced to maximize the likelihood of making errors in spelling the misspelled exemplar terms. Second, each slide was projected for eight seconds, instead of five seconds. This change was introduced to give each S more time to think through his responses. 9Minor deviations in the spelling of the categorical terms were the only errors which were tolerated in this scoring. loMost apt to be true for intervals following the third, fourth, and fifth test trials. 49 III. Experiment II Subjects A total of 244 Ss—-94 males and 150 females—-participated in Experiment II. These 55 were enrolled in one of six sec- tions of educational psychology or four sections of education— al philosophy at Michigan State University. Four of the sec- tions of educational psychology were considerably smaller than the others. These four groups were therefore arbitrarily treated as two experimental groups. These two condensed groups, as well as the other six sections with typical enroll- ment, were then randomly assigned to one of the eight list conditions. Materials The eight lists of triads were identical to those employ- ed in Experiment I. Procedure With the exception of two changes in directions, the pro— cedure adopted for Experiment II was identical to that utilized for the first learning and first test trials of Experiment I. As noted earlier, the list of triads was presented for a single learning trial at a 4-second rate, followed by a single test trial using a 5-second exposure. All 88 completed the Level of Principle Formation Questionnaire immediately after the test trial. 50 The first change in directions involved the general in- structions which were given prior to each experimental session. Although the general content of these instructions was identical for both experiments, some of the statements which were included between the first learning and first test trial of Experiment I were made prior to the first learning trial of Experiment II. (These differences are depicted in Sections I and II of Appendix A.) The purpose of this change was to reduce the length of the intertrial interval as well as the extent of interference caused by reading directions at this point. In short, an attempt was made to reduce the extent of influence which either of these factors may have had on the relationship between the levels of principle form- ation derived during the first learning trial and performance on the first test trial. Whether or not this change in directions resulted in a loss of generality between the results of Experiment I and Experiment II is contingent upon the dependent variable which is being considered. It is likely, for example, that the number of words recalled on the first test trial will be some- what higher for Experiment II than for Experiment I. But this change should have no effect on the per cent of 85 who dis- cover the principle during the first learning trial of either experiment. In general, the minor difference in directions between Experiment I and Experiment II does not preclude attempts to formulate generalizations based on the combined results of these two experiments. 51 The second change in directions involved the instructions which were given to each S as he attempted to complete the Level of Principle Formation Questionnaire (L.P.F.Q.). Whereas Ss in Experiment I received only those directions which are printed on the cover of this instrument, Ss in Experiment II were given a special set of oral instructions immediately after the questionnaires had been distributed (see Section V of Appendix A). These instructions emphasized that 85 would not answer questions on every page of the L.P.F.Q. Thus, this change in directions was introduced in an attempt to increase the reliability of the L.P.F.Q. IV. Level of Principle Formation Questionnaire A reexamination of the hypotheses in Chapter I clearly suggests that the level of principle formation variable occu- pies a central role in this investigation. In nearly every hypothesis this characteristic functions as either an inde- pendent or a dependent variable. The instrument designed to determine values along this variable is referred to as the Level of Principle Formation Questionnaire (L.P.F.Q.). Presented in Appendix B, the L.P.F.Q. is designed some- what like a branched program in that the particular manner in which a S responds to a given question determines what ques- tion he shall answer next. By examining a given S's pattern of "yes" and "no" responses, it is possible to operationally assign him to one of four levels of principle formation. 52 This operational classification scheme is presented in Table 2 below: Table 2. The operational criteria used to assign individuals to one of four levels of principle formation accord— ing to their responses to the Level of Principle Formation Questionnaire. Operation underlying assign- Level Descriptions ment to this level I Neither looks for, Answers "no" to first two nor formulates any questions (page 1). first-order coding schemes. II Formulates at least Answers "yes" to one question some first—order on page 1, but "no" to the coding schemes. question on page 2. III Discovers some common Answers "yes" to questions on first-order coding page 1 and 2, but "no" to schemes, but does not question on page 5. Provides discover the princi- evidence for assertion on ple. page 4. IV Discovers the under- Answers "yes" to all relevant lying principle. questions on first three pages, and correctly performs at least one of the two tasks on pages 5 and 6. (Incorrect completion of both tasks will result in assign— ment to Level III) Numerous potential assets of the questionnaire, the operational scheme for classifying 53, such as led to the decision to adopt the L.P.F.Q. during this investigation despite the fact that it had never been tested with a group of undergraduate 55 students. But, in practice, the questionnaire proved to have two inherent weaknesses. First, despite the inclusion of an example to clarify the question, the query on page two showed definite signs of ambiguity. Several Ss who answered this question "no" (thereby indicating that they had not dis— covered the principle), nevertheless, produced nine verbal reports on page four which all conformed to statements of the principle. Second, the general directions on the first page did not sufficiently emphasize that 55 would not answer questions on every page. As a result, the specific directions at the end of each question (which revealed the next question to be answered) were often ignored. The fact that many 85 in Experiment I were not allotted ample time to complete the questionnaire may have also contributed to this general fail— ure to follow directions. Whatever the source, an alarming number of errors occurred. Ninety of the 218 Ss in Experi— ment I, or roughly forty-one per cent, made one or more errors in following directions. Fortunately, the number of errors made by Ss in Experi- ment II was substantially reduced by two changes in procedure. First, 55 in this experiment were given unlimited time in which to complete the form. Second, these Ss were also given Special verbal instructions which emphasized that they would 11 not answer questions on every page. Despite these changes, 11These directions are presented in Section V of Appendix A. 54 the performance of Ss in Experiment II was far from perfect. Twenty-seven of the 244 Ss in this group, or roughly twelve per cent, made one or more errors in completing the question- naire. The obvious bias which would result from the deletion of 117 Ss from the sample forced the investigator to make two basic changes in the classification scheme which is pre- sented in Table 2 above. First, only two levels of principle formation were determined instead of four. Second, verbal reports were used as a supplementary basis for identifying Ss who had discovered the principle. These two changes circumvented the problems arising from the two inherent weak— nesses of the L.P.F.Q. while in no way altering the major purpose of this study. The first major change was to collapse levels I, II, and III into a single category; namely, those who did not discover the principle. This decision was based upon the apparent failure of the L.P.F.Q. to make any useful discriminations between $3 at the lower levels of principle formation. In brief, no 8 could be classified at level I under any set of criteria, and Ss classified at levels II and III according to a somewhat modified scheme (identification sometimes based on verbal reports) did not differ in performance on any of the following five dependent variables: total number of spelling errors; number of trials to criterion; number of words correct on the test for long-term retention; and, ”nu-IV “but" 0? on». Miss . 'vrfi ,1 «J . v . c C t. . A v r a «C v l l : a.» in l . l u). A s n .4 Au A HI. .1 A A n. r? be in x r. L. Q D . n. . n A Pv rfi V l r u e; . rs. hv . Q Mel» ; Y “ ‘ VI A n 0 h V t It I s y . . a? “UL s V L a.» . at v A: Q» s C , Cy ¢ Q» .n, u LL Ru Av .nu 55 number of words correct on the first test trial of Experiment II. The series of one-way analysis of variance tests upon which this statement is based are summarized in Table 18 of Appendix C. In View of this finding, it would seem reason— able to assume that levels I, II, and III would have been collapsed into a single category even if every 8 in the sample had followed directions.12 The second, and perhaps more basic change, was to classi— fy 83 into the two categories on the basis of verbal reports on page four, or on the basis of the performance measures on pages five and six, depending on which they completed first. The pattern of "yes" and no" responses which were basic to the original classification system were to a large extent ig- nored in the final system. This change, which circumvented problems arising from the frequent failure to follow specific directions is depicted in Table 5 on the following page. Those Ss who met any of the four criteria which are listed in this table were included in the group which discovered the principle. All others were classified as members of the group which did not discover the principle. Two analyses were undertaken in an attempt to provide empirical support for the final classification scheme. 12The apparent failure of the L.P.F.Q. to discriminate between Ss at the lower levels eliminated any opportunity to test a model of principle formation which this author had derived as a supplemental phase of the investigation. In View of these non-test conditions, reference to this model was deemed extraneous to the major purpose of the investigation, and has therefore been deleted from this report. 56 Table 5. The four criteria ultimately adopted to identify Ss who discovered the principle during acquisition. Criteria Improvement Description I circumvents the Answers "no" to the question problem arising on either page 2 or 5, yet for from the ambiguity all nine verbal reports on of the question on page 4 the S: (1) stated that page 2. two of the words were exemp- lars of the third or categor- ical term, and (2) indicated the appropriate changes for the misspelled exemplar term(s). II circumvents many Same as Criterion I, except of the problems that S answered questions on surrounding the first three pages "yes", but failure to follow then completed the verbal re- specific direct- ports on page 4 prior to the tions at the end performance tasks on pages of each question. 5 and 6. . III None Answers questions on first three pages "yes" and turns directly to page 5 where com— pletes pp§_of the two tasks correctly. IV None Same as Criterion IV, but completes both performance tasks correctly. First, it was necessary to show that the first two criteria listed in Table 5 above which were added to the list were as effective in identifying 83 who discovered the principle as the third and fourth criteria which were carried over from Second, because it was the original classification system. no longer possible to classify every S on the basis of 57 operational criteria, the reliability of the final scheme was determined. A careful appraisal of Table 5 above reveals that basic differences exist among the four criteria, perhaps the most important of which is that Criteria I and II involve an analysis of verbal reports, while Criteria III and IV are based on the successful completion of at least one perform- ance task. In view of these differences, it seemed desirable to demonstrate that the performance of 85 identified as dis— coverers of the principle would not vary significantly across the four criteria. Only if this equality of performance prevails, for example, is it safe to assume that verbal re- ports are as valid as performance tasks in identifying 83 who have discovered the principle. In an attempt to confirm this assertion, the author hypothesized that the means for all four groups of 85 (corresponding to the four criteria) would be equal on each of the following dependent variables: number of words correctly recalled over all eight test trials; total number of misspelled exemplar terms; number of trials to criterion; and, number of words correctly recalled on the test of long-term retention. The obtained sample means and sample variances for the four groups of 55 identified as discoverers of the principle by each of the four criteria are depicted in Table 19 of Appendix C.13 Examination of this table reveals that the 13The one and two homonym lists have been analyzed inde- pendently in View of the differences in both terminology and 58 four means are extremely close on each of the four dependent variables. This apparent equality prevails on both one- and two-homonym lists. Nevertheless, a one-way, fixed-effects, analysis of variance test was computed for each measure, the results of which are also depicted in Table 19. Available tables in Guenther (1964) reveal that none of these differ— ences is significant when alpha is equal to 0.10. In fact, only one difference is significant when alpha is equal to the extremely low value of 0.25.14 In view of this apparent lack of difference in performance among the four groups, it seems reasonable to conclude that the four criteria are equally effective in identifying 83 who have discovered the principle. The low values of alpha upon which these decisions to accept the null hypothesis are based lends further support to this conclusion. Under these conditions, the probability of making a type II error is quite small. As a second phase in the evaluation of the final classi- fication scheme, interjudge reliability was determined. Two different individuals were involved in this analysis. The first classified 85 in Experiment I, while the second classified underlying principles on these two lists. Because of these differences, the distinction between the two lists shall be maintained throughout the dissertation. 14When this test was followed by the Scheffé’test of individual comparisons, the results suggested that the mean number of words recalled by Ss classified on the basis of verbal reports (criterion I) surpassed the corresponding mean for individuals classified on the basis of performance tasks (criterion IV). This result occurred on the test of long-term retention of one-homonym lists (0.10 < p < 0.25). 59 Ss in Experiment II. Training for these two independent judges consisted of general instructions, followed by practice on one of the eight lists selected at random. Disagreements in classification on the practice list were carefully analyzed by the principal investigator and the independent judge. Once this training was complete, the two judges made independent classifications on the remaining seven lists in Experiment I and Experiment II respectively. In View of the classification of Ss into only two cate- gories, Phi coefficients were computed for both experiments. In order to compute this statistic, it was necessary to assign every individual in the sample to one of the two categories:15 This procedure yielded 15 disagreements among the independent judge and principal investigator for 190 Ss in Experiment I, and seven disagreements for 210 Ss in Experiment II. The corresponding Phi coefficients were 0.840 and 0.914.16 15During later analyses, however, the principal investi- gator treated 15 Ss in Experiment I and 15 Ss in Experiment II as "unclassifiable." The reasons for excluding these 53 from the final sample included: ambiguous verbal reports; failure to answer all relevant questions; and, gross inconsistencies between verbal reports and performance tasks. As will be shown in Chapter IV, the levels of performance of these indi- viduals consistently fell between Ss who either did or did not discover the principle. Therefore, exclusion of these 83 from the sample probably had little effect on the overall results. 16As a double check on the extent of inter-judge agreement, many of the analyses in Experiment I were repeated for classi- fications made by the independent judge (see Table 20 in Appen- dix C). The single instance in which the results of the inde- pendent judge differed from those of the principal investigator is noted in the text of Chapter IV. With this single exception, the results of both sets of data led to the same statistical decision. 60 In summary, the frequent failure of 85 to follow specific directions on the L.P.F.Q. forced this investigator to make two revisions in the classification scheme. First, Ss were assigned to only two groups (discoverers vs. non-discoverers of the principle) instead of four levels of principle formation. Second, verbal reports were used as supplementary sources in identifying 55 who had discovered the principle. The final scheme, involving four distinct criteria, was then evaluated along two dimensions. First, statistical tests with a high level of power revealed that $5 classified as discoverers of the principle, on the basis of verbal reports, exhibited the same mean level of performance across five dependent variables as 85 classified on the basis of performance tasks (criteria used:h1the original scheme). Second, the inter-judge reli- ability was shown to be sufficient for the purposes of this study. Phi coefficients of 0.840 and 0.914 were obtained for classifications made by the principal investigator and an independent judge in Experiment I and Experiment II, respec- tively. In short, the final classification system not only circumvented problems occurring when Ss failed to follow di- rections, but it also proved to be adequate for the purposes of this study. Finally, it should be noted that these changes in pro- cedure resulted in changes in the statement of certain hypotheses. In Chapter IV, those hypotheses which were orig- inally phrased in terms of differences among individuals at 61 the four levels of principle formation will be rephrased as a simpler contrast between 85 who either did or did not dis- cover the principle. CHAPTER IV RESULTS I. Definition of the Dependent Variables The following dependent variables were examined at some point during either Experiment I or Experiment II: (1) level of principle formation--redefined as per cent of 85 who discovered the principle. Section IV of Chapter III describes how this measure was determined. (2) number of words correct on a given test trial--Each response on the answer sheet was scored as either "one" or "zero", depending on whether or not it occurred in the triad from which the stimulus word was taken. Spelling of the two exemplar terms had to be exact, whereas phon- etically equivalent spellings of the categori- cal term were scored as "one“. The total number of words correct on a given test trial ranged from 0-18. (5) total number of words correct over all eight test trials--This measure was a simple sum of the number of words correct on test trials 1-8. Those Ss who reached criterion prior to the eighth test trial were scored as 18 for every trial beyond the point where they reached criterion.17 l7The scorers in Experiment I made twelve errors. Two 85 had three consecutive errorless trials before being told that they reached criterion. Ten others were informed that they reached criterion at the end of their first errorless test trial. It was therefore necessary to interpolate scores for the latter ten 85 along two variables. First, each S was given credit for 18 correct responses on each of the remaining test trials. This interpolation is probably a reasonable estimate of the total number of words recalled since, without exception, 62 65 (4) trials to criterion--Criterion was defined as two (5) (6) (7) consecutive perfect trials (scores of 18). Because it was possible to complete only 8 test trials during the class period, this measure was interpolated for several 85. Those 83 who received scores of less than 18 on the eighth test trial were scored as 10 trials to criteri- on. 83 who recalled all 18 terms on the eighth test trial, but not the seventh, were scored as 9 trials to criterion.18 spelling distortions--total number of responses which were written as homonyms of the correct exemp- lar terms. intra—listgintrusions-—total number of responses which were on the list, but not as a member of the triad from which the relevant stimulus word was taken. extra-list intrusions--total number of responses which were meaningful words but which were not members of any stimulus triad. (8) improper plurals--number of plural responses to terms (9) total which were singular on the list (e.g., "appen- dages" instead of "appendage") or singular responses to terms which were plural on the list (e.g., "eye" instead of "eyes"). Plurals in- volving forms other than simple "s" changes were treated as extra-list intrusions (e.g., "feet" instead of "foot"). number of intrusions--the sum of all four types of errors (numbers 5-8, above). the ten Ss had scores of at least 16 on the preceding test trial. Second, one test trial was added to the number of trials to criterion. This interpolation is probably subject to serious error since a single mistake on the following trial would have resulted in a much higher value. 18Ibid. 64 _II. Higher-Order Coding Schemes and Short-Term Retention Hypotheses one through four predict that the short-term retention of Ss who discover the principle will be superior to that of Ss who do not discover the principle. This pre- diction is expressed for the performance of 85 over all eight test trials of Experiment I (Hypotheses one and two) as well as the performance of $5 on the first test trial of Experiment II (Hypotheses three and four). Hypothesis 1:19 85 who discover the principle at some point during the learning task will form the re- quired associations more rapidly than those who do not. The corollary hypotheses stated in null form are as follows: Corollary 1a: The mean number of trials to criterion for 83 who discover the prinCiple will be greater than or equal to the corresponding mean for 85 who do not discover the principle. Corollary 1b: The mean total number of words correctly recalled over all eight test trials by 85 who discover the principle will be less than, or equal to, the corresponding mean for 85 who do not discover the principle. Hypothesis 1: With one exception, 85 who discover the principle at some point during the learning task will make fewer intrusions in recall than Ss who do not. The single exception is number of improper plurals, where the opposite prediction is made.20 The corollary hypotheses, stated in null form, are as follows: 19In accord with considerations advanced in section IV of Chapter III, Hypothesis 1 and all subsequent hypotheses which were phrased in terms of "level of principle formation" in Chapter I have been modified to read as a simple contrast be- tween Ss who either did or did not discover the relevant principle. 20This exception is based on a consideration of the gram— matical structure of given expressions of the principle. Many of these expressions are apt to include improper plurals; for 65 Corollary 2a: The mean number of spelling distortions made by Ss who discover the principle will be greater than, or equal to, the mean number of spelling distortions made by 85 who do not dis- cover the principle. Corollary 2b: The mean number of extra-list intrusions made by 83 who discover the principle will be greater than, or equal to, the mean number of extra—list intrusions made by 85 who do not dis- cover the principle. Corollary 2c: The mean number of intra-list intrusions made by Ss who discover the principle will be greater than, or equal to, the mean number of intra-list intrusions made by 85 who do not dis- cover the principle. Corollary 2d: The mean number of improper plurals formed by 85 who discover the principle will be less than, or equal to, the mean number of im- proper plurals formed by Ss who do not discover the principle. Corollary 2e: The mean total number of intru- sions made by Ss who discover the principle will be greater than, or equal to, the mean total number of intrusions made by 85 who do not discover the principle. Hypotheses one and two are concerned with measures of performance in Experiment I. Hypotheses three and four, on the other hand, are concerned with performance on the first test trial of Experiment II. Therefore, the dependent vari- ables in hypotheses one and two represent total scores over all eight test trials of Experiment I, while the corresponding variables in hypotheses three and four represent scores on the single test trial of Experiment II. A second distinction between these two sets of hypotheses is that $8 in Experiment I example, "the 'eyes' and 'knows' are both 'appendagegf." Discoverers of the principle, using such statements as medi- ators between the stimulus and overt response, might therefore be expected to form many improper plurals. Ss who did not dis- cover the principle, on the other hand, were expected to derive first order relations which somehow encompassed the three terms in their original form; for example, "An 'appendage' 'knows' what the 'eyes' see." 66 completed the L.P.F.Q. immediately after they reached cri- terion or after the eighth test trial (whichever came first), while 53 in Experiment II completed this form immediately after the first test trial. Therefore, "discoverers of the principle" in hypotheses one and two represent 35 who dis- covered the principle at some point during the learning task, while the corresponding Ss in hypotheses three and four repre- sent those Ss who discovered the principle during the first learning trial. Hypothesis 5 (stated in null form): The mean number of words recalled on the first test trial by Ss who discover the principle during the first learning trial will be less than, or equal to the correspond- ing mean for 85 who do not discover the principle. Hypothesis 4: With one exception, 85 who discover the principle during the first learning trial will make fewer intrusions in recall on the first test trial than 85 who do not discover the principle. (The single exception is the number of imprOper plurals, where the opposite prediction is made.) If scores on the first test trial of Experiment II are inter- changed with scores over all eight test trials of Experiment I, the five corollaries of hypothesis four are identical to those of hypothesis two. For purposes of convenience, therefore, the corollary hypotheses of hypothesis four are expressed as follows: Corollary 4a: Same as Corollary 2a. Corollary 4b: Same as Corollary 2b. Corollary 4;: Same as Corollary 2c. Corollary 4d: Same as Corollary 2d. Corollary 4e: Same as Corollary 2e. Figure 1 on the following page depicts the learning curves for Ss who either did or did not discover the principle at some point during Experiment I. The curves for Ss presented one- and two—homonym lists are portrayed independently due to the differences in both terminology and underlying principles on Mean Number of Words Recalled 67 18‘ 16: 14‘ 12- 8‘ I I / One homonym (discovered I principle) I . I Two homonyms (discovered I principle) / A Aprinciple) F A___AOne homonym (,did not discover Two homonyms (did not dis- ..0 . O 0- cover pr1nc1ple) Figure 1. I l 4 5 6 Test Trials “H 04-1 #- m—J Mean number of words correctly recalled over eight test trials by Ss who either did, or did not, discover the principle (Experiment I). 68 these two lists.21 Inspection of Figure 1 reveals that the performance of 85 who discovered the principle was clearly superior to the corresponding performance of 35 who did not discover the principle. This superiority is also apparent in the scores of these two groups along other dependent variables of Experiment I. Table 4 on the following page presents the sample means and variances for the seven dependent variables of this experi— ment. Inspection of this table reveals that, without ex- ception, the means are distributed in the predicted direction. Furthermore, scores of 88 who could not be classified, which are also presented in Table 4, generally support the assertion that exclusion of these 85 from the sample did not have a significant impact on the results. This follows from the fact that the performance of this group of Ss was usually somewhere between that of 85 who either did or did not dis- cover the principle. In view of these findings, a series of one-tailed t-tests were computed as tests of hypotheses one and two.22 The re- sults of these tests are also presented in Table 4. These 21Because of these differences, this distinction shall be maintained throughout most subsequent analyses in this disser- tation. 22There are two distinct formulas which may be used in computing the denominator of student's t-ratio. As a general rule of thumb, the author used the formula for unequal popu- lation variances whenever the ratio of the obtained sample variances exceeded two. Otherwise, the pooled estimate formula was used. This rule was followed throughout the dissertation. .H0.0 V m** 69 .H quEflHmmxm mCHHDU unflom mEOm .UHO HOEUHO 033 mm HO ma0>ma UUGMEHONHOQ CUTE USU Cmm3uofl mOUCmeMNHU MO mummB **m¢.m **do.elr*em.e md.ayr*md.m **>o.m **em.ml Anz .m> my mumwuup x oh.mm 0>.0H om.m om.o om.> om.mmm om.m mm m AQZV pmamammmHu om.m om.e om.d ow.o ow.m o¢.moa om.m m. on you pasoo 0:3 mm x am.mm mm.me em.md mm.a mm.m mm.aoe me.m mm mm Aazv mamflocflum onu Om.¢a mm.m om.m em.o >a.¢ m¢.moa am.> .W Hm>oomwo uo: can 053 mm x >d.NN mm.m mo.d om.m ¢>.w em.mma am.m mm be Any mamflocflnm mm.m md.m mm.o mN.H mm.m bd.mmd oh.m m. on» omum>oumwo 0:3 mm mumHJVEMdoEomloza .m **oo.m *toa.a r*ma.m me.u ;*ms.m **ao.m **sm.au 1oz .m> as mummuuu x am.mm mm.m em.fi_ mm.m m>.w mm.mm ¢H.m mm m AUZV powwflmmmao m>.oa mm.m md.m om.a mm.m mm.mma om.m .m on no: UHDOU 0:3 mm x mm.>¢ Ha.ma mm.o mm.d em.m mm.mmd Hm.m mm mm AQZV mHmHUCAHm mnu mm.oa m¢.m mm.n m>.o em.m No.>da mm.m .M Hm>oomflp uoc.owp 0:3 mm x I mm.nfl m>.N mm.e 05.0 mm.m mo.mm Na.m mm he AQV mamfioeflum mw.e mm.d om.d mm.o wm.a >m.>md me.m .m can pmuo>oomap 0:3 mm mumflq E>coEomeco .4 rLI rim ism“ man" Pun o.m .4mN W. N "no u_m "um. T.m Trm nvo nvm a 3.+ q+ 1 n s 1.4 6 Tie 1 1 1 q 54 Tie o e n_L n a n e 9.9 o a e.L 1.a u s s I 8.1 I71 1.1 0 tin P T... u T: T... S 1. . 3. u .4. o o r. o e r. r.s e.+ "um. u n max m OAu m. 7‘1 s s 3 3.4 Au u e 1;! e 1 1 1 set a o e 1 e e o .L 1 u.L I TL d r. s r. r. a u S S J 5 3 3 um OHQHUCHHQ mnu Hm>00mflo .uoc pap no .a means 70 tests indicate that the differences between means for 85 who either did or did not discover the principle are sig- nificant across all but one of the dependent variables (P < 0.01). The only exception occurred for the measure of number of improper plurals. Differences on this variable fell far short of significance. In view of these results, both corollaries of Hypothesis One as well as corollaries "a", "b", "c", and "e" of Hypothesis Two are rejected as stated in the null form (P < 0.01). Corollary 2d, on the other hand, must be accepted as stated in the null form. Two of the above decisions must be regarded as tentative; namely, the decisions to reject corollaries 1a and 2a. Due to the extensive number of interpolations which were made in deriving a measure of trials to criterion, the decision to reject corollary 1a is based on the magnitude of the obtained differences, rather than on a strict statistical test. The decision to reject Corollary 2a is also tentative. When this analysis was repeated for classifications made by the inde- pendent judge, the difference between the mean number of spelling errors made by 85 who either did or did not discover the principle failed to reach a significant level for one- homonym lists (see Table 20 in Appendix C).23 Tests of Hypotheses Three and Four followed the same for- mat as tests of Hypotheses One and Two. Means and variances 23This was the only instance in which an analysis carried out for classifications made by the independent judge failed to confirm the corresponding analysis for classifications of the principle investigator! 71 for 55 who either did or did not discover the principle are depicted in Table 5 on the following page. Examination of this table reveals that, with one exception, the means are distributed in the predicted direction. The single exception was that 85 presented two-homonym lists who discovered the principle during the first learning trial made more spelling errors than 85 who did not discover the principle. One-tailed t-tests were therefore computed to determine which of the obtained differences were significant. The re- sults of these tests are also depicted in Table 5. Here there are decided differences between the results for Experiment I and those for Experiment II. In sharp contrast to Experiment I, differences along only one of the variables are signifi- cant for both one- and two-homonym lists; namely, number of words correctly recalled on the first test trial (P < 0.01). Differences along two other variables--number of improper plurals and number of intra—list intrusions--are significant for two-homonym lists (P < 0.05), but the corresponding dif- ferences on one-homonym lists are extremely small. This con- trast raises the question of whether or not these apparently significant differences might not also have been due to chance. In general, therefore, it would appear that Ss who discovered the principle did not differ from 53 who did not discover the principle as far as the number of intrusions made on the first test trial is concerned. 72 H0.0 V m ** m0.0 V m * oo.H *mm.a mN.o *mm.dn >m.on **mm.m AQZ .m> 0V mummulu . . . . . . x hazy an A ed M mm N no 0 mm H mm 0a mm mm mamwucflum mnu Hm>oo md.¢ >m.a mm.d 00.0 NH.H No.m .M Imflp uo: UHU 0:3 mm x mm.m ma.fi mm.d mm.o «N.H m>.> .M owum>oumwp 0:3 mm mumflq ENdOEomIOBB .m mm.d >H.O mm.d oa.o: no.0 **am.m Anz .m> av mummulu . . . _ . . a . x. 35 mm m em H em m NH 0 mm o m m m mm meHUCHHm on“ Hm>oo >m.m mH.d mo.N mH.o Hm.o >¢.w .M Ime no: UHU 0:3 mm x mm m mm d we m am 0 mm 0 mm b mm Hm ADV mamflucflum onu mo.m md.d Ne.a mfi.o mm.o mm.m .M omum>oomwo 0:3 mm mumflq Ehcosomloco .4 II T. N TN d N PN )N n. N no Fm Tm Tm Tm In a 144 s s n s a.w. 6 1 e nwo. 3 1 «+o. 3 e n I a e 2.5 0.5 e u s 1.1 4:1 711 T11 4.1 P T;u u u s 3 1 o m 4.; «+9 1. 4:3 T;o u 1 u 1 x m oAu nto s a. n 1. n 3. d u a II a s 1 s 1 1 s.L (\1 1 1.2 7:2 0 T. e u u e u 1 s s 1, 6 .HH ucmEHnmmxm mo Hmauu maficumoa umuflm oru mafiuop onHUGHHm mzu Hm>oomflp no: pap Ho pap Honuflm 0:3 mm mo mocmfiuomumm HMHHD ummu umuflm orb CH mooanGMMHp mo mummfi .m magma 75 These results therefore lead to the following decisions: Hypothesis Three is rejected as stated in the null form (P < 0.01). But Hypothesis Four must be accepted in toto, as stated in the null form. The combined results of tests of Hypotheses One through Four lead to the following generalizations. It would appear that Ss who discover the principle during the first learning trial recall more words correctly than those who do not, as early as the first test trial. However, these Ss do not ap- pear to make fewer intrusions in recall on this trial than 55 who do not discover the principle. Beyond this initial test trial, Ss who discover the principle apparently maintain their superior recall and also begin to make fewer intrusions than 83 who do not. These conclusions, which admittedly in- volve free generalization between Experiments I and II, will receive considerable attention in Chapter V of this disser- tation. III. Configural Perception and the Encoding Process Hypotheses Five and Six predict that the per cent of Ss who discover the principle will be larger for lists contain- ing positive perceptual cues than for lists which do not contain these cues. Hypothesis 5: Positive perceptual cues will enhance the discovery of the principle as early as the first learning trial. The corollary hypotheses, stated in null form, are as follows: 74 Corollary 5a: The per cent of individuals who discover the principle during the first learning trial when presented lists with one homonym will be less than, or equal to, the per cent of individuals who discover the principle when presented lists with two homonyms. Corollary 5b: The per cent of individuals who discover the principle during the first learning trial when presented lists in which the categorical terms are capitalized will be less than, or equal to, the per cent of individuals who discover the princi- ple when presented lists in which the categorical terms are not capitalized. Corollary 5c: The per cent of individuals who discover the principle during the first learning trial when presented lists in which the categorical terms are fixed in the first position will be less than, or equal to, the per cent of individuals who discover the principle when presented lists in which the categorical terms are not fixed in the first position. Data from Experiment II were used to test Hypothesis Five. All Ss in this experiment completed the level of principle formation questionnaire immediately after the first test trial. Thus, responses to this form should determine whether or not a given S discovered the principle during the first learning trial. Hypothesis 6: Positive perceptual cues will enhance the discovery of the underlying principle throughout the learning task. The corollary hypotheses, stated in null form, are as follows: Corollary 6a: The per cent of individuals who discover the principle at some point during the learning task when presented lists with one homonym ‘will be less than, or equal to, the per cent of in- dividuals who discover the principle when presented lists with two homonyms. Corollary 6b: The per cent of individuals who discover the principle at some point during the learn- ing task when presented lists in which the categori- cal terms are capitalized will be less than, or equal to, the per cent of individuals who discover the principle when presented lists in which the categori- cal terms are not capitalized. 75 Corollary 6c: The per cent of individuals who discover the principle at some point during the learning task when presented lists in which the categorical terms are fixed in the first position will be less than, or equal to, the per cent of individuals who discover the principle when pre- sented lists in which the categorical terms are not fixed in the first position. Data from Experiment I were used to test Hypothesis Six. Ss in this experiment completed the level of principle forma— tion questionnaire immediately after they reached criterion, or after the eighth test trial, whichever came first. There- fore, responses to the questionnaire should determine whether or not a given S discovered the principle at some point during the learning task. The per cent of Ss who discovered the principle during the first learning trial of Experiment II is depicted in Column I of Table 6 on the following page. The per cent of 85 who discovered the principle at some point during Experi— ment I is shown in column II of the same table. In general, the per cents are distributed in the predicted direction for all three independent variables. A one—tailed test of difference in per cent was therefore computed to determine whether or not the obtained differences were statistically significant. This test, which is des- cribed by Garrett (1958), is based on the central limit theorem, i.e., the binomial distribution closely approximates the normal distribution for large values of N. Thus, the "3" which is obtained from a given test, represents the number of standard deviations an obtained difference in Table 6. 76 Per cent of 85 who discovered the principle during the first learning trial of Experiment II (column I) or'at some point during the eight learning trials of Experiment II (column II) when presented lists with varying perceptual cues. t J Experiment II: First Test Trial Experiment I: Test Trials 1-8 33 presented lists with 85 presented lists with 83 presented lists in which the categorical term is capitalized 85 one homonym two homonyms presented lists in which the categorical term is not capital- ized 28.18% (N = 110) 19.51% (N = 125) 28.44% (N = 109) 19.56% (N = 124) 44.76% (N = 105) 47.00% (N = 100) negative value 50.96% (N = 104) 40.59% (N = 101) 85 presented lists in which the categorical term is fixed in Ss presented lists in which the categorical term is not fixed in first position Z-tests first position 25.48% (N = 115) 25.75% (N = 118) negative value 49.51% (N = 101) 42.51% (N = 104) 1.04 Value needed for significance = 1.645 (alpha = 0.05) 77 per cent falls above or below the hypothesized difference of zero per cent. The obtained "Z's" are also depicted in Table 6. In order to be significant at the 0.05 level, the "Z" resulting from this test must be equal to, or greater than, 1.645. Although none of the obtained "Z's" exceeded 1.645, the results of three tests were extremely close to this value. Differences between Ss presented lists in which the cate- gorical terms were either capitalized or not capitalized approached significance in both Experiment I and Experiment II. The same is true of the difference between Ss presented one and two homonym lists during the single learning trial of Experiment II. However, since none of the obtained "B's" exceeded 1.645, Hypotheses Five and Six must be accepted as stated in the null form. Despite this decision, the three differences which approached significance will receive further consideration in Chapter V. Hypotheses Seven through Ten predict that the performance of Ss presented lists with positive perceptual cues will be superior to the corresponding performance of 83 presented lists which lack these cues. These predictions are based on the logical combination of the hypotheses which have thus far been considered. Hypotheses Five and Six, for example, predict 78 that a greater number of 85 will discover the principle when presented lists with positive perceptual cues than when presented lists which lack these cues. This prediction, therefore, takes the form of an "if A, then B" premise. According to hypotheses one through four, discovery of the principle should, in turn, result in improved performance. This prediction, therefore, takes the form of an "if B, then C" premise. Since hypotheses seven through ten predict that the performance of 85 presented lists with positive perceptual cues will be superior to the corresponding performance of 85 presented lists which lack these cues, these hypotheses represent the deductive conclusion to the above premises; namely, "A, therefore C". The validity of this conclusion was examined for the following four dependent variables: number of words recalled on the first test trial (Hypothesis 7), number of trials to criterion (Hypothesis 8), total number of intrusions (Hypothe— sis 9), and total number of words recalled over all eight test trials (Hypothesis 10). Hyppthesis 7: Because of their effect on the formation of principles, positive perceptual cues will aid in the formation of associations as early as the first learning trial. The corollary hypotheses, stated in null form, are as follows: Corollary 7a: The mean number of words cor- rectly recalled on the first test trial by Ss presented lists with one homonym will be less than, or equal to, the corresponding mean for 85 presented lists with two homonyms. Corollary 7b: The mean number of words cor- rectly recalled on the first test trial by Ss pre— sented lists in which the categorical terms are capitalized will be less than, or equal to, the 79 corresponding mean for 85 presented lists in which the categorical terms are not capitalized. Corollary 7c: The mean number of words cor- rectly recalled on the first test trial by Ss presented lists in which the categorical terms are fixed in the first position will be less than, or equal to, the corresponding mean for 85 presented lists in which the categorical terms are not fixed in the first position. Hypothesis 8: Because of their effect on the discovery of the relevant principle, lists containing positive perceptual cues will be learned more rapidly than lists which lack these cues. The corollary hypothe- ses, stated in null form, are as follows: Corollary 8a: The mean number of trials to , criterion for Ss presented lists with one homonym “ will be greater than, or equal to, the corresponding mean for 83 presented lists with two homonyms. Corollary 8b: The mean number of trials to criterion for 85 presented lists in which the cate- gorical terms are capitalized will be greater than, or equal to, the corresponding mean for 83 presented lists in which the categorical terms are not capi- talized. Corollary 8c: The mean number of trials to criterion for 85 presented lists in which the cate- gorical terms are fixed in the first position will be greater than, or equal to, the corresponding mean for 85 presented lists in which the categorical terms are not fixed in the first position. Hypothesis 9: Because of their effect on the discovery of the relevant principle, lists with positive per- ceptual cues will elicit fewer intrusions in recall than lists which lack these cues. The three corol- lary hypotheses, stated in null form, are as follows: Corollary 9a: The mean total number of intru- sions made by 53 presented lists with one homonym will be greater than, or equal to, the corresponding mean for 85 presented lists with two homonyms. Corollary 9b: The mean total number of intru— sions made by 83 presented lists in which the cate- gorical terms are capitalized will be greater than, or equal to, the corresponding mean for 83 presented lists in which the categorical terms are not capi- talized. Corollary 9c: The mean total number of intru- sions made by 85 presented lists in which the cateb gorical terms are fixed in the first position will be greater than, or equal to, the corresponding mean for 85 presented lists in which the categorical terms are not fixed in the first position. 80 Hypothesis 10: Because of their effect on the formation of principles, positive perceptual cues will en- hance the formation of verbal associations through- out the learning task. The corollary hypotheses, stated in null form, are as follows: Corollary 10a: The mean total number of words recalled by 35 presented lists with one homonym will be less than, or equal to, the corresponding mean for 83 presented lists with two homonyms. Corollary 10b: The mean total number of words recalled by 85 presented lists in which the cate- gorical terms are capitalized will be less than, or equal to, the corresponding mean for 85 presented lists in which the categorical terms are not capi- talized. Corollary 10c: The mean total number of words recalled by Ss presented lists in which the cate- gorical terms are fixed in the first position will be less than, or equal to, the corresponding mean for 85 presented lists in which the categorical terms are not fixed in the first position. The tests of hypotheses seven, eight, and nine followed approximately the same format. In each case, means for the relevant dependent variable were computed for each condition along the independent or treatment variables. These means-- number of words recalled on the first test trial; number of trials to criterion; and, total number of intrusions--are shown in Column II of Tables 7, 8, and 9, respectively. As an examination of Table 7 reveals, the number of words re— called on the first test trial was determined for both Experiment I and Experiment 11.24 Examination of Tables 7, 8, and 9 reveals that, without exception, the obtained means in Experiment I were distributed 24Although this would appear to represent a replication of the test of Hypothesis 7, the difference in directions for these two experiments suggests that this analysis might better be viewed as a repeated test under two distinct test condi- tions. 81 Ho.o_m 4*; 50m Hmuoa H>.m 00m CHsqu mm.o om.m H m x U x m 00.0 $0.0 H m x U em.m 0m.mm H m x m mm.0 mo.w H o x m umGOHuumumucH 0N.0 Hm.m H HH.> .m> mm.m Amv cOHuHmom EOUGMm .m> pmme no.0 mo.m H mm.> .m> H@.@ 100 memo oz .m> memo *4mm.HH om.mHH H mm.m .m> we.» Arc mascOEOm 039 .m> mac HH ucwEHHmQXm mo HMHHB ummB umHHm .m mnH Hmuoa m>.m mmH GHSUHK Hm.m mm.mm H m x U x m 0H.0 00.0 H m x o hH.0 H0.H H m x m Hm.0 0m.m H O x m umcoHuomumucH mm.H em.mH H 00.0 .m> No.0 Amy cOHuHmom Eopcmm .m> pmme as.m mo.sm H ms.m .m> mm.m 100 memo oz .m> memo 4*mm.mm ms.mmm H no.4 .m> we.» Irv mascoeom 039 .m> mco m ucmfiHummxm mo HMHHB umma umnHm .4 mOHumulm .m.z .m.p memo: 0Hmamm condom ; .wmoo Hodummuumm mcH Imum> £UH3 mumHH poucmmmum mm an HMHHU ummu umHHm may :0 UmHHmomH hHuomHHoo mUHOB mo Hmflfidc one m0 ummu 00CMHHM> mo mHmmHmcm .muowmm0IUQXHm .NXNxN < .5 DHQMB 82 05H Hmuoa 00.00 00H 4 cHnuHS 05.0 00.00 H m x O N m 00.N 00.H>H H . . m x O 00.0 H0.0m H 100 0060 co m x m **H0.H 40.04H H OH oHnmu mmmv o x m quOHuumnmucH 00.0 000.0 H 00.0 .m> 00.0 A00 cOHDHmom Eoocmm .m> Umme mm.o «0.0H H Hm.m .m> 00.0 100 memo oz .m> memo *4Hm.HH 00.000 H H0.HH .m> 00.> Irv newcosom 039 .m> mco OHDMHIM .0.z .m.o meow: mousom .AH ucmfiHnmmxmv 0050 Hmsummuumm 0GHhum> cuHB mumHH omucmmmum mm an 0005 mGOHm louucH mo Hones: Hmuou 0:» mo ummu 00GMHHM> mo mHmmHmcm muommmm Umme NxNxN d .0 0HQMB H0.0 V m** 0>H Hmuos 00.0 00H GHEHHS | | I II II II I;- II. II ' II II III II- II l :- al- I I ' II- | II- II II I | 'I cl II I. n' I | II | ' 00.0 bN.N H m x O x m 00.0 5N.N H m x O 00.0 >N.N H m x m 00.0 0H.0H H U x m umGOHuUmnmucH No.0 0H.0 H 00.0 .m> H0.0 Amy QOHuHmom Eowcmm .m> Umme mm.m 00.0 H 00.0 .m> HH.0 100 memo oz .m> memo **mH.m Hm.mm H 00.» .m> Hm.m Amy maxeosom 639 .m> mao OHDMHtm ...0.2 .0.0 meow: 00H500 . c m cmxmu mHMMMW HMDHQGUHGQ mCHMHm> SUH3 mumHH Umucwwmum mm %0 COHHMMHWOTWMMMHXWW -. w HQQESC 03w MO umwu 00C0HHM> MO MHUNHMGM muowmwm ©0XHM NXNXN G .m QHQMB 85 in the predicted direction. Regardless of the dependent variable on which performance was determined, Ss presented lists with one homonym,capitalized categorical terms, or the categorical terms in a fixed position, outperformed Ss pre- sented lists which lacked these cues. Examination of "Part B“ of Table 7, on the other hand, reveals that differences in the mean number of words recalled on the first test trial of Experiment II were distributed in the predicted direction for only one of the three independent variables; namely, the dif- ference between 85 presented one and two homonym lists. Despite this apparent conflict between the results of Experiment I and Experiment II, a 2x2x2 fixed effects analysis of variance was computed for each dependent variable. Prior to this computation the sample size in each of the eight cells was equated by randomly eliminating Ss from the sample. In Experiment I the cell sample was equated at 22, while in Experiment II the "n" in each cell was set equal to 26. The advantage of equating the number of Ss in each cell over computations involving unequal "n's" rests not only in greater ease of computation, but in greater robustness to violations in the assumption of homogeneity of variance as well (Guenther, 1964). Following this adjustment, the analysis of variance tests were computed. Mean square variances and F-ratios resulting from these tests are also depicted in Tables 7, 8, and 9. The results of these tests are remarkably consistent across 84 all three dependent variables. In each case, the difference in performance between 85 presented lists with either one or two homonyms was statistically significant (P < 0.01). Dif— ferences along the other two independent variables, on the other hand, were not statistically significant. In most cases, these differences did not even approach significance. Only one other result reached a statistically significant level; namely, the interaction between number of homonyms and capitalization of the categorical terms when total number of intrusions served as the dependent variable. Means for the four conditions in this interaction are shown in Part "A" of Table 10 on the following page. Examination of this table reveals that capitalizing the categorical terms reduces the number of intrusions on one-homonym lists, but slightly in— creases the number of intrusions on two-homonym lists. The corresponding interaction for various other dependent variables is also shown in Table 10. Although the interaction on these measures fails to reach significance when alpha is equal to 0.05, each distribution suggests that capitalizing the categorical terms has a greater impact on one-homonym than on two-homonym lists. Further, many of these interactions approach statistical significance. The only evidence which runs counter to the above was ob- tained for number of words correctly recalled on the first test trial of Experiment II. Here, differences were somewhat larger for two-homonym than for one-homonym lists. But this v Q.\q.-- c :- ,i~ adkupyv J H..J Lin-z. .q‘; manure 7.! AHJ '3‘ I l l .- lwfdhh \ A name.“ a !.NJ..1 ~15}; i. AMP-.1. - n~.v.- Vilflh‘.fl.rvskd-\~ ”EU 0.9 ~ 1‘ l..‘r 85 00.04 am.mm 00.9H am.Hm 080:0800 EN:oEom 039 0:0 0000 02 0000 AH 0:0EHu0mxmv 01H 0H0Huu 0:H:H00H 0:HH:0 0H0H0:Hum 0:HH0>000H0 0:00 :00 A00 00.0 00.3 09mm memo 62 00.0 99.0 L am.mm eo.Hm memo 00.9 HH.m _ 080:0500 EN:oEom 08>:oEom E0:0&om 039 0:0 039 0:0 AHH #:0EHu0mxmv H0Huu mcH:u00H umuHm 0:HH:0 0H0H0:Hum 0:HH0>000H0 0:00 :00 A00 0Humnlm 0000 02 0000 ":OHH0DHH0 ou 0H0HH9 A00 N0.N 0Hu0ulm H0.> 0Hu0nlm 9N.HHH 09.0HH 0000 oz 0N.0H 90.0 0000 oz 00.0HH No.0mH 0000 00.NH N0.0 0000 mEN:oEom EN:oEom 059:0500 EN:oEom 039 0:0 039 0:0 uu00uuoo “098:: H0u09 Amy 0:0 05%:OEOE mo M0955: :003009 :OHUUMH0DGH 0£u :H 0:008 MO COHUSQHHumHQ «MCOHWDHHZHH H0QES: H0009 A00 .Eu0u H0UHH000000 0:0 mo :OHu0NHH0UH000 .OH 0Hnt 86 finding presents an interesting contrast to the per cent of Ss who discovered the principle during the first learning trial of this same experiment (depicted in Table 10 above). In short, the overwhelming majority of evidence supports the assertion that the effect of capitalizing the categori- cal terms is more pronounced on one-homonyms lists than on two-homonym lists. This finding will therefore receive further consideration in Chapter V. The results of the four analysis of variance tests de- picted in Tables 7, 8, and 9 above, lead to the following decisions. Corollaries 6a, 7a, and 8a, are all rejected when stated in the null form (P < 0.01). Regardless of which dependent variable is considered, the performance of Ss pre- sented one-homonym lists appears to be superior to the cor- responding performance of Ss presented two-homonym lists. Corollaries "b" and "c" of hypotheses six, seven, and eight, on the other hand, must all be accepted as stated in the null form. Regardless of which dependent variable is considered, the performance of Ss presented lists in which the categori- cal terms are fixed in the first position does not appear to differ from the corresponding performance of 85 presented lists in which the position of the categorical terms is varied from triad to triad. Further, if capitalizing the categorical term does have any impact on performance, this effect is probably limited to one—homonym lists. 87 The test of hypothesis ten was somewhat more elaborate than that of hypotheses seven through nine. The decision to use a different analysis stemmed from the author's inspection of the learning curves for each of the list conditions where number of words correctly recalled served as the dependent variable. These curves, which are depicted in Figure 2 on the following page, reveal that differences between the two conditions on each treatment variable are larger on the first two test trials than on any subsequent trials. In View of these observations, it seemed desirable to determine whether or not the interactions between test trials and any of the independent variables were significant. The existence of a significant interaction would, in turn, suggest that the effect of the relevant variable was dependent upon the particular test trial on which retention was determined. These considerations led the author to select a four-way analysis of variance test, with test trials 1-8 serving as the fourth variable. However, the inclusion of the test trials variable gives rise to an important methodological consideration. Simply stated, observations on this variable are not independent, i.e., scores for any two trials are based on the same group of individuals. Including this variable in the analysis therefore violates the assumption of independent observations which underlies any analysis of vari- ance test . 88 .HH 0:0EHH0mxmv 00:0 H0sum0on0m 0:H»H0> :uHB mumHH 000:00000 00 How A00HH000H 0H000HH00 mUHOB mo H0AE::V m0>u50 0:H::00A 0H0HH9 0009 0 0 p b p b b p w 0 H 800:0: old 099: old H1 «0:000AH [NH .HH r0H 6MHHH :Haqu cOHuHmom .6 .00. .NH. 0H0HH9 0009 0 9 0 0 e 0. m H p H H r H H p 0 00NH . 0-09. 0 IH0UH000 uoz Clio H .00. [WOH 00NHH0DH00O 1 HH "0:0004 \ H -NH 10H 10H :0H mau09 H00Hu000u00 I u:HHm 0NHO .m 0 0 2.331.103 .IBCUHUN UEGW 0 H .N 0H50Hm 0H0HH9 0009 b 0 0 H 0 N p p . p p b muqu H 50:0500 039 0!:0 H w ,r mumflfl HH 40 80:01:00 0:0 E H H ”0:000Aw 1 mE>:oSom mo :0neoz .0 00.0 00.5 0H NH HH 0H 1031103 1eqmnN ueew 0H 89 Fortunately, Geiser and Greenhouse (1958) have described a computational procedure which takes full account of this violation of the assumption of independence. This procedure involves partitioning the within variance into two compon- ents--a within groups variance and a treatments x individuals 'within groups variance. The within groups variance is a pooled estimate of the variance of observations about the in- dependent treatment means, while the treatments x individuals within groups variance is a pooled estimate of the variance of observations about the correlated treatment means. Thus, the within groups variance is used in the denominator of F-ratios for tests of differences between independent means while the treatments x individuals within groups variance is used in the denominator of F-ratios for tests of differences between correlated means. But this computational revision does not fully account for the correlations which exist among observations whenever repeated measurements are made on the same group of indi- viduals. As Geiser and Greenhouse (1958) point out, these correlations must fall somewhere between zero and one. If the correlations are zero, the observations are independent and the analysis of variance test is apprOpriate. But if the correlations are equal to one, it is possible to generate tire complete set of observations from knowledge of observa- tions on only one condition of the variable, thereby reducing the degrees of freedom to one. 90 In view of these considerations, Geiser and Greenhouse suggest that two tests should be used for every analysis based on non—independent observations--a liberal test in which the obtained F-ratios are compared with values needed for significance when the degrees of freedom in the numerator are set equal to K-1, and a conservative test in which the obtained F-ratios are compared with values needed for sig- nificance when the degrees of freedom in the numerator are reduced to one. If the F-ratios exceed the conservative value, one may safely conclude that the means are significantly dif- ferent. But if the F-ratios do not exceed the liberal value, one must conclude that the means are not significantly dif- ferent. In view of these considerations, a 2x2x2x8 fixed effects analysis of variance test was computed using the revisions suggested by Geiser and Greenhouse. As in preceding analyses, the sample size was reduced to 22 in each cell prior to the computation. The results of this analysis are presented in Table 11 on the following page. F-ratios for interactions between trials and the three perceptual cues were subjected to both the liberal and conservative tests. Results of these comparisons are also shown in Table 11. An examination of Table 11 reveals that two of the inter- actions are significant--trials x number of homonyms (P < 0.01-- conservative test), and trials x capitalization of the cate- gorical terms (P < 0.05--liberal test). The distribution of 91 5009 HOHOQHA 00.0 V 0* 5009 0>H50>n00:00 H0.0 v 0** 9oeH Hmuoa mm.m 09HH mmsoum :Hequ 0H050H>H00H N 00:0850009 «0.0 00.m 9 9 x 0 x m mo.H 90.0 9 a x m x 0 00.H 09.9 9 9 x m x 0 00.0 9m.m 9 9 x 0 x m NH.H 00.0 9 a x m *00.0 00.0 9 0 x 9 **00.0 00.00 9 9 x m "0:0H000H00:H **Hm.000 00.0H0m 9 H90 0H0HH9 om.Hm, 00H mmsouo 00H: 0H.H 700.00 H m x 0 x m 00.H 00.00 H 0 x 0 000.0 H0.0 H 0 x m 00.0 00.90 H 0 x 0 "0:0H500H00:H 100 09.0 HH.mH H 09.0HH .m> 00.9HH :OHuHmom soecmm .m> emme HH.H 00.00 H 00.0HH .m> 00.0HH H00 0000 oz .m> 0000 **H0.mm 00.009 H 90.0HH .m> NH.NNH H00 meacosom 039 .m> mqo 0Hu0ul0 0H0500 :00: .0.0 0:002 0H0E00 00H500 :H 0:0H0H>0m .H000Hv 0050::0050 0:0 H00H00 :0 0000A :0Hu0u50200 .0050 H05u000000 0:H>u0> :uHB mumHH 000:00000 00 09 00HH000H 00503 00 500:5: H0500 0:5 00 0000 00:0H:0> mo 0H09H0:0 0xmxmxm 0 mo 00H500m .HH 0HQ09 92 means in these two interactions is represented by learning curves "A" and "B" in Figure 2 above. Although the inter- actions may be in large part attributed to decreases in the differences between the two conditions as the 33 approach the low-ceiling in the task, these differences also display some variability across the first three test trials where the low-ceiling effect is not pronounced. Differences between Ss presented either one- or two- homonym lists, for example, are fairly constant across the first two test trials, but show some decrease on the third test trial. Differences between 55 presented lists in which the categorical term is either capitalized or not capital- ized, on the other hand, reach an asymtote on test trial two, and then decrease dramatically on test trial three. These results suggest that the effect of the two relevant treatments is sizable across the first two test trials, but is then counteracted by the low-ceiling effect during test trials three through eight. The results also suggest that the author should have focused his analysis on the second test trial where differences reached a peak, rather than during the first test trial where they had not yet reached this level. Regarding the effects of the three treatment variables, the results of this test are highly consistent with those obtained in tests of Hypotheses seven through nine. Once again, the means for the three treatments fall in the pre- dicted direction, but only the difference between one- and 95 two-homonym lists is significant (P < 0.01). The only other result which approaches statistical significance is that ob- tained for the interaction between number of homonyms and capitalization of the categorical terms. The distribution of means in this interaction, which are presented in Table 10 above, suggests that the effect of capitalizing the cate- gorical terms is more pronounced for one-homonym than for two—homonym lists. Because the F—ratios obtained in the above test are algebraically equivalent to those obtained in a standard, three-way analysis of variance test of the total number of words recalled over all eight test trials, it is possible to use these results to make the following decisions regarding hypothesis ten. Corollary 10a is the only corollary which should be rejected as stated in the null form (P < 0.01). Corollaries 10b and 10c must be accepted as stated in the null form. The inferences which may be made from this test are therefore identical to those of hypotheses seven through nine. Further Analysis of the Effect of Configural Perception on the Derivation of Higher-Order Coding Schemes Three variables have been shown to affect the short—term retention of a given 8 in this study--whether or not he dis- covered the relevant principle; whether he was presented a list with one-homonym or a list with two-homonyms; and, whether he received a list in which the categorical terms were 94 capitalized or a list in which these terms were not capital- ized. In an attempt to determine the basic source of these effects, the author undertook the following exploratory analysis. First, each correct response over the first three test trials was identified by type--categorical, misspelled ex- emplar, or exemplar.25 The first three test trials were selected for this analysis because differences in recall were largest on these trials, and, because all three stimulus lists had been presented by this point in the task. "Type of re- sponse“ was then treated as an independent variable in a four-way analysis of variance test. The three variables list- ed above constituted the other three treatments in this analy— sis. By examining the interactions between "type of response" and each treatment it is possible to gain some insight into the source of the obtained differences in performance between the two conditions on each variable. Suppose, for example, that there is a significant inter- action between "type of response" and discovery of the prin— ciple. Further suppose that the distribution of means in this interaction suggests that differences between 33 who either did or did not discover the principle were small for both exemplar and misspelled exemplar terms, but were quite large fcm'categorical terms. This would suggest that the categori- cal terms play a critical role in the discovery of the assee Table 1 in Chapter III for the meaning of these three types of response. 95 principle. This, in turn, would explain why capitalizing the categorical terms increased retention; namely, differen- tial attention to the categorical terms resulting from capitalizing these terms resulted in an increased likelihood of discovering the principle. In contrast to this explanation, suppose that the inter— action between "type of response“ and capitalization of the categorical terms is significant. Further suppose that the distribution of means in this interaction suggests that dif- ferences between Ss presented lists in which the categorical terms are either capitalized or not capitalized are large for categorical terms, but are small for exemplar and misspelled— exemplar terms. This result would suggest that the differen- tial recall between these two groups stemmed from simple dif— ferences in the recall of the categorical terms and not from differences in the likelihood of discovering the principle as the above example suggests. In short, it should be pos- sible to test several potential explanations of the effect of each of the three treatment variables by examining the inter- actions between "type of response" and each treatment. A 2x2x2x5 analysis of variance test was therefore com— puted with "type of response" as the fourth independent vari- able. The dependent variable in this analysis was number of words correctly recalled over the first three test trials. Since observations on the "type of response“ variable are based on the same group of Ss, the modifications suggested 96 by Geiser and Greenhouse (1958) were adopted. By randomly eliminating Ss from the sample, the "n" in each cell was equated at nineteen. This had the further effect of equating the number of Ss who discovered the principle for the two conditions on both list variables. The results of this analysis are presented in Table 12 on the following page. The first finding of some importance is that the difference between Ss presented either one- or two-homonym lists remains significant despite the fact that, in this analysis, the per cent of Ss who discovered the principle was equal for these two groups (P < 0.01). Differ— ences between 33 presented lists in which the categorical terms were either capitalized or not capitalized, on the other hand, were severely reduced by these conditions. A serendipidous finding was the significant interaction be- tween discovery of the principle and number of homonyms (P < 0.01). Reexamination of Figure 2 in Section II of this Chapter reveals that discovery of the principle is more beneficial to Ss presented lists with two homonyms than to 35 presented lists with one homonym. Differences in the mean number of words correctly recalled over the first three test trials were 1.69 and 5.16 for one- and two-homonym lists, respectively. But the interactions between "type of response" and each of the three treatments failed to even approach significance. Although this result was greeted with initial disappointment, 97 d0.0 V 0** 009 H0009 00.0 000 000000 000003 0H05UH>HU:H x 00:0E00009 0N.o No.N N m x U x m x 0 00.0 No.0 N 0 x 0 x m 00.0 N0.0 N 0 x U x 0 00.0 0N.o .N 0 x m x 0 Nm.o om.N .N m x U 9H.o 09.H N m x m 9H.o 00H 0 m x 0 “0:0H000000:H H9.90 .00H050x0 0000000005 HN.mm 00H0E0x0 90.00 H00H0om0000 00.0 NH.0 N ”Amy 00:00000 00 0009 09.9 99H 000000 000003 N9.N 0H.0H H 0 x m x 0 00.0 HH.0N H U x m 00.0 90.H H o x 0 **0H.0 0H.H0 H m x 0 «0:00000000:H 00.0 90.9H H 90.00 .0> H9.90 A00 0000 oz .0> 0000 *000.90 0H.HH9 H 00.00 .0> 00.00 000 05000000 039 .0> 000 **N0.mm 00.090 H 0H.00 .0> 99.90 A0v 0H0 I00:000 0oz .0> 0H000:000 0000010 000500 :00: .0.0 0:002 000800 000500 ”f .0HQ0H00> £00500 0:0 00 =00:00000 00 0000: £003 .H 0:0EH000xm 00 0H0H00 0000 00000 00000 0:0 00>o 00000000 00003 00 . 0005:: 0:0 00 0000 00:0000> 00 000000:0 0000000u00x90 mexNxN 0 mo 0003000 m9 00009 98 it was nevertheless useful in eliminating several potential explanations of the effects of the three treatment variables. These results will therefore receive considerable attention in Chapter V. In view of this failure to obtain significant inter— actions, the author decided to analyze correct responses in still another way. Each correct response over the first three test trials was classified by the "type of association" in- volved: categorical-exemplar; categorical-misspelled exemplar; and, exemplar-misspelled exemplar. Categorical-exemplar associations, for example, involved giving an exemplar term as a response to a categorical stimulus or vice-versa. The other two types of association were identified in an analogous manner. "Type of association" was then treated as the fourth variable in a repetition of the 2x2x2x5 analysis of variance test presented above. The results of this test are presented in Table 15 on the following page. Once again, the interactions failed to ap- proach statistical significance. Although this result is somewhat discouraging, the test was nevertheless useful in eliminating other potential explanations of the effect of the three treatment variables. Therefore, this analysis will also be considered in more detail in Chapter V. 99 H0.0 V 0** mm0 00009 00.0 000 000000 000003 0H050H>HUCH .x 00:0E0000B mm.m N 4x00000000 HN.o m9.H N d x U x m 9N.o 9N.N N fl x U x 0 mm.o Nm.N N 0.x 0 x 0 00.0 00.0 N < x U H9.o mm.m N 0 x m H0.0 00.0 N d x 0 "00000000000H O9.Nm 00H0E0x0 00HH0000HEI00H0E0X0 H9.0m 00H0E0x0 U0HH000008100UH0000000 mo.¢m 00H0E0x0IH0UH0om000U mH.N ON.mH N "000 00H00H0000¢ 00 0009 09.9 09H 000000 000003 09.0 0H.0H H 0 x 0 x 0 00.0 HH.mN H U x m 00.0 0m.H H 0 x U **®H.m md.fim H m X 0 "00000000000H 00.0 90.9H H 90.00 .0> H9.90 000 0000 oz .0> 0000 **00.90 0H.HH9 H 00.o0 .0> 00.00 000 00000000 030 .0> 000 0000.00 00.000 0 00.00 .0> 09.00 000000000000 0oz .0> 000000000 0H000I0 000500 0002 .m.© 00002 0H0500 000500 0003 .H 00000000xm 00 000000 0000 00000 00000 0:0 00>o Umaaouv0 . wfiflo 3 w 000500“ .0HQ0H00> 00000000000 £00500 0:0 00 :00H00000000 mo fina* 000000 000 00 0000 0000000> 00 00000020 0000000a00x00 .00 00005 0000000 0 0o 100 IV. Higher-Order Coding Schemes and Long-Term Retention Exactly three weeks after the initial learning session, 85 in Experiment I were presented with a single test trial. Measures of performance on this trial were used to test Hypotheses Eleven through Fourteen. With the exception of Hypothesis Fourteen (which will be considered in more detail later in this section), this set of hypotheses predicts that long-term retention will be influenced by the same variables that affect short-term retention. Hypothesis 11 (gtated_in null form): The mean number of words correctly recalled on the test of long-term retention by Ss who discovered the principle during acquisition will be less than or equal to the mean number of words recalled by Ss who did not discover the principle. Hypothesis 12: With one exception, the number of intru- sions made on the test of long-term retention will be lower for those 85 who discovered the principle during acquisition than for those Ss who did not discover the principle. The single exception is number of improper plurals, where the opposite pre- diction is made. The corollary hypotheses, stated in null form, are as follows: Corollaryiiza: The mean number of spelling dis- tortions made by 85 who have discovered the princi— ple during acquisition will be greater than, or equal to, the mean number of spelling distortions made by Ss who have not discovered the principle. Corollary 12b: The mean number of extra-list intrusions made by Ss who have discovered the principle during acquisition will be greater than, or equal to, the mean number of extra-list intru- sions made by Ss who have not discovered the princi- ple. Corollary 12c: The mean number of intra-list intrusions made by Ss who have discovered the prin- ciple will be greater than, or equal to, the mean number of intra-list intrusions made by Ss who have not discovered the principle. 101 Corollary 12d: The mean number of improper plurals formed by Ss who have discovered the prin— ciple during acquisition will be less than, or equal to, the mean number of improper plurals formed by 83 who have not discovered the principle. Corollary 12e: The mean total number of intru- sions formed by 85 who have discovered the principle during acquisition will be greater than, or equal to, the mean total number of intrusions made by 85 who have not discovered the principle. Hypothesis 15: Because of their effect on the discovery of the prinCiple, “positive" conditions on each of three perceptual variables will aid in long-term retention. The three corollary hypotheses, stated in the null form, are as follows: Corollagy 15a: The mean number of words cor- rectly recalled on the test for long-term retention by 85 presented lists with one homonym will be less than, or equal to, the mean number of words recalled on this test by 83 presented lists with two homonyms. Corollary 15b: The mean number of words cor- rectly recalled on the test of long-term retention by 53 presented lists in which the categorical terms are capitalized will be less than, or equal to, the mean number of words recalled on this test by 85 presented lists in which the categorical terms are not capitalized. Corollary 13c: The mean number of words cor— rectly recalled on the test for long-term retention by 55 presented lists in which the categorical terms are fixed in the first position will be less than, or equal to, the mean number of words recalled on this test by 85 presented lists in which the categori— cal terms are not fixed in the first position. Tests of Hypotheses Eleven through Thirteen followed the same format as the corresponding tests of short-term retention. However, the sample was slightly modified for the analysis of long-term retention. In order to control for differences in the initial level of acquisition, only those Ss who reached Criterion were included in this analysis. Sample means and variances were computed for each depend- ent variable in Hypotheses Eleven and Twelve. These values 102 are depicted in Table 14 on the following page. Inspection of this table reveals that, with the exception of number of improper plurals on one-homonym lists, the means are dis- tributed in the predicted direction across the six dependent variables. One-tailed t-tests were therefore computed to determine whether or not the obtained differences were statistically significant. The results, which are also shown in Table 14, reveal that none of the differences even approach statistical significance. In View of these findings, Hypotheses Eleven and Twelve must be accepted as stated in the null form. The performance of Ss who have discovered the principle during acquisition does not appear to differ from the corresponding performance of 85 who have not discovered the principle. The test of Hypothesis Thirteen also followed the same format as earlier tests of the effect of list variables. The mean number of words recalled for each condition on the three independent variables is shown in Part "A" of Table 15 on page 104. With the exception of the "position within each triad" variable, these means are distributed in the opposite direction from that predicted. Therefore, the null statement of Corollaries 15a and 13b must be accepted without any further analysis. Part "B" of Table 15 depicts the mean total number of intrusions for each condition on the three treatment variables. Due to certain theoretical considerations advanced in the 105 033 mm.a 0a.a 00.a 0w.0u 05.0 00.0 A02 .m> 00 mpmmnuu x 00.ma mm.0 00.5 00.0 mm.m 00.0 mm m 1020 emanammmHo 00.¢ mm.0 00.m 00.0 50.« 5m.mfi .m mg no: casoo 0:3 mm x H0.5 m¢.0 5m.a ma.0 $0.0 H0-0 mm om 1020 mamflocaum ma.m 00.0 00.0 ma.0 00.a mo.aa mp Ho>oomao “on can 0:3 mm x 0H.¢ «H.0 m5.0 0¢.0 0~.m ma.ma mm mm 100 mamfloawum 00.0 06.0 mm.0 Hm.0 0¢.a 55.afi .m umum>oumH0 0:3 mm mama...“ ENGOEOmIOBB .fl Ha.a em.a mm.0 50.0 50.0 00.0 102 m> no mummuuu x m¢.m 5m.0 aa.m mm.o mm.m mm.m mm 5 1020 emanammmao 5m.e fi5.0 00.6 m«.0 5m.a 5m.m .m mg no: wagon 0:3 mm X . 5m.m 00.0 00.0 mm.o 00.0 ea.m mm mm Aazo mamaocflum 5m.m 00.0 mo.m 00.0 H~.H 00.0H m. um>oomflc no: can 0:3 mm x e«.¢ 00.0 00.0 50.0 mm.0 00.5 mm an 100 mamflocaum 00.0 ma.0 «5.fi mm.0 ma.a am.aa _m nmnm>oomH0 0:3.mm mumflq Ecofiomlmao .4 r. L I N I N d N T N 1 N w. N U 0 I.“ T.“ Tum "um. 8 m. 9 10+ S S n 5 D 5 I P 1.Q. 1.0. I a. E 9 n T. a 9 B 9 n39 T53 U s 1.1 1.1 T. J 1.1 T. 1 P vtu u U S S a o m .4 t. 1 a I. .4 s o. M u wAU IVA m OAu O s a. n a. n 1. d I a I a S J s I I 3 TL p 1 Te Te 0 TI s O . 0.. .d O _ u u 3 u L BunnnflflulhbT HM xim .COHUHmHSUUM mCHHDU mHmHUCHHm wnu H®>OUmHU HOG UHU HO UHU HGSUHM .wfi mHQmB mm @mOSU HON COHUCQUQH EHGUIECOH MO ummu mflu CO mH®>®H @UGMEHOMHmm Emma 104 m0.0 V m* moa HmuOB mm.m mm a0£u03 mm.0 mo.m 0 m x U x m no.0 00.0 0 m x 0 00.0 v 05.0 0 m x m 00.0 v 00.0 0 o x m "0:00uomnmucH mm.0 mo.m 0 m¢.m .m> om.m Amv c00u0mom Eoocmm .m> Umx0m 00.0 v 00.0 0 00.0 .m> 00.0 on 0000 oz .m> 0000 *00.0 00.00 0 00.0 .m> 05.0 Amy mamcoaom 030 .m> 000 mQOHmsuucH mo Hmnfisz 00009 .mfi mo0 00009 0m.m mm :05003 mm.o «m.m 0 m x U x m 00.0 om.m 0 m x 0 00.0 00.0 0 m x m 5N.m m¢.m0 0 U x m "mc00uomnmucH 00.0 mm.m0 0 55.00 .m> Nw.00 Amv GOHuHmom Eoocmm .m> pmx0m 00.0 00.0 0 50.00 .m> 00.00 A00 0000 oz .m> 0000 00.0 00.00 0 50.00 .m> 05.00 Amv 05000500 030 .0> mco Umaamomm quomuuoo monoz.mo Hmnfisz .< mO0umulm wumnvm G002 .m.© mammz mamsmm wousom .mwdu Hodumwoumm mcflmum> £003 mum00 Umucmmmum mm mo GOHucmumH Eumu ImQOH How ummu m£u GO mUGmEHOMHGQ m0 ummu wUGMHHm> MO mflmhamcm am no mudsmmm .mfi mHQmB 105 following discussion of Hypothesis Fourteen, it was not pos- sible to make a priori predictions of direction on this de- pendent variable. This analysis is therefore included for exploratory purposes rather than as a test of a formulated hypothesis. It is interesting to note, however, that the distribution of means for the number of homonyms treatment has reversed from that which occurred in short-term retention. A 2x2x2 fixed-effects analysis of variance'test was com- puted for both dependent variables. The sample size in each cell was equated at 15 prior to the analysis. The results of this test for number of words recalled and number of intru— sions are also presented in Parts "A" and "B“ of Table 15. The difference in total number of intrusions made by Ss pre- sented either one or two homonym lists was the only difference which reached statistical significance (P < 0.05). All the other differences may be properly attributed to chance. In view of these findings, Corollary 15c must also be accepted as stated in the null form. Unlike all preceding hypotheses in this study, Hypothesis Fourteen has direct theoretical ties. Both the concept of a "stable" trace and/or Ausubel's theory of "obliterative sub- sumption” (Ausubel, 1965), predict that generalizations will be better retained than specifics. If this is true, one might expect retention of the principle to persist over an interval 106 of three weeks, while retention of specific details will probably have dropped out by this point. The principle which underlies two-homonym lists contains a derivative rule for spelling; namely, spell both exemplar terms as homonyms. The principle underlying one-homonym lists, on the other hand, contains a correlative rule for spelling; namely, spell one of the exemplar terms as a homonym. If the principle is retained, while specifics drop out, then it follows that 55 presented two-homonym lists who discover the principle will not make any spelling errors on the test for retention, while the corresponding 53 presented one- homonym lists should make several errors of this type. This conclusion follows from the differences in the nature of the spelling rules for the two types of lists as Well as from the fact that remembering which term to spell as a homonym (in the case of one-homonym lists) involves recall of specifics. Hypothesis Fourteen therefore reflects these theoretical considerations: Hypothesis 14 (stated in null formL; Among 83 who dis- cover the principle during acquisition, the mean number of misspelled exemplar terms on the test for long-term retention made by Ss presented one- homonym lists will be less than, or equal to, the corresponding mean for 83 presented two-homonym lists. The mean number of spelling distortions made by Ss who discovered the principle and who reached criterion was de- termined for both one- and two-homonym lists. These means were 1.15 and 1.46 for one- and two-homonym lists, respectively. The obtained means were therefore distributed in the opposite 107 direction from that predicted. Thus, the null statement of Hypothesis Fourteen must be accepted without further analysis. In short, tests of Hypotheses Eleven through Fourteen all point toward one conclusion. Those 53 who discovered the principle during acquisition were apparently unable to utilize the principle during the test of long-term retention. CHAPTER V DISCUSSION AND CONCLUSIONS I. Higher-Order Coding Schemes and Short-Term Retention The effect of higher-order coding schemes on short-term retention was examined in tests of Hypotheses One through Four. These tests confirm the following general conclusions: (1) Approximately 50% of the Ss in this study were able to perceive and to utilize the principle by which each triad was constructed. (2) Those 85 who discovered the principle formed the necessary associations more rapidly than 83 who did not discover the principle. (a) 88 who discovered the principle on the first learning trial recalled more words correctly on the first test trial (Experiment II). (b) 83 who discovered the principle at some point during learning trials 1-8 recalled more words correctly over all 8 test trials, and took fewer trials to reach criterion than Ss who did not discover the principle (Experiment I). (5) In general, 85 who discovered the principle at some point during the task made fewer intrusions in recall over test trials 1-8 than 83 who did not (Experiment I). (4) But there was little or no difference in number of intrusions on the first test trial made by 85 who either did or did not discover the principle during the first learning trial (Experiment II). 108 109 There is a notable agreement between the percent of 85 who discovered the principle in this study and the reported proportion of 55 who discovered that they could transform letter trigrams into words in Underwood and Keppel's study (1964). In both cases, the proportion of 85 who discovered the relevant "built-in" encoding scheme was roughly 50%. Furthermore, in both tasks this discovery lead to enhanced performance. In Underwood and Keppel's study (1964), for example, when 83 were allowed to recall the letters in each trigram in any order, transforming the letters into words resulted in sizeable gains in recall. The conclusions of this study which are listed above also suggest that discovery of the relevant higher-order coding scheme led to superior performance scores across all measures of immediate retention. However, a more detailed analysis of the results of this study suggests that this superiority held only for measures of recall and number of intrusions over all eight test trials of Experiment I, as well as recall on the single trial of Experiment II. Differences in number of intra- and extra-list intrusions made by Ss who either did or did not discover the principle failed to reach statistical significance on the single test trial of Experiment II. A full account of these findings must therefore begin with a discussion of the nature of intra- and extra-list intrusions. Gibson's somewhat controversial discussions of generalization and differentiation (1940) seem most appropriate to this 110 discussion. Gibson maintains that a major necessity of verbal learning is the formation of discriminations among the items to be learned. When a response is formed to a given stimulus, there is a concurrent tendency for all similar stimuli to evoke the same response (stimulus generalization). There is also an increased tendency for the given stimulus to evoke similar responses (response generalization). Thus, prior to the development of discriminations among stimulus items on the list, a response which should be given to Stimulus "a", may be given to Stimulus "b" because of the similarity be- tween "a" and "b". Stimulus generalization would therefore seem to account for intra-list intrusions. Furthermore, response "b", instead of the correct response "a", may be given to a certain stimulus because of a similarity between responses "a" and "b". Thus, response generalization would seem to account for extra-list intrusions.25 Although both intra- and extra-list intrusions are overt responses stemming from a lack of differentiation between appropriate stimuli and apprOpriate responses respectively, inappropriate 2sExtra-list intrusions may also stem from the perceived relations among the terms in each triad. Such relations, or encoding schemes, may suggest certain incorrect responses. An example of this is the response "animal" to a stimulus selected from the triad mammal-dear-bear. This response is likely to occur if the S perceives that all three terms in the triad are animals. Unfortunately, it is impossible to differentiate between this source of errors, and response generalization since "animal" is also highly similar to the term "mammal". Number of misspelled exemplar terms therefore provides the only direct index of the strength of this factor. 111 responses may also occur covertly. Under these conditions, the covert response may merely inhibit the correct response which also tends to occur. Errors in spelling may also result from a lack of differ- entiation. Here the source of the error rests in improper discrimination between the spelling of each term as it ap- pears on the list and the spelling which is suggested by the encoding scheme adopted by the S. If a S notes only that the exemplar terms fit some category, for example, then there will be a strong tendency to spell these terms as exemplars of that category rather than as homonyms of the correct exemplar terms. 85 who discover the one-homonym principle are also faced with the task of discriminating between the term which is spelled as an exemplar and the term which is spelled as a homonym within each triad. This explanation is supported by the results of at least three experimental studies (Underwood, 1949; Underwood and Hughes, 1950; and, Deese, 1959a). Perhaps the strongest evidence was provided by Underwood and Hughes (1950) who found that on a test of one-week retention of nonsense syllable- adjective pairs, the more similar the extra-list errors were to the correct response, the greater the frequency with which these errors occurred. A rather crude analysis which was conducted in this study also lends support to this explanation of extra-list intrusions.25 A simple frequency count was made 28Underwood and Hughes (1950) note the complexities in- volved in an elaborate analysis and recommend that no one 112 of the number of times a given term appeared as an extra-list intrusion in this task. In general, the most frequently occurring terms bore a strong similarity to the correct re- sponse (e.g., "foot" instead of "£eet"). These results are depicted in Table 21 of Appendix C. The above discussion suggests that a critical phase of any verbal associational task is the formation of certain discriminations. Each S must not only learn to discriminate between similar stimuli which appear on the list, but he must also learn to distinguish between the correct response to a given stimulus and similar responses which also tend to be elicited by the stimulus. Prior to the establishment of these discriminations, incorrect responses are apt to appear either covertly, resulting in response competition, or covertly as intra-list and extra-list intrusions, respectively. When viewed in terms of this discussion, Underwood's paradigm of associational learning (1962), appears to be some— what incomplete. Underwood states, The first stage is the response-learning stage during which a S must acquire the response-terms so they are readily available in recall. The second stage is the associative stage during which an associ- ation is formed between the stimulus and the response term. ‘(Underwood §p_§l,, 1962, p. 555). Thus, Underwood gives no attention to discrimination learning. By modifying Underwood's model to include a third, or discrimination stage, it is possible to account for the attempt to analyze extra-list intrusions without somehow structuring the task so that errors of a certain type are apt to occur. This advice was followed by Deese (1959a). 115 findings of this study. The modified model differs from Underwood's paradigm only by the addition of a third stage. Thus, the above quotation adequately describes the first two stages of the revised paradigm. The third stage, however, involves discrimination learning: namely, differentiation between similar stimuli as well as differentiation between correct responses and similar response terms which do not appear on the list. This revised paradigm suggests that 85 who discover the principle in this task may enjoy advantages over those who do not during both the second and third stages of associ- ational learning. During the associational stage, for example, discovery of the principle results in the acquisition of a single coding scheme which may be used to relate the terms in all nine triads. 83 who do not discover the principle, on the other hand, must not only devise none distinct relationships among the terms in each triad, but they must somehow retain these nine coding schemes as well. Thus, 85 who discover the principle approach each early test trial effectively armed with nine associational aids, while Ss who do not discover the principle probably approach each early test trial with less than nine associational aids. This would explain.why Ss who discover the principle recall more words correctly than $5 who do not as early as the first test trial. Ss who discover the principle also enjoy an advantage dur- ing the discrimination stage. Discovery of the principle 114 brings with it the immediate termination of the second, or associational stage. At this point, a S who discovers the principle may devote his full attention to the unique attributes of the terms in each triad in an attempt to master the neces- sary discriminations. Ss who do not discover the principle, on the other hand, must give some attention to determining relations among the terms in those triads for which they have not developed verbal codes. Thus, these Ss are able to give only partial attention to the unique attributes of the terms in each triad. This would explain why Ss who discover the principle make fewer intrusions in recall over all eight test trials. The fact that discrimination learning follows the formation of associations would also explain why the two groups did not differ in number of intrusions on the first test trial. Further support for the three stage model is derived from a study by Spiker (1960) as well as from one further analysis of the data obtained in this study. Spiker provided an ex- perimental group of fifth graders with four sentences by which they could relate four pairs of words. This group was then encouraged to develop their own sentences in a second list consisting of four different pairs of words. The experi— mental group probably proceeded rapidly through the associ- ational stage and were therefore faced primarily with discrimi- nation learning. It is therefore not surprising that Ss in this group made only about one-third as many intra—list 115 intrusions as $5 in the control group who did not receive this training. An analysis of the spelling distortions made by Ss in this study also provides support for the three stage model. The above discussion suggests that errors for 85 who discover the principle should be concentrated in the early test trials, while errors for 85 who did not discover the principle should be more evenly distributed across all eight test trials. The cumulative proportion of spelling distortions was therefore determined for these two groups. Due to the concentration of spelling errors in the first few test trials, the author predicted that the curve for 55 who discovered the principle would be above the corresponding curve for Ss who did not discover the principle. Figure 5 on the following page depicts the cumulative frequency distribution for these two groups for both one- and two-homonym lists. Inspection of this figure reveals that the curves are distributed in the predicted direction. A one- tailed Kolmogorov-Smirnov test was therefore computed to determine whether or not the two distributions were signifi- cantly different (Siegel, 1956). Using the maximum difference of 0.254 on the fourth test trial for one-homonym lists, and the maximum difference of 0.256 on the second test trial for one-homonym lists, the obtained Chi-squares were 9.072 and 19.158 for one-homonym and two-homonym lists, respectively (df = 2). Thus, differences between the two groups are 116 .AH 0:08000mev 00m0oc00m 000 00>0000p 00: p00 00 600 000000 023 00 >0 0008 00000000000 0:0000m0 mo 00000om00m 0>0u005§50 000009 0005 ©000>0000p 033 00 m 5 m m w m N 0 b p p _ n p L L 0.0 00m0oc00m 00>oo .1000 000 000 on; 00 oulb 0 0.309009 4 d \ 1N.o "0000A E>GOEomIo3B .m 000009 0009 5 m m d m p p h h b .m 00sm0m $10113 burtteds go uorqxodoxa eArqunmna 00m0oa00m 00>oo 1000 000 p00 0:3 00 00Q0UC00m p000>0000© 0:3 00 culao 0.0 4IIAV\p. 0.0 \ - 0.0 rm.o um.o -00 "0000A Smcofioml0co .0 810113 Burtteds go nomnlodozd eAtqunan 117 statistically significant for one-homonym lists (P < 0.05), as well as for two-homonym lists (P < 0.01). This finding follows logically from the three-stage model of associational learning, but could not be predicted from Underwood's two stage model. It therefore appears that the three stage model more adequately accounts for the findings of this study. However, statements regarding the generality of the paradigm await further investigation. II. Configural Perception and the Encoding Process The effect of configural perception on the development of higher-order coding schemes was examined in tests of Hypotheses Five and Six. These tests confirm the following general con- clusions: (1) If perceptual cues do affect the likelihood that a given S will discover the principle, then this effect is generally limited to the early test trials: (a) Differences in the per cent of Ss who discovered the principle on the first learning trial ap- proached significance for one- and two-homonym lists as well as for capitalization or non- capitalization of the categorical terms (0.05 < P < 0.10). (b) But only the difference between capitalization and non-capitalization remained at this level by the completion of the task. (c) Order of presentation within each triad had little or no effect on discovery of the principle at any point during the learning task. The effect of perceptual cues on short-term retention was examined in tests of Hypotheses Seven through Ten. These tests confirm the following general conclusions: 118 (2) The short-term retention of 83 presented one-homonym lists is clearly superior to that 0f.Ss presented two-homonym lists. Differences between these two groups were statistically significant on every measure of short—term retention (P < 0.01). (5) No conclusive statement may be made in regard to the effect of capitalizing or not capitalizing the cate— gorical terms on short-term retention. It would ap- pear that if this variable does have an impact on performance, then this effect is almost totally limited to one-homonym lists. (a) Differences between the two conditions on this variable failed to reach a statistically signifi- cant level on any measure of short-term retention. (b) But the interaction between number of homonyms and capitalization was significant when total number of intrusions served as the dependent variable (P < 0.01) and approached significante on nearly every other dependent variable. The distribution of means in each of these interactions suggests that differences between the two condi- tions were sizeable on one-homonym lists, but extremely small on two-homonym lists. (4) Order of presentation within each triad has little or no effect on short-term retention. None of the dif— ferences between the two conditions on this variable even approached statistical significance. The most consistent differences in performance occurred for one and two-homonym lists. These two lists vary according to the number of terms which must be transformed into their homonyms before thé underlying principle is apparent. They might thus be said to vary in the degree of embeddedness of the underlying principle. Since the two groups did vary in performance, the obvious conclusion would be to say that this difference arose from the difference in the perceptual attri- butes of the two lists. Unfortunately, however, the two lists have two other systematic differences which may also have contributed to the obtained difference in performance-- 119 frequency of occurrence of the terms, and degree of estab— lished association among the terms in each triad. First, 18 of the 27 terms on the two lists are identical; namely, the categorical and misspelled exemplar terms. But the other nine terms are unique to each list. As noted in Chapter III, the unique terms on one-homonym lists occur somewhat more frequently in printed texts than the correspond- ing terms on two-homonym lists. Since frequency of occurrence is related to recall, this difference would tend to favor performance on one-homonym lists. But if this factor has a significant bearing on the obtained difference between one— and two-homonym lists, then there should be a sizeable dif- ference in number of unique terms recalled and little or no difference in the recall of the 18 identical terms, i.e., there should be a significant interaction between number of homonyms and type of response. But when this interaction was examined over the first three test trials, the resulting F- ratio equalled 0.17 (see Table 12 in Chapter IV). This finding suggests that the difference in recall of identical terms between Ss presented one- and two-homonym lists was approximately equal to the corresponding difference in recall of unique terms. Therefore, the obtained difference between one- and two-homonym lists cannot be solely attributed to the difference in frequency of occurrence of the terms which appeared in each. 120 Second, the degree of established association among the terms in each triad also seems to favor performance on one- homonym lists. The association between the categorical and exemplar terms on one-homonym lists is well established prior to the time the S enters the task (e.g., mammal-bear). But the corresponding association on two-homonym lists is extreme- ly arbitrary (e.g., mammal-bare). On the other hand, differences in degree of established association are minimal for the other two types of associ- ations. The exemplar-misspelled exemplar associations are arbitrary for both types of lists (e.g., bear-dear and bare- dear) and the categorical-misspelled exemplar associations are identical in both lists (e.g., mammal-dear). Therefore, if the difference in degree of established association plays a critical role in the obtained difference in performance on one- and two-homonym lists, there should be a significant interaction between number of homonyms and type of association. Table 16 on the following page depicts the means in this interaction. Examination of this Table reveals that differ- ences were actually smallest for the categorical-exemplar associations. One must therefore conclude that the difference in degree of established association between the terms in one- and two-homonym lists cannot fully account for the obtained difference in performance on these two lists. Since neither of these variables can fully account for the obtained difference in performance, it would seem 121 Table 16. The mean number of words correctly recalled over the first three test trials (total possible 0 18) for each condition in the interaction between number of homonyms and type of association. F T Categorical- Exemplar- misspelled Categorical- misspelled exemplar exemplar exemplar ,One:Homonym Lists (N = 113) 12.50 11.85 11.84 Two—Homonym Lists (N = 105) 10.52 10.47 9.66 Difference in Means for the two lists 2.18 1.57 2.18 reasonable to attribute this difference to the relative degree of embeddedness of the underlying principle on one- and two- homonym lists. As described earlier in this section, the fact that the principle underlying one-homonym lists is apa parent following the transformation of a single term in each triad into its homonym suggests that this principle should be discovered fairly rapidly. The difference in the proportion of Ss who discovered the principle during the first learning trial which was very nearly significant offers some support for this assertion. This difference should, in turn, give rise to the obtained difference in performance on one- and two-homonym lists. 122 Before accepting this explanation, however, one further finding must be considered. According to the above logic, the difference in performance on one- and two-homonym lists should disappear if the per cent of 85 who discovered the principle is equated for both groups. Yet in an earlier analysis when this occurred, the mean number of words recalled by Ss presented one-homonym lists remained significantly larger than the corresponding mean for two-homonym lists (P < 0.01).27 Therefore, the obtained difference in perform- ance on one- and two-homonym lists does not stem solely from the difference in proportion of Ss who discovered the princi- ple during the initial learning trials. In short, none of the three variables which have thus far been considered can fully account for the decisive superiority of performance by Ss presented one-homonym lists. The most reasonable conclusion therefore appears to be that this superiority stemmed from an additive combination of all three factors. However, an equally strong possibility is that this difference stemmed from an interaction between two or more of the factors. Consider the interaction between degree of embeddedness and degree of established association. The difference in degree of embeddedness of the princi- ple underlying one- and two-homonym lists stems from the fact that two terms must be transformed into homonyms before the principle underlying two-homonym lists becomes apparent, while 27$ee Table 12 in Chapter IV. 125 only one term must be transformed on one-homonym lists. The principle underlying one-homonym lists is therefore more obvious simply because fewer terms have to be transformed in the derivation of this coding scheme. In addition to this difference in number of terms which must be transformed, a second factor may also contribute to the difference in degree of embeddedness of the underlying relationship; namely, degree of established association among the terms in each triad. The existence of a previously es- tablished association among two of the terms in each triad on one-homonym lists may have a decisive impact on perform- ance because it contains a strong suggestion that the third term should be transformed into its homonym. Suppose, for example, that a S notes that gip§_and beverage both involve drinking, or that mammals and bears are both animals. Conscious recognition of this relationship may also suggest that a simple transformation of the third term makes it an exemplar of the category as well. If pig; is changed to pggp, for example, it is also a drink. Thus, one-homonym lists contain an implicit hint that a transformation of terms should be made; namely, the established association among two of the terms in each triad. Two-homonym lists, on the other hand, lack this suggestion that any transformations should occur. Ss may therefore con- tinue to react to each term as it is written on the list over a longer interval of time. If one reacts to whine as a form 124 of crying, for example, then it is difficult to see how this term might be included in a common category with bever ge. Since there is no hint of this categorical relationship, there is also no reason to perceive that pip; should be trans- formed to pppp. In the absence of any suggestion of a cate- gorical relationship, it is even possible that many 85 in this group began to look for syntactical relationships such as "He whined in his beverage as the pi§£_passed by." An early commitment to this strategy would clearly interfere with the likelihood that a S would discover the principle. In short, there are two interacting factors which may contribute to the ease with which a given S may derive the underlying principle-~number of homonyms which must be trans- formed and the extent to which the utility of making these transformations is suggested by previously established associ— ations among the terms. Ss presented one-homonym lists are at an advantage over Ss presented two-homonym lists on both of these factors. Perhaps even more important is the fact that these two factors probably also contribute to the ease with which Ss may derive any first-order coding schemes among the three terms in each triad. Ss presented two-homonym lists are handicapped by the fact that unless two terms in each triad are trans- formed into their homonyms, the derivation of any relationship, other than a syntactical relationship, is extremely arbitrary. And, as is evident from the above discussion, there is no 125 implicit hint that such transformations should be made. 35 presented one-homonym lists, on the other hand, are faced with the task of incorporating only one arbitrary term into each first-order coding scheme. Furthermore, the existence of an established association between two of the terms sug- gests how this incorporation might be easily accomplished; namely, transforming the term into its homonym so that it also becomes an exemplar of some category. One might therefore expect that among those 85 who do not discover the principle, 85 presented one-homonym lists will have established first order coding schemes more rapidly than the corresponding Ss presented two-homonym lists. The results of several studies, such as Martin §p_§l. (1965a), suggest that this early advantage should, in turn, contribute to the obtained difference in performance between these two groups of Ss. In short, the interaction between number of terms which must be transformed and degree of previously established associations among the terms in each triad would seem to ac- count for the obtained difference in performance between 83 presented one- and two-homonym lists. This follows from the fact that differences along these two variables give rise to differences in the prOportion of Ss who discover the principle early in the learning process as well as to differences in the Speed with which first—order coding schemes are derived by those Ss who do not discover the principle. 126 Perhaps more important, these same two variables may also account for the significant interaction between number of homonyms and capitalization of the categorical terms. The two easily related terms in each triad on one-homonym lists suggest not only that a transformation of one term in each triad should be made, but they also suggest the nature of the first order relations which should be established; namely, categorical relationships. It is possible that the differential attention drawn to the categorical terms by capitalization may suggest the specificity of this relationship. Suppose, for example, that a S notes that bevegaqe and wipp are both drinks, and that when pig; is changed to pppp, it is also a drink. Capitalizing a categorical term may cause the S to take another look at this term. In the process he might note that ppp£_and wig; are both beverages. If a closer analysis of the other triads results in a similar discovery of the more specific category, then the S is well on his way to discovery of the relevant principle. In the absence of capitalization, on the other hand, the S might remain content with the derivation of first-order coding "28 schemes such as "They are all drinks. In short, capitalie zation of the categorical terms on one-homonym lists may have 281t is clear that if a S derived an extra-list category for each triad, such as "drinks," he would have derived a higher- order coding scheme. The likelihood of this occurring is ex- tremely small, however, due to the lack of an obvious extra-list category for such triads as "weather-son-rain." Therefore, in all probability, those few Ss who derived a first-order relation for each triad on the list used some extra-list and some intra-list categories, thus precluding the use of the same coding scheme for each triad. 127 frequently led to the discovery of the principle. Turning to two—homonym lists, the absence of an estab— lished association among two of the terms in each triad may have eliminated the effect of capitalizing the categorical terms. First, the absence of this easily recognized relation- ship reduces the likelihood that Ss will initially attempt to derive categorical relationships. Should a S attempt to de- rive syntactical relationships, for example, then differential attention to the categorical terms would have absolutely no impact on performance. Second, unless the S perceives that two of the terms in each triad should be transformed into their homonyms, the categorical terms are no more suggestive of any first-order coding schemes than either of the other two terms. Since two-homonym lists lack any implicit hints that such transformations should be made, the impact of capi— talizing the categorical terms may be minimal during the initial test trials. In short, this explanation suggests that capitalizing the categorical term should enhance the probability of discovering the principle on one-homonym lists, while registering little or no impact on the likelihood of discovering the two-homonym principle. Differential discovery of the principle, should in turn, yield the obtained differences in performance. 128 Although the explanation which seems to parsimoniously account for the major findings of this phase of the research is post hoc in nature, it is nevertheless highly suggestive of further research. Such research might take the form of free-associations to the terms in each triad, presented either individually or in pairs, in an attempt to determine which terms elicit the greatest number of relevant categorical re- sponses. These critical terms might then be perceptually isolated using either differences in the size of print (simi- larity) or differences in the physical distance between the terms in each triad (contiguity). If the above explanation is sound, the resulting differential attention to these terms should have an impact on the derivation of first-order coding schemes as well as discovery of the principle which, in turn, will enhance performance. The primary improvement of this study over that described in this dissertation would rest in the fact that the critical terms would be empirically determined, while in this study, it was assumed that the categorical terms were more highly suggestive of the principle than either of the other two. And, as described above, the results tend to raise a serious question regarding the validity of this assumption. III. Higher-Order Coding Schemes and Long-Term Retention The effect of the derivation of higher-order coding schemes on long—term retention was examined in tests of Hypotheses 129 Eleven through Fourteen. These tests support the following two conclusions: (1) Those 83 who discovered the principle during acquisi- tion did not perform at higher levels on the test of long-term retention than Ss who did not discover the principle. Differences between 85 who either did or did not discover the principle did not approach sig— nificance on any measure of long-term retention. (2) None of the three perceptual variables had the pre- dicted effect on long-term retention. Not only did the difference between the twaconditions on each variable fail to reach a statistically significant level, but in several tests the two means were dis- tributed in the Opposite direction from that which was predicted. These two findings may probably be combined into a single conclusion; namely, that the vast majority of 85 who discovered the principle during acquisition were not able to utilize this principle on the test for long-term retention. Two further findings also point to this conclusion. First, Ss presented two-homonyms lists made more spelling distortions than 83 presented one-homonym lists despite the fact that the princi- ple underlying the former lists contained a derivative rule for spelling, while the corresponding spelling rule for one- homonym lists was correlative in nature. Second, an indirect measure of retention of the principle for 85 presented two- homonym lists revealed that at least 72.5% of the 83 who dis- covered the two-homonym principle during acquisition did not utilize this principle on the test for long-term retention. (See Table 25 in Appendix C for an explanation of this finding.) Before accepting this explanation, however, two alterna- tives should be considered; namely, that the differences were 150 not significant due to the selectiveness of the sample used in tests of long-term retention, or to the greater opportunity for overlearning among those 83 who did not discover the principle. First, comparisons of short-term retention were based on every S in the sample, while comparisons of long-term retention were based on only those Ss who reached criterion. It is there- fore possible that differences in short-term retention did not exist among Ss in this more select sample. In lieu of this possibility, the entire analysis of short-term retention was replicated for those Ss who reached criterion. The results of this series of t-tests are shown in Table 22 in Appendix C. Examination of this Table reveals that whereas differences were smaller among Ss in this select sample, they were never- theless significant on all but two measures: number of mis- spelled exemplar terms and number of improper plurals. Since differences on the latter measure were not significant for the total sample, this explanation is limited to measures of the number of spelling distortions. A second alternative is that differences in retention were minimized by differences in the extent of overlearning during acquisition. Ss who did not discover the principle took more trials to reach criterion than Ss who discovered the principle, yet both groups were near the ceilinglevel of performance as early as the third test trial. Thus, it is likely that the extent of overlearning was greater for Ss who did not discover 151 the principle, and it is possible that this, in turn, mini- mized differences in long-term retention. But if the extent of overlearning exerts a strong influence on retention, then the correlation between number of trials to criterion and number of words correctly recalled on the test for long-term retention should at least approach a positive direction. But when this correlation was computed, it was not only negative, but significantly different from zero (P < 0.01). Correlations between number of trials to criterion and number of words correctly recalled on the test for long-term retention were -0.292 and -0.590 for 94 Ss in the one-homonym group and 95 Ss in the two-homonym group, respectively. The correlation for the two groups combined was -0.471. These correlations are nearly as high as the correlations betWeen number of words correctly recalled during acquisition, and number of words correctly recalled on the test of long-term retention. The combined correlation in this case equalled 0.558. It would therefore appear that the extent of overlearning did not have a significant impact on the obtained results. Since neither alternative accounts for the findings, the most plausible explanation would seem to be that the vast majority of 85 did not utilize the principle on the test for long-term retention. If this is true, then according to Postman, one would expect no differences in recall between 85 who either did or did not discover the principle since, as this author states: 152 Recoding can enhance retention only to the extent that the recoding symbols are recalled and accurate decoding follows. (Postman, 1965, p. 46.) Since this finding came as a complete surprise, the de— sign of this study does not permit the author to choose be- tween two potential explanations of why so many 55 failed to utilize the principle on the test for long-term retention. First, it is possible that $5 simply failed to see the utility of reinstating the principle during recall. This possibility may follow from the function of encoding schemes during acquisition. Woodworth and Schlosberg, for example, maintain that the function of extraneous associations (encoding schemes) is to, "hold certain items together until a direct association has been established between them" (Woodworth and Schlosberg, 1958, p. 54). Evidence from studies by Reid (1958), O'Brien (1921), and Barnes and Underwood (1959) suggests that once these direct associations are formed, the relevant mediator or coding scheme tends to drop out. This author conjectured, however, that when the direct associations were weakened over time, the S would again interject the coding scheme between a given stimulus term and the two relevant responses. It is possible, however, that most of the Ss simply did not use this strategy during recall. A second, and perhaps more reasonable, explanation is that 85 failed to use the principle on the test of retention; because they could not remember it. Since this possibility has some bearing on the dominant theories of retention, a cursory examination of each is in order. Although this study 155 was not designed as a critical test of any theory of retention, the expressed predictions regarding long-term retention were consistent with two theories: Gestalt Theory and Ausubel's Theory of Obliterative Subsumption. According to Gestalt Theory, acquisition results in the establishment of memory traces which are isomorphically equiva- lent to the incoming information. Forgetting results from changes which occur in these traces over time, these changes, in turn, result from two primary mechanisms--assimilation and autonomous disintegration. Regardless of what the organism _does or learns, a memory trace will undergo a type of spon- taneous decay or autonomous disintegration with resulting decrements in retention. Changes in a memory trace also occur as a result of complex interactions between the established traces and incoming traces from subsequent learning, i.e., assimilation. This interaction tends to favor incoming traces, since they are relatively more stable (Ausubel,.1964). The nature of both assimilation and autonomous disintegration gives rise to the fundamental premise of Gestalt Theory: namely, that structural or organizational traces are more stable than isolated traces. Turning to this study, if one assumes that discovery and utilization of the principle results in a structural trace, while the derivation of nine independent coding schemes re— sults in nine isolated traces, then it follows from Gestalt Theory that long-term retention should be superior for those Ss who discovered the principle. Since the major focus of 154 Gestalt Theory is on the establishment of stable traces dur- ing encoding, this prediction should hold regardless of whether of not the S utilizes the principle during recall. According to Katona's (1940) description of structural traces, the above assumption seems feasible. Katona states: The individual traces may be translated into every- day language as the knowledge of more-or-less uncon- nected facts, while the knowledge of laws and principles, of meaning and significance, of the setting and general forms, is the equivalent of structural traces. (Katona, 1940, p. 206.) Therefore, unless Katona's statement has been misinterpreted by this author, it would appear that the results of this study clearly do not support the Gestalt theory of retention. In short, discovery of the principle, in and of itself, does not insure improved long-term retention as the Gestalt theory of retention would seem to imply. The second theory, which is consistent with the predic- tions expressed in this study is Ausubel's theory of obliter- ative subsumption (Ausubel and Blake, 1958; and Ausubel, . 1965). Ausubel maintains that as new information enters the cognitive field of the learner, it interacts with and is apprOpriately subsumed by a relevant and more inclusive con- ceptual cluster which he terms a subsumer. As a result, both the new information and the subsumer are somewhat modified. However, an established subsumer ordinarily undergoes consider- ably less modification than the subsumed element. Forgetting, or Obliterative subsumption, may be expressed in the following equation where "A'" represents the modified 155 conceptual cluster or subsumer; "a‘" represents the modified elements of incoming information; and "A'a'" represents the product of the interaction between "a'" and "A'": Ala! ? Al + a! As suggested by this equation, there is a temporary equilibrium among the three components. Therefore, soon after acquisition, the individual is able to regenerate both the specific elements and their relevant subsumers (as represented on the right-hand side of the equation). Over time, however, the temporary equilibrium begins to shift toward the left. The interaction product represented in this portion of the equation suggests that the individual is no longer able to disassociate the incoming elements from their relevant subsumer. In other words, because it is more economical to retain a single inclusive concept than to remember a large battery of specific items, the individual is ultimately capable of re- calling only the more generalized interaction product. More— over, because the subsumer usually undergoes less modification than the subsumed elements, this product usually takes on the dominant characteristics of the subsumer. A fundamental premise of this theory is, therefore, highly similar to Gestalt Theory: namely, that hierarchical or organizational cognitive content (subsumers) is usually much more stabile than specific details of incoming information. 156 It is unfortunate from this author's point of view that Ausubel has relied on somewhat vague descriptions, rather than a specific definition in his development of the concept of a subsumer. Inferences made by this author suggest that this concept is roughly equivalent to Goss's (1961) notion of a "conceptual scheme'I and that it represents the cognitive counterpart to Bartlett's (1952) "schemata". But even this does not help, since these terms are also described in a highly ambiguous fashion. The author is, therefore, forced to somewhat equivocally assume that the principles underlying one- and two-homonym lists represent appropriate subsumers. However, Ausubel's reference to "a recently learned abstract concept" in describing subsumers provides some justification for this assumption (Ausubel, 1965, p. 56). With this assumption in mind, Ausubel's theory seems to imply that recall of the principle should persist over a fairly long interval of time, while recall of details such as which term to spell as a homonym should dissipate fairly rapidly. The fact that 85 who discovered the principle dur- ing acquisition failed to recall more words correctly or make fewer intrusions in recall than those who did not seems to raise some question regarding the validity of Ausubel's theory. The further finding that 85 who discovered the two-homonym principle made fewer spelling errors than 83 who discovered the one-homonym principle poses a more central challenge to advocates of Ausubel's theory. 157 It should be evident, however, that the design of this study did not represent a direct test of either Gestalt or Ausubel's theory. Several alternatives are therefore avail- able to proponents of these two theories. To cite only one example, both might point to the highly artificial nature of the principle by which each triad was constructed. In doing so, Ausubel might argue that the principle does not represent an appropriate subsumer, and Gestalt psychologists might argue that the principle does not give rise to a stable trace. The importance of this cursory review is, therefore, not to provide a coup de grace of the two relevant theories, but rather to show that neither theory, no matter how extensively modified, would predict that the principle itself will be rapidly forgotten. Since neither Gestalt nor Ausubel's theory yields a simple explanation of why Ss may have forgotten the principle, about the only remaining alternative is interference theory. Proponents of this theory maintain that retention is nothing more than a response produced by a stimulus (Osgood, 1955). Forgetting therefore results when stimuli lose their capacity to evoke previously associated responses. In simple terms, this loss results from the fact that one learns similar associations both prior to the critical task on which reten- tion is determined (proactive interference) and subsequent to this task (retroactive interference). The negative impact of this additional learning stems from the introduction of 158 competing responses to the critical set of stimuli. In short, when the relevant set of stimuli is re-presented, the S is forced to discriminate between the appropriate set of responses and the irrelevant set of responses which were acquired during the additional learning activity. Due to stimulus generalization, the two sets of stimuli (critical and additional learning task) need not be identical for interference to occur. But the greater the similarity, the greater the extent of generalization between the two sets of stimuli, and therefore the greater the interference. Finally, retroactive interference may also stem partially from the fact that the critical stimulus—response associations have undergone a certain amount of unlearning or extinction during the interpolated learning activity (McGeoch and Under- wood, 1945 and Barnes and Underwood, 1959). Although consider— ation of this unlearning factor would seem to suggest that retroactive interference will result in greater decreases in retention than proactive interference, recent evidence points to the opposite conclusion: namely, that proactive interfer- ence is a more critical factor in forgetting (Underwood, 1957). Evidence favoring interference theory has been acquired primarily from tasks involving nonsense syllables or other stimulus materials where relations between stimuli are highly arbitrary and difficult to establish. A controversy has therefore arisen regarding the generalization of this theory to connected discourse or "meaningful learning." Some evidence, 159 for example, suggests that the extent of retroactive inhibi- tion is minimal for this type of material (Hall, 1955: Ausubel, Robbins and Blake, 1957; and Ausubel and Blake, 1958). But other authors have provided evidence suggesting that re- troactive inhibition does occur for connected discourse (McGeoch and McKinney, 1954, and Slamecka, 1960a). Postman (1965) seems to advocate a compromise between these two ex— tremes; he maintains that the basic assumptions of interference theory remain tenable for most forms of learning. But he also suggests that these assumptions must be supplemented by cer- tain "principles of conservation" such as recoding (encoding) which, under some conditions, have been shown to systematically reduce the amount of forgetting. By carrying Postman's analysis one step further, it may be possible to account for all types of forgetting within an interference framework without citing exceptions to the basic assumptions. This parsimony may be accomplished by shifting from the current concern with similarities in the systematic properties of the stimulus and response terms to a concern with similarities between the encoding schemes adopted during the critical and additional learning tasks. In short, forget- ting may result from an interference between highly similar coding schemes which are adopted during these two tasks. There is some evidence to support this assertion. Martin's (1965a) research, for example, suggests that strategies (encoding schemes) adopted during paired-associate tasks 140 tend to involve systematic properties of the stimulus and response terms, e.g., the same combination of letters in the two terms. As a result, one might mistakenly assume that interference results from similarities in these proper— ties themselves when it actually results from the fact that similar properties give rise to highly similar coding schemes. Turning to tasks in which relations between the stimulus and response terms are not arbitrary, but rather involve meaning- ful and/or well-established relations, the evidence is less equivocal. Slameka (1959 and 1960b), for example, has'shown that the extent of retroactive inhibition is a direct func- tion of the similarity between the themes or topics of criti- cal and interpolated passages of connected discourse. Since Bartlett (1952) has shown that retention of connected or meaningful material is based on translations of this material into basic themes, it follows that the critical element in these studies is interference among the adopted coding schemes. Finally, two studies have shown that both proactive and retro- active interference occurs when two sets of materials are based on closely related principles (Postman, 1954; and Entwisle and Huggins, 1964).29 Turning to this study, a high degree of similarity does exist between the principle (coding scheme adopted by those Ss who discovered the principle) and the well-established 23The results of Postman's (1954) study were not statis- tically significant. 141 method of classifying those terms which appeared on the list. The following statement by Postman therefore seems relevant. To the extent that the prescribed associations are consistent with pre-existing language habits, positive transfer, and facilitation at recall are to be expected. When there is competition between the required response sequences and prior language habits, there should be negative transfer and interference at recall. (Postman, 1965, p. 40.) - It is therefore possible that the similarity between the established means of classifying the terms on each list and the general nature of the principle gave rise to a high degree of proactive interference between these two coding schemes. This interference would in turn account for the fact that many 83 forgot the principle. In summary, there are two potential explanations for the conclusion that the vast majority of 85 who discovered the principle did not utilize this principle on the test for long- term retention. First, it is possible that Ss simply did not adopt this strategy during recall, despite the fact that they could have recalled the principle. Studies showing that mediators tend to drop out during acquisition provides some evidence for this position. Second, Ss may not have utilized the principle during recall, because they could not recall' the principle. As shown above, a simple modification of inter- ference theory could account for this alternative. Unfortunately, the design of this study does not provide conclusive evidence for either alternative. However, further research could readily provide this evidence. Suppose, for 142 example, that immediately prior to the recall trial Ss are instructed to use whatever relations they derived during acquisition as aids in recall. If differences fail to occur under these conditions, one may conclude that many Ss have forgotten the principle. But if differences do occur, then the relevant conclusion is that 85 did not realize the utility of reinstating the principle during the test for retention. IV. Implications for Education Although this investigation would be classified as "pure research" according to Hilgard's scale (1964), the major find- ings may nevertheless raise certain questions regarding edu- cational practice. The first finding which should be considered is that those 55 who derived higher-order coding schemes in this task per- formed at a higher level during acquisition than those who did not. This result.seems to suggest that the coding activi- ties of the learner will have a significant impact on the acquisition phase of some learning tasks. If this generali- zations is valid, then two implications follow. First, a teacher may better understand the source of differences among learners if he determines what coding schemes have been adopted by each. Second, it may be possible to increase learning efficiency in certain tasks through instruction aimed at im- proving the coding skills of students. The first of these implications may pose a problem for teachers; namely, how might they efficiently determine what 145 coding schemes have been adopted by a given student? Fortunately, this investigation suggests a solution. Due to frequent errors in completing the L.P.F.Q., two independent methods were used to identify those 85 who had deriVed a higher—order coding scheme: verbal reports and performance tasks involving transfer of the principle. As described in Chapter III, there were no differences in performance among Ss identified by either of these methods. Thus, verbal re- ports seemed to be as effective as transfer of training tasks in identifying Ss who had derived a higher-order coding scheme. This finding clearly implies that teachers need not be reluctant to ask students what process they have used to master a given task.30 This method of identifying the coding schemes adopted by a given student is not only more direct than interences of process based on transfer of'training, but it also has a wider range of application. Thus, it would seem to be to the teacher's advantage to ask each student how he has gone about mastering a given task. This teaching‘ strategy may, in turn, yield better understanding of the source of differences in acquisition among students. 80It should perhaps be noted that this suggestion directly counters the position which a majority of psychologists take; namely, that any form of introspection should be avoided in attempts to determine the process by which a S has mastered a given taSk. But, no matter how reasonable any argument posed by the anti-introspection position may seem, it fails to counter the consistent finding that verbal reports account for a significant proportion of the variance in performance among Ss (see Chapter II). The results of this investigation further suggest that in certain situations, verbal reports are as effective in identifying process variables as methods which avoid the use of verbal reports, i.e., inferences based on transfer of training. 144 The second implication suggests that learning efficiency may be improved in certain tasks through the use of instruc- tion aimed at improving the coding skills of the learner. It should be noted, however, that the results of this investi- gation provide logical, rather than empirical support for this assertion. Unfortunately, there is no clear empirical evidence, either pro or con, which is relevant to this impli- cation. A definite need for further research therefore exists. A corollary implication is that both teachers and students should receive some instruction regarding the limitations of human memory. Such training may suggest to teachers that it is unreasonable to expect students to "know everything." This instruction may also show students that there is a need to efficiently condense whatever they are attempting to re- member. The provision of this training might therefore prove to be a valuable antecedent to training in coding skills. The second major finding of this investigation which should be considered is that perceptual cues had little or no effect on the likelihood that a given S would discover the principle. Even the presentation of one- and two-homonym lists where one "built-in" relationship seemed far more obvious than the other had virtually no effect on the per cent of 83 who ultimately discovered the principle. In a general sense, this finding seems to suggest that it may be virtually impos- sible to exert control on the cognitive activities of the learner through variations in the manner of presenting a given 145 set of stimulus materials. Specifically, it implies that at- tempts to perceptually isolate those elements which are deemed most critical to a given cognitive process will have little or no effect on the likelihood that a given S will adopt that process during the learning activity. Suppose, for example, that a teacher capitalizes those letter combinations within each word which are known to gener— ate the most errors in spelling. The results of this investi- gation suggest that the perceptual isolation of these "critical elements" will have little impact on the process by which the student will learn to spell. It is therefore not surprising that this teaching strategy has been found to be no more effective than the presentation of Spelling lists in which all letters are written in small print (TenBrink, 1966). As a second example, suppose that a teacher capitalizes all prefixes and suffixes in an attempt to increase the likeli- hood that a student will use "intra-word context cues" in pronouncing unfamiliar words.31 The above implication suggests that this perceptual isolation will have little effect on the likelihood that a given S will use these cues as an aid in pronunciation. Capitalizing prefixes and suffices will there- fore have little effect on a student's ability to pronounce unfamiliar words. alCarroll (1964) discusses a number of cues which a reader might adopt in his attempts to pronounce an unfamiliar word. 146 These two examples are consistent with the assertion that attempts to perceptually isolate critical elements will prob- ably not have a significant impact on the coding activities of the learner. However, further research must be undertaken to determine the validity and/or legitimate range of this generalization. It is possible, for example, that the effect of perceptually isolating certain elements may depend upon the particular coding scheme which one is attempting to facili- tate, or upon the particular elements which have been isolated. The final result which should be considered is that the long-term retention of 83 who derived higher-order coding schemes during acquisition did not exceed the corresponding performance of Ss who initially derived a set of first-order coding schemes. This finding seems to imply that during tests of long-term retention, a majority of 83 will not be able to utilize those encoding schemes which they have derived during acquisition. Admittedly, however, this may be a gross overgenerali- zation. The higher—order coding scheme (principle) which some Ss derived during this investigation was not only highly artificial, but it was also appropriate to only one labora- tory task. Thus, Ss may have Seen little advantage in attempt— ing to remember it. Despite this limitation, this result poses at least two questions which should be answered by empirical research. First, is the facilitating effect of coding schemes restricted to acquisition or short-term retention in all learning tasks? 147 If not, is there any relation between the specificity of a given coding scheme and the likelihood that a S will be able to remember that scheme over long intervals of time? Second, are higher forms of learning such as principles subject to the same basic laws of interference as lower forms such as isolated facts? Answers to both of these questions have clear implications for education. V. Conclusions The results of this investigation provide partial support for at least six major conclusions. The author is convinced that future research will confirm each of these generaliza— tions. (1) When presented with an associational task, some, but not all, 33 will look for general relationships which hold for more than one pair (or triad) of stimulus and response terms, i.e., will attempt to derive higher—order coding schemes. Therefore, investigators who focus their attention on relationships or strategies which are derived for each pair of terms and who ignore the potential impact of inter- relationships among these strategies, may be overlooking an important source of inter-individual variance. (2) In those associational tasks where it is possible to derive a single higher-order coding scheme, Ss who discover 148 this scheme will not only form the required associations more rapidly than those who do not, but they will also make fewer intrusions in short-term recall. (5) Associational learning should be viewed as a three— stage process--response learning, associational learning, and discrimination learning. The addition of discrimination learning to Underwood's (1962) two-stage model stems from an analysis of the intrusions in recall which occurred in this task. This analysis suggests that 85 must learn to differen— tiate between relevant and irrelevant stimuli, as well as between relevant and irrelevant responses. Failure to form these discriminations will result in either covert-competing responses or overt responses in the form of intra-list and extra—list intrusions. (4) Although the results of this study do not provide conclusive evidence regarding the role of perception in the formation of higher-order coding schemes, the methodo- logical technique which was employed does show promise. Future investigations which utilize lists with "built-in" relationships between the stimulus and response terms, to- gether with the perceptual isolation of those terms (or pairs of terms) which are most suggestive of this relationship should provide more affirmative evidence regarding this rather com—~ plex topic. 149 (5) Even in those associational tasks where it is pos- .sible to derive a single higher-order coding scheme, the discovery and utilization of this scheme during acquisition does not insure improved long-term retention. Rather, this superiority is contingent upon the recall and utilization of the relevant coding scheme during the test for long-term retention. (6) Retention of the relevant coding scheme may, in turn, be dependent upon whether or not an individual has derived similar coding schemes in learning activities which occur either prior to or subsequent to the relevant task. Inter- ference among similar coding schemes may not only account for the forgetting of a given coding scheme, but it may also account for forgetting generally. Although this conclusion must be regarded as tentative pending further research, the author is convinced that similarities among coding schemes represent a more critical source of interference than simi- larities among the formal properties of the stimulus and response terms in the relevant and irrelevant learning activi- ties. BIBLIOGRAPHY Articles and Periodicals Aborn, Murray and Rubenstein, Herbert. "Information Theory and Immediate Recall." Journal of Experimental Psy- chology 44: 260-266; October, 1952. Anderson, G. Lester. "Quantitative Thinking as Developed Under Connectionist and Field Theories of Learning." Lparninq Theopy inpSchool Situations, E. J. Swenson, (ed.) Minneapolis: University of Minnesota Press, 1949. Asche, S. E. "Forming Impressions of Personality." Journal of Abnormal and Social Psychology 41: 258-290: 1946. Astin, Helen S. and Ross, Sherman. "Learning of Numbers in Differentially Shaped and Positioned Matrices." American Journal of Psychology 71: 764-765; December, Ausubel, D. P., Robbins, L. and Blake, E., Jr. "Retroactive Inhibition and Facilitation in the Learning of School Materials." Journal of Educational Psychology 48: 554- 545; October, 1951. Ausubel, David P. and Blake, Elias, Jr. "Proactive Inhibition in the Forgetting of Meaningful School Material." Journal of Educational Research 52: 145-149: December, 1958. Ausubel, David P. "The Use of Advance Organizers in Learning and Retention of Meaningful Verbal Material." Journal of Educational Psychology 51: 267-272; August 1960. Baddeley, A. D. "Stimulus-Response Compatibility in the Paired-Associate Learning of Nonsense Syllables." Nature 191:.1527-1528: September, 1961. BarClay, Allan. "Objective Mediators in Paired-Associate Learning." American Journal of Psychology 74: 575-585; September, 1961. .150 151 Barnes, J. M. and Underwood, B. J. "Fate of First-List Associ- ations in Transfer Theory." Journal of Experimental Psychology 58: 977105; August, 1959. Battig, W. F. "Evidence for Coding Processes in 'Rote' Paired-Associate Learning.“ Journal of Verbal Learnigg and Verbal Behavior 5: 117-181: April, 1966. Belbin, E. "The Influence of Interpolated Recall Upon Recog- nition." Quarterly Journal of Experimental Psychology 2: 165-169: Part 4, 1950. Bolles, R. C. "The Effect of Altering the Middle of the List during Serial Learning.“ American Journal of Psychology 72: 577-580: December, 1959. Bousfield, W. A. and Cohen, B. H. "The Effects of Reinforce— ment on the Occurrence of Clustering in the Recall of Randomly Arranged Associates." Journal of Psychology 56: 67-81: July, 1955. Bousfield, W. A. "The Occurrence of Clustering in the Recall of Randomly Arranged Associates." Journal of General Psychology 49: 229-240: October, 1955. Bruner, J. S. "Going Beyond the Information Given." Contemporary Approaches to Cognition. Cambridge, Massa- chusetts: Harvard University Press, 1957. Bruner, J. 8. "Learning and Thinking." Harvard Educational Review 29: 18 -192: Summer,.1959. Bruner, J. S. "The Act of Discovery." Harvard Educational Review 51: 21-52; Winter, 1961. Bruner, J. S. "Some Theorems on Instruction Illustrated with Reference to Mathematics." Theories of Learning and Ipstruction. N.S.S.E. Yearbook: Vol. 65. University of Chicago Press, 1964. Bryan, W. L. and Harter, H. "Studies in the Physiology and Psychology of the Telegraphic Language." Psychological Review 4: 27-55: January, 1897. Bugelski, B. R. and Scharlock, D. P. "An Experimental Demon- stration of Unconscious Mediated Association." Journal of Experimental Psyghology 44: 554-558: October, 1952. Carmichal, L.: Hogan, H. P.: and Walter, A. A. "An Experi- mental Study of the Effect of Language on the Reproduction of Visual Perceived Form." Journal of Experimental Psychology 15: 75-86: February, 1952. 152 Carroll, John B. "The Analysis of Reading Instruction: Perspectives from Psychology and Linguistics." Theories of Learning and Instruction. N.S.S.E. Yearbook: Vol. 65, 1964. Clark, L. L.: Lansford, T. G.; and Dallenbach, K. M. "Repetition and Associative Learning." American Journal of Psychology 75: 22-40; March, 1960. Cofer, C. N. "A Study of Clustering in Free Recall Based on Synonyms." Journal of General Psychology 60: 5-10; January, 1959. Cofer, C. N. "An Experimental Analysis of the Role of Context in Verbal Behavior." Transactions of the New York Academy of Science 22: 541-547: Series II, 1960. Cofer, C. N. “Verbal Behavior in Relation to Reasoning and Values." Groups, Leadership, and Men, H. Guetzkow, (ed.) New York: Russell and Russell, 1965. Cofer, Charles N. "Some Evidence for Coding Processes Derived from Clustering in Free Recall." Journal of Verbal Learning and Verbal Behavior 52: 188-192: April, 1966. Cohen, Burton H. "An Investigation of Recoding in Free Recall." Journal of Experimental Psychology 65: 568-576: April, 1965. Cohen, Burton H. "Some-or-None Characteristics of Coding Behavior." qurnal of Verbal Learning and Verbal Behavior 5: 182—187: April, 1966. Coleman, E. B. "Approximations to English: Some Comments on the Method." American Journal of Psychology 76: 259-248; June, 1965. Corman, B. R. "The Effect of Varying Amounts and Kinds of Information as Guidance in Problem Solving." Psychological Monographs 71, No. 451: 1957. Craig, R. C. "Directed vs. Independent Discovery of Estab— lished Relations." Journal of Educational Psychology 47: 225-254; April, 1956. Dallett, Kent M. "Category Information in Free Recall." Journal of Experimental Psychology 68: 1-12: July, 1964. Davidson, Robert E. "Mediation and Ability in Paired- Associate Learning." Journal of Educational Psychology 55: 552-556: December, 1964. 155 Deese, J. and Hardman, G. W., Jr. "An Analysis of Errors in Retroactive Inhibition of Rote Verbal Learning." American Journal of Psychology 67: 299-507: June, 1954. Deese, J. and Kaufman, R. A. "Serial Effects in the Recall of Unorganized and Sequentially Organized Verbal Material." Journal of Experimental Psychology 54: 180-187; September, 1957. Deese, J. "On the Prediction of Occurrence of Particular Verbal Intrusions in Immediate Recall." Journal of Experimental Psychology 58: 17-22; July, 1959Ka). Deese, J. "Influence of Inter-item Associative Strength upon Immediate Free Recall." Psychological Reports 5: 505- 512: 1959(b). Duncker, K. "On Problem-Solving." (Trans., L. S. Lees) Psychological Monggraphs 58, No. 270: 1945. Eagle, Morris and Leiter, Eli. "Recall and Recognition in Intentional and Incidental Learning." Journal of Experi— mental Psychology, 68: 58-65: July, 1964. Earhard, Marcia. "The Facilitation of Memorization by Alphabetic Instructions." Canadian Journal of Psychglogy. 21: 15-24; February, 1967. Entwisle, Doris R. and Huggins, W. H. "Interference in Mean- ingful Learning." Journal of Educational Psychology 55: 75-78; April, 1964. Epstein, W.; Rock, I.; and Zuckerman, C. B. "Meaning and Familiarity in Associative Learning." Psychological Monographs 74, No. 491: 1960. Epstein, W. "The Influence of Syntactical Structure on Learning." American Journal of Psychology 74: 80-85; March, 1961. Ewert, P. H. and Lambert, J. F. "Part II: The Effect of Verbal Instructions upon the Formation of a Concept." Journal of General Psychology 6: 400-415: 1952. Farber, I. E. "The Things People Say to Themselves." American Psychologist 18: 185-197; April, 1965. Feldman, S. M. and Underwood, B. J. "Stimulus Recall Follow- ing Paired-Associate Learning." Journal of Experimental Psychology 55: 11-15: January, 1957. 154 Foley, J. P., Jr., and Cofer, C. N. "Mediated Generalization and the Interpretation of Verbal Behavior: Experimental Study of Certain Homophone and Synonym Gradiants.“ Journal of Experimental Psychology 52: 169-175: February, 1945. French, R. S. "The Effect of Instructions on the Length- Difficulty Relationship for a Task Involving Sequential Dependency." Journal of Experimental Psycholggy 48: 89- 97; August, 1954. Gagne, R. M. and Brown, L. T. "Some Factors in the Program— ming of Conceptual Learning." Journal of Experimental Psychology 62: 515-521; October, 1961. Gagné: R. M. and Smith, Ernest C., Jr. "A Study of the Effects of Verbalization on Problem-Solving." Journal of Experi— mental Psychology 65: 12-18; January, 1962. / Gagne, R. M. "Problem-Solving." Catsgories of Human Learnipg, Arthur W. Melton (ed.) New York: Academic Press, 1964. Geiser, S. and Greenhouse, S. W. "An Extension of Box's Results on the Use of the F-Distribution in Multivariate Analysis." Annals of Mathematical Statistics 29: 885-891; 1958. Gibson, E. J. "A Systematic Application of the Concepts of Generalization and Differentiation to Verbal Learning." Psychological Review 47: 196-229: May, 1940. Goss, Albert E. "Acquisition and Use of Conceptual Schemes." Verbal Learning and Verbal Behavior, C. N. Cofer and B. S. Musgrave (eds.) New York: McGraw-Hill, 1961. Goulet, L. R. "Retroaction and the 'Fate' of the Mediator in Three Stage Mediation Paradigms." Journal of Verbal Learning and Verbal Behavior 5: 172-176: April, 1966. Hall, J. F. "Retroactive Inhibition in Meaningful Material." Journal of Educational Psychology 46: 47-52: January, 1955. Hall, J. F. and Ugelow, Alvin. "Proactive, Retroactive, and Coactive Inhibition with Meaningful Material." Psychological Reports 5: 515-517: 1959. Haslerud, G. M. and Meyers, S. "The Transfer Value of Given and Individually Derived Principles." Journal of Edu— cational Psychology 49: 295-298: December, 1958. 155 Hendrickson, G. and Schroeder, W. N. "Transfer of Training in Learning to Hit a Submerged Target." Journal of Educational Psychology 52: 205-215: March, 1941. Hendrix, G. "A New Clue to Transfer of Training." Elementary School Journal 48: 197-208; December, 1947. Hilgard, E. R.; Irvine, R. D.: and Whipple, R. P. "Rote Memori- zation, Understanding, and Transfer: An Extension of Katona's Card Trick Experts." Journal of Experimental Psychology 46: 288-292; October, 1955. Hilgard, E. R.: Edgren, R.; and Irvine, R. D. "Errors in Transfer Following Learning with Understanding." Journal of Experimental Psychology 47: 457—464: June, 1954. Hilgard, E. R. "A Perspective on the Relationship Between Learning Theory and Educational Practice." ,Tpeories of Lparning and Instruction, N.S.S.E. Yearbook: Vol. 65, Part I: 402—415. Chicago: University of Chicago Press, 1964. Horton, D. L. and Kjeldergaard, P. M. "An Experimental Analysis of Associative Factors in Mediated Generaliza- tion." Psychological Monographs 75: No. 515; 1961. Howes, D. and Osgood, C. E. "On the Combination of Associative Probabilities in Linguistic Contexts." American Journal of Psychology 67: 241-258; June, 1954. Jenkins, J. J.: Mink, W. D.: and Russell, W. A. "Associative Clustering as a Function of Verbal Association Strength." Psychological Reports 4: 127-156: 1958. Jenkins, J. J. "Stimulus 'Fractionation' in Paired—Associate Learning." Psychological Reports 15: 409-410: October, 1965. Jensen, Arthur R. and Rohwer, William D., Jr. "Verbal Medi- ation in Paired-Associate and Serial Learning." Journal of Verbal Learning and Verbal Behavior 1: 546-552: February, 1965(a). Jensen, A. R. and Rohwer, W. D. "The Effect of Verbal Medi- ation on the LearninganuiRetention of Paired-Associates by Retarded Adults." American Journal of Mental Qeficiengy 68: 80-84: July, 1965(b). Judd, C. H. "The Relation of Special Training to General Intelligence." Educational Review 56: 28-42: June, 1908. 156 Kendler, T. S. and Kendler, H. H. "Effect of Verbalization on Reversal Shifts in Children." Science 154: 1619-1620; November, 1961. Kendler, Howard H. "Coding: Associationistic or Organiza- tional?" Journal of Verbal Learning and Verbal Behavior 5: 198-200, April, 1966. Kersh, B. Y. "The Adequacy of 'Meaning' as an Explanation for the Superiority of Learning by Independent Discovery." Journal of Educational Psychology 4: 282-292; October, 1958. Kersh, B. Y. "The Motivating Effect of Learning by Directed Discovery." Journal of Educational Psychology 55: 65-71; April, 1962. Kimble, G. A. and Dufort, R. H. "Meaningfulness and Isolation as Factors in Verbal Learning." Journal of Experimental Psychology 50: 561-568: December, 1955. Kittell, J. E. "An Experimental Study of the Effect of External Direction During Learning on Transfer and Retention of Principles." Journal of Educational Psychology 48: 591- 405: November, 1957. Lindley, R. H. "Effects of Controlled Coding Cues in Short- Term Memory." Journal of Experimental Psychology 66: 580-587: December, 1965. - McGeogh, J. A. and McDonald, W. T. "Meaningful Relation and Retroactive Inhibition." American Journal of Psychology 45: 579-588: October, 1951. McGeogh, J. A. and McKinney, F. "The Susceptibility of Prose to Retroactive Inhibition." American Journal of Psye chology 46: 429-456; July, 1954. McGeogh, J. A. and Underwood, B. S. "Tests of the Two-Factor Theory of Retroactive Inhibition." Journal of Experimental Psychology 52: 1-17: January, 1945. Mandler, G. and Earhard, B. "Pseudo-Mediation: Is Chaining an Artifact?" Psychonomic Scienge 1: 247-248; September, 1964. Marks, M. R. "Problem-Solving as a Function of the Situation." Journal of Experimental Psychology 41: 74-80; January, 1951. Marks, M. R. and Jack, C. "Verbal Context and Memory Span for Meaningful Material." American Journal of Psychology 65: 298-500: April, 1952. 157 Martin, Clessen J.; Boersma, Frederick J.; and Cox, David, L. "Associative Strategies in Paired—Associate Learning." Paper read at the Midwestern Psychological Association Convention: Chicago, Illinois; 1965(a). Martin, Clessen J.; Cox, David, L.; and Boersma, Frederick J. "The Role of Associative Strategies in the Acquisition of P-A Material; An Alternate Approach to Meaningfulness." Psychonomic Science 5(10): 465-464: 1965(b). Martin, C. J.: Cox, D. L.: and Bulgarella, R. "Verbalization of Associative Strategies at Three Developmental Levels." Paper read at the American Psychological Association Convention: New York, New York, 1966. Miller, George A. and Selfridge, Jennifer A. "Verbal Context and the Recall of Meaningful Material." American Journal pf Psychology 65: 176-185; April, 1950. Miller, G. A. "Human Memory and the Storage of Information." I.R.E. Transcriptions of Information Theory 2: 129-157; 1956(a). Miller, G. A. "The Magican Number 7, Plus or Minus 2: Some Limits on Our Capacity for Processing Information." Psychological Review 65: 81-97; March, 1956(b). Miller, G. A. "Information and Memory." Scientific American 195: 42-46; August, 1956(c). Miller, George. "Free Recall of Redundant Strings of Letters." Journal of Experimental Psychology 56: 485-491; December, 1958. Murdock, B. B., Jr. "The Effects of Failure and Retroactive Inhibition on Mediated Generalization." Journal of Experi— mental Psychology 44: 156-164: September, 1952. Newman, Edwin B. "Forgetting of Meaningful Material During Sleep and Waking." American Journal of Psychology 52: 65-71; January, 1959. O'Brien, Francis J. "A Qualitative Investigation of the Effect of Mode of Presentation Upon the Process of Learning." American Journal of Psychology 52: 249-285: April, 1921. Olander, H. T. Transfer of Learning in Simple Addition and Subtraction." Eiementary School Journal 51: 558-569, 427-457; January and February, 1951. Osgood, C. E. "Meaningful Similarity and Interference in Learning." Journal of Experimental Psychology 56: 277- 501: August, 1946. 158 Paul, Irving H. "Studies in Remembering: The Reproduction of Connected and Extended Verbal Material." Psychological Issues 1: No. 2: 1959. Peters, H. N. "Mediate Association." Journal of Experimental Psychology 18: 20-48; February, 1955. Pillsbury, W. B. and Raush, H. L. "An Extension of the Kohler- Restorff Inhibition Phenomenon." American Journal of Psychology 56: 295-298: April, 1945. Pollack, I. "Assimilation of Sequentially Encoded Information." American Journal of Psychology 66: 421-455: July, 1955. Postman, L. "Learned Principles of Organization in Memory." Psychological Monographs 68: no. 5: 1954. Postman, L.: Adams, P. A.: and Bohm, A. M. "Studies in Inci- dental Learning: V, Recall for Order and Associative Clustering.“ Journal of Experimental Psychology 51: 554-542; May, 1956. Postman, L. and Adams, P. A. "Studies in Incidental Learning: VII, The Effects of Contextual Determination." Journal of Experimental Psychology 59: 155-164: March, 1960. Postman, L. "The Effects of Language Habits on the Acquisition and Retention of Verbal Associations." Journal of Experi- mental Psychology 64: 7-19: July, 1962. Postman, Leo. "Does Interference Theory Predict Too Much Forgetting." Journal of Verbal Learning and Verbal Behavior 2: 40-48: July, 1965. Postman, L. and Phillips, L. W. "The Effects of Variable Contexts on the Acquisition and Retention of Paired- Associates." American Journal of Psychology 77: 64-74: March, 1964. Pyles, Marjorie. "Verbalization as a Factor in Learning." Child Development 5: 108-115: June, 1952. Ray, W. E. "Pupil Discovery vs. Direct Instruction." Journal of Expegimental Education 29: 271-280; March, 1961. Reed, H. B. "Associative Aids: I. Their Relation to Learn- ing, Retention, and Other Associations." Psychological Review 25: 128-155: March, 1918. 159 Reid, L. 8.; Lloyd, K. E.: Brackett, N.R.; and Hawkins, W; F. "Short-term Retention as a Function of Average Storage Load and Average Load Reduction." Journal of Experi- mental Psychology 62: 518-522; November, 1961. Reid, L. 8.: Brackett, R.; and Johnson, R. B. "The Influence of Relationships Among Items to be Recalled Upon Short- term Retention." Journal of Verbal Learning and Verbal Behavior 2: 86-92; July, 1965. Richardson, Jack. "The Learning of Concept Names Mediated by Concept Examples." Journal of Verbal Learning and Verbal Behavior 1: 281-288: January, 1965. Richardson, Jack. "The Effect of B-C Presentation and Anti- cipation Interval on Mediated Transfer." Journal of Verbal Learning and Verbal Behavior 5: 119-125: April, 1966. Richardson, Patricia and Voss, J. E. "Replication Report: Verbal Context and the Recall of Meaningful Material-" Journal of Experimental Psychology 60: 417-418: December, 1960. Rock, J. "The Role of Repetition in Associative Learning." American Journal of Psychology 70: 186-195: June, 1957. Rothkopf, E. Z. and Coke, E. U. "The Prediction of Free Recall from Word Association Measures." Journal of Experimental Psychology 62: 455-458; November, 1961. Rubenstein, H. and Aborn, M. "Immediate Recall as a Function of Degree of Organization and Length of Study Period." Journal of Experimental Psychology: 146—152; August, 1954. Russell, W. A. and Storms, L. H. "Implicit Verbal Chaining in Paired-Associate Learning." Journal of Experimental Psychology 49: 287-295: April, 1955. Sakoda, J. M. "Individual Differences in Correlation Between Clustering and Recall of Meaningful Words." Journal of General Psychology 54: 185-190: April, 1956. Saul, E. V. and Osgood, C. E. "Perceptual Organization of Materials as a Factor Influencing Ease of Learning and Degree of Retention." Journal of Experimental Psy- chology 40: 572-579; June, 1950. Schaub, G. Rolf and Lindley, Richard H. "Effects of Subject- Generated Recoding Cues on Short-term Memory." Journal of Experimental Psychology 68: 171-175: August, 1964. 160 Schulz, R. W. and Lovelace, E. A. "Mediation in Verbal Paired—Associate Learning: The Role of Temporal Factors." Psychonomic Science 1: 95-96: May, 1964. Shanmugan, A. V. and Miron, M. S. "Semantic Effects in Mediated Transfer." Journal of Verbal Learning and Verbal Behavior 5: 561-568: August, 1966. Shepard, R. N.; Hovland,‘C. 1.; and Jenkins, N. M. "Learning and Memorization of Classifications." Psychological Monographs 75: No. 517: 1961. Shepard, Roger N. "Learning and Recall as Organization and Search." Journal of Verbal Learning and Verbal Behavior 5: 201-204; April, 1966. Shepard, W. O. and Shaeffer, Maurice. "The Effect of Concept Knowledge on Discrimination Learning." Child Development 27: 175-178: June, 1956. Siegel, Irving. "Structure Effects Within a Memory Series." Journal of Experimental Psychology 55: 511-516: October, ,1945. Slamecka, N. J. "Studies of Retention of Connected Discourse." American Journal of Psychology 72: 409-416; September, 1959. Slamecka, N. J. "Retroactive Inhibition of Connected Discourse as a Function of Practice Level." Journal of Experimental Psychology 59: 104-108: February, 1960(a). Slamecka, N. J. "Retroactive Inhibition of Connected Discourse as a Function of Similarity of Topic." Journal of Experimental Psychology 60: 245-249; October, 1960(b). Spielberger, Charles D. and Levin, Saul M. "What is Learned in Verbal Conditioning?" Journal of Verbal Learning and Verbal Behaviog 1: 125-152; September, 1962. Spiker, Charles C. "Associative Transfer in Verbal Paired- Associate Learning." Child Development 51: 75-87; March, 1960. Stacey, Chalmers L. "The Law of Effect in the Retained Situ- ation with Meaningful Material." Learning Theory in School Situations, E. J. Swenson, G. L. Anderson, and C. L. Stacey. Minneapolis: University of Minnesota Press, 1949. 161 Suppes, Patrick. "Towards a Behavioral Psychology of Mathe- matics Thinking." Unpublished Manuscript, Stanford University,.1965. Reported in Bruner, J. S. "Some Theorems on Instruction Illustrated with Reference to Mathematics.“ Theories of Learning and Instruction, N.S.S.E. Yearbook: Vol. 65. Chicago: University of Chicago Press, 1964. Swenson, Esther J. "Organization and Generalization as Factors in Learning, Transfer, and Retroactive Inhibition." Learning Theorypin School Situations, E. J. Swenson, G. L. Anderson, and C. L. Stacey. Minneapolis: University of Minnesota Press, 1949. Thiele, C. L. "Arithmetic in the Middle Grades." The Teaching of Arithmetic, N.S.S.E. Yearbook: Vol. 50, Part II, Chicago: University of Chicago Press, 1951. Tulving, E. and Patkau, Jeannette E. "Concurrent Effects of Contextual Constraint and Word Frequency on Immediate Recall and Learning of Verbal Material." Canadian Journal of Psychology 16: 85-95: June, 1962. Tulving, E. "Subjective Organization in Free Recall of 'Unrelated' Words." Psychological Review 69: 544-554; July, 1962(a). . Tulving, E. "The Effect of Alphabetical Organization on Memorizing Unrelated Words." Canadian Journal of Psy- chology 16: 185-191; September, 1962(b). Tulving, Endel. "Subjective Organization and EffeCts of Repetition in Multi-Trial Free-Recall Learning." Journal of Verbal Learning and Verbal Behavior 5: 195-197; April, 1966. Tulving, Endel and Pearlstone, Zera. "Availability Versus Accessibility of Information in Memory for Words." Journal of Verbal Learning and Verbal Behavior 5: 581-591; August, 1966. Underwood, B. J. "The Effect of Successive Interpolations on Retroactive and Proactive Inhibition." Psychological Monographs 15: no. 275: 1945. Underwood, B. J. "Proactive Inhibition as a Function of Time and Degree of Prior Learning." Journal of Experimental Psychology 59: 24-54; February, 1949. Underwood, B. J. and Hughes, R. H. "Gradiants of Generalized Verbal Responses." American Journal of Psychology 65: 422-450: July, 1950. 162 Underwood, Benton J. "Interference and Forgetting." Psychological Review 64: 49-60; January, 1957. Underwood, B. J. and Schulz, R. W. "Studies of Distributed Practice: XXI, Effects of Interference from Language Habits." Journal of Experimental Psychology 62: 571-575; December, 1961. Underwood, B. J.: Rehula, R.; and Keppel, G. "Item Selection in Paired-Associate Learning." American Journal of Psye chology 75: 555-571: September, 1962. Underwood, B. J. and Keppel, G. "Coding Processes in Verbal Learning." Journal of Verbal Learning and Verbal Behavior 1: 250- 257, January, 1965. Underwood, B. J. "Stimulus Selection in Verbal Learning." Verbal Behavior and Learning, C. N. Cofer and B. S. Musgrave (eds.) New York: Russell and Russell, 1965. Underwood, B. J. "The Representativeness of Rote Verbal Learning." Categories of Human Learning, A. W. Melton (Ed.) New York: Academic Press, 1964. Underwood, B. J. and Erlebacher, Adrienne H. "Studies of Coding in Verbal Learning." Psychological Monographs: General and Applied 79: No. 606; 1965. Von Restorff, H. "Ueber die Wirkung von Bereichsbildungen in Spurenfeld,“ Psychol. Forsch 18: 299-542; 1955, as reported in Organizing and Memorizing: Studies in the Psychology of Learning and Teaching, S. Katona. New York: Columbia University Press, 1940. Waters, R. H. "The Influence of Tuition upon Ideational Learning." Journal of General Psychology 1: 554-547: July-October, 1928. Books Anderson, R. C. and Ausubel, D. P. Readings in the Psychology of Cognition. New York: Holt, Rinehart, and Winston, 1965. 690p. Ausubel, David P. The Psychology of Meaningful Verbal Learning. New York: Grune and Stratton, 1965. 255p. Bartlett, F. C. Remembering: A Study in Experimental and §pcial Psychology. Cambridge, England: Cambridge University Press, 1952. 517p. 165 Cofer, C. N. and Musgrave, B. S. (eds.) Verbal Learningiand Verbal Behavior. New York: McGraw-Hill, 1961. 241p. Cofer, C. N. and Musgrave, B. S. (eds.) Verbal Behavior and Verbal Learning. New York: McGraw-Hill, 1965. 597p. Cox, David L. Effects of Strategy Aids in Paired-Assgpiate Learning: a Developmental Study. Unpublished MaSter's Thesis: Michigan State University, 1965. Dixon, Wilfrid J. and Massey, Frank J. Introduction to Statistical Analysis. New York: McGraw-Hill, 1957. 488p. Gagnéz Robert. Conditions of Learning. New York: Holt, Rinehart and Winston, 1965. Garrett, Henry E. Statistics in Psychology and Education. New York: Longmans, Green and Co., 1958. 478p. Guenther, William C. Analysis of Variance. Englewood Cliffs: Prentice-Hall, 1964. 199p. Johnson, Donald M. The Psychology of Thought and Judgment. New York: Harper and Bros., 1955. 515p. Katona, George. Organizing and Memorizing: Studies in the Psychology of Learning‘and Teaching. New York: Columbia University Press, 1940. 518p. Koffka, K. Principles of Gestalt Psychology. New York: Harcourt, Brace, 1955. 720p. Melton, Arthur W. Categories of Human Learning. New York: Academic Press, 1964. 556p. Miller, G. A.; Galanter, E.: and Pribram, K. H. Plans and the Structure of Behavigp. New York: Holt, 1960. 226p. Osgood, C. E. Method and Theopy in Experimental Psychology. New York: Oxford University Press, 1955. 800p. Siegel, Sidney. Nonparametric Statistics for the Behavioral Sciences. New York: McGraw-Hill, 1956. 512p. Swenson, E. J.; Anderson, G. L.; and Stacey, C. L. Learning Theory in School Situations. Minneapolis: University of Minnesota Press, 1949. 105p. TenBrink, Terry. A Study of Isolation Effects Duringythe _Acquisition and Recall of Spelling Words. Unpublished Master's Thesis: Michigan State University, 1966. 164 Thiele, Carl Louis. The Contribution of Generalization to the Learning of the Addition Facts. New York: Bureau of Publications, Teachers College Columbia University, 1958. 84p. Thorndike, E. L. and Lorge, I. The Teacher's Word Book of 50,000 Words. New York: Columbia University Press, 1944. 274p. Underwood, B. J. and Schulz, R. W. Meaningfulness and Verbal Learning. Chicago: Lippincott, 1960. 450p. Wertheimer, Max. Productive Thinking. New York: Harper and Bros., 1945. 224p. Wertheimer, Max and Wertheimer, Michael (eds.) Productive Thinkipg, revised edition. New York: Harper and Bros., 1959. 502p. Woodworth, Robert S. Experimental Psychology. New York: Henry Holt and Co., 1958. 889p. Woodworth, Robert S. and Schlosberg, Harold. Experimental Psyphology. New York: Holt, revised edition, 1954. 948p. APPENDICES 165 APPENDIX A Instructions Given at Some Point During the Experimental Session 166 167 I. General Instructions Given at the Beginning of Each Experimental Session A. Experiment I "This study involves learning associations among groups of three words. First, you will view a series of nine slides, each of which contains three words. The words on every slide are arranged in rows like this....(write the words 'automobile', 'table', and 'consider' on the board). These slides will be projected for very brief periods, during which time you should sit quietly and concentrate on remembering which words appear on each. Later, I will present one word from each slide, and ask you to write the other two words. In short, your task during the presentation of the first nine slides is to learn to associate the three words on each in such a way that if I present any one of them later, you will be able to write the other two. After you have seen the entire list of three word slides, I will present a second series of slides. These slides will contain one of the words from each of the original groups of three words. As each of these words is presented, I want you to write the two words which appeared with it on the original slides. Notice that each item on your answer sheet has two blanks. Thus, if the first slide contained the word, 'table', you would write the words, 'automobile' and 'consider' in the two blanks of item one like this....(illustrate on board). After this task, there will be a two minute rest period, at which time the assistants will check your answers. I must insist that you remain quiet during these periods. After this rest, I will present the list containing three words again. Then you will see a second set of slides containing a single word. But these words will be different from those presented on the first trial, and will also be in a different order. Thus, in the example, the word which is presented on the second trial might be the word 'consider' instead of 'table' and it might be the fifth word presented, instead of the first. You would then write the words 'automobile' and 'table' in the fifth answer blank. When you can write the two words correctly for all nine slides twice in a row, you will be through learning the list. Your answers must be perfect, however, including spelling, so you might want to concentrate on the spelling of the words as they are shown in the original groups of three. There will be a short break at the end of each trial. Are there any questions? In short, your task is to learn to associate the three words on each slide in such a way that if I present any one of them, you will be able to write the other two. Is everyone ready?" 168 B. Experiment II "This study involves learning associations among groups of three words. First, you will view a series of nine slides, each of which contains three words. The words on every slide will be arranged in rows like this....(write the words, 'automobile', 'table', and 'consider' on the board). These slides will be projected for very brief periods, during which time you should sit quietly and concentrate on remembering which words appear on each. Later, I will present one word from each slide, and ask you to write the other two words. In short, your task during the presentation of the first nine slides is to learn to associate the three words on each of these slides in such a way that if I present any one of them later, you will be able to write the other two. As you might expect from this description, the list of three word slides will be followed by a second series of slides. These slides will contain one of the words from each of the original groups of three. As each of these slides is presented, I want you to write the two words which appeared with it on the original slide. You will have to write your answers quickly, however, as each slide will be shown for only a brief period. Remember that the words which are selected from each of the original groups may be any one of the three words. Another important point is that the words will not be in the same order as the original slides containing three words. Let us suppose. for example, that the triad of words I have put on the board is the first slide that you view. After seeing it, and eight others like it, I will present the list containing single words. The fifth word on this slide might be the word 'table', in which case you would write the words 'automobile' and 'consider' in the fifth answer blank like this....(illustrate on board). 0n the other hand, the word selected might be the word 'consider', instead of 'table', and it might be presented third in order instead of fifth. If this were the case, you would write the words 'automobile' and 'table' in the third answer blank. After you have written the words which you can remember for all nine slides, there will be a two minute rest period. Then we will start the cycle over again. You will see the original groups of three words, followed by the list containing one word from each group, and you will have another two minute rest. This cycle will continue until you can write the two words correctly for all nine slides twice in a row. When you can do this, you will be through learning the list. Your . answers must be perfect, however, including Spelling, so you might want to concentrate on the spelling of the words as they are shown in the original groups of three. Listen carefully, because this is a helpful hint. Other studies have Shown that those people who write the most words in situations like this, even if they are all incorrect, tend to learn the lists fastest. So it will be to your advantage 169 to write whatever comes into your mind when the list of single words is presented, even if you aren't sure of the correct answer.... Are there any questions? In short, your task is to learn to associate the three words on each slide in such a way that if I present any one of them, you will be able to write the other two.... Is everyone ready?" II. Instructions Given Between the First Learning and First Test Trial A. Experiment I "We are now ready to see how much you can remember. Take out your answer sheet marked trial number one. I will give you the number of the space in which you Should write your answers immediately before each slide is shown. In other words, if I say 'one' and then show a slide, you put the two words which go with the word on the slide in the Space marked 'one'. Whatever you do, write your answers quickly, as you will not have much time on each slide. But please print your answers so that the scorers will be able to read them." "Other studies have shown that those people who write the most responses in situations like this usually learn the list fastest. So even if you aren't sure of the answers, write whatever comes into your mind as each Slide is Shown." B. Experiment I; "We are now ready to see how much you can remember. All of you should have the answer Sheet labeled trial number one. Remember, it is to your advantage to write whatever comes into your mind, even if you aren't sure of the correct answer.... Ready?" III. Instructions Given at the End of the First Test Trial in Both Experiment I and Experiment II "Turn your answer sheet face down, and one of the assist— ants will collect it. You will now have a two—minute rest period. Please do not talk. When you have two trials in a row correct, the assistant will read your subject number. At this point, get up as quickly as you can, and go into the ad- joining hall where your instructor will give you further di- rections." 170 IV. Instructions Given Immediately Prior to the Test for Long-Term Retention "In an attempt to see how much you have remembered over the past three weeks, I am going to Show you the list of single words one more time. You will not see the original slides containing three words. Rather, only the list of Single words will be shown. As each word is projected, I want you to write the two words which went with it on the original Slides in the appropriate places on your yellow answer Sheets. As soon as I finish showing the list, I will briefly explain the purpose of the study.... Are there any questions? It is important that all of you do the best you can, even if you have forgotten most of the words. Since those people who write the most words generally do best on these tests for retention, please write whatever comes into your mind as each slide is shown, even if you are not sure of the correct answer. This will insure that your retention score will be the highest possible.... Ready?" V. Instructions Given Prior to the Completion of the Level of Principle Formation Questionnaire in Experiment II . . . . “On the pages of the questionnaire which follow, you will be asked series of questions concerning what went on while you were looking at the groups of words. These questions are arranged like a program similar to the one you used in your natural science course. That is, depending on the particular way in which you answer a given question, you will be directed to another question. If your answer to the first question is 'yes', for example, the instructions tell you to go to page 2. If your answer to this same question is 'no', however, the instructions tell you to answer the second question on page 1. Whatever you do, ANSWER ONLY THOSE QUESTIONS ON THE PAGES TO WHICH YOU ARE REFERRED. In all likelihood, you will not answer all the questions on the questionnaire.. Thus you Should watch very carefully for your instructions regarding the next ques- tion you are to answer. Whatever you do, read the directions at the top of each page very carefully before answering the questions on that page. Finally, please answer the questions as accurately and as hon- estly as you can. Remember, watch carefully for the instruc- tions regarding which questions you are to answer, and answer only those questions to which you are referred. You may begin now." APPENDIX B The Level of Principle Formation Questionnaire 171 172 List Subject Number Name Student Number DO NOT OPEN UNTIL INSTRUCTED TO DO SO EVEN NUMBERED LISTS .175 Directions: Check the apprOpriate box for each question. Then turn to the page which your answer directs you to, and answer the question :on thst page in the same manner. Answer only those ggestions on the pages to which you are referred! 1. Beginning with the first trial, did you attempt to find relationships between the words in‘ppy of the triads? (‘7 yes - turn to page 2. (ignore question #2 below) ’ ID no - answer question 2 below. 2. Despite the fact that you were not actively looking for relationships among the words in the triads, did you nevertheless discover that relationships existed on one or more of the triads? :3 yes - turn to page 2 D no - turn to page I. .174 Consider the following two triads: Dab Bad Evil Tac Cat Feline If you observe closely enough, you will see that the words in the first triad are related to each other in exactly the same way that the words in the second triad are related. (Note: "dab" is "bad" spelled backwards and "bad" is synonymous with "evil".) They might thus be said to have a "common" relationship. While you were looking for relationships among the words in the triads on the list you just completed, did you ever discover a common relationship between words in two or more of the triads? [:3 yes - turn to page 3 {:3 no - turn to page 4 I «ET-12:0 175 l. were you able to find a single relationship which held for all of the triads? [:7 yes - TURN TO PAGE 5 a no - answer question 2 below. (ignore #3) 2. About how many common relationships did you discover? (Indicate the approximate number of triads which shared each of the common relationships you discovered in the space provided.) I 7 one This relationship was common to about triads. [:3 two The first relationship was common to about triads. [::3 three The first relationship was common to about triads. The second was common to about.________ triads. The third was comon to about triads. l ) four or more TURN TO PAGE 4 1176 Directions: DO NOT TURN THE PAGE NOR REMOVE THE CLIP BEFORE READING ms , DIRECTIONS CAREFULLY: On the pages which follow, you will find the list of triads which you were presented in this study. After reading the directions, expose the pages, gps‘s§.g time. Write any relationship which you found among the words in the exposed triad in the space provided to the right of each. Do this as quickly and as completely as you can. Example: Dab Bad Evil: "bad" is "dab" spelled backwards and means about the same thing as "evil". It is important that you write only those relationships which you discovered while going through the initial learning experience. Do us; write relationships which ysg discover while completipg‘shis_ task. This would Invalidate the study. .0“ 7 ‘ If you were not able to find a relationship between the words in a given triad while learning the list, simply write the word "none" in the space provided. Remember, write the relationship as quickly and as completely as you can. Finish each triad as it is exposed. Do not turn back to earlier pages! When you have finished, TURN TO PAGE 7. Turn the page and begin writing. 177 In the questionnaire which was presented to each subject, each triad was presented on an individual Sheet. For pur- poses of convenience, however, these are depicted as follows: 1. appendage feat tale: 2. beverage whine bier: 5. food foul meet: 4. study rite reed: 5. metal steal led: 6. weather rein son: 7. group heard teem: 8. receptor ayes knows: 9. mammal dear bare: TURN TO PAGE 7 :178 Directions: Keep the single relationship which you feel holds for all triads firmly in mind. Then check the boxes to the left of two statements below which best describe the relationship which you have in mind. Example: The relationship which held for both of the triads in the example presented earlier (Dab, Bad, Evil) was that the middle word was the first word spelled backwards, and was synonymous with the third word. Thus the following two statements regarding this relationship have been properly checked belaw.f0t the example. : TX? When one of the words is spelled backwards, it is identical '7 with another word in the triad. 5;: Two of the words have opposite meanings. Now check two statements below which are true of the relationship which you feel holds for all triads in the list you learned earlier. CHECK.THE TWO STATEMENTS WHICH BEST DESCRIBE THE RELATIONSHIP! L’J 9. The relationship which holds for all triads is that the words in each may be easily connected in a sentence. Tvo of the words in each triad have opposite meanings. Two of the words in each triad sound alike. If the spelling of one of the ponds in each triad were changed into a word which sounds more like it, the overall relationship between the words would be more apparent. If the spelling of one of the words in each triad were changed into a word which sounds like it, it would be identical with another word in the triad. If the spelling of two of the voids in each triad were changed into words sound like then» the overall relationship between the words would be more apparent. With certain modifications, all of the words in the triad mean about the same thing. With certain modifications, all of the words in the triads are examples of a fourth categorical word. With certain modifications, two of the words in the triads are examples of the third word. TURN TO PAGE 6 179 Directions: Suppose you are asked to construct a list of triads similar to the one which you have just completed. Bearing in mind the relationship which you feel holds for all of the triads, select three words from each of the lists below which conform to this relationship. In other words, construct two triads which are as similar as possible to those on the list which you have just completed. CIRCLE THE THREE WORDS IN EACH LIST. QgHQQE look back on page 5 before completing this task! ,_Lia£_l_ _Li££_2_ peach triple two nine . intelligent for -cherry three pair against dumb ate fruit sixty died triad plumb number apple triangle If you learned list 1, 3, 5, or 7, ANSWER QUESTION #1 BELOW. (Ignore #2) If you learned list 2, 4, 6, or 8,-ANSWER.QUESTION #Z'BELOW. (Ignore #1) 2. Briefly examine your answers on page 5. Were the two statements which you checked #6 and #9? m ~*fl yes - turn to page 7 and follow the directions. ~“”T “0 - turn back to page 4 and follow the directions. ‘0“— Briefly examine your answers on page 5. Were the two statements which you checked #4 and #9? CIZL yes - turn to page 7 and follow the directions. .2 no - turn back to page 4 and follow the directions. (. 18C) Please state any additional information which might be helpful in analyzing your performance so this test: Other than allowing the investigater to complete his dissertation, what do you feel is the purpose of this study? After you have completed the two questions above, you have finished your part of the study. Please hand these sheets to your instructor. I sincerely appreciate your cooperation in this study. APPENDIX C Table 17 through 25 Additional Analyses of the Data 181 182 Amv mom.m Amy mmo.¢ m3ocxlAmmmmvmmmmunoummomu m ANV who.¢ Adv wmm.w cowlAcamuvofimuuumsummB 0 Add emm.m Ame mme.a ammouxummokumouameems a Amy mom.m Amv Hhm.m mama:AummmvumwMImmmocmmmm m Amv mm¢.m Amy oam.m ommHIAmufluzvmuHHImosum m AHV mmm.¢ ANV mom.o “can:Amofl3vocwn3lmmmnm>mn o Amv ¢H>.N Any m¢>.ma EmmuIAUHmfivoummfinmsoum m Amv mam.m on mmm.o Hmoumuhomoavowaunmuwe m Ame eao.m “my mom.~ humanxasoevasoeuoooe a xcom cows zoom com: omens unm>oamm umHH mop ,mumflq ENdoEOmIOBB woman ENdoEom moo CH cowuwmom I a II .Aoo.m u manflmmom Hououv Umauu some now mamanu ummu mounu umuflm mnu um>o ooaamomh hauumuuou mono? mo HdflEdn new: .ha magma 185 J _ AHH ucmfiflummxmv Admv Anmv AHNV Acv Amway umnflm mw.> dd.m mm.dd Nxm co fiomuuoo .m.z .m.z .m.z mm.d Hm.m em.> mm.> MW mouoz Hogans Ammo Acme Adam Ant Aaoeugmumue so.e mm.m. mm.ea mxm bumunoo .m.z .m.z .m.z ma.o mw.oa mm.0a no.0a mm mouo3 Hanan: Ammo lose lame lav a>.m mm.m m~.ms mam, . coaumuauu on *x .m.z .m.z **aa.m e¢.m mm.m mm.m m_ masses cones: Ammo Amav Aamv Ace meoaunoumeo om.m m>.m mN.N Nxm mcflaammm .m.z .m.z .m.z am.a mm.a mm.m, e¢.d .m “moses Ammv Amov AHNV . Amv uumunou mm.mm om.ddm mm.¢Hd.m m m mUHOS .m.z .m.z .m.z mo.m mm.mma mm.maa me.oma mm Hones: Hmuou "mumflq E>cofiomlmco efi >HuHHH >HuHH HHHaHH capsule >H HHH HH moaumaumum moacmaum> mcomflummfiou HMSUH>HUGH coaumfiuomlmHmHUCHHm m0 Hw>mq mamfimm ucmwcmmwn mo ummB.Wmmmnum «H F .oofiumEuom mamflocaum mo mam>ma quHGMMHU no Umflmammoao mm mcoEo mmucmHmMMAU ecu mo ummu mucmwno> mo mflmhamcm hmanco m W0 muadmmm .mfi mHQMB 184 «0.0 v m mo.o v m uGMUflmacmHm uo: u .m.z ** * AHH ucmEHHomxmv Amdv thv Ammv Aav HMAHU umnfim oo.¢ mm.m mo.md Nxm :0 uomuuoo omez . emez smsz stH ”WaN. OHem mNem lxl. mcHOQV ngg Ammv Abav Addy Acv A:0fluomumuv Hm.md am.0d mm.nd xm uoounoo .m.z .m.z .m.z mm.a ae.aa mm.oa em.ma mm mouoz Hones: fiend Amev Ammv Acv ma.n >m.m mm.¢ xm coaumuauo oy 1 .m.z .m.z £392 mad Exp 2.6. mm 335 noose: Aomv Andy Ammv sz Aaowuwmasvuov em.m uo.m mm.ma mxm mcoauuoumao .m.z .m.z .m.z 3:: mad 33., min m 923% $85 Awmv _ Andy shmmv any Acoauflmwsvomv am.mme mm.mmm em.emm mxm somuuoo ** .m.z .m.z **mo.md d>.mma NO.¢OH mm.mdd .X mvuo3 Hmnfid ”mumflq.fiua080m103a . 185 mm.o v m v oa.o * Amm.o u menace HGMUflmwcmHm #02 u .m.z maumaano ou mafiouoood definammmau mam56w>flocH .m.z Adda Ame Adda awed Ace me.aa om.> m¢.ma mm.om mxm Acoaucmusuv No .0 mp in.“ om . a.“ am 13“ mm . .3” mm uomunov mono} .m.z Nome Ame Ammv Amwv Ace em.m mo.m em.m mm.» am coaumuauo 0» ma.o m¢.m ss.m ma.m mm.m «m. meshes Hones .m.z imam Ame Amos fined Ame am.» s.,me.o« mo.m mm.m m m meoausouma ma.o mm.a e¢.m Hm.m mm.m mp meaaamom amass .m.z Amae Ame “new Anae Ace me.am mm.mma mo.oha mo.mma xm homuuo mm.o mm.mma mm.mma om.mma. om.ama mm. moses amass "mmmmmdmwcoawnloza . * Ame u at an n 2. Am u av Am u av Ace am.m om.e eo.am mm.a New Aeoauemumuv E .H em .oa om .3 S .9“ ma .3 m oomuuoo mono .m.z lam n at “a n no it u no Add u at Age mm.m mm.d om.m >¢.m Nxm coaumufluo o ma.o em.m ma.m ah.m me.m mp mamas» amass .m.z Adm u av Aw u av As H av Awd n CV any mm.n am.« om.« ma.a New meoauuouma ow .0 mm .a me A mm 4 mm .o m 933% 3&5 .m.z Adm n av Aw u av A5 n no Ada n no Add mm.m> om.md hm.>m >m.ooa NXm uuouuo am.o am.>ma m>.mma a>.ema mm.mma _m mono; Hones ulwmfla E>coEomlwco . oflmuum >H HHH HH H & Gammon manoflnm ucwvcmmm I {It .m magma ca Umnfluommo maumufluu poem any 0» mawououom UDHMfimmmao mHmSUH>HUcH maoEm mwocmummmflo mo ummu mUGMHHm> mo mammdmcm >m31mqo m mo muasmmm .mfi wanes 186 **Ho.m **md.m **>N.ml **¢Om.m Anz m> QV mummulu Ame Ame ~ma Am. : me.¢m am.e we.o $5.0m mxm omaeflmmmao so mm.ma ad.s mm.m mm.>m m_ no: cases 0:3 mm Aug Ammo ~nmv Anne Ammv He Aczv em.mm am.od we.«s em.mae xm Hoaoeauo was em>oo mm.ma mm.m em.e em.moa am. an no: see 0:3 mm Ame Away capo “mac imam m A am.mm mo.» om.m em.aea m m Ace maoaoeauo was oe.m mm.m o~.m em.mma mm omnm>oomao 0:3 mm Ace "coaummflsoofi J woman E>cofiomlo3a .HH **mm.m Om.H **mH.MI **OO.¢ AQZ m> 0v mumwulu Adv Add Aav Adv m u n I I m m omaeammmau ma. oo.¢ oo.o oo.m oo.mma .m soc oases 0:3 mm Aug Aamv Aamv Aamv Aamv a Amzv mm.om «0.0a mm.« am.mom mxm mamaocauo we» um>ou am.oa mm.m Na.e mm.aaa _m. imao so: can 0:3.mm Ame Ammo ~wma Anne Ammo c mm.ea om.m ea.m ma.mm Nxm Ana mamaoeeum was 8d 24 ohm mmdma m 33883 93 mm a: ”COHtflmHSUU¢ I mumwn E>cofiom moo .H m:0Hm5HucH macauuoumflo coaumufluu uomuuoo Gammon Hmnfidz moflaawmm on Hones! , Hmuoe Hmnfisz mHoHHB Hmuoa . .H ucmfiflnmmxm How macaw ucmocmmoocH any an some mGOHuMUHMHmmMHU ou moflouooom mamflucflum ecu Hm>oomflo uo: Uflo no UHU Hooves 0:3 mm mo mam>ma muomfiuomnmm some mnu :fl mmUGmHmMMHQ .ON OHQMB 187 H0.0 V m ** mw.d Ho.d . do.d Anz m> av mumwulu. Amv Am. Amv c 00.0 00.0 00.0 xm omaeammmao mo oo.s oo.m om.> mm. you wanna one mm on Aamv Aemv Aemv c Aazv om.> om.m mm.0d Nxm mamwucflum mnu Hm>oo am.m me.« paeaa Hm. Imao uoc who 0:3 mm Amy Ammo Ammo Ammo c mm.e maIN mm.aa xm Ana maouoeuam was mm.m mm.a mo.ma mm omsm>oomflo 0:3 mm Ace "cowuamuom EMmBImaoq I mumme>GOEomIowH_.>H em.a sa.d m>.o Anz_m> av mumoqu Adv «we Adv a II II. II xm omamemmmao on oo.m oo.m oo.m mm. uo: cases 0:3 mm flue Ammo Ammo Ammo; e Amzv ms.m mm.a oo.oa xm maoaoeaso on» um>ou em.m aa.a om.oa mm. Imao son who 0:3 mm Ame 38 ammo $8 a ¢N.N «m.o ¢O.m xm Any mamflocflum ecu; ea.m mo.a so.aa mm. omum>oomao 0:3 mm Ace "coaucmuwm EHmBIwGOA 1"! I I mumflfl EMGOEOEIDGO .HHH 188 Table 21. The most frequently occurring extra-list intrusions for one- and two-homonym lists. Intrusion Occurrence Frequency of Relevant Triad I. One Homonym Lists foot 59 drink 22 iron .15 animal 9 learn 9 feed 9 fish __9. Total 150 Total number different terms II. Two Homonym Lists drink 17 animal 10 ale 7 foot 7 leg 4 iron __$ Total 49 Total number different terms Total number extra-list intrusions N Total number extra-list intrusions NI appendage-tale-feet beverage-bier-wine (19) food-meet-fowl (5) metal-steal-lead mammal-dear-bear study-read—write food-meet—fowl food-meet-fowl = 285 = 82 105 beverage-bierdwhine (15) food-meet-foul (4) mammal-dear-bare beverage-bierdwhine appendage-feat-tale appendage-feat-tale metal-steal-led = 214 = 127 I100 189 H0.0 V m ** m0.0 v m * . **mm.N **Om.N **m>.N m¢.dl Hm.H **Nd.m **mm.¢ Am a dv mumwulu mm.mm mm.ma mm.m 00.0 mm.a 00.m 00.oem mxm Am u 20 omamemmmeu mm.» mm.s mm.a 00.0 em.a 00.m 00.0ma mm on no: canon 0:3 mm A00 mm.mm mm.aa am.s Naps Ism.¢ mm.a 0m.mm mxm ism u 20 madaucenm was em.m sm.m ma.m no.0 mm.m mm.m mm.maa .m_um>oumao no: can 0:3.mm Amv ma.ma _ sm.m Na.a ew.m ma.m 0m.a am.mm mum 50¢ n 20 maoeoeauoI , aa.m ms.a am.0 mm.a mm.« mm.« mm.mma mm. was omuw>oomflowoo3_mm Ace . "woman EthEomIomH..m **mo.m **Nm.d **mm.m mm.OI m¢.H **Om.m **Om.N Am a oouao you one 0:3 mm Amy IIamIMH e0.m mm.m me.0 as.a mm.a mo.a0 mam Ame u 20 macamcaua m0.e mm.« 0a.a 50.0 mm.0 so.m >0.mma mw was omum>oumao one mm lav "muqu E>coaom oco .dJ rim TM Tm dim PM 3N DI w. unu .I r. iL r. o moo e 3. .m 9%. nm Sm m f. 5 JD. 4 1. I 1. D In. 8 HT 8 8 Ba 08 13 ST. u s Ill 7:: TI: 1.4 TI: is P TIu u u s .+ 1 1.u om «4.7.. 178 To 1.5 a; H In IX m 0d 11 m. cad. n.+ nl+ .d u 9 TL? 8 SJI SJ 1 ST 09 a 1 Tre 7:2 0 T. u.L 1 O. 0. d T. S u u e u S S I 5 .oflc mobbed 0:3 mm mcoEm coflpcouou Enoquuocm CH moocouommwc mo muonIu ooaflmqucO .oamfiwm onu cw cocoaocfl COHHoUHHU cocumou hamumEHuHs 033 mm mmonu haco £ua3 .mamflucfium.mnu Hm>oomflo .uo: @flo no .mm magma cm>dm IsHocoo Eoym you oHOMoHony com enzymc ca 00: ymom my cowyyommm many you oococy>o smoocy co>o .oamflocwym may coaaooou mm may no meow ymmoa yo ymcy o>oHHoQ oy coupon boom mscy my chose .Aao.o v mv coaycoyou EyoyImcoH mo monomoofi you ycooflmycmflm adamoyymHyMym one Ham .GOHycoy Ion Enonyyonm mo monomooa co ycmoymycmym mum moocoyommyc 03y yo ococ mmoHoQB ymsy mayooaccfl mymoy omony mo myaomoy ony mo GOHymcHonm .msoym Aoamyocwym yocIonyocHymv may com mooym AssocxcoIonHocwymv may coo3yon mymoyly coaamyloco mo myaomoy ocy myoflmoc omam mm canoe .oamflocwum many Hamooy 0y canons mum onHocHym may coyo>oomwc 0:3 mm co£3 oEHy ca ycflom may yo ycommmmflc waoyoamfioo GOflycoyoy Eyonyyonm mo monomomfi co mmooym 03y mmony coo3ymn ymflXo hoe QUH£3 moocoyomwflo wcm yony mymommsm mamcoyym oucocfl>m mane .ycoymflxococ maamoayomym one? coyycoyou EyoyImcoa mo monommofi How moocoyommwc mcflocommoyyoo may .cowycoyou Eyoylyyocm mo monommoe you Hodycmymnsm who: mmsoum 03y omosy coo3yon oUGMEuowyom ca moocouom IMHU mmoyonk .mooym Aoamflocflum yocloamwocflym yocv ony com msoym AmamwOCHym yocIonHUGHHmV may coo3yon omony oym mcomwymmEoo ycmyyomsy ymoE one .ommm mcflzoaaom may no mm dance ca C30£m mum mmooum ooyny any mo some now mcooz .cOHyoyHyo 0y maoflyy mo Hones: .ccm “mGOHmsyycfl ymflHImuyxo com ImyyCfl mo Hones: “ooaaoooy mayooyyoo mcyo3 mo Hogans "moHQMHHm> ycoocomoo mcHBOHHom ony co msoym some now cocHono cony oyo3 coaycoyoy Eyoylmcoa com Eyoylyyozm SyQfl mo monomooz no mu..AonHocHym HOGIonHUcflHm yocv maofiyy ymoy yzmflo GHSyHB coayoyflyo conocoy mmoHoSyHo>oc 0:3 ySM .coayfimasoom mcfluoc mamwocflum 03y yo>oomflc yoc cat 0:3 mm omosy oHoB coyooflmcoo mp3 £0H£3 moon Hmcflm one .onHOCflym may vOHoQEwEmH o>mn .yoc hoe yo .mmE mooym mflny CH mm ycny yumm mm» mo o>Hym0HUcH ma msoym ocooom any mom Henna any CH.Eyoy =csocxco= use .ACBOGMGSIonwocHH v coaycoyoy How ymoy may no muoyyo o: oomE 0:3 omony com .Aoamflocflym yocloamwucflymu coyycoyoy you ymoy any co mcoayyoymflo mcflaaomm oHoE no moo come 033 omocyllmmooym;03y oycH UoUfl>HU whom onony myo3 coayflmflooom moayoc onHocHHm may coyo>oomHU 0:3 mymHH EhcoEonIOBy oncomoym mm .memcoaoc mo mEHoy ymamfioxo Syon Haomm .MHoEmc “mcflaaomm you many o>Hym>Hymo o mcflmycoo mymHH omosy moHHyoUcS coass oHQflUCHym may yccy yomy may Eoym mBOHHom coayywmmm wane .oamwucflym may mNHHHys yoc vac coHy Icoyoy Eyoylmcoa you ymoy ony co muoyyo mcflaaomm once no 020 come 0:3 mymfla excoEonIo3y nyfiz coycomoym mm moose .COHycoyoH EuoylmcoH mo ymoy ony co camyocwym many oNHHflys yo: tab 033 yon .coyyymyoooo mcflyoc camflocanm any coyo>oomwo 0:3 mm mo msoym Hanan o mwwycoofl maoymuooom oy oanflmmom my ya tyo>o3oc .maoymcoyyom .mcoym mflcy mo cmwmoc osy CH Body HOnmE m mm mocoym many mamflocyym may mo coaycoyou mo ouommofi yooyflc m mo xoma oSB .cowycoyoy Eyoylmcoa mo ymoy may co mamflocflym many Hamoou oy magmas oHoB GOHymeovom awayflcw mcwyso onHUCHHm may coyo>oumflo 0:3 mm omony .maoamc “GOHmDHUcoo Hmyocom oco cym30y ycflom Ham calm mmmocyomms mo mymoa .coayoocoyycfl Homecow .mm canoe 191 H0.0 V m ** **mw.m **>¢.m ,**wd.m mm.a 50.0 ¢¢.o mm.a Am .m> «V ymmyly low. Tom. low. .am. lam. lam. .sm. in." n¢.0 em.a Hm.m mm.0a 0m.e mm.a 0m.mm mxm mamaoeyum om.o 00.0 mo.dd m¢.m mm.d mm.m mm.maa mp yoclmamyocfiym yoz ADV Ammv Ammv Ammv Ammv Ammv Ammv Ammv Adv Ayouym ma.0 00.0 mm.aa we.» ea.a m>.a mm.mm mxm meaaaodm mco ymmma umv ma.0 ae.0 ma.0a om.m me.0 ea.m m0.mma .m. waoaocauc uoeImHoyoeyum Amv Aadv Aaav Aaav Aadv Add Adav Aaav Acv 00.0 ma.0 mm.ma mm.m em.0 00.a em.ma Nxm Amuounm medaamom 0:0 00.0 md.o Hm.md wd.d mm.o oo.m m¢.mmd mm c3ocxcolmamyocwym Adv T..N IN 01 TN TN 3N 0L 1 um mm mm um mm 0m. 3. a. 4 3,c. 1 2 1+0. nnq 0 1.e e e e 9.1 e .e 1 e e.L u Til 1;. 3 fl. r;l 1:4 0 P u. u «wu _‘ u u 3 «en 4.! «+e 1+1. 1.u e 3 Tin Jx m ..u ..x 1.1 m n.+ n.+ q nun. n.+ 1;! 3.. 5.4 e s 1 _s.l o e e 1.2 7:2 I To. To. UT. 1 o _ o _ o . o _ S u. u u u s . s s s . HyQMymm .m Goflyfimflsood .< in . . P! "NW N [Ihl “u .coyycmymy Emmylmcoa mom ymmy may no myOHHm mcflaammm mxmfi .yoc can no .UHU ymnyam 0:3 cam mamaocfiym may omym>oomflo 033 mocha Emcofioclozy mzy cw mm wcoEm mmocmymmmac mo mymmyly cmaflmyImco .mm magma M'TITI'I‘II‘I’IILIMLHEIflififlflfl'fflflflflfliflflmfl'ES