UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Free recall: organization and long-term memory Schwartz, Robert Marc 1972

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-UBC_1972_A1 S38.pdf [ 8.11MB ]
Metadata
JSON: 831-1.0101411.json
JSON-LD: 831-1.0101411-ld.json
RDF/XML (Pretty): 831-1.0101411-rdf.xml
RDF/JSON: 831-1.0101411-rdf.json
Turtle: 831-1.0101411-turtle.txt
N-Triples: 831-1.0101411-rdf-ntriples.txt
Original Record: 831-1.0101411-source.json
Full Text
831-1.0101411-fulltext.txt
Citation
831-1.0101411.ris

Full Text

FREE RECALL: ORGANIZATION AND LONG—TERM MEMORY by ROBERT MARC SCHWARTZ A.B. University of California, 1969 M.A., University of British Columbia, 1970 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in the Department of Psychology We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA July, 1972 In p r e s e n t i n g t h i s t h e s i s in p a r t i a l f u l f i l m e n t o f the r e q u i r e m e n t s f o r an a d v a n c e d d e g r e e a t the U n i v e r s i t y o f B r i t i s h C o l u m b i a , I a g r e e t h a t t h e L i b r a r y s h a l l make i t f r e e l y a v a i l a b l e f o r r e f e r e n c e and s t u d y . I f u r t h e r a g r e e t h a t p e r m i s s i o n f o r e x t e n s i v e c o p y i n g o f t h i s t h e s i s f o r s c h o l a r l y p u r p o s e s may be g r a n t e d by t h e Head o f my D e p a r t m e n t o r by h i s r e p r e s e n t a t i v e s . I t i s u n d e r s t o o d t h a t c o p y i n g o r p u b l i c a t i o n o f t h i s t h e s i s f o r f i n a n c i a l g a i n s h a l l not be a l l o w e d w i t h o u t my w r i t t e n p e r m i s s i o n . D e p a r t m e n t o f Psychology The U n i v e r s i t y o f B r i t i s h C o l u m b i a V a n c o u v e r 8, Canada D a t e August 3. 1972 i ABSTRACT Free recall was examined In order to (1) examine the differences between the clustering of items from a l i s t with an a priori structure and the subjective organization from a l i s t of unrelated words and to (2) describe organizational processes in terms of theoretical constructs. Three sections of experiments found: (1) that two manipulations of or-ganization of lists of unrelated words had asymmetric effects on recog-nition and recall performance; (2) that order of output and recall performance from two lists of unrelated words was predicted by similarity judgments data; and (3) that order of output from two specially-con-structed lists was described by a hierarchical arrangement of categories. The data were interpreted as showing three commonalities between the clustering of items from a l i s t with an a priori structure and the sub-jective organization from a l i s t of unrelated words. These were: (1) asymmetric effects on recognition and recall can be produced by manip-ulations of both categorized and unrelated lists;'(2) order of output can be predicted by similarity judgments data for both categorized and unrelated l i s t s ; and (3) free recall performance can be predicted by conformity to category structures for both categorized and unrelated l i s t s . An informal model of free recall was proposed. The model had the following properties which were abstracted from the experiments: (1) the internal structure of organization was described as hierarchical; (2) the processing which results in this structure was based on similari-ties between items; and (3) the output was described as representing dif-ferent processes dependent upon whether recall or recognition tests are used. 11 TABLE OF CONTENTS Chapter T i t l e Page 1 A Review o f the Data 1 2 A Review of the Theories 29 3 O r g a n i z a t i o n : A Storage o r R e t r i e v a l E f f e c t ? . 48 4 C l u s t e r i n g and S u b j e c t i v e O r g a n i z a t i o n . . . . . 73 5 An Examination of Order of Output . . . . . . . 100 6 Summary 127 7 T h e o r e t i c a l Notes . , 1 3 6 i i i LIST OF TABLES Table T i t l e Page 'I V a r i a n c e s , Means and Transformed Means and Variances f o r the Four Conditions i n Experiment I . . . 56 I I ANOVA Source Table f o r Conditions i n Experiment I. . . . 57 I I I ANOVA Source Table f o r Imagery-Recall and Standard-Recall Conditions as a Function of S e r i a l P o s i t i o n i n Experiment I . . 60 IV Graeco-Latin Square Design Used i n Experiment I I . . . 64 V Means and Variances and Transformed Means and Variances f o r the Four Conditions i n Experiment I I . . 66 VI ANOVA Source Table f o r Conditions i n Experiment I I . . 67 V I I ANOVA Source Table f o r S i l e n t - R e c a l l and Overt-R e c a l l Conditions as a Function of S e r i a l P o s i t i o n i n Experiment I I 70 V I I I S e l e c t i o n Values as a Function of Number of Categories f o r the C l u s t e r A n a l y s i s of the S o r t i n g and Free R e c a l l Data of Experiment IV 86 IX Category S t r u c t u r e s f o r the S o r t i n g and the Free-R e c a l l Data i n Experiment IV 87 X HGroup S e l e c t i o n Values f o r Group Data and Data f o r Two I n d i v i d u a l Subjects as a Function of Number of Categories i n Experiment VI 108 XI Category S t r u c t u r e f o r Mammal Names Reported by Henley (1969) and Found i n Free R e c a l l f o r the Group Data and Data f o r Two I n d i v i d u a l Subjects i n Experiment VI . HO X I I Chi-squared Values and Contingency C o e f f i c i e n t s f o r R e l a t i o n s h i p s between C l u s t e r S t r u c t u r e s f o r Henley's (1969) Data and Group Data and Two I n d i v i d u a l Subjects i n Experiment VI 113 Iv LIST OF FIGURES Figure Title . Page 1 Probability of Recall as a Function of Serial Position for the Imagery-Recall and the Standard-Recall Conditions in Experiment I . . 59 2 Probability of Recall as a Function of Serial Position for the Silent-Recall and the Overt-Recall Conditions in Experiment II 69 3 Recall and Organization Scores as a Function of Trials for a l l Subjects in r&periment IV 89 4 Upper: Recall as a Function of Trials for High, Middle, and Low Organizers in Experiment IV Lower: Organization as a Function of Trials for High, Middle, and Low Organizers in Experiment IV . . . 92 5 Recall and Z-Scores as a Function of Trials for Experiment VI 106 6 Hierarchical Category Structure for the Group Data in Experiment VI . . . . . . . . . . . . . . . i * l 1 1 7 Hierarchical Category Structure for Individual < Subject 1 in Experiment VI 115 8 Hierarchical Category Structure for Individual Subject 2 in Experiment VI 116 9 Recall and Two-Way Subjective Organization Scores as a Function of Trials in Experiment VII 121 V ACKNOWLEDGMENT I thank the members of my dissertation committee for their careful criticisms of this manuscript. Dr.Michael S. Humphreys, my committee chairman, was instrumental throughout the preparation of this manuscript* and he deserves special thanks not only for help with this dissertation but also for providing a stimulating academic environment throughout my graduate career. I thank my friends, Duncan Hi l l and Keith Borkman, for help with preparation of the figures. I also thank my parents for moral and financial support throughout my education, and i t is to them that I dedicate this dissertation. CHAPTER 1 A REVIEW OF THE DATA Methodologically, the free recall paradigm i s perhaps the most simple of the verbal learning paradigms. In a typical free recall ex-periment the subject i s presented a l i s t of words, usually in a random order. He Is then asked to recall as many of the words as he can, in any order that he desires. In the case of sin g l e - t r i a l free r e c a l l , the ex-periment ends when the subject can recall no more words. Unlike the sin g l e - t r i a l free recall experiment,which may be clas-s i f i e d as a memory paradigm, mu l t i t r i a l free recall can be regarded as a learning paradigm. In multitrial free r e c a l l , the subject i s presented a l i s t of words and i s asked to recall them, just as in s i n g l e - t r i a l free r e c a l l . However, after the subject has finished recalling so many words as he can, he i s again presented the l i s t , usually in a different random order. After the second presentation, the subject i s again asked to recall the words. This process of presentation followed by recall i s continued for a fixed number of t r i a l s or u n t i l a preset criterion of performance is reached. A third, albeit less frequently used, variation of the free recall paradigm involves multiple presentations of the words to be recalled but only one t r i a l in which to recall them. That i s , the subject is presented the l i s t of words for a fixed number of t r i a l s ; during these presentation t r i a l s he may be required to perform some operation involving the words ([e.g., sorting). However, he is asked to rec a l l the words only after the fi n a l presentation t r i a l . 2 Although the free re c a l l task is methodologically more simple than paired-associate or serial-learning tasks, free recall requires a d i f -ferent terminology. Traditionally, paired-associate and serial-learning tasks have been described in terms of stimuli and responses. For ex-ample, for a given response in paired-associate learning, the exper-imenter-defined or nominal stimulus is the stimulus term which has been paired with that response. In the case of rote paired-associate learn-ing, the stimulus term is probably the effective stimulus, or the stimulus that the subject uses to produce the response. Even when the production of a response is said to be mediated by a learned association (cf, Kjeldergaard, 1968) or an image (cf. Paivio, 1969), the nominal stimulus must e l i c i t the mediator which is the effective stimulus for the response. In s e r i a l learning, the nominal stimulus is the word preceding the required response in the ser i a l l i s t . The effective stimulus may be identical to the nominal stimulus (e.g., Bugelski, 1950), may be a rep-resentation, e l i c i t e d by the nominal stimulus, of the ordinal position of the required response (e.g., Mueller, 1970), or may be a combination of the two (e.g., Schwartz, 1970). In any case, i t is clear that in s e r i a l learning, as in paired-associate learning, the nominal stimulus is easily identified and directly linked or identical to the effective stimulus. In a free recall task in which the subject is allowed to recall the words in any order that he desires and no cue i s presented to e l i c i t a particular response, there is no nominal stimulus. If free recall i s to be described in stimulus-response terminology, any concept of an effective stimulus must be a subject-produced stimulus. To describe a subject-3 p r o d u c e d s t i m u l u s n e c e s s i t a t e s a t h e o r y about t h e n a t u r e o f t h a t s t i m -u l u s and i t s f o r m a t i o n . T hus, t o d e s c r i b e f r e e r e c a l l w i t h o u t s u b s c r i -b i n g t o a p a r t i c u l a r t h e o r y r e q u i r e s a l a n g u a g e o t h e r than t h a t o f s t i m -u l i and r e s p o n s e s . One possible a t h e o r e t l c a l t e r m i n o l o g y w o u l d d e s c r i b e free r e c a l l as c o n s i s t i n g o f i n p u t and o u t p u t p e r i o d s , r e f e r r i n g t o p r e -s e n t a t i o n and r e c a l l p e r i o d s , r e s p e c t i v e l y . M o t h e r p o s s i b l e t e r m i n o l o g y , w h i c h i s somewhat t h e o r e t i c a l , w o u l d d e s c r i b e f r e e r e c a l l as a p r o c e s s o f s t o r i n g and r e t r i e v i n g t h e i t e m s i n the l i s t . C l o s e l y a l i g n e d t o t h e p r o b l e m o f t h e s t i m u l u s i n f r e e r e c a l l i s the p r o b l e m of o r d e r o f o u t p u t . As t h e s u b j e c t i s a l l o w e d t o r e c a l l t h e words i n any o r d e r , o r d e r o f o u t p u t i s a s u b j e c t - d e t e r m i n e d v a r i a b l e . T h i s s u b j e c t - d e t e r m i n e d v a r i a b l e w o u l d p r e s e n t l i t t l e p r o b l e m i f s u b j e c t s r e c a l l e d t h e words i n a random f a s h i o n o r i f t h e i r r e c a l l o r d e r c o r r e -sponded d i r e c t l y w i t h the p r e s e n t a t i o n o r d e r . W i t h r a n d o m l y - o r d e r e d r e -c a l l , i t c o u l d be assumed t h a t p r o p e r t i e s o f i n d i v i d u a l i t e m s c o n t r i b u t e t o t h e " s t r e n g t h s " o f t h e i t e m s and t h a t t h e " s t r o n g e s t " i t e m s a r e r e -c a l l e d f i r s t . W i t h p r e s e n t a t i o n o r d e r r e c a l l , f r e e r e c a l l might be de-s c r i b e d as a nalogous t o s e r i a l l e a r n i n g . However, o r d e r o f o u t p u t i s n o t random, n o r does i t d i r e c t l y c o r r e s p o n d t o t h e p r e s e n t a t i o n o r d e r . Thus, u n l i k e t h e o r i e s o f p a i r e d - a s s o c i a t e o r s e r i a l l e a r n i n g , a t h e o r y o f f r e e r e c a l l demands a d e s c r i p t i o n of the o r d e r o f o u t p u t . T h i s de-s c r i p t i o n may be i n d e p e n d e n t o f t h e d e s c r i p t i o n o f t h e " s t i m u l u s " i n f r e e r e c a l l > o r i t may t r e a t the p roblems o f o r d e r o f o u t p u t and t h e s t i m u l u s as one p r o b l e m . A l t h o u g h l a c k of r e s t r i c t i o n on o r d e r o f o u t p u t i n c r e a s e s t h e 4 complexity of describing free recall, i t also increases the fascination of the paradigm. Obvious and interesting questions have yet to be answered satisfactorily. Why does the subject order his output? On what basis does he order his output? How can the structure of his output order be described? These questions provide the motivation for the cur-rent investigation; however, they cannot be answered in their present form. In order to operationalize these questions, i t is worthwhile to review the free recall literature. Two reviews will be undertaken. In this chapter, the data and nsthodology concerning organization in free recall will be described. In the next chapter, theories which have been proposed to account for organization in free recall will be evaluated in terms of the data presented in this chapter. Two Types of Organization Two types of organization in free recall, clustering and subjective  organization, have been identified. The major methodological differences are that clustering is experimenter-defined and may be measured in single-t r i a l free recall, while subjective organization is subject-defined and can only be measured in multitrial free recall. These two types of organization will be discussed in separate sections. Clustering Bousfield and Sedgewick (1944) asked subjects to l i s t items belonging to specific categories, e.g., birds, and found that many se-quences of related items appeared in the subjects' listing protocols, e.g., duck and goose, hawk and eagle. Because he was unable to quantify 5 this "clustering" of related items in the listing protocols, Bousfield (1953) developed a method whereby the experimenter was able to control both the items to be listed and the relatedness of those items. He prepared a l i s t of 60 two-syllable nouns; these were 15 nouns from each of four categories: animals, vegetables, names, and occupations. The nouns were placed in a random order and read to groups of subjects at a rate of one word every three seconds. After presentation was comple-ted, the subjects were asked to l i s t , in any order that they desired, as many of the words as they could remember. This procedure i s , of course, identical to the procedure described as single-trial recall. The measurement of clustering that Bousfield used involved the number of category repetitions in a given subject's recall protocol, i.e., the number of occurrences of a word from a given category followed by a word from the same category. In order to determine an "expected" number of repetitions to compare with each subject's observed number of repetitions, Bousfield ran an " a r t i f i c i a l " experiment in which he drew capsules at random and without replacement from a bo:c containing 15 blue, 15 green, 15 orange, and 15 white capsules. This procedure was repeated for each subject in the experiment, drawing a number of cap-sules that corresponded to the number of words that subject recalled (mean recall was about 25 words) and then counting the number of colour repetitions. Thus, Bousfield was able to compare the number of repeti-tions for subjects in his "real" experiment with those of his " a r t i f i -c i a l " experiment; he found that his "real" subjects clustered more than his " a r t i f i c i a l " subjects. 6 A second a n a l y s i s i n v o l v e d measuring changes i n the c l u s t e r i n g t e n -dency during the course of r e c a l l . To do t h i s , B o u s f i e l d " V i n c e n t i z e d " the s u b j e c t s ' r e c a l l p r o t o c o l s ; that i s , he d i v i d e d each su b j e c t ' s r e -c a l l p r o t o c o l i n t o s u c c e s s i v e d e c i l e i n t e r v a l s i n terms of the order i n which the items had been r e c a l l e d . He found t h a t the c l u s t e r i n g tendency was i n i t i a l l y above chance (chance was def i n e d as the r e s u l t s of the a r -t i f i c i a l experiment), rose to a maximum a t about the f o u r t h d e c i l e , and p r o g r e s s i v e l y dropped to the chance l e v e l by the tenth d e c i l e . B o u s f i e l d ' s (1953) experiment was concerned w i t h one type of conformity of sub j e c t ' s order of output to an a p r i o r i s t r u c t u r e of the l i s t to be r e c a l l e d , i . e . , the c l u s t e r i n g of members of a taxonomic category. Since B o u s f i e l d ' s o r i g i n a l experiment, a v a r i e t y of other types o f c l u s t e r i n g has been reported. For example, Jenkins and R u s s e l l (1952) found c l u s t e r i n g of d i r e c t associates as measured by word asso-c i a t i o n norms, and Tu l v i n g (1962a) found c l u s t e r i n g o f words having the same i n i t i a l l e t t e r . There i s no need to review a l l experiments i d e n t i f y i n g new types of c l u s t e r i n g as both B o u s f i e l d and B o u s f i e l d (1966) and S h u e l l (1969) have made p a r t i a l l i s t i n g s of such experiments. One f u r t h e r p o i n t should be made. Most experiments designed to provide evidence f o r c l u s t e r i n g i n accordance w i t h an a p r i o r i s t r u c t u r e have used a s i n g l e - t r i a l f r e e r e c a l l paradigm. However, i n some cases, c l u s t e r i n g has been examined as a f u n c t i o n o f t r i a l s i n a m u l t i t r i a l f r e e r e c a l l experiment (e.g., P u f f , 1970). The t y p i c a l f i n d i n g i s that c l u s t e r i n g i n c r e a s e s as a f u n c t i o n of t r i a l s and that the c o r r e l a t i o n between c l u s t e r i n g scores and r e c a l l scores over t r i a l s i s very h i g h . 7 Subjective Organization There i s substantial evidence that subjects' re c a l l protocols conform with an a p r i o r i structure of the l i s t to be recalled. Do sub-jects' r e c a l l protocols also show organization when the l i s t to be re-called does not have an a p r i o r i structure? Tulving (1962b) attempted to observe organization of a l i s t of "unrelated" words, words chosen so that there are no systematic intra-l i s t relationships apparent to the experimenter* He reasoned that in a l i s t of unrelated words, there was no method for determining the degree of organization in s i n g l e - t r i a l free r e c a l l , so Tulving used mu l t i t r i a l free recall i n order to observe the organization of a given subject. Organization scores were defined i n terms of the number of t r i a l s on which the subject recalled word x_ followed by word y_. That i s , Tulving*s measure of organization involved the number of t r i a l s on which two words appeared adjacently in the subject's recall protocols. This measure Tulving termed subjective organization,with subjective Indicating that the more frequently occurring adjacencies may differ from subject to subject. Tulving used a 16-word l i s t and 16 presentation t r i a l s . The words were presented at a rate of one word per second, and after each presen-tation, the subjects were allowed 90 seconds to record their r e c a l l . As did Bousfield, Tulving used a r t i f i c i a l subjects whose performance was matched with his real subjects; these a r t i f i c i a l subjects provided a baseline for determining the degree of organization used by his real subjects. 8 Tulving's experiment yielded four main results. First, overall organization scores for his real subjects were far greater than for his art i f i c i a l subjects. Second, organization scores and recall scores in-creased over trials. The correlation between the mean recall and log mean organization scores as a function of trials was .96. Third, total organization scores and total recall scores over subjects ware found to be significantly correlated, r_ = .63. That i s , subjects who showed high levels of organization tended to recall more than did subjects who showed low levels of organization. Fourth, there was a certain degree of commonality among subjects in the more frequently occurring adjacen-cies. This commonality was shown to increase as a function of trials. Tulving's results have frequently been replicated, even in experi-ments in which different measures of organization have been used (e.g., Bousfield, Puff, & Cowan, 1964). It seems safe to conclude that corre-lation between organization measures and recall performance have a high degree of generality. At this point, two types of organization have been discussed. For the remainder of this paper, organization of a l i s t of words with an ji priori structure will be referred to as clustering, and organization of a l i s t of words with no a priori structure w i l l be referred to as sub- jective organization. Measurement of Organization Until this section, organization has been defined on a fairly abstract level. However, operational definitions of organization do 9 exist in terms of the indices that have been proposed for both clustering and subjective organization. In this section, the indices that are used in the experiments to' be reported will be discussed. For a review of other indices, the reader is referred to recent papers by Shuall (1969) and Frankel and Cole (1971). Measurement of Clustering For measurement of clustering, Frankel and Cole's (1971) adaptation of the multiple-category runs test was used. The important statistic for this test is the number of runs of items from the same category that appears in a subject's recall protocol. For example, a hypothetical sub-ject recalls eight items in the following order: vegetable, vegetable, vegetable, animal, occupation, occupation, vegetable, name. The number of runs in this string is five, i.e., one more than the number of times that the category of the recalled item was changed. Wallis and Roberts (1957) have presented formulae for the expected mean and variance of the multiple-category runs test, and Frankel and Cole have described these formulae In terms of category clustering. As the me?n number of runs is normally distributed, knowledge of the expected mean and variance and the observed mean can lead to the calculation of a JZ score. When a z_ statistic indicates that the number of observed runs was significantly fewer than the number of expected runs, evidence for clustering can be said to have been found. Frankel and Cole recommend that the z_ score be used as a measure of the magnitude of clustering because a _z score is inversely related to the probability that the observed clustering is due to chance. 10 The r u n s measure p r e s e n t s t h e f o l l o w i n g advantages ( F r a n k e l & C o l e , 1 9 7 1 ) : 1. the d i s t r i b u t i o n o f t h e s t a t i s t i c i s known, so t h a t f o r a g i v e n s u b j e c t , s i g n i f i c a n c e o f c l u s t e r i n g can be d e t e r m i n e d ; 2. the d i s t r i b u t i o n o f a s e t o f JZ s c o r e s i s known, so t h a t f o r a group o f s u b j e c t s , t h e s i g n i f i c a n c e o f c l u s t e r i n g can be d e t e r m i n e d (by t4T x where n i s t h e number o f s u b j e c t s ) ; 3. t h e s t a t i s t i c i s i n d e p e n d e n t o f the number o f i t e m s r e c a l l e d , so t h a t any o b s e r v e d r e l a t i o n s h i p between z_ s c o r e s and r e c a l l p e r f o r m a n c e w i l l n o t r e s u l t f r o m an a r t i f a c t o f t h e c l u s t e r i n g measure; 4. t h e c a t e g o r i c a l c o m p o s i t i o n o f t h e r e c a l l p r o t o c o l i s i n c o r p o r a t e d i n t o t h e d e t e r m i n a t i o n o f t h e e x p e c t e d mean and v a r i a n c e , so t h a t a g i v e n s u b j e c t ' s z_ s c o r e i s i n d e p e n d e n t o f t h e p a r t i c u l a r i t e m s t h a t he r e c a l l s . F r a n k e l and C o l e have r e v i e w e d o t h e r measures o f c l u s t e r i n g and f o und t h a t t h e s t a t i s t i c i s t h e o n l y measure w h i c h e x h i b i t s a l l f o u r o f the p r e c e d i n g a d v a n t a g e s . Measurement o f S u b j e c t i v e O r g a n i z a t i o n As S h u e l l (1969) has p o i n t e d o u t , v a r i o u s i n d i c e s o f s u b j e c t i v e o r g a n i z a t i o n a r e much the same? t h e y a l l i n v o l v e t h e number o f i n t e r t r i a l r e p e t i t i o n s and d i f f e r m a i n l y i n t h e i r c o r r e c t i o n s f o r chance o r g a n i z a -t i o n . As a l l o f t h e measures a r e h i g h l y c o r r e l a t e d and, u n l i k e t h e runs t e s t , none o f t h e measures has a d i s t r i b u t i o n w h i c h i s known, c h o i c e o f a measure f o r t h e p r e s e n t e x p e r i m e n t s was somewhat a r b i t r a r y . The B o u s f i e l d and B o u s f i e l d (1966) d e v i a t i o n measure was c h osen b e c a u s e i t 11 i s the most frequently used measure of subjective organization and because a computer programme was av a i l a b l e f o r i t s c a l c u l a t i o n . For measurement of subjective organization, Bou3field and Bousfield defined an observed i n t e r t r i a l r e p e t i t i o n (OITR) as the case where word y_ follows word x on both t r i a l n and t r i a l n + 1 . This measure can be ex-tended by def i n i n g an OITR as the case where word y_ follows word x on t r i a l n and e i t h e r word y_ follows word x or word x follows word y_ on trial n + 1 . I t i s t h i s extansion to a faro-way index that w i l l be used i n the present paper. Bou s f i e l d and Bousfield also present formulae f or determining the expected number of i n t e r t r i a l r e p e t i t i o n s (EITR). The formulae involve the following parameters: (1) h, the number of items r e c a l l e d on t r i a l n; (2) k, the number of items r e c a l l e d on t r i a l n + 1 ; and (3) c^  the number of items ttfhich were r e c a l l e d on t r i a l n and also r e c a l l e d on t r i a l n + 1 . Given an observed and an expected number of i n t e r t r i a l r e p e t i t i o n s , the measure of organization i s simply, OITR-EITR. I t i s apparent that this measure involves only the deviation of the observed from the ex-pected value and, unlike the runs t e s t , does not take into account the expected variance. Thus, no test f o r the s i g n i f i c a n c e of organization of a s i n g l e protocol e x i s t s , and the r e s u l t i n g values may not be com-parable for l i s t s of d i f f e r e n t lengths. Comparisons of the Two Types of Organization As measures of c l u s t e r i n g are not equivalent to measures of sub-j e c t i v e organization, d i r e c t comparisons of the two types of organization are d i f f i c u l t . The f a i l u r e to have equivalent measures re s u l t s from the 12 f a c t t h a t g i v e n t h e r e c a l l o f word J C , a d j a c e n t r e c a l l o f o n l y one p a r t i -c u l a r word ( t h e word a d j a c e n t t o word x on t h e p r e c e d i n g t r i a l ) w i l l l e a d t o an i n c r e m e n t o f a s u b j e c t i v e o r g a n i z a t i o n s c o r e , w h i l e a d j a c e n t r e c a l l o f a number o f words (words w h i c h a r e members o f t h e same c a t e g o -r y as word x) w i l l l e a d t o an i n c r e m e n t o f a c l u s t e r i n g s c o r e . One a t t e m p t t o compare t h e two t y p e s o f o r g a n i z a t i o n was r e p o r t e d by P u f f ( 1 9 7 0 a ) . P u f f p r e s e n t e d s u b j e c t s a l i s t o f e i t h e r 18 u n r e l a t e d words o r a l i s t c o n t a i n i n g s i x i t e m s f r o m each o f t h r e e t axonomic c a t e -g o r i e s ; t h e words were p r e s e n t e d f o r f i f t e e n t r i a l s a t a r a t e o f 2.5 seconds p e r w o r d , and one m i n u t e was a l l o w e d f o r each r e c a l l p e r i o d . A l t h o u g h s y l l a b l e l e n g t h and T h o r n d i k e - L o r g e (1944) f r e q u e n c y were s i m i l a r f o r b o t h l i s t s , P u f f f ound t h a t t h e c a t e g o r i z e d l i s t was e a s i e r t o l e a r n t h a n t h e u n r e l a t e d l i s t . P u f f ' s c o m p a r i s o n s o f t h e two t y p e s o f o r g a n i z a t i o n i n v o l v e d t h e e x p r e s s i o n o f b o t h t y p e s as p r o p o r t i o n s o f the maximum p o s s i b l e o r g a n i -z a t i o n s c o r e . H i s f i n d i n g s showed t h a t c l u s t e r i n g s c o r e s i n the c a t e g o -r i z e d l i s t were f a r g r e a t e r t h a n s u b j e c t i v e o r g a n i z a t i o n s c o r e s o f t h a t l i s t and t h a t s u b j e c t i v e o r g a n i z a t i o n s c o r e s f o r t h e c a t e g o r i z e d l i s t were n o t s i g n i f i c a n t l y d i f f e r e n t from s u b j e c t i v e o r g a n i z a t i o n s c o r e s f o r t h e u n r e l a t e d l i s t . These f i n d i n g s a r e open t o t h r e e p o s s i b l e i n -t e r p r e t a t i o n s . One i n t e r p r e t a t i o n i s t h a t c l u s t e r i n g i s a s t r o n g e r phenomenon t h a n s u b j e c t i v e o r g a n i z a t i o n . A second i n t e r p r e t a t i o n i s t h a t t h e r e i s as much s u b j e c t i v e o r g a n i z a t i o n i n a l i s t o f r e l a t e d words as t h e r e i s i n a l i s t o f u n r e l a t e d words. A t h i r d i n t e r p r e t a t i o n i s t h a t s u b j e c t i v e u n i t s c o n t a i n more t h a n two x;ords and t h a t t h e r e i s 13 l i t t l e organization within a subjective u n i t . The f a c t that none or a l l of these i n t e r p r e t a t i o n s may be correct points to the d i f f i c u l t y involved i n d i r e c t l y comparing the two types of organization. Perhaps the only r e l i a b l e method of comparing c l u s t e r i n g (of categorized words) to subjective organization (of unrelated words) would be to compare the e f f e c t s of manipulations of independent variables on the r e c a l l of categorized and unrelated l i s t s . Four such comparisons follow. Order of Input A number of studies have shown that when a categorized l i s t i s presented i n a blocked fashion, so that a l l members of one category are presented before a l l members of another category, r e c a l l i s superior to random presentation (e.g., Ueingartner, 1964; Cofer, Bruce, & Reicher, 1966). Of course, no d i r e c t analogue of blocked presentation can be found for the case of an unrelated l i s t . However, Tulving (1965), c a p i t a l i z i n g on the relatedness of h i s "unrelated" words, has shown that presentation order can a f f e c t r e c a l l performance of an uncategorized l i s t . He arranged two presentation orders of a l i s t used i n a previous i n v e s t i g a t i o n (Tulving, 1962b). In one of the presentation orders, ad-jacent pairs had been frequently r e c a l l e d adjacently i n h i s e a r l i e r study. In the other presentation order, adjacent pairs had been i n f r e -quently r e c a l l e d adjacently i n h i s e a r l i e r study. M u l t i t r i a l free r e c a l l was used, and subjects were run to a c r i t e r i o n of one perfect r e c i t a t i o n . Median t r i a l s to c r i t e r i o n f o r the group with frequently occurring ad-jacencies was 3.4, and median t r i a l s to c r i t e r i o n f or the group with 14 i n f r e q u e n t l y o c c u r r i n g a d j a c e n c i e s xvas 7.05 t h e d i f f e r e n c e was s t a t i s t i -c a l l y s i g n i f i c a n t . Thus, p r e s e n t a t i o n o r d e r seems t o a f f e c t the amount r e c a l l e d i n l i s t s o f words w i t h an a p r i o r i s t r u c t u r e and l i s t s o f words w i t h no SL p r i o r i s t r u c t u r e ; i n b o t h t y p e s o f l i s t s , a d j a c e n c i e s i n p r e s e n t a t i o n o r d e r can be a r r a n g e d so t h a t r e c a l l p e r f o r m a n c e i s i m p r o v e d . P r e s e n t a t i o n R ate C o f e r , B r u c e , and R e i c h e r (1966) have shown t h a t when t h e s t i m u l i a r e c a t e g o r i z e d w o r d s , amount r e c a l l e d and c l u s t e r i n g i n c r e a s e w i t h s l o w e r p r e s e n t a t i o n r a t e s . S h a p i r o and Ponce ( 1 9 7 0 ) , i n an e x p e r i m e n t u s i n g m u l t i t r i a l f r e e r e c a l l o f a l i s t o f u n r e l a t e d w o r d s , f o u n d g r e a t e r s u b j e c t i v e o r g a n i z a t i o n s c o r e s and i n c r e a s e d r e c a l l p e r f o r m a n c e w i t h s l o w e r p r e s e n t a t i o n r a t e s . Thus, f o r b o t h a l i s t o f c a t e g o r i z e d words and a l i s t o f u n r e l a t e d w ords, o r g a n i z a t i o n measures and r e c a l l p e r f o r m -ance a r e r e l a t e d t o p r e s e n t a t i o n r a t e . C u e i n g A f t e r having been p r e s e n t e d a l i s t o f c a t e g o r i z e d w o r d s , s u b j e c t s may be g i v e n t h e c a t e g o r y names as r e c a l l a i d s . T h i s p r o c e d u r e , c u e i n g , has sometimes been found t o improve r e c a l l p e r f o r m a n c e . When measures a r e t a k e n t o i n s u r e t h a t s u b j e c t s use the c a t e g o r y s t r u c t u r e , by p r o v i -d i n g t h e c a t e g o r y names a l o n g s i d e t h e words d u r i n g p r e s e n t a t i o n ( T u l v i n g & P e a r l s t o n e , 1966) o r by u s i n g b l o c k e d c a t e g o r y p r e s e n t a t i o n (Wood, 1969), c u e i n g has a l m o s t always been found t o improve r e c a l l p e r f o r m a n c e . T u l v i n g and P e a r l s t o n e (1966) p r o v i d e d c a t e g o r y names a t p r e s e n t a t i o n and, 15 f o r subjects i n the "cueing" groups, also at r e c a l l . They used l i s t s of 12, 24, and 48 words i n length and each l i s t contained 1, 2, or 4 words per category. In a l l but one condition, cueing was found to s i g -n i f i c a n t l y increase r e c a l l and even i n that condition (a 12-item l i s t with 4 words per category), cueing resulted i n s l i g h t l y b e t t e r r e c a l l performance. R e c a l l of categorized l i s t s may be analyzed i n t o two components, r e c a l l of a category and r e c a l l of words within a category. Tulving and Pearlstone reported data on both of these components, and they used r e c a l l of one or more words from a category as t h e i r d e f i n i t i o n of category r e c a l l . Subjects i n cueing conditions r e c a l l e d s i g n i f i c a n t l y more categories than did subjects i n control conditions. However, once a category was r e c a l l e d , the number of items r e c a l l e d w i t h i n that cate-gory was not d i f f e r e n t f or cued and control subjects. Thus the apparent e f f e c t of cueing i s to allow subjects to r e c a l l more categories. To examine the e f f e c t of cueing of subject-determined categories, Dong & K i r t s c h (1968) used a v a r i a t i o n of a paradigm f i r s t used by Handler (Mandler, 1967; Mandler & Pearlstone, 1965). In t h e i r experi-ment, Dong and Kintsch had each subject sort three l i s t s , each contai-ning 25 randomly-chosen words, For each l i s t , the subject sorted the words i n t o between two and seven categories; number of categories and basis of category organization were subject-determined. Subjects sorted each l i s t u n t i l they reached a c r i t e r i o n of two consecutive i d e n t i c a l s o r t s . A f t e r having reached the s o r t i n g c r i t e r i o n on a given l i s t , subjects i n two groups were required to provide names for t h e i r categories. 16 A f t e r having provided names for t h e i r t h i r d l i s t , subjects were asked to r e c a l l as many of the words from the three l i s t s as they could. During r e c a l l , subjects i n the cueing (Group RC) condition were provided with t h e i r category names, while subjects i n the control (Group NC) condition were not provided with t h e i r category names. The r e s u l t s i n -dicated that cued subjects r e c a l l e d more than did control subjects. As Tulving and Pearlstone found f o r categorized l i s t s , the advantage of cueing was i n the number of categories r e c a l l e d ; cued subjects r e c a l l e d more categories than did control subjects but did not r e c a l l more words per category. Thus, cueing seems to have the same e f f e c t on both ex-perimenter-determined and subject-determined categories. Number of Categories Examinations of the e f f e c t of number of categories of a l i s t with an a p r i o r i structure on free r e c a l l performance presents two problems. F i r s t , as was the case with cueings i t i s important that the subject use the category structure provided by the experimenter. Second, i n l i s t s of a f i x e d length, the number of items per category i s inv e r s e l y r e l a t e d to the number of categories. This perfect confounding of items per category and number of categories presents many d i f f i c u l t i e s i n i n t e r -p r e t a t i o n . Tulving and Pearlstone's (1966) f i n d i n g that the advantage of cued r e c a l l i s to increase the number of categories r e c a l l e d but not to increase r e c a l l w i t h i n a category suggests that when number of Items per category i s held constant, the more categories the subject r e c a l l s , the more words he w i l l r e c a l l . Thus,number of categories would appear to be 17 an important determinant of r e c a l l performance. In Tulving and P e a r l -stone's experiment, subjects who were presented l i s t s of 48, as opposed to 24, words did not d i f f e r i n items per category r e c a l l e d but did d i f f e r i n r e c a l l performance. Obviously, the advantage of subjects r e c e i v i n g the longer l i s t x<ras i n the number of categories r e c a l l e d . Within a l i s t of a given length, under conditions of cueing, Tulving and Pearlstone found that subjects who were presented more categories r e c a l l e d more words. Again, number of categories appears to be an important deter-minant of r e c a l l performance. However, both of Tulving and Pearlstone's findings i n v o l v i n g num-ber of categories are questionable. F i r s t , although i t i s i n t e r e s t i n g that r e c a l l within categories did not d i f f e r between 24- and 48-word l i s t s , i t i s not i n the le a s t s u r p r i s i n g that subjects who were presented the 48-word l i s t r e c a l l e d more. Furthermore, i n the Tulving and P e a r l -stone experiment, subjects who were presented 12-word l i s t s r e c a l l e d more items per category than did subjects who were presented 24- or 48-word l i s t s . Second, i n the control conditions i n which no cue was presented during r e c a l l , the r e l a t i o n s h i p between number of categories and r e c a l l performance was not a d i r e c t function. In f a c t , Tulving and Pearlstone's data i n d i c a t e that, i n the control conditions, r e c a l l performance may have been an inverse function of the number of catego-r i e s . The exact r e l a t i o n s h i p between number of categories and r e c a l l performance appears to be a function of l i s t length. Also, i t seems as i f the control conditions may be more representative of free r e c a l l than are the cueing conditions. For instance, i n the 24-item l i s t , cueing 18 r e s u l t e d i n highest r e c a l l when there was only one item per category. This, of course, would be an unrelated l i s t , and r e c a l l of an unrelated l i s t has often been shown to be worse than r e c a l l of a categorized l i s t (e.g., Tulving and Pearlstone's 24-item l i s t s containing 12 or 6 cate-g o r i e s ) . Thus, although i t appears that the number of categories has an e f f e c t on r e c a l l performance, the e f f e c t i s confounded with l i s t length, items per category, and cueing. Mandler (1967) has presented an extension of the e f f e c t of number of categories to a l i s t of unrelated words. In a t y p i c a l experiment, subjects sort a l i s t of e i t h e r 52 or 100 words in t o categories; they are usually allowed to choose the number of categories they use with the r e s t r i c t i o n that they use no fewer than two and no more than seven cate-gories. The s o r t i n g process i s continued u n t i l stable categorization i s reached. Then, the subjects are asked to r e c a l l as many of the words as they can. Generally, a l i n e a r r e l a t i o n s h i p between number of categories used and amount r e c a l l e d i s obtained. Mandler's r e s u l t s suggest that number of subjective categories i s d i r e c t l y r e l a t e d to r e c a l l performance. However, there are two possible methodological a r t i f a c t s i n Mandler's paradigm. F i r s t , s o r t i n g i s s e l f -paced and, as i t i s continued to c r i t e r i o n , number of s o r t i n g t r i a l s i s not experimentally c o n t r o l l e d . Second, as subjects are allowed to choose the number of categories they use, number of categories i s not experimen-t a l l y c o n t r o l l e d . Mandler (1968) has stated that he has c o n t r o l l e d f o r these possible a r t i f a c t s . F i r s t , he has shown that the category— r e c a l l r e l a t i o n s h i p i s obtained when s o r t i n g i s experimenter-paced and continued for a f i x e d number of t r i a l s (Mandler, 1967, Experiment H). Second, he 19 has shown that when subjects are assigned a number of categories i n a between-sul?jects (Handler, 1967, Experiment B) or a within-subjects (Mandler, 1968) design, the r e l a t i o n s h i p i s s t i l l found. However, Mandler did not assign number of categories to h i s subjects i n h i s ex-perimenter-paced task, nor did he con t r o l time per t r i a l and number of t r i a l s i n the studies i n which subjects were assigned number of categories. Schwartz and Humphreys (1972a) have assigned subjects number of categories i n an experimenter-paced task, and they f a i l e d to f i n d a s i g -n i f i c a n t r e l a t i o n s h i p between number of categories and r e c a l l performance. In a d d i t i o n , both Schwartz and Humphreys (1972a) and Nelson, McRae, and Sturges (1970) have found only small r e l a t i o n s h i p s between number of cate-gories and r e c a l l performance i n experimenter-paced tasks i n which sub-j e c t s were allowed to choose t h e i r number of categories. Thus, Mandler's r e s u l t s , much l i k e Tulving and Pearlstone's, can be severely questioned on methodological grounds. I t must be concluded that both the number of taxonomic and the number of subjective categories can influence re-c a l l performance, but i n neither case, can the d i r e c t i o n or amount of influence be c l e a r l y predicted. Miscellaneous In the previous sections of t h i s review, two types of organization were defined, t h e i r measurement was discussed, and the two types were compared. For purposes of this paper, there remains three other areas for review. Unfortunately, these do not f i t quite so n i c e l y into the previous format. The three sections include (1) i n h i b i t i o n , (2) i n d i v i d u a l d i f f e r e n c e s , and (3) short- and long-term memory components of r e c a l l . 20 Inhibition Until this section, the review has only dealt with studies i n which subjects have learned one l i s t . However, as with other types of learning paradigms, free recall has been studied as a function of prece-ding (proactive inhibition) and intervening (retroactive inhibition) l i s t s . Proactive inhibition. In examining the effects of proactive inhibition, this review considers only the case in which some or a l l of the words from the f i r s t l i s t are contained in the second l i s t . Three types of proactive inhibition paradigms include an overlap of f i r s t - and second-l i s t items. In part-whole free r e c a l l , a l l of the f i r s t - l i s t items plus some additional items are contained in the second l i s t . In whole-part free r e c a l l , a l l of the second-list items are contained i n the f i r s t l i s t , but not a l l of the f i r s t - l i s t items are contained i n the second l i s t . In whole-whole free r e c a l l , the two l i s t s are identical. For a l l three paradigms, the control group learns two l i s t s which have no overlap of items. Tulving (1966), in two part-whole experiments, reported three main findings: (1) there was no difference between experimental and control subjects in total second-list r e c a l l : (2) there was an inter-action between groups and t r i a l s on the second l i s t , such that control subjects learned the second l i s t at a faster rate than experimental sub-jects; and (3) on later t r i a l s of second-list learning, control subjects were performing slightly, but not significantly, better than experimental subjects. Tulving's results have been replicated by Novinski (1969) and 21 Bower and L e s g o l d ( 1 9 6 9 ) . , S i n c e T u l v i n g ' s (1966) o r i g i n a l e x p e r i m e n t , a number o f f a c t o r s have been shown t o l e a d t o i m p r o v e d x v h o l e - l i s t p e r f o r m a n c e i n t h e p a r t -whole p a r a d i g m . F i r s t , Wood and C l a r k (1968) have shown t h a t s u b j e c t s ' w h o l e - l i s t p e r f o r m a n c e i n c r e a s e s when t h e y a r e i n s t r u c t e d about t h e r e l a t i o n s h i p between t h e p a r t l i s t and t h e w h ole l i s t . Second, O r n s t e i n (1970) has shown t h a t b l o c k i n g p a r t - l i s t and new words d u r i n g p r e s e n t a -t i o n l e a d s t o b e t t e r w h o l e - l i s t p e r f o r m a n c e . T h i r d , B irnbaum (1968) has shown t h a t w h o l e - l i s t p e r f o r m a n c e i s i m p r o v e d when th e p a r t l i s t i s composed o f c a t e g o r i e s w h i c h a r e t a k e n i n t a c t f r o m t h e whole l i s t . F o u r t h , S c h w a r t z and Humphreys (1972b) have shown t h a t when s u b j e c t i v e o r g a n i z a t i o n i s measured on t h e p a r t l i s t , h i g h o r g a n i z e r s r e c a l l more w h o l e - l i s t i t e m s t h a n do low o r g a n i z e r s . S c h w a r t z and Humphreys have p o i n t e d out t h a t i n a l l f o u r o f t h e p r e c e d i n g s t u d i e s , i m p r o v e d w h o l e -l i s t r e c a l l p e r f o r m a n c e was c o r r e l a t e d w i t h an i n c r e a s e d tendency t o c l u s t e r p a r t - l i s t words d u r i n g w h o l e - l i s t l e a r n i n g . The d a t a on b o t h w h o l e - p a r t and w h ole-whole f r e e r e c a l l a r e s i m i l a r t o t h o s e on p a r t - w h o l e f r e e r e c a l l . T u l v i n g and O s i e r ( 1 9 6 7 ) , i n a W h o l e - p a r t f r e e r e c a l l e x p e r i m e n t , found t h a t e x p e r i m e n t a l s u b j e c t s p e r f o r m w o r s e , i n terms o f t o t a l r e c a l l , t h a n do s u b j e c t s i n an a p p r o p r i -a t e c o n t r o l g roup. A l t h o u g h the e x p e r i m e n t a l s u b j e c t s have an i n i t i a l a d v a n t a g e , t h e y a r e s h o r t l y ( i n terms o f t r i a l s ) a t a d i s a d v a n t a g e w h i c h remains t h r o u g h o u t s e c o n d - l i s t l e a r n i n g . N o v i n s k i (1969) has r e p l i c a t e d t h e s e f i n d i n g s . T u l v i n g (1966) gave e x p e r i m e n t a l s u b j e c t s i n c i d e n t a l l e a r n i n g o f a l i s t o f words w h i l e c o n t r o l s u b j e c t s p e r f o r m e d an u n r e l a t e d 22 task. Both groups were then given i n t e n t i o n a l learning of the same l i s t . Tulving found that the experimental subjects had no advantage i n terms of t o t a l r e c a l l , during i n t e n t i o n a l learning. In f a c t , Tulving's graphs for experimental and control group performance suggest the pre-sence of a cross-over e f f e c t . That i s , i n c i d e n t a l learners had a small advantage on the f i r s t t r i a l , and control subjects had a small advantage on the l a t e r t r i a l s . Tulving did not report any s t a t i s t i c s concerning h i s Groups X T r i a l s i n t e r a c t i o n . However, Schwartz and Humphreys (1972b) gave subjects two t r i a l s of i n t e n t i o n a l learning on a l i s t of unrelated words, then ten t r i a l s of i n t e n t i o n a l learning on the same l i s t . Sub-j e c t s i n the experimental group had a s i g n i f i c a n t f i r s t - t r i a l advantage, but did not r e c a l l s i g n i f i c a n t l y more words over the ten t r i a l s than the c o n t r o l groups. Also, the Groups X T r i a l s i n t e r a c t i o n was s i g n i f i c a n t , and c o n t r o l subjects had a s l i g h t , but non-significant, advantage on the l a t e r t r i a l s . In summary, the xrork on proactive i n h i b i t i o n v/ith overlap between f i r s t - and s econd-list words appears to be quite s i m i l a r . Learning of a l i s t containing elements of a previously-learned l i s t gives subjects an i n i t i a l advantage. However, c o n t r o l groups which have no overlap of items i n the two l i s t s generally r e c a l l s l i g h t l y more than the experi-mental groups on l a t e r t r i a l s . Usually, there i s no difference between experimental and c o n t r o l groups i n t o t a l r e c a l l . Retroactive I n h i b i t i o n . Retroactive i n h i b i t i o n i n free r e c a l l has been less frequently studied than has proactive i n h i b i t i o n . However, three 23 recent studies (Shuell, 1968; Wood, 1970; Tulving & Psotka, 1971) have examined retroactive inhibition* Shuell (1968) had experimental sub-jects learn tx*o categorized l i s t s . For the "same" experimental condi-tion, both lists contained the same conceptual categories; for the "dif-ferent" experimental conditions, the two lists contained different con-ceptual categories. Control subjects learned only one l i s t . First-list recall performance, after interpolated second-list learning, was the main independent variable. The results indicated that control subjects recalled more than did subjects in both experimental groups, and "dif-ferent" experimental subjects recalled more than "same" experimental subjects. The superior performance of control as compared to experi-mental subjects was due to both a greater loss of category recall and loss of items within categories for the experimental subjects. The superior performance of "different" experimental as compared to "same" experimental subjects was due to a greater loss of category recall, but not items within categories, for the "same" experimental subjects. Wood (1970), in three experiments, manipulated the similarity of organization between the two lists by instructing subjects how to organize the lis t s . He found greater retroactive effects when his instructions in-creased the similarity of organization between the two list s . Tulving and Psotka (1971) studied retroactive inhibition by having subjects learn and subsequently recall from one to six categorized list s . In a l l instances, each l i s t contained different categories than any of the preceding lis t s . They found: (1) that the number of words recalled was an inverse function of the number of intervening lists; 24 (2) that most of the loss as a function of intervening l i s t s was due to loss of category r e c a l l — contrary to Shuell's (1968) fin d i n g s , there was no apparent e f f e c t of intervening l i s t s on r e c a l l within categories; and (3) that cueing with category names eliminated r e t r o a c t i v e e f f e c t s . The three studies on r e t r o a c t i v e i n h i b i t i o n suggest the following generalizations: (1) that the r e t r o a c t i v e e f f e c t may be on the number of categories r e c a l l e d rather than on the number of words r e c a l l e d from wit h i n a category: (2) that r e t r o a c t i o n i s a d i r e c t function of the s i m i l a r i t y of organization between the l i s t s ; and (3) that r e t r o a c t i o n increases as a function of the number of intervening l i s t s . I n dividual Differences There i s evidence that differences between subjects i n organiza-t i o n scores are consistent across l i s t s . Gorfein, B l a i r , and Rowland (1968) had subjects learn four l i s t s , each of a d i f f e r e n t type of stim-ulus material — consonant trigrams, nonsense s y l l a b l e s , unrelated words, and categorized words. They then correlated the subjects' organization scores between each of the four l i s t s to produce s i x c o r r e l a t i o n s . A l l of the c o r r e l a t i o n s were p o s i t i v e and the magnitude of a given c o e f f i -cient was a d i r e c t function of the s i m i l a r i t y between the stimulus mate-r i a l s . In an unpublished study, the present author had nine subjects learn two l i s t s of unrelated words; the type of stimulus material was f e l t to be more s i m i l a r to the type used i n studies of subjective organ-i z a t i o n than was the material used by Gorfein et a l . The rank-order c o r r e l a t i o n between organization scores on each of the two l i s t s was s i g -n i f i c a n t , rho = .62, p_ < .05, one-tailed. 25 As differences i n organization scores are consistent, these scores may represent i n d i v i d u a l differences i n verbal learning a b i l i t y . Earhard (Earhard, 1967; Earhard & Endicott, 1969) has shown that sub-j e c t i v e organization scores can be used to p r e d i c t performance on both s e r i a l and paired-associate l i s t s . The general procedure used by Earhard i s to have subjects learn a free r e c a l l l i s t , measure organiza-t i o n scores on that l i s t , and i d e n t i f y groups of high and low organizers. Then, the subjects learn a s e r i a l or paired-associate l i s t . Earhard's general f i n d i n g i s that high organizers learn the second l i s t f a s t e r than do low organizers. Gorfein (1971) has ref i n e d Earhard's findings by showing that when performance on the free r e c a l l l i s t i s c o n t r o l l e d by p a r t i a l c o r r e l a t i o n procedures, organization scores do not correlate with paired-associate performance. When organization scores are con-t r o l l e d i n the same way, free r e c a l l performance s i g n i f i c a n t l y Correlates with paired-associate performance. From these r e s u l t s i t can be con-cluded that organization scores predict i n d i v i d u a l differences i n other verbal learning tasks only to the extent that they r e f l e c t differences i n free r e c a l l performance. Schwartz and Humphreys (1972a) have shown that the number of cate-gories a subject uses may also be a measure o^ : i n d i v i d u a l differences i n verbal learning a b i l i t y . They showed that across s o r t i n g tasks, subjects were consistent i n the number of categories they used. Then, they found some evidence for a r e l a t i o n s h i p between the number of categories a sub-j e c t chooses to use on one s o r t i n g task and the number of words he r e -c a l l s from a completely d i f f e r e n t task. 26 In summaryj the data concerning individual differences i n sub-jective organization are far from conclusive. However, the data do warrant that extreme care be taken when drawing causal inferences from observed relationships between organization scores and recall performance. Short- and Long-Term Memory Components Single-trial free r e c a l l generally yields a bow-shaped se r i a l position curve with greater recency than primacy effects. Recent studies have provided plausible explanations f o r the b e t t e r recall of items presented at the ends of the l i s t . The recency effect has been identi-fied as a short-term memory component, i n that i t is affected by manipu-lations which have been shown to affect r e c a l l in the Brown-Peterson (Brown, 1954; Peterson & Peterson, 1959) paradigm. For instance, Kintsch and Buschke (1969) have shown that manipulations of acoustic, but not semantic, similarity affect the recency portion of the ser i a l position curve. Glanzer and Cunitz (1966) have shown that 30 seconds of interpolated activity can eliminate the recency effect. As Kintsch (1970) has indicated, the recency portion of recall may be independent cf organization. He has used interpolated activity f o l -lowing l i s t presentation in studies in which he desired to examine the structure of organization. Following Kintsch (1970), some of the ex-periments reported in the present paper a l s o used interpolated activity following l i s t presentation. Explanation of the primacy effect has been more elusive than ex-planation of the recency effect; however, Rundus and Atkinson (1970) have presented evidence that the primacy portion may be due to the 27 extra opportunity f o r rehearsal of the e a r l i e r items. They had subjects rehearse aloud during presentation of the l i s t to be r e c a l l e d and found a s t r i k i n g p a r a l l e l between number of rehearsals and p r o b a b i l i t y of r e -c a l l , except f o r the recency portion of the l i s t . Thus, there may be two components to s i n g l e - t r i a l free r e c a l l . The recency portion represents a short-term memory component which may be eliminated by s u f f i c i e n t i n t e r p o l a t e d a c t i v i t y . The primacy and middle portions probably represent a long-term memory component i n which the p r o b a b i l i t y of r e c a l l i s rel a t e d to rehearsal a c t i v i t y . Summary Two v a r i a t i o n s of the free r e c a l l paradigm, s i n g l e - t r i a l and m u l t i -t r i a l free r e c a l l , have been described. Two important features of free r e c a l l were i d e n t i f i e d . F i r s t , there i s no nominal stimulus i n free r e -c a l l and second, organization may be observed i n subject's r e c a l l protocols. Two types of organization, c l u s t e r i n g and subjective organization, have been described. Clustering was defined as conformity of r e c a l l out-put with an a p r i o r i s t r u c t u r e , while subjective organization was defined as the t r i a l - t o - t r i a l r e p e t i t i o n of adjacencies i n the output of unrelated words. Clustering and subjective organization have been compared as a function of v a r i a t i o n s of presentation order, presentation rate, cueing, and number of categories. F i r s t , presentation order can be arranged so that both structured ("clustered") and unrelated ("subjectively-organ-ized") l i s t s can be bett e r r e c a l l e d . Second, slower presentation rates increase r e c a l l performance and organization measures of both structured 28 and unrelated l i s t s . T h ird, cueing with category names serves to i n -crease the amount r e c a l l e d from both structured and unrelated l i s t s by increasing the number of categories, but not number of words w i t h i n a category, r e c a l l e d . Fourth, number of categories appears to be an important v a r i a b l e a f f e c t i n g r e c a l l performance of both structured and unrelated l i s t s ; however, because of methodological d i f f i c u l t i e s , i n neither case can the nature of the e f f e c t be c l e a r l y determined. I n h i b i t i o n , i n d i v i d u a l differences, and short- and long-term memo-ry components of r e c a l l have been discussed. Proactive i n h i b i t i o n was described i n part-whole, whole-part, and whole-whole free r e c a l l . Learning of two l i s t s which have an overlap of items does not seem to have a f a c i l i t a t i v e e f f e c t , but on l a t e r t r i a l s , i t seems to have an i n h i b i t o r y e f f e c t . Retroactive i n h i b i t i o n was found to be a d i r e c t function of s i m i l a r i t y of organization of the int e r p o l a t e d l i s t s and the number of interpolated l i s t s . Also, most of the r e t r o a c t i v e e f f e c t was found i n category r e c a l l , not r e c a l l w i t h i n categories. Experiments i d e n t i f y i n g organization scores as a measure of i n d i v i d u a l differences i n verbal learning a b i l i t y were described. P o s i t i v e c o r r e l a t i o n s were reported between a subject's organization score on one free r e c a l l l i s t and h i s performance on another free r e c a l l l i s t or on a paired-associate or s e r i a l l i s t . Experiments i d e n t i f y i n g two components of free r e c a l l , a short-term memory component and a long-term memory component, were reviewed. A procedure f or eliminating the short-term memory component was described. 29 CHAPTER TWO A REVIEW OF THE THEORIES In Chapter 1, the major experimental findings concerning organi-zation and free r e c a l l were discussed. This chapter w i l l consider the theories x^hich have been proposed to account for these f i n d i n g s . The theories w i l l be c r i t i c a l l y reviewed i n terms of t h e i r a b i l i t y to ac-count f o r the data presented i n Chapter 1. Perhaps e s s e n t i a l to understanding most of the theories i s an o f t - c i t e d paper by George M i l l e r (1956). In t h i s paper, M i l l e r was con-cerned with an integer, or a c t u a l l y a small range of integers, 7±2. M i l l e r ' s concern was r e l a t e d to two types of experimental paradigms, psychophysical judgments and the d i g i t span. The two paradigms w i l l be discussed separately. In one type of psychophysical task, the subject i s presented a number of s t i m u l i varying along one dimension. Each of the s t i m u l i i s assigned a l a b e l or name, and upon presentation of one of the s t i m u l i , the subject must produce i t s l a b e l or name. Important to understanding the performance of a subject In a task of t h i s type i s the concept of " b i t s " of information. A b i t represents the amount of information necessary to decide between two equally l i k e l y a l t e r n a t i v e s (a binary d e c i s i o n ) . Two b i t s , then, represents the information necessary to de-cide between four equally l i k e l y a l t e r n a t i v e s , while three b i t s repre-sents the information necessary to decide between eight equally l i k e l y a l t e r n a t i v e s , etc. Thus, a subject's performance i n a psychophysical task, such as that described above, may be measured i n terms of b i t s of information. 30 M i l l e r c i t e d a number of experiments i n which absolute judgments of unidimensional s t i m u l i were made. He drew from experiments of audi-tory, gustatory, and v i s u a l discriminations and found the range of b i t s s u c c e s s f u l l y processed (channel capacity) by subjects across these sen-sory modes was 1.6 to 3.9, representing from 3 to 15 items. What most impressed M i l l e r about these findings was the small variance across ex-periments i n the number of b i t s s u c c e s s f u l l y processed; the mean number of b i t s was 2.6, which corresponds to a mean of 6.5 items, while the standard deviation was 0.6 b i t s , so that one standard deviation from the mean includes a range of 4 to 10 items. Hedging s l i g h t l y , M i l l e r concluded that performance i n a unidimensional psychophysical task could be described i n terms of discriminations between 7+2 s t i m u l i . How can a subject discriminate between more than seven stimuli? The obvious answer i s to have the s t i m u l i d i f f e r along more than one dimension. For instance, i f there were two stimulus dimensions, i t would be expected that subjects could discriminate between 49 s t i m u l i , etc. M i l l e r , also, has argued that increasing the number of dimensions should increase channel capacity. He c i t e d a number of experiments which provide evidence for t h i s point, but the experiments i n d i c a t e that the function r e l a t i n g the number of items to the number of dimensions i s not a simple m u l t i p l i c a t i v e one; rather, the addition of stimulus d i -mensions adds to the channel capacity at a decreasing rate. A second task which concerned M i l l e r was the d i g i t span task. In t h i s task, the subject i s presented a number of s t i m u l i and i s asked to repeat them i n the order i n which they were presented. For various types of s t i m u l i , the span has been found to be about 7±2 items. 31 M i l l e r pointed out that subjects remember about 7±2 s t i m u l i i n the d i g i t span task regardless of the amount of information that the s t i m u l i convey. For instance, one binary d i g i t represents one b i t of information, and one decimal d i g i t represents over three b i t s of information. Yet, subjects remember 7±2 binary d i g i t s and 7±2 decimal d i g i t s . Thus, i n the d i g i t span information i s not constant, while the capacity f o r the number of items that can be r e c a l l e d appears to be constant. One way to increase capacity i n the d i g i t span task i s to increase the amount of information conveyed by each stimulus. For instance, M i l l e r reports experiments i n which subjects recode binary d i g i t s i n t o o c t a l d i g i t s and show a t h r e e f o l d increase i n the amount of information re-c a l l e d . This recoding process has been referred to as chunking. Free r e c a l l t h e o r i s t s have drawn analogies between M i l l e r ' s (1956) paper and free r e c a l l experiments i n t r y i n g to determine how subjects can learn to r e c a l l more than seven items. Two analogies can be drawn, one to the psychophysical task and one to the d i g i t span task. The analogy to the psychophysical task requires that the words to be r e c a l l e d can be described i n terms of multiple dimensions. This as-sumption appears to be i n t u i t i v e l y p l a u s i b l e because words have semantic, acoustic, s y n t a c t i c , etc. properties. This analogy needs one of two further assumptions to describe output i n free r e c a l l . F i r s t , the pro-cess may be passive such that words are coded along various dimensions at input and c l u s t e r i n g i s a product of some systematic search of the dimensions. Second, the process may be active such that words are chunked or coded along only a s p e c i f i e d number of dimensions, and e i t h e r the search i s l i m i t e d or the r e c a l l of the code f o r a chunk precedes r e c a l l of the words wit h i n that chunk. This second assumption, then i s a mixture of d i g i t span chunking and psychophysical coding, and thus i s not a d i r e c t analogy; however, i t may s p e c i f y a stimulus for the free r e c a l l of the items. The f i r s t assumption i s perhaps a more d i r e c t analogy, but Ttfithout a d d i t i o n a l assumptions does not solve the problem of the stimulus i n free r e c a l l . The second analogy has been drawn to the d i g i t span task. Here, the subject i s assumed to a c t i v e l y chunk the items. R e c a l l of addi-t i o n a l items may proceed by e i t h e r adding information to a l i m i t e d number of chunks, or by forming a h i e r a r c h i c a l arrangement of chunks, subchunks, etc. This second analogy provides a stimulus i n terms of a code f o r a chunk, but i t presents three other problems. F i r s t , why does order of output correspond to the properties of the stimulus words, (e.g., Bousfield, 1953)? To answer t h i s question, further assumptions concer-ning the r e l a t i o n s h i p of chunking to some storage structure i s necessary. Second, the analogy ignores possible d i s t i n c t i o n s between short- and long-term memory. Th i r d , decoding i n a free r e c a l l task i s not analo-gous to decoding i n a d i g i t span task. For instance, i f , i n a d i g i t span task, a subject decodes three binary d i g i t s i n t o an o c t a l d i g i t , when he r e c a l l the o c t a l d i g i t , only one possible set of three binary d i g i t s w i l l be suggested. That i s , there i s a one-to-one mapping between the new code and the o r i g i n a l s t i m u l i . In a free r e c a l l task, however, i f the subject recbdes three animal names in t o the code 'animal," the r e c a l l of "animal" can suggest many more items than the three animal 33 names that were presented. Thus, i f a theory of free recall i s based on an analogy to Miller's discussion of the digit span, i t must add assump-tions about how the subject discriminates which items have been pre-sented from those which have not been presented. Having argued that analogies relating free recall to Miller's suggestions based on the "magical number seven" are rather limited, the remainder of this chapter w i l l be devoted to three theories based on these' analogies: (1) a theory based on a chunking process where there are a limited number of chunks, and improvement in re c a l l i s a function of increasing the amount of information per chunk; (2) a theory i n which chunks are assumed to be hierarchically arranged; and (3) a theory in which multiple dimensions are used to describe free r e c a l l . A Chunking Model ~ Tulving (1968) Tulving (1968) recognized both the limited capacity of immediate memory and the differences between paired-associate or s e r i a l learning and free r e c a l l learning. He raised two questions. F i r s t , how can a subject show increases in r e c a l l , as a function of practice, beyond the limitations of immediate memory? Second, how can interitem associations in free re c a l l be described, and what is the relationship between these associations and free re c a l l performance? To answer these questions, Tulving (1968) described two types of units of analysis, E-units and S-units. E-units, or experimenter units, may be defined as the units that the experimenter uses in scoring the data. For instance, E-units may be the number of words recalled or, i n the case of categorized l i s t s , the number of categories or the number of 34 words w i t h i n a category r e c a l l e d . S-units, or subjective u n i t s , on the other hand, are the fu n c t i o n a l units employed by the memory system. In-creases i n r e c a l l , as measured i n terms of E-units,are assumed to r e f l e c t the increasing s i z e of S-units. Tulving's d i s t i n c t i o n between E-units and S-units i s a d i r e c t analogy to M i l l e r ' s (1956) hypothesis concerning the l i m i t e d capacity of immediate memory. I f a subject can r e c a l l only a given number of items or E-units, he must chunk these items into larger items or S-units i f he i s to show increases i n r e c a l l performance. Thus, increases i n r e c a l l r e f l e c t the increasing s i z e of S-units. What i s the nature of S-units? Tulving (1968) described two pos-s i b l e types of S-units. The f i r s t type of S-unit i s based on item-to-item associations. That i s , an item i s organized i n t o a given S-unit. according to i t s strengths of association with other items i n the S-unit. The second type of S-unit i s composed of associations to a common media-tor . That i s , items that share the same superordinate category vr±ll be grouped i n t o the same S-unit. Tulving points out that i t i s premature to d i s t i n g u i s h between the two possible types of S-units. An item that i s a d i r e c t associate to another item w i l l most c e r t a i n l y share some superordinate category with that item. Conversely, items that are i n the same superordinate category w i l l have some associative strength. Regardless of which type of associations describes S-units, Tulving contends that increases i n asso c i a t i v e strength w i t h i n S-units leads to increases i n r e c a l l . This contention, then, allows free r e c a l l to be described i n a manner s i m i l a r to paired-associate or s e r i a l learning 35 where i t i s possible to describe learning i n terms of the strengthening of associations between s t i m u l i and responses. This b r i e f sketch of Tulving's theory presents as systematic an account as can be drawn. Tulving prefers to modify or embellish h i s theory i n order to describe the free r e c a l l data. A d e s c r i p t i o n of such modifications w i l l be made i n terms of the data discussed i n the f i r s t chapter. The Data Tulving (1968) did not d i s t i n g u i s h between mechanisms underlying c l u s t e r i n g as opposed to subjective organization; rather, he considers the differences to l i e s o l e l y i n terms of experimental operations used to examine the two types of organization. Thus, h i s t h e o r e t i c a l con-siderations of the free r e c a l l data apply to both c l u s t e r i n g and sub-j e c t i v e organization. Variables which a f f e c t free r e c a l l performance are assumed to af-fect the subject's opportunity to discover relevant associations and to use these associations i n forming S-units. Thus, presenting related words i n contiguous input order should increase the subject's a b i l i t y to perceive relevant associations, use these associations i n forming S-u n i t s , and lead to increased r e c a l l . S i m i l a r l y , slowing the presentation rate would give the subject more time to f i n d relevant associations. In both cases, measures of organization as w e l l as r e c a l l performance should show improvement. To explain the e f f e c t of cueing and number of categories, Tulving (1968) distinguished between the " a v a i l a b i l i t y " and " a c c e s s i b i l i t y " of 36 S-units. The terms "av a i l a b i l i t y " and "accessibility," may be looked at as analogous to the terms ''storage" and "retrieval," respectively. That i s , an S-unit may ba stored and be available to the subject, but because there i s a limited capacity retrieval system, that S-unit may not be accessible. If the number of S-units exceeds the retrieval ca-pacity, some S-units w i l l f a i l to ba recalled. Providing cues should allow the subject access to these S-units, and recall should show im-provement. Accessibility, apparently, only applies to S-units, as the number of items recalled within an S-unit is independent of whether the subject recalls the S-unit or whether he must be cued to recall the S-unit. Following the argument that re c a l l of items within categories is independent of the number of categories recalled, number of categories should be an important determinant of recall performance. If number of categories exceeds retrieval capacity, recall performance should be i n -hibited. Subjects would have to combine or rearrange categories i n order to reach perfect re c a l l performance. Thus, a subject's re c a l l should improve as a function of the number of categories un t i l the point at which the retrieval capacity is reached. Tulving's most impressive theoretical predictions involve the part-whole and whole-part paradigms, although his interpretation has been recently questioned (Schwartz & Humphreys, 1972b; Slamecka, Moore, & Carey, 1972). His predictions involve his assumptions about the relat-edness of associations and the limited retrieval capacity. For part-whole free recall of unrelated words (Tulving, 1966), the subject is 37 assumed to form S-units f o r the part l i s t s . These S-units, on the average, w i l l be inappropriate for w h o l e - l i s t organization. Because the subject can only r e c a l l a l i m i t e d number of S-units, he cannot simply maintain h i s p a r t - l i s t S-units and form new S-units for the new who l e - l i s t words. Thus, he w i l l have to modify and reorganize h i s p a r t -l i s t S-units. As this reorganization i s assumed to take time, the i n -h i b i t o r y e f f e c t s of p a r t - l i s t learning should not occur u n t i l l a t e r i n learning. A s i m i l a r argument has been made for the whole-part paradigm (Tulving & Osier, 1967). Tulving's (Tulving & Psotka, 1971) d e s c r i p t i o n of re t r o a c t i v e i n h i -b i t i o n i s not so cl e a r as h i s d e s c r i p t i o n of proactive i n h i b i t i o n . Ra-ther than explain why re t r o a c t i v e i n h i b i t i o n occurs, he uses h i s d i s -t i n c t i o n between a v a i l a b i l i t y and a c c e s s i b i l i t y of S-units to explain how r e t r o a c t i v e i n h i b i t i o n occurs. That i s , he states that r e t r o a c t i o n should represent the i n a b i l i t y to r e t r i e v e S-units rather than a loss of items from storage. I f an S-unit from an early l i s t i s r e t r i e v e d , there should be no loss of items within the S-unit. I f cues are presented f o r S-units from an e a r l i e r l i s t , there should be no decrement i n r e c a l l . His data (Tulving & Psotka, 1971) support t h i s point of view. To the author's knowledge, Tulving has made no published statement concerning the r e l a t i o n s h i p between organization scores and learning a b i l i t y . However, as organization scores i n free r e c a l l have been shown to be r e l a t e d to paired-associate and s e r i a l learning scores, a possible explanation i s apparent. S e r i a l learning and paired-associate learning can be described i n terms of the strengthening of associations, and 38 Tulving has described free recall learning in terms of the strengthening of associations. Thus, a subject who had high organization scores would be assumed to be good at forming associations and should do well in paired-associate or s e r i a l learning. Tulving does not agree that the bow-shaped s e r i a l position curve found in s i n g l e - t r i a l free re c a l l represents two storage systems. He argues, instead, that the items presented at the ends of the l i s t may be more accessible. The increased accessibility may be due to some auxil-iary information about the end items, such as temporal dating or acoustic traces. Evaluation The f l e x i b i l i t y of Tulving's theory is probably i t s greatest strength, and the f l e x i b i l i t y makes the theory d i f f i c u l t to c r i t i c i z e in terms of the data. However, there are some possible inconsistencies be-tween theory and data. For instance, Shapiro and Bell's (1970) finding that some subjects show increases in recall without corresponding i n -creases i n organization scores is d i f f i c u l t to reconcile with the pro-posed necessity of forming S-units. Schwartz and Humphreys' (1972b) finding that subjects who do not reorganize their f i r s t - l i s t S-units perform better at part-whole free recall suggest that the hypothesized limit to the number of S-units which can be retrieved needs to be fur-ther investigated. Shuell's (1968) finding of a loss of items within categories as a function of retroactive inhibition i s d i f f i c u l t to rec-oncile with the independence of catetory recall and r e c a l l of items within categories. 39 However, the main c r i t i c i s m s of Tulving's theory l i e not i n i t s i n a b i l i t y to explain some aspects of the data but instead i n the absence of c e r t a i n mechanisms. For instance, items i n S-units are s a i d to be rela t e d . How i s th i s "relatedness" extracted from long-term memory? There i s a l i m i t e d number of S-units which can be r e c a l l e d . Is there a l i m i t on the number of items which can be included i n an S-unit? I f so, how could the theory handle t h i s l i m i t ? Other problems involve an ab-sence of data. For instance, cueing i s supposed to serve as a r e t r i e v a l aid. Can i t , then, be demonstrated that the e f f e c t of organization i n noncued r e c a l l i s a r e t r i e v a l rather than a storage e f f e c t ? I t i s as-sumed that the same processes underlie both c l u s t e r i n g and subjective organization. Can t h i s assumption be demonstrated? These are some of the questions that w i l l be the concern of the experiments reported i n the next three chapters. A H i e r a r c h i c a l Model — Mandler (1967) Mandler (1967) has proposed a model of organization which i s very s i m i l a r to that offered by Tulving (1968). The main differ e n c e between the two models i s the h i e r a r c h i c a l arrangement of higher-order memory units i n Mandler's model as opposed to the chunking i n Tulving's model. Like Tulving, Mandler, by extension from M i l l e r ' s (1956) argument con-cerning the l i m i t e d capacity of immediate memory, hypothesized that organization i s necessary f o r improving f r e e r e c a l l performance. How-ever, Mandler also provided an answer f or a question raised i n regard to T u l v i n g ! s model. That i s , i s there a l i m i t to the number of items contained i n an S-unit? Mandler suggested that there i s such a l i m i t , 40 and i f the number of items w i t h i n an S-unit surpasses t h i s l i m i t , sub-j e c t s w i l l form subordinate S-units. Thus, a h i e r a r c h i c a l structure w i l l develop. Handler (1967) also proposed a numerical value f o r the l i m i t to the number of u n i t s , items w i t h i n u n i t s , etc. — 5±2. The difference between the 5±2 that Mandler proposed and the 7±2 offered by M i l l e r (1956) r e s u l t s from Mandler's recognition of the differences between short- and long-term memory. Using the r e s u l t s of an experiment con-ducted by Waugh and Norman (1965), Mandler estimated the l i m i t of short-term memory as 3±1. The remainder, 4±1, was offered as Mandler's es-timate of the long-term memory component. His experiments on the cate-g o r y - r e c a l l function (see Chapter 1) have led him to revise t h i s e s t i -mate to 5+2. As Mandler's (1967) theory has not been extended to cover as many aspects of the free r e c a l l data as has Tulving's (1968) theory, Mandler's theory w i l l not be evaluated i n terms of the data. In any case, because of the s i m i l a r i t y between Handler's approach and Tulving's approach, i t i s l i k e l y that Mandler would handle many aspects of the data i n the same way that Tulving has. However, i t should be pointed out that Schwartz and Humphreys (1972a) have severely questioned, on methodological grounds, Mandler's cate g o r y - r e c a l l function (see Chapter 1). Though Handler has not presented a systematic account of the data, he should be credited with two t h e o r e t i c a l improvements of Tulving's theory. F i r s t , Mandler presents an approach for handling a possible l i m i t e d capacity of items w i t h i n S-units. Second, he distinguishes 41 between short- and long-term memory components i n dealing with h i s analogy from M i l l e r ' s (1956) review. A Multidimensional Model — Kintsch (1970) Kintsch (1970) has presented a model of free r e c a l l which may be c l a s s i f i e d as a multidimensional model. Kintsch's model, however, i s not based on a d i r e c t analogy to the psychophysical judgment tasks discussed by M i l l e r (1956), nor does the model d i r e c t l y r e f e r to a multidimensional structure. Nevertheless, Kintsch does describe words i n long-term memory as represented by a serie s of "markers." I f these markers are allowed to assume continuous values, the serie s of markers f o r a given word xirould i n d i c a t e i t s l o c a t i o n i n a multidimensional space. Kintsch re f e r s to three primary types of markers, S_, I_, and P_. J3 markers are syntactic-semantic markers and are of greatest importance i n free r e c a l l . I_ markers are sensory feature markers, and P_ markers are phonetic markers. markers may be broken down Into l e x i c a l , as-s o c i a t i v e , and sy n t a c t i c f i e l d s . Here, l e x i c a l f i e l d s are of the greatest importance. L e x i c a l f i e l d s may be of two types. The f i r s t type, which Kintsch c a l l s antonymy, i s given by oppositions, e.g., good-bad. In some cases, these oppositions can be seen as extremes of a dimension. The second type of l e x i c a l f i e l d i s c a l l e d h i e r a r c h i c a l or sequential. Markers i n t h i s type of f i e l d r e f e r to re l a t i o n s h i p s of class i n c l u s i o n or category membership. When a word i s presented i n a free r e c a l l task, the subject stores the information that the word has occurred. He does not store the word per se because the word i s already present i n long-term memory. Instead, 42 he tags the marker l i s t that corresponds to the word i n order to i n d i c a t e that the word was presented. To follow the course of input and output i n free r e c a l l , Kintsch uses the Atkinson and S h i f f r i n (1968) "buffer" model. A word, when presented, has some p r o b a b i l i t y of entering a short-term memory b u f f e r (for v ariables which determine whether a word w i l l enter the b u f f e r , see Atkinson & S h i f f r i n , 1968). I f the word does enter the b u f f e r , some "cognitive" work w i l l be performed on i t . This cognitive work can be of two types. F i r s t , the word may be added to the marker l i s t of another word i n short-term memory. Second, and both more e f f i c i e n t and more i n l i n e with a multidimensional model, a marker may be found that i s held i n common with another word. The e f f i c i e n c y of t h i s cognitive work w i l l vary d i r e c t l y with the amount of time tha item remains i n the b u f f e r . In r e t r i e v a l , items from the short-term memory b u f f e r are f i r s t r e c a l l e d . Then, another word i s chosen i n a random manner. The markers associated with t h i s word are examined. I f a marker for t h i s word has been associated with another item, that item i s examined. I f that item's markers i n d i c a t e that the item was on the l i s t , the item i s r e c a l l e d . The process can then recycle with t h i s new item. I f no item i s found which was on the l i s t , the search stops or a new item i s chosen to re-cycle the search. A model of t h i s type predicts that s i m i l a r i t y of items should be the best pr e d i c t o r of output order i n free r e c a l l . That i s , the more s i m i l a r two items are, the more l i k e l y they are to share common markers, and the " c o g n i t i v e " work performed i n the b u f f e r should ba more e f f i c i e n t 43 i n f i n d i n g these common markers. Thus, c l u s t e r i n g of rela t e d items i s r e a d i l y predicted. Also, the greater r e c a l l of categorized l i s t s , as opposed to unrelated l i s t s , can be predicted as a function of the prob-a b i l i t y of the search stopping. That i s , with unrelated l i s t s , i t i s less l i k e l y that the set of markers associated with a r e c a l l e d item w i l l include a marker which i s shared i n common with another item on the l i s t . The Data Kintsch's (1970) model i s i n i t s formative stages, and no system-a t i c attempt has been made to use the model i n explaining a large v a r i e t y of free r e c a l l phenomena. In th i s s e c t i o n , an attempt w i l l be made to show how a model of th i s type could handle some of the data d i s -cussed i n the f i r s t chapter. F i r s t , Kintsch does seem to d i f f e r e n t i a t e between c l u s t e r i n g and subjective organization. He refe r s to subjective organization as i d i o -s y n c r a t i c (1970, p. 350). However, i n his model, he does not in d i c a t e that the two types of organization represent d i f f e r e n t processes. Thus, i n the analysis of the data, the two types of organization w i l l be treated as i f they were the same. Order of input, presentation rate, cueing, and the s e r i a l p o s i t i o n curve should be handled e a s i l y by Kintsch's model. Any manipulation which would allow the subject increased opportunity to discover common markers should r e s u l t i n increased organization and increased r e c a l l . Thus, when items which share common markers are presented adjacently, they are l i k e l y to reside i n the bu f f e r at the same time and the common 44 markers can be extracted. When the pressentation rate i s slowed, there i s more time f o r the "cognitive" work which r e s u l t s i n find i n g common markers. The search process could stop without c e r t a i n markers having been found i n the l i s t s of markers that were examined. A cue word, then, could serve the purpose of a marker and aid r e t r i e v a l of items as-sociated with that marker. Kintsch's use of the Atkinson and S h i f f r i n (1968) b u f f e r concept provides a mechanism for describing the s e r i a l p o s i t i o n curve. As Kintch's model does not include a concept of categories, i t would be d i f f i c u l t to explain the e f f e c t of number of categories on free r e c a l l performance. Also, added assumptions would be necessary to ex-p l a i n i n h i b i t i o n and i n d i v i d u a l d i f f e r e n c e s . Evaluation I t i s d i f f i c u l t to evaluate Kintsch's (1970) model i n terms of the data because the model has not been applied to many aspects of the data. However, i t i s possible to compare K i n t s c h 7 s model with a model such as that offered by Tulving (1968). Kintsch's approach o f f e r s answers to questions of relatedness i n r e c a l l protocols. Unlike Tulving, Kintsch suggested a structure f o r long-term memory and provided a mechanism for using that structure i n free r e c a l l learning. However, the structure presents other problems. For instance, what prevents the subject from using most of h i s search time fi n d i n g (but not r e c a l l i n g ) items that were not presented? That i s , a marker which describes an item which was presented should also describe many items which were not presented. 45 Kintsch also described subjective organization as being i d i o s y n -c r a t i c while c l u s t e r i n g was described as being based on s i m i l a r i t i e s . The d i s t i n c t i o n between c l u s t e r i n g and subjective organization was not made by Tulving. As yet, no data are available x^hich examine sub-j e c t i v e organization i n terms of s i m i l a r i t i e s . Like Tulving, Kintsch described the e f f e c t of organization to be i n r e t r i e v a l rather than i n storage. I f an item enters the b u f f e r , i t s marker l i s t i s tagged. Thus, whether or not an item i s stored i s a function of i t s having entered the b u f f e r and not of organization. How-ever, organization, i n terms of common markers, i s an important deter-minant of r e c a l l p r o b a b i l i t y . Unlike the models offered by Tulving and Mandler, Kintsch's model does not imply that organization i s necessary for improving free r e c a l l performance. Conclusion Because the three theories presented i n t h i s chapter d i f f e r i n th e i r degree of completion and the types of data they were formulated to explain, i t would be d i f f i c u l t to choose the 'better'' of the three. However, there are differe?ices between the theories, and some of these differences motivate the research presented i n the next three chapters. The major diffe r e n c e between the theories i s that two of them (Tulving, 1968: Mandler, 1967) consider increases i n organization neces- sary for increases i n r e c a l l performance, and the t h i r d (Kintsch, 1970) considers increases i n organization h e l p f u l for increases i n r e c a l l perform-ance. I t was argued i n Chapter 1 that i t i s impossible to d i s t i n g u i s h be-tween these a l t e r n a t i v e views of organization. However, i t i s possible to ask 46 a question r e l a t e d to these a l t e r n a t i v e views. That i s , given that organization a f f e c t s r e c a l l , at what stage of the r e c a l l process i s the e f f e c t f e l t ? Both Tulving (1968) and Kintsch (1970) argue that the ef-f e c t i s i n the r e t r i e v a l , rather than the storage, stage. However, analogies from M i l l e r ' s (1956) notion of l i m i t e d "channel" capacity do not necessitate a r e t r i e v a l , rather than storage, e f f e c t , and there i s l i t t l e evidence which i s d i r e c t l y r e l a t e d to the s t o r a g e - r e t r i e v a l question. In Chapter 3 two experiments d i r e c t l y related to the storage-r e t r i e v a l question w i l l be reported. Another difference between the theories i s that Tulving (1968) does not d i s t i n g u i s h between c l u s t e r i n g and subjective organization, but Kintsch (1970) may make such a d i s t i n c t i o n . Also, Kintsch's model has a mechanism f o r organizing according to s i m i l a r i t y between items, but Tulving's model does not have such a mechanism. The experiments re-ported i n Chapter A were designed to examine both of these d i f f e r e n c e s . Here, Kintsch's notion of s i m i l a r i t y was used to examine the organiza-t i o n of unrelated words. I f order of output of unrelated words could be predicted from s i m i l a r i t y judgments data, i t could be argued (1) that the same processes underlie both c l u s t e r i n g and subjective organ-i z a t i o n and (2) that a mechanism i n v o l v i n g s i m i l a r i t y can be used as an explanatory device for both types of organization. The three theories also d i f f e r i n the i n t e r n a l structures used to generate order of output. Tulving's (1968) model uses a chunking struc t u r e , and Mandler's (1967) model uses a h i e r a r c h i c a l arrangement of chunks. Kintsch's (1970) model i s more analogous to a "cue-directed'' 47 model o f m u l t i d i m e n s i o n a l s e a r c h . I n C h a p t e r 5 o r d e r o f o u t p u t was examined i n t h e hope t h a t w i t h an a p p r o p r i a t e s e l e c t i o n o f m a t e r i a l s , an adequate d e s c r i p t i o n o f o r d e r o f o u t p u t m ight be made. Such des-c r i p t i o n m i g h t r e f l e c t on t h e i n t e r n a l s t r u c t u r e used t o g e n e r a t e s u c h o u t p u t . I n C h a p t e r 6, t h e d a t a f r o m C h a p t e r 3, 4, and 5 were summarized and d i s c u s s e d . I n C h a p t e r 7, the d a t a p r e s e n t e d i n t h e f i r s t c h a p t e r , t h e t h e o r e t i c a l c o n s i d e r a t i o n s r a i s e d i n t h i s c h a p t e r , and the d a t a f r o m the e x p e r i m e n t s r e p o r t e d i n t h e n e x t t h r e e c h a p t e r s were i n c o r p o -r a t e d i n t o an i n f o r m a l model o f f r e e r e c a l l . S u g g e s t i o n s as t o how t h e model c o u l d h a n d l e t h e d a t a p r e s e n t e d i n C h a p t e r 1 were made. 43 CHAPTER 3 ORGANIZATION: A STORAGE OR RETRIEVAL EFFECT? Theories of f r e e r e c a l l have g e n e r a l l y i n v o l v e d the assumption that the e f f e c t of o r g a n i z a t i o n a l processes i s on the amount r e t r i e v e d r a t h e r than the amount s t o r e d . For i n s t a n c e , T u l v i n g (1968) d i f f e r e n -t i a t e s between the " a v a i l a b i l i t y ' ' and " a c c e s s i b i l i t y " of items and argues that o r g a n i z a t i o n i n t o S-units a f f e c t s a c c e s s i b i l i t y . His argument f o l -lows from h i s n o t i o n of a l i m i t e d c a p a c i t y r e t r i e v a l system and from data he has c o l l e c t e d on the e f f e c t s of cueing ( T u l v i n g & P e a r l s t o n e , 1966) and r e t r o a c t i v e i n h i b i t i o n ( T u l v i n g & P s o t k a , 1971). K i n t s c h (1970) a l s o argues that o r g a n i z a t i o n a f f e c t s r e t r i e v a l r a t h e r than storage. K i n t s c h ' s argument i s an extension of h i s two process theory of r e c o g n i t i o n and r e c a l l . His model of r e c o g n i t i o n i s based on the assumption that storage i s determined by whether an item enters the short-term memory b u f f e r where the p r o b a b i l i t y of an item e n t e r i n g the b u f f e r i s not r e l a t e d to o r g a n i z a t i o n a l processes. However, the prob-a b i l i t y of r e c a l l i s dependent upon the extent of the " c o g n i t i v e work" or o r g a n i z a t i o n that takes place i n the b u f f e r . Although T u l v i n g (1968) and K i n t s c h (1970) agree that o r g a n i z a t i o n a f f e c t s the amount r e t r i e v e d but not the amount s t o r e d , t h e i r t h e o r i e s do not e l i m i n a t e the p o s s i b i l i t y of storage e f f e c t s . For i n s t a n c e , the types of o r g a n i z a t i o n described by both T u l v i n g (1968) and K i n t s c h (1970) a f f e c t the form of the s t o r e d i n f o r m a t i o n . T u l v i n g views o r g a n i z a t i o n as analogous to the chunking process described by M i l l e r (1956). Thus, s e v e r a l items i n the l i s t are recoded, and the code i s s t o r e d . In 49 Kintsch's model, an item Is represented by a series of markers which describe that item. The "cognitive work" performed on that item is represented by a change in i t s l i s t of markers. For instance, an item may be added to the marker l i s t of another item. Thus, in both Kintsch's and Tulving's models, organization affects the form of item storage. As these theories do not preclude storage effects, the storage-retrieval question is a useful point of departure for an empirical i n -vestigation of organizational processes in free r e c a l l . In this chapter, two experiments which examine the storage-retrieval question are reported. The paradigm for both experiments is similar and takes the form of a 2 X 2 factorial design. The f i r s t factor is a manipulation of organiza-tion so that there are two groups of subjects who dif f e r in the extent to which they have organized the material. The second factor is the form of testing, i.e., whether recognition or recall tests are used. It is assumed that recognition measures the amount stored and recall measures the amount retrieved. This assumption appears intuitively valid, and support for this assumption can be found in a paper by Kintsch (1970) and in a review article by McCormack (1972). If the organizational manipulation affects recognition performance, i t can be concluded that organization does affect the amount stored. If the organizational manipulation affects r e c a l l performance but not recognition performance, i t can be concluded that organization affects the amount retrieved but not the amount stored. Both Kintsch (1968) and Earhard (personal communication) have ex-amined recognition and r e c a l l performance as a function of differences in 50 organizational l e v e l . Kintsch's manipulation involved the type of ma-t e r i a l presented. A l l subjects were presented l i s t s of 10 words belonging to each of four conceptual categories. For some subjects, the items were high associates to the category names; f o r other subjects, the items were law associates to the category names. Subjects were tested using e i t h e r recognition or r e c a l l t e s t s , and the r e s u l t s showed that subjects pre-sented l i s t s of high category associates r e c a l l e d but did not recognize more than subjects presented l i s t s of low category associates. Kintsch's r e s u l t s have been r e p l i c a t e d by Bruce and Fagan (1971). However, Kintsch's r e s u l t s cannot be c l e a r l y interpreted as i n d i -cating that organization a f f e c t s r e c a l l performance but not recognition performance. F i r s t , to reach such a conclusion, i t must be assumed that the only d i f f e r e n c e between presenting a l i s t of high category associates as opposed to a l i s t of low category associates i s the amount of organi-zation the subject uses. This assumption i s not necessarily v a l i d . Puff (1970b) has found that i n s i n g l e - t r i a l free r e c a l l , subjects who clustered a categorized l i s t more than would be expected by chance did not r e c a l l more than subjects who clustered the same l i s t at the chance l e v e l , and that both groups of subjects r e c a l l e d more than subjects who learned a l i s t of unrelated words. Thus, i t i s not c l e a r that the d i f -ference between r e c a l l of a l i s t of high category associates and a l i s t of low category associates i s due to l e v e l of organization. Second, as the material presented varied between groups, recognition tests could not be equated. That i s , the type of d i s t r a c t o r used to t e s t the recognition 51 of a l i s t of low category associates i s not equivalent to that used to test the recognition of a l i s t of high category associates. Earhard (personal communication) has i d e n t i f i e d groups of high and low subjective organizers and compared these subjects on recognition t e s t s . She found that high organizers performed better on the recognition tests than low organizers. One possible conclusion from her r e s u l t s i s that organization a f f e c t s the amount stored. However, as discussed i n the f i r s t chapter, organizational l e v e l may be a measure of some verbal learning a b i l i t y , and i t i s c l e a r that Earhard's study does not separate the degree of organization from t h i s a b i l i t y . Thus, a second possible conclusion i s that the locus of the i n d i v i d u a l a b i l i t y which correlated with organization scores i s i n the amount stored. Experiment I In the experiments reported i n t h i s chapter, an attempt was made to avoid the confounding of organization with e i t h e r type of material or the subject's a b i l i t y . In Experiment I t h i s attempt was r e a l i z e d by i n s t r u c t i n g subjects i n the experimental groups to use v i s u a l imagery to organize the items i n the order i n which they were presented. The e f f e c t of i n s t r u c t i o n s to use v i s u a l imagery i n paradigms other than free r e c a l l has been reviewed by Paivio (1969). Insofar as "organization" may include the grouping of items at input, i n s t r u c t i o n s to use v i s u a l imagery i n order to group items can be interpreted as a manipulation of organization. Bower, Lesgold, and Tieman (1969) have used imagery instructions to encourage subjects to group items; however, unlike a typical free recall situation, the Bower et al. experiments presented items in groups. In the present experiment, to create a more typical free recall situation, items were presented one at a time and experimental subjects were asked to use visual imagery to group items according to input order. Subjects in the control groups were read standard instructions which did not mention visual imagery. Subjects in both experimental and control groups were given recognition and recall tests. Method Stimuli and materials. The stimuli were 120 concrete (C > 5>50) and frequent (F > 20) nouns chosen randomly and without replacement from the Paivio, Yuille, and Madigan (1969) norms. Of these 120 nouns, 60 were randomly chosen to comprise the l i s t to be presented to a l l of the subjects. The other 60 nouns served as distractors for the subjects in the recognition conditions. Each of the 60 nouns on the l i s t to be presented was mounted as a slide. The slides were then randomly arranged to determine order of presentation for a l l of the subjects. Presentation was via a Kodak carousel projector (Model 850) with an external timer. Answer sheets were prepared for subjects in the recognition and recall conditions. For the recall conditions, the answer sheets contained two columns of 30 blank lines. At the top of the answer sheet was an instruction asking the subject to write down in any order he desired, as many words as he could remember from the l i s t he had been presented. For the 53 recognition conditions, the 60 presentation items and the 60 distractor items were randomly arranged and typed in four columns of 30 items each. The instruction at the top of the answer sheet asked the subject to put a line through a l l of the words he thought were contained in.the l i s t he had been presented. Design. A 2 X 2 factorial design, with both factors varying be-tween groups of subjects, was used. The f i r s t factor was an instructional manipulation, i.e. whether the subject was given standard learning instructions or was given instructions to use visual imagery to organize the l i s t . The second factor was type of test, i.e. whether the subject was given a recognition or recall test. Thus, there were four conditions which were labelled Standard-Recall, Standard-Recognition, Imagery-Recall, and Imagery-Recognition. Subjects. The subjects were 57 volunteers from introductory psychol-ogy classes at the University of British Columbia. They were tested in groups of 1-3 with a l l subjects in a group receiving the. same instructional manipulation. There were 14 subjects i n each condition except the Stau- dard-Recognition condition which had 15 subjects. Procedure. Subjects in a l l conditions were told that they would be presented with 60 nouns that would be familiar to them. Upon seeing each noun, they were to try to remember i t as best they could. They were also told that after seeing a l l 6 0 nouns, they would be tested on how well they had remembered them, although they were not informed of the nature of the test. Subjects in the imagery condition were given additional instructions 54 about learning the l i s t . P i l o t work had indicated that these i n s t r u c t i o n s were e f f e c t i v e f or increasing performance i n a free r e c a l l task. The ad-d i t i o n a l i n s t r u c t i o n s follow: Some students report that they can remember words of t h i s sort b e t t e r i f they form v i s u a l images containing a number, usually f i v e or fewer, of the words. This idea of forming v i s u a l images may best be explained by example. Suppose the f i r s t f i v e words presented were monkey, tennis, s h e l f , sandwich, and rock. You might imagine two monkeys playing tennis and eating sand-wiches on one s h e l f of a giant bookcase. Instead of a tennis b a l l , they were playing with a large rock. Then the next f i v e words would be presented, and you would form another image with those f i v e words and so on. Usually, the more unusual or b i z a r r e an image you form, the easier i t w i l l be to remember the words. We want you to t r y to use t h i s -method of forming v i s u a l images to help you remember the words. Try not to put more than f i v e words i n an image, because i f you use more than f i v e , you may f i n d yourself confused. Remember, the words w i l l always be presented at a constant rate; there w i l l be no break a f t e r each set of f i v e words to t e l l you when to stop forming one image and to s t a r t working on another. You w i l l have to keep track of the number of words i n each of your images on your own. If the subjects had no questions, the l i s t was then presented. Rate of presentation was four seconds per item. Immediately following presentation and according to t h e i r conditions assignment, the subjects were handed e i t h e r a recognition or a r e c a l l answer sheet. The subjects were t o l d that the Instructions were at the top of the sheet and they would have four minutes to complete the task. Results Following Kintsch (1968), the score for each subject was the d i f f e r -ence between the number of correct and the number of i n c o r r e c t responses. The means of these scores were 45.29, 42.27, 20.71, and 16.29 for the 55 Imagery-Recognition, Standard-Recognition, Imagery-Recall, and Standard- R e c a l l conditions, r e s p e c t i v e l y . Means and variances of these scores f o r each of the four conditions appear i n Table I. To approximate homogeneity of variance, each subject's score was transformed using a square root transformation. The transformed means and variances are also shown i n Table I. The transformed data were subjected to analysis of variance. Table II contains a source table f o r that a n a l y s i s . As Table II i n d i c a t e s , the imagery i n s t r u c t i o n resulted i n s i g n i f i c a n t l y improved r e c a l l perform-ance but not s i g n i f i c a n t l y improved recognition performance. A second analysis was performed on the data of the r e c a l l conditions to assure that the s u p e r i o r i t y of the Imagery-Recall condition could be att r i b u t e d to the i n s t r u c t i o n a l manipulation. As the subjects i n the imagery conditions were t o l d to form images of groups of f i v e words ac-cording to the input order, i t was predicted that the protocols of the subjects i n the Imagery-Recall condition would show more c l u s t e r i n g ac-cording to input order than those of the subjects i n the Standard-Recall condition. To te s t t h i s p r e d i c t i o n , the presentation l i s t was treated as i f i t contained 12 categories of f i v e words each. Each category was a group of f i v e words presented adjacently, s t a r t i n g with words 1-5, 6-10, etc. Then each subject's r e c a l l protocol was analyzed according to Frankel and Cole's (1971; see Chapter 1) adaptation of the runs t e s t , and a z;-score was determined for each subject. The mean z_-scores f o r the Imagery-Recall and Standard-Recall conditions were 3.53 and 1.93,respecti-vely. Both means indicated s i g n i f i c a n t l y more c l u s t e r i n g according to 56 TABLE I Variances, Means and Transformed Means and Variances for the Four Conditions in Experiment II Group Measure Recognition Recall Imagery Standard Imagery Standard ,45.29 42.27 20.71 16.29 S2x 98.38 80.50 47.91 16.99 6.68 6.46 4.50 4.00 s 2 ^ r 0.67 0.49 0.67 0.31 57 TABLE II ANOVA Source Table f o r Conditions i n Experiment I Source df SS ms F p Tota l 56 99.68 -Recognition vs. Re c a l l 1 76.64 Imagery Recognition vs.Standard Recognition 1 0.35 Imagery-Recall vs. Standard R e c a l l Error 1 1.72 53 20.97 76.64 191.60 <.001 0.35 <1 n.s. 1.72 4.30 <.05 0.40 58 input order than would be expected by chance (_z = 13.20 for Imagery- Recall and = 7.21 for Standard-Recall, both p's < .001). However, the Imagery-Recall condition showed significantly more clustering according to input order than the Standard-Recall condition, = 4.24, p < .001. Thus, i t can be concluded that the Imagery-Recall subjects were organ-izing in accordance with the instructions. A third analysis examined the effect of the imagery instruction on recall as a function of ser i a l position at input. Each subject's recall protocol was examined for probability of recall of items presented in seri a l position 1-15, 16-30, 31-45,and 46-60. The mean probabilities of recall as a function of these four blocks of input positions for the Imagery-Recall and Standard-Recall conditions are presented i n Figure 1. The data shown in Figure 1 were subjected to analysis of variance. The dependent measure was the number recalled at each of the four blocks of input positions for each subject. A source table for this analysis is contained in Table III. The results of this analysis indicate a sig-nificant effect of s e r i a l position but no significant Imagery-Recall vs. Standard-Recall X Serial Position interaction. Thus, the effect of the imagery instruction was not restricted to a certain portion of the seri a l position curve. Experiment II Experiment I indicated that the effect of the organizational manipulation was to increase the amount recalled but not the amount recognized. However, concluding that organization affects the retrieval 59 FIGURE 1 P r o b a b i l i t y of R e c a l l as a Function of S e r i a l P o s i t i o n f o r the Imagery-Recall and the Standard-Recall Conditions i n Experiment I 60 TABLE HI ANOVA Source Table for Imagery-Re call and Standard-Recall Conditions as a Function of Serial Position in Experiment I Source df SS IDS F Total 111 872.28 - -Between Subjects 27 210.43 - -Imagery-Recall vs. Standard-Recall (1) 1 31.08 31.08 4.51 Error 26 179.35 6.89 -Within Subjects 692.93 - -Serial Position (2) 3 202.39 67.46 10.93 1 X 2 3 9.02 3.01 <1 Error 78 481.92 6.17 — n.s. 61 capacity but not the storage capacity requires accepting the n u l l hypothesis concerning the recognition data. In order to gain more con-fidence i n accepting the n u l l hypothesis., Experiment II was designed to r e p l i c a t e the r e s u l t s of Experiment I. In Experiment I I , a d i f f e r e n t manipulation was used to create differences i n organization. This manipulation, unlike the imagery i n s t r u c t i o n which resulted i n higher organization, was intended to depress organization. In h i s doctoral d i s s e r t a t i o n , Rundus (1970) had subjects rehearse aloud during l i s t presentation and found that items which had been rehearsed together were l i k e l y to be r e c a l l e d together. From t h i s r e s u l t , i t can be i n f e r r e d that subjects' rehearsal strategies are part of the organizational process. Thus, i f the subject i s i n s t r u c t e d to rehearse aloud only the item being presented ( F i s c h l e r , Rundus, & Atkinson, 1970), i t i s possible that h i s l e v e l of organization would be depressed. In Experiment I I , subjects i n the experimental group were required to rehearse aloud the item which was being presented. Sub-j e c t s i n the c o n t r o l group were not required to rehearse aloud and were not asked to use any p a r t i c u l a r rehearsal strategy. I f the e f f e c t of organization i s on r e t r i e v a l capacity, i t could be predicted that sub-j e c t s required to rehearse aloud would r e c a l l less than control subjects but would not d i f f e r from co n t r o l subjects i n recognition performance. Method Sti m u l i and M a t e r i a l s . The s t i m u l i were 480 words, three from each of 160 conceptual categories i n the B a t t i g and Montague (1968) and Shapiro and Palermo (1969) norms. The words were high associates to the category 62 names. One word from each of the categories was chosen to be on the lists to be presented; the remaining two words from each category served,as distractors for the recognition conditions. The 160 words to be presented were typed on 3 X 5 inch notecards. Th2 notecards were then randomly divided into four lists of 40 words each. The lists were arbitrarily designated Lists A, B, C, and D. Within each l i s t , the notecards were randomly arranged to form a presen-tation order for that l i s t . Recall answer sheets were prepared in the same manner as in Ex-periment I, except there were only 20 blank lines in each of two col-umns. Four separate recognition answer sheets were prepared, one for each of the four lis t s . A recognition answer sheet for a given l i s t contained a random arrangement of the 40 words on that l i s t and the two distractor words from the categories of those 40 words. The recognition answer sheets were arranged in four columns of 30 xjords each, and the instructions at the top of the recognition answer sheets were the same as those used In Experiment I. Design. As Experiment I, Experiment II could be conceptualized as a 2 X 2 factorial design. The first factor would be rehearsal con-ditions, silent or overt, and the second factor would be type of test, recognition or recall. Tnus, there would be four conditions, Silent- Recognition , Overt-Recognition, Silent-Recall, and Overt-Recall. However, unlike Experiment I, the four conditions were varied within subjects so that each subject served in each of the four condi-tions. Each subject was presented each of the four lists and served in 63 one of the four conditions on each l i s t . To counterbalance order of conditions and presentation of l i s t s , a Graeco-Latin square design was used. The exact Graeco-Latin square i s i l l u s t r a t e d i n Table IV. As can be seen from Table IV, the Graeco-Latin square resulted i n four arrange-ments of l i s t s and conditions f o r presentation. Subjects. The subjects were 20 volunteers from introductory psychology classes. Five subjects served i n each of the four Graeco-L a t i n square arrangements. The subjects were tested i n d i v i d u a l l y . Procedure. Each subject was read two sets of i n s t r u c t i o n s , one set before he served i n each of the two rehearsal conditions. In n e i -ther set of i n s t r u c t i o n s was the subject informed of the type of test which would follow the presentation of the l i s t . For both sets of i n s t r u c t i o n s , the subject was t o l d that he would be presented 40 words that were f a m i l i a r to him. He was shown a deck of notecards on which one of the l i s t s of words was typed and was t o l d that he would be presented the l i s t , one word at a time, by the experi-menter turning over the notecards. The subject was t o l d that a card would be turned over every f i v e seconds and that he should try to remember as many words as he could. In addition, f o r the s i l e n t re-hearsal condition, the subject was t o l d he could study the words i n any way that he wished. For the overt rehearsal condition, the subject was t o l d to repeat aloud the word he was currently being presented. He was to repeat the word throughout the f i v e seconds that he saw i t and repeat i t at a constant rate, about once every second. A f t e r the subject was read the set of i n s t r u c t i o n s appropriate f o r 64 TABLE IV Graeco-Latin Square Design Used in Experiment II Order of Presentation Arrangement List Condition List Condition L i s t Condition L i s t Condition C Overt-Recall B Overt- A Silent- D Silent-Recognition Recall Recognition B Silent-Recall C Silent- D Overt- A Overt-Recognition Recall Recognition A Silent- D Silent- C Overt- B Overt-Recognition Recall C Recognition Recall 4 D Overt- A Overt Recognition Recall B Silent- C Silent-Recognition Recall 65 his f i r s t two l i s t s , the f i r s t l i s t was presented. The c l i c k of an e l e c t r i c timer paced the experimenter's presentation. Immediately f o l -lowing the completion of f i r s t l i s t presentation, the subject was handed an answer sheet appropriate to the condition and l i s t under which he was being tested. He was given two minutes to complete the answer sheet. Then, he was t o l d that he would be presented another l i s t and he was to rehearse t h i s l i s t i n the same way as he rehearsed the f i r s t l i s t . F o l -lowing the t e s t of the second l i s t , the subject was read i n s t r u c t i o n s appropriate f o r rehearsal of the t h i r d and fourth l i s t s . The t h i r d and fourth l i s t s were then presented and tested. Results The subject's protocols were scored as i n Experiment I. The means (number correct minus number incorr e c t ) x^ere 27.60, 28.60, 12.75, and 10.00 for the Silent-Recognition, Overt-Recognition, S i l e n t - R e c a l l and Overt-Recall conditions, r e s p e c t i v e l y . Means and variances f o r the four conditions are contained i n Table V. Again, the data were trans-formed using a square root transformation. Transformed means and variances f o r the four conditions are also contained i n Table V. Analysis of variance was used to examine the transformed data i n Table V. Variables that were examined x<rere L i s t s (A, B, C, and D), Graeco-Latin square Arrangements (1, 2, 3, and 4), Presentation Order (the f i r s t , second, third, or.fourth l i s t learned by the subject), and Conditions (recognition vs. r e c a l l , s i l e n t vs. overt rehearsal). The r e s u l t s of t h i s analysis are contained i n Table VI. The s i g n i f i c a n t e f f e c t of presentation order indicates that subjects performed best on 66 TABLE V Means and Variances and Transformed Means and Variance for the Four Conditions i n Experiment II Group Measure Recognition R e c a l l S i l e n t Overt S i l e n t Overt x 27.60 28.60 12.75 10.00 s 2 x 41.73 40.15 23.99 14.84 v^T 5.22 5.31 3.50 3.10 s2v£"~ 0.40 0.42 0.50 0.45 67 TABLE VI ANOVA Source Table for Conditions in Experiment II Source df SS ms F P Total 79 112.04 - - -BETWEEN SUBJECTS 19 21.83 - - -Arrangements (1) 3 3.60 1.20 1.05 n.s. ERROR 16 18.23 1.14 - -WITHIN SUBJECTS 60 90.22 - - -Presentation Order (2) 3 0.92 0.31 3.59 <.025 1 X 2 9 86.16 86.16 - -Lists 3 2.50 0.83 9.79 <,001 Recognition vs. Recall 1 76.93 76.93 905.04 <.001 Silent Recall vs. Overt Recall 1 1.58 1.58 18.57 <.001 Silent Recognition vs. Overt Recognition 1 0.08 0.08 <1 n.s. Residual 3 5.07 1.69 21.12 <.001 ERROR 48 4.06 0.08 mm 68 the f i r s t l i s t presented; t h i s f i n d i n g probably indicates the e f f e c t of proactive i n h i b i t i o n between l i s t s . The s i g n i f i c a n t e f f e c t of l i s t s i ndicates that L i s t s B and C were easier to learn than L i s t s A and D. The i n s t r u c t i o n to rehearse overtly resulted i n a s i g n i f i c a n t decrement i n r e c a l l performance but no s i g n i f i c a n t decrement i n recognition per-formance. Thus, Experiment II has r e p l i c a t e d Experiment I by showing an e f f e c t of organization i n r e c a l l but not i n recognition. (The s i g -n i f i c a n t r e s i d u a l term contains too few degrees of freedom for further a n a l y s i s . However, i t does indi c a t e that a between-subjects, rather than a ttfithin-subjects design be used i f the experiment i s to be r e p l i -cated.) A second analysis examined r e c a l l as a function of s e r i a l p o s i t i o n for the Overt-Recall and S i l e n t - R e c a l l conditions. P r o b a b i l i t y of r e c a l l as a function of four blocks of 10 s e r i a l p o s i t i o n s i s graphed i n Figure 2. Analysis of variance was used to examine the data shown i n Figure 2. The dependent messure was the number r e c a l l e d at each of the four blocks of input positions for each subject. The r e s u l t s of the analysis are contained i n Table VII. The s i g n i f i c a n t e f f e c t of t r i a l s indicates the presence of a s e r i a l p o s i t i o n e f f e c t . The s i g n i f i c a n t S i l e n t - R e c a l l vs. Overt-Recall X S e r i a l P o s i t i o n i n t e r a c t i o n indicates that the better r e c a l l of the S i l e n t - R e c a l l condition was l i m i t e d to the primacy and middle portions of the. s e r i a l p o s i t i o n curve. This f i n d i n g i s i n accord with the r e c a l l data reported by F i s h i e r et a l . (1970). General Discussion The data from the two experiments c l e a r l y suggest that the e f f e c t 69 FIGURE 2 Probability of Recall as a Function of Serial Position for Silent-Recall and Overt-Recall Conditions in Experiment II .50 r - — -SERIAL POSITION * 'jf 70 TABLE VII ANOVA Source Table for Silent-Recall and Overt-Recall Conditions as a Function of Serial Position in Experiment II Source df SS ms F P TOTAL 159 642.84 - - -BETWEEN SUBJECTS 19 98.22 - - -WITHIN SUBJECTS 140 544.63 - - -Silent Recall vs. Overt Recall (1) 1 18.91 18.91 7.66 <.01 Serial Position (2) 3 170.02 56.67 22.96 <.001 1 X 2 3 27.48 9.16 3.71 <.025 Error 133 328.23 2.47 _ 71 of the organizational manipulations was on r e c a l l but not on recognition performance. The conclusion from t h i s f i n d i n g i s that organization a f f e c t s r e t r i e v a l capacity but not storage capacity. It i s also c l e a r that the organizational manipulations used i n the present experiments, e s p e c i a l l y the imagery i n s t r u c t i o n , affected the form of the stored information; that i s , the imagery i n s t r u c t i o n encouraged subjects to organize the items In groups of f i v e and to use a v i s u a l code for these groups. However, the form of the stored inform-ation apparently had no e f f e c t on storage capacity as recognition scores did not vary as a function of the organizational manipulations. Thus, the evidence supports the contentions of Tulving (1963) and Kintsch (1970) and suggests that organization serves as a r e t r i e v a l a i d . In addition to supporting the free r e c a l l theories of Tulving and Kintsch, the present data support Kintsch's (1970) two process theory of recognition and r e c a l l . Kintsch's two process theory includes the ar-gument that there are q u a l i t a t i v e , rather than q u a n t i t a t i v e , differences between recognition and r e c a l l , and i t i s counter to the notion that the s u p e r i o r i t y of recognition as compared to r e c a l l performance l i e s i n the greater "strength" of items necessary for r e c a l l of those items. Data contrary to the "strength" notion would be showing the e f f e c t of an independent va r i a b l e to d i f f e r as a function of whether recognition or r e c a l l tests are used. The present data, which showed that the organ-i z a t i o n a l manipulations affected r e c a l l but not recognition performance, i s an example of the type of data Kittsch (1970) c i t e s as support for h i s two process theory. 72 One major diffe r e n c e between the two experiments involved the i n t e r a c t i o n between the r e c a l l conditions and s e r i a l p o s i t i o n . In Ex-periment I, the imagery i n s t r u c t i o n resulted i n greater r e c a l l across a l l blocks of s e r i a l p o s i t i o n . In Experiment I I , overt rehearsal resulted i n depressed r e c a l l only i n the primacy and middle portions of the s e r i a l p o s i t i o n curve. This difference could imply that the imagery i n s t r u c t i o n affected the recency or short-term memory component of the input order. The appropriate t e s t of th i s i m p l i c a t i o n would involve showing improved performance i n a Brown-Peterson (Brown, 1954; Peterson & Peterson, 1959) short-term memory paradigm when imagery i n s t r u c t i o n s are used. Although such an experiment has not been conducted, Wickens and Engle (1970) have found that words rated high i n imagery are r e c a l l e d b e t t e r i n the Broxm-Peterson paradigm than are words rated low i n imagery. The s e r i a l p o s i t i o n curves found i n Experiment II may be described i n terms of opportunity for rehearsal. The elim i n a t i o n of the primacy e f f e c t f o r the overt rehearsal condition may be due to the loss of op-portunity f o r extra rehearsal of e a r l y items. The f a i l u r e to f i n d d i f -ferences between overt and s i l e n t rehearsal groups i n r e c a l l of recent items may r e f l e c t no loss of opportunity for rehearsal of those items. Similar arguments have been advanced by Fundus and Atkinson (1970) and Fisc h l e r et a l . (1970). In conclusion, the two experiments reported i n th i s chapter i n -dicate that the e f f e c t of organization i s to increase r e t r i e v a l , but not storage, capacity. Implications of th i s f i n d i n g i n terms of a theory of free r e c a l l w i l l be discussed i n Chapter 7. 73 CHAPTER 4 CLUSTERING AND SUBJECTIVE ORGANIZATION In Chapter 1, a major distinction between two types of organiza-tion, clustering and subjective organization, was made. Operationally, clustering and subjective organization differ in the types of items presented and the methods of measurement employed. For clustering, the items presented have an a priori structure, and clustering is measured in terms of the amount of conformity of the subject's recall protocols to that structure. For subjective organization, the items presented are unrelated, and subjective organization is measured in terms of the sub-ject's tendency to repeat adjacencies from one t r i a l to the next. Fur-thermore, subjective organization can be measured only in multitrial free recall, while clustering may be measured in single-trial free recall. Similarities between the effects of experimental manipulations on clustering and subjective organization were also discussed in Chapter 1. Slowing presentation rate will result in higher recall and higher organ-ization scores regardless of whether the materials presented have an a_ priori structure. Also, order of input can be manipulated so that in-creases in recall and organization scores can be observed for both categorized and unrelated l i s t s . Cueing leads to increased recall from both lists of taxonomic categories and lists of subjective categories, and number of categories appears to have an effect on recall performance regardless of whether the categories are experimenter-determined or subject-determined. The data reported in Chapter 3 may also be inter-preted as showing that an experimental manipulation has similar effects 74 on both categorized and unrelated l i s t s . I n s t r u c t i n g subjects to organ-i z e (Experiment I) or not to organize (Experiment II) l i s t s of unrelated words a f f e c t s r e c a l l and recognition performance i n the same way that varying the strengths of category associates a f f e c t s r e c a l l and reco-g n i t i o n performance (Kintsch, 1968: Bruce & Fagan, 1970). In Chapter 2, c l u s t e r i n g and subjective organization x?ere d i s -cussed i n terms of Tulving's (1968) and Kintsch's (1970) theories of free r e c a l l . Tulving, apparently, does not make a d i s t i n c t i o n between the processes underlying c l u s t e r i n g and subjective organization; how-ever, he does not describe how properties of items are abstracted, or even what properties are abstracted, so that appropriate S-units are formed. Kintsch, on the other hand, appears to d i s t i n g u i s h between c l u s t e r i n g and subjective organization. C l u s t e r i n g i s based on the s i m i l a r i t i e s between the items on the l i s t , while subjective organiza-t i o n may be i d i o s y n c r a t i c . However, Kintsch does not suggest d i f f e r e n t mechanisms underlying c l u s t e r i n g and subjective organization. Kintsch's assertion that s i m i l a r i t y underlies c l u s t e r i n g has been tested i n an experiment reported by Kintsch, M i l l e r , and Hogan (1970). In t h i s experiment, some subjects were presented a l i s t of 40 c a t e g o r i -zed words ( f i v e words from each of eight categories) and were asked to make s i m i l a r i t y judgments on those words. S i m i l a r i t y judgments were made by having the subjects sort words in t o categories, and subjects vrere allowed to use anywhere between two and eight categories. The basic data were the number of subjects who sorted each p a i r of words int o the same category; the more subjects who sorted each p a i r of words 75 into the same category, the more similar that pair was considered to be. Two self-paced sorting t r i a l s were used. Other subjects in the Kintsch et a l . experiment were asked to free re c a l l the same categorized l i s t of words. The basic data from the free recall experiment were the number of times each pair was recalled adjacently. The subjects were given five free re c a l l t r i a l s , and data from the f i r s t and f i f t h t r i a l s were analyzed. The free recall task was self-paced, and each t r i a l was followed by 15 seconds of counting backwards. The data from both the sorting and free r e c a l l tasks were used to generate four proximity matrices: one matrix each for the f i r s t sort, second sort, f i r s t r ecall t r i a l , and f i f t h r e c a l l t r i a l . The proximity matrices contained, for each pair of items, the number of subjects who, in the case of sorting, sorted that pair together or, in the case of free r e c a l l , recalled that pair adjacently. Each of the four proximity matrices was subjected to cluster analysis (BCTry: Tryon & Bailey, 1970). Marked similarity was found between the cluster structures produced from the sorting matrices, and both sorting and recall matrices were highly similar to the a p r i o r i structure of the l i s t , thus suppor-ting Kintsch's (1970) assertion that order of output i n the free recall of a categorized l i s t reflects similarity. In the present chapter, an attempt was made to determine whether Kintsch's notion of similarities could be extended to a l i s t of unrelated words. If order of output from a l i s t of unrelated words can be predicted by similarity judgments, i t would be additional evidence that the same processes underlie both clustering and subjective organization. In this 76 dissertation, similarity w i l l be regarded as a measure of proximity from tasks i n which subjects are asked to make "similarity 1' judgments. Experiment III Experiment III was designed to determine whether i t is worthwhile to pursue the notion of similarity as a predictor of output order i n free r e c a l l when the l i s t of items does not have an a p r i o r i structure. In Experiment I I I , a nine-item l i s t of "unrelated" words served as stimuli. Some subjects made similarity judgments on the nine words, and other subjects were given a multitrial free recall task x^ith the same nine words serving as stimuli. Because of the short length of the l i s t , a measure of similarity x^hich yields more data per subject than does the card sorting task used by Kintsch et a l . was employed. The measure x*as that of triads judgments (e.g., Wexler, 1970). For triads judgments, the subject is presented a l l possible sets of three of the stimulus words. The subject's task i s , for each set, to select the word x*hich he thinks i s least similar to the other two x^ords in that set. Method Stimuli and materials. The stimuli were nine xjords chosen ran-domly from Lists 3 and 4 of Tulving's (1964) l i s t s of unrelated words, x-Tith the restriction that the six most infrequent words in each l i s t not be chosen. For the triads task, a l l 84 possible sets of three words were used. The order of words within each set was randomly deter-mined, and the entire set of 84 triads x^ as randomly arranged for presen-tation. This arrangement of triads appeared in the ansx*er booklets, xtfith 24 triads on each of three pages and the remaining 12 triads on 77 a fourth page. For the free r e c a l l task, the nine words were mounted as s l i d e s f o r presentation v i a a carousel p r o j e c t o r . Answer booklets contained 12 pages with nine blank l i n e s per page. Design. The design was c o r r e l a t i o n a l i n nature. The basic data for both tasks were proximity matrices. The proximity matrix for the t r i a d s task contained the number of times, collapsed over subjects, each p a i r of words was chosen as more s i m i l a r than the t h i r d x^ord i n the t r i a d . The proximity matrix f o r the free r e c a l l task contained the number of times, collapsed over subjects and t r i a l s , each p a i r of words was r e -c a l l e d adjacently. The measure of the r e l a t i o n s h i p between the two matrices was a Pearson product-moment c o r r e l a t i o n c o e f f i c i e n t . Subjects. For the t r i a d s task, the subjects were 17 volunteers from introductory psychology classes. The subjects were tested i n groups of 1-3. For the free r e c a l l task, the subjects were 17 volunteers from undergraduate education classes. These subjects were tested i n one group session. Procedure. For the t r i a d s task, each subject was given an answer sheet and was t o l d that, on each l i n e of that sheet, there xrere three words. His task was to cross ^ut the word on each l i n e that he thought was l e a s t s i m i l a r to the other two xrords on that l i n e . The meaning of s i m i l a r i t y x<ras not defined. The task was self-paced. For the free r e c a l l task, the subjects were read standard m u l t i -t r i a l free r e c a l l i n s t r u c t i o n s which emphasized that the words could be r e c a l l e d i n any order. Words xvere presented at a one second rate, and there were 12 t r i a l s of free r e c a l l learning. There was a one minute interval between t r i a l s during which the subjectswere to recall as many words as they could. During this interval, the experimenter shuffled the slides in the slide tray to determine order of presentation for the next t r i a l . Results Mean recall for the f i r s t t r i a l was 6.56 items, and mean recall for the f i n a l t r i a l was 9.00 items. The mean number of items recalled per t r i a l ranged, across subjects, from 7.57 to 8.92 items, with an average of 8.42 items. Proximity matrices for both tasks contained 36 c e l l s , i.e. one c e l l for each possible pair of items. For the triads task, each pair was presented in seven triads. As there were 17 subjects in the triads task, the maximum possible c e l l entry was 119. For the free recall task, there were 12 t r i a l s and 17 subjects. The maxi um possible c e l l entry, then, was 204. The correlation between the proximity matrices for the free recall and triads tasks was r_ = .70. This correlation was highly significant, p_ < .001. However, the relationship between order of output and the triads judgments is underestimated. F i r s t , the data from the recall task included early t r i a l s in which short-tern memory and input order are important determinants of output order. That i s , the length of the l i s t was l i t t l e beyond the memory span, and a good portion of ea r l y - t r i a l re-c a l l may have reflected the order of input. As a l l subjects had the same presentation orders, the effect of input order was not controlled. Second, although the Instructions did not define similarity, the data 79 indicated that the triads subjects were rating mostly on similarity of "meaning." However, the recall protocols indicated that other types of similarity were used by the free recall subjects. For example, xylem and zenith were the only too words on the l i s t with the same i n i t i a l sound, and this pair was ranked second in frequency of adjacent recall. However, xylem-zenith was the fifteenth ranked pair In frequency of judged similarity. Elimination of xylem-zenith from the analysis raised the correlation from .70 to .76. Experiment IV Experiment IV was designed as a replication and extension of Exper-iment III. Experiment IV differed from Experiment III in that a longer l i s t of words was used and similarity judgments were obtained from a card sorting task. For the card sorting, the basic data were, for each pair of items, the number of subjects who placed that pair In the same category. For the free recall task, the basic data were, for each pair, the number of times that pair was recalled adjacently on the f i n a l three t r i a l s . As in the Kintsch et a l . (1970) study, proximity matrices were generated from which cluster structures were derived. Descriptions of the cluster analysis programme used and a method for comparing cluster structures follow. Cluster analysis. The cluster analysis programme used was HGroup (Veldman. 1967). The form of the data for HGroup is an ordered vector for each item. For example, in the present analysis, the vector for a given word is the number of times i t was sorted together (in the case of card sorting) or recalled adjacently (in the case of free recall) with 80 every other word. The method of HGroup begins by t r e a t i n g each word as a category. At each step, the number of categories i s reduced by one. For the f i r s t reduction, the two words which are c l o s e s t , i n terms of the minimum of the squared differences between t h e i r vectors, are combined. Then, the items which have been combined are given a vector which i s the mean of the two vectors, and the process recycles. At the end of the method, there are two categories. HGroup also y i e l d s s e l e c t i o n values f o r optimal number of catego-r i e s . These s e l e c t i o n values, although not e x p l i c i t l y explained by Veldman, r e f l e c t the cost, i n terms of squared deviations, of making the next reduction i n the number of categories. The s e l e c t i o n values are read so that the larger the s e l e c t i o n value at a p a r t i c u l a r number of categories, the more r e l i a b l e that category structure i s . Cluster comparison. HGroup w i l l produce category structures from both the card s o r t i n g and free r e c a l l data. Then, i t Is necessary to compare the two c l u s t e r structures i n order to determine i f they are r e l a t e d . Schwartz and Humphreys ( i n preparation) have proposed a method for t h i s comparison. B a s i c a l l y , two variables are involved i n the anal-y s i s : (1) the number of s p e c i f i c p a i r s of items which are contained i n both c l u s t e r structures: and (2) the number of s p e c i f i c pairs of items contained i n the f i r s t structure but not i n the second plus the number of s p e c i f i c p a i r s of items contained i n the second structure but not i n the f i r s t . The analysis r e s u l t s i n a chi-squared s t a t i s t i c with one degree of freedom. The formula f o r the chi-squared s t a t i s t i c involves the 81 d e v i a t i o n of the observed from the expected values of (1) and (2) described i n the previous paragraph. Again, Schwartz and Humphreys should be r e f e r r e d to for the s p e c i f i c formula. As the analysis r e s u l t s i n a chi-squared value, the contingency c o e f f i c i e n t (see, e.g., S i e g a l , 1956) may be used to estimate the magnitude of the r e l a t i o n s h i p . Extension of Experiment I I I . Besides r e p l i c a t i n g Experiment III with a longer l i s t of items and a d i f f e r e n t measure of s i m i l a r i t y , Exper-iment IV xtfas also designed to extend the conclusions of Experiment I I I . This extension was an attempt to p r e d i c t r e c a l l performance as a function of the amount of conformity with the c l u s t e r structure produced by the card s o r t i n g task. That i s , i f the l i s t of items were treated as i f i t had an a p r i o r i structure and organization were measured i n terms of con-formity to that structure, could the same re l a t i o n s h i p s between organiza-t i o n and r e c a l l observed by Tulving (1962b) and Shapiro and B e l l (1970), who used subjective organization scores, be demonstrated? The r e l a t i o n -ships observed by Tulving were (1) that increases i n organization scores predicted increases i n r e c a l l over t r i a l s and (2) that differences be-tween subjects i n organization scores predicted differences i n r e c a l l . The r e l a t i o n s h i p s observed by Shapiro and B e l l (1970) xjere that organiza-t i o n and r e c a l l increased over t r i a l s f o r high organizers, but r e c a l l increased over t r i a l s , without increases i n organization, for low organ-i z e r s . Simulations of the Tulving (1962b) and Shapiro and B e l l (1970) r e s u l t s would provide a d d i t i o n a l evidence that the same processes under-l i e both c l u s t e r i n g and subjective organization.1 1. The term simulation i s used because the differences i n measures of organization make the use of r e p l i c a t i o n inappropriate. 82 Method Sti m u l i and materials. The s t i m u l i were 40 words chosen randomly from Tulving's (1964; L i s t s 1-4) l i s t s of unrelated words, with the r e s t r i c t i o n that the s i x most infrequent words i n each l i s t not be chosen. For use i n the card s o r t i n g task, each word was typed on a 3 x 5 inch card. Three sets of 40 cards each were prepared. Answer sheets for subjects In the card s o r t i n g task contained eight blank columns. For the free r e c a l l task, the 40 words were mounted as s l i d e s . Also, ten subtraction problems using t h r e e - d i g i t mumbers were chosen, and these problems were mounted as s l i d e s . Answer booklets contained ten pages, with a space f or the answer to the subtraction problem and 40 blank l i n e s f o r the r e c a l l of words on each page. A carousel projector was used f or presentation. Design. The design contained a number of components, a l l of which were c o r r e l a t i o n a l i n nature. The f i r s t of these components i n -volved the r e p l i c a t i o n of Experiment I I I . Proximity matrices were gen-erated from both the s o r t i n g task and the free r e c a l l task. The basic data f o r these matrices were the number of times each p a i r of words was sorted i n t o the same category or r e c a l l e d adjacently for the s o r t i n g and r e c a l l tasks, r e s p e c t i v e l y . The second component involved the simulation of Tulving's (1962b) study of subjective organization i n free r e c a l l when organization was measured as conformity to the categories produced by the c l u s t e r analysis of the s o r t i n g data. For each subject, a z-score was determined by trea t i n g each subject's ten t r i a l s as a s i n g l e p r o t o c o l . Then, c l u s t e r i n g 83 and r e c a l l performance as a function of t r i a l s , collapsed over subjects, were compared. A comparison of ^ -scores and r e c a l l performance as a function of subjects, collapsed over t r i a l s , was also made. The t h i r d component involved a simulation of Shapiro and B e l l ' s (1970) study of subjective organization i n free r e c a l l when organization was measured as conformity to the categories produced by the c l u s t e r anal-y s i s of the sorting data. On the basis of t o t a l z-scores, the subjects were divided i n t o groups of high, middle, and low c l u s t e r e r s , and com-parisons of z-scores and r e c a l l performance as a function of t r i a l s were made among these groups. Subjects. For the s o r t i n g task, 40 volunteers from introductory psychology classes served as subjects. They were seen i n groups of 1-3. For the free r e c a l l task, the subjects were 47 introductory psychology students. They were seen i n one group session, and the experiment x«;as run during t h e i r regularly-scheduled l e c t u r e period. Procedure. For the s o r t i n g task, each subject was given a set of cards on which the 40 words were typed. The subject was t o l d to go through the cards, one at a time and at h i s own pace, so that he could get a " f e e l 1 ' f o r the types of words that were on the cards. Then, he was to go through the deck again, t h i s time s o r t i n g the words in t o categories. He was tol d to sort according to the meanings of the *:ords, and he was to use anywhere between two and eight categories. The subject was allowed to change words or categories u n t i l he f e l t that he had made the best sort that he could. Then, he was to write down a l l the words i n each category on h i s answer sheet. He was to use the columns to 84 represent categories. The subject was also t o l d : (1) that he would not be tested on the words; (2) that there i s no preconceived "correct" s o r t ; and (3) that there i s no preconceived " c o r r e c t " number of categories. The order of the cards was randomized before the cards were given to the subject. For the free r e c a l l task, the subjects were read standard m u l t i -t r i a l free r e c a l l i n s t r u c t i o n s which emphasized that the words could be r e c a l l e d i n any order. They were t o l d that, on each t r i a l , they would be presented 40 words, one at a time, and that the l a s t word would be followed by a subtraction problem. They were to answer the subtraction problem, and they could use the margin of the answer sheet to perform the arithmetic. A f t e r they recorded the answer to the subtraction pro-blem, they were to write down as many words from the l i s t as they could remember. The words were presented at a one second rate, and there was a two minute i n t e r v a l between t r i a l s . During t h i s i n t e r v a l , the subjects (1) answered the subtraction problem, (2) r e c a l l e d the words, and (3) turned the page3 of t h e i r answer booklets, and the experimenter s h u f f l e d the s l i d e s i n the s l i d e tray to determine the order of presentation for the next t r i a l . There were ten free r e c a l l learning t r i a l s . Results Cluster analysis. The mean and median nurribers of categories used by the subjects i n the card s o r t i n g task were 6.73 and 6 . 2 9 ,respectively. For each word, the number of subjects who sorted i t with every other word formed the vector which was the input for the c l u s t e r analysis programme. For the entry which indicated the number of subjects who sorted a word 85 with i t s e l f , the value 40, which was the number of subjects i n the card s o r t i n g task, was used. The data were analyzed both with and without the HGroup option to standardize the column scores. I d e n t i c a l solutions were obtained from both analyses, and the analysis with the standardiza-t i o n option i s reported i n t h i s section. The s e l e c t i o n values f o r the standardized solutions as a function of the number of categories are contained i n Table VIII. The s e l e c t i o n values i n d i c a t e that the most r e l i a b l e s o l u t i o n with a reasonable number of categories i s the f i v e category s o l u t i o n , which has the highest s e l e c t i o n value of any s o l u t i o n less than 24 categories. The category structure f o r the f i v e category s o l u t i o n appears i n Table IX. For the free r e c a l l task, the data from the l a s t three free r e c a l l t r i a l s were used. Mean r e c a l l scores f o r T r i a l s 8, 9, and 10 were 27.51, 29.21, and 29.56, r e s p e c t i v e l y . The l a s t three t r i a l s were used because It was desired to obtain not only the most stable data but also s u f f i c i e n t data f o r the analysis. The data from four subjects were discarded from this and subsequent analyses because these subjects r e c a l l e d i n alpha-b e t i c a l order. Although r e c a l l i n alphabetical order i s one form of organization (Tulving, 1962a), i t does not correspond with the card sorting subjects' i n s t r u c t i o n s to sort according to s i m i l a r i t y of meaning. In a s i m i l a r manner, Kintsch et a l . discarded data of subjects who r e -c a l l e d s t r i c t l y according to input order. For each word, the number of times I t was r e c a l l e d adjacently to every other word, collapsed over subjects and the f i n a l three t r i a l s , formed the vector which was the input f o r the c l u s t e r analysis programme. 86 TABLE VIII Selection Values as a Function of Number of Categories for the Cluster Analysis of the Sorting and Free Recall Data of Experiment IV Selection Values Number of Categories Sorting Free Recall 39 14.97 9.22 38 15.98 259.78 37 9.94 15.77 36 9.91 28.13 35 6.04 1.82 34 0.00 1.87 33 24,05 10.07 32 0.30 1.30 31 39.95 2.10 30 1.52 3.91 29 8.64 0.80 28 0.54 2.16 27 0.63 0.55 26 2.12 0.04 25 8.92 1.65 24 8.27 1.01 23 0.02 0.17 22 1.29 0.40 21 3.88 0.53 20 1.53 0.47 19 1.19 0.18 18 1.08 0.07 17 2.54 0.72 16 1.35 0.47 15 1.65 0.55 14 1.07 0.21 13 1.18 0.57 12 0.73 0.46 11 0.92 0.50 10 1.78 1.30 9 0.02 0.57 8 0.94 0.04 7 1.84 0.43 6 3.95 0.51 5 7.10 0.27 4 1.05 0.12 3 4.76 0.28 2 -2.00 -2.00 87 TABLE IX Category Structures for the Sorting and the Free Recall Data In Experiment IV Category Number Word Sorting Free Recall Action 1 1 Effort 1 1 Impact 1 1 Answer 1 2 Question 1 2 Finding 1 7 Rumor 1 7 Treason 1 7 Waiver 1 7 Express 1 9 Accent 1 8 Bridle 2 3 Flower 2 3 Gable 2 3 Noodle 2 3 Ether 2 7 Water 2 7 Vulture 2 8 Letter 2 10 Hermit 3 4 Miser 3 4 Gambler 3 7 Orphan 3 7 Voter 3 7 Despot 3 7 Walker 3 7 Novice 3 10 Buyer 3 9 Cherub 3 8 Legion 4 5 Union 4 5 Maxim 4 7 Zenith 4 7 Office 4 10 Quarter 4 7 Centre 4 9 Island 5 6 Ocean 5 6 Jungle 5 6 Valley 5 6 88 For the entry which indicated the number of times a word was r e c a l l e d adjacently to i t s e l f , the value 80 was used. Again the standardization option was employed. The s e l e c t i o n values for solutions as a function of a number of categories are contained i n Table VIII. The s e l e c t i o n values i n d i c a t e that the most r e l i a b l e s o l u t i .n with a reasonable number of categories i s the ten category s o l u t i o n , which has the highest value of any s o l u t i o n less than 25 categories. The category structure f o r the ten category s o l u t i o n appears i n Table IX. The number of s p e c i f i c p a i r s of items shared i n common by the two category structures was 36. The number of pair s contained i n the card so r t i n g structure but not i n the free r e c a l l structure plus the number of p a i r s contained i n the free r e c a l l structure but not i n the card s o r t i n g structure was 183. Under (he hypothesis that there i s no r e l a -tionship between the two structures, the expected number of pa i r s shared i n common i s 19.87 and the expected number unique to e i t h e r structure i s 163.13. Deviations of the observed from the expected leads to a s i g n i f -icant chi-squared s t a t i s t i c , x 2 (1) - 14.68, p_ < .001. Thus, there i s a r e l a t i o n s h i p between the two cl u s t e r s t ructures. Tulving (1962b)-type analysis. Conformity to the category struc-ture produced by the c l u s t e r analysis of the s o r t i n g data (see Table IX) was analyzed according to the multiple category runs test (Frankel & Cole, 1971). A z-score was obtained f o r each subject f o r each t r i a l . The mean z-score per t r i a l i s i l l u s t r a t e d i n Figure 3 . As with 43 subjects, a mean ^-score of 0.39 i s needed for s i g n i f i c a n c e at the .01 l e v e l , c l u s t e r i n g on a l l t r i a l s except the f i r s t was s t a t i s t i c a l l y s i g n i f i c a n t . Mean r e c a l l per t r i a l i s also i l l u s t r a t e d i n Figure 3. A 8 9 FIGURE 3 R e c a l l and Organization Scores as a Function of T r i a l s for A l l Subjects i n Experiment IV T i i i i i i i i i » 1 2 3 4 5 6 7 8 9 10 TRIAL 90 c o r r e l a t i o n between mean r e c a l l and mean ^-scores as a function of t r i a l s was T = .94, p_ < .001. This c o r r e l a t i o n i s very s i m i l a r to that of .96 reported by Tulving (1962b) who measured organization by subjective organ-i z a t i o n scores. Thus, the increase i n r e c a l l of a l i s t of unrelated x«>rds can be predicted by increases i n conformity to a category structure derived from s i m i l a r i t y data. To c a l c u l a t e a t o t a l c l u s t e r i n g score f o r a subject, his ten r e -c a l l protocols were treated as a s i n g l e protocol. From t h i s p r o t o c o l , a 2^-score was obtained. The mean of these ^-scores was 4.65 which i s s t a t i s t i c a l l y s i g n i f i c a n t . T o t a l r e c a l l scores were also obtained f o r each subject. The c o r r e l a t i o n between t o t a l r e c a l l and t o t a l z-scores as a function of subjects was s t a t i s t i c a l l y s i g n i f i c a n t , r = .49, JJ < .002. This c o r r e l a t i o n i s s i m i l a r to that reported by Tulving (1962b) ttfho found a c o r r e l a t i o n of .62 between t o t a l r e c a l l and t o t a l subjective organization scores. Thus, differences among subjects i n the r e c a l l of unrelated words can be predicted from t h e i r conformity to a category structure derived from s i m i l a r i t y data. Shapiro & B e l l (1970)-type a n a l y s i s . On the basis of the t o t a l c l u s t e r i n g scores, subjects were divided i n t o three groups. The 13 sub-j e c t s with the highest z-scores were designated as high organizers, the 13 subjects with the middle ^-scores as middle organizers, and the 13 subjects with the lowest ^-scores as low organizers. Mean ^-scores f o r the high, middle, and low organizers were 5.28, 2.37, and 0.38, respec-t i v e l y . The ^z-score f o r the low organizers indicates that they did not cl u s t e r more than would be expected by chance, J J = 1.39, £ > .05. High 91 organizers recalled a mean of 253.31 words, middle organizers recalled a mean of 229.92 words, and low organizers recalled a mean of 194.46 words. Recall and ^ -scores as a function of t r i a l s were examined for the three groups. Mean recall per t r i a l for the three groups i s shown in the upper part of Figure 4, and mean j2-scores as a function of t r i a l s for the three groups is shown in the lower part of Figure 4. Correlations between recall and ^ -scores as a function of t r i a l s were obtained for the three groups. The correlations between the rec a l l curves and the ^ -score curves for the high and middle organizers were r_ = .96 and r_ = .91, respectively. The correlation between the two curves for the low organizers was _r = .07, which i s not s t a t i s t i c a l l y significant. Thus, recall increased over t r i a l s for high, middle, and low organizers and organization increased over t r i a l s for high and mid-dle organizers. However, the increase in recall over t r i a l s for low organizers was not accompanied by a corresponding increase i n their organization scores. These results are similar to those of Shapiro and Bell (1970) , who measured subjective organization rather than conformity to a category structure. Experiment V Experiment IV replicated the most important results from correla-tional studies of the subjective organization and recall of "unrelated' words by treating a l i s t of unrelated words as a categorized l i s t , with categories derived from similarity judgments. However, one aspect of clustering according to these categories, as compared to clustering 92 FIGURE 4 Upper: Recall as a Function of Trials for High (lio), Middle (Mo), and Low (Lo) Organizers in Experiment IV Lower: Organization as a Function of Trials for High (Ho), Middle (Mo), and Low (Lo), Organizers in Experiment IV J 1 1 1 I I l i t ' 1 2 3 4 5 6 7 8 9-10 TRIAL 93 according to categories taken from category norms, was not reproduced in Experiment IV. That i s , significant clustering was not observed i n the f i r s t t r i a l of Experiment IV, while i t Is usually observed in the single-t r i a l free r e c a l l of l i s t s taken from norms. Experiment V was an attempt to demonstrate that significant clustering, defined as conformity to the categories derived from the cluster analysis of the sorting data in Ex-periment IV, could be observed in a si n g l e - t r i a l free r e c a l l . The failure to observe significant f i r s t - t r i a l clustering in Ex-periment IV could have been due to the small mean number of items re-called, 9.53. This mean Is l i t t l e beyond the span of immediate memory, and, thus, f i r s t - t r i a l r ecall might have reflected presentation order. As the presentation order did not vary across subjects, any effect of f i r s t - t r i a l clustering may have been obscured. In line with the preceding argument, Experiment V differed from Experiment IV i n two ways. F i r s t , the presentation rate was slowed from one second to five seconds. Second, the order of presentation was randomized for each subject i n the experi-ment. Method Stimuli and materials. The l i s t of 40 words used in Experiment IV served as stimuli. For presentation, a set of 40 3 x 5 inch cards which contained those words was used. An answer sheet containing 40 blank lines was prepared for each subject. Design. No experimental manipulation was used. A l l subjects were presented the same l i s t of words. It was desired to obtain a ^ -score, which x*as the measure of category clustering, for each subject, and to 94 determine whether the mean £-score indi c a t e d that subjects were c l u s t e -ring beyond the chance l e v e l . Subjects. The subjects were ten volunteers from introductory psychology classes. They were tested i n d i v i d u a l l y . Procedure. The subjects were read standard s i n g l e - t r i a l free r e -c a l l i n s t r u c t i o n s which emphasized that the words could be r e c a l l e d i n any order. The words were then presented at a f i v e second rate. The c l i c k of an e l e c t r i c timer paced the experimenter's presentation of the words. Following the presentation of the l a s t word P the experimenter handed the subject an answer sheet on itfhich the subject was to write down as many words as he could remember. The subject was given two minutes to r e c a l l the words. Order of presentation was randomized for each subject. Results Mean r e c a l l was 14.50 words. This was greater than the mean of 9.53 observed on the f i r s t t r i a l of Experiment IV. The mean j^-score, 1.10, was also greater than that observed on the f i r s t t r i a l of Experi-ment IV, 0.09. The mean z;-score i r dicated s i g n i f i c a n t c l u s t e r i n g accor-ding to the category structure derived from the s o r t i n g task i n Experi-ment IV, js = 3.48, p_ < .001. Thus, s i g n i f i c a n t c l u s t e r i n g i n a l i s t of unrelated words which has b 6 6 H Celt £ gorized according to s i m i l a r i t y judgments has been observed i n s i n g l e - t r i a l free r e c a l l . Discussion The experiments reported i n Chapter 4 have served to reduce the d i s t i n c t i o n between the c l u s t e r i n g of items from a categorized l i s t and the subjective organization of items from an unrelated l i s t . F i r s t , 95 Experiments I I I and IV have shown that some agreement can be reached by subjects who judge the s i m i l a r i t y of items i n an "unrelated* 7 l i s t , and that these s i m i l a r i t y judgments can predict order of output when other subjects free r e c a l l these items. Thus, Experiments III and IV have r e p l i c a t e d Kintsch et a l . ' s (1970) findings of a c o r r e l a t i o n between s i m i l a r i t y judgments and output order, except that the l i s t s used by Kintsch et a l . had an a p r i o r i structure taken from category norms while the l i s t s used i n Experiments I I I and IV d i d not have such a structure. Second, Experiment IV showed that the amount of c l u s t e r i n g accor-ding to the category structure derived from s i m i l a r i t y judgments of a l i s t of "unrelated" words was a p r e d i c t o r of free r e c a l l performance. Mean r e c a l l and mean c l u s t e r i n g scores as a function of t r i a l s were found to be highly correlated. A s u b s t a n t i a l c o r r e l a t i o n was also found between r e c a l l scores and c l u s t e r i n g scores as a function of sub-j e c t s . Thus, amount of organization predicted both increases i n r e c a l l over t r i a l s and differences among subjects i n t o t a l r e c a l l , simulating the r e s u l t s of Tulving's (1962b) experiment. However, Tulving, who also used unrelated words, measured organization according to subjective or-ganization scores while Experiments III and IV measured organization ac-cording to c l u s t e r i n g scores. The difference i s that Tulving's measure tapped the amount of consistency i n a subject's output order from one t r i a l to the next, while the measure used i n Experiment IV tapped the amount of conformity of a subject's output order with a category s t r u c -ture derived from s i m i l a r i t y judgments. Th i r d , differences between high, middle, and low organizers were 96 found i n Experiment V. For high and middle organizers, there were systematic increases i n r e c a l l but no systematic increases i n c l u s t e r i n g scores over t r i a l s . Thus, Experiment IV has simulated the r e s u l t s of Shapiro and B e l l (1970) who found that the r e c a l l of low organizers i n -creased over t r i a l s without corresponding increases i n organization scores. The differences between Experiment T.V and the Shapiro and B e l l study are the same as the differences between Experiment TV and the Tulving (1962b) study. Fourth, Experiment IV found that the organization of unrelated words could be observed, i n a s i n g l e - t r i a l free r e c a l l . As one of the operational differences between the subjective organization of unrelated words and the c l u s t e r i n g of words which have an a p r i o r i structure i s that the former can only be observed i n m u l t i t r i a l free r e c a l l while the l a t t e r may be observed i n s i n g l e - t r i a l free r e c a l l , the r e s u l t s of Experiment V reduce the d i s t i n c t i o n between c l u s t e r i n g and subjective organization. Although the r e s u l t s of the experiments reported i n t h i s chapter co respond with the r e s u l t s of previous studies, some differences are ap-parent. P a r t i c u l a r l y , the correspondence between the c l u s t e r structures derived from r e c a l l and s o r t i n g data reported by Kintsch et a l . i s much greater than that found i n Experiment IV. Also, middle organizers i n Ex-periment IV showed systematic increases i n organization over t r i a l s , but middle organizers i n the Shapiro and B e l l study did not show such increases. There are a number of major differences between the Kintsch et a l . study and Experiment IV which may have led to the greater correspondence 97 between the c l u s t e r structures derived from s o r t i n g and those derived from r e c a l l observed by Kintsch et a l . F i r s t , Kintsch et a l . probably observed greater r e c a l l and thus, more stable output orders, than those observed i n Experiment IV. Kintsch et a l . used a self-paced presentation rate and Experiment IV used a one second presentation rate. Also, the obvious category structure of the l i s t s used by Kintsch et a l . probably led to greater r e c a l l than from the unrelated words used i n Experiment IV. I f one chooses to view the l i s t used i n Experiment IV as composed of the categories derived from the s o r t i n g data, then the words used by Kintsch et a l . should be thought of as more obvious category members than the words used i n Experiment IV. The use of more obvious categories may have led to a greater agreement between card s o r t i n g subjects i n the Kintsch et a l . study. That i s , the categories used i n the Kintsch et a l . study may have been mutually exclusive, such that a word sorted into one category would almost never be sorted i n t o any other category. On the other hand, an examination of the category structure found i n Experiment IV (see Table IX) suggests a number of words which may have been sorted into a category other than that i n which they appear. Thus, the r e c a l l protocols could have reflected;, to some extent, these other possible categorizations. A t h i r d reason for the superior correspondence found i n Kintsch et a l ' s study may have been that the e f f e c t s of input order were c o n t r o l l e d i n t h e i r study, but not i n Experiment IV. That i s , f o r each subject on each t r i a l , Kintsch et a l . randomized the presentation order; i n Exper-iment IV, on each t r i a l a l l of the subjects received the same presentation 9 8 order. Thus, the category structure derived from the free r e c a l l data i n Experiment IV may have r e f l e c t e d input order more than that derived from the Kintsch et a l . data. A second discrepancy between Experiment IV and a previously-pub-l i s h e d experiment xvas that middle organizers showed systematic increases l n organization, over t r i a l s i n Experiment IV, but middle organizers may not have shown such systematic increases i n Shapiro and B e l l ' s (1970) study. Thus, the present findings question Shapiro and B e l l ' s conclusion that middle organizers show increases i n r e c a l l over t r i a l s without corre-sponding increases i n organization scores. The major differ e n c e between Shapiro and B e l l ' s study and Experiment IV was that organization was measured by subjective organization scores i n the former, but organiza-t i o n was measured by c l u s t e r i n g scores i n the l a t t e r . C lustering scores may be more s e n s i t i v e than subjective organization scores because, given the r e c a l l of a p a r t i c u l a r item on t r i a l n, an increase i n subjective organization requires the adjacent r e c a l l of the s p e c i f i c item which was r e c a l l e d adjacently on t r i a l n-1, while an increase i n c l u s t e r i n g requires the adjacent r e c a l l of any item that belongs to the same category as the r e c a l l e d item (see Chapter 1). I f the argument that a more s e n s i t i v e measure of organization was used i n Experiment IV than i n the Shapiro and B e l l study i s accepted, an implication about experiments i n v e s t i g a t i n g the necessity of increases i n organization for increases i n r e c a l l performance i s suggested. That i s , i f the s e n s i t i v i t y of the organization measure used determines whether increases i n organization scores are found, studies i n which increases i n 99 r e c a l l performance without corresponding increases i n organization scores have been observed may have f a i l e d to use a s u f f i c i e n t l y s e n s i t i v e meas-ure of organization. By extension, then, the c l u s t e r i n g measure used i n Experiment IV may not have been s e n s i t i v e enough to detect increases i n organization for low organizers. A s i m i l a r argument concerning the s e n s i -t i v i t y of the organization measure has been made both by Handler (1967) and i n the f i r s t chapter of the present paper. In conclusion, the experiments reported i n the present chapter have r e f l e c t e d s t r i k i n g p a r a l l e l s between the subjective organization of a l i s t of unrelated words and the c l u s t e r i n g of a l i s t of categorized words. Although the present experiments have not proved that s i m i l a r processes underlie both c l u s t e r i n g and subjective organization, they strongly sug-gest a conclusion i n v o l v i n g s i m i l a r processes. Such a conclusion would not necessitate the exclusion of i d i o s y n c r a t i c organization i n the r e c a l l of unrelated words, but i d i o s y n c r a t i c organization has not been excluded In the r e c a l l of categorized words. For present purposes, the c l u s t e r i n g of categorized words and the subjective organization of unrelated words w i l l not be treated as q u a l i t a t i v e l y d i f f e r e n t , and the informal model presented i n Chapter 7 w i l l not describe d i f f e r e n t processes f o r the two types of organization. xoo CHAPTER 5 AN EXAMINATION OF ORDER OF OUTPUT The research reported i n t h i s chapter was conducted f o r four pur-poses. The f i r s t of these purposes was to extend the findings reported i n Chapter 4 and by Kintsch et a l . (1970) . The data reported i n Chapter 4 Indicated that s i m i l a r i t y judgments can p r e d i c t order of output from a l i s t of unrelated words, and the data reported by Kintsch et a l . i n d i -cated that s i m i l a r i t y judgments can p r e d i c t order of output from a l i s t of categorized words. In the present chapter, two types of items d i f - ' ferent from those employed e i t h e r i n Experiments I I I , IV, and V or by Kintsch et a l . were used, and order of output was compared with data from s i m i l a r i t y judgments. In Experiment VI the items were a l l taken from what could be considered one category, and i n Experiment VII the items were pai r s of x-rords, with both members of a p a i r belonging to the same category. The second purpose of the experiments reported i n t h i s chapter x*as to compare order of output i n free r e c a l l xvd.th data from s i m i l a r i t y judg-ments as a function of the type of analysis used. The items used i n Ex-periment VI had been s t i m u l i i n various s i m i l a r i t y judgments tasks from which c l u s t e r structures and multidimensional structures were derived (Henley, 1969). In Experiment VI, order of output i n free r e c a l l x^as used to generate proximity matrices. From these matrices, c l u s t e r s t r u c -tures and multidimensional structures were derived, and these structures were compared with those derived from the s i m i l a r i t y judgments data. A t h i r d purpose of the experiments x^as to provide a d e s c r i p t i o n of 101 order of output i n free r e c a l l and to examine t h i s d e s c r i p t i o n i n terms of theories of free r e c a l l . That i s , theories such as Tulving's (1968), Mandler's (1967),and Kintsch's (1970) describe a subject's i n t e r n a l structure which accounts for h i s order of output. Kintsch's theory even contains a mechanism for extracting information from long-term memory so that s i m i l a r items are r e c a l l e d adjacently. However, none of these theories i s a process theory i n terms of describing order of output; that i s , although the theories could be modified so that they are able to describe which items are most frequently r e c a l l e d together, they can-not f u l l y p r e d i c t the order of output i n a given subject's r e c a l l proto-c o l s . The de s c r i p t i o n of output order for some i n d i v i d u a l subjects i n Experiment VI may help provide a l i n k between output and i n t e r n a l structure. A fourth purpose of the experiments i n t h i s chapter was to d i f f e r -e ntiate between the three types of i n t e r n a l structures of organization: c l u s t e r s , h i e r a r c h i c a l c l u s t e r s , and multidimensional spaces. That i s , c e r t a i n r e s u l t s from the c l u s t e r a n a l y s i s , multidimensional s c a l i n g , or examination of output order may be incompatible with one of the i n t e r n a l structures or may suggest a modification of one of the i n t e r n a l structures. Rather than review the r e s u l t s which would suggest po s s i b l e modifications, the discussion of the implications f o r types of i n t e r n a l structures w i l l be made i n terms of the re s u l t s of Experiment VI. Also, Experiment VII was designed to determine whether subjects w i l l use h i e r a r c h i c a l categories, and whether the formation of h i e r a r c h i c a l categories can p r e d i c t r e c a l l performance. 102 Experiment VI Experiment VI was similar to Experiments III and IV i n that data from similarity judgments tasks were compared xtfith order of output from a free recall task. Hox<rever, Experiment VI differed from Experiment III and Experiment IV in the following ways: (1) the items were draxra from what might be considered to be one category: (2) the similarity judgments were from a previously published study; (3) the comparison was made as a function of two types of analyses; and, (4) a different distance measure x<ras used in examining order of output in free r e c a l l . The stimuli, the distance measure, and the types of analyses w i l l be discussed separately. Stimuli. The items were 30 mammal names xjhich had been used in ex-periments reported by Henley (1969). Using data from various similarity judgments tasks, Henley derived cluster structures and multidimensional structures for these items. The exact items and a typical cluster struc-ture are contained i n Table XI. The typical multidimensional solution contained three dimensions. These dimensions might loosely be labeled size, ferocity, and human-ness (Henley, 1969). The extremes of the size dimension were mouse and elephant; the extremes of the ferocity d i -mension were cox? and tiger; and the extremes of the human-ness dimension were pig and g o r i l l a . Distance measure. In Experiments III and IV the measure of dis-tance in free r e c a l l protocols was the number of times each pair of items was recalled adjacently. In Experiment VI the measure used was one of relative distance (see, e.g., Buschke & Hinrichs, 1968). For any recalled item, there are tiro items x^hicli are recalled closest to i t ; these two 103 items are given a relative distance measure of one. The next two closest items to the recalled item are given a relative distance measure of two, and so on. In Experiment VI the relative distance between each recalled item and every other recalled item was determined for each subject's final recall protocol. The relative distance between each item and a l l other items was then averaged across subjects, and a proximity matrix was obtained. This proximity matrix was used as input for the cluster analysis programme, and the matrix was collapsed into lower diagonal form for the multidimensional scaling programme. As, so far as the present author is aware, relative distance has not been used in previous studies of order of output in free recall, the validity of the relative distance measure can only be described as a function of the results of Experiment VI. Relative distance was used because i t provides more information from a single protocol than does number of adjacencies. Analysis. The cluster analysis programme used was HGroup, which was described in Chapter 4. For a hierarchical cluster analysis, HGroup was also used. That i s , although HGroup yields a selection value for the most reliable number of categories, the programme will continue to com-bine categories past the number of categories which yields the highest selection value. This combination results in a hierarchical cluster structure. The multidimensional scaling programme used was SSA-1 which is contained in the Guttman-Lingoes package of scaling programmes. SSA-1 has been described by Guttman (1968), and an empirical comparison between SSA-1 and other multidimensional scaling programmes has been reported by 104 Lingoes and Roskam (1971). The results of the multidimensional scaling of the group data did not sufficiently resemble the results of Henley's (1969) analysis. Because the failure to find sufficient resemblance can-not be traced to one specific source, the multidimensional scaling w i l l not be further discussed. Method Stimuli and materials. The stimuli were 30 mammal names taken from Henley (1969). Each of these was typed on a 3 X 5 inch card. Seven three-digit numbers were also chosen to be used as starting numbers for counting backwards following each presentation of the l i s t . Answer sheets were prepared and contained seven pages with 30 blank lines per page. Design. No experimental manipulation was used. Each subject was presented the same l i s t of items for seven t r i a l s of presentation and re-c a l l . Relative distance measures tcere obtained from each subject's seventh re c a l l t r i a l , and these distances were averaged over subjects for use as input for the cluster analysis programme. Comparisons between the solutions obtained from the free r e c a l l protocols and Henley's data were made. Also, an examination of recall protocols for two subjects who recalled a l l 30 items on Trials 4-7 was made. The relative distance between each pair of items for each of the last four t r i a l s was obtained, and these relative distances were averaged over the four t r i a l s to form a proximity matrix for each subject. Subjects. The subjects were 20 volunteers from introductory psychology classes. They were tested individually. Procedure. The subject was told that he would be given a deck of 105 30 cards and that each card contained a mammal name. He was to go through the deck at his own pace, reading each card aloud, studying i t , and then turning i t face down. Me was not allowed to go back to a card once he had turned i t face down. The subject was told that after he turned over the last card, he would be read a three-digit number. He was to count backward by ones from this number for 15 seconds. After 15 seconds he would be told to stop counting and asked to write down on the f i r s t page of his answer sheet, in any order he desired, as many words from the l i s t as he could remember. When he could think of no more words, he was to turn the page of his answer sheet. Then, the same procedure would be repeated, as i t would for a total of seven t r i a l s . During the intervals in which the subject recalled, the experimenter shuffled the cards to determine order of presentation for the next t r i a l . Results Performance. Mean rec a l l per t r i a l i s shown in Figure 5. The mean recall of over 28 items on T r i a l 7 indicates that f a i r l y complete data were available for the relative distance measure. The subject recalling the fewest items on T r i a l 7 recalled 25 items on that t r i a l . The amount recalled by subjects ranged from 18.6 items per t r i a l to 29.1 items per t r i a l , over the seven t r i a l s . The cluster structure derived from Henley's (1969) data, shown in Table XI, was used to treat the l i s t as i f i t had an a p r i o r i structure. Clustering was measured for each subject on each t r i a l according to con-formity to that structure. The clustering measure used was the z^-score (see Chapter 1). Mean ^-scores per t r i a l are shown in Figure 5. As can 106 FIGURE 5 Recall and Z-scores as a Function of T r i a l s f o r Experiment VI 107 be seen from Figure 5, the jz-scores indicate that clustering was s i g n i f i -cant at least at the .001 level on each t r i a l . A correlation between the jz-scores and recall scores was found to be highly significant, _r = .96. A total z-score was calculated for each subject by treating the seven t r i a l s as a single protocol. The range of the z-scores thus ob-tained was from 3.20 to 15.02, and the mean was 8.15, These z-scores i n -dicate that a l l subjects were clustering at less than the .001 level. A correlation between the total z-scores and recall performance was positive but did not reach significance, _r = .36, £ > .05. The conclusion from this section is that subjects clustered accor-ding to the categories derived from similarity judgments and that the amount of this clustering increased over t r i a l s along with increases in recall performance. The failure to find a significant correlation between recall performance and z-scores may be due to the self-paced presentation or the small number of subjects, rather than to the nature of the materials. Group cluster analysis. Each item was described by a vector which contained the mean T r i a l 7 relative distances between that item and every other item. The value 1.0, the minimum of possible observed relative distances, was used to describe the diagonal entries indicating the relative distance between an item and i t s e l f . These vectors formed the proximity matrix which was the input for HGroup. As in Experiment IV, the HGroup option to standardize the column entries was used. Table X contains the selection values for solutions as a function of the number of categories. These selections values indicate that the eight category solution was the most optimal solution containing fewer than 18 categories:. 108 TABLE X HGroup Selection Values for Group Data and Data for Two Individual Subjects as a Function of Number of Categories in Experiment VI Selection Values Number of Categories Group Data Subject 1 Subject 2 29 56.51 6.41 6.30 28 4.22 2.77 27.27 27 3.12 11.59 7.70 26 1.81 1.00 6.75 25 16.10 6.56 0.63 24 2.13 4.19 2.96 23 2.84 10.64 1.07 22 2.24 0.31 0.07 21 2.71 3.10 0.48 20 0.38 1.65 7.06 19 0.36 0.36 5.18 18 23.77 0.78 9.16 17 0.86 2.61 4.66 16 0.41 9^ 08 10.80 15 2.38 13.33 3.20 14 3.59 1.28 1.62 13 3.61 4.73 3.29 12 1.05 0.89 0.20 11 1.66 2.09 1.64 10 0.63 7.56 1.01 9 0.11 1.43 3.02 8 5.29 1.79 7.77 7 1.35 2.79 0.80 6 3.13 0.00 2.33 5 0.06 0.20 0.95 4 1.25 7.36 6.39 3 0.44 3.53 5.37 2 -2.00 -2.00 -2.00 109 The category structure f o r the eight category s o l u t i o n i s i l l u s -t rated i n Table XI. This category structure was compared with Henley's category structure using the Schwartz and Humphreys' (In preparation) method for category comparison. A chi-squared s t a t i s t i c indicated that the free r e c a l l category structure and Henley's category structure were s i g n i f i c a n t l y r e l a t e d x 2 ( l ) =71.27, p_ < .001. The associated contin-gency c o e f f i c i e n t was C_ = .46. The conclusion from these s t a t i s t i c s i s that the c l u s t e r structures derived from s i m i l a r i t y judgments data and from order of output i n free r e c a l l show a high degree of resemblance. The combination of categories into fewer than eight categories r e s u l t s i n a h i e r a r c h i c a l c l u s t e r structure. The h i e r a r c h i c a l c l u s t e r structure for the group data i s i l l u s t r a t e d i n Figure 6. Although Henley did not report a h i e r a r c h i c a l arrangement of her categories, the h i e r a r -c h i c a l arrangement shown i n Figure 6 appears to be i n t u i t i v e l y pleasing as the nodes are f a i r l y obvious. Possible names f o r nodes are i l l u s t r a t e d i n parentheses i n Figure 6. For the most part, with the exception of the chipmunk-squirrel category, the items contained i n higher-order nodes are f a i r l y w e l l described by those nodes. The conclusion i s that a h i e r a r -c h i c a l c l u s t e r structure may provide a good d e s c r i p t i o n of order of out-put i n free r e c a l l . However, t h i s conclusion must be viewed with caution because the h i e r a r c h i c a l arrangement may simply r e f l e c t the d i f f e r i n g p r o b a b i l i t i e s of items being contained i n categories. That i s , the ob-servation of a sensible h i e r a r c h i c a l structure may have occurred as a function of averaging r e l a t i v e distances across subjects. This p o s s i b i l i t y x*as more d i r e c t l y tested i n Experiment VII which was designed to determine 110 TABLE XI Category Structures for Mammal Names Reported by Henley (1969) and Found in Free Recall for the Group Data and Data for Two Individual Subjects in Experiment VI Category Numbor Stimulus Henley Group Data Subject 1 Subjec Chimpanzee 1 1 1 1 Gorilla 1 1 1 1 Monkey 1 1 1 1 Beaver 2 2 2 2 Racoon 2 2 2 2 Rabbit 2 2 3 4 Chipmunk 2 3 4 1 Squirrel 2 3 4 1 Bear 3 2 5 4 Elephant 3 4 5 2 Giraffe 3 4 5 5 Leopard 3 5 6 3 Lion 3 5 6 3 Tiger 3 5 6 3 Mouse 4 6 7 4 Rat 4 6 7 4 Cat 4 6 7 3 Dog 4 6 7 3 Fox 4 7 8 6 Wolf 4 7 8 6 Antelope 5 4 9 2 Deer 5 2 9 4 Cow 5 8 7 7 Pig 5 8 10 6 Sheep 5 8 10 8 Gcat 5 8 10 8 Camel 5 4 6 2 Donkey 5 8 10 7 Horse 5 8 10 7 Zebra 5 4 8 2 I l l FIGURE 6 H i e r a r c h i c a l Category Structure f or the Group Data i n Experiment VI 112 i f s u b j e c t s w i l l f orm h i e r a r c h i c a l c l u s t e r s t r u c t u r e s . I n d i v i d u a l C l u s t e r A n a l y s i s . Two s u b j e c t s r e c a l l e d a l l 30 i t e m s on each o f T r i a l s 4-7. R e l a t i v e d i s t a n c e measures were o b t a i n e d f o r t h e s e two s u b j e c t s on each o f the f o u r t r i a l s , and t h e measures were av e r a g e d a c r o s s t h e f o u r t r i a l s . Thus, a p r o x i m i t y m a t r i x was o b t a i n e d f o r e a ch o f t h e two s u b j e c t s , and t h e s e p r o x i m i t y m a t r i c e s s e r v e d as i n -p u t f o r t h e c l u s t e r a n a l y s i s programme. As i n p r e v i o u s a n a l y s e s , t h e v a l u e o f 1.0 was used f o r t h e d i a g o n a l e n t r i e s , and the s t a n d a r d i z a t i o n o p t i o n was employed. S e l e c t i o n v a l u e s f o r the two s u b j e c t s as a f u n c t i o n o f number o f c a t e g o r i e s a r e c o n t a i n e d i n T a b l e X . F o r one s u b j e c t , t h e s e l e c t i o n v a l u e s i n d i c a t e t h a t t h e t e n c a t e g o r y s o l u t i o n was t h e b e s t s o l u t i o n c o n -t a i n i n g fewer t h a n 15 c a t e g o r i e s . The t e n c a t e g o r y s o l u t i o n f o r t h i s s u b -j e c t i s shown i n T a b l e X I . F o r t h e o t h e r s u b j e c t , t h e s e l e c t i o n v a l u e s i n d i c a t e t h a t t h e e i g h t c a t e g o r y s o l u t i o n was t h e b e s t s o l u t i o n c o n t a i -n i n g fewer t h a n 16 c a t e g o r i e s . The e i g h t c a t e g o r y s o l u t i o n f o r t h i s s ub-j e c t i s shown i n T a b l e X I . C h i - s q u a r e d v a l u e s were computed f o r the r e l a t i o n s h i p s between t h e c l u s t e r s t r u c t u r e f o r t h e two s u b j e c t s , t h e group d a t a , and H e n l e y ' s (1969) d a t a ; w i t h each c h i - s q u a r e d v a l u e , a s s o -c i a t e d c o n t i n g e n c y c o e f f i c i e n t s were o b t a i n e d . The c h i - s q u a r e d v a l u e s and c o n t i n g e n c y c o e f f i c i e n t s a r e r e p o r t e d i n T a b l e X I I . A l l c h i - s q u a r e d v a l u e s f o u n d i n T a b l e X I I a r e s i g n i f i c a n t a t l e s s t h a n t h e .025 l e v e l . Thus, t h e c l u s t e r s t r u c t u r e s d e r i v e d from s i m i l a r i t y judgments d a t a c a n , t o some e x t e n t , p r e d i c t t h e c l u s t e r s t r u c t u r e s d e r i v e d from i n d i v i d u a l r e c a l l p r o t o c o l s . 113 TABLE XII Chi-squared Values and Contingency Coefficients for Relationships between Cluster Structures for Henley's (1969) Data and Group Data and Two Individual Subjects in Experiment VI Measure Structures Chi-squ?.red Contingent Coefficient Henley - Group 71.27 .46 Henley - Subject 1 62.89 .58 Henley - Subject 2 5.75 .20 Group - Subject 1 115.92 .75 Group - Subject 2 59.61 .61 Subject 1 - Subject 2 5.38 .24 114 The combination of categories i n t o h i e r a r c h i c a l categories for the two subjects i s i l l u s t r a t e d i n Figures 7 and 8. The h i e r a r c h i c a l arran-gement shown i n Figure 7 seems to be s e n s i b l e , and i t appears to c o r r e -spond, to some extent, with the h i e r a r c h i c a l arrangement for the group data. The h i e r a r c h i c a l arrangement shown i n Figure 8 seems to be less sensible and has l e s s correspondence to the group data. The conclusion Is that the i n d i v i d u a l data can be described by a model of h i e r a r c h i c a l l y -arranged categories, although the p a r t i c u l a r h i e r a r c h i e s may vary across subjects. Order of output. Examination of the r e c a l l protocols for the two subjects indicated that both subjects were r e c a l l i n g i n approximately the same order on the four t r i a l s that were analyzed. That r e c a l l was i n a f a i r l y constant order over the l a s t four t r i a l s f o r these subjects has an important i m p l i c a t i o n about a model used to describe free r e c a l l . That i s , a model would require a mechanism which would r e t r i e v e items or groups of items from memory i n a fixed order. Such a mechanism would be compa-t i b l e with a h i e r a r c h i c a l arrangement of items or groups of items. Experiment VII The data reported i n Experiment VI point to the need for further i n v e s t i g a t i o n of h i e r a r c h i c a l l y - a r r a n g e d categories. In Experiment VII, three questions about h i e r a r c h i c a l c l u s t e r i n g were asked. F i r s t , w i l l subjects form h i e r a r c h i c a l clusters? Second, w i l l increases i n h i e r a r -c h i c a l c l u s t e r i n g vary with increases i n r e c a l l as a function of t r i a l s ? T h i r d , can the arrangement of categories i n t o higher-order categories be predicted by s i m i l a r i t y judgments data? 115 116 FIGURE 8 H i e r a r c h i c a l Category S t r u c t u r e f o r I n d i v i d u a l Subject 2 i n I Experiment VI < 117 Method Stimuli and materials. The s t i m u l i were 12 p a i r s of items; both members of a p a i r were high associates to the same category. The catego-r i e s xirere chosen randomly from the B a t t i g and Montague (1969) norms, and there were no obvious r e l a t i o n s h i p s among the categories. For presentation i n the free r e c a l l task, two decks of 12 3 X 5 inch cards were used. Each deck contained the 12 pairs with one p a i r on each card. A given card contained one member of the p a i r typed above the other member of that p a i r . The d i f f e r e n c e between the decks was that the top member of the p a i r was interchanged between decks. That i s , i f Deck A contained red typed above blue, Deck B contained blue typed above red. Answer booklets for the r e c a l l task contained seven pages, and there were 12 pa i r s of blank l i n e s on each page. Seven t h r e e - d i g i t numbers were chosen to be used as s t a r t i n g numbers f o r counting backwards following each presentation of the l i s t . For use i n the t r i a d s task, a l l 220 possible combinations of three of the p a i r s were typed on 3 X 5 inch cards. The p a i r s were typed from l e f t to r i g h t on the cards, and one word i n each p a i r was typed above the other word i n that p a i r . Each p a i r was a r b i t r a r i l y assigned a number from 1-12, and the number assigned to a p a i r appeared above that p a i r . The order of the p a i r s was randomly arranged from l e f t to r i g h t on each card, and the member of a p a i r which appeared on top x^as randomly deter-mined for each p a i r on a card. The cards were numbered from 1-220, and the number of a card appeared i n i t s upper right-hand corner. Answer sheets for the t r i a d s task contained 220 p a i r s of blank l i n e s , and the 118 pairs of lines were consecutively numbered. Design. No experimental manipulation was used. The design was correlational i n nature and contained three components. The f i r s t of these components was to determine whether the free recall subjects orga-nized the pairs. For each subject, both two-way and one-way indices of subjective organization (Bousfield & Bousfield, 1966) were measured, and two-tailed _t-tests were used to determine whether overall organization scores indicated more observed than expected repetitions of adjacencies. The second component was to determine whether increases i n recall over t r i a l s were correlated with increases in organization. Mean subjective organization scores and mean recall scores were compared as a function of t r i a l s . A Pearson product-moment correlation coefficient was used to make this comparison. The third component was an examination of the relationship between the data from the triads task and the data from the free re c a l l task. For the free recall task, the proximity matrix contained the number of times, collapsed over subjects and t r i a l s , each pair was recalled adja-cently. For the triads task, the proximity matrix contained the number of times, collapsed over subjects, each pair was judged as similar to every other pair. A Pearson product-moment correlation coefficient was used to compare the two proximity matrices. Subjects. The subjects were 20 volunteers from introductory psychology classes. Ten subjects served i n the free re c a l l task, and the other ten subjects served in the triads task. A l l ths subjects were tested individually. 119 Procedure. For the recall task, the subject was told that he would be given a deck of cards and that each card contained a pair of words. He was to go through the deck at his own pace, reading each card aloud, studying i t , and then turning i t face down. He was not to go back to a card once he had turned i t face dorm. The subject was told that after he turned over the last card, he would be read a three-digit number. He was to count backwards by ones from this number for fifteen seconds. After fifteen seconds, he would be told to stop counting and asked to write down on the f i r s t page of his answer sheet, i n any order he desired, as many pairs from the l i s t as he could remember. He was to write down both members of a pair on adjacent lines, and he was told to write down both members of a pair before writing down the next pair. After he could think of no more pairs, he x^ as to turn the page of his answer sheet. Then the same procedure would be repeated, as i t would for a total of seven t r i a l s . The txjo decks of cards were alternated as stimuli on each t r i a l , and the experimenter shuffled the decks during the interval i n which the subject recalled the pairs. For the triads task, the subject x<?as told that he would be given a deck of cards and that each card contained three pairs of words. His task x-ras to decide, for each card, which two of the pairs x^ere most similar. He x^ as to go through the deck, one card at a time. For each card, he x/as to read the number in the upper right-hand corner and then read the numbers of the tx?o pairs he thought were most similar. The ex-perimenter recorded the subject's responses on the answer sheet. The deck was shuffled before being given to the subject. 120 Results Organization. The mean number of pairs recalled per t r i a l , col-lapsed over subjects, was 10.90, with a range from 10.14 to 11.57. For the two-way index of subjective organization, the mean numbers of observed and expected repetitions per t r i a l were 2.03 and 1.55, respectively. Thus, the mean difference between observed and expected repetitions per t r i a l was 0.48, with a range from -0.46 to 1.51. To determine whether these scores indicated significant organization of the pairs, a t-test was used to compare the mean of 0.48 with a test mean of 0. Tha result indicated that organization approached the accepted level of significance, _t (9) = 2.11, £ < .10. For the one-way index of subjective organization, the mean numbers of observed and expected repetitions per t r i a l were 1.13 and 0.77, respectively. Thus, the mean difference between observed and expected repetitions per t r i a l was 0.36, with a range from -0.01 to 0.74. To determine whether these scores indicated significant organization of the pairs, a _t-test was used to compare the mean of 0.36 with a test mean of 0. The result indicated that organization was significant beyond the accepted level, t_ (9) = 3.15, p < .02. Figure 9 contains the curves for mean recall and mean differences scores for the two-way index of subjective organization as a function of t r i a l s . As can be seen from Figure 9, both re c a l l and organization scores showed systematic increases over t r i a l s . The correlation between the organization and recall scores as a function of t r i a l s was significant, r = .85, p < .05. 121 122 The conclusion from this section i s that subjects can form hierar-chical categories and that the formation of these hierarchical categories correlates with recall performance. That the difference scores showed more consistent organization when the one-way, rather than the two-way, index was used, suggests that the rec a l l of categories within a hierar-chical node tended to be in a constant order. That i s , the one-way index is sensitive to only constant-order adjacencies, while the two-way index i s sensitive both to constant-order and changed-order adjacencies. Thus, the present data confirm, to some extent, the observation of constant order recall in Experiment VI. Similarities. The proximity matrices for the recall and the triads data contained 66 cells each, one c e l l for each possible set of two pairs. As there were ten subjects i n the triads task and the number of times each set of two pairs appeared i n a triad was ten, the maximum possible c e l l entry for the triads proximity matrix was 100. As there were ten subjects in the rec a l l task and there were seven t r i a l s of free r e c a l l , the maximum possible c e l l entry for the recall matrix was 70. A Pearson product-moment correlation coefficient was used to compare the proximity matrices for free r e c a l l adjacencies and triads judgments. The correlation was s t a t i s t i c a l l y significant, _r = .35, p_ < .01. Thus, i t can be concluded that the formation of hierarchical categories can, at least to a limited extent, be predicted by similarity judgments data. Discussion As there were four purposes for conducting the experiments reported 123 i n t h i s chapter, four sections w i l l be used to discuss the r e s u l t s . The discussion of the four purposes w i l l be i n the same order as they were introduced. S i m i l a r i t y . The data of Kintsch et a l . (1970) and Experiments I I I and IV were s u c c e s s f u l l y extended. Kintsch et a l . found that s i m i l a r i t y judgments can predict order of output from l i s t s of categorized words, and In Experiments I I I and IV s i m i l a r i t y judgments predicted order of out-put from l i s t s of "unrelated words. In Experiment VI, the correspondence between s i m i l a r i t y judgments data and order of output was found when the l i s t was composed of words which might be considered to belong to a s i n -gle category, and i n Experiment VII, the correspondence was found when the l i s t was composed of p a i r s of words, with both members of a p a i r belonging to the same category. Thus, the correspondence between order of output i n free r e c a l l and judged s i m i l a r i t y i s very general. In a d dition, i t was found i n Experiment VI that, when c l u s t e r i n g was measured as conformity to a category structure derived from sim-i l a r i t i e s data, increases i n c l u s t e r i n g over t r i a l s correlated with i n -creases i n r e c a l l performance. Also, there was some evidence that sub-j e c t s who were better c l u s t e r e r s were also better r e c a l l e r s . Thus Ex-periment VI r e p l i c a t e d the c o r r e l a t i o n s , which were found i n Experiment IV, between r e c a l l performance and c l u s t e r i n g . However, the s t i m u l i used i n Experiment VI were a l l "mammal" names. Thus, i t can be conclu-ded that a model of free r e c a l l should not only provide a mechanism f or organizing " s i m i l a r " items, but i t should also explain why organizing s i m i l a r items correlates with r e c a l l performance. Types of analyses. In Experiment VI, s i m i l a r i t y judgments and order of output x^ere compared using c l u s t e r a n a lysis. The data f o r the 124 analysis of the similarity judgments were gathered from a number of tasks by Henley (1969) and the measure of relationship i n free r e c a l l was the relative distance between each pair of items in the subjects' f i n a l re-c a l l protocols. The correspondence between the cluster structures derived from similarity judgments and from free recall was found to be high. The cluster analysis programme also allowed an examination of hie-rarchical clustering i n free r e c a l l . The hierarchical cluster structure found in Experiment VI appeared to be sensible in that the nodes were easily labelled, but Henley did not report a hierarchical cluster struc-ture which would allow comparison with that derived from the present re-c a l l protocols. Also, one possible artifact of the hierarchical cluster structure reported in Experiment VI was that the recall data were collap-sed over subjects. To examine this possibility, in Experiment VII cate-gories were defined for subjects in order to determine whether their or-ganization of categories into higher-order categories could be predicted by similarity judgments data. The results of Experiment VII indicated that such predictions could, i n fact, be made. The conclusions from this analysis are specific to a model of free recall which invokes similarity as a mechanism underlying organization. The results of the experiments support a model of hierarchically-arranged chunks. Order of Output Examination of re c a l l protocols for two individual subjects i n Ex-periment VI indicated that asymptotic re c a l l tended to be in constant order. The constant order of output for the individual subjects in 1 2 5 Experiment VI tended to conform x^rith the s i m i l a r i t y judgments data reported by Henley. That i s , c l u s t e r structures for the two subjects were correlated with the c l u s t e r structures reported by Henley. Also, h i e r a r c h i c a l c l u s t e r structures for the two subjects indicated some degree of s e n s i b i l i t y i n the arrangement of the c l u s t e r s . The observation of constant order of output does have some implica-tions about types of i n t e r n a l structures used to account for order of out-put i n free r e c a l l . I f a chunking structure i s used, some higher-order taechanlsm f o r d i r e c t i n g order of output of chunks i s necessary. Such a higher-order mechanism would be s i m i l a r to a h i e r a r c h i c a l arrangement of chunks. I f a multidimensional structure i s used, some form of cueing to d i r e c t a systematic search i s necessary. I f a cue i s used to describe the l o c a t i o n of more than one item, such a multidimensional structure would be s i m i l a r to a chunking structure. Type of Structures The data reported i n Experiment VI pointed to the need for further i n v e s t i g a t i o n of hi e r a r c h i c a l l y - a r r a n g e d chunks. Such i n v e s t i g a t i o n was ca r r i e d out i n Experiment VII. Experiment VII included three major findings. F i r s t , subjects w i l l organize between categories. That i s , the evidence indicated that subjects can form h i e r a r c h i c a l category structures. Second, the formation of h i e r a r c h i c a l categories correlated with r e c a l l performance. That i s , organization among categories increased over t r i a l s as the p r o b a b i l i t y of r e c a l l of the categories increased. T h i r d , s i m i l a r i t y among categories can be used to p r e d i c t the organiza-t i o n among categories. That Is, s i m i l a r i t y judgments data were 126 significantly correlated with order of output of categories i n free re-c a l l . Thus, organization of categories into larger units resembles the organization of x^ords into categories. Although the results of Experiments VI and VII have not eliminated independent categories or multidimensional spaces as possible internal structures to account for order of output in free r e c a l l , they do suggest that a hierarchical arrangement is an attractive internal structure. Such a structure, then, w i l l be pursued i n the informal model of recall presented in the f i n a l chapter. 127 CHAPTER 6 SUMMARY In Chapter 2 three questions which have formed the basis for the present investigation were asked. These were: (1) Given that organiza-tion affects r e c a l l , at what stage of the recall process is the effect felt? (2) Do the same processes underlie both clustering and subjective organization? and (3) Can a description of order of output be made which would reflect on the internal structure used to generate such output? In this chapter the results of the experiments w i l l be reviewed in terms of the three questions. The question concerning clustering and subjective organization w i l l be discussed f i r s t because most of the present experiments are relevant to this question. Clustering and Subjective Organization The present dissertation is unusual in that the research has not appeared to concentrate on a specific aspect of free r e c a l l , nor has i t been designed to test a specific contention or hypothesis. Yet, the research can be regarded as having made a specific point or as having an underlying theme. That i s , various types of organization seem to represent the same processes. In terms of the present experiments, this point may be regarded as having three subpoints. F i r s t , various manipu-lations of organization have asymmetric affects on recognition and re-c a l l . Second, there are correlations between similarity judgments data and order of output in free r e c a l l which extend over categorized and un-related l i s t s . Third, the amount of organization, as measured by a va-riety of techniques, can be used to predict recall performance. These three 128 subpoints w i l l be discussed separately. Organization and Recognition. Data reported by both Kintsch (1968) and Bruce and Fagan (1970) have shown that varying the strengths of cate-gory associatesj which a f f e c t s r e c a l l performance, does not a f f e c t recog-n i t i o n performance. Thus, I f the manipulation used by these experimenters i s regarded as a manipulation of organization, i t can be argued that the organization of a categorized l i s t may have asymmetric e f f e c t s on r e c a l l and recognition. Experiments I and II were designed to extend Kintsch's and Bruce and Fagan's findings to l i s t s which did not have an a p r i o r i s tructure. In Experiment I, i n s t r u c t i o n s to use v i s u a l imagery to organize items i n t h e i r input order were found to lead to a s i g n i f i c a n t increase i n r e c a l l performance but not a s i g n i f i c a n t increase i n recognition performance. In Experiment I I , i n s t r u c t i o n s to rehearse overtly the item being pre-sented were found to decrease r e c a l l performance but to have no e f f e c t on recognition performance. Thus, the findings of Kintsch and Bruce and Fagan have been r e p l i c a t e d with d i f f e r e n t types of materials and d i f f e r -ent manipulations of organization. Organization and S i m i l a r i t i e s . The data gathered i n Experiments I I I , IV, VI, and VII are a l l consistent with the data gathered by Kintsch, M i l l e r , and Hogan (1970). Kintsch et a l . observed c o r r e l a t i o n s between judged s i m i l a r i t y and order of output i n free r e c a l l when the materials had an a p r i o r i s t r u c t u r e . Correlations between judged s i m i l a r i t y and order of output were also observed i n Experiment I I I , IV, VI, and VII, but other types of materials were used. 129 In Experiments III and IV, the s t i m u l i were "unrelated" words. In Experiment I I I , there was a nine-item l i s t , and number of adjacencies of pairs was the measure of order of output', s i m i l a r i t y judgments were from a t r i a d s task. The measure of r e l a t i o n s h i p was a c o r r e l a t i o n between the proximity matrices. In Experiment IV, there was a 40-item l i s t and num-ber of adjacencies was again the measure of order of output; s i m i l a r i t y judgments were from a card-sorting task. Correspondence between c l u s t e r structures derived from the proximity matrices was usad to measure the r e l a t i o n s h i p . In Experiment VI, the s t i m u l i were 30 items which might be consid-ered to belong to a s i n g l e category. Relative distance between each p a i r of items on the subjects' f i n a l r e c a l l protocols was the measure of order of output. S i m i l a r i t y judgments data had been gathered i n a previously-published study (Henley, 1969). Again, correspondence between c l u s t e r structures derived from the proximity matrices was used to measure the r e l a t i o n s h i p . In Experiment VII, the s t i m u l i were 12 pai r s of words, with both members of a p a i r belonging to the same category. Number of adjacencies between the p a i r s was the measure of order of output, and a t r i a d s task was used to measure s i m i l a r i t y . The Erasure of r e l a t i o n s h i p was a c o r r e l a t i o n between the proximity matrices. Experiments I I I , IV, VI, and VII a l l showed that s i m i l a r i t y judg-ments data can pr e d i c t order of output i n m u l t i t r i a l free r e c a l l . In Experiment V, s i m i l a r i t y judgments data were used to p r e d i c t order of output In s i n g l e - t r i a l free r e c a l l . The s t i m u l i were the 40 unrelated 130 words used i n Experiment IV, and the category structure was derived from the s o r t i n g data i n that experiment. In Experiment V s i g n i f i c a n t c l u s t e r i n g according to the category structure was observed i n s i n g l e -t r i a l free r e c a l l . As one of the operational d i s t i n c t i o n s between sub-j e c t i v e organization of unrelated words and c l u s t e r i n g of categorized words i s that the former cannot be measured i n s i n g l e - t r i a l free r e c a l l , the r e s u l t s of Experiment V serve to reduce the d i s t i n c t i o n between the two types of organization. A conclusion that judged s i m i l a r i t y can be used to predict order of output appears to be w e l l supported. Correspondences have been found over d i f f e r e n t types of materials (categorized l i s t s and unrelated l i s t s ) , a large range i n number of items (9-40), d i f f e r e n t types of judged s i m i l a r i t y (card-sorting and t r i a d s judgments), two measures of order of output (adjacencies and r e l a t i v e distance), and two types of comparisons (correlations between proximity matrices and between c l u s t e r s t r u c t u r e s ) . It seems appropriate to conclude that the r e l a t i o n s h i p between judged s i m i l a r i t y and order of output i n free r e c a l l i s robust. Organization and Performance. The major findings which support the contention that increases i n organization are necessary f o r increases i n r e c a l l performance are the c o r r e l a t i o n s which have been observed between measures of organization and amount r e c a l l e d (e.g., Tulving, 1962" Mandler, 1967). Regardless of whether organization i s necessary for re-c a l l , the present experiments have y i e l d e d further evidence that organi-zation correlates with r e c a l l . The usual measures of organization are c l u s t e r i n g scores, which are 131 used when the l i s t i s composed of categorized words, and subjective organization scores, which are used when the l i s t i s composed of "unre-lated" words (see Chapter 1). In Experiment IV, the l i s t was composed of unrelated words, but clustering scores were used to measure organiza-tion. That i s , the results of the cluster analysis of the similarity judgments data were used to treat the l i s t as i f i t were composed of categories, and correlations between recall performance and clusterering scores were determined. Correlations between increases in recall and i n -creases in organization as a function of t r i a l s and between total recall and total organization scores over subjects were significant. In Experiment III, the l i s t consisted of associates to a single category name. Again, results of similarity judgments data, collected by Henley (1969), were used to treat the l i s t as i f i t were composed of associates to multiple categories i Clustering was scored according to the amount of conformity of the subjects' recall protocols to these categories. A high correlation between increases In clustering scores and increases in rec a l l performance over t r i a l s and a positive, although not s t a t i s t i c a l l y significant, correlation between total r e c a l l and total organization scores over subjects were observed. In Experiment VII, the l i s t was composed of unrelated items and, as i s typical with unrelated items, subjective organization scores were used to measure organization. However, the unrelated items used were unusual in that each item was a pair of words, with both members of a pair belonging to the same category. Even with these pairs, increases In recall over t r i a l s were found to correlate with increases in organization. 132 Another finding related to organization and performance was the simulation of the Shapiro and Bell (1970) study in Experiment IV. Here, relationships between organization and rec a l l over t r i a l s for high, middle, and low organizers were observed. In accordance with the Shapiro and Bell results, recall increased over t r i a l s for a l l three groups, and high organizers, but not low organizers, shot-red correspon-ding increases in organization scores. However, although in both Ex-periment IV and the Shapiro and Bell study unrelated words were used, organization was measured as conformity to a category structure derived from similarity judgments data in Experiment IV, xrtiile organization was measured by subjective organization scores in the Shapiro and Bell experiment. Thus, the results of experiments which used unusual types of materials and/or unusual combinations of organization measures and types of materials showed positive correlations between organization and recall performance. It can be concluded that, over a range of materials and organization measures, increases zn organzzatzon over t r i a l s predict increases in recall performance. It can also be con-cluded that differences in organization among subjects can, to some extent, predict differences in rec a l l performance. Conclusion. In Chapter 1, the clustering of categorized words was compared with the subjective organization of unrelated words. Differences between these two types of organization were found to be limited to the type of material used and the unit of measurement employed. Four manipulations were discussed which produced differences in organization 133 scores and r e c a l l performance independently of xdiether categorized or unrelated l i s t s were used. One possible conclusion i s that c l u s t e r i n g and subjective organization represent the same processes. In t h i s s e c t i o n , three more commonalities between the c l u s t e r i n g of categorized words and the subjective organization of unrelated words were noted. F i r s t , manipulations of organization of unrelated words and categorized words may produce asymmetric e f f e c t s on recogni-t i o n and r e c a l l . Second, s i m i l a r i t y judgments can p r e d i c t order of output In the free r e c a l l of both unrelated words and categorized x^ords. Th i r d , c l u s t e r i n g measures can be used to p r e d i c t r e c a l l performance when the l i s t i s composed of unrelated x^ords. Thus, the data reported i n t h i s d i s s e r t a t i o n can be viex^ed as supporting the notion of common processes underlying two operationally d i s t i n c t types of organization. Storage and R e t r i e v a l The observation i n Chapter 3 of asymmetric e f f e c t s of organiza-t i o n a l manipulations on r e c a l l and recognition performance was i n t e r -preted as evidence that organization a f f e c t s the amount re t r i e v e d but not the amount stored. The assumption underlying t h i s i n t e r p r e t a t i o n was that recognition measured storage and that r e c a l l measured storage plus r e t r i e v a l . Again, t h i s i s not the appropriate place for re-hashing the argument of whether r e t r i e v a l i s involved i n recognition, and the reader i s r e f e r r e d to papers by Kintsch (1970) and McCormack (1972) f o r data pertinent to t h i s argument. I t should be pointed out that one manipulation of organization has been shown to a f f e c t recognition performance. That manipulation 134 is one of presentation order, such that blocked presentation of category members results in better recognition performance than random presentation of category members (D'Agostino, 1969; Jacoby, 1972). A possible explanation for the discrepancy between this manipulation and the manipulations used in Chapter 3 and by Kintsch (1968) and Bruce and Fagan (1970) w i l l be advanced in the next chapter. Besides reflecting on the storage-retrieval question, the results of the experiments reported in Chapter 3, particularly Experiment I, have another implication about the nature of organizational processes. That implication i s that organization can be a control process. Ex-periment I showed that subjects w i l l use instructions to form visual images i n order to organize a l i s t of unrelated words, and that using these instructions wd.ll lead not only to increases i n recall performance, but also to increases i n clustering according to input position. As the l i s t used in Experiment I did not have an a p r i o r i structure and presentation order was randomized, i t i s li k e l y that the organization resulting from use of the imagery instruction was not the organization that the subjects would normally use. Thus, i t can be concluded that there i s some f l e x i b i l i t y , which i s under the subject's control, in organizational processes. Order of Output and Internal Structure The major finding of the experiments reported in Chapter 5 i n -volved hierarchical organization. In Experiment VI, the cluster analysis of the group recall data produced a hierarchical cluster structure which was intuitively pleasing. That i s , the combination 135 of categories i n t o higher-order categories appeared reasonable, and the nodes of the higher-order categories could be e a s i l y l a b e l l e d . In Experiment VII, h i e r a r c h i c a l c l u s t e r i n g was examined more d i r e c t l y than i n Experiment VI. The s t i m u l i were tx^o-word categories, and organization was measured as the combination of these two-word categories i n t o higher-order categories. The data indicated that the subjects xtfere organizing i n t o higher-order categories and that t h i s organization increased over t r i a l s . A t h i r d f i n d i n g i n v o l v i n g h i e r a r c h i c a l c l u s t e r i n g x^ ras the ana-l y s i s of order of output for two i n d i v i d u a l subjects i n Experiment VI. Both of these subjects r e c a l l e d i n a constant order over the last-four t r i a l s . In the context of a c l u s t e r i n g model, constant order of output x-jould necessitate a mechanism for ordering the r e c a l l of various categories. Such a mechanism could be s i m i l a r to higher-order nodes i n a h i e r a r c h i c a l c l u s t e r i n g model. Thus, the experiments reported i n Chapter 5 i n d i c a t e that h i e r -a r c h i c a l c l u s t e r s may be an adequate representation of asymptotic free r e c a l l protocols. This does not shox<r that a hierarchical, structure i s the appropriate structure to represent the set of r e l a t i o n s among words which r e s u l t s from a s i n g l e subject learning a free r e c a l l l i s t . However, i t appears convenient to explore the p o s s i b i l i t y that there i s a r e l a t i v e l y d i r e c t correspondence between the structure of organ-i z a t i o n observed i n sets of free r e c a l l protocols and a single subject's i n t e r n a l representation x/nich produces that structure. Thus, a model employing a h i e r a r c h i c a l i n t e r n a l structure w i l l be pursued i n the next chapter. 136 CHAPTER 7 THEORETICAL NOTES This chapter contains two major sections. In the f i r s t section an informal model of free re c a l l i s proposed. In the second section, the data reviewed i n the f i r s t chapter i s described i n terms of the model and some experiments suggested by the model are b r i e f l y described. Towards a Model In this section, some thoughts about a model of free recall are presented. The model i s described i n three subsections. In the f i r s t subsection, an internal structure of organization which i s consistent with the description of output order extracted from the results of the present experiments i s described. In the second subsection, the pro-cess by which the internal structure develops i s described; the process i s similar to that outlined by Kintsch (1970). In the third subsection, the relationship between organization and tests of re c a l l and recogni-tion i s described, and a possible explanation of why organization may have asymmetric effects on recognition and r e c a l l i s advanced. Internal Structure In relation to free r e c a l l , the term structure may be thought of as having three referents. The f i r s t of these i s to l i s t structure, that i s , the structure that may be manipulated by the experimenter. For instance, one type of l i s t structure may be described as items belonging to a number of mutually exclusive categories. Another type of l i s t structure may be described by pairwise measures of judged similarity among the items. The second type of structure i s that found i n subjects' 137 recall protocols, and In the present paper this type of structure has often been referred to as organization. For Instance, recall structure may be described as the amount of conformity In the subjects' recall protocols to l i s t structure. Another example of recall structure may be described as a given subject's tendency to repeat specific adjacen-cies from one tr i a l to the next. A major thesis of the present paper has been that, to some extent, recall structure reflects l i s t structure (see Chapter 4). that i s , the subject processes items from the l i s t , and the product of this processing is a mechanism which generates recall structure. In this section, the mechanism which results from the processing is of interest, and this mechanism is the third type of structure, internal structure. As a mechanism, internal structure should be thought of as a structure which results from the subject's processing of the items on the l i s t and from which the recall structure is generated. As a structure, internal structure should be thought of as the subject's representation of the relationships, or patterns of sets of relationships, among the items on the l i s t to be recalled. In the present section, these relationships are described. A model of internal structure should account for the data con-cerning order of output. That i s , the model should explain the ob-served relationship between order of output and judged similarity. Also, the model should produce an order of output which can be des-cribed as a hierarchical combination of categories. Finally, a model of internal structure should allow the development of a highly 138 consistent order of output of categories and of items within categories. The present model of internal structure is a hierarchical combina-tion of categories and items within categories. The hierarchical struc-ture develops because of some limitations on processing which are dis-cussed in the next section. Constant order of output of categories and of items within categories is related to a concept of associative strength between higher-order category codes and category codes and between category codes and items. Strength of association, here, may be thought of as the ability of a higher-order category code to e l i c i t a category code and as the ability of a category code to e l i c i t the items within that code. Because a code is described as a characteristic common to the items within that code, a code serves as a mechanism by which order of output is related to judged similarity. The present model of internal structure is one of a static structure. A static structure may be thought of as the end product of the subject's processing of the items on the l i s t , and i t is the structure from which asymptotic recall is generated. Within this structure there is some limit to the number of categories and to the number of items within a category. If the number of categories ex-ceeds this limit, i t is likely that the subject combines categories into higher-order categories. This combination keeps the items within a subcategory distinct so that a l l of the items within one subcategory are recalled before a l l of the items within a second subcategory. That such a structure is possible was suggested by the results of Experiment VII, The process by which this structure develops and the 139 l i m i t a t i o n s which r e s u l t i n the formation of " i n t e r n a l " c a t e g o r i e s are described I n the next s u b s e c t i o n . A category i s considered to be some s e r i e s of a s s o c i a t i o n s r a t h e r than as a "chunk" i n M i l l e r ' s (1956) sense. A "chunk" i m p l i e s some process of recoding or r e l a b e l l i n g two or more Items i n t o a l a r g e r item. This " u n i t i z a t i o n " of items leads to increased r e c a l l c a p a c i t y ; however, along w i t h t h i s i n c r e a s e , there i s a l o s s of i n f o r m a t i o n about the o r i g i n a l s t i m u l i . This l o s s of in f o r m a t i o n w i t h u n i t i z a t i o n was c l e a r l y demonstrated by Horowitz and Manelis ( i n p r e s s ) . T h e i r sub-j e c t s were presented w i t h three types of adjective-noun p a i r s : u n i t i z e d phrases (such as hot and dog which could be recoded as hotdog), mean-i n g f u l phrases (e.g., blue and shoes), and anomalous phrases (e.g., hot and shoes). They found t h a t r e c a l l was best f o r the u n i t i z e d p a i r s , but that r e c o g n i t i o n f o r i n d i v i d u a l items was worst f o r these p a i r s . T h e i r c o n c l u s i o n was that although u n i t i z a t i o n r e s u l t e d i n greater r e -c a l l , i t caused a l o s s of in f o r m a t i o n about the i n d i v i d u a l items. That t y p i c a l o r g a n i z a t i o n i n f r e e r e c a l l does not i n v o l v e a l o s s of i n f o r m a t i o n about i n d i v i d u a l items i s f a i r l y c l e a r . Recognition memory f o r i n d i v i d u a l items has not been shown to be worse, and i s some-times shown to be b e t t e r , f o l l o w i n g manipulations which are hypothesized to i n c r e a s e o r g a n i z a t i o n . The d i s c u s s i o n of p r e v i o u s l y p u b l i s h e d ex-periments and some o r i g i n a l experiments i n Chapter 2 i n c l u d e s a v a r i e t y of manipulations of o r g a n i z a t i o n , none of which depressed r e c o g n i t i o n performance. Thus, there i s some e m p i r i c a l evidence to r e j e c t the n o t i o n 140 of chunking as descriptive of categories which are formed in free recall. Other reasons to reject the chunking notion involve the problems of analogy from the immediate memory paradigm to the free recall task. The reader is referred to Chapter 2 for discussion of these problems. Although chunking does not seem to be an accurate description of internal structure, the idea that a subject produces a code is at-tractive. The cluster analysis of the data presented in Experiment VI indicated that the nodes were easily labelled, and informal reports of subjects who participated in that experiment indicated that they were actively naming their categories. Thus, although the evidence does not necessitate a coding process, i t is not unreasonable to postulate such a process. In this way, categories may be described by codes, and higher-order categories may be described by higher-order codes. The structure, then, may be referred as a coding hierarchy. One possible description of a coding hierarchy is multiple as-sociations between a category name and the members of that category. Recall of items within a category is ordered by some probabilistic function of the varying strengths of association between the category code and the items within that category. For instance, one possible decision rule is to have the subject, upon recall of a category code, recall the strongest associates to that code before recalling the weaker associates. Thus, one would expect certain orders of recall of items within categories to be more prevalent than other orders. As 141 an Illustration, such prevalence wee investigated for the three jungle cats used as stimuli in Experiment VI. Here, the author's intuition was that free association to the category name jungle cat would have lion as a higher associate than tiger which would be a higher associate than leopard.1 In fact, when these three items were recalled adjacently in Experiment VI, the order lion, tiger, leopard appeared 41% of the time, while the other five orders appeared a total of only 59% of the time ( X 2(l) - 32.9, p_<.001). A similar process ls postulated to describe the combination of codes into higher-order codes. Further, the highest-order codes are associated with a code such as list-membership, and the highest-order codes are recalled in an order depending upon the strength of association with l i s t member- ship. Strength of association is defined in terms of two constructs. First, associative strength between an item and code can exist for the subject prior to the experiment. Second, the strength may vary as a function of the amount of practice during learning of the l i s t to be recalled. There are other possible representations of a structure similar to that described above. For example, categories could be cued by a category examplar, rather than by a category code, and that exemplar Would exist in some series of item-to-item associations with other The author's intuition does not necessarily reflect normative as*-sociative strength, and no norms exist to the category jungle cat. How-ever, lion is a stronger associate to wild animal than is tiger which is a stronger associate than is leopard (Shapiro & Palermo, 1970). 142 members of that category. The types of associations could be des-cribed so they could account for constancy in order of output. It would be difficult to discriminate between this type of representation and a representation of category codes. As Tulving (1968) has pointed out, items which share a common code probably are associates of each other, and items which are associates of each other probably share a common code. Processing A model of processing should account for the development of the internal structure described in the previous section. There should be a reason why the subject forms categories, and this reason, in the present model is related to some limitations on processing. The dev-elopment should involve similarity among items, so that more "similar" items are likely to be contained in the same category (see Chapter 4). The model should also account for the serial position curve found in free recall (see Chapter 1), and It should describe different pro-cesses for recognition and recall (see Chapter 3). The description of processing involves two memory stores, both of which are related to the internal structure described in the previous section. The first of these stores is a long-term memory store. Here, the subject has pre-experimental characteristics of items at his disposal. The second store is a short-term memory store and is called a limited capacity processing mechanism (LCPM) . This store not only accounts for the short-term memory data described in Chapter 1 but also is the center for processing of the items. The stores are 143 r e l a t e d t o t h e i n t e r n a l s t r u c t u r e i n t h a t i n f o r m a t i o n f r o m l o n g - t e r m memory i s p r o c e s s e d i n t h e LCPM and t h i s p r o c e s s i n g p r o d u c e s t h e i n -t e r n a l s t r u c t u r e . The i n t e r n a l s t r u c t u r e i s l a b e l l e d t h e f r e e r e c a l l memory s t r u c t u r e o r FP.KS, and i t was d e s c r i b e d i n t h e p r e v i o u s s e c t i o n . D e s c r i p t i o n s o f l o n g - t e r m memory and t h e LCPM f o l l o w . The LCPM i s s i m i l a r t o t h e " b u f f e r " used by b o t h A t k i n s o n and S h i f f r i n (1965, 1963) and K i n t s c h (1970). However, i t i s a f f o r d e d p r o p e r t i e s t h a t t h e s e t h e o r i s t s d i d n o t u s e , so t h e t e r m b u f f e r i s n o t employed. The LCPM i s v i e w e d as a s h o r t - t e r m memory i n p u t - o u t p u t mechanism w h i c h has f o u r m a j o r p r o p e r t i e s . F i r s t , i n i n p u t , r e c e n t i t e m s may e n t e r and r e s i d e i n the LCPM. There i s a l i m i t t o t h e number o f i t e m s w h i c h can r e s i d e i n t h e L C P I 2 , and t h e v a r i a b l e s w h i c h a f f e c t t h e LCPM i n t h i s way have been d e s c r i b e d by A t k i n s o n and S h i f f r i n (1965, 1968). T h i s f i r s t p r o p e r t y a c c o u n t s for t h e s h o r t -t e r m memory component i n f r e e r e c a l l . The second p r o p e r t y o f t h e LCPM i s t h a t " c o g n i t i v e x^ork" i s p e r f o r m e d i n i t . T h i s c o g n i t i v e x^ork i s s i m i l a r t o t h a t d e s c r i b e d b y K i n t s c h (1970). The a s s u m p t i o n h e r e , i s t h a t when an i t e m i s i n t h e LCPM, i t s l o n g - t e r m memory c h a r a c t e r i s -t i c s a r e t r a n s f e r r e d t o t h e LCPM and a r e a v a i l a b l e f o r p r o c e s s i n g . The p r o c e s s i n g x * i l l be d e s c r i b e d l a t e r i n t h i s s e c t i o n , and t h e p u r -pose o f t h e p r o c e s s i n g i s t o f i n d a code w h i c h i s s h a r e d b y two o r more Items i n t h e LCPM. The t h i r d p r o p e r t y o f t h e LCPM i s t h e s t r e n g t h -e n i n g o f a s s o c i a t i o n s between a code and i t e m s w i t h i n t h e code so t h a t a code i s more l i k e l y t o e l i c i t i t s a s s o c i a t e d i t e m s . H e r e , a f t e r a code has been found f o r two o r more i t e m s , t h e more t i m e t h o s e i t e m s 144 r e s i d e I n t h e LCPM, t h e s t r o n g e r t h e s t r e n g t h o f a s s o c i a t i o n between code and i t e m s . A s s o c i a t i o n s may be s t r e n g t h e n e d by some p r o c e s s s u c h as c o n t i g u o u s r e h e a r s a l (see Rundus, I S 7 0 ) , a l t h o u g h i t i s n o t n e c e s s a r y t o d e s c r i b e t h e e x a c t n a t u r e o f t h e s t r e n g t h e n i n g p r o c e s s a t t h i s t i m e . The f o u r t h p r o p e r t y o f t h e LCPM i n v o l v e s r e c a l l o f i t e m s r e p r e s e n t e d i n t h e FRMS. H e r e , i t e m s r e p r e s e n t e d i n t h e FRMS a r e t r a n s f e r r e d t o t h e LCPM f o r o u t p u t . The l i m i t e d c a p a c i t y o f the LCPM r e q u i r e s t h a t o n l y a c e r t a i n number o f i t e m s may be t r a n s f e r r e d a t any one t i m e , and, t h u s , a l i m i t e d number o f c a t e g o r i e s and i t e m s p e r c a t e g o r y may be d e r i v e d . T h i s f o u r t h p r o p e r t y w i l l be p u r s u e d i n t h e d i s c u s s i o n o f r e c a l l i n t h e n e x t s e c t i o n . I n agreement w i t h t h e t h e o r i e s o f K i n t s c h (1970) and A n d e r s o n and Bower ( 1 9 7 2 ) , t h e r e p r e s e n t a t i o n o f an i t e m f o r p r o c e s s i n g c o n s i s t s o f a l i s t o f t r a i t s , c h a r a c t e r i s t i c s , o r markers r a t h e r t h a n t h e i t e m i t s e l f . The markers r e p r e s e n t a v a r i e t y o f t y p e s o f c h a r a c t e r i s t i c s o f t h e i t e m s , s u c h as s u p e r o r d i n a t e s , i t e m - t o - i t e m a s s o c i a t i o n s , and a c o u s t i c and f o r m a l p r o p e r t i e s . T hat p r o p e r t i e s o t h e r t h a n common s u p e r o r d i n a t e s w i l l s e r v e as a b a s i s f o r o r g a n i z a t i o n has been demon-s t r a t e d i n a number o f e x p e r i m e n t s , some o f w h i c h were d i s c u s s e d i n C h a p t e r 1. A l t h o u g h an i t e m may have many m a r k e r s , n o t a l l o f t h e s e markers a r e e q u a l l y a v a i l a b l e . T h i s n o t i o n o f d i f f e r e n t i a l a v a i l a b i l i t y s e r v e s as t h e b a s i s f o r t h e p r o c e s s i n g model. A v a i l a b i l i t y , h e r e , may be th o u g h t o f i n a s s o c i a t i o n i s t i c terms s u c h t h a t a more a v a i l a b l e marker i s more e a s i l y e l i c i t e d by the i t e m . Which t y p e s o f markers a r e more 1 4 5 a v a i l a b l e varies as a function of the s p e c i f i c item and the item's context. In the f i r s t case, the d i f f e r e n t i a l a v a i l a b i l i t y of an item's markers may r e s u l t from the subject's pre-experimental experience with that item. For instance, the marker animal may be r a p i d l y found upon presentation of the word dog, where formal properties, such as the i n i t i a l sound Is d, w i l l be less r a p i d l y found. Conversely, the marker the i n i t i a l sound i s z may be more r a p i d l y found f o r an item such as xylem, an item with which the subject may not have had much "semantic" experience. Thus, i t could be predicted that less frequent words would more l i k e l y c l u s t e r according to formal or acoustic pro-p e r t i e s than more frequent words. In the second case, a v a i l a b i l i t y of markers varies with the item's experimental context. That i s , c e r t a i n experimental manipulations may make a marker more a v a i l a b l e . For instance, Instructions about how to organize a f f e c t the a v a i l a b i l i t y of markers. When t o l d to organize a l p h a b e t i c a l l y , a subject w i l l use the markers associated with the items' i n i t i a l l e t t e r s rather than the markers associated with the items' meanings. A second example of contextual a v a i l a b i l i t y takes the form of a p r e d i c t i o n . For instance, i f e a r l y i n the l i s t , the subject i s presented oboe and fo l k adjacently, i t i s l i k e l y that he would f i n d markers which correspond to music to use as a code.(Marker w i l l be used to define information about an item which e x i s t s i n long-term memory; a f t e r a common marker i s found, i t w i l l be r e f e r r e d to as a code. A code, here, i s what was described i n the section on i n t e r n a l structure.) I f l a t e r i n the l i s t he i s presented singer, the l i k e l i h o o d 146. that he would find the marker music rather than occupation should be increased. Processing which can account for this type of contextual availability is described later in this section, and a direct experi-mental test of the prediction is suggested in the last section. The discussion of the cognitive work which is performed In the LCPM only concerns itself with the case where the LCPM is fi l l e d . The LCPM is assumed to be fill e d except at the beginning portions of the serial order of presentation, so the description should hold for most of the l i s t presentation. Descriptions of processing for the first few items, particularly as the processing relates to the primacy effect in single-trial free recall, can be found in Rundus (1970). When an item x^hich is presented first enters the LCPM, its l i s t of markers is extracted from the subject's long-term memory. The l i s t of markers is found by some process of matching, and this matching in-volves direct access to the item. That i s , the subject need not en-gage in a complex search to find that item's markers (see Kintsch, 1970). Two operations are accorded this l i s t of markers. First, the li s t is given a time tag similar to that described by Kintsch (1970). The time tag acts like a marker in that i t is elicited by the item, and i t is available for tests of recognition. Second, the tagged l i s t of markers is processed in the LCPM. In the LCPM, this l i s t of markers is compared with other lists of markers already in the LCPM. The comparison involves a "reading off" of the markers for the nex* l i s t and the markers are read in some order, with the most available markers being read first. Each marker that is read from the new l i s t of markers 147 i s compared with the other marker l i s t s i n the LCPM. This comparison with other items may proceed i n some s e r i a l manner, with the i n i t i a l comparison being made with the most recent item to enter the b u f f e r . I f the most av a i l a b l e marker from the new l i s t f a i l s to be found i n the other l i s t s of markers, the next most a v a i l a b l e marker from the new l i s t i s extracted, and the process recycles. I f a match of markers i s found, the marker which i s shared i n common with the two items i s compared with the remaining marker l i s t s i n the LCPM to determine whether that marker i s also shared by any of these remaining items. Then, the associations between the common marker and the items are strengthened; the contextual strength of as-s o c i a t i o n betxtfeen the items and the common marker i s a function of the amount of time, a f t e r the common marker i s found, that the items reside i n the LCPM. When a marker common to two or more items i s found, that marker becomes a code for those items, and, thus, the code and the items are sa i d to be represented In the FRMS. The strength of association be-tween the code and i t s items i s a function of both t i p a v a i l a b i l i t y of the code and the amount of time the items and the code have resided i n the LCPM. In some cases an item w i l l enter the LCPM and no marker common to another item i n the LCPM w i l l be found. That i s , the more ava i l a b l e markers for that item x * i l l not be common to other Items i n the LCPM, and the item may not reside i n the LCPM long enough f or some less a v a i l a b l e markers to be processed. In th i s case, when the item leaves 148 leaves the LCPM, Its time-tagged marker l i s t is available to the sub-ject for tests of recognition. After the first common marker has been found and along with its associated items is represented in the FRMS, the processing of new items entering the LCFrf proceeds in a slightly different manner. First, as was the case when no codes have been established, the marker l i s t fir the item is extracted from the subject's long-term memory, is time-tagged, and enters the LCPM. Then, however, this l i s t of markers is not compared with the marker lists for other items in the LCPM; rather, i t is compared with the codes for items which are represented in the FRMS. If the marker l i s t for the new item is found to contain one of these codes, the items in the FRMS which share this code are brought back into the LCPM. Here, the associations between items and codes are strengthened. If the new item does not contain a marker that is common to a code for items represented in the FRMS, the comparison process proceeds as i f there were no items represented in the FRMS. That i s , the marker l i s t for the new item is compared with the marker lists for other items in the LCPM. Regardless of whether a comparison leads to a match in either the FRMS or the LCPM, when the item leaves the LCPM, its time-tagged marker l i s t is available for tests of recognition. The strength of the category code is defined as the strength of association with a higher-order code, that code being list-membership. The strength of the code is determined by the amount of time that code resides in the LCPM. Thus, the strength of a category code should 149 be some monotonically Increasing function of the strengths of the items associated with that code. The hierarchical nature of organization may result i n one of two possible ways, and i t i s not unlikely that both ways occur i n typical free r e c a l l situations. F i r s t , i n I n i t i a l stages of free r e c a l l lear-ning, some code which describes a number of items i n the l i s t may enter the FRMS. As the l i s t i s presented, many items may become associated with that code, and the total number of items so associated may exceed the capacity of the LCPM. That i s , during r e c a l l , the items within a code are assumed to channel through the LCPM, and there i s a limit to the number of items within a code which can be transferred to the LCPM (see the description of rec a l l i n the next section). Thus, the sub-ject may limit the number of items within a code by finding new codes which describe only a subset of the items. To do this, as free re c a l l learning progresses, he w i l l compare marker l i s t s of items which share this common code i n search of a more available code. This search proceeds i n the same manner as that which led to the i n i t i a l code, and the strengthening of associations with the new code w i l l also proceed i n the same manner. In this way, an item w i l l be associated with more than one code, and the codes w i l l be associated with that item i n a hierarchical manner. Thus, this f i r s t way i n which a hierarchy dev-elops can be described as development from higher levels to lower levels. It i s also possible that hierarchies may develop firm lower levels to higher levels. Within the structure of the proposed model, i t i s possible that the subject w i l l form more than the number of codes that 150 can be processed in the LCPM. That i s , codes are assumed to channel through the LCPM and there is a limit to the number of codes which can be transferred to the LCPM. In this case, on subsequent trials, the subject could process the codes as i f the codes were items. That i s , comparison of codes could lead to the matching of a marker shared in common by two or more codes, and these codes could be combined and as-sociated in the same way items within a code become associated. Thus, as an item can become associated with a code and that code can become associated with a higher-order code, a hierarchical structure can dev-elop. That a hierarchy can develop from lower levels to higher levels was demonstrated in Experiment VII. Recognition and Recall A model of output should describe why manipulations of organiza-tion may have asymmetric effects on recognition and recall (see Chapter 2 ) , but i t should also explain when manipulations (i.e., blocking) may have symmetric effects. A model of output should also describe the transfer of information from the internal structure to the recall protocol. In the present section three types of recognition memory failures are described. One of these, horaographic encoding, can be affected by manipulations of organization. The other two are failure to store and decay or interference. Recall is described as transfer from the FRMS through the LCPM. The description of recall accounts for consis-tent order of output. The description of recognition memory proceeds firs t . 151 The description of recognition memory is similar, in some as-pects, to that presented by Winograd and Raines (1972). Two basic as-sumptions are necessary for this description. First, some items are assumed to have two or more lists of markers in memory. This assump-tion may best be explained in terms of homographs, i.e., words with two or more distinct meanings. For instance, orange may have a l i s t of markers which corresponds to orange as a color, and i t may have a second l i s t of markers which corresponds to orange as a fruit. Al-though the marker l i s t for orange as a color may contain a reference to the marker l i s t for orange as a fruit, the two marker lists are distinct. It should also be assumed that for an item with two or more marker l i s t s , one of these markers lists is more available. That i s , when the item is presented, there are differential probabilities of extracting the various marker lists from memory. Operationally, the more available marker l i s t for known homographs can be defined in a free association paradigm. For instance, i f orange is a stimulus for free association, and the subject responds apple, i t is likely that orange as a fruit is more available than orange as a color. If he res-responds red, i t is likely that orange as a color is more available. The availability here is assumed to be probabilistic; that i s , i f the subject is presented the same homograph on two different trials, there is a chance that he may extract different marker lists on each of these trials. It should be pointed out that some recent studies have col-lected norms of association to homographs (e.g., Kausler & Kollasch, 1970). 152 The second assumption i s th a t r e c o g n i t i o n memory does not i n -v o l v e complex search processes. An item which i s presented i n a r e c o g n i t i o n memory task i s d i r e c t l y found i n the su b j e c t ' s long-term memory. Here, the su b j e c t checks the marker l i s t of that item f o r i t s time-tag. I f the time-tag i n d i c a t e s that the item was on the l i s t , he can c o r r e c t l y recognize i t . I f no time-tag i s found, the subj e c t r e p o r t s t h a t the item i s a new item. Thus, some of the f a i l u r e s t o recognize items which were on the l i s t can be a t t r i b u t e d to f a i l u r e s to time-tag that i t e m i n the LCPM. Two other types of r e c o g n i t i o n memory f a i l u r e s can be i d e n t i f i e d . F i r s t , the time-tags are given some decay (or i n t e r f e r e n c e ) c h a r a c t e r -i s t i c . T h i s decay c h a r a c t e r i s t i c i s e s s e n t i a l given t h a t r e c o g n i t i o n performance does s u f f e r over delays from p r e s e n t a t i o n t o t e s t (see e.g., Wickelgren & Norman, 1966). A l s o , a decay i s necessary i n order to p r o v i de f o r f a l s e alarms, i . e . , responses of " o l d " t o d i s t r a c t o r items. That l s , i f the su b j e c t l s responding to decaying time-tags at a low enough c r i t e r i o n , he may respond t o items which have not been presented. A more complete d i s c u s s i o n of time-tag decay and r e c o g n i t i o n i s presented i n K i n t s c h (1970). Second, r e c o g n i t i o n memory f a i l u r e s occur as a f u n c t i o n of the p r o b a b i l i s t i c encoding of homographic words. For example, on the study t r i a l , the su b j e c t may see orange and tag the l i s t o f markers which cor-responds to orange as a f r u i t . On the t e s t t r i a l , w i t h some p r o b a b i l i t y he may see orange and e x t r a c t the l i s t of markers which corresponds t o orange as a c o l o r . T h i s l i s t of markers w i l l not c o n t a i n a time-tag 153 indicating that orange was on the l i s t , and thus, the subject will f a i l to correctly recognize orange. That such failures to recognize as a function of changed meaning occurs has been demonstrated in a number of studies which have manipulated the meanings of homographs on study and/or test trials (e.g., Light & Carter-Sobell, 1970). The presentation of items in blocked fashion should bias which marker l i s t is chosen. Also, the use of category distractors should bias which marker l i s t is chosen for recognition. For blocked presen-tation, both presentation and recognition bias should be the same. For spaced presentation, there should be no bias at presentation, al-though there will be a bias at recognition. Thus, blocked presentation should produce better recognition performance. Manipulations of organization other than blocking have not been found to affect recognition performance, probably because they have not differentially biased meaning between groups with different levels of organization. For example, in Experiment I, the items were "unrelated" and the distractor items were randomly chosen. In Experiment II, al-though the distractor items belonged to the same categories as the presented item, there were no systematic relationships between the items in the presentation l i s t . Thus, both rehearsal groups should experience the same difficulty in terms of the distractor items. Choosing distractors from the same categories as the presentation items should lead to increased difficulty in recognition (cf. Bower, Clark, Lesgold, & Winzenz, 1969). While the recognition process involves direct matching with time-154 tagged marker li s t s , the recall process involves retrieval from the FRMS which has been hypothesized to be a hierarchical structure of the presentation items. The hierarchy is assumed to be a set of associa-tions such that list-membership is associated with category codes and category codes are associated with either lower order codes or with the items. Recall from this structure is assumed to follow a path from the top to the bottom, and probability of recall of both codes and items within codes is assumed to be a function of the associative strength. Recall of the items for this structure is assumed to channel through the LCPM. First, especially early in learning, the LCPM is emptied of recently presented items (see e.g., Kintsch, 1970). Then, the code list-membership is transferred to the LCPM along with the highest-order codes. The number of codes which can be transferred at any one time is limited by the capacity of the LCPM. Then one code, probably the highest associate of list-membership, is chosen. Remember that the strength of association of list-membership to a code is a function of the amount of LCPM time that the code has had along with its associated items. When this code is chosen, the items within that code (or lower order sub-codes) are transferred to the LCPM. These items are recalled in some order, possibly dependent upon the strengths of association of the code to the different items. Here, strength of association is a function of both amount of LCPM time and the availa-bili t y of the code. Availability of a code has been described earlier. After the LCPM has been emptied of items within the strongest 155 code, the next str o n g e s t code i s used to t r a n s f e r items from the h i e r a r c h y to the LCPM and the process i s assumed to r e c y c l e throughout the h i e r a r c h y . I n t h i s way, r e c a l l may approach a constant order. As the r e c a l l o f the code i s assumed to precede r e c a l l of items w i t h i n t h a t code, f a i l u r e t o r e c a l l a code w i l l preclude r e c a l l of i t s as-2 s o c i a t e d items. Experiments and P r e d i c t i o n s In t h i s s e c t i o n , the model i s discussed i n two ways. F i r s t , an attempt i s made to show how the model can handle the f r e e r e c a l l data discussed i n the f i r s t chapter. Second, some p o s s i b l e experimental t e s t s of the model's assumptions and of p r e d i c t i o n s that can be made from the model are suggested. E v a l u a t i o n The model makes no d i s t i n c t i o n between the c l u s t e r i n g of l i s t s w i t h an a p r i o r i s t r u c t u r e and the s u b j e c t i v e o r g a n i z a t i o n of "unre-l a t e d " words i n terms of u n d e r l y i n g processes. The major theme of the experiments reported i n t h i s d i s s e r t a t i o n has been t h a t , regard-l e s s of the types of items used, s i m i l a r manipulations of o r g a n i z a t i o n produce s i m i l a r r e s u l t s i n performance. The model can handle the data which i n d i c a t e that l i s t s w i t h h i g h a s s o c i a t i v e s t r u c t u r e are r e c a l l e d , but not recognized, b e t t e r I t i s p o s s i b l e , as Mandler (1967) suggests, t h a t an item could have more than one r e p r e s e n t a t i o n on the h i e r a r c h y . This p o s s i b i l i t y c o u l d be handled by the model; however, i t w i l l not be pursued. 156 than lists with low associative structure (e.g., Kintsch, 1968). The discussion of the recognition process indicated that, unless the manipulations of organization produced differential homographic enco-ding for different groups of subjects, there should be no change in recognition performance. If both high and low associatives lists are presented in a random order, i t is difficult to see how encoding could be affected; again, i f both are presented in blocked order, encoding bias should be approximately the same for both groups. The notion of marker availability, however, can explain why highly structured lists are better recalled. Highly structured lists are assumed to contain more available common markers than lists with l i t t l e structure. The advantage to the subject, then, can occur at two stages of the recall process. First, as more available common markers take less LCPM time to find, the subject will have more LCPM time to strengthen the associations between the items and the codes. Second, as the strength of association between a code and its asso-ciated items has some pre-experimental value, more available markers should be better at eliciting their associated items. As the model does not distinguish between the clustering of words with an a priori structure and the subjective organization of "unrelated" words, the remainder of this section will treat the two types of organization as the same. The description that follows is ordered in the same way as the review of the data in Chapter 1. Order of Input. If items with more available common markers reside in the LCPM at the same time, there will be less time necessary to 157 f i n d common markers and more t i m e a v a i l a b l e f o r s t r e n g t h e n i n g o f a s -s o c i a t i o n s . As b l o c k e d p r e s e n t a t i o n s h o u l d i n c r e a s e t h e p r o b a b i l i t y t h a t i t e m s r e l a t e d b y more a v a i l a b l e common markers r e s i d e i n t h e LCPM a t t h e same t i m e , i t s h o u l d be p r e d i c t e d t h a t b l o c k i n g w o u l d l e a d t o g r e a t e r a s s o c i a t i v e s t r e n g t h between codes and i t e m s w i t h i n c o d e s . Thus, r e c a l l p e r f o r m a n c e s h o u l d be i m p r o v e d . H i g h e r o r g a n i z a t i o n s c o r e s can a l s o be p r e d i c t e d . I n random p r e s e n t a t i o n , s u b j e c t s may e x t r a c t codes w h i c h a r e l e s s a v a i l a b l e t h a n t h o s e b u i l t i n t o t h e l i s t s t r u c t u r e . F u r t h e r p r o c e s s i n g o f i t e m s w i t h i n t h e s e c o d e s , t h e n , may be n e c e s s a r y t o l i m i t t he s i z e o f c a t e g o r i e s ( s e e t h e d i s c u s s i o n o f how o r g a n i z a t i o n d e v e l o p s f r o m t h e b o t t o m o f t h e h i e r a r c h y t o t h e t o p ) . W i t h t h i s f u r t h e r p r o c e s s i n g , t h e s u b j e c t ' s o r g a n i z a t i o n w i l l change. I n b l o c k e d p r e s e n t a t i o n c o n d i t i o n s , no s u c h f u r t h e r p r o c e s s i n g may be n e c e s s a r y . I f i t i s , i t i s l i k e l y t h a t t h e p r o c e s s i n g w o u l d p r o c e e d f r o m t h e b o t t o m o f the h i e r a r c h y t o the t o p , and l i t t l e change i n ad-j a c e n c i e s a t r e c a l l w o u l d o c c u r . P r e s e n t a t i o n R a t e . The arguments c o n c e r n i n g p r e s e n t a t i o n r a t e f o l l o w t h o s e c o n c e r n i n g b l o c k e d p r e s e n t a t i o n . T hat i s , s l o w i n g p r e s e n t a t i o n r a t e s h o u l d i n c r e a s e t h e LCPM t i m e a v a i l a b l e t o t h e s u b j e c t and t h u s , more t i m e x<rould be a l l o w e d f o r s t r e n g t h e n i n g o f a s s o c i a t i o n s . I n t h i s way, i n c r e a s e d r e c a l l and i n c r e a s e d o r g a n i z a t i o n s c o r e s can be p r e d i c t e d . C u e i n g . The model a l l o w s some i n d e p e n d e n t " f o r g e t t i n g " o f t h e code and i t s a s s o c i a t e d i t e m s . When m a n i p u l a t i o n s i n s u r e t h a t t h e s t r e n g t h s o f a s s o c i a t i o n between codes and t h e i r a s s o c i a t e d i t e m s a r e about t h e same o v e r a l l c a t e g o r i e s , s u c h i n d e p e n d e n t " f o r g e t t i n g " w o u l d be 158 e x p e c t e d . P r e s e n t a t i o n o f c o d e n a m e s a l o n g w i t h i t e m s o r b l o c k e d p r e s e n t a t i o n o f i t e m s w i t h i n a c o d e s h o u l d , t o s o m e e x t e n t , e q u a l i z e s t r e n g t h s o f a s s o c i a t i o n b e t w e e n c o d e s a n d i t e m s w i t h i n c o d e s . T h u s , i f a c o d e i s " f o r g o t t e n " a n d s u c h p r e s e n t a t i o n c o n d i t i o n s a r e u s e d , w h e n t h e s u b j e c t l a t e r " r e m e m b e r s " t h a t c o d e , t h e n u m b e r o f i t e m s t h a t h e r e c a l l s s h o u l d n o t b e l e s s t h a n t h e n u m b e r t h a t h e w o u l d h a v e r e -c a l l e d h a d h e n o t f o r g o t t e n t h e c o d e . C o d e s m a y b e " f o r g o t t e n " i f t h e n u m b e r o f c o d e s e x c e e d s t h e n u m b e r w h i c h c a n b e t r a n s f e r r e d f r o m t h e F R M S t o t h e L C P M . T h e p u r -p o s e o f a c u e , t h e n , i s t o b y p a s s t h e n e c e s s i t y t o t r a n s f e r m o r e t h a n o n e c o d e f r o m t h e F R M S t o t h e L C P M a t a n y o n e t i m e . A c u e i s e n v i -s i o n e d a s p r o v i d i n g d i r e c t a c c e s s t o t h e F R M S , m u c h i n t h e s a m e w a y t h a t a f t e r t h e F R M S d e v e l o p s , t h e s u b j e c t c o m p a r e s m a r k e r s l i s t s f o r n e w i t e m s w i t h t h e c a t e g o r y c o d e s i n t h e F R M S . A s t h e c u e i s d i r e c t l y c o m p a r e d w i t h t h e c a t e g o r y c o d e s , i t i s n e c e s s a r y t h a t , i n o r d e r t o p r o v i d e a " m a t c h , " t h e c u e b e i d e n t i c a l t o t h e c a t e g o r y c o d e . T h u s , f o r c u e i n g t o b e e f f e c t i v e , m a n i p u l a t i o n s w h i c h i n s u r e t h a t t h e s u b j e c t u s e s c o d e s w h i c h w i l l e v e n t u a l l y b e u s e d a s c u e s a r e n e c e s s a r y . B l o c k i n g , p r e s e n t i n g t h e c u e a t p r e s e n t a t i o n , a n d a s k i n g t h e s u b j e c t t o n a m e h i s c a t e g o r i e s a r e s u c h m a n i p u l a t i o n s . T h a t t h e s e m a n i p u l a t i o n s a r e e f f e c t i v e i n p r o d u c i n g a d d i t i o n a l r e c a l l w i t h c u e i n g w a s d i s c u s s e d i n t h e f i r s t c h a p t e r . N u m b e r o f C a t e g o r i e s . A s t h e m o d e l p o s t u l a t e s t h a t t h e r e i s a l i m i t t o t h e n u m b e r o f c o d e s t h a t c a n b e t r a n s f e r r e d f r o m t h e F R M S a t a n y o n e t i m e , n u m b e r o f c a t e g o r i e s s h o u l d b e a n e f f e c t i v e v a r i a b l e . 159 However, the optimal number of categories should also be a function of l i s t length, as the number of items within a category is also assumed to have some limitation. Thus, i t is not surprising that an effect of number of categories has been found but that a function describing this effect is not clear. At present, the author will not attempt to formulate such a function. Inhibition. The model which has been presented has not been constructed to handle learning of two or more list s . However, i t is not unlikely that i t could be modified for such situations. Elsewhere, i t has been suggested that l i s t differentiation may account for some effects of pro-active inhibition (Schwartz and Humphreys, 1972b), and this suggestion will be pursued in refinements of the model. Individual Differences. If, as Earhard (1967) hypothesizes, individual differences in free recall represent individual differences in the ability to form associations, the present model can handle such differ-ences. That i s , recall is viewed as sets of associative bonds between items, category codes, and higher-order codes, such that individual differences in ability to form these bonds will result in differences in both recall and organization scores. However, this explanation really avoids the more interesting questions why are there individual differences in ability to form associations? Short-term Memory. As the present model employs a buffer-like construct (LCPM), such findings as the large recency effect are easily explained. The model would handle short-term memory phenomena in the same way as Kintsch's (1970) model. 160 Future Research The model has many aspects which are speculative. Ilany of these might be open to experimental i n v e s t i g a t i o n . Also, the a p p l i c a t i o n of the model to the data may suggest new l i n e s of i n v e s t i g a t i o n . In t h i s s e c t i o n , some experiments suggested by the model w i l l be discussed. There i s no contention that a l l , or even a good percentage of, possible experiments may be represented. Rather, a number of studies which i n t e r e s t the author w i l l be b r i e f l y described. Structure of Organization. Tne data indicated that a h i e r a r c h i c a l category arrangement i s a good d e s c r i p t i o n of the structure of organi-zation. However, there are no data which prove t h i s contention. So f a r as the author knows, there i s no way to prove or disprove any representation of structure for free r e c a l l memory. However, an i n t e r e s t i n g question can s t i l l be asked: i s a h i e r a r c h i c a l arrange-ment the most e f f i c i e n t structure f o r r e c a l l ? A possible approach to answering t h i s question would be to have d i f f e r e n t groups of subjects learn concept i d e n t i f i c a t i o n tasks with solutions being based on c l u s t e r i n g , h i e r a r c h i c a l categorization, or multidimensional designs. The items would be words and they would be the same for a l l groups, and the groups would be tested on r e c a l l of the words. I f a s u i t a b l e design can be found, the author would l i k e to pursue t h i s l i n e of research, The notion of constant order r e c a l l of categories and items with-i n categories as a function of strength of association can be tested. As strength of association i s defined i n two ways, two manipulations 161 suggest themselves. F i r s t , pre-experimental strength of association of items to a code can be manipulated, within a category, by using low and high category associates. It would be predicted that when the category i s recalled, high associates would tend to be recalled before low associates. Second, contextual strength of association may be varied by the amount of LCPM time afforded an item and i t s category name. Here, an item which belongs to a category might be repeated twice i n the same l i s t presentation in order to give i t more LCPM time. It would be expected that, when the category is recalled, the item with multiple presentations w i l l tend to be recalled f i r s t , provided that i t i s represented at only one location in the structure. Processing. Three lines of research are suggested by the discussion of processing. F i r s t , the notion that processing proceeds i n the LCPM in a se r i a l manner makes a prediction about presentation order. Here, the amount recalled should decrease when related items reside in the LCPM at the same time, but have not been presented adjacently. In order to test this prediction, a l i s t of two-word categories could be used. Items within the same category would be presented either adja-cently, with one intervening item, or with two intervening items. It would be predicted that the number of categories recalled would vary inversely with the number of intervening items. If the prediction i s not substantiated, the model may have to be modified to include parallel processing in the buffer. The second line of research involves xjhether a l l marker comparison takes place in the LCPM or whether there i s f i r s t a comparison with 162 categories represented in the FRMS. Here, i f a category i s represented in the FRMS, the present model predicts that exemplars need not be in the LCPM for a new item to be associated with that category. To test this assumption, imagine two l i s t s of 30 items. In the f i r s t l i s t con-dition, these 30 items are three members from each of ten categories. The f i r s t two items are from the f i r s t category as i s the twenty-first item; the third and fourth items are from the second category as is the twenty-second item, etc. The second l i s t condition i s the same as the f i r s t - l i s t condition except that the f i r s t ten items in odd-numbered presentation positions have been replaced by ten new, unrelated items. Recall of the last ten items, which are the same for both conditions, is the dependent variable. The present model would predict better re-c a l l of the last ten items for the f i r s t - l i s t condition, but a model which contends that a l l comparison i s done in the LCPM (e.g., Kintsch, 1970) would predict no differences between the conditions i n re c a l l of these items. The third line of research involves the differential availability of markers. No particular experiments are suggested, but an interesting research programme might centre around identifying various types of markers, seeing which types are more available as a function of the type of stimuli, and testing the effects of various manipulations in changing the availability of markers. Recognition and Recall. The description of recognition differences as a function of homographic encoding can be tested. The typical paradigm of spaced versus adjacent presentation of two-item categories can be used. 163 Two a d d i t i o n a l groups w o u l d be p r e s e n t e d the same s p a c e d and a d j a c e n t l i s t s ; however, c a t e g o r y names w o u l d be p r o v i d e d a l o n g s i d e t h e i t e m s f o r t h e two g r o u p s . T h i s m a n i p u l a t i o n s h o u l d reduce t h e p o s s i b i l i t y o f d i f f e r e n t i a l bia3 between the two gr o u p s . Thus, an i n t e r a c t i o n between s p a c i n g and w h e t h e r t h e c a t e g o r y names a r e p r e s e n t e d w o u l d be e x p e c t e d . 164 REFERENCES A n d e r s o n , J.R., & Bower, G.H. R e c o g n i t i o n and r e t r i e v a l p r o c e s s e s i n f r e e r e c a l l . P s y c h o l o g i c a l Review, 1972, 79, 97-123. A t k i n s o n , R . C , & S h i f f r i n , R.M. M a t h e m a t i c a l models f o r memory and l e a r n i n g . T e c h n i c a l R e p o r t No. 79, I n s t i t u t e f o r M a t h e m a t i c a l S t u d i e s i n t h e S o c i a l S c i e n c e s , S t a n f o r d U n i v e r s i t y , 1965. A t k i n s o n , R . C , & S h i f f r i n , R.M. Human memory: A p r o p o s e d s y s t e m and i t s c o n t r o l p r o c e s s e s . I n K.W. Spence and J.T. Spence ( e d s . ) , The p s y c h o l o g y o f l e a r n i n g and m o t i v a t i o n , V o l . 11. New Y o r k : Academic P r e s s , 1968. B a t t i g , W.F., A l l e n , M., & J e n s e n , A.R. P r i o r i t y o f f r e e r e c a l l o f n e w l y l e a r n e d i t e m s . J o u r n a l o f V e r b a l L e a r n i n g and V e r b a l  B e h a v i o r , 1965, 4, 175-179. B a t t i g , W.F., & Montague, W.E. C a t e g o r y norms f o r v e r b a l i t e m s i n 56 c a t e g o r i e s : A r e p l i c a t i o n and e x t e n s i o n o f t h e C o n n e c t i c u t c a t e g o r y norms. J o u r n a l o f E x p e r i m e n t a l P s y c h o l o g y , 1969, 80, ( 3 , P t . 2 ) . Birnbaum, I.M. F r e e - r e c a l l l e a r n i n g as a f u n c t i o n o f p r i o r - l i s t o r g a n i z a t i o n . J o u r n a l o f V e r b a l L e a r n i n g and V e r b a l B e h a v i o r , 1968, 2 . 1037-1042. B o u s f i e l d , A.K., & B o u s f i e l d , W.A. Measurement o f c l u s t e r i n g and s e q u e n t i a l c o n s t a n c i e s i n r e p e a t e d f r e e r e c a l l . P s y c h o l o g i c a l  R e p o r t s , 1966, 19, 935-942. B o u s f i e l d , W.A. The o c c u r r e n c e o f c l u s t e r i n g i n the r e c a l l o f r a n -domly a r r a n g e d a s s o c i a t e s . J o u r n a l o f G e n e r a l P s y c h o l o g y , 1953, 49, 229-240. B o u s f i e l d , W.A., P u f f , C R . , & Cowan, T.M. The development o f c o n s t a n c i e s i n s e q u e n t i a l o r g a n i z a t i o n d u r i n g r e p e a t e d f r e e r e c a l l . J o u r n a l o f V e r b a l L e a r n i n g and V e r b a l B e h a v i o r , 1964, 3, 449-459. B o u s f i e l d , W.A., & Sedgewick, C.H.W. An a n a l y s i s o f sequences o f r e s t r i c t e d a s s o c i a t i v e r e s p o n s e s . J o u r n a l o f G e n e r a l P s y c h o l o g y , 1944, 30, 149-165. Bower, G.H., C l a r k , M.C, L e s g o l d , A.M., & W i n z e n z , D. H i e r a r c h i c a l r e t r i e v a l schemes i n the r e c a l l o f c a t e g o r i z e d l i s t s . J o u r n a l  o f V e r b a l L e a r n i n g and V e r b a l B e h a v i o r , 1969, 8, 323-343. Bower, G., & L e s g o l d , A.M. O r g a n i z a t i o n as a d e t e r m i n a n t o f p a r t - t o -whole t r a n s f e r i n f r e e r e c a l l . J o u r n a l o f V e r b a l L e a r n i n g and  V e r b a l B e h a v i o r , 1969, 8, 501-506. 165 Bower, G.H., Lesgold, A.M., &/Tieman, D. Grouping operations in free recall. Journal of Verbal Learning and Verbal Behavior, 1969, j$, 481-493. Brown, J. The nature of set to learn and of intra-material interferences in immediate memory. Quarterly Journal of Experimental Psychology, 1954, $, 141-148. Bruce, D., & Fagan, R.L. More on recognition and free recall of organized li s t s . Journal of Experimental Psychology, 1970, 85, 153-154. Bugelski, B.R. A remote association explanation of the relative difficul-ty of learning nonsense syllables in a serial l i s t . Journal of Ex- perimental Psychology, 1950, 40, 336-348. Buschke, H., & Hinrichs, J.V. Relative vulnerability of item information in short-term storage for the missing scan. Journal of Verbal  Learning and Verbal Behavior, 1968, 2» 1043-1048. Cofer, C.N., Bruce, D.R., & Reicher, G.M. Clustering in free recall as a function of certain methodological variations. Journal of Exper- imental Psychology, 1966, 71, 858-866. D'Agostino, P.R. The bioeked-random effect in recall and recognition. Journal of Verbal Learning and Verbal Behavior, 1969, 8, 815-820. Dong, T., & Kintsch, W. Subjective retrieval cues in free recall. Journal of Verbal Learning and Verbal Behavior, 1969, 813-816. Earhard, M. Subjective organization and l i s t organization as determinant of free-recall and serial-recall memorization. Journal of Verbal  Learning and Verbal Behavior, 1967, 6, 501-507. Earhard, M., & Endicott, 0. Why are there individual differences in subjective sequential organization during free-recall memorization? Journal of Verbal Learning and Verbal Behavior, 1969, 8., 316-319. Fischler, I., Rundus, D., & Atkinson, R.C. Effects of overt rehearsal procedures on free recall. Psychonomic Science, 1970, 19_, 249-250. Frankel, F., & Cole, M. Measures of category clustering in free recall. Psychological Bulletin, 1971, ]J±, 39-44. Glanzer, M., & Cunitz, A.R. Two storage mechanisms in free recall. Journal of Verbal Learning and Verbal Behavior, 1966, 5_, 351-360. Gorfein, D.S. Are good subjective organizers good paired-associate learners? Psychonomic Science, 1971, 22, 340. Gorfein, D.S., Blair, C., Rowland, C. The generality of free recall: 1. Subjective organization as an ability factor. Psychonomic  Science, 1968, 11, 279-280. 166 Guttman, L. A general nonmetric technique for finding the smallest co-ordinate space for a configuration of points. Psychometrika, 1968, 33, 469-506. Henley, N.M. A psychological study of the semantics of animal terms. Journal of Verbal Learning and Verbal Behavior, 1969, 8i, 176-184. Horowitz, L.M. & Manelis, L. Towards a theory of redintegrative memory. In G.H. Bower and J.T. Spence (eds.), The psychology of learning  and motivation, Vol. VI, New York: Academic Press, in press. Jacoby, L.J. Effects of organization on recognition memory. Journal of Experimental Psychology, 1972, 92, 325-331. Jenkinsj J.J., & Russell, W.A. Associative clustering during recall. Journal of Abnormal and Social Psychology, 1952, 47, 818-821. Kausler, D.H., & Kollasch, S.F. Word associations to homographs. Journal of Verbal Learning and Verbal Behavior, 1970, 9_, 444-449. Kintsch, W. Recognition and free recall of organized l i s t s . Journal  of Experimental Psychology, 1968, 7J3, 481-487. Kintsch, W. Models for free recall and recognition. In D.A. Norman (ed.), Models of human memory. New York: Academic Press, 1970. Kintsch, W., & Buschke, H. Homophones and synonyms in short-term memory. Journal of Experimental Psychology, 1969, 80, 403-407. Kintsch, W., Miller, R.F., & Hogan, R.M. Semantic structure and clustering in free recall. Paper presented at CO.B.R.E. research workshop on cognitive organization and psychological processes. Huntington Beach, August, 1970. Kjeldergaard, P.M. Transfer and mediation in verbal learning. In T.R. Dixon and D.L. Horton (eds.), Verbal behavior and general  behavior theory. Englewood Cliffs: Prentice-Hall, 1968. Light, L.L., & Carter-Sobell, L. Effects of changed semantic context on recognition memory. Journal of Verbal Learning and Verbal  Behavior. 1970, 9, 1-11. Lingoes, J.C, & Roskam, E. A mathematical and empirical study of two multidimensional scaling algorithms. Technical Report No. 71-1, Michigan Mathematical Psychology Program, University of Michigan, 1971. Mandler, G. Organization and memory. In K.W. Spence and J.T. Spence, (eds.), The psychology of learning and motivation, Vol., 1, New York: Academic Press, 1967; 167 Mandler, G. Organized recall: Individual functions. Psychonomic  Science, 1968, 13, 235-236. Mandler, G., & Pearlstone, Z. Free and constrained concept learning and subsequent recall. Journal of Verbal Learning and Verbal  Behavior, 1966, 5_, 126-131. McCormack, P.D. Recognition memory: How complex a retrieval system? Canadian Journal of Psychology, 1972, 26, 19-41. Miller, G.A. The magical number seven, plus or minus two: Some limits on our capacity for processing information. Psychol- ogical Review, 1956, 63, 81-97. Mueller, J.H. Response properties of the position indicant in serial learning. Journal of Experimental Psychology, 1970, 84, 35-39. Nelson, A.K., McRae, B.C., & Sturges, P.T. Recall as a function of instructions and level of organization. Paper presented at the meeting of the Western Psychological Association, Los Angeles, April, 1970. Novinski, L. Part-whole and whole-part free recall learning. Journal of Verbal Learning and Verbal Behavior, 1969, JJ, 152-154. Ornstein, P.A. Role of prior-list organization on a free-recall transfer task. Journal of Experimental Psychology, 1970, 86, 32-37. Paivio, A. Mental imagery in associative learning and memory. Psychological Review, 1969, 76_, 241-263. Paivio, A., Yuille, J.C, & Madigan, S.A. Cone re teness, imagery, and meaningfulness values for 925 nouns. Journal of Experimental  Psychology, 1968, 76, (1, pt. 2). Peterson, L.R., & Peterson, M.J. Short-term retention of individual verbal items. Journal of Experimental Psychology, 1959, 58, 193-198. Puff, CR. An investigation of two forms of organization in free recall. Journal of Verbal Learning and Verbal Behavior, 1970, 9, 720-724. (a) Puff, CR. Role of clustering in free recall. Journal of Exper- imental Psychology, 1970, 86, 384-386. Rundus, D. An analysis of rehearsal processes in free recall. Technical Report No. 159, Institute for Mathematical Studies in the Social Sciences, Stanford University, 1970. 168 Rundus, D., & Atkinson, R.C. Rehersal processes in free recall: A Procedure for direct observation. Journal of Verbal Learning  and Verbal Behavior, 1970, 9_, 99-105. Schwartz, R.M. Dual process of serial learning. Psychonomic Science, 1970, 214-215. Schwartz, R.M., & Humphreys, M.S. Examminations of the category-recall function. American Journal of Psychology, 1972, in press, (a) Schwartz, R.M., & Humphreys, M.S. List differentiation in part-whole free recall. American Journal of Psychology, 1972, in press, (b) Shapiro, S.I., & Bell, J.A. Subjective organization and free recall: Performance of high, moderate, and low organizers. Psychonomic  Science, 1970, 21, 71-73. Shapiro, S.I., & Palermo, D.S. Conceptual organization and class mem-bership: Normative data for representatives of 100 categories. Psychonomic Monograph Supplements, 1969, 11, (Whole No. 43). Shapiro, S.I., & Ponce, I . Free recall and organization as functions of paced and unpaced responding and presentation rate. Psychonomic  Science, 1970, 18, 329-330. Shuell, T.J. Retroactive inhibition in free-recall learning of cate-gorized l i s t s . Journal of Verbal Learning and Verbal Behavior, 1968, 2, 797-805. Shuell, T.J. Clustering and organization in free recall. Psychologi- cal Bulletin, 1969, 72, 353-374. Slamecka, N.J., Moore, T., & Carey, S. Part-to-whole transfer and its relation to organization theory. Journal of Verbal Learning and  Verbal Behavior, 1972, 11, 7i-82. Thorndike, E.L., & Lorge, I. The teacher's word book of 30,000 words. New York: Teachers College, Columbia University, Bureau of Publications, 1944. Tryon, R.C, & Bailey, D.E. Cluster analysis. New York: McGraw H i l l , 1970. Tulving, E. The effect of alphabetical subjective organization on memorizing 'unrelated' words. Canadian Journal of Psychology, 1962, 16, 185-191. (a) Tulving, E. Subjective organization in free recall of 'unrelated' words. Psychological Review, 1962, 69_, 344-354. (b) 169 Tulving, E. Intratrial and i n t e r t r i a l retention: Notes towards a theory of free recall verbal learning. Psychological Review, 1964, 71, 219-237. Tulving, E. The effect of order of presentation on learning of 'unrela-ted' words. Psychonomic Science, 1965, _3, 337-338. Tulving, E. Subjective organization and effects of repetition i n multi-t r i a l free-recall learning. Journal of Verbal Learning and Verbal  Behavior, 1966, 5_, 193-197. Tulving, E. Free re c a l l . In D.L. Horton and T.R. Dixon (eds.), Verbal  behavior and general behavior theory. Englewood C l i f f s ; Prentice-Hall, 1968. Tulving, E., & Osier, S. Transfer effects in whole/part free-recall learning. Canadian Journal of Psychology, 1967, Z l , 253-262. Tulving, E., & Pearlstone, Z. Availability and accessibility of inform-ation in memory for words. Journal of Verbal Learning and Verbal  Behavior, 1966, 5_, 381-391. Tulving, E., & Psotka, J. Retroactive inhibition in free r e c a l l : In-accessibility of information available in the memory store. Journal of Experimental Psychology, 1971, 87, 1-8. Veldman, D.J. Fortran programming for the behavioral sciences. New York: Holt, Rinehart and Winston, 1967. Wallis, W.A., & Roberts, H.V. Statistics: A new approach. New York: Free Press, 1957. Waugh, N.C., & Norman, D.A. Primary memory. Psychological Review, 1965, 72, 89-104. Weingartner, H. The free recall of sets of associatively related words. Journal of Verbal Learning and Verbal Behavior, 1964, _3, 6-10. Wexler, K.N. Semantic structure: Psychological evidence for hierarchi-cal features. Paper presented at C.O.B.R.E. research workshop on cognitive organization and psychological processes. Huntington Beach, August, 1970. Wickelgren, W.A., & Norman, D.A. Strength models and ser i a l position in short-term recognition memory. Journal of Mathematical Psychology, 1966, 3_, 316-347. Wickens, D.D., & Engle, R.W. Imagery and abstractness in short-term memory. Journal of Experimental Psychology, 1970, 84, 268-272. 170 Winograd, E., & Raines, S.R. Semantic and temporal variation in recognition memory. Journal of Verbal Learning and Verbal Behavior, 1972, 11, 114-119. Wood, G. Retrieval cues and the accessibility of higlier-»order memory units in multitrial free re c a l l . Journal of Verbal Learning and  Verbal Behavior, 1969, 8, 782-789. Wood, G. Organization, retroactive inhibition, and free-recall learning. Journal of Verbal Learning and Verbal Behavior, 19 70, 9_, 327-333. Wood, G., & Clark, D. Instructions, ordering and previous practice in free-recall learning. Psychonomic Science, 1969, 14, 187-188. 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0101411/manifest

Comment

Related Items