ABSTRACT Supervisor: Dr. B e v e r l y A. Timmons
The relationship between m e mory organization skills and performance was investigated. The purpose was to
investigate o r ganization and training differences that might result in different performance under the recall TES effect. The recall test e x p e c t a n c y effect (TES) states that people who expect and receive a recall test perform significantly better than do people who expect a multiple choice test but receive a recall test (Balota & Neely, 1980). In Phase 1, half of the 96 female u n d e rgraduate participants were
randomly assigned to a c a tegorized word list while the other half were assigned to an u n r e l a t e d word list (Tulving, 1962) condition. The subjects studied and recalled the same list of words over four study-test trials. The participants were divided into high and low organizer groups depending on
their organization scores (Tulving & Sternberg, 1977) from Phase 1. Before Phase 2 the high and low organizer groups were randomly assigned to either a training or no training condition. The subjects in the no training condition went directly to Phase 2. The training period consisted of
suggestions for o rganizing materials for better recall. In Phase 2 all of the subjects received three different word lists for each of three study-test trials. The first two tests were multiple choice (recognition) and the last was a surprise recall test. Thus, the subjects were led to expect a recognition test through practice. The results from all
The results from all of the condition groups were in support of the traditional recall TES effect. This
reflected qualitatively different processes used for
recognition versus recall tests (Underwood, 1972). There was a significant relationship between organizational abilities from Phase 1 and performance on the surprise recall test in Phase 2 for the high organizers but not for the low organizers. Significant group differences were found when training was not provided; however, the training period eliminated organizational group differences ir. Phase 2. The results were interpreted within the theoretical frameworks of Anderson and Bower (1972, 1974) and Gillund and Shiffrin (1984). Future research should be aimed at
identifying the practical organization strategies used by test-takers so that recall performance of students,
especially low organizers, can be facilitated. Examinees:
D r . Beverly
X)
TimmonsDr. C. Brian Harve^
f
Dr. M^ITbnore France
Dr. ^R^ljayd B. May
TABLE OF CONTENTS PRELIMINARY PAGES
Title Page
Abstract ii
Table of Contents iv
List of Tables vii
List of Figures viii
Acknowledgements ix
Dedication x
CHAPTER 1 INTRODUCTION 1
CHAPTER 2 LITERATURE REVIEW
Test Expectancy Research 4
List Studies 4
Recall Tests 4
Recognition Tests 5
Other Factors Investigated In List Studies 6
Memory Organization Research 8
Clustering 10
Other Variables That Affect Clustering 16
Subjective Organization 25
Organizational Skills 32
The Measurement of Secondary Organization 36 Theoretical Explanations Of Memory And Organization 37
CHAPTER 3 STATEMENT OF THE PROBLEM AND EXPERIMENTAL PROCEDURES Rationale 48 Hypotheses 51 Subjects 52 Study Materials 52
Categorized Word List 52
Unrelated Word List 53
Test Expectancy Word Lists 53
Test Materials 54
Phase 1 54
Phase 2 55
Data Collection And Procedures 55
Analysis Of The Data 58
CHAPTER 4 RESULTS AND DISCUSSION
Results 60
Phase 1 60
Organizational Scores (PF) Across Trials 60
Hypotheses 1 And 2 64
Phase 2 66
The Test Expectancy Effect 66
Hypotheses 3 And 4 (Between Group
Differences) 70
The Relationship Between Organizational
Hypotheses 5 And 6 74
Summary of Hypotheses and Results 77
Discussion 79
Phase 1 79
Hypotheses 1 And 2 79
Phase 2 8C
The Test Expectancy Effect 80
Hypotheses 3 And 4 81
The Relationship Between Organizational Ability
And Recall Performance 84
Hypotheses 5 And 6 84
CHAPTER 5 SUMMARY AND CONCLUSIONS
Major Conclusions 87
Limitations 90
Future Research 93
LITERATURE CITED 96
APPENDICES
Appendix A: Operational Definitions 112 Appendix B: Study And Training Materials 116
Appendix C: Procedural Protocol 121
Appendix D: Sum of the Ranks Method 123
List of Tables
Table 1
Table 2
Table 3
PF Means And Standard Deviations (SD)
For Trials 1 To 4 in Phase 1 61 Mean Correct (Out Of Eighteen) And
Standard Deviations (SD^ In Phase 2
For Phase 1 Lists. PF Groups. Trials 67 Mean Correct (Out of Eighteen! And
Standard Deviations (SD) In Phase 2 For Phase 1 Lists. Training. PF Groups.
List of Figures
Figure 1 Figure 2
Figure 3
Research Design. 50
PF Groups X Phase 1 Lists X Trials
Interaction for Phase 1. 62
Phase 1 Lists X PF Groups X Trials
The author wishes to acknowledge the assistance of her committee members, Drs. Timmons, Harvey, France, May, and Ollila. Special appreciation is given to her supervisor, Dr. Beverly Timmons, for her support and guidance.
DEDICATION
Chapter One: Introduction
The recall test expectancy effect (TES) has proven to be a robust and replicable effect in word list studies
(Balota & Neely, 1980; Connor, 1977). When the recall TES occurs, people who expect a recall test and get a recall test perform significantly better than do people who expect a multiple choice (recognition) test but receive an unexpected recall test. Some (Leonard & Whitten, 1983; Underwood, 1972) believe that the
performance decrement accompanying the surprise recall test is due to qualitatively different processes that occur during study and recall. Authors of pertinent theoretical frameworks (Anderson & Bower, 1972, 1974; Gillund & Shiffrin, 1984) have explained that studying for a recognition test includes only part of the
processes required for good recall. Thus, when subjects are faced with a recall test when they expected a
recognition test they experience declined performance. Even though the recall TES effect has been replicated many times with word lists, few studies (Goldsmith, 1985) have investigated individual differences that might
affect performance in test expectancy conditions.
Other researchers have discovered that those who have poor organization skills require specific
instructions (Shuell, 1983), appropriate advance organizers (Tyler et al., 1983), meaningful learning
questions (Rickards & Hatcher, 1977-78), and highly structured materials (James & Brown, 1973). It seems that low achievers may require instructions to use
specific study strategies for recall versus recognition ta~*cs. The organizational data (James & Brown, 1973; Rickards & Hatcher, 1977-78; Shuell, 1983; Tyler et al.,
1983) indicates that poor or low organizers might not spontaneously select and employ different strategies for different types of tests which could result 1*; lower recall performance.
Research in memory organization has shown that a significant relationship exists between the degree to which someone can organize material on his or her own and the amount of information he or she can recall. This has been shown in studies concerning clustering (Bousfield, 1953; Puff, 1970) and subjective organization (Tulving, 1964; Mandler, 1967). See Appendix A for operational definitions.
The purpose of the present study was to investigate individual differences that may exist when considering susceptibility to part of the recall TES effect. How a person performs on a recall test when they expect a recognition test was the part of the recall TES effect examined. Chapter Two consists of a review of the
literature. Included in the review are sections on test expectancies in list learning as well as subjective
organization and clustering in memory organization. In addition, research on organizational skills, the
measurement of subjective organization and theoretical explanations of memory and organization are reviewed. In Chapter Three, the rationale and hypotheses of the
present study are presented. The methodology of the study is also described in Chapter 3. This includes details about the subjects, study materials for the categorized, unrelated, and test, expectancy sections, test materials, data collection and procedures, and analysis of the data. The results are presented in Chapter Four. The relationship between organizational ability and recall is addressed. The results are
discussed in terms of theoretical frameworks and other research findings. As well, the Hypotheses are addressed in the discussion section of Chapter Four. In Chapter Five major conclusions and limitations of the study are summarized. Suggestions for future research in the area of test expectancies} and individual differences are
Chapter Two: Review of the Literature
Even though there has been interest in classical testing effects for over fifty years (Jersild, 1929), investigations and the retical explanations regarding test expectancy did not increase noticeably until the early 1970's (Anderson & Bower, 1972, 1974; Kintsch,
1970). Since Underwood (1972) noted the lack of research concerning tna effects of test expectancy on both
learning processes and the retention of verbal material, various researchers have begun to investigate the area. Most of the early work on test expectancy focused on classroom applications and student performance in field settings (Meyer, 1936; Sax & Collet, 1968). Some of the mor’ recent studies have maintained this focus
(Duchaste.i, 1981) ; however, several researchers are now investigating the effects of test expectancy on recall and recognition of word lists (Balota & Neely, 1980; May & Sande, 1982). Please see Appendix A for operational definitions.
Test Expectancy Research List Studies
Recall tests.
In list studies using recall tests, the main finding has been the recall test expectancy superiority (TES) effect. When this effect occurs, people who expect a recall test and get a recall test perform significantly
better than do people who expect a multiple choice test but receive an unexpected recall test. When a
significant test expectancy is found, it is termed a recall TES effect. This effect is robust and replicable in list learning (Balota & Neely, 1980; Connor, 1977).
Recognition tests.
To date, a significant recognition TES effect has not been reported in the literature on recognition tests. The results concerning a recall or recognition test
expectancy on recognition tests have been considerably mixed. Only two studies (Hall, Grossman & Elwood, 1976; Neely & Balota, 1981) were found by this writer in which significance has been reported and that was for a free recall TES. In both of these studies high imagery and high frequency words were used. Conversely, May and Sande (1982) did not find a significant test expectancy by word frequency interaction. In their results rare words were retained better than common words and subjects expecting and receiving a recall test retained more words
(rare and common) than those subjects expecting a
recognition test and receiving a recall test. Jacoby's (1973) results indicated a significant cued recall test expectancy, suggesting that cueing may be an important variable. D'Ydewalle's (1981) results showed a
non-significant recognition test expectancy effect. The results concerning a recognition TES effect are
inconclusive; however, it appears that word-frequency is an important factor in retention.
Other factors investigated in list studies.
Categorized material and interpolated activities may be associated with non-significant trends for recognition test expectancy (Connor, 1977; Jacoby, 1973). Neely and Balota (1981) believed the number of study-test trials was important in inducing a test expectancy effect.
Their claim was that more study-test trials would produce a more robust test expectancy than would fewer study-test trials. As a point of interest, the test expectancy
effect has reached significance with as few as one study-test trial (d'Ydewalla, 1981; Leonard & Whitten 1983) and as many as six study-test trials (Balota &
Neely, 1980; Neely & Balota, 1981). The number of trials does not seem to be a major factor in inducing test
expectancy, at least with college-aged subjects. Free versus discrete study trials may be an important factor in test expectancy. Out of five experiments using free presentation of the study
materials (Connor, 1977; Hall, Grossman, & Elwood, 1976), only one showed a significant recall test expectancy effect (Hall, Grossman, & Elwood, 1976). The other studies showed a non-significant trend in favor of a
recognition TES effect for a recognition test. The study by Hall, Grossman, and Elwood (1976) differed from other
researchers' in that it did not use categorized material. In experiments using discrete presentation of study
materials, researchers have found a significant free recall TES effect (Neely & Balota, 1981) and a
significant cued recall test expectancy effect (Jacoby, 1973). Aside from these studies the results concerning free versus discrete study trials have been mixed
(Connor, 1977; d'Ydewalle, 1981; Hall, Grossman, & Elvrood, 1976, Experiment 3) .
Encoding and retrieval processes were used in theoretical explanations of recall and recognition expectancies (Anderson & Bower, 1972; Kintsch, 1970).
One problem with the theoretical explanations, which will be discussed later, is that they were based on post hoc methods of analysis. Because the data collection
paradigms did not investigate the encoding and retrieval strategies while the subjects were studying, researchers have examined serial position effects in recall
performance (d'Ydewalle, 1981; May & Sande, 1982). Serial position effects refer to one part (beginning, middle, or end) of a study passage being recalled better or worse than another part of a study passage. May and Sande (1982) found that serial position curves were very different for recall versus recognition expectancies. The recall expectancy position curves showed a strong primacy effect (words from the beginning of a list were
remembered better than words from the end of the list) and a linear slope whereas recognition expectancy
position curves showed no primacy effect and a flattened serial position curve. From this, May and Sande (1982) proposed that subjects expecting a recognition test will use different study strategies than will subjects
expecting a recall test.
Memory Organization Research
The study of associative learning and retention has been of interest for over a century with the publication of Ebbinghaus1 (1885) early nonsense syllable
experiments. At that time in history Ebbinghaus had a great deal of influence on the academic and experimental
"Zeitgeist” . Other researchers (Calkin, 1894;
Kirkpatrick, 1894) introduced a free recall paradigm similar to ones used today, only to be met by critical comments from Ebbinghaus (1902). He claimed that the free recall paradigm was not the most appropriate for the study of associative learning. Ebbinghaus (1885, 1902) endorsed research that studied associative learning via nonsense syllables because he felt it eliminated transfer of past learning. This basic tenet led researchers away from today's common free recall paradigm for decades. It was not until the early 1950's that there was a
significant interest in free recall experiments again. Endel Tulving (1968) claimed that when a person
recalls words in a free recall situation there are two different methods of organizing these words. The first method, which he claimed is the associatively weaker of the two, is called primary organization. This occurs when there are "consistent discrepancies between input and output orders that are independent of the subjects' prior familiarity with a set of input items" (Tulving,
1968, p. 15). An example of primary organization is serial position effects (Murdock, 1962). Conversely, secondary organization was described by Tulving (1968) as ocurring when, "the output order of items is governed by semantic or phonetic relations among items or by the
subjects' prior, extra-experimental or intra-experimental acquaintance with the items constituting the list"
(p.16). An example of secondary organization is
clustering on the basis of meaning. It is the method of secondary organization that is most often measured today and related to other independent and dependent variables. Many experiments have been conducted in the past (Shuell,
1969) which relate secondary organization to variables for the purposes of illuminating underlying processes of free recall. For these reasons secondary organization, and not primary organization, is included in this
literary review and the research study to follow. Two kinds of secondary organization have been identified by researchers (Tulving, 1968): clustering
and subjective organization. Clustering occurs when
words from a list are recalled in experimenter determined subsets. Words within each subset or category are
perceived as more similar than words in other subsets or categories. Empirically, words from a particular subset are recalled together more often that would be expected by chance. An advantage of this method is that
scientifically, this method of recall can be quantified from a single trial. However, this advantage can be limiting since some organizational processes may not be discovered by only one recall trial. This is one of the advantages of using subjective organization techniques. Subjective organization is measured when a subject
recalls the same material for more than one recall trial. The extent of subjective organization is quantified by the proximity of the recalled words across trials. This method does not require the experimenter's knowledge of subsets or categories in advance.
Clustering
In a somewhat restricted word association task, Bousfield and Sedgewick (1944) asked their subjects to give continuous verbal associations to a given category name, e.g., birds. They noticed that the subjects' responses tended to be in clusters or bursts spaced by temporal gaps. For example, domestic birds would be noted, and then birds of prey. Even though Bousfield and
Sedgewick (1944) had several explanations for this "clustering," they lacked a method of objectifying or quantifying their findings. Because of these problems, they abandoned restricted word association tasks and a few years later Bousfield (1953) reported an experiment on free recall and categorical clustering.
Bousfield (1953) aurally presented 100 subjects with a randomized list of 60 words made up of 15 words from each of four categories: animals, names, professions, and vegetables. The subjects were given 10 minutes to recall the words immediately after presentation. Bousfield
(1953) found that the subjects clustered the words in categories mere than could be expected by chance. The words were recalled in an inverted "U" pattern related to the sequence of presentation. That is, words were
recalled better in the middle of the test period than at the beginning or end (Bousfield, 1953; Bousfield & Cohen, 1953). He used his "ratio of repetition" which was a ratio of within category repetitions over total number of responses and compared the experimental results to an artificial "Monte Carlo" type study. This clustering was explained by Bousfield (1953) os attributable to habit strength and relatedness, or clustering, increments. He used the term habit strength to imply that a study word is more likely to be recalled be- ause it occurred in the study list, and relatedness to mean a word is more likely
to be recalled if an associated word is in the study
list. Based on these results, Bousfield (1953) satisfied the purpose of his study:
The theoretical significance of this
undertaking derives in part from the assumption that clustering is a consequence of
organization in thinking and recall. If
clustering can be quantified, we are provided with a means for obtaining additional
information on the nature of organization as it operates in the higher mental processes
(p. 229}.
It is important to note that the validity and reliability of Bousfield's (1953) -ratio of repetition" measurement index, along with other measurement formulas, were
scrutinized by Sternberg and Tulving (1977); however, it was the launching point for later quantification of
secondary organization.
One of the most important questions that Bousfield's (1. 1) experiment raised was that concerning the
relationship between categorical clustering and item recall (Kausler, 1974). To discover whether categorical clustering facilitated item recall, researchers (Cofer, 1967; Puff, 1970) also used an uncategorized list with words matched for frequency count and other important attributes. Puff (1970) found that in a single-trial
free recall task there was no difference in the amount recalled from the categorized list by subjects who were high clusterers and showed significant clustering
(mean=17/30) and those subjects who were low clusterers and whose clustering did not exceed chance
(mean=16.26/30); however, both categorized list groups recalled significantly more words (mean=16.66/30) than the group given the non-categorized list (mean=13.20/30). The finding that subjects in the categorized condition recalled more words than those in the non-categorized condition led Puff to comment that this superior recall "must involve some mechanism other than clustering on the basis of broad conceptual categories at the time of
output" (Puff, 1970, p. 386). Later in his paper, Puff (1970) stressed associative principles.
Another method of investigating whether clustering facilitates recall of list items is by determining
correlation coefficients. Intuitively, it would seem that there would be a moderate to high correlation
between clustering and item recall. However, Bousfield and Cohen (1955) indicated that the correlation may be low. Forrester and King (1971) reported the expected moderately high correlation coefficient (r = .71) between categorical clustering scores and total number of items recalled across categories. Other researchers (Kausler, 1974) believe the reason for this correlational
discrepancy is due to the type of methodology employed. For example, Forrester and King (1971) used a word list that had mixed or scrambled words across categories
rather than unmixed lists containing categorized items in one condition and uncategorized items in another
condition. Forrester and King presented 36 words, 12 were categorized with three words per category, while the remaining 24 words were acoustically related or unrelated words. It seems that the presentation of a mixed list may result in a recall-cluster relationship. Another contrast between Forrester and King's study and those of Puff (1970) and Bousfield and Cohen (1955) is that the former researchers had several input or study periods prior to the recall period while the latter researchers only had one study period prior to recall. Because there were only low correlations between clustering and number of items recalled when there was only one input or study period, it could be that multiple study periods play an important role in cluster-recall relationships,
especially with unmixed lists.
Shuell (1968) demonstrated that there is a
consistent increase in both number of words recalled and amount of clustering with practice. He showed his
subjects five words from each of seven categories over four trials. There were regular increments across trials with the following increases: mean number of words
recalled increased from 14.88 on Trial 1 to 27.46 on
Trial 4; mean number of words per category increased from 2.53 on Trial 1 to 4.13 on Trial 4; and Bousfield's
(1953) ratio of repetition clustering increased from .43 on Trial 1 to .70 on Trial 4. This consistent covariance seemed to suggest the possibility of a causal
relationship. Shuell's (1968) findings were supported by those of Thompson, Hamlin, and Roenker (1972). Their subjects were presented with three lists, three study periods per li-‘t. Each list contained 48 words with 12 words from each of four categories. The subjects were divided into two groups, one of high clusterers and one of low clusterers, based on their recall performance on the first three recall trials (List 1). The high
clusterers outscored the low clusterers on every trial, but the number of words recalled by both the high and low clusterers increased across trials on every list.
Thompson et al. (1972) attributed this deviation from Puff's (1970) findings as being due to high versus low clusterer group assignment after multiple trials and not a single trial as in Puff's (1970) design. Weist (1970, 1972) also found results similar to Thompson et al.
(1972) when he calculated correlation coefficients between clustering scores and it^.m recall scores. It seemed that given reliable group assignment, after several trials instead of a single trial, a covariation
between clustering scores and item recall scores could be demonstrated.
other variables affecting clustering.
Researchers have tested Bousfield's (1953) classic clustering paradigm using different types of categories. Among the most investigated are those that are composed of acoustical clusters (rhyming words), grammatical clusters (adjectives, nouns), and alphabetical clusters
(defined by initial letters in the words).
Those who have used acoustical clusters have reported mixed results. Bousfield and Wicklund (1969) and Fagan (1969) found greater than chance amounts of clustering in unmixed lists while Dolinsky (1972) found the same effect with mixed lists. Examinations of
categorized versus uncategorized lists or investigations of clustering and recall were not included in these
studies. Dolinsky reported that acoustical clustering is reduced when other means of clustering are present, e.g., associative. However, Long and Allen (1973) found that acoustical clustering was more likely than semantic clustering in their study. Forrester and King (1971)
failed to find acoustical clustering when their subjects were presented with mixed lists embedded with
acoustically similar words. In addition they
investigated the relationship between clustering and recall of acoustical words and did not find it to be
significant (r = .27). When this study was followed by one using an unmixed list, Forrester (1972) did not find acoustical clustering.
Cofer and Bruce (1965) did not find significant grammatical clustering when they presented 12 words from each of the following categories: ui.relat.ed nouns,
verbs, and adjectives. The words were presented in four study-recall periods in a random fashion. Later, when they replicated the study and presented the words in a blocked non-random fashion the results were the same. Koplin and Moates (1968) reported the same results even when they prompted the subjects by telling them to use the grammatical categories as a learning aid. It must be noced that the absence of experimenter suggested
categories does not necessarily imply that the subjects did not organize the materials with other methods.
Freund and Underwood (1969) found clustering due to alphabetical categories in the subjects that were
forewarned, but not in those who were unprompted. For those who were forewarned 63.6 percent of the
alphabetical pairs were clustered together compared to only 19.3 percent for those subjects who were not
instructed. When the relationship between alphabetical clustering and recall was explored, there was not a
significant difference between those who were instructed (mean = 13.00) and the control group (13.38). Earhard
(1967) reported a relationship between number of
alphabetical categories and recall but she did not report analyses on the relationship between use of alphabetical categories and recall.
Other category characteristics have been
investigated to determine the relationship between
clustering and recall. Cohen (1963, 1966) found that the relationship was greater when exhaustive categories were used, for example, the seasons of the year, rather than non-exhaustive categories such as animals. Cohen (1963) proposed that this was due to stronger superordinate
(names of categories) to subordinate (category members) relations. In his experiment, he asked his subjects to recall the category names rather than the category
members. Even though there were only two categories
present in the study list, many more names were given for both categories. Cohen concluded that superior detection of superordinates should have been accompanied by
superior recall of the category names. Garner (1962) also attempted to explain the better recall and
clustering for exhaustive categories than nonexhaustive categories. He claimed that this difference was due to "redundancy" or the necessity of having to re-learn a list of words. He suggested that exhaustive categories only require knowledge of the category because the
According to Garner the lack of redundancy for exhaustive categories accounts for better clustering and recall
because less time and effort is spent on re-learning the list of items.
Category size has also been used to investigate the relationship between clustering and recall. Slamecka
(1967) had three groups of subjects learn lists
containing 20 items each. The population from which the 20 items were drawn varied for each group of subjects. They were 20 of the 50 U.S. states, 20 of the digits from 10 to 90, or 20 of 260 possible single letter-single
digit combinations. Slamecka found that as the size of the population decreased recall increased.
Unfortunately, the stimuli for each condition group were so vastly different that his conclusions were confounded. A second study that concerned category size was conducted by Thompson and Poling (1969). They presented their
subjects with lists of words from 15 different categories. Another group of subjects ranked the
category according to how many members they believed each category had in its population. Thompson and Poling
(1969) found that recall covaried inversely with size of category. This was found in Handler's (1967) study also. Researchers have investigated high and low frequency words and their relationship to recall and clustering.
found that word recall and clustering was better for high frequency than low frequency words. This was replicated later by Cofer, Bruce, and Reicher (1966) who departed from the traditional randomized sequence of word list presentation and longer item exposure times of three to
four seconds. Cofer et al. (1966) conducted three
experiments that manipulated high and low frequency words and the following variables: blocked versus unblocked presentation (high and low frequency words grouped separately or mixed together), duration of item
presentation, and immediate versus delayed recall. They reported that block presentation facilitated clustering for both high and low frequency words and aided recall in high frequency words. Clustering also increased as the duration of the exposure interval increased and when immediate versus delayed recall was used. Other
researchers have found that clustering and recall are facilitated by blocked presentation (Dallett, 1964? Puff, 1966).
Jenkins and Russell (1952) conducted a study that used words of varying degrees of associations. The 24 words from the pairs were presented in random order.
When asked to write down the words the researchers judged how many of the pairs were recalled in an adjacent
forward order (stimulus word first) and how many were recalled in an adjacent reverse order (response word
first). Jenkins and Russell (1952) found a mean of 3.41 pairs recalled in the forward order, and a mean of 2.85 pairs recalled in the backward order. When they compared their results to the mean expected in a "Monte Carlo" study (.26) their results were easily significant. They concluded that associative strength is an important
factor in the relationship between recall and clustering. In later studies this relationship was replicated with adults using word pairs (Bousfield, Whitmarsh, &
Berkowitz, 1960; Jenkins, Mink, & Russell, 1958;
Marshall, 1967; Walker, 1971) as well as in studies using unpaired words of var .ng degrees of association (Deese,
1959; Matthews & Hoggart, 1970; Weingartner, 1964). The study by Cofer et al. (1966) was one of the first to depart from the traditional Bousfield (1953) experimental format by varying presentation styles, durations, and recall delays. Other researchers soon followed suit. Bousfield's (1953) traditional
methodology (Kausler, 1974) consisted of four components: (a) the subjects were not told about the categorical nature of the list they received, (b) words were
presented in random order, (c) categorical names were not given during the study period, and (d) recall was uncued. These standards stood as a control procedure for later experimental procedures.
D'Agostino (1969) manipulated both instructional and presentation formats in a 2 x 2 design. The subjects were either told or not told about the categorical structure of the lists. The lists were presented in a blocked or unblocked format. D'Agostino (1969) measured mean correct responses, mean number of categories, and mean items per category. The number of categories
recalled was significantly higher for the instructional variable. It seemed that knowledge about the categories in the list had little effect on item recall. One
explanation for this finding could be that more than one study-recall trial was needed for item recall to be
affected as Cofer (1965) found. D'Agostino (1969) also found that blocking the categorized words significantly increased the number of words recalled when compared to performance in a random or unblocked condition (means = IS.23 versus 16.73 ana 7.47 versus 16.13). These
results supported tho^.e of Cofer et al. (1966) who also found that blocked presentation increased recall. In addition, when Puff (1966) varied his blocking of list words from none at all, to some promixity, to total
blocking, he found that recall of words increased (means = 15.20, 15.73, 16.93, 18.40) and clustering increased
(means = 3 . 7 0 , 5.24, 7.99) as the proximity of category membership increased in the list words.
Segal (1969) had one group of subjects study to remember a random list of 44 high and low frequency words. As his experimental group he had subjects study to remember a similar list of 55 words (the words plus the 11 category names) that included category names. He had an additional group of subjects test effects of list length. All of the subjects had six study-test trials. Segal (1969) reported that category names had a
significant effect on clustering. Category names seemed to have a negative effect on recall for high frequency words and a positive effect on low frequency words. Segal (1969) surmised that the category names may have caused the subjects unproductive memory work as they attempted to remember them as both words and category names. Because recall for the high frequency words was already good, trying to remember the category names may not have been worth the trouble, Segal explained.
Conversely, memory of the category names for the low frequency words may have facilitated recall enough to make the effort worthwhile.
Tulving and Pearlstone (1966) conducted an
experiment to investigate recall and clustering when the words were cued or uncued during the test period. In the study phase of both conditions the list words were
blocked according to category, and the category names were contained in the study list. The subjects were told
that they did not have to remember the category names. In the test phase of the cued condition, the blank test sheet had the names of the categories printed across it. The uncued condition group received only a blank test sheet. Tulving and Pearlstone (1966) found that cued recall significantly exceeded uncued recall. Other researchers used the study as as basis to determine the effect of the presence of category names during the study period. Both Wood (1967) and Crouse (1968) found that recall of high frequency words was facilitated by
retrieval cues if the category names were or were not in the study phase. Conversely, retrieval cues aided the recall of low frequency words only if the category names were present in the study phase.
Investigators have considered the relationship between recall, clustering, and number of categories. When Earhard (1967, 1972) used cued and uncued recall procedures she found that the use of categorized lists was only effective when the number of words per category was fewer than six or seven items. Because of the
interest in Miller's (1956) "magic number" of 7 plus or minus two, much attention was paid to Earhard's (1967, 1972) studies. She measured both cued and uncued recall across 20 trials for seven different list conditions. The results were that cued recall varied with the number of categories on each list. Interestingly, the mean
number of categories for best cued recall was 6.6, a close approximation to Miller's (1956) "magic number". Earhard (1967, 1972) concluded that cued recall exceeded uncued recall when the number of categories is less than 6.6 but is less than uncued recall when the number of items per category is greater than 6.6. Tulving and Pearlstone (1966) and Dallett (1964) also found that the relationship between number of categories and recall was positive with cued recall and inverse with uncued recall. Subjective Organization
In his classic study Tulving (1962) defined the concept of subjective organization as "...the subject's tendency to recall items in the same order on different trials in the absence of any experimentally manipulated sequential organization among items in the stimulus list"
(p.354). Tulving's (1962) subjects learned 16 so-called unrelated nouns presented in a different order on each of
16 trials. He noticed that the subjects tended to group the recalled words together in the same order on repeated trials. When he compared his subjects' organization of recalled words to statistical subjects in a Monte
Carlo-type study, the organization was clearly above
chance as it increased with each trial. In addition, the number of words recalled increased with the amount of subjective organization (r=.78). He concluded there was a "positive correlation between organization and
performance" (p. 345). Other researchers have replicated this relationship (Allen, 1968y Mayhew, 1967; Tulving, 1964).
Cofer (1967) argued against the use of xnultitrial procedures because he did not believe the procedure
parallelled reality. Conversely, Tulving (1968) believed that multitrial procedures were crucial to understanding subjective organization processes. Shuell (1969)
addressed this issue by stating that single-trial and multitrial procedures address different issues. He explained that single-trial procedures are usually used with well-defined categories in the manipulation of a variety of independent variables. Conversely, multitrial procedures are used to investigate subjective
organization processes as they develop and become stable in the absence of well-defined item relationships.
Tulving (1964) pursued these findings in a later study. He investigated the possibility of a causal
relationship between subjective organization and recall. Tulving (1964) had an experimental group of 12 subjects learn a list of 22 common nouns. Before the study period they were shown the 22 nouns over 6 trials. They read each word aloud. The results of these subjects were compared to those of a control group who did not see the words prior to the study session. There were no
experimental and control group. Tulving (1964) showed that repeating words over and over again does not
facilitate recall.
Tulving (1966) demonstrated that under certain
conditions subjective organization, and therefore recall, can be inhibited. He showed two groups of subjects a list of nine words to recall over each of 12 trials. After the 12 trials with the nine word list, all of the subjects were given 12 learning trials with a second list of 18 words. For one group the second list was composed of all new words. For the other group, nine of the words had been learned from the prior list. The group who had learned part of the list previously started out with higher recall, but by seven trials the group who had all new words outscored them. Tulving (1966) concluded that the subjects who learned part of the list previously were unwilling to abandon or change their subjective units. He claimed that recall is due more to subjective
organization than to habit strength as Bousfield (1953) suggested.
In addition to replicating Tulving's (1966)
findings, Novinski (1972) also found that the negative effect of part to whole list learning can be eliminated by informing the subjects of the overall nature of the list. After her subjects studied the part list she told them it would be merged with new words to form a new
list. The result was that practice on the part list
resulted in better performance on the whole list than was found in a control group condition. Novinski (1972)
argued that providing her subjects with this information helped them subjectively organize the whole list.
Bower, Lesgold, and Tiem.m (1969) found similar
results when they interfered with subjective organization processes. All of the subjects were encouraged to make visual images to aid later recall. They presented 24 nouns in groups of four to their subjects. For half of the subjects the same "four-tuples" were shown over four trials. The other half of the subjects were shown
four-tuples across four trials also except that the words in the foursomes changed for each trial, thus interfering with subjective organization processes. The researchers discovered that the subjects who were not permitted to form stable organizations from trial to trial recalled significantly fewer words than the subjects who had the subjective organization opportunity.
The instructions that are given to study
participants seem to play an important role on subjective organization and recall. Allen (1968) allowed a control group to study their words in a self-paced fashion, while the experimental group was presented their words one at a time and were required to spell each word out loud at it was presented. As expected the experimental group
suffered significantly lower recall and subjective
organization than the control group. It seemed that the experimental group was handicapped in their ability to
form associations or units to aid recall. They did not demonstrate the subjective organization -• recall
relationship possible under more typical conditions (Mayhew, 1967; Tulving, McNulty, & Ozier, 1965).
Mandler and Pearlstone (1966) also interfered with subjects' organizational processes. They gave their subjects a deck of 52 cards, each with a word on it, and asked them to sort them into categories of two to seven words. The control group could sort the words using any method they wished, while the experimental group was required to sort the words into experimenter-defined categories. Each subject was asked to re-sort the word cards until they sorted them exactly the same way twice in a row. To counterbalance the difficulty of this task, each experimental group subject had tr* sort the cards in the same way as their yoked control group subject. Once each subject sorted to criterion he or she was asked to recall as many words as possible. Mandler and Pearlstone
(1966) found that the control group needed fewer trials to reach criterion. The experimental group subjects took approximately twice as long to reach criterion. Both groups could recall approximately the same number of words (20), however, the experimental group had more
rehearsal prior to reaching criterion. The results
showed that it is not necessarily the amount of learning trials that determines recall but rather the type of organization allowed during study. Mandler and
Pearlstone (1966) also found that there was a significant correlation between recall and the number of categories used in sorting (r=.95). Their subjects remembered a mean of five words per category, close to Miller's (1956)
"magic number". Interestingly, the correlations between trials to criterion in the sorting task and recall did not correlate significantxy (r=.16), supporting their conclusion that the type of organization facilitated recall. In a slightly modified version of Mandler and Pearlstone's (1966) study, Dong and Kintsch (1968) also found that more words were recalled under cued versus uncued conditions when the cued condition involved the subjects providing their own category names.
Later, Mandler (1967) conducted a similar experiment but held the number of sorting trials constant at five. Half of the subjects were given recall instructions and half were rot. In addition, half of the subjects sorted the cards into two to seven subjective categories, and the other half placed the words one after another into seven columns for the five trials. The subjects who recalled about the same number of nords had received either organizational or recall instructions. The
subjects who recalled significantly fewer words had not received instructions for recall or organization of the words. These results supported the notion that free
recall is facilitated by subjective organization be it by instructions to recall cr by instructions to organize. As Postman (1964) suggested, it is not the intent to
learn that has an effect on learning, but rather the correct responses to learning the material.
Earhard (1967) reported her interesting results concerning subjective organization ability and recall. She used the recall protocols of Tulving's (1962) study to construct two lists of words. One of the lists
consisted of word pairs that Tulving's (1962) subjects had paired together during recall, the other consisted of word pairs that his subjects did not pair together.
Earhard (1967) called her two lists the preferred-order list and the non-preferred-order list respectively. To complete her 2 x 2 design Earhard had poor and good subjective organizers study and recall the list items. As hypothesized, Earhard (1967) found that her subjects recalled more words from the preferred-order list than from the non-preferred-order list. However, the poor organizers showed a significantly greater difference between the two types of lists than did the high
organizers. The interaction between the type of list and ability level was significant. The poor organizers had
difficulty making associations between the unrelated word pairs but they performed much better when they had to recall the related word pairs. In other words, once the poor organizers were given a study aid or mnemonic their recall ability improved. The good crganizers did not benefit similarly from the presence of the more related words indicating an ability to form these relationships themselves. Earhard and Endicott (1969) and Rogers and Battig (1972) supported these findings in later studies. However, there may be reason to suspect that
organizational processes may be different for high and low organizers. Shapiro and Btll (1970) found
significant correlations between subjective organization and recall for high organizers only. The subjective organization scores of their low organizers remained low throughout learning even though they eventually recalled all of the words.
Organizational Skills
Shuell (1983) investigated whether or not instructing students to organize recall would
differentially facilitate the performance of "fast" and "slow" learners. In Experiment 1 he asked the learners to organize their list word recall alphabetically. The number of words recalled by both the fast and slow
learners increased significantly. Shuell (1983) proposed that the type of organization suggested was either
overlooked or not normally used by the learners. In Experiment 2 the learners were asked to organize their word recall categorically. The results were facilitated performance for the slow learners and little difference
in performance for the fast learners. When the learners were informed of the organization of their study
materials and encouraged to use appropriate study strategies, the difference in performance between the fast and slow learners was cut in half. From this, Shuell concluded that the difference between fast and slow learners may lie in their ability to use
organizational strategies most appropriate for the material being learned. Because the organizational
instructions had a facilitative effect of the performance of only the slow learners, Shuell proposed that the
learning strategy was more effective than the one
normally used by slow learners and either less effective or, at least, no more effective than the one used by average and fast learners.
Tyler, Delaney, and Kinnucan (1983) investigated good and poor readers' abilities to use advance
organizers, which the authors described as a short paragraph used to bridge the gap between what the
students already knew and information they were about to learn. The suojects were asked to read short passages preceded by one of two types of advance organizers; one
was a summary of the passage, and the other contained definitions and explanations of major concepts needed to understand the article. The second type of advance
organizer followed Ausubel's (1968) criteria of relating the information to what the reader should already know. The researchers found that good readers showed greater recall of detail given either type of advance organizer, while poor readers displayed enhanced recall of detail for the Ausubel-type advance organizer and not the summary organizer. The authors suggested that the
Ausubel-type organizer may have compensated for the poor readers' organizational deficits (Tyler et al., 1983).
Rickards and Hatcher (1977, 1978) studied the effect of interspersing meaningful learning questions in text prose as semantic cues for poor comprehenders. They had the subjects read 800-word passages, and, after every two paragraphs, they received a meaningful learning question requiring the organization of facts under given concepts, a rote learning question requiring literal recall of
passage information, or no learning question of any kind. For good comprehenders the type of treatment did not have a significant effect on text recall; however, for poor comprehenders the meaningful learning questions
facilitated significantly greater recall of facts than did the rote learning questions or no question at all. Rickards and Hatcher claimed their results supported the
usefulness of Ausubel's theory (1968) in constructing questions appropriate for poor comprehenders. It
appeared that poor comprehenders did not spontaneously organize paragraph details around main ideas of a
passage. Conversely, good comprehenders seemed to
generate a context while reading text, thereby decreasing the need for context-providing cues.
James and Brown (1973) studied the effects of prose organization and individual differences on free recall. They investigated three main topics: (a) the effect of paragraph organization on free recall of sentences and on the selection of clustering strategies; (b) the effect of different organizational skills on a person's recall and strategy selection; and (c) the effect of
organizational skills on other cognitive abilities. The students were asked to study paragraphs organized by concept names, concept attributes, or at random. The group that studied paragraphs organized by concept names recalled the most correct statements, although clustering by names was predominant for all groups. A relationship between cognitive factors and recall scores was found for each group. More importantly, superior recall emerged when the students learned a highly organized passage and a preferred recall strategy. Highly organized students were not greatly influenced by the organization of the paragraphs, whereas low organizers were. The results
indicated that students low in organizational skills may have required highly structured materials, while students high in organizational skills performed similarly on
materials that were high and low in structure.
These studies point out the importance of providing low organizers with specific instructions (Shuell, 1983), appropriate advance organizers (Tyler et al., 1983),
meaningful learning questions (Rickards & Hatcher,
1977-78), and highly structured materials (Jamas & Brown, 1973). If these results can be generalized to a test expectancy situation, low achievers may require
instructions to use specific study strategies for recall versus recognition tasks. The organizational data
indicates that low organizers might not spontaneously select and employ different strategies for different types of tests which could result in lower recall performance.
The Measurement of Secondary Organization Since the onset of empirical studies concerning secondary memory organization, both clustering and subjective organization, various mathematical formulae have been considered to quantitatively represent the phenomenon (see reviews by Shuell, 1969 and Sternberg & Tulving, 1977). The method that has proven to be the most robust and relevant to the present study focuses on participants' subjective organization of the words,, With
this method, the subject is given the same list of words to study, in a different order each time, over several trials. After each study period, the subject takes a free recall test, where he or she is asked to recall the words in any order. In their article, Sternberg and Tulving (1977) reviewed several methods used to quantify the measurement of subjective organisation. After
considering the strengths and weaknesses of each method, their conclusion was that the bidirectional intertrial repetition (ITR) or pair frequency (PF) method is the preferred method. The PF measure, initially proposed by Bousfield and Bousfield (1966), proved to be the most reliable and empirically valid measure of subjective organization in their review. The PF score assesses the amount of subjective organization for each subject across trials, consequently, each subject obtains one PF score for every two successive recall trials. Recalled words that are submitted to the PF formula are measured
according to the pairs of words that were recalled adjacently on consecutive trials. The result of this mathematical formula is the PF score. This procedure is called a multitrial free recall task and is appropriate for categorized and "unrelated" lists of words.
Theoretical Explanations of Memory and Organization Since the early studies (Bousfield, 1953; Bousfield & Cohen, 1953) in secondary organization there have been
several attempts to produce theoretical explanations of the underlying processes. As Bousfield (1953) stated in his paper, "The progessive change in the clustering
tendency is explained on the basis of two assumptions, namely, habit strength and clustering increment"
(p. 240). Bousfield (1953) proposed that habit strength was increased by the reinforcement an item receives
before and during the experiment. Once the word is
recalled, an increment is added to words that are related to that item. Habit strength and increments were what accounted for clustering and the likelihood of the next recalled word being from a related or different category. This theoretical explanation was accepted throughout the
1950's .
However, in the 19605s more research began to be conducted in the area of secondary organization and thus mere theoretical explanations were presented and
scrutinized. Researchers such as Tulving (1968) pointed out that it may not be habit strength or increments that explain clustering but rather recency effects. Other researchers (Deese, 1968) began to draw on Hebb's (1949) proposal. Hebb claimed that relatedness increments were a result of superordinate perceptions. In other words presentation of several stimuli results in a
items). Later, the presentation of a single item would cue or activate the other items.
An explanation that is similar to the
superordinate-subordinate explanation is the coding
interpretation. Miller (1956) explained that when chunks are formed in memory, the subject recodes the stimulus items into categories and then stores the category label. This idea is supported by the finding that if one word from a category is recalled, the number of words from that category that are recalled can be consistent across varying conditions (Miller, 1956). The important notion
in this approach is that items are encoded into categories upon presentation and recalled as a unit.
Bousfield and Puff (1965) chose to emphasize the associations between the items more than the
superordinate structures. They wanted to discover if associative relations were enough to explain clustering. They revealed that both the associative and the coding approaches were required to explain their free-recall clustering. This was supported by other studies as well
(Marshall 1967; Cofer, 1965, 1966). Tulving and
Pearlstone (1966) added that items can be present for recall in storage, but will not be recalled unless sufficient cues are provided. This was replicated in later studies (Dong & Kintsch, 1968; Tulving & Osier, 1968).
