Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Differences Between Cues In Effectiveness As Retrieval Aids
(USC Thesis Other)
Differences Between Cues In Effectiveness As Retrieval Aids
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
This dissertation has been
microfilmed exactly as received ^
LOEB, Jane Watkins, 1938-
DIFFERENCES BETWEEN CUES IN EFFECT
IVENESS AS RETRIEVAL AIDS.
University of Southern California, Ph.D., 1969
Psychology, general
University Microfilms, Inc., Ann Arbor, Michigan
DIFFERENCES BETWEEN CUES IN EFFECTIVENESS AS RETRIEVAL AIDS
by
Jane Watkins Loeb
A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(Psychology)
January 1969
THE GRADUATE SCHOOL
UNIVERSITY PARK
LOS ANGELES, CALIFORNIA 9 0 0 0 7
This dissertation, written by
......... Jajae..W.ajtkins..LQsJ?..........
under the direction of hzr.... Dissertation Com
mittee, and approved by all its members, has
been presented to and accepted by The Gradu
ate School, in partial fulfillment of require
ments for the degree of
D O C T O R OF P H I L O S O P H Y
f j Dean
Date January. .1.9.63.
DISSERTATION COMMITTEE
......
. . . M d l A
ACKNOWLEDGMENTS
The author wishes to thank the members of her com
mittee, Professors Newton Metfessel, Robert Priest, and
Henry Slucki, for their valuable assistance and to extend
special thanks to her committee chairman, Professor Norman
Cliff, for his particularly excellent training and guidance.
To the staff of the University of Southern Califor
nia Computer Sciences Laboratory and especially to Profes
sor Richard Wolf are due thanks for their assistance in the
analysis of the data.
Warm thanks are extended to Mrs. Alice Stowell for
her helpfulness.
Finally, the author is grateful to her husband,
Peter, without whose cheerful support this research would
not have been accomplished.
ii
TABLE OF CONTENTS
Page
LIST OF TABLES v
LIST OF FIGURES vii
Chapter
I. INTRODUCTION 1
II. ORGANIZATION IN FREE RECALL 6
Clustering in Free Recall
Subjective Organization
The Category-Recall Relationship
The Power of a Retrieval Cue
Differences among Cues in
Retrieval Power
Subjects
Apparatus ---
Stimulus List
Design and Procedure . ] .
Digit Span
IV. RESULTS..................................... 31
Hypothesis 1
Hypothesis 2
Hypothesis 3
Hypothesis 4
Hypothesis 5
Hypothesis 6
Hypothesis 7
V. DISCUSSION.................................. 76
VI. SUMMARY................................ 90
REFERENCES..........' ............... 94
III. METHOD
22
iii
Page
APPENDIXES............................................ 97c
APPENDIX A. LIST WORDS AND CUES........ 99
APPENDIX B. INSTRUCTIONS, GROUP C, A AND F. . 101
APPENDIX C. RECALL SHEET, GROUP C, A AND F. . 105
APPENDIX D. LATIN SQUARE FOR ORDERS OF
CATEGORY SIZES ON CUED
RECALL SHEETS ................... 109
APPENDIX E. INSTRUCTIONS FOR DIGIT SPAN TEST. Ill
APPENDIX F. DIGIT SPAN RECALL SHEET ......... 113
Iv
LIST OP TABLES
Table Page
1. Number of Words in Small and Large Categories
Above and Below Median Taxonomic and
Thorndike-Lorge Frequency for the List . . . . 26
2. Mean Length of Words in All Categories......... 26
3. Analysis of Variance of Number of Words
Recalled by All Three Groups . ............. 33
4. Analysis of Variance of Number, of Words
Recalled by Groups C and P ................... 33
5. Analysis of Variance of Number of Words
Recalled by Groups P and A ................... 34
6. Analyses of Variance of Ratio of Repetition
Based on Conceptual and on Alphabetical
Categories, Group P .................. 37
7. Analysis of Variance of Mean Nonzero Pro
portional Recall for Trials 1-3,
Groups C and A ................................ 44
8. Analysis of Variance of Mean Nonzero Pro
portional Recall for Trials 4-6,
Groups C and A .....................';.... 44
9. Analysis of Variance of Mean Nonzero Pro
portional Recall for Trials 7-9*
Groups C and A . . ....................... 45
10. Analysis of Variance of Mean Nonzero Pro
portional Recall for Trials 10-12,
Groups C and A ................................ 45
11. Analysis of Variance of Mean Proportional
Recall, Including Zero Recall, for
Trials 1-12, Groups C and A ................. 46
12. Analyses of Variance of Mean Proportional
Nonzero Recall, for 3-Trial Blocks,
from Categories of Different Lengths,
Group C ................................. 49
v
Table Page
13. Analyses of Variance of Total Recall,
Trials 1-12, from Short and Long
Categories, Groups F and A ................. 50
14. Median Test for Number of Words In Non
recalled Categories, Groups P and A ... . 52
15. Variances of Words Recalled •................. 52
16. Analysis of Variance of Total SO for
2-Trial Blocks, Groups C and A . ........ 55
17. Analysis of Variance of Total SO for
6-Trial Blocks, Groups C and A ............. 57
18. Analyses of Variance of SO for 2-Trial
Blocks within Categories of from
4 to 16 Words, Group C ...................... 6l
19. Analyses of Variance of SO for 2-Trial
Blocks within Categories of from
4 to 16 Words, Group A ..................... 62
20. Analyses of Variance of SO for 6-Trial
Blocks within Categories of from
6 to 16 Words, Group C ..................... 66
21. Analyses of Variance of SO for 6-Trial
Blocks within Categories of from
6 to 16 Words, Group A ...................... 67
22. Correlations of Log SO for 6-Trial Blocks
and Recall within Categories of
Different Sizes, Group C ................... 70
23. Correlations of Log SO for 6-Trial Blocks
and Recall within Categories of
Different Sizes, Group A ................... 70
24. Number of Categorical Intrusions over
12 Trials, Groups C and A ................. 72
25. Correlations of Digit Span with NC,
Groups A, C, and P .......................... 74
26. Correlations of Digit Span with Recall,
Groups A, C, and P .......................... 74
vi
LIST OF FIGURES
Figure Page
1. Mean Words Recalled as a Function of
2-Trial Blocks, Groups A, C, and F .. 32
2. Mean RR Based on Alphabetical and Conceptual
Categories, Group F ....................... 36
3. Mean Proportional Recall for Trials 1-3 from
Categories with Nonzero Recall ............. 39
4. Mean Proportional Recall for Trials 4-6 from
Categories with Nonzero Recall ........ 40
3. Mean Proportional Recall for Trials 7-9 from
Categories with Nonzero Recall ............. 4l
6. Mean Proportional Recall for Trials 10-12
from Categories with Nonzero Recall .... 42
7. Mean Proportional Recall for Trials 1-12
from All Categories of a Given Length . . . 43
8. Mean Total SO within 4 to 16 Words Categories
for. 2-Trial Blocks................. 5^
9. Mean Total SO within 6 to 16 Word Categories
for 6-Trial Blocks................... 56
10. Mean SO, 2-Trial Blocks, within Categories,
Group C .................................... 59
11. Mean SO, 2-Trial Blocks, within Categories,
Group A .................................... 60
12. Mean SO, 6-Trial Blocks, within Categories,
Group C .................................... 64
13.. Mean SO, 6-Trial Blocks, within Categories,
Group A .................................... 65
vil
CHAPTER I
INTRODUCTION
Miller (1956) has suggested that human memory is
limited in its capacity, and that this limitation requires
the human learner to organize groups of stimuli into
"chunks" which are stored and retrieved as units. The num
ber of such units that he can retrieve is limited, but by
the process of "chunking," each unit comes to carry an in
creased amount of information. In this way the subject
overcomes the limitation of his immediate memory capacity.
It is indeed well documented that in the multi-trial
free recall situation the human S does impose organization
on the material which he learns. In multi-trial free re
call, stimuli are presented to S on each of a number of
trials and he is asked to recall as many of the items as
possible, in any order on each trial. Several investiga
tors (e.g., Tulving, 1962a; Bousfield, Puff and Cowan,
1964) have shown that words which are presented to S in
different orders on each trial are recalled by S in increas
ingly similar orders as learning progresses. This growing
similarity in S's order of recall is known as subjective
organization (SO) and degree of SO has been shown to be
strongly correlated with the number of words recalled.
If, as Miller's theory and the evidence on SO sug
gest, learning progresses via the grouping of items together
into "chunks" or "subjective units," then the following are
two possible sources of intersubject variability in free
recall performance: (l) the number of units which an S
uses, and (2) the number of items which he can store and
retrieve per unit.
There is a growing body of evidence that number of
subjective units can indeed function as a potent variable
in recall performance. In the original demonstration of
this phenomenon (Mandler and Pearlstone, 1966) Ss were
asked to sort 52 words into from 2 to 7 categories of their
own choosing. After they had sorted all items consistently
on 2 successive trials, they were given an incidental re
call test. For Ss who were allowed to choose and learn
their own category schemes the correlation of number of
words recalled with number of categories, or units, was
.955. In a recent series of experiments with this categor
izing task, Mandler (1967) found consistent confirmation of
the linear relationship between number of categories and
recall. Extrapolating from these results, Mandler con
cludes that free recall performance probably varies with the
number of categories or units which S uses.
The number of items which can be stored and re
trieved per unit has been studied primarily through the cued
recall experiment, in which S is given cues during recall.
The list is constructed so that a cue, e.g., a category
name or a letter of the alphabet, is relevant to a certain
number of list items. Proportional recall can then be com
pared for categories or units of different sizes. Using
alphabetical cues, i.e., first letters, Earhard (1967)
found that proportional recall decreased as number of items
per cue increased and that free recall was actually superior
to cued recall when cues were relevant to more than 6 items.
This suggests a rather low limit to the number of items a
cue can effectively retrieve; in fact, it suggests that the
number of items retrievable per unit is about the same as
the number of units which S can handle.
Although this type of experiment has shown that for
a particular type of cue, recall is negatively related to
number of items per cue, there have been no comparisons re
ported of different types of cues. It is reasonable to ex
pect that for a given set of words, cues can be devised
which have varying probabilities of retrieving the words
in recall. If this were shown to be the case, it would sug
gest that free recall performance varies as a function of
the particular cues or units used, as well as the number of
these units.
4 .
The research reported here is addressed to the prob
lem of demonstrating differences between sets of cues in
the number of items retrievable per cue. That is, a list
of words was devised for which two sets of cues were rele
vant, and the recall of Ss given the different cues was
compared. The dimension on which the sets of cues differed
was the frequency with which they evoked the relevant words
from the list as free associates. If the items a cue is to
retrieve are high frequency associates of the cue, more
items should be recallable than if they are relatively low
frequency associates of the cue. That is, words which in
general occur frequently as responses to a cue word will be
more likely to be evoked by the cue during recall than will
words which occur less frequently as responses to their
cues. Consequently, a list was devised such that for one
set of cues, the relevant words on the list were relatively
low frequency’ associates, and thus the probability of the
cues evoking the list words during recall was low; for the
other set of cues, the relevant list words were relatively
high frequency associates and thus the probability of the
list words occurring as responses to their cues during re
call was relatively high. The basic assumption made in de
vising the two sets of cues used here is that on the average
a word will be given as a response to a conceptual category
name relevant to It more frequently than it will be given as
a response to its first letter. For example, it is assumed
that "dog" will occur more frequently as a response to
"animal" than to "d." The two sets of cues compared, then,
are conceptual category names and first letters. There are
two primary hypotheses to be tested: (l) recall by Ss given
conceptual cues will be superior to recall by Ss given al
phabetical cues, and (2) the number of items which a con
ceptual cue can retrieve will exceed that which an alphabet
ical cue can retrieve; thus, for conceptual cues, as number
of items per cue increases, proportional recall will de
crease less quickly than it will for alphabetical cues.
CHAPTER II
ORGANIZATION IN FREE RECALL
Clustering in Free Recall
In the first demonstrations of organization in free
recall, it was noted that during recall of randomized lists
of words, clusters of related items appear. Bousfield
(1953) found that on a single trial, words belonging to the
same logical class, i.e., animals, names* professions, and
vegetables, tend to be recalled in related clusters even
though presented in random order. This tendency toward as
sociative clustering can be measured by means of the Ratio
of Repetition or RR (Cohen, Sakoda, and Bousfield, 195*0'
RR = (n - 1)
where r = the number of words which immediately follow
another member of the same category in recall, and
n = the total number of words recalled.
For experimenter-defined categories, then, RR indicates the
proportion of S's total recall which is recalled in clusters
of at least two items from the same category.
6
Clustering in recall has been shown to occur in
lists made up of stimulus-response pairs from the Kent-
Rosanoff Word Association Test (Jenkins and Russell, 1952)
as well as in categorized lists. Minimal clustering occurs
in lists made up of groups of synonyms (Cofer, 1959)• Bous
field, Cohen, and Whitmarsh (1958) have shown that the tend
ency to cluster varies according to the frequency with which
list items occur as associates to the names of the cate
gories present on the list: lists made up of high frequency
associates of the category names show much greater cluster
ing than do lists made up of low frequency associates of
the same names. These findings coupled with the more re
cent data on subjective organization have led to general
acceptance of the interpretation that clustering in recall
results from S's search for relations among list items which
he can use to aid his recall (Cofer, 1965; Mandler, 1967)•
The more obvious the relationships among list items, then,
the greater will be the clustering in recall.
Subjective Organization
When lists are constructed which do not contain ob
vious or systematic categorical or other relationships,
similarity in order of recall increases over trials. Tul-
ving's (1962a) measure of this subjective organization (SO)
begins with a recall matrix for an individual subject. The
rows and columns represent individual words in recall. For
a given block of at least two trials the number of times
each word follows each other word and the number of times
each word is first or last in recall is tabulated in the
appropriate cell. Each row total represents the number of
times a given word is recalled during the trial block and
each cell contains the number of times a given word is re
called in a certairTposition relative to the other words.
Then, for the case in which the number of trials is at
least as great as the number of words,
E n log n±J
so = --------
I] n^ log ni
where n. is the total for row i and n. . is the total for the
i ij
cell in the i^*1 row and column. Since n^ is the total
number of times a word is recalled and n.is the number of
ij
times it is recalled in the same position, SO is the ratio
of actual organization to maximum organization possible in
the particular S's recall.
Clearly, the number of trials in the block will af
fect the amount of SO recorded. With 2-trial blocks a word
has to be in the same position on adjacent trials in order
to add to measured organization, but with a 3-trial block
it can be recalled in the same position on only the first
and third trial and still add to SO.
9
Tulving (1962a) found, as mentioned earlier, a cor
relation of +.96 between mean log SO for 3-trlal blocks and
mean performance over trial blocks. More Important, how
ever, Is his finding of a positive correlation for individ
ual subjects between SO for l6-trial blocks and recall for
8-trial blocks; the correlation was + .45 for the first re
call block and +.78 for the second. Clearly, performance
and SO vary together, both for the individual and the group.
Tulving (1962b) has shown that performance is a function of
SO as well. The list used was a set of English nouns all
beginning with a different letter. After 3 trials half of
the Ss were told to organize the list by alphabet and the
other half were instructed simply to recall the items in any
order. The final number recalled by the alphabetical group
was increased significantly beyond that of the uninstructed
Ss. Tulving concludes that learning requires SO and that
the amount of learning is a function of the degree of SO
imposed.
The relation of SO to performance is'further eluci
dated by Tulving's (1964) trial-to-trial analysis of recall.
In this analysis, each S's recall on trial n is divided
into four components:
1. CC consists of the items recalled on both Trials
n-1 and n. This is regarded as an estimate of
intertrial retention.
10
2. NC consists of the items recalled on Trial n which
were not recalled on Trial n-1. This is regarded
as an estimate of intratrial retention.
3. CN consists of items recalled on Trial n-1 hut not
on Trial n. It is used as an estimate of inter
trial forgetting.
4. NN consists of items recalled on neither trial
and is thought of as an estimate of intratrial
forgetting.
In data for 32 Ss learning 22 words, over 22 trials
CC was found to be a monotonically increasing negatively ac
celerated curve paralleling the recall curve. NC, on the
other hand, was found to be nearly constant over trials,
decreasing only slightly on later trials. The correlation
of Log SO, 7-trial blocks, with performance ranged from
+.506 to +.843 over three 7-trial blocks. Log SO was also
highly correlated with mean CC, the correlations for the
same trial blocks ranging from +.584 to +.862. For the
first trial.block, Log SO and mean NC had a correlation of
-.075; mean CC and mean NC had a correlation of -.018; Mean
CC and Mean Performance had a correlation of +.948. In
other words, CC, performance, and SO are all highly related
and none of them are related to NC. Since NC, or intratrial
retention, remains almost constant over trials while inter
trial retention (CC) and SO grow together, Tulving suggests
that CC reflects the growth in size but not in number of
11
the subjective units into which S is organizing the list
items. If S is considered to add roughly one word to each
unit on each trial, the constant, NC, then reflects the
number of units he is using. If this is the case, NC
should be related to S's immediate memory capacity, whereas
CC should be related to long term storage.
The Category-Recall Relationship
The effect of the number of organizational units on
recall has been studied by manipulating the number of ex
perimenter-defined categories on a list to be recalled
either with or without cues. When list length is constant,
as number of categories increases the number of items per
category simultaneously decreases. Thus number and size of
categories have generally been confounded.
In two experiments with free (uncued) recall, Bous
field and Cohen (1956) used lists of 40 words drawn from
2, 4, or 8 categories. In one experiment recall improved
with increasing numbers -of categories but in the other, re
call decreased as categories increased. In another experi
ment with uncued recall, Cohen and Bousfield (1956) com
pared two 40-word lists, each with 4 categories but one a
list in which each of the 4 categories could be readily sub
divided into 2 subcategories, e.g., animals: feline and
canine. The dual-level list, which can be viewed as con
taining more categories, showed greater recall.
12
Dallett (1964) performed a series of experiments
with lists of varying numbers of categories. For 12 item
lists of 1, 2, 3, 4, or 6 categories, free recall increased
up to 4 categories but decreased for 6 categories. When
Ss were given the category names as cues during recall, re
call in the 6-category condition became superior to recall
in the 4-category condition. Dallett found for 24 word
lists of 2, 4, 6, 8, and 12 words, that uncued recall de
creased with increasing numbers of categories. His data
for cued recall suggest that in this and other cases of un
cued recall, when a larger number of categories has not led
to increased recall it is because of failure to remember
categories during recall.
Mathews (195^) studied cued recall of 24 item lists
drawn from 2, 4, and 6 categories. She found that recall
increased with number of categories. Thus, for cued recall
of lists containing experimenter-selected categories, the
data rather consistently show increased recall with in
creased number of categories.
For several of Mathew's categories, the general
tendency was reversed so that recall by the 2-category
group was actually superior to that of the 3-category
group. This led Mathews to examine the particular cate
gories involved and to speculate that the size of the parent
category from which list words were drawn is the important
variable. That is, for large parent-categories, short
13
list-categories (as in lists of greater numbers of catego
ries) are superior to long list-categories so that recall
increases with number of categories. But for small parent-
categories, long list-categories become superior to short
list-categories and recall is better with fewer categories.
Mathews speculated that a differential interference effect
is responsible. With a.large parent-category there are
more words in the category which aren't on the list which
can then interfere with recall of list words. With a
smaller parent-category there is less interference/ particu
larly with a list-category which approaches the size of the
parent-category so that most members of the parent-category
are on the list. Mathews' hypothesis suggests that the ef
fect of category length may depend on the category, and it
is also directly relevant to the present study. Since the
parent-categories in the present experiment are of differ
ent sizes, differential interference effects must be con
sidered as an alternate possible reason for any differences
noted in recall between the two types of categories.
The category-recall relationship has been studied
for subject-imposed categories, as opposed to experimenter-
imposed categories, by means of Mandler and Pearlstone's
(1966) sorting task described earlier. In a series of ex
periments with this categorizing task, Mandler (1967) found
that the linear relationship between categories and recall
holds under the following conditions:
1. The correlation holds when total time to criterion
is partialled out as well as when number .of unpaced
sorting trials is partialled out. Consequently,
the relationship can not be due to Ss' taking
longer to become consistent when they use more
categories.
2. Ss instructed (a) to learn the words, (b) to cate
gorize the words, or (c) to learn and categorize
the words show extremely similar recall after 5
trials, and they recall significantly more words
than Ss instructed neither to recall nor to cate
gorize. In both groups which categorized, the
correlation held; thus the correlation exists
under intentional as well as incidental learning
conditions.
3. When Ss are instructed on how many categories
they should use (e.g., "divide the words into 4
categories of your own choosing") the linear
category-recall relationship is still found. Con
sequently, the relationship is not merely correla
tional.
The results of these experiments, in which number
of categories has been varied by either the experimenter or
the subject, clearly indicate that for either experlmenter-
or subject-defined categories recall increases with number
of categories.
The Power of a Retrieval Cue
In order to study the number of items retrieved per
cue or unit., Cohen (1966) analyzed data from one free recall
trial with an experimenter-categorized word list in terms
of number of categories from which at least one item from
the category was recalled. He found that number of items
per recalled category was almost invariant under a number of
conditions, including list length, presentation rate, and
sex differences. These variables affected how many catego
ries were recalled, but not mean number of words recalled
per recalled category.
Tulving and Pearlstone (1966) used one trial with
an experimenter-categorized list and gave some Ss category
names as cues during recall, while other Ss were given
standard free recall. They showed that when list length
varies (from 12 to 48 words) and number of words per cate
gory varies (l, 2, and 4 items) cued recall is higher than
free recall, but the difference is due to cued recallers
remembering more categories, not more words within cate
gories. That is, cued recallers recall at least one word
from more categories than do free recallers, but mean num
ber of words recalled per category from which at least one
item is recalled does not vary with recall conditions.
Using subjective categories, Dong and Kintsch (1967)
got similar results. Ss first sorted words into categories
of their own choosing and then recalled them under either
16
free or cued recall conditions. Although cued Ss recalled
more total words and also more categories, mean words per
category recalled did not differ for the two groups.
Earhard (1967), as cited earlier, found that recall
decreased as number of items per alphabetical cue increased
and that free recall was superior to cued recall with more
than 6 items per cue. Finally, Earhard found that lists
made up of smaller categories were learned significantly
faster than lists made up of larger categories.
Mandler (1967) tentatively applies the results on
the category-recall relationship and on the power of a re
trieval cue to free recall as follows. Early in learning
S sets up a category scheme consisting of a number of cate
gories which does not exceed his immediate memory span,
usually around 5. He then "fills up" the categories by
adding roughly one word per trial to each category. The
number of categories he uses is reflected, then, by NC.
Each category is itself limited in the number of words it
can hold, the limit again a function of S's immediate mem
ory capacity. If the list is too long to be accommodated
in this fashion, S may subdivide some or all categories.
Each subcategory can then hold as many items as S can handle
in immediate memory. Final recall is then a function of
the number of units, categories or subcategories, which S
uses. Thus to Mandler, number of units is the basic organ
izational variable in free recall.
17
Differences among Cues In Retrieval Power
The category-recall relationship is clearly predict
able from the following: (l) the more categories there are
on a list of given length, the smaller will be the average
number of words per category, and (2) on a given trial, a
list with a smaller number of (larger) categories has been
learned less well than a list of a larger number of (small
er) categories, because of the difference found by Earhard
in rate of learning. In short, basic to the category-recall
relationship is the limit in the number of items retriev
able, on a given trial, per cue. If different cues, or
groupings, differ in the number of words which they allow
S to retrieve, then the effect of number of categories can
be mitigated by variations among categories. For example,
if there can be highly effective groupings or categories
from which a large proportion of words can be recalled, and
very ineffective groupings from which few words can be re
called, then a small number of effective categories would
be capable of yielding a higher level of recall than a
larger number of ineffective categories. If some cues can
recall more words than others, the limit on the number of
items retrievable by a cue may depend on the particular cue
used, rather than being strictly a function of the immedi
ate memory capacity. Indeed, if there are sets of catego
ries or cues which vary in their effectiveness as retrieval
aids, then free recall performance may vary not only accord-
18
ing to how many units a subject uses but also according to
the effectiveness of the particular units which he uses in
order to organize his recall. In particular, it is to be
expected that a cue for which the relevant list words are
relatively high frequency associates will be capable of re
trieving more words than will a cue for which the relevant
words are relatively low frequency associates.
In order to address the question of whether the
limit on the power of a cue varies with the cue, a list of
words was constructed for which two sets of six categories
each were relevant: the words could be categorized either
according to their initial letters or according to their
membership in a conceptual category. Thus the words could
be cued by either initial letters or conceptual category
names. These two sets of categories were chosen to repre
sent, respectively, relatively low and relatively high fre
quency of association between the cues and their relevant
words. The following were the two major experimental hy
potheses:
1. It was predicted that recall by Ss using the
conceptual cues would be superior to recall by Ss using the
alphabetical cues.
2. It was hypothesized that the number of words
retrievable per cue would vary with the cue. If so, for
cues relevant to increasing numbers of items, Ss using con
ceptual cues should show less reduction In proportional
19
recall than Ss using alphabetical cues. That is, the dif
ference in overall recall between two groups of Ss using
the two sets of cues should be due primarily to greater re
call in the larger categories by the group using the concep
tual cues. Thus it was predicted that there would be a
significant Groups by Category Lengths interaction in pro
portional recall within categories.
In addition to the two major hypotheses, the follow
ing five subsidiary questions were investigated:
3. It was hypothesized that the number of items
retrievable per unit or cue is an important organizational
variable contributing to variability in free recall. Since
this hypothesized source of variance was controlled for Ss
given cues but not for free recall Ss, it was predicted
that the variances of the recall of groups given cues would
be smaller than the variance of a free recall comparison
group.
4. If SO is necessary to learning, a set of cues
which facilitates learning can be expected to facilitate SO
as well. Thus It was predicted that SO within categories
would be higher for Ss using conceptual cues than for Ss
using alphabetical cues.
5. The function of subjective organization is
presumably to overcome the limitation of the Immediate mem
ory span. Consequently, categories which do not exceed this
20
capacity in length should be less likely to require grouping
of items within them. Thus it was expected that SO would
not increase over trials in the smaller categories but that
it would increase in the larger ones.
6. Words drawn from a large parent category have
more potential sources of interference in recall than do
words drawn from smaller parent categories. In the present
experiment, the parent alphabetical categories are larger
than the parent conceptual categories, so that Mathews'
position, discussed earlier, implies greater interference
from category members not on the list for Ss using alphabet
ical cues than for Ss using conceptual cues. Her position,
then, predicts more categorical intrusions for the former
than for the latter Ss. However, conceptual cues are viewed
here as effective because category members are relatively
high frequency responses to them. Conceptual category in
trusions, then, would be expected to have a higher probabil
ity of occurrence than alphabetical category intrusions.
Consequently it was expected that conceptual category intru
sions would occur more frequently in the recall of concep
tually cued Ss than would alphabetical category intrusions
in the recall of alphabetically cued Ss.
7. Finally, since NC has been linked with number
of units and since number of units is thought to be deter
mined by immediate memory span, a digit span test was in
21
eluded in the experiment in order to correlate score on
this test, as an estimate of immediate memory capacity, with
NC. A positive correlation was expected.
CHAPTER III
METHOD
Subjects
Seventy-five undergraduate students at the Univer
sity of Southern California, all native speakers of English,
were conscripted from a variety of undergraduate liberal
arts courses and were paid to serve as Ss. Ss ranged in
age from 16 to 24 years, with all but 2 Ss falling between
18 and 22 years of age, inclusive. The mean age was 19.2
years. At each experimental session, lasting 60 minutes,
Ss were randomly divided among the three experimental
groups, with the restriction that the sex ratio in each
group was kept approximately constant, being 15 males and
10 females in each of the cued groups and 16 males and 9
females in the free recall group. Ss were run in groups of
from 2 to 17.
Apparatus
General instructions to Ss were recorded by E on a
Wollensak tape recorder. The tape recording was used to
deliver instructions and to control the recall interval.
Stimuli were typewritten on transparent paper, mounted as
22
23
slides on Kodak 36 mm. Ready-Mounts, and projected on the
wall of the experimental room by a Sawyer Automatic 707AQ
slide projector. The rate of presentation was controlled
by an Industrial Timer Corporation clock which caused the
remote control circuit on the slide projector to change the
slide every two seconds. Wing-desk chairs were arranged in
rows in the experimental room. Each row in use in a given
session contained Ss from each experimental group, so that
average distance from the stimuli being presented and from
the tape recorder was kept constant for the three groups.
Stimulus List
Forty-eight words were chosen from 6 categories in
the taxonomic norms for category associates (Cohen, Bous-
field, and Whitmarsh, 1957)• The words chosen all had one
of 6 initial letters, so that each word was a member of both
a conceptual and an alphabetical category. That is, each
word was a member of (a) the conceptual category from which
it was chosen, and (b) the alphabetical category defined by
its initial letter. Both the conceptual and alphabetical
categories varied in length, having 2, 4, 6, 8, 12, or 16
members. The alphabetical and conceptual categories of a
given size had words in common, but they were not synony
mous. For example, the l6-word alphabetical category, "c,"
contained at least one item from each of the 6 conceptual
categories, as did the 12-word alphabetical category. The
24
relationship between the two sets of categories and the
words themselves are specified in Appendix A.
Since the frequency with which a category’s members
occur as responses to the category name was considered an
important variable in determining amount of recall within
that category, it was important to control this frequency
in comparing two sets of categories for recall. For ex
ample, if the large conceptual categories had more high
frequency members than did the large alphabetical catego
ries, then Ss cued with conceptual category names would be
expected to recall more words from the large categories
than Ss given alphabetical cues. This would spuriously
cause the shape of the curve of proportional recall as a
function of category size to differ between the two groups.
It was necessary that large and small alphabetical catego
ries have the same number of high frequency words as did,
respectively, large and small conceptual categories. Cate
gories of from 2 to 6 words were defined as "small" and
categories of 8 to 16 words as "large."
Thorndike-Lorge general count was used to approxi
mate the frequency with which a word occurs as a response
to its first letter. That is, it was assumed that the fre
quency with which "lynx" would occur as a response to "1"
would be proportional to the frequency with which "lynx" is
used in the language. For the words chosen, the number of
members of small and large alphabetical categories which
' 25'
were above and below the median Thorndike-Lorge frequency '
I
for the list was the same as the number of members of small
and large conceptual categories which were above and below
the median taxonomic frequency, as shown in Table 1. Median
Thorndike-Lorge general count frequency for the list was
26.5} median taxonomic frequency was 8.5. Since word length
was expected to affect recall, the mean length of words in
small and large alphabetical and conceptual categories was
also examined, and appeared sufficiently close, as seen in
Table 2.
Design and Procedure
Two main experimental groups were used, differing
in which cues they received: alphabetical cues (A), and
conceptual cues (C). In addition a noncued, free recall
comparison group (F) was used. All Ss learned the same
words. For both sets of cues the stimulus list consisted
of 6 categories, one each of length 2, 4, 6, 8, 12, and 16
words. Thus, number and size of categories was the same for
both cued groups, and the experimental groups differed only ;
in the conditions of recall. Following the main experiment,
i
Ss performed a digit span task which was identical for all i
i
groups. I
Learning instructions and recall sheets were pre- |
!
sented to Ss in a booklet. Instructions to turn pages, !
I
read learning instructions, prepare for presentation of
TABLE 1
NUMBER OF WORDS IN SMALL AND LARGE CATEGORIES ABOVE AND BELOW
MEDIAN TAXONOMIC AND THORNDIKE-LORGE FREQUENCY FOR THE LIST
Category Size
’ THORNDIKE-LORGE F
Below Median Above Median
Small
5 7
Large
19 17
Category Size
TAXONOMIC F
Below Median Above Median
Small
5 7
Large
19 17
TABLE 2
MEAN LENGTH OF WORDS IN ALL CATEGORIES
CATEGORY TYPE CATEGORY SIZE
Small
(2-6 words)
Larce
( 8-16 words)
Conceptual 5.66 letters 5 .6 3 letters
Alphabetical 5•75 letters 5.61 letters
istlmuli., or recall words were on tape. First Ss were in
structed by tape to read their written instructions, which
are reproduced fully in Appendix B. Written instructions
for Group C informed them that certain words descriptive of
words on the list would be printed on the recall sheet as
cues to aid their recall. They were told to write each re
called word on a line underneath the cue to which it be
longed. Group A was informed that the initial letters of
the stimulus words would be printed on the recall sheet as
cues to aid their recall. They were told to write each re
called word on a line underneath the cue to which it be
longed. Group F was told to learn the words and to write
them, one on a line, in the order in which they recalled
them. Recall sheets for each cued group contained the cues
appropriate to that group and underneath each cue the cor
rect number of lines for that category. Free recall sheets
contained two columns of 24 lines each. Examples of the re
call sheets are given in Appendix C. Before each of the 12
presentations, all Ss turned to a blank page, so that the
cued groups could not see the cues during list presentation.
Twelve different orders of presentation were pre
pared, one for each trial. Straightforward randomization
of the words was avoided because 12 randomizations might
have presented adjacent to each other more conceptual than
alphabetical category mates, or vice versa. Consequently,
the words were put into two basic orders, one relative to
28
each category system. The conceptual order listed the 2- !
i
I
word conceptual category, then the 4-, 6-, 8-, 12-, and 16- |
word categories. The alphabetical order listed the alpha
betical categories, also In order of increasing size.
Groups of words in common between two categories were kept
in the same order for both listings, e.g., "banker," "butch
er," and "boss" were in that order in both "b" and "occu
pation." Then, six randomizations were applied separately
to the two basic orders, yielding 12 word orders which
placed the same number of category mates adjacent to each
other for either set of categories. If the two word orders =
yielded by applying the first randomization to the concep
tual and then the alphabetical basic order are designated
Cl and Al, respectively, then the sequence in which the
various word orders were arranged over trials was Cl, Al,
A2, C2, A3, C3j C4, a4, C5, A5, A6, C6, so that word orders
derived from the two basic orders were counterbalanced over i
i
trials.
i
Ss were given 2-1 minutes for recall on each trial. :
H i
i
The recall periods were started and stopped by taped in
structions, and the length of the recall interval was thus j
controlled by the pre-recorded tape. Since the order of thej
j
categories on the recall sheet for cued Ss might have af- j
fected number recalled within categories, a Latin Square for;
)
i
orders of categories of different sizes was prepared, as in
Appendix D. Each cued S was given two trials with each of
29
these six orders on the recall sheet. So that growth of
recall over trials could not be affected in these groups by
the order of the cues, a Latin Square for orders of orders
was prepared. On any one trial, then, each cued group con
tained Ss whose cues were arranged in each of the orders
listed in Appendix D.
Digit Span
The digit span test followed immediately after the
word learning task. The method used followed that of Jensen
(1964) as closely as possible. Instructions and stimuli
were recorded by E and presented by tape recorder. Ss were
instructed, as in Appendix E, that a tap would sound immedi
ately before a digit series was read. Similarly, the end
of the series was signalled by a tap, after which Ss wrote
down the digits in the order in which they had been pre
sented. Each series was written in its own row of boxes on
the recall sheet, beginning at the left.
There were four sets of seven series of digits.
Each of the four sets consisted of series which ranged from
three to nine digits in length. The recall sheet for each
set contained seven rows of nine boxes, each box being ^
inch square. A recall sheet is reproduced in Appendix P.
Digit stimuli were picked from a table of random
numbers, with the restriction that no digit appeared more
than once in any one series. The stimuli were presented at
the rate of one digit per second. The recall interval
30
allowed for a series depended on the length of the series.
This recall interval consisted of 4 seconds plus 1 second
for each digit in the series. For example, a three digit
series had a 7 second recall period and a four digit series
had an 8 second recall period.
CHAPTER IV
RESULTS
Results will be reported as significant if p<.05.
Hypothesis 1
The first of the two major hypotheses was that con
ceptual cues would lead to higher recall than would alpha
betical cues. Indeed, Group C was superior to both Groups
P and A in words recalled throughout learning, as seen in
Figure 1. An analysis of variance performed on these data
and summarized in Table 3 revealed that both the effects of
Groups and Trials are significant, as is their interaction.
That is, the three groups differ in amount of recall, learn
ing is occurring, and the groups differ somehow in rate of
learning. In order to discover the locus of the difference
among the groups in words recalled, Group C was compared to
Group F, and likewise Group F was compared to Group A.
These analyses of variance, summarized in Tables 4 and 5,
reveal that Group C recalled significantly more words than
Group F, but the difference between Groups F and A fell
short of significance, with ,10>p>.05. In neither case
was the Groups by Trials interaction significant. However,
31
32
§
10
, - l t
5-6 7-8 1-2 9-10 11-12
Two-Trial Blocks
Figure 1. Mean words recalled as a function of
2-trial "blocks, Groups A, C, and F.
33
TABLE 3
ANALYSIS OF VARIANCE OF NUMBER OF WORDS RECALLED BY ALL THREE GROUPS
Source MS df F
Total 350.0825 449
Between Ss 986.3927 74
Groups 8399.1822 2 IO.762***
Error (G) 780.4829 72
Within Ss 224.5173
375
Trials 14119.8542
5
399.341***
-Groups "by Trials 86.5902 10 2.449**
Error (T)
35.3579
360
**p < .01
***p < .001
TABLE I f
ANALYSIS OF VARIANCE OF NUMBER OF WORDS recalled BY GROUPS C AND F
Source MS df F
Total
357.3571 299
Between Ss 966.1113
49
Groups 5843.2533 1
6.759*
Error (G) 864.5042 48
Within Ss 238.0413 250
Trials 9940.6693
5
254.488***
Groups by Trials 86.4453
5
2.213
Error (T) 39.0615 240
*p < .05
***p> < .001
34
TABLE 5
ANALYSIS OF VARIANCE OF NUMBER OF WORDS RECALLED BY GROUPS F AND A
Source
MS df
F
Total 319.9487
299
Between Ss 896.5510
49
Groups 2748.2133 1 3.203
Error (G)
857.9747
48
Within Ss 206.9347 250
Trials
8371.7173 5 207.441“
Groups by Trials
37.8773 5 • 939
Error (T) 40.3571 240
***p < .001
35
for'Groups C and F this interaction has .06>p>.05, whereas
for Groups P and A the interaction has p>.50. This sug
gests that the significant difference among the three groups
in rate of learning noted in Table 3 is largely located be
tween Groups C and P.
It is reasonable to expect that Ss free to choose
between the two sets of categories would tend to use the
superior set more than the inferior one. In order to inves
tigate this question, Group P's recall was examined and HR
was calculated separately for the two sets of categories.
Mean RR based on the alphabetical categories and mean RR
based on the conceptual categories for 2-trial blocks in
Group F are shown in Figure 2. Mean conceptual FIR is higher
than mean alphabetical RR for each trial block and alphabet
ical RR appears not to increase over trials. Consequently
a one-way analysis of variance was performed for each type
of RR, as reported in Table 6. Indeed, conceptual RR In
creases significantly as a function of trials, but alphabet
ical RR does not. Evidently, there is no significant tend
ency for the free recall Ss to use an alphabetical organiza
tion In recall, but there is a pronounced tendency for these
Ss to use instead the more effective conceptual categories.
Hypothesis 2
The second major hypothesis was that of the two
types of categories used, the conceptual categories would
show less reduction in proportional recall within a category
Mean Ratio o f Repetition
36
• 55 -
.25
20
10
• Conceptual RR
A Alphabetical RR .05
5-6 7-8 11-12 1-2 9-10
Two-Trial Blocks
Figure 2. Mean RR based on alphabetical and
conceptual categories, Group F.
37
TABLE 6 .
ANALYSES OF VARIANCE OF RATIO OF REPETITION BASED ON CONCEPTUAL
AND ON ALPHABETICAL CATEGORIES, GROUP F
RR Source MS df F
Alphabetical Total .0170 1 1 4 - 9
Trials .0093
5 .719
Error (T) .0129 120
Conceptual Total .1366 149
Trials .91403
5 28.963***
Error (T)
.0325
120
***p < .001
as category size increased. Figures 3 through 6 depict mean
proportional recall over 3-trial blocks from categories of
various lengths for which at least one item is recalled.
That is, in arriving at an S’s mean proportional recall for
a given trial block, only those trials on which he recalled
at least one word from a category were considered. Figure
7 depicts mean proportional recall over all 12 trials, from
all categories of each size, including those for which no
word is recalled on a given trial. For Group F, the concep
tual categories were used as the basis for calculating pro
portional recall within categories, since Group F Ss used
primarily the conceptual categories in organizing their re
call. The data for Group F are included merely for purposes
of visual comparison. Statistical' comparisons of the pro
portional recall of Groups C and A for each block of three
trials, omitting zero recall, are summarized in Tables 7
through 10. Comparison of proportional recall of Groups C
and A averaged over all twelve trials, and including zero
recall, is summarized in Table 11. In all these analyses,
over 3- or 12-trial blocks, and either including or exclud
ing zero recall, the Groups by Category Lengths interaction
is significant, as predicted. These results are consistent
with the hypothesis, and inspection of Figures 3-7 suggests
that these significant interactions are due to a difference
in rate of decrease in proportional recall for Increasing
category lengths, particularly for Trials 7-9 and Trials
Mean Proportional Recall
1.00
39
• 90
.80
.70
.60
.50
.Uo
.30
<
• F
0 L J I 1 L _ J
16 I f 6 8 12
Category Length in Words
Figure 3. Mean proportional recall for trials 1~3
from categories with nonzero recall.
Mean Proportional Recall
1.00
40
.90
.80
.70
.60
• 50
,4o
.50
1
2 4 6 8 12
Category Length in Words
Figure 4. Mean proportional recall for trials 4-6
from categories with nonzero recall.
16
Mean Proportional Recall
1.00
4l;
• 90
.80
• 70
.60
• 50
. 4 o
.3 0
I
A C
• F
■ A
...I_ _ 1 1 1
2 4 6 8 12
Category Length in Words
Figure 5. Mean proportional recall for trials 7~9
from categories with nonzero recall.
16
Proportional Recall
1.00
.90
.80
.70
.60
.50
.40
42
.30
<
0
I
A C
• F
■ A
1 1 1 1
Figure 6.
4 6 8 12
Category Length in Words
Mean proportional recall for trials 10-12
from categories with nonzero recall.
16
A
m a
2 4 6 8 12 16
Category Length in Words
Figure 7- Mean proportional recall for trials 1-12
from all categories of a given length.
44
TABLE 7
ANALYSIS OP VARIANCE OF MEAN NONZERO PROPORTIONAL RECALL FOR
TRIALS 1-3, GROUPS C AND A
Source MS
df F
Total .0389
293
Between Ss .0787 48
Groups 1.3183 1 2 5. 186***
Error (G) .0523
47
Within Ss .0311 245
Category Length
.9571 5
85.146***
Grps by Cat. Length
.0399 5
3. 550**
Error (C.L.) .0112
235
**p < .01
***p < .001
TABLE 8
ANALYSIS OF VARIANCE OF MEAN NONZERO PROPORTIONAL RECALL FOR
TRIALS 4-6, GROUPS C AND A
Source MS df F '
Total .0413
299
Between Ss .1183
49
Groups 2.2716 1 30.922***
Error (G)
.0735
48
Within Ss .0262 250
Category Lengths
.7297 5 ■
6 4. 196***
Grps by Cat. Length .0341
5
3. 000*
Linear Interaction .0044 1 .386
Error (C.L.) .0114 240
*p < .05
***p < .001
45
TABLE 9
ANALYSIS OF VARIANCE OF MEAN NONZERO PROPORTIONAL RECALL FOR
TRIALS 7-9, GROUPS C AND A
Source MS df F
Total .0382
299
Between Ss .1121
49
Groups I.9W 1 26.296***
Error (G)
.0739
48
Within Ss
.0237
250
Category Lengths .6122
5 6 2. 289'***
Grps by Cat. Length .0944
5
10.114***
Linear Interaction .1288 1 13. 1 1 4- 3***
Error (C.L.) .OO98 240
***p < .001
TABLE 10
ANALYSIS OF VARIANCE OF MEAN NONZERO PROPORTIONAL RECALL FOR
TRIALS 10-12, GROUPS C AND A
Source MS df F
Total .0312
299
Between Ss .1013
49
Groups 1.4137 1 1 9. 108***
Error (G) .O7I 4O 48
Within Ss
.0175
250
Category Length .4027
5
i + 6.384***
Grps by Cat. Length .0542
5
6.246***
Linear Interaction .1489 1 17.115***
Error (C.L.) .0087 240
***p < .001
46
TABLE 11
ANALYSIS OF VARIANCE OF MEAN PROPORTIONAL RECALL, INCLUDING
ZERO RECALL, FOR TRIALS 1-12, GROUPS C AND A
Source MS
df
F
Total .0302
299
Between Ss .1052
49
Groups I.785O 1 25.4o8***
Error (G) .0703 48
Within Ss .0154 250
Category Length .5002
5
95.102***
Grps by Cat. Length .0194
5 3.689**
Linear Interaction .0015 1 .283
Error (C.L.)
.0053
240
**p < .01
***p < .001
47*
10-12, as shown in Figures 5 and 6. In order to test the I
hypothesis that the significant Groups by Category Lengths
interactions were due to such a difference in rate, the
linear components of the interactions were tested for sig
nificance. The hypothesis tested was that the two lines
describing the proportional recall of Groups A and C for
categories of different sizes differed significantly in
their slopes.
This analysis was not performed for Trials 1-3, be
cause on this trial block Group A had 24 Ss and Group C had
25 Ss. One S had had to be excluded from Group A's data
for this trial block because he had failed to recall any
word from one of the categories on all three trials. Veld-
man’s program Anovar, used in performing the groups by
trials analysis of variance on these data, allows unequal
numbers of subjects in multiple groups (Veldman, 1967).
However, in partitioning the interaction into linear and
nonlinear components, the problem of unequal n's in the two ;
!
1
groups was avoided by omitting this trial block from the j
1
1
analysis. The results of the tests for the remaining three ;
3-trial blocks and for Trials 1-12 are given in Tables 8-11.j
|
The results indicate that for Trials 7-9 and 10-12, the lin-j
ear component of the interaction is significant. In view j
1
____ j
of the nonsignificance of the linear component for Trials
4-6 and Trials 1-12, it appears.probable that early in j
learning the groups do not differ in the rate at which pro-
48
portional recall decreases as category length increases,
but later in learning these rates do differ significantly.
Group C1s mean proportional recall data for 3-trial
blocks were subjected to one-way analyses of variance in
order to discover whether the decrease in proportional re
call for larger categories is significant within this
group. These analyses are summarized in0Table 12. For
each of the four blocks of trials, the effect of category
length on proportional recall is significant.
Examination of Figure 7 suggests that Group F is
superior to Group A in recall from longer categories but
not from shorter categories. It is reasonable to expect
that Group F would not be superior to Group A in recall
within the shorter categories, because Group F lacks cues
and the shorter categories are likely not to be recalled at
all when they are not cued. These two possibilities were
tested. Total number of words recalled over twelve trials
from the three shorter categories and 12-trial total recall
from the three larger categories were separately compared
for Groups F and A. As seen in Table 13, the groups did not
differ in their recall from the shorter categories, but
Group F recalled significantly more words than Group A from
the longer categories.
Next the possibility was examined that Group F had
a higher number of words in nonrecalled categories than the
cued groups, i.e., that categories from which no word is
49
TABLE 12
ANALYSES OF VARIANCE OF MEAN PROPORTIONAL NONZERO RECALL, FOR
3-TRIAL BLOCKS, FROM CATEGORIES OF DIFFERENT LENGTHS, GROUP C
Trials Source MS df F
1-3
Total
.0399
1 * 4 - 9
Category Length .6203
5
6 8. 221***
Error (C.L.) .0091 120
* 4 —6 Total .0302 1 * 4 - 9
Category Length • 3528
5
4 3. 078***
Error (C.L.) .0082 120
7-9
Total .020*4-
1 * 4- 9
Category Length .1711
5
2 8. 353***
Error (C.L.) .0060 120
10-12 Total .01*44 1 * 4 -9
Category Length .0970
5
1 9. 011***
Error (C.L.) .0051 120
***p < .001
50
TABLE 13
ANALYSES OF VARIANCE OF TOTAL RECALL, TRIALS 1-12,
FROM SHORT AND LONG CATEGORIES, GROUPS F AND A
Category
Size Source MS df F
2-, 4—, and
6-vord
Categories
Total
Groups
Error (G)
Group A Mean = 98*32
Group F Mean = 97*40
362.4086
1 0. 5 8 0 0•
369.7383
49
1
48
.029
8-, 12-, and
l6-word
Categories
Total
Groups
Error (G)
Group A Mean = 236.72
Group F Mean = 273*96
3344.2290
17335*2200
3052.7500
49
1
48
5. 679*
*p < .05
51
recalled account for more nonrecalled words in Group F than
in Groups C and A. A median test was performed on the num
ber of words in nonrecalled categories by Group P and the
cued groups, yielding significant results, as listed in
Table 14. Thus, the free recall group omitted more words ’
from recall as a result of nonrecall of a whole category
than did the cued groups.
Hypothesis 3
According to this hypothesis the variances of the
recall of the cued groups were expected to be smaller than
that of the free recall group. These variances were com-
, pared for three measures of recall: total recall for Trials
1-6, total recall for Trials 7-12, and total recall for
Trials 1-12. The variances are listed in Table 15- All
differences are in the predicted direction, both cued
groups displaying less variation in recall than the free
recall group for both 6-trial blocks and for the 12-trial
block. However, only one of the six comparisons was signif
icant: for Trials 7-12, Group C was significantly less vari
able in words recalled than was Group F.
Hypothesis 4
According to this hypothesis, Group C was expected
to develop a higher level of subjective organization within
categories than Group A. In testing this hypothesis, two
comparisons were made. In the first, SO was calculated
52
TA BLE l4
MEDIAN TEST FOR NUMBER OF WORDS IN NONRECALLED CATEGORIES,
GROUPS F AND A
Number of Words in Nonrecalled Categories
Group 2 and Below 4 and Above Total
F
5
20
25
A and C 29 21 50
Total 34 k l
75
x2 = 9.71**
**p < .01
TABLE 15
VARIANCES OF WORDS RECALLED
Trials Group F Group A Group C
1-6 1001.912 643.840 959.822
7-12 2762.974 1448.944 III8.367*
1-12 6621.077 3674.663
3755.033
* S ig n ific a n tly sm aller than the comparable variance fo r Group F.
within all but the 2-word category for blocks of two trials.
For a given 2-trial block, a Total SO measure, was derived by
adding together the SO for each of the 5 categories for that
trial block. Total SO for 2-trial blocks for the two cued
groups is depicted in Figure 8. A groups by trials analysis
of variance, summarized in Table 16, was then performed.
The effect of trials was significant, but not the difference
between the groups. SO for 2-trial blocks increased over
trials, but the tendency noticeable in Figure 8 for Group C
to organize more than Group A-was not significant.
As a second test of this hypothesis, SO for 6-trial
blocks was calculated within all but the 2- and 4-word cat
egories. The 4-word category was omitted from the analysis
because when the number of trials, in this case 6, exceeds
the number of words, in this case 4, the calculation of SO
becomes more difficult. Total SO for 6-trial blocks was
derived as before, by adding SO for each category involved
for a trial block. Figure 9 displays these data. The anal
ysis of variance summarized in Table 17 indicated that the
effect of trials was significant, and the difference be
tween the groups was significant, but there was no signifi
cant interaction. Thus, SO for 6-trial blocks increased
with learning at the same rate for Groups C and A, but
Group C showed a higher level of SO than did Group A.
Total Subjective Organization
1.60
1.50
1.1*0
l.JO
1.20
1.10
1.00
o L
■ A
1 1
1-2 3-1* 5-6 7-8
Two-Trial Blocks
I
J
9-10 11-12
Figure 8. Mean Total SO within 4-16 word
categories for 2-trial blocks.
55
TABLE 16
ANALYSIS OP VARIANCE OF TOTAL SO'FOR 2-TRIAL BLOCKS, GROUPS C AND A
Source
MS
df
F
Total
.3817 299
Between Ss .6288 49
Groups I.656I 1 2.727
Error (G) .6073 48
Within Ss
• 3555
250
Trials 1.1249
5
3.524**
Groups by Trials
.2185 5 .685
Error (T) .3192 240
**p < .01
Total Subjective O rganization
1.50
l.ltO
1.30
1.20
1.10
1.00
I
1
A C
■ A
1-6 7-12
S ix -T r ia l Blocks
Figure 9* Mean Total SO within 6-16
word categories for
6-trial blocks.
57
TABLE 17
ANALYSIS OF VARIANCE OF TOTAL SO FOR 6-TRIAL BLOCKS, GROUPS C AND A
Source MS df F
Total .07IH
99
Between Ss .0709 k9
Groups . 3^27 1 5. 251*
Error (G)
.0653
bd
Within Ss .0772 50
Trials 1.5988 1 34.OI16***
Groups by Trials .0060 1 .128
Error (T) .0^70 kd
*p < .05
***p < .001
Hypothesis 5
It was expected that SO would not increase within
the smaller categories but that it would increase within
larger categories. Both SO calculated for 2-trial blocks
and for 6-trial blocks were separately examined. Level of
SO within categories of different sizes was not directly
compared because small categories are expected to show more
organization due simply to chance than large categories.
Hence any comparison of quantity of SO in different sized
categories would be meaningless. Instead, the effect of
trials on SO within each category size was separately
assessed by means of a one-way analysis of variance. It
was predicted that SO within the small categories of 4 and
6 words would not increase over trials but that SO would
increase within the larger categories, particularly the 12-
and l6-word ones. The position of the 8-word category is
not clear, since 8 items may or may not fall within an S’s
immediate memory span. The SO data for 2-trial blocks for
Group C are depicted in Figure 10 and those for Group A in
Figure 11. The five analyses of variance for these data
for Group C are jointly summarized in Table 18 and those
for Group A, in Table 19. Inspection of the graphs reveals
no general tendency for SO to increase with trials, but for
several of the categories there appears to be an increase,
notably the 6-word category in Group C and the 4-, 6-, and
12-word categories in Group A. Of the ten analyses of
Subjective Organization, 2-Trial Blocks
59
• 55
.50
45
.bo -
• 35
• 30
.25
.20
• 15
.10
• 05
▼ b—-word category
• 8-word category
# 6-word category
>12-word category
A 36-word category
1
1-2 3-4 5-6 7-8
Two-Trial Blocks
J
9-10 11-12
Figure 10. Mean SO, 2-trial blocks,
within categories, Group C,
Subjective Organization, 2-Trial Blocks
60 !
.55
.50
.to
•35
.30
.25
.20
.15
.10
.05
▼ toword category
# 8-word category
# 6-word category
■ 12-word category
A l6-word category
1 1 1
1-2 3-4 5-6 7-8 9-10 11-12
Two-Trial Blocks
Figure 11. Mean SO, 2-trial blocks,
within categories, Group A.
6l
TABLE 18
ANALYSES OF VARIANCE OF SO FOR 2-TRIAL BLOCKS WITHIN CATEGORIES OF
FROM 4 TO 16 WORDS, GROUP C
Category
Size Source MS df F
Total .1181 " ll *9
Trials
.1985 5 1.719
1 * Words Linear . 17l f 6 1
1.513
Nonlinear .201*5 ' 1 *
1.772
Error (T) . 1151* 120
Total .O656
11*9
Trials .081*1*
5
I.36O
6 Words Linear
.2573
1 l*.ll* 3*
Nonlinear .01*12 1 *
.663
Error (T) .0621 120
Total .01*86 ll +9
Trials .0576
5
1.388
8 Words Linear .031*7 1 .836
Nonlinear .0633 1 *
1.525
Error (T) .01*15 120
Total
.0373
ll *9
Trials .0191*
c 5
• 577
12 Words Linear .0131 1
.390
Nonlinear .0210 l * .625
Error (T) .0336 120
-
Total .0285
ll *9
Trials .0273
5
1.016
16 Words Linear .0201* 1 .761
Nonlinear .0290 1 * 1.082
Error (T) .0268 120
< .05
62
TABLE 19
ANALYSES OF VARIANCE OF SO FOR 2-TRIAL BLOCKS WITHIN CATEGORIES OF
FEQM 4 TO 16 WORDS, GROUP A
Category
Size Source
MS df F
Total .1088 149
Trials .1311
5
1.230
4 Words Linear .2708 1 2.540
Nonlinear .0962 4 .902
Error (T) .1066 120
Total
.0737
149
Trials .0863
5 1.336
6 Words Linear
.3727
1
5. 769*
Nonlinear .0147 4 .227_ _
Error (T) .0646 120
Total .0638 149
Trials .0744 •
5
1.210
8 Words Linear .0189 1 .308
Nonlinear .0883 4 1.438
Error (T) .0614 120
Total .0385
149
Trials .1478
5
5. 022***
12 Words Linear
• 5739
1 19.520***
Nonlinear .0413 4 1.407
Error (T) .0294 120
Total .0280 149
Trials .0311
5 1.307
16 Words Linear .0074 1 .311
Nonlinear .0370 4
1.555
Error (T) • .0238 120
*p < .05
***p < .001
o
variance performed, only that for the 12-word category in
Group A showed a significant effect of trials. A test for
linear trend was then performed for all ten categories, as
summarized in Tables 18 and 19. The 6-word category in
Group C and the 6- and 12-word categories in Group A were
the only ones which showed a significant linear trend over
trial blocks and none of the nonlinear trends were signifi
cant.
SO for 6-trial blocks was subjected to a similar
analysis. These data for Group C are displayed in Figure
12, and for Group A, in Figure 13. The four analyses of
variance for Group C are summarized in Table 20 and those
for Group A, in Table 21. These results indicate that in
Group C, all but the l6-word category increased signifi
cantly in SO from the first 6-trial block to the second. In
Group A, all but the 8-word category increased significantly
in SO. Both 6-word categories showed a significant increase
in SO, and one of the l6-word categories did not. Clearly,
these results do not support the hypothesis.
Since the small categories did show an increase in
SO, the question arises whether SO in these categories af
fected recall. That is, SO might exist in these categories
and yet be unnecessary to and therefore uncorrelated with
recall. Consequently, the correlation of SO and recall
was examined.
64 '
jh
?
vo
• >
r t
o
•H
• 8
N
bO
0
1
•P
0
0 )
•^3
1
.50
• 35
.30
•25
.20
• 15
• 6-word category
# 8-word category
■ 12-word category
A l6-word category
.10
.05
1-6 7-12
Six-Trial Blocks
Figure 12. Mean SO, 6-trial blocks,
within categories, Group C.
65
.lK)
vo
.25
.20
• 15
# 6-word category
# 8-word category
■ 12-word category
A 16-word category
.10
.05
1-6 7-12
Six-Trial Blocks
Figure 13. Mean SO, 6-trial blocks,
within categories, Group A.
66
TABLE 20
ANALYSES OP VARIANCE OF SO FOR 6-TRIAL BLOCKS WITHIN CATEGORIES
OF FROM 6 TO 16 WORDS, GROUP C
Category
Size Source MS df • F
Total .0154
49
6 Words Trials .0834 1 7.908**
Error (T) .0105 24
Total .0111
^9
8 Words Trials .1017 1 12.000**
Error (T) .0085 24
Total .0091
49
12 Words Trials .0376 1 5.846*
Error (T) .0064 24
Total .0088
49 •
16 Words Trials .0119 1 2.590
Error (T) .0046 24
*p < .05
**p < .01
67
TABLE 21
ANALYSES OP VARIANCE OP SO FOR 6-TRIAL BLOCKS WITHIN CATEGORIES
OF FROM 6 TO 16 WORDS, GROUP A
Category
Size Source MS df F
Total .0217
49
6 Words Trials .1268 1 8.275**
Error (T)
.0153
24
Total .0098 b9
8 Words Trials .0017 1 .212
Error (T) .0081 24
Total .0068 49
12 Words Trials .0866 1 19.450***
Error (T) .oo44 24
Total .0060
49
16 Words Trials
.0530
1 11.479**
Error (T) .00*16 24
***p < .001
68
The correlation of SO and recall involves a spurious
factor, since the'two measures involved are not independent.
SO is the ratio of actual to maximum organization, as de
fined earlier:
t log n±J
SO = -------------
Z n. log n.
i 1 1
Here n^ represents the number of times the i^*1 word is re
called in a given trial block, So £ ni is simply S's total
recall for the trial block. Similarly, n ^ represents the
number of times the i^h word is recalled in a given position
in a given trial block, so n^j is simply total recall for
the trial block. Since log n^j and log n1 are used as mul
tipliers, neither the numerator nor the denominator repre
sents simply recall. However, they are both clearly corre
lated with recall. Since log 1=0, and since many words
are.recalled only once in a given position, quite a few
words drop out of the numerator, particularly when actual
organization is low. However, over 6-trial blocks, like
those used here,0 words are not often recalled only once.
Hence, where SO is low, the denominator is more highly cor
related with recall than is the numerator. This tends to
force a negative correlation between SO and recall. When
SO is very high, however, the numerator approaches the de
nominator in value and in its correlation with recall, which
69
tends to lessen the pull toward a negative correlation.
Thus it seems that the likelihood of bias decreases as SO
increases, but the problem exists in any case. Since there
is no way to obtain independent estimates of SO and recall,
the correlation was calculated anyway. The nonindependence
of the two measures involved must be considered in inter
preting the results.
For purposes of this analysis, the 8-word category
was arbitrarily defined as small. Then the log of the total
SO from the 6- plus the 8-word category for 6-trial blocks
was correlated across Ss with total recall from these two
categories for the same six trials. Similarly Log Total SO
for the 12- and l6-word categories was correlated with total
recall from these categories. The correlations for Group C
are given in Table 22 and those for Group A, in Table 23.
In neither the small nor large categories did Group A show
a correlation between log SO and recall. On the other hand,
Group C showed significant correlations, .409 and .607* be
tween log SO and recall for Trials 7-12 for the small and
large categories, respectively. Since none of the differ
ences between the correlations for small and large catego
ries approaches significance it appears that category size
has nothing to do with the correlation of log SO and recall
in these groups.
T A B L E 22
CORRELATIONS OF LOG SO FOR 6-TRIAL BLOCKS AND RECALL
WITHIN CATEGORIES OF DIFFERENT SIZES, GROUP C
TRIALS
Category Size
6 + 8 12 + 16
1-6 .231 .161
7-12 .409* . 607**
< .05
**p < .01
TABLE. 23
CORRELATIONS OF LOG SO FOR 6-TRIAL BLOCKS AND RECALL
WITHIN CATEGORIES OF DIFFERENT SIZES, GROUP A
TRIALS
Category Size
6 + 8 12 + 16
1-6 .217 -.115
7-12 .081 .306
Hypothesis 6
It was expected that Group C would have more in
trusions from the conceptual categories than Group A would
have from the alphabetical categories. Extralist intrusions
in the cued groups were scored according to which type of
category they belonged to: just alphabetical, just concep
tual, neither, or both. To qualify as a member of a concep
tual category a word had to be (l) a clearcut member, (2)
mentioned in the taxonomic norms for that category, or (3)
recalled by a member of Group C under that category name.
The number of each type of intrusion was totaled over the
twelve trials for all cued Ss. The mean number of each
type of intrusion over the twelve trials is tabulated in
Table 24. Group C showed slightly more categorical intru
sions than Group A. The difference between the groups in
Total Categorical Intrusions, consisting for Group A of Al
phabetical and Both, and for Group C of Conceptual and Both,
was tested by means of a t-test, but did not approach sig
nificance. The hypothesis received no support.
Hypothesis 7
Performance on the digit span test was correlated
with the number of words recalled on a trial which were not
recalled on the previous trial (NC) to see if a positive
correlation existed. The digit span measure used was the
total number of digits recalled, in the correct position,
72
TABLE 24
NUMBER OF CATEGORICAL INTRUSIONS OVER 12 TRIALS, GROUPS C AND A
Type of Intrusion Mean for Group A Mean for Group C t p
Just Alphabetical 0.76 0.00
Just Categorical 0.00 0. 96
Ne ither 0.00 0.00
Both 2.40 2.72
Total 5.16 3.68 .889
73
throughout the 28 series given. Three measures of NC were
used, because NC can be independent of number recalled on
the previous trial only when recall is not close to the to
tal possible. As recall gets close to the list length, NC
eventually will be determined not by digit span or any
other such factor, but merely by recall on the prior trial.
Since many Ss recalled nearly the whole list on the last
trials, NC was separately totaled for the first and last 6-
trial blocks. It was expected that a correlation between
the two measures would exist on Trials 1-6. The correla
tions of NC with digit span for the two 6-trial blocks and
for the 12-trial block are presented in Table 25. For none
of the three groups is NC significantly correlated with
digit span for the first six trials or for all twelve trials.
However, when the three groups' correlations for Trials 1-6
are averaged, the resulting r = +.24 is significant. This
is consistent with the hypothesized relation between NC and
digit span. On the other hand, in Group C, r = -.416, p <
.05, for Trials 7-12. Evidently Ss with higher digit span
tend to add less new words on these later trials, suggest
ing that they have reached a high level of recall and that
for them digit span should be positively correlated with
recall. Indeed it was found that in Group C, digit span
and recall on Trials 1-6 were significantly correlated (r =
+.427), as shown in Table 26. The mean correlation coeffi
cients for digit span and recall in the two cued groups were
74
TABLE 25
CORRELATIONS OF DIGIT SPAN WITH NC, GROUPS A, C, AND F
TRIAL BLOCKS
Trials 1-6 Trials 7-12 Trials 1-12
Group A
.276 -.01*3
.155
Group C
.11*7 -.1 + 16* -.392
Group F
.281 .173 .237
Mean r for
All Groups
. 21**
*p < .05
TABLE 26
CORRELATIONS OF DIGIT SPAN WITH RECALL, GROUPS A, C, AND F
Trials 1-6
TRIAL BLOCKS
Trials 7-12 Trials 1-12
Group A
1 —1
CM
•
.319
.301
Group C .1*27*
.295 •577
Group F
-.0 3 5
.003 -.012
Mean r for
Grps A & C
.34* . .30* .34*
*p < .05
then obtained, as shown in Table 26. For all three trial
blocks, the mean correlation between recall and digit span
performance was significant for the cued groups. For these
groups, it appears that digit span and recall were related
throughout learning.
CHAPTER V
DISCUSSION
The major experimental hypotheses were supported:
the two sets of cues led to differing recall, and later in
learning the rate at which recall decreased as items per
cue increased was slower for the conceptual cues. Mandler
(1967) found that recall was about 100 percent for catego
ries up to about 3 words, then dropped to 75 percent at
about 7 words, then dropped to 50 percent at about 12 words,
and to 25 percent for very large categories. In the present
experiment, proportional recall for both cued groups on
Trials 1-3 was lower than that of Mandler’s Ss. That is,
the superior recall of Group C was 88 percent for the 2-word
category, about 68 percent for 4- to 8-word categories,
and it then dropped to about 48 percent for 12 words. Pro
portional recall for Group A was even lower, of course: 69
percent for the 2-word category, about 45 percent for the
4- to 8-word categories, and 36 percent for 12 words. On
Trials 4-6, however, Group C's performance began to exceed
that of Mandler's Ss, and by Trials 9-12 proportional re
call for Group C was close to 90 percent for 8 words, as
opposed to Mandler's 75 percent, and dropped only as far as
86 percent for 12 words., as opposed to Mandler1 s 50 percent.
In other words, early in learning proportional recall for
both groups in the present experiment was lower for most
category sizes, small and large, than that of Mandler's Ss.
On later trials, however, both groups showed almost 100 per
cent recall in the 2-word category, and differing propor
tional recall for larger categories. The fact that Group A
showed lower recall in the small categories on the early
trials than did Group C was probably responsible for the
failure of the two groups to show a difference in rate of
decrease of proportional recall as category length in
creased. That is, Group A's recall was lower than Group
C's for all category lengths on early trials. By Trials
7-9 j however, Group A recalled 98 percent from the 2-word
category, and the significant difference in rate of decrease
emerged. The evident superiority of the proportional recall
of the present Ss on later trials to that of Mandler's Ss is
probably due to the fact that the present Ss had a greater
opportunity for practice. It will be recalled that Mand-
ler's Ss learned their words only to a criterion of two suc
cessive identical sorts, whereas the present Ss had twelve
trials, and the opportunity to build up SO within catego
ries. This difference probably underlies the superiority
of the present Ss in proportional recall. In any case, the
results of the present experiment strongly suggest that
proportional recall from within categories is determined
not only by the size of the categories and the stage of
learning at which recall Is tested, but also by the particu
lar manner In which a given set of Items Is grouped Into
categories.
It has been seen that the decrease In proportional
recall as number of Items per cue increases is one of the
main reasons for the category-recall relationship. Ear-
hard* s (1967) finding that small categories are learned
more rapidly than large categories suggests that after a
sufficiently large number of trials large categories might
be recalled as well as small categories. If this were the
case, the category-recall relationship would not hold late
in learning. However, even though Group C's overall recall
was extremely high on Trials 9-12, this group still showed
a significant decrease in proportional recall as category
size increased on these trials. Thus, the category-recall
relationship is expected to hold even after words have been
well learned. However, the considerably smaller decrease
in proportional recall shown by the present Ss compared to
Mandler*s, coupled with the large differences in recall and,
on later trials, in the rate of decrease in proportional re
call shown by Ss using the two sets of cues, strongly sug
gest that along with number of categories or units the ef
fectiveness of the particular organizational units chosen
will affect an S's performance in the free recall situation.
79
In order to aid extrapolation from cued to free re
call, the hypothesis was Investigated that the. cued groups
would be less variable in recall than would be the free
group, because their organizational units were determined
for them, and hence this source of variance was removed.
Had the hypothesis been supported, it would have bolstered
the position that variance in free recall is partly due to
variability in the organizational units used. The hypothe
sis received only the weakest support, however, in the find
ing that Group C's variance was smaller than Group F's for
Trials 7-12. On these trials Group C's recall was extremely
r \
high so that their smaller variance is probably due to a
ceiling effect rather than the effect of cueing as such. In
hypothesizing the difference in variance, however, it had
been anticipated that Group F would tend to use both pos
sible organizations, at least to some extent, and conse
quently would be more variable than the cued groups who had
no choice. The overwhelming use by Group F of the concep
tual organization, as evidenced in Figure 2 and in the fail
ure of alphabetical HR to increase significantly over tri
als, probably means that Group F was almost as consistent in
their use of the categories as were the cued groups. Conse
quently, no differences in variance would be expected. The
extrapolation to free recall, then, must be made without the
support which variance differences would have offered. Fur
ther research is required in order to relate the present re-
80
suits directly to free recall. For example, if the effec
tiveness of Ss' groupings of words does play a role in free
recall, then demonstrable differences- should exist between
fast and slow learners in the manner in which they sort a
set of words.
Earhard (1967) found that free recall was superior
to alphabetical cued recall with more than 6 items per cue,
but in the present study free recall was not superior to
alphabetical cued recall, totalled over 12 trials, with a
mean of 8 items per cue. This difference between the two
studies probably results from the use in the present study
of mixed lists. For a given group, Earhard's categories
were all the same size, whereas the present Ss were given
categories ranging from 2 to 16 words in length. The free
recallers here were clearly using the conceptual categories,
and they were forgetting significantly more words due to
nonrecall of categories than were the cued groups, as shown
in Table 14. These forgotten categories undoubtedly tended
to be the smaller ones; Tulving and Pearlstone (1966) found
that in free recall, category recall became more probable as
category size increased. Hence it seems likely that with
unmixed lists of 8 items per cue the present study's free
recall Ss would have performed better than the alphabetical
group Ss. This position is, of course, supported by their
superior recall of words from the 8-, 12-, and 16-word cate
gories seen in Table 13.
81
The average correlation of digit span and NC for
the three groups, r = .24, was low but significantly differ
ent from zero. This finding is consistent with the reason
ing by which NC is related to immediate memory capacity,
but it argues against a strong relationship between the two
variables. At first glance it might appear that the size
of the correlation could have been reduced by any one of
several possible factors. First, the words were presented
visually but the digits were presented auditorily. This
difference in mode of presentation should not lessen the
correlation however, as Jensen (1964) has shown that in
normal college Ss, like those used here, the correlation
between auditory and visual digit span is virtually perfect
when corrected for attenuation. Another factor which could
conceivably lower the correlation is the categorical nature
of the list. Because of the nature of the list, it could
be argued,'the number of units used is determined and cannot
vary with S's immediate memory capacity. Of course, all Ss
are free to use more than six units by subdividing larger
categories. Besides this, the free recall group, which was
free to deviate at least to some extent from the number of
units involved in the cues, actually tended to show lower
rather than higher correlations of digit span with NC than
o
did the cued groups. A more serious problem, however, may
be the serial nature of recall in the digit span task and
the use of free recall in the learning task. It can be
82
argued that serial recall requires different skills and
capacities than free recall. Tenopyr (1966) indeed found
that performance on a digit span task tended to he corre
lated more with performance on tasks requiring ordered re
call than with tasks requiring free recall. It must be con
cluded then that there is minimal support for the hypothesis
that NC is related to immediate memory capacity, but that
use of a free recall test of immediate memory might lead to
stronger support for the hypothesis.
The most puzzling finding in the present research
is that the size of a category was not related to the
growth of SO or to the correlation of SO with recall within
the category. SO for 2-trial blocks did not increase re
liably over trial blocks in most categories. SO for 6-trial
blocks, on the other hand, did increase significantly from
the first to the second trial block in most categories, re
gardless of their size. It will be recalled that the sensi
tivity of the SO measure increases with the size of the
trial blocks used. For example, the recall of a word in
the same order on both Trials 1 and 3 would not add to SO
for 2-trial blocks, but it would add to SO for 6-trial
blocks. It can be concluded, then, that when the less sens
itive measure is used, SO does not generally increase within
categories used in cued recall; when 6-trial blocks ar'e
used, however, SO within categories generally does Increase,
irrespective of the size of the categories. The correla-
tion of SO with recall within categories was also found to
be independent of the category size. If SO functions to
overcome the limit on immediate memory., for categories
whose length does not exceed this limit, it is not clear why
SO need occur. That immediate memory is involved in re
trieval of items from within categories is indeed suggested
by the low but significant mean correlation of digit span
with recall throughout learning in the cued groups. It is
possible, of course, that the shorter categories used in
the comparison here were too long to fall clearly within
immediate memory span. Because 6 is the least number of
words for which SO can be measured easily for a 6-trial
block, the 6- and 8-word categories were considered small
and were compared to the larger 12- and l6-word categories.
It may be that if smaller categories were used in such a
comparison, categories which more clearly fall within the
memory span, the original hypothesis might be supported.
That Group A apparently learns a fair number of
words without showing a correlation of recall with log SO
within the 6- to l6-word categories is surprising in view
of the large correlations Tulving (1962a, 1964) has ob
tained. It is tempting to conclude that, contrary to gen
eral belief, under some circumstances organization may be
unnecessary to learning. However, this conclusion would be
unwarranted in view of the fact that when SO is low' there
is a bias toward a negative correlation between SO and re-
84
call. Group A does show a lower level of SO than Group C,
and this relatively low SO may have been sufficient to
cause the zero correlation. Consequently the meaning of
the zero correlation is unclear, and no conclusion is pos
sible.
The comparison of categorical intrusions in Groups
A and C failed to support Mathews' position, that Group A's
inferiority in recall was due to their having to deal with
more interference than Group C. However, this comparison
also failed to support the prediction that Group C would
have to deal with more interference than Group A because
Group C's interfering items were higher frequency associ
ates of their cues than were Group A's. Thus neither posi
tion concerning why the two groups differed in recall re
ceived support from the intrusion data. It is reasonable
to suppose, however, that the difference in frequency of
the list words as responses to their cues was the variable
which determined the difference in recall. Deese (1959)
found that lists made up of high, low, or zero frequency as
sociates of Kent-Rosanoff stimulus words vary not only in
average frequency of the list words as responses to their
common stimulus, but also in inter-item associative strength.
That is, the average relative frequency with which all the
items in a list tend to elicit all the other list items as
free associates varies along with the average frequency of
the words as free associates to the stimulus word. Recall
from such lists was correlated +.88 with inter-item associ
ative strength in Deese's experiment. It is highly prob
able that the sets of categories used here varied in inter
item associative strength as well as in average frequency
as responses to their cues. With categories high in both
frequency and inter-item associative strength, S has two
advantages. The cue is likely to evoke the category mem
bers, and the category members are likely to evoke each
other. Furthermore, since the category members are highly
interrelated, a relatively high level of SO, as found in
Group C, is likely to develop. The difference in average •
frequency of the category members as responses to the two
sets of cues, and an attendant difference in inter-item
associative strength, is consequently held to be the most
probable source of the difference observed in recall.
Yet another type of difference between the two. sets
of groupings of words used here is suggested by Guilford's
structure-of-intellect model (Guilford, 1967). In this
model three dimensions of intellect are hypothesized: ope
rations are the major intellectual activities, contents are
the types of information dealt with by the organism, and
products are the forms in which the Information is dealt
with. The operations are (l) cognition, which is comprehen
sion, (2) memory, (3) divergent production, defined as the
generation of a variety of information from other informa
tion, (4) convergent production, defined as generation of
certain particular information which is called for by the
given information, and (5) evaluation, or judgment about
the criterion satisfaction of information. The two types
of contents of interest here are symbolic and semantic.
Symbolic information is in the form of signs which-have no
meaning in and of themselves, such as letters or even words,
when the meanings of the words are not considered. Semantic
information is in the form of meanings. The two types of
products of interest here are units and classes. Units are
defined as items of information with ''thing" character and
classes are the conceptions underlying groupings of similar
items.
There are a number of theoretical relations between
the structure-of-intellect model and processes involved in
verbal learning. For example, in free recall of experi
menter-categorized word lists, the ability known as cogni
tion of semantic classes should be involved in the process
of discovery of the categories on the list. The free re
call group in the present experiment presumably needed to
use this ability, whereas the cued groups did not. Once
recognized, the categories must be remembered as must the
words. Consequently, the abilities known as memory for
semantic classes and memory for semantic units would pre
sumably be brought into play. On the other hand, when free
recall of "unrelated" words is the task, S is free to gener
ate his own groupings of words. The present experiment
suggests that his ability to group words effectively will
be involved in determining his recall. The ability called
for in this case would appear to be evaluation of semantic
classes. Thus performance on a test of this ability should
be positively correlated with free recall of unrelated
words. Finally, when cued recall of categorized lists is
the task set for S, it appears likely that S would tend to
use the cue to generate category members which he then must
evaluated for list membership. Divergent production and
evaluation, then, are the operations he would perform. The
contents, however, would depend on the cues. If the cues
name semantic classes, divergent production and evaluation
of semantic units would be involved. If, on the other hand,
the cues name symbolic classes, divergent production and
evaluation of symbolic units would be involved. In the
present experiment the conceptual cues can be viewed as
calling for divergent production and evaluation of semantic
units. The alphabetical cues, on the other hand, called
for divergent production and evaluation of symbolic units.
The two sets of groupings used in the present experiment re
quire S to learn and recall different types of information.
In terms of the structure-of-intellect model, then, it ap
pears that words considered as symbolic units may be more
difficult to recall than words which are stored and re
trieved according to their meanings. The extent to which
an S uses groupings which use the meanings of words, then,
would be expected to play a role in determining his recall.
Since inter-item associative strength is considered to be
an index of similarity of associative meaning, it appears
that groupings of words with greater similarity of associ
ative meaning will be recalled more readily than groupings
of relatively low similarity of associative meaning. Prom
either point of view, then, the extent to which S groups
words according to similarity in their meanings would be
expected to affect his recall.
It is well documented that Ss faced with the free
recall task tend to subdivide the list into groups of re
lated items. If strong relations among list words exist,
as in the present experiment, Ss will tend to use these re
lationships. If the list is made up of relatively unre
lated words, S will need to search harder for the bases of
such subdivisions. In any case, the results of the present
experiment suggest that S's recall within whatever subdivi
sions he uses will be a function of the same variables which
determine recall from a list. If the members of a subdivi
sion or category can be readily evoked by a cue, and if
they have a relatively high level of Inter-item associative
strength, they will be recalled relatively well. If not,
if they are poorly related words, proportional recall from
within the category will be lower. Thus, the same list of
words can be subdivided a number of different ways. S's
skill in identifying sublists which share associations with
either a cue or with each other will determine, in part, his
free recall performance.
CHAPTER VI
SUMMARY
Learning is often considered to progress by means
of the grouping of items together into informationally
richer units or "chunks." Two Implied sources of individ
ual differences In recall are then (l) the number of units
S uses, and (2) the number of Items per unit which he can
retrieve from storage. An experiment was designed to In
vestigate the second of these possibilities. It was hypoth
esized that groups of words which are high frequency re
sponses to relevant cue words would allow recall of more
items per grouping than would groups of words which are
relatively low frequency responses to their cues. Conse
quently, a list of words was devised which could be divided
into 2 different sets of 6 categories each. One set was de
fined by the Initial letters of the wordsj the other set
consisted of conceptual categories to which the words be
longed, e.g., animals. It was assumed that the words were
higher frequency responses to their relevant conceptual
category names than they were to their initial letters. It
was hypothesized that (l) Ss given conceptual cues during
recall would recall more words than would Ss given alphabet-
90
91
leal cues, and (2) of the 2 sets of groupings, the concep
tual categories would lead to a slower rate of decrease in
proportional recall as number of items per category in
creased. In addition, it was possible to compare the growth
of subjective organization within the categories named’by
the 2 sets of cues and to compare the growth of SO within
categories of different sizes. Finally, an attempt was made
to relate score on a digit span test to the number of items
recalled on a trial which were not.recalled on the previous
trial (NC).
Three groups of Ss were given 12 trials to learn
the same 48 words. Words were drawn from 6 conceptual cate
gories and each had one of 6 initial letters. Thus each
word was relevant to one conceptual and one alphabetical
category. The 2 sets of categories were not synonymous;
each conceptual category contained members of as many alph
abetical categories as possible. Two groups were cued dur
ing recall, one with alphabetical (first letter) cues and
one with conceptual category name cues. The third group
was not cued. Both cued groups were given 6 cues, one each
relevant to 2, 4, 6, 8, 12, and 16 list items. Stimuli were
presented visually at the rate of 2 seconds per word and 2i
minutes were allowed for recall on each trial.
Conceptual cues led to greater recall throughout
learning than either free recall or alphabetical cues. Free
recall was not superior to alphabetical cued recall in gen-
eral, but it was for words in the 3 larger categories. For
the last 6 trials, conceptual cues led to a slower rate of
decrease in recall as number of items per cue increased than
did alphabetical cues. The size of a category had no ef
fect on either the growth of SO within the category or the
correlation of SO with recall from the'category. The aver
age correlation for the 3 groups of digit span and NC was
low but significant (+.24).
The major conclusion was that the particular units
into which S groups list items during free recall are prob
ably a significant source of variance in free recall per
formance. Items grouped together will be recalled rela
tively well if they are readily evoked by a common cue or
by each other. It is probable that S’s skill in identifying
and using such groupings will determine in part his free
recall performance.
REFERENCES
93
REFERENCES
Bousfield, W. A. The occurence of clustering in the recall
of randomly arranged associates. Journal of General
Psychology, 1953, 49, 229-240.
Bousfield, W. A., & Cohen, B. H. Clustering as a function
of the number of word-categories in stimulus-word lists.
Journal of General Psychology, 1956, 54, 95-106.
Bousfield, W. A., Cohen, B. H., & Whitmarsh, G. A. Associa
tive clustering in the recall of words of different
taxonomic frequencies of occurrence. Psychological Re
ports, 1958, 4, 39-44.
Bousfield, W. A., Puff, C. R., & Cowan, T. M. The develop
ment of constancies in sequential organization during
repeated free recall. Journal of Verbal Learning and
Verbal Behavior, 1964, 3, 439-495-
Cofer, C. N. A study of clustering in free recall based on
synonyms. Journal of General Psychology, 1959, 60,
3-10.
Cofer, C. N. On some factors in the organizational charac
teristics of free recall. American Psychologist, 1965,
20, 261-272.
Cohen, B. H. Some-or-none characteristics of coding be
havior. Journal of Verbal Learning and Verbal Behavior,
1966, 6, 162-167.
Cohen, B. H., & Bousfield, W. A. The effects of a dual
level stimulus-word list in the occurrence of cluster
ing in recall. Journal of General Psychology, 1956,
55, 51-58.
Cohen, B. H., Bousfield, W. A., 8c Whitmarsh, G. A. Cul
tural norms for verbal items in 43 categories. Techni-
cal Report No. 22, ONR Contract Nonr-631 (00), Univer
sity of Connecticut, 1957*
94
95
Cohen, B. H., Sakoda, J. M., & Bousfield, W. A. The statis
tical analysis of the incidence of clustering in the
recall of randomly arranged associates. Technical Report
No. 10, ONR Contract Nonr-631 (00), University of Con
necticut, 1954.
Dallett, K. M. Number of categories and category informa
tion in free recall. Journal of Experimental Psychology,
1964, 68, 1-12.
Deese> J. Influence of inter-item associative strength up
on immediate free recall. Psychological Reports, 1959*
5* 305-312.
Dong, T., & Kintsch, W. Retrieval cues in subjective organ
ization. Paper read at Western Psychological Associa
tion meetings, 1967.
Earhard, M. Cued recall and free recall as a function of
the number of items per cue. Journal of Verbal Learn
ing and Verbal Behavior, 1967, b, 257-253.
Guilford, J. P. The nature of human intelligence. New
York: McGraw-Hill, 1967.
Jenkins, J. J., & Russell, W. A. Associative clustering
during recall. Journal of Abnormal and Social Psychol
ogy, 1952, 47, 818-821. ;
Jensen, A. R. Individual differences in learning: inter
ference factor. Final report, Cooperative Research
Project No. I867, U.S. Office of Education, 1965.
Mandler, G. Organization and memory. In K. W. Spence and
J. T. Spence (Eds.) The psychology of learning and
motivation. New York: Academic Press, 1967, Vol. 1.
Mandler, G., & Pearlstone, Z. Free and constrained concept
learning and subsequent recall. Journal of Verbal
Learning and Verbal Behavior, 1966, 5, 126-131.
Mathews, R. Recall as a function of number of classifica-
tory categories. Journal of Experimental Psychology,
1954, 47, 241-247.
Miller, G. A. The magical number seven, plus or minus two:
Some limits on our capacity for processing information.
Psychological Review, 1956, 63, 81-96.
Tenopyr, M. A. A factor-analytic study of symbolic-memory
abilities. Unpublished doctoral dissertation, Univer
sity of Southern California, 1966.
Thorndike, E. L., & Lorge, I. The teacher's word book of
30,000 words. New York: Teacher's College, Columbia
University, 1944.
Tulving, E. Subjective organization in free recall of "un
related" words. Psychological Review, 1962, 69, 344-
354. (a)
Tulving, E. The effect of alphabetical subjective organiza
tion on memorizing unrelated words. Canadian Journal of
Psychology, 1962, 16, 185-191* (b)
Tulving, E. Intratrial and intertrial retention: Notes to
ward a theory of free recall verbal learning. Psycho-
logical Review, 1964, 71, 219-237.
Tulving, E., & Pearlstone, Z. Availability versus access
ibility of information in memory for words. Journal of
Verbal Learning and Verbal Behavior, 1966, 5, 361-391.
Vel'dman, D. J. Fortran programming for the behavioral sci
ences . New York: Holt, Rinehart, and Winston, 1967.
APPENDIXES
97
APPENDIX A
LIST WORDS AND CUES
98
APPENDIX A
LIST WORDS AND CUES
Conceptual Categories
Alphabet
ical
Categories
Musical
Instruments Trees Animals Weather Occupations Furniture
Alphabet ical
Category
Length
S servant sofa 2
T
turtle typhoon tailor trunk k
L
lemon lynx lightning lawyer lounge
lamp
6
D
dogwood deer dry dentist
dean
desk
dresser
drawer
8
B
bassoon beech bear blizzard
breeze
banker
butcher
boss
bureau
bed
bench
bookcase
12
C
cello chestnut camel
chipmunk
cold
calm
cloudy
chief
clerk
captain
carpenter
couch
cabinet
cot
clock
chest
16
Conceptual
Category
Length
2 k 6 8 12 16
vo
vo
APPENDIX
INSTRUCTIONS,
INSTRUCTIONS,
INSTRUCTIONS,
B
GROUP C
GROUP A
GROUP F
100
APPENDIX B
INSTRUCTIONS— GROUP C
This Is an experiment In verbal learning. A list
of 48 words will be shown to you on each of a series of
trials. You are to learn these words. After each trial
you will be asked to recall as many of the words as you can.;
To help you recall the list words, certain other words which
describe groups of words on the list will be printed on the !
recall sheet as cues. You are to write your recalled words ;
underneath the cue to which they belong. For example, if
you remembered "apple" and "orange" were on the list, you
would write them under "Fruit."
Fruit:
apple
!
orange j
Do not write anything while the words are being
shown to you. Do not ask questions or talk for any other
reason during the experiment.
101
APPENDIX B--Continued
INSTRUCTIONS— GROUP A
This is an experiment in verbal learning. A list
of 48 words will be shown to you on each of a series of
trials. You are to learn these words. After each trial
you will be asked to recall as many of the words as you
can. To help you recall the list words, the letters with
which the words begin will be printed on the recall sheet
as cues. You are to write your recalled words underneath
the cue to which they belong. For example, if you remem
bered "apple” and "arithmetic" were on the list, you would
write them under "A":
A:
apple_____
arithmetic
Do not write anything while the words are being
shown to you. Do not ask questions or talk for any other
reason during the experiment.
APPENDIX B— Continued
INSTRUCTIONS--GROUP P
This is an experiment in verbal learning. A list
of 48 words will be shown to you on each of a series of
trials. You are to learn these words. After each trial
you will be asked to recall as many of the words as you can,
in any order. You are to begin at the upper left of the
recall sheet and fill the first column before beginning the
second column. For example, if you remembered "orange" and
"arithmetic" were on the list, you would write "orange"
first and then underneath it you would write "arithmetic."
orange
arithmetic
Do not write anything while the words are being
"shown to you. Do not ask questions or talk for any other
reason during the experiment.
103
APPENDIX C
RECALL SHEET, GROUP C
RECALL SHEET, GROUP A
RECALL SHEET, 'GROUP P
o
o
%
104
Tree:
APPENDIX C
RECALL SHEET— GROUP C
Animal:
Weather:
Furniture:
Occupation:
Musical Instrument:
105
APPENDIX C— Continued
RECALL SHEET— GROUP A
T: L:
D:
C:
Bs
S:
106
APPENDIX C— Continued
RECALL SHEET— GROUP' P
107
APPENDIX D
LATIN SQUARE FOR ORDERS OF CATEGORY SIZES
OF CUED RECALL SHEETS
108
APPENDIX D
Column 1
Column 2
LATIN SQUARE FOR ORDERS OF CATEGORY SIZES
ON CUED RECALL SHEETS
Order 1 Order 2 Order 3 Order ^ Order 5 Order 6
2 16 k 12 6 8
16 k 8 6 2 12
8 6 12 k 16 2
4 2 6 8 12 16
6 12 16 2 8 k
22 8 2 16 i t 6
109
APPENDIX E
INSTRUCTIONS FOR DIGIT SPAN TEST
110
APPENDIX E
INSTRUCTIONS FOR DIGIT SPAN TEST
Here are your instructions for the rest of the ex
periment. Listen carefully. In the second part of the ex
periment you will be asked to remember short series of dig
its. A tap will sound, like this (tap), immediately before
a series of digits are read to you. After the last digit
the tap will sound again (tap). This second tap is your
signal to write down the digits in the same order in which
they were read to you. Notice the rows of boxes on page A.
After the first series is read to you, you will write it
down in the first row of boxes, beginning at the left.
Each digit then goes in a separate box, in the same order
in which you heard it. The next series goes in the second
row, and so on. * Do not write anything yet, but here is-an
example. Suppose the series went: "tap 2 4 tap." You
would write 2 in the first box and 4 in the second. During
this task do not write anything while the digits are being
read— wait for the second tap and then write them. Okay,
get ready, here is the first series.
Ill
APPENDIX F
DIGIT SPAN RECALL SHEET
112
APPENDIX P
DIGIT SPAN RECALL SHEET
H3
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Effects Of Different Treatment Procedures On Reading Ability And Anxiety Level In Children With Learning Difficulties
PDF
Semantic-Memory And Creative (Divergent-Production) Abilities Of Senior-High-School Students
PDF
The Effect Of Discriminability On The Partial Reinforcement Effect In Human Gsr Conditioning
PDF
The Effects Of Feedback On The Communication Of Medical Prescription To Diabetic Patients
PDF
A Factor Analysis Of The Semantic-Evaluation Abilities
PDF
Human Operant Eye Blink Conditioning, Awareness, And The Extraversion-Introversion Dimension Of Personality
PDF
The Influence Of Communality And N On The Sampling Distributions Of Factor Loadings
PDF
Use Of College Students In A Social Therapy Program With Chronic Schizophrenics To Produce Changes In Mood And Social Responsiveness
PDF
Performance Of Retardates On Piagetian Tasks As A Function Of Ethnicity
PDF
The Effect Of Conditions Of Risk, Internal Versus External Control Of Reinforcement, And Sex On Binary Choice Probability Learning
PDF
Effects Of Group Behavior Therapy Imagery On Basketball Performance
PDF
Personality Variables And Intellectual Abilities As Determinants Of Concept Learning
PDF
The effects of an RNA polymerase in the improvement of verbal learning and memory
PDF
Cognitive, Linguistic, And Perceptual Handicaps: Discriminant Patterns Inlearning Discorders
PDF
A Comparison Of Two Methods Of Adapting Self-Instructional Materials To Individual Differences Among Learners
PDF
The Relationship Of Personal Data Items To Vocational Rehabilitation
PDF
The Effects Of Delay In Knowledge Of Results On The Amount Learned In Teaching Machine Programs Of Differing Cue Content
PDF
The Subjective Utility Function As An Estimator Of Optimal Test Weights For Personnel Selection
PDF
The Effect Of Dissonance In Self-Esteem On Susceptibility To Social Influence
PDF
Factorial Stability As A Function Of Analytical Rotational Method, Type Of Simple Structure, And Size Of Sample
Asset Metadata
Creator
Loeb, Jane Watkins
(author)
Core Title
Differences Between Cues In Effectiveness As Retrieval Aids
Degree
Doctor of Philosophy
Degree Program
Psychology
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
OAI-PMH Harvest,psychology, general
Language
English
Contributor
Digitized by ProQuest
(provenance)
Advisor
Cliff, Norman (
committee chair
), Metfessel, Newton S. (
committee member
), Priest, Robert F. (
committee member
), Slucki, Henry (
committee member
)
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c18-706163
Unique identifier
UC11362205
Identifier
6917884.pdf (filename),usctheses-c18-706163 (legacy record id)
Legacy Identifier
6917884.pdf
Dmrecord
706163
Document Type
Dissertation
Rights
Loeb, Jane Watkins
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA
Tags
psychology, general