ATTREBUSES 9F MEMGRY, RETROAC‘TWE ENHEBETEON,
QUEENS MED FREE RECALL LEQREEENG

“2.1222323 f3? 2222 3921222253 2 2‘ E25 2’2.
2 ECH’EGAN 3‘2 UTE UMU’ER‘EW
U‘2 23:2 LOU. 3E PEWE‘EGTON

5"}
it)”;

 

 

ABSTRACT

ATTRIBUTES OF MEMORY, RETROACTIVE INHIBITION, CUEING
AND FREE RECALL LEARNING

By

Joyce Louise Pennington

Underwood (1969) hypothesized that encoding multiple attributes
of an event can facilitate memory for that event by providing more
means of access to the event and that interference between two events
can be decreased by encoding multiple attributes of the events and by
decreasing the number of attributes the two events have in common.

The present study tested these hypotheses by varying the attributes
encoded during List 1 learning (semantic, acoustic, or dual) and the
attributes encoded during List 2 learning (semantic, acoustic, or dual).
Following List 2 learning subjects recalled List 1 once without cues
and once when cued with the attributes they had used for encoding

list words during original List 1 learning.

The attribute encoded was manipulated by varying the organization
of the lists, and all subjects learned the same List 1 and List 2 words.
For subjects who encoded two attributes of the words, the organization
of the list was changed from Trial 1 to Trial 2. The List 1 learning
data revealed that semantic organization subjects recalled more than
acoustic or dual organization subjects. Thus, multiple encoding does
not appear to facilitate learning. In recalling List 1 after List 2
learning, there was no effect due to the number of attributes encoded.
However, retroactive inhibition was less for those subjects who

received different organizations of successive lists (i. e., List 1

Joyce Louise Pennington

semantic organization-List 2 acoustic organization or List 1 acoustic-
List 2 semantic organization) than for those who received the same
organization of the two lists (i. e., semantic organization of both
lists or acoustic organization of both lists). The implications of
the data from subjects encoding multiple attributes were not clear.
Finally, the cueing effect was larger for the subjects who encoded
semantic attributes of List 1 than for those who encoded acoustic
attributes or multiple attributes. The results were discussed in
terms of the plausibility of the consistent organization versus

multiple organization views of retrieval processes.

ATTRIBUTES OF MEMORY, RETROACTIVE INHIBITION, CUEING
AND FREE RECALL LEARNING

By

Joyce Louise Pennington

A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

MASTER OF ARTS

Department of Psychology

1972

; ’4'

ACKNOWLEDGMENTS

The author wishes to express her appreciation to the graduate
committee members, Drs. Donald Johnson, Lester Hyman, and Gordon
Wood, for their assistance and support in the preparation of this
thesis. A special note of thanks is due Dr. Gordon Wood for his

continued guidance and interest in the development of my career.

ii

TABLE OF CONTENTS

Page

List of Tables iv
Introduction 1
Method 9
Design 9
Lists 9
Procedure 10
Results 12
List 1 Learning and Organization 12

List 2 Learning and Organization 17
Stage 3: Noncued Retroactive Inhibition 21
Stage 3: Cued Retroactive Inhibition 25
Discussion 29
References 34
Appendix 36

iii

LIST OF TABLES

Mean Number of List 1 Words Recalled by the Nine Groups
for Trials 1 and 2 and the Mean Total Recall for the
Two Trials.

Mean Clustering of List 1 Words by the Nine Groups for
Trials 1 and 2 and the Mean Total Clustering for the
Two Trials.

Mean Number of List 2 Words Recalled by the Nine Groups
for Trials 1 and 2 and the Mean Total Recall for the
Two Trials.

Mean Clustering of List 2 Words by the Nine Groups for
Trials 1 and 2 and the Mean Total Clustering for the
Two Trials.

Mean RI Scores for the Nine Groups (Stage 3).

Mean Clustering Scores for the Nine Groups for the
Stage 3 Recall of List 1 Words.

Mean Number of Additional Words Recalled by the Nine
Groups after Cueing during Stage 3.

The Mean Cued RI Scores for the Nine Groups (Stage 3).

iv

Page

13

15

18

20

22

24

26

27

INTRODUCTION

An important issue in memory research pertains to the character-
istics of the information stored in memory. Underwood (1969) has
adressed himself to this issue using the construct of attributes of
memory. Underwood proposes that a memory consists of a number of
attributes and that one memory is distinguished from another by its
specific collection of attributes. The attributes which he identifies
are temporal, spatial, frequency, modality, orthographic, associative
nonverbal (acoustic, visual, affective, contextual), and associative
verbal (class). Whereas certain attributes function primarily to
distinguish one memory from another (e. g., frequency), others act as
a means of retrieval of the memory (e. g., associative verbal).

Within this framework Underwood emphasizes the importance of
encoding multiple attributes of the to-be-recalled material as a means
both of facilitating retrieval processes and of minimizing interference.
More specifically, if multiple attributes are encoded, there are
presumably more ways to "gain access" to the to-be-recalled event.
Similarly, the greater the number of attributes encoded for a to-be-
recalled event and the less these attributes are a part of other to-be-
recalled events, the less the interference. Although Underwood implies
that the attributes of a single event are "stored together," it is
possible that under some conditions there is multiple, independent

representation of single events. The purpose of this research is to

determine if encoding multiple attributes facilitates recall and
decreases the amount of retroactive inhibition.

There is considerable evidence supporting the view that attributes
of the to-be-recalled material are stored and not the specific content
of the material (Bobrow, 1970; Bower, 1970; Light and Carter-Sobrel,
1970). A particularly useful approach is to hold the to-be-recalled
material constant and encourage subjects to encode different attributes
of the same material by varying the context or organization of the
material. Bower (1970) reported an unpublished study by Bobrow and
Light in which ambiguous words were used. Each ambiguous word (e. g.,
cardinal) could be classified into two categories (e. g., bird or
clergy). By varying the context of the ambiguous words subjects were
encouraged to encode different attributes of the words. If the study
context had encouraged the classification of the word according to one
category (e. g., priest, cardinal, minister), the category label
appropriate to that context (e. g., clergy) was an effective retrieval
one while the other category label (e. g., bird) was not.

In a later study using ambiguous noun pairs as the subject and
direct object of sentences, Bobrow (1970) manipulated the attribute
encoded by varying the sentence context. There were three context
conditions: exact repetition of the sentence (e. g., dog bark--baseball
pitcher); changed context-same semantic attribute (e. g., animal bark--
big—league pitcher); and changed context-different semantic attribute
(e. g., medicinal bark--porcelain pitcher). At recall subjects were
given the subject of the sentence and were asked to give as much of the
rest of the sentence as possible. There was little difference in

performance between the two groups who were encouraged to encode the

same semantic attribute on successive trials; therefore, simply
changing the sentence context does not impair performance. However,
when subjects were encouraged to encode different semantic attributes
on successive trials, there was almost no improvement in performance
over trials.

These studies strongly support the position that attributes of the
to-be-recalled material are stored rather than specific content of the
material. If specific content were stored, no difference in performance
should have been observed between the two groups receiving changed
sentence contexts. Of greater interest for the present study, however,
is the evidence provided by Bobrow's study on the effect of processing
more than one attribute of the material. Contrary to the hypothesis,
Bobrow's data suggest that encoding more than one attribute (i. e.,
more than one meaning) is detrimental to learning. However, because
this study was not designed as a test of Underwood's hypothesis, it is
open to certain criticisms as a valid test of the hypothesis. If
multiple attributes facilitate recall by providing more retrieval
routes to the memory, a design in which retrieval cues are given
(paired-associate task) destroys the advantage of multiple encoding.
Nevertheless, this argument does not explain the low degree of learning
attained by the encoding variability group.

Other evidence also suggests that inconsistent encoding results in
poorer learning than consistent encoding. Bower, Lesgold, and Tieman
(1969) presented subjects with a twenty-four word list in six groups
of four words each. The four-word groups were presented visually on
cards and subjects were instructed to imagine a scene in which the

objects interacted. One group received the same grouping of the words

on successive trials. Presentation of the words for the second group
was such that no two words belonged to the same group more than once.
Performance of the first group was far superior to that of the second
group, who showed little improvement over trials. The authors
interpret the data in terms of the importance of a consistent organiza-
tion and stable higher-order memory units. However, because context
is an important determiner of the attribute encoded, it is also likely
that subjects who received changed groupings of the words were encoding
different attributes on successive trials. In this case, the results
indicate that consistent encoding of one attribute, or set of attributes,
results in better learning than variable encoding.

The final study to be considered relating to multiple encoding
and free recall learning provides data inconsistent with that already
presented. Wood (unpublished study) used a word list which could be
organized according to four criteria: one set of semantic categories,
a second set of semantic categories, first letter, and number of
letters in the word. Prior to recalling the words, all subjects sorted
the words four times. One group sorted the words into the same set of
semantic categories four times (one—half of the subjects using each set
of semantic categories). The second group sorted the words twice into
each of the two sets of semantic categories for the four sorts. The
third group sorted the words in all four ways, in the order listed
above. (There was also an incidental vs. intentional learning manipul-
ation which will not be considered as the results are similar for both.)
The data are somewhat ambiguous. There was no difference in the amount
recalled by subjects who sorted according to one or two criteria

(semantic categories), but both of these groups were superior to the

group using four sorting criteria. The inferiority of the group
sorting according to four different attributes may be due to the fact
that the two final sorts involved attributes (first letter and number
of letters) which are probably not very effective either as organizers
or as retrieval cues. These final sorts may have had a detrimental
effect on the processing which had already occurred. If so, the
results of the first two groups indicate that encoding multiple
attributes does not impair learning.

The relationship between the number of attributes encoded and free
recall learning is certainly not clear. The data of Bobrow (1970) and
Bower, Lesgold, and Tieman (1969) indicate that consistent encoding of
the same attribute is the better method of processing material. However,
neither of these studies was intended as a test of the hypothesis and
there are important methodological differences between these studies
and Wood's data which are more supportive of the hypothesis. The two
most important differences involve the amount of control exercised by
the experimenter over the attributes encoded by the subjects and the
orthogonality of the attributes encoded. In order to complete the
sorting task, the subject must attend to the attribute intended by
the experimenter and must actively process the material. Because
associative verbal attributes are dominant with meaningful material,
it is necessary to encourage subjects to encode other attributes in
order to manipulate the number of attributes encoded. Secondly, and
more importantly, the use of orthogonal attributes in Wood's study
probably led to less interference between the attributes encoded on
successive trials. The word list used by Wood was designed such that

the attributes were probably independent of one another. These

methodological considerations have been incorporated into the design
of the present research.

The facilitative effect of encoding multiple attributes may
depend on the number of categories used. The recall of categories and
the recall of items within categories have been treated as independent
processes (cf. Mandler, 1967; Tulving and Pearlstone, 1966). This
distinction seems especially appropriate when considering the recall
of categorized word lists. Mandler (1967) has also suggested that the
number of categories which subjects can "spontaneously" recall is equal
to five, plus or minus two. In Wood's study only six categories for
each of the attributes were used. Therefore, subjects processing only
one attribute were probably able to retrieve all of the categories.
However, if the number of categories is increased beyond the number
which can be "spontaneously" recalled, then the importance of having
multiple retrieval routes should be increased. Therefore, a greater
number of categories was used in the present design.

Tulving and Pearlstone (1966) have demonstrated that information
in storage can be made accessible through cueing. Multiple encoding
is expected to effect retrieval processes. At retrieval, however, it
is hypothesized that multiple encoding subjects have more retrieval
routes and so should recall more. Cueing should destroy this advantage
by facilitating the retrieval processes of all subjects. In the
present study all subjects will be cued only after a standard free
recall trial.

0n the other hand, if encoding multiple attributes does not
increase recall, cueing may have a greater effect for the multiple

encoding subjects. It has already been established that a retrieval

cue is only effective when it is consistent with the encoded attribute
of the to-be-recalled event (cf. Bower, 1970; Thomson and Tulving, 1970).
Thus, those subjects encoding a single attribute have fewer effective
retrieval cues than those subjects encoding multiple attributes.

While the effect of cueing is not clear, the results of its manipulation
should be interesting.

The second purpose of the present study involves the relationship
between the encoding of multiple attributes and retroactive inhibition.
To date, no research has investigated the effect of single versus
multiple encoding of attributes on retroactive inhibition. The studies
most closely related to attributes and retroactive inhibition have
examined the effect of organizational processes on retroactive inhibition.
The principal finding has been that retroactive inhibition is greater
when successive lists have similar organizations than when the organi-
zations are different (Shuell, 1968; Thompson and Poling, 1969; Watts
and Anderson, 1969; Winograd, 1968; Wood, 1970, 1971). Since organi-
zation influences the encoding of attributes, these results indicate
that retroactive inhibition is greater when the same attribute is
processed on successive lists. Thus, these studies offer partial
support for the hypothesis: the less the attributes of a to-be-recalled
event overlap with the attributes of a second to-be-recalled event, the
less the interference between the two events. However, none of these
studies has manipulated the number of attributes encoded for the to-be-
recalled list (List 1).

In summary, there is very little evidence relating to the encoding
of multiple attributes and learning or retroactive inhibition. In

relation to learning, the data do not clearly support or refute the

position that multiple encoding facilitates learning. Two methodological
considerations important for evaluating these studies are the amount

of control over the attributes encoded (learning task) and the
orthogonality of the attributes encoded. Also, it appears that the
number of categories may be an important factor in determining the
effect of multiple encoding. In addition, if the facilitory effect of
multiple encoding is the result of better retrieval processes, then
cueing all subjects should destroy the advantage by improving the
retrieval processes of single encoding subjects. In terms of
retroactive inhibition, there is no evidence concerning the effect of
encoding multiple attributes. Therefore, the proposed research deals
with the effect of encoding multiple attributes on free recall learning,

retroactive inhibition, and cueing.

METHOD

Design. The design is a 3 X 3 factorial in which List 1
organization (acoustic, semantic, or dual) and List 2 organization
(acoustic, semantic, or dual) were the variables manipulated. All
subjects received two study and test trials on List 1 (Stage 1), two
study and test trials on List 2 (Stage 2), a noncued recall trial of
List 1 and a cued List I recall trial (Stage 3).

The organization of the list words was used to manipulate the
attribute encoded. Words belonging to the same category were presented
consecutively along with the category name. For those subjects
receiving both the acoustic and the semantic organizations of the words,
half received the acoustic organization on the first trial and the
semantic organization for the second trial while for the other half
of the subjects this order was reversed.

‘Ligtg. List 1 consisted of 48 words selected so that they could
be organized both according to twelve rhyming categories (e. g., sleigh,
bay; rain, plane) and according to twelve conceptual categories (e. g.,
transportation, weather) with from three to five words in each
category. The 48 List 2 words were selected so that they could be
organized into the same rhyming and conceptual categories used for
List 1, again with from three to five words in each category. The
words in both lists were concrete nouns, and the two lists contained

no words in common. In as far as was possible, no two words belonging

10

to the same conceptual category also belonged to the same rhyming
category. The two lists appear in Appendix 1, organized both into
the rhyming and conceptual categories.

Procedure. A total of 144 subjects were randomly assigned to the
nine experimental conditions such that 16 subjects were in each
condition. All subjects were tested in groups composed of from four
to twenty individuals. All subjects were enrolled in introductory
psychology classes at Michigan State University and received class
credit for their participation.

Upon arrival at the experimental session, subjects were randomly
assigned to the conditions. At the beginning of the session each
subject received a booklet containing a cover page, instruction page,
List 1 study page, test page, filler sheet, List 1 study page, test
page, instruction page, List 2 study page, test page, filler sheet,
List 2 study page, test page, instruction page, recall sheet, filler
page, and a cued recall sheet with instructions.

Prior to the presentation of each list, subjects were informed of
the organization of the list (i. e., the type of categories to be
presented) and were told that although the category names might help
them learn the words, they would not be asked to recall the category
names. Subjects receiving both organizations of a list were informed
that the organization of the list would change from Trial 1 to Trial 2.
For subjects studying a single organization of the list, the input
order was constant. All subjects were told that there would be two
trials and that on each test trial they were to recall as many words
as possible in any order. After subjects had read the instructions

and the experimenter had answered any questions, the experiment was begun.

11

There were two alternating study and test trials for List 1
learning. Subjects were allowed 150 seconds to study the words and
four minutes to recall the words on each trial. Instructions prior
to List 2 learning were almost identical to those for List 1 learning.
Those subjects who received the same organization of the two lists
were told that the same categories would be used but that there would
be new words in the categories. Those subjects receiving a different
organization were informed of the change. The List 2 study and recall
times were the same as those used for List 1 learning.

Immediately following List 2 learning, subjects read the instructions
requesting them to recall the List 1 words again. In addition, the
experimenter verbally emphasized the fact that List 1 was to be recalled.
Four minutes was allowed for List I recall. For the cued recall of
List 1 subjects were provided with the category labels which had been
present when they had studied List 1. Thus, subjects who received both
organizations of List 1 received 24 cues while all other subjects
received either 12 conceptual cues or 12 rhyming cues. Since Thomson
and Tulving (1970) have demonstrated that cues, to be effective, must
be encoded with the to-be-recalled material, there was no reason to
provide all subjects with all possible cues. Subjects were informed
that the cues might help them recall additional words. Subjects were
allowed to refer back to their last recall sheet to check on which
words they had already recalled, but they were instructed to write
any additional words which they recalled on the page with the cues.

Following the experiment subjects were informed of the purpose

of the experiment.

RESULTS

List 1 Learning and Organization

Recall protocols were scored for the number of words correctly
recalled on Trials 1 and 2 of Stage 1. The mean number of words
correctly recalled by the nine groups on Trials 1 and 2 and the total
mean recall for the two trials are presented in Table 1. An analysis
of variance revealed a List 1 Organization effect, §_(2, 135) = 61.39,
p-£.001; a Trials effect, E (l, 135) = 349.91, p (.001; and a Trials
by List 1 Organization interaction, §_(2, 135) = 11.33, p‘<.001. It
is clear from inspection of Table 1 that the List 1 Organization effect
is due to the higher recall of the semantic organization subjects
compared to both the acoustic and the dual organization subjects. The
Trials by List 1 Organization interaction reflects the fact that
semantic subjects show the greatest improvement over trials while the
dual organization subjects show the least improvement. The fact that
the List 2 Organization effect was nonsignificant indicates that the
three groups comprising each of the List 1 organization conditions did
not differ prior to the introduction of the List 2 organization
manipulation.

Recall protocols were also scored for the amount of clustering on
each trial. The clustering measure is the ratio of the obtained to
maximum possible clustering, which can vary from 0.0 to 1.0. A cluster

is counted every time two words belonging to the same category are

12

13

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

No.5m NH.~N om.mH oo.me mo.o~ mm.wH em.om No.HN Hm.mH H<bn
mu.wm oo.¢~ mH.¢H NH.mm qm.- mH.NH mN.wm mm.mm ww.¢H UHHmDoo<
om.~o oo.nm ¢¢.mm mo.om mH.mm om.HN mn.oo mo.nm oo.m~ UHHz<zmm
Hmuoa N H Hmuoa N H Hauoa N H
GOHumNHnmwuo
H < D Q o H H m a o o < o H H z ¢ 2 m m H umHH
GOHuoNHamwuo N umHH

 

 

.mHmHHH 035 onu pom HHmoom Hmuoa
cam: onu was N van H mHMHHH now wmsouo oon can an voHHwomm wwHoB H uwHH mo Hmnasz dam: .H oHan

 

l4

recalled consecutively. The mean clustering scores for the nine

groups on Trials 1 and 2 and the total mean clustering for the two
trials are presented in Table 2. The categories used to determine

the amount of clustering were those categories given for study on

the preceding study trial. Thus, conceptual categories were considered
for semantic subjects and rhyming categories for acoustic subjects.
Similarly, for those subjects receiving both organizations of List 1,
rhyming categories were used to determine clustering on one trial and
conceptual categories on the other, depending on which set of categories
had been presented for study on the preceding trial. An analysis of
variance was performed on the clustering data and there were three
significant effects: List 1 Organization, §_(2, 135) = 15.00, p (.001;
Trials, §_(1, 135) = 11.81, p <.001; and Trials by List 1 Organization
interaction, §_(2, 135) = 8.26, p (.001. The results of the recall and
clustering data were very consistent.

The List 1 Organization effect is attributable to the fact that
clustering is higher for semantic subjects than for acoustic and dual
organization subjects. There is also some tendency for the acoustic.
subjects to cluster better than the dual organization subjects, but
this effect is not obtained consistently (i. e., the group receiving
dual organizations of both lists has higher clustering than any
acoustic group). The Trials effect reflects the tendency for
clustering to decrease slightly over trials. The Trials by List 1
Organization interaction is due to the fact that subjects receiving
the dual organization of List 1 decrease in their clustering over
trials while acoustic and semantic subjects tend to have fairly stable

clustering scores.

15

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

mw.H mm. mm. mo.H on. mm. oo.H on. em. H<DD

NN.H mw. Nm. NN.H mm. «m. wN.H mm. om. UHHmDou<

em.H omr mm. om.H mm. mm. Nm.H No. om. UHHZ¢ZMm

Hmuoa N H Hmuoa N H Hmuoy N H DOHumNHammuo

H < D o H H m D o u H H z < z_m m
D o < H umHH
aOHuwuHamwuo N umHH
.mHMHHH 039 mSu How waHHmumoHu HMuoa

amoz onu was N wow H mHMHuH mom mmsouu maHz oau hp mwuoz H umHH Ho waHuoumsHu sow: .N oHDmH

 

16

There are at least two explanations for the decrease in clustering
over trials exhibited by subjects receiving both organizations of List 1.
The dual organization subjects may have experienced some confusion due
to changing the organization of the list of words even though they had
been informed of the change. Secondly, there is some tendency for
subjects in the dual organization groups who studied conceptual
categories on Trial 1 to retain their semantic organization on Trial 2
in spite of the fact that they studied rhyming categories on Trial 2.
This lowers their clustering scores based on rhyming categories.

This difference in clustering indicates that the effect of encoding
multiple attributes appears to depend on which attribute was encoded
first. There are several distinctions which can be made between
subjects who received the semantic organization followed by the acoustic
organization [Both (semantic-acoustic)] and those who received the
acoustic organization followed by the semantic organization [Both
(acoustic-semantic)]. The following observations are based on the
Both (semantic-acoustic) vs. Both (acoustic-semantic) distinction,
regardless of List 2 organization condition. Thus, there are 24
subjects in each of these groups. First, consider the mean recall on
Trial 2. The mean number of words recalled by the Both (semantic-
acoustic) subjects was 21.21 while the Both (acoustic-semantic) subjects
recalled a mean of 25.75. Furthermore, eight of the 24 subjects in the
Both (semantic-acoustic) condition had lower recall scores on Trial 2
than on Trial 1 while none of the Both (acoustic-semantic) subjects
decreased from Trial 1 to Trial 2. But, if the total mean recall on
List 1 is considered, the order of these groups is reversed [41.62 for

the Both (semantic—acoustic) and 38.12 for the Both (acoustic-semantic)].

17

The clustering data yield a similar picture. The mean clustering
on Trial 2 by Both (semantic-acoustic) subjects is .65 while the
corresponding figure for the Both (acoustic-semantic) subjects is .92.
While acoustic organization led to lower clustering than semantic
organization in the independent groups, this fact cannot account for
this great of a difference in the clustering of the dual organization
subjects.

In summary, if one examines the data for subjects in the dual
organization condition, the evidence suggests that the effect on
learning of encoding two attributes depends partially on the order in
which the attributes were encoded. If the more easily encoded attribute
is encoded first, the introduction of a second attribute is more
detrimental than it is if the less easily encoded attribute is encoded
first. However, in either case the amount of learning exhibited by
subjects encoding multiple attributes is much lower than the degree of
learning attained by subjects who consistently encode the dominant
attribute while multiple attribute encoding subjects and subjects
encoding the less dominant attribute consistently exhibit about the

same amount of learning.

List 2 Learning and Organization

Stage 2 recall protocols were scored for the number of List 2
words recalled on each trial of List 2 learning. The mean number of
words recalled on each trial and the total mean recall for each group
are presented in Table 3. An analysis of variance was performed and
the results strongly parallel those of List 1 learning. There was

an effect due to List 1 Organization, F_(2, 135) = 8.25, p (.001, again

18

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Hm.wm mo.ON mN.wH mH.N¢ mm.NN Hw.mH q<.m¢ mo.mN mN.wH H<DD
oo.¢q mo.oN mm.NN 90.nq om.NN on.NH mH.mn ¢¢.mm mN.HN UHHmDou<
mN.¢e om.qN mo.mH oo.nq om.NN cm.NH mN.mm mn.mm om.mN UHHZ¢ZMm
Hmuoa N H Hmuoy N H Hmuoa N H
GOHuwuHawwuo
H 4 D D o H H m D o o < o H H z < z m m H umHH
doHumuHameo N umHH

 

 

 

.mHmHHH 039 mnu How HHmomm HauoH
amoz onu cam N bum H mHmHHH pom mnsouu oaHz onu ma umHHmoom mvuoD N umHH mo Honaaz sums .m mHDMH

 

19

indicating the higher recall of the semantic organization subjects
compared to the acoustic and dual organization subjects who differ

very little. The Trials effect was significant, §_(l, 135) = 224.72,

p ‘.001. Finally, there was a Trials by List 2 Organization interaction,
§_(2, 135) = 18.72, p“.001. The semantic organization subjects improve
their recall more over trials than do the acoustic groups, and both of
these groups increase more than the groups receiving both organizations
of List 2. The group which received the dual organization of both lists
has particularly low recall and very little increase in free recall

from Trial 1 to Trial 2.

The mean clustering scores for Trials 1 and 2 and the total mean
clustering for List 2 learning for the nine groups are presented in
Table 4. An analysis of variance revealed results similar to those
for List 1 clustering and List 2 learning. The List 2 Organization
effect, §_(2, 135) = 8.25, p ‘.001, shows that semantic subjects
cluster more than acoustic subjects who in turn have higher clustering
than dual organization subjects. Once again there is a tendency for
clustering scores to decrease from Trial 1 to Trial 2, §_(l, 135) =
19.29, p ‘.001. The Trials by List 2 Organization interaction,
§_(2, 135) = 17.78, p (.001, indicates that subjects receiving both
organizations of List 2 decrease in their degree of clustering over
trials while semantic and acoustic subjects have fairly stable
clustering scores.

In general, the results found in List 2 learning are consistent
with those based on List 1 learning. It should be noted that there was
no List 1 by List 2 Organization interaction, although the group

receiving both organizations of both lists does exhibit low List 2 learning.

20

 

 

 

 

 

 

 

 

 

 

 

oN.H mm. mm. mw.H Nm. mm. Hm.H ma. om. H<DD

mo.H ow. om. «w.H Nm. Nm. mw.H no. No. 0H9mDoo<

om.H em. om. Hm.H «a. mm. e¢.H Na. Na. 0H9z<ZMm
Hmu09 N H H0u09 N H H0009 N H

:0Humuchwuo

H < D D o H 9 m D o 0 ¢ 0 H 9 z ¢ 2 m m H umHH

 

 

 

 

aoHuMNHamwuo N umHH

 

 

.mHMHu9 039 050 How wsHuoumsHo H0u09
000: one 000 N van H mHmHu9 How mmaouo mcHz 050 he mwu03 N umHH H0 waHumumsHo 000: .q 0Hn09

 

21

Stage 3: Noncued Retroactive Inhibition

The Stage 3 noncued recall protocols were scored for the number
of List 1 words recalled. A Retroactive Inhibition (RI) score was
obtained by taking the difference between the number of words recalled
on the second trial of Stage 1 and the number of words recalled during
noncued recall of List 1 during Stage 3. The mean RI scores for each
of the nine groups are presented in Table 5. An analysis of variance
revealed no significant effects. The groups encoding multiple
attributes of List 1 words do not show less retroactive inhibition than
the other groups. The notion of multiple attribute encoding producing
less forgetting by providing more retrieval cues was not supported.

Excluding all groups receiving both organizations of either list,
the results show that subjects receiving different organizations of
the two lists show less retroactive inhibition than do subjects who
receive the same organization of both lists. That is, the groups
receiving either a semantic organization or an acoustic organization of
both lists (mean RI scores of 9.00 and 9.44, respectively) exhibit
greater retroactive inhibition than the group receiving List 1 semantic
organization and List 2 acoustic organization or the group receiving
List 1 acoustic organization and List 2 semantic organization (mean
RI scores of 5.25 and 6.62, respectively), §_(l, 60) = 6.83, p (.05.

Wood (1971) found that subjects receiving the same organization of
both lists tended to lose words within categories which were recalled
while different organization subjects tended to lose more whole
categories. That is, the words lost between the last trial during
Stage 1 and Stage 3 can occur in two ways: whole categories may be

lost (1. e., no words belonging to a category which was recalled on

 

22

Table 5. Mean RI Scores for the Nine Groups (Stage 3).

 

List 2 Organization

 

 

 

 

List 1

Organization SEMANTIC ACOUSTIC DUAL
SEMANTIC 9.00 5.25 5.31
ACOUSTIC 6.62 9.44 10.25
DUAL 8.50 6.50 8.12

 

 

 

 

 

23

the last trial of Stage 1 are recalled at Stage 3) or parts of
categories may be lost (i. e., at least one but not all of the words
belonging to a category recalled on the last trial of Stage 1 is
recalled during Stage 3). This analysis was made but the results do
not provide a better understanding of the data so the results will
not be reported.

The noncued recall protocols were also scored for the amount of
clustering. The categories used to determine clustering were the
categories which the subject had studied during Stage 1 for the List 1
semantic and acoustic organization subjects. For the subjects
receiving both organizations of List 1, clustering was scored for
both semantic and acoustic categories and the higher clustering score
was used for this analysis. The mean clustering scores for the nine
groups for noncued Stage 3 recall are presented in Table 6. There was
a significant effect due to List 1 Organization, §_(2, 135) = 10.33,

p (.001, indicating that semantic subjects have the highest clustering
scores and dual organization subjects the lowest.

It is interesting to note that 39 of the 48 subjects encoding
multiple attributes of List 1 words show greater semantic clustering
than acoustic clustering during Stage 3. It appears that the semantic
organization was retained for long term recall. Probably because of
this dominant semantic organization, the List 1 dual organization groups
have RI scores similar to List 1 semantic groups. That is, the dual
(List l)-semantic (List 2) organization group and the dual (List 1)-
dual (List 2) organization group have RI scores similar to the semantic
(List l)-semantic (List 2) organization group. Similarly, the dual

(List l)-acoustic (List 2) organization group shows about the same

 

24

Table 6. Mean Clustering Scores for the Nine Groups for

Stage 3 Recall of List 1.

 

List 2 Organization

 

 

 

 

List 1

Organization SEMANTIC ACOUSTIC DUAL
SEMANTIC .95 .97 .98
ACOUSTIC .89 .83 .85
DUAL .76 .84 .80

 

 

 

 

 

25

amount of retroactive inhibition as the semantic (List l)-acoustic
(List 2) organization group. In long term memory, then, the effect
of encoding multiple attributes appears to be similar to the effect

of encoding the more dominant attribute.

Stage 3: Cued Retroactive Inhibition

The mean number of additional words recalled after cueing is
presented in Table 7. Cueing was effective for all conditions. An
analysis of variance on the number of additional words recalled
revealed an effect due to List 1 Organization, F_(2, 135) = 8.64,
p".001. Semantic cues were more effective than acoustic cues. The
acoustic and dual organization groups recalled about the same number
of additional words.

The fact that subjects encoding multiple attributes of List 1
words did not show the greatest cueing effect does not support the
hypothesis that multiple encoding facilitates retrieval by providing
more access routes to the to-be-recalled event. It could be, however,
that having had only one learning trial with each of the attributes,
the association between the retrieval cue and the to-be—recalled event
was weak.

A cued RI score was computed for each subject by taking the
difference between the number of words recalled on Trial 2 or Stage 1
and the total number of List 1 words recalled during Stage 3 (noncued
List I recall plus the number of additional words recalled after cueing).
The mean cued RI scores for each of the nine groups are presented in
Table 8. A negative score indicates that more List 1 words were

recalled during Stage 3 than were recalled on the second trial of Stage 1.

 

26

Table 7. Mean Number of Additional Words Recalled by the
Nine Groups after Cueing during Stage 3.

 

List 2 Organization

 

 

 

 

List 1

Organization SEMANTIC ACOUSTIC DUAL
SEMANTIC 6.62 7.69 7.44
ACOUSTIC 4.19 2.94 4.75
DUAL 3.94 6.31 4.38

 

 

 

 

 

Table 8. The Mean Cued RI Scores for the Nine

27

Groups (Stage 3).

 

 

 

 

 

 

List 2 Organization
List 1
Organization SEMANTIC ACOUSTIC DUAL
SEMANTIC 2.38 -2.43 -2.13
ACOUSTIC 2.44 6.50 6.50
DUAL 4.36 0.19 3.74

 

 

 

 

 

28

An analysis of variance on the cued RI scores revealed a List 1
Organization effect, F_(2, 135) = 12.80, p (.001, and a List 1
Organization by List 2 Organization interaction, §_(4, 135) = 3.70, p <.05.
The List 1 Organization effect reflects the fact that List 1
semantic subjects show the least amount of retroactive inhibition while
the List 1 acoustic subjects show the most. In fact, there is essentially
no forgetting with semantic cues, even though the retrieval cues
(category names) in this study were not as obvious as those used by
Tulving and Psotka (1971). The cueing effect once again verifies the
fact that there is information available in storage which is not
accessible without the appropriate cues. Semantic cues appear to be
more effective in providing accessibility to the event in storage than
are acoustic cues.
The List 1 Organization by List 2 Organization interaction is more
difficult to interpret. If all groups receiving the dual organization
of either list are excluded from consideration, the groups receiving
the different organization of the two lists have less retroactive
inhibition than the groups receiving the same organization of both
lists. An analysis was made of whether the additional words recalled
after cueing belonged to new categories or to categories which had
been recalled on the noncued List 1 recall trial during Stage 3. Since
the additional words belong to both new and old categories, the analysis

did not enhance the understanding of the data and will not be presented.

DISCUSSION

The results of this experiment did not support the hypotheses
suggested by Underwood (1969). Considering either the learning or
the cueing data, the conclusion is the same: encoding multiple
attributes does not facilitate memory for an event by providing more
access routes to the stored event. The degree of original learning
resulting from encoding two attributes was less than the degree of
learning following semantic encoding but was approximately equal to
that following acoustic encoding. The cueing data provide the most
direct test of the proposed facilitation mechanism of multiple
encoding. If multiple attribute encoding provides more means of
access to the stored material, the List 1 dual organization groups
should have had the largest cueing effect. This was not the case,
however. The semantic organization subjects recalled the greatest
number of additional words after cueing despite the fact that they
exhibited less retroactive inhibition.

The implications of the retroactive inhibition data are not as
clear. Subjects who received the dual organization of List 1 followed
by a Single organization of List 2 should have shown less forgetting
than subjects who received a single organization of List 1 and the dual
organization of List 2. In the first case subjects had one retrieval
route (1. e., attribute) which was unique to List 1 and was, therefore,

subject to less interference than the attribute common to the two lists.

29

30

This "unique" attribute should have been an effective means of
retrieval of List 1. Subjects in the latter case did not have a
retrieval route which was unique to List 1 and so should have had
more difficulty in retrieving List 1 after List 2 learning. However,
there were no differences in the amount of forgetting exhibited by
these groups.

On the other hand, the retroactive inhibition data on the same vs.
different organizations of the two lists (if all dual organization
groups are excluded) lend some support to the hypothesis that the less
the attributes of one event are a part of another event, the less the
interference between the two events. These data approach the limiting
case where the attributes of the two events are entirely distinct.

The present finding that similar organizations of two lists leads to
greater forgetting than different organizations has already been well
documented (Shuell, 1968; Thompson and Poling, 1969; Watts and Anderson,
1969; Winograd, 1968; Wood, 1970, 1971).

The cueing data also replicated the well known finding that
information is available in storage but not accessible unless retrieval
cues are provided (cf. Tulving and Pearlstone, 1966). The cueing
effect in the present study was fairly large. In fact, when semantic
cues were provided, there was essentially no forgetting. Another way
of looking at the availability-accessibility distinction lies in
trace-dependent and cue-dependent forgetting. Trace-dependent
forgetting is a consequence of the deterioration of event information
in storage while cue-dependent forgetting results from faulty access
mechanisms. There is very little evidence for trace-dependent

forgetting in the data under consideration. The fact that there was

31

essentially no retroactive inhibition in the List 1 semantic organization
groups after cueing reflects cue-dependent forgetting. It appears
that the major variable in the failure to retrieve stored event
information is the effectiveness of the cues which can be generated
by the subject or are provided by the experimenter.

It is possible that the effect of encoding multiple attributes
depends on the specific attributes encoded. Since most adult
processing involves associative verbal attributes, these attributes are
probably more salient and may be more easily encoded. Thomson and
Tulving (1970) have shown that weak associates of a word can be
effective retrieval cues if they are encoded with the word during study
trials. If, however, an event is encoded in the presence of two cues
differing greatly in associative strength to the event, it is doubtful
that the weak cue would be utilized as a retrieval route to the to-be-
recalled event. Perhaps if two attributes had been chosen which did not
differ so much in salience or ease of encodability, both cues would have
been more effective than a single retrieval cue.

An issue of considerable interest is why multiple cues are not
more effective than a single cue. Bower (1970) has argued that a
consistent organization (single set of retrieval cues) is important
while Postman (1972) indicates that subjects may use multiple retrieval
routes. If multiple retrieval routes are used, it seems likely that
the different retrieval routes are independent of each other. Martin
(1971) has recently argued that for any stimulus A, there are a number
of attributes which may be sampled (encoded) on any trial. If a subject
learns to respond with B after seeing A and then has to learn to respond

with C after seeing A, the attributes of A that are encoded in the two

32

situations will be different. Moreover, the evidence indicates that the
"B" attribute of A is independent of the "C" attribute of A. Thus, when
different attributes of an event are processed on successive trials,
the encoding of the event on Trial n is likely to be independent of the
attribute encoded on Trial n + 1. This conclusion is supported by the
present study in that there was almost no improvement in learning over
successive trials when multiple attributes were encoded. In fact,
subjects who processed acoustic attributes after semantic attributes
often had lower recall on the second trial than on the first.
Unfortunately, accepting the view that different retrieval routes
to the same event are independent of each other does not afford any
insight, to this writer, regarding the plausibility of the consistent
organization versus multiple organization view of retrieval processes.
The evidence seems to favor a consistent organization view if retrieval
cues are not provided (cf. Bower, 1970) since a single presentation
order for a free recall list generally yields higher recall than having
multiple presentation orders. If a cued recall procedure is used there
is no evidence to support the view that multiple cues are better than
single cues (cf. Tulving and Osler, 1968). The present investigation
of the number of attributes encoded is consistent with the earlier
cueing studies and the studies in which presentation order was
manipulated. Two opportunities to encode a single attribute result
in as much or more recall than the amount of recall obtained when a
subject encodes a different attribute of the same event on successive
trials. Although there is no question that subjects can develop multiple

retrieval paths for a given set of events or encode multiple attributes

33

of the same event, there is no good evidence to recommend this

strategy over a single retrieval plan or single attribute strategy.

REFERENCES

REFERENCES

Bobrow, S. A. Memory for words in sentences. Journal of_Verbal
Learning_and Verbal Behavior, 1970, 9, 363-372.

 

Bower, G. H. Organizational factors in memory. Cognitive Psychology,
1970, 1, 18-46.

 

Bower, G. H., Lesgold, A. M., and Tieman, D. Grouping operations in
free recall. Journal of_Verbal Learning_and Verbal Behavior,
1969, 8, 481-493.

 

Light, L. L. and Carter-Sobrel, L. Effects of changed semantic
context on recognition memory. Journal o£_Verbal Learning_and
Verbal Behavior, 1970, 9, 1-11.

 

 

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, Pp. 327-372.

 

 

Martin, E. Verbal learning theory and independent retrieval phenomena.
Psyohological Review, 1971, 78, 314-332.

 

Postman, L. A pragmatic view of organization theory. In E. Tulving
and W. Donaldson (Eds.) Organization of_Memory. New York:
Academic Press, 1972, Pp. 3-48.

 

Shuell, T. J. Retroactive inhibition in free-recall learning of
categorized lists. Journal o£_Verbal Learning and Verbal Behavior,
1968, 7, 797-805.

 

Thompson, C. P. and Poling, D. R. Free recall of successive lists:
Effect of category size and category repetition. Psychonomic
Science, 1969, 16, 201-202.

 

Thomson, D. M. and Tulving, E. Associative encoding and retrieval.
Journal of Experimental Psychology, 1970, 86, 255-262.

 

Tulving, E. and Osler, S. Effectiveness of retrieval cues in memory
for words. Journal of_Experimental Psychology, 1968, 77, 593-601.

 

34

35

Tulving, E. and Pearlstone, Z. Availability versus accessibility of
information in memory for words. Journal of_Verbal Learnigg and
Verbal Behavior, 1966, 5, 381-391.

 

Tulving, E. and Psotka, J. Retroactive inhibition in free recall:
Inaccessibility of information available in the memory store.
Journal of Experimental Psychology, 1971, 87, 1-8.

 

Underwood, B. J. Attributes of memory. Psychological Review, 1969,
76, 559-573.

Watts, G. H. and Anderson, R. C. Retroactive inhibition in free recall
as a function of first-and second—list organization. Journal of_
Experimental Psyohology, 1969, 81, 595—597.

 

Winograd, E. List differentiation, recall, and category similarity.
Journal of_Experimental Psychology, 1968, 78, 510-515.

 

Wood, G. Organization, retroactive inhibition, and free recall learning.
Journal of_Verbal Learning and Verbal Behavior, 1970, 9, 327-333.

Wood, C. Activity, organization, retroactive inhibition, and free recall.
Journal of_Verbal Learning_and Verbal Behavior, 1971, 10, 116-122.

APPENDIX

FOOD

steak
bean
pie
grape

TRAVEL

plane
taxi
sleigh
submarine

BODY PART

thigh
brain
spleen
elbow
knee

APPENDIX 1

List 1

Semantic Organization

LANDSCAPE

sky
ravine
bay
valley

BIRD

sparrow
crane
cuckoo
flamingo

ANIMAL

snake
bear
whale
donkey

CLOTHING

shoe
negligee
veil

HOUSEHOLD

chair
shampoo
diaper
drape
patio

WEATHER

hail

rain
earthquake
thunder

PEOPLE

rabbi
millionaire
male
chauffeur

COUNTRY

Paraguay
Mexico
Chile
Peru

@2112.

mohair
crepe
leather

A (SAY)

Paraguay
sleigh
bay
negligee

AR (STARE)

mohair

bear

chair
millionaire

ER (STIR)

leather
thunder
diaper
chauffeur

Acoustic Organization

AN (CHAIN)

 

rain

plane
crane
brain

E (BEE)

taxi
Chile
donkey
valley
knee

0 FLOW)

Mexico
sparrow
elbow
flamingo
patio

37

List 1

2L (JAIL)

hail
veil
whale
male

EN (DEAN)

bean
spleen
submarine
ravine

00 NEW

Peru
cuckoo
shoe
shampoo

2K (MAKE)

steak
earthquake
snake

crepe
drape
grape

I (CRY)

rabbi
thigh
pie
sky

FOOD

cake
pear
sardine
grain
stew

TRAVEL

train
ferry
subway
canoe

BODY PART

nape
eye
hair

Semantic Organization

LANDSCAPE

lake
scene
trail
fir
sea

BIRD

crow
bluejay
quail

ANIMAL

ape
mare
kangaroo
otter

List 2

CLOTHING

cape
tuxedo
tie
slipper

EQEEEEEE

screen
stair
pail
chimney

WEATHER

hurricane
gale
tornado
dew

PEOPLE

fake
SPY
nephew
queen
padre,

COUNTRY

Norway
Monaco
Italy
Spain

CLOTH

jersey
calico
pique

A SAY)

pique
Norway
subway
bluejay
padre

KR (STARE)

 

mare
stair
pear
hair

ER (STIR)

slipper
fir
Otter

Acoustic Organization

AN (CHAIN)

 

hurricane
train
Spain
grain

E (BEE)

jersey
ferry
Italy
chimney
sea

0 (FLOW)

calico
crow
Monaco
tornado
tuxedo

39

List 2

‘AL (JAIL)

gale
quail
trail
pail

‘

EN DEAN

sardine
scene
screen
queen

00 (NEW)

dew
canoe
nephew
kangaroo
stew

'AK MAKE)

cake
lake
fake

a"

AP TAPE)

ape
nape
cape

I CRY)

SPY
eye
tie

2 8679

I
I
I
.ll’
I
I!
ll
'4
I
I
I,
[I
I!
I!
I
’l
I
1!
l
'5'
II
I,
I
I
!
II"
I,

3 1293 0314

llIIHHHIHUIIIHIHHH