THE EFFECTS OF IMPLICIT ASSOCIATIONS AND
INSTRUCTIONAL SET ON INTERLIST TRANSFER
IN VERBAL DISCRIMINATION LEARNING

Thesis for I‘he Degree of M. A.
MICHIGAN STATE UNIVERSITY

CareI Boice

I966

[J LIBRARY

M iphigan Sun
Unfit-tit]

   
    
   

THESIS

  

 

 

33“.} II éI. {LU

 

THEE

 

 

ABSTRACT

THE EFFECTS OF IMPLICIT ASSOCIATIONS AND
INSTRUCTIONAL SET ON INTERLIST TRANSFER
IN VERBAL DISCRIMINATION LEARNING

by CAROL BOICE

This experiment was designed to determine whether
implicit associations between words in successive lists
of word pairs would be utilized as cues in responding
in a verbal discrimination task. In addition, the
importance of instructions in determining the set of the
gs was investigated.

Eighty introductory psychology students were first
required to learn to a criterion of three errorless
trials a list of 15 pairs of words. in which one word of
each pair had been arbitrarily designated correct by g.
They were then required to learn a second list which
contained some words which were bidirectional or
unidirectional associates of the correct words from the
first list. For one eXperimental condition the associates
in the second list were correct, for the other they were
incorrect. A control group learned two unrelated lists.
Half of the gs in each experimental condition were
instructed about the presence of the associates in the
second list and informed that these words would be correct

or incorrect. depending on the condition. The other

Carol Boice

§s were simply told they would be learning another list.
All gs were given 10 trials on list 2, and the number
of errors per trial served as the measure of performance.
For the instructed §s correct associate words in
list 2 led to slight positive transfer while incorrect
associates produced considerable negative transfer.
The same effects occurred in the uninstructed conditions
but were less pronounced. These differences between the
groups were apparent on the first trial of the second
list, and the error curves converged over the 10 trials.
The set provided by the complete instructions was
apparently crucial in determining the effects of the
associate words, since the uninstructed §s did not
perform significantly differently from the control §s.
These findings were discussed in terms of a frequency

hypothesis of verbal discrimination learning.

Approved WC. M

Committee Chairman

DateW 9, /955

 

mssr

 

l
I
i
n ....i

THE EFFECTS OF IMPLICIT ASSOCIATIONS AND
INSTRUCTIONAL SET ON INTERLIST TRANSFER
IN VERBAL DISCRIMINATION LEARNING

By

'Carol Boice

A THESIS

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

MASTER OF ARTS

Department of Psychology

1966

ACKN OWLEDGE NENTS

The author gratefully acknowledges the assistance
given her by her major professor, Doctor David C. Baskin.
Without his assistance in the development of this
thesis this study would not have been possible. The
author is also indebted to the other members of her
guidance committee for their help and encouragement:

Doctors William T. Stellwagen and Terrence M. Allen.

11

mes:-

 

 

 

ITABLE OF CONTENTS

Page

INTRODUCTION . . . . . . . . . . . . . .
PROCEDURE . . . . . . . . . . . . . . .
Subjects . . . . . . . . . . . . . .

Materials and Apparatus . . . . . .

CDVVVH

Procedure and Instructions . . . . .
RESULTS . . . . . . . . . . . . . . . . 11
DISCUSSION . . . . . . . . . . . . . . . 20
SUMMARY . . . . . . . . . . . . . . . . 2h
REFERENCES . . . . . . . . . . . . . . . 26

APPEDIDICES O O O O O O O O O O O O O O O 28

111

TABLE
I.

II.

III.

IV.

VI.

VII.

VIII.

LIST OF TABLES

Summary of Analysis of Variance
Comparing the Four Experimental
Groups on Performance on List 2

Summary of Multiple Comparisons
between Experimental Groups
Using Duncan's Multiple Range
TeStO O O O O O O O O O O O O 0

Summary of Analysis of Variance
Comparing the Uninstructed
and Control Groups. . . . . . .

Summary of Analysis of Variance
Comparing the Instructed and
Control Groups. . . . . . . . .

Analysis of Variance of Number
of First Trial Errors for All
Groups. . . . . . . . . . . . .

Summary of Multiple Comparisons
on Number of First Trial Errors
for All Groups Using Duncan's

Multiple Range Test . . . . . .

Number of Errors on Bidirectional

and Unidirectional Associates
for the Experimental Groups . .

Summary of Analysis of Variance

Comparing Bidirectional and
Unidirectional Associates . . .

iv

12

1h

15

16

17

18

19

THESI!

 

 

LIST OF FIGURES

FIGURE Page

1. Mean errors for the four Experi-
mental groups on List 2. . . . . . 13

2. Mean errors for U and C condi-
tions on List 2. . . . . . . . . . 15

3. Mean errors for C and I condi-
tions on List 2. . . . . . . . . . 16

LIST OF APPENDICES

APPENDIX Page

A. Associate Word Pairs. . . . . . 29

B. Instructions for List 1 . . . . 32

vi

INTRODUCTION

Verbal discrimination has frequently been viewed as
the first stage in a two process model of paired-associate
and serial learning (Runquist and Freeman, 1960).

As such it supposedly consists of a familiarization

with the available responses. The second stage of the
process involves learning the order of the responses

or pairing each with the apprOpriate stimulus. From

this model it has been hypothesized (Battig, Williams

and Williams, 1962) that verbal discrimination (VD)
experience with a list should facilitate paired- associate
(PA) learning of the same words. since the first part

<of the process had already been accomplished before
switching to the PA list.

Several studies have tested this hypothesis. but
the results have been inconsistent. Battig. Williams
arui Williams (1962) report that they found no positive
transfer from VD to PA learning, although VD words paired
incorrectly for the PA list produced some negative
trazmsfer. A VD retention test indicated that more
rlgrrt words than wrong words were learned. On the
other hand. Spear. Ekstrand and Underwood (1961») found
not only negative transfer with inappropriate pairings
but Positive transfer with appropriate pairings on the
PA 1131:. They suggest that the differences between

their' results and those of Battig. Williams and Williams

1

was:

 

 

may be due to the fact that Eattig, Williams and Williams
used nonsense syllables and Spear, Ekstrand and
Underwood used low frequency words. With low meaningful
pairs of nonsense syllables, VD involves recognition
with very little association developing between
syllables, but with meaningful word pairs a much
stronger association develops between the words. This
association between words resulted in positive transfer
for appropriate pairs and negative transfer for
inappropriate pairs on the PA task. It is clear,
however, that the incorrect words were not independently
learned in the VD list in either study.

McClelland (1942) tested the possibility that both
correct and incorrect reSponses were learned, but
incorrect words were forgotten more quickly. His gs
learned a VD list to a 15 out of 20 correct criterion
and were then shifted, without being informed of the
change, to a new list containing the words from list 1
paired with new words. Words correct on list 1 were
still correct, and those that had been wrong were still
wrong. On the second list, pairs with transferred wrong
words were significantly more difficult to learn than
those with transferred correct words. McClelland
concluded that the only acquired response connected with
the incorrect words was a weak avoidance--only the
correct words were actually recognized and remembered.

Similarly, Saltz (1964) has recently suggested that

masts

 

 

 

 

learning a VD list may involve not only acquisition of
the correct words but also an inhibition of the
tendency to respond with the incorrect words.

The most clearly elaborated hypothesis about VD
learning has been presented by Underwood, Jesse and
Ekstrand (1964) and Ekstrand, Wallace and Underwood (1966).
This hypothesis is based on data collected in an eXperi-
ment (Underwood, Jesse and Ekstrand, 1964) which
differed from McClelland's (1942) in that 1) the §S
were informed of the changes made between the first and
second lists, 2) the VD lists were not mixed (instead of
introducing both correct and incorrect new words in
list 2, the new items were either all correct or incorrect),
and 3) SS learned the first list to three successive
errorless trials. The results indicated that replacing
the incorrect words produced no performance decrement
on the second list, and that replacing the correct
words resulted in a small decrement on trial 1 of list
2 but extremely slow improvement over the 10 trials.

Underwood, Jesse and Ekstrand interpreted these
these results as indicating that discriminations depend
initially on the relative frequency of the members of
the pairs. Each time S looks at, or responds with, a
word one frequency unit is added to that word. On the
first trial of list 2 the transferred word has more
frequency units than the new word, but words that were

Correct on list 1 have more units than those that were

THES‘

 

 

 

u,

incorrect. This differential frequency provides a cue
for responding. If the second list includes new

wrong words the differential between the transferred
correct words and the new words is maximal. Thus, there
is no decrement on trial 1 of list 2 and each further
trial only increases the difference by adding more units
to the correct word than to the new word. When new
correct words are introduced on list 2, there is a
smaller but still substantial difference between the
frequencies of the old and new words and the decrement is
small. However, the trials on list 2 in this case add
more units to the new than the old words and eventually
differential frequency breaks down as a cue and the S
must turn to some other means of making the discrimination.
An analysis of VD learning data to determine if guessing
correctly or incorrectly on trial 1 affected performance
on following trials has provided moderate support for the
frequency hypothesis (Dominowski, 1966).

Bousfield, Whitmarsh and Danick (1958) have
suggested that the presentation of a verbal unit will
elicit implicit associated reSponses related to the
given unit, and such implicit associations have been
shown to mediate positive transfer in PA learning
(Bugelski and Scharlock, 1952; Russell and Storms, 1955).
In VD learning associated words have been used to add
frequency units to some of the words in the list

(Ekstrand, Wallace and Underwood, 1966). When both words

THES

 

 

 

5

in an associate pair were correct in the VD task learning
was facilitated, presumably because each time one word
of the associate pair appeared a frequency unit was also
added to the other word through implicit association.
If the associated words were both correct and incorrect
the association produced interference and performance was
inferior to that of a control group which learned a list
without any associate pairs. These effects, although
significant, were small,and a group receiving incorrect
associates did not differ from the control group.

Ekstrand, Wallace and Underwood did not inform
their S3 of the presence of the associates in the VD list,
and because these associations are apparently automatically
produced this should not have affected the results.
However, even when some identical words are used in
two VD lists complete instructions may produce greater
transfer than switching to the second list without
instructions (Underwood, Jesse and Ekstrand, 1964;
McClelland, 1942). This difference may have also been
at least partly due to the fact that McClelland used
a mixed list, but it seems probable that set factors are
important. Schwartz (1963), for example, discusses the
importance of instructional set in PA learning, pointing
out that such learning is not fixed by the classical
tYpes of variables, but is also influenced by the set
§S are given.

This experiment attempted to test the frequency

6
hYpothesis of VD learning by using implicit association
to manipulate the differential frequency between words
(n1 trial 1'of a second list constructed partly of words
vflqich were associates of words from a first VD list.
TUde importance of instructional set in this type of task

was also investigated.

THE

 

NETR OD

Subjects - The §S in this experiment were 80 under—
graduate students enrolled in an introductory psychology
course at Mighigan State University. None had previously
participated in a verbal learning study. One _S_ who
failed to learn the first list after 25 trials was

excluded from the experiment.

Materials and Apparatus - Part of the words used in

It 4. L'v.-&-.fli¢‘w--m-1-'c-IuJaj-"rm
.I
‘x - -
'. ;

constructing the VD lists for this study consisted of 10
bidirectional and 10 unidirectional associate pairs. These
pairs were taken from Bilodeau and Howell (1965) and Mink
(1957) and are presented, with their associative strengths,
in Appendix A. The remainder of the words were from
Bilodeau and Howell (1965) and were not associated with words
in the associate pairs or each other (no associative value
Was greater than .04).

Four different lists were constructed, each consisting of
15 pairs of words. Lists 1A and 1B contained 10 pairs of words
that were comprised of a word from one of the associate pairs
(5 unidirectional and 5 bidirectional in each list to control
for possible differential effects of directionality) and a
nonassociate word. For both groups the associate words were
deSfignated correct. The other five pairs, which were added
to increase the length of the list contained two nonassociate
IWards. List 1C was composed of 15 pairs of nonassociate
words. In an effort to keep these three lists as

comparable as possible the same nonassociate words
7

aDpeared in all three lists, although 10 additional pairs
were necessary to complete list 10.
List 2, which was the same for all _S_s, also
contained 15 pairs, 10 of which were composed of the
second halves of the associate pairs. For each of these

pairs the associate of a 1A word was coupled with the

.1‘““'
q.

associate of a 1B word, and the associate of the 1A word

was always correct. The other five pairs in list 2

- --— 7—-..w1 n—l—nwuzu

consisted of 10 new nonassociate words. For the

unidirectional pairs, the stronger response word always

II

appeared in the second list.

The lists were presented on an MTA 100 Scholar,
modified for presentation of a continuous loop and with
the off/on control mounted on a clipboard so E could
Control the machine from a position behind the _S_. The
Window on the machine was reduced in size and divided
SO that there were four 1" x 3/4" windows. Three of

1'SI'Iese were always covered so § could see only one list at

a time.

Procedure and Instructions - Each pair of words was
presented, with one word typed directly above the other
for a two second period. Next, the correct word
appeared alone in the window for two seconds. A two
SeBond ITI followed presentation of all 15 pairs, and
during this period a row of asterisks appeared in the

Wl1‘1<3.ow. There were two randomly arranged rotations of

each list and each word appeared once on the top and
once on the bottom for the two rotations.

Ss were assigned to one of five groups in order of
their appearance for the eXperiment. After being
seated in front of the teaching machine, all Ss were
read identical instructions describing the task, which
was simply to learn which word in each pair was correct
(see Appendix B). Each S then learned the appropriate
first list to a criterion of three successive errorless
trials. At this point the card covering the second list
was moved to cover the list S had just learned, and
S was read one of two different sets of instructions.

All the S3 who had learned list 1C and half of
those who had learned 1A or 1B received minimal
information about list two as follows:

The second part of the eXperiment is similar

to the first. You will be asked to learn a

new list of words which will be presented in

exactly the same way as the first list.

Once again, please tell me which word you

think is correct before it appears alone.

This time we will stop after 10 trials.
fThe remainder of the 1A and 1B SS were given the
:following detailed instructions about the relation
‘between the words in the first and second lists:

The second part of the experiment is

similar to the first. You will be asked to

learn a new list of words which will be presented

in exactly the same way as the first list.

Some of these new pairs of words will contain

words that are associated with the words that

were correct in the first list. By associated,

I mean any kind of connection between the
words. For example, they can be synonyms

lO

"ocean-sea", opposites "boy-girl", or words

that commonly appear together "salt-pepper".

These associated words that will appear in

this next list will always be correct (incorrect).
Do you understand?

Once again, please tell me which word

you think is correct before it appears alone.

This time we will stOp after 10 trials.

The performance measure used was the number of
errors per trial for the 10 second-list trials.

The design of the experiment, thus, involved the
two variables of Instructions and Correctness of
associates on list 2, each of which had two levels
which were combined factorially giving the following
four experimental groups with 16 SS in each:

I-AC -- Instructed with second-list associates

correct.

I-AW -- Instructed with associates wrong.

U-AC -- Uninstructed about relation between first

and second lists with associates correct.

U-AW -- Uninstructed with associates wrong.

In addition there was a control group (C) with 16 SS

who learned two unrelated VD lists. There were 9 male

and 7 female subjects in each of the five groups.

RESULTS

The mean number of trials to the criterion of three
successive errorless trials on the three first lists
ranged from 7.93 to 8.81, with an overall mean of 8.22.
An analysis of variance indicated that these differences
were not significant (_F_ = .44, d.f. = 2, 79).
Apparently the different lists were of comparable
difficulty.

For all analyses on list 2 responses on the five
pairs that contained nonassociate words were not
included since they were used merely as filler pairs to

increase the length of the list.

Comarison of experimental groups -- To determine
the effects of the Association and Instruction variables

the control Ss were excluded and the four experimental

groups were compared. An analysis of variance (Table I)

indicated that the correctness of the associate words in
list 2 had a significant effect on performance (p_<.01).
Although the main effect of the Instructions variable
was not significant (S(1.0), the Instructions x
Association interaction was significant beyond the .01
level. Thus, the effect of the instructions was to
accentuate the differences between the AC and AW conditions.
Inspection of Figure 1 will clarify the nature of

the significant Trials x Instructions and Trials x

11

12

Table I. Summary of Analysis of Variance Comparing
the Four Experimental Groups on Performance on

 

 

 

 

List 2.

Source d.f. Mean Square 2

Between Ss
Association (A) 1 49.0 22.3*
Instructions (I) 1 .50 .23
A x I 1 22.8 10.4*
Error 60 2.2

Within Ss
Trials (T) 9 99.9 153.7*
T x A 9 2.7 4.2*
T x I 9 3.7 5.7*
T x A x I 9 1.1 1.7
Error 540 .65

*3 (.01

Aissociation interactions. Aside from the first two
ixrials, the I-AW group made considerably more errors than
the other three groups, and did not reach the performance
level of the other groups in the 10 second list trials.
The I-AC group. on the other hand, made fewer errors
thari the other groups although this difference was

small and disappeared when the U groups reached

the same, nearly errorless,level of performance on trial 4.

13

5.0 4 Ar———1§ I-Aw

A—A I—AC
\ o---—--o U-AW
4.0 I

MEAN
ERRORS 3'0

 

 

 

TRIALS

Figure 1. Mean errors for the four Experimental
groups on List 2.

frhe U groups generally fell between the two I groups with
'the U-AC group having made slightly fewer errors than
the U-AW.

A Duncan multiple range comparison (Winer, 1962),
0f"these eXperimental groups indicated that all four
SITNJPB were significantly different from each other

(2 (.01) (Table II).

jgpinstructed and control groups -- Learning curves

cOmpeu-ihg; the U-AC and U-AW groups with the Control group

14

Table II. Summary of Multiple Comparisons between
Experimental Groups Using Duncan's Multiple

Range Test.

 

 

Group I-AC U-Ac U-Aw I-AW
Total 76 128 156 225
Errors

I-AC -- 52* 80* 149*
U-AC -— 28* 97*
U-AW -- 69*
*p_<.01

on List 2 learning are shown in Figure 2. An analysis
of variance, summarized in Table III, indicated that
these groups were not significantly different from each
other, but that the Trials x Association interaction
was significant beyond the .01 level. On trial 1 of
list 2 both the U-AC and U-Aw groups showed some
negative transfer when compared to the C group, but

‘these differences had disappeared by trial 3.

Instructed and control groups -- A comparison of
tune I-AC, I-AW, and C groups is shown graphically in
Figure 3. Analysis of variance yielded an 1: value for
the Association main effect that was significant beyond
the: .01 level. The summary of this analysis appears in
Talile IV. Again the Association x Trials interaction
'WaS :significant, with the I-AC group showing positive

traxusfer on trial 1 and the I-AW group showing not only

15

c~---o U-AN

  
    

e-u—e U-AC

D————fl C

MEAN
ERRORS

 

 

Figure 2. Mean errors for U and C conditions
on List 2.

Table III. Summary of Analysis of Variance Comparing
the Uninstructed and Control Groups.

 

Source d.f. Mean Square E
Between Ss
Association (A) 2 2.8 1.83
Error 45 1.6
Within Ss
Trials (T) 9 89.9 INS-5*
T x A 18 1.6 2.5*
Error 405 .63

__

*g<.01

16

A----A I-AW
A----‘ I-AC

MEAN 3-0 I
ERRORS

 

 

 

Figure 3. Mean errors for C and I conditions
on List 2.

'Table IV. Summary of Analysis of Variance Comparing the
Instructed and Control Groups.

 

 

 

Source d.f. Mean Square S
Between Ss
Association (A) 2 37.5 21.6*
Error 45 1.7
‘Within Ss _
Trials (T) 9 57.3 91.0*
T x A 18 1.6 2.5*
Error 405 .63

*
p‘(.01

17

initial negative transfer, but also a continued
inferior performance over the 10 list 2 trials when
compared to the I-AC and C groups.

First trial differences -- The nature of the
significant Association x Trials interactions can be
further clarified by an analysis of the number of errors
on trial 1 for each group. This analysis was significant
beyond the .05 level (Tablei’).

Table V. Analysis of Variance of Number of First Trial
Errors for All Groups.

 

 

 

 

Source d.f. Mean Square ‘S
Between Groups 4 7.7 2.9*
Error 75 2.6

*p(.05

A Duncan multiple range comparison (Winer, 1962)
Iwavealed that only the first trial difference between
the I-AC and U-Aw groups was significant (TableVI ).

AE! has already been mentioned, the I-AC group was

the only one to show positive transfer when compared
with the C group. The U-AW group, in addition to
‘Uaiang unaware of the presence of the associates, was
unable to use any cues that might have been unverbalized
b{icause the contingencies were reversed. As a result

thJJS group made the greatest number of errors on trial 1.

Table VI. Summary of Multiple Comparisons on Number
of First Trial Errors for All Groups Using
Duncan's Multiple Range Test.

 

 

Group I-AC C I-AN U-AC U-Afl

Total 47 60 65 68 77

Errors

I-AC _- 13 18 21 30*

C -- 5 8 17

I-Aw -- 3 12

U-AC -_ 9

*2 (.05

Bidirectional vs. unidirectional associates —- The
lTumber of errors on bidirectional and unidirectional
:associates for each group is shown in Table VII.

'Taifle VII. Number of Errors on Bidirectional and Uni-
directional Associates for the Experimental

 

 

Groups.
93232 Bidirectional Unidirectional
I-AC 41 35
I-AW 100 125
U-AC 57 71
U-AW 86 70

An Exnalysis of variance (Table VIII) indicated that the
Association x Instructions x Directionality interaction
was significant (2 (.Cl). Examination of Table VII
Suggests that the unidirectional associations were stronger

they: the bidirectional associations in that the unidirec-

19

Table VIII. Summary of Analysis of Variance Comparing
Bidirectional and Unidirectional Associates.

 

 

 

Source d.f. Mean Square E

Between SS
Association (A) 1 24.48 22.5I
Instructions (1) 1 .23 .21
A x I 1 11.43 10.5*
Error 60 1.09

Within‘Ss
Directionality (D) 1 .23 .72
Trials (T) 9 46.59 145.6*
A x D 1 0.00 0.0
A x T 9 1.79 5.6*
I x D 1 .34 1.1
I x T 9 2.34 7.3*
A x I x D 1 2.91 9.1*
A x I x T 9 .50 1.6
A x D x T 9 .27 .84
I x D x T 9 .27 .84
A x I x D x T 9 1.85 5.78“
Error 1140 .32

{-

p(.01
151<>nal associates led to fewer errors in the I-AC group
andmore errors in the I-Aw group. This trend, however,

was reversed for the uninstructed groups.

«x.—
on»

nNn
s‘ I

 

v

I‘”
'4:

fi..|

II.-

the
.n.

A-
.n‘

DISCUSSION

Generally the effects of manipulating frequency by
implicit association were in line with predictions that
could be made on the basis of the frequency hypothesis
of Underwood, Jesse and Ekstrand (1964). If implicit
associations occurring to words on list 1 were effective
in producing differential frequencies of the words on
list 2, this difference should have been apparent on the
first trial of list 2. 0n trial 1 of list 2 the large
frequency differential between the new and associate
words should make discrimination relatively easy. When
the associate words are correct on list 2 performance
should be nearly perfect, if the associates are
incorrect a slight decrement might be expected. Over
the 10 trials on list 2 the group with the correct
Vassociates should make almost no errors, but if the
associates are incorrect the frequency differential
b6tween the correct and incorrect words should decrease
and fflrmlly reverse. This group, then, should make
few errors on trial 1, an increasing number of errors
on tune next several trials as discrimination becomes more
difficult, and finally should. begin to improve slowly
as the correct words build up more frequency units than
the iducorrect words.

'This is essentially what did happen, except that

the ilrltial decrement for all four experimental groups was

20

 

21

greater than was predicted. It is probable that this
first trial decrement was due to the failure of some of
the implicit associations to occur in each S.
Simply considering the associative strengths of the pairs
(all less than .75), it is unlikely that all the associate
words would be effective in eliciting an association
in any given S. The fact that the instructed groups
made fewer errors on trial 1 of list 2 than either of the
uninstructed groups suggests that this might be the case.
That the U-AW group made more errors on trial 1 than
the other groups can be explained as the result, not
only of failures to associate, but of having also to learn
that the contingencies were reversed.

As expected, the I-AC appeared to show some positive
transfer, although the difference between the I-AC and
C groups on the first trial did not reach significance.
The U-AC group did not show any significant transfer effects,

but after three trials performance for both AC groups and

the C group was nearly perfect. Elimination of errors was

Significantly slower for the I-AW group than for the
other groups, but the increase in errors after trial 1
predicted by the frequency hypothesis did not occur.
It is likely, however, that this effect was masked by
the large initial decrement, so that this deviation

from predicted results should not be construed

22

as a breakdown of the frequency analysis. It is
reasonable to assume that there are factors other than
differential frequency that may influence VD performance.

Instructions are apparently critical in determining
the effects of associated words in VD learning. The
uninstructed groups, although showing the expected trends,
did not differ significantly from the C group. In this
experiment Ss were shifted from one list to another and
detailed instructions were necessary for the S'to view
the two tasks as related. After the experiment some of
the uninstructed Ss were asked if they were aware of
any connection between the two lists, and all but one
reported that they had not noticed any relation and had
not expected one. That the U-Aw and U-AC groups were
significantly different from each other probably indicates
that unverbalized associations were somewhat effective
in influencing performance, but clear cut results appear
to depend on appropriate instructional set. Similarly,
the failure to inform SS of the presence of the associated
words within the VD list may account for the weak effect
of the Association variable in the Ekstrand, Wallace and
Underwood (1966) study.

The finding that, for the instructed Ss, unidirec-
tional associates tended to provide somewhat more
information (i.e., led to fewer errors in the I-AC group
and more in the I-AW group) is paradoxical, especially

considering that in all unidirectional pairs the word used

23

in list 2 was the word that had a frequency of

association back to the first list word that was no

greater than .02. Especially in these instructed
conditions, it would have been eXpected that the bidirec-
tional associates, which presumably facilitated associating
back to the list 1 words, would provide the strongest

cues for responding. For the uninstructed groups

the bidirectional associates did, in fact, tend to provide

the most information.

24
SUMMARY

This eXperiment was designed to determine whether
implicit associations between words in successive lists
of word pairs would be utilized as cues in responding
in a verbal discrimination task. In addition, the
importance of instructions in determining the set of
the SS was investigated.

Eighty introductory psychology students were first
required to learn to a criterion of three errorless
trials a list of 15 pairs of words, in which one word of
each pair had been arbitrarily designated correct by S.
They were then required to learn a second list which
contained some words which were bidirectional or
'unidirectional associates of the correct words from the
'first list. For one experimental condition the associates
in.the second list were correct, for the other they
Ivere incorrect. A control group learned two unrelated
lists. Half of the SS in each experimental condition
Ivere instructed about the presence of the associates
:in.the second list and informed that these words would
Ice right or wrong, depending on the condition. The
(ather Ss were simply told they would be learning another
list. All Ss were given 10 trials on list 2, and the
number of errors per trial served as the measure of

Performance .

For the instructed Ss correct associate words in

25
list 2 led to slight positive transfer while incorrect
associates produced considerable negative transfer. The
same effects occurred in the uninstructed conditions
but were less pronounced. These differences between the
groups were apparent on the first trial of the second list,
and the error curves converged over the 10 trials.
The set provided by the complete instructions was
apparently crucial in determining the effects of the

associate words, since the uninstructed SS did not

 

If.

II

y.
'2'». -

perform significantly differently from the control Ss.
These findings were discussed in terms of a frequency

hypothesis of verbal discrimination learning.

REFERENCES

Battig, W.F., Williams, J.M., and Williams, J.G. Transfer
from verbal-discrimination to paired-associate

learning. g; exp. Psychol., 1962, S2, 258-268.

Bilodeau, E.A. and Howell, D.C. Free association norms
py discrete and continued methods. Washington, D.C.:
Office of Naval Research, 1965.

 

Bousfield, W.A., Whitmarsh, G.A., and Danick, J.J.
Partial response identities in verbal generalization.

Psychol. Rep., 1958, 3, 703-713.

Bugelski, B.R. and Scharlock, D.P. An experimental
demonstration of unconscious mediated association.

g; exp. Psychol., 1952, 33, 334-338.

Dominowski, R.L. First trial guessing and verbal-
discrimination learning. Psychon. Sci., 1966,

‘5, 231-233.

Ellis, H.C. The transfer 2: learning. New York:
Macmillan, 1965.

Ekstrand, B.R., Wallace, W.F., and Underwood, B.J.
A frequency theory of verbal-discrimination
learning. Psychol. Rev.. In press.

McClelland, D.C. Studies in verbal discrimination
learning. II. Retention of responses to right
and wrong words in a transfer situation. gg'exp.
Psychol., 1942, 21, 149-162.

Mink, W.D. Semantic generalization as related to word
association. Unpublished doctoral dissertation,
University of Minnesota, 1957.

Runquist, W.N. and Freeman, M. Roles of associative
value and syllable familiarization in verbal
discrimination learning. g; exp. Psychol., 1960,

59, 396-401 .

Russell, W.A. and Storms, L.H. Implicit verbal chaining
in paired-associate learning. g; exp. Psychol.,
1955. 22. 287-293.

£3altz, E. The precriterion phase in verbal discrimination.
J. verbal Learn. verbal Behav., 1964, 2, 166-170.

 

Ehchwartz, H.A. Influence of instructional set and response
frequency on retroactive interference. g; exp.

26

27

Psychol., 1963, SS, 127-132.

Spear, N.E., Ekstrand, B.R., and Underwood, B.J.
Association by contiguity. g; exp. PSychol..
1964, 6 , 151-161.

 

Underwood, B.J., Jesse, F., and Ekstrand, B.R.
Knowledge of rights and wrongs in verbal-discrimi-
nation learning. g; verbal Learn. verbal Behav.,

Winer, B.J. Statistical rinci les in experimental
design. Chicago: McGraw-H 1,-I962.

 

rm

 

If.” sacs-r -

APPENDICES

APPENDIX A

ASSOCIATE WORD PAIRS

The associate pairs used in the eXperiment are

presented below according to their source and the

eXperimental condition in which they were used.

word in each pair appeared in one of the first lists,

the second word in list

2.

The first

The number beside each word

indicates the frequency with which that word occurs as a

response when the other word in the pair is given as the

stimulus word.

EXperimental Condition

Associates Correct

Unidirectional Sglpp

From Mink (1957)

.00 BLOSSOM FLOWER
.00 TOBACCO SMOKE

.00 MUTTON LAMB

lFrom Bilodeau and Howell (1965)

.01 WIFT FAST

. 02 INFANT BABY

Bidirectional Pairs
From Mink (1957)
-?4 QUEEN KING

- 61 BLACK WHITE

068 001
.51 .02
.36
0’49 .00
056 001
.01
.71 .56
.74 .64
.47

EAGLE
HEAVY

THIRSTY
SCISSORS

STOMACH

SWEET
HIGH

HAMMER

Associates Wropg

BIRD

LIGHT

WATER
CUT

ACHE

SOUR
LOW

NAIL

~55

.48
.60
.35

31

Associates Correct Associates flappg
From Bilodeau and Howell (1965)
.41 MAN WOMAN .73 .44 HARD SOFT .56
.46 HOT COLD .72 .34 TALL SHORT .49
.21 SKY BLUE .56

APPENDIX B
INSTRUCTIONS FOR LIST 1

__..i_ _. _ mnlfifln-&—--_-‘ h‘.--‘

We are interested in certain complex relationships
involved in learning to discriminate words that are
common to all peOple and are not concerned with your
personal performance.

In the window you see on the machine a list of
words will be presented. First a pair of words will
appear, one of which is correct. This pair will be
presented for two seconds and then the correct word

will be presented alone for two seconds. There are

 

15 pairs in the list. Every time we have gone through
the list completely you will see the row of asterisks
that is now in the window. This will mean that we
have completed one trial.

Your task is to guess which word in each pair is
correct. Since there is no reason that the correct word
is correct (i.e., we have arbitrarily designated one
word in each pair as correct), you will only be
guessing the first time we go through the list but
please choose one of the words anyway. It is important
that you tell me which word in each pair you think is
correct before it appears alone in the window. We will
Ikeep going through the list until you have made no errors
Ifor three consecutive trials.

Please do not ask any questions about the purpose

Of the eXperiment until we are finished. Do you have

arm'questions regarding your task?

 

 

 

AAAAA

 

”TIM!!!U!Mull/IllIIIILI'IEIIZITEI'IMIEEHW'55

3 1293 O