You are on page 1of 7

Journal of Experimental Psychology

1963, Vol. 66, No. 5, 454-460

WORD ASSOCIATION AS A FUNCTION OF


CONDITIONED MEANING 1
HOWARD R. POLLIO 2
University of Michigan

2 groups of 40 and 37 Ss, respectively, were exposed to the Staats'


language conditioning procedure involving 3 nonsense syllables as CSs
and 3 sets of 9 meaningful words each, as UCSs. At the conclusion of
training Ss gave word associations to the CS syllables. A 3rd group
of 25 Ss then rated these associates on the evaluative scale of the
semantic differential. Results indicated that SD ratings of the as-
sociates produced were congruent with the SD ratings of the paired
UCS word group. This was true whether or not the associates were from
the original UCS words or whether they were cued by the connotative
characteristics of the CS syllable. The results indicate that in the
absence of specific word-word associative habits, word association may
be determined by the connotative characteristics of the stimulus word.

Although word association is a nenbaum, 1957) and Mowrer (1954)


classical psychological procedure, the mediational hypothesis of word mean-
critical factors which lead to the ing acquisition, has assumed that
evocation of a specific word as an while the frequency of occurrence of
association have yet to be described an individual word, or its frequency of
in any detail. The similarity between co-occurrence with another word,
word association and memory experi- provide some basis for predicting overt
ments has been noted by Woodworth verbal associative behavior, the con-
(1938), and it is, therefore, not sur- ditioned meaning response aroused by
prising to find that most explanations stimulus words must also strongly
of word association involve considera- affect the course of overt association.
tions of such factors as frequency and As evidence for this hypothesis, Staats
recency which have been shown to has examined the correlation between
have a strong effect on memory for the evaluative semantic differential
words. Thus, Johnson (1956), Os- (SD) rating of a stimulus word and
good and Anderson (1957), Howes the mean evaluative SD rating of the
(1957), and Underwood and Schulz first 20 associates to that stimulus
(I960) all assume some variety of a word. For 10 words, the rank-order
frequency of occurrence explanation correlation between these ratings was
as sufficient to explain why one word .90 (p < .01).
produces another in the word associa- A more extensive analysis of the
tion procedure. semantic relationship between a stim-
Staats (1961; Staats & Staats, ulus word and its first associates has
1959), in his elaboration of the Osgood been done (Pollio, 1962) and the
(Osgood, 1953; Osgood, Suci, & Tan- results show a significant positive
1
The present research was done while the
correlation between the SD rating of
author was on a Predoctoral United States 50 Kent-Rosanoff stimulus words and
Public Health Service Research Fellowship. the SD ratings of their primary
Arthur W. Melton offered many helpful associates on the evaluative, potency,
suggestions during the course of this in-
vestigation. and activity dimensions of the seman-
2
Now at the University of Tennessee. tic differential. Meaningful media-
454
WORD ASSOCIATION AND CONDITIONED MEANING 455

tion would seem to be involved in with positive evaluative SD ratings as


determining overt word associations determined by the Jenkins, Russell, and Suci
(1958) norms, one set of nine with negative
although its specific contribution is evaluative SD ratings, and one set of nine
not completely clear. with neutral evaluative SD ratings.
The purpose of the present experi- The nine words serving as the positive
ment is to examine word associations UCS words and the nine words serving as the
negative UCS words were so selected as to
given in response to an originally have fewer within-set common associates than
neutral language stimulus such as a between-sets common associates. There are
nonsense syllable of low association 36 intragroup associates common to two or
value which has had a specific mean- more words in the set of positive UCS words
ing response conditioned to it. The (i.e., M = 4.00); while there are 42 intragroup
associates in the set of negative UCS words
procedure successfully used by Staats (M = 4.07). The communality between
(Staats, 1961; Staats & Staats, 1959) these two sets yields 87 cases (M = 4.85).
permits the conditioning of specific This control on the interword associative
meaning responses to nonsense syl- tendencies of the UCS words serves to guar-
antee that if the SD profiles of associations
lables without the concomitant condi- emitted by 5s in response to one of the
tioning of any or many words as direct nonsense syllables are similar to the SD
word associates. We have assumed profiles of the UCS words paired with that
that after language conditioning takes nonsense syllable, this similarity is the result
place: (a) The 5s will be able to of communality in meaning response, rather
than the result of communality in intralist
provide a significantly greater number associative relations such as have been
of meaningful word associations to described by Deese (1959) and by Rothkopf
nonsense syllables which initially and Coke (1961).
aroused few if any associates. (6) The The specific words employed in the various
SD ratings of associates given in sets of UCS words will be found in Table 1.
The 1958 SD ratings were taken from the
response to a nonsense syllable having Jenkins, Russell, and Suci (1958) norms while
meaning responses established through the second set of SD ratings presented was
this procedure, will be similar to the obtained from the 25 5s in our word rating
mean SD ratings of the words used group. As can be seen from Table 1 the
present ratings, when compared to those
as the UCS words. obtained by Jenkins et al., produce both less
polarized mean SD values and larger within-
METHOD group sigmas for all three sets of UCS words.
The differences in within-group variability are
Subjects particularly noticeable for neutral UCS words.
There were two experimental groups; While the mean difference between ratings is
Group I had 40 5s while Group II had 37 5s. not great, many of the individual ratings in
A third group of 25 5s was used to rate this group deviate considerably from the
material obtained from Groups I and II. All original norms (e.g., GLOW).
5s were female introductory psychology
students at the University of Michigan and Procedure
served in this experiment as part of their
course requirement. Meaning conditioning,Two different con-
ditioning sessions were run, each with a
Materials different group of 5s. For Group I, GEJ was
paired with positive UCS words; XAD was
The materials consisted of the following paired with negative UCS words; and QIH
three trigrams, with their respective associa- was paired with neutral UCS words. For
tion norms as restandardized by Archer Group II, XAD was paired with positive UCS
(1960); GEJ (9%), QIH (6%), and XAD (7%). words, GEJ with negative UCS words, and
These words will be referred to as CS words. QIH was again paired with neutral UCS
Three sets of nine words each were used as words. The specific pairing of CS syllables
UCS words and had the following character- and UCS words took place as follows: A CS
istics: There was one set of nine UCS words syllable was flashed on a screen for 5 sec.
456 HOWARD R. POLLIO

TABLE 1
UCS WORDS AND THEIR EVALUATIVE SEMANTIC DIFFERENTIAL RATINGS

Positive Words Neutral Words Negative Words

UCS 1958 Present UCS 19S8 Present UCS 1958 Present


Word Rating" Rating Word Rating' Rating Word Rating8 Rating

MUSIC 1.60 1.60 ARCHED 4.07 3.70 SICKNESS 6.50 6.55


EATING 1.57 2.75 BOULDER 4.17 4.00 COLD 5.27 4.55
HOUSE 1.77 2.60 HEAVY 4.10 4.80 ROUGH 5.00 4.80
COMFORT 2.03 2.05 BRISTLY 3.93 4.75 ANGER 5.57 4.63
SMOOTH 2.57 2.60 GLOW 3.93 1.70 SOUR 5.37 5.30
SWEET 1.93 1.90 LONG 3.87 4.45 TROUBLE 6.10 5.45
BEAUTIFUL 2.00 1.80 RUN 3.87 3.55 HUNGRY 5.17 5.15
HEALTH 1.97 1.25 MALLET 3.80 4.75 THIEF 6.37 5.75
SWIFT 2.73 2.25 BLOCK 3.67 4.40 AFRAID 5.47 5.50
M 2.00 2.09 3.93 4.01 5.62 5.30
SD .35 .48 .14 .93 .49 .59

Taken from the Jenkins, Russell, and Suci (1958) norms.

After 1 sec., E read one of the UCS words permutations being randomized across 5s and
aloud, and all 5s pronounced the word after groups. Association was paced, so that 5s
him. This procedure was followed until each were given approximately 20 sec. in which to
of the three CS syllables had been paired with respond to any set of three syllables. Only
nine appropriate UCS syllables. This made those words given in the first 20 sec. associa-
a total of 27 pairings in all. The order of tion are considered in subsequent analyses.
presentation of the UCS-nonsense-syllable These words are most appropriate for demon-
pairings was random for both groups. During strating the effects of conditioned meaning on
this conditioning period 5s were told not "to word association uncontaminated by the
make any special attempt to memorize either nature of the specific word associates already
the words or the syllables." produced. Choosing the associates produced
Word association.At the conclusion of in any other association period, or the mean
the training sessions, 5s in both groups were value for all six periods, adds in the con-
asked "to write down, in the booklets given founding effect of these preceding associates.
to you as you entered, the first word that Since the purpose of this experiment is to
comes into your mind when you see the examine the effects of conditioned meaning
nonsense syllable." These booklets contained and not of prior associative responses, on
6 pages. On each of these pages the three overt word association only those associates
CS syllables appeared in one of the six produced in the first 20 sec. association period
permutations possible for three elements. provide appropriate data for this analysis.
For every S, each of the six possible orders Independent rating of associates.All of
was presented; the specific order of the the word associations given by the 77 5s of

TABLE 2
MEANS AND 5Ds OF THE SD RATINGS OF THE ASSOCIATIONS GIVEN IN RESPONSE
TO CS SYLLABLES AS A FUNCTION OF CONDITIONED MEANING PAIRING

Paired UCS Word Group

Experimental Positive Neutral Negative N


Group

M SD M SD M SD

i 2.76 1.00 3.10 1.13 4.72 1.27 40


ii 2.79 1.48 3.67 1.25 4.39 1.37 37
WORD ASSOCIATION AND CONDITIONED MEANING 457

Groups I and II during this first 20-sec. TABLE 3


period were given to a different group of 25
female 5s for rating of each word on the good- ANALYSIS OF VARIANCE OF THE DIFFERENCES
bad evaluative scale of the semantic differ- IN SD RATINGS OF ASSOCIATIONS AS A
ential. The specific order of words to be rated FUNCTION OF CONDITIONED MEANING
was randomly determined in booklets pre-
sented to this group of 5s. In each of these Source df MS F
booklets, the associates produced by 5s in Between Ss 76
Groups I and II were typed in capital letters. Groups (G) 1 .47
A 7-step evaluative SD scale, good-bad, was Error (b) 75 1.35
Within 5s 154
printed under each of these words. The 5s Conditioned Meanings (M) 2 63.61 36.35***
indicated their ratings of a word by placing a M XG 2 3.96 2.26
Error (w) 150 1.75
mark on the scale where they thought ap- 230
propriate. The instructions given to this
rating group were essentially similar to those
used by Jenkins, Russell, and Suci (19S8).

RESULTS An analysis of variance was per-


formed on these data, and the only
Evaluative characteristics of all as- significant source of variation among
sociates.-After SD ratings were se- means was attributable to conditioned
cured for the word associates given in meanings. Thus a CS syllable paired
response to the CS syllables for with meaningful UCS words tends
Groups I and II, mean SD values were to produce associates having the
obtained for each of these associations. same SD ratings as the UCS words
Table 2 shows that for both groups, themselves.
the SD rating of the word associates Evaluative characteristics of cued
given in response to the CS syllables associates.At this point it may be
were similar to the SD ratings of the argued that the SD ratings of the
paired UCS words. Thus, CS syl- associates to the CS syllables were
lables paired with positive UCS words similar to the SD ratings of the UCS
produced word associates which had words because the associates by and
positive evaluative meaning, while CS large, were the UCS words themselves.
syllables paired with negative UCS An examination of the associates
words produced word associates that produced, however, revealed that for
had negative evaluative meaning. the 40 5s in Group I, 54% (64 of a
The SD ratings of associates to the possible 120) of these associates were
neutral CS syllable fell between the words not contained in any of the
ratings of the positive and negative UCS word groups, while the com-
groups. parable figure for the 37 5s in Group

TABLE 4
NUMBER AND NATURE OF CUED ASSOCIATES AND MEANS AND 5Ds OF THE SD RATINGS
OF THESE CUED ASSOCIATES AS A FUNCTION OF PAIRED UCS WORDS

Paired UCS Word Group

Experimental Positive Neutral Negative


Group

M SD N M SD N M SD 2V

i 3.10 .93 22 3.28 1.14 20 4.70 1.50 22


ii 3.25 1.85 15 3.54 1.17 20 4.00 1.28 18
458 HOWARD R. POLLIO

II was 48% (53 of a possible 111). TABLE 5


Thus, about half of all of the associ- ANALYSIS OF VARIANCE OF THE DIFFERENCES
ates given could not be attributed to AMONG THE MEAN SD RATINGS
S learning a specific UCS word as a OF CUED ASSOCIATES
paired associate to a CS syllable. A
non-UCS word, which was given as an Source df MS F
associate to a CS syllable, will be Meanings (M) 2 16.460 9.058***
called a cued associate in subsequent Groups (G) 1 .250
discussion. A word was considered as M XG 2 2.735 1.505
Error 111 1.817
a cued associate if it was not spelled Total 116
exactly the same as any one of the
UCS words. For example, ARCHERY, *** f < .001.

despite some formal similarity to the


UCS word ARCHED, was considered a 4 are dependent on each S's responses
cued associate. and since the number of responses per
Table 4 presents the number and condition are confounded over 5s,
nature of the cued associates given as there is no completely adequate
a function of the paired UCS word statistic to test the significance of the
group, as well as the means and SDs differences among these mean SD
of the SD ratings for these cued ratings. Consequently, a simple 3 X 2
associates. As can be seen, the means factorial analysis of variance, which
are ordered as expected on the basis in effect disregards within-S correla-
of the experimental hypothesis, i.e., tion, was performed on these data.
cued word associations given in re- The results of these analyses indicate
sponse to the CS syllable paired with that the evaluative SD rating of a
positive UCS words tend to be posi- cued associate to an originally neutral
tive, those paired with the negative CS nonsense syllable is a function of
UCS words tend to be negative, and the evaluative SD rating of the UCS
those paired with neutral UCS words words paired with that CS syllable.
tend to fall between these extreme Evaluative characteristics of recalled
groups. associates,Approximately 50% of
Since the means presented in Table the associates given in response to

TABLE 6
NUMBER AND NATURE OF THE SPECIFIC UCS WORDS PRODUCED AS ASSOCIATES AS A FUNCTION
OF THE CONDITIONED MEANING CHARACTERISTICS OF THE CS SYLLABLE

Conditioned
Experimental Meaning Charac- Number and Nature of Specific UCS Words Total
Group teristics of CS Produced as Associates to CS Syllables
Syllables

Positive Neutral Negative

i Positive 17 0 1 18
Neutral 13 6 1 20
Negative 1 5 12 18
ii Positive" 18 2 1 21
Neutral" 5 6 5 16
Negative" 2 2 14 18

* A different S in each of these conditions failed to produce an associate in the 20-sec, association period,
WORD ASSOCIATION AND CONDITIONED MEANING 459

each of the three nonsense syllables SD ratings of CS syllables after language


for both groups were the UCS words conditioning as a dependent variable, is,
themselves. Table 6 presents the therefore, open to the criticism that it is
semantic characteristics of the specific the UCS words which are recalled, and
are being rated, rather than the CS
UCS word recalled as a function of syllable itself. It is, of course, possible
the connotative characteristics of the to argue that UCS syllables are recalled
CS syllables. With the sole exception because of the meaning response condi-
of the UCS words recalled in response tioned to the CS syllable, i.e., if associa-
to the CS syllables conditioned neu- tions are determined by the meaning
tral, the UCS words recalled had the response aroused by the CS syllable then
same connotation as was conditioned the UCS words are produced because
to the CS syllable. This effect is quite they most nearly approximate the mean-
strong for both the positively and ing response now conditioned to the CS.
negatively conditioned syllables. A The results presented in Table 6 offer
some support for this hypothesis by
chi square test, done on the 3 X 3 showing that there are few intrusions in
table obtained by collapsing across recall from UCS word groups having
experimental groups, yielded a value different evaluative characteristics than
of 63.46 (df = 4; p < .001). These those conditioned to the CS syllable.
results suggest that the semantic Where specific UCS syllables are re-
characteristics of the UCS word re- called, they are congruent in connotation
called as a paired associate to a CS to the evoking CS syllable. For syl-
syllable tends to have the same con- lables conditioned neutral, however,
notation as was conditioned to the CS intrusions from other UCS lists did occur.
syllable. Discrimination of the evalu- The single UCS word most often inap-
propriately intruded as an associate to
ative characteristics of the UCS words the neutral CS syllable was HOUSE. An
seems to have occurred, such that examination of many of the words con-
intrusions from the opposite evalu- tained in the neutral UCS word group
ative UCS word class do not occur. (ARCHED, BOULDER, HEAVY, MALLET,
BLOCK) indicates that semantic attributes
DISCUSSION other than evaluative (e.g., potency)
may have been conditioned to this
The results of the present study indi- syllable, and that the recall or association
cate that the meaning response aroused of the word HOUSE may have been made
by a verbal stimulus is an important in response to these attributes. The
factor affecting the evocation of associa- recall of UCS words given in response to
tive responses. We have seen that the the neutral CS syllable then follows the
evaluative meaning of a word associate same rule as does recall given in response
is similar to that of its evoking stimulus. to the positive and negative CS syllables.
This implies, as both Osgood (1953) and Although this analysis provides a
Staats (1961) have suggested, that possible explanation for the pattern of
meaning responses often precede, and to recall of UCS words in the present study,
a large extent, determine the connotative it does not answer the methodological
characteristics of associated responses. criticism raised earlier. At present, the
Contrary to Staats' supposition that operations involved in the language con-
the language conditioning procedure ditioning procedure do not permit us to
provides a technique for the conditioning determine if changes in the SD ratings
of meaning without the concomitant of CS syllables are the result of condi-
conditioning of specific words to CS tioned meaning responses or are the
syllables, the present results indicate result of simple paired-associate recall
that about 50% of the UCS words are of UCS words.
recalled by 5s. Using differences in the The results involving cued word
460 HOWARD R. POLLIO

associates, however, clearly demonstrate mediate recall. /. exp. Psychol., 1959, 58,
the effect of conditioned meaning re- 17-22.
sponses on word association. Here, HOWES, D. On the relationship between the
associative responses were not contained probability of a word as an association and
in any of the original UCS word lists, its general linguistic usage. /. abnorm. soc.
although they were similar in evaluative Psychol., 1957, 54, 75-85.
connotation to these words. It seems JENKINS, J. J., RUSSELL, W. A., & Suci, G.
An atlas of semantic profiles for 360 words.
that when 5s are unable to utilize specific Amer. J. Psychol., 1958, 71, 688-699.
word-word connections (as occurs in JOHNSON, D. Word association and word
in paired-associate recall) they pro- frequency. Amer. J. Psychol., 1956, 69,
duce word associations largely on the 125-126.
basis of conditioned meaning. Perhaps MOWRER, O. H. The psychologist looks at
the absence of direct word-word con- language. Amer. Psychologist, 1954, 9,
nections is a necessary condition for 660-692.
the utilization of conditioned meaning OSGOOD, C. E. Method and theory in experi-
as the primary determiner of overt mental psychology. New York: Oxford
word association. Univer. Press, 1953.
These analyses suggest that both OSGOOD, C. E., & ANDERSON, L. Certain
recall and association are affected by the relations among experienced contingencies,
meaning response common to a group associative structure, and contingencies in
encoded messages. Amer. J. Psychol., 1957,
of words paired with an originally neutral 70, 411-420.
stimulus. In terms of recall, the present OSGOOD, C. E., Suci, G. J., & TANNENBAUM,
results indicate that intrusion errors P. H. The measurement of meaning.
occur only infrequently if the to-be- Urbana: Univer. Illinois Press, 1957.
recalled material is homogeneous with POLLIO, H. R. Word association as a function
respect to a shared connotative dimen- of semantic structure. Unpublished doc-
sion. In terms of word association, the toral dissertation, University of Michigan,
present results indicate that in the 1962.
absence of strong word-word connections ROTHKOPF, E., & COKE, E. The prediction
among words, stimulus-aroused meaning of free recall from word association meas-
responses precede and determine sub- ures. /. exp.Psychol,, 1961, 62, 433-438.
sequent associated responses. Under this STAATS, A. W. Verbal habit families, con-
condition associative responses are largely cepts, and the operant conditioning of word
classes. Psychol. Rev., 1961, 68, 190-204.
the result of similarity in meaning be-
tween the stimulus and response words. STAATS, A. W., & STAATS, C. K. Meaning
and m\ Correlated but separate. Psychol.
Rev., 1959, 66, 136-144.
REFERENCES UNDERWOOD, B. J., & SCHULZ, R. W. Mean-
ARCHER, J. A re-evaluation of the meaning- ingfulness and verbal learning. New York:
fulness of all possible CVC trigrams. Lippincott, 1960.
Psychol. Monogr., 1960, 74(10, Whole No. WOODWORTH, R. S. Experimental psychology.
497). New York: Holt, 1938.
DEESE, J. On the prediction of occurrence
of particular verbal intrusions in im- (Received November 23, 1962)

You might also like