ATTITUDES ESTABLISHED BY CLASSICAL CONDITIONING1

ARTHUR W. STAATS AND CAROLYN K. STAATS

Arizona State College at Tempe

stimuli, unlike nonsense syllables, would be

O
SGOOD and Tannenbaum have stated,
"... The meaning of a concept is its expected to evoke attitudinal responses on the
location in a space denned by some basis of the pre-experimental experience of the
number of factors or dimensions, and attitude 5s. Thus, the purpose of the present study is to
toward a concept is its projection onto one of test the hypothesis that attitudes already
these dimensions defined as 'evaluative' " (9, elicited by socially significant verbal stimuli
p. 42). Thus, attitudes evoked by concepts can be changed through classical conditioning,
are considered part of the total meaning of using other words as unconditioned stimuli.
the concepts.
A number of psychologists, such as Cofer and METHOD
Foley (1), Mowrer (5), and Osgood (6, 7), to Subjects
mention a few, view meaning as a response—an Ninety-three students in elementary psychology
implicit response with cue functions which participated in the experiments as 5s to fulfill a course
may mediate other responses. A very similar requirement.
analysis has been made of the concept of
Procedure
attitudes by Doob, who states, " 'An attitude
is an implicit response . . . which is considered The general procedure employed was the same as in
the previous study of the authors (12).
socially significant in the individual's society' " Experiment I,—The procedures were administered
(2, p. 144). Doob further emphasizes the to the 5s in groups. There were two groups with one half
learned character of attitudes and states, "The of the 5s in each group. Two types of stimuli were used:
learning process, therefore, is crucial to an national names which were presented by slide pro-
understanding of the behavior of attitudes" (2, jection on a screen (CS words) and words which were
presented orally by the E (US words), with 5s required
p. 138). If attitudes are to be considered to repeat the word aloud immediately after E had
responses, then the learning process should be pronounced it. Ostensibly, 5s' task was to separately
the same as for other responses. As an example, learn the verbal stimuli simultaneously presented in the
the principles of classical conditioning should two different ways.
Two tasks were first presented to train the 5s in the
apply to attitudes. procedure and to orient them properly for the phase of
The present authors (12), in three experi- the experiment where the hypotheses were tested. The
ments, recently conditioned the evaluative, first task was to learn five visually presented national
potency, and activity components of word names, each shown four times, in random order. 5s'
meaning found by Osgood and Suci (8) to learning was tested by recall. The second task was to
learn 33 auditorily presented words. 5s repeated each
contiguously presented nonsense syllables. The word aloud after E. 5s were tested by presenting 12
results supported the conception that meaning pairs of words. One of each pair was a word that had
is a response and, further, indicated that word just been presented, and 5s were to recognize which
meaning is composed of components which can one.
The 5s were then told that the primary purpose of
be separately conditioned. the experiment was to study "how both of these types of
The present study extends the original learning take place together—the effect that one has
experiments by studying the formation of upon the other, and so on." Six new national names
attitudes (evaluative meaning) to socially were used for visual presentation: German, Swedish,
significant verbal stimuli through classical con- Italian, French, Dutch, and Greek served as the C5s.
These names were presented in random order, with
ditioning. The socially significant verbal exposures of five sec. Approximately one sec. after the
stimuli were national names and familiar CS name appeared on the screen, E pronounced the US
masculine names. Both of these types of word with which it was paired. The intervals between
exposures were less than one sec. 5s were told they
1
This study is part of a series of studies of verbal could learn the visually presented names by just
behavior being conducted by the authors at Arizona looking at them but that they should simultaneously
State College at Tempe, The project is sponsored by the concentrate on pronouncing the auditorily presented
Office of Naval Research (Contract Number NONR- words aloud and to themselves, since there would be
2305 (00)), Arthur W. Staats, principal investigator. many of these words, each presented only once.
37
38 ARTHUR W. STAATS AND CAROLYN K. STAATS
The names were each visually presented 18 times in design: conditioned meaning (pleasant and unpleasant);
random order, though never more than twice in C5 names (Dutch and Swedish, or Tom and Bill); and
succession, so that no systematic associations were groups (1 and 2). The scores on the semantic differential
formed between them. On each presentation, the CS given to each of the two CS words were analyzed in a
name was paired with a different auditorily presented 2 x 2 latin square as described by Lindquist (4, p. 278)
word, i.e., there were 18 conditioning trials. CS names for his Type II design.
were never paired with US words more than once so
that stable associations were not formed between them.
Thus, 108 different US words were used. The CS RESULTS
names, Swedish and Dutch, were always paired with US The 17 5s who indicated they were aware of
words with evaluative meaning. The other four CS
names were paired with words which had no systematic either of the systematic name-word relation-
meaning, e.g., chair, with, twelve. For Group 1, Dutch ships were excluded from the analysis. This
was paired with different words which had positive was done to prevent the interpretation that
evaluative meaning, e.g., gift, sacred, happy; and the conditioning of attitudes depended upon
Swedish was paired with words which had negative
evaluative meaning, e.g., bitter, ugly, failure, 2 For awareness. In order to maintain a counter-
Group 2, the order of Dutch and Swedish was reversed balanced design when these 5s were excluded,
so that Dutch was paired with words with negative four 5s were randomly eliminated from the
evaluative meaning and Swedish with positive meaning analysis. The resulting Ns were as follows: 24
words. in Experiment I and 48 in Experiment II.
When the conditioning phase was completed, 5s
were told that E first wished to find out how many of Table 1 presents the means and SDs of the
the visually presented words they remembered. At the meaning scores for Experiments I and II.
same time, they were told, it would be necessary to The table itself is a representation of the 2 X 2
find out how they/eW about the words since that might design for each experiment. The pleasant
have affected how the words were learned. Each S was
given a small booklet in which there were six pages.
On each page was printed one of the six names and a TABLE 1
semantic differential scale. The scale was the seven- MEANS AND SDs OF CONDITIONED ATTITUDE SCORES
point scale of Osgood and Suci (8), with the con-
tinuum from pleasant to unpleasant. An example is as Names
follows:
um^u
Dutch Swedish
German Expen-
pleasant: : : : : : : .'unpleasant ment Group Mean SD Mean SD
The 5s were told how to mark the scale and to
indicate at the bottom of the page whether or not the I 1 2.67 .94 3.42 1.50
word was one that had been presented.
2 2.67 1.31 1.83 .90
The 5s were then tested on the auditorily presented
Tom Bill
words. Finally, they were asked to write down anytlu'ng
they had thought about the experiment, especially the ment Group Mean SD Mean SD
purpose of it, and so on, or anything they had thought II 1 2.71 2.01 4.12 2.04
of during the experiment. It was explained that this 2 3.42 2.55 1.79 1.07
might have affected the way they had learned.
Experiment //.—The procedure was exactly re- Note.—On the scales, pleasant is I, unpleasant 7.
peated with another group of 5s except for the CS
names. The names used were Harry, Tom, Jim, Ralph, TABLE 2
Bill, and Bob. Again, half of the 5s were in Group 1 and
half in Group 2. For Group 1, Tom was paired with SUMMARY OF THE RESULTS OF THE ANALYSIS OF
positive evaluative words and Bill with negative VARIANCE FOR EACH EXPERIMENT
words. For Group 2 this was reversed. The semantic
differential booklet was also the same except for the Exp. I Exp. II
C5 names. Source
df MS F df MS F
Design
The data for the two experiments were treated in the Between 5s
same manner. Three variables were involved in the Groups 1 7.52 4.36* 1 15.84 5.00*
Error 22 1.73 46 3.17
2
The complete list of CS-US word pairs is not pre- Within
sented here, but it has been deposited with the American Conditioned 1 7.52 5.52* 1 55.51 10.47**
Documentation Institute. Order Document No. 5463 attitude
Names 1 .02 .01 1 .26 .05
from ADI Auxiliary Publications Project, Photo- Residual 22 1.36 46 5.30
duplication Service, Library of Congress, Washington Total 47 95
25, D. C., remitting in advance §1.25 for microfilm or
SI.25 for photocopies. Make checks payable to Chief, ' t < .05.
Photoduphcation Service, Library of Congress. "p < .01.
 ATTITUDES ESTABLISHED BY CLASSICAL CONDITIONING 39

extreme of the evaluative scale was scored 1, since they occur only once and are followed by
the unpleasant 7. other associations which are inhibitory. The
The analysis of the data for both experi- direct associations indicated in the figure
ments is presented in Table 2, The results of between the name and the individual words
the analysis indicate that the conditioning would also in this way be inhibited.
occurred in both cases. In Experiment I, the It was not thought that a rating response
F for the conditioned attitudes was significant was conditioned in this procedure but rather
at better than the .05 level. In Experiment II, an implicit attitudinal response which medi-
the F for the conditioned attitudes was signifi- ated the behavior of scoring the semantic
cant at better than the .01 level. In both differential scale. It is possible, with this con-
experiments the F for the groups variable was ception, to interpret two studies by Razran
significant at the .05 level. (10, 11) which concern the conditioning of rat-
ings. Razran found that ratings of ethnically
DISCUSSION labeled pictures of girls and sociopolitical slo-
It was possible to condition the attitude gans could be changed by showing these stimuli
component of the total meaning responses of while Ss were consuming a free lunch and, in
US words to socially significant verbal stimuli, the case of the slogans, while the 5s were
without Ss' awareness. This conception is presented with unpleasant olfactory stimula-
schematized in Fig. 1, and in so doing, the way tion. The change in ratings could be thought
the conditioning in this study was thought to to be due to the conditioning of an implicit
have taken place is shown more specifically. evaluative response, an attitude, to the CSs by
The national name Dutch, in this example, is means of the lunch or the unpleasant odors.
presented prior to the word pretty. Pretty That is, part of the total response elicited
elicits a meaning response. This is schematized by the food, for example, was conditioned to
in the figure as two component responses; an the pictures or slogans and became the
evaluative response rpy (in this example, the mediation process which in turn elicited the
words have a positive value), and the other positive rating.
distinctive responses that characterize the It should be stated that the results of the
meaning of the word, Rp. The pairing of present study do not show directly that Ss'
Dutch and pretty results in associations between behavior to the object (e.g., a person of Dutch
Dutch and rpv, and Dutch and Rp. In the fol- nationality) has been changed. The results
lowing presentations of Dutch and the words pertain to the Ss' attitudinal response to the
sweet and healthy, the association between signs, the national names themselves. However,
Dutch and r P v is further strengthened. This is Kapustnik (3) has demonstrated that a re-
not the case with associations RP, Rs, and RH, sponse generalized to an object when the re-
sponse had previously been conditioned to the
CS verbal sign of the object. Osgood states,
The aggressive reactions associated with Nazi and Jap
DUTCH ^.-=u-_ _ . on a verbal level certainly transferred to the social
objects represented under appropriate conditions.
"PRETTY. Similarly, prejudicial behaviors established while read-
ing about a member of a social class can transfer to the
class as a whole . . . (7, p. 704).

The results of this study have special rele-

vance for an understanding of attitude forma-
tion and change by means of verbal communi-
cation. Using a conception of meaning as a
mediating response, Mowrer (5) has suggested
DUTCH
that a sentence is a conditioning device and
that communication takes place when the
-HEAI/DHY
meaning response which has been elicited by
FIG. 1. THE CONDITIONING or A POSITIVE ATTITUDE.
the predicate is conditioned to the subject of
THE HEAVINESS OF LINE REPRESENTS STRENGTH the sentence. The results of the present study
OP ASSOCIATION and the previous one of the present authors
40 ARTHUR W. STAATS AND CAROLYN K. STAATS

(12) substantiate Mowrer's approach by sub- words were different, they all had an identical
stantiating the basic theory that word meaning evaluative meaning component. In Experi-
will indeed condition to contiguously pre- ment I, one national name was paired with
sented verbal stimuli. In the present study, positive evaluative meaning and another was
the meaning component was evaluative, or paired with negative evaluative meaning. In
attitudinal, and the CSs were socially signifi- Experiment II, familiar masculine names
cant verbal stimuli. The results suggest, there- were used. In each experiment there was sig-
fore, that attitude formation or change through nificant evidence that meaning responses had
communication takes place according to these been conditioned to the names without 5s'
principles of conditioning. As an example, the awareness.
sentence, "Dutch people are honest," would
condition the positive attitude elicited by REFERENCES
"honest" to "Dutch"—and presumably to any 1. COEER, C. N., & FOLEY, J. P. Mediated generaliza-
person called "Dutch," If, in an individual's tion and the interpretation of verbal behavior:
history, many words eliciting a positive atti- I. Prologemena. Psychol. Rev., 1942, 49, 513-
tude were paired with "Dutch," then a very 540.
positive attitude toward this nationality would 2. DOOB, L. VV. The behavior of attitudes. Psychol.
Rev., 1947, 54, 135-156.
arise. 3. KAPUSTNIK, 0. P. The interrelation between direct
The reason for the group differences in each conditioned stimuli and their verbal symbols.
of the experiments is not clear. These differ- (Trans, from Russian title) Psychol. Abstr.,
ences could have arisen because there were 1934, 8, No. 153.
4. LINDQUIST, E. F. Design and analysis oj exper-
actual differences in the 5"s composing each iments in psychology and education. Boston:
group, or in some condition of the procedure Hough ton Mifflin, 1953.
occurring to one of the groups. Nothing the 5. MOWRER, O. H. The psychologist looks at lan-
authors were aware of seem to indicate this guage. Amer. Psychologist, 1954, 9, 660-694.
as the explanation, and in the previous experi- 6. OSGOOD, C. E. The nature and measurement of
meaning. Psychol. Brill, 1952, 49, 197-237.
ments of the authors (12) there were no group 7. OSGOOD, C. E. Method and theory in experimental
differences. Since in a 2 x 2 latin square the psychology. New York: Oxford Univer. Press,
interactions are entirely confounded with the 1953.
main effects, the group differences could also 8. OSGOOD, C. E., & Suci, G. J. Factor analysis of
meaning. J. exp. Psychol., 1955, 60, 325-338.
have arisen as a result of the interaction of the 9. OSGOOD, C. E., & TANNENBAUM, P. H. The
other two main effects (i.e., direction of con- principle of congruity in the prediction of
ditioning and names). attitude change. Psychol. Rev., 1955, 62, 42-55.
10. RAZRAN, G. H. S. Conditioning away social bias
SUMMARY by the luncheon technique. Psychol. Bidl., 1938,
35, 693.
Two experiments were conducted to test 11. RAZRAN, G. H. S. Conditioned response changes in
the hypothesis that attitude responses elicited rating and appraising sociopolitical slogans.
by a word can be conditioned to a contigu- Psychol. Bidl, 1940, 37, 481.
ously presented socially significant verbal 12. STAATS, C. K., & STAATS, A. W. Meaning estab-
lished by classical conditioning. /. exp. Psychol,
stimulus. A name (e.g., Dutch) was presented 1957, B4, 74-80.
18 times, each time paired with the auditory
presentation of a different word. While these Received June 12, 1957.