Journal o!

Applied Psychology
1976, Vol. 61, No. 3, 348-353

Studying Creativity by Means of Word Association Tests

Harrison G. Gough
Institute of Personality Assessment and Research,
University of California, Berkeley
The Kent-Rosanoff Word Association Test and a new scientific word list were
given to 66 senior honor students in engineering and 4s industrial research
scientists. Criterion ratings of creativity were available for all subjects. Asso-
ciations were classified into five categories, by decreasing frequency of occur-
rence. Moderately infrequent responses were more strongly related to creativity
than were responses of extreme remoteness or common occurrence. A weighted
summation of the five Kent-Rosanoff categories correlated .30 with the
creativity criterion for engineers and .24 for scientists. A similar summation
for the scientific list correlated .35 with the engineers' criterion and .36 with
that for scientists. Both word association indices were relatively uncorrelated
with other measures often used in studies of creativity.

The word association method, first intro- and Guilford (Guilford, Wilson, Christensen,
duced by Francis Gallon (1879), is one of & Lewis, 1951) also stressed this factor. In
the most venerable measuring devices of his review of methods for assessing creative
present-day psychology. Gallon's original aim potential, Tryk (1968) cited the associa-
was to investigale imaginalion, but he also tional tests of Mednick, Guilford, and others
recognized that the technique could shed light (e.g., Torrance, 1966) as among the most
into "the deepest recesses of the character." useful of the devices now available. Barren
Gallon's discovery was quickly noted by (1965, 1969) has also emphasized the ability
other psychologists. Cattell (1889), for ex- to form unusual associations and aesthetically
ample, while still in Wundt's laboratory, com- appropriate symbolic equivalents as signifi-
piled relative frequencies of associations lo cant determiners of creative attainment. The
standard lists of words. Jung (1910) proposed most direct use of word association technique
a set of 100 words, to aid in the detection of in studies of creativity is that of Maltzman
"complexes" of worry and neurotic involve- (1960), who evaluated the effects of different
ment. In 1910 Kent and Rosanoff introduced sets and training methods in producing more
their list of 100 words along with frequency diversified associations to standard lists.
tables for the responses given by 1,000 nor- There is no doubt that the word associa-
mal subjects. The Kent-Rosanoff list is still tion method can provide evidence on the re-
in use today, although the method is dis- moteness and infrequency of an individual's
linctly less popular among clinical psycholo- responses to the stimulus words. However, a
gists and psychiatrists than it was between, difficulty in equating idiosyncralic responses
say, 1910 and 194S (see Vernon, 19S3, pp. with crealive thinking is occasioned by the
172-175). fact that very unusual associations are or-
Another line of development, relevant to dinarily considered to be diagnostic of psy-
the empirical work to be reported below, chiatric disturbance and aberrations of per-
stems from the relationships between associa- ception. Associations of moderate but not
lional fluency and creativity. Mednick (1962) extreme atypicalitjr may furnish a beller in-
postulated the capacity to establish remote dex of creative potential than very rare or
associations as a key element in creativity, distanl responses. In the study to be reported
below an examination will be made of this
possibility. Another issue to be investigated
Requests for reprints should be sent to Harrison is the contention that stimulus material
Gough, Institute of Personality Assessment and
Research, University of California, Berkeley, Cali- drawn from a specific domain of work will
fornia 94720. be more relevant to creative accomplishment
348
 CREATIVITY AND WORD ASSOCIATIONS 349

in that domain than will more general or non- "C" responses were those given by from 10% to
specific word lists. 24.9%, "D" responses were those given by from
1% to 9.9%, and "E" responses were those given by
less than 1%. The protocol for any respondent
METHOD would therefore yield five part scores, depending on
Samples and Measures the distribution of his associations over these five
categories.
In accordance with the considerations just set To develop the Scientific Word Association Test
forth, a battery of assessment devices was admin- an initial listing of over 200 scientific words was
istered to a sample of 45 professional research scien- assembled, trying to identify terms that would have
tists studied at the Institute of Personality Assess- generally positive affective connotations for scien-
ment and Research. Twenty-eight of these men held tists and engineers. 2 This was attempted so as to
PhD degrees, nine had MA degrees or equivalent make the final device more attractive and ego-
graduate study, and eight had lesser levels of train- syntonic to scientists than a purely random collec-
ing. Mean age was 35.69 years (S£) = S.46). Fields tion of words would have been. Examples are:
of specialization include physics, mathematics, elec- binary, density, feedback, loop, spectrum, and
trical engineering, aerodynamics, and related dis- resonance.
ciplines. Further discussion with Institute colleagues and
Tests administered included (a) the 100-item Kent- scientific consultants led to the reduction of this list
Rosanoff Word Association Test, scored by means of to 100 terms. One criterion for dropping words was
the Russell-Jenkins (Note 1) Minnesota norms, (b) to eliminate those that might elicit purely definitional
Guilford's test of unusual uses, scored for the mean replies. An example here is btu. Another was to
judged quality of uses suggested (see Wilson, Guil- drop terms formed from proper names, for example,
ford, & Christensen, 1953), (c) the Minnesota Engi- Gaussian, and a third was to delete terms designat-
neering Analogies Test (Dunnette, 1955), used as a ing specific apparatus, such as tachometer and X-ray.
measure of general technical knowledge, and (d) the These 100 words were then given to the 45 scien-
Terman Concept Master)' Test (Terman, 1956), cm- tists and to the 66 senior honor students in engineer-
ployed as a measure of intellectual aptitude. In ad- ing. Frequency tallies were made of the responses to
dition, a new Scientific Word Association Test (to each word. Two examples may be provided.
be described below) was given. The same set of five
tests was also given to a sample of 66 senior honor For neutron, responses were: proton (28), atom,
students in engineering at the University of Cali- (15), physics (9), electron (8), particle (4), cyclo-
fornia. 1 The mean age for these students was 23.92 tron (3), nucleus ( 2 ) , and (1 each) baseball,
A creativity criterion for the scientists was de- charge, diffraction, free, glow, grams, head, mass,
rived Trom ratings furnished by peers and super- meson, neutral, no charge, positron, question,
visors. All 45 scientists were asked to rate each other, small, speed, and value.
insofar as they were acquainted with one another's For fission, responses were: fusion (14), nuclear
work. The number of ratings per scientist ranged (12), splitting (9), atoms (6), energy (6), atomic
from 5 to 21, with a mean of approximately 8. The (5), bomb (4), uranium (4), reaction (3), heat
corrected interjudge reliability for these ratings was (2), material (2), physics ( 2 ) , product ( 2 ) , and
.72. Each scientist was also rated for scientific cre- (1 each) blast, blow, boom, break, cleavage, ele-
ativity by two supervisors; the corrected interjudge ment, explosion, fish, fuel, H-bomb, how, power,
reliability coefficient for the supervisors was .53. The process, and separate.
two arrays were then standardized and summed.
Scoring classifications were next set up, using the
The corrected reliability of this final criterion was
same A-B-C-D-E categories already mentioned for
.77.
For the engineers, ratings were requested from the Kent-Rosanoff list. Five examples of the full set
faculty members. No student was rated by less than of 100 stimulus words with scoring instructions for
two faculty members, and the mean number of A, B, C, and D responses are given below. 3 Any un-
raters per student was approximately 6. The cor- listed response is tallied as an "E" association.
rected interjudge reliability coefficient for these
ratings was .72. 1
Werner Goldsmith was kind enough to put the
writer in contact with these honor students,
The Word Association Tests years, (SD -3.77).
2
A scoring method for the Kent-Rosanoff Word The help of Franklin Hurlbut and Donald Wood-
Association Test was developed on the basis of the worth in developing this initial list of words is grate-
Minnesota norms previously mentioned. A five-step fully acknowledged.
3
classification of responses was defined so that dis- The complete 100-item scientific list, with scor-
tinctions could be made between moderately and ing instructions, may be obtained from the National
extremely remote associations. Responses to any Auxiliary Publications Service, NAPS Document No.
word given by 50% or more of the Minnesota sub- 02780, 5 pages. Order from ASIS/NAPS, c/o Micro-
jects were defined as "A" replies. "B" responses were fiche Publications, 440 Park Avenue South, New
those given by from 25% to 49.9% of the subjects, York, N.Y. 10016.
350 HARRISON G. GOUGH

TABLE 1
WORD ASSOCIATION RESULTS TOR THE SCIENTISTS AND ENGINEERS

Engineers Scientists

Word association variable M SD r M SD r

Kent-Rosanoff
A response 13.92 5.88 -.23 14.04 5.48 -.20
B response 10.88 4.84 -.10 10.96 4.60 -.10
C response 16.20 5.01 -.18 14.13 4.06 -.10
D response 27.80 5.67 .20 28.69 4.97 .17
E response 31.02 12.72 .12 32.09 11.72 .10
Scientific Word
Association Test
A response 6.67 2.50 -.01 6.24 2.22 -.11
B response 7.59 3.04 -.14 8.11 3.00 -.07
C response 15.82 4.44 .16 16.06 4.61 -.20
D response 31.27 7.17 .35** 36.22 5.02 .29*
E response 38.29 10.36 -.26* 33.00 9.68 .00
*p < .05.
**/> < .01.

Solid. A = none; B — state, liquid; C = none; D — Scientific Word Association Test, with the
hard, metal, fluid, ice, light, body, mass, soft, rock. D tally for both samples giving rise to the
Hyperbola. A = curve; B = none; C = parabola,
geometry; D — ellipse, conic, mathematics.
highest correlations with the creativity rat-
Unstable. A=nonc; B i s t a b l e ; C = equilibrium; ings. Although the differences between the D
D = atom, falling, explosion, wobbly, fall, oscilla- and E correlations were in the anticipated
tion, decompose, moving. direction in all four instances, the only con-
Fidelity. A = high; B = n o n e ; C = music; D — trast attaining a statistically significant level
truth, sound, true, hi-fi, faithful, reproduction.
Catalyst. A = none; B = reaction, chemistry; C — (P < .05) was for the engineers on the scien-
agent; D = chemical, help, platinum, reagents, tific list.
actions, heat. It was also anticipated that the domain
RESULTS specific list would produce stronger relation-
ships than the general word list. This trend
A, B, C, D, and E tallies were made of the
is apparent in the correlations for the D cate-
responses given by the 111 subjects to the
gories, .35 versus .20 for the engineers and
Kent-Rosanoff words and to the scientific
.29 versus .17 for the scientists. The differ-
for these scores are given in Table 1, and also
ences within each pair of coefficients, how-
stimulus list. Means and standard deviations
ever, are not statistically significant.
their correlations with the creativity cri-
terion for each sample. The individual correlations in Table 1 as
On the Kent-Rosanoff list there is a very just reviewed provide one line of evidence on
clear progression of relationships for both the relevance of the two tests and their part
samples. More common responses, that is, scores. Another method of examining these
those in the A, B, and, C categories, are nega- issues is to look at the beta weights that
tively correlated with creativity whereas less would be attached to each category in a mul-
common responses are positively related. The tiple regression analysis. For this analysis, the
relationships for words in the D category two groups were combined into a single sam-
(association of moderate but not extreme fre- ple of 111 subjects, and then stepwise regres-
quency) were stronger in both samples than sions were conducted on the Kent-Rosanoff
those for responses in the E cell. A similar and Scientific Word Association Test tallies.
progression of relationships to creativity was The resulting beta weights for the five Kent-
found for the five scoring categories of the Rosanoff categories were as follows: A =
 CREATIVITY AND WORD ASSOCIATIONS 351

TABLE 2
INTERCORRELATIONS AMONG THE VARIABLES INDICATED IN SAMPLES or 66 STUDENT
ENGINEERS AND 45 RESEARCH SCIENTISTS

Criterion
Variable 5 rating M SD

1. Kent-Rosanofi Word Associ-

ation Test
Engineers .23 .13 .18 .23 .30* 49.53 2.84
Scientists -.05 .17 .10 .07 .24 50.19 2.42
2. Scientific Word Associ-
ation Test
Engineers -.05 .24* -.05 .35** 48.76 3.82
Scientists •.07 -.03 -.14 .36* 50.88 2.59
3. Guilford Unusual Uses Test
Engineers .10 .07 -.02 9.74 3.57
Scientists -.22 .02 .22 10.96 2.97
4. Minnesota Engineering Analo-
gies Test
Engineers .36* .34** 34.74 4.33
Scientists .17 .11 35.02 6.52
5. Terman Concept Mastery Test
Engineers .33** 78.39 26.49
Scientists -.05 118.22 29.38
* p < .05.
**P < .01.

-.19, B = .16, C = -.08, D = .17, and E Test the computing equation was S9.74 —
= .06. These weights show that when re- .22A - .61B -.33C + ,28D - ,22E.
sponses to all five classes are to be combined As an independent marker for the creativ-
into a single scoring, B, D, and E words ity dimension, Guilford's test for unusual uses
should be weighted positively, and that nega- was utilized. To assess general scientific apti-
tive weights should be assigned to A and C tude, Dunnette's engineering analogies test
reactions. was employed. Terman's Concept Mastery
A similar analysis of the scientific word list Test was selected as an indicator of general
generated these beta weights: A = —.05, B = intellectual ability. Table 2 presents inter-
-.18, C=-.1S, D = .18, and E = -.22. correlations among the five measures and the
Here the contribution of the D category is creativity criterion for the engineers and
clearly highlighted, as it is the only one with scientists.
a positive sign, The Kent-Rosanoff weighted index corre-
lated positively with the criterion ratings of
Comparison with Other Variables creativity for both samples, although the co-
In studies of creativity a word association efficient for scientists fell just short of sta-
list would seldom be used alone; evidence tistical significance (p = .06). Except for
concerning the relationship between this moderate relationships to the scientific word
method and others used to appraise creative list and Terman test in the student sample
potential is therefore needed. To obtain scores (r = .23), the Kent-Rosanoff index was more
for the Kent-Rosanoff and scientific lists the or less uncorrelated with any of the other
two regression analyses were employed. The variables, The contribution of this variable
two equations are viewed as heuristic only, to the estimation of creative potential, in
awaiting modification and improvement as other words, was reasonably independent of
new and larger samples become available. For other tests of creativity and of measures of
the Kent-Rosanoff list, the scoring formula either special or general aptitude.
was 43.29 - .33A + .34B — .17C + .30D + The weighted index on the Scientific Word
.OSE. For the Scientific Word Association Association Test revealed a similar pattern
352 HARRISON G. GOUGH

of relationships. It was significantly corre- scores gives rise to indices that appear to
lated with the creativity ratings for both sam- offer promise as indicators of creative poten-
ples, but only slightly related to scores on the tial. For both word lists it is the part score
Guilford test and the Terman Concept Mas- based on the number of moderately infre-
tery. It did correlate positively (r = .24) quent responses that contributes the strong-
with scores on the Engineering Analogies est positive weighting to the estimation of
Test in the student sample, but in the sample creativity.
of scientists this coefficient dropped to —.03. From a theoretical perspective, the finding
The Guilford Unusual Uses Test was mod- that moderately unusual replies are critical
erately, although insignificantly, correlated is worthy of attention. Associations that go
with the creativity criterion for scientists (r more or less completely afield, as in the E
— . 2 2 ) , but in the sample of engineers had a category, are known from past work with the
correlation of —.02. The two measures of apti- Kent-Rosanoff list to be indicative of per-
tude were positively correlated with the cri- sonal disturbance and psychiatric malfunc-
terion ratings for student engineers, but not tioning. But associations that are less atypical
for the research scientists. The ratings for the seem to be indicative of the kind of new per-
students were supplied by faculty members, spectives and original reactions that are al-
whose major contacts with the students oc- most by definition a part of the creative
curred in classroom and laboratory settings. process.
It might also be mentioned that the faculty The distinction between these two cate-
ratings for creativity correlated .72 with fac- gories is quite clearly seen in the scientific
ulty ratings of general scientific competence list. D responses tend generally to be within
and .71 with grade point averages. There is the domain of scientific concerns and nomen-
clearly a component of general academic com- clature, even if infrequently given. E re-
petence in the criterion ratings for the stu- sponses are often those that depart from the
dent engineers. This may explain the signifi- domain of reference, as suggested by the fol-
cant relationships found for the engineering lowing examples:
analogies and concept mastery tests. The near
zero correlations for the scientific and gen- Stimulus D response E response
eral aptitude measures found in the sample
Neutron cyclotron baseball
of scientists are more in accord with the find- nucleus head
ing from prior studies of creativity (Mac- particle small
Kinnon & Hall, 1972). feedback loop chaos
The differences in mean scores between servo machine
signal whistle
engineering students and scientists are in the matrix determinant complex
expected direction in all instances; that is, clement honeycomb
the scientists attained higher scores. The only tensor trix
variable for which this difference was statis-
tically significant was the Terman Concept From a practical standpoint one always
Mastery Test, where the t ratio was 7.22 wonders to what extent a respondent can
( / > < .001). manipulate his replies to a test so as to
achieve a desired effect. The weighting of D
DISCUSSION responses on both measures offers a difficult
The Kent-Rosanoff and Scientific Word problem to anyone wanting to appear as
Association lists can be administered as group maximally "creative'' on these devices: By
tests in from 10 to IS minutes, and either going a bit too far toward the unusual, points
can be given alone in less than 10 minutes. are lost on both word lists. It would be in-
Respondents typically find the procedures un- teresting to conduct a study of faking on
objectionable and easy to complete. By tally- these devices, to see if scores on the two
ing for responses varying in commonality, equations would go up or down when dis-
a convenient set of part scores is obtained, simulated protocols are contrasted with those
and appropriate weighting of these part gathered under normal testing instructions.
 CREATIVITY AND WORD ASSOCIATIONS 353

Of course, the most important evidence on Guilford, J. P., Wilson, R. C., Christensen, P. R.,
the merit of the word association method as & Lewis, D. J. A factor-analytic study of creative
thinking. I. Hypotheses and description of tests.
a diagnostic indicator must come from repli- Reports from the Psychological Laboratory (No.
cation. New findings showing a positive rela- 4). Los Angeles: University of Southern Cali-
tionship between the proposed scoring indices fornia, 1951.
for the two lists and nontest criteria of crea- Jung, C. G. The association method. American Jour-
nal of Psychology, 1910, 21, 219-269.
tive attainment would lend credence to the Kent, G. H., & Rosanoff, A. J. A study of associa-
belief that the word association technique can tion in insanity. American Journal of Insanity,
provide valuable information concerning sci- 1910, 67, 37-96; 317-390.
entific as well as other kinds of creativity. MacKinnon, D. W., & Hall, VV. B. Intelligence and
creativity. In H. W. Peter (Chair.) Colloquium
17: The measurement of creativity. Proceedings,
REFERENCE NOTE XVII International Congress of Applied Psychol-
ogy, Liege, Belgium, 25lh-30th July 1971 (Vol.
1. Russell, W. A., & Jenkins, J. J. The complete
2). Brussclls: Editest, 1972.
Minnesota norms for responses to 100 words from
Maltzman, I. On the training of originality. Psycho-
the Kent-RosanofJ Word Association Tesl (studies
on the role of language in behavior; Tech. Rep. logical Review, 1960, 67, 229-242.
No. 11, Contract No. N8-ONR-66216). Minneap- Mednick, S. A. The associative basis of the creative
olis: University of Minnesota, 1954. process. Psychological Review, 1962, 69, 220-232,
Terman, L. M. Manual for the Concept Mastery
Test. New York: Psychological Corporation, 1956.
REFERENCES Torrance, E. P. Torrance tests of creative thinking:
Norms-technical manual, research edition. Prince-
Barren, F. The psychology of creativity. In T. M. ton, N.J.: Personnel Press, 1966.
Newcomb (Ed.), New directions in psychology II, Tryk, H. E. Assessment of creativity. In P. Mc-
New York: Holt, Rinehart & Winston, 196$. Reynolds (Ed.), Advances in psychological assess-
Barren, F. Creative person and creative process. New ment (Vol. 1). Cupertino, Calif.: Science & Be-
York: Holt, Rinehart & Winston, 1969. havior Books, 1968.
Cattell, J. M. Mental association investigated by Vernon, P, E, Personality tests and assessments.
experiment. Mind, 1889, 14, 230-250. London: Mcthucn, 1953.
Dunnette, M. D. Minnesota Engineering Analogies Wilson, R. C., Guilford, J. P., & Christensen, P. R.
Test: Preliminary manual. New York: Psychologi- The measurement of individual differences in orig-
cal Corporation, 1955. inality. Psychological Bulletin, 1953, 50, 362-370.
Galton, F. Psychometric facts. Nineteenth Century,
1879, 5, 425-433. (Received August 28, 1975)