the associative process of creativity

13
Psychological Review 1962, Vol. 69, No. 3, 220-232 THE ASSOCIATIVE BASIS OF THE CREATIVE PROCESS SARNOFF A. MEDNICK University of Michigan The intent of this paper is the pres- entation of an associative interpreta- tion of the process of creative thinking. The explanation is not directed to any specific field of application such as art or science but attempts to delineate processes that underlie all creative thought. The discussion will take the follow- ing form, (a) First, we will define creative thinking in associative terms and indicate three ways in which crea- tive solutions may be achieved—seren- dipity, similarity, and mediation, (b) This definition will allow us to deduce those individual difference variables which will facilitate creative perform- ance, (c) Consideration of the defi- nition of the creative process has sug- gested an operational statement of the definition in the form of a test. The test will be briefly described along with some preliminary research results. (d) The paper will conclude with a discussion of predictions regarding the influence of certain experimentally manipulable variables upon the crea- tive process. Creative individuals and the proc- esses by which they manifest their creativity have excited a good deal of 1 The essence of this paper was written while the writer was a Visiting Research Psychologist at the Institute of Personality Assessment and Research, University of California, Berkeley. The author wishes to acknowledge his indebtedness to Benton J. Underwood for his encouragement and the stimulation of his paper, "Orientation to research on thinking" (1952). Work on this material has been supported by the Coopera- tive Research Program of the Office of Education (Project No. 1073) and the Na- tional Science Foundation (Grant G-3855). interest and curiosity. There are ex- tended analyses of novels and novel- ists, poems and poets, mathematics and mathematicians, both biographical and autobiographical. Perusal of the intro- spections of manifestly creative in- dividuals uncovers a surprising vein of similarity in the processes they de- scribe (Ghiselin, 1952). Thus we find Albert Einstein's self-searching to suggest that "The psychical entities which seem to serve as elements in thought are certain signs and more or less clear images which can be com- bined . . . This combinatory play seems to be the essential feature in productive thought." Samuel Taylor Coleridge is described as having de- veloped his ideas in the following man- ner : "Facts which sank at intervals out of conscious recollection drew together beneath the surface through the almost chemical affinities of common ele- ments." In the field of art, we find Andre Breton referring to a collage by Ernst as being distinguished by a "marvelous capacity to grasp two mu- tually distant realities without going beyond the field of our experience and to draw a spark from the juxtaposi- tion." Most explicit, however, is the oft-quoted statement by the mathema- tician, Poincare, who talks about an evening when "ideas rose in crowds; I felt them collide until pairs interlocked so to speak, making a stable combina- tion. By next morning I had estab- lished the existence of a class of Fuch- sian functions." From these experi- ences, Poincare felt that he could state that "to create consists of making new combinations of associative elements 220

Upload: api-26890925

Post on 10-Apr-2015

144 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: The Associative Process of Creativity

Psychological Review1962, Vol. 69, No. 3, 220-232

THE ASSOCIATIVE BASIS OF THE CREATIVE PROCESS

SARNOFF A. MEDNICK

University of Michigan

The intent of this paper is the pres-entation of an associative interpreta-tion of the process of creative thinking.The explanation is not directed to anyspecific field of application such as artor science but attempts to delineateprocesses that underlie all creativethought.

The discussion will take the follow-ing form, (a) First, we will definecreative thinking in associative termsand indicate three ways in which crea-tive solutions may be achieved—seren-dipity, similarity, and mediation, (b)This definition will allow us to deducethose individual difference variableswhich will facilitate creative perform-ance, (c) Consideration of the defi-nition of the creative process has sug-gested an operational statement of thedefinition in the form of a test. Thetest will be briefly described along withsome preliminary research results.(d) The paper will conclude with adiscussion of predictions regarding theinfluence of certain experimentallymanipulable variables upon the crea-tive process.

Creative individuals and the proc-esses by which they manifest theircreativity have excited a good deal of

1 The essence of this paper was writtenwhile the writer was a Visiting ResearchPsychologist at the Institute of PersonalityAssessment and Research, University ofCalifornia, Berkeley. The author wishes toacknowledge his indebtedness to Benton J.Underwood for his encouragement and thestimulation of his paper, "Orientation toresearch on thinking" (1952). Work on thismaterial has been supported by the Coopera-tive Research Program of the Office ofEducation (Project No. 1073) and the Na-tional Science Foundation (Grant G-3855).

interest and curiosity. There are ex-tended analyses of novels and novel-ists, poems and poets, mathematics andmathematicians, both biographical andautobiographical. Perusal of the intro-spections of manifestly creative in-dividuals uncovers a surprising vein ofsimilarity in the processes they de-scribe (Ghiselin, 1952). Thus wefind Albert Einstein's self-searching tosuggest that "The psychical entitieswhich seem to serve as elements inthought are certain signs and more orless clear images which can be com-bined . . . This combinatory playseems to be the essential feature inproductive thought." Samuel TaylorColeridge is described as having de-veloped his ideas in the following man-ner : "Facts which sank at intervals outof conscious recollection drew togetherbeneath the surface through the almostchemical affinities of common ele-ments." In the field of art, we findAndre Breton referring to a collageby Ernst as being distinguished by a"marvelous capacity to grasp two mu-tually distant realities without goingbeyond the field of our experience andto draw a spark from the juxtaposi-tion." Most explicit, however, is theoft-quoted statement by the mathema-tician, Poincare, who talks about anevening when "ideas rose in crowds; Ifelt them collide until pairs interlockedso to speak, making a stable combina-tion. By next morning I had estab-lished the existence of a class of Fuch-sian functions." From these experi-ences, Poincare felt that he could statethat "to create consists of making newcombinations of associative elements

220

Page 2: The Associative Process of Creativity

THE ASSOCIATIVE BASIS OF THE CREATIVE PROCESS 221

which are useful. The mathematicalfacts worthy of being studied . . . arethose which reveal to us unsuspectedkinships between other facts wellknown but wrongly believed to bestrangers to one another. Amongchosen combinations the most fertilewill often be those formed of elementsdrawn from domains which are farapart." An exceptionally compellingillustration of a useful combination ofelements "drawn from domains whichare far apart" occurs in a line from thepoem, "The Monkey Puzzle" by Mari-anne Moore (1951), "The lion's fero-cious chrysanthemum head."

We will state our basic hypothesisregarding the nature of creative think-ing in the form of a definition. Withthese introspective statements servingas background, we may proceed to de-fine the creative thinking process as theforming of associative elements intonew combinations which either meetspecified requirements or are in someway useful. The more mutually re-mote the elements of the new combina-tion, the more creative the process orsolution. An additional criterion ofthe level of creativeness of a productis described below.

Creative thinking as defined here isdistinguished from original thinkingby the imposition of requirements onoriginality. Thus, 7,363,474 is quitean original answer to the problem"How much is 12 + 12?" However,it is only when conditions are such thatthis answer is useful that we can alsocall it creative. There are many orig-inal ideas expressed in institutions forthe mentally ill and mentally retarded;few of these are likely to be creative.There are many fields of creative en-deavor in which the usefulness of prod-ucts would be difficult to measure re-liably. While these difficulties musteventually be faced, for the present ourresearch efforts have been concentrated

on laboratory situations in which cri-teria for usefulness can be arbitrarilyexperimenter-defined and unequivo-cally explained to the subject. Theoriginality of a response is simply in-versely related to its probability in agiven population.

It should be pointed out that thisdefinition of creativity is quite similarto basic notions advanced by Britishassociationists from Locke (1690) toBain (1855), and by those psycholo-gists whose work is based in largemeasure on their speculations. Freud(1938), Hollingsworth (1928), andBinet (1899) may serve as examples.

WAYS OF ACHIEVING A CREATIVESOLUTION

In terms of associative theory, wemay point to three ways of achievinga creative solution. Generally, anycondition or state of the organismwhich will tend to bring the requisiteassociative elements into ideationalcontiguity will increase the probabilityand speed of a creative solution.Therefore, the following three ways ofattaining creative solutions are allmethods of bringing the requisite asso-ciative elements together.

Serendipity. The requisite associa-tive elements may be evoked contigu-ously by the contiguous environmentalappearance (usually an accidental con-tiguity) of stimuli which elicit theseassociative elements. This sort of cre-ative solution is often dubbed seren-dipitous. This is the manner of dis-covery to which is popularly attributedsuch inventions as the X ray and suchdiscoveries as penicillin. One physi-cist has described how he has reducedserendipity to a method by placing ina fishbowl large numbers of slips ofpaper, each inscribed with a physicalfact. He regularly devotes some timeto randomly drawing pairs of these

Page 3: The Associative Process of Creativity

222 SARNOFF A. MEDNICK

facts from the fishbowl, looking fornew and useful combinations. Hisprocedure represents the operationalembodiment of this method of achiev-ing creative solutions.

Similarity. The requisite associativeelements may be evoked in contiguityas a result of the similarity of theassociative elements or the similarityof the stimuli eliciting these associativeelements. This mode of creative solu-tion may be encountered in creativewriting which exploits homonymity,rhyme, and similarities in the structureand rhythm of words or similarities inthe objects which they designate. Thecontiguous ideational occurrence ofsuch items as alliterative and rhymingassociates may be dependent on a fac-tor such as primary stimulus general-ization. It seems possible that thismeans of bringing about contiguity ofassociational elements may be of con-siderable importance in those domainsof creative effort which are less di-rectly dependent on the manipulationof symbols. Here we might includecertain approaches to painting, sculp-ture, musical composition, and poetry.

Mediation. The requisite associa-tive elements may be evoked in con-tiguity through the mediation of com-mon elements. This means of bringingthe associative elements into contigu-ity with each other is of great impor-tance in those areas of endeavor wherethe use of symbols (verbal, mathemati-cal, chemical, etc. . . .) is mandatory.For example, in psychology, the ideaof relating reactive inhibition and cor-tical satiation may have been mediatedby the common associates "tiredness"or "fatigue" (Kohler & Fishback,19SO).

INDIVIDUAL DIFFERENCES

From the definition given above, thefactors that will make for individualdifferences in the probability of achiev-

ing creative solutions may be deduced.Any ability or tendency which servesto bring otherwise mutually remoteideas into contiguity will facilitate acreative solution; any ability or ten-dency which serves to keep remoteideas from contiguous evocation willinhibit the creative solution.

Listed below are several illustrativepredictions concerning individual dif-ferences that one may make fromthis theoretical orientation.

Need for Associative Elements

It should be clear that an individualwithout the requisite elements in hisresponse repertoire will not be ableto combine them so as to arrive at acreative solution. An architect whodoes not know of the existence of anew material can hardly be expectedto use it creatively.

Associative Hierarchy

The organization of an individual'sassociations will influence the proba-bility and speed of attainment of acreative solution. There is a wholefamily of predictions that one maydraw from this concept of the associa-tive hierarchy. As an initial example,let us take the question of the mannerin which the associative strengtharound ideas is distributed. If wepresent an individual with the word"table," what sort of associative re-sponses does he make? The individ-ual who tends to be restricted to thestereotyped responses, such as "chair,"may be characterized as having an as-sociative hierarchy with a steep slope(see Figure 1). That is, when you getpast the first one or two conventionalresponses to the stimulus, the individ-ual's associative strengths to otherwords or ideas (lower in the hier-archy) drops rapidly. We can alsoconceive of a second sort of individualwhose associative hierarchy is charac-

Page 4: The Associative Process of Creativity

THE ASSOCIATIVE BASIS OF THE CREATIVE PROCESS 223

terized by a rather flat slope. This isan individual who perhaps also has ashis strongest response the conventionalchair. But for him this response is notoverly dominant and so it is morelikely that he will be able to get to theless probable, more remote kinds ofassociations to table. It is among thesemore remote responses that the requi-site elements and mediating terms fora creative solution will be lurking.This slope factor may be related tothe mathematical analysis of associa-tive production developed by Bous-field, Sedgewick, and Cohen (1954).It probably is closely approximatedby their constant, m, measuring rate ofdepletion of the associative reservoir.They found a high negative correlationbetween rate of association and totalnumber of associations. It would bepredicted from Figure 1 that the highcreative subject (flat hierarchy) wouldrespond relatively slowly and steadilyand emit many responses while thelow creative subject (steep hierarchy)would respond at a higher rate butemit fewer responses.

It would be predicted that thegreater the concentration of associativestrength in a small number of stereo-typed associative responses (steephierarchy) the less probable it is thatthe individual will attain the creativesolution. Thus, the word associationbehavior of the high creative individualshould be characterized by less stereo-typy and commonality. This last pre-

AssociatlMRaicoflMStrength

Steep Asseeiatiw Hierarchy

Flat Associative Hierarchy

FIG. 1. Associative hierarchies around theword "table."

diction is supported by a study byMednick, Gough, and Woodworth(Mednick, 1958). Research scientistsrated for creativity were divided intorelatively high (N = 15) and rela-tively low (N = 15) groups. The lowcreatives gave more stereotyped re-sponses on 80% of a group of 36 testwords from the Kent-Rosanoff list.(These test words were chosen fortheir tendency to elicit stereotyped re-sponses. Stereotypy was denned bythe Minnesota Kent-Rosanoff WordAssociation Norms, Russell & Jen-kins, 1954). It should be pointed outthat these results lend themselves toanother possible interpretation. Thehighly creative individual may alsohave a steep hierarchy but a deviantone. That is, his most dominant asso-ciative response may be quite strongbut quite different from the popular,dominant associative response. Thereare different predictions that can bemade for the flat-associative-hierarchycreative and the steep-deviant associa-tive-hierarchy creative. The latter ismore likely to be the one-shot pro-ducer (a not uncommon phenomenonamong novelists). If he does createfurther products, they will tend to re-semble closely the first product. Theformer is more likely to be a multi-producer ; he is more likely to producein a variety of avenues of creative ex-pression.

The prediction suggesting an expec-tation of less creativity from an in-dividual with a high concentration ofassociative strength in a few responsesleads to another prediction. Thegreater the number of instances inwhich an individual has solved prob-lems with given materials in a certainmanner, the less is the likelihood ofhis attaining a creative solution usingthese materials. Such an individualwill "know the meaning" of the ele-ments of the subject matter. That is, he

Page 5: The Associative Process of Creativity

224 SARNOFF A. MEDNICK

will have a steep associative hierarchyaround these elements. An exampleof the operation of this principle re-cently occurred to the writer whileteaching an honors freshman introduc-tory course in psychology at the Uni-versity of Michigan. I was giving awell known interpretation of a wellknown experiment in stimulus general-ization when interrupted by a studentwho calmly stated that the interpreta-tion was in error. After a few minutesof blustering I asked him to explain.His explanation proved him to be cor-rect. I had been dealing with thismaterial for years and "knew" the"correct" interpretation; for him thismaterial was new, he had a low, flatassociative hierarchy. Thus, if a new-comer to a field has the requisite in-formation, he is more likely to achievea creative solution than a long-timeworker in the field. This may be thereason that theoretical physicists andmaster chess players are often said tohave passed their prime by the age of25.

Number of Associations

The greater the number of associa-tions that an individual has to the re-quisite elements of a problem, thegreater the probability of his reachinga creative solution. This variable isnot independent of the preceding onesince an individual with a high con-centration of associative strength infew associative responses is not likelyto have a proliferation of associations.The more associates which are evokedby a requisite element of a problem,the more likely it is that an associatewill exist which will serve as a medi-ating bridge to another requisite ele-ment, facilitating combination. Itseems likely that this variable will notbe related to speed of creative solutionsince it may take a good deal of timeto get to the mediating links.

Cognitive or Personality Styles

Previously learned or innately pre-disposed methods of approaching prob-lems will influence the probability ofa creative solution. If the requisiteassociational elements of a new anduseful combination are probable asso-ciates of the concrete representationsof relevant aspects of the problem, anindividual with a predominately "per-ceptual" approach will be more likelyto reach a creative solution. If, how-ever, the requisite associational ele-ments are not elicited as responses tothese concrete representations or ifthere is no concrete representation thenan individual with a "conceptual" ap-proach will be more likely to reach acreative solution.

Another cognitive style of impor-tance may lie along the "visualizer-verbalizer" dimension. The visualizeris one who tends to call up relativelycomplete memorial sensory representa-tions of the relevant concrete aspectsof problems. If the problem deals withhorses, he tends to picture a horse interms of its sensory qualities. On theother hand, the verbalizer explores theproblem by associating with wordsaround the word "horse." If the req-uisite elements are high in his verbalassociative hierarchy to the wordhorse, the verbalizer will be more likelyto attain a creative solution; the visual-izer may be thrown off or at least de-layed by many false leads. On theother hand, if a requisite verbal asso-ciative response to the word horse isvery low, or not present in the ver-balizer's hierarchy, then the visualizerwill be more likely to attain the crea-tive solution. It is therefore clear thatsome types of problems will be solvedmore easily by the visualizer and someby the verbalizer.

Factors such as these (admittedlyvery poorly denned) may be partly re-

Page 6: The Associative Process of Creativity

THE ASSOCIATIVE BASIS OF THE CREATIVE PROCESS 225

sponsible for differential aptitudes forcreative work in differing fields.

Selection of the Creative Combination

The creative combination of elementsis only one among the many which maypresent themselves to the subject. Howor why is the creative combination se-lected? Some speculations regardingthis problem follow. The explanationof the process of selection may be re-lated to the nature of the problem.Problems either entail a specific andrelatively objective set of testable cri-teria (Paint a realistic portrait of thisindividual. Design a refrigerator sothat it will be automatically free offrost.) or they do not (The chemistmixes two liquids out of curiosity. Thepainter dabs hopefully at a fresh canvaswaiting for an idea. The psychologisttosses a new test into a correlationmatrix). When specific criteria areprovided, they form an important partof the stimulus set which is determiningwhich associative elements are beingelicited and thus becoming eligible forentering into combination with otherelements. Important sets of associa-tions to each of these combinations arethe consequents of the combinations.The set of consequents for each com-bination (If I put x, y, and z together,a and b will happen) is continuallycompared with the set of requirementsof the problem. When the set of conse-quents of a new combination achievesa close fit with the set of problem re-quirements, this combination is selected.When there is complete overlap of sets,"search behavior" is terminated. Aswith the other requisite elements of theproblem, individual differences in thiscase will vary with (among otherthings) the structure of the associa-tional hierarchies around the require-ments of the problem. When therefrigerator-defroster problem was pre-

sented to an undergraduate class almostall of the preferred solutions were basedon the principle of ridding the refriger-ator of already heavily accumulatedfrost. A couple of individuals (possi-bly familiar with the defroster principlespresently in use) suggested methodswhich disposed of the frost before itbuilt up to an overly annoying level.In addition to these there were twounique responses, i.e., a "new" methodof preventing moisture from condens-ing in the freezer compartment, and amethod of allowing frost to accumulatebut limiting the location of accumula-tion to a small box which could be regu-larly and conveniently removed andemptied. Thus it may be seen that anindividual's associations to the require-ments may be characterized as to theirstereotypy; the imposed requirementsof the problem may be viewed as partof the requisite elements in the situa-tion. The earlier theoretical statementsconcerning these elements may be seenas being relevant here. The foregoingsuggests an explanation of the selectionprocess for the case where the subjectmust hunt for a combination of ele-ments which will satisfy given criteria.In the case where no criteria are speci-fied, the subject is typically producingrandom combinations of elements; thetask of selection in this case consists infinding relevant criteria for the givenpartial products.

If we may continue along a bit fur-ther with this example of the defroster,we may begin to see some glimmeringsof a solution to the most serious prob-lem in research on creative thinking—how may we determine to what degreebehavior is creative? We have sug-gested one criterion in our hypothesis.In the following an additional criterionis developed. To begin with let us ex-amine the requirements as originallystated—"Design a refrigerator so thatit is automatically free of frost." The

Page 7: The Associative Process of Creativity

226 SARNOFF A. MEDNICK

first thing that strikes us is that whilesome requirements have been stated,there are even more that are stronglyimplied and essential, many that aredesirable, and a number that we wouldonly become aware of after somemethod of satisfying them had beensuggested.

Let us examine some possible so-lutions :

1. Simply refraining from openingthe refrigerator door would solve theproblem as stated since this would pre-vent moisture from entering and con-densing as frost. This solution meetsmany of the implied requirements. Itis cheap, convenient, effective, doesnot require special training, etc. . . .However, it is not an optimal solutionsince it violates one essential, impliedrequirement—the usefulness of therefrigerator must not be impaired.(This is the cutting-off-your-nose-to-spite-your-face solution.)

2. A primitive solution is thehammer-and-screwdriver method. Thisis tried and true and meets many of theessential requirements. It falls downin that it is inconvenient, messy, un-economical (when caked with frost, therefrigerator unit is very inefficient), en-dangers the mechanism, and is hardlyautomatic.

3. In a refrigerator we once ownedanother solution was used. The open-ing and closing of the refrigerator dooroperated a counter. At a certain countthe refrigerator unit was automaticallyheated and the melted water evaporatedoutside the refrigerator. The superi-ority of this solution is immediately ap-parent. The source of this superior-ity lies in the number of requirementswhich it meets. It is economical, auto-matic, convenient, peculiarly appropri-ate (the operation of the heatingelement is contingent upon the numberof door openings. The amount of frostaccumulated is also in part dependent

on the number of door openings.), doesnot interfere with the normal use of therefrigerator, and does not require spe-cial training. Note that the principlebehind this highly creative solution(not allowing massive build-ups offrost) was infrequently suggested inthe classroom group. However, thissolution is not wholly successful atmeeting some criteria. The frequentheating and cooling may injure frozenfood stored near the heating element.Secondly, since the heating processmust be brief and mild, it is inevitablethat not all frost is removed. Whilethis solution does effectively curtail thenumber of defrostings, it does not elim-inate them completely. It is clear thata method which would encompass all ofthe advantages of the "counter" method,but which would, in addition, eliminatedefrosting altogether would be evenmore creative. What is suggested bythis discussion is that the creativenessof a product is some function of thenumber of requirements that the prod-uct meets. The most ready applicationof this definition will be in laboratoryresearch in which tasks, solutions, andrequirements may be arbitrarily con-structed and varied.

A TEST OF CREATIVITY

The definition of the creative processhas suggested a way of testing for indi-vidual differences in creativity. Thetest items are intended to require thetestee to perform creatively. That is,he is asked to form associative elementsinto new combinations by providingmediating connective links. Since thetest situation is contrived, the combina-tion must meet specified criteria thatare experimenter imposed.

The definition dictates the structureof the test. We must provide stimulusitems from two mutually distant reali-ties and ask the subject to "draw a

Page 8: The Associative Process of Creativity

THE ASSOCIATIVE BASIS OF THE CREATIVE PROCESS 227

spark from their juxtaposition." Tostate it more usefully, we must providestimulus elements from mutually re-mote associative clusters and have thesubject find a criteria-meeting medi-ating link which combines them. Afirst problem concerns the type of ma-terial of which the stimulus item shouldbe composed. If the test is to be ap-propriate for all fields of creative en-deavor, the material must either benonsensical so as to avoid bias favoringany specific means of creative expres-sion, or it must be so common in so-ciety that familiarity could be assumedto be high across fields of interest. Theproblems involved in constructing thenonsense materials so as to avoid favor-ing any interest groups soon proved tobe apparently insurmountable. Thisleft us searching for materials withwhich most individuals in the culturecould claim acquaintance; this, in turn,brought us to verbal materials.

While it may be true that certainoccupational groups have extensive ex-perience in dealing with words, thereare some verbal associative habits thatcould reasonably be assumed to be fa-miliar to almost all individuals thathave been brought up in this (USA)culture. Among such habits are theassociative bonds between words like"ham and eggs," "bed-bug," "pool-hall,""hound-dog," "whole-wheat," "chorus-girl," "kill-joy," and "red-hot." Thesebecame the materials for the test.

Having decided on the materials, thetest almost constructed itself in accord-ance with the definition. Several wordsfrom mutually distant associative clus-ters must be presented to the subject;his task must be to provide mediatinglinks between them. Further, (a factorof extreme importance), the mediatinglink must be strictly associative ratherthan being of a sort that follows elabo-rate rules of logic, concept formation,or problem solving. In their final (or

at least present) form, the test itemsconsist of sets of three words drawnfrom mutally remote associative clus-ter. One example might be:

Example 1: rat blue cottage

The subject is required to find a fourthword which could serve as a specifickind of associative connective link be-tween these disparate words. The an-swer to Example 1 is "cheese.""Cheese" is a word which is present inthe word pairs "rat-cheese," "blue-cheese," and "cottage-cheese." Thesubject is presented with several ex-amples so that he has an adequate op-portunity to achieve the specific setnecessary for the task.

Example 2: railroad girl classExample 3: surprise line birthdayExample 4: wheel electric highExample 5 : out dog cat2

(None of these examples is a test itemfrom any form of the actual test.) Thetwo college level forms of the test (onecoauthored by Sharon Halpern and theother by Martha T. Mednick) have 30items each; the subject is allowed 40minutes; his score is the number right.

The test, called the Remote Associ-ates Test (RAT), has some interestingcorrelations with other measures.

Comparisons with Criteria. A studywas conducted at the College of Archi-tecture, University of California, Berke-ley, by the writer and Sharon Halpern.Ratings of creativity by faculty mem-bers of the College who taught the De-sign courses were correlated with RATscores. These ratings form an unusu-ally excellent criterion of creative per-formance since the raters had beenadvising and evaluating the students inthe creation of new designs and modelsof structures. They had been workingwith these students for at least a year

2 Answers to sample RAT items: 2.working; 3. party; 4. chair or wire; 5.house.

Page 9: The Associative Process of Creativity

228 SARNOFF A. MEDNICK

and in many cases two or more. Theratings and RAT scores correlated sig-nificantly (r = .70, df -19, p< .01)In this study an early form of the RATwas used.

The RAT was administered to agroup of first year psychology graduatestudents at the University of Michiganwhose native language was AmericanEnglish (JV = 35). Faculty researchsupervisors (who had been directingthe independent research efforts of thestudents), rated the eight highest andeight lowest RAT scorers either "high"or "low" in research creativity (nomiddle category allowed). Researchcreativity was defined as being demon-strated if the student developed newresearch methods and/or pulled to-gether disparate theory or researchareas in useful and original ways. Ofthe 16 research supervisors, one feltthat he had not had enough contactwith his student to make the judgment.His student was a low RAT scorer.Of the eight high RAT scorers, sixwere rated high on research creativityand two were rated low; of the sevenlow RAT scorers, only one was ratedhigh, the other six being rated low.By Fisher's exact test the probabilityof these events occurring by chance isless than .05. Miller Analogies Test(MAT) scores were available for thesestudents. Of the seven high MATscorers, three were rated high on re-search creativity; of the eight lowMAT scorers, four were rated high inresearch creativity.

Reliability. The Spearman-Brownreliability of the RAT was .92 in onesample (289 women, almost all thestudents at an Eastern women's college,tested as part of a project under the di-rection of Theodore Newcomb) and.91 in another (215 men tested at theUniversity of Michigan as part of aproject under the direction of WarrenT, Norman),

Correlation with Grades. One of thepresent forms of the RAT correlatednegatively with the first-two-year gradepoint averages of a group of under-graduates at a large Eastern technologycollege, (r = -.27, N = 74, p < .05).This same correlation was obtainedwith the summer grades of a smallergroup, N = 34, of summer students ata large Eastern liberal arts college (notstatistically significant in this case).In a study by Miller (1960) it wasfound that high RAT scorers tended toget higher grades from teachers ratedas flexible than from teachers rated asdogmatic. Low RAT scorers receivedhigher grades from teachers rated dog-matic than from teachers rated asflexible.

Correlation with Social Attitudesand Occupational Interests. It is clearthat creative individuals must haveaccess to improbable associative re-sponses. Kowalski (1960) hypothe-sized that this is a general tendencywhich also manifests itself in their atti-tudes and interests. She tested andinterviewed 15 high RAT scoring and15 low RAT scoring undergraduatewomen. The two groups had radicallydiffering views on sexual morality andwomen's rights. The views of the highcreatives were more atypical and"liberal" (£7 = 37, / > < .001). Onthe Strong Vocational Interest Blank,Metis' Form (SVIB), the high creativegroup showed "significantly higher in-terest on the artist (p < .05), psycholo-gist (/> < .005), physician (p < .025),mathematician (p < .025), and author-journalist (p < .05) keys. The lowcreative group showed higher intereston the farmer (p < .05), math-physicalscience high school teacher (p < .05),office man (p < .05), and pharmacist(p < .01) keys" (p. 19). (These arethe probability values of obtained chisquares.) The only one of these keysrelated to ACE scores was that of

Page 10: The Associative Process of Creativity

THE ASSOCIATIVE BASIS OF THE CREATIVE PROCESS 229

physician. "The commonality of theseinterest patterns was evaluated by-noting the per cent overlap of thespecific key with the general popula-tion of interest expression. For ex-ample, the Farmer key overlaps 45%with the general population, while theArtist key overlaps 20%" (p. 20).The significant keys of the highercreatives had significantly less com-monality than the significant keysof the low creatives. These differ-ences were independent of the influenceof intelligence as measured by the ACE,

Associative Behavior. In the discus-sion of illustrative predictions it wassuggested that highly creative individu-als would be characterized by a flat as-sociative hierarchy rather than a steepassociative hierarchy. Further, it wasproposed that the greater the numberof associations that an individual has tothe requisite elements of a problem, thegreater the probability of his reachinga creative solution. From these twoindependent statements it may be de-duced that when required to display hisreservoir of associations to single stim-ulus words, the highly creative indi-vidual will have greater access to lessprobable associates and therefore pro-duce a greater number of associates. Astudy by Craig and Manis (1960 un-published 3) supports this deduction.Thirty-eight college students had theRAT and an associative task admin-istered to them. In the associative taskthey were given 1 minute to write asmany associates as they could to eachof 20 words. The correlation of thenumber of such associates with RATscores was .38 (/> < .01).

In two related studies, Karp (1960)and Kowalski (1960) found RATscores to be directly related to the or-

3 Craig, M., & Manis, M. Prediction ofscores on the Remote Associates Test bysize of response repertoire. Unpublishedmanuscript, I960.

iginality and quantity of anagrams con-structed using the test word "Genera-tion." In the Karp study 40 under-graduates were given 5 minutes to pro-duce as many four letter anagrams fromthe test word as they could. The pro-ductions were scored for quantity(number of acceptable answers) andoriginality (a weighted score for eachresponse was developed from the fre-quency with which the response wasgiven by the 40 subjects). The correla-tion of the RAT with the quantityscores was .44 (p < .01); the correla-tion of the RAT with the originalityscore was .37 (p < .05). Kowalskipresented the anagrams task to 15 highRAT scorers and to 15 low RATscorers, giving them 5 minutes to pro-duce words of any length from the testword "Generation." In this study orig-inality was measured by computing thepercentage of responses given by an in-dividual which had not been given byany other of the 30 subjects. The dif-ference on this measure between highand low RAT scorers was significant(U = 68, p < .04). "Only four sub-jects in the low creative group gaveany original responses at all whileeleven subjects in the high creativegroup did" (p. 19).

At the Institute of Personality As-sessment and Research, University ofCalifornia, Berkeley, the RAT was in-cluded as part of the assessment batteryadministered to a group of 40 highlyeminent architects. The RAT corre-lated .31 with the Originality (O-I)Scale of the IPAR Questionnaire Scaleand —.31 with the total ConformityScore obtained in the CrutchfieldConformity Experiment (Crutchfield,1955). Interviewers rated high scorersas significantly higher in "graceful andwell-coordinated in movement" and"reticent and taciturn in speech." Thecollege grade point average which thesubjects reported correlated —.34 with

Page 11: The Associative Process of Creativity

230 SARNOFF A. MEDNICK

RAT score, a result which tends toconfirm findings reported above.

EXPERIMENTALLY MANIPULABLEVARIABLES

While only one experimental study(described below) which makes use ofthis general framework has been com-pleted in this laboratory, it may be use-ful to indicate briefly the kinds of ex-perimental investigation it suggests.

Massed vs. Distributed Work Ses-sions. Total time of work being equal,massed sessions of creative work shouldbe more successful than distributed ses-sions. There are two reasons why thiswould be so. The first is that the indi-vidual making use of the massed sessiontechnique is more likely to achieve tem-poral contiguity of the requisite associ-ative elements within a single intensivework period than is an individual whohas distributed his work in shorterperiods over several days. Secondly, itmay take some time for an individualto work on a problem enough to gobeyond its obvious aspects. In the firsthour of work, he may get through onlythe conventional and stereotyped asso-ciations to the elements of the problem,while it is perhaps in the later stages ofintensive work on a problem that onecan begin to entertain the more remoteassociations that are evoked by ele-ments of the problem. It is, of course,among these remote associations thatthe key to the creative solution will lie.

Warmup. In creative work a warmupsession should serve to arouse the moreremote associations to the requisite ele-ments of the problem. While theirwork has gone considerably beyond theproblem of warmup, Maltzman, Bo-gartz, and Breger (1958) have demon-strated that the repeated elicitation ofdifferent word associations to the samestimulus words does indeed tend toproduce remote associations to thesestimulus words. Further, this induced

originality tends to transfer to otherrelatively unrelated tasks presentedafter this associative warmup. Associ-ative warmup of this type should be-come more effective as the warmupstimuli are more similar to the task ma-terials. It may be that the effects ofwarmup will prove to be a further ad-vantage that massed sessions have overdistributed sessions for creative pro-ductivity.

Stereotyping Associative Responses.As stated above, if an individual's as-sociative response to a stimulus elementof a creative problem is of excessivestrength, this will tend to reduce thelikelihood of occurrence of more remoteassociative responses. This will reducethe probability and speed of creative so-lution. It would therefore be predictedthat extensive training of a specific re-sponse other than a requisite one to astimulus element of a problem requiringa creative solution should retard laterattempts at solution of the problem.This prediction is related to the con-cept of "functional fixedness" intro-duced by Duncker (1945). Birch andRabinowitz (1951) and Adamson andTaylor (1954) completed experimentswhich are related to this prediction.Their test situation was the two stringproblem. The subjects are asked totie together two strings suspended fromthe ceiling. When the subject graspsone string he finds that the other stringis hanging out of his reach. The solu-tion to the problem requires the sub-ject to attach a weight to one of thestrings, get the weight swinging andcatch it while holding the other string.Various objects can be used as aweight. The subjects that had beenpretrained by having them use a switchin its usual manner of functioningtended to be unlikely to use it as aweight. They had developed strong re-sponse strength for the association"switch-close circuit" which had re-

Page 12: The Associative Process of Creativity

THE ASSOCIATIVE BASIS OF THE CREATIVE PROCESS 231

duced the probability of the remote as-sociation "switch-weight."

Another feasible experimental ap-proach would make use of the RATitem as a creative task and test the in-fluence upon it of certain variables.For example, the words of which anitem is composed may be presented atvarying rates to test the massed trialshypothesis. In addition, various pre-training conditions may be evaluated interms of their effectiveness in increasingor decreasing RAT performance.

Another possible experimental ap-proach would entail separating out highand low RAT scorers and observingthe differential effect of certain vari-ables upon their behavior. In an ex-periment just completed Houston andMednick (in press) postulated that animportant motive impelling the be-havior of the creative individual was aneed for improbable associative stimu-lation. It was reasoned that if suchstimulation were supplied, it wouldtend to satisfy this need and be rein-forcing. Further, if such stimulationregularly followed a given response thehigh creative individual should tend tolearn that response. Thirty high and30 low RAT scorers were asked toread aloud only one of two typedwords on a 3 x 5 card. Exceptingbuffer items and including 40 pairsaimed at gauging the free operantlevel of noun-choice, there were 160pairs of words, each pair consisting of anoun and a nonnoun (verb, adjective,adverb, etc.). If a subject in the experi-mental group (IS high RAT subjects,15 low RAT subjects) responded withthe noun member of a pair, the experi-menter responded with an improbableassociation; if the subject chose thenonnoun, the experimenter respondedwith the most probable association. Inthe control group (15 high RAT, 15low RAT) both the nouns and thenonnouns were invariably followed by

their most probable associate. Asso-ciative probabilities were obtained fromthe Russell and Jenkins (1954) andDeese (1960) norms. If the improba-ble response was satisfying a need,the probability of noun-choice shouldincrease over the 160 trials. It didsignificantly in the high RAT experi-mental group; the low RAT experi-mental group showed a decrease. Thehigh and low RAT control groupsshowed no reliable change.

Some of the positions which havebeen taken in this paper are assump-tions and not deductions. As moredata are gathered some of these as-sumptions will assume the status offacts, some will be revised. For ex-ample, the opening paragraph suggeststhat the paper is not meant to applyonly to one field of creative endeavorbut attempts to delineate processesthat underlie all creative thought. Thismay require modification. The expla-nation may fit the process of scientificdiscovery and not be appropriate todiscussions of painting or music. Forthe present (paradoxically enough), themore encompassing assumptions seemmore parsimonious. It may eventuallyturn out (as is hinted at in the body ofthe paper) that the differences betweenthe fields are more determined by dif-ferences in suitability of the threemeans of achieving contiguity, i.e.,serendipity, similarity, and mediation.

SUMMARY

An associative theory of creativethinking has been outlined. Differencesbetween high creatives and low crea-tives have been predicted along speci-fied dimensions. Predictions have beenmade regarding the effect on the crea-tive process of some experimentallymanipulable variables.

The associative definition of the crea-tive process has taken the operational

Page 13: The Associative Process of Creativity

232 SARNOFF A. MEDNICK

form of a test. Some preliminary re-search with this test is described.

REFERENCES

ADAMSON, R. E., & TAYLOR, D. W. Func-tional fixedness as related to elapsed timeand to set. /. exp. Psychol, 1954, 47,122-126.

BAIN, A. The senses and the Intellect. 1855.BINET, A. The psychology of reasoning.

Chicago: Open Court, 1899.BIRCH, H. G., & RABINOWITZ, H. S. The

negative effect of previous experience onproductive thinking. /. exp. Psychol.,1951, 41, 121-125.

BOUSFIELD, W. A., SEDGEWICK, C. H. W., &COHEN, B. H. Certain temporal character-istics of the recall of verbal associates.Amer. J. Psychol, 1954, 57, 111-118.

CRUTCHFIELD, R. Conformity and character.Amer. Psychologist, 1955, 10, 191-198.

DEESE, J. Word association norms. Author,1960. (Ditto)

DUNCKER, K. On problem-solving. Psy-chol. Monogr. 1945, 58(5, Whole No.270).

FREUD, S. Wit and its relation to the un-conscious. In, The basic zvritings of Sig-mund Freud. New York: Modern Library,1938.

GHISELIN, B. The creative process. Berke-ley : Univer. California Press, 1952.

ROLLINGS WORTH, H. L. Psychology: Itsfacts and principles. New York: Appleton,1928.

HOUSTON, J. P. & MEDNICK, S. A. Creativ-ity and the need for novelty. /. abnorm.soc. Psychol., in press.

KARP, S. A validity study of a measure ofcreativity. Senior honors thesis, Univer-sity of Michigan, 1960.

KOHLER, W., & FISHBACK, J. The destruc-tion of the Muller-Lyer illusion in repeatedtrials: II. Satiation patterns and memorytraces. /. exp. Psychol., 1950, 40, 398-410.

KOWALSKI, J. Attitudes and occupationalinterests of creative individuals. Seniorhonor thesis, University of Michigan, 1960.

LOCKE, J. Essay concerning the human un-derstanding. 1690.

MALTZMAN, I., BOGARTZ, W., & BREGER, L.A procedure for increasing word associa-tion originality and its transfer effects./. exp. Psychol, 1958, 56, 392-398.

MEDNICK, S. A. An orientation to researchin creativity. (Res. Memo. No. 2) Berke-ley, Calif.: University of California, Insti-tute of Personality Assessment and Re-search, 1958.

MILLER, B. A study of creativity in collegestudents and teaching method types. Sen-ior honor thesis, University of Michigan,1960.

MOORE, MARIANNE. The monkey puzzle. In,Collected poems of Marianne Moore. NewYork: Macmillan, 1951.

RUSSELL, W. A., & JENKINS, J. J. Thecomplete Minnesota norms for responses to100 words from the Kent-Rosanoff WordAssociation Test. Technical Report No.11, 1954, University of Minnesota, ContractN8 onr 66216.

UNDERWOOD, B. J. An orientation to re-search on thinking. Psychol. Rev., 1952,59, 209-220.

(Received September 22, 1960)