Individual Differences in Need for Cognitive Closure PDF

PERSONALITY PROCESSES AND INDIVIDUAL...

PERSONALITY PROCESSES AND INDIVIDUAL DIFFERENCES Individual Differences in Need for Cognitive Closure Donna M. Webster and Arie W. Kruglanski This article introduces an individual-difference measure of the need for cognitive closure. As a dis- positional construct, the need for cognitive closure is presently treated as a latent variable manifested through several different aspects, namely, desire for predictability, preference for order and structure, discomfort with ambiguity, decisiveness, and close-mindedness. This article presents psychometric work on the measure as well as several validation studies including (a) a "known-groups" discrimi- This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. nation between populations assumed to differ in their need for closure, (b) discriminant and con- This document is copyrighted by the American Psychological Association or one of its allied publishers. vergent validation with respect to related personality measures, and (c) replication of effects obtained with situational inductions of the need for closure. The present findings suggest that the Need for Closure Scale is a reliable and valid instrument of considerable potential utility in future "motivated social cognition" research. In this article, we describe a dimension of individual differ- effortful or otherwise costly, the need for closure may be, there- ences related to persons' motivation with respect to information fore, elevated. The need for closure may also be aroused when processing and judgment. This motivation is referred to as the the judgmental task appears intrinsically dull and unattractive need for cognitive closure. As used here, the term need denotes to the individual. Under such circumstances, closure may serve a motivated tendency or a proclivity rather than a tissue deficit as a means of escaping an unpleasant (hence, a subjectively (for a similar use see Cacioppo & Petty, 1982). In previous the- costly) activity. ory and research (Kruglanski, 1989, 1990b; Kruglanski & Web- Functionally opposite to the need for closure is the need to ster, 1991) "need for closure" was defined in terms of a desire avoid closure. Those two needs are conceptualized as ends of a for "an answer on a given topic, any answer,... compared to continuum ranging from strong strivings for closure to strong confusion and ambiguity" (Kruglanski, 1990b, p. 337). Such resistance of closure (Kruglanski, 1989). The need to avoid clo- need was referred to as "nonspecific" and was contrasted with sure may stem from the perceived costs of possessing closure needs for "specific closure," that is, for particular (e.g., ego-pro- (e.g., envisioned penalties for an erroneous closure or perceived tective or enhancing) answers to one's questions. drawbacks of actions implied by closure) and the perceived ben- The need for (nonspecific) cognitive closure is assumed to be efits of lacking closure (e.g., immunity from possible criticism proportionate to the perceived benefits of possessing closure, of any given closure). the perceived costs of lacking closure, or both. For instance, clo- The foregoing discussion suggests that need for closure may sure affords predictability and a base for action. Thus, need for vary as a function of the situation. Indeed, situational induc- closure may arise where predictability or action seem tions of need for closure have often been used in past research. important. Thus, Kruglanski and Freund (1983) found that elevating the Similarly, the absence of closure may seem costly in various need for closure through time pressure increased subjects' ten- circumstances. Thus, under time pressure the absence of clo- dency to succumb to primacy effects in impression formation, sure may imply the danger of missing an important deadline. render stereotypically driven judgments, and anchor judgments Hence, time pressure should elevate the need for closure. A on initial estimates, all presumed to represent various effects of different cost of lacking closure may stem from perceived labors of further information processing. Where processing is seen as the need for closure on the judgmental process. Similar time- pressure effects were obtained in research by Freund, Kruglan- ski, and Schpitzajzen (1985), Heaton and Kruglanski (1991), Jamieson and Zanna (1989), and Sanbomatsu and Fazio (1990). Donna M. Webster and Arie W. Kruglanski, Department of Psychol- Webster (1993) manipulated the need for closure through ogy, University of Maryland, College Park. varying the perceived attractiveness of an attitude-attribution This research was supported by National Institute of Mental Health task (Jones & Harris, 1967). When the task was perceived as Grant 5R01MH 4612-02. We wish to thank Alan Heaton for his assis- unattractive (rendering extensive processing of relevant infor- tance in the early development of the scale and Tom Ford for his assis- tance with data analyses. mation costly), subjects were more likely to exhibit the "corre- Correspondence concerning this article should be addressed to spondence bias" than when the task was perceived as moder- Donna M. Webster, who is now at the Department of Psychology, Uni- ately attractive. Furthermore, when the task was perceived as versity of Florida, Gainesville, Florida 32611. highly attractive (reducing the perceived costs of information Journal of Personality and Social Psychology, 1994, Vol. 67, No. 6, 1049-1062 Copyright 1994 by the American Psychological Association, Inc. 0022-3514/94/S3.00 1049  1050 DONNA M. WEBSTER AND ARIE W. KRUGLANSKI processing), the tendency to exhibit the correspondence bias Study 1: Psychometric Properties of the Need for was all but eliminated. Closure Scale (NFCS) Heightened need for closure may dispose persons to react negatively to individuals who disrupt closure (e.g., to opinion In developing the NFCS, our initial item-generation process deviates in a group). Indeed, Kruglanski and Webster (1991) attempted to capture a broad sense of the construct. In accor- found that in proximity to the decision deadline, or in the pres- dance with the underlying theory (Kruglanski, 1989, 1990a, ence of environmental noise, both assumed to enhance the de- 1990b), we reasoned that the need for closure may express itself sirability of closure, group members tended more to reject opin- in various ways. Thus, we treated it as a latent variable manifest ion deviates, and to be more evaluatively positive toward con- through different aspects (Carver, 1989). In particular, we iden- forming individuals who made special efforts on behalf of the tified five major such aspects assumed to broadly represent the consensual opinion. universe of the construct and generated diverse items corre- spondent with those aspects. Theoretically, persons with a high need for closure should de- sire definite order and structure in their lives and abhor uncon- This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. Need for Closure as an Individual-Difference Dimension This document is copyrighted by the American Psychological Association or one of its allied publishers. strained chaos and disorder. Accordingly, one subset of items we Though need for closure may vary as a function of the situa- have selected assessed the extent to which individuals professed tion, it may also represent a dimension of stable individual a preference for order and structure in their environment (e.g., differences. This possibility is explored in the present research. "I think that having clear rules and order at work is essential for The development of an individual-difference measure of the success"). We included in this groupfiveitems (namely, NFCS need for closure promises to offer three distinct advantages. items 6, 10, 32, 33, and 35) from a previous instrument based First, it affords a desirable cross validation of our previous, sit- on the lay epistemic theory (Kruglanski, 1989) referred to as the uational, inductions of this motivation. Theoretically, individu- Personal Need for Structure Scale (M. Thompson, Naccarato, als who score high on our personality measure of the need for Parker, & Moskowitz, 1993). A second item subset pertained closure should exhibit similar judgmental patterns to those put to the affective discomfort occasioned by ambiguity, that is, an under need for closure through such situational manipulations absence of closure (e.g., "I'd rather know bad news than stay in as time pressure, environmental noise, or task attractiveness. A a state of uncertainty"). We assumed that individuals with a comparison of results from a personality measure and situa- high need for closure would experience as aversive situations tional inductions represents a rigorous test of the need for clo- devoid of closure, in which their motivation is frustrated. A sure construct. third subset of items tapped the urgency of striving for closure in judgment and decision making (e.g., "I usually make impor- Secondly, a development of an individual-difference measure tant decisions quickly and confidently"). We assumed that per- allows a richer exploration of the various subjective manifesta- sons with a high need for closure would experience an urgent tions of the need for closure. In previous studies, assessment desire to reach closure, reflected in a decisiveness of their judg- of need for closure often amounted to "manipulation checks" ments and choices. Three items in this group (namely, NFCS designed to tap whether the intended experimental conditions items 15, 22, and 37) were taken from M. Thompson, Nacca- (e.g., of task attractiveness or unpleasant noise) were created. rato, Parker, and Moskowitz's (1993) Personal Fear of Invalidity Construction of a specific personality scale affords the opportu- Scale. nity for a more differentiated and complex conceptualization of The fourth and fifth item subsets pertained to the desire for ways in which the need for closure may be subjectively experi- secure or stable knowledge, assumed to increase under high enced. Beyond its methodological significance then, an individ- need for closure. A secure knowledge is one that can be relied ual-difference measure represents an opportunity for theoreti- on across circumstances and is unchallenged by exceptions or cal refinement. disagreements. Specifically, the fourth item subset tapped the Finally, an individual-difference measure of the need for cog- trans-situational-consistency implication of secure knowledge, nitive closure enables the allocation of individual variance to a affording predictability to future contexts (e.g., "I don't like to personality main effect and to the person-situation interaction, go into a situation without knowing what I can expect from it"). reducing error variance and enhancing the statistical power for This subset too contained several items (namely NFCS items assessing situational effects (cf. Eysenck, 1954). 3, 5, and 19) from the Personal Need for Structure Scale (M. In the following paragraphs, we describe a program of re- Thompson, Naccarato, Parker, & Moskowitz, 1993). The fifth search in which an individual-difference measure of the need subset tapped the close-mindedness that the desire for secure for closure is developed and tested. We first describe the item closure may induce, that is, an unwillingness to have one's selection process and the basic psychometric work on the scale. knowledge confronted (hence, rendered insecure) by alternative Subsequently, we report on discriminant validation of the scale opinions or inconsistent evidence (e.g., "I do not usually consult against alternative personality measures. We then report a vali- many different opinions before forming my own view"). dation of the scale through the "known-group" method and a Items in all the foregoing categories were intended to tap di- series of experiments examining the relation of the scale scores verse assumed manifestations of the need for closure. As our to a variety of cognitive measures. This last set of studies at- theoretical interest was in this latent construct as such, its extent tempts to replicate by means of our individual-difference mea- was assessed additively across the different item categories sure a variety of previousfindingsobtained through situational (Carver, 1989). inductions of the need for closure. The 42 items composing the NFCS are presented in Krug-  NEED FOR CLOSURE 1051 lanski, Webster, and Klem (1993). Subjects indicate the extent eliminated. The resultant factor loadings are presented in to which they endorse each item by responding to a 6-point Table 1. Likert scale ranging from 1 (strongly disagree) to 6 (strongly Reliability. Cronbach's alpha was recalculated for the com- agree). Items 2, 5, 12, 15, 17, 18, 19, 20, 22, 24, 27, 28, 34, posite scale after removing each item in turn and dropping 37, 38, and 42 were designed to tap respondents' need to avoid items that substantially reduced internal consistency. As noted closure; hence, these items are reverse scored. Items tapping the earlier, 15 items in all were dropped from the original scale on need for closure were negatively correlated with those tapping the basis of this and the earlier item analyses. the need to avoid closure (r = -.4566, n = 281, p <.01), sup- Additional analyses indicated that the revised, 42-item scale porting a bipolar conceptualization of the need for closure con- possesses high internal consistency (Cronbach's a =.8405) as struct (Kruglanski, 1989). Respondents' composite need for well as high test-retest reliability (r =.8611). Cronbach's alpha closure score is calculated by summing across each of the indi- for each of the item subsets ranged from.62 to.82 and are pre- vidual items (after reverse scoring the appropriate items). sented in Table 2. The original form of the scale included a total of 57 items. Factor structure. Our theory predicts that the need for cog- This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. On the basis of various item analyses, 15 of those items were nitive closure is a unitary latent variable, potentially manifest This document is copyrighted by the American Psychological Association or one of its allied publishers. dropped, leaving 42 items that make up the revised composite in various ways. Thus, we expected that a confirmatory factor scale. The specific analyses, on which basis the original scale analysis would support a single-factor model as the best fit to was revised, are described next. our data. However, because items were generated as part of five The NFCS was administered to two independent, divergent general domains, we expected greater interitem correlation groups of individuals. Thefirstgroup, henceforth referred to as within each of those facets. This expectation is consistent with the student sample, consisted of psychology undergraduates. results of the exploratory factor analysis. Hence, the model we The second group, referred to as the library sample, consisted hypothesized as providing the best fit to our data was a one- of adults at public libraries. Our purpose was to examine factor model that included a specification of correlated errors, whether the psychometric properties of the scale replicate that is, shared domain-specific variance within each of the five across different subject populations, and hence whether they facets. In other words, we expected to find support for a model may be considered of general utility. specifying a single coherent construct withfivefacets. The hypothesized model was evaluated through a confirma- tory factor analysis (through LISREL, SPSSX) where compari- Student Sample sons were made between the hypothesized model and models specifying alternative structural relationships among the scale Our student sample consisted of 146 female and 135 male items. In particular, we were interested in whether the hypothe- undergraduates in an introductory psychology course at the sized model would provide a better fit to the data than a model University of Maryland at College Park. They participated in specifying five correlated orfiveorthogonal factors that corre- the study to fulfill a course requirement. These students com- sponded to thefivedomains within which items were generated. pleted the 57-item form of the NFCS at the start of a 14-week Thus, model comparisons were conducted between the hypoth- semester. Mean composite scale scores did not differ signifi- esized model and each of the twofive-factormodels. Further- cantly for the two gender groups (M for men = 154.9; M for more, we compared the hypothesized single-factor model with a women = 153.94). To provide a means of assessing the test- single-factor model that did not include correlated errors within retest reliability of the scale, a subgroup (n = 49) of the original item domains. sample returned after a 12-13-week interval and completed the (revised) scale a second time. Table 3 summarizes tests of goodness of fit for each of the four competing models. According to goodness-of-fit indexes, the hypothesized model seems to provide the best fit to the data Item Selection among the four competing models. Results of chi-square difference tests between the competing models also indicate the Exploratory factor analysis. As part of the item selection hypothesized model provides a significantly betterfitto the data process, we began by conducting an exploratory factor analysis than any of the other competing models. Taken together, those (SPSSX, varimax rotation). We expected greater interitem cor- results support our hypothesis that the NFCS assesses a single relation within groups of items belonging to the same subset. latent variable, potentially manifest in various ways. Hence, a pattern of factor loadings consistent with a five-factor solution was expected. The scree plot presented in Figure 1 de- picts the pattern of eigenvalues. As noted by the rapid drop in Library Sample magnitude of eigenvalues after the fifth factor, five factors ac- Our second group of respondents consisted of 77 male and 95 count for a substantial amount of the variance (38%) and the female adults recruited at three different public libraries in the remaining factors appear less capable of accounting for vari- suburban Maryland area and who volunteered to participate in ance. Furthermore, the configuration of loadings on each of the the study. As with the first group, the mean composite scale five factors supports the predicted pattern, such that loadings of scores did not differ significantly for the two gender groups (M most items were relatively high on the factor correspondent to for men = 156.78; M for women = 153.35). Subjects ranged in the predicted subset and low or zero on the remaining factors. age from 24 to 56 years and were not affiliated with the Univer- In accordance with the simple structure criterion (Thurstone, sity of Maryland. They all completed the 42-item revised form 1942), items loading.30 or higher on more than one factor were of the NFCS.  1052 DONNA M. WEBSTER AND ARIE W. KRUGLANSKI 9.130 4.900 This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. This document is copyrighted by the American Psychological Association or one of its allied publishers. 3.090 2.161 1.592 1.359 1.144.975 * * * *.733 * * * * *.524 * * * * * * *.350.177 Figure 1. Scree plot for original Need for Closure Scale. Reliability. Reliability analyses replicate the earlier finding tain whether the NFCS taps a unique variable distinct from al- that the revised, 42-item scale has high internal consistency ternative relevant constructs. Essentially the latter constructs (Cronbach's a =.8413). Cronbach's alpha for each of the item fell into three categories: (a) overly inclusive constructs that be- subsets ranged from.63 to.80 and are presented in Table 2. side tapping aspects of the need for closure also tapped other Factor structure. A confirmatory factor analysis (through unrelated variables, (b) overly exclusive constructs that tapped LISREL, SPSSX) was performed on data from this group, and only some but not other aspects of the need for closure, and (c) the model comparisons conducted for the student sample were constructs seemingly unrelated to the need for closure, yet repeated. The results replicate our earlier findings. According whose pervasive relevance to various sociocognitive phenom- to goodness-of-fit indexes, the hypothesized single-factor model ena suggested the advisability of empirically probing their re- with correlated errors within domains seems to provide the best lations to this variable. fit to the data among the four competing models. Results of chi- The overly inclusive category contained scales measuring au- square difference tests between the competing models also indi- thoritarianism, dogmatism, intolerance of ambiguity, cognitive cate that the hypothesized model provides a significantly better complexity, and impulsivity. As we elaborate later, some aspects fit to the data than any of the other competing models. Those of those concepts are partially related to the need for closure, results are presented in Table 3. whereas other aspects are not. In summary, the results from our two samples of respondents The overly exclusive category contained scales measuring the suggest that the NFCS reliably assesses the need for closure con- need for structure and the fear of invalidity. Those scales, also struct. In addition, it seems to capture the single latent variable derived from the lay epistemic theory, tap some aspects of the proposed by lay epistemic theory. Furthermore, the high test- need for closure construct but not others. Finally, the category retest reliability observed over a 12-13-week period indicates of unrelated constructs of general relevance included scales that the personality construct tapped by the scale is relatively measuring social desirability, need for cognition, and stable. Finally, gender does not seem to affect one's score on the intelligence. NFCS. Method Study 2: The NFCS and Other Relevant Tests Subjects. One hundred fifty-seven male and female introductory The purpose of our second study was to examine the associa- psychology students at the University of Maryland participated in the tion between the NFCS and other relevant measures to ascer- study to fulfill a course requirement  NEED FOR CLOSURE 1053 Table 1 Table 2 Factor Loadings for Need for Closure Scale: Cronbach 's Alpha for Need for Closure Scale and Facets Exploratory Factor Analysis Group 2 Group 1 Factor loadings Facet (« = 281) («= 172) Items (n = 281) Total 42-item scale.8405.8413 Facet 1 (Structure).8216.7725 Factor 1: Preference for Order.7181 Facet 2 (Predictability).7867 Item 33.7462 Facet 3 (Decisiveness).7001.7882 Item 35.7139 Facet 4 (Ambiguity).6656.8002 Item 32.6712 Facet 5 (Closed Mind).6152.6166 Item 06.6512 Item 20 (reversed).6358 Item 25.5930 Item 01.4810 tolerance of Ambiguity Scale (Eysenck, 1954); the Personal Need for This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. Item 27 (reversed).4781 Structure Scale (Neuberg & Newsom, 1993; M. Thompson, Naccarato, This document is copyrighted by the American Psychological Association or one of its allied publishers. Item 42 (reversed).4388.3847 Parker, & Moskowitz, 1993); and the Personal Fear of Invalidity Scale Item 10 (M. Thompson, Naccarato, Parker, & Moskowitz, 1993). Subjects re- Factor 2: Preference for Predictability ceived one of four possible packets, each including a different random- ized order of tests. Item 19 (reversed).7244 A second group of tests was administered to 60 other subjects. First, Item 41.6203 the Quick Test (QT; Ammons & Ammons, 1962), which is a measure Item 25.5712.5509 of intelligence, was administered to subjects individually. In addition, Item 05 (reversed) Item 26.5286 subjects worked in groups ranging from 1 to 10 persons and completed Item 18 (reversed).5204 the Modified Bieri REP Test (Bieri, 1966), which assesses cognitive Item 11.4638 complexity, and the Control (vs. Impulsiveness) subscale of the Multi- Item 07.3727 dimensional Personality Questionnaire (Tellegen, 1982). The order in which subjects received each of the three tests was determined ran- Factor 3: Decisiveness domly. Correlations of each of the aforementioned measures with the Item 22 (reversed).7828 NFCS are summarized in Table 4. Item 17 (reversed).7283 Item 16.6771 Item 37 (reversed).6021 Item 13.5343 Table 3 Item 15 (reversed).5178 Goodness-of-Fit Indexes (GFIs) and Chi-Square Difference Item 12 (reversed).4806 Tests From Confirmatory Factor Analyses of the Need for Closure Scale Factor 4: Discomfort With Ambiguity Item 30.5730 Group 1 Group 2 Item 36.5493 Test (n = 281) (n=172) Item 08.5186 Item 31.4672 Test of 1-factor model/correlated errors Item 14.4402 within facets Item 29.4341 x2 1,097.00 1,335.40 Item 21.3977 df 661 661 Item 39.2262 GFI.868.755 Item 03.5559 Test of 5-correlated factor model 2 x 1,731.85 1,813.85 Factor 5: Closed-Mindedness df 809 809 GFI.796.670 Item 28 (reversed).7141 Test of 5-orthogonal factor model Item 24 (reversed).6207 2 x 2,025.31 1,924.66 Item 38 (reversed).6119 df 824 824 Item 40.4940 GFI.761.650 Item 02 (reversed).4698 Test of 1-factor model/uncorrelated errors Item 34 (reversed).4486 within facets Item 04.2290 x2 2,793.24 2,797.18 Item 09.2020 df 819 819 GFI.177.501 X2 difference for 1-factor model/correlated errors and 5-correlated factor model" 634.85 478.45 Procedure. Ninety-seven subjects working in small groups that x2 difference for 1-factor model/correlated ranged in size from 3 to 12 subjects completed a packet of question- errors and 5-orthogonal factor model" 928.31 589.26 x2 difference for 1-factor model/correlated naires including the NFCS; the F Scale, form 40 (Sanford, Adomo, errors and 1-factor model/ Frenkel-Brunswik, & Levinson, 1950); the Dogmatism Scale, form E uncorrelated errors* 1,696.24 1,461.78 (Rokeach, 1960); the Social Desirability Scale (Crowne & Marlowe, 2 1964); the Need for Cognition Scale (Cacioppo & Petty, 1982); the In- " All x differences are significant atp <.001.  1054 DONNA M. WEBSTER AND ARIE W. KRUGLANSKI Table 4 Correlations ofthe Needfor Closure Scale (NFCS) and Facets With Other Relevant Personality Measures Closed- Personality measure NFCS Order Predictability Decisiveness Ambiguity Mindedness Dogmatism Scale.2870**.3376**.2621*.2743*.3268**.1917 F Scale (authoritarianism).2660*.2929**.2312* -.1158.4028**.0979 Intolerance of Ambiguity Scale.2877**.3553**.2261* -.2230.3579**.0841 Fear of Invalidity Scale -.2109*.0379 -.1406 -.3867**.0202 -.1447 Need for Structure Scale.2355**.2763**.2737** -.2156.1316.0362 Need for Cognition Scale -.2831* -.3092** -.1420.1048 -.1291 -.3219** Bieri REP Test (cognitive complexity) -.2952* -.3256* -.1771 -.0750 -.1923 -.3061* MPQ Control Subscale (impulsivity).2683.2349.1507.2471.1429.1968 This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. Social Desirability Scale -.0181.0713.0326.0522 -.1022 -.1887 This document is copyrighted by the American Psychological Association or one of its allied publishers. Quick Test (intelligence) -.1710 -.2747 -.1213.0726 -.0248 -.0097 Note. Order = Preference for Order; Predictability = Preference for Predictability; Ambiguity = Discomfort With Ambiguity; MPQ = Multidi- mensional Personality Questionnaire. *P