Socially Desirable Responses in the Measurement of Need For Cognition

James M. Hunt, Temple University
Karen M. Stevens, Temple University
Anindya Chatterjee, Temple University
Jerome B. Kernan, George Mason University
ABSTRACT - Two studies suggest that subjects may feign a high need for cognition (NC) in a socially-desirable effort to appear "smart." In one study, the need for cognition scale (Cacioppo and Petty 1982; Cacioppo, Petty, and Kao 1984)Cboth long and short versionsCwas found to be correlated significantly with the social desirability scale (Crowne and Marlowe 1964). In a second, the scale was assessed experimentally by manipulating the motivational set under which subjects completed it. Findings revealed that subjects given normal instructions: (1) scored higher in NC than those instructed to "downplay" their intelligence in completing the scale; and (2) scored no higher or lower than subjects instructed to "demonstrate" their intelligence. In that NC is a key concept in Petty and Cacioppo's (1986) elaboration likelihood model (ELM), caution would seem appropriate when the NC scale is employed, particularly in those settings associated with cognitive achievement.
[ to cite ]:
James M. Hunt, Karen M. Stevens, Anindya Chatterjee, and Jerome B. Kernan (1994) ,"Socially Desirable Responses in the Measurement of Need For Cognition", in NA - Advances in Consumer Research Volume 21, eds. Chris T. Allen and Deborah Roedder John, Provo, UT : Association for Consumer Research, Pages: 543-546.

Advances in Consumer Research Volume 21, 1994      Pages 543-546

SOCIALLY DESIRABLE RESPONSES IN THE MEASUREMENT OF NEED FOR COGNITION

James M. Hunt, Temple University

Karen M. Stevens, Temple University

Anindya Chatterjee, Temple University

Jerome B. Kernan, George Mason University

[This research was supported by a Faculty Senate Grant, Temple University. An earlier version of the first of the two studies reported here can be found in James M. Hunt, Anindya Chatterjee and Jerome B. Kernan (1993), Perceptual and Motor Skills, 77 (August), 95-98.]

ABSTRACT -

Two studies suggest that subjects may feign a high need for cognition (NC) in a socially-desirable effort to appear "smart." In one study, the need for cognition scale (Cacioppo and Petty 1982; Cacioppo, Petty, and Kao 1984)Cboth long and short versionsCwas found to be correlated significantly with the social desirability scale (Crowne and Marlowe 1964). In a second, the scale was assessed experimentally by manipulating the motivational set under which subjects completed it. Findings revealed that subjects given normal instructions: (1) scored higher in NC than those instructed to "downplay" their intelligence in completing the scale; and (2) scored no higher or lower than subjects instructed to "demonstrate" their intelligence. In that NC is a key concept in Petty and Cacioppo's (1986) elaboration likelihood model (ELM), caution would seem appropriate when the NC scale is employed, particularly in those settings associated with cognitive achievement.

Attitude theory and research has been profoundly influenced by the elaboration likelihood model of Petty and Cacioppo (1986). This model (ELM) has both enhanced and refined traditional thinking about the attitude construct by providing valuable insights from the area of social cognition. Not only does the model prescribe a rich network of factors and responses, but it also encourages creative approaches to the dynamics of attitude processes.

Among other things, the Petty and Cacioppo model shows how individual difference variables affect a person's susceptibility to persuasive stimuli. One of these variables, need for cognition (Cacioppo and Petty 1982), is the focus of our concern. More specifically, Cacioppo, Petty, and Morris (1983) have demonstrated rather convincingly that an individual's need for cognition can influence the intake and use of information contained in persuasive messages. Their principal findingCembedded in the ELMCis that people with a high need for cognition are more likely to process message argumentation actively, whereas those who are low in this need will focus on the peripheral elements of the presentation (e.g., source attractiveness). Cacioppo and Petty (1982; Cacioppo, Petty, and Kao 1984) use a need for cognition scale (NCS), which they developed based on the work of Cohen (1957; Cohen, Stotland, and Wolf 1955). Our concern arises from the scale's possible transparency (cf. Cacioppo and Petty 1982). Accordingly, we offer two studies designed to ascertain whether the NCS produces an independent assessment of a subject's need for cognition, or whether that scale presents a confound with social desirability (Crowne and Marlowe 1964)Cin effect, subjects (particularly students) attempting to appear more studious than they are in reality. In that NCS scores are typically used as a blocking factor (both a priori and ex post) in ELM-related studies, this concern is far from being an idle one.

BACKGROUND

As initially presented by Cacioppo and Petty (1982), the concept of need for cognition (NC) is defined as an individual's "tendency to engage in and enjoy thinking" (p. 116). The authors note that this conceptualization follows from the work of Cohen (1957; Cohen, Stotland, & Wolf 1955), who viewed this need as one's need for cognitive clarity (Cohen 1964). Cohen (1957) was able to demonstrate that a high need for cognitive clarity seems to be associated with greater attention to a persuasive message as well as a greater desire or ability to organize the information contained in that message. Presumably, such processing on the part of these message recipients facilitates message retrieval, thereby influencing post-message attitude. Since Cohen left no documentation as to how he measured the need he studied, Cacioppo and Petty (1982) undertook a series of studies aimed at developing a measuring instrument. The objective of this work was to develop a need for cognition scale that would "distinguish between individuals who dispositionally tend to engage in and enjoy effortful analytic activity and those who do not" (Cacioppo et al. 1983, p. 806).

Cacioppo and Petty's original work produced an instrument containing 34 opinion statements. Response to each item was measured on a 5-point bipolar scale ranging from "extremely uncharacteristic of me" to "extremely characteristic of me." In subsequent research, this instrument was reduced to 18 items by the authors (Cacioppo, Petty, & Kao 1984). Results from the latter effort indicated the two versions to be closely correlated (r=0.95). Both versions have been shown to have a high reliability as measured by Cronbach's alpha (.91 for the longer version and .90 for the shorter). In both cases, factor analysis of the items has revealed a structure with one dominant factor, explaining from 27% (long version) to 37% (short version) of the variance.

Cacioppo and Petty have employed their scale to observe how personality can systematically influence the intake of persuasive communication and the resulting formation of attitudes. Typically, they have manipulated subjects' issue-related thoughts by exposing them to either a strong set of arguments (those that produce more positive than negative issue-related thoughts) or a weak set (those that produce more negative thoughts). The usual finding obtained by these authors (see Haugtvedt, Petty, Cacioppo, and Steidley 1988) is that greater effects occur for subjects who are high in NC than for those who are low in NC. Presumably, this is a result of greater motivation to process on the part of high NC subjects.

A second common finding of Cacioppo and Petty's workCcouched in their ELMCinvolves the manipulation of peripheral (to the central arguments of the message) persuasion cues. For instance, the authors (Haugtvedt et al. 1988) have manipulated subjects' attraction to the message sourceCeither positive or negative. The usual finding is that low NC subjects are more responsive to this manipulation than are high NC subjects when post-message attitude is used as the response variable. Since low NC subjects supposedly engage in less active message processing, they are thought to rely more on simple heuristics such as using peripheral cues to form issue-related attitudes.

One of Cacioppo and Petty's (1982) stated concerns with their NCS was that it might be confounded with social desirability, or "the disposition to respond in a socially desirable manner" (Crowne and Marlowe 1964, p.10). This concern is particularly acute for any scale purported to measure the value individuals place on thinking. For this reason, Cacioppo and Petty (1982, Study 3) tested for the presence of a relationship between need for cognition and social desirability as measured by the Crowne-Marlowe scale. Results indicated this relationship to be nonsignificant (r=.08, N=104). Although the NCS is somewhat transparent, the authors tentatively concluded that the scale is not confounded by respondents' efforts to convey favorable impressions when their responses are anonymous.

STUDY 1

In light of Cacioppo and Petty's stated concern regarding the potential transparency of the NCS, we first assessed its relationship to social desirability in an associative study carried out in an academic environment (where the appropriateness of establishing a contemplative facade should be relatively strong). Inasmuch as this study was designed to explore a broad set of issues dealing with personality, cognitive style, and generalized values, need for cognition was only one of several variables we studied. To ensure an unbiased assessment, we first estimated the association of each measure and need for cognition.

Subjects

One-hundred-fifty-six students from a large Eastern university agreed to participate in a survey for extra credit in various courses. Sixty-three of them were undergraduates, 93 were graduate students. Most (95%) were business majors. Average age was 26 years. Sixty-seven were female, 89 male.

Materials and Procedures

Subjects were asked to complete a battery of six questionnaires over two sessions. Administration of the social desirability and need for cognition scales took place during the second session, which was conducted after the close of class. In all, five groups of 12 to 41 students participated. At the beginning of each session, respondents were told that they would be participating in a study of general opinions and attitudes, that their responses would be held in confidence, and that their identity would remain unknown to the researchers. (To assure anonymity over both sessions, respondents were identified only by number). Immediately following these instructions, test booklets were distributed. The NCS was distributed first, followed by the Crowne-Marlowe (1964) social desirability scale. The NCS distributed was the longer 34-item (1982) scale, but both it and the shorter 18-item version (Cacioppo et al. 1984) were used in analyses. At the conclusion of subjects' last session, they were thanked and dismissed. Debriefing took place after all groups had been run.

Results

Our results regarding the properties of the NCS parallel those of Cacioppo and Petty discussed above. Alpha coefficients were .89 for the longer version and .88 for the shorter one. Correlation between both versions of the scale was .95 (p=.001). Factor analysis, in which a maximum likelihood extraction procedure was used, confirmed Cacioppo and Petty's (1982) finding that one factor was dominant in explaining scale variance (32.6%). Eigenvalues indicated that each of the other factors explained only a small proportion of the remaining variance. Our results also parallel Cacioppo et al. (1984) in that all but one of the original items loaded heavily (.43 to .73) on the first factor. In contrast to Cacioppo and Petty, however, we found a direct relationship between need for cognition and social desirability. Correlation coefficients were .24 (p<.01) for the longer NCS and .19 (p<.02) for the shorter version. It appears as though the transparent nature of the NCS did encourage, however slightly, some biasing of responses. In elaborating this relationship through regression analysis, we found both age and class status (undergraduate vs. graduate) interacted with NCS in a positive manner. As age--F1,150=9.45, p<.01-and class standing- F1,150=11.84,p<.01 -increased, social desirability biased NCS scores more. These predictive alliances would be expected if NCS is confounded with social desirability in the first place. Graduate students, who are generally older, are also more socialized with academic values. As a consequence, they would be more disposed to "highlight" their cognitive skills. Of course, it also could be argued that as people age, they simply become more cognitively motivated.

Discussion

Our initial evidence suggests the possibility, albeit slight, that subjects' perceived need to give socially desirable answers may bias NCS. Apparently, NCS scores are explained, at least in part, by subjects' belief that it is socially advantageous to appear deliberative or contemplative. Although tentative, these findings reinforce Cacioppo and Petty's caution regarding the potentially transparent nature of their scale. Both the validity and utility may be jeopardized, particularly when administered in contexts involving cognitive achievement. Despite anonymously responding to our questionnaire outside of class, our subjects still seem to have associated the task with their academic environment. To the extent that such associations are made, NCS scores become biased measures of individual differences in cognitive need. We recognize, of course, that our correlations are relatively small; however their significance gives pause for concern. Why our results depart from those reported by Cacioppo and Petty is not readily apparent. Perhaps our student subjects were more "tied" to their academic surroundings. Alternatively, maybe they were less focused on academic achievement since they were required to complete a number of other questionnaires.

STUDY 2

Given the possibility that the NCS is confounded by social desirability, a second study was undertaken to assess the explanation that individuals attempt to project a studious facade, particularly in settings associated with cognitive performance. Our tack was to have subjects simulate (based on the instructions they received) the test behavior we ascribed to those scoring "high" vs. "low" on the NCS. That is, subjects simulated extreme levels of social desirability. We hypothesized differences in NCS results between these groups, but no difference between the "high NC simulators" and a control group given normal instructions.

Subjects

Fifty evening students completed a questionnaire for extra course credit. Thirty-four were MBA students, 16 were undergraduate marketing students. Average age was 30. Twenty-three were female, 27 male.

Materials and Procedures

In an administered procedure, all subjects were asked to complete the long version of the NCS after the closing of class. Each was randomly assigned to one of three treatment groups, constituted by the instructions given. This assignment produced a "high NC simulator group," a "low NC simulator group," and a "usual instruction" control group. Instructions to the high simulator group read: "Please answer these questions as though you want to demonstrate your ability to act intelligently and to be contemplative and methodical." Instructions to the low simulator group were identical to those received by the high group with the exception of these words: "Please answer these questions as though you want to downplay your ability to act intelligently and to be contemplative and methodical." Instructions to the control group read: "Please answer these questions as honestly as possible." The experimenter ensured that all subjects understood their instructions. After completing the questionnaire, subjects were asked what they thought the study was about and whether they could guess the hypothesis. Debriefing occurred the following week.

Results

As in Study 1, analysis of both NCS scales indicated that a single factor dominated the explanation of variance (.47 for the long version and .55 for the short one), with each additional factor contributing little to the explanation. Alpha coefficients for both versions of the scale were .98. Two analyses of variance were run on subjects' NCS scores, one for each version of the scale. Cell sizes were 16 per instruction group. (Of the 50 original subjects, two were dropped from analysis due to incomplete questionnaires.) Our results do not support an NCS free of confound; analyses revealed a significant difference across the three experimental groups, for both the long and short versions of the NCS. For the long version, cell means were: 139 (S.D.=12.3) for the high NC simulator group, 133 (S.D.=13.2) for the control group, and 105 (S.D.=34.7) for the low NC simulator group. The statistical difference produced across these conditions (F2,45=9.91, p<.001) was due to the low NC simulator group (Sheffe's test, p=.05), which scored below the control group and the high NC simulator group. For the short NCS, cell means were: 74 (S.D.=6.9) for the high NC group, 73 (S.D.=7.6) for the controls, and 56 (S.D.=20.3) for the low NC group-F2,45=8.88, p<.001. These differences were a result of the low scores of the low NC group (p=.05). No evidence of demand characteristics was present; none of our subjects was able to guess the study's purpose or hypothesis. Further, it appears that all subjects made a conscientious effort to complete the questionnaire according to instructions. The significantly lower means in the low NC simulator group attest to an effective experimental manipulation. Indeed, the comments of several subjects from the low NC simulator group are noteworthy; these people mentioned that they had difficulty in "downplaying" their cognitive ability, which may explain the larger standard deviations for that group. If this was the case, our results are conservative ones.

Discussion

Our second study implies that people completing the NCS under what can be described as normal test instructions/conditions (our control subjects) react to that scale in the same way they would have had they assumed a socially desirable facade (our high NC simulator group). If this is so, then the transparency of the NCS poses a serious problem. We realize that drawing conclusions from the null hypothesis is always a tenuous proposition, at best. And we should make mention that our failure to find significant differences between the controls and high NC subjects might be due to a "ceiling effect"-a possibility for which we have no data. Nevertheless, when considered in conjunction with our first study, these experimental results caution against the unrestricted use of the NCS.

SUMMARY DISCUSSION

Succinctly stated, the results of our research show little support for the notion that Cacioppo and Petty's (1982) NCS produces an unbiased measure of need for cognition. According to other studies (Cacioppo and Petty 1982; Cacioppo et al. 1984), the magnitude of correlation between NCS scores and social desirability should not have reached statistical significance. In our first study, however, it did. Although the coefficients were relatively small in magnitude, they were significant. Apparently, our subjects responded to the NCS in a manner that reflected their desire to give socially propitious answers, to appear analytical or otherwise "smart." The greater this desire, the higher the NCS. Support for this interpretation also comes from our analysis involving age and level in school (undergraduate or graduate). As age and level increased, so did the relationship between social desirability and NCS. This would be expected if one assumes that older/graduate students are more apt to judge themselves on the basis of cognitive endeavor-i.e., they are more socialized with academic values. As a result, when given a questionnaire that is a transparent measure of those values, these students will likely respond in a manner that mirrors those "socially desirable" values.

The results we obtained in our second study follow nicely the results we obtained from the first one. Based on the earlier study, we had reasoned that if NCS is an unbiased measure of NC (i.e., the scale is unconfounded with respect to social desirability), then its product should be independent of manipulations that encourage subjects to adopt a "studious" facade in completing the scale. Our data did not show this to be the case, however. Subjects who completed the NCS under normal instructions and circumstances responded to the scale in the same way as those who had been instructed to "demonstrate" their cognitive ability. Notwithstanding a possible ceiling effect, this finding implies that the NCS produces scores that are confounded with subjects' desires to exhibit their intellectual prowess.

Thus, both our studies suggest that the NCS is confounded by a form of social desirability. However, this apparent confound is not an entirely unique one, nor is it necessarily a serious one. It is well known that self-report measures can be clouded by socially desirable responding (Crowne and Marlowe 1964). However, a more insidious type of problem could arise when the NCS is used in conjunction with other variables that involve cognitively oriented tasks. This can be seen in the ELM research designs employed by Cacioppo and Petty (1982; Cacioppo et al. 1983; Cacioppo et al. 1986; Haugtvedt et al. 1988; Haugtvedt and Petty 1992). A common objective of their work is to study how individual differences can predict the likelihood that recipients of persuasive communication will extract and cognitively elaborate on the information contained in that communication. To operationalize the ELM framework, subjects are sometimes partitioned a priori into either a high or low level of need for cognition based on NCS scores. This blocked factor is then crossed experimentally with an argument quality factor, which is made up of either strong or weak persuasive information. Results from this design typically demonstrate that subjects who are high in NC process the issue-relevant information more actively than do those who are low in NC. Evidence of this processing difference is normally presented in terms of attitudinal or memory differences between the two types of subjects. High NC subjects exhibit a greater cognitive and attitudinal sensitivity to the argument quality manipulation than do low NC subjects. Presumably, high NC subjects' elevated sensitivity is a reflection of their more elaborate message processing. Although we do not dispute the general conclusions drawn from Cacioppo and Petty's studies, we note the possible confound that their research design produces if an association exists between the NCS and social desirability. If NCS really captures individuals' desire merely to appear cognitively astute, as such an association would imply, then the degree to which NC affects the elaboration of persuasive information is indeterminate. It would be unclear whether (and to what extent) subjects' responses to their cognitive task were a result of a chronic NC or an acute need to present a cerebral facade. The greater the latter, the more likely persuasive information would be elaborated by subjects-but not for the hypothesized reason. Subjects assigned to the high NC group by virtue of their NCS score might exhibit a greater sensitivity to information cues simply as a result of their desire to appear contemplative, and not because of any inherent cognitive need. In short, demand characteristics present an alternative explanation to NC where social desirability has not been ruled out.

A confounded NC measure also could bring about an undesirable interaction between the pre-measured NCS and ELM subjects' subsequent exposure to cognitively oriented treatments. If subjects are required to complete a NCS prior to their exposure to such treatments, they could easily become sensitized to the cognitive demands of the persuasive task. Their desire to appear thoughtful would heighten their attention to the quality of the arguments contained in the experimental stimuli and both internal and external validity would be threatened.

Finally, we emphasize the apparent role played by age in the NCS-social desirability relationship. Our data suggest that the older the student, the greater the "temptation" of social desirability. This should be expected, inasmuch as most of our older subjects were MBA students who identify with middle- to upper-level management mentalities. Cognitive achievement is highly salient for such people, particularly when they are cast into a highly competitive academic environment. All this tells us nothing about the social approval proclivities of experimental subjects older than graduate students, however. One can speculate, but the self-presentation tendencies of, say, middle aged or elderly people are not well understood. But typical ELM experiments use college students and other young adults as subjects, hence the limited age range for which we do have data is probative.

To summarize, our results echo the expressed concerns of Cacioppo and Petty (1982) regarding the potential transparency of their NCS. A confound seems to exist between that scale and people's tendency to respond in a socially desirable way. Age of subject (over a relevant range) appears to accentuate this confound. Of course these findings are tentative and limited to academic settings in which administered questionnaires are used. Moreover, we recognize that none of these conclusions proceeds from a comprehensive replication of any of Cacioppo and Petty's individual studies (cf. Rosenthal 1991). Certainly, a better understanding of the relationship between the focal variables would emerge from such a study. Nevertheless, our findings counsel against the indiscriminate use of the NCS in any setting that is likely to emphasize cognitive skill. The cumulative contribution of the ELM paradigm has been substantial because the model is sound conceptually and because its tests have been executed with skill and care. We wish merely to emphasize what Cacioppo and Petty (1982) recognized about their NCS at the outset-that its use requires particular care.

REFERENCES

Cacioppo, John T. and Richard E. Petty (1982), "The Need for Cognition," Journal of Personality and Social Psychology, 42 (January), 116-131.

Cacioppo, John T., Richard E. Petty and Chuan F. Kao (1984), "The Efficient Assessment of Need for Cognition," Journal of Personality Assessment, 48 (June), 306-307.

Cacioppo, John T., Richard E. Petty, Chuan F. Kao and Regina Rodriguez (1986), "Central and Peripheral Routes to Persuasion: An Individual Difference Perspective," Journal of Personality and Social Psychology, 51 (November), 1032-1043.

Cacioppo, John T., Richard E. Petty and Katherine J. Morris (1983), "Effects of Need for Cognition on Message Evaluation, Recall, and Persuasion," Journal of Personality and Social Psychology, 45 (October), 805-818.

Cohen, Arthur R. (1957), "Need for Cognition and Order of Communication as Determinants of Opinion Change," In Carl I. Hovland (ed.), The Order of Presentation in Persuasion (pp. 79-97), New Haven, CT: Yale University Press.

Cohen, Arthur R. (1964), Attitude Change and Social Influence, New York: Basic Books.

Cohen, Arthur R., E. Stotland and D. Wolfe (1955), "An Experimental Investigation of Need for Cognition," Journal of Abnormal and Social Psychology, 51 (September), 291 -294.

Crowne, Douglas P. and David Marlowe (1964), The Approval Motive, New York: Wiley.

Haugtvedt, Curt, Richard E. Petty, John T. Cacioppo and Theresa Steidley (1988), "Personality and Ad Effectiveness: Exploring the Utility of Need for Cognition," In Michael Houston (ed.), Advances in Consumer Research, (Vol. 15, pp. 209-212), Provo, UT: Association for Consumer Research.

Haugtvedt, Curt and Richard E. Petty (1992), "Personality and Persuasion: Need for Cognition Moderates the Persistence and Resistance of Attitude Change," Journal of Personality and Social Psychology, 63 (August), 308-319.

Hovland, C. I. and Janis, I. L., eds. (1959), Personality and Persuasibility, New Haven, CT: Yale University Press.

Petty, Richard E. and John T. Cacioppo (1986), Communication and Persuasion: Central and Peripheral Routes to Persuasion, New York: Springer-Verlag.

Rosenthal, Robert (1991), "Replication in Behavioral Research," In J. W. Neuliep (ed.), Replication Research in the Social Sciences, (pp. 1-30), Newbury Park, CA: Sage.

----------------------------------------