Title: Need for cognition and cognitive performance from a cross-cultural perspective: Examples of academic success and solving anagrams.(Statistical Data Included)
Date: January 1, 2001 Publication: The Journal of Psychology Author: Gulgoz, Sami
The cross-cultural validity of the Need for Cognition Scale and its relationship with cognitive performance were investigated in two studies. In the first study, the relationships between the scale and university entrance scores, course grades, study skills, and social desirability were examined. Using the short form of the Turkish version of the Need for Cognition Scale (S. Gulgoz & C. J. Sadowski, 1995) no correlation with academic performance was found but there was significant correlation with a study skills scale and a social desirability scale created for this study. When regression analysis was used to predict grade point average, the Need for Cognition Scale was a significant predictor. In the second study, participants low or high in need for cognition solved multiple-solution anagrams. The instructions preceding the task set the participants' expectations regarding task difficulty. An interaction between expectation and need for cognition indicated that participants with low need for cognition perfo rmed worse when they expected difficult problems. Results of the two studies showed that need for cognition has cross-cultural validity and that its effect on cognitive performance was mediated by other variables.
Key words: academic performance, cross-cultural, need for cognition, problem-solving expectation
INDIVIDUALS WHO PROCESS INFORMATION more actively perform better in most of the cognitive tasks that are given to them. Whether they will process information more actively may be determined by the nature of the task, their interest or knowledge related to that task, or the situational demands. Processing information actively can also be construed as an individual tendency. Cacioppo and Petty used the term need for cognition to denote the individual differences in the tendency to seek and enjoy effortful thinking. They also developed the Need for Cognition Scale to measure this construct (Cacioppo & Petty, 1982). We can assume that people who have a high need for cognition would also tend to be more active processors. The Need for Cognition Scale and its short form demonstrated high internal consistency (Cacioppo, Petty, & Kao, 1984; Sadowski & Gulgoz, 1992b), and the test-retest reliability of the short form is also high (Sadowski & Gulgoz, 1992b).
The research by Cacioppo, Petty, and their colleagues has shown that the effectiveness of persuasive messages varies according to how likely the individual is to engage in processing the information presented in the message. For example, Cacioppo, Petty, Kao, and Rodriguez (1986) and Priester and Petty (1995) found evidence to support the argument that people with a higher need for cognition have a higher tendency to elaborate on the persuasive messages presented to them. Over the years there has been an accumulation of research results indicating that the need for cognition is related to performance in a number of tasks that are more directly relevant to cognitive processing (see Cacioppo, Petty, Feinstein, & Jarvis, 1996, for a review). These studies showed significant relationships between need for cognition and various cognitive performance measures with some of them directly related to learning. One example is the correlation between the scores on the Need for Cognition Scale and the grades that student s get in college-level courses (e.g., Dornic, Ekehammar, & Laaksonen, 1991; Gulgoz, 1996; Sadowski & Gulgoz, 1992a; Srull, Lichtenstein, & Rothbart, 1985).
There have also been discussions on the specificity of personality constructs for the cultures that produced them. In a provocative article, Markus and Kitayama (1991) emphasized the ways cultures determine how their members construe themselves, and they indicated that the way individuals construe themselves would have a critical influence on cognition and motivation, among other things. Recent research has demonstrated both the presence and absence of consistencies across cultures in basic personality patterns or constructs (Di Bias & Forzi, 1999; Eysenck, Barrett, & Barnes, 1993; McCrae & Costa, 1997; Okeke, Draguns, Sheku, & Allen, 1999; Pervin, 1999).
One question that I wanted to confront in the studies reported here was whether the Turkish version of the Need for Cognition Scale (Gulgoz & Sadowski, 1995) yielded results similar to those of the original scale in terms of psychometric qualities as well as relationships with other measures. My major interest was its relationship with measures of cognitive performance, because that is a particular area that would be directly influenced by a tendency to process information extensively.
The studies reported here build on the previous study by Gulgoz and Sadowski (1995) who translated the original 18-item Short Need for Cognition Scale (Cacioppo, Petty, & Kao, 1984) into Turkish and tested it with 156 participants. The final participant sample consisted of 142 undergraduate students who gave their consent for the investigators to access their records of academic performance. They participated in two sessions of 71 students each, allowing a measure of test-retest reliability. The test-retest reliability coefficient was significant but lower than the one obtained by Sadowski and Gulgoz (1992b). Cronbach's alpha coefficients were .69 for the first stage of the study and .78 for the second stage. The factor structure of the scale revealed one factor on which 16 items of the 18-item scale loaded; each of the remaining two items (Items 7 and 16) loaded on separate factors. Item-total correlations were significant (p < .005) for all items except Items 7 and 16.
The correlations between the Need for Cognition Scale scores and the scores on the four subtests of the university entrance examination (science, math, Turkish, and social sciences) were between -.15 and .10, and none were significant. The correlations between need for cognition and course grades produced coefficients ranging between .07 and .35, and none were significant.
These results demonstrated that the Turkish version of the Need for Cognition Scale had acceptable levels of test-retest reliability and internal consistency but that there was no correlation with measures of academic performance. In the studies reported here, I gave the Turkish version of the short form of the Need for Cognition Scale to new groups of Turkish participants to test further relationships that may have led to the results in the Gulgoz and Sadowski (1995) study. I focused particularly on the possibility that the relationship between need for cognition and performance on cognitive tasks may be mediated by other factors.
Study 1
This study was very similar to that of Gulgoz and Sadowski (1995) with the exception that two other scales, created for this study, were given to the participants together with the Turkish version of the Need for Cognition Scale. These new scales included a social desirability scale and a study skills scale. To observe whether similar findings would be observed, the participants in this study were individuals who had not participated in previous aforementioned studies.
In earlier studies, it is likely that the effects of need for cognition were overshadowed by other variables that influenced course performance. For this reason, in this study, in order to predict course performance, I used the new study skills scale as well as measures such as participants' scores on their university entrance examinations. It is also possible that what Gulgoz and Sadowski (1995) were measuring was socially desirable responding rather than a true expression of the participants' self-perception. To investigate the degree that social desirability could account for scores on the Need for Cognition Scale, we asked the participants to respond to the social desirability scale as well.
Method
Participants. The participants were 97 students at Koc University who were given course credit. All subjects participated in one data collection session in which they responded to all three scales.
Materials. The Turkish version of the short form of the Need for Cognition Scale was used alongside two additional scales. The new social desirability scale included a number of items similar to the items on the Marlowe-Crowne Social Desirability Scale (Marlowe, 1962), but some of the items in that scale did not lend themselves to the Turkish culture. For that reason I created a culturally appropriate version. In addition, a newly created scale intended to measure the study skills of the students was included in the study. Both scales demonstrated high internal consistency and reliability (Gulgoz, 1996).
Procedure. The participants were first given a consent form requesting permission to obtain their records, which included course grades and university entrance examination scores. The three scales were then administered at the end of a class meeting after which the participants were given the booklet containing the three scales. We used student numbers to match the booklets and the records containing the student scores and then discarded the student numbers from the data set to preserve anonymity.
Results
The mean score on the Need for Cognition Scale was 25.08 (SD = 20.33). The mean was not significantly different from the mean obtained by Gulgoz and Sadowski (1995), t(226) = 0.85, p > .05, [alpha] = .88; item-total correlations ranged between .415 and .812 (all ps < .001) except for Item 16, which had a lower but significant correlation, r(89) = .24, p <.05. Factor analysis procedure using the iterative principal axis factor analysis showed that all items except 16 and 17 loaded on one factor and that Items 16 and 17 each loaded on separate factors. There was no gender difference in the Need for Cognition Scale scores, t(89) = .39, p >.05.
Correlations were calculated between the subtests of the university entrance examination scores and the need for cognition scores, but none of these correlations were significant. Grades from courses were also used in correlations. There were 16 courses, in which at least 8 of the participants had been enrolled. The need for cognition scores were not significantly correlated with the grades in any of the 16 courses or with the average of all course grades.
When men and women were treated separately, the correlations with course performance were nonsignificant with one exception. That exception was the correlation between the Need for Cognition Scale scores and the grade on the computer skills course for men, r(34) = .44, p < .01. I performed another group of analyses to investigate the correlations among the Need for Cognition Scale scores and scores on the new social desirability scale and the study skills scale. Both the social desirability scale scores, r(79) = .39, p <.001, and the study skills survey scores, r(89) = .40, p >< .001, were significantly and positively correlated with the Need for Cognition Scale.
In a final analysis, a stepwise multiple regression analysis, the predicted variable was the average course grade. The predictor variables initially entered were scores on the general aptitude test composed of verbal and quantitative abilities (used as the initial screening test for admission to universities), scores on Turkish, mathematics, and social sciences subtests of the university placement examination, the study skills scale, and the Need for Cognition Scale. The model resulting from this analysis included four significant variables: a Turkish subtest of the placement test, a general aptitude test, the study skills survey, and the Need for Cognition Scale scores, accounting for an adjusted squared multiple R of .60, F(4, 72) = 28.95, p < .001, MSres = 0.132.
Discussion
The results provided a higher internal consistency level for the Need for Cognition Scale than was obtained by Gulgoz and Sadowski (1995). The coefficient obtained in the present study was very close to values obtained in the U.S. study by Sadowski and Gulgoz (1992b) and in the same range of coefficients reported in Cacioppo et al. (1996). The results of the factor analysis revealed a one-factor solution. Two items loaded on two separate factors and one of these items (16) was also incongruent with the rest of the items in Gulgoz and Sadowski (1995). Although it may have been more feasible to remove the item from the Turkish version of the scale, removing it did not alter the alpha coefficient or the relationship of the scale scores with other variables in any significant way. Therefore, Item 16 was kept in the scale.
The results, combined with the results of the study by Gulgoz and Sadowski (1995), support the reliability and the internal consistency of the scale when used with the Turkish sample and in Turkish. However, the lack of significant relationships with course performance and university entrance examination scores point to some difficulty in validating the Turkish version of the scale. Other studies have also used course performance as a means of measuring the relationship between the need for cognition and cognitive performance. Leone and Dalton (1988) found an overall nonsignificant relationship but discovered that individuals with a high need for cognition perform better on materials demanding cognitive effort. They found no difference between individuals high or low in need for cognition on tasks requiring little cognitive effort.
In the Sadowski and Gulgoz study (1992a), performance in a portion of the course correlated significantly with the need for cognition scores although the correlation was rather low, r(47) = .28, p < .05. In a previous discussion of the relationship between measures of cognitive performance and the Need for Cognition Scale (Gulgoz & Sadowski, 1995), it had been argued that the correlations with cognitive performance could not be expected to be very high. This is simply because course performance is also affected by many other factors, including a variety of cognitive skills and strategies, situational variables, and motivational factors other than the need for cognition.
The results of the stepwise regression analysis on average course grade confirmed this argument. The Need for Cognition Scale was retained as part of the regression equation in addition to the study skills scale, the general aptitude test, and the Turkish subtest of the university placement examination. The same is true for the study skills survey; it is included in the regression equation even though its correlation with course performance was not significant.
Another important result is the correlation with the tendency for socially desirable responding. Cacioppo et al. (1996) reported five studies in which social desirability was used in correlations with the need for cognition. In two of these studies (Cacioppo et al., 1984; Olson, Camp, & Fuller, 1984), the correlations were considered significant but they were very low, r(96) = .21, p < .05, and r(139) = .16, p < .10, respectively, accounting for no more than about 4% of the variance. Our finding in the present study is considerably higher, with a shared variance of about 15%.
Sadowski and Cogburn (1997) have shown that need for cognition scores increase as respondents become more conscientious. Conscientiousness, in turn, has been shown to be associated with social desirability in a meta-analysis performed on social desirability literature by Ones, Viswesvaran, and Reiss (1996). Their participants' characteristics may have been one reason for a significant and moderate correlation coefficient, as their particular sample of students was attending a university with high educational aspirations and standards. This may have created social pressure on students to value thinking and cognitive activity. Also, the high level of desirability for education in the Turkish culture (Kagitcibasi, Bekman, & Sunar, 1993) may have reflected on the correlation between the need for cognition and social desirability. These can be only tentative explanations, because there is not a direct comparison with control groups to support higher levels of social desirability for cognitive functions in the part icipants of this study.
Study 2
In the second study my purpose was to investigate the relationship between the need for cognition and cognitive performance. Participants were asked to solve problems of a novel kind. I introduced multiple-solution anagrams because they are the types of problems that benefit more from a tendency to enjoy and engage in thinking. Multiple-solution anagrams can yield more than a single solution as long as the player works on it. There is no external indicator other than the player him- or herself that signals the completion of the problem.
I hypothesized that individuals with higher need for cognition would expend more effort on these problems and would find a larger number of solutions than individuals with lower need for cognition. I also manipulated the expectation of the participants by giving advance information about the difficulty of the problems. The exact impact of manipulating expectations on performance and response certitude has been observed to vary from no impact to impact on one but not the other (Feather & Saville, 1967; Stock, Winston, Behrens, & Harper-Marinick, 1989). The effect of such manipulation on motor performance has shown that setting expectations has an effect on performance only when the task is below a certain level of difficulty (Eyal, Bar-Eli, Tenenbaum, & Pie, 1995).
In the present study we manipulated performance expectations by telling half the participants that their peers generally solved the problems in the experiment easily. The other half of the participants were told that very few people have been able to solve these problems. We expected that the expectations created in the participants would affect the performance but that the participants with high need for cognition would be less vulnerable to a decrease in performance when they were told that these problems were too difficult for them.
Method
Participants. The participants were 57 undergraduate students (33 women, 24 men) at Koc University who had not participated in the first study. They were all native speakers of Turkish and participated for extra credit.
Materials. The students were given a booklet containing the Turkish version of the Need for Cognition Scale followed by an instruction sheet that included information on the problems and instructions about how to indicate the time they began working on the problems and the time they finished. The instruction sheet also contained information about the difficulty of the problems in order to manipulate performance expectations. In half the booklets it was indicated that most university students can solve the problems easily and in the other half the instructions read that most university students have difficulty solving these problems.
The last page of the booklet contained the multiple-solution anagram problems in Turkish that were developed specifically to be used in the present study. The page began with an example of a four-letter multiple-solution anagram with all of its solutions and 18 five-letter anagrams with the number of solutions ranging from one to nine words. The clock on the Microsoft Windows 3.1 was projected onto a large screen in a digital fashion with the hour, minute, and seconds showing for the participants to indicate the time they began and stopped working on the anagrams.
Procedure. The participants were tested in one large group seated apart from each other. They were told that they were participating in a study to establish norms on a set of problems. They were also told that these problems had been tested on other university students earlier but that a larger and more representative sample was needed. They were told to read the instructions very carefully, respond correctly to all the scale items and solve the problems to the best of their ability. The verbal instructions did not contain any information about the difficulty of the items. The participants were told that they could work on the problems as long as they wished.
Results
The data from 7 of the participants were considered unusable either because they missed the second page of items in the Need for Cognition Scale or because they spent less than 150 s on the anagram problems. The analyses were performed on the data from the remaining 50 participants.
The mean need for cognition score was 26.16 (SD = 14.50). The mean for anagrams solved was 29.86 out of a possible 66 words (SD = 6.20). The mean amount of time spent in seconds was 794.82 (SD = 305.02). The correlations among these three variables indicated that only the correlation between the total number of anagram solutions and the time spent on solving the problems was significant, r(49) = 0.51, p < .001.
To observe the effects of expectation and need for cognition on performance on the anagram problems and their interaction, I conducted an analysis of variance (ANOVA). For this purpose the participants were separated into two groups, high and low need for cognition, by applying a median split. The ANOVA on the number of anagram solutions revealed a marginal main effect of difficulty expectation, F(1, 46) = 3.95, p < .06, MSE = 28.88, and an interaction between need for cognition and difficulty expectation, F(1, 46) = 15.78, p < .001. The means of this interaction are shown in Table 1.
The Tukey test (p < .05) conducted to further understand the interaction showed two significant differences. One was the difference between the low need for cognition participants who were told that the task was easy and those who were told that it was difficult. The students who had a low need for cognition performed as well as high need for cognition students when they were told that the task was easy. However, their performance dropped significantly when they believed the task to be too difficult for them. The other significant difference in anagram solutions was between high and low need for cognition groups when they were told that the task was difficult.
I conducted a similar ANOVA on time spent solving the anagrams to investigate whether this interaction could be explained to a certain extent by the amount of time the participants spent in trying to solve the problems. This analysis did not reveal any significant effects or interaction but the main effect of difficulty expectation was marginal, F(1, 46) = 3.35, p < .08, MSE = 91107.22, such that the participants who expected easy problems spent slightly more time (about 140 s more) than those who expected difficult problems.
Discussion
Unlike in earlier studies, I used a more direct measure of performance in the form of multiple-solution anagrams. The results of the second study using the multiple-solution anagrams renewed earlier findings indicating no correlation between the Need for Cognition Scale and measures of cognitive performance. However, the manipulation of expectations about the difficulty of the task revealed a more complicated relationship. When the problems were presented as a difficult task, the relationship between need for cognition and cognitive performance emerged. Participants with a high need for cognition performed equally well when they expected the problems would be difficult and when they expected the problems would be easy. Participants with low need for cognition, on the other hand, produced fewer anagram solutions when they expected the problems would be difficult. Both the high need for cognition group and the low need for cognition group performed equally well when the problems were presented as easy. The dif ference between the low need for cognition group and the high need for cognition group became apparent when the problem was presented as a difficult task.
This relationship did not exist in the amount of time that participants of each group or expectation spent on solving anagrams. Therefore, the result cannot be attributed to the amount of time per se; rather, we must assume that there must be a different process taking place during the time spent on the problem for participants high in need for cognition.
Overall Discussion
The purpose of the two studies presented here was to Lest the appropriateness of the Need for Cognition Scale in another culture (and language) and to provide evidence that would support the use of the construct in other cultures. The results of these studies show that the Turkish version of the Need for Cognition Scale is a consistent and reliable instrument. A difficulty arose when the correlations of the Need for Cognition Scale with academic performance measures that were ordinarily obtained in the U.S. samples were not obtained with the Turkish samples. Following that outcome in Gulgoz and Sadowski (1995), I performed the studies presented here with the expectation of reproducing the same results with new samples.
I observed once again that there were no correlations between need for cognition and academic performance measures such as course grades and university entrance examination scores. However, an interesting finding emerged in Study 1, when the need for cognition was entered into a regression equation with other predictor variables in an attempt to predict academic performance. When the verbal score and the general aptitude score for the university entrance examination, and the score on the study skills scale were used together with the need for cognition, the need for cognition remained a significant predictor variable. Thus, it appears that there is a relationship between need for cognition and academic performance when it is combined with other variables.
A lack of direct correspondence between need for cognition and other cognitive performance measures was observed in Study 2 as well. There was no correlation between need for cognition and performance on multiple-solution anagrams. However, the experimental manipulation in this study included setting different expectations for different groups of students. Those students who expected to have difficult problems were able to produce fewer solutions than those who were told that the problems would be easy.
Participants' need for cognition did not affect the two groups in the same way. The difference between the participants with low need for cognition and those with high need for cognition became apparent when the participants were told that the anagram problems would be difficult for them to solve. The participants with high need for cognition were not discouraged by their expectation of difficult problems, and they performed equally well on problems described as difficult or easy. However, the participants with low need for cognition produced fewer results when they were told that the problems would be difficult. The low need for cognition group was the only group affected by the experimental manipulation. The gender of the participants was tested as a possible explanation, although that was not included in the design of the study, but it did not show any significant effect or interaction.
I might have argued that participants who were set up to expect more difficult problems or participants with a lower need for cognition may have spent less time working on the anagrams simply because they were discouraged or they had a tendency to avoid such problems. For example, Gulgoz, Aktunc, Kumkale, and Eskenazi (1999) observed that readers with a low need for cognition read texts much faster than those with a high need for cognition. However, that interpretation is not completely supported in the studies reported here. Although there was a correlation between the time spent on the problems and the number of solutions produced, the ANOVA that examined the effects of need for cognition and expectation on the time spent on the problems did not show any effect of either variable. Therefore, the cause of the interaction cannot be attributed to time spent on the problem, and the real issue seems to be how the participants used the time rather than how much time they spent.
There is an inherent problem when one measures the time spent on a particular task. In this study, time was measured through self-report of beginning time and finishing time for each anagram. Although the participants were clearly instructed to refrain from going back to a problem once they thought they had produced all the solutions that they could, this was difficult to control. In addition, they reported the times but there was no control over what sort of activity took place between reported times. For these reasons, I consider the measure of time an approximation rather than a precise measurement.
In the studies reported here, I observed need for cognition to be related to other variables that may interact with need for cognition or suppress its effects. The variables included in the regression analysis and expectations about problem difficulty could be considered in that group. In addition, I observed that for the specific sample I had selected, the correlation between need for cognition and social desirability was higher than obtained in other studies. This may be due to the characteristics of the sample or of the culture in which the sample was embedded. However, the fact that the correlation was high does not immediately imply socially desirable responding to the need for cognition items. An alternative interpretation would be that the participants who tended to seek social approval also tended to be high in need for cognition.
When we consider other findings in this article and findings by other researchers that need for cognition correlates with conscientiousness (Sadowski & Cogburn, 1997), knowledge, and information learned from text (Gulgoz et al., 1999), the latter interpretation seems to be the more favorable one. The effect of need for cognition on learning from text is particularly relevant to the findings here because evidence shows that readers with high need for cognition learn better than others when they lack prior knowledge of the text they are reading or the text is too difficult. Otherwise, their performance is similar to that of readers with low need for cognition.
My conclusion, based on the given evidence, is that there is support for the cross-cultural validity of the Need for Cognition Scale as well as for the construct itself. The relationship between need for cognition and cognitive performance is not direct. Rather, need for cognition asserts its effect under circumstances that make the task particularly difficult. TABLE 1 The Means (and Standard Deviations) for the Number of Anagram Solutions According to Difficulty Expectation and Need for Cognition Need for cognition Expected difficulty LowHigh Easy M33.15 28.75 SD2.345.67 Difficult M24.00 31.80 SD5.946.49

0 commenti: