COGNITIVE DIFFERENTIATION AND SOCIO POLI

COGNITIVE DIFFERENTIATION A N D SOCIO-POLITICO IDEOLOGY: A N EXPLORATORY STUDY l JIM SIDANIUS AND BO EKEHAMMAR University of Stockholm Summary.-Cognitive differentiation was studied in

COGNITIVE DIFFERENTIATION A N D SOCIO-POLITICO

IDEOLOGY: A N EXPLORATORY STUDY l

JIM SIDANIUS AND BO EKEHAMMAR

University of Stockholm Summary.-Cognitive differentiation was studied in relationship to socio- politico ideology using 105 Swedish university students Three indices of cog- nitive differentiation were defined by use of similarity estimates among nine Swedish political parties and Ekman's multidimensional scaling method Socio- politico ideology was operationalized by asking the subjects to describe themselves

as either Radicals, Liberals, or Conservatives Cross-classification and likelihood ratio analysis disclosed statistically significant relationships between two of the three indices of cognitive differentiation and socio-politico ideology The re- sults were discussed in terms of three theories and/or hypotheses, authoritarian personality theory, the extremism hypothesis and the context hypothesis The significant results were congruent with the context hypothesis

Investigation of the interrelationship between cognitive functioning and the nature of one's socio-politico beliefs has been occupying psychological re- searchers for some time now Frenkel-Brunswik ( 1949, 1954) was perhaps the first to explicitly postulate a connection between ethnocentrism and cognitive style, e.g., premature closure, intolerance of ambiguity, rigidity, etc This con- nection between ethnocentrism and cognitive functioning was seen as part of a general personality syndrome, the authoritarian personality, of which general political conservatism was considered merely one manifestation (see Adorno, Frenkel-Brunswik, Levinson, & Sanford, 1950) A good deal of empirical sup- port, subsequent to the publication of The Authoritariart Personality (Adorno,

e.g., racism or ethnocentrism, and cognitive variables Most of these empirical results have been congruent with authoritarian personality theory

Much of the research within the authoritarian personality tradition has been criticized on the basis that authoritarianism, usually measured by the F-Scale, is only sensitive to authoritarianism of the right and does not reflect authoritarian- ism of the left This kind of criticism has usually been launched by researchers adhering to what Sidanius ( 1976b) has referred to as the extremism hypothesis The extremism hypothesis postulates that whereas "extreme" leftists and "ex- treme" rightists can both be characterized as possessing simplistic, rigid and dog- matic cognitive styles, "normals," or persons of the ideological middle range, can

be characterized as possessing relatively complex, flexible and open-minded cog- nitive styles (see Shils, 1954; Taylor, 1960; Eysenck & Coulter, 1972) By 'This study was supported by a research grant to the authors from the Swedish Council for Social Science Research for a project on Political perception Requests for reprints should be addressed to Jim Sidanius, Department of Psychology, University of Stockholm, Box 6706, S113 85 Stockholm, Sweden

"extremists" most workers within this tradition have meant either "fascists,"

"communists," or "extreme liberals."

A third hypothesis concerning the relationship between cognitive function- ing and socio-politico ideology has recently been designated as the context hy- pothesis (Sidanius, 1976b, 1977) The context hypothesis also postulates a con- nection between cognitive functioning and socio-politico ideology However, whereas the extremism hypothesis expects less cognitive complexity and flexibil- ity with increasing socio-politico extremism, the concext hypothesis expects just the opposite The context hypothesis assumes greater cognitive complexiry, flexibility, etc., with increasing socio-politico extremism It should be noted, however, that the context hypothesis regards the term "extremism" as having rela- tive and not absolute meaning This means that certain political ideas could be either considered as extremist or moderate depending upon the given CUltural and historical context

Multidimensional scaling analysis has recently come into use in the effort

to delineate differences in the number and/or nature of the latent dimensions underlying the cognition of political stimuli between individuals with differing socio-politico ideologies By use of such scaling, individuals' experienced simi- larity or dissimilarity among an array of stimuli forms the basis for making infer- ences concerning their underlying cognitive functioning The results of some multidimensional scaling studies have indicated differences in cognitive function- ing between persons with differing socio-politico ideologies (Abelson, 1954; Sherman & Ross, 1972) while others have not (Messick, 1956, 1961; Stone & Coles, 1972; Yellig & Wearing, 1974) As specified by Sidanius and Eke- hammar ( 1976), there are at least nine reasons which might contribute to these conflicting results from such studies

The present study was restricted to an examination of the relationship be- tween one aspect of cognitive functioning, namely, cognitive differentiation, and socio-politico ideology By cognitive functioning we mean the manner in which individuals structure and process information By cognitive differentiation most psychologists usually mean the number of dimensions or categories of meaning

an individual uses in representing his physical and social universe (see Bieri, Atkins, Briar, Learnan, Miller, & Tripodi, 1966; Schroder, Driver, & Streufert, 1967)

Against this background, this study attempted to answer the following two questions: ( 1 ) Is there any relationship between cognitive differentiation and socio-politico ideology? ( 2 ) Can any support be found for authoritarian person- ality theory, the extremism and the concext hypotheses?

The subjects were 105 undergraduate psychology students at the University

of Stockholm who participated in the study as part of their course requirements There were 34 males and 71 females

Procedzre

Each subject was given a questionnaire containing, among other things, the S4 Conservatism Scale (see Sidanius, 1976a, 1976c), a scale developed for a Scandinavian context and of the Wilson-Patterson C-Scale item format type The subjects were also asked to rate the similarity among Swedish political parties

Socio-politico Ideology

Socio-politico ideology was operationalized by asking the subjects to describe rhemselves as ( 1 ) Radical, ( 2 ) Liberal or ( 3 ) Conservative Of the 105 sub- jects, 59 described themselves as radicals, 37 as liberals and 9 as conservatives

T o test whether these three groups really had significantly different socio-politico viewpoints, the mean S4 Conservatism scores were computed for each group and submitted to a one-way analysis of variance The results showed significant and very strong ideological differences between the groups (F2/102 = 74.49, fl <

eta = .77) Furthermore Schefft pair-wise post hoc comparisons showed that all of the pair-wise comparisons among the three groups were significant

( p < .Ol)

Stimuli-political Parties

The political stimuli used in the similarity rating task were 9 Swedish political parries or organizations (Sidanius & Ekehammar, 1976) The parties were chosen so as to cover a wide range of political viewpoints

Similarity Ratings

The subjects were instructed to rate the perceived over-all similarity among all pairs ( 3 6 ) of political parties on a numerical scale ranging from 0 ("no per- ceived similarity at all") to 10 ("perceived as identical") Three different presentation orders of stimulus pairs were used and randomly distributed to sub- jects The similarity ratings were linearly transformed to a scale from 0 to 1 All non-complete matrices were dropped from further analyses This left a total

of 97 subjects (55 Radicals, 33 Liberals, 9 Conservatives) with completely usable matrices

Multtdzmen~ional Scaling Method

The similarity matrix for each subject was submitted to multidimensional scaling analysis according to the Ekrnan model ( 1965, 1970) The simplest variant (Ekehammar, 1972) within this model was used, by which the similarity ratings are interpreted as cosines between stimulus vectors I n principle, this makes possible the use of conventional factor analytic techniques in the treatment

of the similarity matrices It should be pointed out, however, that similarity matrices, especially individual matrices, can often be non-gramian When non-

gramian matrices are factor analysed certain anomalies may occur, but these are often of little practical importance (Gorsuch, 1974, p 146) In the present case and in accordance with the Ekrnan method, principal components analysis was used Each solution was rotated to simple structure via the Varimax criterion

Indices of Cognitive Differentiation

As mentioned before, Cognitive differentiation refers to the number of dimensions assumed to underlie a person's cognition of stimuli Based upon the results of the individual multidimensional scaling and similarity analyses, Cogni- tive differentiation was operationalized by the use of three indices In ( 1 ) Mani- fest differentiation the raw similarity ratings constituted the basis for this index and was mathematically defined as the arithmetic mean of the 36 similarity rar- ings (after linear transformation) among the 9 political parties This index was conceived on the assumption that the lower the mean perceived similarity among the stimuli, the greater the cognitive differentiation The index, ( 2 ) Latent dif- ferentiation-Eigen, was based directly upon the multidimensional scaling analy- ses and was defined as the number of principal components in each individual's similarity matrix with eigenvalues > 1.00 It was assumed that the larger this number, the greater the cognitive differentiation The ( 3 ) Latent differen- tiation-Scree was also based upon the analyses and was defined as the num- ber of significant dimensions in each individual's similarity matrix as deter- mined by the scree test (Cattell, 1966) The scree test has been found to be a simple and valid test for the number of significant dimensions (Gorsuch, 1974)

Statistical Anuiy~er

In order to ascertain whether any relationship existed becween cognitive differentiation and generalized socio-politico ideology, likelihood ratio esti- mates to the chi-square distribution were computed (Hays, 1970, pp 597-598;

Mood & Graybill, 1963, pp 311-319) The likelihood ratio test was chosen instead of the usual Pearson chi-square because it is less affected by small sample size than is the Pearson chi-square (Mood & Graybill, 1963) T o get a simpler over-all view of the results, disregarding the form of the score distributions within each socio-politico category, the average cognitive differentiation scores, expressed as arithmetic means and medians, were computed for each of the three socio-poli tico groups

RESULTS

The individual analyses disclosed that the number of dimensions lying behind the subjects' similarity ratings of Swedish political parties varied from two to five The results also showed that although the two indices of Latent differentiation were highly related, their absolute agreement was only GO%, i.e., the two indices gave 60% of the subjects the same absolute number of latent dimensions To ascertain the degree of relationship becween Manifest

differentiation, on the one hand, and Latent differentiation-Eigen and Latent differentiation-Scree on the other hand, two one-way analyses of variance were performed using Manifest differentiation as the dependent variable and each of

the other two indices as independent variables The results showed significant

and linear relationships in both cases (Eigen vs Manifest differentiation: F3/93

= 31.36, p 5 10-11, eta = .71; Scree vs Manifest differentiation: F 3 / 0 3 = 4.10,

p 01, eta = .34)

In order to study the association between political ideology and cognitive differentiation, three 3 X 3 contingency tables were constructed, one for each index of cognitive differentiation Three categories of cognitive differentiation (high, medium, low) were cross-classified against three categories of political ide- ology (radical, liberal, conservative) For Latent differentiation (Eigen and Scree), high differentiation was defined as 4 or 5 dimensions, medium as 3 dimensions, and low as 2 dimensions (see Table 1 ) For Manifest differentiation, high dif- ferentiation was defined as the range 0-0.28, medi'um differentiation as the range 0.29-0.37, and low differentiation as the range of 0.38-0.67 This tri- chotomization divided the subjects into approximately 3 equally large groups (see Table 1 )

PERCENTAGE OF EACH SOCIO-POLITICO GROUP DISPLAYING DIFFERENT DEGREES OF COGNITNE DlFPERBNTlATlON FOR EACH OF THRBE INDICES Degree of Differentiation n Socio-politico Ideology

Radical Liberal Conservative

Low

Medium

High

Total

Low

Medium

High

Total

Low

Medium

High

Total

Latent Differentiation-Eigen

100

Latent Differentiation-Scree

45 5 6

15 l3 100

Manifest Differentiation

3 3 20

3 5 40

100

These results of the three likelihood ratio tests showed statistically signifi- cant ( p L 06) relationships between socio-politico ideology and cognitive dif-

ferentiation for two of the three indices of cognitive differentiation ((Ltent dif- ferentiation-Eigen: h = 10.95, df = 4, p = .02; Manifest differentiation:

A = 9.37, df = 4, p = .05)

The relationships between political self-description and Latent differentia- tion-Eigen indicated a much smaller proportion of the radicals (24%) than the liberals and conservatives ( 5 2 % and 44%, respectively) displayed low cog- nitive differentiation Although the differences with respect to the Scree cri- terion were not statistically significant, they were generally in agreement with the differences with respect to the Eigen criterion (see Table 1) Examination

of the nature of the relationship between Manifest differentiation and political self-description disclosed that a much larger percentage of the radicals than the liberals and conservatives displayed high cognitive differentiation (40% vs 27% and 22 % respectively) Furthermore, a much larger percentage of liberals, than radicals and conservatives, showed low cognitive differentiation (48% vs 20% and 22 % respectively, see Table 1 )

According to authoritarian personality theory, we should expect to find a monotonically increasing level of cognitive differentiation from "conservative"

to "radical." According to the extremism hypothesis, we should observe a curvi- linear trend in the relationship between socio-politico ideology and cognitive dif- ferentiation wherein liberals show the highest level of cognitive differentiation while radicals and conservatives show the lowest level of cognitive differentia- tion According to the context hypothesis, we should expect to observe a curvi- linear trend in the relationship between socio-politico ideology and cognitive dif- ferentiation, however, in this case radicals, i.e., leftists, and conservatives should show the greatest level of cognitive differentiation while liberals should show the lowest level of cognitive differentiation

AVERAGE COGNITIVE DIPPEREmnON INDICES, EXPRESSED AS MEDIANS (Mdn) AND Arm-m.fsnc MEANS ( M ) , AS FUNCTIONS OF SOCIO-POLITICO IDEOLOGY

Index of Cognitive Differentiation Soci~politico Ideology

Radical Liberal Conservative

The average cognitive differentiation scores for each socio-politico category (see Table 2 ) are interesting in terms of the light they can shed upon the rela- tive plausibilities of the above theories/hypotheses regarding the relationship be- tween cognitive differentiation and socio-politico ideology

As we will recall, the likelihood ratio analyses have tended to show statisti- cally different distribution forms between radicals, liberals and conservatives over the different levels of cognitive differentiation Examination of Table 1

shows that the distributions for the different political categories are skewed and

sometimes skewed in different directions This fact implies that the median rather than the arithmetic mean is the proper measure of central tendency (Hays, 1970, p 175) However, for the purpose of comparison, both measures have been computed and may be found in Table 2 Examination of this table indicates that when using the median (the proper measure of central tendency with these data), the trends for Manifest differentiation and Latent differentia- tion-Eigen are consistent with the context hypothesis rather than the extremism hypothesis or authoritarian personality theory Furthermore this curvilinear trend was found with just those two indices which showed significant distribu- tion differences between the political groups

DISCUSSION

In this paper we have examined the relationship between socio-politico ideology and cognitive differentiation We have defined cognitive differentia- tion as the number of dimensions inherent in subjects' cognitive space This concept is restricted to the structure of cognitive space rather than the contents

of cognitive space Furthermore we have restricted the operationalization of the concept to the domain of political judgment, basing it solely upon individuals' judgments of the similarity among political parties

The analyses disclosed statistically significant relationships between socio-

politico ideology and two of the three indices of cognitive differentiation Radi- cals, liberals, and conservatives showed significant differences in their distribu- tions over the different levels of cognrtive differentiation The general nature

of these distribution differences disclosed that a much larger proportion of liberals and conservatives than radicals tended to exhibit a low level of cognitive differentiation (Latent differentiation) and that a much larger proportion of radicals than liberals and conservatives tended to exhibit a high level of cognitive differentiation (Manifest differentiation) Examination of over-all median trends tended to be consistent with the context hypothesis rather than the ex- tremism hypothesis or authoritarian personality theory

However, caution must be used in generalizing these results, principally because the subjects were Swedish university students and the definition of cog- nitive differentiation was restricted to one domain of political judgment Nat- urally before any general conclusions can be reached, cross-validation from dif- ferent populations and from different content domains must be performed The results of Warr, Schroder, and Blackman (1969) are pertinent in this regard Among other things, Warr, et al investigated the relationship between cognitive differentiation via a multidimensional scaling technique, and socio- politico ideology among 86 British university srudents The subjects were di- vided into three political groups: ( 1 ) left-wing, ( 2 ) centre and ( 3 ) right- wing The results showed no significant differences in the mean, i.e., arithmetic, number of cognitive dimensions among the groups Furthermore, the non-sig-

nificant trend that was found tended to be consistent with the extremism hypothe- sis rather than the context hypothesis or authoritarian personality theory Their results are in definite conflict with the results reported here

However, there are a number of differences between these two scudies, any one of which might accounc for the result differences: (1) Warr, et al used English subjects while we used Swedish subjects, ( 2 ) W a r , et al used national governments as the stimuli for the multidimensional scaling analyses while we used Swedish political parties, ( 3 ) the multidimensional technique used by Warr, et ul was Kruskal's non-metric analysis while we used Ekman's metric method, ( 4 ) Warr, et al did not compare the median cognitive differentiation

of their political groups as was done here, ( 5 ) the subjects composing Warr,

et al.'s "centre" group consisted of individuals uncommitted to any political party, the ideological makeup of which seems quite unknown The ideological groups used in our study (radicals, liberals, conservatives) were defined on the basis of their ideological positions and were shown to significantly differ on a measure of general conservatism -

Of the five differences above, the last two are perhaps the most fundamental The fact that Warr, et al did not compare the median differentiation scores be- tween the political groups might have had a severe effect upon the nature of the conclusions they reached Analysis of our data here has clearly shown that the use of the median rather than the mean has had a profound effect upon the con- clusions we have reached and can serve as a textbook example of when the arich- metic mean should not be used as a measure of central tendency However, re- gardless of which measure of central tendency one might care to use, i.e., arith- metic mean, median, or mode, the results here are not congruent with the dictates

of the extremism hypothesis The second fundamental difference between this study and the work of Warr, et al lies in the manner in which the latter defined their political groups The procedure used by Warr, et a) does not allow one to ascertain whether any cognitive differences (or lack of the same) are a function

of political ideology as such rather than political commitment The results found here tend to show that there are differences in the cognitive differentiation

of subjects who have ideological differences As already mentioned, this condu- sion is, for the time being, restricted to certain populations and certain cognitive domains Much more work needs to be done in order to ascertain just how far these conclusions can be generalized

REFERENCES ABELSON, R P A technique and a model for multidimensional attitude scaling Aner-

ican PsychoLogisr, 1954, 9, 319

ADORNO, T., FRENKEL-BRUNSWIK, E LEVINSON D., & SANFORD, N T h e authoritarian perronality New York : Harper, 1950

BIERJ, J., ATKINS, A L., BRIAR, S., LEAMAN, R L., MILLER, H & TRIPODI, T Clinical and social judgment: the discrimination o f behavioraL information New York:

Wiley, 1966

CATTELL, R B The scree test for the number of factors Multivariate Behavioral Re- search, 1966, 1, 245-276

EKEHAMMAR, B Multidimensional scaling according to different vector models for sub- jective similarity Acta Psychologica, 1972, 36, 79-84

EKMAN, G TWO methods for the analysis of perceptual dimensionality Perceptual and Motor Skills 1965, 20, 557-572

EKMAN, G Comparative studies on multidimensional scaling and related techniques

Reports f r o m t h e Psychological Laboratories, University of Stockholm, Supplement

3, 1970

EYSENCK, H J., & COULTER T T The personality and attitudes of working-class British communists and fascists Journal o f Social Psychology, 1972, 87, 59-73

FRENKEL-BRUNSWIK, E Intolerance of ambiguity as an emotional and perceptual vari- able Journal o f Personality, 1949, 18, 108-143

FRENKEL-BRUNSWK, E Further exploration by a contributor to The Authoritarian Personality In R Christie & M Jahoda (Eds.), Studies in t h e scope and method

o f t h e authoritarian personality New York: Free Press, 1954 Pp 226-275

GORSUCH, R L Factor analysis Philadelphia: Saunders, 1974

HAYS, W L Statistics New York: Holt, 1970

MESSICK, S The perception of social attitudes J o u r m l o f Abnormal and Social Psy- chology, 1956, 52, 57-66

MESSlCK, S The perceived structure of political relationships Sociometry, 1961, 24, 270-278

MOOD A M., & GRAYBILL, F A Introduction t o t h e theory o f statistics New York: McGraw-Hill, 1963

SCHRODER, H M., DRTVBR, M J., & S ~ U P E R T , S H u m a n information processing

New York: Holt, 1967

SHERMAN, R C., & ROSS, L B Liberalism-conservatism and dimensional salience in the perception of political figures IournuL o f Personality and Social Psychology, 1972,

23, 120-127

SHILS, E A Authoritarianism : " ~ i ~ h t " ' and "Left." In R Christie & M Jahoda (Eds.) ,

Studies in the scope and method o f t h e authoritarian personality New York: Free Press, 1964 Pp 24-49

SWANIUS, J Further tests of a Swedish scale of conservatism Reports f r o m the Depart-

m e n t o f Psychology, University o f Stockholm, No 467, 1976 ( a )

SIDANIUS, J Studies in political-psychology : theoretical and 'empirical speculations toward the derivation of theory Reports f f o m t h e Department of Psychology, University

o f Stockholm, Supplement 34, 1976 ( b )

S1DAN1US9 I A Swedish scale of conservatism Reports from t h e Department o f Psy- cho ogy, University o f Stockholm, No 465, 1976 ( c )

SIDANIUS, J Intolerance of ambiguity and socio- olitico ideology: a multidimensional analysis European Joarnal o f Social ~ s y c h o E ~ y , 1977, in press

SWANIUS, J., & EKEHAMMAR, B Cognitive functioning and socio-political ideology: a multidimensional and individualized analysis Scandinavian Journal o f PsychoG ogy, 1976, 17, 205-216

STONE, L A., & COLBS, G J Multidimensional judgment scaling of well-known political figures Journal o f Social Psychology, 197 2, 87, 127-1 37

TAYLOR, I A Similarities in the structure of extreme social attitudes Psychological Monographs, 1960, 74, 1-36

WARR, P B., SCHRODER, H M., & BLACKMAN, S The structure of political judgment

British Journal o f Social and Clinical Psychology, 1969, 8, 32-43

WILSON, G D., & PATTERSON, J R A new measure of conservatism ~ r i t i s h ' J o u r n a l o f Social and Clinical Psychology, 1968, 7 , 264-269

YELLIG, W F., & WEARING, A J T h e dimensions of political ideology Journal o f Sock! Psychology, 1974, 93, 119-131

Accepted J u n e 6, 1977