Significant differences between comparable groups of men and women might, of course, be found on various
individual
items; this problem has not been systematically explored.
Adorno-T-Authoritarian-Personality-Harper-Bros-1950
29
1. 53 1. 70 1. 61
340 1. 0-7. 0
Mean (total) Mean (A half) Mean (B half)
S. D. (total) S. D. (A half) S. D. (Bhalf)
N
Range
3. 41 3. 77 3. 06
1. 40 1. 68 1. 35
59 110 1. 0-6. 1 1. 6-7. 0
arhe groups taking this form are as follows: Group I: Extension Testing Class Women Group II: San Quentin Men Prisoners
Group III: Psychiatric Clinic Women
Group IV: Psychiatric Clinic Men Group V: Working Class Men and Women
bin obtaining the over-all means, the individual group means were not weighted by N.
crhe reliabilities for Form 45 are not based on odd-even or equivalent halves but on EA vs. E8; they are therefore slightly lower than they would be had equivalent halves been used (see text).
17 This group was not included in the over-all sample because the proportion of students in the sample was already too great. This group was obtained for the primary purpose of making a correlational analysis of the Form 45 scales, particularly the F scale (see Chapter VII).
? Property
Mean
S. D.
Mean D. P. N
Ranged
TABLE 17 {IV} (CONT'D. } RELIABILITY OF' THE E SCALE ~FORMS 45 AND 40)
B. Groups Taking Form 40 (EA) a
Groupb
VI VII VIII IX X XI XII XIII
Over-all c
3. 90 1. 87 4. 87
779
Group VI: Group VII: Group VIII: Group IX: Group X: Group XI: Group XII: Group XIII:
George Washington University Women California Service Club Men Middle-Class Men
Middle-Class Women
working-Class Men working-Class Women Los Angeles Men
Los Angeles Women
4. 04 1. 58 4. . 10
4. 31 3. 89 1. 73 2. 08 4. 54 5. 28
3. 64 1. 96 5. 11
3. 92 3. 91 1. 71 2. 25 4. 53 5. 64
3. 82 1. 89 5. 08
3. 71 1. 78 4. 67
132 63 69 154 61 53 117 130
aThe E scale in Form 40 contained only 5 items, referred to as the EA items and identical to the EA items in Form 45.
brhe groups taking this form are as follows:
ern obtaining the over-all means, the individual group means were not weighted by N. drhe range in every case was 1. 0-7. 0.
? q6
THE AUTHORITARIAN PERSONALITY TABLE 17 {IV) (CONT'D. )
RELIABILITY OF THE E SCALE {FORMS 45 AND 40) a c. Groups Taking both Forms 45 and 40
Property
Over-all
. 80
4. 30 4. 74 3. 85
1. 42 1. 52 1. 54
230 1. 1-6. 6
Mean {EA~
S. D. (EA)
N 55 164 219 Range 1. 0-7. 0
Form 45:
Reliability
Mean {total) Mean {A half) Mean {B half)
S. D. {total) S. D. {Ahalf) S. D. {Bhalf)
Employment Service Men Veterans
? 86
4. 26 4. 67 3. 85
1. 60 1. 63 1. 71
Maritime School Men
? 73
4. 34 4. 82 3. 85
1. 25 1. 40 1. 36
N 51 179
Range 1. 1-6. 6
Form 40:
1. 2-6. 6
4. 21 1. 75
5. 08 1. 47
4. 64 1. 61
aThe total number of cases on Forms 45 and 40 is as follows: Form 45 Form 40 Total
N 570 998 1568
is reported for each group in Table 17 (IV)B. The over-all mean D. P. of 4. 87 suggests what the total E reliability also suggests: that the subjects show a relatively high degree of consistency in response to all items. The mean D. P. in four of the eight groups is over 5. o; this suggests that the distribu- tion of scores is bimodal, that is, that the subjects tend either to agree strongly or to disagree strongly (in contrast to the more common result in which scores cluster around the "uncertain" neutral point). The high S. D. 's and wide range of scores indicate the same thing.
The group differences in average degree of ethnocentrism are of some interest. Among the groups taking Form 45, the three which stand clearly at the head of the list in terms of mean E score are the San Quentin Men
(4. 6r), the Maritime School Men (4. 34), and the Employment Service Men Veterans (4. 26), these means being significantly higher than the others (3. 34-3. 67). That the San Quentin Men are so ethnocentric makes it clear that being in a subordinate group is not a guarantee against ethnocentrism. The results for the San Quentin group, and the psychological affinity be-
tween criminality and fascism, are considered in detail in Chapter XXI.
It is unclear why, in the Veteran and Maritime School groups, the EA
Group
1. 2-7. 0 1. 0-7. 0
? THE STUDY OF ETHNOCENTRIC IDEOLOGY I 37
means should be so different in Form 40 as compared with Form 45 (Table I7(IV) C). Thus, for the Veterans, the EA mean drops from 4. 67 to 4. 2 I, while for the Maritime School it increases from 4. 82 to 5. 08. Although these differences are not statistically significant (at the 5 per cent level), they might, if they were both in the same direction, suggest a general systematic difference between the two Forms. It might he hypothesized, for example, that the presence of the anti-Semitic items in EB makes some people defensive and thus lowers the mean on the entire scale in Form 45? This hypothesis is opposed, however, by the facts that neither difference is significant, that in the Maritime School the EA mean is higher in Form 40 than in Form 45, and that the EA means in the other Form 40 groups (Table I7(IV) B) are of the order of magnitude as in the Form 45 groups. It would appear, in short, that the presence of the EB items in Form 45 produces no systematic increase
or decrease in scores on the other items.
The mean E score of 3. 7, as well as the wide range and the large S. D. , for
the Psychiatric Clinic patients indicates that no simple relationship exists between psychological ill health and ethnocentrism. The degree of ethno- centrism in this group of neurotic and psychotic-primarily the former- individuals just about equals the average of all groups tested. It would appear incorrect, therefore, to assume that there is on the average more pathology, psychologically speaking, in ethnocentrists than in nonethnocentrists or conversely. 18 Evidence to be presented later, however (Chapter XXII), will show that high and low scorers differ significantly with respect to type of pathology. The least ethnocentric groups taking Form 45 and 40 are the Testing Class Women and the Working-Class Men and Women. The low mean for the former group is consistent with previous results on University groups in California and Oregon. The EA mean for the Form 45 group of Working-Class subjects is slightly but nonsignificantly lower than for the larger Working-Class group taking Form 40. This difference is apparently due to the fact that the Form 45 sample contains a greater proportion of sub- jects from the California Labor School, a subgroup with an extremely low E mean. Further discussion of the relation of economic class and politico-eco- nomic ideology to ethnocentrism is reserved for Chapter V. From the results in Table I7(1V), particularly for the groups taking Form 40, it would appear that socioeconomic class, as such, is not a major determinant of dif- ferences in ethnocentrism. The means for the Middle-Class groups are almost identical with those for the Working-Class groups. This is not to
18 This conclusion depends, of course, on the representatives of our sample. What can be stated unequivocally is that every quartile on E contains some psychologically dis- turbed individuals. We may suspect, however, that a truly random sample of seriously disturbed individuals would show a higher average degree of ethnocentrism than is shown by the present sample, which includes, for the most part, individuals who recognize their problems as primarily psychological and who are willing to undergo psychological treatment-personality trends associated, as later chapters will show, with lack of ethno- centrism.
? THE AUTHORITARIAN PERSONALITY
say that economic forces play no role in ethnocentrism, or that class member- ship is unimportant. However, the average amount of ethnocentrism in the two classes appears to be the same, to the extent that the measuring instru- ment is valid and the sample adequate. Moreover, there are wide variations within each class, some groups being very high in ethnocentrism, others very low. Thus, within the middle class, the service clubs are significantly more ethnocentric than the university groups. Individual and group differences in E score within each class are associated with differences in ideology (political, religious, and so forth) and in personality as shown by the chapters which follow.
c. ITEM ANALYsis: FoRMs 45 AND 40. The item means and D. P. 's for the groups taking Forms 45 and 40 are presented in Table r8(IV). While the item means for men average slightly higher than those for women, the rank orders of the individual item means and D. P. 's are similar for the two sexes. Furthermore, the wide range of the over-all item means and D. P. 's suggests that similar consistency exists among the various groups of men and women comprising the total sample. In other words, the relative level of acceptability (mean) and "goodness" (D. P. ) of the items is fairly stable from group to group.
The best items in Form 45 deal with Negroes, Jews, zootsuiters, and foreigners. For the women two items, 32 (Negroes' own fault) and 40 (Jew- ish neighborhoods), had means of below 3. 0 and D. P. 's ranking ro and 9 respectively. Even the lowest D. P. for men and for women (3. 0 in each case) is sufficient to differentiate high from low scorers with a minimum of overlap. The only item in Form 45 with a mean of over 5. 0 for both men and women is number 45 (World organization). While this item dis- criminates very well between low and high scorers on the total scale, the low scorers are apparently less sure of themselves on the issue of national sovereignty than on the other issues; the high scorers almost uniformly rate this item +3, but the low scorers are less emphatic and more divided.
The significantly higher means for men than for women on both forms may not reflect a true sex difference since they are not based on comparable groups of men and women. Thus, the four highest men's groups (San Quen- tin, Veterans, Maritime School, Service Clubs) have no high-scoring coun- terparts among the women. The absence of a significant sex difference is also suggested by the very similar means obtained by comparable sex groups (see Table r7(IV) B): Working-Class, Middle-Class, and Los Angeles Men and Women.
Significant differences between comparable groups of men and women might, of course, be found on various individual items; this problem has not been systematically explored.
The differences in means and D. P. 's between Forms 45 and 40 may also be less significant than they appear at first glance. That the mean D. P. is almost one point higher for both sexes on Form 40 than on Form 45 is
? Mean
4. 14
(Negro rights)
15. (Foreign ideas) 3. 81
D. P. Rank D. P. Mean D. P. Rank D. P. Mean
5. 10.
4. 57
3. 57 (7) 4. 58 (1) 3. 34 (9) 4. 54 (2) 4. 04 (4) 3. 49 (8) 3. 82 (5) 4. 35 (3) 3. 71 (6) 3. 01 (10)
3. 84
4. 10 4. 26 (5) 3. 64 4. 25 4. 92 (2) 3. 93 3. 64 4. 44 (3) 3. 26
4. 05 4. 91 3. 99 4. 60 3. 88 3. 50 2. 99 4. 24 3. 26 3. 50
3. 89
(4) 3. 48 ( 1) 3. 96 (5) 3. 25 (2) 3. 86 (6)
Item
(Zootsuiters)
D. P. R1Ulk D. P. Mean D. P. Rank D. P.
No.
20. (Negro foremen) 24. (J. businessmen) 28. (Marry a Jew)
32. (Negroes live) 36. (Jews alike)
40. (Jewish Neighb. ) 45. (World org. )
4. 50 (4) 5. 21 (1) 4. 49 (5) 5. 15 (2)
TABLE 18 (IV)
MEANS AND DISCRIMINATORY POWERS OF THE E-SCALE ITE1dS (FORMS 45 AND 40)
MEN'S GROUPSa (N = 969) WOMEN'S GROUPBP (N = 599)
Form 45 (N=440) Form 40 (N=529) Form 45 (N=130) Form 40 (N=469)
4. 32 4. 11 3. 25 3. 10 4. 03 3. 55 5. 54
4. 04
4. 07
4. 93
(1)
4. 00 3. 48 3. 24 2. 42 3. 20 2. 96 5. 16
3. 53
4. 97
4. 20
4. 38
4. 59
(4)
(7. 5) (10) (3) (9) (7. 5)
4. 58
3. 83
5. 07 (3)
4. 88
aThe data were obtained from the following groups of men:
Service Men Veterans (N = 106)? Maritime School Men (N = 343). California Service Club Men (N = 63). Psychi- atric Clinic Men (N = 50). Middle-Class Men (N = 69), Working-Class Men (N = 61). Los Angeles Men (N = 117). The Working-Class Men and Women (N = 50) were also included here since 34 of these 50 subjects were men.
bThe data were obtained from the following groups of women: Extension Testing Class Women (N =59)? George Washington University Women Students (N = 132). Psychiatric Clinic Women (N = 71)? Middle-Class Women (N = 154). Working-Class Women (N = 53). Los Angeles women (N = 130).
Cin obtaining the over-all means and D. P,? s, the individual group values were not weighted by N.
San Quentin Men Prisoners (N = 110).
Employment
? THE AUTHORITARIAN PERSONALITY
due in part to the smaller number of items in Form 40 (each item therefore contributing a larger portion of the total score). It is also partly due to sampling factors: the composition of the various samples taking Form 40 was more heterogeneous, resulting in larger S. D. 's (Table 17(1V)), more extreme scorers, and thus higher D. P. 's. Both men and women had slightly lower EA means on Form 40 than on Form 45 (4. 48-4. 20 for men, 4. oo-3. 83 for women). For reasons discussed earlier, these differences in means may be attributed mainly to sampling differences (both systematic and random) rather than to the nature of the forms themselves.
d. CoRRELATIONAL ANALYSIS: FoRM 45? It was possible, using the group of 517 University of California student women mentioned above, to make a correlational analysis of theE scale (Form 45). 19 Only the highlights of these results need be presented here. The group was near the average of the total sample with respect to mean (3. 64), S. D. (1. 52), and reliability (. 79 for EA vs. EB, . 87 for odd vs. even halves). For the single items the means ranged from 2. 25 for Item 32 (Negroes' own fault) to 5. oo for Item 45 (World organization), while the S. D. 's ranged from 1. 77 for Item 32 to 2. 47 for Item 28 (Marry a Jew). The average of the interitem correlations was . 42. The lowest interitem r's, . 25 and . 26, were between Item 15 (Foreign ideas) and Items 40 (Jewish neighborhoods) and 32 (Negroes' own fault), respectively. The highest r's, . 61 and . 62, were between Items 24 (Jewish businessmen) and 36 (Jews alike), and between Items ro (Negro rights) and 20 (Negro foremen), respectively. The correlations between each item and the sum of the remaining items averaged . 59; the two lowest values, ? 43 and . 46, were for Items 15 and 45, the two highest, . 67 and . 69, for Items
ro and 36. Six of the ten items correlated . 6o or higher with the sum of the remaining ones. These results, including the rank order of goodness of items and the general level of magnitude of the correlations, are consistent with the results for the other groups. While there is a tendency for items refer- ring to a given group to cluster somewhat, the predominant trend is toward broad internal consistency. That the consistency is incomplete is shown by the fact that the correlations are far from perfect. In terms of statistical rigor, the scale shows about the same degree of unidimensionality (consistency) , as the standard intelligence tests.
e. AGE AND ETHNOCENTRISM. The total sample from which the above data were obtained was not randomly distributed with respect to age. Its mem- bers were predominantly in their twenties and thirties, a disproportion- ately small number being in their forties or older. It was hypothesized that younger people tend to be less conservative and less ethnocentric than their elders, and that the mean E scores for the present sample might consequently
19 We wish to express our thanks to the Social Science Research Council for the funds which made this aspect of the research possible.
? THE STUDY OF ETHNOCENTRIC IDEOLOGY
141
be lower than for the population at large. As a partial check on this hypoth- esis, correlations between age and E score (Form 45) were computed for the Psychiatric Clinic Men and W omen (N = I 2 I ) . This group, despite its atypicality with respect to psychological health, appeared to be the most diverse group taking Form 45, and its E-scale results (mean, reliability, in- ternal consistency, and correlations with other scales) were fairly representa- tive of the total sample. Approximately So per cent of this group was between I 8 and 40 years old, the mean (and median) being 34 years. The figures for men were very similar to those for women.
The correlation between age and E score for both men and women was . I 9? This value for men and women combined is significantly above zero at the 5 per cent (lowest acceptable) level of confidence. It suggests that there is a slight but consistent tendency for younger adults to be less ethnocentric than those of middle or old age. That the correlation is not likely to be higher for the general population is indicated by the fact that very high
E-scale means were made by such young adult groups as the Employment Service Veterans and the Maritime School Men. The sampling bias in favor of younger age levels appears, then, to be of minor importance in our final results: the average degree of ethnocentrism found is slightly but significantly lower than would be found in a more representative sample.
3. A SUGGESTED FINAL E SCALE
The above data indicate that even the brief ten-item E scale in Form 45 can be used as a dependable measure of ethnocentrism. Whenever pos- sible, however, it would appear advantageous to use a longer form which is likely to be more reliable and which can cover in a more systematic fashion various aspects of ethnocentric ideology. Table I9 (IV) contains a suggested finalE scale of 20 items-short enough to meet most demands of practicality, yet sufficiently long to meet theoretical and statistical requirements. These items were selected, on the basis of both statistical adequacy and theoretical significance, from previous forms of the Ethnocentrism and Anti-Semitism scales. The wording of several items has been shortened and simplified. Sub- scales have been reinstated: six items dealing with Jews, six with Negroes, and eight with other Minorities and Patriotism.
1. 53 1. 70 1. 61
340 1. 0-7. 0
Mean (total) Mean (A half) Mean (B half)
S. D. (total) S. D. (A half) S. D. (Bhalf)
N
Range
3. 41 3. 77 3. 06
1. 40 1. 68 1. 35
59 110 1. 0-6. 1 1. 6-7. 0
arhe groups taking this form are as follows: Group I: Extension Testing Class Women Group II: San Quentin Men Prisoners
Group III: Psychiatric Clinic Women
Group IV: Psychiatric Clinic Men Group V: Working Class Men and Women
bin obtaining the over-all means, the individual group means were not weighted by N.
crhe reliabilities for Form 45 are not based on odd-even or equivalent halves but on EA vs. E8; they are therefore slightly lower than they would be had equivalent halves been used (see text).
17 This group was not included in the over-all sample because the proportion of students in the sample was already too great. This group was obtained for the primary purpose of making a correlational analysis of the Form 45 scales, particularly the F scale (see Chapter VII).
? Property
Mean
S. D.
Mean D. P. N
Ranged
TABLE 17 {IV} (CONT'D. } RELIABILITY OF' THE E SCALE ~FORMS 45 AND 40)
B. Groups Taking Form 40 (EA) a
Groupb
VI VII VIII IX X XI XII XIII
Over-all c
3. 90 1. 87 4. 87
779
Group VI: Group VII: Group VIII: Group IX: Group X: Group XI: Group XII: Group XIII:
George Washington University Women California Service Club Men Middle-Class Men
Middle-Class Women
working-Class Men working-Class Women Los Angeles Men
Los Angeles Women
4. 04 1. 58 4. . 10
4. 31 3. 89 1. 73 2. 08 4. 54 5. 28
3. 64 1. 96 5. 11
3. 92 3. 91 1. 71 2. 25 4. 53 5. 64
3. 82 1. 89 5. 08
3. 71 1. 78 4. 67
132 63 69 154 61 53 117 130
aThe E scale in Form 40 contained only 5 items, referred to as the EA items and identical to the EA items in Form 45.
brhe groups taking this form are as follows:
ern obtaining the over-all means, the individual group means were not weighted by N. drhe range in every case was 1. 0-7. 0.
? q6
THE AUTHORITARIAN PERSONALITY TABLE 17 {IV) (CONT'D. )
RELIABILITY OF THE E SCALE {FORMS 45 AND 40) a c. Groups Taking both Forms 45 and 40
Property
Over-all
. 80
4. 30 4. 74 3. 85
1. 42 1. 52 1. 54
230 1. 1-6. 6
Mean {EA~
S. D. (EA)
N 55 164 219 Range 1. 0-7. 0
Form 45:
Reliability
Mean {total) Mean {A half) Mean {B half)
S. D. {total) S. D. {Ahalf) S. D. {Bhalf)
Employment Service Men Veterans
? 86
4. 26 4. 67 3. 85
1. 60 1. 63 1. 71
Maritime School Men
? 73
4. 34 4. 82 3. 85
1. 25 1. 40 1. 36
N 51 179
Range 1. 1-6. 6
Form 40:
1. 2-6. 6
4. 21 1. 75
5. 08 1. 47
4. 64 1. 61
aThe total number of cases on Forms 45 and 40 is as follows: Form 45 Form 40 Total
N 570 998 1568
is reported for each group in Table 17 (IV)B. The over-all mean D. P. of 4. 87 suggests what the total E reliability also suggests: that the subjects show a relatively high degree of consistency in response to all items. The mean D. P. in four of the eight groups is over 5. o; this suggests that the distribu- tion of scores is bimodal, that is, that the subjects tend either to agree strongly or to disagree strongly (in contrast to the more common result in which scores cluster around the "uncertain" neutral point). The high S. D. 's and wide range of scores indicate the same thing.
The group differences in average degree of ethnocentrism are of some interest. Among the groups taking Form 45, the three which stand clearly at the head of the list in terms of mean E score are the San Quentin Men
(4. 6r), the Maritime School Men (4. 34), and the Employment Service Men Veterans (4. 26), these means being significantly higher than the others (3. 34-3. 67). That the San Quentin Men are so ethnocentric makes it clear that being in a subordinate group is not a guarantee against ethnocentrism. The results for the San Quentin group, and the psychological affinity be-
tween criminality and fascism, are considered in detail in Chapter XXI.
It is unclear why, in the Veteran and Maritime School groups, the EA
Group
1. 2-7. 0 1. 0-7. 0
? THE STUDY OF ETHNOCENTRIC IDEOLOGY I 37
means should be so different in Form 40 as compared with Form 45 (Table I7(IV) C). Thus, for the Veterans, the EA mean drops from 4. 67 to 4. 2 I, while for the Maritime School it increases from 4. 82 to 5. 08. Although these differences are not statistically significant (at the 5 per cent level), they might, if they were both in the same direction, suggest a general systematic difference between the two Forms. It might he hypothesized, for example, that the presence of the anti-Semitic items in EB makes some people defensive and thus lowers the mean on the entire scale in Form 45? This hypothesis is opposed, however, by the facts that neither difference is significant, that in the Maritime School the EA mean is higher in Form 40 than in Form 45, and that the EA means in the other Form 40 groups (Table I7(IV) B) are of the order of magnitude as in the Form 45 groups. It would appear, in short, that the presence of the EB items in Form 45 produces no systematic increase
or decrease in scores on the other items.
The mean E score of 3. 7, as well as the wide range and the large S. D. , for
the Psychiatric Clinic patients indicates that no simple relationship exists between psychological ill health and ethnocentrism. The degree of ethno- centrism in this group of neurotic and psychotic-primarily the former- individuals just about equals the average of all groups tested. It would appear incorrect, therefore, to assume that there is on the average more pathology, psychologically speaking, in ethnocentrists than in nonethnocentrists or conversely. 18 Evidence to be presented later, however (Chapter XXII), will show that high and low scorers differ significantly with respect to type of pathology. The least ethnocentric groups taking Form 45 and 40 are the Testing Class Women and the Working-Class Men and Women. The low mean for the former group is consistent with previous results on University groups in California and Oregon. The EA mean for the Form 45 group of Working-Class subjects is slightly but nonsignificantly lower than for the larger Working-Class group taking Form 40. This difference is apparently due to the fact that the Form 45 sample contains a greater proportion of sub- jects from the California Labor School, a subgroup with an extremely low E mean. Further discussion of the relation of economic class and politico-eco- nomic ideology to ethnocentrism is reserved for Chapter V. From the results in Table I7(1V), particularly for the groups taking Form 40, it would appear that socioeconomic class, as such, is not a major determinant of dif- ferences in ethnocentrism. The means for the Middle-Class groups are almost identical with those for the Working-Class groups. This is not to
18 This conclusion depends, of course, on the representatives of our sample. What can be stated unequivocally is that every quartile on E contains some psychologically dis- turbed individuals. We may suspect, however, that a truly random sample of seriously disturbed individuals would show a higher average degree of ethnocentrism than is shown by the present sample, which includes, for the most part, individuals who recognize their problems as primarily psychological and who are willing to undergo psychological treatment-personality trends associated, as later chapters will show, with lack of ethno- centrism.
? THE AUTHORITARIAN PERSONALITY
say that economic forces play no role in ethnocentrism, or that class member- ship is unimportant. However, the average amount of ethnocentrism in the two classes appears to be the same, to the extent that the measuring instru- ment is valid and the sample adequate. Moreover, there are wide variations within each class, some groups being very high in ethnocentrism, others very low. Thus, within the middle class, the service clubs are significantly more ethnocentric than the university groups. Individual and group differences in E score within each class are associated with differences in ideology (political, religious, and so forth) and in personality as shown by the chapters which follow.
c. ITEM ANALYsis: FoRMs 45 AND 40. The item means and D. P. 's for the groups taking Forms 45 and 40 are presented in Table r8(IV). While the item means for men average slightly higher than those for women, the rank orders of the individual item means and D. P. 's are similar for the two sexes. Furthermore, the wide range of the over-all item means and D. P. 's suggests that similar consistency exists among the various groups of men and women comprising the total sample. In other words, the relative level of acceptability (mean) and "goodness" (D. P. ) of the items is fairly stable from group to group.
The best items in Form 45 deal with Negroes, Jews, zootsuiters, and foreigners. For the women two items, 32 (Negroes' own fault) and 40 (Jew- ish neighborhoods), had means of below 3. 0 and D. P. 's ranking ro and 9 respectively. Even the lowest D. P. for men and for women (3. 0 in each case) is sufficient to differentiate high from low scorers with a minimum of overlap. The only item in Form 45 with a mean of over 5. 0 for both men and women is number 45 (World organization). While this item dis- criminates very well between low and high scorers on the total scale, the low scorers are apparently less sure of themselves on the issue of national sovereignty than on the other issues; the high scorers almost uniformly rate this item +3, but the low scorers are less emphatic and more divided.
The significantly higher means for men than for women on both forms may not reflect a true sex difference since they are not based on comparable groups of men and women. Thus, the four highest men's groups (San Quen- tin, Veterans, Maritime School, Service Clubs) have no high-scoring coun- terparts among the women. The absence of a significant sex difference is also suggested by the very similar means obtained by comparable sex groups (see Table r7(IV) B): Working-Class, Middle-Class, and Los Angeles Men and Women.
Significant differences between comparable groups of men and women might, of course, be found on various individual items; this problem has not been systematically explored.
The differences in means and D. P. 's between Forms 45 and 40 may also be less significant than they appear at first glance. That the mean D. P. is almost one point higher for both sexes on Form 40 than on Form 45 is
? Mean
4. 14
(Negro rights)
15. (Foreign ideas) 3. 81
D. P. Rank D. P. Mean D. P. Rank D. P. Mean
5. 10.
4. 57
3. 57 (7) 4. 58 (1) 3. 34 (9) 4. 54 (2) 4. 04 (4) 3. 49 (8) 3. 82 (5) 4. 35 (3) 3. 71 (6) 3. 01 (10)
3. 84
4. 10 4. 26 (5) 3. 64 4. 25 4. 92 (2) 3. 93 3. 64 4. 44 (3) 3. 26
4. 05 4. 91 3. 99 4. 60 3. 88 3. 50 2. 99 4. 24 3. 26 3. 50
3. 89
(4) 3. 48 ( 1) 3. 96 (5) 3. 25 (2) 3. 86 (6)
Item
(Zootsuiters)
D. P. R1Ulk D. P. Mean D. P. Rank D. P.
No.
20. (Negro foremen) 24. (J. businessmen) 28. (Marry a Jew)
32. (Negroes live) 36. (Jews alike)
40. (Jewish Neighb. ) 45. (World org. )
4. 50 (4) 5. 21 (1) 4. 49 (5) 5. 15 (2)
TABLE 18 (IV)
MEANS AND DISCRIMINATORY POWERS OF THE E-SCALE ITE1dS (FORMS 45 AND 40)
MEN'S GROUPSa (N = 969) WOMEN'S GROUPBP (N = 599)
Form 45 (N=440) Form 40 (N=529) Form 45 (N=130) Form 40 (N=469)
4. 32 4. 11 3. 25 3. 10 4. 03 3. 55 5. 54
4. 04
4. 07
4. 93
(1)
4. 00 3. 48 3. 24 2. 42 3. 20 2. 96 5. 16
3. 53
4. 97
4. 20
4. 38
4. 59
(4)
(7. 5) (10) (3) (9) (7. 5)
4. 58
3. 83
5. 07 (3)
4. 88
aThe data were obtained from the following groups of men:
Service Men Veterans (N = 106)? Maritime School Men (N = 343). California Service Club Men (N = 63). Psychi- atric Clinic Men (N = 50). Middle-Class Men (N = 69), Working-Class Men (N = 61). Los Angeles Men (N = 117). The Working-Class Men and Women (N = 50) were also included here since 34 of these 50 subjects were men.
bThe data were obtained from the following groups of women: Extension Testing Class Women (N =59)? George Washington University Women Students (N = 132). Psychiatric Clinic Women (N = 71)? Middle-Class Women (N = 154). Working-Class Women (N = 53). Los Angeles women (N = 130).
Cin obtaining the over-all means and D. P,? s, the individual group values were not weighted by N.
San Quentin Men Prisoners (N = 110).
Employment
? THE AUTHORITARIAN PERSONALITY
due in part to the smaller number of items in Form 40 (each item therefore contributing a larger portion of the total score). It is also partly due to sampling factors: the composition of the various samples taking Form 40 was more heterogeneous, resulting in larger S. D. 's (Table 17(1V)), more extreme scorers, and thus higher D. P. 's. Both men and women had slightly lower EA means on Form 40 than on Form 45 (4. 48-4. 20 for men, 4. oo-3. 83 for women). For reasons discussed earlier, these differences in means may be attributed mainly to sampling differences (both systematic and random) rather than to the nature of the forms themselves.
d. CoRRELATIONAL ANALYSIS: FoRM 45? It was possible, using the group of 517 University of California student women mentioned above, to make a correlational analysis of theE scale (Form 45). 19 Only the highlights of these results need be presented here. The group was near the average of the total sample with respect to mean (3. 64), S. D. (1. 52), and reliability (. 79 for EA vs. EB, . 87 for odd vs. even halves). For the single items the means ranged from 2. 25 for Item 32 (Negroes' own fault) to 5. oo for Item 45 (World organization), while the S. D. 's ranged from 1. 77 for Item 32 to 2. 47 for Item 28 (Marry a Jew). The average of the interitem correlations was . 42. The lowest interitem r's, . 25 and . 26, were between Item 15 (Foreign ideas) and Items 40 (Jewish neighborhoods) and 32 (Negroes' own fault), respectively. The highest r's, . 61 and . 62, were between Items 24 (Jewish businessmen) and 36 (Jews alike), and between Items ro (Negro rights) and 20 (Negro foremen), respectively. The correlations between each item and the sum of the remaining items averaged . 59; the two lowest values, ? 43 and . 46, were for Items 15 and 45, the two highest, . 67 and . 69, for Items
ro and 36. Six of the ten items correlated . 6o or higher with the sum of the remaining ones. These results, including the rank order of goodness of items and the general level of magnitude of the correlations, are consistent with the results for the other groups. While there is a tendency for items refer- ring to a given group to cluster somewhat, the predominant trend is toward broad internal consistency. That the consistency is incomplete is shown by the fact that the correlations are far from perfect. In terms of statistical rigor, the scale shows about the same degree of unidimensionality (consistency) , as the standard intelligence tests.
e. AGE AND ETHNOCENTRISM. The total sample from which the above data were obtained was not randomly distributed with respect to age. Its mem- bers were predominantly in their twenties and thirties, a disproportion- ately small number being in their forties or older. It was hypothesized that younger people tend to be less conservative and less ethnocentric than their elders, and that the mean E scores for the present sample might consequently
19 We wish to express our thanks to the Social Science Research Council for the funds which made this aspect of the research possible.
? THE STUDY OF ETHNOCENTRIC IDEOLOGY
141
be lower than for the population at large. As a partial check on this hypoth- esis, correlations between age and E score (Form 45) were computed for the Psychiatric Clinic Men and W omen (N = I 2 I ) . This group, despite its atypicality with respect to psychological health, appeared to be the most diverse group taking Form 45, and its E-scale results (mean, reliability, in- ternal consistency, and correlations with other scales) were fairly representa- tive of the total sample. Approximately So per cent of this group was between I 8 and 40 years old, the mean (and median) being 34 years. The figures for men were very similar to those for women.
The correlation between age and E score for both men and women was . I 9? This value for men and women combined is significantly above zero at the 5 per cent (lowest acceptable) level of confidence. It suggests that there is a slight but consistent tendency for younger adults to be less ethnocentric than those of middle or old age. That the correlation is not likely to be higher for the general population is indicated by the fact that very high
E-scale means were made by such young adult groups as the Employment Service Veterans and the Maritime School Men. The sampling bias in favor of younger age levels appears, then, to be of minor importance in our final results: the average degree of ethnocentrism found is slightly but significantly lower than would be found in a more representative sample.
3. A SUGGESTED FINAL E SCALE
The above data indicate that even the brief ten-item E scale in Form 45 can be used as a dependable measure of ethnocentrism. Whenever pos- sible, however, it would appear advantageous to use a longer form which is likely to be more reliable and which can cover in a more systematic fashion various aspects of ethnocentric ideology. Table I9 (IV) contains a suggested finalE scale of 20 items-short enough to meet most demands of practicality, yet sufficiently long to meet theoretical and statistical requirements. These items were selected, on the basis of both statistical adequacy and theoretical significance, from previous forms of the Ethnocentrism and Anti-Semitism scales. The wording of several items has been shortened and simplified. Sub- scales have been reinstated: six items dealing with Jews, six with Negroes, and eight with other Minorities and Patriotism.
