The
correlation
between the D.
Adorno-T-Authoritarian-Personality-Harper-Bros-1950
29 2.
48 3.
21 3.
16 2.
47 3.
66 2.
59 3.
16 3.
60 ,4.
59 3.
99 2.
67 1.
88 3.
85 3.
20 ,2.
47
Mean
5. 41 5. 16 4. 20 4. 69 3. 47 3. 80 5. 22 4. 59 3. 75
4. 53
(7) 2. 32 (22. 5) (22) 3. 19 (5)
(4) 3. 91 (4) (1) 2. 67 (23) '(6) 2. 89 (19) (16)
3. 31 (24) 2. 92 (27) 4. 34 (12) 2. 58 (28) 3. 38 (23) 4. 04 (17) 4. 88 (5)
2. 79
2~54
2. 98 1. 76 2. 90 2. 48 2. 84 1. 97 3. 25 1. 58 3. 13 2. 20 1. 70
(13) (17. 5) (8) (28) (10) (20. 5) ( 11) (26) (4) (30) (6)
( 25) (29)
(17) (5) (12) (24) (30) (13) (21) (27. 5)
2. 93 3. 10 2. 54 3. 36 1. 88 3. 18 2. 90 2. 29
4. 42 (11) 3. 56 (20) 2. 54 (29)
Men's GrOU(! S Rank D. P .
(1) 2. 70
(4) 2. 48 (15) 2. 55
Women's GrOUI! S
Rank D. P . Rank
(18) 2. 52 (2) 1. 77 (8) 3. 60 (19) 2. 54 (10) 2. 82 (21) 2. 74 (6) 3. 04 (14) 1. 98 (3) 2. 32 (9) 3. 68 (26) 2. 94
(8) 2. 98 (18) 3. 92 (15) 3. 14 (3) 2. 60 (11) 4. 03 (19) 2. 47 (10) 3. 52 (25) 3. 12 (9) 3. 72 (14) 2. 40 (2) 2. 87 (7) 4. 32 (29) 3. 39 (20) 3. 66
(22. 5) 3. 33 (27. 5) 2. 73 (16) 3. 02 (26) 2. 19 (22. 5) 2. 85
(3) (12) (24) (2) (26) (7) (13) (5) (27) (20) (1) (8) (6) (10) (22) (15) (29) (21) (17) (14) (25) (9) (30) ( 11) (18) (28)
(19) (27) (3) (17. 5) (12) (14) (7) (24) (22. 5) ( 2) (9)
(16) 3. 76 (1)
Mean/person/item 4. 00
arhese data are based on all fourteen groups taking Forms 40 and 45 (see Table 8 (VII)).
2. 64
3. 63
3. 08
? 45)a
Men and women combined
Mean
5. 04 5. 05 4. 26
Rank
(2)
(1) ,( 11)
D. P. Rank
3. 31 (6. 5) 2. 58 (22) 2. 72 (17)
(6) 2. 65 (19) . . . , 3. 45 (20) 3. 56 (4) ::q
4. 48
(17) (3. 5) (10) (19) (9)
2. 83 2. 19 3. 82 2. 51 3. 17
(16) M
3. 76
5. 01
4. 31
3. 50
4. 33
3. 31
4. 43 (7) 3. 38
4. 00 (14. 5) 2. 19(26. 5) ~
5. 01 4. 41 2. 77 3. 65 3. 24 2. 70 4. 00 2. 59 3. 27 3. 82 4. 74 4. 16 2. 89 2. 12 4. 14 3. 38 2. 51
3. 81
(3. 5) (8)
( 26) (18) (24) ( 27) ( 14. 5) (28) (23) (16) (5) (12) (25) (30) (13) (21) (29)
(5)
2. 60 (21) 1-<
(22) 2. 93
(14) ~
(26. 5) ( 2) (24) (8. 5)
>
~
::q
1-< . . . ,
4. 00 (1) >
3. 17 (8. 5)
3. 71 (3) "d
M
3. 06 ( 11)
2. 64 (20)
3. 00 (12)
1. 98 (29)
2. 88 (15)
2. 71 (18)
2. 97 (13) -< 2. 26 (25)
3. 31 (6. 5) 1. 73 (30) 3. 16 (10) 2. 55 (23) 2. 00 (28)
2. 85
N
0\ 0
z
~
en
t-<
1-< . . . ,
0 z >
? MEASUREMENT OF ANTIDEMOCRA TIC TRENDS
items which work well for one sex tend, in general, to work well for the other.
The correlation between the D. P. rank orders for the men and those for the women is . 84. This is sufficient justification for averaging the D. P. 's of the two groups to obtain an over-all "order of goodness" for each item. Since the differences between men and women, in the present context, are probably as great as the differences between any two groups of the same sex in the present sample, it is highly probable that a correlation between the D. P. rankings of any two such groups would be in the neighborhood of . 84- There appear to be no general or systematic differences between the items which work better for men and those which work better for women.
Mean scores for the men's groups are somewhat higher on the average than mean scores for the women's groups. This phenomenon would seem to be due primarily to the three male groups discussed above whose scores are particularly high. If men and women of the same socioeconomic class are compared, the means are not significantly different. Moreover, items which appeal most strongly to the men are much the same as those which appeal most strongly to the women, the rank-order correlation between the means for men and those for women being ? 95?
b. CoRRELATIONAL ANALYSIS. As a part of an independent investigation, the E, PEC, and F scales (from Forms 40 and 45) were administered to 900 students in an Elementary Psychology Class at the University of California. It was decided not to include the data from this new college group among the general results of the present study because the total sample of subjects was already weighted too heavily on the side of young and relatively well- educated people. However, the 517 women from this psychology class con- stitute the only group in whose case the scales were subjected to an item-by- item correlational analysis. 4 The results of this analysis will be summarized here.
Each item of the F scale was correlated with every other item. The average of the 435 coefficients was . I 3, the range -. 05 to ? 44? 5 In addition, each item was correlated with the remainder of the scale, the mean r here being . 33, the range . 15 to . 52. In the case of theE scale the mean interitem r was . 42, and the mean item-total scorer, ? 59? Whereas theE scale has about the same degree of unidimensionality as do acceptable intelligence tests (in the case of the I937 Stanford-Binet Revision the average interitem r is about . 38, the average item-total score r, . 6I), the F scale rates considerably lower in this regard. Despite the scale's relative lack of surface homogeneity, however, we are justified in speaking of an F pattern or syndrome, for the items do "hang together" in the sense that each is significantly correlated with the
4 This analysis was made possible by a Grant-in-aid from the Social Science Research Council.
5 Fisher's Zr was used in computing the average r.
? THE AUTHORITARIAN PERSONALITY
scale as a whole. It will be recalled in this connection that in constructing the F scale two purposes were held in mind: (a) to seek over a wide area for diverse responses that belonged to a single syndrome, and (b) to construct an instrument which would yield a reliable prediction of scores on E. It is clear that the first purpose has been in large part realized, although the search for additional items that would help characterize the F syndrome could be continued with profit. The fact that the individual F items correlate . 2 5 on the average with the total E scale augurs well for the fulfillment of the second purpose-a matter to which we shall turn in a moment.
Proof that the variables or groups of items used in thinking about the F scale are not clusters in the statistical sense, is contained in the data from the present group of 517 women. Although the items within each of the Form 45 F-clusters tend to intercorrebte (. I I to . 24), the items in any one cluster correlate with one another no better than they do with numerous items from other clusters. We are justified in using these clusters, therefore, only as a priori aids to discussion.
D. CORRELATIONS OF THE F SCALE WITH E AND WITH PEC
Correlations of F with the E and PEC scales, based on the three question- naire forms and derived from all the groups used in the study, are shown in Table ro (VII). The major result expressed in this table is that the correla- tion between E and F has increased with the successive revisions of the scale until it has reached a point (about ? 75 on the average in Forms 40, 45) where scores on the former can be predicted with fair accuracy from scores on the latter.
The correlation between F and E varies rather widely from one group to another, a matter that seems to depend mainly upon the reliability of the scales themselves. 6 Thus, in the San Quentin group, where the reliability of F is . 87 and that of E only . 65, the correlation between the two scales is at the lowest, ? 59; while in the case of the Working-Class Women, where the reliability of F climbs to . 97,7 the correlation is at its maximum, . 87. It is obvious, therefore, that if the reliabilities of the two scales were increased
(which can be done by increasing the number of items within each) the
6 The correlation between E and F does not seem to depend upon whether the two scales are administered at different times, or at the same time with items from the one scale in- terspersed among those of the other. The correlation obtained in the case of the Universiy of Oregon Student Women, who were given Form 6o in two parts, is not only similar to that obtained, with the use of the regular Form 6o, in the case of the University of Oregon and University of California Student Women, but it is virtually the same as the mean E. F correlation for all groups of subjects.
1 The reliability of the "A" half of the E scale, which was given as a part of Form 40 to that group, was not calculated.
? MEASUREMENT OF ANTIDEMOCRA TIC TRENDS TABLE 10 (VII)
CORRELATIOIIS OF THE F SCALE WITH THE A-S, E, MD PEC SCALES IN THE SEVERAL FORMS OF THE QUESTIONNAIRE
Groups taking Form 78:
Public Speaking Class Women 140 ? 55 ? 58 Public Speaking Class Men 52 . 52 ? 56 Extension Class Women 40 . 49 ? 74 Professional Women 63 ? 57 . 73
Over-a11a: Form 78 295 . 53 . 65
Groups taking Form 60:
Univ. of Oregon Student Women 47 . 72 Univ. o f Oregon and Univ. o f
California Student Women 54 ? 78 Univ. of Oregon and Univ. of
California Student Men 57 ? 58 Oregon Service Club Men 68 ? 69 Oregon Servic~ Club Menb 60
Over-all: Form 60 286 . 69
Groups taking Form 45:
Testing Class Women 59 . 79 San Quentin Men Prisoners 110 ? 59 Psychiatric Clinic Women 71 . 86 Psychiatric Clinic Men 50 ? 76 Working-Clas's Men and Women 50 ? 85 Employment Service Men Veterans 51 ? 67 Maritime School Men 179 . 56
Over-all: Form 45 570 . 73
Groups taking Form 40 e;
George Washington Univ. Women 132 ? 69 California Service Club Men 63 ? 80 Middle-Class Men 69 . 81 Working-Class Men 61 ? 76 Middle-Class Women 154 . 83 Working-Class Women 53 . 87 Los Angeles Men 117 ? 82 Los Angeles Women 130 ? 75 Employment Service Men Veterans 55 . 72 Maritime School Men 165 ? 62
Over-all: Form 40 999 ? 77
Over-all: All Forms 2150 . 53 . 73
~n obtaining the over-all group means, the individual group means were not weighted by N,
bThis group of Oregon service Club Men received a short questionnaire form containing only the F scale and half of the PEC scale,
cFbr the correlations of F with PEC in the Psychiatric Clinic groups, the number of women was 45, the number of men 29, due to a substitution of forms.
dThese F-PEC correlations are based on both Fbrms 40 and 45. Since it was con- sidered highly unlikely that the presence or absence of 5 E items would affect the correlation of F and PEC, the two forms are taken together in order to have the advantage of the larger N's. The total N is 106 for the Employment service Men Veterans, 343 for the Maritime School Men,
ein Fbrm 40, it will be recalled, only the "A" half of the 10-item E seale was used.
N F. A-S F. E
F. PEC
. 52 ? 45 ? 54 ? 65
. 54
. 29 ? 49
. 43 . 29 . 22
. 34
. 52
. 53 ? 59 . 71 ? 60 ? 70 . 72 ? 58 . 61 ? 62 ? 39
. 61 . 52
? THE AUTHORITARIAN PERSONALITY
correlation between E and F would be very high indeed. 8 This is not to say, however, that E and F for all practical purposes measure the same thing. A correlation of ? 775 means that about two-thirds of the subjects who score in the high quartile on the one scale, score in the high quartile on the other, and that there are practically no reversals, i. e. , cases in which a subject is high on one scale but low on the other. If one wished to use the F scale alone in order to single out subjects who were practically certain to be highly ethnocentric, i. e. , in the high quartile on the present E scale, it would be necessary for him to limit himself to those scoring at the very highest extreme on F, perhaps the top I o per cent. As pointed out earlier, there are reasons why some discrepancy between the two scales should be expected. Surely there are some individuals who have the kind of susceptibility to fascist propaganda with which the F scale is concerned but who for one reason or another tend to inhibit expressions of hostility against minority groups (sub- jects high on F but low on E). And we have good reason to believe that there are other people who rather freely repeat the cliches of ethnocentrism- perhaps in accordance with the climate of opinion in which they are living- without this being expressive of deep-lying trends in their personalities (sub- jects high on E but low on F). Such "exceptions" will be taken up in more detail later.
It is to be noted that the correlation between F and E is slightly higher on the average in the case of groups taking Form 40 than for groups taking Form 45? This means that F correlates slightly better with the A half of the E scale than with the total E scale, and that the correlation must be still lower in the case of the B half of the scale. In several groups taking Form 45 the correlations of EA and of EB with F were calculated, in addition to the cor- relation of total E with F. The results appear in Table 1 I (VII). In each
TABLE 11 (VII)
CORRELATIONS OF THE F SCALE WITH EACH HALF AND WITH THE WHOLE OF THE E SCALE
~
San Quentin Men Prisoners Employment Service Men Veterans Maritime School Men
Testing Class Women
Mean
N
110 51 179 59
Correlations
EA-F EB? F EA+B? F
. 56 ? 45 . 59 ? 66 . 61 . 67 ? 61 . 40 . 56 ? 77 ? 66 . 79
? 65 . 53 . 65
8 The correlation coefficient which, theoretically, would result if two scales were per~ fectly reliable, i. e. , if the average obtained r were corrected for attenuation, is about . 9. This indicates a striking correspondence, though not a complete identity, of what is meas- ured by the two scales.
? MEASUREMENT OF ANTIDEMOCRA TIC TRENDS
group EA. F is notably higher than EB. F, and about the same as EA+B? F. It may be recalled that the A half of the scale refers to highly generalized ethnocentrism and contains no A-S items, while the B half is made up of four A-S items and one Negro item. It happened that this Negro item was a relatively poor one in the statistical sense (rank order, 5 for men, 10 for women), but this is not enough to account for the superiority of the EA. F correlations. It seems, rather, that the F syndrome is actually more closely related to general ethnocentrism than to anti-Semitism. This is in keeping with the finding, reported earlier, that in Form 78 the F scale correlated more highly with theE scale than with the A-S scale. Although anti-Semitism is still to be understood primarily as an aspect of general ethnocentrism, there can be no doubt but that it has some special features of its own. Some of these features are described in Chapter XVI.
The F syndrome bears only a moderately close relation to politico- economic conservatism, the average correlation for Forms 45 and 40 being ? 57? Our interpretation is that high scores on PEC may proceed either from genuine conservatism or from pseudoconservatism, and that it is the latter which is most expressive of the personality trends which the F scale measures. This is in keeping with the finding that E, which is closely related to F, also shows only moderate correlation with PEC. The E. PEC correlation is about the same as the F. PEC correlation. It would appear that general ethnocen- trism, as measured by the present scales, is mainly an expression of those personality structures which the F scale measures; politico-economic con- servatism, while it may have this same source, may be more dependent than E upon factors in the individual's contemporary situation.
E. DIFFERENCES IN MEAN F-SCALE SCORE AMONG V ARIOUS GROUPS
W e may turn now to a consideration of the mean F-scale scores of different groups.
Mean
5. 41 5. 16 4. 20 4. 69 3. 47 3. 80 5. 22 4. 59 3. 75
4. 53
(7) 2. 32 (22. 5) (22) 3. 19 (5)
(4) 3. 91 (4) (1) 2. 67 (23) '(6) 2. 89 (19) (16)
3. 31 (24) 2. 92 (27) 4. 34 (12) 2. 58 (28) 3. 38 (23) 4. 04 (17) 4. 88 (5)
2. 79
2~54
2. 98 1. 76 2. 90 2. 48 2. 84 1. 97 3. 25 1. 58 3. 13 2. 20 1. 70
(13) (17. 5) (8) (28) (10) (20. 5) ( 11) (26) (4) (30) (6)
( 25) (29)
(17) (5) (12) (24) (30) (13) (21) (27. 5)
2. 93 3. 10 2. 54 3. 36 1. 88 3. 18 2. 90 2. 29
4. 42 (11) 3. 56 (20) 2. 54 (29)
Men's GrOU(! S Rank D. P .
(1) 2. 70
(4) 2. 48 (15) 2. 55
Women's GrOUI! S
Rank D. P . Rank
(18) 2. 52 (2) 1. 77 (8) 3. 60 (19) 2. 54 (10) 2. 82 (21) 2. 74 (6) 3. 04 (14) 1. 98 (3) 2. 32 (9) 3. 68 (26) 2. 94
(8) 2. 98 (18) 3. 92 (15) 3. 14 (3) 2. 60 (11) 4. 03 (19) 2. 47 (10) 3. 52 (25) 3. 12 (9) 3. 72 (14) 2. 40 (2) 2. 87 (7) 4. 32 (29) 3. 39 (20) 3. 66
(22. 5) 3. 33 (27. 5) 2. 73 (16) 3. 02 (26) 2. 19 (22. 5) 2. 85
(3) (12) (24) (2) (26) (7) (13) (5) (27) (20) (1) (8) (6) (10) (22) (15) (29) (21) (17) (14) (25) (9) (30) ( 11) (18) (28)
(19) (27) (3) (17. 5) (12) (14) (7) (24) (22. 5) ( 2) (9)
(16) 3. 76 (1)
Mean/person/item 4. 00
arhese data are based on all fourteen groups taking Forms 40 and 45 (see Table 8 (VII)).
2. 64
3. 63
3. 08
? 45)a
Men and women combined
Mean
5. 04 5. 05 4. 26
Rank
(2)
(1) ,( 11)
D. P. Rank
3. 31 (6. 5) 2. 58 (22) 2. 72 (17)
(6) 2. 65 (19) . . . , 3. 45 (20) 3. 56 (4) ::q
4. 48
(17) (3. 5) (10) (19) (9)
2. 83 2. 19 3. 82 2. 51 3. 17
(16) M
3. 76
5. 01
4. 31
3. 50
4. 33
3. 31
4. 43 (7) 3. 38
4. 00 (14. 5) 2. 19(26. 5) ~
5. 01 4. 41 2. 77 3. 65 3. 24 2. 70 4. 00 2. 59 3. 27 3. 82 4. 74 4. 16 2. 89 2. 12 4. 14 3. 38 2. 51
3. 81
(3. 5) (8)
( 26) (18) (24) ( 27) ( 14. 5) (28) (23) (16) (5) (12) (25) (30) (13) (21) (29)
(5)
2. 60 (21) 1-<
(22) 2. 93
(14) ~
(26. 5) ( 2) (24) (8. 5)
>
~
::q
1-< . . . ,
4. 00 (1) >
3. 17 (8. 5)
3. 71 (3) "d
M
3. 06 ( 11)
2. 64 (20)
3. 00 (12)
1. 98 (29)
2. 88 (15)
2. 71 (18)
2. 97 (13) -< 2. 26 (25)
3. 31 (6. 5) 1. 73 (30) 3. 16 (10) 2. 55 (23) 2. 00 (28)
2. 85
N
0\ 0
z
~
en
t-<
1-< . . . ,
0 z >
? MEASUREMENT OF ANTIDEMOCRA TIC TRENDS
items which work well for one sex tend, in general, to work well for the other.
The correlation between the D. P. rank orders for the men and those for the women is . 84. This is sufficient justification for averaging the D. P. 's of the two groups to obtain an over-all "order of goodness" for each item. Since the differences between men and women, in the present context, are probably as great as the differences between any two groups of the same sex in the present sample, it is highly probable that a correlation between the D. P. rankings of any two such groups would be in the neighborhood of . 84- There appear to be no general or systematic differences between the items which work better for men and those which work better for women.
Mean scores for the men's groups are somewhat higher on the average than mean scores for the women's groups. This phenomenon would seem to be due primarily to the three male groups discussed above whose scores are particularly high. If men and women of the same socioeconomic class are compared, the means are not significantly different. Moreover, items which appeal most strongly to the men are much the same as those which appeal most strongly to the women, the rank-order correlation between the means for men and those for women being ? 95?
b. CoRRELATIONAL ANALYSIS. As a part of an independent investigation, the E, PEC, and F scales (from Forms 40 and 45) were administered to 900 students in an Elementary Psychology Class at the University of California. It was decided not to include the data from this new college group among the general results of the present study because the total sample of subjects was already weighted too heavily on the side of young and relatively well- educated people. However, the 517 women from this psychology class con- stitute the only group in whose case the scales were subjected to an item-by- item correlational analysis. 4 The results of this analysis will be summarized here.
Each item of the F scale was correlated with every other item. The average of the 435 coefficients was . I 3, the range -. 05 to ? 44? 5 In addition, each item was correlated with the remainder of the scale, the mean r here being . 33, the range . 15 to . 52. In the case of theE scale the mean interitem r was . 42, and the mean item-total scorer, ? 59? Whereas theE scale has about the same degree of unidimensionality as do acceptable intelligence tests (in the case of the I937 Stanford-Binet Revision the average interitem r is about . 38, the average item-total score r, . 6I), the F scale rates considerably lower in this regard. Despite the scale's relative lack of surface homogeneity, however, we are justified in speaking of an F pattern or syndrome, for the items do "hang together" in the sense that each is significantly correlated with the
4 This analysis was made possible by a Grant-in-aid from the Social Science Research Council.
5 Fisher's Zr was used in computing the average r.
? THE AUTHORITARIAN PERSONALITY
scale as a whole. It will be recalled in this connection that in constructing the F scale two purposes were held in mind: (a) to seek over a wide area for diverse responses that belonged to a single syndrome, and (b) to construct an instrument which would yield a reliable prediction of scores on E. It is clear that the first purpose has been in large part realized, although the search for additional items that would help characterize the F syndrome could be continued with profit. The fact that the individual F items correlate . 2 5 on the average with the total E scale augurs well for the fulfillment of the second purpose-a matter to which we shall turn in a moment.
Proof that the variables or groups of items used in thinking about the F scale are not clusters in the statistical sense, is contained in the data from the present group of 517 women. Although the items within each of the Form 45 F-clusters tend to intercorrebte (. I I to . 24), the items in any one cluster correlate with one another no better than they do with numerous items from other clusters. We are justified in using these clusters, therefore, only as a priori aids to discussion.
D. CORRELATIONS OF THE F SCALE WITH E AND WITH PEC
Correlations of F with the E and PEC scales, based on the three question- naire forms and derived from all the groups used in the study, are shown in Table ro (VII). The major result expressed in this table is that the correla- tion between E and F has increased with the successive revisions of the scale until it has reached a point (about ? 75 on the average in Forms 40, 45) where scores on the former can be predicted with fair accuracy from scores on the latter.
The correlation between F and E varies rather widely from one group to another, a matter that seems to depend mainly upon the reliability of the scales themselves. 6 Thus, in the San Quentin group, where the reliability of F is . 87 and that of E only . 65, the correlation between the two scales is at the lowest, ? 59; while in the case of the Working-Class Women, where the reliability of F climbs to . 97,7 the correlation is at its maximum, . 87. It is obvious, therefore, that if the reliabilities of the two scales were increased
(which can be done by increasing the number of items within each) the
6 The correlation between E and F does not seem to depend upon whether the two scales are administered at different times, or at the same time with items from the one scale in- terspersed among those of the other. The correlation obtained in the case of the Universiy of Oregon Student Women, who were given Form 6o in two parts, is not only similar to that obtained, with the use of the regular Form 6o, in the case of the University of Oregon and University of California Student Women, but it is virtually the same as the mean E. F correlation for all groups of subjects.
1 The reliability of the "A" half of the E scale, which was given as a part of Form 40 to that group, was not calculated.
? MEASUREMENT OF ANTIDEMOCRA TIC TRENDS TABLE 10 (VII)
CORRELATIOIIS OF THE F SCALE WITH THE A-S, E, MD PEC SCALES IN THE SEVERAL FORMS OF THE QUESTIONNAIRE
Groups taking Form 78:
Public Speaking Class Women 140 ? 55 ? 58 Public Speaking Class Men 52 . 52 ? 56 Extension Class Women 40 . 49 ? 74 Professional Women 63 ? 57 . 73
Over-a11a: Form 78 295 . 53 . 65
Groups taking Form 60:
Univ. of Oregon Student Women 47 . 72 Univ. o f Oregon and Univ. o f
California Student Women 54 ? 78 Univ. of Oregon and Univ. of
California Student Men 57 ? 58 Oregon Service Club Men 68 ? 69 Oregon Servic~ Club Menb 60
Over-all: Form 60 286 . 69
Groups taking Form 45:
Testing Class Women 59 . 79 San Quentin Men Prisoners 110 ? 59 Psychiatric Clinic Women 71 . 86 Psychiatric Clinic Men 50 ? 76 Working-Clas's Men and Women 50 ? 85 Employment Service Men Veterans 51 ? 67 Maritime School Men 179 . 56
Over-all: Form 45 570 . 73
Groups taking Form 40 e;
George Washington Univ. Women 132 ? 69 California Service Club Men 63 ? 80 Middle-Class Men 69 . 81 Working-Class Men 61 ? 76 Middle-Class Women 154 . 83 Working-Class Women 53 . 87 Los Angeles Men 117 ? 82 Los Angeles Women 130 ? 75 Employment Service Men Veterans 55 . 72 Maritime School Men 165 ? 62
Over-all: Form 40 999 ? 77
Over-all: All Forms 2150 . 53 . 73
~n obtaining the over-all group means, the individual group means were not weighted by N,
bThis group of Oregon service Club Men received a short questionnaire form containing only the F scale and half of the PEC scale,
cFbr the correlations of F with PEC in the Psychiatric Clinic groups, the number of women was 45, the number of men 29, due to a substitution of forms.
dThese F-PEC correlations are based on both Fbrms 40 and 45. Since it was con- sidered highly unlikely that the presence or absence of 5 E items would affect the correlation of F and PEC, the two forms are taken together in order to have the advantage of the larger N's. The total N is 106 for the Employment service Men Veterans, 343 for the Maritime School Men,
ein Fbrm 40, it will be recalled, only the "A" half of the 10-item E seale was used.
N F. A-S F. E
F. PEC
. 52 ? 45 ? 54 ? 65
. 54
. 29 ? 49
. 43 . 29 . 22
. 34
. 52
. 53 ? 59 . 71 ? 60 ? 70 . 72 ? 58 . 61 ? 62 ? 39
. 61 . 52
? THE AUTHORITARIAN PERSONALITY
correlation between E and F would be very high indeed. 8 This is not to say, however, that E and F for all practical purposes measure the same thing. A correlation of ? 775 means that about two-thirds of the subjects who score in the high quartile on the one scale, score in the high quartile on the other, and that there are practically no reversals, i. e. , cases in which a subject is high on one scale but low on the other. If one wished to use the F scale alone in order to single out subjects who were practically certain to be highly ethnocentric, i. e. , in the high quartile on the present E scale, it would be necessary for him to limit himself to those scoring at the very highest extreme on F, perhaps the top I o per cent. As pointed out earlier, there are reasons why some discrepancy between the two scales should be expected. Surely there are some individuals who have the kind of susceptibility to fascist propaganda with which the F scale is concerned but who for one reason or another tend to inhibit expressions of hostility against minority groups (sub- jects high on F but low on E). And we have good reason to believe that there are other people who rather freely repeat the cliches of ethnocentrism- perhaps in accordance with the climate of opinion in which they are living- without this being expressive of deep-lying trends in their personalities (sub- jects high on E but low on F). Such "exceptions" will be taken up in more detail later.
It is to be noted that the correlation between F and E is slightly higher on the average in the case of groups taking Form 40 than for groups taking Form 45? This means that F correlates slightly better with the A half of the E scale than with the total E scale, and that the correlation must be still lower in the case of the B half of the scale. In several groups taking Form 45 the correlations of EA and of EB with F were calculated, in addition to the cor- relation of total E with F. The results appear in Table 1 I (VII). In each
TABLE 11 (VII)
CORRELATIONS OF THE F SCALE WITH EACH HALF AND WITH THE WHOLE OF THE E SCALE
~
San Quentin Men Prisoners Employment Service Men Veterans Maritime School Men
Testing Class Women
Mean
N
110 51 179 59
Correlations
EA-F EB? F EA+B? F
. 56 ? 45 . 59 ? 66 . 61 . 67 ? 61 . 40 . 56 ? 77 ? 66 . 79
? 65 . 53 . 65
8 The correlation coefficient which, theoretically, would result if two scales were per~ fectly reliable, i. e. , if the average obtained r were corrected for attenuation, is about . 9. This indicates a striking correspondence, though not a complete identity, of what is meas- ured by the two scales.
? MEASUREMENT OF ANTIDEMOCRA TIC TRENDS
group EA. F is notably higher than EB. F, and about the same as EA+B? F. It may be recalled that the A half of the scale refers to highly generalized ethnocentrism and contains no A-S items, while the B half is made up of four A-S items and one Negro item. It happened that this Negro item was a relatively poor one in the statistical sense (rank order, 5 for men, 10 for women), but this is not enough to account for the superiority of the EA. F correlations. It seems, rather, that the F syndrome is actually more closely related to general ethnocentrism than to anti-Semitism. This is in keeping with the finding, reported earlier, that in Form 78 the F scale correlated more highly with theE scale than with the A-S scale. Although anti-Semitism is still to be understood primarily as an aspect of general ethnocentrism, there can be no doubt but that it has some special features of its own. Some of these features are described in Chapter XVI.
The F syndrome bears only a moderately close relation to politico- economic conservatism, the average correlation for Forms 45 and 40 being ? 57? Our interpretation is that high scores on PEC may proceed either from genuine conservatism or from pseudoconservatism, and that it is the latter which is most expressive of the personality trends which the F scale measures. This is in keeping with the finding that E, which is closely related to F, also shows only moderate correlation with PEC. The E. PEC correlation is about the same as the F. PEC correlation. It would appear that general ethnocen- trism, as measured by the present scales, is mainly an expression of those personality structures which the F scale measures; politico-economic con- servatism, while it may have this same source, may be more dependent than E upon factors in the individual's contemporary situation.
E. DIFFERENCES IN MEAN F-SCALE SCORE AMONG V ARIOUS GROUPS
W e may turn now to a consideration of the mean F-scale scores of different groups.