The higher mean and discriminability of the former item are probably due to its greater
indirectness
and distance from crude anti- Semitism.
Adorno-T-Authoritarian-Personality-Harper-Bros-1950
42
Mean:
Means for total. scale:
4 . 1 9 4. 27
Number: Total group =144; H. Q. =36: L. Q. =37.
? Range of total scores: Total group: 52-286; H. Q. : 183-286; L. Q. : 52-89.
5. 89 2. 97 4. 89 4. 17 4. 47 3. 39 5. 28 5. 05 3. 22 6. 58 4. 36 3. 89 4. 97
4. 64 3. 86 4. 03 4. 78 2. 89 2. 89
? 8o THE AUTHORITARIAN PERSONALITY
with few scores of 6 or 7 (+z and +3 responses). And group means of over 5. 0, conversely, indicate relatively uniform agreement.
The Discriminatory Power, on the other hand, is a measure of the variabil- ity of the high and low scorers around the group mean, and of their average difference in response. How large must a D. P. be in order to indicate almost no overlap between highs and lows? This depends on the form of the dis- tribution and the size of the group mean. An item with a group mean of z. o, a low quartile mean of I. o, a high quartile mean of 3. 0, and a D. P. of z. o, is undoubtedly very discriminating; the low scorers responded unanimously with -3, and the high scorers probably varied but little around the -1 re- sponse. In general, the more extreme the group mean (especially below 3. 0 or above 5. 0) the lower the D. P. can be and still adequately separate the low from the high scorers. From a broader point of view, however, the best items should have means nearer to 4-o; when the item mean is above 5. 0 or below 3. 0, the item should be reworded so that fewer people or more people, re- spectively, will agree. 2
For iteins with group means in the approximate range 3. 0 to 5. 0, Dis- criminatory Powers may be evaluated according to the following general standards: a D. P. of over 4. 0 is very high and indicates almost uniform agree- ment by the high scorers, disagreement by the low scorers, with almost no overlap. D. P. 's of 3. o-4-o are very satisfactory and indicate a clear-cut dif- ference between high and low scorers. D. P. 's of z. o-3. 0, while statistically significant, indicate greater variability in the responses of low and high scorers and a fair amount of overlap. A D. P. between 1. 0 and z. o involves considerable agreement by the low scorers and disagreement by the high scorers, but it still indicates a statistically significant difference between the low mean and the high mean. 3 As the D. P. decreases below 1. 0, the possi- bility of significance decreases rapidly.
With these considerations in mind we can examine the data in Table 9? In general the Discriminatory Powers are quite satisfactory, averaging 2. 85 for the entire group. 4 For the 52 items, 5 D. P. 's are over 4-0, 21 are between
2 A minimum item mean of 2. 5 ought probably to be set for this group, since various studies have shown college students to be less prejudiced than the general population. For other groups studied in the present research, many item means were as much as a point higher.
3 While standard deviations have not been obtained for all items, it can be shown that (with group N = roo to r 50) the standard error of the difference between the means for low and high scorers is almost never above . so, seldom below . 2 5. In terms of the critical ratio, then, a D. P. of over r. o is statistically significant, that is, the means are different
though the distributions are partially overlapping.
4 While correlations between items or between each item and the total scale have not
been computed for this group, later data on similar scales suggest that the average inter- item correlation is about -4, while between each item and the sum of the remaining items the average correlation is about . 6. (See Chapter IV . )
? THE STUDY OF ANTI-SEMITIC IDEOLOGY 8r
3. 0 and 3? 9, and 15 are betWeen z. o and 2. 9. Only II D. P. 's are between r. o and 1. 9, the lowest being 1. 2.
All of the D. P. 's are therefore above a minimum standard of acceptability. The 26 items with D. P. 's of over 3. 0 are statistically very satisfactory. Why were the other items less adequate? The answer is indicated by the group means on these items. Of the II items with D. P. 's of less than z. o, IO have means below 2. 1. Conversely, almost all of the items with means of over 3. 0 have D. P. 's of over 3. 0. The mean for the low quartile is very low (below 2. o) on almost every item. The mean of the high quartile, on the other hand, varies greatly from item to item. The items with low D. P. 's were, in almost every case, statements with which the high quartile tended predominantly to disagree. This result seems to be due partly to a lack of pseudodemocratic coloring in these items, partly to their obvious illogicality or lack of truth, and partly to a lack of extreme anti-Semites among these subjects. Thus, the mean of the high quartile on all 52 items averages 4? 3 and varies from 2. 2 to 6. 6. For the IO items with the highest D. P. 's, however, the high quartile means average well over 5. 0.
The dependability of the item means and D. P. 's is indicated by a com- parison of the present group with the group of college students previously tested and reported on. The latter group responded to the entire scale (and other questionnaire material) at one sitting; the reliability was . 98 and the subscale intercorrelations were also slightly higher than in the present group. The group mean per item was 3. 0 as compared with 2. 7 for the present group, and the average D. P. was 3-4 as compared with 2. 85 here. The main differ- ence between the two groups seems to lie in the greater number of high scorers in the first group tested. The over-all mean of the low quartile was almost identical for the two groups: I. 39 then, I-42 now. But the high quartile averaged 4. 8o then as compared to 4. 27 now. In noncollege groups a larger number of high scorers, and larger D. P. 's, have been found (see pp. 76, I4o). Despite the over-all scale differences between the two groups, however, the adequacy of the individual items was very similar. Thus, the rank-order cor- relation between the D. P. 's was . 78, while the item means correlated . 92. In short, the most discriminating items for one group were also the best for the other group, and similarly for the poor items. The general conclusions about item means and discriminabilities to be drawn from the present group, then, are generally true for the previous group as well.
Table 9(III) reveals that the best items pertain to a variety of topics. Stereotypy in anti-Semitic imagery is shown in the tendency to overgen- eralize Jewish faults, and in the fact that Item 1-I 3 ("Jews are all alike") is one of the most discriminating, with a D. P. of 3. 83. The idea of Jews as a political threat (radicalism: Items 11-6, -I4) was much less prevalent than the idea of Jews as an economic threat (wealth and power: Items l-5, -8,-22,
? THE AUTHORITARIAN PERSONALITY
II-5, -II) or as a moral threat (immoralit~': Items l-1, II-I, -7, -q). Ac- cusations on religious grounds were seldom supported," as shown by the low group means and D. P. 's on Items 1-zo and Il-19. A variety of discriminatory attitudes (exclusion, restriction, suppression: most items in the "Attitude" subscale) found considerable support and \vere very discriminating.
The importance of careful formulation of items is shown by a comparison of good with poor items. The most discriminating items are usually the most pseudodemocratic ones according to criteria discussed above (Section B, I). Consider, for example, the two items dealing with intermarriage. Item 11-ro ("I can hardly imagine myself marrying a Jew") has a group mean of 4. 2 and a D. P. of 4. 3, with a high quartile mean of 6. 6 (almost all +2 and +3 responses). On the other hand, Item 1-I 5 ("It is wrong for Jews and Gentiles to intermarry") has a group mean of only 2. 6, a D. P. of 3. 0, and a high quartile mean of 4. 2.
The higher mean and discriminability of the former item are probably due to its greater indirectness and distance from crude anti- Semitism. By what criterion is this item anti-Semitic? The criterion is the fact that it correlates well with the total scale, that is, it differentiates very well between subjects who score high and subjects who score low on the total scale. (The fact that the correlation is not perfect indicates that re- sponse to any single item is not a clear-cut sign of anti-Semitism nor of its opposite; the criterion must be the total scale score).
Similar reasoning applies to items dealing with housing restrictions. The following items had very low group means (r. 5 to 2. 5): Item II-IS (con- sistent exclusion from hotels), l-3 (exclusion from neighborhoods), I-2I
(occasional exclusion from apartment houses). Item Il-7 (Jews give a neigh- borhood a "typical Jewish atmosphere"), which is more indirect and pseudo- democratic, had a higher mean (3. 2) and D. P. (4. 0). It would appear that many individuals who are not now willing actively to support anti-Semitic programs have nevertheless a negative imagery and an underlying hostility that constitute a definite potentiality for such action. Even the more open and crude items on housing had significant D. P. 's, and the high quartile means of 4? 3 and over (except on 11-IS), seem to indicate only weak re- sistance to these ideas.
The same considerations hold for the items dealing with occupations. Items which urge explicit policies of suppression and restriction of Jews (1-6, --<), II-<), -I 5) tend to have low means. But items which emphasize gen-
5 It is frequently held that Sunday School training is a major cause of anti-Semitism, which is then regarded as a form of "religious prejudice. " In this group, at least, rejection on religious grounds was infrequent. From the generality and irrationality of anti-Semitic ideology, it is clear that many diverse accusations are almost always involved, and that there are many sources for the underlying hostility which makes a given individual recep- tive to anti-Semitism. For a discussion of the role of religion in prejudice see Chapters VI and XVIII.
? THE STUDY OF A"N"TI-SEMITIC IDEOLOGY
eralized Jewish faults and? which introduce subtly discriminatory practices tend to have higher means; thus, Jewish businessmen are regarded as unfair and conniving (Item l-5), and they have too much financial power (1-8). While active suppression is not so desirable (low mean on I-<)), it is unwise for an employer to hire many Jews (higher mean on l-24, ll-3). The simplest solution-one that eliminates the need for suppression-would be for Jews to form their own nation (ll-24).
The pessimism of the high scorers regarding the solution to this problem is brought out by a number of items. On the one hand, they take the view that anti-Semitism has been entirely or primarily brought by the Jews on themselves and that any solution of the problem is a matter of Jewish re- sponsibility (Items ll-1, li-z 1). Non-Jews are simply the victims of Jewish faults; if Jews would improve, become as good as "other people," anti- Semitism would be eliminated. On the other hand, the Jews seem to be in- corrigible, and any apparent change only masks the Jewishness beneath (Items l-7, -q, ll-4, -8, -22). The contradiction is therefore complete: anti-Semitism is due to Jewish faults, but the Jews are unable to improve; the Jews should make sincere efforts to change, but their "basic Jewishness" is unchangeable. For the antidemocratic anti-Semite the only answer is open and direct suppression; for the pseudodemocrat it is subtle exclusion and "resigned tolerance" toward a bad state of affairs. The pseudodemocrats seem to betray a sense of threat and some antidemocratic potential by their doubts that democratic methods can solve the problem (Item l-12).
D. THE SHORT FORM OF THE A-S SCALE
It was a regular policy of the present research to contract the proven techniques in order to introduce new ones measuring additional trends of theoretical importance. In line with this policy, and in view of the high reli- ability and internal consistency of the original sz-item A-S scale, a short form of ten items was used in the first revision of the questionnaire.
The short form is presented in Table 1o(lll). The ten items were selected from the original fifty-two on the basis of both statistical and theoretical con- siderations. Since statistical adequacy (Discriminatory Power) was a neces- sary-but not sufficient-condition for inclusion, the new items were selected from the fifteen or twenty which had been most discriminating on the two administrations of the long form. Among these, selection was determined by the following qualitative considerations. Each item should be as rich in mean- ing as possible. There should be a minimum of duplication of meaning or con- tent among items. They should cover most of the subscales and most of the areas of accusation and discrimination. These desiderata have not been realized entirely; there were other items that seemed to merit inclusion, and for certain
? Old No.
l-24
II-I7 II-I
11-21
11-IO 11-7
1-7
1-I3 1-II
New No. ?
I I.
I6.
21.
26.
33?
40. 49?
62.
72.
TABLE IO (III)
THE TEN-ITEM A-S ScALE (Form 78)
Anyone who employs many people should be careful not to hire a large percentage of Jews.
One trouble with Jewish businessmen is that they stick to- gether and connive, so that a Gentile doesn't have a fair chance in competition.
The Jewish districts in most cities are results of the clannish- ness and stick-togetherness of Jews.
Persecution of the Jews would be largely eliminated if the Jews would make really sincere efforts to rid themselves of their harmful and offensive faults.
Jewish leaders should encourage Jews to be more incon- spicuous, to keep out of professions and activities already overcrowded with Jews and to keep out of the public notice. I can hardly imagine myself marrying a Jew.
The trouble with letting Jews into a nice neighborhood is that they gradually give it a typical Jewish atmosphere.
No matter how Americanized a Jew may seem to be, there is always something different and strange, something basically
Jewish underneath.
There may be a few exceptions, but, in general, Jews are pretty much alike.
There are too many Jews in the various federal agencies and bureaus in Washington, and they have too much control over our national policies.
THE AUTHORITARIAN PERSONALITY
? "New number" refers to the numbering of the items in Form 78.
"Old number" refers to numbering in the long form discussed previously. Slight revi-
sions will be noted in the wording of several items.
purposes they would probably be superior. The high internal consistency of the long form indicates that several statistically adequate short forms might be constructed. Nevertheless, the present form was expected to provide an adequate tool for most purposes of measurement. The slight revisions in the wording of some items were intended to make them simpler and clearer in meaning. The manner of presentation of this form was different from that previously used. Whereas previously each scale had been presented "all of a piece," on a page or pages of its own, in this and all successive forms of the questionnaire the various scales were presented interspersed with each other, so that no single scale was particularly prominent or focal, and adjacent items dealt with widely varying topics. 6
The new questionnaire, identified as Form 78 (on the basis of its having
6 The other scales in this form, to be discussed in the chapters that follow, deal with general prejudice (Negroes, other minorities, patriotism), with politico-economic liberal- ism and conservatism, and with potentially antidemocratic personality trends. There were 78 items in all. This form of the questionnaire, like all the other forms, contained in addi- tion other questions dealing with group memberships, personality, and so on.
? THE STUDY OF ANTI-SEMITIC IDEOLOGY
78 scale items), was administered in the spring of I945 to the following groups. Two of the groups comprise undergraduate students at the Univer- sity of California: the full membership, save for absentees, of the introduc- tory Public Speaking class. Here, as in al~ groups, men and women were separated in the statistical treatment and analysis was limited to native-born, white, non-Jewish Americans. The first two groups, then, are the Public Speaking Women (N = I 40) and the Public Speaking Men (N = sz). The third group comprised forty women, the entire feminine membership of an Extension Division class in Psychology at the University of California. Most subjects of this group were in their thirties and late twenties, and hence were somewhat older on the average than those of our college sample. The fourth group, Professional Women (N = 63), is actually a combination of three smaller groups: (a) Twenty-four public health nurses, the entire nursing staff of a nearby health department (the director of this department was generally liberal in his outlook and had tried to select younger nurses with more advanced ideas about public health); (b) public school teachers; and (c) social workers, woo were reached through the mails. In the latter two cases, only about 20 per cent of those appealed to sent in their question- naires, and this sampling technique was not tried again.
The reliability data for the short A-S scale are presented in Table II (III). Reliabilities of . 89-. 94 are extremely satisfactory, especially for a 10-item scale, and they are similar to those obtained on the long form. 7 The means of 3? 3 to 3? 4 for University and Extension Class students are substantially the same as the mean of 3? 55 on these ten items for the previous class taking the long form of the scale. However, the mean of 2. 6 for the Professional Women is significantly lower than the others (above the I per cent level statistically). This difference may be due partly to sampling errors; the teachers and social workers responded voluntarily by mail, and the tendency to cooperate in filling out a questionnaire dealing with prejudice and with personal feelings is probably correlated with lack of prejudice. 8 The slightly greater reliability
(. 94) of the scale for this group may reflect a greater ideological consistency in older age groups.
The Discriminatory Power method of item analysis was again carried out, and the results are presented in Table 12(III). The average D. P. of 3. 68 is very satisfactory and indicates that on most items there were very few low- quartile members who agreed, few high-quartile members who disagreed.
7 The fact that these reliabilities are similar to those obtained on the long form argues against the hypothesis that the high reliability of the latter was due to a "set" for all-nega- tive items.
8 This hypothesis is supported by questionnaire and clinical material on personality trends (opposition to "prying" and to "being analyzed" in the prejudiced subjects). Also, fewer high-scoring than low-scoring subjects in the groups tested were willing. to be interviewed.
? 86 THE AUTHORITARIAN PERSO~ALITY TABLE 11 (III~
RELIABILITY OF THE A-S SCALE (FORM 78)
Propertyc Gr. Aa Reliability . 89
Gr. Ba
093
3. 36 3. 30 3. 42
1. 48 1. 51 1. 56
1. 1-6. 3
Gr. Ca
. 90
3. 40 3. 20 3. 63
1. 36 1. 38 1. 48
1. 2-6. 1
Gr. Da
0 94
2. 57 2. 34 2. 83
1. 37 1. 27 1. 58
1. 0-6. 2
OV er-all . 92
3. 16 2. 96 3. 38
1. 41 1. 40 1. 56
1. 0-7. 0
b
Mean (total) Mean (odd half) Mean (even half)
S. D. (total) S. D. (odd. half) S. D. (even half)
Range
3. 33 2. 98 3. 66
1. 43 1. 42 1. 62
1. 0-7. 0
aThe four groups on which these data are based are: Group A, U. C. Public Speaking Class Women {N =140); Group 8, U. C. Public Speaking Class Men
{N =52); Group C, U. C. Adult Extension Clas54Wornen {N =40); Group D, Professional women {nurses, teachers, social- workers, N = 63).
bin obtaining the over-all means, the indtvidual group means were not weighted by N.
cThe values of the means, Standard Deviations, and ranges are given in terms of mean/person/item. If multiplied by 10 {the number of items), they are translated into values representing total scale score per ' peraon.
The best items deal with such varied topics as conniving businessmen, Jews being all alike, intermarriage, exclusion from neighborhoods.
How much influence did the form of presentation of the items have on their individual means and D. P. 's? Does it matter whether the items are presented in a solid block, as in the first form, or randomly dispersed through a -longer series of extremely varied items, as in Form 78? Evidence bearing on this question was obtained by comparing the results on these ten items for the two types of presentation. The mean for the Psychology Class women on these ten items (first form, excluding the remaining forty-two items) was 3? 55, as compared with 3? 32 for the Public Speaking Class women, the most
comparable group taking Form 78, and the average D. P. 's were 3. 76 and 3. 68 respectively. The differences are not statistically significant. Furthermore, the rank-order correlations between the individual item means for these two groups was . 62, while the D. P. 's correlated . 90. These correlations seem even more significant when one considers that the wording of some items was changed, and that the two groups were not systematically equated. The results? on the first form were also compared with the over-all averages for all four groups taking Form 78. The individual item means correlated .
Mean:
Means for total. scale:
4 . 1 9 4. 27
Number: Total group =144; H. Q. =36: L. Q. =37.
? Range of total scores: Total group: 52-286; H. Q. : 183-286; L. Q. : 52-89.
5. 89 2. 97 4. 89 4. 17 4. 47 3. 39 5. 28 5. 05 3. 22 6. 58 4. 36 3. 89 4. 97
4. 64 3. 86 4. 03 4. 78 2. 89 2. 89
? 8o THE AUTHORITARIAN PERSONALITY
with few scores of 6 or 7 (+z and +3 responses). And group means of over 5. 0, conversely, indicate relatively uniform agreement.
The Discriminatory Power, on the other hand, is a measure of the variabil- ity of the high and low scorers around the group mean, and of their average difference in response. How large must a D. P. be in order to indicate almost no overlap between highs and lows? This depends on the form of the dis- tribution and the size of the group mean. An item with a group mean of z. o, a low quartile mean of I. o, a high quartile mean of 3. 0, and a D. P. of z. o, is undoubtedly very discriminating; the low scorers responded unanimously with -3, and the high scorers probably varied but little around the -1 re- sponse. In general, the more extreme the group mean (especially below 3. 0 or above 5. 0) the lower the D. P. can be and still adequately separate the low from the high scorers. From a broader point of view, however, the best items should have means nearer to 4-o; when the item mean is above 5. 0 or below 3. 0, the item should be reworded so that fewer people or more people, re- spectively, will agree. 2
For iteins with group means in the approximate range 3. 0 to 5. 0, Dis- criminatory Powers may be evaluated according to the following general standards: a D. P. of over 4. 0 is very high and indicates almost uniform agree- ment by the high scorers, disagreement by the low scorers, with almost no overlap. D. P. 's of 3. o-4-o are very satisfactory and indicate a clear-cut dif- ference between high and low scorers. D. P. 's of z. o-3. 0, while statistically significant, indicate greater variability in the responses of low and high scorers and a fair amount of overlap. A D. P. between 1. 0 and z. o involves considerable agreement by the low scorers and disagreement by the high scorers, but it still indicates a statistically significant difference between the low mean and the high mean. 3 As the D. P. decreases below 1. 0, the possi- bility of significance decreases rapidly.
With these considerations in mind we can examine the data in Table 9? In general the Discriminatory Powers are quite satisfactory, averaging 2. 85 for the entire group. 4 For the 52 items, 5 D. P. 's are over 4-0, 21 are between
2 A minimum item mean of 2. 5 ought probably to be set for this group, since various studies have shown college students to be less prejudiced than the general population. For other groups studied in the present research, many item means were as much as a point higher.
3 While standard deviations have not been obtained for all items, it can be shown that (with group N = roo to r 50) the standard error of the difference between the means for low and high scorers is almost never above . so, seldom below . 2 5. In terms of the critical ratio, then, a D. P. of over r. o is statistically significant, that is, the means are different
though the distributions are partially overlapping.
4 While correlations between items or between each item and the total scale have not
been computed for this group, later data on similar scales suggest that the average inter- item correlation is about -4, while between each item and the sum of the remaining items the average correlation is about . 6. (See Chapter IV . )
? THE STUDY OF ANTI-SEMITIC IDEOLOGY 8r
3. 0 and 3? 9, and 15 are betWeen z. o and 2. 9. Only II D. P. 's are between r. o and 1. 9, the lowest being 1. 2.
All of the D. P. 's are therefore above a minimum standard of acceptability. The 26 items with D. P. 's of over 3. 0 are statistically very satisfactory. Why were the other items less adequate? The answer is indicated by the group means on these items. Of the II items with D. P. 's of less than z. o, IO have means below 2. 1. Conversely, almost all of the items with means of over 3. 0 have D. P. 's of over 3. 0. The mean for the low quartile is very low (below 2. o) on almost every item. The mean of the high quartile, on the other hand, varies greatly from item to item. The items with low D. P. 's were, in almost every case, statements with which the high quartile tended predominantly to disagree. This result seems to be due partly to a lack of pseudodemocratic coloring in these items, partly to their obvious illogicality or lack of truth, and partly to a lack of extreme anti-Semites among these subjects. Thus, the mean of the high quartile on all 52 items averages 4? 3 and varies from 2. 2 to 6. 6. For the IO items with the highest D. P. 's, however, the high quartile means average well over 5. 0.
The dependability of the item means and D. P. 's is indicated by a com- parison of the present group with the group of college students previously tested and reported on. The latter group responded to the entire scale (and other questionnaire material) at one sitting; the reliability was . 98 and the subscale intercorrelations were also slightly higher than in the present group. The group mean per item was 3. 0 as compared with 2. 7 for the present group, and the average D. P. was 3-4 as compared with 2. 85 here. The main differ- ence between the two groups seems to lie in the greater number of high scorers in the first group tested. The over-all mean of the low quartile was almost identical for the two groups: I. 39 then, I-42 now. But the high quartile averaged 4. 8o then as compared to 4. 27 now. In noncollege groups a larger number of high scorers, and larger D. P. 's, have been found (see pp. 76, I4o). Despite the over-all scale differences between the two groups, however, the adequacy of the individual items was very similar. Thus, the rank-order cor- relation between the D. P. 's was . 78, while the item means correlated . 92. In short, the most discriminating items for one group were also the best for the other group, and similarly for the poor items. The general conclusions about item means and discriminabilities to be drawn from the present group, then, are generally true for the previous group as well.
Table 9(III) reveals that the best items pertain to a variety of topics. Stereotypy in anti-Semitic imagery is shown in the tendency to overgen- eralize Jewish faults, and in the fact that Item 1-I 3 ("Jews are all alike") is one of the most discriminating, with a D. P. of 3. 83. The idea of Jews as a political threat (radicalism: Items 11-6, -I4) was much less prevalent than the idea of Jews as an economic threat (wealth and power: Items l-5, -8,-22,
? THE AUTHORITARIAN PERSONALITY
II-5, -II) or as a moral threat (immoralit~': Items l-1, II-I, -7, -q). Ac- cusations on religious grounds were seldom supported," as shown by the low group means and D. P. 's on Items 1-zo and Il-19. A variety of discriminatory attitudes (exclusion, restriction, suppression: most items in the "Attitude" subscale) found considerable support and \vere very discriminating.
The importance of careful formulation of items is shown by a comparison of good with poor items. The most discriminating items are usually the most pseudodemocratic ones according to criteria discussed above (Section B, I). Consider, for example, the two items dealing with intermarriage. Item 11-ro ("I can hardly imagine myself marrying a Jew") has a group mean of 4. 2 and a D. P. of 4. 3, with a high quartile mean of 6. 6 (almost all +2 and +3 responses). On the other hand, Item 1-I 5 ("It is wrong for Jews and Gentiles to intermarry") has a group mean of only 2. 6, a D. P. of 3. 0, and a high quartile mean of 4. 2.
The higher mean and discriminability of the former item are probably due to its greater indirectness and distance from crude anti- Semitism. By what criterion is this item anti-Semitic? The criterion is the fact that it correlates well with the total scale, that is, it differentiates very well between subjects who score high and subjects who score low on the total scale. (The fact that the correlation is not perfect indicates that re- sponse to any single item is not a clear-cut sign of anti-Semitism nor of its opposite; the criterion must be the total scale score).
Similar reasoning applies to items dealing with housing restrictions. The following items had very low group means (r. 5 to 2. 5): Item II-IS (con- sistent exclusion from hotels), l-3 (exclusion from neighborhoods), I-2I
(occasional exclusion from apartment houses). Item Il-7 (Jews give a neigh- borhood a "typical Jewish atmosphere"), which is more indirect and pseudo- democratic, had a higher mean (3. 2) and D. P. (4. 0). It would appear that many individuals who are not now willing actively to support anti-Semitic programs have nevertheless a negative imagery and an underlying hostility that constitute a definite potentiality for such action. Even the more open and crude items on housing had significant D. P. 's, and the high quartile means of 4? 3 and over (except on 11-IS), seem to indicate only weak re- sistance to these ideas.
The same considerations hold for the items dealing with occupations. Items which urge explicit policies of suppression and restriction of Jews (1-6, --<), II-<), -I 5) tend to have low means. But items which emphasize gen-
5 It is frequently held that Sunday School training is a major cause of anti-Semitism, which is then regarded as a form of "religious prejudice. " In this group, at least, rejection on religious grounds was infrequent. From the generality and irrationality of anti-Semitic ideology, it is clear that many diverse accusations are almost always involved, and that there are many sources for the underlying hostility which makes a given individual recep- tive to anti-Semitism. For a discussion of the role of religion in prejudice see Chapters VI and XVIII.
? THE STUDY OF A"N"TI-SEMITIC IDEOLOGY
eralized Jewish faults and? which introduce subtly discriminatory practices tend to have higher means; thus, Jewish businessmen are regarded as unfair and conniving (Item l-5), and they have too much financial power (1-8). While active suppression is not so desirable (low mean on I-<)), it is unwise for an employer to hire many Jews (higher mean on l-24, ll-3). The simplest solution-one that eliminates the need for suppression-would be for Jews to form their own nation (ll-24).
The pessimism of the high scorers regarding the solution to this problem is brought out by a number of items. On the one hand, they take the view that anti-Semitism has been entirely or primarily brought by the Jews on themselves and that any solution of the problem is a matter of Jewish re- sponsibility (Items ll-1, li-z 1). Non-Jews are simply the victims of Jewish faults; if Jews would improve, become as good as "other people," anti- Semitism would be eliminated. On the other hand, the Jews seem to be in- corrigible, and any apparent change only masks the Jewishness beneath (Items l-7, -q, ll-4, -8, -22). The contradiction is therefore complete: anti-Semitism is due to Jewish faults, but the Jews are unable to improve; the Jews should make sincere efforts to change, but their "basic Jewishness" is unchangeable. For the antidemocratic anti-Semite the only answer is open and direct suppression; for the pseudodemocrat it is subtle exclusion and "resigned tolerance" toward a bad state of affairs. The pseudodemocrats seem to betray a sense of threat and some antidemocratic potential by their doubts that democratic methods can solve the problem (Item l-12).
D. THE SHORT FORM OF THE A-S SCALE
It was a regular policy of the present research to contract the proven techniques in order to introduce new ones measuring additional trends of theoretical importance. In line with this policy, and in view of the high reli- ability and internal consistency of the original sz-item A-S scale, a short form of ten items was used in the first revision of the questionnaire.
The short form is presented in Table 1o(lll). The ten items were selected from the original fifty-two on the basis of both statistical and theoretical con- siderations. Since statistical adequacy (Discriminatory Power) was a neces- sary-but not sufficient-condition for inclusion, the new items were selected from the fifteen or twenty which had been most discriminating on the two administrations of the long form. Among these, selection was determined by the following qualitative considerations. Each item should be as rich in mean- ing as possible. There should be a minimum of duplication of meaning or con- tent among items. They should cover most of the subscales and most of the areas of accusation and discrimination. These desiderata have not been realized entirely; there were other items that seemed to merit inclusion, and for certain
? Old No.
l-24
II-I7 II-I
11-21
11-IO 11-7
1-7
1-I3 1-II
New No. ?
I I.
I6.
21.
26.
33?
40. 49?
62.
72.
TABLE IO (III)
THE TEN-ITEM A-S ScALE (Form 78)
Anyone who employs many people should be careful not to hire a large percentage of Jews.
One trouble with Jewish businessmen is that they stick to- gether and connive, so that a Gentile doesn't have a fair chance in competition.
The Jewish districts in most cities are results of the clannish- ness and stick-togetherness of Jews.
Persecution of the Jews would be largely eliminated if the Jews would make really sincere efforts to rid themselves of their harmful and offensive faults.
Jewish leaders should encourage Jews to be more incon- spicuous, to keep out of professions and activities already overcrowded with Jews and to keep out of the public notice. I can hardly imagine myself marrying a Jew.
The trouble with letting Jews into a nice neighborhood is that they gradually give it a typical Jewish atmosphere.
No matter how Americanized a Jew may seem to be, there is always something different and strange, something basically
Jewish underneath.
There may be a few exceptions, but, in general, Jews are pretty much alike.
There are too many Jews in the various federal agencies and bureaus in Washington, and they have too much control over our national policies.
THE AUTHORITARIAN PERSONALITY
? "New number" refers to the numbering of the items in Form 78.
"Old number" refers to numbering in the long form discussed previously. Slight revi-
sions will be noted in the wording of several items.
purposes they would probably be superior. The high internal consistency of the long form indicates that several statistically adequate short forms might be constructed. Nevertheless, the present form was expected to provide an adequate tool for most purposes of measurement. The slight revisions in the wording of some items were intended to make them simpler and clearer in meaning. The manner of presentation of this form was different from that previously used. Whereas previously each scale had been presented "all of a piece," on a page or pages of its own, in this and all successive forms of the questionnaire the various scales were presented interspersed with each other, so that no single scale was particularly prominent or focal, and adjacent items dealt with widely varying topics. 6
The new questionnaire, identified as Form 78 (on the basis of its having
6 The other scales in this form, to be discussed in the chapters that follow, deal with general prejudice (Negroes, other minorities, patriotism), with politico-economic liberal- ism and conservatism, and with potentially antidemocratic personality trends. There were 78 items in all. This form of the questionnaire, like all the other forms, contained in addi- tion other questions dealing with group memberships, personality, and so on.
? THE STUDY OF ANTI-SEMITIC IDEOLOGY
78 scale items), was administered in the spring of I945 to the following groups. Two of the groups comprise undergraduate students at the Univer- sity of California: the full membership, save for absentees, of the introduc- tory Public Speaking class. Here, as in al~ groups, men and women were separated in the statistical treatment and analysis was limited to native-born, white, non-Jewish Americans. The first two groups, then, are the Public Speaking Women (N = I 40) and the Public Speaking Men (N = sz). The third group comprised forty women, the entire feminine membership of an Extension Division class in Psychology at the University of California. Most subjects of this group were in their thirties and late twenties, and hence were somewhat older on the average than those of our college sample. The fourth group, Professional Women (N = 63), is actually a combination of three smaller groups: (a) Twenty-four public health nurses, the entire nursing staff of a nearby health department (the director of this department was generally liberal in his outlook and had tried to select younger nurses with more advanced ideas about public health); (b) public school teachers; and (c) social workers, woo were reached through the mails. In the latter two cases, only about 20 per cent of those appealed to sent in their question- naires, and this sampling technique was not tried again.
The reliability data for the short A-S scale are presented in Table II (III). Reliabilities of . 89-. 94 are extremely satisfactory, especially for a 10-item scale, and they are similar to those obtained on the long form. 7 The means of 3? 3 to 3? 4 for University and Extension Class students are substantially the same as the mean of 3? 55 on these ten items for the previous class taking the long form of the scale. However, the mean of 2. 6 for the Professional Women is significantly lower than the others (above the I per cent level statistically). This difference may be due partly to sampling errors; the teachers and social workers responded voluntarily by mail, and the tendency to cooperate in filling out a questionnaire dealing with prejudice and with personal feelings is probably correlated with lack of prejudice. 8 The slightly greater reliability
(. 94) of the scale for this group may reflect a greater ideological consistency in older age groups.
The Discriminatory Power method of item analysis was again carried out, and the results are presented in Table 12(III). The average D. P. of 3. 68 is very satisfactory and indicates that on most items there were very few low- quartile members who agreed, few high-quartile members who disagreed.
7 The fact that these reliabilities are similar to those obtained on the long form argues against the hypothesis that the high reliability of the latter was due to a "set" for all-nega- tive items.
8 This hypothesis is supported by questionnaire and clinical material on personality trends (opposition to "prying" and to "being analyzed" in the prejudiced subjects). Also, fewer high-scoring than low-scoring subjects in the groups tested were willing. to be interviewed.
? 86 THE AUTHORITARIAN PERSO~ALITY TABLE 11 (III~
RELIABILITY OF THE A-S SCALE (FORM 78)
Propertyc Gr. Aa Reliability . 89
Gr. Ba
093
3. 36 3. 30 3. 42
1. 48 1. 51 1. 56
1. 1-6. 3
Gr. Ca
. 90
3. 40 3. 20 3. 63
1. 36 1. 38 1. 48
1. 2-6. 1
Gr. Da
0 94
2. 57 2. 34 2. 83
1. 37 1. 27 1. 58
1. 0-6. 2
OV er-all . 92
3. 16 2. 96 3. 38
1. 41 1. 40 1. 56
1. 0-7. 0
b
Mean (total) Mean (odd half) Mean (even half)
S. D. (total) S. D. (odd. half) S. D. (even half)
Range
3. 33 2. 98 3. 66
1. 43 1. 42 1. 62
1. 0-7. 0
aThe four groups on which these data are based are: Group A, U. C. Public Speaking Class Women {N =140); Group 8, U. C. Public Speaking Class Men
{N =52); Group C, U. C. Adult Extension Clas54Wornen {N =40); Group D, Professional women {nurses, teachers, social- workers, N = 63).
bin obtaining the over-all means, the indtvidual group means were not weighted by N.
cThe values of the means, Standard Deviations, and ranges are given in terms of mean/person/item. If multiplied by 10 {the number of items), they are translated into values representing total scale score per ' peraon.
The best items deal with such varied topics as conniving businessmen, Jews being all alike, intermarriage, exclusion from neighborhoods.
How much influence did the form of presentation of the items have on their individual means and D. P. 's? Does it matter whether the items are presented in a solid block, as in the first form, or randomly dispersed through a -longer series of extremely varied items, as in Form 78? Evidence bearing on this question was obtained by comparing the results on these ten items for the two types of presentation. The mean for the Psychology Class women on these ten items (first form, excluding the remaining forty-two items) was 3? 55, as compared with 3? 32 for the Public Speaking Class women, the most
comparable group taking Form 78, and the average D. P. 's were 3. 76 and 3. 68 respectively. The differences are not statistically significant. Furthermore, the rank-order correlations between the individual item means for these two groups was . 62, while the D. P. 's correlated . 90. These correlations seem even more significant when one considers that the wording of some items was changed, and that the two groups were not systematically equated. The results? on the first form were also compared with the over-all averages for all four groups taking Form 78. The individual item means correlated .
