In fact, if the trend as discussed for Table 7 (IX) should be
representative
of the entire sample, interrater agreement for the remaining 85 per cent of the inter- viewees would be close to ?
Adorno-T-Authoritarian-Personality-Harper-Bros-1950
p6 THE AUTHORITARIAN PERSONALITY
the numerical results are not meant to yield conclusive evidence for the validity of the personality differences found between our high and low scorers. They do, however, describe in a relatively systematic, organized, and controlled way the impressions formed about these personality differ- ences in the course of intensive studies of individual cases.
This agreement between interview scoring and case studies justifies in- creased confidence in the figures presented in the next four chapters. The entire approach should be considered, however, as an initial attempt to com- bine the stu? y of variables on a group basis with the study of individuals.
2. BROAD OUTLINE OF CATEGORIES IN THE INTERVIEW SCORING MANUAL
The list of categories as well as the more formal part of the explanations accompanying each of them, which together make up the Scoring Manual, will be presented in full, together with the discussion of the specific results, in Chapters X, XI, and XII. There are sixty-two main classifications. Sub- divisions of some of them bring the total number of scoring categories used for women to ninety, the total being slightly less for men. These categories cover ( r) predominantly factual material such as childhood events or family structure, along with (2) data dealing with attitudes toward oneself, one's parents, the opposite sex, or people in general, and (3) highly interpretative dimensions exemplified by technical psychodynamic terms (such as "counter- cathectic rejection" of certain drives) or else by more "formal" character- izations (such as "rigidity," "intolerance of ambiguity," "anti-intraceptive- ness," and so forth).
The factual material and the evidence on attitudes are presented under the four headings of "Family" (Chapter X) and of "Sex," "People," and "Self" (Chapter XI).
The dynamic and formal categories are especially emphasized in a fifth and concluding part of the scoring scheme, under the heading of "Character Structure and Personality Organization" (Chapter XII). Although these categories were to a considerable extent inspired by psychoanalysis, they should not be considered as psychoanalytic in the narrower sense of the word, since classification of our material is done primarily on the basis of present personality structure rather than on the basis of psychogenetic data. The entire framework, length, and condition of our interviews made it impossible directly to obtain material of a depth-level comparable to that of genuine psychoanalytic material. At the same time, however, there was enough spon- taneous material at hand to make it possible for raters trained in dynamic psychology to infer some of the major structural problems and types of defense mechanisms in our subjects, in accordance with the categories pro- vided by the Scoring Manual.
A certain parallelism, although by no means an exact duplication, between
? INTERVIEWS AS APPROACH TO PREJUDICED PERSONALITY 327
the Interview Schedule and the Scoring Manual will be noted when com- paring the two in detail.
3. THE INTERVIEW RATING PROCEDURE AND THE RATERS
Since our major purpose in analyzing the interview material was to gain additional evidence concerning the relationship of prejudice to personality, it was important to conceal from the rater the explicit stand of the subject with respect to ethnic tolerance or prejudice as well as with respect to polit- ical ideology in general. All references to these topics throughout the inter- view were thus carefully deleted before the protocols were handed to the raters. The diagnosis of the subjects' personality was thus rendered "blind. " The raters did know, of course, that their subjects had scored either high or low on the scales for measuring prejudice, but they did not know which were the high and which the low scorers.
In all other respects the interview protocols remained unchanged.
a. RA TING BY CA TEGORIES. The rating of the interviews was done for each of the categories separately. The score for the category in question, how- ever, was obtained in a synoptic rather than a piecemeal fashion. The major source for the assignment of a score was the clinical part of the interview, but evidence was utilized from any part of the interview which might be brought to bear on each category.
Scores were in terms of a three-point scale. Since, as was mentioned above, a careful study of the interviews had preceded the construction of the Scor- ing Manual, certain more or less definite expectations as to what might constitute the personality aspects of a prejudiced as contrasted with an un- prejudiced subject had been developed in the way of advance hypotheses. In view of this, the two opposites within each category were tentatively designated as the presumably "High" and "Low"2 variants or alternatives, i. e. , those expected to be typical of prejudiced and of unprejudiced persons, respectively. In the Manual the left column was used for the presumably high and the right column for the presumably low variant.
The third rating, "Neutral," comprised two distinct possibilities: (I) the existing evidence was too colorless or self-contradictory within the category in question to warrant assignment to either the "High" or the "Low" alterna- tive; or (2) there was no evidence at all pertinent to this category. The for- mer case is much less frequent among the "Neutral" ratings actually given than one might expect beforehand. In some of the protocols possibilities (I) and (2) were scored separately; their proportion was found to be about three to seven. In particular, factual questions were somewhat more likely to yield
2 Note that the initial letter of the terms "High," "Low," (and "Neutral") is capitalized when referring to interview ratings in order to distinguish these ratings from the actual "high" or "low" scores of the subjects on prejudice and on the other scales of the question- naire.
? p8 THE AUTHORITARIAN PERSONALITY
"no evidence" than interpretative categories, but even in the case of the group of categories subsumed under character structure the number of cases in which the material was too ambiguous to make a decision possible remained within comparatively low limits.
In absolute terms the number of "Neutrals" is considerable, especially in the case of some of the categories dealing with childhood events (see Chap- ters X to XII). For many variables lack of information is~by no means always due to the impossibility of gathering evidence but rather to the sportiness of either the spontaneous responses of the subject or of the inquiry on the part of the interviewer who, as stated above, could not possibly cover the entire ground in each case. Both the Interview Schedule and the Scoring Manual make an attempt to cover systematically as many as possible of the very numerous areas, but it could not be hoped that each case would furnish material on all of the questions involved.
b. INTUITIVE OvER-ALL RATINGs. Besides the ratings on each of the cate- gories, the raters were asked also to make intuitive over-all ratings. They were instructed, that is, to give their conclusive impression as to whether the subject involved was prejudiced or not. One of two alternatives, "High"
(H) or "Low" (L), had to be chosen (for data see final column of Tables I (IX) and 2 (IX); discussion in Chapter XIII).
c. THE INTERVIEW RATERS. The ratings were made by two members of the staff of the study, here to be called M and R, one of them male and the other female. Both are well-trained psychologists and were thoroughly familiar with the nature of the categories and the underlying implications as to personality theory. These raters had participated actively in numerous conferences at which the scoring procedure was thoroughly discussed, prior to making the ratings.
Each of the raters scored approximately half of the men and half of the women, high and low scorers being distributed at random in about equal proportions within each group. (Concerning added duplicate ratings to check on reliability, see below. )
In particular, the interviews of cases Jl,fz to M2o and F22 to F39a among those scoring high, and of cases M2 to Mzg and of F29 to F39 among the low-scoring interviewees were evaluated by rater M, and those of the re- maining interviewees, listed farther down in the respective subdivisions of Tables 1 (IX) and 2 (IX), were evaluated by rater R. It should be added that the code numbers used were distributed at random among the various groups so that each rater rated not only men and women, low scorers and high scorers, but also approximately equal proportions of subjects who had been given Forms 78 and 45 of the questionnaire (see also Table 3 (IX)).
4. RELIABILITY OF THE INTERVIEW RATINGS
There were three ways in which some light was thrown upon the difficult question of the reliability of the interview ratings, although only the second
? INTERVIEWS AS APPROACH TO PREJUDICED PERSONALITY 329
of these deals directly with reliability in the technical sense of the term. The other two refer to aspects which are merely more or less closely related to this problem.
a. PROPORTIONs OF RATINGs GIVEN. First, the proportion of High, Low, and Neutral ratings within each category was compared for the two raters and found to be in fairly good agreement. One method used in computing an index for this agreement was the following: the frequencies of "High" responses, as scored by the first of the raters on each of the variables was plotted on a scatter diagram against the frequencies of such responses as scored by the second rater. In this comparison, the two raters are repre- sented by the different nonoverlapping groups they were assigned to rate. Similar scattergrams were obtained for "Low" and "Neutral" ratings, and in each of the three cases men and women were plotted separately. With one exception, the correlation coefficients computed from the six scatter- grams were between . 70 and . 82. This indicates that the two raters tended to concur fairly well in giving either a relatively large or a relatively small number of "High," or of "Low," or of "Neutral" ratings within any of the approximately ninety categories, showing a certain uniformity at least for one aspect of the rating policy.
(It may be added that in absolute terms there is also good agreement, the range of frequencies of "High" scores being o to I2, and o to I4, for the two raters, respectively, the various categories being considered for each of the sexes separately. For "Low" scores the corresponding ranges were o to I I and o to I 3? Thus, while both the raters neglect to use some of the alter- natives offered by the Manual (as indicated by "o"), neither of them uses the opposite alternatives indiscriminately, i. e. , in the characterization of all or nearly all the interviews analyzed by them. )
In view of the fact that the two staff members rated different samples of subjects, the coefficients and other data given above suggest that the char- acter and distribution of ratings given for the various categories are to a considerable degree intrinsic to the category in question, at least within our specific combination of raters.
b. INTERRATER AGREEMENT. Secondly, we turn to reliability proper. Nine interviewees in the group assigned to rater M were, in an additional checking procedure not used for the main analysis or for the survey in Tables I (IX) and 2 (IX), also rated by rater R under the identical detailed set of aspects. Two of these nine subjects were deliberately chosen from the relatively small group-12 of the So interviewees, i. e. , I 5 per cent-of those for whom the composite standing based on the detailed ratings of the original rater had missed the correct diagnosis as to prejudice. These subjects were M19 and F39. As may be further seen from Tables I (IX) and 2 (IX), the remaining seven are likewise mostly not from among the clearest cases as far as inter- view ratings are concerned.
A comparison of t:he gross results of the two rating procedures is shown
? Interviewees
Prejudice score (from
TABLE 7 (IX)
RELIABILITY OF INI'ERVIEW RATINGS: INTERRATER AGREEMENT ON NINE SUBJEX:TS
Percentage of High ratings on interview
Rater M Rater R
Intuitive rating on interview Rater M Rater R
LL H H LL HH H ? HH LL HH HL
questionnaire). M3low19. 236. 1LL
M4 high 77. 3 78. 1 H H M5low31. 431. 7LL M6 high 72. 7 74. 4 H H Ml9 low 70. 4 31. 1 H L F24 high 70. 5 70. 0 H H F29 low 33. 3 34. 4 L L F31 high 79. 4 77. 7 H H F39 low 72. 2 27. 2 H L
Composite standing on interview Rater M Rater R
? INTERVIEWS AS APPROACH TO PREJUDICED PERSONALITY 33 I
in Table 7 (IX). The prejudice scores based on the questionnaire are taken from Tables I (IX) and 2 (IX). However, the numerical scores appearing on these tables for the subjects listed here, as based on the ratings of rater M, are limited to a selection of the more discriminating categories (see below, Section 6). The figures in Table 7 (IX) are based upon the ratings on all categories and are further given as percentages of "High" ratings relative to the possible maximum of ratings as given by the total number of cate- gories. Thereby the number of Neutral ratings-easily obtained by subtract- ing both "High" and "Low" ratings from the total number of categories, 86 for men and 90 for women (see below)-has been added half and half to the "High" and the "Low" ratings. Composite standing as indicated by a percentage score of over, or of under, so, and finally intuitive over-all ratings of the interview make up the remainder of the table.
Percentage scores show excellent interrater agreement for six of the nine subjects. Of the remaining three, those with really striking discrepancies are the same two mentioned above as having been misjudged by the original rater, M, namely M19 and F39. In both cases, the second rater has rectified the error very clearly by establishing percentage scores in the neighborhood of 30 which contrast sharply with those in the neighborhood of 70 as obtained from the first rater. The correlation coefficient between the columns repre- senting the two raters-not very meaningful under the circumstances-is . 6I. It would be raised to about . 8 if one of the two "controversial" cases just mentioned were eliminated so as to adjust the proportion of such cases more closely to that referred to above as existing in the total sample of inter- viewees, namely Is per cent. Such a figure, if verified on a broader basis, would be quite satisfactory for the kind of material involved.
In terms of composite standing and intuitive ratings, agreement is perfect save for the two cases mentioned. (Intuitive ratings on one of the controversial subjects, M 19, is incomplete due to the fact that rater R, contrary to instruc- tions, declared herself as unable to make up her mind in this particular case. )
It may be added that the means of the percentage scores for the two raters are quite close to one another as well as to the ideal value of so. They are s8. 4 for M and so. 9 for R. This augments the evidence brought forward above under (a) with respect to the proportion of ratings given by the two raters. The slight preponderance of "High" ratings in rater M is also reflected in his intuitive over-all ratings. In fact, it is concentrated in the two cases where he makes his mistakes and where the second rater evens out the score.
A breakdown for the six major areas covered by the Scoring Manual, namely family patterns (see Chapter X), attitude toward sex, other people, and self (see Chapter XI), and dynamic character structure and cognitive personality organization (see Chapter XII) is given in Table 8 (IX). The number of categories for each area is also indicated. Considering the small- ness of these numbers, pairs of averages from raw scores in terms of number
? Areas in Scoring Manual
Family pattern (parents, etc) A ttitude toward sex
Number
of four high scorersc
categories Rater M Rater R
five low scorersc Rater M Rater R
A ttitude toward other
Attitude toward self
Dynamic character structure Cognitive personality organization
Totals
H-L 28 6. 8-1. 3
7 4. 5-0. 3 11 6. 3-0. 0 16 9. 3-1. 3 22 15. 3-0. 3
6 5. 0-0. 0
goa 47. 2-3. 2 49. 9-4. 4
H-L
3. 8-5. 6 1. 8-2. 2 2. 2-3. 2 3. 0-6. 6 5. 8-6. 2 2. 0-2. 8
TABLE 8 (IX)
INl'ERRATER AGREEMENl' ON INl'ERVIEW RATiroS FOR SIX MAJOR AREAS
people
aFor men the total is only 86; no adjustment to this sl~ght difference has been made in the present table in the case of the men subjects.
bRounded to one decimal place csee Table 7 (IX)
Average number of rating. sb received by
H-L
9. 3-1. 3 5. 3-0. 3 6. 3-0. 3 9. 5-l. 5
H-L
3. 0-8. 6 1. 0-3. 4 0. 4-5. 0 1. 0-9. 8 1. 6-8. 8 0. 4-5. 0
14. 5-0. 5 5. 0-0. 5
18. 6-26. 6 7. 4-40. 6
? INTERVIEWS AS APPROACH TO PREJUDICED PERSONALITY 333
of "High" and "Low" ratings (H-L) are given for the four high scorers as contrasted with the five low scorers listed in Table 7 (IX). No indices of reliability were computed here; but comparison of the first with the second and of the third with the fourth pair of figures in each row of Table 8 (IX) reveals a good deal of agreement between the two raters. The fact that this agreement is less pronounced in the case of the low scorers as shown in the columns containing the third and fourth pairs of figures, and that, further- more, the values of H and L within these pairs ofteri show less clear-cut differentiation than they do in the left part of the table, is due-as was revealed in more detailed analysis not presented here-to the fact that both cases with controversial diagnosis, M 19 and F39, happen to be in this group. Perhaps with the exception of family pattern and attitude toward self, this lack of differentiation for the group of low scorers, especially in rater M, and the ensuing disagreement with rater R, is about evenly distributed over the various areas; for high scorers differentiation and agreement is about equally good for all the areas, and the "totals" are in excellent agreement with one another.
Discounting the controversial cases, i. e. , the 15 per cent in our total sample of interviewees for which the original rater arrived at a diagnosis op- posite to that given by the defining score on the prejudice scales, the results of this fragmentary analysis of reliability are quite encouraging.
In fact, if the trend as discussed for Table 7 (IX) should be representative of the entire sample, interrater agreement for the remaining 85 per cent of the inter- viewees would be close to ? 9? For the other 15 per cent one may contemplate challenging the validity of the defining prejudice score along with doubting the validity of the interview rating. The "questionnaire-high" may after all be considered the product of an approach that is by definition less con- cerned with underlying dynamics than is the diagnosis of the "personality- high. "
Further data on interrater agreement on the interview will be presented in Chapter XIII.
The problem of agreement of various types of ratings among themselves, such as in our present context especially of composite standing and intuitive over-all rating, will be discussed in Chapter XIII. In a broader sense such aspects are also included within the general concept of reliability.
A third avenue of scrutiny somewhat akin to reliability problems is through the study of "halo-effects," to be discussed next.
5. MINIMIZING HALO-EFFECTS IN RATING THE INTERVIEWS
We return now to the problem of the carry-over from one category to another, much in the way of the "halo-effect" known in social and educa- tional psychology. One way of preventing or minimizing the halo-effect would have been to use designations other than "(presumably) High" and
? 334
THE AUTHORITARIAN PERSONALITY
"(presumably) Low" to characterize the two opposites within the various categories. But such a procedure would have prevented only raters un- familiar with the underlying hypotheses from succumbing to halo-effects. Such raters, on the other hand, would have been undesirable from other, more crucial points of view. It was thus decided to leave control of halo- effects to special analytical attitudes the raters were asked to maintain, and to ascertain the degree of relatedness in a statistical analysis of the completed ratings. It must be noted that-as in all cases of halo-effects-a certain amount of correlation may be fully justified by fact, i. e. , by existing correlation of real traits. Exactly how much of the halo is realistic would require further intensive study for all combinations of categories involved.
Both the variability of the discriminatory power of the single categories (see below) and the variability of the proportion of "High" and "Low" rat- ings ascribed to the various subjects seem to indicate that the raters succeeded, at least in part, in keeping the halo-effect within reasonable bounds in rat- ing the subjects. Evidence on the second of these points is contained in Tables I (IX) and 2 (IX) in the first two of the columns relating to ratings of the interviews. As is readily seen, hardly any of the subjects are classified as "High" or as "Low" on all categories even when the "Neutral" ratings are excluded from consideration, and for some of them there is an approximately equal number of "High" and "Low" ratings. This shows that the raters were quite capable of separating the various issues involved, and of rating a person
as "High" in one respect and as "Low" in another.
6. TABULATION OF INTERVIEW RATINGS BY CATEGORIES: STATISTICAL SIGNIFICANCE
A tabulation of interview ratings by single categories was obtained simply by counting the instances of High (H), of Low (L), and of Neutral (N) ratings, on a given category, among subjects of each of four groups-high- scoring men, low-scoring men, high-scoring women, and low-scoring women (scoring, that is, extremely high or low on the direct prejudice ques- tionnaire).
Analysis of the figures for a given category, with a view to their bearing upon the underlying hypotheses, could have any one of four outcomes-two "positive" and two "negative. " The "positive" instances are confirmatory of the original hypothesis. They include "High" ratings-i. e. , those designat- ing reactions presumed to occur more frequently in the prejudiced person- when given to subjects scoring high on the prejudice scales, as well as "Low" ratings given to relatively unprejudiced, i. e. , our so-called "low-scoring," subjects, in short, the hH and lL combinations. The remaining two figures, indicating the frequency of prejudiced subjects receiving a "Low" and of un- prejudiced subjects receiving a "High" rating, in short, the? hL and the lH combinations, constitute the "negative," nonconfirmatory instances.
? INTERVIEWS AS APPROACH TO PREJUDICED PERSONALITY 33S
The series of tables in Chapters X to XII (Tables I (X) to 2 (XII)) pre- sents, for the two sexes separately, the numbers of prejudiced and of un- prejudiced interviewees who give the presumably "High" and the pre- sumably "Low" responses in the interviews, for each of the scoring categories. The number of "Neutrals" may be obtained by subtraction of their sums from the total number of subjects in the respective prejudiced or unprejudiced groups. The four raw figures are followed by two sums which contrast the total number of positive with the total number of negative in- stances. All "positive," i. e. , confirmatory, evidence is italicized. Whenever a category is defined in the Scoring Manual by only one variant, H or L, rather than by an opposition of an H with an L, only the presence of the trait in question is registered in the tables and the remaining space is left blank.
The final column of the tables indicates the level of statistical significance of the difference, on the category in question, between prejudiced and unprej- udiced extremes as defined in the present study. It thus refers to the "dis- criminatory power" and the importance of the category as a personality correlate-and therefore as a potential measure-of overtly expressed ethnic prejudice. Indications of significance are given in terms of whether or not the sper cent (satisfactory), the 2 per cent, or the I per cent level (highly satisfactory, since in this latter case there is a 99 per cent probability that the obtained difference is not due to chance factors) has been reached or surpassed, without specifying whether or by how much they have been surpassed. Significance was computed after evenly dividing the "Neutrals" among the "High" and the "Low" ratings. It must not be forgotten, in this connection, that dividing the Neutral ratings tends to lower the index of significance, the more so the larger the proportion of these ratings. Thus, in treating the Neutral ratings as we do, we are keeping on the safe side, since, as was pointed out above, the Neutral scorings are based, in a con- siderable proportion of the cases, on lack of information rather than on lack of actual discriminability.
Wherever the proportion of Neutrals for the total sample of interviewees is larger than so per cent, the statistical significance of the category in question was not computed and therefore there was no entry under any circumstances in the last column of the tables. The category was also omitted from the survey of interview scores in Tables I (IX) and 2 (IX). For men the total number of categories is 86, somewhat less than for women, in whose case a few more subdivisions were introduced into the Scoring Manual. Of these 86 categories, 72 yielded less than so per cent Neutrals, and thus re- mained for full treatment. For women only 6s of the original 90 categories yielded less than soper cent Neutrals and were thus retained for full treat- ment. The fact that the categories were generally somewhat more dis- criminatory in the case of the men than in the case of the women may be
? THE AUTHORITARIAN PERSONALITY
accounted for by the fact that most of the men were interviewed at a later stage of the study and that therefore their records were more complete.
All the calculations discussed above were performed separately for men and women. Corresponding figures for each appear closely adjacent to one another in the tables, those for men (M) in the upper left-hand and those for women (W) in the lower right-hand area of each of the "cells" that one may imagine at the intersection of vertical columns with horizontal rows, the latter defined by the various categories.
? CHAPTER X
P ARENTS AND CHILDHOOD AS SEEN THROUGH THE INTERVIEWS Else Frenkel-Brunswik
A. INTRODUCTION
In turning to the specific results of the interviews we begin with the organization of the family. Many of the attitudes and underlying needs dis- cussed in this volume must be assumed to originate, as far as the individual is concerned, in the family situation. Here the growing child learns for the first time to handle interpersonal relations. Some of the members of the family are in an authoritative, others in an equalitarian or in a weaker position than himself. Some are of the same, others are of the opposite sex. It soon becomes evident to the youngster what kind of behavior is considered appropriate and will lead to reward and what kind of behavior will be punished. He finds himself confronted with a certain set of values and certain expectations which he has to meet.
Within the general common framework of the white American popula- tion, families vary greatly as to the rigidity or flexibility of the roles defined within the family, as well as to values in general. We shall encounter families in which considerably more emphasis is placed on obedience than it is in others. In some cases discipline is harsh and threatening, in others intelligible and mild. Or there may be rigorous adherence to conventional rules and customs rather than to more flexible and more intrinsic values which lead to greater tolerance for individual variations. Or smooth functioning within the family may depend either more on exchange of well-defined obligations and "goods," or else on an exchange of genuine affection. These and other differences in the organization of the family are under scrutiny in the light of their possible implications with respect to the personality structure of the individual and his social and political beliefs.
Although no striking relations between these patterns and gross economic factors have been uncovered in the present study, systematic investigations of a more distinctly sociological nature would undoubtedly reveal broader
337
? THE AUTHORITARIAN PERSONALITY
cultural and subcultural determinants of these differences, say, the greater frequency of one or the other type of family organization in different national subgroups, or a dependence on the relative stability or instability of the socioeconomic family history. ?
While this goes to press, data from a separate project (Frenkel-Bruns- wik, for an advance report see 30)1 seem to indicate that parents of ex- tremely prejudiced children are relatively often the children of foreign-born parents or show preoccupation as to social and national insecurity.
Specific rating categories from the Interview Scoring Manual in the area of family organization, followed by tabulation of quantitative results and eventually by a discussion and the presentation of pertinent quotations from the interviews themselves, are given in Sections B to D.
B. A TTITUDES TOW ARD P ARENTS AND CONCEPTION OF THE FAMILY
1. DEFINITION OF RA TING CA TEGORIES AND QUANTITATIVE RESULTS
In line with the over-all subject matter of the present volume, the discus- sion concerning attitudes toward parents will, among others, center about the following questions: Is the general tendency toward glorification and lack of critical evaluation of ingroups on the part of the ethnically prejudiced also mirrored in . their attitudes toward their parents? Is there a tendency toward submission, and how are the problems of rebellion, hostility, and guilt handled in our two opposite groups? How are the feelings of genuine love related to conformity as contrasted with independence?
The definitions of the specific categories of the Scoring Manual dealing with problems of this nature are listed here in the form in which they were used by the raters. It should be remembered, however, that the Manual represents merely a summary of what was developed, and discussed with the two raters, in extended conferences preceding the actual rating procedure. The subsequent text makes occasional use of the more prominent of the con- notations thus established which were not formally incorporated in the Manual in order not to overload it in actual use.
Before starting the presentation of the various parts of the Interview Scoring Manual, attention must be called to the fact that strict opposi-
1 The study on social discrimination in children referred to here is being carried out at the Institute of Child Welfare of the University of California. The project was initiated by the present writer, in cooperation with Harold E. Jones and T. \V. Adorno, and spon- sorship was at first by the Scientific Department of the American Jewish Committee. In developing the tests and experiments the present writer was aided primarily by Claire Brednor, Donald T. Campbell, Joan Havel, Murray E. Jarvik, and Milton Rokeach.
? PARENTS AND CHILDHOOD SEEN THROUGH INTERVIEWS 339
tion or near-opposition of the presumably "High" and the presumably "Low" variants is to be assumed only where the numbers or number-letter combinations appearing on the right side of the page are identical with those on the left. Thus Category 2, "victimization," stands in a somewhat oblique relation to Categories za to zc. In some cases pairings of this kind reflect the fact that there is more than one opposite to a given variant. Beginning with Section C of this chapter, different sets of letters are sometimes used on the two sides to stress an absence of one-by-one correspondence of a series of alternatives listed on the right with a series of items on the left, although the lists in their entirety define a more clear-cut pattern of oppo- sition. (Concerning the lopsided evaluation of asymmetrical categories in the tables to follow, see the concluding pages of the preceding chapter. )
The first subdivision of the Interview Scoring Manual follows. As in the Interview Schedule, italics are used to represent key phrases which had been emphasized to the raters by underscoring.
INTERVIEW SCORING MANUAL: ATTITUDES TOWARD PARENTS AND CONCEPTION OF FAMILY
(to Table r(X)).
I .
2.
PRESUMABLY "HIGH" VARIANTS
Conventional idealization of parent(s): Overestimation of qualities and status, expressed in behavioral (essentially exter- nal), conventionalized general- ities, or undifferentiated "all's
well" attitude
Victimization (quasi-persecu- tory) byparent(s): Neglect, in- cluding failure to give proper discipline, unjust discipline;
"picked on"; unfair: resents pre- ferring of rival sib or spouse (or foster-sib or step-spouse); etc.
PREsUMABLY "Low" VARIANTs I. Objective appraisal of parents
2a. Principled open rejection
2b. Genuine positive affect: some reference to (positive) psycho- logical qualities; individualized
characterizations
2c. Blocked affect (Presumably
mutually exclusive with 2a) 3? Principled independence
4? Love-seeking succorance-nur- turance-affiliation toward par- ents
3a. Submission to parental author- ity and values: respect based on fear
3b. Capricious rebellion against parents; delinquency
4a. Ego -alien dependence- for - things and suppon on parents: essentially exploitive-manipula- tive-"getting"; an externalized relationship
4b. Sense of obligation-and-duty to
? THE AUTHORITARIAN PERSONALITY
parents: Desire to "make it up
to them"
5? lngroup orientation to family as 5? Individualized approach to
a whole; e. g. , emphasis on fam- members of the family ily heredity and "background";
homogeneous-totalitarian fam-
ily vs. rest of world; aristocratic
superiority of family, etc.
In Table I (X), the results for both men and women interviewees are presented in the manner described in the concluding section of the preceding chapter. Abbreviated formulations of the categories just listed are presented, along with category numbers, for purposes of easier identification.
2. IDEALIZA TION VS. OBJECTIVE APPRAISAL OF P ARENTS
In view of their general tendency toward conventionality and submission toward ingroup members, it is not surprising to find in the prejudiced sub- jects a tendency toward "idealization of the parents. " This idealization is expressed characteristically in generalized and undifferentiated, convention- alized terms which primarily glorify external features of physical appear- ance or overt behavioral conduct rather than involving the more internal aspects of their personality. "Objective appraisal" of parents, referring to an ability for critical evaluation of the parents in specific and psychologically conceived terms, on the other hand, is predominant in the unprejudiced subjects.
The detailed results presented in Table I(X) (under Category I) reveal the striking fact that none of the low-scoring women interviewed shows the glorification of the parents just described; instead, I I of them show objective, critical appraisal. Of the high-scoring women, 9 show glorification and 6 ob- jective appraisal of parents. For the total group of women, there are 20 "positive" as contrasted with 6 "negative" instances in the sense defined at the end of the preceding chapter. The statistical significance of the differ- ence between the positive evidence (i. e. , that confirming the original hypoth- esis underlying the distinction between the "High" and the "Low" variant) and the negative (nonconfirmatory) evidence cited is found to be at the " I per cent level," and thus highly satisfactory (see Chapter IX, Section F, 6).
A good illustration of the "High" attitude in women is given by the fol- lowing quotation from the interview of one of the high scorers: "Mother- she amazes me-millions of activities-had two maids in years ago, but never since-such calmness-never sick, never-beautiful woman she really is. " The reference to external dimensions, both behavioral ("million ac- tivities") and physical ("never sick, beautiful"), can be seen clearly in the foregoing record. It must be emphasized that the subjects were asked, in this connection, "What kind of person is your father (mother)? " without further
? 1. Conventional idealization(H) vs. objective appraisal(L) of parents Women
. H.
9 6 0 ! ! 20 6
2. Victimization by parents (H) vs. a. Principled open rejectton of
parents (L}
b. c. Genuine affection or blocked Men
affect for parents(L)
aa. Submission to parents (H) vs. principled independence(L)
women 3 10 10 3 Men 14 2 2 10 24 4
women J1. . 0 1 8 17 1 Men 9 4 9 4
women6 1 61
1 2
1 5
3b. Capricious rebellion(H)
4a. Dependence for things on parents Men 13
1 5
1 2
1 1
2
14 27 2 8 2f 7
(H) vs. love-seeking affiliation(L)
4b. Sense of obligation and duty(H) 5. Ingroup conception of family(H)
Women T3 Men 5
women1
--
!
20 men and 25 women "high scorers"
20 men and 15 women "low scorers"
Men
1
l
Men women
_? _ 4 8 4
. . ? . 4
8 4 -
-
TABLE 1 (X)
INTERVIEW RATINGS ON ATTITUDE TOWARD PARENTS AND CONCEPT OF FAMILY
FOO 80 SUBJECTS SCORING EXTREMELY "HIGH" OR "LOW" ON THE ETHNIC PREJUDICE QUESTIONNAmE SCALE
Interview rating categories (abbreviated from Manual)
s~x I
Number of "High"(H) and "Low"(L) ratings received by
Sums of instances Level of statistical
? positive" "negative" significance reached
L H L
11 1 2 13 24 3
(percentage)
. . .
Men
Wo111en 2 ~ 3 2
Men 6
Women7 1
i1 11
62 71
1 A . A. 1 4 12 12 4
? THE AUTHORITARIAN PERSONALITY
specification (see the Interview Schedule as presented in the preceding chap- ter).
The overestimation of parents in more general terms is especially clear in the record of another prejudiced woman interviewee (F24): "Father-he is wonderful; couldn't make him better. He is always willing to do anything for you. He is about years old, six feet tall, has dark brown hair, slim build, young-looking face, good-lookirig, dark green eyes. " The same sub- ject gives further evidence of the stereotypical conceptoion of parents in the high scorers by repeating the same description for her mother, differentiat- ing only the physical characteristics. After repeating the first two sentences she continues: "She is about 5 feet 5 inches, neither slim nor stocky. She's reducing. She has dark hair, blue eyes, is nice looking. She is years old. " These illustrations are quite typical of the responses of our high-scoring subjects.
It should be added that none of the 6 high-scoring women who show "ob- jective appraisal" of parents was considered to manifest "positive affect" toward the parents (Category 2). This will be discussed below.
Very characteristic of high-scoring subjects is an initial statement of great admiration for parents, followed by some criticism which is not, how- ever, recognized as such by the subject. The comparison of reactions to general questions with reactions to specific questions proved especially fruitful, e. g. , when parents were characterized in general positive terms but the specific episodes and traits referred to were mostly of a negative charac- ter. (Regarding related mechanisms of self-deception, see 33, 34? ) The sub- ject is aware only of admiration toward the parents, and the reservations seem to enter the picture against "better" intention and knowledge, thus in- jecting into the statement an element of ambivalence.
Examples of this attitude from the protocols of high-scoring women are quoted in the following. It should be mentioned here that throughout Chapters X to XII the special code numbers used in rating the interviews are retained for greater anonymity; for the same reason, localities, occupa- tions and related personal data are either left out or disguised.
F31: (Father? ) "He has a marvelous personality and gets along well with people. He has a hot temper. "
Or F79: "Mother was, of course, a very wonderful person. She was very nervous. Irritable only when overdoing. "
Or again, Fu: "Father is quiet and calm. He never shows irritation.
the numerical results are not meant to yield conclusive evidence for the validity of the personality differences found between our high and low scorers. They do, however, describe in a relatively systematic, organized, and controlled way the impressions formed about these personality differ- ences in the course of intensive studies of individual cases.
This agreement between interview scoring and case studies justifies in- creased confidence in the figures presented in the next four chapters. The entire approach should be considered, however, as an initial attempt to com- bine the stu? y of variables on a group basis with the study of individuals.
2. BROAD OUTLINE OF CATEGORIES IN THE INTERVIEW SCORING MANUAL
The list of categories as well as the more formal part of the explanations accompanying each of them, which together make up the Scoring Manual, will be presented in full, together with the discussion of the specific results, in Chapters X, XI, and XII. There are sixty-two main classifications. Sub- divisions of some of them bring the total number of scoring categories used for women to ninety, the total being slightly less for men. These categories cover ( r) predominantly factual material such as childhood events or family structure, along with (2) data dealing with attitudes toward oneself, one's parents, the opposite sex, or people in general, and (3) highly interpretative dimensions exemplified by technical psychodynamic terms (such as "counter- cathectic rejection" of certain drives) or else by more "formal" character- izations (such as "rigidity," "intolerance of ambiguity," "anti-intraceptive- ness," and so forth).
The factual material and the evidence on attitudes are presented under the four headings of "Family" (Chapter X) and of "Sex," "People," and "Self" (Chapter XI).
The dynamic and formal categories are especially emphasized in a fifth and concluding part of the scoring scheme, under the heading of "Character Structure and Personality Organization" (Chapter XII). Although these categories were to a considerable extent inspired by psychoanalysis, they should not be considered as psychoanalytic in the narrower sense of the word, since classification of our material is done primarily on the basis of present personality structure rather than on the basis of psychogenetic data. The entire framework, length, and condition of our interviews made it impossible directly to obtain material of a depth-level comparable to that of genuine psychoanalytic material. At the same time, however, there was enough spon- taneous material at hand to make it possible for raters trained in dynamic psychology to infer some of the major structural problems and types of defense mechanisms in our subjects, in accordance with the categories pro- vided by the Scoring Manual.
A certain parallelism, although by no means an exact duplication, between
? INTERVIEWS AS APPROACH TO PREJUDICED PERSONALITY 327
the Interview Schedule and the Scoring Manual will be noted when com- paring the two in detail.
3. THE INTERVIEW RATING PROCEDURE AND THE RATERS
Since our major purpose in analyzing the interview material was to gain additional evidence concerning the relationship of prejudice to personality, it was important to conceal from the rater the explicit stand of the subject with respect to ethnic tolerance or prejudice as well as with respect to polit- ical ideology in general. All references to these topics throughout the inter- view were thus carefully deleted before the protocols were handed to the raters. The diagnosis of the subjects' personality was thus rendered "blind. " The raters did know, of course, that their subjects had scored either high or low on the scales for measuring prejudice, but they did not know which were the high and which the low scorers.
In all other respects the interview protocols remained unchanged.
a. RA TING BY CA TEGORIES. The rating of the interviews was done for each of the categories separately. The score for the category in question, how- ever, was obtained in a synoptic rather than a piecemeal fashion. The major source for the assignment of a score was the clinical part of the interview, but evidence was utilized from any part of the interview which might be brought to bear on each category.
Scores were in terms of a three-point scale. Since, as was mentioned above, a careful study of the interviews had preceded the construction of the Scor- ing Manual, certain more or less definite expectations as to what might constitute the personality aspects of a prejudiced as contrasted with an un- prejudiced subject had been developed in the way of advance hypotheses. In view of this, the two opposites within each category were tentatively designated as the presumably "High" and "Low"2 variants or alternatives, i. e. , those expected to be typical of prejudiced and of unprejudiced persons, respectively. In the Manual the left column was used for the presumably high and the right column for the presumably low variant.
The third rating, "Neutral," comprised two distinct possibilities: (I) the existing evidence was too colorless or self-contradictory within the category in question to warrant assignment to either the "High" or the "Low" alterna- tive; or (2) there was no evidence at all pertinent to this category. The for- mer case is much less frequent among the "Neutral" ratings actually given than one might expect beforehand. In some of the protocols possibilities (I) and (2) were scored separately; their proportion was found to be about three to seven. In particular, factual questions were somewhat more likely to yield
2 Note that the initial letter of the terms "High," "Low," (and "Neutral") is capitalized when referring to interview ratings in order to distinguish these ratings from the actual "high" or "low" scores of the subjects on prejudice and on the other scales of the question- naire.
? p8 THE AUTHORITARIAN PERSONALITY
"no evidence" than interpretative categories, but even in the case of the group of categories subsumed under character structure the number of cases in which the material was too ambiguous to make a decision possible remained within comparatively low limits.
In absolute terms the number of "Neutrals" is considerable, especially in the case of some of the categories dealing with childhood events (see Chap- ters X to XII). For many variables lack of information is~by no means always due to the impossibility of gathering evidence but rather to the sportiness of either the spontaneous responses of the subject or of the inquiry on the part of the interviewer who, as stated above, could not possibly cover the entire ground in each case. Both the Interview Schedule and the Scoring Manual make an attempt to cover systematically as many as possible of the very numerous areas, but it could not be hoped that each case would furnish material on all of the questions involved.
b. INTUITIVE OvER-ALL RATINGs. Besides the ratings on each of the cate- gories, the raters were asked also to make intuitive over-all ratings. They were instructed, that is, to give their conclusive impression as to whether the subject involved was prejudiced or not. One of two alternatives, "High"
(H) or "Low" (L), had to be chosen (for data see final column of Tables I (IX) and 2 (IX); discussion in Chapter XIII).
c. THE INTERVIEW RATERS. The ratings were made by two members of the staff of the study, here to be called M and R, one of them male and the other female. Both are well-trained psychologists and were thoroughly familiar with the nature of the categories and the underlying implications as to personality theory. These raters had participated actively in numerous conferences at which the scoring procedure was thoroughly discussed, prior to making the ratings.
Each of the raters scored approximately half of the men and half of the women, high and low scorers being distributed at random in about equal proportions within each group. (Concerning added duplicate ratings to check on reliability, see below. )
In particular, the interviews of cases Jl,fz to M2o and F22 to F39a among those scoring high, and of cases M2 to Mzg and of F29 to F39 among the low-scoring interviewees were evaluated by rater M, and those of the re- maining interviewees, listed farther down in the respective subdivisions of Tables 1 (IX) and 2 (IX), were evaluated by rater R. It should be added that the code numbers used were distributed at random among the various groups so that each rater rated not only men and women, low scorers and high scorers, but also approximately equal proportions of subjects who had been given Forms 78 and 45 of the questionnaire (see also Table 3 (IX)).
4. RELIABILITY OF THE INTERVIEW RATINGS
There were three ways in which some light was thrown upon the difficult question of the reliability of the interview ratings, although only the second
? INTERVIEWS AS APPROACH TO PREJUDICED PERSONALITY 329
of these deals directly with reliability in the technical sense of the term. The other two refer to aspects which are merely more or less closely related to this problem.
a. PROPORTIONs OF RATINGs GIVEN. First, the proportion of High, Low, and Neutral ratings within each category was compared for the two raters and found to be in fairly good agreement. One method used in computing an index for this agreement was the following: the frequencies of "High" responses, as scored by the first of the raters on each of the variables was plotted on a scatter diagram against the frequencies of such responses as scored by the second rater. In this comparison, the two raters are repre- sented by the different nonoverlapping groups they were assigned to rate. Similar scattergrams were obtained for "Low" and "Neutral" ratings, and in each of the three cases men and women were plotted separately. With one exception, the correlation coefficients computed from the six scatter- grams were between . 70 and . 82. This indicates that the two raters tended to concur fairly well in giving either a relatively large or a relatively small number of "High," or of "Low," or of "Neutral" ratings within any of the approximately ninety categories, showing a certain uniformity at least for one aspect of the rating policy.
(It may be added that in absolute terms there is also good agreement, the range of frequencies of "High" scores being o to I2, and o to I4, for the two raters, respectively, the various categories being considered for each of the sexes separately. For "Low" scores the corresponding ranges were o to I I and o to I 3? Thus, while both the raters neglect to use some of the alter- natives offered by the Manual (as indicated by "o"), neither of them uses the opposite alternatives indiscriminately, i. e. , in the characterization of all or nearly all the interviews analyzed by them. )
In view of the fact that the two staff members rated different samples of subjects, the coefficients and other data given above suggest that the char- acter and distribution of ratings given for the various categories are to a considerable degree intrinsic to the category in question, at least within our specific combination of raters.
b. INTERRATER AGREEMENT. Secondly, we turn to reliability proper. Nine interviewees in the group assigned to rater M were, in an additional checking procedure not used for the main analysis or for the survey in Tables I (IX) and 2 (IX), also rated by rater R under the identical detailed set of aspects. Two of these nine subjects were deliberately chosen from the relatively small group-12 of the So interviewees, i. e. , I 5 per cent-of those for whom the composite standing based on the detailed ratings of the original rater had missed the correct diagnosis as to prejudice. These subjects were M19 and F39. As may be further seen from Tables I (IX) and 2 (IX), the remaining seven are likewise mostly not from among the clearest cases as far as inter- view ratings are concerned.
A comparison of t:he gross results of the two rating procedures is shown
? Interviewees
Prejudice score (from
TABLE 7 (IX)
RELIABILITY OF INI'ERVIEW RATINGS: INTERRATER AGREEMENT ON NINE SUBJEX:TS
Percentage of High ratings on interview
Rater M Rater R
Intuitive rating on interview Rater M Rater R
LL H H LL HH H ? HH LL HH HL
questionnaire). M3low19. 236. 1LL
M4 high 77. 3 78. 1 H H M5low31. 431. 7LL M6 high 72. 7 74. 4 H H Ml9 low 70. 4 31. 1 H L F24 high 70. 5 70. 0 H H F29 low 33. 3 34. 4 L L F31 high 79. 4 77. 7 H H F39 low 72. 2 27. 2 H L
Composite standing on interview Rater M Rater R
? INTERVIEWS AS APPROACH TO PREJUDICED PERSONALITY 33 I
in Table 7 (IX). The prejudice scores based on the questionnaire are taken from Tables I (IX) and 2 (IX). However, the numerical scores appearing on these tables for the subjects listed here, as based on the ratings of rater M, are limited to a selection of the more discriminating categories (see below, Section 6). The figures in Table 7 (IX) are based upon the ratings on all categories and are further given as percentages of "High" ratings relative to the possible maximum of ratings as given by the total number of cate- gories. Thereby the number of Neutral ratings-easily obtained by subtract- ing both "High" and "Low" ratings from the total number of categories, 86 for men and 90 for women (see below)-has been added half and half to the "High" and the "Low" ratings. Composite standing as indicated by a percentage score of over, or of under, so, and finally intuitive over-all ratings of the interview make up the remainder of the table.
Percentage scores show excellent interrater agreement for six of the nine subjects. Of the remaining three, those with really striking discrepancies are the same two mentioned above as having been misjudged by the original rater, M, namely M19 and F39. In both cases, the second rater has rectified the error very clearly by establishing percentage scores in the neighborhood of 30 which contrast sharply with those in the neighborhood of 70 as obtained from the first rater. The correlation coefficient between the columns repre- senting the two raters-not very meaningful under the circumstances-is . 6I. It would be raised to about . 8 if one of the two "controversial" cases just mentioned were eliminated so as to adjust the proportion of such cases more closely to that referred to above as existing in the total sample of inter- viewees, namely Is per cent. Such a figure, if verified on a broader basis, would be quite satisfactory for the kind of material involved.
In terms of composite standing and intuitive ratings, agreement is perfect save for the two cases mentioned. (Intuitive ratings on one of the controversial subjects, M 19, is incomplete due to the fact that rater R, contrary to instruc- tions, declared herself as unable to make up her mind in this particular case. )
It may be added that the means of the percentage scores for the two raters are quite close to one another as well as to the ideal value of so. They are s8. 4 for M and so. 9 for R. This augments the evidence brought forward above under (a) with respect to the proportion of ratings given by the two raters. The slight preponderance of "High" ratings in rater M is also reflected in his intuitive over-all ratings. In fact, it is concentrated in the two cases where he makes his mistakes and where the second rater evens out the score.
A breakdown for the six major areas covered by the Scoring Manual, namely family patterns (see Chapter X), attitude toward sex, other people, and self (see Chapter XI), and dynamic character structure and cognitive personality organization (see Chapter XII) is given in Table 8 (IX). The number of categories for each area is also indicated. Considering the small- ness of these numbers, pairs of averages from raw scores in terms of number
? Areas in Scoring Manual
Family pattern (parents, etc) A ttitude toward sex
Number
of four high scorersc
categories Rater M Rater R
five low scorersc Rater M Rater R
A ttitude toward other
Attitude toward self
Dynamic character structure Cognitive personality organization
Totals
H-L 28 6. 8-1. 3
7 4. 5-0. 3 11 6. 3-0. 0 16 9. 3-1. 3 22 15. 3-0. 3
6 5. 0-0. 0
goa 47. 2-3. 2 49. 9-4. 4
H-L
3. 8-5. 6 1. 8-2. 2 2. 2-3. 2 3. 0-6. 6 5. 8-6. 2 2. 0-2. 8
TABLE 8 (IX)
INl'ERRATER AGREEMENl' ON INl'ERVIEW RATiroS FOR SIX MAJOR AREAS
people
aFor men the total is only 86; no adjustment to this sl~ght difference has been made in the present table in the case of the men subjects.
bRounded to one decimal place csee Table 7 (IX)
Average number of rating. sb received by
H-L
9. 3-1. 3 5. 3-0. 3 6. 3-0. 3 9. 5-l. 5
H-L
3. 0-8. 6 1. 0-3. 4 0. 4-5. 0 1. 0-9. 8 1. 6-8. 8 0. 4-5. 0
14. 5-0. 5 5. 0-0. 5
18. 6-26. 6 7. 4-40. 6
? INTERVIEWS AS APPROACH TO PREJUDICED PERSONALITY 333
of "High" and "Low" ratings (H-L) are given for the four high scorers as contrasted with the five low scorers listed in Table 7 (IX). No indices of reliability were computed here; but comparison of the first with the second and of the third with the fourth pair of figures in each row of Table 8 (IX) reveals a good deal of agreement between the two raters. The fact that this agreement is less pronounced in the case of the low scorers as shown in the columns containing the third and fourth pairs of figures, and that, further- more, the values of H and L within these pairs ofteri show less clear-cut differentiation than they do in the left part of the table, is due-as was revealed in more detailed analysis not presented here-to the fact that both cases with controversial diagnosis, M 19 and F39, happen to be in this group. Perhaps with the exception of family pattern and attitude toward self, this lack of differentiation for the group of low scorers, especially in rater M, and the ensuing disagreement with rater R, is about evenly distributed over the various areas; for high scorers differentiation and agreement is about equally good for all the areas, and the "totals" are in excellent agreement with one another.
Discounting the controversial cases, i. e. , the 15 per cent in our total sample of interviewees for which the original rater arrived at a diagnosis op- posite to that given by the defining score on the prejudice scales, the results of this fragmentary analysis of reliability are quite encouraging.
In fact, if the trend as discussed for Table 7 (IX) should be representative of the entire sample, interrater agreement for the remaining 85 per cent of the inter- viewees would be close to ? 9? For the other 15 per cent one may contemplate challenging the validity of the defining prejudice score along with doubting the validity of the interview rating. The "questionnaire-high" may after all be considered the product of an approach that is by definition less con- cerned with underlying dynamics than is the diagnosis of the "personality- high. "
Further data on interrater agreement on the interview will be presented in Chapter XIII.
The problem of agreement of various types of ratings among themselves, such as in our present context especially of composite standing and intuitive over-all rating, will be discussed in Chapter XIII. In a broader sense such aspects are also included within the general concept of reliability.
A third avenue of scrutiny somewhat akin to reliability problems is through the study of "halo-effects," to be discussed next.
5. MINIMIZING HALO-EFFECTS IN RATING THE INTERVIEWS
We return now to the problem of the carry-over from one category to another, much in the way of the "halo-effect" known in social and educa- tional psychology. One way of preventing or minimizing the halo-effect would have been to use designations other than "(presumably) High" and
? 334
THE AUTHORITARIAN PERSONALITY
"(presumably) Low" to characterize the two opposites within the various categories. But such a procedure would have prevented only raters un- familiar with the underlying hypotheses from succumbing to halo-effects. Such raters, on the other hand, would have been undesirable from other, more crucial points of view. It was thus decided to leave control of halo- effects to special analytical attitudes the raters were asked to maintain, and to ascertain the degree of relatedness in a statistical analysis of the completed ratings. It must be noted that-as in all cases of halo-effects-a certain amount of correlation may be fully justified by fact, i. e. , by existing correlation of real traits. Exactly how much of the halo is realistic would require further intensive study for all combinations of categories involved.
Both the variability of the discriminatory power of the single categories (see below) and the variability of the proportion of "High" and "Low" rat- ings ascribed to the various subjects seem to indicate that the raters succeeded, at least in part, in keeping the halo-effect within reasonable bounds in rat- ing the subjects. Evidence on the second of these points is contained in Tables I (IX) and 2 (IX) in the first two of the columns relating to ratings of the interviews. As is readily seen, hardly any of the subjects are classified as "High" or as "Low" on all categories even when the "Neutral" ratings are excluded from consideration, and for some of them there is an approximately equal number of "High" and "Low" ratings. This shows that the raters were quite capable of separating the various issues involved, and of rating a person
as "High" in one respect and as "Low" in another.
6. TABULATION OF INTERVIEW RATINGS BY CATEGORIES: STATISTICAL SIGNIFICANCE
A tabulation of interview ratings by single categories was obtained simply by counting the instances of High (H), of Low (L), and of Neutral (N) ratings, on a given category, among subjects of each of four groups-high- scoring men, low-scoring men, high-scoring women, and low-scoring women (scoring, that is, extremely high or low on the direct prejudice ques- tionnaire).
Analysis of the figures for a given category, with a view to their bearing upon the underlying hypotheses, could have any one of four outcomes-two "positive" and two "negative. " The "positive" instances are confirmatory of the original hypothesis. They include "High" ratings-i. e. , those designat- ing reactions presumed to occur more frequently in the prejudiced person- when given to subjects scoring high on the prejudice scales, as well as "Low" ratings given to relatively unprejudiced, i. e. , our so-called "low-scoring," subjects, in short, the hH and lL combinations. The remaining two figures, indicating the frequency of prejudiced subjects receiving a "Low" and of un- prejudiced subjects receiving a "High" rating, in short, the? hL and the lH combinations, constitute the "negative," nonconfirmatory instances.
? INTERVIEWS AS APPROACH TO PREJUDICED PERSONALITY 33S
The series of tables in Chapters X to XII (Tables I (X) to 2 (XII)) pre- sents, for the two sexes separately, the numbers of prejudiced and of un- prejudiced interviewees who give the presumably "High" and the pre- sumably "Low" responses in the interviews, for each of the scoring categories. The number of "Neutrals" may be obtained by subtraction of their sums from the total number of subjects in the respective prejudiced or unprejudiced groups. The four raw figures are followed by two sums which contrast the total number of positive with the total number of negative in- stances. All "positive," i. e. , confirmatory, evidence is italicized. Whenever a category is defined in the Scoring Manual by only one variant, H or L, rather than by an opposition of an H with an L, only the presence of the trait in question is registered in the tables and the remaining space is left blank.
The final column of the tables indicates the level of statistical significance of the difference, on the category in question, between prejudiced and unprej- udiced extremes as defined in the present study. It thus refers to the "dis- criminatory power" and the importance of the category as a personality correlate-and therefore as a potential measure-of overtly expressed ethnic prejudice. Indications of significance are given in terms of whether or not the sper cent (satisfactory), the 2 per cent, or the I per cent level (highly satisfactory, since in this latter case there is a 99 per cent probability that the obtained difference is not due to chance factors) has been reached or surpassed, without specifying whether or by how much they have been surpassed. Significance was computed after evenly dividing the "Neutrals" among the "High" and the "Low" ratings. It must not be forgotten, in this connection, that dividing the Neutral ratings tends to lower the index of significance, the more so the larger the proportion of these ratings. Thus, in treating the Neutral ratings as we do, we are keeping on the safe side, since, as was pointed out above, the Neutral scorings are based, in a con- siderable proportion of the cases, on lack of information rather than on lack of actual discriminability.
Wherever the proportion of Neutrals for the total sample of interviewees is larger than so per cent, the statistical significance of the category in question was not computed and therefore there was no entry under any circumstances in the last column of the tables. The category was also omitted from the survey of interview scores in Tables I (IX) and 2 (IX). For men the total number of categories is 86, somewhat less than for women, in whose case a few more subdivisions were introduced into the Scoring Manual. Of these 86 categories, 72 yielded less than so per cent Neutrals, and thus re- mained for full treatment. For women only 6s of the original 90 categories yielded less than soper cent Neutrals and were thus retained for full treat- ment. The fact that the categories were generally somewhat more dis- criminatory in the case of the men than in the case of the women may be
? THE AUTHORITARIAN PERSONALITY
accounted for by the fact that most of the men were interviewed at a later stage of the study and that therefore their records were more complete.
All the calculations discussed above were performed separately for men and women. Corresponding figures for each appear closely adjacent to one another in the tables, those for men (M) in the upper left-hand and those for women (W) in the lower right-hand area of each of the "cells" that one may imagine at the intersection of vertical columns with horizontal rows, the latter defined by the various categories.
? CHAPTER X
P ARENTS AND CHILDHOOD AS SEEN THROUGH THE INTERVIEWS Else Frenkel-Brunswik
A. INTRODUCTION
In turning to the specific results of the interviews we begin with the organization of the family. Many of the attitudes and underlying needs dis- cussed in this volume must be assumed to originate, as far as the individual is concerned, in the family situation. Here the growing child learns for the first time to handle interpersonal relations. Some of the members of the family are in an authoritative, others in an equalitarian or in a weaker position than himself. Some are of the same, others are of the opposite sex. It soon becomes evident to the youngster what kind of behavior is considered appropriate and will lead to reward and what kind of behavior will be punished. He finds himself confronted with a certain set of values and certain expectations which he has to meet.
Within the general common framework of the white American popula- tion, families vary greatly as to the rigidity or flexibility of the roles defined within the family, as well as to values in general. We shall encounter families in which considerably more emphasis is placed on obedience than it is in others. In some cases discipline is harsh and threatening, in others intelligible and mild. Or there may be rigorous adherence to conventional rules and customs rather than to more flexible and more intrinsic values which lead to greater tolerance for individual variations. Or smooth functioning within the family may depend either more on exchange of well-defined obligations and "goods," or else on an exchange of genuine affection. These and other differences in the organization of the family are under scrutiny in the light of their possible implications with respect to the personality structure of the individual and his social and political beliefs.
Although no striking relations between these patterns and gross economic factors have been uncovered in the present study, systematic investigations of a more distinctly sociological nature would undoubtedly reveal broader
337
? THE AUTHORITARIAN PERSONALITY
cultural and subcultural determinants of these differences, say, the greater frequency of one or the other type of family organization in different national subgroups, or a dependence on the relative stability or instability of the socioeconomic family history. ?
While this goes to press, data from a separate project (Frenkel-Bruns- wik, for an advance report see 30)1 seem to indicate that parents of ex- tremely prejudiced children are relatively often the children of foreign-born parents or show preoccupation as to social and national insecurity.
Specific rating categories from the Interview Scoring Manual in the area of family organization, followed by tabulation of quantitative results and eventually by a discussion and the presentation of pertinent quotations from the interviews themselves, are given in Sections B to D.
B. A TTITUDES TOW ARD P ARENTS AND CONCEPTION OF THE FAMILY
1. DEFINITION OF RA TING CA TEGORIES AND QUANTITATIVE RESULTS
In line with the over-all subject matter of the present volume, the discus- sion concerning attitudes toward parents will, among others, center about the following questions: Is the general tendency toward glorification and lack of critical evaluation of ingroups on the part of the ethnically prejudiced also mirrored in . their attitudes toward their parents? Is there a tendency toward submission, and how are the problems of rebellion, hostility, and guilt handled in our two opposite groups? How are the feelings of genuine love related to conformity as contrasted with independence?
The definitions of the specific categories of the Scoring Manual dealing with problems of this nature are listed here in the form in which they were used by the raters. It should be remembered, however, that the Manual represents merely a summary of what was developed, and discussed with the two raters, in extended conferences preceding the actual rating procedure. The subsequent text makes occasional use of the more prominent of the con- notations thus established which were not formally incorporated in the Manual in order not to overload it in actual use.
Before starting the presentation of the various parts of the Interview Scoring Manual, attention must be called to the fact that strict opposi-
1 The study on social discrimination in children referred to here is being carried out at the Institute of Child Welfare of the University of California. The project was initiated by the present writer, in cooperation with Harold E. Jones and T. \V. Adorno, and spon- sorship was at first by the Scientific Department of the American Jewish Committee. In developing the tests and experiments the present writer was aided primarily by Claire Brednor, Donald T. Campbell, Joan Havel, Murray E. Jarvik, and Milton Rokeach.
? PARENTS AND CHILDHOOD SEEN THROUGH INTERVIEWS 339
tion or near-opposition of the presumably "High" and the presumably "Low" variants is to be assumed only where the numbers or number-letter combinations appearing on the right side of the page are identical with those on the left. Thus Category 2, "victimization," stands in a somewhat oblique relation to Categories za to zc. In some cases pairings of this kind reflect the fact that there is more than one opposite to a given variant. Beginning with Section C of this chapter, different sets of letters are sometimes used on the two sides to stress an absence of one-by-one correspondence of a series of alternatives listed on the right with a series of items on the left, although the lists in their entirety define a more clear-cut pattern of oppo- sition. (Concerning the lopsided evaluation of asymmetrical categories in the tables to follow, see the concluding pages of the preceding chapter. )
The first subdivision of the Interview Scoring Manual follows. As in the Interview Schedule, italics are used to represent key phrases which had been emphasized to the raters by underscoring.
INTERVIEW SCORING MANUAL: ATTITUDES TOWARD PARENTS AND CONCEPTION OF FAMILY
(to Table r(X)).
I .
2.
PRESUMABLY "HIGH" VARIANTS
Conventional idealization of parent(s): Overestimation of qualities and status, expressed in behavioral (essentially exter- nal), conventionalized general- ities, or undifferentiated "all's
well" attitude
Victimization (quasi-persecu- tory) byparent(s): Neglect, in- cluding failure to give proper discipline, unjust discipline;
"picked on"; unfair: resents pre- ferring of rival sib or spouse (or foster-sib or step-spouse); etc.
PREsUMABLY "Low" VARIANTs I. Objective appraisal of parents
2a. Principled open rejection
2b. Genuine positive affect: some reference to (positive) psycho- logical qualities; individualized
characterizations
2c. Blocked affect (Presumably
mutually exclusive with 2a) 3? Principled independence
4? Love-seeking succorance-nur- turance-affiliation toward par- ents
3a. Submission to parental author- ity and values: respect based on fear
3b. Capricious rebellion against parents; delinquency
4a. Ego -alien dependence- for - things and suppon on parents: essentially exploitive-manipula- tive-"getting"; an externalized relationship
4b. Sense of obligation-and-duty to
? THE AUTHORITARIAN PERSONALITY
parents: Desire to "make it up
to them"
5? lngroup orientation to family as 5? Individualized approach to
a whole; e. g. , emphasis on fam- members of the family ily heredity and "background";
homogeneous-totalitarian fam-
ily vs. rest of world; aristocratic
superiority of family, etc.
In Table I (X), the results for both men and women interviewees are presented in the manner described in the concluding section of the preceding chapter. Abbreviated formulations of the categories just listed are presented, along with category numbers, for purposes of easier identification.
2. IDEALIZA TION VS. OBJECTIVE APPRAISAL OF P ARENTS
In view of their general tendency toward conventionality and submission toward ingroup members, it is not surprising to find in the prejudiced sub- jects a tendency toward "idealization of the parents. " This idealization is expressed characteristically in generalized and undifferentiated, convention- alized terms which primarily glorify external features of physical appear- ance or overt behavioral conduct rather than involving the more internal aspects of their personality. "Objective appraisal" of parents, referring to an ability for critical evaluation of the parents in specific and psychologically conceived terms, on the other hand, is predominant in the unprejudiced subjects.
The detailed results presented in Table I(X) (under Category I) reveal the striking fact that none of the low-scoring women interviewed shows the glorification of the parents just described; instead, I I of them show objective, critical appraisal. Of the high-scoring women, 9 show glorification and 6 ob- jective appraisal of parents. For the total group of women, there are 20 "positive" as contrasted with 6 "negative" instances in the sense defined at the end of the preceding chapter. The statistical significance of the differ- ence between the positive evidence (i. e. , that confirming the original hypoth- esis underlying the distinction between the "High" and the "Low" variant) and the negative (nonconfirmatory) evidence cited is found to be at the " I per cent level," and thus highly satisfactory (see Chapter IX, Section F, 6).
A good illustration of the "High" attitude in women is given by the fol- lowing quotation from the interview of one of the high scorers: "Mother- she amazes me-millions of activities-had two maids in years ago, but never since-such calmness-never sick, never-beautiful woman she really is. " The reference to external dimensions, both behavioral ("million ac- tivities") and physical ("never sick, beautiful"), can be seen clearly in the foregoing record. It must be emphasized that the subjects were asked, in this connection, "What kind of person is your father (mother)? " without further
? 1. Conventional idealization(H) vs. objective appraisal(L) of parents Women
. H.
9 6 0 ! ! 20 6
2. Victimization by parents (H) vs. a. Principled open rejectton of
parents (L}
b. c. Genuine affection or blocked Men
affect for parents(L)
aa. Submission to parents (H) vs. principled independence(L)
women 3 10 10 3 Men 14 2 2 10 24 4
women J1. . 0 1 8 17 1 Men 9 4 9 4
women6 1 61
1 2
1 5
3b. Capricious rebellion(H)
4a. Dependence for things on parents Men 13
1 5
1 2
1 1
2
14 27 2 8 2f 7
(H) vs. love-seeking affiliation(L)
4b. Sense of obligation and duty(H) 5. Ingroup conception of family(H)
Women T3 Men 5
women1
--
!
20 men and 25 women "high scorers"
20 men and 15 women "low scorers"
Men
1
l
Men women
_? _ 4 8 4
. . ? . 4
8 4 -
-
TABLE 1 (X)
INTERVIEW RATINGS ON ATTITUDE TOWARD PARENTS AND CONCEPT OF FAMILY
FOO 80 SUBJECTS SCORING EXTREMELY "HIGH" OR "LOW" ON THE ETHNIC PREJUDICE QUESTIONNAmE SCALE
Interview rating categories (abbreviated from Manual)
s~x I
Number of "High"(H) and "Low"(L) ratings received by
Sums of instances Level of statistical
? positive" "negative" significance reached
L H L
11 1 2 13 24 3
(percentage)
. . .
Men
Wo111en 2 ~ 3 2
Men 6
Women7 1
i1 11
62 71
1 A . A. 1 4 12 12 4
? THE AUTHORITARIAN PERSONALITY
specification (see the Interview Schedule as presented in the preceding chap- ter).
The overestimation of parents in more general terms is especially clear in the record of another prejudiced woman interviewee (F24): "Father-he is wonderful; couldn't make him better. He is always willing to do anything for you. He is about years old, six feet tall, has dark brown hair, slim build, young-looking face, good-lookirig, dark green eyes. " The same sub- ject gives further evidence of the stereotypical conceptoion of parents in the high scorers by repeating the same description for her mother, differentiat- ing only the physical characteristics. After repeating the first two sentences she continues: "She is about 5 feet 5 inches, neither slim nor stocky. She's reducing. She has dark hair, blue eyes, is nice looking. She is years old. " These illustrations are quite typical of the responses of our high-scoring subjects.
It should be added that none of the 6 high-scoring women who show "ob- jective appraisal" of parents was considered to manifest "positive affect" toward the parents (Category 2). This will be discussed below.
Very characteristic of high-scoring subjects is an initial statement of great admiration for parents, followed by some criticism which is not, how- ever, recognized as such by the subject. The comparison of reactions to general questions with reactions to specific questions proved especially fruitful, e. g. , when parents were characterized in general positive terms but the specific episodes and traits referred to were mostly of a negative charac- ter. (Regarding related mechanisms of self-deception, see 33, 34? ) The sub- ject is aware only of admiration toward the parents, and the reservations seem to enter the picture against "better" intention and knowledge, thus in- jecting into the statement an element of ambivalence.
Examples of this attitude from the protocols of high-scoring women are quoted in the following. It should be mentioned here that throughout Chapters X to XII the special code numbers used in rating the interviews are retained for greater anonymity; for the same reason, localities, occupa- tions and related personal data are either left out or disguised.
F31: (Father? ) "He has a marvelous personality and gets along well with people. He has a hot temper. "
Or F79: "Mother was, of course, a very wonderful person. She was very nervous. Irritable only when overdoing. "
Or again, Fu: "Father is quiet and calm. He never shows irritation.
