18
THE EFFECT OF ETHNICITY ON CRIME ATTRIBUTION GIDEON FISHMAN, ARYE RATTNER, AND GABRIEL WEIMANN University of Haifa, Israel This study investigates the impact of ethnic origin on the attribution of criminal oflenses to faces seen in portrait photographs. The underlying hypothesis is that ethnic stereotypes are a result of the way people perceive themselves and other groups (that is, the relation of in-groups to out- groups) and that ethnic stereotypes directly affect a person’s readiness to attribute particular offenses to given portraits. Thus, this study explores cross-ethnic and intra-ethnic offense attribution. This study explores the effect of ethnicity on crime attribution. Ethnicity was the variable that was presented in the form of photographic portraits of people of diverse ethnic origins. The task was to attribute various criminal offenses to the people in the photographs, and this study is an analysis of the results. The relationship between facial appearance and criminality has intrigued criminologists for a very long time. One avenue of research has been to inves- tigate whether certain types of faces characterize people who commit certain types of crimes. However, research of this sort has been quite rare (Bull, 1982). The most common procedure has been to present photographs of the faces of actual criminals and to ask subjects to indicate what crime they thought the pictured criminal had committed (Kozney, 1962; Thornton, 1939; Bull and Green, 1980). There may be an adequate explanation for why certain types of crimes are attributed to certain types of people belonging to particular social or ethnic groups; this attribution seems to be a consequence of the general cognitive process by which people structure and simplify their experience and give meaning to their environment (Brewer, 1977: 24). According to Stephan and Rosenfield (1982: 97), stereotypes are created in order to predict how other people will behave, and stereotyping is a generic process applicable to any socially defined group. The studies also point out that the content of stereo- types depends, at least in part, on factors unique to the relations between the groups in question. Accordingly, one of the basic processes by which stereo- types are created and maintained is the categorization of the world into dis- tinct racial and ethnic groups (Stephan and Rosenfield, 1982: 103). From an attribution-theory perspective, it has been suggested that people have a well-organized set of categories-that is, have stereotypes that provide CRIMINOLOGY VOLUME 25 NUMBER 3 1987 507

THE EFFECT OF ETHNICITY ON CRIME ATTRIBUTION

  • Upload
    haifa

  • View
    1

  • Download
    0

Embed Size (px)

Citation preview

THE EFFECT OF ETHNICITY ON CRIME ATTRIBUTION

GIDEON FISHMAN, ARYE RATTNER, AND GABRIEL WEIMANN

University of Haifa, Israel

This study investigates the impact of ethnic origin on the attribution of criminal oflenses to faces seen in portrait photographs. The underlying hypothesis is that ethnic stereotypes are a result of the way people perceive themselves and other groups (that is, the relation of in-groups to out- groups) and that ethnic stereotypes directly affect a person’s readiness to attribute particular offenses to given portraits. Thus, this study explores cross-ethnic and intra-ethnic offense attribution.

This study explores the effect of ethnicity on crime attribution. Ethnicity was the variable that was presented in the form of photographic portraits of people of diverse ethnic origins. The task was to attribute various criminal offenses to the people in the photographs, and this study is an analysis of the results.

The relationship between facial appearance and criminality has intrigued criminologists for a very long time. One avenue of research has been to inves- tigate whether certain types of faces characterize people who commit certain types of crimes. However, research of this sort has been quite rare (Bull, 1982). The most common procedure has been to present photographs of the faces of actual criminals and to ask subjects to indicate what crime they thought the pictured criminal had committed (Kozney, 1962; Thornton, 1939; Bull and Green, 1980).

There may be an adequate explanation for why certain types of crimes are attributed to certain types of people belonging to particular social or ethnic groups; this attribution seems to be a consequence of the general cognitive process by which people structure and simplify their experience and give meaning to their environment (Brewer, 1977: 24). According to Stephan and Rosenfield (1982: 97), stereotypes are created in order to predict how other people will behave, and stereotyping is a generic process applicable to any socially defined group. The studies also point out that the content of stereo- types depends, at least in part, on factors unique to the relations between the groups in question. Accordingly, one of the basic processes by which stereo- types are created and maintained is the categorization of the world into dis- tinct racial and ethnic groups (Stephan and Rosenfield, 1982: 103).

From an attribution-theory perspective, it has been suggested that people have a well-organized set of categories-that is, have stereotypes that provide

CRIMINOLOGY VOLUME 25 NUMBER 3 1987 507

508 FISHMAN, RATTNER, AND WEIMANN

a framework for making attributions about other people and situations (Bris- lin, 1981). These stereotypes are formed on the following bases: (1) the dif- ferentiation between various conspicuously and easily discernible groups; (2) the attribution of familiar characteristics to another group; (3) the fulfill- ment of social functions, such as keeping members of an in-group apart from an out-group; (4) the maximization of the positive view one holds of one’s own group; (5) the projection of one’s own feelings; (6 ) the shared beliefs of members of the same group; (7) the personality traits of people that encourage “approach” or “avoidance”; and (8) the salience or immediate per- sonal relevance of important information (Brislin, 198 1 : 105).

Published studies on the subject of stereotypes have demonstrated that attribution of character traits to a face in a photograph is often only remotely related to the objective information in one’s possession. Attribution of mean- ing to any situation, scene, or face depends on a stereotype established in the mind, and this produces a particular behavioral response (Alport and Post- man, 1947; Loftus and Palmer, 1974).

The present study investigates the impact of ethnic origin on the attribu- tion of criminal offenses to faces seen in portrait photographs. The underly- ing hypothesis is that ethnic stereotypes are a result of the way people perceive themselves and other groups (that is, the relation of in-groups to out- groups) and that ethnic stereotypes directly affect a person’s readiness to attribute particular offenses to given portraits. Thus, this study explores cross-ethnic and intra-ethnic offense attribution.

Israel provides an ideal social laboratory for this kind of study. Israeli society consists of a variety of social and ethnic groups. However, the society can be divided into two major groups: Jews, who comprise 83% of the popu- lation, and Arabs, who comprise the remaining 17%. This is the major divi- sion since these two groups divide along ethnic, national, and religious fronts. Nevertheless, the Jewish population is also far from being homogeneous. The mass immigration of Jews from all over the world has produced a highly ethnically mixed population of Jews in Israel.

This population can be broadly classified into two principle ethnic groups: “Ashkenazic” Jews who are of European origin, and make up 40% of the country’s Jewish population; and “Sephardic” Jews, who are mainly of Asian and North African origin, and account for 60% of Israeli Jews. As men- tioned, there is a large Arab minority (17%) in addition to the country’s two major Jewish ethnic groups.

Taken as a whole, the Jewish majority is the country’s norm setter and, understandably, its dominant social group. Unavoidably, there are tensions between Jews and Arabs in Israel that are rooted in the history of Jewish- Arab relations in this region. These tensions can be traced from the turn of the century, through the establishment of Israel in 1948, to the armed con- flicts between Israel and the Arab nations which followed. As a consequence,

ETHNICITY AND CRIME ATTRIBUTION 509

the Arab minority in Israel occupies a position of marginality in regard to almost every aspect of the country’s social life.

Ethnically grounded tensions and conflicts also exist among Israeli Jews, namely between Sephardics and Ashkenazics. Israel’s founding fathers were for the most part Ashkenazic Jews, and they established the supremacy of Western culture in the country. This circumstance made the absorption of the Ashkenazic immigrants to the country relatively easy, but had the oppo- site effect in regard to Sephardic immigrants. Historically, among Israeli Jews, Ashkenazics have, by and large, attained a high social status, whereas Sephardics have tended to occupy the bottom of the social ladder (Smooha and Peres, 1974).

This process of ethnic and social differentiation was attended by alienation, prejudice, and the creation of social distance between the country’s ethnic groups despite various changes in the demography of Israel’s population over the years. When Israel was founded, the Sephardics were an insignificant minority, but now they are the nation’s largest ethnic group. Nevertheless, there is still an unfavorable encounter between Ashkenazics and Sephardic immigrants in the process of their absorption into the country. Schwarzwald and Amir (1984) have pointed out that one of the persisting effects of this encounter has been a pattern of asymmetry in the acceptance and rejection processes between these two ethnic groups. Thus, Ashkenazics tend to evalu- ate members of their own group positively, whereas they evaluate Sephardics negatively. And, perhaps surprisingly, Sephardics are inclined to evaluate members of their own ethnic group less positively than they do Ashkenazics. It would appear, then, that the attitudes of Sephardics in this matter reflect those of the dominant group.

Such was the finding among elementary-school children, who were asked to evaluate ethnic traits of the “typical Israeli” (Rim, 1968). The same was found to be the case among high-school students when they were asked to evaluate members of ethnic groups (Peres, 1976), or when they reported on the extent of their willingness to engage in activities with teenagers of other ethnic groups (Amir, Sharan, Ben Ari, Bizman, and Rivner, 1978). It seems likely that this asymmetry involves more than the mere preferences of certain groups, and that a process of self-deprecation is at work, as well; this ulti- mately leads to the adoption of a negative self-image. These are typical processes in encounters between strong majority groups and weak minorities, as has been shown in a number of studies conducted in the United States on the subject of relations between whites and blacks in America (Ashmore, 1970; Brand, Ruiz, and Padilla, 1974). In these studies, too, subjects who were members of a minority group considered themselves to be less worthy than persons belonging to the majority, because they have adopted toward themselves the attitude the majority group had toward them.

The influence of ethnic stereotypes is not, however, confined to attitudes of

5 10 FISHMAN, RATTNER, AND WEIMANN

mind; it extends to actions as well. Jaffe (1984) reports that, when Israeli respondents were asked to identify criminals among a series of portrait photo- graphs of Ashkenazics and Sephardics, 97% of the Ashkenazic respondents and 96% of the Sephardic respondents attributed criminality to the portraits of Sephardics, whereas only 3% of the Ashkenazic respondents and 4% of the Sephardic respondents attributed criminality to the portraits of Ashkenazics.

The present study dealt with three Israeli ethnic groups: Ashkenazic Jews, Sephardic Jews, and Israeli Arabs. Subjects were shown portrait photographs of individuals from all three ethnic groups; they were then asked to assign to these portraits a criminal offense selected from a five-item list of offenses. The aim of the study was to determine whether the subjects’ attributions of crimi- nality would be in line with earlier research findings or different results would be obtained because the introduction of a third ethnic group had altered the dynamics of the situation. Thus, in this experimental setup, Ashkenazic Jews unquestionably enjoyed the highest social status, whether all three groups were considered together, or merely the two Jewish groups alone. Sephardics, although they occupy a less exalted position in relation to Ashkenazics, are nevertheless better placed socially than Israeli Arabs. Another question this experiment attempted to answer was whether Israeli Arabs, who as a group are relatively segregated socially and undoubtedly occupy an inferior position in the country’s social hierarchy, also experience a sense of divided identity and act accordingly.

DESIGN AND METHODS

Included in this study were 513 subjects randomly selected from the stu- dent body of the University of Haifa in Israel. The sample population con- sisted of three independent groups: one group of Ashkenazic students (n = 185), a second of Sephardic students (n = 123), and a third of Israeli Arab students (n = 205).

The study consisted of two parts. In one part subjects viewed a sequence of nine photographs of male faces1 on a television screen; the sequence was repeated for the subjects as long as they needed in order to complete their attribution assignment.

The subjects were asked to look at the portraits and to attribute any one of five offenses to the nine faces. The listed offenses were homicide, rape, armed

1. The stimulus photographs were randomly chosen from a large pool. All portraits with distinctive features such as glasses, beards, and so on, were omitted. The remaining photographs were presented to a group of 60 subjects who were asked to determine the ethnic origin of the portraits. Only those photographs which received over 75% agreement regarding their ethnic origin were selected as appropriate stimulus photos. This procedure enabled the selection of faces with clear ethnic distinctiveness.

ETHNICITY AND CRIME ATTRIBUTION 51 1

robbery, breaking and entering (burglary), and fraud. No time limit was set for answers, and subjects were allowed to change their attributions until they felt satisfied with their choices. They were also free to attribute “no offense” to any of the portraits. In the other segment the subjects were all asked to fill out a structured questionnaire regarding their own cross-ethnic attitudes and the frequency of their contact with members of the other groups. The ques- tionnaire included two scales. The first was a 66-item scale of attitudes toward the other ethnic groups (Bogardus, 1933). The second was a 14-item measure of the extent of experience and contact with members of other ethnic groups at places of work, in the neighborhood, and so on (Bogardus, 1959). Between the two parts there was a 30-minute break during which a diversion- ary task was presented to the subjects in order to minimize any possible influ- ence of one part on the other.

Nine stimulus portraits were used: three portraits of Ashkenazics, three of Sephardics, and three of Arabs. The three photographs of each ethnic group were then collapsed into the single relevant ethnic category and yielded a distribution of offense attribution. This distribution was by ethnic origin of the respondent and by ethnic identity of the portrait. Then contingency tables of three by three, based respectively on the ethnic identities of the respondent and the portrait, were constructed for each assigned offense.

The offense assignment served as a dependent variable, and the independ- ent variables consisted of identities (that is, Ashkenazic, Sephardic, or Arab) of the portrait and respondent. Two additional independent variables which were included were the indices of cross-ethnic attitude and contact. These scales are additive in nature, but were refined by not assigning an equal weight to each item. Instead, the two scales were factor analyzed, and two weighted scales were constructed.* The weights were calculated and added in the case of each respondent for each of the two scales.

The continuous variables of contact and attribution were to be cross-tabu- lated with the categorical variables of ethnic origin of the identifying subjects, ethnic origin of the portrait, and the offense attributions. However, the con- tinuous categories of contact and attitudes had to be collapsed in order to avoid the problem of creating too many empty cells in a contingency table of the four independent variables with the dependent variable. The collapsing of the contact and attitude variables into “low” and “high” followed the statisti- cal rejection of their impact on attribution of offenses still at their precol- lapsed form.3

2. The weighted scores were calcylated by the following formula: W=(Xi-%)* C/sd, where Xi is the item’s raw score, X is the item’s mean, C is the item’s factor coeffi- cient score, and sd is the item’s standard deviation.

A log linear logit model was tested where the predictor was a continuous variable (Contact and Attitude interchangably) and the dependent variable was a dichotomous one (attribution of specific offenses). The procedure was repeated for each ethnic group (that of

3.

512 FISHMAN, RATTNER, AND WEIMANN

The four independent variables (of which the variables of contact and atti- tude were collapsed into low and high, using the median as a criterion) were cross-classified with the dependent variables of offense attribution. The result was a 2 x 2 ~ 3 ~ 3 ~ 2 contingency table. This table was in turn submitted for a log linear analysis, a procedure appropriate to a multivariate analysis in which most of the variables are nominal (Goodman, 1970, 1972, 1978; Upton, 1978; Knoke and Burke, 1980). Such a procedure enables the analysis of multivariate data, especially when the variables are dichotomized or polytomized nominal variables, as in this study.

ANALYSIS AND FINDINGS Interesting differences can be observed between the ethnic groups in regard

to offense assignments. The probability of a person having a given type of offense attributed to him differs markedly for each ethnic group. As Table 1 shows, Arabs have the greatest chance (40%) of being identified as homicidal

Table 1. Probability of Attribution of Offense to Each of the Major Israeli Ethnic Groups

Offense Ashkenazic Sephardic - Arab

Homicide .29 .31 .40

Rape .28 .23 .49 Armed Robbery .27 .41 .32

Breaking and Entering (Burglary) .24 .36 .40

Fraud SO .30 .20

offenders and as burglars. There is an even greater chance rape would be attributed to them (49%). Sephardics have the highest chance of being iden- tified as armed robbers, and the probability of burglary attributed to them is only slightly lower (36%) than of its being attributed to Arabs (40%). The probability of Ashkenazics having the offense of fraud attributed to them is very high (50%), but their chances of being identified as offenders in other categories is low relative to the other ethnic groups.

This study also examined the effect of a set of four independent variables

the identifiers and that of the identified photos). All the chi square likelihood ratios were very high (ranging from 910 to 1,200 with 30 to 35 degrees of freedom); thus, they did not include any significant effect of the continuous measures of attitude and contact on offense attribution. Consequently, when their interaction and joint effect with the other independ- ent variables had to be tested, collapsing of the continuous variables was required.

ETHNICITY AND CRIME ATTRIBUTION 513

on one dependent variable. This was done by employing log linear analysis. The independent variables are the ethnicity of a stimulus portrait, denoted by (P); the ethnicity of the identifier (E); his attitudes toward the ethnic groups (A); and his contact with the ethnic group (C). The dependent variable con- sists of the attribution of any of five different criminal offenses (0) to the portrait: homicide, rape, armed robbery, burglary, and fraud. Accordingly, the log linear analysis will be repeated five times. However, because of the great quantity of data, all possible models will not be presented. Instead, only the best models will be presented in an aggregated table to which reference will be made as each offense is analyzed in turn.

Table 2 presents the best of the tested possible models that express the

Table 2. Testing Selected Models Relating Attribution of Offense (0) to Ethnicity of the Picture (P), Ethnicity of the Identifier (E), Level of Contact (C), and Cross-Ethnic Attitude (A)"

Homicide Armed Robbery Rape B & Eb Fraud

LR: LR: LR: --- Model Fitted

--- df LR; LRX2 -~ Marginals

HI H2 H3 H4 H5 H6 H7 H8 H9 HI0 HI 1 H12

(PCEA)(O) 35 (PCEA)(PO) 33 (PCEA)(EO) 33 (PCEA)(CO) 34 (PCEA)(AO) 34 (PCEA)(PO)(EO) 3 1 (PCEA)(PO)(AO) 32 (PCEA)(PO)(EO)(AO) 30 (PCEA)(PO)(EO)(CO) 30 (PCEA)(PO)(EO)(CO)(AO) 29 (PCEA)(PEO) 21 (PCEA)(P AO) 30

169.31 122.07 111.51 155.52 168.80 63.90

121.86 63.43 5 1.23 42.12 46.9 1

91.64

146.90 91.34

128.82 146.90 145.58 73.03 91.13 73.01 72.56 12.53 67.95 89.90

214.04 129.65 121.75 84.90 64.81 98.18

193.52 121.56 11 1.05 213.94 126.54 118.93 213.71 128.12 98.58

63.81 56.73 81.36 84.90 61.78 65.08 63.81 52.95 54.41 63.19 52.95 74.44 63.75 51.11 54.40 59.92 41.39 70.94 70.41 47.56 61.95 , .

Only selected models are presented here, although all possible models were tested in the

Breaking and entering analysis.

potential relations under consideration. The letters in brackets signify the set of relations that is being tested. Note that (0) is here the dependent variable in each of the models, and the relations among the independent variables (PCEA) are fixed and fitted for each model being tested.

In each of the analyses, model H1 represents the first possibility; namely, that none of the independent variables is significantly related to the depen- dent variable. A test of the goodness of fit of this model indicates that it

514 FISHMAN, RATTNER, AND WEIMANN

should be rejected for every type of offense. The examination of the models for goodness of fit suggests that in the case of all of the offenses save one, the ethnicity of the portrait interacts with the ethnicity of the identifier and that this in turn influences the attribution of a specific offense.

ATTRIBUTION O F OFFENSES

An examination of the models reveals one model to be superior to all others; it is the best-fitting model for four of the five attributed offenses. Model H11 seems to be the best-fitting model for four of the offenses: homi- cide, rape, armed robbery, and burglary (see Table 2). This model, which contains the (PEO) relations, seems to suggest that ethnicity plays a highly significant role in the attribution of these four offenses. Thus, when the eth- nic origins of the subject and the identified portrait are considered jointly, the effect is very noticeable. On the other hand, the model seems to give little importance to cross-ethnic attitudes and contact. It seems that this aspect of ethnic relations plays a very minor or even negligible role in the process of offense attribution.

When closely examining the phenomenon of homicide attribution, model H11 yields a likelihood ratio (LRx’) of 46.91. When one compares models H2 and H3 with model H6, one finds both relations (EO) and (EP) individu- ally to have a similar impact in reducing the likelihood ratios. In the case of attribution of homicide, the variable of contact (C) also seems to have some effect; this can be seen in a comparison of models H6 and H9, where the LRx’ is reduced by a score of 12.67, which is statistically significant (df = 1, p < .001). However, the parsimonious model is undoubtedly model H11, which utilizes only the independent variables of ethnicity.

In the instance of attribution of rape, model H11 yields a LRx2 of 59.92 (df = 27) and is the model which fits best. Neither the contact (C) nor the atti- tude (A) variables contribute anything in this model. This even includes con- tributions within any interactive model as well as their contributions as a single variable.

However, unlike the attribution of homicide, the impact of (PO) relation is very significant even when not placed in an interaction. This is revealed in a comparison of models H6 and H3, wherein the introduction of (PO) reduces the likelihood ratio from 193.52 in model H3, to 63.81 in model H6. In this way, the relative importance of the ethnic origin of the identified portrait in the attribution process is revealed.

In attribution of armed robbery, again the interactive model H11 seems to be the best-fitting model (LRx’ = 67.95, df = 27). Also, here the other independent variables (attitude and contact) are of no significance in deter- mining attribution of robbery to stimulus photos.

When models H9 and H10 are compared with model H6, it is obvious that

ETHNICITY AND CRIME ATTRIBUTION 515

the addition of the (CO) relation, or of both the (CO) and (AO) relations, contributes only very marginally to improving the goodness of fit of the model. Such an addition reduces the LRx2 value from 73.03 in model H6 to only 72.56 in model H9, and to 72.53 in model H10.

The effect of the ethnic identity of the portrait alone was found to have the most profound effect on attributions of the crime of armed robbery. This conclusion emerged from the comparison of models H6 and H3 which iso- lated the direct effect of (PO). Although the best model includes the variables of both the ethnicity of the portrait and the ethnicity of the identifier in inter- action (PEO), it is the ethnic identity of the stimulus portrait that exerts the strongest influence. It alone reduces the LRx2 from 128.82 in model H3 to 73.03 in model H6; hence, the significance of the ethnic identity of the por- trait in the case of attributions of the crime of armed robbery.

Almost identical results are found with regard to attribution of burglary. Here, model H11 produces a LRx2 of 41.39 (df = 27) and, as in the case of rape and armed robbery, the (PO) relation-that is, the ethnic identity of the portrait-seems to make the single strongest contribution to the model. This is evident from the reduction of the LRx2 from 121.56 in model H3 to 56.73 in model H6.

The attribution of fraud seems to deviate from the above pattern. Here model H8 seems parsimoniously to fit the data (see Table 2, “Fraud”). Accordingly, the attribution of fraud is affected by three relationships: the ethnic identity of the stimulus portrait (PO); the ethnic origin of the identifier (EO); and the identifier’s attitude toward the relevant ethnic group (AO). However, a comparison of the models reveals that the H10 is a better model. Although model H8 produces a negligibly higher LRx’ (54.41), it must be considered the preferable model since it employs only three variables to arrive at an improved goodness of fit almost equal to that of model H10, which does so with four variables. Thus, although the variables of the ethnic identity of the portrait, the ethnic origin of the identifier and the identifier’s attitude are each important to the process of attributing the offense of fraud, it is the ethnicity of the stimulus portrait and the identifier’s attitude that seem to be singly more significant to the process than is the ethnicity of the identifier.

MAIN EFFECTS AND INTERACTION EFFECTS

So far, the analysis reveals that the ethnic origin of the identified person and, to a lesser extent, the ethnic origin of the identifier, are the two major variables determining the attribution of crimes. Therefore, the nature of these relationships will be closely examined; this will involve an examination of which ethnic origins are related to which crimes and the strength of these relationships. To measure these effects, the log linear procedure includes the computations of the standardized lambda coefficients.

516 FISHMAN, RATTNER, AND WEIMANN

A positive lambda coefficient will indicate the existence of a greater than average chance for a given offense to be attributed to a given group of stimu- lus portraits by identifiers belonging to the specified ethnic groups. A nega- tive coefficient will indicate the existence of a less than average chance for such an assignment to be made. Finally, a zero coefficient will indicate that no difference exists in the rate of attributions of a specific offense from that which is expected by chance-when the outcome is independent of the tested parameter. The calculations of the standardized lambda coefficients are derived from the “saturated” model-that is, from the model that includes all possible effects (Goodman, 1978).

MAIN EFFECTS

All the offenses, apart from fraud, are most likely to be attributed to por- traits of Israeli Arabs than to portraits of any other ethnic member; this includes armed robbery, which received a positive, albeit low lambda score (see Table 3).

Portraits of Ashkenazic Jews are invariably the least likely to have any offense attributed to them, except for fraud (which received only a low lambda coefficient).

Ashkenazic identifiers would seem to be the least eager to attribute offenses to stimulus portraits, since they show the lowest probability among the groups of identifiers in attributing any of the offenses.

The more contact a person has had with a particular group, the less likely he is to attribute a crime to a member of that group (in all cases of “high” contact, the coefficients are negative, thereby indicating a reluctance to attri- bute an offense).

A similar tendency is discernible in the case of the attitude variable: the more positive a person’s attitude is toward a group, the less likely he is to attribute an offense to a person belonging to that group. However, the magni- tude of the effect is not strong in the case of either attitude or contact (that is, the coefficients are closer to zero than those of the preceding two variables of the ethnicity of the stimulus portrait and the ethnicity of the identifier).

INTERACTION EFFECTS

Since the (PEO) relation was found to be the best predictor, an examina- tion of the indirect effects of its components as they apply across offenses seems to be required (the coefficients are presented in Table 3).

In the case of the attribution of the offense of homicide to a portrait, the coefficients in Table 3 are all very low, indicating a weak effect of any specific combination of the ethnicities of identifier and stimulus portrait.

With regard to rape, Sephardics seem to attribute this offense most often to Ashkenazic stimulus portraits (A = 1.09), whereas they tend to refrain from

ETHNICITY AND CRIME ATTRIBUTION 517

Table 3. Effects of Offense Attribution by Ethnicity of Portrait (P), Ethnicity of Identifier (E), Cross- Ethnic Contact (C), and Attitude (A)

A. Main Effects

Portrait Ashkenazic Sephardic Arab Identifier Ashkenazic Sephardic Arab Contact High Low

Standardized Lambda (A) for Relevant Offenses Homicide - Rape Robbery - B&E

- 1.28 -2.24 - 1.45 - 1.85 0.27 -0.74 1.88 0.45 1.01 2.98 -0.43 1.40

-0.83 -0.07 -0.93 0.74

2.99 2.52 3.01 0.86 -2.16 -2.45 -2.08 - 1.50

- 1.37 -0.35 -1.13 -0.19 1.37 0.35 1.13 0.19

Attitude

Negative 1.20 0.74 0.20 1.59

B. Interaction Effects

Ethnicity Of Identifier

Positive - 1.20 -0.74 -0.20 - 1.59

Ethnic Origin Of Portrait By Type Of Offense

Ashkenazic Sephardic

Ethnicity of Identified Portrait Ashkenazic Homicide 0.38 -0.47

Robbery 0.06 -0.64 B&E -0.19 - 1.47

Rape 0.40 1.09

Fraud - 1.36 0.31

Sephardic Homicide 0.57 0.36 Rape 0.3 1 -0.07 Robbery 1.15 -0.52 B&E 0.67 -0.19 Fraud 1.03 0.61

- Arab Homicide -0.95 0.11 Rape -0.71 - 1.02 Robbery -1.21 1.16 B&E -0.48 1.66 Fraud 0.33 -0.91

Fraud -

0.68 -0.34 -0.34

-0.52 -2.35

2.87

-0.93 0.93

-2.20 2.20

Arab -

0.09

0.58 1.66 1.05

- 1.49

-0.93 -0.24 -0.63 -0.48 - 1.64

0.84 1.73 0.05

0.58 - 1.18

attributing this category of crime to Arabs (A = - 1.02); and their likelihood of attributing it to a Sephardic stimulus portrait is close to chance probability (A = -0.07). Arabs, on the other hand, tend to attribute rape to stimulus

518 FISHMAN, RATTNER, AND WEIMANN

portraits of Arabs (A = 1.73), but refrain from attributing it to Ashkenazic portraits (h = - 1.49).

The assignment of armed robbery represents an exceptional instance in which Ashkenazic identifiers exhibit a clear tendency to attribute an offense to the stimulus portraits. Although they attribute robbery to Ashkenazic portraits only randomly, they attribute the same crime to Sephardic portraits with a degree of probability that is beyond mere chance; (A = 1.15) they evidently refrain from attributing this crime to the Arab portraits (h = - 1.21). It is Sephardic identifiers who are inclined to attribute armed rob- bery to the portraits of Arabs (A = 1.16).

In the case of burglary, the stronger coefficients indicate that Sephardics tend to attribute the offense to portraits of Arabs (h = 1.66). Arabs tend also to attribute this crime primarily to the Ashkenazic portraits (h = 1.66), and much less so to Sephardic portraits (A = -0.48).

And finally, in the instance of fraud attribution, Ashkenazic respondents attributed fraud to portraits of Sephardics (h = 1.03), whereas Sephardic respondents refrained from making the same attribution to Ashkenazics. Arab respondents, for their part, seemed to refrain from attributing fraud to Sephardics but willingly attributed it to the Ashkenazic portraits.

These findings can be roughly summed up in a set of predictions based on the magnitude and direction of the lambda values for interaction effects.

Most Likely To Be Assigned Identifier Portrait the Offense of

Ashkenazic

Ashkenazic Sephardic ----> Robbery (h=1.15); Fraud (A= 1.03)

Arab

Ashkenazic ---- > Rape (A = 1.09)

Sephardic Sephardic

Arab ---- > Robbery (h= 1.16); Burglary (A= 1.66)

Ashkenazic ---- > Burglary (h= 1.66); Fraud (1.05)

Arab Sephardic

Arab ----> Rape (A=1.73)

CONCLUSION

The findings of this study have theoretical, methodological, and practical

ETHNICITY AND CRIME ATTRIBUTION 519

implications. On the theoretical level, the findings reveal much about the effect of ethnicity on labeling behavior. The ethnicity of a portrait was the most powerful variable in determining the attribution of offenses; it was the major factor in attributing crimes, even more powerful than either the expressed attitudes of subjects toward other groups or the extent of their con- tact with them. This supports the argument, suggested by the attribution theory, that people have well-established sets of social categories which they use in making attributions both to other people and to situations (Brislin, 198 1). This “routinization” of the process of attribution simplifies encounters with new people and new situations and facilitates the process of giving mean- ing to them. This study highlights the important role that ethnicity plays in the formation of such categorizations.

The impact of ethnicity as a form of categorization clearly reflects the social climate. The authors found that Israeli Arabs were more likely than were other groups to be identified as criminals, and that more serious crimes were attributed to them. At the other end of the spectrum, Ashkenazic Jews were found less likely than the other groups of having a crime attributed to them, and the only crime that was attributed to them was fraud. This pattern of offense assignment does not correspond with the real distribution of crimes in Israel, and is best explained by the stratification of Israeli society. Statistics regarding the ethnic distribution of offenders in Israel are not available. However, it is estimated by the prison authorities that Sephardics constitute more than two-thirds of the total inmate population in correctional facilities and almost 90% of the Jewish inmates. Israeli Arabs who account for less than a third of the total criminal inmate population, constitute about 50% of the homicide offenders, a little over 25% of the armed robbers, 15% of the burglers, and less than 10% of the fraud offenders. The exaggeration and disproportional offense assignment (except in the case of homicide) seem to correspond with the general nature of stereotypes where reality and facts are of little or no relevance to action. But even if the assignments would fit the statistical distribution, the negative stereotype is deduced and attributed also to the individual, regardless of his guilt or innocence. The role of ethnicity in the social structure of Israel has been elucidated in a number of studies (for a review, see Smooha, 1978) which suggest that Ashkenazic Jews are highly ranked socially in Israel relative to Sephardic Jews; and that the Arab minor- ity is perceived as constituting the lowest class in the country. The findings of this study show this social order to be fully correlated with the attribution of crimes to stimulus portraits. The probabilities of attribution are at variance in a strong negative correlation with the ethnic status of the portrait. More- over, the attribution theory points to the importance of in-group and out- group categorization as a basis of evaluations and judgments. The study shows that at least in the case of the two Jewish groups, this distinction affects offense attributions. The members of all three groups were less likely

520 FISHMAN, RATTNER, AND WEIMANN

to assign an offense to persons belonging to their own group, and generally attributed crimes to out-group members. Arabs, however, behaved differ- ently by attributing crimes to Arab stimulus portraits at a rate that far exceeded mere chance probabilities. This finding may indicate the internal- ization of an “inferior” social image to be at work. This process was revealed in several studies which have focused on encounters between majority and minority groups (Ashmore, 1970; Brand et al., 1974). These studies found that members of the lower social groups tend to consider themselves less wor- thy because they adopt the image the norm-setter group has of them.

In regard to the interactions between the ethnicities of the identifier and the identified persons, the findings reveal a number of interesting patterns. Ash- kenazic Jews appeared to make fewer stereotypical attributions. Their assign- ments of crimes to the stimulus portraits were random and “fair” in comparison to the other two groups. Sephardic Jews, who are the middle group among the three groups in the study, appeared to be negatively biased against Israeli Arabs and for the most part positively biased toward Ashkenazics. This finding is consistent with those of various studies of the political attitudes of Israelis in which Sephardic Jews were revealed to be more “hawkish” and anti-Arab than Ashkenazic Jews (Peres, 1976). This is often explained as being the consequence of the special position of Sephardics as a group in Israeli society; they occupy a place between Ashkenazics, who are at the top of the social ladder, and Arabs, who are at the bottom. As a consequence of their middle position, they have developed a positive attitude toward the group at the top, which serves as a positive-reference group, and they have adopted a negative attitude toward the group at the bottom, which serves as a negative-reference group. Here, too, one can observe social cate- gorization at work; it is seen in the way “desirable” and “undesirable” traits are attributed to social groups on the basis of prejudicial social images and stereotypes, and of normative social stratification.

Thus, the process of attributing crimes to faces is affected by ethnicity and intergroup relations in Israeli society. The assignment of crimes to faces is directly related to the ethnic divisions in the country and reflects perfectly the social images that are associated with these divisions in the minds of the iden- tifiers. However, this behavior is by no means confined to Israeli society. There have been a number of studies which have examined the relations between ethnicity and stereotyping in other societies (for a review, see Brig- ham, 1971) and its impact on judgments concerning guilt and innocence (for example, Shoemaker, South, and Lowe, 1973). The main contribution of the present study consists in its highlighting of the correspondence between eth- nic dominance and attribution processes and its examining of the effect of various ethnically based identifier-identified compositions on the attribution of specified crimes.

Concerning the methodological implications, the findings point out the

ETHNICITY AND CRIME ATTRIBUTION 52 1

advantages of studying actual judgments in comparison to expressed atti- tudes. The fact that the reported attitudes were not especially effective in predicting the attribution of crimes to stimulus portraits makes doubtful the validity of such measurement. The situation of actual exposure and decision was found to be more effective and fruitful. The finding that the attributions of respondents clearly matched the ethnic division not only demonstrates the powerful effect of ethnicity, but shows that this factor is best measured by a realistic situation of decision making. As to the measurement of cross-ethnic contact, the findings indicate such contact to have a very weak impact on offense attribution. This should not be surprising in view of the findings of past research. As Brigham and Malpass conclude (1985: 152), “The research to date provides only weak evidence of relationships between cross-race accu- racy and interracial experience.” The findings of this study are in line with this trend and, therefore, offer further encouragement to those who would seek to refine the measurement of cross-ethnic contact (that is, to find meas- ures that are more objective than self-reports), or would even undertake to rethink the issue by focusing on the nature of the contact rather than on its extent.

Finally, in respect to the practical implications of the present study, the findings are of special relevance to situations involving disputed eyewitness testimony and identification in criminal cases. Drawing on a database of nearly 500 cases of wrongful conviction in the United States, Huff, Rattner, and Sagarin (1985) concluded that the single most important factor leading to wrongful conviction was eyewitness misidentification. The revelation of the present study, namely that the attribution of a crime is affected by the ethnic categorization of faces, and its resultant activation of preexisting judgments, would seem to fit in with a whole body of research on the impact of race and ethnicity on misidentification (for example, Loftus, 1979; Malpass and Kra- vitz, 1969; Luce, 1974; Brigham and Barkowitz, 1978; Weimann, Fishman, and Rattner, in press). The findings of this study, in regard to the attribution of criminal offenses, may be further developed and pursued in order to explore the impact such ethnic stereotypes may have on the criminal justice process at the level of suspect gathering and the eyewitness identification and prosecution of suspects.

REFERENCES Alport, Gordon W. and Leo J. Postman

The Psychology of Rumor. New York: Henry Holt. Amir, Yehuda, Shlomo Sharan, Rachel Ben Ari, Ahron Bizman, and Miriam Rimer

Asymetry, academic status, differentiation, and the ethnic perceptions and preferences of Israeli youth. Human Relations 3 1 : 99- 1 16.

1947

1978

522 FISHMAN, RATTNER, AND WEIMANN

Ashmore, Richard D. 1970 Solving the problem of prejudice. In Barry E. Collins (ed.), Social

Psychology: Social Influence, Attitude Change, Group Processes and Prejudice. Reading, MA: Addison-Wesley.

A social distance scale. Sociology and Social Research 17: 265-271. Social Distance. Yellow Springs, OH: Antioch Press.

Bogardus, Emory S. 1933 1959

Brand, Elaine S., Rene A. Ruiz, and Amado M. Padilla 1974 Ethnic identification and preference: A review. Psychological Bulletin 8 1 :

860-890.

Brewer, Marilynn 1977 Perceptual processes in cross-cultural interaction. In David S. Hoopes, Paul

Pedersen, and George Renwick (eds.), Overview of Intercultural Education, Training, and Research. Washington, D.C.: SIETAR (Georgetown Univer- sity).

Ethnic stereotypes. Psychological Bulletin 76: 15-38.

Do they all look alike? The effects of race, sex, experience, and attitudes in the ability to recognize faces. Journal of Applied Psychology 8: 306-318.

The role of experience and contact in the recognition of faces of own-and other-race. Journal of Social Issues 41: 139-155.

Brigham, John C.

Brigham, John C. and Paul Barkowitz

197 1

1978

Brigham, John C. and Roy S. Malpass 1985

Brislin, Richard W. 198 1

Bull, Ray 1982

Cross-Cultural Encounters. New York: Pergamon.

Physical appearance and criminality. Current Psychological Reviews 2: 269- 282.

Bull, Ray and J. Green 1980 The relationship between physical appearance and criminality. Medicine,

Science and Law 20: 79-83.

Dion, Karen 1972 Physical attractiveness of children’s transgressions. Journal of Personality

and Social Psychology 24: 207-2 13.

The multivariate analysis of qualitative data. Journal of the American Statistical Association 63: l,091-l,131. A general model for the analysis of surveys. American Journal of Sociology

Analyzing Qualitative/Categorical Data. Cambridge, MA: Abt Books.

Huff, Ronald C., Arye Rattner, and Edward Sagarin Guilty until proved innocent: Wrongful conviction and public policy. Crime and Delinquency 32: 518-544.

Ethnic Preferences Among Israelis. Tel-Aviv: Cherikover.

Goodman, Leo A. 1970

1972

1978 77: 1,035-1,086.

1986

Jaffe, Eliezer 1984

ETHNICITY AND CRIME ATTRIBUTION 523

Knoke, David and Peter J. Burke 1980 Log Linear Models. Beverley Hills: Sage.

Experimentelle Untersuchungen zur Ausdruckskundemittel photographichs- statistischer Methode. Archiv fur die gesamte Psychologie 114: 55-71.

Eyewitness Testimony. Cambridge, MA: Harvard University Press.

Reconstruction of automobile destruction: An example of interaction between language and memory. Journal of Verbal Learning and Verbal Behavior 13: 585-589.

Kozney, Erik D. 1962

Loftus, Elizabeth F.

Loftus, Elizabeth F. and John C. Palmer

1979

1974

Luce, Terrence S. 1974 Blacks, whites and yellows: They all look alike to me. Psychology Today 8:

106- 108.

Malpass, Roy S . and Jerome Kravitz 1969 Recognition for faces of own and other race. Journal of Personality and

Social Psychology 13: 330-334.

Ethnic Relations in Israel. Tel Aviv: Sifriat Poalim. Peres, Yohanan

1976

Rim, Yehuda 1968 National stereotypes in children. Megamot 16: 45-50.

Interethnic relations and education: An Israeli perspective. In Norman Miller and Marilynn Brewer (eds.), Groups in Contact: The Psychology of Desegregation. New York: Academic Press.

Shoemaker, Donald J., Donald R. South, and Jay Lowe Facial stereotypes of deviants and judgments of guilt or innocence. Social Forces 51: 427-433.

Schwartzwald, Joseph and Yehuda Amir 1984

1973

Smooha, Sami 1978 Israel: Pluralism and Conflict. London: Routledge and Kegan Paul.

Smooha, Sami and Yohanan Peres 1974 Ethnic inequality in Israel. Megamot 20: 5-22.

Racial and ethnic stereotypes. In Arthur G. Miller (ed.), The Eye of the Beholder: Contemporary Issues in Stereotyping. New York: Praeger Publishers.

Stephan, Walter G. and David Rosenfield 1982

Thornton, G.R. 1939 The ability to judge crimes from photographs of criminals. Journal of

Abnormal and Social Psychology 34: 378-383.

The Analysis of Cross Tabulated Data. Chichester: Wiley. Upton, Graham J.

Weimann, Gabriel, Gideon Fishman, and Arye Rattner

1978

In Press Social distance and identification. International Journal of Comparative Sociology.

524 FISHMAN, RATTNER, AND WEIMANN

Gideon Fishman is a Senior Lecturer in the Department of Sociology at the University of Haifa, Israel. His research interests include crime seriousness, juvenile delinquency, violent behavior, and victimology.

Arye Rattner is a Lecturer of Sociology at the University of Haifa. His major research interests include the functioning of the criminal justice system, wrongful conviction, and crime trends.

Gabriel Weimann is a Senior Lecturer in the Department of Sociology at the University of Haifa. His major research interests are in the field of communication and include media effects, terrorism and the media, and violence in the media.

The three authors contributed equally to this article.