You are on page 1of 14

Journal of Pcnonality and Social Piycholoty 1984, Vol. 47, No.

2, 263-276

Copynihi 1984 by the American Piychological Ajsooauon, Inc.

Impact of Newspaper Crime Reports on Fear of Crime: Multimethodological Investigation


Linda Heath
University of Minnesota In this study, the impact of components of newspaper crime reports on fear of crime was investigated. Key newspaper components were based on theorizing in the areas of downward comparison, attribution, and normative behavior. In Study 1 (afieldquasi experiment) 36 newspapers were classified according to the proportion of crime reports that involved local crimes, sensational crimes, and random crimes. The levels of fear of crime among a random sample of readers of these newspapers (N = 335) were assessed through telephone interviews. Respondents who read newspapers that printed a high proportion of local crime news reported higher levels of fear if the crimes were predominantly sensational or appeared to be random, whereas respondents whose newspapers printed a low proportion of local crime news reported lower levels of fear if the crimes were predominantly sensational or random (p < .05). Thesefindingswere replicated in Study 2 (a laboratory experiment). College students (N = 80) who read reports of local crimes expressed higher levels of fear if the crimes were random rather than precipitated, whereas the opposite pattern emerged in regard to nonlocal crimes (p < .05). Implications for psychologists and journalists are discussed.

In the late 1960s, social policymakers and researchers realized that the problem of crime was compounded by the problem of fear of crime. People perceived crime as even more prevalent than it actually was and consequently experienced fear that was disproportional to their actual danger (Erskine, 1974). The mass media were identified as a primary force (along with direct and indirect experience) in shaping people's views of crime (Dominick, 1978; Quinney, 1970), and research on the media flourished. Most of the media research focused on effects of televised crime, although a substantial amount examined effects attributable to newspaper presentations of crime. Researchers documented how much attention was given to crime in the newspapers (e.g., Cirino,

Thisresearchwas supported by National Science Foundation Grant DAR 8011225 and by a University of Minnesota Graduate School Grant. The author would like to thank Amy Blythe, Teresa Crimmins, Alison Langley, and Kristen Ragozzino for assistance in data collection and Ellen Berscheid, Eugene Borgida, Richard Petronio, and Mark Snyder for helpful comments on earlier drafts of this manuscript. Requests for reprints should be sent to Linda Heath, Department of Psychology, Loyola University of Chicago, 6S2S North Sheridan Road, Chicago, Illinois 60626.

1974; Quinney, 1970; Ryan & Owen, 1976), compared the amount of crime presented in newspapers with crime rates reported by the Uniform Crime Reports (UCR) (e.g., Jones, 1976), and investigated the effect of newspaper crime reporting on perceptions of crime (e.g., Baker, Nienstedt, Everett, & McCleary, 1983; Davis, 1951; Gordon & Heath, 1981). Most of the previous research on the effects of newspaper crime coverage on perceptions of crime has shared the same conceptual biasit is based on global rather than component analysis of crime reports. That is, researchers have treated crime news as a largely unidimensional entity, ignoring differences in degree of detail, style of presentation, and relevance of the crime for the average reader. Newspaper crime coverage has most often been operationalized as the number of crime stories (e.g., Cirino, 1974), the amount of space devoted to crime stories (e.g., Otto, 1962), and the percentage of the news devoted to crime (e.g., Gordon & Heath, 1981). Some researchers have made gross distinctions between types of crimes, differentiating, for example, between violent and property crimes (e.g., Cirino, 1974; Heath, Gordon, & LeBailly, 1981). Others have noted the presence or absence of accompanying graphics and the prominence of display

263

264

LINDA HEATH

within the newspaper (e.g., Gordon & Heath, 1981). But even these finer distinctions do not allow discrimination between newspaper accounts of a sensational gangland slaying and a random streetcorner murder. Nor can these analyses differentiate between reports of crimes that occurred in the newspaper's locality and those that happened thousands of miles away. Global analysis of newspaper crime coverage entails the assumption that all crime reports are equally fear provoking. Component analysis, on the other hand, is based on the assumption that detail and perspective in crime articles can influence the level of fear aroused by reading about crime. This assumption is tested in this article by an analysis of three components of newspaper crime coverage: sensationalism, apparent randomness of victim selection, and crime location. These three key components are based on the social psychological theorizing concerning attribution of control, normativeness, and downward comparison. Attribution ofcontrol. People have a general tendency to prefer to be in control of their environments, in areas as diverse as noise (Glass & Singer, 1972), bureaucracy (Glass & Singer, 1972), crime (Janoff-Bulman, 1979; Medea & Thompson, 1974), spinal injury (Bulman & Wortman, 1977), and plant watering (Langer & Rodin, 1976; Rodin & Langer, 1977). Although there is some individual variation in the extent to which people perceive themselves as being in control (Rotter, 1966), having control and perceiving control (at least in situations involving little or no disconfirmation of control) are generally adaptive for intellectual (Glass & Singer, 1972) and emotional (Brown & Heath, in press; Heath, 1980; Janoff-Bulman, 1979; Taylor, 1982) functioning. Walster (1966) posited that people engage in self-protective attributions to convince themselves that they will not be victims of negative events. According to her formulation, in threatening situations, onlookers can gain control by attributing responsibility for the outcome to some characteristic or action of the victim. Then, onlookers can convince themselves that they are not like the victims or would not take similar actions and are therefore safe from harm. Without information concerning victim characteristics or pre-

cipitating actions, onlookers cannot engage in self-protective attributions. Newspapers differ in the amount of information they provide concerning crime victims, particularly in regard to precipitating factors involving the victims. The presence of precipitating factors indicates that actions taken by the victim increased the likelihood that he or she would become a crime victim, not that the victim caused the crime or deserved to be victimized. The total lack of victim precipitation indicates that the victim was chosen virtually at random. A crime that occurred during our data collection period illustrates differences between various newspapers' attentiveness to precipitating details: A man and a woman were accosted and severely beaten in their home in a middle-class neighborhood. No robbery was committed, and the assailants were still at large. One local newspaper ended the account there. The other newspaper added the information that the male victim had recently been indicted on several counts of promoting juvenile prostitution. In the first account, the crime appeared to be random and lacking motivation. The second account included a possible precipitating factor and made the crime seem less random, providing readers with the information necessary for self-protective attributions. From the perceived control theoretical framework, the following hypothesis is derived:
HI: The absence of information about precipitating events in newspaper crime reports will be associated with high levels of fear among readers.

Normativeness. Criminal acts are violations not only of our legal code but also of the normative code that governs behavior in society. Criminal acts vary in the number of social norms they violate and the severity of the violations. Infanticide, for example, violates both the norms against murder and also the norms that dictate protection and nurturance of the young. Trying to put one's murdered wife down the garbage disposal violates not only the norm against murder but also the norm that proscribes desecration of the dead. Sensational crimes achieve their spine-chilling quality from either severe violations of social norms or from violations of deeply engrained social norms.

MEDIA CRIME AND FEAR

265

Sensational crimes are also nonnormative in the sense of being unusual or unexpected, both factors that contribute to the memorability of events (Hastie, 1981; Taylor & Crocker, 1981). Although the exact cognitive process that leads to increased recall of distinctive, unexpected events is not clearly understood, reports of bizarre, violent crimes should be better remembered than reports of mundane crimes, and this memorability should, according to previous research (Tyler, 1980), lead to perceptions of crime danger in the environment. Clearly, sensational crimes contain many confounded dimensions, ranging from unexpectedness to extremity of concern with sex and violence. All of these confounded dimensions, however, could lead to greater impact on perceptions of crime. These various theoretical statements concerning normativeness all lead to the following hypothesis:
H2: The prevalence of sensational crimes in newspaper reporting will be associated with high levels of fear among readers.

porting, resulting in increased coverage of local crimes because of the greater accessibility of neighbors, lovers, and families of local suspects and victims. Still other staffs decide to devote major resources to other sections of the newspaper, which could result in greater reliance on the less expensive wire (and therefore nonlocal) crime information. These forces result in substantial variation in the proportions of local crime coverage, as will be detailed later in this article. The crime environment in other locales can appear worse not only in regard to number of crimes (evidenced by the proportion on nonlocal crimes reported in the newspaper) but also in regard to the type of crime that occurs in other places. Environments that are characterized by crime that is sensational and/or random could be deemed worse than environments with nonsensational, nonrandom crimes. The worse the crime environment appears to be elsewhere, the safer a reader should, according to downward comparison theory, feel in his or her own environment. This leads to the last hypothesis:
H3: Factors that increase fear when occuring in the immediate locale (i.e., randomness and sensationalism) decrease fear when they occur elsewhere.

Downward comparison. Downward comparison theory (Wills, 1981) posits that in negative situations people can reduce anxiety and increase their subjective well-being by comparing themselves with others in more dire straits. Although people prefer comparisons with similar others in positive situations (Festinger, 1954), comparisons with dissimilar others are preferred in negative situations (Brickman & Bulman, 1977; Brickman & Campbell, 1971). The reduction in tension and anxiety associated with such downward comparisons has been demonstrated in negative situations rangingfromreceiving electric shock (e.g., Zimbardo & Formica, 1963) to doing poorly on a test (e.g., Brickman, 1975) to contracting cancer (e.g., Taylor, 1982). Newpaper crime reports afford opportunity for downward comparisons in regard to crime victimization, because newspapers present not only reports of local (and therefore somewhat similar) crime victims, but also reports about nonlocal crimes. The ratio of local to nonlocal crime reports in a newspaper is largely the result of editorial decisions rather than local crime environment. Some editorial staffs openly decide to downplay local crime news. Other staffs decide to "featurize" crime re-

These hypotheses will be examined with data from two sources: a quasi experiment based on a telephone survey of readers of newspapers that vary in the proportion of news about local, random, and sensational crimes, and a laboratory experiment. Study 1: Nonequivalent Control Group Quasi Experiment Method
Overview. Perceptions of crime among readers of newspapers with different crime reporting styles were examined within a reversed treatment, nonequivalent control group design (Cook & Campbell, 1979).' Newspapers were classified through content analysis into a three-way factorial design according to their proportion of local, random, and sensational crime reports. The perceptions of crime among readers of these newspapers were assessed through a telephone survey.

1 This design differs from classical experimental design because respondents are not randomly assigned to condition, resulting in potential nonequivalencies between respondents in various conditions.

266

LINDA HEATH Interviews were conducted primarily during evenings and on weekends to obtain more equal samples of women and men. Potential respondents were asked to participate in a half-hour interview about media habits and views on crime. Interviews were completed only for respondents who read either a target newspaper or no newspaper at all. Respondents who read both target papers in cities with two local newspapers were included only if they clearly read one of the papers as a primary paper and the other paper only occasionally. Twelve respondents were sampled from each target city with one local newspaper, and 18 respondents were sampled from cities with two local newspapers. In all, 372 interviews were completed.3 Respondents who read no local newspaper comprised 10.2% of this sample, and their data were removed from these analyses (resulting N = 335). Women were overrepresented in our sample, comprising 67% of the respondents who were newspaper readers. Respondents ranged in age from 16 to 83, with the median age at 36 years. All income levels and education levels were represented, with high school graduates representing 42% of the overall sample as well as the median education level. Most respondents were either employed (42%) or homemakers (29%), with the rest being students (8%), retired (14%), unemployed (6%), or disabled (1%). Respondents with household incomes under $20,000 per year comprised 48% of the sample. Experimenters. Interviews were conducted by a team of 7 female undergraduates who had been trained in telephone survey techniques. Interview order was balanced so that each interviewer conducted an equal proportion of interviews for all target cities. Materials. A standard interview schedule was constructed for this study. Key variables assessed newspaper habits (e.g., frequency, time, percent of newspaper read, identification of primary and secondary newspapers), perceptions of crime, victimization experiences (e.g., violent and property crimes committed against self or others close to respondent), and demographics (e.g., sex, age, income, education). Three different aspects of fear of crime were measured: fear in neighborhood, fear downtown, and fear for teen. Previous research has shown that fear levels vary depending on location (Skogan & Maxfield, 1981)andonthepreceived vulnerability of the target (Gordon, Riger, LcBailly, & Heath, 1980). Consequently, measures of these various aspects of fear were included in the survey. The items measuring fear of crime were preceded by the lead, "Now I'd like to ask you some questions about crime in (target city)." The perception of neighborhood safety was measured by the following item: "How safe do you feel out on the streets in your neighborhood alone at night? Very Safe, Somewhat Safe, Somewhat Unsafe, or Very Unsafe?" In addition, respondents' perceptions of safety in the downtown areas of their cities were assessed
2 These three variables were coded by two independent raters. Coding disagreements occurred on fewer than two percent of the items. In cases of disagreement, the coders conferred with the author to determine the appropriate code for the variable. 3 Completion rate among potential respondents was 61%. The demographics for this sample match the overall demographics for the cities we contacted, except for an oversampling of women.

Target newspaper selection. All crime articles that appeared in the local press of 42 American cities during one constructed week (i.e., one random Monday, one random Tuesday, etc.) over a one-month period were content analyzed. Cities were roughly matched on size, minority composition, socioeconomic status, and crime rate. Twenty of the cities had two local newspapers; 22 cities had only one local newspaper In all, 62 newspapers (1,926 articles) were content analyzed. On the basis of this content analysis, newspapers were classified according to the percentage of crime articles dealing with local crime, random or nonprecipitated crime, and sensational crime. Locale refers to geographic location of the crime inception, with local crimes including those in the city of publication, surrounding suburbs, and small towns within the newspaper's local distribution area. Randomness refers to the lack of information that indicates the victim took some action that made him or her more vulnerable to criminal victimization. For example, if the newspaper article stated that the victim was in a dangerous place (e.g., a bar renown for brawls), engaged in a dangerous occupation (e.g., drug dealer, Mafia member, night clerk at a convenience store), or took a dangerous action (e.g., hitchhiking, leading a protest), the newspaper article was coded as indicating some degree of victim precipitation. If no victim precipitation was evident, the crime was coded as nonprecipitated or random. Sensationalism refers to sensationalism intrinsic to the crime rather than sensationalistic reporting style. Sensational crimes are those that are extremely violent and/or bizarreJ Newspapers were classified into a three-way factorial framework, consisting of the Locale, Randomness, and Sensationalism factors. Each factor has two levels, derived through median split Newspapers from cities with only one local press were excluded if scores on two or more factors fell within one percentage point of the median for those factors. Six cities with one local newspaper did not meet this criterion and were excluded. In addition to the above criterion, newspapers from cities with two local newspapers were also excluded if both newspapers fell within the same cells for the Sensationalism and Randomness factors. Twenty newspapers (representing ten cities) did not meet these criteria. Table 1 presents the 36 newspapers that met the criteria for inclusion in the final sample, accompanied by each newspaper's classification on the Locale, Randomness, and Sensationalism factors. Participants Respondents for the telephone survey were selected through a variant of Random Digit Dialing. Telephone numbers were selected from telephone directories for each target city. These numbers were increased or decreased by increments of one until one interview was completed. Because each respondent was identified by a separate randomly generated base number, this procedure results in a random sample of respondents. This sampling method has the advantage of including respondents with unlisted numbers, while being more cost efficient than straight Random Digit Dialing. The disadvantage of this method is that areas of a city that received telephone numbers sequentially (for example, a new housing development that was composed of several housing units that opened simultaneously) and that has a high percentage of unlisted numbers would be underrepresented in the sample. Because cities rather than the more recently developed suburbs were the target areas, this risk was considered to be minimal.

MEDIA CRIME AND FEAR by the following: "How about downtown in your city at night? How safe do you feel when alone at night out on the street? Would you say you feel Very Safe, Somewhat Safe, Somewhat Unsafe, or Very Unsafe?" Finally, in order to tap views about general safety, as opposed to personal safety, we also asked "How safe do you think a teenage girl would be out alone on the streets of your city at night? Very Safe, Somewhat Safe, Somewhat Unsafe, or Very Unsafe?"

267

in large part, on the absence of other variables that co-vary with the independent variables under consideration. Three types of possible confounding variables were examined in relation to the newspaper stylistic components: demographic variables, victimization variables, and number of crime stories. Demographic variables examined for possible confounding relations included sex, age, Results and Discussion income, and education. Age was found to be Examination of possible confounding vari- totally unrelated to the newspaper component ables. The interpretability of results from a variables, so any age effect on perceptions of nonequivalent control group design depends, crime would be equally distributed across cells
Table 1 Newspaper Identification and Classification on the Locale, Sensationalism, and Randomness Variables
City Akron, OH Charleston, WV Charleston, WV Chattanooga, TN Chattanooga, TN Cleveland, OH Cleveland, OH Dallas, TX Dallas, TX Denvei; CO Denver, CO Flint, MI Ft. Worth, TX Fremont, CA Fresno, CA Grand Rapids, MI Jersey City, NJ Las Vegas, NV Las Vegas, NV Lubbock, TX Meridian, MS Nashville, TN Nashville, TN Omaha, NE Peoria,IL Rockford, IL San Bernadino, CA Scranton, PA Scranton, PA Tacoma, WA Toledo, OH Tulsa,OK Tulsa,OK York, PA York, PA Youngstown, OH
1

Newspaper Journal Daily Mail Gazette Times Free Press Plain Dealer Press Times Morning News Post Rocky Mt. News Journal Star Argus

%Locala 19 (low) 27 (low) 22 (low) 18 (low) 20 (low) 55 (high) 53 (high) 29 (low) 15 (low) 50 (high) 20 (low) 22 (low) 33 (low) 12 (low) 41 (high) 35 (high) 59 (high) 44 (high) 28 (low) 44 (high) 53 (high) 52 (high) 24 (low) 17 (low) 47 (high) 45 (high) 31 (low) 46 (high) 35 (high) 55 (high) 33 (low) 30 (low) 26 (low) 38 (high) 33 (low) 57 (high)

% Sensational" 52 61 43 60 37 77 61 48 53 44 35 58 38 50 67 49 27 35 48 48 21 29 42 41 25 41 68 30 50 13 32 58 21 54 47 (high) (high) (low) (high) (low) (high) (high) (high) (high) (high) (low) (high) (low) (high) (high) (high) (low) (low) (high) (high) (low) (low) (low) (low) (low) (low) (high) (low) (high) (low) (low) (high) (low) (high) (high)

% Random" 81 (high) 76 (high) 77 (high) 79 (high) 59 (low) 61 (low) 72 (high) 58 (low) 75 (high) 58 (low) 69 (low) 79 (high) 63 (low) 76 (high) 68 (low) 69 (low) 80 (high) 70 (low) 64 (low)

Number of crime articles

23 27 21 33 46 23 23
44

30 34 32
31 31 43 46

Bee
Press Journal Review

39
75 41

Sun
Journal Star Banner Tennessean World Herald Star Star

65 (tow)
85 92 67 78 79 78 74 67 100 65 33 57 58 79 100 57 (high) (high) (low) (high) (high) (high) (high) (low) (high) (low) (low) (low) (low) (high) (high) (low)

36 42
19

Sun Tribune
Times News Tribune Blade Tribune Daily World Dispatch Daily Record Vindicator

15 (tow)

29 31 54 32 42 36 32 23 34 29
49

24 28 25 32

Note. The words in parentheses indicate each newspaper's classification on each variable. Median = 33; range = 12-59; n (low) = 19; n (high) = 17. "Median = 44; range = 13-77; n (low) = 17; n (high) = 19. 'Median = 70; range = 33-100; n (low) = 18; n (high) = 18.

268

LINDA HEATH

in the major analyses, perhaps increasing the error term but not biasing the results. Sex, however, was significantly related to the locale variable, F\l, 302) = 3.95, p < .05, with women being overrepresented among readers of newspapers with high proportions of local crime reports (71% women in the high locale, compared with 63% women in the low locale cell). Uncorrected, this pattern could result in an overrepresentation of the level of fear among readers of newspapers with high proportions of local crime reports (because women generally report higher levels of fear of crime than men do). Therefore, sex was entered as a variable in the newspaper analyses. Income level of respondents snowed a significant relation to the interaction between the sensationalism and randomness components of newspaper style, F\l, 302) = 4.16, p < .05. Although none of the comparisons of individual cells in this analysis reached standard levels of significance, the general pattern indicated that among readers of newspapers with a lot of local coverage of crime, higher levels of sensationalism were related to lower income levels among readers. Education showed a marginally significant relation to the interaction between the randomness and locale newspaper components, F\ 1,326) = 3.84, p < .051, although, again, none of the comparisons between individual cells was significant. The general pattern here indicated that newspapers that present a lot of local, random crime are read by people with lower education levels than are newspapers with a lot of local, nonrandom crime. This relation is problematic because previous research has shown that people with less education report higher levels of fear than do more educated people, and, consequently, this pattern could introduce bias in the hypothesized direction. Even given the marginal significance of this pattern, we adopted the conservative approach and included education (based on a median split) as a factor in our analyses of newspaper component effects. Victimization variables were examined for possible confounding effects because some previous research has found crime victimization to be related to fear of crime (although this relation is not consistently found). Our target cities were roughly matched on crime rate based on the UCR data. Because the UCR

data are not perfect indicators of actual crime rate, we also collected victimization data during our telephone survey. Respondents were asked about serious and less serious (e.g., theft) crimes that had happened to themselves or to people close to them. Serious victimization to either the respondent or to someone close to the respondent was totally unrelated to the newspaper variables, as was less serious victimization to the respondent. Less serious victimization to others, however, was significantly related to all three main newspaper component variables. The significant relation with the randomness variable, F{1, 326) = 6.64, p < .05, indicated that people who knew victims of less serious crimes were more likely to read newspapers with low proportions of random crimes than were people who did not know crime victims. The relation with sensationalism, F{\, 326) = 5.69, p < .05), indicated that people who knew crime victims were likely to read newspapers that printed less sensational crime news, and the relation with locale, F{1, 326) = 8.90, p < .005) revealed that people who knew victims of less serious crimes were more likely to read newspapers that presented little local crime news than were people who did not know victims. Because previous research has presented conflicting results concerning the relation between victimization and fear, and because this victimization measure is both indirect (that is, victimization of another rather than self) and concerned with less serious crimes, the relationship between the victimization measure and fear of crime in this sample was analyzed. No significant relations were revealed between this victimization variable and the main fear of crime variables in these data (fear downtown, f(326) = .88, ns; fear for teen, f(329) = .59, ns; fear in neighborhood, 4330) = .61, ns). The lack of relation between this victimization variable and the fear measures indicates that this variable would not produce bias in the main analyses. The final variable examined for possible confounding effects was the number of crime stories published in the newspaper. On the basis of the content analysis data, number of stories does not correlate significantly with any of the three main newspaper component variables (number of stories, sensationalism, r =

MEDIA CRIME AND FEAR

269

.20, p> .10; number of stories, locale, r = .03, p > .40; number of stories, randomness, r = .17, p > .15). Analysis of variance with number of stories as the dependent variable showed no main effects for newspaper stylistic components, but significant 2-way interactions among all the stylistic components did emerge, 7=1(1, 327) = 4.92, p < .05 (Sensationalism X Locale), F{\, 327) = 23.39, p < .005 (Sensationalism X Randomness), F{\, "ill) = 19.11, p < .005 (Locale X Randomness). Although the significance level on these interactions might be inflated because number of stories is a newspaper-level rather than respondent-level variable, causing nonrandom error to be introduced into the data, these relations are still problematic. Number of stories (based on a median split) was therefore entered as a factor along with the other stylistic component variables in analyses of all the fear of crime variables. In no instance did the main effect for number of stories or any interaction involving number of stories reach the standard level of significance. Differences among the number of crime articles appearing in different types of newspapers is therefore less plausible as an alternative explanation for results obtained in the main analyses. Newspaper stylistic component effects on perceptions of crime. Five-way ANOVAS were conducted on the fear of crime variables,4 with the three component variables (Randomness, Sensationalism, and Locale) and two demographic variables (Sex and Education) included as independent variables. The three fear of crime variables included fear in neighborhood, fear downtown, fear for teen.5 Concerning how safe the respondents felt on the streets in their own neighborhoods at night, analysis indicated that women are more fearful than men are in their own neighborhoods, fU, 305) = 30.56, p < .001 (Ms = 2.25 for women and 1.58 for men, with a high score indicating high fear). Additionally, less educated respondents reported higher levels of fear than did more educated respondents, F\ 1, 305) = 4.41, p < .05 (Ms = 2.19 for less educated and 1.85 for more educated respondents). None of the hypothesized main effects or interactions involving newspaper variables was statistically significant. These analyses replicated previous findings regarding demo-

Table 2 Mean Levels of Fear Downtown by Randomness and Locale


Local Random 3.19 (.94) (77) Nonrandom 2.97 (1.10) (87) 3.11 (.98) (81) 2.73 (1.13) (82) Nonlocal

Note. Fear downtown has a 4-point range, with a high score indicating high fear level. SD and cell ns in parentheses.

graphic differences in perceptions of crime and indicate that factors other than newspaper presentations of crime are responsible for levels of fear of crime in immediate residential areas. Perceptions of safety on the streets in the downtown areas at night also showed a sex difference, F{1, 301) = 75.15, p < .001, with women reporting higher levels of fear than men (Ms = 3.33 for women and 2.32 for men). In addition, the interaction between the Randomness and Locale variables, F{1, 301) = 4.61, p< .05 indicates that newspaper stylistic components, although not influencing neighborhood fear levels, do influence the level of fear associated with downtown areas (see Table 2 for cell means). Cell-wise comparisons indicate that, as predicted, an above median
4 Four- and five-way interactions were not computed because cell ns fell below acceptable limits for these analyses. This reduces the number of F ratios generated per analysis and thus reduces the possibility of Type I error using uncorrected F values. 5 Thefirstorder correlations among these variables are as follows: fear in neighborhood, fear downtown, r = .46, p < .001; fear in neighborhood, fear for teen, r = .38, p < .001; fear downtown, fear for teen, r = .62, p < .001. These three items were designed to tap different aspects of fear and are therefore presented separately rather than combined into a scale. Analysis of an additive scale composed of these three fear items, however, reveals the same pattern of interaction between locale and randomness, f{\, 292) = 6.76, p < .01) and locale and sensationalism, F(l, 292) = 4.48, p < .05) as do the fear downtown and fear for teen items.

270

LINDA HEATH

proportion of newspaper articles about local crimes was associated with higher levels of fear among readers if most of the crimes were presented as being random rather than precipitated, F[ 1,301) = 5.73, p < .05. Among readers of an above median proportion of nonlocal crimes, however, random crimes were associated with less fear than were nonrandom crimes, F(\, 301) = 5.62 p < .05. As predicted, aspects of crimes that are frightening when they occur in the immediate locale are reassuring when they occur elsewhere. No other cell-wise comparisons resulted in significant differences. As with randomness, the effect of sensationalism also depends on the locale in which the crime occurred, F{\, 301) = 6.83, p < .01 (see Table 3). The decomposition into simple main effects indicate that, as predicted, reading about sensational crimes is more frightening if the crimes occurred in the reader's locale than if the crimes occurred elsewhere, F{1, 301) = 10.36,/? < .005. Reading articles about nonsensational crimes, however, is associated with lower levels of fear if the crimes occurred in the immediate locale than if they occurred elsewhere, F{1, 301) = 4.54, p < .05. Finally, as predicted, readers of newspapers that present mainly articles about sensational crimes that happened elsewhere exhibit less fear than do readers of articles about nonsensational crimes that happened elsewhere, F(l, 301) = 9.69, p < .005. Again, the worse things are elsewhere, the better we feel about our immediate environment. In order to unconfound personal vulnerability from perception of the crime environment, we asked respondents how safe they thought a teenage girl would be alone on the streets downtown in their cities at night. The main effect for sex, F{1, 304) = 22.24, p < .005 indicates that women judge the teenage girl to be less safe than men do (Ms = 3.52 and 3.07, respectively, with higher scores indicating less safety). The interaction between the locale and randomness variables, F\l, 304) = 8.25, p < .005, indicates that, as hypothesized, reading about random crimes that occurred elsewhere is associated with less fear than reading abut nonrandom crimes that occurred elsewhere, f{ 1,304) = 11.64, p < .005. Additionally, there is a marginally significant simple effect indicating that random crimes

Table 3 by Mean Levels of Fear Downtown the Sensationalism and Locale


Local Sensational 3.12 (1.03) (74) Nonsensational 3.03 (1.04) (90) Note SD and cell ns in parentheses. 3.17 (.94) (71) 2.73 (1.12) (92) Nonlocal

occurring locally are more frightening than precipitated crimes that occur locally, F(l, 304) = 3.54, p < . 10 (see Table 4 for means). In addition, analysis of the fear for teen variable revealed an unpredicted interaction between sex and sensationalism, P(l, 304) = 8.32, p < .005. Female respondents who read newspapers that presented a high proportion of sensational crime news reported higher levels of fear than did women who read newspapers with less sensational crimes news, F{1, 304) = 9.10, p < .005 (Ms = 3.61 and 3.43, respectively). For men, however, the pattern was reversed. Men who read newspapers that presented a high proportion of sensational crime news thought that a teenage girl would be safer on the streets than did men who read newspapers with less sensational crime news, Fll, 304) = 9.15, p < .005 (Ms = 2.89 and 3.27, respectively). The most pronounced cellwise difference exists between men and women who read newspapers with high proportions of sensational crime news, with women judging the teenage girl to be much less safe than the men do, F{1, 304) = 37.87, p < .001. Perhaps the use of female interviewers, combined with the higher level of fear among women and the use of a female target person, lead to his pattern. Were this pattern obtained for a personal fear variable, we could invoke some self-defensive dynamic to try to account for it. Because this variable concerns safety of a hypothetical teenage girl, however, we have neither a theoretical base nor adequate data to explain this puzzling finding. Overall, these findings indicate that com-

MEDIA CRIME AND FEAR

271

Table 4 Mean Levels of Fear for Teen by Locale. and Randomness


Local Random 3.42 (.78) (76) Nonrandom 3.31 (.87) (87) 3.55 (.70) (82) 3.22 (.83) (85) Nonlocal

These findings further indicate that crime reporting is not a unidimensional entity that can be adequately captured by simply counting or measuring crime articles. Even when the amount of crime information is controlled, variations in levels of sensationalism, victim precipitation, and, most important, crime locale are associated with variations in fear level. Study 2: The Laboratory Experiment Although quasiexperimental research done in field settings generally has greater external validity than laboratory experiments, the fear that lurks in every quasiexperimentalist's heart is that some unaccounted for variable is producing spurious effects. To allay this fear, a laboratory experiment was conducted in which the same conceptual variables were examined under controlled conditions.

Note Fear for teen has a 4-point range, with a high score indicating high fear. SD and cell ns are in parentheses.

ponents of newspaper crime articles do not influence fear of crime in the neighborhood but do influence fear associated with downtown areas. These analyses revealed the hy- Method pothesized interactions among newspaper Participants. Forty female undergraduates components in relation to fear of crime. Spe- from the University of male and 40participated in this exMinnesota cifically, the effects of the randomness and periment as an option in their introductory psychology sensationalism evident in newspaper reports course. Materials Fictional crimes were constructed that either of crime depend on the locale in which the to be random or precipitated and that were crime occurred. Among newspapers that re- appearedor distant from the respondent's social andeither close to geoport a high proportion of local crime, the effect graphic spheres, creating a 2 X 2 factorial design. We of randomness matches predictions from per- expanded the locale variable from the quasi experiment ceived control formulations. That is, random to include both social and geographic distance for several predictions crimes are scary. Without some indication of reasons (theoretical and pragmatic). First, should hold derived from downward comparison theory for victim precipitation evident in the newspaper comparisons with others who are socially as well as geoarticle, readers cannot convince themselves graphically removed from the respondent The worse things that they could not also be victimized by such are for others who are different from the respondent socioeconomic status, a crime. This high level of fear is evident both (whether the difference is based onrespondent should feel age, or geography), the better the in regard to personal fear and fear for others about his or her own situation. Second, we could not plau(i.e., a hypothetical teenage girl). sibly maintain that we had access to records and interview When newspaper articles concern crimes information from crimes that occurred at substantial disstates). Finally, because the underthat occur elsewhere, however, the tables are tances (i.e., in other is composed largely of commuters graduate population turned. As we predicted based on downward from the metropolitan area, we could not be certain that comparison theory, the worse crimes seem to locations within the metropolitan areas were, in fact, distant be elsewhere, the better things look in the im- from the respondents' homes. Two types of information about each crime were premediate environment. Reading about a lot of nonlocal random crimes is associated with sented to the respondents: police incident reports and copies The lower levels of fear (both personal and general) of pages ostensibly from a reporter's notepad. with police incident reports were official looking forms inforthan is reading about nonlocal precipitated mation concerning the date of incident, incident, report, crimes. And, although respondents found and response times (in military hours); decription of the reading about sensational crimes frightening incident, names and addresses of victims, suspects, and witnesses; and officers' names, badge numif the crimes happened near home, reading bers, and unit the respondingincrease the versimilitude of numbers. To about sensational crimes that happened some- this form, we included instructions regarding other forms where else was associated with lower levels of (with three letter and eight number identifiers) that were to be completed if the incident involved physical injury fear.

272

LINDA HEATH given a short survey of all three areas so we can make sure that there are not drastic differences among the general opinions of the people who get the three kinds of information. Respondents were then taken to individual research rooms and asked to complete a survey with five items concerning the environment, five concerning crime, and five concerning the budget. On completion of this survey, respondents were given the materials for the condition to which they had been "randomly" assigned. In fact, all respondents received materials for newspaper articles relating to crimes. The purported materials concerning the budget and the environment were included to allow us to collect pretest information concerning perceptions of crime in order to test the adequacy of our randomization procedure. When the respondents had completed the research materials, they were fully debriefed and asked not to discuss the study with others.

or more than one witness. In addition, the street addresses and phone numbers of suspects, victims, and witnesses were blacked out on the research materials. The second source of information concerning each crime consisted of copies of 5 X 7 notebook sheets containing handwritten notes concerning the crime. The notes for each crime were in different handwriting. The notes basically restated and elaborated on the information in the police incident report. Six basic crime scenarios were constructed. Each scenario was then altered to fit the four conditions. For example, one scenario involved a couple who was assaulted in a parking lot after an evening's entertainment The alleged assailant was either unknown to the couple (random condition) or had been involved with the couple in an altercation over a spilled drink earlier in the evening (precipitated condition). In addition, the entertainment was either a dinner theaterfrequentedby middle-aged clientele in a distant suburb (distal condition) or a rock concert near the university (proximal condition). Each respondent was presented descriptions of six crimes, representing one of the four factorial cells. Following each set of crime information, the respondent completed a one-page information rating form that assessed how informative each type of information had been, which facts were most important for understanding the crime, and the merit of including information gained from the interview. This form was included to ensure that respondents carefully read all the information presented in the crime reports. The dependent measures were presented following the six crime scenarios. First, respondents completed a Current Mood Assessment Scale that consisted of eight 9-point bipolar scales. The key scales (Fearful/Safe, Relaxed/Nervous, Agitated/Calm, Upset/Relieved, Angry/Not Angry) were interspersed with filler items (e.g., Tired/Energetic, Amused/Somber). Respondents then replied to general items about perceptions of crime. Procedure. Participants were scheduled in small groups (ranging in size from 2 to 5). A female experimenter met with the group and delivered the cover story script and the instructions. In order to ensure that respondents read the materials carefully, the study was presented as an evaluation of the importance of personal interview information in news reporting. Experimenters delivered the following introductory speech: Journalists have two main sources of information for stories they write; official public statements and personal interviews. The interviews are very expensive, so we'd like to find out how effective interviews are and whether they are necessary. You will be rating different types of information concerning the environment, the Minnesota budget, and crime, and also rating the effects of this information on people. The information you will read is from official public documents and from interviews that took place in April of this year. None of these stories ever made it into the paper, so you most likely haven't heard about them. We will divide you randomly into three groups. Onethird of you will be reading environmental information, one-third will be reading crime information, and onethird will read about the Minnesota budget Before you begin reading the newspaper information, you will be

Results6 and Discussion Mood scale analysis. Included in the Current Mood Assessment Scale that respondents completed immediately after rating the crime reports werefiveitems that related to general emotional arousal. Only three of these items, however, produced a fear scale with adequate reliability (Safe/Fearful, Calm/Agitated, Not Angry/Angry; a = .77). A 2 X 2 X 2 (Randomness X Proximity X Sex) ANOVA was performed, with the three6 Adequacy of randomization. The gain from a trueexperimental (as opposed to a quasi experimental) design is that the randomization procedure allows the researcher to presume that the treatment groups are comparable prior to the manipulation. Not wanting to be presumptuous, we examined the pretest scores for possible differences among respondents in the different conditions prior to our manipulation. Five items concerning perceptions of crime were included in the pretest materials. These items concerning perceptions of crime were included in the pretest materials. These items assessed general risk of victimization, personal risk of victimization, personal safety in the neighborhood and downtown, and the extent to which crime is perceived to be a problem locally. Five 2 X 2 X 2 (Randomness X Proximity X Sex) ANOVAS were performed with pretest items serving as the dependent variables. Sex of respondent was significantly related to responses concerning personal risk, F{\, 72) = 17.38, p < .001, and personal safety, both downtown, f{l, 72) = 24.86, p < .001, and in the neighborhood F[\, 72) = 16.96, p < .001, with women expressing greater fear and risk than men. Neither of the newspaper variables entered significant (or even marginally significant) relations with pretest perceptions of crime. From this, we conclude that the randomization was successful and the different treatment groups were comparable in regard to crime perceptions prior to the manipulation.

MEDIA CRIME AND FEAR

273

item Current Mood Assessment Scale serving as the dependent variable. The predicted twoway interaction between randomness and proximity was marginally significant, F\\, 72) = 3.88, p = .053. Given the lack of dennitiveness of the above analysis, we created a composite of the two mood items that correlated most highly (Safe/Fearful and Not Angry/Angry). Entering this composite as the dependent variable, we found that women reported marginally higher levels of fear and anger than did men, F(\, 72) = 3.23, p < .10 (Ms = 7.48 and 5.98, respectively). Additionally, the significant interaction between randomness and proximity, JFU, 72) = 4.65, p < .05, revealed that respondents who had rated scenarios that presented crimes that were apparently random and that happened to people socially and geographically close to the respondents reported marginally higher levels of fear and anger than did respondents who rated crimes that were close but not random, F(\, 72 = 2.87, p < .10, or random but not close, F(l, 72) = 2.83, p < .10).7 (See Table 5 for means). This essentially replicates the findings from Study 1. Crimes that happen to people who are socially and geographically close to us are more frightening if they are random than if they are precipitated. Random crimes that happen to people distant from us are much less frightening. General crime perceptions. The Crime Perception Survey included both items about personal safety and more general items about perceptions of crime. Responses to the personal safety items (such as, "How safe do you feel out on the streets in your neighborhood alone at night?") appeared to be unaffected by the newspaper component manipulation. Only a sex difference consistently emerged from analysis of the safety items, with women reporting higher levels of fear than men. Responses to more general crime perception items, however, were affected by characteristics of the crime reports. Respondents were asked to indicate whether they thought crime in thencity was a big problem, somewhat of a problem, or not much of a problem. A three-way ANOVA (Proximity X Randomness X Sex) revealed once again that women reported higher levels of concern than men, F\l, 72) = 5.75, p < .05 (Ms = 2.10 and 1.87, respectively). In addition, the significant interaction between

Table 5 Mean Levels of Fear and Anger by Randomness and Proximity Proximal Random 7.75 (3.58) Nonraodom 5.75 (3.89) 7.50 (3.93) 5.90 (3.57) Distal

Note. High scores indicate high levels of fear and anger. Cell ns = 20. SD in parentheses.

randomness and proximity, F\\, 72) = 5.87, p < .05, indicated that respondents who had read about random crimes that happen to victims socially and geographically close to themselves rated crime as more of a problem than did respondents who read about crimes that were close but not random, or random but not close, Fs(l, 72) = 4.60, 5.11,/? < .05, respectively. In addition, as predicted, respondents who rated scenarios about random crimes that were distant thought the local problem was less acute than did respondents who read about precipitated distant crimes, F\\, 72) = 5.1 \,p < 0.5. Finally, respondents who rated scenarios concerning precipitated crimes rated local crime as less of a problem if the crimes were proximal rather than distal, F(\, 72) = 4.60, p < .05, (see Table 6). Respondents indicated their agreement or disagreement with the following assertion "Crime is a random event. It can happen to anyone." Respondents who had rated scenarios concerning random crimes were more likely to agree with the assertion than were respondents who had rated scenarios involving precipitated crimes, F(\, 71) = 4.55, p < .05 (Ms = 4.00 and 3.69, respectively, with a high score indicating agreement). Also, respondents who rated crimes that were socially and geographically proximal were more in agreement than respondents who had rated distal crimes,
7 The fear and anger items correlate .55 (p < .001) and, analyzed separately, produce the same interactive pattern as the combined analysis at marginal levels of significance (for fear, F{\, 72) = 3.70, p = .059; for anger, I\l, 72) = 3.62, p = .061).

274
Table 6

LINDA HEATH

Mean Assessment of Crime Problem by Randomness and Proximity


Proximal Random 2.10 (.45) (20) Nonrandom 1.95 (.22) (19) 2.10 (.55) (20) 1.80 (.41) (20) Distal

and geographic terms) and reassuring when they are distal. These patterns are revealed in both mood measures and measures of general crime environment perceptions. In regard to the general nature of crime, men appear to be willing, given the appropriate cues, to see crime as nonrandom, whereas women appear more likely, given appropriate cues, to see crime as random. General Discussion Two general conclusions can be drawn from these findings. First, social psychological theories fare pretty well in the mass media world. All three theoreticalframeworksused to identify key media components in this research received support from thefindings.As downward comparison theory predicts, comparison of one's own situation with that of someone in more dire straits results in more satisfaction with the immediate situation. The more newspapers print articles about criminals in other places running amok, picking victims at random, and trampling social norms, the more secure readers feel in their own environments. In essence, readers like the grass to be browner on the other side of the fence, and the browner the better. Far from being frightening, reports of grizzly, bizarre crimes in other cities are reassuring. Readers are still exposed to some

Note. High scores indicate that crime was rated as a bigger problem. SD and cell ns are in parentheses.

F\l, 71) = 5.01, p< .05 (Ms = 4.03 and 3.48, respectively), and women were more in agreement than men, F{1, 71) = 4.56, p < .05 (Ms = 4.00 and 3.49, respectively). These three main effects, however, were subsumed under a three-way interaction, F{1, 71) = 7.26, p < .01, (see Table 7 for means). Men who had rated scenarios that presented precipitated, distal crimes disagreed more strongly with the assertion than did men who rated scenarios about precipitated proximal crimes or random, distal crimes Fs(l, 71) = 9.24 and 9.89, p < .01, respectively. A different pattern is apparent among women, howeyer. Women who rated scenarios concerning random, proximal crimes were marginally more likely to agree with the assertion than were women who read about Table 7 precipitated, proximal or random, distal Mean Endorsement of Crime As a Random Event crimes, i=s(l, 72) = 2.82 and 3.56, p < .10, by Sex, Randomness, and Proximity respectively. That is, men appear to be most Male Female influenced by reading about crimes that portend safety (i.e., precipitated, distal), whereas Proximal Distal Proximal Distal women appear to be most influenced by reading about crimes that portend danger (i.e., Random random, proximal). 3.80 . 3.90 4.60 3.70 These findings indicate, as did the survey (1.32) (1.10) (.70) (.95) (10) findings, that media effects are not simple ef- (10) (10) (10) fects. The proximity of the crime victim to Nonrandom the respondent interacts with the apparent randomness of the crime in determining mood 3.90 3.89 2-40 3.80 (1.43) (.63) (1.10) and general perceptions of crime. As predicted, (1.17) (10) (10) (10) (9) factors that are associated with increased fear when the crime is close at hand are reassuring Note. A high score indicates strong agreement with the when the crime is distant. Random crimes are assertion. frightening when they are proximal (in social SD and cell ns are in parentheses.

MEDIA CRIME AND FEAR

275

reports of crimes that occur locally, but the severity and outrageousness of such crimes appear to be judged in comparison with crimes from other places. When the crimes are occurring on the local turf, however, the tables are turned. As predicted from perceived control theorizing, readers do not appreciate criminals choosing their victims at random (or, at least, media accounts that make it appear so). Reports of crimes that lack rhyme or reason are frightening. If the victim apparently did nothing to precipitate the crime, then the reader can do nothing to avoid the crime. If, on the other hand, the victim took some action that made him or her more vulnerable to the victimization, then the reader can avoid that action and presumably remain safe. And,finally,the predictions regarding nonnormative behavior are confirmed. Reports of bizarre, violent crimes are frightening if the crime occurred locally. This fear is not attributable to sensationalistic reporting style on the part of the newspaper, but, rather, to the sensationalism intrinsic to the crime itself. The unexpected, the quirky, the heinous crimes that are reported in newspapers increase fear of crime among readers in that crime locality, even if the reporting style itself is nonsensational. A second general conclusion that can be drawn from this research is that not all crime reports are equally fear provoking, which has conceptual, methodological, and practical implications. Conceptually, researchers need to pay attention to aspects or components of crime reports, rather than treating crime news as a unidimensional entity. Methodologically, researchers need to be aware that global operationalizations of newspaper crime reporting (i.e., simple counts or measurements of crime news) can lead to the erroneous conclusion that newspaper crime reports do not influence fear of crime. In this research, for example, the number of crime stories printed by the newspapers showed no relation to readers' perceptions of crime, although components of the crime reports did. And, finally, on the practical level, policy implications derived from analysis of the components of crime coverage rather than crime coverage in general would probably be more useful and palatable to news personnel. Analyses of the components of crime coverage allow journalists options for

altering crime reporting without sacrificing it entirely. Consequently, component analysis is more likely to be attended to and acted on than is global analysis. (See Cook, Kendzierski, & Thomas, 1983, for a similar analysis in regard to research on television and aggression.) The implications of these research findings for journalists are three-fold. First, decisions concerning the mix of local and nonlocal crime news should be made carefully, because the relative proportions of local and nonlocal news influences the impact of these stories on readers. In addition to being readily available and relatively inexpensive, wire news reports mollify the fear aroused by reports of local crimes. Second, the inclusion of information concerning precipitating actions taken by the victim, although more difficult to obtain and verify, reduces the fear aroused in readers, particularly in regard to local crimes. Of course, the decision to include such information should take into account factors other than fear potential (e.g., harm to victim resulting from publication of such information), but the exclusion of such information in regard to local crimes causes readers to experience high (and probably unwarranted) levels of fear. Finally, the sensational aspects of local crimes should be presented with caution, because reports of sensational local crimes arouse fear among readers. The effects of reports of sensational local crimes are not removed by reporting information concerning victim precipitation. Nonlocal, sensational crimes, on the other hand, are reassuring to readers regarding the safety of the local crimes environment and could be used to offset the frightening effects of local sensational crimes. Overall, the current research shows that details in newspaper articles do make a difference in the effect of the article on perceptions of crime. Similar effects emerged in both the naturalistic setting (with all its problematic third variable explanations) and the laboratory setting (with its tight control and weaker manipulation). The theoretical formulations (attribution of control, social comparison, and norm violations) proved useful for selecting key variables and, in turn, received support from thefindings.In sum, psychological theory can contribute to media research, and media messages can be used to test psychological theory.

276
References

LINDA HEATH toward and behaviors concerning rape. Unpublished doctoral dissertation, Northwestern University, Evanston, Illinois. Heath, L., Gordon, M. T, & LeBailly, R. (1981). What newspapers tell us (and don't tell us) about rape. Newspaper Research Journal, 2, 48-55. Janoff-Bulman, R. (1979). Characterological versus behavioral self-blame: Inquiries into depression and rape. Journal ofPersonality and Social Psychology, 37, 17981809. Jones, E. T. (1976). The press as metropolitan monitor. Public Opinion Quarterly, 40. 239-244. Langer, E. J., & Rodin, J. (1976). The effects of choice and enhanced personal responsibility for the aged: A field experiment in an institutionalized setting. Journal of Personality and Social Psychology, 34, 191-198. Medea, A., & Thompson, K. (1974). Against rape. New York: Farrar, Straus & Giroux. Otto, H. A. (1962). Sex and violence on the American newsstand. Journalism Quarterly, 40, 19-26. Quinney, R. (1970). The social reality of crime. Boston: Little, Brown. Rodin, J., & Langer, E. J (1977). Long-term effects of a control-relevant intervention with the institutionalized aged. Journal of Personality and Social Psychology, 35, 897-902. Rotter, J. B. (1966). Generalized expectancies for internal versus external control of reinforcement Psychological Monographs. 80 (Whole No. 609). Ryan, M., & Owen, D. (1976). A content analysis of metropolitan newspaper coverage of social issues. Journalism Quarterly, 53. 634-640. Skogan, W. G., & Maxfield, M. G. (1981). Coping with crime: Individual and neighborhood reactions. Beverly Hills, CA: Sage. Taylor, S. E. (1982). Social cognition and health. Personality and Social Psychology Bulletin, 8, 549-562. Taylor, S. E., & Crocker, J. (1981). Schematic bases of social information processing. In E. T. Higgins, C. P. Herman, & M. P. Zanna (Eds.), Social cognition. The Ontario symposium, (Vol. 1, pp. 89-134), Hillsdale, NJ: Erlbaum. Tyler, T. (1980). Impact of directly and indirectly experienced events: The origin of crime-related judgments and behaviors. Journal of Personality and Social Psychology. 39. 13-28. Walster, E. (1966). Assignment of responsibility for an accident. Journal of Personality and Social Psychology, 3. 73-79. Wills, T. A. (1981). Downward comparison principles in social psychology. Psychological Bulletin, 90, 245-271. Zimbardo, P. J., & Formica, R. (1963). Emotional comparison and self-esteem as determinants of affiliation. Journal of Personality, 31, 141-162.

Baker, M. H., Nienstedt, B. G, Everett, R. S., & McCleary, R. (1983). The impact of a crime wave: Perceptions, fear, and confidence in the police. Law and Society Review. 17, 319-335. Brickman, P. (197S). Adaptation level determinants of satisfaction with equal and unequal outcome distributions in skill and chance situations. Journal of Personality and Social Psychology, 32, 191-198. Brickman, P., & Bulman, R. J. (1977). Pleasure and pain in social comparison. In J. M. Suls & R. L. Miller (Eds.), Social comparison processes: Theoretical and empirical perspectives (pp. 149-186). Washington, DC: Hemisphere. Brickman, P., & Campbell, D. T. (1971). Hedonic relativism and planning the good society. In M. H. Appley (Ed.), Adaptation-level theory (pp. 287-302). New York: Academic Press. Brown, S., & Heath, L. (in press). Coping with critical life-events: An integrative cognitive-behavioral model for research and practice. In S. Brown & R. Lent (Eds.), Handbook of counseling psychology. New York: Wiley. Bulman, R. J., & Wortman, C. (1977). Attributions of blame and coping in the "real world": Severe accident victims react to their lot. Journal of Personality and Social Psychology, 35, 351-363. Cirino, R. (1974). Power to persuade. New York: Bantam. Cook, T. D., & Campbell, D. T. (1979). Quasi-expenmentation. Design and analysis issues for field settings Chicago: Rand McNally. Cook, T. D., Kendzierski, D. A., & Thomas, S. V. (1983). The implicit assumptions of television research: An analysis of the NIMH report on television and behavior. Public Opinion Quarterly, 47, 161-201. Davis, F. (1951). Crime news in Colorado newspapers. American Journal of Sociology, 57, 325-330. Dominick, J. R. (1978). Crime and law enforcement in the mass media. In C. Winick (Ed.), Deviance and mass media (pp. 105-128). Beverly Hills, CA: Sage. Erskine, H. (1974). The polls: Fear of violence and crime. Public Opinion Quarterly, 38, 131-145. Festinger, L. (1954). A theory of social comparison processes. Human relations, 7, 117-140. Glass, D. G, & Singer, J. E. (1972). Urban stress New York: Academic Press. Gordon, M. T, & Heath, L. (1981). The news business, crime, and fear. In D. Lewis (Ed.), Reactions to crime (pp. 227-250). Beverly Hills, CA: Sage. Gordon, M. T, Riger, S., LeBailly, R. K., & Heath, L. (1980). Crime, women, and the quality of urban life. Signs: Journal of Women in Culture and Society, 5, (3, Suppl.), S144-S160. Hastie, R. (1981). Schematic principles in human memory. In E T. Higgins, C. P. Herman, & M. P. Zanna (Eds.), Social cognition. The Ontario symposium, (Vol. 1, pp. 39-88). Hillsdale, NJ: Erlbaum. Heath, L. (1980). A multi-methodological examination of the effects of perceptions ofcontrol on women's attitudes

Received September 26, 1983 Revision received February 6, 1984

You might also like