Power Point Help

Full Terms & Conditions of access and use can be found at https://www.tandfonline.com/action/journalInformation?journalCode=utsq20

The Sociological Quarterly

ISSN: 0038-0253 (Print) 1533-8525 (Online) Journal homepage: https://www.tandfonline.com/loi/utsq20

“Kick 'Em When They're Down”: Explanations of the Relationship Between Stress and Interpersonal Aggression and Violence

Richard B. Felson

To cite this article: Richard B. Felson (1992) “Kick 'Em When They're Down”: Explanations of the Relationship Between Stress and Interpersonal Aggression and Violence, The Sociological Quarterly, 33:1, 1-16, DOI: 10.1111/j.1533-8525.1992.tb00360.x

To link to this article: https://doi.org/10.1111/j.1533-8525.1992.tb00360.x

Published online: 12 Dec 2016.

Submit your article to this journal

Article views: 5

View related articles

Citing articles: 25 View citing articles

“KICK ’EM WHEN THEY’RE DOWN”: Explanations of the Relationship Between Stress

and Interpersonal Aggression and Violence

Richard B. Felson” State University of New York at Albany

The latest version of the frustration-aggression hypothesis (here, FA) posits that any form of negative affect or distress is likely to increase the likelihood of aggression. Stressful life events could thus produce aggression and violence because they create negative affect. In contrast, the social interactionist (SI) approach interprets many acts of aggression as expressions of grievances and informal social control. Stressful life events cause people to behave in ways that lead others to attack them. This study examines these approaches using (1) an adult sample of ex-criminal offenders, ex- mental patients, and the general population, and ( 2 ) longitudinal data obtained from a national sample of high school boys. The results suggest that being a target mediates the negative life-eventsiaggression relationship, thus favoring SI over FA. Also, anger affects other forms of delinquency as strongly as it affects aggressive behavior, imply- ing that some forms of delinquency, not usually associated with aggression, have aggressive goals.

TWO THEORIES OF AGGRESSION AND VIOLENCE

The Frustration-Aggression Hypothesis

Perhaps the most influential theory of aggression and violence has been the frustration- aggression hypothesis (here, FA). According to the original theory, aggression is likely when an external event impedes some goal-response (Dollard, Doob, Miller, Mowrer, and Sears 1939). This interference produces aggressive energy that is released through ag- gressive behavior toward the frustrating agent or “displaced” against unassociated targets.

Later, this formulation was modified, as it was realized that other responses to frustra- tion exist and that some aggression is instrumental and not related to thwarting (e.g., Miller 1941). Also argued was that frustration only elicits aggression when the external costs are not too great (Berkowitz 1962), when aggression-facilitating cues exist (Berkowitz 1964), and when the frustration is arbitrary or illegitimate (Pastore 1952). This

*Direct all correspondence to: Richard B. Felson. Department of Sociology, State University of New York at Albany, Albany. NY 12222.

The Sociological Quarterly, Volume 33, Number 1, pages 1-16. Copyright @ 1992 by JAI Press, Inc. All rights of reproduction in any form reserved. ISSN. 0038-0253.

2 THE SOCIOLOGICAL QUARTERLY Vol. 33 iNo . 1 /?992

last qualification FA could not easily explain and therefore presented a theoretical challenge.

In the most recent FA formulation, Berkowitz (1989) argues that aversive events in general lead to aggression. He suggests that psychological discomfort, depression, anx- iety, and physical pain, as well as goal-blockage, instigate aggression. Aversive events lead to aggression because they produce negative affect. By substituting aversive condi- tions and negative affect for frustration, Berkowitz explains why arbitrary and illegitimate thwartings are most conducive to aggression: they produce the greatest negative affect. In other words, people who do not receive what they expect experience the most negative affect.

The experimental evidence for FA and Berkowitz’s reformulation is mixed. The evi- dence catharsis can dissipate aggressive energy is mostly negative (see, e.g., Geen 1990 for a review). While some evidence for displacement exists, so do alternative interpreta- tions of the results (Tedeschi and Norman 1985). A number of laboratory studies show that attack is a much stronger predictor of aggression than is frustration (see, e.g., Buss 1963; Geen 1968). Since these studies show small frustration effects, they suggest that it may be of more limited importance than originally thought. Finally, evidence exists that aversive experiences, more generally, can lead to aggression (e.g., Carlson and Miller 1988; see Berkowitz 1989 for a review). However, these studies do not examine whether negative affect-a subjective state-mediates the aversive events/aggression relationship.

Despite weak experimental support, FA has greatly influenced correlational studies of violence outside the laboratory. These studies focus on effects of chronic rather than situational frustration. FA is used to explain the relationship between lynching and the price of cotton in the South (Hovland and Sears 1940), between relative deprivation and political violence (Feierabend and Feierabend 1966), and between income inequality and the homicide rate (e.g., Messner 1982; Krahn, Hartnagel, and Gartrell 1986). Blau and Schwartz (1984), for example, argue that inequality can produce “pent-up aggression which manifests itself in diffuse hostility and violence” (p. 180).

FA also plays a central role in the interpretation of the environmental stressors/aggres- sion relationship. For example, Mueller (1983) suggests that stress affects aggression because it impedes behavior, produces stimulus overload, and creates feelings of an- noyance, irritability, and discomfort. These factors are associated with frustration or negative affect. In addition, he suggests that many stressors produce arousal, a gener- alized drive for behavior. This impacts the predominant response mode, which is some- times aggression. However, Zillman (1983), after reviewing the evidence, suggests a much more limited role of arousal in energizing aggressive behavior.

FA also figures prominently in family violence research. Stan (1988) and O’Leary (1988) review the evidence on the effects of stress on child abuse and marital violence, respectively (see also Linsky and Straus 1986). Most relevant to the present research are studies that find a relationship between stressful life events and child abuse, marital violence, and courtship violence (Straus 1980a, 1980b; Makepeace 1983; Seltzer and Kalmuss 1988; see Farrington 1986 for a review). Levinson and Ramsay (1979) report that assaultive behavior is associated with stressful life events among persons evaluated for involuntary civil commitment. Finally, Masuda, Cutter, Hein, and Holmes (1978) show a similar relationship in a prison population.

These studies measure stress by external events. However, what is stressful for one person may not be for another. Subjective reactions to events are relevant as well as the

”Kick ’em When They‘re Down” 3

events themselves. This is consistent with the stress literature which treats psychological distress as a subjective reaction to stressful events (e.g., Dohrenwend and Dohrenwend 1974). It is also consistent with Berkowitz’s formulation of FA, since distress is another term for negative affect, which, he argues, mediates the effect of aversive events on aggression. Thus, FA can be tested by examining the effects of either external events or the psychological response to them. Previous research neglects subjective states, probably because causal inferences about them are difficult with cross-sectional data.

The Social lnteractionist Approach

What I term the “social interactionist” perspective (here, SI) on aggression emphasizes the role of social interaction, as opposed to individuals’ psychological characteristics (Tedeschi 1970; Tedeschi, Gaes, and Rivera 1977; Luckenbill 1977; Black 1983; Goode 1971; Felson 1978, 1984; Averill 1982, 1983). This approach interprets aggressive behavior as goal-oriented rather than as an involuntary response to aversive stimuli. Aggressive actions seek to compel and deter others, achieve status, and obtain justice, as defined by the actor. For example, Felson (1984) emphasizes social control processes in the sequence of events in aggressive interactions. He finds that these incidents usually begin when someone perceives a rule violation. The social control agent may then accuse or admonish the alleged offender, who is likely to be punished unless he or she gives an account to explain the violation. This punishment constitutes the first attack and often results in retaliation and escalation.2 People retaliate both to deter the antagonist and others from future attacks and to achieve justice. They also retaliate in order to save face or maintain favorable identities (Felson 1978). Thus, evidence suggests that retaliation is more likely when an audience is present (Felson 1982). In general, SI emphasizes the antagonists’ perceptions, third party influence, and the dynamic interchange between participants in an aggressive encounter (Luckenbill 1977). Note that SI as described here only loosely relates to symbolic interaction theory.

Black (1983) takes a similar approach in his discussion of “crime as social control.” He argues that much criminal violence reflects decisions to personally settle grievances (“self-help”) rather than seek the help of the police. While not a surprising interpretation of homicides and assaults, Black argues that much theft and vandalism also reflects grievances. For example, he interprets some thefts between people who know each other as a form of debt collection; however, his only evidence is that offenders sometimes know the targets. His approach, then, has not adequately been tested.

Stress effects on aggression are difficult for SI to explain unless the actor blames and then attacks the “guilty” party for the event. However, many stressful events cannot be blamed on anyone, and many of them are associated with displaced aggression, that is, aggression against targets uninvolved in producing the stress. For example, SI cannot readily explain why a man comes home and screams at his children after a tough day at work. If he has any grievance, it is with someone at work. Therefore SI has difficulty explaining the relationship between most stressful life events and violence, just as it does explaining experimental effects of aversive stimuli on displaced aggression (Berkowitz 1989; however, see Tedeschi and Norman 1985). FA seems to better explain these effects.

One possible S1 explanation is that stressful events indirectly affect aggression through their effect on the target’s behavior. Considerable evidence shows that stress depresses school and work performance (see, e.g., Holmes and Masuda 1974; Motowidlo, Packard,

4 THE SOCIOLOGICAL QUARTERLY Vol. 33/No. 1 il992

and Manning 1986; Cohen 1980). If distressed persons perform less competently, violate expectations, or annoy others, these others are likely to express grievances. This may foster aggressive interactions in which the distressed person is often, initially, the target.3 While either party might aggress first, data suggest that it is more often the aggrieved party (Felson 1984).

Felson (1978) similarly argues that distressed people more often violate rules of defer- ence (see Goffman 1956). A central rule of social interaction prescribes polite and friendly exchange, with a show of mutual support. However, upset people may less competently perform this ritual. They may find it difficult to feign positive emotions, and their mood may reduce interest in showing deference, creating behavior others may consider inap- propriate or even aggressive. If others express a grievance, an aggressive interaction can result.

In contrast to FA, SI does not posit a direct link between most forms of negative affect and aggression. The only type of negative affect with a direct impact is anger, as described by Averill (1982, 1983). His evidence suggests that people become angry when they blame someone for a misdeed. In other words, anger is the subjective response of people when they have a grievance. Thus, it mediates the relationship between provocations and aggression.

PRESENT STUDIES

The present research examines the interrelationship between stressful life events, negative affect, and aggression. The first study focuses on the effects of stressful events; the second, their subjective counterpart-negative affect.

The first study explores competing FA and SI hypotheses. Using cross-sectional data on adults, I examine the effects of stressful life events on aggression and being the target of aggression. FA claims that negative life events affect aggression by first producing nega- tive affect. Stressful events should not affect being a target unless stress results in aggres- sion and aggression leads others to retaliate. Such an indirect relationship might produce a weak relationship between stress and being targeted but that should disappear when aggression is controlled. SI, on the other hand, predicts that negative life events affect the distresseds’ being a target of aggression more than they affect their initiating aggression. Stress leads persons to engage in behaviors that elicit others’ social control. This culmi- nates in aggressive behavior, usually first by the party who feels aggrieved by and targets the distressed person. A relationship may thus result between stress and aggression, but it should be mitigated when being a target is controlled.

The second study explores some additional hypotheses using longitudinal data on high school boys. It does not permit theory testing by examining competing hypotheses, but does allow us to test hypotheses derived from each theory, a weaker, but more common method of theory testing.

These analyses consider whether anger affects delinquent behavior not usually thought to have an aggressive motive. This follows Black’s (1983) assertion that much nonviolent crime expresses grievances against the victim. If delinquency reflects blaming others then anger should affect it. That is, anger should affect theft, vandalism, and deviance in school, as well as more obvious types of aggression. Sometimes these behaviors may express grievances against parents and teachers.

I also examine whether other forms of negative affect or distress affect aggressive

“Kick ‘em When They‘re Down” 5

behavior. According to Berkowitz’s reformulation of FA, any kind of negative affect should result in aggression. For example, anxious or depressed persons should more often aggress. While SI could also possibly account for these effects, this is one of Berkowitz’s central hypotheses and therefore worth testing.

METHODS

Adult Data

The cross-sectional analyses are based on interviews in Albany County, New York, with persons ages 18-65 from three types of populations. First, multistage sampling was used to obtain a representative general population sample (N = 245). Streets in each census tract and then the dwellings on those streets were randomly sampled. Male and female respondents were then chosen in equal number. Second, social clubs for ex-mental patients provided an ex-mental patient sample (N = 148).4 Only those both living in the area and not hospitalized for at least six months in the preceding year were selected. Finally, the sample of ex-criminal offenders (N = 141) includes parolees and local offen- ders contacted by mail, who had been identified by the Department of Corrections as living in the community for at least six months. To obtain more female offenders, we contacted a community day program for women released from the state prison or a local jail.

The questionnaire is part of a larger study of situational factors in aggression and violence. It asks how often in the last year respondents engaged in various aggressive acts against their children, their spouse, other family members, persons they knew, and strang- ers. Each type of target involves an item on screaming and shouting, one on pushing, shoving, and slapping, and one on hitting with a fist or object, with response categories: “never; 1 or 2 times a year; 3 or 4 times a year; twice a month; monthly; weekly.” These are coded 1 through 6 and summed for each target categ01-y.~ The scores for strangers and people known (outside the family) are collapsed because of low frequency.

In addition, respondents are asked how often in the past year they were the target of physical violence-how often someone “pushed or shoved or slapped you?”; “hit you with a fist or an object?”; “threatened you with or hurt you with a gun or knife?” These items, with the same response categories as before, are summed.6

Stressful life events are measured by 26 items from the well-known Holmes and Rahe scale (1967). For the major analyses, these items are summed to form separate scales for negative versus positive life events. I incorporate the latter because they might produce stressful changes that affect aggression.

High School Data

The high school sample is based on longitudinal data from the Youth in Transition project, a well-known data set based on interviews with a nationally representative sample of high school boys (Bachman 1970). 1 use the first two panels from the longitudinal file constructed by Bachman and his colleagues (N = 1886). The first data wave was collected in the fall from tenth grade students, the second wave a year and a half later, in the spring of their junior year.

Causal inference is particularly problematic when interpreting the relationship between

6 THE SOCIOLOGICAL QUARTERLY Vol. 33/No. 111992

subjective states and reported behavior. For example, respondents may judge their level of anger by how aggressive they have been. A longitudinal design is thus particularly useful for these analyses.

Two aggression scales are available on this file. The first, 3-item scale measures verbal aggression and asks how often respondents had fought or argued with students and teachers, or had purposely made the teacher angry.

The second, 8-item scale measures physical violence and asks how often students had threatened or hurt someone, hit their parents or teachers, engaged in gang fights, or used weapons.

Two scales measure delinquency. A 9-item scale indexes theft and vandalism; and a 5- item scale, school deviance, such as cheating on tests, coming late, and skipping class.

Five scales on the longitudinal file reflect various forms of distress or negative affect: a 7-item scale measures anxiety by the extent to which the respondent worries about things; a 5-item scale measures tension by the respondent’s nervousness; a 6-item scale measures depression; an 1 8-item scale assesses somatic symptoms; and a 7-item scale measures anger. The anger scale items ask respondents, for example, whether they lose their temper easily, carry a chip on their shoulder, are easily irritated or annoyed, and feel “like a powder keg ready to explode.” Following Averill’s (1982) research, I see this as a measure of the tendency to perceive events as provocations and respond emotionally to them. The reliability of each distress scale is computed using the tenth grade data. Cronbach’s alpha is .72 for anxiety, .51 for tension, .83 for somatic symptoms, .58 for depression, and .63 for anger.

RESULTS

Adult Data

Table 1 presents means, standard deviations, and zero-order correlations; Table 2 the standardized regression coefficients. To estimate these equations I regressed aggression measures against various targets on negative and positive life events, education, race, sex, age, and dummy variables for ex-mental patients and ex-offenders. The general popula- tion is the omitted dummy category. Race is also a dummy, whites coded 1 and nonwhites zero. Six percent of the nonwhites are Puerto Rican and the rest black. Finally, gender is coded 2 for females, 1 for males.

Regression results suggest that negative life events affect aggression against all targets except children, for whom the effect is opposite that predicted. The effect on spousal aggression is statistically nonsignificant but in the predicted direction. These results generally back FA and previous research. However, respondents with negative life events are also more likely to be the targets of aggression (see Table 2, last line). That negative life events most strongly affect aggression against the respondent is more consistent with SI than FA since the former predicts that being the target of aggression mediates the relationship between stressful events and aggression.

To test FA more directly, I control for being a target when examining negative life events’ effects on aggression. To test SI more directly, 1 control for aggression in an equation in which being the target of aggression is the dependent variable. The other controls are the same as before. Note that the causal interpretation of the relationship between being a target and aggression does not affect the coefficient representing the effect

ro’ 3

Ta bl

e 1

M ea

ns , S

ta nd

ar d

D ev

ia tio

ns ,

an d

Ze ro

-O rd

er C

or re

la tio

ns (A

du lt

Sa m

pl e)

1. A

gg re

ss io

nl C

hi ld

.2

9 .0

9 .0

7 .0

8 -

2. A

gg re

ss io

n/ Sp

ou se

.3

4 .2

6 .4

1 3.

A gg

re ss

io n/

O th

er F

am ily

.SO

.4

2 4.

A gg

re ss

io n/

N on

fa m

ily

.5 6

5. A

gg re

ss io

n/ R

es po

nd en

t 6.

N eg

at iv

e E

ve nt

s 7.

P os

iti ve

E ve

nt s

8. E

du ca

tio n

9. G

en de

r 10

. A ge

11

. R ac

e 12

. E x-

m en

ta l P

at ie

nt s

13 . E

x- of

fe nd

er s

-. 03

.0

6 .1

0 .1

5 .2

3 .1

7 -.1

3 .0

3 .3

3 .2

2 -.

23

-.0 3

.2 9

.3 8

-. I5

-. I7

.3

4 .2

2 -.

I4

-. 04

.3

6 -.1

5 -.

04

-. 02

-.

12

.0 2

-.3 7

.I 2

-. I4

-.

34

-. 02

.0

1 -.

35

-.o o

.1 1

-.3 7

-. 09

-.

04

-. 29

-.0

4 .0

2 -.

26

-. lo

.2

0 -.4

9 -.

I8

-. 01

-.

07

.2 1

-. I3

.0

9 .1

1 .I

0 .I

8 .0

7 .I

4

- .0

2 .3

0 .1

7 .3

3 .2

6 .1

8 .3

3 -.1

8 - .2

2 -.3

1 -.

32

-. 38

3. 18

2.

77

1. 78

2.

97

1. 16

1.

57

I . 92

11

.9 3

1. 43

33

.9 2

.7 9

,2 92

.2

59

3. 41

2.

44

2. 57

3.

90

2. 01

1.

71

1. 67

2.

83

.5 0

13 .3

6 .4

1 .4

55

.4 38

B 3

8 THE SOCIOLOGICAL QUARTERLY Vol. 33/No. 1 /1992

Table 2 Standardized Coefficients Representing Determinants of Aggressive Behavior

Against Various Targets (Adult Sample)

Target of Aggression

Children Spouse Other Family Nonfamily Respondent

Negative Events

- .05 . I2 .20* .13* .25*

Positive Events Education Gender Age

-.07 -.03 .15* -.48* -.08 -.I0 .I0 -.29* -.01 -.23* -.OO -.32*

.17* -.12* -.lo* -.21* - .01 -. 10* .01 -.19*

EX- mental Ex-

Race Patient offender

.13* -.19* -.I4

.10 .07 .28*

.07 . 1 1 * .05

.07 . 00 .17*

.08 .03 .18*

Note: * p < .05.

of life events, which is our interest.’ In these analyses, the aggression measure is based on items assessing physical violence against all targets except spouses and children. The focus on physical violence makes the measure comparable to that for the respondent as target. I ignore aggression against spouse and children since these were nonsignificant in the previous analyses. In addition, this preserves sample size since unmarried or childless respondents can be included.8

Including being a target in the regression equation eliminates the effect of negative life events on aggression (Beta = .04, n.s.). However, the effect of negative life events on being a target is preserved when aggression is controlled (Beta = .17, p < .0001). The pattern of results, then, supports SI and contradicts FA: stressful events’ effect on aggres- sion is apparently indirect, through their effect on being targeted.

Might there be some artifactual explanation of why being targeted, but not being aggressive, relates to stressful events?9 For example, perhaps some response set leads persons to report more negative events and more incidents of being a target, but not incidents in which they aggress. This might also affect reports about witnessing aggres- sion. I examine this by controlling for witnessing aggression, asking respondents how many times in the past year they saw other people “threaten to use or actually use a knife on each other,” “hit each other,” or “scream and shout at each other” (alpha = .76). This reduces the standardized coefficient for negative life events from .25 to .19, still highly significant ( p < .001). This suggests that the relationship between negative life events and being the target of aggression is not artifactual. lo

Another possible interpretation of the stressful life eventdaggression relationship is that certain types of stressful events are the result, not the cause of aggressive behavior. For example, an aggressive person may be more likely to be fired from a job. The relationship could also be spurious, if people with interpersonal problems are more likely to fight with others and to have negative events involving others occur to them. Either of these pro- cesses could make aggression more strongly relate to stressful events that can be influ- enced by how people relate to others. I therefore examine the relationship between aggression and specific life events. Negative life events more likely affected by a person’s aggressive behavior or interpersonal problems are divorce; separation; break-up of a love relationship; and being fired/demoted/changed to a less responsible job. Negative events

“Kick ’em When They’re Down” 9

less likely affected include school or training program failure; movement to a worse neighborhood; widowhood; serious physical illness/injury; loved one’s death or serious injury/physical illness; business failure; lay-off. Partial correlations between aggression and each negative event, with the same controls as before, reveal no evidence that aggression more highly relates to events on the first than second list. Thus no evidence suggests that aggression is more likely to relate to stressful events dependent on the person’s behavior.

Other results in Table 2 should be noted as well. First, there are strong age effects, indicating that aggression is much more frequent among younger respondents. That these results are consistent with numerous other studies suggests that the measures of aggression in this study are valid. Note that younger respondents are also more likely to be targets, presumably because they tend to associate with each other. Second, more education generally reduces involvement in aggressive interactions, either as actor or target. Third, sample differences appear. In general, ex-offenders more frequently aggress. Ex-mental patients are slightly more aggressive against family members, except spouses and chil- dren, but otherwise no more than the general population. I suspect ex-offenders and ex- mental patients are less aggressive against their children because they are less likely to be involved with them. Fourth, consistent with the literature, women more frequently ag- gress against their children (e.g., Straus 1980b), presumably because they are more likely to be the primary caretaker. On the other hand, they less often aggress against persons outside the family. Also, as the literature shows, the frequency of wives’ aggression against their husbands is as great as the converse (see Steinmetz and Lucca 1988). Fifth, only one significant race difference appears: whites are more likely to aggress against their children. Finally, positive life events affect extra-familial aggression but have no other significant effect. l2

No significant interactions between negative life events and the control variables emerge. In particular, no evidence shows that the stress/aggression relationship differs for ex-mental patients, ex-offenders, and the general population, which justifies pooling samples. The generality of these effects is also demonstrated since they appear for three radically different samples.

High School Data

Table 3 presents zero-order correlations, means, and standard deviations for the high school boys, Table 4 the lagged regressions. For the latter, grade 11 behavior is regressed on grade 10 distress or anger and grade 10 behavior. This technique helps control for spuriousness due to response sets and other factors (Kessler and Greenberg 1981). Esti- mates are not biased unless an unmeasured variable affects the independent variable and change in the dependent variable. Note that lagged coefficients are usually small, particu- larly given substantial time lag. Note also the need for caution about attaching importance to statistical significance with such a large sample.

Table 4 shows anger significantly affects both aggression measures and both delinquen- cy measures. The delinquency effects support Black’s contention that some nonviolent criminal behavior reflects grievances. The impacts of other forms of negative affect are slight but 15 out of 16 coefficients are in the predicted direction. For physical aggression and theft/vandalism, 7 out of 8 effects are statistically significant. The evidence suggests that negative affect slightly influences these more serious forms of delinquency.

10 THE SOCIOLOGICAL QUARTERLY Vol. 331No. 1 /1992

Table 3 Zero-Order Correlations, Means, and Standard Deviations (High School Sample)

Time 2 Variables

Verbal Physical School Theft/ Time I Variables Aggression Aggression Deviance Vandalism x SD

Anxiety Tension Somatic Symptoms Depression Anger Verbal Aggression Physical Aggression School Deviance TheftNandalism x SD

. l l

.09

.I7

.12

.29

.47

.33

.28

.26 210.9

64.8

.08

.09

.20

.12

.20

.27

.39

.26

.22 123.8 44.7

.10

.07

.I8

.15

.22

.36

.37

.47

.32 198.7 66.2

. l l

.09 -.I5

.15

.23

.25

.37

.29

.44 140.4 56.5

289.1 264.8 220.3 240.17 261.2 219.4 153.9 180.8 154.1

70.2 75.0 58.9 67.9 63.2 71.2 58.4 64.4 59.0

DISCUSSION

Since its introduction, FA has been criticized on both conceptual and empirical grounds. Such skepticism suggests a need for an alternative explanation of the relationship between aversive events and aggression, which SI provides.

The adult data are consistent with previous studies and FA in showing a correlation between negative life events and aggression. Further, that positive life events are uncorre- lated with aggression reveals that only aversive events are related to aggression. However, negative life events correlate slightly more strongly with being a target than agent of aggression. More important, the effect of negative life events on aggression disappears when being a target is controlled. On the other hand, the effect of negative life events on being targeted is not affected when aggression is controlled. These key results suggest that being a target mediates the stressful life eventdaggression relationship. Apparently the mediating process involves conflict and social control rather than some internal psycho- logical state. l4 This is more consistent with SI than FA. At the very least, these results create skepticism about the usual theoretical interpretation of the stressful life

Table 4 Standardized Coefficients Representing Effects of Distress and Anger (Tl)

on Behavior (T2) and Controlling for Behavior (Tl)

Dependent Variuble Anxiety Tension Symptoms Depression Anger

Verbal Aggression .02 -.01 .01 .03 . lo* Physical Aggression .04 .06* .11* .04* .lo* School Deviance * 02 .01 .01 .03 .lo* TheftNandalism .06* .05* .08* .07* .13*

Note: * p < .05

“Kick ‘em When They‘re Down” 11

eventsiaggression relationship. This skepticism should remain even if the particular Sl explanation here favored proves incorrect.

I argue that it is the behavior of stressed persons that results in their being aggressed against. However, failure to measure such behavior represents a study limitation. Also at issue is whether the objects of the respondent’s attacks are those who attacked the respon- dent, and who first attacked. However, the causal interpretation offered here is the only convincing explanation of why being a target mediates the relationship between stressful events and aggression.

An alternative explanation not tested is that the respondents are themselves blamed for the negative events and then punished. This seems improbable, as the stressful events/aggression relationship is unrelated to whether the event is likely to be affected by the respondent’s behavior. Also possible is that distressed persons are perceived as vul- nerable and safer to attack. However, the opposite argument also obtains: someone upset may seem more dangerous, and “hitting a person when they are down” is antinormative.

I considered other explanations for why stress might directly affect being aggressed against but not aggressing. For example, perhaps the aggression variable has more mea- surement error than the target variable because the former is more socially undesirable, hence underreported. However, that the aggression variable correlates more highly with age than the target variable does suggests this is not the case, since age is the best predictor of aggression in this as well as other studies. It seems unlikely that some response set can explain why stressful events more strongly relate to being a target than to aggression (or to witnessing aggression).

No findings indicate that those with troubled interactions are more likely to experience stressful life events. Were this true one would expect a stronger relationship between aggressive behavior and events a person’s behavior might affect. This pattern is not observed. In addition, evidence not presented suggests that aggression at the earlier time period does not produce later distress.

This study is consistent with research that emphasizes the importance of a target’s behavior and social control processes in instigating violent interactions. For example, many child abuse studies suggest that children with behavioral problems are more iikely to be abused (for reviews of the literature, see, e.g., Kadushin and Martin 1981; Wolfe 1985). Social control processes are also shown important in criminal violence (e.g., Luckenbill 1977; Felson and Steadman 1983; Black 1983) as well as less serious ag- gressive interactions (Felson 1984; Averill 1982, 1983). In addition, research on the source of police brutality emphasizes the officer’s attempt to control the citizen and the citizen’s defiance (e.g., Toch 1969; Westley 1970). Finally, evidence shows that bar violence often begins when bartenders refuse to serve customers (Felson, Baccaglini, and Gmelch 1986). Bartenders’ social control behavior leads patrons-often young and intox- icated-to respond with violence.

Negative Affect

With high school data 1 examined some central hypotheses suggested by FA and Sl. The emphasis on subjective states rather than external events as determinants of aggres- sion responds to a limitation of the life events literature by recognizing that people respond to aversive events differently. The longitudinal design helps disentangle the causal relationships involving these variables. Not surprisingly, the results suggest that anger predicts change in aggressive behavior over time. Also some evidence shows that other

12 THE SOCIOLOGICAL QUARTERLY Vol. 33/No. 1 / I 992

forms of negative affect slightly affect physical aggression and theftivandalism. This accords with Berkowitz’s argument that any form of negative affect produces aggressive behavior, although the weakness of these effects suggests that negative affect figures minimally. Further, the theoretical interpretation of any such effect is unclear. The first study’s analyses suggest that negative affect may lead someone to be the target of aggression.

Thus far I have presented FA and SI as competing perspectives. An alternative ap- proach suggests that both theories are valid but each is only part of the picture. Aggression may reflect social interaction processes in some instances while in others it may result from frustration. In fact, many scholars in the area would take this stance, as indicated by the wide acceptance of the distinction between angry and instrumental aggression. How- ever, the assumptions of these two perspectives are so different that eclecticism may be inappropriate.

Our focus on long-term effects of negative affect and stressful life events is consistent with the correlational literature. However, the short-term situational effects of aversive stimuli observed experimentally might not be strong enough to appear with any consisten- cy in a study focused on individual variation and long-term effects. If situational effects are weak, then individual differences (based on variation in exposure) are likely to be weaker still, particularly given the time lag in longitudinal analyses. A study of short-term effects might show stronger and more consistent effects of negative affect. Note however that angers produces lagged effects. Further, all of the subjective states are fairly stable over the two time periods (all above r = .50), suggesting that treating them as chronic states is reasonable. l 5

While longitudinal analysis of negative affect using survey data has limitations, so does experimental study. The latter requires manipulating affect indirectly, probably impossible without also influencing subjects’ cognitions as well (Neiss 1988). To manipulate situa- tional variables (e.g., the arbitrariness of the frustration) and simply assume that negative affect mediates their effect on aggression is not enough. The theory cannot be discon- firmed if one assumes levels of negative affect on an ad-hoc basis. For this and other reasons, Berkowitz’s FA reformulation is difficult to test.

Anger, Grievances and Rebellion

It is interesting that anger affects change in delinquency over time as strongly as it does aggression. This suggests that some of these acts aim to harm someone who has aggrieved the person. This supports Black’s (1983) contention that grievances are expressed in a variety of ways, including crimes like theft and vandalism, usually associated with other motives. This claim has never been tested, to my knowledge. The targets of this delin- quency could be, for example, teachers, parents, or other students. Unfortunately, the high school data do not specify the target of the anger and aggression. In particular, our measures of anger focus on the respondent’s disposition. However, Averill’s evidence (1982) shows that anger targets individuals. While one can argue that certain people are angry at society or people generally, this may so stretch the approach as to be vague and untestable.

These results suggest an SI interpretation of “rebellious” behavior among youth. At least some acts so interpreted may actually express grievances, that is, be attempts by youths to punish adults they feel wronged by. This accords with the finding that angry youths engage in various forms of delinquency.

“Kick ’em When They‘re Down” 13

Another explanation of why anger affects delinquency as well as aggression is that the anger scale actually measures respondents’ self-control or impulsiveness. Gottfredson and Hirschi (1990) suggest that low self-control is the key individual variable predicting aggression and delinquency as well as other impulsive behaviors. They base their argu- ment on evidence that violent and other criminal offenders rarely specialize in a particular type of offense.

In sum, the evidence suggests that the causal mechanism FA proffers does not explain the relationship between stressful events and aggression. The evidence is more consistent with the idea that persons under stress behave in ways that lead them to be punished, and this results in aggressive interactions. In addition, some delinquent acts appear to involve expression of grievances. These explanations implicate social interaction rather than inter- nal biological or psychological processes in human aggression and violence.

ACKNOWLEDGMENTS

This research was supported by a grant from the National Institute of Mental Health. I wish to thank Nan Lin, Allen Liska, Steve Messner, James Tedeschi, and Stewart Tolnay for their comments on earlier drafts of this article.

NOTES

1. Different theories in this tradition focus on different factors. For example Black (1983) and Felson (1984) stress aggression as social control; Felson (1978) as impression management; Luck- enbill (1977) as a dynamic interchange involving antagonists’ definitions of the situation; and Tedeschi (1970) as a social influence tactic.

Harm-doing tends to be called punishment when approved of, and aggression, violence, or abuse when not. To avoid this value judgment I define aggression as an act that intentionally harms or threatens to harm another, for whatever reason. This includes imprisonment and grounding children, as well as violent crime.

The approach does not assume a grievance is legitimate, only that it is perceived to be so. I use the word “target” rather than “victim” to avoid judgments about blame. This approach does not “blame the victim” since a causal statement does not necessarily imply blame (Felson 1991). People’s characteristics and behavior may affect whether they are targets of aggression yet they are not morally or legally culpable.

Seventeen are from a list released to me by the New York State Department of Mental Hygiene. For more detailed description of the sampling, see Steadman and Felson (1984).

Dollard and associates’ (1939) FA account suggests the different forms of aggression are interchangeable. In general, the relationships here observed are similar for verbal and physical aggression.

Whether it is appropriate to compute reliabilities for these scales is unclear since they involve counts of aggressive behavior. Some items’ distribution makes them resemble check lists, like life events scale items. The basic problem is that some items, particularly those involving hitting with a fist or object, are rare and thus do not correlate highly with other items, lowering reliability measures. For the reader’s information, the Cronbach’s alpha for these scales (correcting for number of items) are .57 for aggression against children; .63 against spouse; .72 against other family members; .77 against persons outside the family; and .63 for physical violence against the respondent.

I assume that the aggressionkarget correlation primarily reflects retaliation: those who have attacked others are more likely to be attacked, and vice versa. However, a correlation could also emerge if aggressive people tend to associate with each other.

2 .

3.

4.

5.

6.

7.

14 THE SOCIOLOGICAL QUARTERLY Vol. 33/No. 1 /1992

8. 9.

Cronbach’s alpha for this scale is .75. The measures of aggression and being a target could correlate because of response sets or

because they tend to occur during the same incidents due to retaliation. However, these relationships would not affect their relationship with stress.

When being a target is controlled, negative life events have a weak effect on witnessing aggression (Beta = .09, p = .02).

A reviewer points out that some stressful events result from being aggressed against (e.g., serious injury) while others can increase the opportunity to be targeted (e.g., moving to a worse neighborhood). However, few aggressive interactions cause serious injury. Further, no evidence shows that events of these types more highly correlate with being a target of aggression.

12. In analyses not presented, the relationship between negative fife events and aggression is unaffected when violent upbringing, alcohol use, and attitudes related to aggression are controlled.

13. Analyses not presented on whether aggression at TI affects anger at T2, controlling for the lagged variable, show that verbal (but not physical) aggression has an effect on reported anger.

14. This interpretation does not depend on the time order of these attacks or on who attacks first. Presumably both the respondent and the antagonist engage in aggressive behavior in many incidents.

15. The correlations between the subjective states over the two time periods are .54 for anger, .52 for somatic symptoms, .50 for tension, .49 for depression, and 3 8 for anxiety.

10.

1 1.

REFERENCES

Avcrill. James R. 1982. Anger and Aggression: An Essay on Emotion. New York: Springer-Verlag. . 1983. “Studies on Anger and Aggression: Implications for Theories of Emotion.” Ameri- can Psychologist 38: 1145-1 160.

Bachman, Jerold G. 1970. Youth in Transition II: The Impact of Family Background and Intel- ligence on Tenth Grade Boys. Ann Arbor: Institute for Social Research.

Berkowitz, Leonard. 1962. Aggression: A Social Psychological Analysis. New York: McGraw-Hill. . 1964. “Aggressive Cues in Aggressive Behavior and Hostility Catharsis.” Psychological Review 71: 104-122. . 1989. “The Frustration-Aggression Hypothesis: An Examination and Reformulation.” Psychological Bulletin 106: 59-73.

Black, Donald. 1983. “Crime as Social Control.” American Sociological Review 48: 34-35. Blau, Peter M. , and Joseph E. Schwartz. 1984. Crosscutting Social Circles: Testing a Macrostruc-

turd Theory of Intergroup Relations. New York: Academic Press. Buss, Arnold. 1963. “Physical Aggression in Relation to Different Frustrations.” Journal ofAbnor-

ma1 and Social Psychology 67: 1-7. Carlson, Michael, and Norman Miller. 1988. “Bad Experiences and Aggression.” Social Science

Research 72(3): 155-157. Cohen, S . 1980. “After-Effects of Stress on Human Performance and Social Behavior: A Review of

Research and Theory.” Psychological Bulletin 88: 82- 108. Dohrenwend, B.S., and B.P. Dohrenwend, eds. 1974. Stressful Life Events: Their Nature and

Ejjfects. New York: Wiley. Dollard, J., L.W. Doob, N.E. Miller, 0. Mowrer, and R.R. Sears. 1939. Frustration and Aggres-

sion. New Haven: Yale University Press. Farrington, Keith. 1986. “The Application of Stress Theory to the Study of Family Violence:

Principles, Problems and Prospects.” Journal of Family Violence 1: 131-147. Feierabend, lvo K. , and Rosalind L. Feierabend. 1966. “Aggressive Behavior Within Polities,

1948-1962: A Cross-National Study.” Journal of Conflict Resolution 10: 249-271, Felson, Richard B. 1978. “Aggression as Impression Management.” Social Psychology 41 : 205-

213.

“Kick ‘em When They’re Down“ 15

. 1982. “Impression Management and the Escalation of Aggression and Violence.” Social Psychology Quarterly 45: 245-254. . 1984. “Patterns of Aggressive Interaction.” Pp. 107-126 in Social Psychology of Aggres- sion: From Individual Behavior to Social Interaction, edited by Amelie Mummendey. Berlin: Springer-Verlag. . 1991. “Blame Analysis: Accounting for the Behavior of Protected Groups.” American Sociologist Spring: 5-23.

Felson, Richard B., William Baccaglini, and George Gmelch. 1986. “Bar-room Brawls: Aggres- sion and Violence in Irish and American Bars.” Pp. 153-166 in Violent Transactions. edited by Anne Campbell and John J. Gibbs. Oxford: Basil Blackwell.

Felson, Richard B., and Henry J. Steadman. 1983. “Situations and Processes Leading to Criminal Violence.” Criminology 21: 59-74.

Geen, Russell G. 1968. “Effects of Frustration, Attack, and Prior Training in Aggressiveness upon Aggressive Behavior.” Journal of Personality and Social Psychology 9: 316-32 I , . 1990. Human Aggression. Pacific Grove, CA: BrooksKole.

Goffman, Erving. 1956. “The Nature of Deference and Demeanor.” American Anthropologist 58:

Goode, William J. 1971. “Force and Violence in the Family.” Journal of Marriage and the Family

Gottfredson, Michael, and Travis Hirschi. 1990. A General Theory of Crime. Stanford: Stanford University Press.

Holmes, Thomas H., and Minoru Masuda. 1974. “Life Change and Illness Susceptibility.” Pp. 45- 72 in Stressful Life Events: Their Nature and Effects, edited by B .S. Dohrenwend and B. P. Dohrenwend. New York: Wiley.

Holmes, Thomas H., and R.H. Rahe. 1967. “The Social Readjustment Rating Scale.” Journal oj Psychosomatic Research 11: 213-218.

Hovland, C., and R. Sears. 1940. “Minor Studies in Aggression VI: Correlation of Lynchings with Economic Indices.” Journal of Psychology 9: 301-3 10.

Kadushin, Alfred, and Judith A. Martin. 1981. Child Abuse: An Interactional Event. New York: Columbia University Press.

Kessler, Ronald C. , and David F. Greenberg. 198 1 . Linear Panel Analysis. New York: Academic Press.

Krahn, H., T.F. Hartnagel, and J.W. Gartrell. 1986. “Income Inequality and Homicide Rates: Cross-National Data and Criminological Theories.” Criminology 24: 269-295.

Levinson, R.M., and G. Ramsay. 1979. “Dangerousness, Stress, and Mental Health Evaluations.” Journal of Health and Social Behavior 20: 178-187.

Linsky, A.S., and M.A. Straus. 1986. Social Stress in the United States: Links to Regional Patterns in Crime and Illness. Dover, MA: Auburn House.

Luckenbill, David F. 1977. “Criminal Homicide as a Situated Transaction.” Social Problems 25: 176- 1 86.

Makepeace, James M. 1983. “Life Events Stress and Courtship Violence.” Family Relations 32:

Masuda, M., D.L. Cutter, L. Hein, and T.H. Holmes. 1978, “Life Events and Prisoners.” Archives of General Psychiatry 35: 197-203.

Messner, Steven F. 1982. “Societal Development, Social Equality, and Homicide: A Cross-Na- tional Test of a Durkheimian Model.” Social Forces 61: 225-240.

Miller, Neal E. 1941, “The Frustration-Aggression Hypothesis.” Psychological Review 48: 337- 342.

Motowidlo, Stephan, John S. Packard, and Michael R. Manning. 1986. “Occupational Stress: Its Causes and Consequence for Job Performance.’’ Journal of Applied Psychology 7 1 : 6 18- 629.

473-502.

33: 624-635.

10 1 - 109.

16 THE SOCIOLOGICAL QUARTERLY Vol. 33/No. 1 i l 9 9 2

Mueller, Charles W. 1983. “Environmental Stressors and Aggressive Behavior.” Pp. 5 1-76 in Aggression: Theoretical and Empirical Reviews, vol. 2, edited by R.G. Geen and E.I. Donnerstein. New York: Academic Press.

Neiss, Rob. 1988. “Reconceptualizing Arousal: Psychobiological States in Motor Performance.” Psychological Bulletin 103(3): 345-366.

O’Leary, K . Daniel. 1988. “Physical Aggression Between Spouses: A Social Learning Theory Perspective.” Pp. 31-56 in Handbook of Family Violence, edited by Vincent B. Van Hasselt, Randall L. Morrison, Alan S. Bellack, and Michel Hersen. New York: Plenum.

Pastore, N . 1952. “The Role of Arbitrariness in the Frustration-Aggression Hypothesis.” Journal ($Abnormal and Social Psychology 47: 728-731.

Sears, Robert R., E.E. Maccoby, and H. Levin. 1957. Patterns of Child Rearing. Evanston, IL: Row Peterson.

Seltzer, Judith A., and Debra Kalmuss. 1988. “Socialization and Stress Explanations for Spouse Abuse.” Social Forces 67: 473-491.

Starr, Raymond H . , Jr. 1988. “Physical Abuse of Children.” Pp. 119-156 in Handbook of Family Violence, edited by Vincent B. Van Hasselt, Randall L. Morrison, Alan S. Bellack, and Michel Hersen. New York: Plenum.

Steadman, Henry J., and Richard B. Felson. 1984. “Self-Reports of Violence: Ex-mental Patients, Ex-offenders, and the General Population.” Criminology 22: 321-342.

Steinmetz, Suzanne K. , and Joseph S . Lucca. 1988. “Husband Battering.” Pp. 233-246 in Hand- book of Family Violence, edited by Vincent B . Van Hasselt, Randall L. Morrison, Alan S. Bellack, and Michel Hersen. New York: Plenum.

Straus, Murray A. 1980a. “Social Stress and Marital Violence in a National Sample of American Families.” Annals of the New York Academy of Science 347: 229-250. . 1980b. “Stress and Child Abuse.” Pp. 86-103 in The Battered Child, 3rd ed., edited by C.H. Kempe and R.F. Helfer. Chicago: University of Chicago Press.

Tedeschi, James T. 1970. “Threats and Promises.” Pp. 155-192 in The Structure ofConjlicr, edited by P. Swingle. New York: Academic Press.

Tedeschi, James T., J. Gaes, and A.N. Rivera. 1977. “Aggression and the Use of Coercive Power.” Journal of Social Issues 33: 101-125.

Tedeschi, James T., and Nancy M. Norman. 1985. “A Social Psychological Interpretation of Displaced Aggression.” Advances in Group Processes 2: 29-56.

Toch, Hans. 1969. Violent Men: An Inquiry into the Psychology of Violence. Chicago: Aldine. Westley, W.A. 1970. Violence and the Police: A Sociological Study of Law, Custom and Morality.

Cambridge: M.I.T. Press. Wolfe, David A. 1985. “Child-Abusive Parents: An Empirical Review and Analysis.” Psychologi-

cal Bulletin 97: 483-496. Zillman, Dolf. 1983. “Arousal and Aggression.” Pp. 75-101 in Aggression: Theoretical and

Empirical Reviews, vol. 1, edited by R.G. Geen and E.1. Donnerstein. New York: Academic Press.