Violent Men in Intimate Relationships: An Analysis of Risk Markers

~--~ ~ ,...

itit.Jiit.i

II

ViolentMenIn IntimateRelationships:An Analysisof Risk Markers1

DAVIDB.SUGA~Rhode Island College

GERALDT. HOTALING

Dtptzrtmmt of CrimiPurJ JusticeUni'Dmity 0{ LDwtlI

In the family vio1enceliterature, a number of risk markers associated withmen's use of violence against women have been identified. Using the 1975Na-tional FamilyViolenceSurvey,a multivariate analysis was performed to examinewhich risk markers best clifferentiatedamong nonvi01entmen. verbally aggres-sive men, men who exhibit minor physical violence, and eevere1yphysicallyviolent men. High levels of marital a>nflic:t.lower socioeconomic status, andgreater exposure to violent role models in the home of origin emerged as primarycorrelates of levels of violence between intimates. Theoretical and re!le8n:him-plications are discussed.

Over the past two decades, a considerable amount of research hasbeen conducted on a wide number of risk markers associated with the

use of violence by men against women in conjugal relationships. In thiscontext, a risk marker refers to an atb"ibute or exposure that is as-sociated with an increased probability of involvement in spousalviolence (Last, 1983).It is a predictor or correlate of this vioIenceand notan outcome of violence.

Recently,Hotaling and Sugannan (1984)~sed the knowledge baseon risk markers in this area examining only case-control or contrastedgroup studies in evaluating risk. This review revealed six characteristicsof male perpetrators, five characteristics of the women whom they vic-timized, and eight characteristics of the marital relationship as consis-tent risk markers in predicting men's use of violence toward women.

I

.,I

IAn earlier draft of this paper was presented at the Eastern Psychological As!Iodationmeetings, New York, NY, April. 1986. This research was supported under Contract No. 299-84-C755 from the Department of Health and Human Services through the ViolenceEpidemiology Branch at the Centers for Di.seJIseControl, Atlanta. GA and by the NationalInstitute of Mental Health Grant No. MH 15161-07 through the University of NewHampshire. The authors are also indebted to M. A. Straus, D. Finkelhor, L. Williams, G.Kantor. J. Suitor, J. Ellison and G. king for CI01J'Iments01\ earlier drafts of this paper. 'Iheviews and conclusions ~~~ in this paper are !Io1e1ythose of the authors and do notnecessarily reflect the views 01 any of the above-mentioned organizations or individuals.

2c:0rresp0ndence conceming this article should be sent to the first author at the Dep8rt-ment of Psychology, Rhode Island College, Providence, RI az908.

1034

Journal of Applied Social Psychology, 1989, 11,12, pp. 1034-1048.Copyright e 1989 by V.H. WlRSton &Son, Inc. ABrights reserved.

-~ ---

- ~ ~. _._, ~=,.T ~ --~.~~._~. - ~------ n__~ -- -----

~~-=-" - -~---.:.

--.;

- - -- --,

VIOLENT MEN 1035

Battering men were more likely than nonviolent men to have lower self-esteem, lower income levels and lower occupational status, to more fre-quently abuse alcohol, to have been physically abused as a child, and tohave witnessed parental violence while growing up. Battered womenwere also more likely to have lower self-esteem,to have more traditionalsex role expectations, to use drugs more often, to have been physicaUyabused by their parents, and to have witnessed physical violence be-tween their parents while growing up than nonvictimized women. Fin-aUy, regarding dyad characteristics, physically violent men were inintimate relationships marked by higher levels of marital conflict ormaladjustment, higher levels of both educational and religious incom-patibility, lower family income levels, higher levels of verbal abuse, andan increased likelihood that the couple is separated or divorced thantheir nonviolent counterparts.

Hotaling and Sugannan's (1984) review points out a number ofproblems with the family violence literature. First, a number of theoriescan account for risk markers because of the unclear operationalization ofcurrent theoretical constructs. For example, the psychiatric model (Faulk,1974;Gandolf, 1985;Hamberger & Hastings, 1986)assumes that certaininternal and stable traits of the man predisposes him to engage in violentbehavior toward his partner. 1bese traits can be attributed to either in-herited factors (i.e., temperament) or experiential factors (i.e.,past learn-ing contingencies or socialization processes). The assumption of thismodel is that the locus of causation for the person's behavior is internaland the result of factors in the individual's past Oearly, the lower self-esteem of the battering man and his past history of experiencing and wit-nessing violence in his family of origin would support this model.However, the last two markers (witnessing and experiencing violence inthe family of origin) can also be accounted for by a social learningperspective (Peterson, 1980). Th~ model argues that husband-to-wifeviolence results primarily because of the past exposure to the vicariousreinforcement of violent role models as well as the present reinforcement(or nonpunishment) of the individual's own violent behavior. Essentially,the individual not only witnessed his father physically aggress againsthis mother and be reinforced for this behavior (the individual's mothergave into the father's demands), but his own violent behavior goes un-punished and is reinforced by his partner's surrendering to his will.Consequently, the risk markers of experiencingand witnessing family-of-origin violence could be accounted for by either the psychiatric or sociallearning models.

f.,

--_u ___n-,-

,.. - -' 00' -'---'n

VIOLENT MEN 1035

:nen to have lower self-lnal status, to more fre-)used as a child, and tog up. Battered women0 have more traditional0 have been physicallyi physical violence be-lictimized women. Fin-violent men were in

, of marital conflict or11and religious incom-?ls of verbal abuse, andrated or divorred than

)ints out a number of;t, a number of theories-9r operationalization of.sychiatric model (Faulk,~) assumes that certainhim to engage in violent~attributed to either in-I factors (i.e.,past leam-:me assumption of this,:>n'sbehavior is internal

Oearly, the lower self-of experiencing and wit-ld support this model.,experiencing violence inor by a social learning,?Sthat husband-to-wife',q>Osureto the vicarious"1epresent reinforcementent behavior. Essentially,'ysically aggress against(the individual's motheriolent behavior goes un-urrendering to his will.nd witnessing family-of-the psychiatric or social

T-

------.-.- -_. '--'-

I,

1036 SUGARMAN AND HOTAUNG

In a similar vein, resource theory (Goode, 1971) is not conceptuallydistinct from stress theory (Farrington, 1980,1986).Goode argues that aDsocial systems are based to some degree on force or the threat of force. Itis the individuals within the system who have the greatest resources whoare able to command the greatest influence because they can use theseresources to influence others. Those individuals who are deficient inthose resources that are not based on physical strength (e.g., economicresources, intellectual capabilities, interpersonal skills, etc.) may have torely on their greater physical resources (e.g.,physical threat and violence)to promote their interests. Consequently, marital violence emerges as apower tactic utilized by an individual who has relatively few resourcesin other realms. On the other hand, social stress theory (Farrington, 1980,1986) argues that the source of violence is not a personal resourceproblem but rather an individual's inability to meet the demands ofstressful life events. Associations of risk attributes such as low income,employmentstatus, low occupationalstatus,and low educationallevel .with husband to wife violence, are used by both models as direct sup-port. This creates a problem in that a .single operationalization (e.g.,employment status) may represent a number of hypothetical constructs(e.g., stress level or resource availability). Consequently, models that areemployed to explain husband-to-wife violence appear to employ similaroperationalization to confirm their hypotheses and make it difficult toeliminate alternative hypotheses.

Second, aside from the conceptual redundancy of the varioustheoretical models of husband-ta-wide violence, Hotaling and Sugarman(1984)suggests a number of the risk markers are highly correlated witheach other (e.g., family income and husband's occupational status; wit-nessing and experiencing violence as a child). Consequently, it is quitepossible that the various risk markers are not independent of each otherand may represent underlying latent variables. These multivariate vari-ables may aid in clarifying some of the conceptual ambiguities of thetheories noted above.

Third, Hotaling and Sugarman (1984) noted that .much of the case-control or contrasted group research employed a basic two group designcomparing a battering sample to a nonbattering sample. Unfortunately,only a few attempts have been made to evaluate whether the observedgroup differences may be due to the level of marital discord with therelationship (d., Rosenbaum & O'Leary, 1981;Teich & Lindquist, 1984).Furthermore, no reported research has attempted to discriminate thosemen who employ severe forms of aggression (e.g.,beating their partner)

-- -- .--'

-- ~ ,.-- -. iii

~

........~--~.--.' -. - -.. -.. - -. ->'~-= ""'-~

-, i_-" '&

"".

-

~" ~,.~'..<~tii

VIOLENT MEN 1037

from those men who use less severe violent behaviors (e.g., slappingtheir partner).

The purpose of the present study is an attempt to circumvent some ofthese problems and to assess the utility of specific, risk markers in under-standing male violence toward women as well as differences in the.severityof this violence.The present design employs four groups of menwhich vary along a continuum of violence. In addition to a nonviolentcontrol group, a group of men who are verbally aggressive but not physi-cally violent towards their partners is included. Two groups of men arephysically violent: Men who reported severe physical violence againsttheir partners and men who reported only minor physical violenceagainst their partners.

Method

Description ofSample

A secondary analysis of the National Family VIOlenceStudy <NfVS;Straus, Gelles, IeSteinmetz, 1980)was performed. The data come from anarea probability sample of 2143families in which one adult family mem-ber was interviewed (960 men; 1183 women). 1be present analysis in-cluded only male respondents and because of the listwise deletion ofrespondents with missing data, the original sample of 960 was reducedto 608.

In order to determine whether the Iistwise deletion of respondentswith missing data was a threat to the representativeness of the sample, acomparison of the 608 men used in the present analysis (present sub-sample) with the 352 men who were not included (missing subsample)was performed. No significant differences between these two groupswere found with respect to marital status (;(2) =3.46,P=ns), religiouspreference (;(4) = 5.93,P = ns), and number of childrenin the family(t(958) =1.36, P =ns). However, the missing subsample was less likely to

be high school graduates (;(3) =20~ P < .001),to be employed(;(2) =11.82,P = 0.003),and to be nonblack (r-(2) =6.58,P= .()4)than thepresentsample. Furthermore, the missing subsample was older (t(953)=4.96,P <.001;45.2years vs. 40.6years) and was married longer (t(955)= -3.90,P<.001; 18.92years vs. 15.71years) than the present subsample. This sug-gests that the sample used in the present analysis is more representativeof a white middle-class population.

-- ---. ,- '---' ---.- '-

,

VIOLENT MEN 1037 1038 SUGARMAN AND HOTAUNG

ehaviors (e.g., slapping Operationalization of Group Membership

)tto circumvent some ofc.risk markers in under-II as differences in the

loys four groups of menddition to a nonviolentggressivebut not physi-Two groups of men are:tysicaJviolence againstlinor physical violence

The data provide for a more refined measure of violent behavior thanis usually found in studies of domestic violence. Violent behavior wasmeasured using the Conflict Tactics Scale (CTS;Straus, 1979).The ~male respondents were divided into four groups based on their reportsof the types of behaviors which they employ in resolving conflicts withintimate others. These behaviors range from the use of discussio~debate, and reasoning to resolve the conflict (Reasoning scale) to theemployment of verbal acts that threaten or harm the partner (VerbalAg-gression scale) and further includes behaviors that directly attempt tophysically hurt the partner <Violencescale).Each of these behavioral tac-tics has a 7-point frequency scaJe associated with it, ranging from 0("never") to 6 ("more than twenty times").

Respondents were classified into violence groups based on whetherthey ever employed any of these tactics or not. Nonviolent men reportedthat they never employed any of the verbal or physical aggression iacticsof the crs. While verbally aggressive respondents did report using thesenonviolent tactics, they also employed at least one of the verbal aggres-sion tactics. The minor physical violence group reported that they, atleast once, had pushed, grabbed, shoved, slapped, or threw something attheir partner. The severe violence group reported that they kicked, bit, hitwith a fist, threatened to or did employ a knife or gun against theirpartner, or beat up their partner. It is important to note that these groupclassifications are hierarchical in nature. The respondents who reportedemploying more severe forms of violence to resolve conflict with apartner also may have employed less severe tactics. However, the non-violent and verbal aggression groups will not include any respondentwho would have employed a physically violent tactic. Each respondentwas classified on the basis of responses to the Conflict TacticSScaJeintoone of four mutually exc1usive groups: No Violence group (n =153); Ver-bal Aggression OnJy group (n =369); Minor Physical Violence group (n =62);and SeverePhysicalViolencegroup(n=24). °

This classification of physicaJly violent respondents into either aminor violence group or severe violence group is not to suggest that twoseparate factors underlie the items of the violence subscale of the crs. Infact, factor analytic studies of the crs would support a single factor in-terpretation of this subscale (Barling, O'leary, Jouriles, Vivian, & Mac-Ewen, 1987). Rather, Straus's (1979) distinction of minor and severe

Violence Study <NFVS;o' The data come from an, one adult family mern-The present anaJysis in-the listwise deletion of

'pIe of 960 was reduced

deletion of respondentstiveness of the sample, a.t analysis (present sub-led (missing subsample)ween these two groups=3.46, P =ns), religiousIf chi1dren in the family;ample was less likely to.to be employed (;(2) =p =.04) than the presentsolder (t(953) =4.96, P <onger (t(955) = -3.90,P<nt subsample. This sug-is is more representative

t,-- --,~- ~-~

--- ,---o_-eopr o~o_' ".<- °

~_o~_C.=--,=.o~ _o o ~ o~._-~~-~-_.~--o_-- _0 _0__'- .~o-.

,-.-."}~oii.ii,.

VIOLENT MEN 1039

violence items was employed onJy as a heuristic to classify respondentsin to this hierarchically organized and mutually exclusive typology.

Risk Marker Inclusion and OperationalizJztion

An advantage of the NFVSdata set was that the interview was verycomprehensive allowing for the measurement of 16 of the 17 riskmarkers previously identified. The single missing risk marker waswomen's drug abuse. The omission of this marker did not directly affectour analysis. There is some evidence that the woman's use of drugs is anoutcome of victimization rather than a predisposing factor (Stark et al.,1981;Straus, Gelles,& Steinmetz, 1980).

Because of the nature of the data set and the present research design,four other risk markers had to be excluded from the analysis. The NFVSonly requested infonnation on experiencing and/or witnessing ofviolence in the respondent's family of origin. Consequently, the twomarkers of the wife's experiencing and witnessing of violence in herfamily of origin had to be omitted from the analysis because the presentstudy focused solely on male respondents. The third excluded riskmarker, marital status, was dropped for several reasons. First, the NFVSsurveyed only households in which the adults were married orcohabitating. Thus, none of the respondents were presently separated ordivorced from their present partners. Second, atthough there werecohabitating respondents, only Yllo and Straus (1981)reported higherrates of violence for cohabitants than married individuals and this was

from an analysis of this same data set. Consequently, it is questionable toemploy this married-cohabitating variable as a risk marker given thelimited research that supports it. The fourth risk marker that had to beexcluded was the frequency of verbal aggression. This variable wasomitted because it was employed in the operationalization of the group-jng variable. Two risk markers, witnessing violence as a child and ex-periencing violence as a child were broken down by the sex of the parentthus yielding four separate risk markers. As a'result of these exclusionsand additions, 14 risk markers remained in the analysis. 'Theseare of-fereij in Table1.

Scales were constructed for five risk markers: martial conflict, sexrole traditionalism, husband's occupational status, male self-esteem,andperceived female self-esteem. The marital conflict marker requestedrespondents to indicate how often the couple agreed on five. dyadic

,,..--- --- ''"''~'::;'~.s.~~~,:'::i.'C~~c..::._~:t:.'~...~..,- -~=...,.". -- H.. ."" -

ill ~- ~~.' ._-,"~,.,'--_..

VIOLENT MEN 1039

to dassify respondentsxclusivetypology.

the interview was veryof 16 of the 17 risk

,sing risk marker was~ did not directly affect-nan's use of drugs is an,sing factor (Stark et at,

present research design,the analysis. The NFVSand / or witnessing ofConsequently, the two

;sing of violence in herysis because the ~t"he third excluded riskreasons. First, the NFVSjults were married or-e presently separated ori, although there were5 (1981) reported higherndividuals and this was

'ntly, it is questionable to1 risk marker given thek marker that had to besion. This variable was:>naJ.jzationof the grou~lence as a child and ex-n by the sex of the parent-esult of these exclusionse analysis. These are of-

ers: martial confljct, sexus, male self-esteem, andInfljct marker requestede agreed on five. dyadic

-~r"'"

'.r~~'- ~~

1040 SUGARMAN AND HOTAUNG

issues (managing the money; cooking, cleaning, or repairing the house;social activities; affection and sex relations; and other things about thechildren). Each of the items within this marker was assessed on a 5-pointUkert scale with 1anchored as "Always"and 5, ''Never.''

The sex role traditionalism marker was operationalized by a powernonn index in which the respondents indicated who within the relation-ship they would say would have the final say about six issues (e.g.,buying a car, having children, what job your spouse should take). A 5-point Ukert scale was used for each of these items. The scale's anchorswere: 1, wife only; 2, wife more; 3, husband and wife same; 4, husbandmore; and 5, husband only. These items were then repeated, requestingfrom the respondent what their spouse would answer. Adifference scoreon each item was computed by subtracting the response for the wifefrom the response for the husband and an average difference score wasderived. Higher scores indicated that the couple's decision-makingpower was within the husband's purview.

The husband's occupational status marker was computed through aprocedure suggested by Treiman (1977).The two self-esteem markers in-volved the respondent's evaluation of himself and his spouse on threeitems: (a) being wann and affectionate, (b) smart and intelligent, and(c)being physically attractive. It is important to note that the wife's scaleis the man's evaluation of his partner's self-eteem. Cronbach alpha mef-fidents for the five markers ranged from 58 to .87.

Results

Factor Analysis of Risk Markers

In order to examine the redundancy among the 14 measured riskmarkers, a factor analysis employing Varimax rotation was perfonned.This procedure resulted in six latent variables which are presented inTable1.A 30 criteria for inclusion of a marker in a factor was used. Based

on this analysis, factor scores were computed for each of the six latentvariables.

Factor 1 appeared to represent a socioeconomicstatus factor that badhigh loadings on family income, husband's income, and husband's occu-pational status. High factor scores on this latent variable denote highersocioeconomic status. The husband's self-esteem index and his pe~tion of his wife's self-esteem loaded positively on Factor n suggesting a

..-- - . ."- --.- --- ,".'-"~- . "-'""""--' ,~._"-" " ',,,.,,_..,"',,'

Table 1

,

VIOlENT MEN 1041

Vllr1max-Rotated FllCforAnalysis of the Risk Markers

VariablesFactors

I IV V VIII III

Husband's education relative to wife'seducation (XI15)"

Husband's religious participation relative .07 .09 -.03 -.05 .13 .32to wife's participation (XE2)

Husband's occupational prestige:Treiman scale (XE26)

Family income (Q119)Husband's income (Q120)Husband's self-esteem index (XE1H)Wife's self-esteem index (XE1W)

Frequency of respondent's father hittingmother (Q106R)

Frequency of respondent's mother hitting .00 .01 .10 .11 .03 .12father (Q108R)

Frequency of father's use of physicalpunishment (Q103)

Frequency of mother's use of physicalpunishment (Q104)

Marital conflict (X2)Frequency of husband's drunkenness

(QE6H)Power norm index (XPZl1)

.01 -.03 -.05 -.00 -.04 -.36

.36 .19 -.04 -.09 .01 ~

.94 .11

.85 .11

.16 .90

.11 .71-.04 -01

-.03 -.04 .02 .07.00 -.02 .01 m.01 -.02 -.02 .15.01 -.00 -.09 -.01.81 .15 -02 -.03

..

-.01 -.03 .os .10 -.03 -.13

-.08 .00 .18 .15 .{)6 .03

.09 -.11 .07 .11 ~4.2 .13

.02 .09 .04 -.06 50 -.14

-.04 -.06 -.03 -.00 .34 .05

~e labels within the parentheses represent the variables' labels in theNfVS data set.

self-esteem factor. Consequently, a higher factor score represented ahigher level of self-esteem. Factor ill involved markers which assessedthe frequency of witnessing physical violence in the family of origin forthe husband (father's violence towards mother and mother's violencetowards father), while Factor IV tapped into experiencing physical

------ .-...--

.

.

VIOLENT MEN 1041

Factors

II III IV V

.03 -.05 -.00 -.04 -.36

.09 -.03 -.05 .13 32

.19 -.04 -.09 .01 .36

.11 -.03 -.04 .02 .07

.11 .00 -.02 .01 .07

.90 .01 -.02 -.02 .15

.71 .01 -.00 -.09 -.01

.01 .81 .15 .02 -.03

.01 .11 .03 .12

.03

.70

.08 .70 -.03 -.13

.00 .18 .75..06 .03

.1l. .07 .11 :..4.2.13

.09 .04 -.06 50 -.14

.06 -.03 -.00 34 .05

variables' labels in the

::>rscore represented anarkers which assessedthe family of origin forand mother's violenceexperiencing physical

r-"---'." ~.-.,-~. -...

VI

1042 SUGARMANANDHOTAUNG

~

f.

violence in the family of origin (father's and mother's violence towardsthe male respondent).' High scores on both of these factors indicatedhigher frequencies of observing or experiencing violence as a child. Threemarkers load highly on the fifth factor. These included frequency of thehusband's drunkenness, amount of marital conflict, and power nonnindex. Factor V seems to be assessing the milieu of dissension that ex-isted in the relationship, and was labelled a marital conflict factor withhigher scores being indicative of greater conflict. Three markers loadedon the final factor.These were the husband's educational attainment rela-tive to his wife's educational attainment, the husband's occupationalprestige, and the husband's religious participation relative to his wife'sreligious participation. Thus, Factor VIappears to denote a relative statusfactor.Factor scores were computed for each latent variable.

Be~-group AnIllysis

)J

A four-group analysis of variance was performed on each of the sixfactor scores using the type of aggression exhibited (no violence, verbalaggression only,minor physical violence, severe physical violence) as thegrouping factor.The means and standard deviations for these groups aswell as the F values are presented in Table 2. No significant between-group differences were found when the self-esteem factor scores <FactorII) and the relative status factor scores (Factor VI) were analyzed. Theother four factors revealed significant findings. In order to obtain a morcsensitive assessment of these group differences, post hoc Scheffe testswere perfonned.

The marital conflict factor <FactorV) emerged as the most powerfuldiscriminator of these four groups. Examination of these factor scoresrevealed that both severely and slightly physical1yviolent men reportedsignificantly greater amounts of marital conflict than the other twogroups. Similarly, verbally aggressive men had higher marital conflictscores than nonviolent men. On Factor n, severely violent men had sig-nificantly lower socioeconomic factor scores than the other three groupsand the same group had significantly higher scores than the other threegroups on the witnessing physical violence latent variable (Factor m).Final1y,the significant effect revealed from an analysis of the experienc-ing physical violence factor scores (Factor IV) resulted from the verballyaggressive men scoring lower on this factor than men who reportedminor physical violence.

" '-"'-'-- . -- '-----~ ..._- ..''''' ='II!II'

~,--:.,;':t"'~:'7' 411 :!.-'-"--'~''''''-- .._~-.." - ~ ' ~

n --

~1..

'7he labels associated with the respective factor are the following: Fac-tor I - Socioeconomicstatus factor; Factor n - Self-esteemfactor; Factor III-Witnessing physical violence factor; Factor IV -Experiencing physicalviolence factor; Factor V -Marital conflict factor; and Factor VI -Relativestatus factor.

..The degrees of freedom associated with each analysis were 3,604.Note.Means that do not share the same letter superscript are statisticallydifferent at the .05 level using a Scheffe procedure.

Discussion

The present analysis suggests that a number of factors discriminatevarying degrees of husband-to-wife violence.First, it appears that an in-creased likelihood of violence is associated with higher levels of maritalconflict or disruption. Interestingly,marital conflictdoes not differentiate

,i_. ~~-~-=-~~~~~:~~:s~.¥.~~~:-¥..2;~~~£~~::.~~-".-

VIOLENT MEN 1043

Table 2

Mean Factor Scoresfor Each Aggression Group

Group

No Verbal Minor SevereFactott Aggression Aggression Violence Violence PO' p

I M -.oif' .13A -.11A -1.00B 12.43 <.001sd .98 .90 .97 .89

n M -.01 .01 .01 .02 .02 nssd 1.03 .89 .69 1.02

III M -.1SA .02A .04A .66B 6.28 <.001sd .46 .91 .88 1.67

IV M .ooAB _.06A .34B .15AB 4.44 .004sd .93 .74 1.01 1.05

V M -.4oA .07B .43c .44c 37.58 <.001sd .46 .63 .78 .70

VI M .02 -.01 .03 -.12 .42 nssd .54 .60 .68' .66

N 153 369 62 24

- .---------.--...

is a weaker predictor of?!Sing of violence. One-ushment as a child-rear-iUS,Gelles, &:Steinmetz,Thissuggests that future~sons for this differen-e.11\to emerge. First, theter understanding of the:)lence.It is possible that11,socioeconomic statusikely to employ violenceherisk markers offereda,hol in marital violence.'or of constraints against..vithinthe marriage,Jhateases the probabiJity ofru>'c E", au nE n..ni... "

&UUI!I:: UAt:JY given me re18DVeIYJOWcosts associated with aggressiontoward intimates (Sebastian, 1984).However, many of the interactionsbetween husband and wife may be viewed as ambiguous in motive andintent, especially during episodes of conflict. For example, an un-employed steel worker may perceive his wife's encouragement to find anew job as a threat to his identity in that it reminds him that he is un-employed and that he will require giving up a part of personal identity(i.e., being a steel worker). On the other hand, the wife's behavior mayhave resulted from a positively motivated desire to have her husbandregain a sense of self-worth. Consequently, as the number of interper-sonal conflicts between husband and wife increase, so does thelikelihood of physical violence.

Hotaling (1980) appears to imply that battering husbands havedeveloped a type of aggressive attributional style. Supporting evidencefor this position is offered by Dodge and Richard (1986).They found thataggressive children have a greater tendency to impute malevolent intentto accidental or ambiguous aggressive acts than nonaggressive du1drenand as a consequence are more likely to retaliate against the perceiveda~rc Thic. jl:.lla_tc th..at ~ic:. <aHftb.utinn.,.} hi",,, h.s.. d 1n l .. ...

,_."

VIOLENT MEN 1043 1044 SUGARMAN AND HOTAUNG

of factors discriminate;t, it appears that an in-higher levels of marital:t does not differentiate

between those husbands who employ either severe violence from thosethat exhibit moderately violent behavior; but does discriminate betweenhusbands who use verbal aggression or nonaggression strategies. Unfor-tunately, it is important to point out that this result may be tautologicalbecause the conceptual distinction between marital conflict and interper-sonal aggression (verbal or physical) is hard to draw. Still, given thespecificmarkers that loaded on the marital conflict factor, it appears thatphysically violent husbands have a significantly larger number of issuesover which they are in conflict with their wives.

Two factors differentiate severely violent men from the other threegroups. First, severely violent husbands appear to have lowei' ~economic status. Two interpretations can be made of this finding. Onepossibility is that men who have lower SESare exposed to greater socialstress; and they possess less resources (e.g., economic security, educa-tional capabiJities) to cope with this stress. A second intet'pretation sug-gests that the relationship between lower socioeconomic status andincreased likelihood of husband-to-wife assault can be due to the exist-ence of a culture of violence (Wolfgang &:Ferracutti, 1961).Thus in-dividuals who have lower incomes and educational levels may holddiffering cultural values regarding the use of violence against women.These two explanations are not mutually exclusive. Third, it is possiblethat the higher socioeconomic status respondents may have been moreaware of the social stigma attached to intimate violence and consequent-ly responded to the interviewer's queries in a socially appropriate man-ner. However, another interpretation is possible. A comparison of therespondents in the present sample with those who had missing data andwere eliminated from the analysis suggests that this latter group is morecharacteristically seen as being from the lower socioeconomic strata.Consequently, lower socioeconomic respondents may be just as sensitiveto these socialcues and instead of deceiving the interviewer, they just ter-minated the interview. As a result, the present analysis may have under-estimated the amount of violence in the lower socioeconomicstrata.

The second finding that discriminated severely violent males from theother three groups involved the greater frequency of Witnessing violencein their family of origin. These husbands are from families where rolemodels who employ violence against women are present.

Surprisingly, experiencing violence in the family of origin did notdiscriminate the severely violent husbands from the other groups; rather,it onJy differentiated verbally aggressive husband from husbands whoemploy minor physical aggression. While this finding is difficult to

are the following: Fac-esteem factor; Factor IIIExperiencing physical

md Factor VI -Relative

ilysis were 3, 604.-erscript are statistically

...,--, -,--

~-... ::~:--"- , !'!!'c -.'.-.r'

---' '"'~' -.",

- ~- -.- ,-.----,..-, -- - '-'

SevereViolence F- p

-1.00B 12.43 <.001.89.02 .02 ns

1.02.66B 6.28 <.001

1.67.1SAB 4.44 .004

1.05.44c 37.58 <.001.70

-.12 A2 ns.6624

c,