Elsevier

Aggression and Violent Behavior

Volume 7, Issue 4, July–August 2002, Pages 313-351
Aggression and Violent Behavior

Sex differences in physically aggressive acts between heterosexual partners: A meta-analytic review

https://doi.org/10.1016/S1359-1789(01)00061-1Get rights and content

Abstract

Meta-analyses are reported of sex differences in acts of physical aggression to heterosexual partners, derived from the Conflict Tactics Scale (CTS) [J Marriage Fam 8 (1979) 75], using methods based on standardized mean differences, and on odds ratios. Women were more likely than men to throw something at the other, slap, kick, bite, or punch, and hit with an object. Men were more likely than women to beat up, and to choke or strangle. Differences ranged from very small to medium. Samples selected for marital problems showed large effects in the male direction, and student samples showed effects more in the female direction than community samples. Effect sizes derived from partners' reports were more in the male direction than those derived from self-reports, but the overall pattern of results was similar. A similar pattern of findings resulted from the use of odds ratios to derive effect sizes, although the magnitude of effects was greater. Limitations of the current CTS measures and the database are discussed.

Introduction

Family conflict researchers (e.g., Straus, 1990, Straus & Gelles, 1988a) argue that physical aggression between partners involves both sexes to an approximately equal extent, and arises from conflicts caused by the many everyday frustrations and stresses of living together. In contrast, feminist (e.g., Pagelow, 1984, Walker, 1989) and evolutionary (e.g., Shackleford & Buss, 1997, Wilson & Daly, 1992) researchers emphasize the imbalance between male perpetrators and female victims of relationship aggression, which they view as arising either from patriarchy or from the evolved proprietary motives of men.

Evidence for the family conflict researchers' position is derived from questionnaires inquiring about the ways people solve relationship conflicts, typically administered to samples not selected for their high level of violence (e.g., Magdol et al., 1997, Morse, 1995, Straus & Gelles, 1988b). Evidence for the alternative view is typically derived from crime surveys (e.g., Dobash & Dobash, 1977–1978, Gaquin, 1977–1978, Schwartz, 1987), from female victims' reports (e.g., Mooney, 1994), or from male perpetrators identified by law enforcement agencies (e.g., Claes & Rosenthal, 1990). The difference in the conclusions reached from these different sources has been attributed either to the different nature of the samples Johnson, 1995, Straus, 1997, Straus, 1999 or to the different nature of the measures. Two opposing arguments have been advanced about the methods. One is that the family conflict researchers' questionnaire measure is unreliable, in that it neglects the context and consequences of acts of physical aggression Dobash et al., 1992, Romkens, 1997. The other is that the crime surveys are unreliable Mihalic & Elliott, 1997, Straus, 1997, Straus, 1998, Straus, 1999.

Meta-analyses of sex differences in physical aggression to heterosexual partners (Archer, 2000a) have revealed little difference in the proportion of men and women who used one or more acts of physical aggression, and in the composite frequency of such acts. Typically, these studies involve questionnaires consisting of specific acts of physical aggression (usually the Conflict Tactics Scales, or CTS: Straus, 1979) administered to dating or community samples of married or cohabiting people. When measures of the consequences of physical aggression were obtained from the targets of such aggression, men were more likely than women to inflict an injury. Even so, around a third of those injured were men. These findings partially support the claims that different measurement methods have been responsible for conflicting findings. Nevertheless, the sample involved was an important moderator of effect size. Reports from women in refuges, and from men selected for high levels of partner violence, produced high effect sizes in the male direction using the CTS. Younger-aged samples of dating couples, and samples with lower rates of male aggression, showed effect sizes in the female direction.

Despite the finding that, overall, men and women show a similar likelihood and frequency of using any act of physical aggression, it is widely believed (e.g., Fagan & Browne, 1994, Straus et al., 1980) that serious acts of physical aggression are perpetrated mainly by men against their women partners. In assessing whether this is so, it is important to weigh both the severity of different acts of physical aggression and their consequences. Consequences were analyzed previously in the form of injuries (Archer, 2000a). The present paper reports further meta-analyses of sex differences in partner aggression, using the individual acts of physical aggression from the CTS, to assess whether there is increased male involvement for more severe acts. Parallel meta-analyses of the same data were undertaken in view of the use of two different methods for calculating effect sizes for dichotomous data (Haddock et al., 1998, Johnson, 1989: see below).

The CTS is the questionnaire most commonly used by family conflict researchers. It involves asking which of a number of acts of physical aggression the person and his or her partner have used in seeking to solve conflicts. Various ways in which conflicts can be solved are listed: These involve constructive problem solving, verbal or indirect aggressive acts, and physically aggressive acts. It is the third of these that are scored in studies of physical aggression towards partners. Although there are more studies providing aggregate measures of the occurrence and frequency of CTS acts for men and women, there is still a substantial number providing data for specific acts. Following Straus (1977–1978), many researchers have divided the CTS items into minor acts (the first three, such as “slapped the other one”), and severe (the remaining five or six1, starting with “hit, bit, or hit with a fist” and ending with “used a knife or gun”). The distinction is made on the basis that the first three are less damaging than the others. If it is men who typically perpetrate damaging acts, we should expect such acts to show a different pattern from the first three, “minor” acts.

The present paper uses data on the occurrence of each individual CTS act, to address this issue of whether serious acts of physical aggression are mainly perpetrated by men against their women partners, with less serious acts, such as throwing something at or slapping, being more frequently used by women. If so, we could conclude that the apparent lack of sex differences (or higher female levels) found for aggregate measures hides a different pattern of distribution of acts of physical aggression among men and women. This would go some way towards supporting the view of critics of conclusions derived from aggregate CTS measures (e.g., White, Smith, Koss, & Figueredo, 2000; see Archer, 2000b) that these do not account for the more damaging nature of men's physical aggression.

To illustrate this point, consider the following case reported in the British press in 1998. Police were called to a violent domestic incident, and photographed the woman's injuries to reveal severe bruising to her face, and fingernail and hand marks on her neck. The subsequent account of the event stated that it involved the following. First, a struggle occurred, in which the woman grabbed at and scratched the man's cheek and jaw; the man squeezed her throat, and then caused the injuries shown in the photograph. The man claimed that these resulted from pushing his partner, who then fell, but it seems more likely that the bruising resulted from several punches. CTS measures of the numbers of men and women showing any form of physical aggression would count this incident as one each for the man and woman. Measures of “cause visible injury” might also score one for each partner, if we assume that the woman's scratches left marks. Scoring the incident in terms of individual CTS acts would produce a different picture. The man would score positively for “push, grab, shove,” “kick, bite, punch,” “beat up,” and “choke or strangle,” and the woman for “push, grab, shove” only (scratching not being included in the CTS). Owing to the serious nature of the last two acts scored for the man, separating out the individual CTS acts provides a more realistic account of this incident than relying on one measure, whether based on acts or consequences.

The following prediction can be made: Although the threshold between physical and verbal or indirect acts of aggression is just as likely to be crossed by women as men, men will be more likely than women to exhibit more serious acts of physical aggression. In this context, serious refers to levels where there is greater risk of injury to the recipient, such as beating them up, or choking them, or threatening them with a weapon.

Acts of physical aggression used in the CTS and modifications of it vary in their potential for damaging the other, and in the risks they entail to the self. The first act, throwing something at the other, does not involve close proximity, and therefore to some extent avoids the immediate danger of physical retaliation associated with acts delivered by a part of the protagonist's body. The next category (“push, grab, or shove”) involves physical contact but is not aimed at inflicting pain on a specific part of the opponent's body. However, “slap,” and the potentially more damaging “kick, bite or punch,” are intended to inflict pain on a particular part of the body. “Hit with an object” typically has more potential for damage than if an object is not used (although there is some ambiguity here, since “object” could range from a rolled-up newspaper to a baseball bat).

The category “beat up” is different from those preceding it, in that it does not specify the action involved, and is defined more by its consequences than the way it is achieved. Thus a person may be “beaten up” by being punched, kicked, pushed and hit with an object, or some combination of these. It also carries the implication that the damaging acts are one-sided, and they are repetitive in form. “Beat up” is particularly relevant to the hypothesis that men are much more likely than women to perpetrate serious acts, since it implies that there are damaging consequences. This category is not subject to the charge that only actions and not their consequences are included in the CTS (e.g., Dobash et al., 1992, Romkens, 1997).

Choke or strangle involves specific actions that are by definition potentially damaging, in that one of them is life threatening. Since this is well known, it must be assumed that choking or strangling a partner involves the perpetrator acting recklessly regarding the other's life. Choke and strangle also involve the ability to overwhelm the other physically, and therefore we should expect the most frequent perpetrators to be men.

Threaten with a knife or gun, or using one of these, are the remaining two categories. Because they involve artificial weapons, they increase the danger to the opponent without increasing the (immediate) cost2 to the perpetrator. A gun can be used at a distance, thus further eliminating the immediate danger of retaliation from an unarmed opponent. Knives are more likely to be carried by men, and are involved in many acts of intermale violence; on the other hand, they are also readily available as kitchen and other household utensils, and may also be used by women as a result. Guns have been referred to as “the great equalizer” but again their widespread possession (particularly in the US) is a mainly male occurrence. For these reasons, it is difficult to predict whether we should expect men to be the predominant sex using or threatening to use knives or guns.

The data set used here is different from that used to analyze aggregate measures and injuries (Archer, 2000a). Of the total number of studies used for these meta-analyses (shown in the Appendix of the previous paper), 48 are included in the present analyses because they also provided data that enabled effect sizes to be calculated for specific acts. A further 10 studies not used in the previous analyses were included in the present ones because they provided data for individual acts (but not overall values). Sixty-nine of the studies used in the previous analyses could not be used in the present ones because they only provided data for aggregate CTS measures.

Most of these studies provide separate measures for self-reports and reports by partners, enabling effect sizes to be calculated for the two data sources. Self-reports are comparable to the many individual difference variables used in personality, social and clinical psychology. Partner reports are similar to peer reports used in studies of children's aggression (e.g., Bjorkqvist, Lagerspetz, & Kaukiainen, 1992), and also to reports from significant others used in studies of adults (e.g., Buss & Perry, 1992; McCrae & Costa, 1990; O'Connor, Archer, & Wu, 2001). A meta-analysis of aggregate measures comparing self- and partner reports (Archer, 1999) found that while there was evidence of high correlations between the two sources, self-reports were overall lower than partner reports, and that this was more pronounced for men than for women. This pattern explained the finding from the meta-analyses of aggregated data that effect sizes were more in the male direction for partner than for self-reports. The most obvious explanation for this finding is systematic underreporting of one's own physical aggression compared to that of a partner, and for this tendency to be more pronounced for men than women.

This issue may be more complex in the case of specific CTS acts. Riggs, Murphy, and O'Leary (1989) asked people to rate how likely they would be to report items on the CTS: they found that willingness to report one's own aggression was not only less than willingness to report partners' aggression, but it also decreased with the seriousness of the CTS act. In a meta-analysis, Sugarman and Hotaling (1997) also found that the largest negative correlation between social desirability and physical aggression occurred for serious CTS acts. In the present study, discrepancies between reports from the self and the partner were calculated, to assess their possible association with the seriousness of the acts concerned.

The present meta-analyses provided an opportunity to examine the impact of moderator variables on the sex differences in acts of partner aggression. The rationale for choosing the variables was as follows. (1) The source of data was included to assess whether published studies were a representative sample of all available studies, in view of claims that publication of some studies finding assaults by women on their partners had been suppressed (Straus, 1997). (2) The measurement instrument was coded to enable a comparison between effect sizes from studies using the CTS (see above) and other, more rarely used, measures. (3) The country of origin enabled a preliminary comparison across nations. (4) Age category allowed examination of whether sex differences were more in the female direction at younger ages, as was found for aggregate measures. (5) The sample was included to assess whether those selected for relationship problems, including marital violence, showed large effect sizes in the male direction, and whether student samples showed effect sizes more in the female direction. (6) Comparing dating with married or cohabiting samples enabled an assessment of whether effect sizes were more in the female direction for less permanent liaisons. (7) Nominal and interval level were compared to assess whether effect sizes differed according to whether they were derived from numbers of each sex or frequencies of each act. (8) There was variation in the period used when asking about partner aggression, and since higher rates would be expected over longer time periods, this was coded. (9) Sex of author was included because, for the aggregate measure, the sex difference was more in the female direction for male than female authors (Archer, 2000a).

In the analysis of values representing the aggregate from self- and partner reports, dichotomous data, mainly in the form of frequencies of occurrence for men and for women, were converted into mean values using the DSTAT program (Johnson, 1989), a method that is usually used for meta-analyses in psychology. However, Haddock et al. (1998) have argued that this method underestimates the magnitudes of effect sizes unless the marginal distributions are similar to one another. Instead, they recommend computing odds ratios from the proportions of two categories that did and did not fulfil a particular criterion (in the present case commit a particular act of physical aggression to a partner). These odds ratios can be converted into effect sizes (d values) by a simple calculation. Haddock et al. provided some examples from individual studies of divergent effect sizes calculated by the two methods, those from odds ratios being considerably higher than those calculated from treating the proportions as means.

In the present study, since nearly all the data was in the form of proportions committing specific acts, it was possible to calculate effect sizes from odds ratios and to compare these with values from the standardized mean difference (DSTAT) method. The dichotomous data in each study involved exactly the same categories (men and women) and the same measures (did or did not commit a specific act). It was therefore also possible to aggregate the data from individual studies, to produce overall values for men and women who did and did not commit each act (cf. Shadish & Haddock, 1994, pp. 272–273). These aggregate values could then be used to calculate effect sizes using both the DSTAT, and the odds ratio methods. The present study therefore also provides a comparison between effect sizes calculated in two different ways from the same data, and also between effect sizes calculated from individual studies and from combined frequencies across all studies. It will become apparent that effect sizes calculated from odds ratios derived from aggregate proportions overcomes a problem encountered when using this method for acts that have zero values in individual studies.

Section snippets

Sample of studies

The main search involved PsycLIT(r) on CD-ROM for the years 1976 to October 1998, using the key words “marital or dating” and “aggression or violence” but excluding “sexual,” “rape” and “pornography” because sexual forms of aggression are typically studied separately and have only recently been included in studies involving the CTS and related measures (see Discussion). This search produced 581 titles. Dissertations were searched via DISS (Dissertation Abstracts International Online), from 1979

Study characteristics

Appendix A lists each study used for the meta-analyses, together with the effect sizes for each act (from the DSTAT method) and the study characteristics. Table 1 summarizes the study characteristics, indicating that most were carried out in the 1980s and 1990s, in the USA. Over half involved college or high school students in dating relationships. These statistics alone limit the generalizations that can be made from the subsequent analyses. There are, however, sufficient numbers of community

Summary of the main findings

Four of the first five CTS acts showed mean weighted effect sizes in the female direction, ranging from medium to very small in terms of Cohen's (1988) criteria, according to which act and which measure was used. Small differences were found for weighted mean effect sizes derived from standardized mean differences, both before and after a correction had been made for the impact of large samples, and when outliers were removed. The same pattern was found for self- and partner reports analyzed

Acknowledgements

I thank Nicola Graham-Kevan, Jane Ireland, and Samantha Gilling for invaluable assistance with the analysis and literature search; Blair Johnson and Alice Eagly for answering inquiries about meta-analytic procedures; Norman Birbeck, Marion Seed, and Andrea Burch for additional help with the literature search. I also thank Alice Eagly and Frank Fincham for constructive critical comments on an earlier version of the paper.

References (114)

  • G.L. Roberts et al.

    Prevalence study of domestic violence victims in an emergency department

    Annals of Emergency Medicine

    (1996)
  • J. Archer

    Assessment of the reliability of the Conflict Tactics Scales: a meta-analytic review

    Journal of Interpersonal Violence

    (1999)
  • J. Archer

    Sex differences in aggression between heterosexual partners: a meta-analytic review

    Psychological Bulletin

    (2000)
  • J. Archer

    Sex differences in partner aggression: a reply to Frieze (2000), O'Leary (2000), and White, Smith, Koss and Figueredo (2000)

    Psychological Bulletin

    (2000)
  • J. Archer et al.

    Game theory models and escalation of animal fights

  • K. Bjorkqvist et al.

    Do girls manipulate and boys fight? Developmental trends in regard to direct and indirect aggression

    Aggressive Behavior

    (1992)
  • J.R. Bohannon et al.

    Using couple data to determine domestic violence rates: an attempt to replicate previous work

    Violence and Victims

    (1995)
  • J. Bookwala et al.

    Predictors of dating violence: a multivariate analysis

    Violence and Victims

    (1992)
  • Breen, R. N. (1985). Premarital violence: a study of abuse within dating relationships of college students. Unpublished...
  • F.C. Breslin et al.

    Family precursors: expected and actual consequences of dating aggression

    Journal of Interpersonal Violence

    (1990)
  • M. Brinkerhoff et al.

    Interspousal violence

    Canadian Journal of Sociology

    (1988)
  • J. Browning et al.

    Assessment of wife assault with the Conflict Tactics Scale: using couple data to quantify the differential reporting effect

    Journal of Marriage and the Family

    (1986)
  • A.H. Buss et al.

    The aggression questionnaire

    Journal of Personality and Social Psychology

    (1992)
  • A.L. Cantos et al.

    Injuries to women and men in a treatment program for domestic violence

    Journal of Family Violence

    (1994)
  • M. Carrado et al.

    Aggression in British heterosexual relationships: a descriptive analysis

    Aggressive Behavior

    (1996)
  • J.A. Claes et al.

    Men who batter: a study in power

    Journal of Family Violence

    (1990)
  • J. Cohen

    Statistical power analysis for the behavioral sciences

    (1988)
  • M. de Vries Robbe et al.

    Prevalence of domestic violence among patients attending a hospital emergency department

    Australian and New Zealand Journal of Public Health

    (1996)
  • R.E. Dobash et al.

    Wives: the “appropriate” victims of marital violence

    Victimology: An International Journal

    (1978)
  • R.E. Dobash et al.

    Violence against wives: a case against the patriarchy

    (1980)
  • R.P. Dobash et al.

    The myth of sexual symmetry in marital violence

    Social Problems

    (1992)
  • D.G. Dutton

    Patriarchy and wife assault: the ecological fallacy

    Violence and Victims

    (1994)
  • A.H. Eagly

    The science and politics of comparing women and men

    American Psychologist

    (1995)
  • A.H. Eagly et al.

    Explaining sex differences in social behavior: a meta-analytic perspective

    Personality and Social Psychology Bulletin

    (1991)
  • J. Fagan et al.

    Marital violence, physical aggression between women and men in intimate relationships

  • M.S. Fiebert et al.

    College women who initiate assaults on their male partners and the reasons offered for such behavior

    Psychological Reports

    (1997)
  • J.L. Fleiss

    Statistical methods for rates and proportions

    (1981)
  • L. Foo et al.

    A multivariate investigation of dating aggression

    Journal of Family Violence

    (1995)
  • O. Frenkiel

    If our women are unfaithful — we kill them

    The Guardian (London): The Editor

    (1999, January 23)
  • D.A. Gaquin

    Spouse abuse: data from the National Crime Survey

    Victimology

    (1978)
  • R.J. Gelles

    The violent home

    (1972)
  • R.J. Gelles et al.

    Intimate violence in families

    (1990)
  • M.J. George

    Riding the donkey backwards: men as the unacceptable victims of marital violence

    The Journal of Men's Studies

    (1994)
  • George, M. J. (1998). Beyond all help? Comments upon ‘Commonly held unhelpful beliefs’ in ‘Domestic violence: a health...
  • J. Giles-Sims

    Wife battering: a systems theory approach

    (1983)
  • F.E. Gryl et al.

    Close dating relationships among college students: differences by use of violence and by gender

    Journal of Social and Personal Relationships

    (1991)
  • P.A. Gwartney-Gibbs et al.

    Learning courtship aggression: the influence of parents, peers, and personal experiences

    Family Relations

    (1987)
  • C.K. Haddock et al.

    Using odds ratios as effect sizes for meta-analysis of dichotomous data: a primer on methods and issues

    Psychological Methods

    (1998)
  • L.V. Hedges et al.

    Statistical methods in the meta-analysis of research on gender differences

  • L.V. Hedges et al.

    Statistical methods for meta-analysis

    (1985)
  • W.W. Hudson et al.

    The assessment of spouse abuse: two quantifiable dimensions

    Journal of Marriage and the Family

    (1981)
  • B.T. Johnson

    Software for the meta-analytic review of research literatures

    (1989)
  • M.P. Johnson

    Patriarchal terrorism and common couple violence: two forms of violence against women

    Journal of Marriage and the Family

    (1995)
  • K.-I. Kim et al.

    Epidemiological survey of spousal abuse in Korea

  • K.E. Lane et al.

    Violence in the context of dating and sex

    Journal of Family Issues

    (1985)
  • M.R. Laner

    Courtship abuse and aggression: contextual aspects

    Sociological Spectrum

    (1983)
  • M.R. Laner

    Unpleasant, aggressive, and abusive activities in courtship: a comparison of Mormon and non-Mormon college students

    Deviant Behavior: An International Journal

    (1985)
  • M.R. Laner et al.

    Abuse and aggression in courting couples

    Deviant Behavior: An International Journal

    (1982)
  • S.D. LeBlanc

    Abuse in dating relationships: an urban/rural comparison

    Masters Abstracts International

    (1995)
  • R.F. Lehr

    Characteristics of couples in violent relationships: implications for predicting successful marital therapy

    Dissertation Abstracts International

    (1988)
  • Cited by (330)

    • Development and assessment of the Partner Exploitation Inventory

      2022, Current Research in Behavioral Sciences
    View all citing articles on Scopus
    View full text