Conflict and Control:
Images of Symmetry and Asymmetry in Domestic Violence
Michael P. Johnson
To appear in Alan Booth, Ann C. Crouter and Mari Clements (Eds.), Couples in Conflict.  Hillsdale, NJ: Lawrence Erlbaum, 2000.

I love the Bradbury, Rogge, and Lawrence paper for its brazenness—to come to a conference on conflict and argue that conflict isn’t all that important—and for its thoughtful working out of that thesis (Bradbury, Rogge, & Lawrence, 2000).Their approach, however, does take us away from the initial questions posed for this session of the symposium.They are (1) “What are the interpersonal roots of couple conflict?” and (2) “What are its consequences for individuals and couples?” I want to get back to those questions, and in order to do so I am going to present some of my own work.My major point will be that neither of these questions can be answered without making distinctions among types of conflict.
 
I spent much of this morning’s session (Wilson & Daly, 2000) musing about the dangers of treating violence (even wife-murder) as if it were a unitary phenomenon.I work primarily in the area of domestic violence, and my mission these days is to convince others who work in that area that the domestic violence literature is full of serious mistakes, mistakes that arise because most researchers do not make what I consider to be some absolutely necessary distinctions among types of violence (Johnson, 1995; Johnson, 1998; Johnson, 1999; Johnson & Ferraro, 2000).The most visible of these errors, and perhaps the one with the most serious implications in terms of public misunderstanding of the nature of domestic violence, is the mistaken idea that women are as violent as men in intimate relationships.They are not, and I hope that the next 15 minutes will convince you of that.

 

From the very first page of Bradbury, Rogge and Lawrence’s paper, I felt uneasy.Their ostensibly quite reasonable definition of marital conflict is “those social interactions in which the spouses hold incompatible goals (p. 2).”It sounds so symmetrical and almost benign—the interaction of two people with incompatible goals.But when I read it, it certainly didn’t seem to fit with the data I had just been looking at from a woman who described her husband as beating and humiliating her so regularly that she could no longer keep track of how many times it had happened.How many times?All she could say was, “Oh, at least 500 times.”
 

The conflicts that are described by women entering shelters or filing Protection from Abuse Orders do not have the symmetrical feel of“incompatible goals.”Instead, they seem to represent the single-minded commitment of one person to completely dominate and control another (Dobash & Dobash, 1979; Kirkwood, 1993; Pence & Paymar, 1993).Admittedly, since these “others” do not wish to be dominated and controlled, we could characterize the situation as one of “incompatible goals,” but something very important seems to be missing in that description, and I think the heart of the problem is to be found in the implied symmetry of the definition.
 

In the domestic violence literature, symmetry has been at the center of a decades-long, unusually acrimonious debate over the nature of marital violence.It all started when the first National Family Violence Survey (NFVS) became available in the late 1970s.Suzanne Steinmetz (Steinmetz, 1977-78) published an infamous paper on “husband-battering,” in which she used the NFVS data to support her argument that there was a problem of husband-battering that was perhaps as serious as that of wife-battering.Although the NFVS data did in fact show almost perfect gender-symmetry of partner violence, other scholars quickly responded with rebuttals, arguing that all other studies had found that domestic violence is almost entirely male, and attacking the validity of the data. But study after study has appeared in the intervening decades, right up to last month, that ostensibly show gender-symmetry in partner violence.The talk shows love it.
 

The scholarly debate involves the two major groups of sociologists who study relationship violence.Here is an over-simplification of their differences.One group is usually referred to as the “family violence researchers,” because they have studied a variety of types of familyviolence in addition to partner violence.Murray Straus and Richard Gelles, who designed the National Family Violence Surveys, are the best known members of this group.The major methodology of this tradition is the large-scale survey, assessing violence by means of a set of survey questions called the Conflict Tactics Scales.In general, it is this group that argues that men and women are equally violent in intimate relationships.The other major group is usually referred to as the “feminist researchers,”and the best known of them are Rebecca and Russell Dobash.This group uses data collected primarily in qualitative research focusing on women who are clients of public agencies such as shelters, the courts, hospitals, and so on.They argue that partner violence is male and rooted in patriarchal traditions.
 

What is striking about this debate is that both the family violence researchers and the feminist researchers are able to cite multiple studies to support their position.Of course, each side sees fatal flaws in the other group’s research strategies.On the one hand, the feminists present a measurement critique of survey research, arguing that the Conflict Tactics Scales ignore everything about violence that is important, merely counting number of incidents instead of attending to motives and consequences.On the other hand, the family violence folks present a sampling critique of feminist research, arguing that shelter and court samples have obvious selection biases.As Murray Straus puts it, they deal with “clinical samples,” while the surveys deal with “general samples”(Straus, 1990).
 

Well, I think they’re both right.I’ll get to measurement later, but my primary argument hinges on the sampling issue, with a different take on bias, however, than the one argued by Straus and his colleagues.The biases of shelter samples may be more obvious than those of general surveys, but the samples in so-called random-sample surveys are equally biased.We do not in fact interview random samples.We interview those who do not refuse to be interviewed.And the refusal rate is not trivial—the 1985 National Family Violence Survey actually had a refusal rate of 40%, not the 16% usually reported (Johnson, 1995).
 

I am going to argue, therefore, that the results of these two types of research differ, not because one is biased and the other is not, but because both are seriously biased, and each gets at only one of the two major types of partner violence.What might those two types be?Well, if you look at those few characteristics of the violence that are measured in both types of research, here’s what you see in the literature.First, there is the gender difference that started the argument.Feminist researchers using public agency samples find male violence.Family violence researchers, using general survey samples, find gender-symmetric violence.Second, the per-couple frequency of violent incidents is dramatically higher in shelter samples than it is in general samples (on the order of ten times higher).Third, shelter data show almost universal escalation of the violence; general survey research shows very little escalation and considerable de-escalation.Finally, in shelter data it is unusual for the woman to fight back, while general survey data show considerable so-called “reciprocity” of violence.
 

On the basis of this literature review, I published a paper in 1995 in which I argued that these two sampling tactics provide access to decidedly different, virtually non-overlapping populations of violent couples, that there are two quite different types of partner violence, one gender-symmetric, the other decidedly, if not entirely, male (Johnson, 1995)
 

This is where we return to the initial question for this session, the one about the interpersonal roots of conflict.What are the interpersonal roots of these two types of violence?I have argued that the characteristics of the heavily male type of violence are consistent with a general motive to control one’s partner, a motive that is rooted in patriarchal ideas about relationships between men and women.The violence is used often because it needs to be, in order to subdue one’s partner or to display one’s power and control.Similarly, it escalates as required, an escalation that is likely to be necessary in a culture in which women in most cases are not willing to concede all power and control to their husbands.Finally, such a general pattern of power and control is likely in the long-run to subdue physical resistance.I labeled this type of partner violence “patriarchal terrorism.”
 

The other type of violence, which is more gender-symmetric, is consistent with a more specific, narrowly-focused motive to get one’s way in a particular conflict situation, within a relationship in which there is not a general pattern of power and control, but in which specific arguments sometimes escalate into violence.Since it does not involve a general motive to control, it is less frequent, it does not escalate over time (in fact it is likely to de-escalate), and the violence is more likely to be reciprocated.I labeled it “common couple violence.”
 

The defining difference between these two types of violence is a difference of motives and the interpersonal dynamics that those motives produce across the many interactions that comprise a relationship.The defining characteristic of patriarchal terrorism is a general motive to control that activates a range of power and control tactics in addition to the use of violence.One model of such a pattern is the “Power and Control Wheel” that is used to characterize domestic violence in the shelter movement in which I am active.The model was developed in the Duluth batterers-education project on the basis of reports from shelter clients regarding the nature of their relationships with their abusers (Pence & Paymar, 1993).The general point of the model is that the violence in this pattern is only one of many control tactics employed in the service of a motive to exert general control over one’s partner.Thus, if we want to distinguish this type of violence from common couple violence, our surveys need to ask questions not just about violence, but about a variety of control tactics, and this is where the measurement issue returns.
 

All of this post hoc speculation is based only on the differences in the gender, frequency, escalation, and reciprocity of the violence uncovered bythe two types of sampling strategies.The literature that I reviewed in 1995 presented no direct evidence regarding the general use of a variety of control tactics.To test these ideas, I needed data that (1) included questions regarding a variety of control tactics in addition to violence, (2) were likely to include both patriarchal terrorism and common couple violence, and (3) included information regarding both spouses.I turned to the internet, and in response to my query to a feminist list, Irene Frieze replied that she had a late 1970s data set that might meet my needs, and she sent me the data.
 

The data come from interviews with married or formerly-married women living in southwestern Pennsylvania in the late 1970s.The mixed sampling design seemed likely to include both types of violence, since it included on the one handwomen from shelters and the courts, on the other hand a matched sample of women who lived in the same neighborhoods (Frieze, 1983; Frieze & Browne, 1989; Frieze & McHugh, 1992).My hope was that the court and shelter samples would include patriarchal terrorism, and that the neighborhood sample would include some cases of common couple violence.
 

From the lengthy interview protocols, I identified items tapping seven non-violent control tactics: threats, economic control, use of privilege, using children, isolation, emotional abuse, and sexual control.Each of these measures was standardized and they were entered into a cluster analysis (for methodological details see Johnson, 1999).The results indicated a two-cluster solution as optimal, and as you can see in Table 1, the pattern is quite simple, with one cluster exhibiting a high average on all seven of the control tactics, the other being relatively low on all seven.




Table 1: Control Tactics by Cluster

(Reports on both men and women from wives, n = 274)

 
 

Control Tactics

 
Economic
Use of
Using
Emotional
Sexual
Threats
Control
Privilege
Children
Isolation
Abuse
Control
High Control
Mean
3.95
1.61
2.73
1.41
3.82
1.48
1.48
(n=109)
(Z)
(1.19)
(.97)
(1.04)
(.96)
(.95)
(1.07)
(1.29)
Low Control
Mean
1.96
1.19
1.79
1.09
2.77
1.09
1.03
(n=439)
(Z)
(-.29)
(-.26)
(-.26)
(-.27)
(-.24)
(-.28)
(-.32)
Eta
.59
.49
.52
.49
.48
.53
.65


Are there violent people in both clusters?Remember, the clusters are defined by patterns of non-violent control tactics.By simply combining each individual’s cluster membership with information regarding whether they had ever been violent in their marriage, we can distinguish three types of spouses: (1) non-violent spouses, (2 those involved in non-controlling violence, and (3) those involved in controlling violence.You can see in Table 2 that the controlling violence is almost entirely male, while the non-controlling violence is not.

Table 2: Type of Individual Violence by Gender

(Data on 272 husbands and 271 wives, as reported by wives)

Husbands
Wives
N
Nonviolent
42%
58%
212
Non-Controlling Violence
38%
62%
224
Controlling Violence
98%
2%
107

Now, how do we identify terrorism and common couple violence?They are dyadic phenomena that require attention to the behavior of both partners in the marriage.Table 3 presents data only on violent individuals, that is, those who had been violent in their relationship at least once.It places individual violence within its dyadic context, distinguishing among four types of violent behavior.The first row, “mutual violent control,” refers to

Table 3: Individual Violent Behavior in a Dyadic Context
(Violent individuals only, as reported by wives)
Husbands
Wives
N
Mutual Violent Control
50%
50%
10
(Patriarchal) Terrorism
97%
3%
97
Violent Resistance
4%
96%
77
Common Couple Violence
56%
44%
146

controlling violence in a relationship in which both spouses are violent and controlling.There were only five such couples in this data set.The second row, “terrorism” with the “patriarchal” in parentheses, refers to relationships in which only one of the spouses is violent and controlling, the other is not.In this data set, that violent and controlling spouse is the husband in 94 of the 97 cases, cases in which the term “patriarchal terrorism” is appropriate.The third row refers to cases in which the focal spouse is violent but not controlling, and his or her partner is both violent and controlling.I call it “violent resistance,” and it is almost entirely a female form of violence in this sample of heterosexual relationships.Of course that is because in these marriages almost all of the terrorism is male, and in some cases the wives do reply with violence.Finally, we have “common couple violence,” individual non-controlling violence in a dyadic context in which neither of the spouses is violent and controlling.
The relationship between gender and these types of violence supports the hypothesis that “terrorism” is indeed an almost exclusively male phenomenon in heterosexual marital relationships, thus appropriately referred to as “patriarchal terrorism,” while common couple violence is close to gender-symmetric, at least by these crude criteria.(Data on the frequency and severity of male and female common couple violence—not shown—indicate that by other criteria men are more violent than women even within common couple violence.)
Now, let’s take a quick look at the characteristics ofterrorism and common couple violence, to see if they it fall in line with what I found in my 1995 literature review.(1) The gender differences are clear.Patriarchal terrorism is 97% male, common couple violence only 56%.Since there are only three female terrorists in these data, I am going present the last three pieces of data looking only at violent men.It was, of course, male violence that was evaluated in the literature review cited above.(2) The violence in patriarchal terrorism is quite frequent; the median number of incidents in these marriages is 58.It is much less frequent in common couple violence with a median of 14 violent events.(3) Patriarchal terrorism is reported to escalate in 76% of the cases, common couple violence in only 28%.(4) In most violent incidents in patriarchal terrorism, wives do not physically resist; the median of the ratio of number of times the wife has been violent to the number of times the husband has been violent is .17.In common couple violence they are much more likely to resist, the median ratio being .40.These data do not leave much doubt that patriarchal terrorism and common couple violence are not the same phenomenon.
Finally, let me nail down my sampling argument, that general survey samples allow conclusions only about common couple violence, while “shelter” studies give access only to patriarchal terrorism.Again, looking only at male violence, Table 4 shows that the general sample accesses almost nothing but common couple violence, only 11% of the violence being patriarchal terrorism.In stark contrast, the court and shelter samples that are typical of feminist research get at violence that is predominantly patriarchal terrorism.



Table 4: Who’s Finding Whom?
(Violent husbands only, as reported by wives)
Survey Sample
Court Sample
Shelter Sample
(n = 37)
(n = 34)
(n = 43)
Mutual Violent Control
0%
3%
0%
Patriarchal Terrorism
11%
68%
79%
Violent Resistance
0%
0%
2%
Common Couple Violence
89%
29%
19%

What kind of mistakes can this generate?The original Steinmetz article is an excellent example of the danger of thinking that a general sample provides information about what we conventionally mean by domestic violence (i.e., patriarchal terrorism).Steinmetz tells anecdotes about true husband-battering, women who are terrorists in the sense described above.She then cites general survey data, which we now know represents only common couple violence, to make her case that such battering is as serious a problem as is wife-battering.The data she presents in fact have nothing to do with husband-battering.Serious as husband-battering is in the particular case, as a general phenomenon it is dramatically less frequent than wife-battering.
Similarly, one can err by assuming that the patterns observed in shelter samples describe all partner violence.It is common for shelter workers to argue in our educational programs that violence always escalates, that if he hit you once he’ll do it again and it will get worse.That pattern is, as shown above, much more true of patriarchal terrorism than of common couple violence, and patriarchal terrorism is what we see in our shelter work.However, most couples who experience violence, including those in our audiences, are involved in common couple violence and for those audience members we are providing an inaccurate picture of the likely course of their relationships..

 

Dramatic as these sampling biases are, Table 4 also shows that neither type of male violence is found exclusively in one type of sample, implying that it is possible to study the differences between these two types of violence in a variety of research settings.First, the 11% of male violence in the general sample that is patriarchal terrorism indicates that with large enough samples, it may bepossible to study both common couple violence and patriarchal terrorism with survey data.In order to do that, of course, we need to include questions that will allow us to distinguish one from the other.Second, since women do bring cases of common couple violence both to the courts (29%) and to shelters (19%), researchers in those contexts will be able to study the effects of various intervention strategies on the two types of violence.Again, however, it won’t happen unless we gather information that will allow us to make these distinctions.
 

Why bother?Well, I hope you are convinced by now that we have to make these distinctions if we want to understand partner violence.The evidence you have just seen regarding the dramatic differences between patriarchal terrorism and common couple violence in terms of gender, per couple frequency of incidents, escalation, and reciprocity should serve as a warning that until further notice we have to assume that the answers to all of our important questions may be different for the two different forms of violence.
 

For example, “What are the interpersonal roots of couple conflict?”For common couple violence, we may need to look to issues of effective communication or anger management.The roots may be found in the interaction processes involved in the current relationship.For patriarchal terrorism, we probably need to look somewhere other than the current relationship, at the origins of the husband’s desperate need to control his wife.A number of studies suggest that the origins of such needs may be different for different men (Holtzworth-Munroe & Stuart, 1994; Jacobson & Gottman, 1998).
 

Or what about our second question for today: “What are the consequences for individuals and couples?”On the one hand, general survey studies, which I have shown tap primarily common couple violence, often report a surprising lack of any relationship between violence and marital satisfaction or stability.On the other hand, shelter studies, which tap primarily patriarchal terrorism, suggest deep dissatisfaction, and almost inevitable dissolution.But, of course, the ultimate consequences depend in some cases on the effectiveness of interventions.
 

Which brings me to my final point, and to one of the important questions addressed in the Bradbury, Rogge, and Lawrence paper: “What are likely to be the most effective intervention strategies?”For common couple violence, the prognosis may be fairly positive, and anger management approaches and couples therapy of various kinds might be reasonable strategies.
 

The prognosis for intervention in patriarchal terrorism, however, is not good.So far, batterers reform programs have a dismal record of success.Intervention in patriarchal terrorism has to focus instead on the woman’s safety.Women almost always do leave such relationships, as soon as they can put together the information and the financial resources they need to escape to a reasonably safe life for themselves and their children.Of course, that is what the women’s shelter movement is all about.So, my final plea is (1) make distinctions, and (2) support your local women’s shelter.



Bibliography

Bradbury, T., Rogge, R., & Lawrence, E. (2000). Reconsidering the role of conflict in marriage.  In A. Booth, A. C. Crouter, & M. Clements (Eds.), Couples in Conflict . Hillsdale, NJ: Lawrence Erlbaum.
 

Dobash, R. E., & Dobash, R. P. (1979). Violence Against Wives: A Case Against Patriarchy. New York: Free Press.
 

Frieze, I. H. (1983). Investigating the causes and consequences of marital rape. Signs, 8(3), 532-553.
 

Frieze, I. H., & Browne, A. (1989). Violence in marriage. In L. Ohlin & M. Tonry (Eds.), Family violence (Vol. 11, pp. 163-218). Chicago: The University of Chicago Press.
 

Frieze, I. H., & McHugh, M. C. (1992). Power and influence strategies in violent and nonviolent marriages. Psychology of Women Quarterly, 16(4), 449-465.
 

Holtzworth-Munroe, A., & Stuart, G. L. (1994). Typologies of male batterers: Three subtypes and the differences among them. Psychological Bulletin, 116(3), 476-497.
 

Jacobson, N., & Gottman, J. (1998). When Men Batter Women: New Insights into Ending Abusive Relationships. New York: Simon & Schuster.
 

Johnson, M. P. (1995). Patriarchal terrorism and common couple violence: Two forms of violence against women. Journal of Marriage and the Family, 57(May), 283-294.
 

Johnson, M. P. (1998, June). Commitment and entrapment. Paper presented at the Ninth International Conference on Personal Relationships, Saratoga Springs, NY.
 

Johnson, M. P. (1999, November). Identifying patriarchal terrorism and common couple violence. Paper presented at the National Council on Family Relations, Irvine, CA.
 

Johnson, M. P., & Ferraro, K. J. (2000). Research on domestic violence in the 1990s: The discovery of difference. Journal of Marriage and the Family, 62(4).
 

Kirkwood, C. (1993). Leaving Abusive Partners: From the Scars of Survival to the Wisdom for Change. Newbury Park, CA: Sage.
 

Pence, E., & Paymar, M. (1993). Education Groups for Men Who Batter: The Duluth Model. New York: Springer.
 

Steinmetz, S. K. (1977-78). The battered husband syndrome. Victimology, 2(3-sup-4), 499-509.
 

Straus, M. A. (1990). Injury and frequency of assault and the 'representative sample fallacy' in measuring wife beating and child abuse. In R. J. Gelles & M. A. Straus (Eds.), Physical Violence in American Families: Risk Factors and Adaptations to Violence in 8,145 Families (pp. 75-91). New Brunswick, NJ: Transaction.
 

Wilson, M., & Daly, M. (2000). The evolutionary psychology of couple conflict in registered versus de facto marital unions. In A. Booth, A. C. Crouter, & M. Clements (Eds.), Couples in Conflict . Hillsdale, NJ: Lawrence Erlbaum.