Elevated sexual activity in violent marriages: hypersexuality or sexual extortion?

The eroticization of violence is a frequent theme in modern film. Imagery linking sexual excitement to control, domination, exploitation, and rape is common not only in pornographic videos but in mainstream movies as well (Cowan, Lee, Levy, & Snyder, 1988; Palys, 1986; Yang & Linz, 1990). The extent to which such imagery mirrors the dynamics of sexual attraction in real life is unclear. On the one hand, researchers have found that sexual attraction and sexual frequency are highest in an atmosphere characterized by strong affective bonds and an absence of conflict (Blumstein & Schwartz, 1983). On the other hand, recent studies also show marriages characterized by violence on the part of one or both partners to be more sexually active than others (Apt & Hurlbert, 1993; Donnelly, 1993). Do such findings suggest that aggression fans sexual excitement, or perhaps that sexual excitement elicits aggression? Although both are plausible, two other hypotheses appear to be more compelling, based on previous theory and research.

The first is what I will term the "hypersexuality hypothesis." It presumes that an elevated sexual appetite (i.e., hypersexuality) is linked to aggressivity via some underlying proclivity for impulsive, unconventional, or risk-taking behavior. The existence of such an orientation is predicted by several lines of research and theory, including studies of problem behavior in adolescence and young adulthood, general theories of deviance, and investigations of the impact of hormones--in particular, testosterone--on a wide range of behavioral phenomena. In the process of assortative mating, individuals with a proclivity for aggression and an elevated sexual appetite are likely to pair up with similar members of the opposite sex (for example, see Simons, Johnson, Beaman, & Conger, 1993, for evidence that aggressive women tend to marry similar men). These persons' marriages are then subsequently characterized by both a greater intensity of conflict and violence and greater levels of sexual activity. In this case, the association of violence with elevated sexual activity would be purely spurious.

A more sinister possibility is that men, in particular, are able to use violence--whether consciously or not--to exercise complete control over sexual access to their wives. Although marital rape has been the focus of considerable attention in recent years (Finkelhor & Yllo, 1985; Russell, 1982), sexual coercion can take place without the occurrence of such a direct assault. Wives who have been the victims of their husbands' violence at some point in their marriage have reported often "giving in" to their husbands' sexual demands largely out of the fear of displeasing them (Finkelhor & Yllo, 1985; Russell, 1982). In that wives' sexuality is being "extorted" in this fashion with the implicit--or perhaps even explicit--threat of violence in the event of noncompliance, I term this explanation of the violence--sexual frequency association the "sexual extortion" hypothesis. In this case the role of violence is causal: A husband's history of violence leads wives to engage in sexual activity at higher rates than would otherwise occur.

In this study, I employed a large, recent national probability survey of the United States population to investigate whether hypersexuality or sexual extortion accounts for the association between marital violence and elevated sexual activity. Both married and unmarried cohabiting couples are included in the study. However, the terms husband, wife, and marriage are used throughout to refer generically to unmarried as well as married intimate partners and their relationships. Although the survey entailed a rather extensive set of questions about the quality of the marital relationship, including questions regarding sexual frequency, sexual satisfaction, and the occurrence of violence, respondents were not asked about sexual assault. For this reason, my approach in this article is more indirect. I argue that each hypothesis--hypersexuality versus sexual extortion--would lead one to expect specific patterns among other individual or relationship factors about which data were collected. I begin by reviewing relevant theory.

Theoretical Considerations

Relationship Violence and Sexual Frequency

The regularity of sexual relations in marriage has been found in previous research to be governed by diverse factors. Perhaps foremost among these are the age and health of the partners. Sexual activity tends to decline with the advancing age of either partner (Udry, Deven, & Coleman, 1982; Udry & Morris, 1978) or deteriorating health--particularly among males (Rao & DeMaris, 1995). A decline is also seen to occur with increasing relationship duration (James, 1981; Jasso, 1985; Rao & DeMaris, 1995; Udry, 1980), which is unrelated to either health or age. Several scholars have attributed this change to the loss of novelty associated with the newlywed phase of a relationship (Greenblat, 1983; James, 1981). Other predictors of enhanced sexual activity are an absence of children--or at least young children--in the household (James, 1974; Jasso, 1985), wives who are not currently pregnant (Jasso, 1985; Rao & DeMaris, 1995), cohabiting unmarried (Huey, Kline-Graber, & Graber, 1981; Newcomb, 1986; Rao & DeMaris, 1995), and high marital satisfaction (Blumstein & Schwartz, 1983; Rao & DeMaris, 1995).

Recently, two studies have shown that the presence of physical violence between partners is associated with elevated sexual frequency. Donnelly (1993), investigating sexually inactive marriages using the National Survey of Families and Households 1987-88 (NSFH), found that marriages in which there had been "physical arguments" within the past year were more likely than others to be sexually active. Apt and Hurlbert (1993) compared a sample of physically abused married women with a demographically matched sample of nonabused wives. The abused women rated themselves as having a lower degree of sexual assertiveness, arousability, and satisfaction. Nonetheless, they reported significantly greater frequency of sexual intercourse with their husbands than did the nonabused wives.

Both studies controlled for obvious correlates of both sex and violence, most notably, age. Younger men and women are both more sexually active and more likely to be violent. The question therefore remains as to why sex and violence are so associated. Apt and Hurlbert (1993) maintained that this correlation is strictly due to the power and control exercised by abusing husbands over their wives in all aspects of the marriage, including sex. Hence, in violent marriages, the husband controls when and how often the couple has sex. This is what I refer to as the sexual extortion explanation. Donnelly (1993), on the other hand, entertained this, as well as an alternative possibility. She theorized that some couples are simply more likely to express their emotions physically, whether love or anger. Hence, the association between sex and violence is spurious. I call this the hypersexuality explanation.

Donnelly's (1993) third explanation for this association involves what is referred to as the "cycle of violence" (Deschner, 1984, p. 12; Donnelly, 1993, p. 177), a phenomenon observed by both researchers and clinicians. In relationships characterized by violence, there is first a tension-building phase, followed by an outburst of aggression, and finally a "honeymoon" or repentance phase in which partners are especially loving to each other. Donnelly suggested that higher rates of sexual activity among violent couples may be due to the honeymoon stage of this cycle.

None of these explanations has yet been subjected to test. In the following sections I provide the theoretical underpinnings for tests of the hypersexuality and sexual extortion hypotheses. Unfortunately, the data preclude evaluation of the cycle-of-violence hypothesis because they do not allow a precise determination of the timing of sexual congress in relation to instances of violence. Therefore, this hypothesis will not be tested in the current study.

Noncausal Linkages Between Violence and Sexual Activity

To suggest ways in which heightened sexual activity and violence might be noncausally linked, I will briefly review three lines of research. Each relies on an "underlying" factor to establish associations among diverse behaviors, including aggression and sexual activity. Insofar as this factor is antecedent to both hypersexuality and hyperaggressivity and leads to their association across individuals, the relationship between violence and elevated sexual frequency is spurious.

Problem-behavior theory. Jessor and Jessor (1977, p. 33) defined problem behavior as "behavior that is socially defined as a problem, a source of concern, or as undesirable by the norms of conventional society and the institutions of adult authority, and its occurrence usually elicits some kind of social control response." These scholars theorized that the tendency to engage in problem behaviors is the result of a balance between tendencies toward and inhibitions against performing these behaviors in the individual's social world. That world, moreover, involves the interaction of forces belonging to three major systems of psychosocial variables: the personality system, the perceived environment system, and the behavior system. Variables within each system represent either "instigations" or "controls." Instigations, such as attitudinal tolerance of deviance, enhance the likelihood of engaging in a particular behavior. Controls, such as religiosity, on the other hand, reduce that likelihood. The interplay of instigations and controls in each psychosocial system results in a dynamic state called "psychosocial proneness." This reflects the overall likelihood of occurrence of problem behavior (Jessor, Donovan, & Costa, 1991; Jessor & Jessor, 1977).

In several studies of adolescents and young adults, Jessor and his colleagues have focused on problem drinking, substance abuse, cigarette smoking, "precocious" or early sexual experimentation, and general deviance--including violence--as the primary problem behaviors exhibited by youth. These behaviors have been consistently shown to be intercorrelated across individuals in samples of both adolescents and young adults, with correlations ranging as high as .67, suggesting an identifiable "syndrome" of problem behavior that characterizes certain people (Donovan & Jessor, 1986; Jessor et al., 1991; Jessor & Jessor, 1977). That is, participation in a particular behavior usually implies having engaged in other, similar behaviors. Confirmatory factor analyses involving continuous measures of these behaviors indicate a single underlying factor, which Jessor and his colleagues have labeled unconventionality (Donovan & Jessor, 1985; Jessor et al., 1991).

Control theory. Gottfredson and Hirschi's (1990) general theory of crime relies on a single personality trait, self-control, as the most proximate factor in the genesis of criminal and analogous behaviors. Low self-control doesn't necessarily impel people to commit crime. Rather, it is held to enable criminal behavior by virtue of removing the normal constraints against crime experienced by others. Compared to those with high self-control, those low on this attribute tend to attach more value to the immediate gains and have less dread of the long-term costs associated with destructive behavior. Self-control is developed early in an individual's biography via the techniques of parental socialization. Once developed, this trait is held to be relatively stable throughout an individual's life.

Low self-control is manifested in a variety of behaviors, many of which are imprudent, but not particularly criminal. The characteristics of those with low self-control, as described by Gottfredson and Hirschi, are a tendency to seek immediate gratification of desires; lack of diligence, tenacity, or persistence in a course of action; a tendency to be adventuresome, active, and physical; a tendency to have unstable marriages, friendships, and job profiles; a tendency to have little interest in, and be unprepared for, long-term occupational pursuits; a tendency to be self-centered and indifferent or insensitive to the suffering and needs of others; a tendency to pursue immediate pleasures that are not criminal, such as smoking, drinking, using illicit drugs, gambling, having children out of wedlock, and engaging in illicit sex; a tendency to have minimal tolerance for frustration and little ability to deal with conflict except through violence; and a tendency to be tolerant of physical pain or indifferent to physical discomfort.

Tests of Gottfredson and Hirschi's (1990) formulation have generally been supportive of the theory. Several scholars have found measures of self-control, as articulated by Gottfredson and Hirschi, to be adequately described by a one-factor model (Arneklev, Grasmick, Tittle, & Bursik, 1993; Grasmick, Tittle, Bursik, & Arneklev, 1993; Wood, Pfefferbaum, & Arneklev, 1993), suggesting that this trait is unidimensional. Research also supports the notion that imprudent, as well as delinquent, behaviors are correlated with measures of self-control in the manner hypothesized by Gottfredson and Hirschi (Arneklev et al., 1993; Grasmick et al., 1993; Wood et al., 1993). In that sexual adventurousness (e.g., having children out of wedlock, engaging in illicit sex) and violence are both described as manifestations of this construct, self-control is a second factor that could render the association between these variables spurious.

Testosterone, aggression, and sexuality. A number of studies have shown that the presence of elevated testosterone levels is associated with enhanced levels of libido, aggression and dominance, sensation seeking, depressed occupational achievement, and antisocial behavior in both sexes (see Dabbs, 1992a, for a comprehensive review of this research). For example, Dabbs and his colleagues (Booth & Dabbs, 1993; Dabbs, 1992b; Dabbs & Morris, 1990) measured testosterone levels in 4,462 former military servicemen in 1985-1986. Higher testosterone concentrations were associated with lower occupational status. This relationship was further mediated by the association of testosterone level with antisocial behavior, low intelligence, and low educational attainment (Dabbs, 1992b). Individuals higher in testosterone, compared to others, more often reported having trouble with parents, teachers, and classmates; being assaultive toward other adults; going AWOL in the military; using hard drugs, marijuana, or alcohol; and having numerous sexual partners (Dabbs & Morris, 1990). Marital experience was also related to testosterone levels. Higher concentrations of this hormone were associated with men being less likely to marry, more likely to have separated or divorced, more likely to report having engaged in extramarital sex, and more likely to have hit or thrown things at their spouses (Booth & Dabbs, 1993). In sum, violence and hypersexuality could be linked in both sexes via a biological predisposition for both behaviors.

Causal Linkages Between Violence and Sexual Activity

In contrast to violence and sexual activity being linked by an underlying predisposition for both, the occurrence of violence may itself enhance subsequent levels of sexual activity. In this instance, the role of violence is causal. In particular, it is possible that a history of physical aggression on the part of husbands leads wives to accede to sexual activity at higher rates than they would otherwise prefer. This may be because violence is used directly to obtain sex, as in marital rape. Or the potential for dissatisfied husbands to be violent may simply provide wives with a strong incentive to accede to their husbands' sexual advances.

Many wives report feeling compelled to have unwanted sex, even when the degree of force or coercion involved is not sufficient for the act to qualify officially as rape (Finkelhor & Yllo, 1985; Russell, 1982). Finkelhor and Yllo (1985), in fact, distinguished four types of coerced sex in marriage. Social coercion is the pressure women feel to have sex with their husbands because of social expectations. Interpersonal coercion is experienced by wives who have sex with their husbands in the face of nonviolent threats. Such threats take the form of threatening to leave the marriage, have affairs with other women, or withhold money or other resources desired by the wife. Threatened physical coercion involves gaining compliance through either direct threats of violence or the potential for violence to occur if the wife does not cooperate. Finally, physical coercion involves husbands physically restraining their wives or being violent with them to accomplish sexual intercourse. I use the term sexual extortion to refer to the last two types of coercion. In either case, sexual access is being obtained either by direct force or the implicit or explicit threat of violence.

Although the focus of this article is on sex forced upon women by men in the context of an intimate relationship, men also report giving in to unwanted sexual demands under various types of pressures. However, the nature of such "force" is nowhere near as severe as it can be for women. Given differences in size, strength, and fighting ability, it is very unlikely that men fear violent recriminations from their partners in the event of noncompliance. Men are also more likely than women to report feeling good about unwanted sexual experiences, and to say that these had no lasting effects on their lives (Muehlenhard, 1989).

Hypotheses

The first hypothesis to be tested in this study pertains to the association between relationship violence and sexual frequency: Sexual activity is higher among couples in which either the husband or the wife has been physically aggressive with the spouse.

Two alternative sets of hypotheses will be examined to account for the association of violence with more frequent sexual intercourse. The first set encapsulates the hypersexuality hypothesis. This entails the notion that certain individuals are simply more prone to both a more active schedule of sexual activity and a higher likelihood of interpersonal violence. If this is the case, the impact of the husband's or the wife's violence on a couple's sexual frequency should be minimal once factors that index a proclivity for both sex and violence (hereafter called hypersexuality factors) are held constant. The second hypothesis is therefore Controlling for the hypersexuality factors, violence on the part of either husband or wife is unrelated to the couple's sexual activity.

The second possible explanation for an association between relationship violence and sexual activity is that wives are being physically intimidated into having sex more regularly than they prefer. The key, in this case, is that under this hypothesis, only husbands' violence should be predictive of elevated sexual frequency. Controlling for husband's violence, I would not expect wives' violence to be predictive of greater rates of intercourse. The first hypotheses suggested by this sexual extortion explanation are therefore (a) Husband's violence predicts more frequent sexual activity, even after holding wife's violence constant, and (b) Holding husband's violence constant, wife's violence is unrelated to sexual frequency.

Moreover, under the sexual extortion thesis, I expect husbands' violence and sexual frequency to interact in their effects on the sexual satisfaction of wives. Ordinarily, there should be a positive association between sexual frequency and wife's sexual satisfaction. Yet when women are compelled to have sex by either force or the fear of violent reprisals for noncompliance, more sexual activity should be predictive of less sexual satisfaction. The next hypothesis is therefore An interaction between husband's violence and sexual activity in their effects on sexual satisfaction occurs such that sexual frequency is positively related to wife's sexual satisfaction when a husband has not been violent, but negatively related to wife's sexual satisfaction when a husband has been violent.

The next set of sexual extortion hypotheses pertains to the impact of sexual activity on wives' depressive symptomatology. Whether one would expect an association between these variables depends on which factor is antecedent to the other. Wives who become depressed might be expected to experience a reduction in subsequent sexual desire and concomitant sexual activity. Therefore, if depression is antecedent, these two variables might be negatively related. Normally, however, one would not expect a couple's sexual frequency to affect the wife's subsequent level of depression. That is, if sexual frequency is the antecedent variable, one would expect no association between it and depression. In either case, it would not be reasonable to expect a positive relationship between sexual frequency and depressive symptomatology. But when wives' compliance in sexual activity is being extorted through the use, or the threat, of husbands' violence, such an association would indeed be likely. Forced sex essentially implies a loss of control over one's body, leaving women feeling vulnerable and powerless. The inability, in turn, to exercise control over either good or bad outcomes in one's life has been shown to be a consistent predictor of depressive symptomatology (Mirowsky & Ross, 1990). The hypotheses are therefore (a) In the absence of husband's violence, sexual frequency has either no effect or a negative effect on the wife's depressive symptomatology, and (b) Given the occurrence of violence by the husband, the association between sexual frequency and depression is positive.

The last set of sexual extortion hypotheses concerns how physical injury to the wife influences both interaction effects just referred to. The notion of sexual extortion rests on the idea that a husband's violence enhances sexual frequency in marriage because it creates a pervasive climate of fear for his wife. The greater that fear, the more likely she is to deny her sexual appetites in an attempt to please her husband. To the extent that she experiences physical injury as a result of her husband's assaults, she should be more fearful than otherwise (see, for example, DeMaris & Swinford, 1996). Fear, moreover, should be at its zenith when the husband is the only violent partner and the wife has been injured. When wives are also (or exclusively) violent, self-efficacy theory (Bandura, 1992; Ozer & Bandura, 1990) suggests that they should experience less fear arousal. In many cases they are, in fact, the initiators of the physical confrontation (Straus, 1993). However, even if they were not the initial aggressors, evidence suggests that women who can "hit back" tend to feel more confident in their ability to defend themselves when attacked. These women are correspondingly less fearful of future attacks than their more defenseless counterparts (Ozer & Bandura, 1990). Higher levels of fearfulness should, in turn, both interfere more with sexual satisfaction and induce higher levels of depressive symptomatology, The hypotheses are (a) Sexual frequency is more negatively related to the wife's sexual satisfaction when only the husband was violent and the wife was injured than when both partners were violent and the wife was injured or only the husband was violent and the wife was not injured, and (b) Sexual frequency is more positively related to the wife's depressive symptomatology when only the husband was violent and the wife was injured than when both partners were violent and the wife was injured or only the husband was violent and the wife was not injured.

Method

Participants

The data for this study were drawn from the National Survey of Families and Households 1987-88 (NSFH), a multistage probability sample of 13,008 adults in the noninstitutionalized U.S. population. Information was, for the most part, obtained from one adult randomly selected from all adults (aged 19 and older) in the household. This primary respondent was interviewed face to face, although portions of the interview schedule--for example, those containing sensitive questions--were self-administered. A shorter, self-enumerated questionnaire was given to spouses and cohabiting partners when present. Several subpopulations were oversampled: unmarried cohabitors, recently married couples, single-parent families, families with stepchildren, and households whose heads were Black, Mexican-American, or Puerto Rican. Further details concerning the design and content of this study can be found in Sweet, Bumpass, and Call (1988).

Several selection criteria were employed to delimit the subsample for the current analysis. I restricted the sample to female primary respondents because wives' depression was a key variable in the analysis, and this was only measured for primary respondents. I further restricted the sample to households in which the primary respondent was married and living with the spouse, or cohabiting unmarried, and the partner's questionnaire was returned completed. Moreover, because my theoretical arguments are couched in terms of gendered behavior, I excluded same-sex couples. Couples in which either partner was in the armed forces were also eliminated. The dynamics of intimate relationships are likely to be different among those who are part of the military subculture. However, with only 68 primary respondents in the armed forces, sample sizes are not sufficient to investigate model differences across groups. Finally, I eliminated couples who were missing data on the key variables of sexual frequency, male and female violence, wife injury, wife's depression, or wife's sexual satisfaction. The final subsample for this analysis therefore consisted of 2,435 couples.

Measures

The dependent variables. Each spouse was asked how many times in the past month the couple had "had sex." Although many would interpret this phrase as referring specifically to sexual intercourse (see, for example, Call, Sprecher, & Schwartz, 1995), this measure captures varieties of sexual interaction other than coitus (Donnelly, 1993). Thus, sexual activity or sexual frequency as used in this article always refers to the wider panoply of behaviors encompassed by the notion of sexuality, including, of course, coitus. Each partner's report of sexual frequency can be considered an indicator of the couple's true frequency of sexual activity--a latent variable. Therefore, I used the mean of the partner's reports as my scale of sexual frequency, with an alpha reliability of .81.

The wife's sexual satisfaction was tapped with a single item. Respondents were asked to imagine that they were to separate from their partners. They were then queried concerning how they expected "things to change" in each of several areas, one of which was their "sex life." Responses ranged from "much worse" to "much better," coded, respectively, from 5 to 1. Higher scores were assumed to indicate a more satisfactory sexual relationship with the current partner.

Wives' depression was measured with a 12-item version of the Center for Epidemiological Studies Depression scale (CES-D). This scale has been used extensively in previous research to tap depressive symptomatology (see, for example, Glass & Fujimoto, 1994; Ross, Mirowsky, & Huber, 1983). It presents the respondents with 12 symptoms and asks them to indicate the number of days during the past week they experienced each symptom. A typical item is "On how many days during the past week did you feel that you could not shake off the blues even with help from your family or friends?" The response format for each item is 0-7. The scale therefore ranges from 0 to 84, with 84 indicating maximum depressive symptomatology. Although a high score does not necessarily indicate a diagnosis of clinical depression, the scale has been shown to discriminate well between clinically depressed patients and others, and it correlates highly with other depression rating scales (Ross et al., 1983). The alpha reliability of the scale was .92 in the current sample. Following Glass and Fujimoto (1994), 1 constructed a depression score for every wife who had at least 8 valid observations among the 12 items. The CES-D score was calculated as the mean of the items answered, times 12. The result is equivalent to the sum of the 12 items.

The focus variables. Both partners were asked whether either spouse had hit , shoved, or thrown things at the other in the past year and whether the wife was cut, bruised, or seriously injured in a fight with her husband. To code whether male aggression occurred, I examined both husband's and wife's reports concerning the husband's Violence. If both responses were missing, male violence was coded as missing for the couple. If both agreed that the husband had not been violent or if one partner reported that the husband had not been violent and the other partner's response was missing, husband's violence was coded, as "no." On the other hand, if at least one partner reported that the husband had been violent, husband's violence was coded as "yes." This coding strategy follows the rationale that violence is more likely to be under-than overreported. Moreover, using both partners' reports concerning the same phenomenon guards somewhat against self-serving bias in the responses. The coding of female aggression and female injury was handled in the same way.

The wife injury variable posed some problems. Among the 309 couples reporting physical arguments in the past year, 71, or about 23%, reported that the wife was injured as a result. (For comparison, 42, or 14% of couples, reported that the husband was injured.) Generally, no injury was reported unless at least one partner indicated that the marriage had been violent. There were nevertheless two couples reporting that the wife was injured even though neither partner was reported to have been aggressive. This is, however, plausible because all of these couples indicated having "physical" arguments, despite some couples (in fact, 69 of them) being reluctant to identify either partner as the aggressor. More importantly, cell sizes are very small for certain combinations of categories of wife injury and aggressor. For this reason, I have assumed that sexual frequency is linear in its interaction with partner violence and wife injury (i.e., the partial slope for sexual frequency becomes a + b * husband violence + c * wife violence + d * wife injury) in their effects on both sexual satisfaction and depression. That is, only the first-order interactions between husband violence, wife violence, and wife injury, on the one hand, and sexual frequency, on the other, are used in the following analyses. Higher-order interactions involving wife injury would involve cell sizes that are too small for reliable estimation of parameters. However, the crossproduct of husband with wife violence was examined in models for sexual frequency, sexual satisfaction, and depression. Also, the second-order interaction effect created by the crossproduct of husband violence, wife violence, and coital frequency was examined in models for sexual satisfaction and depression. In all cases, the results were essentially unchanged from those in Tables 1, 2, and 3 and, consequently, have not been reported.

Table 1
Unstandardized Coefficients for the Regression of Sexual Frequency
on Focus, Control, and Hypersexuality Factors(a)

                                                Model 1
Predictor                                     b      t Ratio
Focus Variable
 Husband has been violent                    2.56        2.68(**)
Control Variables
 Disagreement scale                          -.17       -4.44(***)
 Wife wants to have more children soon(b)    -.86       -1.84
 Wife wants to have more children later(b)  -1.56       -1.80
 Wife currently pregnant(b)                 -2.50       -2.63(**)
 Husband's education                         -.15       -2.71(**)
 Husband 5+ yrs. older than wife(c)         -1.07       -3.01(**)
 Cohabiting couple                           4.14        6.82(**)
 Partners both Hispanic(d)                  -1.73       -2.26(*)
 Other race combination(d)                   1.79        2.96(**)
 Wife's health                                .78        4.22(***)
 Husband's health                             .52        2.86(**)
 Wife's age                                  -.19       -8.39(***)

Hypersexuality Variables
 Wife's occupational prestige                 -           -
 Number of times wife moved in
  last 5 yrs.                                 -           -
R-Squared                                     .22(***)

                                                Model 2
Predictor                                     b         t Ratio
Focus Variable
 Husband has been violent                    2.43        2.52(*)
Control Variables
 Disagreement scale                          -.16       -4.25(***)
 Wife wants to have more children soon(b)   -1.06       -2.20(*)
 Wife wants to have more children later(b)  -1.89       -2.15(*)
 Wife currently pregnant(b)                 -2.88       -3.01(**)
 Husband's education                         -.11       -1.91
 Husband 5+ yrs. older than wife(c)         -1.17       -3.10(**)
 Cohabiting couple                           3.89        6.32(***)
 Partners both Hispanic(d)                  -1.55       -2.02(*)
 Other race combination(d)                   1.92        3.14(**)
 Wife's health                                .86        4.63(***)
 Husband's health                             .57        3.11(**)
 Wife's age                                  -.18       -5.28(***)

Hypersexuality Variables
 Wife's occupational prestige                -.02       -3.14(**)
 Number of times wife moved in
  last 5 yrs.                                 .20        2.12(*)
R-Squared                                     .23(***)

(a) Only predictors that were significant in at least one model are shown.

(b) Reference category is Wife has no intention of having more children.

(c) Reference category is Spouses' ages are within two years of each other.

(d) Reference category is Partners are both White.

(*) p <.05

(**) p <.01

(***) p < .001

Table 2 Unstandardized Coefficients for the Regression of Sexual Satisfaction on Focus, Control, Hypersexuality, and Sexual Extortion Factors(a)

Predictor                       b        t Ratio
Control Variables
 Disagreement scale             -.06    -10.53(***)
 Partners are both
  Black(b)                      -.40     -4.41(***)
 Partners are both
  Hispanic(b)                   -.25     -2.31(*)
 Wife's age                     -.02     -3.22(**)
 Husband's church
  attendance                     .03      3.15(**)
 Wife's church
  attendance                     .03      3.53(**)
Hypersexuality Variables
 Husband married two
  or more times before(c)       -.31     -2.80(**)
 Number of weeks
  husband unemployed
  in past year                  -.01     -3.08(**)
 Number of weeks wife
  unemployed in
  past year                      .00(d)   2.15(*)

 Focus Variable
  Sexual frequency               .02      5.12(***)

R-Squared                        .15(***)

(a) Only predictors that were significant in the model are shown.

(b) Reference category is Partners are both White.

(c) Reference category is Husband has never been married before.

(d) Exact value is .004.

(*) p <.05

(**) p <.01

(***) p <.001

Table 3
Unstandardized Coefficients for the Regression of Depressive
Symptomatology on Focus, Control, and Hypersexuality Factor(a)

                                          Model 1
Predictor                               b       t Ratio
Control Variables
 Disagreement scale                     .90        10.85(***)
 Both partners have no religious
  preference(b)                       -3.91        -2.18(*)
 Wife's health                        -4.54       -11.22(***)

Hypersexuality Variables
 Husband married once before(c)        2.52         2.63(**)

Focus Variable
 Sexual frequency                       .01          .22

Sexual Extortion Effects
 Sexual frequency x husband has been
  violent                               -            -
Sexual frequency x wife has been
  violent                               -            -
Sexual frequency x wife has been
  injured                               -            -

R-Squared                               .18(***)

                                           Model 2
Predictor                               b          t Ratio
Control Variables
 Disagreement scale                     .91       10.97(***)
 Both partners have no religious
  preference(b)                       -3.97       -2.22(*)
 Wife's health                        -4.55      -11.28(***)

Hypersexuality Variables
 Husband married once before(c)        2.50        2.61(**)

Focus Variable
 Sexual frequency                      -.05       -1.07

Sexual Extortion Effects
 Sexual frequency x husband has been
  violent                               .66        2.58(**)
Sexual frequency x wife has been
  violent                              -.44       -1.75
Sexual frequency x wife has been
  injured                               .34        1.50

R-Squared                               .19(***)

(a) See Appendix for a complete listing of the independent variables in both models.

(b) Reference category is Both partners are Protestant, or other combination.

(c) Reference category is Husband has never been married before.

(*) p <.05

(**) p <.01

(***) p <.001

Controls. Several predictors of sexual frequency, gleaned from past research, were employed as controls in the current analyses. Nonviolent relationship conflict was assessed with seven items asked of each spouse concerning the frequency of open disagreements experienced in the past year over household tasks, money, spending time together, sex, having a(nother) child, in-laws, and the children. Responses for each item ranged from "never" (coded 1) to "almost every day" (coded 6). A disagreement index was created for each spouse, provided that he or she had answered at least five items, by taking the mean response and multiplying by seven. Again, the result is equivalent to the sum of the seven items. As with reports of sexual activity, each spouse's answer was an indicator of an underlying reality in the relationship. The disagreement scale used in the analysis is therefore the mean of the spouse's reports, with an alpha reliability of .69.

The measurement of the remaining control variables was quite straightforward. Continuous variables were husband's and wife's education, duration of the current marriage or union in years, total income of the couple in thousands of dollars, number of children (under 18) in the household, husband's and wife's self-reported health status (1 = "very poor," 5 = "excellent"), wife's age, and husbands' and wives' church attendance (1 = "never," 9 = "more than once a week"). Qualitative variables were indicated in the analysis by dummies. The first was the couple's religious preference, coded as both Catholic, both stating no religious preference, and other combinations (the omitted category). The NSFH religious affiliation variable contained 64 separate codes. As Catholics were one of the most numerous categories (18% of couples were Catholic) and, additionally, represent a relatively unified body of religious doctrine, they were coded as a distinct group. Catholics and those with no religious preference were then contrasted with those falling into all other categories. Other dummy variables represented the wife's birth intentions, coded as unsure about having more children, wanting more children soon, wanting more children later, currently pregnant, and wanting no more children (the omitted category); the partners' age difference, coded as husband 2-5 years older than wife, husband more than 5 years older than wife, wife 2 or more years older than husband, and within 2 years of each other's ages (the omitted category); the couple's marital status, coded as cohabiting unmarried and married (the omitted category); and the couple's race status, coded as both White (the omitted category), both Hispanic, both Black, and other race combinations.

Hypersexuality variables. Several variables measured in the NSFH are implicated in one or another, of the explanations enumerated earlier for the co-occurrence of hypersexuality and hyperaggressivity. For example, substance abuse and early sexual involvement are manifestations of an underlying unconventionality that links them with general deviance (that includes fighting) in problem-behavior theory (Jessor et al., 1991; Jessor & Jessor, 1977). According to Gottfredson and Hirschi (1990), marital and employment instability, substance abuse, and having children out of wedlock, along with physical responses to conflict and illicit sexuality, are manifestations of low self-control. Sensation seeking and depressed occupational achievement are linked to elevated libido and aggression, based on studies of the influence of testosterone on behavior (Booth & Dabbs, 1993; Dabbs, 1992a, b; Dabbs & Morris, 1990). With this in mind, I employed depressed occupational achievement, marital instability, substance abuse, sensation seeking, early, sexual involvement, and having children out of wedlock as factors reflecting an underlying propensity toward unconventional, risky, or deviant behavior. Under the hypersexuality hypothesis, the association between hypersexuality and aggressivity occurs because both are also manifestations of this propensity.

These predictors are entered into the analyses as individual variables rather than as part of a composite scale. In that many of these measures are dichotomous, they are not appropriate for factor analysis. More importantly, it is not clear that a single factor necessarily underlies them, as the different items are drawn from competing theoretical formulations. In any event, entering them individually allows maximum flexibility in modeling. To the extent that they are strongly correlated (because of a common underlying factor), this is automatically taken into account by the regression procedure. Moreover, entering individual items in lieu of a summated scale avoids constraining the regression weights to be equal across items. (For example, the equation E(Y) = a + b([X.sub.1] + [X.sub.2]) is equivalent to E(Y) = a + b[X.sub.1] + b[X.sub.2].)

Measures of depressed occupational achievement for each partner were occupational prestige scores (tapped by the socioeconomic indicator scores of Stevens & Cho, 1985) and number of weeks unemployed in the past year. Marital instability was tapped by the husband's and the wife's numbers of previous marriages, dummied for each as married once before, married two or more times before, and having never been married before (the contrast category). Each partner was asked to indicate who in the household has an alcohol or a drug problem. The couple was coded as having a substance abuse problem (dummy variable) if either partner was so identified. Sensation seeking is somewhat more difficult to assess. Two factors in the NFSH, however, bear on the tendency of individuals to move away from home in search of "new adventures" in their lives. These are whether one has lived with both natural parents while growing up (dummy variables for each partner) and the number of moves experienced in the last five years. This last factor was only available for wives in the current sample.

This measure is admittedly only a crude proxy for sensation seeking. Others have suggested that residential migration measures depleted social capital, the latter referring to the network of social relationships established between individuals and the communities in which they live (Hagan, MacMillan, & Wheaton, 1996). However, to the extent that migration is driven by the desire for a change of surroundings, it can be said to tap sensation seeking. A measure of early sexual involvement was also only available for wives: the age of the first cohabitational union, whether marital or not. In that marriage carries with it the expectation of sexual involvement, younger ages at first marriage index earlier involvement. Additionally, cohabiting women become sexually active at earlier ages than noncohabiting women, so cohabitation at an early age also indexes early sexual involvement (Tanfer, 1987). The age of the first cohabitational union was coded as the age at which the wife began her first marital or cohabiting union with any partner--not necessarily the current one. Finally, having children out of wedlock is unconventional behavior in the sense, that it is contrary to the widely shared family ideal of marrying first, then having children. It is particularly so when the mother is also a teenager. Therefore, the last factor in this group is a dummy variable representing whether the wife had experienced a premarital birth while still a teenager. The Appendix presents descriptive statistics for all variables involved in the current study.

Statistical Analysis

The dependent variables--sexual frequency, depression, and sexual satisfaction--are approximately interval. Hence, multivariate analyses were conducted with ordinary least squares (OLS) regression. All inferential analyses (including the bivariate results discussed later) were weighted to correct for oversampling. Weights have additionally been modified to sum to the unweighted number of cases. In that the use of husbands' and wives' characteristics in OLS has the potential for causing multicollinearity problems, variance inflation factors (VIFs) were examined for each estimated equation. VIF values below 10 are usually indicative of nonproblematic associations among predictors (Myers, 1986). All main effect coefficients had VIFs well under 10, except for wife's age and marital duration. The VIFs for these were around 12 and 14, respectively, indicating--as one might expect--that these factors are strongly correlated. Nevertheless, these variables are controls rather than focus variables, so estimates of their effects are not of primary interest in this study. Overall, there was no evidence that collinearity posed any major problems in the analyses.

Missing value imputation. Mean (continuous variables) or modal (qualitative variables) values were substituted for missing values on predictors prior to performing analyses. To examine whether missing value imputation biased the analyses, 1 created a dummy variable indicating whether any missing values had been imputed for any predictors. Adding this dummy to the analyses produced virtually no detectable changes in the coefficient values. Nevertheless, it was possible that the impacts of the predictors might be different for those with values imputed, compared to those with complete data on all variables. I checked this by adding to all multivariate models a set of terms representing the interactions between whether missing values were imputed and all other predictors. However, in all cases, the block of interaction terms was nonsignificant. It appears that missing imputation did not substantially affect my results. The multivariate models are therefore presented without controls for missing imputation.

Results

The average couple in this study reported 7.69 acts of sex per month, whereas around 8% to 9% of couples reported husband and wife violence, respectively (see Appendix). Are these factors related? I began by examining the bivariate associations of both husband's and wife's use of aggression with the couple's sexual frequency, using t tests. As has been confirmed in other studies, both associations were significant. Couples among whom husbands had been violent in the past year reported, on average, 11.08 sex acts in the previous month, compared to 6.75 acts reported by couples with nonviolent husbands (t = 7.20, p [is less than] .0001). Similarly, the average monthly frequency of sex was 10.02 when wives had been violent but 6.79 otherwise (t = 5.67, p [is less than] .0001). Under the hypersexuality hypothesis, both partners may well be predisposed to both greater aggression and greater sexual activity. However, under the sexual extortion hypothesis, only husband's aggression should enhance sexual frequency. The association of wife's aggression with more frequent sex would be due to the strong correlation between husband and wife aggression. Indeed in the current sample, that correlation (phi) is .82. To assess which hypothesis has more credence, we must turn to the multivariate analyses.

The coefficients and their t ratios from a regression of sexual frequency on both partners' aggression plus control variables are shown in the columns headed Model 1 in Table 1. To conserve space, only significant effects are shown. However, all control variables for the prediction of sexual frequency were included in the analysis. (See the Appendix for a complete listing of independent variables in the various equations; complete regression tables are available from the author on request.) As is evident, controlling for each partner's use of violence, only the husband's is significantly related to sexual frequency. The coefficient suggests that where husbands have been violent with their wives, couples have sex about two and a half times more often per month, controlling for other factors.

Several other predictors are also significant. The more frequently the couples have open disagreements, the fewer times per month they have sex, as one would expect. Couples in which wives are pregnant reduce their frequency of sexual activity by about two and a half times per month. Husbands with more education and those who are more than five years older than their wives have sex with their wives less often than others. Controlling for other factors, cohabiting couples have sex about four more times per month than married couples. Hispanic couples have sex less often than White couples, but couples from ethnic groups other than White, Hispanic, or Black, and race-heterogamous couples have sex more often. Sexual frequency goes up with the health of either partner and down with wife's age. About 22% of the variation in sexual frequency is accounted for by this model.

Model 2 adds the block of hypersexuality factors to the equation. Although the effect size for this block is small--the squared semipartial correlation is .228-.219 = .009--it is significant, F(14, 2394) = 2.02, p [is less than] .02, indicating that at least one of these factors has a significant effect on sexual frequency. If the hypersexuality hypothesis were supported, the effect of husband's violence would be reduced and no longer significant. However, this is not the case. The coefficient for husband aggression hardly changes (2.56 vs. 2.43) and remains significant. Of the added factors, two are significant: Wife's occupational prestige is negatively related to sexual activity, whereas the number of times she has moved in the last five years enhances sexual activity. The direction of these effects is consistent with prediction.

To evaluate the sexual extortion hypothesis, I next regressed wife's sexual satisfaction and wife's depression on sexual frequency plus the focus, control, and hypersexuality factors. Wife's injury was also included in the model. I then added to the model the interactions between husband's and wife's violence, and wife's injury, on the one hand, and sexual frequency, on the other, in their effects on wives' sexual satisfaction and wives' depression. This block of interaction terms was not significant for wives' sexual satisfaction; therefore, only the main effects model is shown in Table 2. Interestingly, although open disagreement significantly depresses wives' sexual satisfaction, neither husband's nor wife's violence had any effect once the level of conflict was held constant. Sexual satisfaction appears to be higher among younger couples, higher among White couples than Blacks or Hispanics, and higher among those where either partner attends church frequently. Wives are less satisfied when their husbands have been married two or more times before or when their husbands have been unemployed for a number of weeks in the past year. However, the more weeks the wife has been unemployed, the more sexually satisfied she is. As one would expect, more frequent sexual activity predicts greater sexual satisfaction.

Table 3 presents the main effects (Model 1) and interaction (Model 2) models for depressive symptomatology. Once again, only the significant effects are shown, although the models control for all independent variables shown in the Appendix. About 18% of the variation in depression is explained by the main effects model. As predicted, the main effects model shows no effect of sexual frequency on depression. Interestingly, neither partner's violence has a significant effect on wife's depression, but open disagreements do. When couples have more disagreements, wives are more depressed. Wives in couples among whom both partners profess no religious affiliation are less depressed than others. Wives in better health are significantly less depressed than others. Wives married to husbands who have been married once before are significantly more depressed than those whose husbands are in their first marriages.

Model 2 adds the three crossproduct terms to Model 1. The increment in explained variance is significant, F(3,2389) = 5.21,p [is less than] .01, suggesting that the effect of sexual frequency on depression is a function of violence and injury. To facilitate the explication of the interaction effects, I have used the estimates in Model 2 to recover the simple slopes (Aiken & West, 1991) for the regression of depression on sexual frequency at different levels of partner violence and wife injury, controlling for all other effects. These values are shown in Table 4.

Table 4 Simple Slopes for the Regression of Depressive Symptomatology on Sexual Frequency, by Levels of Spousal Violence and/or Wife's Injury(a)

                                     Simple
Violence Condition                   Slope   t Ratio    p
No violence, wife not injured        -.05    -1.07      >.20
Husband violent, wife not injured     .61     2.38      <.05
Husband violent, wife injured         .95     3.15      <.01
Both violent, wife not injured        .17     1.57      >.10
Both violent, wife injured            .51     2.49      <.05
Wife violent, wife not injured       -.49    -1.95       .05(b)
Wife violent, wife injured           -.16     -.45      >.200
Neither violent, wife injured         .28     1.23      >.200

(a) Based on Model 2 in Table 3.

(b) Exact value is .051.

T-tests for simple slopes are formed by dividing the simple slopes by their standard errors. If the partial slope for the regression of depression on sexual frequency is expressed as a + b * husband violence + c * wife violence + d * wife injury, then the simple slope for the regression of depression on sexual frequency for couples reporting husband violence and wife injury is a + b + d, and its standard error is the square root of V(a) + V(b) + V(d) + 2 * Cov(a,b) + 2 * Cov(ad) + 2 * Cov(b,d). Substituting elements from the covariance matrix of parameter estimates into this term provides the estimated standard error (Aiken & West, 1991).

These tests reveal that sexual frequency has a significant and positive impact on depression when the husband is the only violent partner--whether or not the wife has been injured--and when the wife has been injured in the context of both partners being violent. Although the effect of sexual frequency on depression is also positive when both partners have been violent but the wife has not been injured, it is not significant. On the other hand, the effect for sexual frequency is marginally significant but negative when the wife was the only violent partner and she was not injured as a result.

At the suggestion of one reviewer I investigated whether any high-influence observations were driving the results in Tables 2 and 3. In particular, I examined the following influence diagnostics: DFFITS scores, hat diagonals, externally studentized residuals, and Cook's D values. From criteria outlined in Myers (1986), 1 identified all cases with values on these measures indicative of strong influence on either coefficient estimates or fitted values of the regression. There were 195 high influence observations for the analysis of wives' sexual satisfaction and 266 high influence observations for the analysis of wives' depression. However, none of these cases revealed patterns of responses that were suggestive of coding errors or attempts by respondents to sabotage the survey by giving nonsensical answers. One couple had an unusually high value for sexual frequency in the past month of 93 episodes. If "having sex" referred exclusively to intercourse, this would be a highly questionable result. Given its more general connotation of activities that involve a sexual response, it is not clear that such a value is in error. At any rate, I re-ran all analyses, including those in Table 1, after deleting this observation. No substantive differences emerged compared to the analyses reported in Tables 1-3.

Finally, I tested whether the simple slope for sexual frequency was significantly larger when the husband was the only violent partner and the wife was injured, compared to cases in which the husband was the only violent partner and the wife was not injured, or both partners were violent and the wife was injured. Contrasts among simple slopes are based on tests of the differences between them (Aiken & West, 1991). For example, the coefficient for the crossproduct of wife injury with sexual frequency, d,,represents the difference in simple slopes among couples reporting husband violence and wife injury, compared to couples reporting husband violence and no wife injury. From Table 3, it is evident that this difference is insignificant. Although the simple slope was highest under the first condition, it was not significantly higher than the other two.

Discussion

Two explanations for the association of partner violence with more frequent sexual activity have been examined in this study. The hyper-sexuality hypothesis suggests that the association is spurious, and due to a pre-existing predilection for unconventional or deviant behavior that characterizes both partners. The sexual extortion hypothesis suggests that a husband's violence creates a climate of fear in marriage that leads wives to acquiesce to sex more frequently than they would otherwise consider, The findings appear to be more consistent with the latter explanation, for several reasons.

To begin, although bivariate associations between either partner's aggression and sexual frequency were positive, only husband's violence was predictive of greater sexual activity in the multivariate context. This suggests, as mentioned, that wife's violence is only associated with greater sexual frequency by virtue of its strong link to husband violence. Given a mutual predilection for both hyperaggressivity and hypersexuality among husbands and wives, we would expect each partner's violence to predict sexual activity independent of the other.

Second, the fact that disagreements and husband violence have opposite effects on sexual activity is intriguing. Although conflict and violence are positively correlated (DeMaris, 1993), conflict per se diminished sexual activity. Nevertheless, at a given level of conflict, the use of violence by husbands served to elevate sexual frequency. This appears to make most sense only when one assumes that a husband's violence has a coercive effect. Otherwise, if conflict generally "turns partners off" to sex, it would not be reasonable to assume that violence--often the result of conflict--turns them back on.

Perhaps most important to the sexual extortion hypothesis was the fact that, as predicted, greater sexual frequency generally increased wives' depression whenever husbands had been violent to them. These findings suggest that sexual activity is "depressing" when it occurs in the context of physical confrontations involving violence by husbands. Again, this makes sense if one assumes that wives under these conditions are giving in to sex partly out of fear and that their corresponding sense of powerlessness elevates levels of depression.

Despite the persuasiveness of these observations, several caveats must be kept in mind regarding the tenability of these conclusions. Above all, although several measures were employed in the study to tap a propensity for deviant, unconventional, or sensation-seeking behavior, several other measures were missing. In particular, testosterone levels were not assayed for NSFH respondents, nor did the survey include measures of the personality trait of low self-control. It is certainly possible that these, as well as other factors, could account for the lingering association of male violence with sexual frequency found after hypersexuality variables had been partialed out. At the very least, however, one can say that this association persists even after controlling for a host of background factors that could reasonably be presumed to reduce it to zero. Moreover, of the initial difference in average sexual acts per month of 4.33 between marriages with and without husband violence, the greater part (56%) of this difference remained after all other explanatory variables were in the model.

Another limitation of the study is that sexual coercion has only been inferred but not measured directly. Essentially, the analyses have relied on sociological "detective work" to build a case based on circumstantial evidence alone. I have argued, based on theoretical reasoning, that sexual coercion, or extortion, should be revealed by an interaction between sexual frequency and violence (including injury) in their effects on wives' depressive symptomatology. To the extent that this was found, that reasoning is supported. However, without wives indeed acknowledging that they were coerced into having sex, that inference remains somewhat speculative.

It is also not possible, on the basis of these results, to rule out the cycle-of-violence explanation for the association of violence with elevated sexual activity. However, I find this third hypothesis to be less tenable than the others. Women's sexual response to a partner is said to be especially sensitive to the quality of the interpersonal relationship between partners (Rubin, 1983), in comparison to that of men. Unless one is willing to subscribe to the idea that women are generally "turned on" by being assaulted--an idea that does not accord with the literature on marital rape (Finkelhor & Yllo, 1985; Russell, 1982)--sexual activity should be diminished immediately after an episode of violence. Therefore, couples would have to be especially active sexually during the "repentance phase" to produce rates of sexual activity that are greater than average. Yet, those who subscribe to the idea that violence unfolds in the aforesaid cyclical pattern describe the repentance phase as one in which the batterer--most often the husband--relinquishes power to the wife to win back her allegiance (Deschner, 1984). In that control over sexual activity would pass at this point into her hands, it is unlikely that sexual frequency would be markedly higher than normal during this time period. To assume otherwise suggests, again, that wives' sexual response is dramatically enhanced by the experience of violence at the hands of their husbands. In any case, this argument could not be fully assessed because of the limitations of the current data. With data on the timing of episodes of violence in relation to sexual activity, future researchers may be better able to evaluate the plausibility of this thesis.

In spite of popular imagery to the contrary, the findings of this study suggest that aggression and violence may not be so sexy, after all. Moreover, one of the many destructive consequences of relationship violence may be wives' loss of control over their sexuality. To the extent that this situation leads to depressive symptomatology, it merits our attention as a mental health issue. The findings additionally suggest that sexual and emotional intimacy may well be among the first casualties of spousal aggression. Sexual relationships shoUld be pleasurable and Should help to cement the bonds between partners. When sex takes place in an atmosphere of intimidation or coercion, those bonds are much more difficult to forge. Human sexuality is still a complex affair that is not well understood. More work needs to be done on the relationships among conflict, violence, and sexuality with measures Specifically designed to address the subtleties of these phenomena. Until this is accomplished, much of sexual behavior will remain shrouded in mystery.

References

Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and interpreting interactions. Thousand Oaks, CA: Sage.

Apt, C., & Hurlbert, D. F. (1993). The sexuality of women in physically abusive marriages: A comparative study. Journal of Family Violence, 8, 57-69.

Arneklev, B. J., Grasmick, H. G., Tittle, C. R., & Bursik, R. J., Jr. (1993). Low self-control and imprudent behavior. Journal of Quantitative Criminology, 9, 225-247.

Bandura, A. (1992). Exercise of personal agency through the self-efficacy mechanism. In R. Schwarzer (Ed.), Self-efficacy: Thought control of action (pp. 3-38). Washington, DC: Hemisphere.

Blumstein, P., & Schwartz, P. (1983). American couples: Money, work, sex. New York: William Morrow.

Booth, A., & Dabbs, J. M., Jr. (1993). Testosterone and men's marriages. Social Forces, 72, 463-477.

Call, V, Sprecher, S., & Schwartz, P. (1995). The incidence and frequency of marital sex in a national sample. Journal of Marriage and the Family, 57, 639-652.

Cowan, G., Lee, C., Levy, D., & Snyder, D. (1988). Dominance and inequality in x-rated videocassettes. Psychology of Women Quarterly, 12, 299-311.

Dabbs, J. M., Jr. (1992a). Testosterone measurements in social and clinical psychology. Journal of Social and Clinical Psychology, 11, 302-321.

Dabbs, J. M., Jr. (1992b). Testosterone and occupational achievement. Social Forces, 70, 813-824.

Dabbs, J. M., Jr., & Morris, R. (1990). Testosterone, social class, and antisocial behavior in a sample of,#,462 men. Psychological Science, 1, 209-211.

DeMaris, A. (1993). When disagreements lead to violence in marriage. Proceedings of the American Statistical Association, 756-760.

DeMaris, A., & Swinford, S. (1996). Female victims of spousal violence: Factors influencing their level of fearfulness. Family Relations, 45, 98-106.

Deschner, J. P. (1984). The hitting habit: Anger control for battering couples. New York: Free Press.

Donnelly, D. A. (1993). Sexually inactive marriages. The Journal of Sex Research, 30, 171-179.

Donovan, J. E., & Jessor, R. (1985). Structure of problem behavior in adolescence and young adulthood. Journal of Consulting and Clinical Psychology, 53, 890-904.

Finkelhor, D., & Yllo, K. (1985). License to rape: Sexual abuse of wives. New York: Free Press.

Glass, J., & Fujimoto, T. (1994). Housework, paid work, and depression among husbands and wives. Journal of Health and Social Behavior, 35, 179-191.

Gottfredson, M. R., & Hirschi, T. (1990). A general theory of crime. Stanford, CA: Stanford University Press.

Grasmick, H. G., Tittle, C. R., Bursik, R. J., Jr., & Arneklev, B. J. (1993). Testing the core empirical implications of Gottfredson and Hirschi's general theory of crime. Journal of Research in Crime and Delinquency, 30, 5-29.

Greenblat, C. S. (1983). The salience of sexuality in the early years of marriage. Journal of Marriage and the Family, 45, 289-299.

Hagan, J., MacMillan, R., & Wheaton, B. (1996). New kid in town: Social capital and the life course effects of family migration on children. American Sociological Review, 61, 368-385.

Huey, C. J., Mine-Graber, G., & Graber, B. (1981). Time factors and orgasmic response. Archives of Sexual Behavior, 10, 111-118.

James, W. H. (1974). Marital coital rates, spouses' ages, family size and social class. The Journal of Sex Research, 10, 205-218.

James, W. H. (1981). The honeymoon effect on marital coitus. The Journal of Sex Research, 17, 114-123.

Jasso, G. (1985). Marital coital frequency and the passage of time: Estimating the separate effects of spouses' ages and marital duration, birth and marriage cohorts, and period influences. American Sociological Review, 50, 224-241.

Jessor, R., Donovan, J. E., & Costa, F. M. (1991). Beyond adolescence: Problem behavior and young adult development. Cambridge, England: Cambridge University Press.

Jessor, R., & Jessor, S. L. (1977). Problem behavior and psychosocial development: A longitudinal study of youth. New York: Academic Press.

Mirowsky, J., & Ross, C. E. (1990). Control or defense? Depression and the sense of control over good and bad outcomes. Journal of Health and Social Behavior, 31, 71-86.

Muehlenhard, C. L. (1989). Young men pressured into having sex with women. Medical Aspects of Human Sexuality, 23, 50-62.

Myers, R. H. (1986). Classical and modern regression with applications. Boston: Duxbury Press.

Newcomb, M. D. (1986). Sexual behavior of cohabitors: A comparison of three independent samples. The Journal of Sex Research, 22, 492-513.

Ozer, E. M., & Bandura, A. (1990). Mechanisms governing empowerment effects: A self-efficacy analysis. Journal of Personality and Social Psychology, 58, 472-486.

Palys, T. S. (1986). Testing the common wisdom: The social content of video pornography. Canadian Psychology, 27, 22-35.

Rao, K. V., & DeMaris, A. (1995). Coital frequency among married and cohabiting couples in the United States. Journal of Biosocial Science, 27, 135-150.

Ross, C. E., Mirowsky, J., & Huber, J. (1983). Dividing work, sharing work, and in-between: Marriage patterns and depression. American Sociological Review, 48, 809-823.

Rubin, L. B. (1983). Intimate strangers: Men and women together. New York: Harper & Row.

Russell, D. E. H. (1982). Rape in marriage. New York: Macmillan.

Simons, R. L., Johnson, C., Beaman, J., & Conger, R. D. (1993). Explaining women's double jeopardy: Factors that mediate the association between harsh treatment as a child and violence by a husband. Journal of Marriage and the Family, 55, 713-723.

Stevens, G., & Cho, J. H. (1985). Socioeconomic indexes and the new 1980 census occupational classification scheme. Social Science Research, 14, 142-168.

Straus, M. A. (1993). Physical assaults by wives: A major social problem. In R. J. Gelles & D. R. Loseke (Eds.), Current controversies on family violence (pp. 67-87). Thousand Oaks, CA: Sage.

Sweet, J., Bumpass, L., & Call, V (1988). The design and content of the national survey of families and households. Madison: University of Wisconsin Center for Demography and Ecology.

Tanfer, K. (1987). Patterns of premarital cohabitation among never-married women in the United States. Journal of Marriage and the Family, 49, 483-497.

Udry, J. R. (1980). Changes in the frequency of marital intercourse from panel data. Archives of Sexual Behavior, 9, 319-325.

Udry, J. R., Deven, F. R., & Coleman, S. J. (1982). A cross-national comparison of the relative influence of male and female age on the frequency of marital intercourse. Journal of Biosocial Science, 14, 1-6.

Udry, J. R., & Morris, N. M. (1978). Relative contribution of male and female age to the frequency of marital intercourse. Social Biology, 25, 128-134.

Wood, P. B., Pfefferbaum, B., & Arneklev, B. J. (1993). Risk-taking and self-control: Social psychological correlates of delinquency. Journal of Crime and Justice, 16, 111-130.

Yang, N., & Linz, D. (1990). Movie ratings and the content of adult videos: The sex-violence ratio. Journal of Communication, 40, 28-42.

Appendix
Descriptive Statistics for all Variables in the Analyses

                                                        Mean    SD
Dependent Variables
 Couple's sexual frequency in past month                7.69    7.56
 Wife's depressive symptomatology in past week         14.72   15.88
 Wife's sexual satisfaction                             4.17     .98

Independent Variables
 Focus Variables:
  Husband has been violent                               .08    -.27
  Wife has been violent                                  .09     .28
  Wife has been injured                                  .03     .17
 Control Variables:
  Disagreement scale                                   12.49    4.27
  Both partners have no religious preference(a)          .03     .18
  Both partners are Catholic(a)                          .18     .38
  Wife is unsure whether to have more children(b)        .06     .25
  Wife wants to have more children soon(b)               .18     .39
  Wife wants to have more children later(b)              .04     .20
  Wife is currently pregnant(b)                          .03     .17
  Husband's education                                  13.02    3.23
  Wife's education                                     12.89    2.73
  Duration of current marriage/union in years          14.27   13.81
  Total income of the couple, in thousands             43.95   44.05
  Husband is 2-5 years older than wife(c)                .29     .45
  Husband is more than 5 years older than wife(c)        .26     .44
  Wife is two or more years older than husband(c)        .08     .28
  Couple is cohabiting unmarried                         .09     .28
  Partners are both Black(d)                             .08     .27
  Partners are both Hispanic(d)                          .05     .21
  Partners are other race combination(d)                 .07     .25
  Number of children in the household                   1.20    1.29
  Husband's self-reported health                        4.08     .79
  Wife's self-reported health                           4.05     .79
  Wife's age                                           38.70   14.00
  Husband's church attendance                           4.33    2.70
  Wife's church attendance                              5.26    2.86

 Hypersexuality Factors:
  Husband's occupational prestige                       31.18  22.77
  Wife's occupational prestige                          22.46  23.45
  Husband married once before(e)                         .21     .41
  Husband married two or more times before(e)            .05     .21
  Wife married once before(f)                            .21     .41
  Wife married two or more times before(f)               .04     .19
  Number of weeks husband unemployed in past year       1.97    7.13
  Number of weeks wife unemployed in past year          4.41    9.52
  Number of times wife has moved in last 5 years        1.40    1.90
  Either spouse has alcohol or drug problem              .07     .26
  Husband lived with both parents while growing up       .80     .40
  Wife lived with both parents while growing up          .71     .46
  Wife's age at first marital/cohabitational union     21.24    4.29
  Wife experienced a premarital birth before age 20      .06     .24

(a) Reference category is Both partners are Protestant, or Other Combination.

(b) Reference category is Wife has no intention of having more children.

(c) Reference category is Spouses' ages are within two years of each other.

(d) Reference category is Partners are both White.

(e) Reference category is Husband has never been married before.

(f) Reference category is Wife has never been married before.

Data are from the National Survey of Families and Households 1987-88, which was funded by a grant (HD21009) from the Center for Population Research of the National Institute of Child Health and Human Development.

Address all correspondence regarding this manuscript to Alfred DeMaris, Ph.D., Department of Sociology, Bowling Green State University, Bowling Green, OH 43403. Telephone: [phone]. E-mail: [e-mail]. Fax: [phone].

Manuscript accepted March 18, 1997