is expected of parents when a child persists in misbehaving. Both the law and the .... Obtaining data on husband-to-wife assaults from the husbands results.
CORPORAL PUNISHMENT OF ADOLESCENTS EPT~EIMTOT.OC,Y BY PARENTS: A RISK FACTOR .-.IN THE - --- - -- -- -- - - - - OF DEPRESSION, SUICIDE, ALCOHOL ABUSE, CHILD ABUSE, AND WIFE BEATING -
Murray A. Straus and Glenda Kaufman Kantor ABSTRACT
Over 90% of parents of toddlers spank or use other forms of corporal punishment. Although the rate declines each year from about age five, this study of a large national sample of U.S. adults found that almost half recalled having been corporally punished during their teen years. This high prevalence indicates a need to investigate the possibility that corporal punishment puts adolescents at increased risk of developing mental health and social relationship problems later in life. The analysis, which controlled for a number of possible confounding risk factors such as low socioeconomic status, found that children who experienced corporal punishment in adolescence had an increased risk later in life of depressive symptoms, suicidal thoughts, alcohol abuse, physical abuse of children, and wife heating. The consistent association of corporal punishment with major adult problem behavior, together with the fact that at least: half of all adolescents are victims of corporal punishment by their parents, indicates a need to replicate the study using longitudinal data. If the findings are confirmed,it suggests that a major step in primary prevention of violence and mental health problems can he achieved by a national effort to reduce or eliminate all use of corporal punishment. The terms physical punishment and corporal punishment are used here as synonyms to refer t o an act by parents intended to cause the child physical pain, but not injury, for purposes of correction or control of misbehavior. The 1975 and 1985 National Family Violence Surveys revealed that over 90% of American parents use corporal punishment as just defined. This is consistent with a large number of other studies (Straus, 1991) and with t h e belief that corporal punishment is used by
The data are from the National Family Violence Resurvey, funded by National Institute of Mental Health grant ROlMH40027 (Richard J. Gelles and Murray A. Straus, co-investigators). Dr. Kaufman Kantor's work was supported by grant R01 AA08269-01 from the National Institute of Alcohol Abuse and Alcoholism (Glenda Kaufman Kantor, Principal Investigator). The authors acknowledge the contribution of Carrie Yodanis, who ran the statistical analysis for this paper. Request reprints to Murray A. Straus, Ph.D., Professor of Sociology and CoDirector, Family Research Laboratory, University of New Hampshire, 126 Horton Social Science Center, Durham, NH 03824. ADOLESCENCE, Vol. 29, No. 115, Fall 1994 Libra Publishers, Inc., 3089C Clairemont Dr., Suite 383, San Diego, CA 92117
parents with toddlers or young children. The first issue to be addressed here is the extent to which this pattern continues into adolescence. The second and central issue addressed is whether use of corporal punishment of adolescents is associated with psychosocial problems in adulthood. In contrast to physical "abuse" of a child, which is universally regarded with revulsion, 84% of all Americans regard hitting children as "sometimes necessary" (Straus, 1991) and harmless provided it does not cross the line to physical abuse. The same acceptance seems to be true of social scientists. Numerous studies have investigated the long-term disabling effects of physical "abuse" (Widom, 1989), but despite the prevalence of corporal punishment, only a few studies have considered the possibility that "legitimate" corporal punishment is also a risk factor for psychological maladaptation in adulthood, and most of those have been confined to increased risk of aggression (Lefkowitz, Eron, Walder, & Huesman, 1977). Adult psychological distress, alcohol abuse, and violence toward family members are all problems which share a complex etiology though not necessarily a common etiology. We are not suggesting that these maladaptive behaviors are caused solely by corporal punishment in adolescence, but that corporal punishment needs to be considered as a significant risk factor which increases the probability of psychological disorders. The social and psychological dynamics underlying this association are presumed to resemhle those of other forms of violent victimization in children and adults. Previous research has shown that victimization is a traumatic process which may result in interpersonal deficits such as feelings of powerlessness (Straus & Gimpel, 1994), diminished self-esteem, and self-defeating interpersonal strategies (McGrath, Keita, Strickland, & Russo, 1990; Herman, 1992). These characteristics are also consistent with psychological profiles of depression (American Psychiatric Association, 1987), alcoholism (Cox, 1987; Goluke, Landeen, & Meadows, 19831, and intra-family abuse (Hamberger & Hastings, 1986). One of the reasons so few parents and social scientists question the wisdom of "spare the rod and spoil the child and why so few researchers have investigated the potential adverse effects, is probably that, in legal terms it is permissible; and in terms of informal cultural norms, it is expected of parents when a child persists in misbehaving. Both the law and the informal culture assume that, when done "in moderation," corporal punishment is harmless and sometimes necessary. This research started from assumptions that are almost the opposite of that aspect of American culture. I t tests the hypothesis that the greater the use of corporal punishment, the greater the probability, later in life, of depression, suicidal ideation, alcohol abuse, wife assaults, and child abuse.
The findings reported here must be regarded as tentative because they are based on recall data about corporal punishment, whereas an adequate test of this hypothesis requires prospective data from a large and representative sample. Such research will take many years and large funding, but could have vast implications for primary prevention of family violence and psychological problems. Preliminary evidence is therefore needed as one basis for deciding whether to invest in a prospective study. This evidence is necessary because, as noted, most of the public and most social scientists would otherwise reject the value of such a n investment. METHODS
Sample The findings are based on subsets of the 6,002 families who were studied as part of the 1985 National Family Violence Survey. This is a national probability sample selected by random digit dialing sampling methods. Interviews lasted an average of 35 minutes. The response rate, calculated as "completed portion of eligibles" was 84%. One member of each household, either the husband or the wife, was interviewed. A random procedure was used to select the gender of the respondent. When more than one child under 18 was in the household, a random procedure was used to select the "referent child as the focus of the parent-to-child violence questions. Of the 6,002 families, approximately 4,500 provided information on all the variables needed for analyses in which the dependent variable is depression, suicidal ideation, alcohol abuse, or wife assault. For the analysis using physical abuse of a child as the dependent variable, it was possible to use only the 2,149 cases with a minor child living a t home and complete data on all the variables. Detailed information on the sample and survey methods is given elsewhere (Straus & Gelles, 1990). Measures and Prevalence Rates1 Corporalpunishment in the family of origin. Respondents were asked "Thinking about when you yourself were a teenager, about how often 'The prevalence rate that follows the description of the procedure used to measure each psychosocial problem is for the subset ofcases that had no missing data on any of the variables needed to test the model explaining that variable, as given in Table 1. Since each model included whether there was husband-to-wifeviolence (in order to control for confounding of that aspect of violence with corporal punishment), the subsets are for two-partner families. The resulting subsets have a slightly lower prevalence of psychosocial pmblems than were previously reported for some of these measures. For example, the rate of physical abuse of children is 11.2 per hundred for this subset of cases, compared to 11.8 for the entire sample of families with children, which included single-parent families which have higher rates of physical abuse.
would you say your mother or stepmother used physical punishment, like slapping or hitting you? Think about the year in which this happened the most. Did this happen never, once, twice, 3 to 5 times, 6 to 10 times, 11 to 20 times, more than 20 times?' The question was repeated for corporal punishment by the father. Almost half the sample (49.8%) reported having been physically punished one or more times. This seems an astonishingly high rate for adolescents, but it is consistent with previous research (Straus, 1971; Straus, Gelles, & Steinmetz, 1980). This is far from an ideal measure of corporal punishment because the validity of recall data on events that took place many years earlier is questionable. Nevertheless, the data on prevalence and chronicity in the next section show a close correspondence with the rate based on contemporaneous data obtained by interviewing parents of teenage children (Wauchope & Straus, 1990). Prevalence and chronicity of corporal punishment of adolescents Despite the length of time which typically intervened between the time of interview and when these adult respondents were adolescents, about half the sample recalled having been corporally punished one or more times during their teen years. Among boys, they are the majority (58%) and among girls almost a majority (44%). Moreover, since these data are based on recall over long time periods, they are probably minimum estimates. A 50% prevalence rate may seem extremely high for adolescents, but this rate corresponds closely to the rates based on contemporaneous data obtained by interviewing parents of 13- io 15-year-old children in this sample, and is also consistent with previous research (Straus, 1971; Straus et a)., 1980).This correspondence shows that respondents who reported having been corporally punished during their teen years are not a small and highly deviant subset of teenagers. They are a t least half of all American children and therefore represent the modal situation. Of the parents hitting a teen, almost all (79.5%) did it more than once during the 12 months preceding the interview. The mean was 7.9 times. The distribution is skewed, but even the median of 4 is high, especially because the data on chronicity are even more likely to be underestimates than the prevalence data since they require the respondent to not only remember if there was a n instance of hitting, but also how many times it occurred. While the correspondence between the retrospective data used in this paper and contemporaneous reports of parents of teenagers are not necessarily evidence of validity, it does somewhat alleviate concern about two threats to validity. First, it reduces the possibility that, if the hypotheses are supported, it is because depressed adults remember
more bad things about their childhood, such as having been hit as a teen, even though more such events did not happen. In this case, we know that the percentage who recall the "bad thing" is almost identical to the percentage of parents who reported doing it to an adolescent during the year of the survey. Second, as previously noted, the 50% rate shows that the respondents who reported having been corporally punished during their teen years are not a small and highly deviant subset of all teenagers. Of course, the percentage who hit a child or a teen frequently is much less. This raises the question of whether there are enough cases in the high frequency of corporal punishment categories to justify the regression analysis and whether the analysis will be unduly influenced by a few "outlier" cases. The latter concern was addressed by combining all cases of 30 or more instances into the highest category. The resulting Ns are: Once = 466, Twice = 425, 3 to 5 times = 570,6 to 10 times = 3 1 6 , l l to 20 times = 252,21 to 29 times = 153, and 30 or more times = 115. Thus each category is represented by over 100 cases. Physical abuse measures. The Conflict Tactics Scales or CTS (Straus, 1979, 1990) was used to measure physical abuse of children and spouses. The CTS asks respondents to think of situations in the past year when they had a disagreement or were angry with a specified family member and to indicate how often they engaged in each of the acts included in the CTS. The list starts with the items for the Reasoning Scale of the CTS, such as "Discussed the issue," and proceeds to the items in the Verbal Aggression Scale, such as "Insulted or swore a t himlher." These are followed by the Violence index acts such as slapping, punching, and kicking. The CTS is a well-validated and widely used measure of reasoning, verbal aggression, and violence in families. There is a n accumulation of evidence that the CTS has a stable factor structure, moderate reliability and concurrent validity, and strong evidence of construct validity (Grotevant & Carlson, 1989). Child Abuse. The CTS Severe Violence index measures whether one or more of the following acts by a parent, each of which is almost universally regarded as abusive, occurred during the 12 months preceding the interview: kicking, biting, punching, beating up, scalding, and attacks with weapons. We also added hitting the child with a n object such as a stick or a belt. Although this inclusion is not consistent with informal American social norms or legal norms, hitting with objects does carry considerable risk of physical harm to the child. For the 12 months prior to the interview, 11.2% of the parents reported using one or more of these acts on the "referent child (the child randomly selected as the focus of the interview when there was more than one minor child a t home) and were therefore classified as having physically abused.
Wife assault. The CTS assault index measures whether one or more of the following acts occurred during the preceding 12 months in the course of an argument: threw something a t her; pushed, grabbed or slapped her; kicked, bit, hit with fist; hit with object; beat-up; choked; threatened with knife or gun; used knife or gun. The 12-month prevalence rate for wife assault in this sample was 11.2 per hundred. As noted, one member of each household, either the husband or the wife, was interviewed. This means that in half the families, male respondents report on their assaultive behaviors toward wives, and in the other half, female respondents report on their husband's assaults. Obtaining data on husband-to-wife assaults from the husbands results in an underestimate of the actual rate of wife assault since husbands tend to underreport their own offenses (Jouriles & O'Leary, 1985; Stets & Straus, 1990; Szinovacz, 1983). However, tests of hypotheses concerning the antecedents or consequences of husband-to-wife assaults results in almost identical findings (Straus, 1990). Depressive symptoms. Identifying who is depressed in a large crosssectional sample of Americans is a difficult and controversial task. The method used in the 1985 National Family Violence Survey is based on the Psychiatric Epidemiological Research Instrument or PERI (Dohrenwend, Kranoff, Askenasy, & Dohrenwend, 1978; Newman, 1984). The PERI provides data on a number of different psychiatric problems and is too long to be included in the half-hour interviews conducted. The measure used consists of the following four PERI items that Newman (1984) found to be most indicative of depression: Been bothered by feelings of sadness or depression?; felt very bad and worthless?; have you had times when you couldn't help wondering if anything was worthwhile anymore? Felt completely hopeless about everything? Respondents were asked to indicate how often in the past year each of these occurred using the following categories: 0 (Never), 1 (Almost Never), 2 (Sometimes), 3 (Fairly Often), 4 (Very'Often). These items were factor analyzed using the SPSS Principle Components program. The analysis found a single factor which accounted for 66% of the variance. The Depressive Symptoms Index used for this study is the factor weighted sum of these four items and has an Alpha coefficient of reliability of 22. Depressive symptoms are widespread and, as noted by Jacobson and Revanstorf (1988), the important issue both theoretically and clinically, is the occurrence of a high or chronic level of depressive symptoms. Since some estimates of severe depression have found a prevalence rate of about lo%, for purposes of the logistic analysis, the Depressive Symptoms Index was dichotomized a t the 90th percentile. Suicidal ideation. Suicidal thoughts often accompany depression, but are a separate phenomenon. Because of the unique importance of suicide in medical practice, and because it indicates such a severe level
of psychological distress, a separate analysis of the item "Thought about taking your own life" was carried out. Consequently we asked our respondents if they had "Thought about taking your own life" in the previous 12 months. The 4.6% of the sample who reported one or more instances of thinking about taking their own life were coded as 1, and the remainder of the sample was coded 0. Drinking Index. The Drinking Index was computed from the following questions: "In general, how often do you consume alcoholic beverages-never, less than 1day a month, 1to 3 days a month, 1to 2 days a week, 3 to 4 days a week, 5 to 6 days a week, daily?" and: "On a day when you do drink alcoholic beverages, on average, how many drinks do you have? By a drink we mean a drink with a shot of 1'12 ounces of hard liquor, 12 ounces of beer, or 5 ounces of wine." This measure groups respondents according to six categories of frequency and consumption ranging from abstinence of high-volume drinking and binge drinking. The median frequency of drinking for this sample was 1 to 3 times a month, and the median number of drinks per day was two. Relative to the sample medians, high volume daily drinkers were categorized as those who had 3 or more drinks a day, 3 to 4 times a week, up to daily. Binge drinkers were those who imbibed on infrequent occasions--once a month up to 1 to 2 times a week; 5 or more drinks a day. In order to focus the study on "problem" drinking, the index was dichotomized by assigning a code of 1 to the 8.4%of the sample who were classified as either high-volume drinkers or binge drinkers; all others were coded 0. The Drinking Index differs from the traditional Quantity-Frequency measure developed by Cahalan and associates (1969) because it is not beverage-specific and may underestimate the upper limits of consumption levels. However, the Drinking Index is comparable to measures used by other researchers (Neff & Baqar, 1985) and also seems to have construct validity as shown by the results presented in four papers using this measure (Kaufman Kantor, 1990a, 1990b; Kaufman Kantor & Straus, 1987, 1989). Statistical Analysis Logistic regression. Logistic regression or "logit" (Aldrich & Nelson, 1984) was used to test the etiologic models because it permits use of dichotomous dependent variables. Some of the dependent variables were available only as dichotomies (such as whether a child abuse incident had occurred during the preceding 12 months) and other dependent variables (such as drinking type) were converted to a dichotomy because the critical issue is whether there was a high or chronic level of the problem. Since five psychosocial problems were hypothesized as possible outcomes of corporal punishment, five logit models were tested. 549
Model speczfication. Each logit model included a number of variables in addition to corporal punishment, as listed in Table 1. All models included a specification for the age and gender of the subject, family socioeconomic status (SES), and whether there was violence between the parents of the subject because these characteristics are confounded with both corporal punishment and the incidence of violence and mental health problems. The SES measure is a factor score index constructed with the use of the following items: education of the wife and the husband, their occupational prestige scores, and the combined income of the couple. Violence between the parents was measured by asking the subjects if, when they were a teenager, their father had hit their mother, and if so, how often. The same questions were repeated for whether the mother of the subject had hit their father. If either parent was reported as having hit the other, i t was considered a case of marital violence between the parents. In addition to the four specification variables listed above, which were included in all five logit models, some of the psychosocial problems that were hypothesized as effects of corporal punishment can also be part of the etiology of other problems. For example, being a victim of assault by one's husband is known to be associated with depression (McGrath e t al., 1990; Stets & Straus, 1990). Consequently, in addition to testing the hypothesis that corporal punishment as a teenager is associated with a greater probability of assaulting a wife later in life (model E in Table I), wife assault was included as a specification of model A, which tested the hypothesis that corporal punishment as a teen is associated with depression as an adult. Similar reasoning led to including drinking problems and depressive symptoms in the models where these were not the dependent variable, as indicated in Table 1. Graphs. The conditional plotting technique described in Hamilton (1990) was used to present the results showing the effect of corporal punishment i n adolescence on the probab~lityof violence and mental health problems as an adult. Each graph plots one of the logistic regression coefficients for Corporal Punishment (given in the first row of each panel of Table 1)in the form of the predicted probability of the dependent variable, contingent on another of the independent variables in the model, such as separate plots for men and women, while controlling for all other variables in the model.
probability of depressive symptoms as an adult, even when controlling for the effects of a number of other variables. This is remarkable in view of the fact that Part A shows that five other variables are also significant risk factors for depression. The large t values for SES and Gender in Rows 2 and 3 reflect the higher incidence of depression among lower SES individuals and women found in previous studies (Weissman & Klerman, 1978) though these differences may also be influenced by class- and gender-based diagnostic biases and differences in help-seeking behavior. The row labeled "Parent's Marital Violence" shows that growing up in a family in which there is corporal violence between the parents is associated with a significantly greater probabil-
A. ANNUAL PREVALENCE (One or more instances during year) Total sanpte Boys Girts N d e r of Cases Total BOVS
B. CHRONICITY (Frequency of h i t t i n g atwng those using corporal punishment) Averaae F r w e n c y Mean Median Chronicity Cateqories More than m c e
3 or more times 6 or more times 11 or more times 20 or nore times 30 or twre times
57X 35% 14% 8% 3%
80 62% 35% 22% 12% 5%
Corporal Punishment and Depressive Symptcins The t of 4.427 in the first row of Part A of Table 1shows that corporal punishment in adolescence is associatea with a significantly increased
for the adult r e c a l l data was "...when you were a teenager ...Thetheageyearreferent i n which t h i s happened the twst." Since the 13 and.14 were the
teen ages a t which the most parents reported using corporal pun~shment, the parent data i n t h i s table i s for ages 13 and ?4. Thus, the e n t i r e table can be thought of as r e f e r r i n g t o children I n t h e t r e a r l y teens.
Logistic Regression Analyses* Logit Coeff.
4.42 -8.09 5.98 .i3 3.59
.001 .DO1 .nor .--. .231 .001
A. DEPRESSION (N=4,513) Corporal Punishment SES Gender Age Parent's Marital Violence Uife Assault Drinking Constant
,0977 -.0190 .5760 .OD23 .4284
.0221 .0023 .0963 .0031 .I191
B. SUICIDAL THOUGHTS (N=4,523) Corporal Punishment SES Gender Age Parent's Marital Violence Uife Assault Drinking Constant
1066 .0017 .2920 .0095 .5314 1.2760 .la57 -3.5218
.0357 .OD38 .1577 .0059 .I801 .l625 .2407 .3675
2.98 .44 1.85 -1.62 2.95 7.85 .77 -9.58
.002 .328 .032 .053 .002 .OD1 .220 .001
C. ALCOHOL ABUSE (N=4.513)
Corporal Punishment .0522 SES -.0140 Gender -1.7720 Age - .a332 Parent's Marital Violence .1993 Uife Assault ' .3450 Depression .I263 Constant .I194
.0274 .0032 -1319 .. .0046 .1572 , .I521 .I572 .2647
1.90 -4.43 -13.43 .. . -7.19 1.26 2.26 .80 .45
.028 .001 .nnr .--. .001 .lo2 .011 .211 ,326
0. CHILD ABUSE (N=2,419)
Corporal Punishment SES Gender Age Parent's Marital Violence Uife Assault Drinking Constant
,1822 .0004 .I978 .0001 .4017 .8480 .2975 -2.8062
.a324 .0036 .I433 .0083 .I620 .I577 .2276 ,3834
ity of depressive symptoms. The row labeled Wife Assault indicates that current husband-to-wife assaults are associated with the highest probability of depression in either spouse. Corporal Punishment and Suicidal Ideatzon Since the question on thinking about committing suicide is part of the depression index, it is not surprising that the test of the etiologic model for this variable (Part B of Table 1)shows some similarities in the logit estimates for depression. The first row of Part B shows a strong and significant association of corporal punishment with the probability of thinking about suicide during the 12-month period covered by this study, net of other variables in the model. Perhaps the two most important difference are that SES is not significantly related to suicidal thoughts, and that the most significant predictor of suicidal thoughts is assaultive behavior toward the wife. Because of the confounding of gender with depression and with being a victim of assault by a husband, Figure 1 plots the probability of thinking about committing suicide by corporal punishment separately for men and women. The plot lines show that suicidal ideation increases markedly with the frequency of adolescent corporal punishment for both men and women, but i i more prevalent among women.
FIG 1. - PROBABILITY OF THWKING ABOUT SUICIDE BY CORPOML PUNISHMENT DURING TEEN YEARS PROBABILITY ,35
OF T H I N K I N G
1.38 .O1 2.48 5.37 1.30 -7.32
.001 .457 .084 .495 .007 .001 .095 .001
E. UlFE ASSAULT (N=4,5291 Corporal Punishment SES Gender Age Parent's Marital Violence Drinking Constant
- .I418 .0093 - .I592 .a467 .El6 - .3800 .2384
.023 1 .0026 .I013 .0042 .I183 .I487 .2293
6.14 -3.56 1.57 -11.09 6.18 2.55 -1.04
.001 ,001 .058 .001 .001 .006 .I49
The N1s vary because of Listvise deletion. The N in Part 0 is reduced because this refers only to families vith children.
Once ~w'ice 3-5
CORPORAL PUNISHMENT DURING TEEN YEARS 553
Corporal Punishment and Adult Alcohol Abuse Part C of Table 1shows that increasing amounts of corporal punishment are associated with an increasing probability of alcohol abuse (high daily drinking or high volume binge drinking). The findings for the other variables are consistent with previous research. They show that gender is the most significant predictor of alcohol abuse. Age, which was not significantly related to depression or suicidal thoughts, is strongly related to alcohol abuse, which is also consistent with previous research. The other significant variables in the model are SES and wife assault. The negative signs of the coefficients indicate that the probability of alcohol abuse is greater among men, younger persons, and those of lower socioeconomic status. The positive t value for wife assault indicates that the occurrence is associated with a higher probability of alcohol abuse. The greater prevalence of alcohol abuse among men has been a consistent finding of previous research (Cahalan, Cisin, & Crossley, 1969; Clark & Midanik, 1979). Figure 2 therefore plots the relationship of corporal punishment to alcohol abuse separately for men and women. The fact that the plot line for men is well above that for women shows that the probability of excessive drinking is vastly greater for men,
FIG 2. - PROBABILITY OF ALCOHOL ABUSE BYCORPORAL PUNISHMENT DURING TEEN YEARS
CORPORAL PUNISHMENT DURING TEEN YEARS
and the somewhat steeper slope of the plot line for men shows that the amount of corporal punishment by parents is more closely related to alcohol abuse by men. Corporal Punishment and Child Abuse The results in Parts A, B, and C of Table 1have shown that corporal punishment is significantly related to depressive symptoms, suicidal thoughts, and alcohol abuse, but that other etiological variables in the model were even more closely related. The findings in Part D of Table 1 parallel those findings because they reveal that the more corporal punishment the subjects experienced when they were teenagers, the greater the risk that they will go beyond ordinary corporal punishment to acts that are severe enough to be classified as physical abuse. However, the results in Part D differ because comparison of the t values shows that the frequency of corporal punishment as an adolescent is the etiological variable that is most highly related to the probability of physically abusing a child. A close second is violence by husbands, followed by whether the parents of the subject were violent toward each other. Thus, the model in Part D indicates that the presence of other types of family violence are the most important etiological risk factors for child abuse, and that when these other types of family violence are specified in the model, the remaining risk factors (SES, gender of parent, alcohol abuse, and age of the parent) are not significantly related to child abuse. These results are consistent with the previous analyses of factors influencing child abuse which show that wife assault is strongly related to the occurrence of child abuse. In view of this relationship, Figure 3 plots the effects of corporal punishment on child abuse separately for those subjects who had experienced husband-to-wife violence during the year of survey and those who had not. The parallel plot lines indicate that, regardless of whether there was violence in the current marriage, the more corporal punishment experienced as a n adolescent, the greater the probability of physically abusing a child later in life. The upper plot line (wife assault occurred in the previous 12 months) demonstrates that the probability of child abuse is much greater in families where wives are beaten. The somewhat steeper slope of the upper plot line also indicates an interaction effect in which the effect of corporal punishment is greater when the marriage is characterized by wife assault. Corporal Punishment and Wife Assault Although studies have documented the effects of physical violence in the family of origin on increasing the probability of wife assault, the exclusive focus of those studies on the most severe childhood abuse,
FIG 3. - PROBABILITY OF CHILD ABUSE BY CORPORAL PUNISHMENT DURING TEEN YEARS
FIG 4. - PROBABILITY OF WIFE ASSAULT BY CORPORAL PUNISHMENT DURING TEEN YEARS PROBABILITY OF
None Once Twice 3-5 CORPORAL PUNISHMENT DURING TEEN YEARS t. andlor witnessing their mdther being beaten, implies that only these severe levels of violence are risk factors for wife heating in adult life. The first row of Part E of Table 1shows that ordinary corporal punishment, as experienced by half of all American adolescents, is also a risk factor for assaults on wives later in life, even when five other theoretically important variables are controlled (low SES, youthfulness, violence between the parents of the subjects, and current alcohol abuse are also significant risk factors for wife assault). The interrelation of corporal punishment with these other variables is illustrated by the two plot lines in Figure 4. The lower of the two lines shows that even when there has been no violence between the respondents' parents, a history of corporal punishment significantly increases the risk of husbands' assaultive behavior. At the same time, the fact that the plot line for men whose parents were violent toward each other is higher on the X axis of Figure 4 shows that adding parental marital violence in the family of origin to the equation substantially increases the probability of current wife assault. 2
CORPORAL PUNISHMENT DURING TEEN YEARS LIMITATIONS
As noted earlier, the corporal punishment data are based on recall. This concern is somewhat mitigated by the fact that the prevalence rates are almost the same as those based on contemporaneous data obtained by interviewing parents of teenagers. A second potential problem arises because subjects who reported corporal punishment may also have suffered from abusive parenting. This concern is somewhat mitigated by the results of other research in which it was possible to exclude from the analysis children who were also physically abused. That research found that corporal punishment of children whose parents did not engage in either physical or verbal abuse, is associated with increased rates of such problematic behavior as interpersonal problems with other children, delinquency, and aggression (Straus, 1994; Vissing, Straus, Gelles, & Harrop, 1991). There is the possibility that corporal punishment of teenagers may be part of the symptomatology of multiple-problem families, and i t is only this small group of families within the larger sample of thousands who simultaneously exhibit multiple problems of poverty, alcohol abuse, depression, and family violence. This seems to be the least likely
of t h e threats to validity. First, as already noted, corporal punishment of adolescents occurs i n at least half of all American families. Second, t h e use of multiple regression i n which each of these problematic characteristics was statistically controlled makes it unlikely t h a t the adverse effects of corporal punishment a r e confined to "multiple-problem families." The most serious limitation stems from lack of prospective data. A model tested with cross-sectional data can determine if the d a t a fit a pattern of relationships specified by the theory, but i t cannot establish t h e causal direction. I n t h e present case, it seems likely t h a t t h e processes leading to adult psychological distress or aggression a r e reciprocal i n nature and ". . . i n which t h e child is reinforced for hislher own aggression, and i n which t h e child is t h e object of aggression" (Huesmann & Eron, 1984, p. 243; see also Ammerman, 1991). Although the nature of such reciprocal processes need to be investigated by longitudinal and experimental studies, their existence is not inconsistent with t h e hypotheses and findings of this study.
CONCLUSIONS The results of this study, if replicated by longitudinal studies which do not have t h e limitations noted, suggest t h a t use of corporal punishment by parents of adolescents is a risk factor for depression, suicide, aleohol abuse, physical abuse of chiJdren, and physical assault on wives. The sociopsychological' processes which produced these effects need to be determined to provide a basis for treating persons suffering these consequences. For example, for some children, being hit by those they love and on whom they depend for their very existence, may be a traumatic event with consequences similar to other traumatizing events. Another possibility, and one suggested by a recent study, is t h a t corporal punishment tends to create a feeling of helplessness and powerlessness (Straus & Gimpel, 1992). Perhaps even more important a r e the implications for primary prevention of physical and psychological injury. If, as suggested by other evidence, the harmful effects of corporal punishment apply to toddlers as well as teens (Sears, Maccoby, & Levin, 1957; Larzelere, 1986; Vissing e t al., 1991), ending all use of spanking a n d other corporal punishment can make a n important contribution t o primary prevention of physical abuse of children and spouses, depression, suicide, and drinking problems. As is now t h e national policy in Sweden and several other countries (Deley, 1988; Haeuser, 19881, parents need to be unambiguously informed t h a t spanking or other hitting of children and adolescents is never appropriate because, even though i t may serve t o correct a n immediate problem, i t places the child at risk for serious problems later i n life.
REFERENCES Aldrich, J . H., & Nelson, F. D. (1984). Linear probability, logit, and probit models. Newberry Park, CA: Sage. American Psychiatric Association (1987). Diagnostic and Statistical Manual ofMental Disorders (Third Edition-Reuised). American Psychiatric Asso-
Ammerman, R. T. (1991). The role of the child in physical abuse: A reappraisal. Violence and Victims, 6(2), 87-101. Cahalan, D., Cisin, I. H., & Crossley, H. M. (1969).American drinkingpractices (Monograph No. 6). New Brunswick, NJ: Rutgers Center of Alcohol Studies. Clark, W., & Midanik, L. (1982). Alcohol use and alcohol problems among U.S. adults: Results of a 1979 national survey. In Alcohol and health monographs no. 1: Alcohol consumptron and related problems (DHHS Publication No. ADM 82-1190). US Department of Heslth and Human Services, Natlonal Institute on Alcohol Abuse and Alcoholism. Cox. W. M. (1987). Personalitv theorv and research. In H T. Blane. & K. E. Leonard (Eds.), ~ s ~ c h o l i ~ itieories cal of drinking &dalcohol;sm. New York: The Guilford Press. Deley, W. (1988). Physical punishment of children: Sweden and the USA. Journal of Comparative Family Studies, 19(3),419-431. Dohrenwend, B. S.. Kranoff. L.. Askenasv. A. R.. & Dohrenwend. B. P. (1978). ~xem~lification of a method for sca"ling lif; events: The PERI life events scale. Journal of Health and Social Behauior, 19, 205-229. Goluke, U., Landeen, R., & Meadows, D. (1983). A comprehensive theory of the pathogenesis of alcoholism. In B. Kissin, & H. Begleiter (Eds.), The pathogenesis ofalcoholism (pp. 60&675). New York and London: Plenum Press. Grotevant, H. D., & Carlson, C. I. (1989). Family assessment: A guide to methods & measures. New York: The Guilford Press. Haeuser, A. A. (1988).Reducing uiolence towards US. children: Transferring positive innouations from Sweden. Milwaukee, WI: Department of Social Welfare, University of Wisconsin, Milwaukee. Hamberger, L. K., & Hastinrrs, J . E. (1986). Personalitv correlates of men who abuse their partners: A cross-validation stud< Journal of Family Violence, 1, 323341. Hamilton, L. C. (1990). Statistics with stata. Pacific Grove. CA: BrooksiCole. Herman, J. (1992). Trauma and recouery. New York: FreePress. Huesmann, L. R., & Eron, L. D. (1984).Cognitive processes and the persistence of aggressive behavior. Aggressive Behauior, 10,243-251. Jacobsen, N. S., & Revenstorf, D. (1988). Statistics for assessing the clinical significance of psychotherapy techniques: Issues, problems, and new developments. Behauioral Assessment, 10,133-145. Jouriles, E. N., & O'Leary, K. D. (1985). Interspousal reliability of reports of marital violence. Journal of Consulting and Clinical Psychology, 53, dl . .9 --42.1- ..
Kaufman Kantor, G. I1990af.Purenfal drrnkrng, i,rolvnr.?rind chdd aggression. Paper presented at rhc Annual 3fectings of the Anwric;rn I'syhologxnl Association, Boston, August. Kaufman Kantor G. (1990b). Ethnicity, drinking, and family violence. Paper Presented at the Annual Meetings of the American Society of Criminology. Baltimore, MD.
Kaufman Kantor, G., & Straus, M. A. (1987).T h e drunken b u m theory of wifebeating. Social Problems, 34(3), 213-230. Kaufrnan Kantor, G., & Straus, M. A. (1989).Substance abuse as a precipitant o f wife abuse victimizations. American Journal of Drug and Alcohol Abuse, 15(2),173-189. Larzelere, R. E. (1986). Moderate spanking: Model or deterrent of children's aggression in t h e family? Journal of Family Violence, 1(1), 2 7 3 6 . Lekowitz, M., Eron, L., Walder, L., & Huesmann, L. (1977).Growing up to be vio1ent:A longitudinal study of the development ofaggression. New York: Pergamon. McGrath, E., Keita, G. P., Strickland, B. R., & Russo,, N. F . (Eds.). (1990). Women and depression: Risk factors and treatment issues. Washington, DC: American Psychological Association. N e f f ,J . A,, & Baqar, A. H. (1985).Stress-bufferproperties of alcohol consurnption: The role o f urbanicity and religious identification.Journal ofHealth and Social Behauior, 26, 207-222. Newmann, J . P. (1984). Sex differences in symptoms o f depression: Clinical disorder or normal distress. Journal of Health and Social Behavior, 25, 136159. Sears, R. R., Maccoby, E. C., & Levin, H. (19571. Patterns of child rearing. Evanston, IL: Row, Peterson and Company. Stets, J. E., & Straus, M. A. (1990). Gender differences i n reporting marital violence and its medical and psychological consequences. In M. A. Straus, & R. J . Gelles (Eds.),Physicalviolence i n American families: Risk factors and adaptations to violence in 8,145 families (pp. 151-166). New Brunswick, NJ: Transaction Press. Straus, M. A. (1994).Beating the devil out of them: Physical punishment in A,P American families. Straus, M. A. (1991).Discipline and devianc?: Physical punishment o f children and violence and other crimes i n adulthood. Social Problems, 38, 101-123. Straus, M. A. (1990).T h e Conflict Tactics Scales and its critics: A n evaluation and new data on validity and reliability. In M. A. Straus, & J. R. Gelles (Eds.),Physical violence i n American families: Risk factors and adaptations to violence i n 8,145 families (pp. 49-74). New Brunswick, NJ: Transition Press. Straus, M. A. (1971).Some social antecedents of physical punishment: A linkage theory interpretation. Journal of Marriage and the Family, 33, 659-663. Straus, M. A,, Gelles, R. J., & Steinmetz, S. K . (1980).Behind closed doors: Violence i n the American family. N Y : DoubledayIAnchor. Straus, M. A., & Gelles, R. J. (1990).Physical violence i n American families: Risk ,~~~ factors and ada~tationsto violence in 8.145 families. New Brunswick, NJ: Transaction Books. Straus, M. A., & Gimpel, H. (1994). Corporal punishment and economic achievement: A theoretical model and some preliminary data. Chapter 9 i n Straus (1994). Szinovacz, M. E. (1983).Using couple data as a methodological tool: T h e case of marital violence. Journal of Marriage and the Family, 45, 6 3 3 4 4 4 . Vissine. Y . M.. Straus. M. A.. Gelles. R. J.. & Hamoo. J . W . (1991). Verbal k d p~jchoso&alproble'& of children. Child aggression by Abuse and Neglect, 15, 223-238. ~
Wauchope, B. A,, & Straus, M. A. (1990). Physical punishment and physical abuse of American children: Incidence rates by age, gender, and occupational class. In M. A . Straus, & R. J . Gelles (Eds.), Physical violence i n American families: Risk factors and adaptations to violence i n 8,145 families. New Brunswick, NJ: Transaction Press. Weissman, M., & Klerman, G . (1978).Epidemiology o f mental disorders. Archives of General Psychiatry, 35, 70&712. Widom, C. S. (1989).T h e cycle o f violence. Science, 244(49011, 160-166.