Validation of the Propensity for Abusiveness Scale in Diverse Male Populations

Donald G. Dutton*

Monica A. Landolt

Andrew Starzomski

Mark Bodnarchuk

University of British Columbia

Department of Psychology

Correspondence to Dr Don Dutton, 2136 West Mall, Department of Psychology, University of British Columbia, Vancouver BC, Canada, V6T 1Z4

Tel. (604) 822-2151, Fax (604) 733-6666

Abstract

The Propensity for Abusiveness Scale (PAS: Dutton, 1995) was designed as a self-report perpetrator profile for intimate abusiveness. It was empirically validated through reports of abuse by intimate partners. The original PAS (Dutton, 1995) was given to 144 men in treatment for partner abuse and 44 demographically matched controls. It correlated significantly with partner reports of abusiveness and correctly classified men 82.2% of the time, as one standard deviation above or below the mean partners’ report score for abusiveness. In the present study, the PAS was given to clinical outpatients, gay males, male college students, and a group of spousal assaulters. A criterion measure for abusiveness (the Psychological Maltreatment of Women Inventory: Tolman 1989, or the Psychological Maltreatment Inventory: Kasian and Painter 1992) was collected from intimate partners. In all groups, the PAS correlated significantly with partners’ reports of both physical and psychological abusiveness on subscales of the criterion measures: Dominance\Isolation and Emotional Abuse. For the college students and wife assault groups, a new criterion measure was used: the Severity of Violence Against Women Scale. The PAS correlated significantly with partners’ reports of threats, and violence measured by this scale. The PAS appears to provide a non-reactive assessment instrument that is a strong predictor of intimate abusiveness across a variety of populations.

Validation of the Propensity for Abusiveness Scale in Diverse Male Populations

Perpetrator reports of their own physically abusive acts can be contaminated by a variety of factors such as reactivity and social desirability, leading to under-reporting of abuse (Arias & Beach, 1987). By reactivity, we mean that men assessed specifically because of abusiveness or a potential for abusiveness, react to direct explicit inventories assessing abuse tactics by underreporting abuse. In part, this reactivity is state specific: a product of the assessment circumstances, which generate shame and defensiveness (Dutton & Hemphill 1992). Browning & Dutton (1986) for example, found court-mandated spousal assaulters' self reports of abuse tactics indicated only about 50% of the incidence of spousal abuse reported by the female partners. They speculated that men chosen for assessment because of their assaultiveness, and who are in the early stages of treatment, may demonstrate high levels of minimizing of violence on self-report inventories. This minimizing is generated by, and indicative of shame about the abusiveness ( Dutton, Starzomski & van Ginkel 1995, Dutton 1998a).

Another aspect of under-reporting is related to social desirability. Dutton & Hemphill (1992) found that perpetrators' reports of both emotional and physical abuse were strongly and significantly correlated with two measures of social desirability. Social desirability as assessed by the Marlowe-Crowne Scale (MC: Crowne & Marlowe, 1960) was highly correlated with Emotional Abusiveness (-.57) and Dominance/Isolation (-.50) as assessed by the Psychological Maltreatment of Women Inventory (Tolman, 1989). Physical abuse was more strongly associated (-.32 p < .05) with the Balanced Inventory of Desirable Responding (BIDR: Paulhus 1984), than by the MC (-.20 ns). Overall, this suggests a strong tendency to underreport both physical and emotional aspects of abusiveness. Social desirability effects on reports of physical abuse were smaller, Dutton & Hemphill (1992) argued, because physical abuse scores have a more constricted range and a leptokurtic distribution. Nevertheless, Dutton & Hemphill (1992) found a significant (-.32) correlation between self reports of Physical Abuse and scores on the BIDR.

A third aspect of underreporting is interactional; reflecting how social desirability operates for specific populations in specific circumstances. For example, absolute social desirability scores for men in treatment for spousal assault are not significantly different from control groups (15.8 versus 16.3 respectively, Dutton & Starzomski, 1994), but social desirability scores will differentially correlate with whatever other self reported aspects the respondents find salient to their assessment. Dutton & Starzomski (1994) found that self-referred men had social desirability scores that were significantly more highly correlated (negatively) with reports of trauma symptoms and personality inventories than the scores of court-referred men. Court-referred men, on the other hand, had social desirability scores that were significantly more highly correlated (negatively) with reports of paternal rejection.

The effects of social desirability on attitudes relevant to spousal assault are also very strong. In a sample of incarcerated men taking an anger management program, Dutton (1993) found large correlations (-.85) between the Marlowe-Crowne and the Check & Malamuth (1985) Hostility Toward Women Scale; although the mean absolute Marlowe-Crowne score (16.1) for this population was not atypical when compared to college student norms.

In general, self report inventories of actions or attitudes about women, when obtained from abusive or potentially abusive men, show a strong relationship to social desirability. However, in a series of studies (Dutton, 1993; 1994a; Dutton & Ryan, 1992; Dutton & Starzomski, 1993, 1994) a variety of less reactive scales have been found to be strongly associated with both emotional and physical abusiveness. These include a personality measure assessing identity diffusion and primitive defenses (Borderline personality (BPO):Gunderson, 1984), the presence of chronic trauma symptoms (Trauma Symptoms Checklist (TSC): Briere & Runtz, 1989), recollections of parental rejection (EMBU: Gerlsma, Emmelkamp, & Arrindell, 1990), attachment style (Relationship Style Questionnaire (RSQ): Dutton, Saunders, Starzomski, & Bartholomew,1994), and anger (Multidimentional Anger Inventory (MAI): Siegel, 1986).

Dutton (1995) described a new scale to assess abusiveness empirically derived from the above scales. The face content of the items on the new scale is less likely to generate reactivity than previous abuse inventories. It does not contain explicit reference to abuse tactics, focusing instead on personality trait descriptions, anger expression, experiences of trauma symptoms, and recollections of paternal treatment. The resulting scale was called the Propensity for Abusiveness Scale (PAS). The PAS is a 29 item Likert scale that requires a Grade 10 reading ability. The PAS is completed by the perpetrator and responses from victims can be used for corroboration.

In the original development of the PAS, Dutton (1995) obtained abuse reports from female partners of 96 men (63 who were in treatment for spousal assault, 33 demographically matched controls) and collected comprehensive psychological profile data on the men using the above five measures. Each scale had previously indicated a strong and significant correlation with wives’ reports of psychological and physical abusiveness. Construction of the PAS involved (1) selecting subscales of the above measures with the highest bivariate correlations with abusiveness as reported by female partners, (2) ascertaining the item-whole correlations of the scales, (3) factor analyzing the new scale, and finally (4) determining associations of the PAS with criterion variables. The criterion measure of abusiveness used with female partners of this original group was the Psychological Maltreatment of Women Inventory (PMWI: Tolman 1989), a 58 item scale with two factors: Emotional Abuse and Dominance\Isolation, and the Conflict Tactics Scale (CTS: Straus 1979, Straus & Gelles 1992) which measures physical abuse.

The original results on the PAS produced a scale with a mean score of 62.2 (SD = 17.1) for self-referred spousal assaulters, and a mean score of 44.7 (11.7) for non-violent demographically matched controls. Court-referred spousal assaulters had scores that fell between these two means. The PAS yielded three factors: Recalled Negative Parental Treatment, Affective Lability, and Trauma Symptoms. The PAS, in the original sample, correlated +.51 (p< .001) with female partners’ reports of Dominance/Isolation by their male partner and +.37 (p < .01) with their reports of emotional abuse on the PMWI. Although there is a general linear relationship between the PAS and criterion scores, there is a "jag" in Dominance/Isolation scores when the PAS reaches the 57-63 range.

The discriminant validity of the scale was assessed by a discriminant analysis of wives reports of the man's abusiveness using the Psychological Maltreatment of Women Inventory. Cut-off criteria were set at one standard deviation from the mean for each factor (< 49.6 and > 105.2 for Dominance/Isolation and < 20.7 and > 54.7 for Emotional Abuse). With these criteria, the discriminant analysis achieved significance ( p < .0001) and correctly classifyed 80.0% of men on the Dominance/Isolation factor and 84.4% on the Emotional Abuse factor. In a group of men being treated for spousal assault mean scores on the PAS are 59.2 (17.1), demographically matched control groups score 44.7 (11.7).

Figure I demonstrates the centrality of PAS scores in the initial study between recollections of early upbringing as measured by the EMBU, "associated features" of abusiveness (Dutton, 1994b, 1995) such as anger and trauma symptoms, and abusive behaviors as reported by the men’s female partners.

<<Insert figures I and II>>

Note that in the non-violent demographically matched sample (Figure II), PAS scores still correlate significantly with emotional abusiveness as measured by the PMWI.

Saunders (1991) reported a procedure for adjusting self-reports of violence for social desirability bias. This technique was applied to the PAS. The "corrected" PAS was again correlated with other key scales and with itself. Application of the Saunders procedure to the PAS made no significant difference in the PAS’ predictive ability. Correlations between the PAS, the PAS corrected for social desirability (using Saunders' correction), and other scales are quite similar. Linear comparisons revealed no significant differences between the two scales in this respect. With PMWI, for example, the PAS-PMWI correlations for Factor 1 were +.51 (uncorrected) and +.48 (corrected), for Factor 2, they were +.37 (uncorrected) and +.36 (corrected).

Dutton (1995) concluded that The Propensity for Abusiveness Scale (PAS) was a short non-reactive assessment device with acceptable accuracy in classifying respondents as being one standard deviation above or below the mean on their female partners' reports of abusiveness. The scale appeared to be non-reactive. It contained no overt items about abusiveness. Although significantly correlated with the Marlowe-Crowne (MC), the strength of this association decreased rather than increased for men who knew they were being assessed for spousal assault. More importantly, corrections for social desirability did not significantly alter the scales' strength of association with the criterion variable; male abusiveness as reported by his female partner. For this reason, inclusion of items measuring social desirability as part of the scale was deemed unnecessary. The particularly low correlation of the PAS with the MC for court-referred men

(-.07) indicated that the instrument might be particularly effective as a non-reactive assessment device with this population.

The PAS distills previous research profiling abuse perpetrators into a short scale predictive of abusiveness. The correlations of the PAS scores with female partners' reports of abusiveness are strong and significant, much stronger than the MCMI correlations, for example. The MCMI, a generic assessment instrument, was not designed for the specific tasks of abusiveness prediction (Choca, Shanley, & Van Denburg, 1992) whereas the PAS has been designed for this specific task.

The PAS also correlates significantly with physical abusiveness as assessed by female partners' reports using the Conflict Tactics Scale (CTS: Straus, 1979). The correlation of PAS and womens' reports on the Physical Abuse Scale of the CTS is + .30 (p = .01) for our original sample. In general, since physical abuse is a highly skewed distribution, correlations will be weaker than with the more normally distributed Psychological Maltreatment of Women Inventory scale. Data transformations can be used to overcome this problem.

<< insert Table I >>CURRENT VALIDATIONAL STUDY

Cross-validation of the PAS required administration of the scale to new groups. Changes in the format of the scale, or the base rate level of abusiveness in the criterion group could affect the discriminative ability of the scale. To examine this possibility, we cross-validated the scale with three groups of men that had not presented for treatment of partner assault: (1) a non-violent clinical outpatient population, (2) male college students, and (3) a sample of gay men in long term relationships. These groups were chosen for the following reasons: the clinical group was expected to represent a group with similar psychological dysfunction who were not assaultive, we wondered if the PAS would still predict emotional abusiveness in this group. The college students, apart from their convenience, represented a late teen group of males. We wondered if PAS would manifest itself in correlations to abusiveness at this relatively early age. Prior abusive samples had averaged 31 years of age. The gay group was chosen to ascertain whether personality factors would still predict abusiveness when conventional male-female gender dynamics were removed. We also tested the scale in a new assaultive sample.

The cross-validation with the clinical outpatient group would also help determine whether the scale predicted abusiveness specifically or was merely correlated with general psychological disturbance. If the former were true, we would expect significant correlations between the PAS and PMWI scores in the outpatient group. If this group showed high general PAS scores but these did not correlate significantly with abusiveness the scale was more likely a predictor of general psychological disturbance.

Method

Subjects were recruited from a variety of sources: advertisement in newspapers, psychological clinics, treatment groups in three cities and college classes. One newspaper advertisement specifically requested gay couples in relationships of at least six months duration. This relationship requirement was also made for heterosexual couples in other groups. Clinical group solicitation included an additional specification for those who had in the recent past sought, or were presently seeking, treatment at an outpatient clinic. Only men with intimate anger problems or alcoholism were rejected, the former because they were too similar to our assaultive group, the latter because of the unreliable nature of their responding. Remaining participants in the clinical sample were solicited from 14 outpatient clinics located in the Vancouver area. Clinics were originally selected from a directory of social services published by Information Services of Vancouver. Data was eventually collected from a total of 12 different clinics.

Counselors at each clinic were informed of the conditions for their clients participation in the study as follows: any man who was presenting for a mental health related problem, who had not previously presented there or elsewhere (as revealed to the councilors in their patient history files) for anger or abuse problems, who were in at least a six month heterosexual relationship, and whose partner was also willing to participate. The resulting sample had an average age of 34 (s.d. 5.1) , modal income of ($35,000 CA, s.d. $9,600.00), and average education of Grade 12.

Men in spousal assault treatment groups were recruited as in the original study (see Dutton, 1995). Gay couples were recruited through newspaper advertisements. The Gay sample was the same age (X = 34, s.d 5.8), but more highly educated (average 2 years of university vs. Grade 12) and had higher modal incomes than the original assaultive, control, and clinical group men (Modal couple income was $60,000 CA, s.d. $12,000.00). College students were recruited as part of their research credit. This group was obviously younger than the other groups (average age = 19, s.d. .96)) and comparable to the Gay group in educational attainment. All respondents in all groups had to be in committed relationships of six months duration.

Participants in all groups were predominantly white Euro-Canadians, 20% of the sample was non-white and from a variety of cultural backgrounds.

Psychological Maltreatment of Women Inventory (PMWI). The PMWI (Tolman, 1989) assesses a broad range of intrafamilial, nonphysical aggression that is not identified by the Conflicts Tactics Scale. It contains 58 items (rated from 1 "never" to 5 "very frequently") which comprise forms of Dominance/Isolation and Emotional/Verbal abuse. Dominance/Isolation includes items related to rigid observance of traditional sex roles, demands for subservience, and isolation from resources. In contrast, Emotional/Verbal abuse includes withholding emotional resources, verbal attacks, and behavior that degrades women. Factor analyses support this two-factor structure. In the sample considered in this study, Cronbach's alpha for the Dominance/Isolation subscale was .82, and for the Emotional/Verbal subscale was .93.

Psychological Maltreatment Inventory (PMI: Kasian & Painter 1992). The PMI is a revised gender neutral version of the PMWI developed for dating populations. . This measure was utilized, in the college age sample and with the sample of gay males.

Conflict Tactics Scale (CTS: Straus, 1979) The CTS is a standardized scale designed to measure the frequency and intensity of 19 tactics used in dyads to resolve conflict. The scale includes rational tactics, withdrawal, and a variety of verbally, emotionally and physically abusive strategies. Respondents report both their own use of these tactics and their use by an interactant. This allows independent assessment of both use of, and being a recipient of, various conflict tactics on affective reactions to stimulus materials. Straus, Gelles and Steinmetz (l980) have published population norms for usage of each tactic in a variety of intimate relationships.

The Severity of Violence Against Women Scale, or the Severity of Violence Against Men Scale (SVAWS, SVAMS: Marshall 1992). Marshall developed these scales to augment the repertoire of assessment instruments for abusiveness. The SVAWS (Marshall, 1992) is a 46-item measure consisting of four subscales: threatening behavior, mild physical violence, severe physical violence, and sexual violence. The SVAWS assesses the incidence and rated severity of physical violence towards women in the past 12 months. The SVAMS is a reworded equivalent. The Marshall scales have two advantages over prior criterion scales; they assess a broader scope of abuse than either the CTS or the PMWI (including threats and sexual abuse). Furthermore, by having men and women rate the severity of each scale item, weighted scores can be generated by multiplying frequency with item severity. The Marshall scales yield nine subscale totals measuring a wide range of abusiveness. These scales were not utilized in all of the samples.

<<Tables I and II>>

Results

Table I shows the basic means and standard deviations for all samples. Table II shows the corresponding correlations between the PAS and criterion measures of abuse for all samples.

The new assaultive men demonstrated an even higher association of Propensity for Abusiveness Scale (PAS) scores with wives' reports of abusiveness than the original sample. Figure III schematically presents the pattern of correlations for PAS scores with other relevant scores in the clinical group. As can be seen from Figure III, the pattern of correlations derived from the Figure I assaultive sample is essentially replicated.

<<Insert figures III,IV,V, and VI, and tables III and IV>>

The mean PAS score for both college students and gay men is similar to that for original demographically matched controls; scores of about 45 (SD 12) seem to characterize community samples who are neither batterers nor in clinical treatment. Nevertheless, PAS scores still correlate significantly with emotional abusiveness in these groups. In the original control sample, this correlation was + 42 (p< .01). In the gay sample, it is +.50 (p .01) and in the college sample it is +.51 (.01). Of particular interest with the gay sample, is the ability of psychological factors as assessed by the PAS, to predict abusiveness in couples where there isn’t necessarily a gender based power dynamic in place (Landolt and Dutton, 1997)

The Propensity for Abusiveness Scale (PAS) performed very well with the new assaultive group, showing significant correlations with anger and with several measures of abusiveness: the Psychological Maltreatment of Women Inventory Factors 1 and 2, and the Severity of Violence Scales (Marshall, 1992), which measure severe physical abuse, threats and sexual abuse. The PAS was not correlated with sexual abuse as it was rarely reported.

One difference did emerge between the original and new assaultive sample. The average BPO score dropped to 62.0 (19.8) in the new sample, while the PAS scores increased to 66.1 (33.8). In part for this reason, the predictive advantage of the PAS over the BPO scale was greater with the new sample than with the original sample. Across all samples correlations between PAS and anger are in the range of +.62 to +.71 (clinical outpatients).

Discussion

The overall performance of the PAS was acceptable in cross validational groups. The PAS "predicted" (was correlated with partners’ reports of) abusiveness in a new assaultive sample, a new clinical sample, a sample of college males, and a sample of gay men. Although these groups differ in several important demographic dimensions (age, income, SES, sexual orientation) the PAS retained its’ strong association with partner reports of abusiveness. It significantly predicts Emotional Abuse, use of Dominance\Isolation, Physical abuse and the use of threats to kill or injure.

The PAS was highly predictive of emotional abusiveness in all groups tested. Its’ ability to predict intimate abusiveness in the clinical outpatient group, suggests that it does not merely generic psychopathology but specifically predicts abuse across groups. In general, the PAS has now been applied to about 500 men in six separate and demographically disparate groups (original assaultive, original controls, new assaultive, gay, clinical, college). Its overall performance has been to significantly predict abusiveness as assessed by three different criterion instruments completed by intimate partners.

One advantage in using the PAS is its’ non-reactive nature. Most items appear unrelated to abuse perpetration, even though they are empirically related to it. Most other assessment scales ask about prior offences (Dutton, 1998b) or attempt to predict risk for severe violence in populations that have already demonstrated severe violence, substance abuse problems and criminal histories. Many assessments utilize clinical instruments that have no empirically demonstrated relationship to abusiveness (see Roehl & Guertin 1998). These instruments are often used with men who already have long histories of intimate abusiveness. The PAS could be used much earlier and prior to criminal justice intervention. It could be used, for example, to establish secondary intervention programs in high schools or in programs with an emphasis on prevention. It could be used by private practice clinicians to establish signposts for intervention. Even in identified assaultive groups, the PAS predicts the severity and frequency of assaultiveness, suggesting that it could be relied upon in the absence of corroborating partner reports.

Future research involves the expansion of testing to other groups. Little is known for example, whether the PAS predicts as well in-groups with different ethnic backgrounds that those examined. Some data collection is going on in several sights, including a number of US cities and Buenos Aires to rectify this situation. Also it is not known whether the PAS would predict female abuse perpetration. We are currently testing the PAS in a group of female batterers to see if it predicts abusiveness in this group.

Finally, some evidence is beginning to amass suggesting that additional items assessing self perceptions of conflict inefficacy may further improve predictability of the PAS (Starzomski 1999).

REFERENCES

Arias, I. & Beach, S.R. (1987). Validity of self-reports of marital violence. Journal of Family

Violence, 2(2), 139-150.

Briere, J. & Runtz, M. (1989). The Trauma Symptom Checklist (TSC-33): Early data on a new

scale. Journal of Interpersonal Violence, 4(2), 151-163.

Browning, J.J. & Dutton, D.G. (1986). Assessment of wife assault with the Conflict Tactics

Scale: Using couple data to quantify the differential reporting effect. Journal of Marriage and the Family, 48, 375-379.

Check, J. & Malamuth, N. (1985). The Hostility Toward Women Scale. Paper presented at the

International Society for Research on Aggression, Victoria, B.C.

Choca, J.P., Shanley, L.A. & Van Denburg, E. (1992). Interpretative Guide to the Millon

Clinical Multiaxial Inventory. Washington, D.C.: APA Press.

Crowne, D.P. & Marlowe, D. (1960). A new scale of social desirability independent of

psychopathology. Journal of Consulting Psychology, 24, 349-354.

Dutton, D.G. (1993). William Head Project: Research Report #1. Correctional Service of

Canada: Ottawa.

Dutton, D.G. (1994a). Behavioral and affective correlates of Borderline Personality Organization

in wife assaulters. International Journal of Law and Psychiatry, 17(3), 265-277.

Dutton, D.G. (1994b). The origin and structure of the abusive personality. Journal of Personality

Disorders, 8(3), 181-191.

Dutton, D.G. (1995). Trauma symptoms and PTSD-like profiles in perpetrators of intimate

abuse. Journal of Traumatic Stress, 8(2), 299-315.

Dutton, D.G. (1998a). The abusive personality. New York: Guilford.

Dutton, D.G. (1998b) Risk assessment for domestic violence recidivism: L.A. County

Probation Department, Los Angeles, CA.

Dutton, D.G. & Hemphill, K.J. (1992). Patterns of socially desirable responding among