Archives of Sexual Behavior, Vol. 21, No. 6, 1992

Measurement of Sexual Aggression in College Men: A Methodological Analysis James F. Porter, M.S., 1,2 and Joseph W. Criteili, Ph.D. 1

Researchers have devoted increased attention in recent years to the measurement of sexual aggression in college populations. This review describes and critically examines current methods of measuring sexual aggression which rely on a self-reported history of such behavior. We suggest that the construct validity of these approaches can be enhanced through a systematic consideration of instrumentation and methodological issues. Twenty-six empirical studies were evaluated with regard to sample characteristics such as size, type, representativeness, and subject exclusion criteria, as well as study design, measurement method, statistics reported, use of debriefing, and control for socially desirable response bias. Specific instrumentation methodology to measure a self-reported history of sexual aggression was examined in detail. KEY WORDS: methodology; measurement; sexual aggression; college males; rape.

INTRODUCTION This paper critically examines the measurement of self-reported sexual aggression among American college and university males. Its focus, then, is methodological rather than substantive. The reader is referred to recent reviews for surveys of theories and findings of sexual aggression research with this population (Burkhart and Stanton, 1988; Koss and Leonard, 1984; Malamuth, 1984). Given the social importance of research in this area, it is crucial that empirical investigations proceed on a sound methodological basis. ValidaIUniversity of North Texas. 2To whom correspondence should be addressed at 6326 Shady Brook Lane, #1142, Dallas, Texas 75206. 525 0004-0002/92/1200-0525506.50/0 © 1992 Plenum Publishing

Corporation

Porter and Critelli

526

tion of the construct of sexual aggression in nonoffender populations is still in the early stages. Research that systematically assesses the various linkages in the nomological network of relationships among theoretical and observable entities builds support for construct validity (Cronbach and Meehl, 1955). The quality of measurement of observables represents a key dimension in this process. The present analysis closely examines the forms of measurement of the construct of self-reported history of sexual aggression currently in use in empirical research in this area. Studies selected for review included those of college males who, by their own admission, had committed at least one act of some type of sexual aggression. For the purposes of this paper, data on females, frequently included in these studies, were omitted from consideration. First, general observations on the methods employed in these investigations are made. Next, the instrumentation developed for the measurement of a history of sexually aggressive behavior is examined in some detail. Finally, suggestions for methodological improvements are presented.

GENERAL METHODOLOGICAL CONSIDERATIONS Table I presents a breakdown of the 26 studies that met the inclusion criterion and their standing with regard to several methodological dimensions.

Sampling Issues Sample sizes in these studies varied widely, ranging from 50 in Heilbrun and Loftus' (1986) examination of sexual sadism to 2972 in the normative study on the Sexual Experiences Survey (SES; Koss et aI., 1987). In addition, while the great majority of the studies used college men only, three (Malamuth, 1986, 1988a, 1988b) used heterogeneous samples recruited from several sources. In two of these studies (Malamuth, 1986, 1988a), college men constituted 80% of the sample. In the third (Malamuth, 1988b), however, the characterization of the sample as overlapping "almost 50%" (p. 479) with the earlier study (Malamuth, 1986) provides virtually the only descriptor of subjects, rendering external validity difficult to assess. Finally, only 10 of the studies assessed the representativeness of the sample used, or at least discussed the issue. Yegidis (1986), for example, provided data demonstrating the comparability of the sex and race of her sample with that of the university population from

Measuring Sexual Aggression

527

which it was drawn. Without such comparisons, sample biases may remain undetected. Several of the studies explicitly excluded certain subjects. Five studies excluded subjects to achieve sample homogeneity. Templeman and Stinnett (1991) dropped three men with homosexual experience. Kanin (1983, 1985) asked for volunteers who thought they might have raped, but eliminated the men who felt the woman was not traumatized or became a willing participant. Koss et al. (1985), as well as Lisak and Roth (1988), excluded subjects with no reported history of sexual encounters. Malamuth (1988a) ran analyses with and without such subjects, and his findings in the two analyses were "very similar" (p. 130) to one another. It seems clear that subjects without sexual experience will also be without sexually aggressive experience, and that sexually experienced males will have had more opportunity (Malamuth, 1986) to sexually aggress. Finally, Petty and Dawson (1989) eliminated subjects whose questionnaire responses suggested inconsistency in approach, whereas Craig et aL (1989) omitted their three "assaultive" (used physical force) subjects for statistical reasons. To the extent that subject samples are of sufficient size and homogeneity, and are well-described in research reports, interpretation of external validity is enhanced. In addition, care should be given to equivalence of comparison groups in terms of, for example, extent of sexual experience, number of dating partners, frequency of dating and sexual experience, level of sexual "aspiration" (Kanin, 1983), and so forth.

Type of Design

Depending upon the goals and level of control included, studies may be classified as descriptive, correlational, quasi-experimental, experimental, or mixed in design (cf. Campbell and Stanley, 1963). Descriptive designs do not attempt to assess relationships among variables, and six of the studies are of this type. Eight studies utilized correlation-based methodology, including multivariate elaborations such as multiple regression or discriminant function analyses (cf. McNemar, 1962). Two studies (Mosher and Anderson, 1986; Smeaton and Byrne, 1987) can be considered quasi-experimental in that they combined experimental methodology (random assignment of subjects to groups) with analysis of intact groups. The Craig et al. (1989) and Malamuth (1988a) studies can also be considered quasiexperimental due to their statistical treatment of history of sexual aggression as an independent variable. Eight studies combined two or more design strategies, resulting in a mixed design. Most frequently, descriptive goals were combined with correlational (e.g., Petty and Dawson, 1989) or

528

Porter and Critelli

-~o~

I

-I-

I

+

I

I

I

i

t

I

L

I

I

-I-

I

I

I

-I-

i

+

i

I

I

l

i

I

++

"=! 0

+

+

< -H

-~, ~.~

÷I

0

0

0

[

+

I

o

I

o

o ~ o

+

+ + o

~

+

+

+

+++

÷

+÷+

+

~

~, ~

+

~ ~

.0 e-, ell

e-,

~

N

#

o

~: ~

o

~

0

~

÷,

Measuring Sexual Aggression

I

+ +

+

!

I

+

I

I

529

+

I

+

I

U

I

I

I

I

I

]

I

+

I

+

I

I

-t-

I

[

.=:'~ " t=

+~ [

I

I

I

Jr

I

+1

I

~

I

•~

{

~j ~J

I

+~ ~

4-

+ ~

+

I

~

+

+ l

+

+

+

~

Cq

+1

I

I

I

I

Jr

[

+

I

I

f

I

+

i

I

I

I

I

+

+

I

I

l

+

~.~ ~=~ +i +~ +

v% ~e~ t "

+

t ~-

+

~

+

~'~

+

~0

+

+

~1"~

+

~

+

I~

+

~d"

+

+ "~

~'~

£

>,0o

~0

~.

II

II

530

Porter a n d Critelli

group comparison (Templeman and Stinnett, 1991) approaches. The fact that self-reported history of sexual aggression is not experimentally manipulable may account for the absence of "true" experimental designs in these studies. A fuller understanding of sexual aggression in college males requires more programmatic research, attacking the problem with converging operations, using a variety of designs (see, e.g., Comrey, 1978). Correlational findings, for example, may provide the basis for the development of experimental investigations (Campbell and Stanley, 1963). In addition to the designs mentioned above, a small number of sexually aggressive college men might be examined in some detail using, for example, a clinical approach. This strategy has proved useful in studies of convicted rapists (e.g., Groth and Birnbaum, 1979; MacDonald, 1971). Another approach would involve the close examination of particular episodes of sexual aggression, including, if possible, sensitive interviewing of both perpetrator and victim. Sexual Aggression Measure By far, the instrument most frequently used (in 14 of the 26 studies) to measure self-reported history of sexually aggressive behavior in college males is the SES (Koss and Oros, 1982) or some modification thereof. This, and the nine other measurement methods that have been developed, are examined in detail below. Other Methodological Issues Only 5 of the studies failed to specify, in at least a part of the article, whether the correlation reported was parametric or nonparametric (see, e.g., Garfield, 1978). In these cases, the reader is left to assume the authors selected the appropriate coefficient. However, 14 studies did not report internal consistency coefficients for some or all of the scales used. In some cases (e.g., Koss and Oros, 1982; Muehlenhard and Linton, 1987), this was not as relevant, given the study goals and the way the scales were used, as in other cases (e.g., Peterson and Franzese, 1987; Petty and Dawson, 1989), where the use of summed scale scores clearly implied the assumption of a unitary dimension. Twenty studies did not attempt to check the veracity of some or all of subjects' self-reports. Only 1 study conducted a replication of its findings. Seventeen reports failed to indicate that some type of debriefing was undertaken, despite data indicating the prosocial effects of debriefing in research on sexual aggression (Malamuth and Check, 1984). Finally, only 4

Measuring Sexual Aggression

531

investigations attempted to control for social desirability, despite the intuitively probable relevance of such a variable in the self-reporting of sexual aggression. Heilbrun and Loftus (1986), for example, argued that episodes of sexual aggression may have been underreported due to their socially undesirable nature despite subjects' anonymity and the precaution of embedding the critical items among questions on a variety of dating topics. Porter et aL (1992) found that social desirability was negatively correlated with self-reported sexual coercion in a population of college males, suggesting that those concerned with social evaluation tend to underreport sexual aggression experiences.

MEASUREMENT OF HISTORY OF SEXUAL AGGRESSION While a "proclivity to rape" (Malamuth, 1981) has been frequently studied in terms of self-reported likelihood to rape (Malamuth, 1984) or sexual arousal to rape depictions (Abel et al., 1977), investigations addressing an actual history of sexually aggressive acts would seem to tap more directly a propensity to sexually aggress. The various instruments used to measure sexual aggression history, however, have been confounded by their lack of explicitness regarding several important dimensions of this variable. This section first describes the assessment procedures and then examines their overall success at treating the various potential confounds. These are many dimensions involved in the behavioral history of sexual aggression. In general, these have not been overtly addressed in the empirical investigations. However, it is crucial to the development of construct validity in this area of research that the specification of these dimensions be explicit and precise. M e a s u r e m e n t Methods

The most widely used instrument, the SES (Koss and Oros, 1982), was originally composed of 13 items but has more recently been used in a 10-item format (e.g., Koss et al., 1987). There are male and female forms for use in the assessment of both perpetration by males and victimization experiences of females. Items include information on both the type of sex act that occurred and the tactic or method of coercion used by the male. An example of a question tapping a relatively mild level of sexual aggression is "Has a woman given in to sex play (fondling, kissing, or petting, but not intercourse) when she didn't want to because she was overwhelmed by your continual arguments and pressure?" A sample question repre-

532

Porter and Critelli

senting a severe level of sexual aggression asks "Has a woman had sexual intercourse when she didn't want to because you threatened or used some degree of physical force (twisting her arm, holding her down, etc.) to make her?" Within each item, sexual aggression is treated dichotomously (yesno). Some investigators have then totaled the yes responses to arrive at a total score (e.g., Greendlinger and Byrne, 1987; Peterson and Franzese, 1987), whereas others have divided subjects into various levels according to the most severe aggression endorsed (e.g., Koss and Dinero, 1988; Malamuth, 1988a). In other studies, the SES has been modified to some degree. For example, Lisak and Roth (1988) included a new item to tap fantasy activity involving rape, but without a history of acting on the fantasy. Craig et al. (1989) and Smeaton and Byrne (1987) modified the SES items to assess the frequency of occurrences, rather than treating them dichotomously. The Coercive Sexuality Scale (CSS) devised by Rapaport and Burkhart (1984) consists of 19 items. Eleven items ask about various acts with increasing degrees of intimacy, ranging from "How many times have you held a woman's hand against her will?" to "How many times have you had intercourse with a woman against her will?" The remaining 8 items assess, with increasing degrees of force, the method or tactic used by the man. These range from "attempted to verbally convince a woman" to "used a weapon on a woman." In the original version of the CSS, all questions can be responded to at any of four levels of frequency: never, once or twice, several times, or often. In later studies (Porter et aL, 1992; Tang et al., in press), five frequency options were offered. Two studies (Rapaport and Burkhart, 1984; Smeaton and Byrne, 1987) used a single total score as the index of coercive sexuality, while the two later CSS studies provided two additional scores: intimacy of sexual act and level of force used. The eight remaining procedures for the assessment of sexually aggressive episodes are specific to the particular researchers. Kanin (1983, 1985) collected 71 subjects over a 10-year period by soliciting volunteers from university organizations and classes. He simply recruited males who believed they had or might have engaged in the crime of rape. Subjects participated in an interview and completed a questionnaire. As indicated above, Kanin excluded cases in which, according to the man's report, the victim showed no traumatic consequences or became a willing participant. Since it is clearly possible that the man's perception of the woman's reaction was biased, it is likely that some rape episodes were excluded. In the Heilbrun and Loftus (1986) study of sadistic motivation, a history of sexual aggression was tapped by two items embedded in a 51-item dating questionnaire. This procedure has the advantage of disguising the purpose of the study which, theoretically, would tend to minimize con-

Measuring Sexual Aggression

533

tamination from a socially desirable response bias. The first question established the history; the second question assessed frequency. If the frequency was greater than one, the subject was considered sexually aggressive. As in the Kanin (1983, 1985) studies, the means of assessment leaves open the possibility that some episodes of sexual aggression remain undetected. The question asked whether the subject had had an experience "in which his dating partner had expressed dissatisfaction with him because he had exceeded the sexual limits that she would have preferred" (p. 324). In the case of limited protest by the woman, the man's perceptual threshold becomes a key factor in whether he realized that the woman was indeed expressing "dissatisfaction." Research by Shotland and Goodstein (1983) has indicated the importance of the woman's level of protest to the man's perception of his behavior as sexually aggressive. Incarcerated rapists often discount their crime, rationalizing that "she said no, but she really wanted it" (Scully and Marolla, 1984). For their study on guided imagery of rape, Mosher and Anderson (1986) constructed a scale to assess history of sexual aggression, the Aggressive Sexual Behavior Inventory (ASBI). This 33-item instrument varied level of force and tactics to gain sexual access. Frequency of each behavior was rated on a 7-point Likert scale from never to extremely frequent. There are several items assessing each of several types of tactics, such as verbal manipulation, anger expression, threats, and use of drugs or alcohol. However, some items do not explicitly refer to sex. For example, the item, "I have roughed a woman up a little so that she would understand that I meant business" could be interpreted as referring to a variety of contexts other than a sexual one. A different approach was used by Muehlenhard and Linton (1987) in their study of date rape incidence and risk factors. Subjects responded to a dating questionnaire, one part of which asked if they had engaged in sexual activity that was unwanted by the woman. If so, they were asked to describe characteristics of the date, the type of sex act involved, and the tactics used by the man. Questions were presented in multiple-choice format, but also included "other" as a response choice with space to write in answers. To minimize guessing, a "don't remember" option was also included. Petty and Dawson (1989) developed a Use of Force in Sexual Experiences Scale (UFSE), with 38 items assessing a range of sexually aggressive behaviors up to and including force resulting in the woman's death. Subjects responded on a 4-point Likert-type frequency scale, similar to that utilized in the original CSS. A summed score was used to assign subjects to high and low force groups.

534

Porter and Critelli

The study by Sigelman et al. (1984) on dating violence included two sexual aggression questions among the physical aggression questionnaire items. They were designed to tap two levels of force, although the items were combined in data analyses. Subjects were asked whether they had used either "strong physical force" or "violence" to try to engage in sexual acts against the other person's will. These terms were not explicitly defined or distinguished from one another, however, leaving open the possibility of varying interpretations by subjects. Templeman and Stinnett (1991) used the Clarke Sexual History Questionnaire (SHQ; Paitich et al., 1977). This extensive self-report inventory of various types of sexual behavior includes three items assessing sexually coercive acts (excluding items on frottage not explicitly against the woman's will), and has been used in the evaluation of sex offenders (Langevin, 1983). Sexual activities were subdivided into a number of categories of desire or sexual misconduct. The "coercive sex" category combines rape and attempted rape into a single grouping, acts assessed separately in most of the other systems, Finally, Yegidis (1986) administered a questionnaire asking subjects about experiences with "forced sexual activities while dating" (p. 52). She offered response options involving four types of sexual activity and a range of types of force used by the male, the latter similar to those presented in the CSS. Subjects responded in yes-no fashion to items assessing episodes in the past year and at any time in their lives. Dimensions of Sexual Aggression History Table II presents an analysis of the scales and procedures used to assess a history of sexually aggressive behavior in college men. The methods are arrayed against various dimensions of sexually aggressive events. As shown in the table, the type of sexual act involved and the type of tactic or force used by the man in his attempt to engage in sexual behavior with the woman are both specified in the SES, CSS, and several of the other procedures (Muelenhard and Linton, 1987; Petty and Dawson, 1989; Yegidis, 1986). However, none of the measurement methods fully cross type of act with type of tactic. Thus, the SES, for example, taps the experience of attempted or actual sexual intercourse associated with the tactic of using alcohol or drugs, but there is no item linking "sex play" (i.e., kissing, petting, or fondling, but not intercourse) with the use of alcohol or drugs. The CSS asks separately about both acts and tactics but does not link these with one another. Thus, for example, it is impossible to tell whether having intercourse with a woman against her will was the result of physical ag-

Measuring Sexual Aggression

0

~35

m)

~e

~N

Measurement of sexual aggression in college men: a methodological analysis.

Researchers have devoted increased attention in recent years to the measurement of sexual aggression in college populations. This review describes and...
981KB Sizes 0 Downloads 0 Views