Department of Psychology

California State University, Los Angeles

1. This report is based on a thesis submitted by the first author and supervised by the second to the Department of Psychology, California State University, Los Angeles in partial fulfillment of the requirements for the M. A. degree. We thank committee member Nancy Cobb for her insights, and Annette Ehrlich, Patricia Frazier, Nancy Henley, and James Shanteau for critical comments on an earlier version of the manuscript.

2. Now at University of California, Riverside.

Abstract

Introductory psychology students read summary "police reports" to judge whether an alleged victim consented to sex with a suspect. The factors manipulated were the effects of injury and amount of time between the incident and the filing of a report were manipulated. A functional measurement analysis was used to determine the cognitive algebra underlying the judgments. An additive model was supported by the data. Subjects were more likely to consider the sex consensual if the victim was uninjured and waited three days to report the incident to police. Male subjects assigned higher consent ratings than females for all stimulus combinations.

The publication of Susan Brownmiller's (1975) treatise "Against Our Will: Men, Women and Rape," brought to public attention the unsettling fact that charges of rape are seldom vigorously prosecuted. Women who claim to have been raped are often not believed.

Disbelief takes the form of inferring consent when the woman has claimed nonconsent. These inferences are generally made from the complainant's behavior before, during, and after the rape. Often, they amount to judgments of her character rather than judgments of the assailant's actions (Estrich, 1987; Madigan & Gamble, 1991). Almost any of a victim's actions (or inactions) can be taken as evidence that she was in some way responsible for the incident.

Many studies have examined the connection between aspects of the rape victim's character and judgments about her role in the attack (Calhoun, Selby, & Warring, 1976; Macrae & Shepherd, 1989; McCaul, Veltum, Boyechko, & Crawford, 1990). Issues of consent have rarely been examined experimentally. Consent is a central question in rape trials. If the identity of the suspect is not in doubt, it is essentially the sole issue.

In the present study, two factors were manipulated to investigate the attribution of consent: whether injury was incurred during the incident and how quickly the victim reported to the police. Members of the legal system (police officers, prosecutors, judges and juries) have used both of these factors to determine the credibility of the complainant (Feild, 1978; LaFree, 1980).

Among crimes, rape is unique in requiring "utmost resistance" from its victims in order to establish that the victim submitted rather than consented to intercourse (Harris, 1976; Madigan & Gamble, 1991). In a study of outcomes of court cases involving rape, those involving physical injury to the victim -- one indication of resistance -- were more likely to result in conviction (LaFree, 1980). In the laboratory as well, those victims seen as resisting more strongly were blamed less for the rape (McCaul et al., 1990). Krulewitz and Payne (1978) found that subjects were significantly more likely to define the activity as "rape" when physical force and threats were used against the victim than when nonphysical coercion was used.

The other factor, timeliness of reporting, has also been found to affect the conviction rate for rape. LaFree (1980), in an examination of 124 rape cases filed in a midwestern criminal court (of 912 reported to police), found that women who did not report the incident promptly were less likely to have their complaints come to trial and result in conviction. One prosecutor put it succinctly: "A rape case must be reported immediately to have value" (LaFree, 1980).

Two general approaches have been used in research in this domain: studying outcomes in real rape trials, and simulating rape cases to elicit judgments from subjects. A simulation can be elaborate or simple, either duplicating the substance of a trial (Hastie, 1983) or giving the facts of a case in abbreviated form.

We chose the simulation method for the present work, in part because jury outcomes do not necessarily reflect accurately the thinking and decisions of individual members (Davis, 1992; Davis, Bray & Holt, 1977). The decision is simplified, in that the crime can be fixed without allowing for the possibility of a reduced charge. In a real trial a rape charge may be reduced to one of sexual assault, which carries less severe penalties.

Relieving the subjects of the complexities of a jury setting allows us to focus on the private cognitions underlying reasoning about rape. Such cognitions form the basis for the handling of rape within the judicial system. However, they may be masked by forced governing legal outcomes, such as persuasive attorneys and the pressure to achieve consensus.

Another advantage of studying this topic experimentally is that subjects are not required to make exclusively binary decisions, which is the case in trials where a "guilty" or "not guilty" judgment is the only one available to jurors. This compression of available choices inevitable reduces the information yield, as borderline decisions are indistinguishable from open-and-shut ones. Here, in contrast, a continuous response scale allows subjects to express a broad range of opinion.

In experiments with student subjects, there is an obstacle to capturing the reluctance of participants in the judicial process to label a disputed incident as one of rape. Students have been sensitized to this issue, and the "correct" response is likely to be "rape". This disparity between the campus and the courtroom calls for the researcher to generate sufficient ambiguity to overcome the bias toward a response which is out of character with the real world of a rape case.

In the present investigation, functional measurement methodology was employed. The goal in this approach is to explore how the respondent combines the information presented about the incident to arrive at a judgment (Anderson, 1981). A proposed model of the integration process is evaluated. Here, the two manipulated factors in each study were hypothesized to combine additively. The import of additivity is that the factors contribute independently to the judgment; the subjective impact of injury does not depend upon reporting delay. One of the major advantages of the methodology is that the cognitive rule employed by the subject in making the judgment is observable.1 Functional measurement has been used to study the cognitive algebra of persons making moral and legal decisions (Hommers & Anderson, 1989; Howe, 1991; Leon, 1982), with additivity being the general finding.

Throughout the literature on attributions about rape, male and female subjects have regularly been found to arrive at different judgments. Males are more influenced by characteristics of the victim than females (Macrae & Shepherd, 1989), are more likely to think of rape as less serious (Feldman-Summers & Lindner, 1976; L'Armand & Pepitone, 1982), and are more likely to blame the victim (Calhoun et al., 1976; Kanekar & Nazareth, 1988; McCaul et al., 1990). Females are more likely to consider a given incident to be one of rape (Bridges, 1991). It is of interest whether a gender effect in the present context reflects a difference in judgmental process or is instead a valuation effect, such that males see higher consent regardless of the specifics of an incident.

Introductory psychology students (69 women, 20 men) served as subjects during a regularly scheduled meeting. Participation was voluntary; almost all students chose to participate. The gender imbalance in the sample reflects that in the introductory psychology student population. The mean age was 20.

Each subject responded to written scenarios constructed using a factorial design. A 2x2 design (complainant injured/uninjured x reported immediately/reported three days later) was used. The small design was a compromise dictated by pilot study results. It is customary in functional measurement studies to employ more levels per factor in order to get greater information about the factor's contribution, as well as to enhance statistical power. However, pilot work showed that subjects could not give consistent consideration or independent answers to a large set of lengthy, quite similar summaries. For the same reason, within-subject replication was not feasible. Accordingly, a group analysis was planned rather than the more powerful single-subject analysis usually employed in functional measurement.

The dependent variable was the subject's judgment of whether the victim consented to have intercourse with the defendant. This was indicated by making a mark along a 100-millimeter line to reflect the level of agreement with the statements which anchored each end: "No, she did not consent" at the left end of the line, and "Yes, she did consent" at the right end. This graphic rating instrument avoids biases associated with number habits and artificial attempts to maintain consistency (Weiss, 1980). The line mark was subsequently converted to a number between 0 and 100.

All subjects received a packet consisting of four mock police reports, identical except for the factorial manipulations, and a cover sheet of instructions (see the Appendix for a sample vignette). Each report, along with a rating scale, was presented on a separate page. The order of presentation of the four summaries was counterbalanced, resulting in 24 different orderings.

Each summary contained a description of a sexual incident that the victim characterized as nonconsensual and the defendant characterized as consensual. Pilot work revealed this ambiguity to be vital in eliciting a range of responses from observers. In early pilot phases, the vignettes were unambiguous descriptions of a sexual assault. When there was no question that a rape had occurred, very few subjects assigned culpability to the victim. However, when the element of certainty was removed, as is usual in real rape cases, judgments spanned the entire response scale.

All summaries contained information associated with lower victim credibility. She was divorced (Burt & Albin, 1981; Feldman-Summers & Lindner, 1976; Jones & Aronson, 1973) and she had been raped once before, which made her more likely to be seen as "the kind of person who gets herself in situations like [rape]" (Calhoun et al., 1976). The incident took place in the parking lot of a park, which pilot work showed to generate higher consent ratings than if it occurred in a library parking lot. These elements were held constant.

Verbal instructions included a definition of "consent" and a demonstration of how to use the response measure. Subjects were assured of the anonymity of their responses. The study was described, both verbally and in the written instructions, as an attempt to find out how well the subjects could make a decision from the evidence they would be reading.

The students were run as a group and randomly assigned to one of the different orderings of the stimuli. Subjects finished the experiment in from 10 to 30 minutes, plus about 10 minutes for instructions.

The two manipulated factors produced the expected effects. Lower consent was attributed to an injured complainant (a mean difference of 11.09 on the 100 point response scale), and to one who reported immediately (a mean difference of 18.96). A graphical representation is given in Figure 1; perhaps the most striking features is that in every case, males assigned higher consent than females.

Figure 1. Factorial plot of mean responses, grouped separately for each gender. The lower two curves give ratings for the two scenarios featuring immediate report of the incident, while the higher two are for the scenarios with delayed report. The horizontal axis denotes the injury status of the complainant. The larger the numerical value of the response, the more the subject considers the encounter to have been consensual rather than rape.

The analysis of variance is presented in Table 1. Both substantive factors were significant: "Injury" F(1,87) = 33.92, p < .01; "Delay" F(1,87) = 60.24, p < .01. This result is hardly surprising, as the stimulus factors were selected to yield large effects. The crucial requirement of the additive model is the absence of interaction between these substantive factors. As the obtained F-ratio for the Injury x Delay term was <1, additivity is supported.

Table 1 Analysis of Variance (N = 89)

* p < .05

** p < .01

The graphical counterpart of additivity is parallelism in the cell means, and this may be seen most effectively in Figure 2, in which pooling over gender has been carried out. A cautionary note is in order, in that the small number of stimulus levels used here does not afford as stringent a test of the model as is usual.

Figure 2. Factorial plot of overall mean responses. The lower curve shows ratings for the two scenarios featuring immediate report of the incident; the higher curve is for the scenarios with delayed report.

As Figure 1 would suggest, gender was also significant, F(1,87) = 5.67, p < .05, as was the interaction between gender and injury, F(1,87) = 6.04, p < .05. The proportional N repeated-measures analysis reported in Table 1 was carried out with subjects nested under gender and crossed with the substantive factors. Effects of presentation order are not reported as none was significant.

In addition to these overall results, a post hoc inspection revealed a subgroup worthy of note. Ten subjects, all female, did not conform to the additive model but instead responded "0" ("No, she did not consent") in every condition.

As expected on the basis of previous research, subjects considered a woman who was uninjured and who waited three days to report a sexual assault less credible than one who was injured and reported immediately. Utmost resistance strengthens a woman’s charge. Meanwhile, a delay in reporting the crime led raters to doubt the claim of nonconsent. That these factors proved to be additive means that they contributed independently to the subject’s evaluation. The cognitive rule employed is directly observable. The unique contribution of the functional measurement methodology is that it affords validation of the response instrument, and therefore allows inferences about cognitive strategy in a way that separate items on a questionnaire do not. Here, the factors contributed independently. This means that extent of injury and delay of reporting each affect the subjective decision, with no amplification or compensation. The consistency of the present results with those of previous investigations employing other methodologies underscores the impact of these variables.

Both male and female subjects applied the additive model in their judgments, but consistent with previous studies, significant gender differences were observed despite the reduced power inherent in an unbalanced sample. Mean ratings of consent by male subjects were higher than those of females for all cells of the design. Thus, the cognitive process is the same for both genders, but males assign a higher likelihood of consent than females to both levels of delay and of injury. The gender difference remains even if the data from the post-hoc subgroup of 10 women who responded zero in all conditions are removed.

The question of judged innocence vs. guilt is of practical importance in evaluating rape scenarios. Were the subjects serving on a jury, only those who thought the incident was unambiguously one of rape should vote for conviction. According to the reasonable doubt criterion in American jurisprudence, any opinion that would generate a response other than zero in this experiment should lead to a vote for acquittal.

From the research perspective, the continuous response scale² affords greater insight into the subject's cognition than a forced choice such as guilty vs not guilty. This advantage would be present even without the reasonable doubt issue. If internal evaluations below a particular point (such as the scale midpoint) had to be reported as "yes, it was rape" and those above that dividing point as "no, it was not rape", then the collapsing of responses could lead to distorted estimates. A rating of 49 and one of 52 reflect very similar opinions, yet the 49 would be grouped with very low responses and the 52 would be grouped with very high ones.

In early pilot work, our subjects were reluctant to make any response other than zero. Our attempts to resolve the disparity between the responses we were seeing and what seems to occur in the justice system led us to scrutinize closely the stimulus materials. Initial presentations employed an omniscient reporter to convey the factorial information to the subjects. The suggestion of a factual account caused most respondents to judge the cases to be rape, no matter how the stimulus levels were varied.

By changing the presentation mode to one in which each party to the action gave an independent account, we were able to generate ambiguity. Although the principal facts of the case were not in dispute, the interpretations given them by the complainant and the accused were quite different, requiring the subjects to make inferences. Ambiguity appears to be vital to eliciting a range of responses.

Methodological issues are especially of concern in studies which aim to explain a real-life phenomenon. The accord of the present results with conclusions found using a variety of investigative techniques strengthens the assertion that the findings are valid. The convenience of the functional measurement approach makes it viable as a technology for evaluating such pragmatic efforts as rape sensitivity training for police (Feild, 1978) and prosecutors. With the establishment of a response instrument, the paradigm may be used to explore factors that might not be expected to combine additively, such as ethnicity of the complainant and of the accused.

SUMMARY POLICE REPORT: 25-year-old female called 911 to report a sexual assault at Redwood County Park at approximately 7:00 p.m. Officers arrived at the scene within minutes to take her report and transport her to Monterey View Hospital for an examination.

She was wearing a pair of light running shorts and a short, tight, "crop top," which was torn. Her hair and makeup were disheveled. Visual examination revealed several lacerations and bruises on her face and upper body, including both eyes, her cheeks, her chest and abdomen, and both forearms and wrists. She appeared to be nursing her right shoulder. The hospital examination confirmed these observations of a recent struggle. X-rays were negative for broken bones. Physical exam revealed traces of semen in her vagina, which positively matched the suspect's.

The alleged victim is divorced, lives alone, and has been raped once before. There were no apparent witnesses at the scene.

VICTIM'S STATEMENT: Alleged victim said she had gone to the park to jog after work. While jogging, she said she exchanged pleasantries with the suspect. As she left the park for her car, she reportedly heard a male voice shout "Hey!" in a non-threatening manner. She stopped and turned around. The suspect then caught up to her and pushed her into the bushes. The alleged victim claimed she verbally protested when she realized he wanted to have sex with her. A physical struggle ensued, during which the alleged victim claims she fought to get away from the suspect but he overpowered her with numerous slaps and punches. The alleged victim states that even when he penetrated her vaginally, she continued to struggle and resist. When she sat up and began straightening her clothing, the suspect left the scene on foot.

SUSPECT'S STATEMENT: Suspect's description went out over police radio and he was apprehended in the vicinity of the park. The suspect reported he was "shocked" to be picked up for questioning. He stated that he did have sex with the complainant at the park but that "she wanted to." He said she had been progressively more friendly to him since they first started noticing one another at the park a few months earlier. On this evening, he states that they talked together for nearly his entire visit to the park. According to the suspect, the alleged victim made it clear when she was leaving the park that he should follow her. He said he thought she wanted to have intercourse with him based on her friendly manner and her appearance. He also stated that the alleged victim was "turned on" sexually when he accidentally tore part of her shirt, so he continued to tear it. He said the victim then encouraged him to be "rough" with her and seemed very excited by his playful slaps, which he said supports his claim that she was a willing sexual partner.

QUESTION

We know jurors sometimes acquit defendants in cases like this because it appears that the woman consented to have intercourse. How sympathetic would you be to the defendant's argument that the woman in this case consented to have intercourse?

________________________________________

No, she did Yes, she did

not consent consent

1. Functional measurement employs ordinary factorial stimulus designs and uses analysis of variance as an analytic tool, but the perspective is different from that of a typical search for elements contributing significantly to an outcome. Magnitude of the main effects is seldom an issue, as the levels of each factor are chosen for their expected disparate impact. Rather the question is how the factors combine. The presence of interaction provides evidence against an additive model. Other algebraically expressed cognitive models can also be evaluated empirically using corresponding statistical tests (Anderson, 1981).

2. Although a line mark does not appear to be a sophisticated tool, because it is continuous it is sensitive enough to capture a range of opinions. The question of the validity of the response scale must still be addressed. According to functional measurement logic (Anderson, 1981), success of the additive model via parallelism simultaneously supplies this essential link. The basic argument is that if the scale distorted internal opinions, it would require an unlikely coincidence to make the data appear to conform to the model, namely that bias in the instrument was of just the precise form to cancel out inaccuracy in the model.