Detecting the Scope of Rape: A Review of Prevalence Research Methods

Reviewed by
Priscilla Schulz, LCSW

from an article of the same title by:
Mary P. Koss, Ph.D.
Arizona Health Sciences Center, Tucson, AZ

Published: Journal of Interpersonal Violence,
V. 8 (2), 198-222, June 1993

What is the scope of this article?
How common is rape? Researchers attempt to answer this question in many different ways. This article examines how investigators in the United States have framed and posed questions to randomly sampled subjects, and explains how definitions and methods affect reports of the frequency of rape. In general, "prevalence" is the number of cases that are active during a given period of time. Because rape's impact is long lasting, some studies' prevalence rates count all individuals that have ever been raped during an entire lifetime. Other studies use shorter time periods. In table form the author summarizes twenty studies of rape prevalence for girls and women, and refers to others. Included in her table are short descriptions of the studies' samples, data collection methods, participation rates, types of screening, contexts of questioning, what was being measured, and the prevalence rates found.

Threats to Validity
The accuracy of studies based on self-report is threatened by fabrication, or false reports, and by nondisclosure, or failure to report incidents that did in fact occur. The author's investigations led her to conclude that fabrication is rare. Nondisclosure, though, whether intentional or not, is a significant source of error. Rape victims may expect to be doubted and denigrated if they disclose, or may not recall being raped. Methods aimed at overcoming subjects' reluctance to report and difficulty remembering rape should be employed.

The examined studies found a range of results. A handful of the studies reported that about 20% of adult women have been raped during their lifetimes. Another handful reported figures closer to 10%. For girls up to age 18 the rate was less variable at about 12%. For female psychiatric inpatients and women in jail the rates approached 50% and 100% respectively. Rates reported for all adult men vary from .6% to 7%. For male psychiatric inpatients and men in jail the rates were 0% to 4% and 8% respectively. Since threats to validity suppress rather than inflate rates, the higher end of the range is considered more accurate for the general population.

The Effect of Definitions

One source of variability of rates among studies is in definition of rape. Many studies presented no explicit definition. A group of others showed many similarities since they based their definitions on recent laws that 1) are gender neutral, 2) broadly define penetration, and 3) include assaults by spouses. Only a few studies included men and boys as victims, however. Statutory age, which varies by jurisdiction, may cause similar assaults to be counted as rape in one place but not another because a legal definition was used. Where statutory age was an issue, studies varied in the age used. It is suggested that completed and attempted rape be counted separately to make comparing across studies more valid. To emphasize the numerous sources of variability in rape prevalence estimates, it is noted that in some cases lower rates have been found when rape is more broadly defined than when more narrowly defined, contrary to expectations.

The Effect of Questioning

In one study reviewed, only 27% of female college students who had "forced, unwanted intercourse" referred to the experience as rape. Data suggest that screening adult female subjects by directly asking if they were "raped" yields low prevalence rates. If followed up with thorough explanations, the suppressive effect is reduced.

Single, broadly framed, indirect screening questions are also associated with low rates of prevalence. Presumably subjects are not sufficiently prompted to interpret their own experience as fitting into the particular broad category as presented.

The most promising strategy is to ask a broadly framed question to stimulate a category of memories, then follow up with multiple, narrowly framed questions referring to specific behaviors which will cue the subject to locate targeted memories. This strategy typically yields higher prevalence estimates. Other factors that affect study outcomes include how much time has elapsed since the rape, the order in which survey questions are presented, and whether the subject is prompted to recall events in their life from the past to the present, or from the present to the past.

If rape prevalence is derived from data collected for a larger purpose, the rape data will also be affected. The following are offered as ways to reduce contextual effects in rape data collection:

Isolate the rape items chronologically during an interview or physically on a page.
Label them in a way that separates them from other items.
Provide a special introduction to the items.

The Effect of Data Collection Methods

Interviews, and surveys (in person, by phone, and by mail) have been used to collect data. While there is no clear evidence that prevalence rates will be higher or lower depending on which technique is employed, phone surveys have produced lower rates. In one study that compared interviews by telephone and in person, almost twice as many rapes were reported in person. Allowing the subject to respond in private, assuring confidentiality, and matching the gender of the researcher to the subject are identified as ways to maximize disclosure.


Respondents tend to be younger and better educated than the population as a whole. Why older women report fewer rapes in previous studies is not known. Since reports include a lifetime of experience, and most rapes occur before age 25, rates should only be higher in younger samples if rates are increasing. Rapes may have been less common in the past, may have been forgotten, or may have been reinterpreted. Why better educated women report more rapes is not known. Older respondents and women with less education may simply be less willing to disclose sexual experiences. Within studies, prevalence rates are sometimes very different for different ethnic groups. It is not known whether these are true differences. Other cultural factors, including language ability, may account for the differences. To the extent that subjects are self-selecting, there is a danger that victims are more likely to choose to participate than are non-victims.

Reviewed by Priscilla Schulz, November 1999

Feedback Join Us Site Map VAWPrevention Home
  National Violence Against Women Prevention Research Center © Copyright 2000
(843) 792-2945/telephone       (843)  792-3388/fax