measured for 49 of 57 cases; 17 (34.7 percent) tested alcohol positive, and 4 (8.1 percent) had BACS at or above .10 percent. Although based on a geographically limited sample, Copeland's (1986) results correspond to earlier estimates. Harper and Albers (1964) detected positive BACS in more than 35 percent of pilots involved in the fatal aviation crashes studied, and Modell and Mountz (1990) noted that typical estimates of alcohol involvement by pilots in fatal aviation crashes range from 10 percent to 30 percent. There is general agreement that even low BACS affect pilot safety by impairing a range of perceptual, cognitive, and psychomotor skills (Modell and Mountz 1990). It is less well known that pilot performance may be impaired long after the pilots eliminate detectable blood alcohol, although such impairments would not be detectable through standard blood or breath alcohol analysis. Oosterveld (1970) reported that increased gravity forces associated with flight can produce positional alcohol nystagmus (PAN) up to 48 hours after a pilot consumes alcohol. After reviewing the literature on hangover effects, Gibbons (1988) hypothesized that PAN may contribute to many aviation crashes involving spatial disorientation. Morrow et al. (1990) examined the effects of alcohol and age on radio communication during simulated flight. They noted that radio communication-an essential element of aviation safety and a complex information-processing task that often competes with other flight-related taskscould be particularly susceptible to divided attention deficit, known to accompany alcohol impairment. BACs as low as .04 percent significantly impaired radio communication performance, and other studies have found that the effect is stronger among older pilots. Morrow et al. (1990) also found evidence of long-term hangover effects on pilots' radio communication performance, which was significantly impaired 8 hours after drinking to a BAC of .10 percent. Federal Aviation Administration (FAA) regulations deem it unlawful for any crew member to have a BAC of .04 percent or higher or to fly within 8 hours after consuming alcoholic beverages. In a survey of pilots, Ross and Ross (1990) found that only 37 percent of respondents knew FAA guidelines concerning the .04 percent BAC limit. Respondents were asked to judge the amount of beer, wine, and liquor they could consume in 1 hour before reaching target BACS of .04 percent and .10 percent. Nearly half overestimated the amount of beer or wine they could consume before reaching .04 percent. A smaller percentage overestimated the amount of liquor they could consume before reaching .10 percent. Many pilots, especially heavy drinkers, underestimated the amount of time required for alcohol elimination. Such findings, coupled with experimental evidence about performance impairments at low BACS and potential longterm performance deficits due to hangover effects, present a potentially troublesome picture. There is general agreement that even low BACs affect pilot safety by impairing a range of perceptual, cognitive, and psychomotor skills. Modell and Mountz (1990) recommended modifying FAA regulations to allow a maximum BAC of .01 percent (allowing for errors in measuring devices). They also recommended extending the minimum required time interval between drinking and flying to at least 12 hours and adding a regulation to prohibit flying within 24 hours of consuming five or more standard drinks. Alcohol and Water Safety Estimates suggest that alcohol is associated with between 47 percent and 65 percent of adult drownings (Dietz and Baker 1974; Patetta and Biddinger 1988; Plueckhahn 1982). Howland and Hingson (1988) proposed two reasons why research has not established alcohol's contribution to drownings more fully. First, studies have not included adequate control groups to estimate relative risk. This is true for most studies of alcohol involvement in non-motor vehicle injury events. Second, most studies have not established clear criteria for excluding cases in which the postmortem BAC might not accurately reflect BAC at immersion. Wintemute et al. (1990) analyzed coroner's reports for 442 drownings in Sacramento County, California, between 1974 and 1985. To ensure that their results reflected alcohol involvement in the event, Wintemute et al. analyzed only cases for which BACS had been determined within 24 hours of death and in which death had occurred within 6 hours of immersion (n=234). Forty-one percent of the drowning victims tested. alcohol positive, and 30 percent had BACS of .10 percent or higher. The rate of alcohol involvement in drownings appears to vary by gender, age, and activity type. Alcohol was less likely to be detected in teenage victims than in adults, and victims' BAC was positively associated with age (Wintemute et al. 1990). About 46 percent of victims aged 20 to 64 had BACs of .10 percent or higher, and 38 percent of victims aged 45 to 64 had BACS at or above .20 percent. Alcohol involvement was more likely among male (59 percent) than female (40 percent) drowning victims. The highest rates of alcohol involvement (about 70 percent) were reported for drownings associated with land-based motor vehicles. Alcohol was also present in more than half the drownings associated with swimming, boating, or rafting. Lower rates of alcohol involvement were reported for drownings associated with other activities. Emerging evidence indicates alcohol may be an important factor in water-related spinal cord injuries. The Centers for Disease Control (1988) estimated that between 10,000 and 20,000 spinal cord injuries occur each year in the United States. About 10 percent of spinal cord injuries are diving related (National Committee for Injury Prevention and Control 1989). Using a casecontrol design, Branche et al. (1991) estimated that subjects who experienced diving-related spinal cord injuries were about four times more likely than control subjects to have used alcohol. Although the investigators cited small sample size and difficulties in precisely matching controls to injury victims as important study limitations, their results suggest that alcohol may contribute to diving-related spinal cord injuries. Emerging evidence indicates alcohol may be an important factor in water-related spinal cord injuries. Results from a pilot experiment (Perrine et al. in press) offer corroborating evidence about alcohol involvement in diving-related spinal cord injuries. The investigators examined the effects of alcohol on shallow-water diving performance under controlled conditions. After training, the shallow-water diving performance of 13 subjects was repeatedly measured (videotaped) at 7 testing sessions of 3 dives per session. Subjects consumed premeasured drinks between sessions. (sessions 1 through 5). Performance measurements were obtained at regular intervals on ascending and descending portions of the BAC curve; analysis of these measurements indicated that significant performance deficits were associated with increasing BAC. Although limited by small sample size, this study offers experimental evidence about alcohol-based performance deficits under conditions that closely resemble realworld diving performance. Alcohol and High-Risk The progression of the acquired immunodefi- The emergence of AIDS among male homosexuals and intravenous drug users directed the focus of earlier research on these populations. Several studies found relatively strong associations between substance use (alcohol and recreational drugs) and high-risk sexual behavior among homosexual males (Siegel et al. 1989; Stall 1988; Stall et al. 1986; Valdiserri et al. 1988). In a longitudinal study of homosexually active men, McCusker et al. (1990) found that men who increased their alcohol use during the study period were about twice as likely as those who reduced their alcohol consumption to maintain high-risk sexual practices. Research has begun to focus on the relationship between alcohol and other drug use and high-risk sexual behavior among heterosexuals. To date, the empirical findings remain somewhat equivocal. Ericksen and Trocki (1992) reported that measures of problem drinking behavior were strongly predictive of self-reported sexually transmitted disease. They also reported that highvolume drinking was predictive of a high rate of change in sex partners, which in turn increases the risk of sexually transmitted disease. Temple and Leigh (1992) found that alcohol tended to be associated with encounters involving new sexual partners, but that alcohol use in the event was not predictive of risky sexual behavior. Noting the long incubation period for AIDS and citing recent statistics on the high incidence of AIDS among individuals between 20 and 29 years of age, Hingson et al. (1990) concluded that many in this age cohort probably were infected as teenagers. Accordingly, these investigators used telephone interviews to collect information on sexual beliefs and practices from a randomly selected sample of Massachusetts teenagers (n=1,773). Information on alcohol and drug use was also collected. The results indicated that drug and alcohol use was associated with unsafe sexual practices. Heavy drinkers. (those consuming five or more drinks in a day) were about 2.8 times less likely to report regular condom use than nondrinkers. Respondents who said they had used marijuana during the past month were about 1.9 times less likely than others to use condoms. To more directly examine the influence of alcohol and other drug use on risky sexual behavior, the investigators asked teenagers about condom use in sexual encounters that occurred immediately after drinking or marijuana use (Hingson et al. 1990). Approximately 16 percent of those sampled said they were less likely to use condoms in sexual encounters that immediately followed drinking. Use of marijuana before sex was associated with a 25-percent reduction in reported condom use. These findings suggest a direct association between alcohol and other drug use and unsafe sexual behavior. Similar results were reported in a more recent survey (Strunin and Hingson 1992). In a study of San Francisco-area households, Leigh (1990b) constructed a weighted index of high-risk sexual behavior based on public health guidelines. She reported a strong bivariate relationship between the self-reported frequency of sexual encounters while under the influence of alcohol and risk of AIDS. Multivariate analysis indicated that the interpretation of this association is not straightforward. The observed relationship between alcohol use and risk of AIDS tended to disappear after controlling for the total amount of sexual activity. The best predictor of high-risk sexual behavior was not alcohol or drug use in sexual encounters, but the reported overall frequency of sex. Further analysis indicated that drinking or drug use by the respondent's partner may be associated with high-risk sexual behavior. Leigh (1990b) cautioned that this association could be an artifact of the high correlation between respondents' and partners' drinking. Another methodological caveat about correlational studies is that some third factor, such as personality characteristics, may account for the bivariate correlations between alcohol use and high-risk sexual behavior. Empirical research, using samples representing homosexual and heterosexual populations, has established an association between alcohol use, drug use, and high-risk sexual behavior, but underlying processes and mechanisms that explain this relationship have not been definitively identified. A limitation of several early studies was that overall measures of alcohol and drug use were correlated with overall measures of high-risk sexual activity. Such an approach provides comparatively weak evidence of a direct link between alcohol use during sexual encounters and high-risk sexual behavior. In a longitudinal study of homosexually active men, McCusker et al. found that men who increased their alcohol use during the study period were about twice as likely as those who reduced their alcohol consumption to maintain high-risk sexual practices. Recent studies more directly link substance use and sexual behavior by examining drug and alcohol use during sexual encounters. Evidence strongly suggests that alcohol contributes to behavior that increases the risk of contracting sexually transmitted diseases. Ongoing work indicates that the relationship between alcohol and other drug use and behavior is more complex than was initially anticipated. Earleywine and Finn (1991) found that people who score high on general measures of sensation-seeking were more likely to consume alcohol and to express behavioral disinhibition. Other sociological or psychological attributes may also help explain the relationship between alcohol and other drug use and sexual behavior. Alcohol and Personal Studies of domestic and criminal violence often Alcohol is associated with a substantial proportion of human violence, and perpetrators are often under the influence of alcohol. Welte and Abel (1989) reported that alcohol was most likely to be detected among homicide victims killed by stabbing, among those killed in bars and restaurants, among those killed on Saturday and Sunday nights, and among males killed by females. The prevalence of alcohol involvement was lower among victims killed as a result of other criminal activities. From these findings, Welte and Abel (1989) concluded that alcohol use was most likely to be detected in situations that emerged spontaneously from personal disputes. In a study of 40 male undergraduates, for example, Gustafson (1991) found that alcoholdrinking subjects were more aggressive than placebo-drinking subjects. In another study, Gustafson (1987) divided 28 male undergraduates into drinking and nondrinking groups. After individuals in both groups viewed a violent movie, they wrote stories in response to test pictures and completed semantic differential scales measuring hostility, violence, bloodiness, pleasantness, engagement, positivity, activity, speed, strength, and tragedy. The indirect semantic scales revealed that intoxicated subjects accepted violence more readily and with less negative feeling. Alcohol is associated with a substantial proportion of human violence, and perpetrators are often under the influence of alcohol (Pernanen 1991; Roizen 1982; Welte and Abel 1989). Very high rates of problem drinking are reported among property offenders and violent offenders (Welte and Miller 1987). Wieczorek et al. (1990) used face-to-face interview data collected from a national sample of 1,887 convicted homicide offenders to examine patterns of alcohol and drug use before the crime. More than half the offenders were under the influence of alcohol or drugs at the time of the crime. Alcohol alone was used by about 36 percent, 13 percent used alcohol in combination with other drugs, and only 7 percent used other drugs without alcohol. Wieczorek et al. (1990) found that offenders reported unusually heavy drinking prior to the event. This pattern was consistent across sex and age groups. Based on these findings, the investigators concluded that alcohol was directly implicated in about 50 percent of all homicides. The role of alcohol in violent episodes is not well understood. Alcohol is not the sole cause of violent behavior-many violent episodes occur between sober people with no history of drug or alcohol abuse. The majority of drinkers, even heavy drinkers, never engage in violent behavior (Leonard and Jacob 1988; Shepherd et al. 1989). As Pernanen (1991) stated, "...really dangerous behavior among drinkers is rare. Harmless folly is much more prevalent. On the other hand... such behavior is even rarer in a totally sober state" (p. 212). Personal violence is often associated with drinking by perpetrators. Rates of alcohol involvement and abuse are higher among perpetrators than among victims (Pernanen 1991). Nevertheless, researchers are increasingly examining drinking by the victim. In a study of 3,187 college women, Koss and Dinero (1989) used discriminant analysis to identify characteristics that predict sexual victimization. They found that higher self-reported alcohol consumption was predictive of increased victimization. They speculated that alcohol use could impair women's ability to accurately interpret assault cues and resist unwanted advances. They also noted that women's drinking could act as a cue that triggers responses from sexually aggressive males. In the context of family violence, Miller et al. (1989) compared victimization between two groups of women: alcoholic women from treatment centers and Alcoholics Anonymous groups, and a sample of women from randomly selected households. In this study, the investigators broadly considered spousal violence to include violence that occurred between couples who were not married. Alcoholic women were significantly more likely to have experienced negative verbal conflict with spouses. They were also significantly more likely to have experienced a range of moderate and severe physical violence. Discussing their findings, Miller et al. (1989) stated that drinking by the victim is neither an excuse for nor a direct cause of spousal abuse. Several plausible explanations might explain how victims' drinking could contribute to their increased risk of victimization. Miller et al. (1989) noted that heavy-drinking females are often married to heavy-drinking males, which places them in an environment where their potential exposure to violence is higher. A second possibility is that heavy drinking is more socially stigmatizing for women (Sandmaier 1980). Kagle (1987) noted that inebriated males are often viewed as funny, whereas inebriated females are often perceived as obnoxious or unfeminine. Thus, heavy-drinking females may acquire negative labels and could be perceived as more socially acceptable targets of violence. Dent and Arias (1990) reported findings consistent with this interpretation. They examined the effects of alcohol consumption on attributions of responsibility in marital disputes and found that alcohol consumption by the victim before the event increased negative evaluations of the victim. A third possible explanation is that heavydrinking females may be less sensitive to assault cues and less able to resist assault. Fourth, alcohol abuse by the man, the woman, or both could indirectly contribute to family violence by exacerbating economic problems, child-care difficulties, or other family stressors (Leonard and Jacob 1988). Finally, as Miller et al. (1989) noted, even temporal ordering between victims' drinking and assault has not been clearly established; the experience of victimization could contribute to subsequent heavy drinking. Research has not established which, if any, of these explanations is correct. The relationship between victim and offender drinking and the processes through which family violence emerges is complex and dynamic. In a comprehensive study of violence in a Canadian community, Pernanen (1991) used police records and results from a victimization survey to examine patterns of drinking among victims and perpetrators of violence. Alcohol was impli cated in 42 percent of officially reported violent crimes (table 3). Police records indicated alcohol use by 31 percent of offenders and 26 percent of victims. These rates are lower than those reported in Wolfgang's (1958) classic study of alcohol-related homicide in Philadelphia. Analysis based on official police records appears to underestimate actual alcohol involvement in episodes of violence. Analysis based on official police records appears to underestimate actual alcohol involvement in episodes of violence (Pernanen 1991). Because police records do not systematically include information on alcohol involvement, records that do not explicitly cite it are treated as alcohol negative in calculating base rates, thereby underestimating alcohol involvement. Pernanen (1991) conducted a victimization survey that was supplemented by police reports. Such surveys, which are subject to potential biases, permit an examination of alcohol involvement in many unreported violent episodes. Information on drinking before the event can also be obtained. The victimization survey yielded higher estimated rates of alcohol involvement than those reported in police records. About 54 percent of all reported episodes of violence involved alcohol (table 3). But police records and victimization survey responses suggested somewhat higher rates of alcohol involvement among assailants than among their victims. Larger differences between alcohol use by assailants and victims were observed in the victimization survey than in police records, thus supporting the suggestion of underreporting by police. Compared with official police records, the victimization surveys paint a different pattern of alcohol involvement in violent episodes when gender is considered (table 3). Police records that reported alcohol involvement rates in violent episodes involving female victims were most likely to characterize an episode as alcohol positive when the assailant was male and the victim was female. But the victimization survey suggested that alcohol involvement rates were highest when both victim and assailant were male. This gender discrepancy could involve the stigma against women drinkers. According to |