The studies reviewed indicate that suicidal behavior is infrequent among school children (12%), but increases progressively among junior high (35%), high school (65%), and college students (50-65%). Though considerable, these values are lower than those of similar-age psychiatric populations. Estimates of the rate of actual attempts were 3% for elementary students, 11% for high school students, and 15-18% for college students. Most were low-lethality attempts for which medical or other attention was not sought. Accordingly, the vast majority of suicide attempts will not be uncovered by investigations dealing solely with clinical or medically identified populations. The most commonly identified correlates of suicidal behaviors included depressive symptoms, social problems, family disorganization and problems, life stress, and poor problem-solving skills. Academic problems were not as important. These findings must be interpreted in light of methodological constraints. Response rates have been low, samples have been small, minorities have been underrepresented, nonstudents have been ignored, and volunteers of unreported characteristics have predominated. The representativeness of such groups is questionable. The definition and categorization of behaviors have varied considerably, making between-study comparisons difficult. The differences among thoughts, threats, and attempts (the most frequently used categories) have often not been taken into account. Rather, in analyses, these behaviors have been treated as if they were homogeneous entities and grouped together. Where attempters have been separated from ideators, they too have been treated as a single entity. Many deliberately survived attempts are manipulative and oriented toward the benefits the individual survivors expect. Such attempts probably differ substantially in nature from those attempts where survival is not intended. Understanding of this phenomenon could be enhanced by a comparison of the characteristics of those with high- and low-lethality attempts. Similarly, a third group of behaviors--that of chronic behavior patterns that potentially hasten death (alcoholism, drug abuse, self-mutilation, and risk taking)--has not been adequately addressed in terms of its relationship to more discrete suicidal events. Data have generally been obtained via anonymous self-administered questionnaires. The accuracy of retrospective self-reports on emotionally laden events is suspect. Only in two instances have self-reported questionnaire data been validated via interviews with subjects alone or subjects and their families. The time frames on which the referent behaviors has been collected have included "last week", "last month", "last year", "lifetime", and not specified. The longer the interval between the occurrence of the behavior and the report, the greater the likelihood that some reporting inaccuracies may occur.(ABSTRACT TRUNCATED AT 400 WORDS)