A critical examination of 98 studies on adult children of alcoholics, published from 1986 to 1995, was prompted by suggestions that sampling issues confound research findings. Over half of the studies that were reviewed involved college students. It is important to understand the methodology used in creating a knowledge base that relies heavily on late adolescent and young adult offspring of alcoholics. Thus, two main questions were asked when reviewing these studies: What methods were employed to classify offspring as being children of alcoholics? To what extent does our knowledge about adult children of alcoholics depend on college student samples and clinical samples?
It has been suggested that research findings on adult children of alcoholics are confounded by participant misclassification (Domenico & Windle, 1993); respondent misunderstanding of terminology, such as the meaning of problem drinking (Mathew, Wilson, Blazer, & George, 1993); and reliance on college student and clinical samples (Mintz, Kashubeck, & Tracy, 1995). Since the majority of research on adult children of alcoholics published between 1986 and 1995 used college students, this knowledge base, derived largely from late adolescents, needs to be examined for its methodological validity.
Excluding etiologic reports from the body of research on alcoholics' adult children, there were 100 original investigations between 1986 and 1995. For the present review, 98 empirical studies were found and critically examined. Specifically, two main questions were asked: What methods were employed to identify and differentiate alcoholics' and nonalcoholics' offspring? To what extent does our knowledge about alcoholics' grown children depend on college student and clinical samples? In addition, the question of whether parental alcoholism ought to be treated as a categorical or a continuous variable was considered.
Researchers employed a wide range of techniques to categorize persons who did or did not have alcoholic parents. The Children of Alcoholics Screening Test (CAST; Jones, 1987, 1994; Pilat & Jones, 1984/1985) was the most frequently used instrument. Participants in 48 empirical studies were classified according to CAST scores (most researchers reported using the recommended cutoff score of 6). Nevertheless, only 10 studies provided CAST results, and a few employed altered versions of the CAST (e.g., reduced number of items and modified scoring criteria).
For example, one study classified respondents as adult children of alcoholics if they answered affirmatively on 6 or more of the standard CAST items or 5 of the standard CAST items plus an item about either biological parent. Another identified daughters of alcoholics based on a CAST score of 6 or greater, as well as CAST scores less than 6 if respondents noted parental drug or alcohol problems on other instruments. Still another set the threshold at 4 out of 5 of the CAST items presented, or respondents answered affirmatively to the following question: "Do you think one of your parents is or was alcoholic?" Thus, an identical CAST score (e.g., 5) was used to classify participants as children of alcoholics in some studies but not others.
Further, categorizing respondents based on a single item is not reliable. Single items (used to categorize respondents in 15 investigations) do not provide necessary and sufficient criteria to classify participants, nor do they distinguish degrees of impact on children arising from parents' alcoholism. There is another threat to research validity: operating under the assumption that members of the general public can differentiate social drinking, alcohol problems, abuse, and dependence.
A problem with terminology was demonstrated by the following single-item criterion: "I had a parent while I was growing up who had a drinking problem." The word "alcoholism" may be intentionally avoided because of the concern that some respondents, due to social stigma or denial, would refuse to admit having an alcoholic parent. …