Moreover, findings from studies on abusers in treatment may not be generalizable to the general population. Only a small fraction of alcohol abusers in the general population are registered by clinics [4,8,29,30] and these are likely to have a more severe drinking problem, and more comorbid disorders [31]. Clinical studies may be well-suited for studying the children most affected by parental alcohol abuse, but less severe cases should be studied in population based samples [3,32]. However, studies with non-clinical assessment of alcohol abuse [33,34] often rely on single responders reporting on both their own outcome and, retrospectively, parental alcohol use. Response style and mood-congruent memory may lead to positive or negative responses to both measures, thereby yielding correlated error terms and inflated effect size estimates. Studies which leave the definition of alcohol abuse to the responder [32,35,36] are especially vulnerable to such biases.
