how the results would generalize across age groups or to comparably aged individuals not enrolled in college. Finally, it is important to note 2 limitations of the assessments of AO and AI. First, reports of AO and AI were retrospective, raising concerns about alcohol-induced memory deficits and/or general bias in recalling prior alcohol-related behavior. However, systematic differences in the reliability of reports of AO and AI based on drinker status seem unlikely, given that the data were collected longitudinally from ages 18 to 22 years, an age range that is considerably closer to AO relative to many previous studies. Second, several response options for AO and AI spanned multiple years (“age 9 or before,” “10 to 12,” “13 to 15”). As such, the delay between AO and AI was likely underestimated or overestimated for some individuals. For example, individuals with an AO of 13 years and AI of 15 years would be assigned a delay value of 0 rather than 2 because AO and AI fell within the answer choice “13 to 15.” Reducing concerns about the measurement limitations, analyses examining onset by delay interactions identified no significant interactions in the prediction of the intercepts for heavy drinking or problems.
