Abstract
Data on 876 women from Pennsylvania, USA, aged 21-65, interviewed in 1976, were analyzed by means of the Automatic Interaction Detection algorithm developed by Sonquist to ascertain the effects of a variety of predictors on the dichotomous dependent variable, abstinence/drinking. Drinkers were more likely to belong to religious groups that do not have proscriptions against drinking, or else are ineffective in enforcing such proscriptions; attend religious services less than once a week; have mothers who drink; have European, Black, Puerto Rican or Chinese background; be 21-32 yr old; have 10-17 yr of education; associate with peers who drink; work full time, attend school or, if homemakers, be retired from work outside the home or attend school in addition to being a homemaker.