Levels of caregiver need were used to predict four patterns of continuity and change in social support over a one-year interval among 376 adults caring for a family member with Alzheimer's disease. Canonical correlation analysis was used to identify predictors of each support pattern from selected caregiver characteristics and needs (i.e., demographics, financial resources, physical and mental health, social and recreational activities, and aspects of the caregiving situation). Three significant canonical variates emerged for each type of support, instrumental and perceived adequacy of support, correctly classifying more than half of the caregiver sample. Results suggest that caregiver need does not necessarily elicit support. Needs do, however, predict several patterns of social support, the most common of which is characterized by stability (high or low support). Depending on type of support (instrumental or perceived adequacy), different configurations of predictors emerge.