This paper reports a study of factors associated with the consumption of 22 diverse foods among 2082 students aged 12-15 years in Tasmania, Australia. Students provided data by completing a printed questionnaire. Multiple linear regression was used to assess how well self-reported usual (past) frequency of consumption of a food could be estimated from their perceptions of the qualities of the food (their liking for it, its healthfulness) and their descriptive social norms for the food (its usage by parents and its usage by friends). For all foods, multiple R was significant (p < 0.0001), with values ranging from 0.28-0.64 (median for the 22 foods being 0.49). Standardized regression coefficients for liking and parental usage were highly significant (p < 0.0001) for all foods. However, consistently lower coefficients were found for friends' usage (11 with p < 0.01, but only three with p < 0.0001) and healthfulness (five with p < 0.01, but only one with p < 0.0001). The absolute and relative magnitudes of the regression coefficients varied substantially between the foods. Liking was a stronger predictor than parental usage for 11 of the foods, and parental usage out-ranked liking for the remaining 11. Regression coefficients for friends' usage and healthfulness were smaller than those for liking and parental usage for all foods studied.