We evaluated the consistency of three questionnaire methods of assessing work and leisure activity in the rural biethnic population of the San Luis Valley Diabetes Study. A 7-d physical activity recall (PAR), a ranking of usual activity, and a history of usual participation in vigorous activity were used. Energy expenditure (kJ.kg-1.wk-1) (EE) was estimated from PAR. Subjects were 503 adults [49% non-Hispanic white (NHW), 51% Hispanic]. Physical activity at work rather than leisure-time activity largely determined total energy expenditure. Average EE at work increased with work rank for all subjects combined [mean EE (SEE) for rank 1 (low) = 324.2 (24.4), rank 4 (high) = 874.0 (102.1)] and within sex, ethnic, and occupational subgroups. Leisure EE increased with leisure rank only for NHW men and employed women. Similar patterns were observed in comparisons of PAR data with history of vigorous activity. Further development and validation of instruments appropriate for use across population subgroups are needed.