The validity, reliability and generalizability of diary data.

There is need for reliable and valid behavioral measures for field research. Five sets of data were analyzed, in which diary as well as interview approaches were taken. Results suggest that scores derived from one-week diaries have satisfactory construct validity (convergent, predictive and nomological); that, for salient activities, they are less subjective to retrospective bias than are interview data and that, for this reason, interview-diary discrepancies may be useful for hypothesis testing. For items of neutral value, interview and diary provided remarkably similar values. A one-day diary was not an adequate substitute for a one-week diary. Attrition related to diary-keeping resulted in under-representation of certain groups (e.g., low income, little education, minority ethnicity, poor health), and conclusions would be different if based on diarists' responses or those of the intact sample.