Many previous prevalence studies of polydipsia (PD) have utilized single and often non-biologic measures. In this study we estimated prevalence using specific gravity of urine (SPGU), normalized diurnal weight gain (NDWG), and staff identification (staff ID). Agreement between these two biologic and one behavioral measure was assessed. A total of 572 psychiatric inpatients were assessed for SPGU and NDWG. Unit staff were asked to identify PD patients. Positive and negative PD groups were formed separately based on the SPGU, NDWG, and staff ID data. All three measures were collected on the same day. Prevalence data for the biologic measures varied. The estimate for PD by SPGU (< 1.009 cutoff) was higher (43.4% of sample) than that of NDWG (> 2.5%; 25.4%) or staff ID (21.4%). These prevalence rates did not change substantially after exclusion of medical causes of polyuria. Agreement, assessed by the kappa statistic, was uniformly low among the measures. Weak association between the measures reflects their multidetermined, nonspecific nature, and highlights the lack of a diagnostic standard in the field. The observed prevalence rates must be considered rough approximations. Associations between the measures and certain subject characteristics suggest the measures may identify different types of potential PD patients. These different types of patients are discussed, as are other issues in the measurement of PD. The data suggest estimates of PD are a function of the type of measure used as even biologic measures vary greatly.