Semi-quantitative food checklists and duplicate beverage samples were collected from up to 80 individuals in Maryland in 1995-1996 in as many as six approximately equally spaced sampling cycles as part of a pilot longitudinal exposure investigation. The duplicate beverage samples were homogenized and analyzed for arsenic (As), cadmium (Cd), chromium (Cr) and lead (Pb) using inductively coupled plasma mass spectrometry (ICP-MS). Metal concentrations (microg/kg) and weights of the duplicate beverage samples (kg/day) were used to derive average daily exposure (microg/day) for each metal. Mixed models and generalized linear models were used to evaluate temporal and population variability of the beverage consumption rates, the log-transformed metal concentrations in the beverage samples, and the associated exposures. The mean number of beverage servings consumed per day was 3.4 (SD 1.9). The temporal variability of the total beverage consumption rates was found to be significant (p = 0.0476). As, Cd, Cr, and Pb were present at quantifiable levels in 93.5, 76.0, 93.5, and 96.7% of the beverage samples, respectively. The mean concentration in the samples was 2.0 (SD 4.4) microg/kg for As, 0.9 (1.6) for Cd, 29.2 (138.5) for Cr, and 2.0 (2.4) for Pb. The mean log-transformed concentrations for As, Cr and Pb and exposure for As varied by as much as a factor of 3 across sampling cycles and were statistically significantly different (p<0.05). Concentrations and exposures of all four metals varied significantly among participants. These findings are discussed with respect to the data collection methods, results from comparable studies, and implications for exposure and risk assessment.