We evaluated the ability of a biological marker (nitrogen excretion expressed as protein) to accurately reflect the protein intake of 12 healthy subjects consuming a low protein diet (0.6 g protein/kg standard body wt). In this crossover study, protein intake was confirmed by chemically analyzing a duplicate of the constant diet each subject consumed for 3 d and by calculating protein content of self-selected diets recorded during two additional 3-d periods. Diet analysis matched excretion (difference 0.03 +/- 0.04 g protein/kg standard body wt, means +/- SEM). Self-selected intake manually calculated by subjects using educational materials matched the prescription [0.60 (0.42, 0.86) g protein/kg standard body wt, median (range)], but underestimated excretion by 0.18 +/- 0.02 g protein/kg standard body wt (means +/- SEM). Self-selected intake recalculated by the authors using a computerized database was only +0.05 (-0.08, +0.44) g protein/kg standard body wt higher than subjects' calculations, suggesting that discrepancies between databases and/or subject calculation errors only partially accounted for how greatly self-selected intake underestimated excretion. In a secondary analysis of self-selected intake, the three dietitian subjects consumed more energy and excreted less protein than nondietitians (137 +/- 4.9 vs. 94 +/- 3.5 kJ/standard body wt; 0.72 +/- 0.02 vs. 0.83 +/- 0.02 g protein/kg standard body wt), suggesting that adequate energy intake and/or additional training might improve agreement between intake and excretion. Thus, discrepancies between protein excretion and reported intake may reflect factors other than willful noncompliance.