Assessing the population status and dynamics of species is an important component of monitoring efforts aimed at improving understanding of relationships between freshwater mussels (Bivalvia:Unionidae) and their environment. Most freshwater mussel population assessments are conducted using raw count (density, abundance) or presence/absence data but relatively few studies account for potential biases associated with incomplete detection of individuals or species during sampling. We conducted a capture-mark-recapture study over 7 y to assess survival and recapture probabilities of three federally endangered freshwater mussels in a small southeastern U.S. stream. Although similar numbers of mussels were collected among sampling occasions, only a small proportion of the individuals present within the sampling site were collected on any given occasion. Mean apparent survival was 0.81 and did not vary among species or among years. Modeling results indicated that recapture probabilities varied among species, through time, and among individuals as a function of shell length. Recapture probabilities for Pleurobema pyriforme and Medionidus penicillatus increased with increased shell length, whereas recapture probabilities for Hamiota subangulata decreased with increasing shell length. Under the generally accepted assumption of constant recapture among sampling occasions, apparent survival estimates ranged from 0.89 to 0.99 among the three species and decreased for H. subangulata with increased shell length. Our study demonstrates that failure to account for incomplete recapture of freshwater mussels could bias inferences regarding the status and trajectory of mussel populations, which may in turn result in the implementation of ineffective management and conservation strategies.