Unionid mussels were sampled in the Big South Fork of the Cumberland River, Tennessee and Kentucky, from July to October 1988 with a chain grid of 10 1-m2 quadrats. The chain grid was used to define 100-m2 areas along the stream bed by repeatedly moving the 10-m2 rectangle upstream. Within each 100-m2 area, 30 systematically selected quadrats were sampled to estimate density and size class distribution of mussel populations. Sampling variance within grids reflected the patchiness of mussel distribution and increased with substratum heterogeneity; number of mussels encountered per quadrat ranged from 0 to 29. Among sites, densities ranged from 1 to 8 mussels/m2. Across all sites, precision and estimates of species richness and density did not improve appreciably with sampling effort beyond 15 quadrats. Concurrent density estimates from quadrat and depletion sampling varied significantly among sites. As a percentage of quadrat estimates, depletion sampling consistently underestimated mean density from 8.5% to 87.5% across all sites. Depletion estimates were influenced by mussel size, substratum heterogeneity and observer experience. Also, depletion sampling underestimated smaller size classes (<75 mm) and overestimated larger size classes (>90 mm) by as much as 53% and 56%, respectively. Where substratum variability limits the use of rigid frame samplers and total substratum collection is not an option, the flexible chain grid provides a reliable means of obtaining precise estimates of mussel density and size class distribution.