We measured 926 smallmouth bass (Micropterus dolomieu), 6,935 yellow perch (Perca flavescens), 6,416 rock bass (Ambloplites rupestris), and 4,852 pumpkinseed (Lepomis gibbosus) otoliths recovered from double-crested cormorant (Phalacrocorax auritus) pellets to determine the sizes (total lengths) of these fish consumed by cormorants. Otoliths were recovered from cormorant pellets collected from 1993 to 2002 at six colonies along the eastern Lake Ontario–St. Lawrence River corridor. Otolith –length fish length regressions were used to estimate the length of fish species consumed by cormorants. Only 1.5% of these otoliths had no visible erosion, 33.3% had minor erosion, and 65.2% had moderate erosion. We found that the exclusive use of uneroded otoliths severely limited the sample size available for estimating fish size and likely would cause an overestimation of fish size. Species-specific differences were evident when using erosion criteria to determine fish size and could result in bias when estimating length, especially for species such as smallmouth bass whose otoliths possess a rostrum that is readily eroded. Using a random sample (n = 100) of all intact otoliths recovered in pellets provided a conservative estimate of fish length that was smaller than that derived from uneroded or minimally eroded otoliths. Annual variation in the size of fish consumed by cormorants was more pronounced than seasonal variation for most species. We describe and recommend a new technique that incorporates both chick regurgitant and pellet samples for estimating the size of fish consumed by cormorants.