Purpose: Along with contour interaction, inaccurate and imprecise eye movements and attention have been suggested to contribute to poorer acuity for "crowded" versus uncrowded targets. To investigate the role of eye movements in foveal crowding, we compared percent correct letter identification for short and long lines of near-threshold letters with different separations. Methods: Five normal observers read short (4 to 6 letters) and long (10 to 12 letters) lines of near-threshold, Sloan letters with edge-to-edge letter separations of 0.5, 1, and 2 letter spaces. Percent correct letter identification for the 2 to 4 interior letters in short strings and the 8 to 10 interior letters in long strings was compared with a no-crowding condition. Results: Letter identification was significantly worse than the no-crowding condition for long letter strings with a separation of 1 letter space and for both long and short letter strings with a separation of 0.5 letter spaces. Observers more often reported the incorrect number of letters in long than in short letter strings, even for a separation of 2 letter spaces. Similar results were obtained during straight-ahead gaze and while viewing in 30 to 40 degrees left gaze, where two of the five observers exhibited an increase in horizontal fixational instability. Conclusions: We argue that lower percent correct letter identification and more frequent errors in reporting the number of letters in long compared with short letter strings reflect an eye-movement contribution to foveal crowding.