Surprisingly little research has been done on the detection, recognition, and discrimination of curves. Hubel and Wiesel (2, 3, 4) have suggested that there may not be a very extensive underlying network of cells in cats and monkeys responsive to curves. Hirsch and Spinelli (I) were unable to find any evidence that cats reared in normal environments have curvature-specific cells. The present study required that subjects make a timed decision concerning the shape of a briefly exposed, tachistoscopically presented line. Each line subtended 32' of visual angle in length and 4' in width. The lines were presented at 24 retinal locations, i.e., at 3 eccentricities (0.5, 2.75, 5.0) along 8 meridians corresponding to the major map co-ordinates (north, northeast, etc.). The lines were all slightly (10' of arc) or moderately curved (25). The stimuli were presented for 7 msec. at a luminance of 38 cd/ma. Five subjects participated for six 70-min. sessions each. They received 192 test trials during a session. Half of the stimuli wcre curved and half were straight lines. Subjects had to press one of two telegraph keys to indicate the line's shape. An analysis of variance performed on reaction times on trials where responses were correct did not show a significant effect of shape, straight or curved (PI,, = 2.0). although subjects were slightly faster on straight lines, e.g., 728 msec. compared to 746 msec. for curves. Retinal eccentricity was significant (P2.8 = 67.7, p < .001), accounting for 32% of the variance. The deterioration in performance with increasing retinal eccentricity was approximately linear, with a slope of 21 msec. per degree of visual angle. An analysis of variance performed on the errors (corrected for guessing) showed a significant effect of shape (PI., = 8.7, p < .05). Subjects made 23.5% errors on curves and only 12.2% on straight lines. Additional analyses of degree of curvature and direction of curvature (the arc could point either to the left or right) did not indicate significant effects, although subjects were more likely to make errors and to be slightly slower on 10' curves than on 25 curves. Although not overwhelming, the evidence suggests that curves are somewhat more difficult to recognize than straight lines. Timney and MacDonald (5) have suggested that curves are detected by straightline detectors which respond to the chords of the arcs rather than to the curves themselves.