Fig. 3. Size discrimination in the training experiments. (A) Relative frequency of unrewarded visits to hollow hemispheres of different sizes, for training experiments with five different positive sizes (r=9, 19, 29, 38.5 and 48.5 mm) and a total of nine bats. The frequencies were calculated and averaged for groups of 100 consecutive unrewarded visits each. The frequency of visits to the unrewarded positive size was set to 100%. Values are means ± s.e.m. The broken line marks the 75% level. Data for up to four individuals were pooled (filled circles: one bat, n=2000; filled squares: two bats; n=1300, 2000; filled triangles: four bats, n=300, 800, 2300, 3100; filled diamonds: one bat, n=2900; open squares: one bat, n=1100). (B) r for 75% discrimination as a function of the radius r. r was measured between the training r and the intercept with the 75% discrimination level, and is always plotted over the smaller value of r. The solid black line gives the linear regression (F_1,16=19.65, R²=0.55, P=0.0004), the broken black line gives the linear regression when passing through the origin (F_1,17=181.62, R²=0.91, P<0.0001). The 95% confidence intervals are indicated by black dotted lines.