Fig. 8. Mean body angle for crabs in different flow (high- or low-velocity) and odor (tracking or non-tracking) conditions. Graph shows mean path angle for crabs in each treatment group ± 95% confidence limit. The measure of angular dispersion, r (the mean vector length), is 0.98, 0.98, 0.99 and 0.99 for high-velocity non-tracking, high-velocity tracking, low-velocity non-tracking and low-velocity tracking, respectively. Higher values indicate less variation, and the statistic has a maximum value of 1. The difference between mean angles of tracking vs non-tracking groups is significant only for the low-velocity flow (Watson—Wheeler test; F_1,21=92.24, P<<0.001). The confidence limit is derived from r (Batschelet, 1981). Sample sizes consist of 9, 10, 8 and 14 paths for high-velocity non-tracking, high-velocity tracking, low-velocity non-tracking and low-velocity tracking, respectively.