Fig. 4. Characteristics of the second segment. (A) Direction of second segment measured with respect to the direct path from the turn to F2 (angle in Fig. 1). Histograms plot the direction of the second segment in 5° bins for successive blocks of 10 trials after the switch. 0° on the abscissa marks the direct path to F2. N gives the number of data points in each histogram with data pooled across ants and trials. Paths are only included when there is a turn. % shows the proportion of trials with a turn. The direction of the second segment changed over training (Kruskal–Wallis test: ²=22.9, d.f.=2, P<0.001). Accuracy improved between the first and the second blocks of trials but not afterwards (Mann–Whitney test: first block vs second block, z=–3.965, P<0.001; second block vs third block, z=–0.437, P=0.662). (B) Similar histograms showing changes in the distance of the turn from the ant's starting position. The distance between the start and the turn decreased as training progressed (Kruskal–Wallis test: ²=58.3, d.f.=2, P<0.001; Mann–Whitney test: first block vs second block, z=–2.861, P=0.004; second block vs third block, z=–5.587, P<0.001). (C) Histogram showing the direction of the second segment measured with respect to the direct path between F1 and F2, for all those segments from trial 11–30 that were aimed within ±20° of F2. 0° on the abscissa is the direction from F1 to F2. Here and elsewhere, positive angles are counterclockwise.