A B Figure 9: Experiment 4: Novel Test Views. A, Error Rate in Session 1 vs. Misorientation

D relative to the training view (M: mono, S: stereo). b, Error rate vs. D in session 4. Note that the basic dependency of error rate on D is the same both under mono and stereo conditions. This is another indication that the same recognition strategy for mono and for stereo stimuli may have developed with practice. 46 Figure 8: Experiment 4: novel test views. In BLOCKINuence of stereo on the development of the representation strategy with practice (six novel tube-like objects; four sessions of three trials per view per object each). Prominence of canonical views in intermixed mono and stereo trials was assessed by computing variation of response time and error rate over views. The strongest dierence between mono and stereo conditions experiment 3, the dierences between the two conditions became insignicant in the last session, indicating that the same basic recognition strategy may have developed for mono and for stereo stimuli. 45 Figure 7: Experiment 3: identical training and test views. In BLOCKINuence of stereo on the development of the representation strategy with practice (six novel tube-like objects; four sessions of three trials per view per object each). Prominence of canonical views in intermixed mono and stereo trials was assessed by computing variation of response time and error rate over views. The variation of response time (CV of RT) decreased with session but did not dier signicantly between mono and stereo conditions. The strongest dierence between mono and stereo conditions was in variation of error rate in session 3 (F = 9:4; d:f : = 1; 19; p < 0:006). The dierence in session 4 is n.s. (F < 1), showing that the distribution of error rates tended to become similar in the two conditions. 44 Figure 6: The coecient of variation of response times over views of the stimuli, for the two sessions of the canonical views experiment. Left: The decrease in the variation of response time over views with practice was signicant, indicating that response times became considerably more uniform in the second session. Right: There was no eect of practice on the variation of error rate over views. Values are M ean 6 S EM. 43 View-sphere visualization of E R = f (viewangle) Session 1 Session 2 Figure 5: A stereo plot of the distribution of error rates on the viewing sphere (same object and same format as in Fig. 4). The dierence between …