Abstract
Long-lasting biases in the appearance of ambiguously rotating stimuli can be induced by stereo-disambiguated training stimuli (Haijiang et al., 2006). Does this learning depend on visual attention? Methods: Observers (N=13) participated on three consecutive days. Each session consisted of 480 trials. Each trial contained a 2-sec movie of a rotating Necker cube. Observers fixated a central square, verified by a gaze-tracking camera. An arrow (750 ms) at fixation indicated one of four possible task locations. Two locations were assigned an “attended rotation direction” (ARD) of clockwise and two locations were assigned an ARD of counter-clockwise. On Test trials (128/session) an ambiguous cube appeared at one location. On Training trials (352/session), stereo-disambiguated cubes appeared at all four locations. Training stimuli always rotated according to the ARD when observers attended to them (25% of Training trials). In Experiment 1, equal numbers of ARD and anti-ARD stimuli were shown at each location. Thus, 75% of Training trials at a given location were unattended, of which 1/3 had ARD and 2/3 had anti-ARD. In Experiment 2 the unattended cubes always rotated anti-ARD. On Day 3 the ARDs were reversed at all locations (both experiments) to assess long term learning. Results: In Experiment 1, 81±6% (mean ± SE across observers) of Test trials agreed with the ARD on Day 1, increasing by 6±3% (to 87±6%) on Day 2. These learned biases were robust, dropping by only 9±2 % (to 77±6%) with reverse Training on Day 3. In Experiment 2, however, only 41±3% of Test trials agreed with the ARD on Day 1, increasing to 44±2% on Day 2, dropping 12±5% to 32±5% on Day 3. Conclusions: Long term bias for 3D rotation can be learned with or without attention, but 2-3 unattended trials are needed during training to counteract a single attended trial.
NIH R01-EY-013988, HFSP RPG 3/2006, NSF BCS-0810944.