Abstract
The influential Reverse Hierarchy Theory (RHT, Ahissar & Hochstein, 1997, 2004) of perceptual learning posits that task difficulty determines the plasticity site during training, which in turn determines the amount of transfer to new stimuli on subsequent tests. RHT was formulated on the basis of pop-out detection experiments in which some observers trained and tested with easy stimuli (“EzEz”) and other observers trained and tested with difficult stimuli (“DfDf”). This experimental design confounds the training and test conditions. Recent studies using full factorial designs suggested that the transfer of perceptual learning depended on the test and not on training in orientation discrimination (Jeter et al., 2009) and motion-direction discrimination (Petrov, 2009). Our present goal is to test whether this result holds for Ahissar & Hochstein's (1997) pop-out detection task. Method: The stimuli were 7 × 7 arrays of uniformly oriented lines. Half of the arrays contained a pop-out target whose orientation differed by 22 degrees from the background. Difficulty was manipulated via the target position uncertainty: 2 vs. 36 possible positions. After 4 training sessions (1400 trials each), the target and background orientations were swapped for 3 test sessions. Nineteen observers were assigned to 4 groups: EzEz, EzDf, DfEz, and DfDf. The threshold stimulus-to-mask onset asynchrony (SOA) was tracked using Ahissar & Hochstein's (1997) original adaptive procedure. Results: Thresholds decreased significantly with training in all groups. The learning effect transferred more in the EzEz than in the DfDf group, replicating the classic RHT pattern. However, no statistically significant differences were found at test between EzEz and DfEz or between EzDf and DfDf groups. Conclusion: The amount of transfer seems to depend on the difficulty during test and not during training for a variety of tasks, including the pop-out detection task. This challenges a key assumption of the Reverse Hierarchy Theory.