Abstract
Prior research has shown perceptual learning to be possible in the absence of feedback, but to progress more quickly with feedback (Herzog & Fahle, 1997). These two findings implicate both unsupervised and supervised mechanisms in perceptual learning. Recent work by Frémaux, Sprekeler, and Gerstner (2010) has shown that mixing these two learning mechanisms may potentially lead to synaptic drift and disruption of learning when simultaneously learning two tasks with differing difficulty levels. When the two tasks are equated for difficulty, however, the model predicts no disruption of learning. These modeling results could potentially explain why perceptual learning for two randomly intermingled tasks (termed "roving") has sometimes been found to disrupt learning (Parkosadze, Otto, Malania, Kezeli, & Herzog, 2008; Tartaglia, Aberg, & Herzog, 2009), but not always (Tartaglia et al., 2009). Furthermore, roving has been shown not only to disrupt learning, but also to disrupt task performance for a learned task (Clarke, Grzeczkowski, Mast, Gauthier, & Herzog, 2013). Here we tested the effects of matched or mixed task difficulty levels on performance for a learned task under roving conditions. Participants performed a bisection task where they viewed three vertical lines and had to indicate whether the central line was closer to the left or right of the interval defined by the three lines. Performance improved with training. Following training, we roved the trained bisection stimulus with a narrower bisection stimulus. Crucially, we split subjects into two groups – one where the roved stimuli were equated for difficulty using an adaptive staircase method, and one where the stimuli were made to differ in difficulty levels by using different staircase procedures for each. We found that indeed, roving's effects depend on task difficulty level in the predicted way. Furthermore, training participants over multiple days revealed that roving's deleterious effects decrease with increasing learning.
Meeting abstract presented at VSS 2018