Perceptual Learning 2012

In December 2010, two years after the first Perceptual Learning (PL) conference was held in Beijing, the 2nd meeting was held in Eilat, Israel. It was attended by a large group of researchers of Perceptual Learning who collectively, study its behavioral, computational and physiological characteristics. Sessions were largely informal and left time for discussions. The meeting was followed by this special issue, to which all interested colleagues were invited to submit. PL had been one of the first topics that were systematically studied by psychophysicists. The huge training-induced reduction in the tactile minimal 2-point discrimination thresholds and its dynamics was first reported in the middle of the 19th century (Volkmann, 1858). Yet it was the subsequent association between tuning curves of single neurons along the visual hierarchy and the pattern of PL specificity, suggesting plasticity at primary or secondary visual areas in adults, which increased the salience of this field in the last decades. Fiorentini’s seminal paper (Fiorentini & Berardi, 1980) reported that learning to discriminate between patterns of gratings is specific to their spatial frequency and their orientation. Similar specificity to basic physical attributes of the trained stimuli was subsequently reported for motion discrimination (Ball & Sekuler, 1981), detection of texture orientation (Karni & Sagi, 1991), Vernier acuity (Poggio, Fahle, & Edelman, 1992) and other simple visual tasks. The focus of these papers was the ability to tap plasticity at early sensory cortical areas using behavioral paradigms. Although it was often assumed that low-level plasticity is driven by bottom-up mechanisms, further behavioral studies showed that typical practice based improvement, requires task specific training (e.g. Ahissar & Hochstein, 1993; Fahle & Morgan, 1996; Shiu & Pashler, 1992). Ahissar and Hochstein (1997) manipulated behavioral difficulty of the simple orientation detection task, and observed that learning of easy cases precedes that of difficult cases, and is less stimulusspecific. They proposed the Reverse Hierarchy Theory, which asserts that with increasing demands of spatial accuracy learning progresses gradually from high-level cortices to lower levels. Thus, the bottleneck to naïve performance is not the resolution of low-level representations. Subsequent studies, in the auditory (Ahissar et al., 2009; Nahum, Nelken, & Ahissar, 2008) and visual (Jeter et al., 2009) modalities, indicated that behavioral ‘‘precision’’ may be a better characteristic than ‘‘task difficulty’’. In the past decade, PL studies explored the conditions needed to obtain learning, and the conditions needed to increase generalization. The latter is crucial for any practical purpose, e.g. boosting perceptual skills of clinical populations, where the main limitation is the ability to generalize the gained improvement to novel conditions. Thus, Xiao et al. (2008) found that training task A at position