Amnesic patients show normal patterns of perceptual learning (Fahle and Daum, 2002). A hallmark of perceptual learning is its specificity. As stated by Thorndike’s NVP-BKM120 in vivo law of identical elements, the transfer of any learning from one task to another cannot happen unless the two tasks share identical elements (Thorndike and Woodworth, 1901a, 1901b, 1901c). When applied to perceptual learning, identical elements encompass not only identical stimulus components but also the specific task performed on the stimulus. Training at one visual field location and learning in a task involving discriminating lines of a
particular orientation does not transfer to other locations or other orientations (for review see Sagi, 2011). Moreover, learning is specific for stimulus context or for the configuration of the stimulus. As shown in Figure 6, training on a three-line bisection task can lead to a marked reduction in the threshold, but it does not affect performance on a vernier discrimination task, where the stimulus target is in the same visual field location and has the same orientation, and the tasks involve similar attributes (target
position) but the context (parallel lines versus collinear lines) is different (Crist et al., 1997). Because of the selectivity of early visual cortical neurons for simple stimulus attributes, and because their RFs are restricted to a small visual field area, the specificities of perceptual learning have been attributed to functional changes in early visual cortical areas. Numerous studies have found cortical changes associated with perceptual learning, but the changes have manifested
themselves in very different ways. Some have observed changes in cortical selleck magnification, the amount of cortical territory representing a unit of the sensory input, which can be described as “cortical recruitment.” Animals trained on a tactile vibration frequency discrimination task have a larger representation of the trained digit in primary unless somatosensory cortex, and animals trained on an auditory frequency discrimination task have a larger representation of the trained frequency in primary auditory cortex (Recanzone et al., 1992a, 1992b, 1993). It is plausible that the larger cortical area and larger numbers of neurons that are activated by the stimulus increases the signal-to-noise involved in discriminating the stimulus, a phenomenon referred to as probability summation. But it has been observed that performance does not always correlate with the size of the cortical area (Brown et al., 2004; Recanzone et al., 1992a; Talwar and Gerstein, 2001). The overrepresentation of a particular frequency can reduce the amount of Fisher information around the frequency peak and result in poorer rather than improved performance at that frequency (Han et al., 2007). Interestingly, cortical expansion occurring in the initial phase of training can reverse over time, even though the behavioral effects of training are retained (Yotsumoto et al., 2008).