Perceptual learning and adult cortical plasticity

Abstract

The visual cortex retains the capacity for experience-dependent changes, or plasticity, of cortical function and cortical circuitry, throughout life. These changes constitute the mechanism of perceptual learning in normal visual experience and in recovery of function after CNS damage. Such plasticity can be seen at multiple stages in the visual pathway, including primary visual cortex. The manifestation of the functional changes associated with perceptual learning involve both long term modification of cortical circuits during the course of learning, and short term dynamics in the functional properties of cortical neurons. These dynamics are subject to top-down influences of attention, expectation and perceptual task. As a consequence, each cortical area is an adaptive processor, altering its function in accordance to immediate perceptual demands.

Figure 1. Long range horizontal connections in V1

Cortical pyramidal cells have axons that extend for long distances parallel to the cortical surface. In this figure these connections are visualized by injecting a genetically engineered adenovirus containing the eGFP gene. Neurons transduced with this virus synthesize the eGFP protein, rendering their full dendritic and axonal arbors fluorescent. Combined with optical imaging and electrophysiological recordings, this technique allows us to establish the relationship between the horizontal connections and the cortical functional architecture. A, surface view of visual cortex with a reconstruction of axonal arbors of ∼6000 neurons transduced with eGFP adenovirus, superimposed on optically imaged map of orientation columns. B, the injection site (top right) labelled neurons within a narrow range of orientation columns C, within a 500 μm radius of the injection site axons contacted orientation columns non-selectively. D, beyond the 500 μm radius the axons specifically innervated orientation columns with the same orientation preference as the injection site. E, the visuotopic extent, at the eccentricity of the injection (4 deg, fovea shown at orange circle), is shown in the shaded blue squares. The labelled axons covered an area ∼4 deg in diameter, or extending 2 deg from the injection site. Scale bar = 1 deg. Adapted from Stettler et al. 2002.

Figure 2. Perceptual learning in contour detection

A and B, the saliency of contours formed by collinear line segments embedded in a background of randomly oriented lines increases with the number of lines. C, with practice, subjects improve their detection of contours made of fewer line segments. The neurometric curves (averaged across all recorded neurons) and the psychometric curves (averaged over the corresponding recording sessions) are shown for two different training phases: the first two weeks of training up to the midpoint of training and the following two weeks. Performance in contour detection increases during the period of training, and the amount of facilitation in neuronal responses shows a parallel change. From Li et al. 2008.

Figure 3. Learning and top-down influences in contour detection

Animals were trained on a succession of tasks, and the facilitation in responses to increased contour length was measured at each stage. The first stage involved a fixation task (A), the second a peripheral dimming task at the contour location (B) and the third stage was the contour detection task itself (C). Only when the contour detection task was learned did the facilitation appear. The facilitation was markedly reduced, however, when the animal returned to performing either the central dimming (D) or peripheral dimming tasks (E). Neurometric curves based on ROC analysis of these data show that the contour related responses are not only subject to learning, but are dynamically influenced by top-down influences of specific perceptual tasks. From Li et al. 2008.

Figure 4. Perceptual learning in a shape discrimination task and the role of top-down influences

A, training on a 3-line bisection task involves discriminating between two stimulus configurations – one where the central line of 3 parallel lines is closer to the one on the left or the one on the right. The amount of offset from the central position required to reliably discriminate between the two is the threshold in the task. After thousands of trials, this threshold can decrease by as much as a factor of 3. Training on the 3-line bisection task leads to a marked improvement on that task (left pair of bars) but no change in a related Vernier discrimination task, where the target line is the same but the context is a collinear line as opposed to the 2 flanking parallel lines. Then training on the Vernier task directly leads to a substantial improvement on that task. The lack of transfer in learning between the two tasks demonstrates the specificity of perceptual learning to context. From Crist et al. 1997. B, when presented with an array of 5 lines, a central target and 4 flanking lines (2 near collinear lines and 2 parallel flanking lines), animals are cued to perform either a 3-line bisection task, where the parallel flankers are task relevant, or a vernier discrimination task, where they are irrelevant to the task. The tuning of neurons to the different offsets of the side flanks changes with the task being performed. In this example, there is much more modulation in the neurons’ response when the side bars are task relevant (red curve) than when they are task irrelevant (dashed black curve). C, the difference in the responses between the task-relevant and task-irrelevant conditions is seen from the outside of the neuronal responses. This suggests that the cortical state is set after the animal is given the cue as to which task it will perform but before the stimulus is presented. From Li et al. 2004.