Figure 5.

Illustration of a potential active role of alpha oscillations for perceptual learning, based on recent computational work (Masquelier et al., 2009). (A) STDP-dependent perceptual learning without background alpha oscillations. Learning is illustrated by the ability of a “shape-selective” neuron (or “output neuron”) equipped with STDP to detect the appearance of a target shape (red <), which appears more frequently than two other shapes. As it is illustrated on the left side of the figure, the output neuron receives input from three “orientation-selective” neurons. The three input neurons are connected to the output neuron sharing the same synaptic strength before the stimulation starts. The top row in the central panel illustrates the presentation of the stimuli (shapes), which stimulate the three input neurons. The three input neurons are arranged in a column, responding in this way to specific regions of the shapes (receptive fields). The second row in the central panel indicates that there are no oscillations (LFP) driving the spiking activity of the input neurons. The following rows show the spiking activity of the three input neurons and the output neuron. Since the output neuron is equipped with spike-timing-dependent-plasticity (STDP), the strength of the synapsis connecting the input neurons to the output neuron can change throughout the stimulation according to standard Hebbian rules. During the course of the stimulation, synaptic weights are reinforced whenever input and output spikes coincide within a certain time window, indicated by the gray rectangles. Coincidence of spikes marked in the gray rectangles result in synaptic reinforcement that facilitates the recognition of the target shape (red <). The dashed rectangles indicate some cases in which the spikes of the output neuron don't lead to synaptic reinforcement. At the end of the stimulation, synaptic weights connecting input neurons I and III to the output neuron are reinforced. This reinforcement improves the detection of the target shape. (B) STDP-dependent perceptual learning with background alpha oscillations. Spikes of the input neurons are driven by the stimulus and by ongoing oscillations (alpha LFP), which modulate the membrane potential of the input neurons producing “in-phase” spikes. Output neurons equipped with STDP can learn better to detect coincidences of in-phase spikes, compared to the case where no oscillations are involved. At the end of the stimulation, connections between output neurons and neurons I and II are stronger in the presence of alpha oscillations (B) than without (A), as indicated on the right side of the figure. This implies that the output neuron in case (B) can detect better the presence of the target shape (red <) than the output neuron in (A).