If a particular visual problem has to be dealt with often, the brain will start to build connections so that the problem can be resolved more rapidly. Visual problems that require long and elaborate processing will eventually be resolved in milliseconds. By building new and dedicated connections, elaborate processing steps may be simplified into a fast and short set of interactions. Conscious processing will turn into unconscious processing, because conscious processing has triggered perceptual learning that in turn evokes synaptic changes that create new “dedicated modules” that can do the job unconsciously. This leads to a third thesis:[42]
The LEARN-property of phenomenal representations =Df neural representations that require consciousness and invoke phenomenality, at the same time evoke synaptic plasticity mechanisms and learning, in an attempt to make these representations less dependent on consciousness and invoking less phenomenality.
Indeed, there are several arguments for linking consciousness to perceptual learning. Plasticity in the visual cortex comes in many temporal and spatial scales. There are fast- and short-range adaptations or recalibrations, expressed in altered stimulus-response dependencies (e.g., contrast normalization). But receptive fields may also change in size or feature selectivity when exposed to repeated stimulation. Receptive fields literally grow or shift position when their surrounds are stimulated but the receptive field is not (Gilbert & Wiesel 1992). Prolonged depletion of input leads to the induction of new connectivity via fast axonal sprouting of horizontal connections (Yamahachi et al. 2009). Horizontal connections in particular play an important role in both immediate and longer term plasticity of the visual cortex (Gilbert et al. 1996). The repeated execution of Gestalt grouping via the same connections may therefore induce learning (Gilbert et al. 2001), as, for example, is observed in the learning of texture segregation (Karni & Sagi 1991) or in the gradual improvement of contour integration during childhood development (Kovács et al. 1999). In addition, perceptual learning induces a reorganization of the areas involved in encoding the learned object—a process that is mediated by feedback connections (Sigman & Gilbert 2000; Sigman et al. 2005). It seems that the neural machinery that mediates Gestalt grouping and segregation is also the machinery that mediates perceptual learning.
Furthermore, feedback and horizontal connections have been linked to the molecular mechanisms of neural plasticity. A key component in neural plasticity is the NMDA receptor pathway, and in the monkey, NMDA receptor blocking using APV reduces contextual figure–ground modulation (Self et al. 2012). Similarly, in humans, figure–ground segregation is impaired using Ketamine, an anaesthetic which selectively blocks the NMDA receptor at low doses (Meuwese et al. 2013). Also, it was found that Ketamine at sub-anaesthetic doses interferes with the leaning of Mooney figures. Mooney figures are high-contrast versions of images that are hard to recognize when you don’t know what the image is about. Once you have seen its original natural contrast version, however, the Mooney image is readily recognizable. It was found that the neural representation of Mooney images starts to resemble that of their natural versions once they are learned. Ketamine disrupts this rapid learning process, but only in V1, and not in higher visual areas, indicating that feedback from higher areas to V1 is selectively disrupted by Ketamine (Van Loon et al. submitted).
In sum, there are strong indications that link conscious visual processing and its neural machinery—horizontal and feedback connection—are linked to perceptual learning and the molecular mechanisms involved. This may open up a path to a more molecular understanding of consciousness. In addition, it provides us with a clear idea about the function of consciousness: that of building a new repertoire of visual functions, so that eventually conscious processing is no longer necessary.
It must be noted however, that the link between consciousness and learning is controversial. Many instances of “unconscious” perceptual learning exist (e.g., Gutnisky et al. 2009; Seitz et al. 2009; Seitz & Watanabe 2003; Schwiedrzik et al. 2011). An important issue here, however, is whether these are cases of learning without conscious experience of the stimuli that induce the learning, or whether they are instances of learning without cognitive access or attention to these stimuli (see Meuwese et al. 2013). A further clarification of the role of consciousness in learning is required.