Connolly, K. Perceptual Learning: The Flexibility of the Senses

Abstract

Entering the topic "perceptual learning" into Web of Science returns over 100 publications for every year since 2008, so it is clearly a hot topic. And that probably doesn't include most of the deluge of papers which use machine learning to distinguish stimulus categories-an endeavor which suggests the amazing potential of learning mechanisms and may provide models for human acquisition of perceptual abilities. Kevin Connolly is a philosopher of mind, and he describes the aim of his book to "offer an empirically informed account of perceptual learning for philosophers." The readership of Perception is not therefore his main target audience, but we can ask: how well does he present empirical perceptual science to philosophers? And, does his philosophical approach offer an analysis that is useful for empirical perceptual scientists? Connolly has certainly done his homework on the empirical science-about 50% of the references cited are to studies in experimental psychology and neuroscience. However, in presenting these results, he does not necessarily convey the questions the scientists were asking. The first issue Connolly addresses is: "Does perceptual learning exist," by which he means "Are the changes described by this term truly perceptual?" While he resoundingly refutes some philosophers who have denied that they are, he sets a boundary on "perception" of which we may be sceptical or impatient. For instance, he cites experiments by Law and Gold (2008) which found that the brain changes when macaques learned to improve their motion coherence thresholds were in decision processing structures rather than early visual areas and concludes that this disqualifies the learning in this case from being "perceptual." The result is significant, but the implied dividing line may be less so-the more we know about the brain networks linking vision to decision, action, and recognition, the more we recognize the essential continuity of these processes, especially given the ubiquitous feedback loops that infuse information from "higher" areas into early visual processing. So Connolly cites several lines of evidence that learning modifies functional magnetic resonance imaging and electroencephalography responses in early visual cortex, but doesn't ask from whence these effects find their way into early visual cortex. Questions about how the mechanisms of perceptual learning interface with the detailed processes of vision do not figure strongly in the issues he considers.