Statistical Learning

It is natural to think of perception in terms of the processing of individual features (such as color and shape) and how they are combined into discrete objects (such as animals and bicycles). This simple characterization underestimates the information that is available in perceptual input, though, because there are also massive amounts of information about how these features and objects are distributed in space and time. In time, for example, eating food at a restaurant is more likely to be followed by paying a bill than by climbing a tree—just as (in English) the syllable /sci/ is more likely to be followed by /ence/ than by /on/. And in space, for example, a car is more likely to be next to a bicycle than to ...