Linguistic markedness and category learning

Several psycholinguistic theories have appealed to the linguistic notion of markedness to help explain asymmetrical patterns of behavioural data. We suggest that this sort of markedness is best thought of as a derived rather than a primitive notion, emerging when the distributional properties of linguistic categories interact with general-purpose category learning mechanisms. To test this claim, four category learning experiments were conducted in which the stimuli were spoken nonsense words (Experiments 1 and 4), musical phrases (Experiment 2), or pictures of robots (Experiment 3). In each experiment, the base rates and variability of the categories were manipulated. A broadly similar pattern of markedness-like behaviour was seen in all three experiments. Our findings highlight the potential role of general-purpose learning mechanisms in theorising about linguistic categories.