Words Versus Rules (Storage Versus On-Line Production/Processing) in Morphology
This is an advance summary of a forthcoming article in the Oxford Research Encyclopedia of Linguistics. Please check back later for the full article.
Speakers of most languages comprehend and produce a very large number of morphologically complex words. But how? There is a tension between two facts. First, speakers can comprehend and produce novel words, which they have never experienced and therefore could not have stored in memory. For example, English speakers readily generate the plural form of wug. These novel words often look like they are composed of recognizable parts, such as the plural marker -s. Second, speakers also comprehend and produce many words that cannot be straightforwardly decomposed into parts, such as bought or brunch. Morphology is the paradigm example of a quasi-regular domain, full of only partially productive, exception-ridden patterns, many of which nonetheless appear to be learned and used by speakers and listeners. Quasi-regularity has made morphology a fruitful testing ground for alternative views of how the mind works.
Every major approach to the nature of the mind has attempted to tackle morphological processing. These approaches range from symbolic rule-based approaches to connectionist networks of simple neuron-like processing units to clouds of richly specified holistic exemplars. They vary in their assumptions about the nature of mental representations: particularly, those comprising long-term memory of language. They also vary in the computations that the mind is thought to perform, including the computations that are performed by a speaker attempting to produce or comprehend a word. In challenging all major approaches to cognition with its intricate patterns, morphology continues to provide a valuable window onto the nature of the mind.