"Laissez-Faire" Analogical Change


  • Jordan Kodner




language acquisition, corpora, quantitative, computational, morphology


A wide range of proposals have been put forth to account for the many tendencies of analogical change as well as the typological trends that they induce on morphological systems. Many proposals likely do play some role in analogical change, however, their relative contributions are hard to differentiate, since they do all fit the data and are also correlated with one another. A well-defined baseline to compare against would help to evaluate the range of proposed accounts. The Poverty of the Stimulus itself suggests one such baseline. The evidence for parts of a morphological system may be so sparse in the input that the language faculty, no matter how well-endowed, might not steer all learners towards compatible grammars. This article shows that this input sparsity by itself can account for many observed correlations between frequency, paradigm size, and irregularity in morphological systems prior to the involvement of other factors. The often severe sparsity of morphological input is quantified in terms of saturation and measured in child-directed, adult, and historical corpora. Population-level simulations of linguistic transmission and change confirm the intuitions drawn from the corpora: sparsity in the input drives analogical change in consistent directions prior to the influence of any internal factors.