A View from Language : Growth of language in individuals and populations
Abstract: The main question that this book tries to find answers to is what it is that makes language learnable. Language is one of the most complex human activities, but nearly 100% of humanity have learned a language in childhood. One answer could be that it is something in our brain that makes us learn language better than most other activities. Another answer could be that it is something in language as such that makes it to be learned. Two different simulation models are presented in relation to language learnability. The first simulates an individual learning past tense formation in Swedish, and the second model simulates individuals in a population converging on a distinction. € The first model is a connectionist network set to learn the task of forming the Swedish past tense. This model is evaluated on a difficult task of learning a split paradigm of past tense forms in Swedish. The model is shown to be able to replicate some general trends in the learning of a morphological system, but there were problems with keeping the variation of forms, since the network averaged conflicting formation patterns. The problems that arose inspired a new approach to language learnability. € The second model is a model of how the language is shaped by changes in the population.This model demonstrates a plausible scenario for how Scandinavian languages lost case marking as a result of the plague's effects on the population of Scandinavia. The model considers the individual's gradual acquisition of a distinction, geographical constraints, and increased variation in the population. The model can also demonstrate how a distinction can form in a small population from initial random variation. The preferred solution to the learnability problem is to view language from its own perspective as it adapts to the current conditions for being learned. Thus learnability emerges from an interaction between the individual's ability to learn, and the language that is represented in the population by reproduction of the distinctness of linguistic dimensions. Throughout the book it is argued that language formation is compatible with Darwinian principles applied to an essentially non-biological domain, resulting in the selection of linguistic dimensions without the need for adaptive success of the individual. Language is not necessarily a biological adaptation, but could better be formulated from the view that language results from a complex ecology of language distinctions continually reproduced in a population of speakers.
This dissertation MIGHT be available in PDF-format. Check this page to see if it is available for download.