On how learning mechanisms shape natural languages

Abstract

The goal of this thesis is to argue, on theoretical and empirical grounds, that mechanisms in charge of language acquisition define both synchronic and diachronic patterns of linguistic realisation found across natural languages. This aim leads us to explore, then, how Greenberg's problem is derivative from Plato's problem. An approach to the logical problem of language acquisition in terms of a modified discovery procedure is proposed, which, based on learnability considerations, anchors the acquisition of abstract properties on perceptible ones. In chapter 2, I make the proposal explicit, and I go through some definitions and assumptions that are relevant for the thesis. I also make some comments regarding the methodology that will be used. In chapter 3, I provide the theoretical arguments for the proposal. It is argued that approaching acquisition and linguistic variation as two sides of the same problem has not only desirable theoretical implications, but also allows for a better account of well-known linguistic phenomena. Standard parametric proposals, as well as some of the most famous formal models for parameter setting are reviewed and it is demonstrated that they face some relevant shortcomings. Some guidelines about how to develop a plausible theory of acquisition that also makes relevant predictions for variation patterns is provided. In chapter 4, a learning constraint derived from simplicity considerations on how the learner acquires affixal morphology is proposed. It is argued that some synchronic and diachronic linguistic patterns, namely, those concerning the fusional and the agglutinative nature of affixes, derive from the effects of successive analyses of learners applying this constraint during language acquisition. In chapter 5, a morphophonological mechanism of data analysis is proposed to be active during acquisition. By means of bootstrapping mechanisms, the values obtained by the learner using the data analyser can be postulated to be triggers for the acquisition of some high-order morphosyntactic properties. By using a methodology like this one, some well-known patterns of natural languages can be analysed in the very same terms as those used by the learner when examining his linguistic input. In Chapter 6 I point out some open questions and I provide some research lines for further research.

Date of Award	17 Jun 2014
Original language	English
Awarding Institution	Department of Catalan Studies
Supervisor	María del Carme Picallo Soler (Director)