Cognitive influences on the evolution of new languages.

Thesis (Ph. D.)--University of Rochester. Dept. of Brain and Cognitive Sciences, 2014.As new languages--pidgins and creoles--develop, they start off highly variable, but quickly become more stable and regular. Historical evidence suggests that these languages become more regular at different rates in different lexical and grammatical categories. Open-class categories settle quickly, while closed-class categories remain variable longer. There is debate, though, as to whether individuals reduce variation or whether variation is lost through social processes of coordination and transmission. This work investigates individuals' patterns of learning and producing a variable language and asks how those patterns can account for the observed historical trends. In a series of three experiments, adult participants were trained for 3 days on a miniature language. The language had two labels for each item in one of three grammatical categories: open-class nouns, locative prepositions, or aspect markers. Labels occurred in free variation, but one (dominant) label occurred 67% of the time, and the other (minor) label occurred 33% of the time. At the end of day 1 and day 2, participants' learning was assessed with a vocabulary test, and after training on day 3, participants produced novel sentences, and their use of dominant and minor labels was scored. In the first experiment, participants showed a cline of learning with grammaticization, learning open-class nouns best. They also showed a cline of regularizing the variable categories, stabilizing closed-class categories most readily. To test the relationship between learning and regularizing, the second experiment took the best-learned category from the first experiment, and tested participant behavior after only one day. Although learning suffered, patterns of regularization were not significantly different. Similar previous work (Hudson Kam & Newport, 2009) showed that increased complexity led to increased regularization in adults. The third experiment in this dissertation tested whether increasing overall complexity would alter participant behavior. Here, all three grammatical categories were simultaneously variable. Compared to the first experiment, this made learning much more difficult, but did not change the pattern of regularization. This work concludes that individual patterns of learning and regularization work in tandem with social forces to produce the observed historical trends, and opens up many avenues for future research