Item

Abstract

Speakers learn language from their environment. Th is learning continues throughout speakers' lives as they continuously monitor the enviro nment and adjust their representations accordingly. In general, learning from the environment is better if (a) learners make over-all adjustments to permanent and representative changes in the environment but not to transient changes, and if (b) learners adjust more the poorer the ir fitness in the environment is. We test whether the learning of linguistic input follows these general tenets. Study 1 tests hypothesis (a) with a perceptual learni ng paradigm. Native French speakers listened to native (representative) or Dutch ac cented (unrepresentative) French with imperceptibly manipulated VOTs. Results show that list eners adapt to the VOTs of both native and non-native speakers, but only generalize their le arning to a new native speaker if they learned from the native speaker. This shows sensitivity in learning to the representativeness of the learning environment and its relevance to other l inguistic contexts. Studies 2-3 test hypothesis (b) by examining whether l ess fit speakers, as approximated by poorer performance on a task, learn the linguistic patterns in their environment more. Study 2 tests this by analyzing the grammatical structure used b y players in the TV game show Jeopardy. Players could use one of two grammatical structur es when requesting a clue - with and without a preposition ( Natural Wonders for/ !"## ). Analyses show that the more money players had won, the less likely they were to repeat the structure used by the previous player. Analyses further showed that the preference to repeat t he frequent (with a preposition) over the infrequent ( ) structure is reduced when winning more money, suggesting reduced reliance on linguistic frequency with better fitness. Study 3 prov ided further support from a game task in the lab, where small groups of French speakers played a modifi ed version of the game Go Fish. The modified version required participants to describe t he cards they were requesting. As with Jeopardy, one property of the cards, the flavor of the depicted ice-cream, could be described in one of two grammatical structures - with and without a proposition ( $%!&$%'(!%) * !'+,',$%- ! . Analysis of participants' utterances revealed that part icipants were more likely to repeat the previous player's grammatical choice the more cards their add ressee had taken from them. Studies 2-3 then show that speakers' adaptation to the grammatical structures used in the environment depends on how well they perform in the environment. This dependence occurs even though the linguistic choices are not directly linked to fitness in the environment, as success in neither Jeopardy nor Go Fish depends on grammati cal choices. Together, this set of studies shows that the degree to which individuals adjust their language according to encountered input varies and depend s on their own as well as the environment's fitness. This indicates that the weight gi ven to encountered input depends on the circumstances at which it is encountered. These results have co nsequences for language change as they suggest that innovations are more likely to be adopted during difficult times, when individuals perform more poorly, and by those of a more vulnerable position (studies 2-3) as well as that the learned innovations are more likely to be maintained when these difficult circumstances or poor position seem representative (study 1).