Processing unfamiliar words : a study in the perception and production of native and foreign placenames

• Main Author: Fitt, S. E.
• Published: University of Edinburgh 1997
• Subjects: 410
• Online Access: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.650941

This thesis sets out to examine some of the linguistic processes which take place when speakers are faced with unfamiliar and potentially foreign place names, and the possible psycholinguistic origins of these processes. It is concluded that lexical networks are used to map from input to output, and that phonological rule-based models do not fully account for the data. Previous studies of nativisation have tended to catalogue the phonological and spelling changes which have taken place in historical examples, and explanations have generally been limited to comparison of details of the borrowed and borrowing languages, rather than being set in a solid linguistic framework describing the ways in which speakers and readers process words. There have been psycholinguistic studies of unfamiliar words, but these have generally ignored the foreign dimension, and have been limited in scope. Traditional linguistic work, meanwhile, focuses on descriptions, either abstract or more related to mental processes, of the language that we know and use every day. Studies of foreign language also have a rather different focus from the current work, as they examine what happens when we attempt, over a period of time, to acquire new sounds, vocabulary and grammar. This study takes an experimental approach to nativisation, presenting Edinburgh secondary school pupils with a series of unfamiliar spoken and written European town names, and asking them to reproduce the names either in writing or speech, along with a judgement of origin. The resulting pronunciations and spellings are examined for accuracy, errors and changes, both in perception and production. Different explanations of the output are considered, and it is concluded that models which apply a set of linguistic rules to the input in order to generate an output cannot account for the variety of data produced. Lexicon-based models, on the other hand, using activation of known words or word-sets, and analogy with word-parts, are more able to explain both the details of individual responses and the variety of responses across subjects.