Transfer Of Homonymy And Polysemy With Special Reference To Dutch And English

An investigation of the interlanguage of Dutch learners of English revealed evidence of L1-L2 transfer and interference. 40 pairs of Dutch lexical items, in disambiguating context sentences, were rated on a 7-point scale of semantic similarity by 30 native speakers of Dutch. Next, the same pairs, in the same context sentences, were translated into English by a total of 90 Dutch students of English, subdivided into three proficiency levels. It was hypothesised that the similarity ratings could be used to predict translation performance: in the absence of knowledge, pairs that rated high on perceived similarity would be more likely to be translated by identical English tokens, whereas pairs with low perceived similarity were more likely to be translated by different English tokens. The results showed a strong correlation between perceived semantic similarity and observed translation errors, in the direction predicted by the hypothesis. There were also indications that the lower-proficiency informants were at a relative advantage where cognates were concerned, but that the more proficient informants were better equipped to deal with false friends.