When inferring the meaning of unknown words in a second language, learners make use of a variety of cues including the cross-linguistic formal similarities of loanwords and cognates. However, because learners do not always recognize these cross-linguistic relationships, cognate strategy training has been recommended. The present study investigated the impact of cognate strategy training on Japanese speakers’ accuracy when guessing the meaning of unknown English words. The findings revealed that participants were significantly more accurate in responding to cognates after training than prior to it, while responses to noncognates were similar in both rounds. Moreover, participants were more accurate in identifying cognates with common grapheme–phoneme correspondences in English (e.g. fleece) after training than those that had less-common correspondences (e.g. gauge). Considering error rates, the benefit-cost ratio of applying the strategy was four correct guesses to one incorrect guess. Together with insights from a small-scale qualitative follow-up study, the results reveal that both phonological decoding of English words and conversion into katakana script in Japanese contribute to difficulties in noticing cognate relationships. The primary implication of the study is that cognate strategy training may contribute positively to language teaching pedagogy in Japan and other contexts.

cognate, cross-linguistic similarity, Japanese, lexical inferencing, loanword, word guessing