This paper is concerned with how n-/l- (or r-) variation in loanwords among Cantonese speakers in Hong Kong takes place in different phonetic environments. Previous researches point out that the ongoing merger of n- into l- in a syllable initial position in Cantonese often has influences on the borrowing of words from other languages and the learning of foreign languages. For example, when they introduce English words into Cantonese, Cantonese speakers are likely to realise n- as l- like "notes" /nouts/ as [lok. si]. Our survey with 204 non-sense words written in Japanese hiragana, however, shows that there is no statistical significance between the percentage of errors of n- and l- (or r-) in any phonetic environment except when a nasal precedes r-. In that case, it is realised as n- significantly. The result may suggest that n-/l- (or r-) variation in loanwords among the Cantonese speakers occurs in a two-way manner, namely n- → l- and l- → n-, which is a new finding when compared with the previous researches.
