As a case study of non-linguists' perceptions of accent, this paper investigates how accurately and on what basis Japanese-accented English (JAE) is discernible from other L2 varieties of English in New Zealand (NZ). The paper sheds light on how a feature salient in speech is associated with the perceived sociolinguistic identity of speakers. An analysis of a test with local university students demonstrates that JAE had the highest identification rate among six L2 English accents and that the lack of contrast between English /l/ and /r/ was the most crucial to their identification of major segmental features previously noted as characteristic of JAE. Although the participants also frequently perceived the /l/-/r/ conflation in Cantonese- and Korean-accented speakers, they were more likely to assume that the accent was Japanese rather than any other L2 accent. The conflation served as an indexical cue of JAE in NZ society at the beginning of this century because of New Zealanders' extensive contact with Japanese visitors and familiarity with the Japanese language in secondary education, along with ESL reference books' frequent reference to this feature as peculiar to JAE. The study confirms the applicability of the concept of indexicality to L2 accent investigation.