English-Cantonese cod-mixing, whereby romanized Cantonese linguistic elements are mixed into otherwise English utterances, is prevalent in online chats among bilinguals in Hong Kong who select the English alphabet as the primary inputting method. While this linguistic phenomenon has been noted in a number of previous studies (Cheng 2002; Fung & Carter 2007a, 2007b; Ho, 2003; James, 2001; Lee 2002, 2007), little attention has been paid to code-switching between the English-Cantonese mixed code and other linguistic codes available to the chat participants (i.e. English and Cantonese), as well as its interactional significance. Moreover, studies that have argued for any pragmatic or social functions of code-mixing in Hong Kong tend to analyze the code-mixed utterance without considering its sequential context, the findings of which therefore lack an interactonal basis.

This study explores the interactional functions of code-switching and code-mixing in online chats in Hong Kong and examines what linguistic codes are interactionally relevant in code-switching, using the Gumperzian notion of contextualization cue (Gumperz, 1992) and a sequential approach (Auer, 1995) to analysis. Records of online chatting known as 'chat histories' were collected from 8 subjects, among which 29 'chat history' archives were selected, amounting to an approximately 150,000-word database of chat dialogues. Chat episodes were then analyzed with the sequential approach, and participants' metalinguistic comments provided in the follow-up interviews were used to supplement the sequential analysis.

Analysis of the chat data attests to three interactional functions of code-switching and code-mixing reproted in previous research works on spoken interactions, namely (1) discourse organizaton, (2) conveying affective and attitudinal meanings, and (3) negotiation of social roles and relationships. For discourse organization, it is observed that code-switching helps to separate different discursive units within a single turn, as well as contextualize different stages of the entire chat. Code-switching is also found to contextualize, together with other cues, the display of intense emotions, while the use of the mixed code can convey certain specific attitudinal meanings. Lastly, chat participants are shown to switch codes sometimes to negotiate their social roles or power in relation to other chat participants.

Based on the results of the analysis, this study proposes that the mixed code and Cantonese in its romanized form are contrasted with English in code-switching for interactional purposes, while there is little evidence to suggest a contrast between the mixed code and romanized Cantonese that bears interactional significance. The study also demonstrates the merits of using the sequential approach in examining the interactional aspects of code-swithching. Specifically, it has shown that the interactional meaning arising from code-switching is made locally relevant by the participants, and may depart from the social import conventially associated with a particular linguistic code.