This study examines from a conversation-analytic perspective the prosodic and information-structure properties of a grammatical construction in Cantonese which has conventionally been referred to as ‘right-dislocation’ (RD) or ‘inverted sentence’ (Chao 1968, Lu 1980, Tai & Hu 1991).

Recent studies have shown that ‘right-dislocation’ (RD) serves as a formal device to mark focus domain in Cantonese (Parkard 1986; Law 2003; Luke 2004a, 2004b; Cheung 2005) (henceforth I will refer RD to as ‘focus-first constructions’ (FFCs)); whereas it is well-known that various aspects of information structure in conversation are signalled by means of prosody across languages (Lambrecht 1994, Ladd 1996). In an attempt to investigate how information structure is manifested in the position and ordering of constituents in the grammatical constructions of FFCs; as well as how prosody is employed by speakers to signal the non-focal information status of their further talk and the non-finality of their current turn-unit past the possible syntactic and pragmatic closure of the focal main clause of FFCs, ‘right-dislocated’ constituents (also ‘post-completion constituents’ (PCCs)) of FFCs are studied from spontaneous speech.

A total of 260 FFCs were identified from six recordings of two-party spontaneous Cantonese conversations between 19 and 35 minutes long, totalling 2.7 hours of talk. They were analyzed both auditorily by the analyst and acoustically by PRAAT, except that pitch and loudness were measured within 161 instances (62%) out of 260 instances, leaving out 99 instances (38%) that involve overlaps, laughter and noise.

Prosodic and information-structure analyses of conversational speech reveal that by and large participants orient themselves to a variety of prosodic parameters (including pitch, loudness, tempo and pause) in marking out PCCs as non-focal (including topical) elements; and as unprojected continuations of their focal matrix through absence of prosodic prominence.

The study finds that PCCs are normally produced without any pause (236/260 instances-91%), with lower mean pitch (147/161 instances-91%) and lower mean intensity (149/161 instances-93%) than that of their prior focal matrix. 137 out of 161 instances (85%) satisfy all of the above criteria. Regarding tempo, localized speeding-up effects are found either on final particles (PRT(s)) [47%-122/260 instances] or on the initial constituent of PCCs [41%-107/260 instances]. In utterances without PRT (19/260 instances), 16 out of 19 instances of the initial constituent of PCCs are produced at a faster rate than their immediately preceding constituents. These three varieties account for 94% [(122+ 107+16)/260] of the total instances. It is also found that PRT(s) are particularly prone to compression (21/22 instances) if their subsequent PCCs contain just one syllable that is prolonged by final/pre-boundary lengthening.

In general, the study shows that online utterance-end projection and various aspects of information structure in conversation are cued not only by syntax but also by prosody in Cantonese. The research also supports Couper-Kuhlen & Ono’s (2007) and Luke and Zhang’s (2007) typology of turn continuations.