Multiword sequences are important components of language because they are building blocks that can be used to create long stretches of discourse. They are word combinations that have particular importance because of their co-occurrence and function in discourse that suggest that they are stored and retrieved from memory as a whole rather than as separate word units. The functions that they perform in discourse can vary according to register. In spoken academic discourse, one of the essential functions of multi-word sequences is a discourse organizing function that include introducing a topic and elaborating on an existing topic These varied discourse functions have two main roles in the information structure of discourse: as a major rhetorical organizer or a minor rhetorical organizer (Chaudron & Richards, 1986). However, studies that have examined the discourse-organizing role of spoken linguistic devices, including multi-word sequences, either have examined limited data or have analyzed them from written transcripts only, overlooking an important aspect of speech, i.e., prosody, that has an important communicative role.

This study focuses on one type of multi-word sequence, lexical bundles, which are frequently used recurrent word combinations that are identified computationally in a corpus to understand how their prosodic variations are linked to their discourse function(s). Lexical bundles in spoken academic discourse have been found to have a discourse-organizing function through analyzing spoken text from orthographic transcription. However, what remains to be explored is their prosodic features that have the potential to specify specific discourse-organizing functions more precisely. Therefore, this study focuses on understanding the relationship between the prosodic variation(s) and discourse function(s) of frequently occurring lexical bundles in a corpus.

This study used a corpus-driven framework to analyze the prosodic and discourse function variations of lexical bundles in a spoken academic corpus compiled from YALE open courses. The discourse function of the lexical bundles was analyzed using transcripts and audio files to find emerging patterns in their rhetorical function in information structure. In other words, lexical bundles were classified according to the relationship to preceding and following discourse, i.e., whether it introduced a new topic or expanded, contrasted, or emphasized specific details related to the main topic. Prosodic analysis involved examining pitch movement and prominence within the lexical bundle. Then, the emerging prosodic patterns and their corresponding discourse functions were cross-tabulated to understand the relationship between them.

Findings indicate that some lexical bundles had multiple prosodic variations related to discourse function variations while others had minimal prosodic variation related to one discourse function. The discourse functions were categorized as having a major rhetorical organization role (introducing the main topic for discussion, connecting topics, major contrast) or a minor rhetorical organization role (expanding on a topic through specific details, providing background information, exemplification, or rephrasing, contrasting ideas, emphasizing important information). The variation in discourse function and prosody of lexical bundles in university lectures may indicate that some lexical bundles are more formulaic than others.