Profiling figure legends in scientific research articles: A corpus-driven approach

Abstract Figure legends, a descriptive statement accompanying a graph, is an essential component of science writing, and novice writers often find it difficult to write an adequate and plausible one. However, almost no discourse analysis has been done of this short yet important part of research articles. Taking a corpus-driven approach, this study explores the use of keywords and lexical bundles in figure legends in order to understand how the discourse is constructed by the salient recurrent single and multiple lexical items. Results show that in figure legends writers are most concerned with the textual descriptions of blue and red as markers of contrasting scientific items and bars and images as representative of graphic shapes. Additionally, writers also frequently use shown and indicated to introduce the demonstrative relationship between graphic elements and scientific information. Further, noun/prepositional bundles account for the overwhelming proportion, followed by verb-related and then clause-related forms, while research-oriented functions of bundles are more frequently used than text and participant-oriented functions. The results offer useful pedagogical input and help novice and L2 writers come to grips with this part-genre.