Speech fluency has been extensively researched across a number of disciplines. Its multifaceted nature has led researchers to construe fluency as a core construct that closely interacts with other components of language ability such as pronunciation, lexico-grammatical complexity, and accuracy. In language testing research, speech fluency is often operationalized in terms of macro-level temporal features (e.g., speech rate, number of pauses). In contrast, micro-level disfluency features (e.g., pause position and pause repair) are under-studied, yet these features can provide evidence for the cognitive processes of speech production.

In this talk, I first review research findings on different dimensions of speech fluency. Then, using speech data from local and international language tests, I demonstrate how macro- and micro-level fluency features differ in their relations to (other components of) language proficiency. I argue that while macro-level fluency features are reasonable proxies of overall proficiency, fluency is not a simple collection of holistic temporal features. Instead, fluency is a representation of the cognitive processes during speech production, providing insights into the speakers’ automatic access to and control of various linguistic resources. I will also discuss implications of fluency research in the contexts of language learning, teaching, and assessment.