Student uncertainty is an active research area within the computer tutoring community. Researchers have hypothesized that student uncertainty can signal to the tutor that there is an opportunity for constructive learning to occur, and also that experiencing uncertainty can motivate a student to engage in learning (e.g., VanLehn, Siler, and Murray (2003)). Moreover, correlational studies have shown a link between student uncertainty and learning (e.g., Craig et al. (2004)). In addition, researchers have argued that student uncertainty is associated with various perceptible linguistic cues, many of which are conveyed only through speech, including pausing, loudness, and intonation. A number of these researchers have used such cues to manually annotate student uncertainty (with good levels of inter-annotator reliability), and also to automatically detect student uncertainty (e.g., Litman and Forbes-Riley (2004a/b); Bhatt et al. (2004)). Underlying such research is the intuition that human tutors can readily use such cues to detect when a student is uncertain, and can then adapt their response to address this uncertainty, in a way that can have a positive impact on the learning process. Or as Fox (1993) observes, “[human] tutors use the timing of a student's response, and the way the response is delivered, in addition to what might be called the `literal content' of the response, as a source of diagnostic information”.