Social interdependence is a key concept in CSCL research. However, investigations of students' positive and negative interdependence during collaborative activities have often relied on self-report, rather than dialogue analysis. Bringing together politeness and social interdependence theory, we assessed "dialogue indicators" of positive and negative interdependence from behavioral data (namely, face-saving and face-threatening dialogue moves) and compared the results with those of self-report scales. We analyzed a data set of 30 elementary students learning fractions with an intelligent tutoring system (ITS). Our initial analyses focus on the link between use of language that is face-saving (e.g., marking identity with statements such as "we are great") or face-threatening (e.g., insulting), and students' preferences to collaborate and compete. We found only non-significant correlations between these two broad categories, but found significant correlations between dialogue indicators, such as the use of identity markers and joking, that suggest directions for subsequent studies.