We use language technology to develop corpus measures of lexical and acoustic/prosodic convergence. We show that these measures successfully discriminate randomized from naturally ordered data, and demonstrate both lexical and acoustic/prosodic convergence in our corpus of human/human tutoring dialogs.