Individuals with autism spectrum disorder (ASD) often struggle to use conversational and flexible speech, which complicates their efforts to navigate social relationships and contributes to peer alienation. Effective treatment for communication impairments is hindered by insufficient metrics to measure change. In fact, it is sometime hard to “put your finger on” what exactly differs in the language of a person on the spectrum, even though we easily identify that there is a difference. This study will identify linguistic markers of ASD in naturalistic language samples. Certain markers (e.g., conversational turn-taking) are hypothesized to correlate with real-world social functioning, while others (e.g., contraction use) are hypothesized to correlate with cognitive flexibility (which usually goes hand in hand with issues such as insisting on sameness). The tools developed in this project can be used to establish personalized profiles of linguistic strengths and weaknesses, enhancing our ability to design effective interventions for pragmatic language, and will directly benefit individuals and families by providing quantifiable treatment goals and dimensional measures of change. We hope to expand this line of research and collect conversational language samples remotely, including from toddlers, and use this information as part of a package of tools to more efficiently make diagnoses. Because this project uses “machine learning” and auditory recordings, it is easy to scale up to collect “big data” for research on causal mechanisms and for widespread clinical applications.