I study language recovery in post-stroke aphasia. We use functional MRI and behavioral testing to investigate well-known yet enigmatic features of aphasia. For example, we have examined semantic and phonologically cued picture naming using fMRI and eye-tracking to demonstrate that people with aphasia do indeed 'know what they want to say, even when they can't say it'. Currently, we are developing a test of transactional success in conversation in aphasia, which uses both monologic and conversational discourse to assess how it is that ‚Äúpeople with aphasia communicate better than they talk‚Äù. We are investigating use of natural language processing, including machine learning programs, to analyze aphasic and non-aphasic discourse. We are also investigating the efficacy of various machine learning programs to differentiate between brain activation during naming of action vs. object pictures.