An important challenge for psycholinguistics is to explain the cognitive processes that underlie speakers’ ability to quickly and accurately plan sentences, select words, and to say them out loud. Several influential theories of language production have been proposed that capture important aspects of this skill. But strikingly, they ignore an extremely salient aspect of everday speech: the fact that it is full of disfluencies. This project will test key mechanisms that we hypothesize to play a role in the production of disfluencies. Some of these concern the self-monitoring system, which inspects one’s speech for errors and other problems. Monitoring is assumed to intercept errors (even before they are articulated) but at the expense of introducing disfluency. Therefore, internal errors will lead to disfluency, but so would an excessive amount of attention to monitoring. Further hypothesized mechanisms involve the coordination between planning and execution of speech. If the speaker has executed (said out loud) one word, but hasn’t planned the next one yet, there will be a delay and a disfluency. Once we have established the contribution of these mechanisms, we will extend our existing computer model of word production and self-monitoring, and adapt it so as to produce disfluencies. The mechanisms discovered in the empirical work will inform the model, and we will attempt to simulate our own data, as well as other data in the literature.