The present experiment investigated the relation between guessing (cf. synthetic ability, Jeffers and Barley, 1971) and speechreading performance. Guessing was measured by two types of completion tests: One sentence-completion test (SCT), and one word-completion test (WCT). The results indicated that skilled guessing - in terms of SCT-performance - proved to be critical for longer sentences to be speechread, while skilled guessing - as measured by a WCT - proved to be critical for speechreading situations where a low level of contextual information was offered. The latter result was also obtained for one of the SCT-subtests. The results suggest that speechreading and guessing skill are related to each other, and that different types of guessing tests predict different aspects of the speechreading process.