This paper explores the patterns and processes used by adult students of English as a second language (ESL) when they guess at the meanings of nonsense words in a reading text. The findings suggest that while word-unit processing is a major component of both first- and second-language reading, a major difference between first- and second-language readers is that the latter will not have confidence in their judgment that a word is familiar or unfamiliar, and will guess at its meaning rather than pass it by. In addition, any familiarity in a word may arrest the second-language reader's attention and detract from top-down, contextual processing. This suggests that ESL teachers should recognize the importance of word unit processing and encourage guessing when contextual clues are available; be sensitive to the reduced ability to guess among low proficiency students with limited linguistic knowledge; avoid exclusive use of over-redundant texts; acknowledge that word-analysis is a natural strategy even though sometimes misleading; and provide practice in both guessing from context and word-level graphophonemic accuracy, thus building both flexibility and efficient access to word meaning in memory. (MSE)