Guessing the meaning of unknown words is an essential process in L2 comprehension. At the same time, guessing is considered a nuisance in L2 vocabulary tests. This raises the question of the nature and causes of guessing in L2 vocabulary tests. The traditional claims in language testing are that guessing is (1) a function of proficiency and (2) a function of the test taker's risk attitude. To investigate these claims in the context of standardized vocabulary testing, 135 advanced EFL learners participated in a computer-based yes/no vocabulary test combined with a translation task and a risk test. Stepwise regression analyses suggest that about 60% of the variation in yes/no test guesses are attributable to inappropriate or lack of semantic word knowledge. However, there was no systematic effect of risk attitude and guessing was largely independent of general lexical proficiency level. Methodological and inferential consequences for L2 vocabulary testing are discussed.