Abstract The lexicon plays a fundamental role in reading, but little is known about how it influences reading efficiency. Thus, this study seeks to identify which lexical factors in a lexical decision task are relevant in a semantic decision test. A total of 33 university students were recruited to perform a lexical decision task and a semantic decision task. The results revealed differences between the three types of words in the lexical decision task for all measures, but only in the regressive saccades for the semantic decision task. Ambiguous sentences triggered fewer regressions than sentences related to objects. The only lexical measure found to predict efficiency was average time on regular words, which predicted 24% of the efficiency. We discuss the implications of the use of a lexical decision task and the use of the inverse efficiency score as a semantic measure, and we discuss how the lexicon can predict semantic comprehension.