This study explored the role of word knowledge dimensions in second language (L2) word-meaning inference. College-level L2 learners (N = 121) participated in this study and completed a series of word knowledge tests including vocabulary size, word associates, morpheme-form knowledge, morpheme-meaning knowledge, morpheme discrimination, and morpheme recognition. Two modules of lexical inferencing tasks (pseudoword inference and real-word inference) were also administered to the participants. Through multivariate path analysis, the findings demonstrated that word-knowledge dimensions made a collective contribution to L2 lexical inference after text comprehension ability was controlled for. More critically, the study found that word associates and morpheme-form knowledge had the strongest predicting power among all word-knowledge components. Theoretical justifications were provided to discuss the mechanism of word knowledge and lexical inference. In addition, applied implications were discussed to shed light on L2 vocabulary instruction and learning.