A longitudinal study analyzed (a) which lower- and higher-level semantic components uniquely predicted listening text comprehension and (b) the nature of the relation (i.e., direct and indirect) between the predictors and listening text comprehension in preschoolers. One-hundred and fifty-two children participated in the present study (68 females; mean age = 4;10 years/months and 5;5 years/months at Time 1 and Time 2, respectively). Participants were administered measures to evaluate listening text comprehension (Time 1 and Time 2) and (a) expressive and receptive word knowledge (lower-level semantic components), and (b) inferential skills and the ability to use context (higher-level semantic components) (Time 1). Verbal short-term and working memory were also included as control variables. Data were analyzed using path analysis models. Results showed that both types of semantic components (a) accounted for unique variance in listening text comprehension at Time 2, with lower-level semantic components explaining a larger proportion of variance than higher-level semantic components; (b) were related to later listening text comprehension through direct and indirect relations. Memory resources did not significantly contribute to listening text comprehension. The results are discussed in light of their theoretical relevance and educational/practical implications.