The current study explored whether novel words can be related to pre-existing words in semantic memory via thematic or taxonomic relations. The meaning of the novel words was learned from descriptive actions or perceptual features. In a lexical decision task, the novel words served as primes for their corresponding concepts, taxonomically and thematically related targets. ERP results showed that the taxonomically but not the thematically related words elicited reduced N400s relative to the unrelated words. A control experiment revealed a priming effect between the corresponding concepts and the thematically related words, ruling out the possibility of the inappropriate selection of targets. These results suggest that while adult learners rapidly integrate the learned novel words into semantic network, the taxonomic information is more available than the thematic information for novel words learned from semantic features. (C) 2016 Elsevier Ltd. All rights reserved.