Constituent integration during the processing of compound words: Does it involve the use of relational structures?

Although previous research has suggested that the processing of compound words involves the integration of the constituents, not much is known about what integration entails. Three experiments suggest that integration draws on both linguistic and conceptual knowledge about the constituents and the compound word; ease of processing (as reflected by RT in a sense/nonsense judgment task) is affected by the lemma frequency of the whole compound, as well as by each constituent’s positional family frequency. In addition, the data demonstrate that a compound’s constituents are not just conjointly activated but are bound together in a particular way; responses to a compound (e.g., snowball) were faster when the compound was preceded by a compound using the same relational structure (e.g., snowfort—MADE OF) than when preceded by a compound using a different relational structure (e.g., snowshovel—FOR). This finding suggests that the conceptual representation of a compound word might be based on a relational structure.