Activation of grammatical representations during the reading of Chinese compound words

Abstract

Certain models of the mental lexicon feature a lemma level that stores lexico-syntactic representations such as case, gender and number in inflection-rich languages. However, as Chinese lacks this type of inflections, it remains a question as to whether lexico-syntactic features are represented and activated in the mental lexicon of the Chinese language. As a result, the immediate objective of the present study was to examine if classifiers, a syntactical device common in Chinese, are activated during visual word recognition. Nevertheless, the scope of the study was limited to compound words firstly because of their prevalence, and secondly because they purportedly embody the semantic and also possibly the syntactic features of both internal constituents. This representational dualism would in turn provide valuable insights into the representation of compound words as to whether they are stored as a whole (full-listing), as decomposed parts (full-parsing) or both (dual-route). If a compound is represented as a whole, then it should assume a single classifier representation. Conversely, if a compound is represented as the concatenation of constituents, then they should be associated with separate classifier representations with the representation of the head being dominant. To test the above hypotheses, a grammaticality judgment task in which a classifier matched only the first constituent (related foil), a classifier matched the whole compound (target) and a classifier unrelated to both (control) was conducted on eighteen Cantonese-speaking participants. Consistent with the above hypotheses, latencies for target selection were significantly longer for related foils than the controls, implying activation of the classifier representation of the first constituent and its later competition with that of the whole compound (i.e. two classifier representations are activated for each compound) as predicted under a connectionist framework. A corollary of this interpretation is that compounds are decomposed prior to the activation of the lexico- syntactic features of the constituents. Overall, the above findings have huge implications for existing Chinese reading models in that these models should take into account the role of lexico-syntactic representations in the mental lexicon. Finally, the above findings were evaluated against two competing but unlikely accounts which involved erroneous parsing and the skipping of the second constituent.