Word-superiority effect as a function of semantic transparency of Chinese bimorphemic compound words

Abstract

The word-superiority effect (WSE) describes the superior recognition of word constituents in a word, as opposed to a non-word, context. In this study, the WSE was used as a diagnostic tool to examine the modulatory effect of word semantic transparency on the degree to which Chinese bimorphemic compound words are lexically represented as unitised wholes. Word semantic transparency refers to the degree to which word constituents are semantically transparent/opaque to whole-word meaning. A modified Reicher–Wheeler paradigm and a character identification task were employed. Compounds with at least one opaque constituent (fully and partially opaque) patterned together and displayed a larger WSE (more word-like) than fully transparent compounds. A space occurs naturally between constituents of Chinese compounds. Present results corroborated with recent results on English compound processing in which a space was artificially inserted between constituents of naturally concatenated compounds. Independently, the WSE was larger for high-frequency than low-frequency Chinese compounds.

Keywords:

• Chinese
• Compound words
• Lexical representation
• Semantic transparency
• Word frequency
• Word-superiority effect

Acknowledgements

This paper is based on L.W.M.'s Master's thesis, supervised by Dr Chua, Fook Kee. The research was supported by Grant R-107-000-013-112 from the National University of Singapore (NUS); L.W.M. was supported by the NUS Research Scholarship. The author gratefully acknowledges Dr Chua and anonymous reviewers for their constructive comments on earlier versions of this manuscript.

Notes

¹A well-known definition is that of Bloomfield's (1933) that the word is ‘a minimum free form’. However, this definition has often been felt to be too drastic and several other criteria have been proposed, one of which is that by Greenberg (1957) that a word should be able to accept insertions at the boundaries but not within them, but this is applicable only to languages that do not have infixes.

²There is not much controversy at the orthographic and phonological levels because representations at these levels basically map one-to-one with constituent morphemes (e.g., Zhou, Marslen-Wilson, Taft, & Shu, 1999).

³Hopper and Traugott (1993) used the term ‘lexicalization’ to refer to cases in which linguistic material develops into or is recruited to form lexical items. For the present discussion, we may consider the ‘forming of lexical items’ as the development of corresponding representations in the mental lexicon.

⁴Another way to conceptualise whole-word representations is that they are built upon constituent representations, overlapping to different degrees as a function of the degree of semantic overlap between the whole-word and its constituents (Zhou & Marslen-Wilson, 2000). A more unitised whole-word representation may be interpreted as having a lesser degree of overlap with constituent representations, and vice versa.

⁵Technically speaking, Chinese two-character ‘non-words’ are more like English ‘pseudowords’ than ‘non-words’. They consist of two pronounceable characters that each carries meaning. When proper matching is done, they can look quite similar to the canonical compounds. Character identification in the Chinese ‘non-word’ context may be influenced by one's expectations about the canonical compounds. However, using orthographically familiar pseudowords as controls with which to evaluate the word-superiority effect can actually serve to control for the influence of expectations, when the difference is taken between words and their controls (cf. Hildebrandt, Caplan, Sokol, & Torreano, 1995).

⁶In two cases, the pronunciation of a repeated character had tonal differences across its two occurrences: (1) (send out) – typically pronounced fā in isolation and in ([send out; bright] = invent), but fà in ([hair; point] = bristle with anger); and (2) (just at a time or place) – typically pronounced dāng in isolation and in ([just at a time; now] = now), but dàng(4) in ([convenient; just at] = a take-away meal).

⁷The effect of Occurrence of word items (first vs. second) was later evaluated, taking the set of word stimuli in its entirety, because the matching/controlling of critical item variables was done within the three-way full factorial design employed in constructing the stimuli.

⁸Wherever appropriate, equivalent analyses were conducted by items, without the Nationality factor.