Compound Words, Processing of

Meaning

Early psycholinguistic research regarded the semantics of a compound as the decisive determinant of the processing route. Across languages, opaque compounds (those like hogwash whose combined constituent meanings do not yield the conventional interpretation of the whole) take longer to comprehend and produce and incur more identification errors than transparent compounds (those like doll-house whose meanings arise straightforwardly from the constituent meanings). These studies argued that opaque compounds are preferentially accessed in the mental lexicon as unanalyzed wholes via the storage route, while transparent compounds are accessed via computation.

Other semantic properties have also been shown to influence processing. For example, within a given [Page 164]compound, the semantic transparency of the compound's head (e.g., the similarity of the meanings of ball and snowball) contributes more to the ease of processing than the transparency of the compound's modifier (snow in snowball), though the strength of this effect varies cross-linguistically with the position of the head in the compound (e.g., final in English and Dutch, initial or final in Basque and French). On a finer-grained semantic level, the relation between the compound's head and modifier (snowball, relation: MADE OF) affects how compounds are produced and the ease of recognition and learning of novel compounds for both adults and children.

Nevertheless, the notion that semantics is predominant in compound processing is currently being challenged by a series of experiments that demonstrate that the meanings of constituents are activated via computation in both transparent and opaque compounds. Also, there is evidence that even low-frequency and clearly transparent compounds leave traces in the mental lexicon and activate the storage route. Studies using behavioral measures with fine temporal resolution confirm that semantic effects only emerge at late stages of lexical processing and that the storage versus computation biases are primarily sensitive to properties of compounds' form rather than meaning.

Form and Analogy

Recent cross-linguistic research has shown that compound processing implicates activation both of compounds as wholes and of their constituents. Linguistic properties of compounds and constituents (e.g., the frequency of occurrence or orthographic length) demonstrably affect acoustic characteristics of the production of compounds in speech, as well as the time it takes listeners or readers to recognize a compound during auditory or visual comprehension. Moreover, morphemes deeply embedded in larger constituents of compounds (-er in dishwasher) influence lexical processing, suggesting that morphological parsing reaches lower hierarchical levels of word structure. The order of accessing constituents in opaque as well as transparent compounds appears to be left-to-right in production and comprehension, as if the constituents were presented distinctly as regular words in a sentence.

...