2 - Tuning Language Processing Mechanisms to a Language’s Morphology without Decomposition: The Case of Semantic Transparency

Summary

Psycholinguists often use experimental tasks of word recognition as a window onto understanding how we process words. Here, we review results with the lexical decision task that show sensitivity to morphological structure in that word recognition task. We then highlight the limitations of assuming that evidence of morphological processing is best interpreted as evidence that lexical entries are decomposed into constituent morphemes. Further, when target words follow primes formed from the same stem and presented at brief durations so as to tap into early processing in the lexical decision task, we argue that finding no difference between semantically transparent and opaque pairs in individual priming experiments is not sufficient to conclude that early analysis proceeds without regard to a word’s semantic properties. We familiarize the reader with the intricacies of the priming methodology for the lexical decision task and the claim that target recognition benefits from structural priming based on repetition of a stem morpheme in prime and target in . Inwe then discuss how outcomes change with processing time for the prime and its implication for the claim that when processing time for the prime is curtailed, morphological processing is insensitive to semantics. We argue instead that morphological priming cannot be attributed solely to the letter sequence that constitutes the stem in part because stem repetition accounts downplay the role of differences and similarities of whole-word targets with words other than just its prime. Inwe provide evidence that challenges an account of early morphological processing based on the form but not the semantic consequences of shared morphology between prime and target. In , we summarize meta-analytic results with funnel plots to ascertain the reliability of early effects of semantic similarity among morphological relatives in lexical decision, thus refuting support for a decomposition account that is semantically blind and based on stem form. Finally, in , we touch upon the power of tuning form-with-semantics models across languages and writing systems that differ with respect to their morphological structure and neighborhood density measures by emphasizing patterning distributed across words rather than local decomposition into morphemes. As an alternative throughout, we align results with models in which analysis of wordform and meaning are interdependent, rather than two independent and sequential processes, thus discounting the privileged role reserved for the stem.