CNL Lunch Talks

Robert Fiorentino

Morphological-level segmentation in complex word priming

Thursday October 6th, 2005, 12:30 PM, 3416 Marie Mount Hall

The existence and locus of decompositional processing in complex word recognition, as well as what might constrain decomposition, continues to generate controversy. A series of recent masked priming studies using affixed primes has attracted attention regarding this debate, with results seeming to suggest automatic segmentation into constituent morphemes regardless of semantic relationship, in an environment in which orthographic overlap by itself does not yield similar priming effects (e.g. Longtin et al. 2003, Rastle et al. 2004, among others). If these results are on the right track, they suggest an early and unconstrained morphological-level segmentation process underlying decompositional processing, which challenges explanation based on formal or semantic similarity.

However, serious questions remain about these results. A main concern regarding these studies, which have relied mainly on derivational affixation, is as follows: can the results be explained as epiphenominal, arising due to the salience of a high-frequency, closed-class suffix (e.g. Longtin et al. 2003)?  Further, to what extent are the effects free from constraints based on semantic transparency? (e.g. Diependaele et al. 2005)

Investigating masked priming using compound primes then becomes interesting, since compounds are complex words formed from open-class roots, allowing a direct test of whether the previously observed effects arise from the salience of a closed-class affix, as well as a test of whether any decompositional effects observed are mediated by semantic transparency.

               Thus, we conducted a set of studies to investigate masked priming from semantically transparent (e.g. teacup) and semantically opaque (e.g. bellhop) lexicalized compound primes, to their morphological constituent targets (tea, cup, and bell, hop, respectively). Our results show significant priming both for semantically transparent and for semantically opaque compound words, both in non-head (initial) position and in head (final) position.  We discuss these findings with respect to the masked priming literature mainly involving derivational affixation, as well as with respect to the literature on the processing of compound words in a variety of other paradigms. We consider the implications of these findings for competing models of morphological processing.