Feldman, L. B. 1, 2 , Martín, F. M. d. P. 3, 4 & O’Connor, P. A. 1, 2
1 Dept of Psychology. University at Albany. State University of New York. Albany, NY
2 Haskins Laboratories, New Haven, CT
3 CNRS. Lyon, France
4 Université de Lyon. Lyon, France
Morphemes are units of meaning and units of form yet in recent years it has been suggested that, at an early stage of processing, words composed of multiple morphemes are recognized in terms of constituent morphemes that are unspecified with respect to semantic properties. The argument for semantically blind morpheme-based decomposition rests on two claims (Davis & Rastle, 2010). First is the failure to detect effects of semantic transparency on morphological facilitation in many forward masked priming studies so that semantically transparent pairs like HUNTER-HUNT and semantically opaque or pseudo-derived pairs like CORNER-CORN show equal magnitudes of facilitation. Second is the failure to observe effects of primes whose morphological structure is partially decomposable (stem + illegal affix) under conditions where fully decomposable (stem + legal affix) primes typically facilitate. For example, pairs with partially decomposable primes like BROTHEL-BROTH fail to differ from unrelated controls whereas those with exhaustively decomposable primes like CORNER-CORN tend to facilitate. In the present study we focus on the decomposability claim. We pair the same targets with exhaustively decomposable (BROTHER-BROTH) and partially decomposable morphological structured primes (BROTHEL-BROTH) where neither prime-target type is composed of semantically similar words. Then we compare target decision latencies (accuracy) in the forward masked priming paradigm. Results fail to corroborate reliable facilitation based on decomposition and reveal, instead, inhibition that is stronger for partially than exhaustively decomposable primes. Evidently effects of decomposability can be graded and can lead to inhibition. Similarities and differences between semantically unrelated words that appear to share stems (e.g., BROTH) but to differ with respect to the potential for exhaustive decomposition (e.g., BROTHER, BROTHEL) will be discussed in terms of the purportedly all-or-none process of parallel morpho-orthographic segmentation of stem from affix.