Lehtonen, M. 1 , Monahan, P. J. . 2 & Scharinger, M. J. . 3
1 Cognitive Brain Research Unit, Cognitive Science, Institute of Behavioural Sciences, University of Helsinki, Finland
2 Basque Center on Cognition, Brain and Language (BCBL), Donostia-San Sebastián, Spain
3 Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany
Whether words are stored as morphological constituents or as holistic whole-word representations has important implications for models of language processing. Under masked conditions with very short prime durations, both transparent and morphologically opaque primes facilitate the processing of monomorphemic targets (Rastle, et al. 2000). These results suggest the underlying decompositional mechanism operates on a purely formal basis (Rastle & Davis, 2008). In many reports, however, transparent primes produce greater facilitation than opaque primes, a difference that can be enhanced under certain experimental conditions (Feldman, et al. 2009). In the current experiment, we tested 33 native speakers of English in a masked-priming, lexical decision task. Included were prime-target pairs that were transparent (acidic-ACID), morphologically opaque (beaker-BEAK) or contained only orthographic overlap (spinach-SPIN). Our materials had broad statistical distributions for several lexical parameters (e.g., Baayen et al., 2007) allowing for a correlational analysis of the reaction time data against these lexical statistics. First, as predicted, transparent and opaque primes significantly facilitated the processing of the target word (transparent: +27 ms; opaque: +12 ms), while there was no difference between the orthographic pairs and controls (-6 ms). Second, for the transparent condition, we found a significant negative correlation between prime stem frequency and magnitude of priming, and in the opaque condition, a positive correlation between the prime orthographic neighborhood size and magnitude of priming. When collapsing across conditions, we found a reliable correlation between magnitude of priming and the prime log lemma frequency, as well as a correlation between priming and semantic relatedness of prime and target. Unexpectedly, we did not obtain reliable correlations between the root-suffix transitional probability and the size of the priming effect. These results are discussed within the context of current models of lexical access.