Authors:

FORD, M., MARSLEN-WILSON, W.D., & DAVIS, M.H

Reference:

Year of publication:

2003

CBU number:

5438

Abstract:

Effects of root morpheme and whole-word frequency in lexical decsision have been used to constrain theories of the representation and processing of morphologically complex words. However, results from these studies are unclear and somewhat contradictory, which may in part be due to an over-reliance on factorial designs. We present data from 2 experiments that show the utility of using correlational designs in addition to factorial designs. Experiment 1 used a regression design to explore which of the many counts of word and morpheme frequency best predicts lexical decision response times to monomorphemic words. Results showed that word form frequency was the best predictor, though significant effects of morphological family size, cumulative morpheme frequency and the semantic coherence of the morphological family were also observed. In contrast to previous, factorially designed experiments, no clear effects of lemma frequency were found. Experiment 2 used both factorial and regression designs to probe directly fro lemma frequency effects, using comparative (e.g. stronger) and superlative (e.g. strongest adjectives. The use of the comparative also tested for affix homonymy effects, since er is also an agentive suffix. No effects of affix homonymy were found, but clear lemma frequency effects were found in the regressions. Effects of more than one frequency count on response latencies are interpreted as reflecting the multiple sources of information that are used in making lexical decisions.