Is morphological decomposition limited to low-frequency words?

Abstract

On the basis of data from masked priming experiments, it has been argued that an automatic process of decomposition is applied to all morphologically structured stimuli, irrespective of their lexical characteristics (Rastle, Davis, & New, 2004). So far, this claim has been tested only with respect to low-frequency primes and nonword primes. This is a limitation because some models of morphological processing postulate that only high-frequency complex words are recognized as whole forms. Thus, a more stringent test would be to determine whether high-frequency complex words also show evidence of masked priming. We report an experiment that compares masked-priming effects observed when the primes constitute morphologically structured nonwords (e.g., alarmer–ALARM), low-frequency words with a mean frequency of 2 per million (e.g., notional–NOTION), and high-frequency words with a mean frequency of 60 per million (e.g., national–NATION). These three conditions yielded significant and equivalent effects, lending strong support to the notion of a routine form of decomposition that is applied to all morphologically structured stimuli.

Keywords:

• Morphological processing
• Visual word recognition
• Dual-route models
• Reading
• Masked priming

This research was supported by Grant PTA-0302–0050–0002 from the Economic and Social Research Council (UK). We are grateful to Harald Baayen for valuable comments on an earlier draft of this manuscript.

Notes

¹ In contrast to our previous work (e.g., McCormick, Rastle, & Davis, 2008; Rastle et al., 2004) we did not include an orthographic form condition modelled on the morphological pairs (e.g., brothel–BROTH; electron–ELECT). Indeed, because of the small number of these stimuli available in English, it would have been impossible to match them to the morphological conditions on the critical prime frequency variable. However, we felt confident that our results could safely be attributed to morphological overlap for two reasons. First, the meta-analytic review conducted by Rastle and Davis (2008) demonstrated over 18 experiments that pairs like brothel–BROTH yield average priming effects close to zero. Second, a form priming account of our effects would predict strong inhibition in the high-frequency prime condition (Davis & Lupker, 2006), precisely the opposite to the predicted (and observed) pattern.

² One of our reviewers suggested that the presence of nonword primes in our experiment might have inflated the contribution of the decomposition route, resulting in larger priming effects in the high-frequency condition than would be observed in an experiment comprising high-frequency primes only. We believe that this is very unlikely given that (a) there is little evidence for the modulation of masked priming effects due to the nature of the primes (Brysbaert, 2001; Perea & Rosa, 2002); (b) the related nonword primes constituted only 12.5% of the word trials, thus discouraging rather than encouraging morphological decomposition over the whole set of items; and (c) the priming effect observed with low-frequency primes was exactly the same size as that reported by Rastle et al. (2004) in which no nonword primes were used.