Effects of lexical status and morphological complexity in masked priming: An ERP study

Abstract

Two masked priming experiments examined behavioural and event-related potential responses to simplex target words (e.g., flex) preceded by briefly presented, masked, derived word primes (flexible-flex), complex nonword primes formed by an illegal combination of the target word and a real suffix (flexify-flex), and simplex nonword primes formed by adding a nonsuffix word ending to the target (flexint-flex). Subjects performed a lexical decision task. Behavioural results showed that all prime types significantly facilitated target recognition. Priming effects were reflected in the electrophysiological data by reduced N250 and N400 amplitudes, and these priming effects were statistically equivalent for the three types of prime. The strong priming effects found with simplex primes in the present study, compared with prior research, are thought to be due to the combination of targets always being completely embedded in prime stimuli plus the reduced lexical inhibition that arises with nonword primes. In line with prior behavioural research, however, there was evidence for differential priming effects as a function of prime type in the N400 ERP component in Experiment 2, with greater priming effects for derived and pseudocomplex primes relative to simplex primes at lateral posterior electrode sites.

Keywords:

• Evoked potentials
• Morphology (language)
• Priming
• Word recognition

Notes

¹Note that in psycholinguistic research, pseudocomplex words (e.g., corner) are typically put in the same category as semantically opaque morphologically complex words (e.g., department) as both lack a transparent semantic relationship between the stem and the whole-word.

²Another possible source, not shown in Figure 1, would be inhibitory connections between whole-word representations and semantics (i.e., between the semantic representation of “corner” and the whole-word representation of “corn”.

³The tentative mapping of ERP components onto component processes in word recognition, illustrated in Figure 2, would appear to stand in contradiction with reports of early influences of lexical (i.e., word frequency) and semantic variables on ERPs (e.g., Dell'Acqua, Pesciarelli, Jolicoeur, Eimer, & Peressotti, 2007; Hauk, Davis, Ford, Pulvermüller, & Marslen-Wilson, 2006; Hauk & Pulvermüller, 2004). There are two possible reasons for this apparent discrepancy. One is that the masked priming procedure, used in our studies, produces an overall slowing down of target word processing (a general interference from prime and masking stimuli against which prime relatedness effects emerge). Therefore, our timing estimates will be longer compared with studies using unprimed word recognition. The other possible reason is that major ERP components, such as the widely distributed N250 and N400, may reflect the bulk of processing at a given level of representation (or perhaps resonant activity across two adjacent levels), and not the fastest feedforward processes.

⁴Not significant when corrected for multiple comparisons.