Markedness Effects in Visual Processing of Nonnative Onset Clusters

A variety of experimental methods have provided evidence for knowledge of constraints on wordforms, including sensitivity to markedness distinctions among different types of unattested sequences. For example, in acceptability judgment tasks with both auditory and visual stimuli the rating (or binary judgment) of a form generally decreases with markedness (auditory stimuli: Scholes, 1966; Coleman & Pierrehumbert, 1997; Frisch, Large & Pisoni, 2000; Albright, 2009 [based on Albright & Hayes, 2003]; Kager & Pater, 2012; Hayes & White, 2015; visual stimuli: Bailey & Hahn, 2001; Daland et al., 2011). Many of the relevant studies have focused on markedness relations among onset clusters, for example the graded distinctions of well-formedness due to sonority sequencing (e.g., sonority fall < sonority plateau < small sonority rise < large sonority rise). Unsurprisingly, novel words beginning with onset clusters that are attested in the native language of the participants (e.g., brif) receive higher ratings and endorsement rates than otherwise matched nonwords beginning with unattested clusters (e.g., rbif). More interesting is the fact that judgments distinguish among unattested clusters in a way that reflects degrees of markedness. For example, nonwords beginning with a smallrise onset (e.g., bnif) tend to be judged as superior to nonwords beginning with falling-sonority onsets (e.g., rbif), even by speakers of languages in which neither onset type occurs in lexical forms. Previous studies using auditory stimuli have also found sensitivity to markedness distinctions in online processing. More specifically, there is evidence from many perception and production tasks that nonnative clusters undergo the repair of perceptual epenthesis—the insertion of a reduced vowel that is not presented in the target form—in a way that reflects sonority-based and other markedness constraints (speech production: Davidson, 2006; Haunz, 2007; Wilson et al., 2014; speech perception: Dupoux et al, 1999; Berent et al., 2007; Maïonchi-Pino et al., 2014; Daland et al., submitted). For example, the results of syllable counting tasks essentially mirror those of acceptability judgment: monosyllabic nonwords beginning with more marked clusters are more likely to be identified as disyllabic, plausibly reflecting differences in the rate of perceptual epenthesis (e.g., epenthesis into a sonority plateau [bdif] ➝ [bәdif] applying more frequently than epenthesis into a small sonority rise [bnif] ➝ [bәnif]). While perceptual epenthesis is robustly attested for auditory stimuli, only a small number of studies have reported parallel effects for visually-presented forms (Berent, 2008; Berent & Lennertz, 2010). If the same or similar repair process applies regardless of the modality in which stimuli are presented, this would support a more abstract phonological locus of perceptual epenthesis. On the other hand, failure to find evidence of this repair in visual processing would be consistent with other results indicating that high rates of perceptual epenthesis depend on the presence of particular acoustic cues such as stop releases (e.g., Davidson 2010; Wilson et al., 2014; Lennertz & Berent, 2015; Zhao & Berent, 2015; Daland et al., submitted). In the present paper, we provide new evidence relevant to the question of whether marked onsets are repaired consistently across the auditory and visual modalities. Replicating the results of Berent & Lerentz (2010), we establish that some degree of sensitivity to sonority sequencing constraints can be