Berent, Iris

Much evidence suggests that readers assemble phonology in reading, yet little is
known about the structure of these phonological representations. Linguistic
research suggests that speakers represent prosodic structure via skeletal frames and
that unmarked frames are preferred to marked frames. Seven experiments explore
the role of the skeleton in reading focusing on these three questions: (a) do readers
assemble the skeleton of printed words? (b) do readers prefer certain frames to
others? (c) are skeletal preferences due to grammatical markedness and/or to the
statistical properties of the language? Experiments I and 2 showed that in a forcedchoice
task, readers favor non-words with unmarked eve and cvcc frames (e.g. ,
GOM/ TUSP) to non-words with marked VCC frame (e.g., ELM), regardless of
segment similarity. Li!<ewise, in Experiments 3 and 4, readers were slower and less
accurate in responding to non-words with unmarked CVC and CVCC frames compared to non-words with marked vee frame in both phonological and standard
lexical decision tasks. However, regression analyses revealed that readers'
preferences were due to the statistical properties of the items. Experiments 5
through 7 further investigated skeletal preference in a Stroop-like paradigm. There
was no evidence for the representation of skeletal frames in Experiment 5, a result
attributed to strategic suppression of processing the printed stimulus. Indeed,
Experiment 6 revealed that readers were sensitive only to the skeleton of nonwords
presented in the first half of the experiment; readers named the color "red"
(eVe) faster in the presence of skeletal-congruent eve non-words (e.g., TUP)
compared to both vee and eeve incongruent-skeletal non-words (e.g., UPT/
TWUP). Likewise, in Experiment 7, skeletal-congruent eeve non-words (e.g.,
GROP) facilitated the naming of the color "black" (eeve) compared to
incongruent-unmarked eve non-words (e.g., GOP), but not to incongruent-marked
vee items (e.g., OSP). Again, only the statistical properties of the items
contributed to these results. This demonstrates that readers can distinguish between
non-words with different frames and they favor non-words with unmarked frames
to those with marked ones. Such preferences seem to reflect linguistic knowledge
of the statistical distribution of specific consonant-vowel combinations, rather than
grammatical preferences concerning abstract frames.

Linguistic theory predicts that speakers are equipped with universal sonority principles that restrict the structure of the syllable: onset clusters with sonority rises are preferred to onsets with plateaus, which are preferred to onsets with falling sonority (bn > bd > lb). We investigate the source of this preference in English, a language in which these sonority profiles are unattested. We gauged speakers' preferences for unattested onsets by examining their susceptibility to epenthetic repair. If English speakers are sensitive to onset structure, then onsets that are universally dispreferred should be more likely to elicit repair (e.g., lbif→lebif). Results from a syllable judgment task and an identity task support our predictions: onsets of rising sonority are perceived more accurately compared to onsets with sonority plateaus, which, in turn, are perceived more accurately compared to onsets with sonority falls. Our findings suggest that speakers are equipped with phonological preferences for sonority profiles that are unattested in their language.

Three experiments investigated the source of null phonemic masking effects with homophones under phonology discouraging conditions. Particularly, we examined whether the phonemic masking effect depends on the dominance of the homophones and list structure. For this end, we compared the phonemic masking effect for subordinate and dominant homophones presented in either a blocked or mixed condition. Our findings indicated that neither of these two factors systematically modulated the phonemic masking effect. However, there was some evidence for reliance on phonology in each of the conditions.