This study investigated whether second language learners who began learning Spanish after the onset of puberty acquire the kind of subtle statistical sensitivity that native speakers have to the productivity of a gradient morphophonological pattern in Spanish. Native speakers have been shown to be sensitive to the likelihood of an unstressed diphthong in existing Spanish derivations with particular suffixes. Furthermore, they apply this sensitivity productively, as reflected in their judgments of novel derivations (Eddington 1996, 1998). This pattern is of interest because there is a general constraint against unstressed diphthongs in Spanish, but unstressed diphthongs cooccur to varying degrees with different suffixes in existing derivations. Examining how adult learners respond to this lexical pattern brings novel evidence to bear on similarities and differences between the role of input statistics in language acquisition across age groups. In addition to this primary focus, this project extends prior work in two ways. First, where prior research on this pattern has been conducted using written tasks, this study examines sensitivity to morphophonological probability in the auditory modality. Second, it brings more sensitive measures of online processing to bear on this issue by using the lexical decision paradigm, rather than a metalinguistic judgment task. This aligns the present investigation with a significant literature examining the effects of phonological and morphological structure on (monolingual) lexical processing (e.g. Baayen & Schreuder 2003, Frisch et al. 2000). This study thus compares native and adult learner processing of a subtle, probabilistic pattern in which a phonological alternation is dependent on paradigmatic relations in particular morphological contexts.

A set of neologisms was constructed combining existing Spanish stems containing alternating diphthongs with a set of derivational suffixes. Suffixes represented a continuum from a strong bias toward monophthongs to a strong bias for diphthongs, according to their frequency of cooccurrence with unstressed diphthongs in existing Spanish derivations. Half of the neologisms contained a diphthong, and the other half a monophthong. In native Spanish speakers, neologisms in which the stem allomorph matched the suffix bias were identified as real words more frequently in lexical decision, and a significant facilitation effect was observed in reaction times to the same stimuli. These results are discussed in terms of the dynamic evolution of lexical activation as the auditory stimuli unfolded over time. In the adult learners of Spanish a similar sensitivity to the gradient well-formedness of unstressed diphthongs was observed in the reaction time data, but not in the rate of false positives. Thus, the statistical distribution of unstressed diphthongs across morphological contexts in Spanish impinges on the timecourse of learners' processing of neologisms, but not on their behavior as reflected in errors in lexical decision. Crucially, these results provide compelling support for continued plasticity in adults' sensitivity to statistical patterns in the second language lexicon. The demonstrated sensitivity to subtle yet productive subregularities in morphophonology provides novel evidence for continuity in the role of input in both first language acquisition and later second language acquisition, and argues against strong versions of the critical period hypothesis.