Research

Experimental session in Parets del Vallès, Catalonia

Collaborator: Gemma Repiso Puigdelliura

Catalan phonology exhibits a well-known pattern of final devoicing, but a subset of stems displays an unusual alternation known as saltation, in which a voiced fricative alternates with a voiceless affricate rather than the expected voiceless fricative (e.g., [bɔʒə] ~ [bɔt͡ʃ] and not *[bɔʃ]). Saltatory alternations have been of considerable theoretical interest because they are underivable in classical Optimality Theory and have been argued to be difficult or unstable for learners.

While previous work has shown that both adults and infants with short-term laboratory exposure tend to repair saltation by overgeneralizing simpler phonological processes, it has remained unknown whether infants learning a language that actually contains saltation can acquire and generalize this pattern productively. This project addresses that gap by investigating whether Catalan-learning infants extend saltatory alternations to novel words.

In Fall 2025, I travelled to Catalonia to investigate the learnability of saltation among Catalan speaking infants. Through two central-fixal experiments with 24-month-old Catalan-learning infants, I tested whether infants generalize saltatory alternations to novel words or instead repair them via final devoicing. Results show that infants productively extend saltation to new forms and do not apply phonotactic repair, demonstrating that saltation is learnable in early phonological development.

The manuscript is still in progress. However, the project was featured in two local news outlets which you can access here and here.

Regions where Catalan is spoken

Collaborators: Victoria Mateu and Bruce Hayes

Catalan phonology includes numerous consonant alternations found at the right edge of stems, such as [san-ə] ~ [sa] ‘healthy-fem./masc’. These alternations have been subjected to extensive analysis and have been frequently used as the basis of problem sets and computational studies. They bear on several important research questions including exceptionality, opacity, and saltation. However, no study has yet investigated the extent to which speakers actually generalize the patterns attested in the lexicon. Our study addresses this gap with two "wug test" experiments, one using elicited production, the other using acceptability judgments.

This work was presented as a talk at the Annual Meeting on Phonology (AMP) in November 2024 at Rutgers University in New Brunswick, New Jersey.

This manuscript has recently been submitted for publication. You can read the first submitted draft here.

A baby participating in an experiment in the UCLA Language Lab

Collaborator: Megha Sundara

By 6-months, English-learning infants are able to relate novel words suffixed with -s like babs and teeps that are embedded in passages, with just the stem bab and teep, demonstrating an early sensitivity to morphological relatedness. At the same time, infants are also becoming sensitive to phonotactics. We found that 6-month-old English-learning infants relate nonce forms suffixed with the [-z] allomorph but not the [-s] allomorph of the English -s suffix. Additionally, we found that 6-month-olds only related nonce forms suffixed with [z] when the resulting stem would be phonotactically legal but not when it would be phonotactically illegal. These results show that morpheme decomposition is not obligatory at 6-months; instead, infants’ ability to decompose potentially suffixed words is constrained by their developing knowledge of phonotactic restrictions in English.

This work was presented as a talk at the Conference on Laboratory Phonology (LabPhon) in June 2024 at Hanyang University in Seoul, South Korea, the International Congress of Infant Studies (ICIS) in July 2024 in Glasgow, UK, and the Boston University Conference on Language Development (BUCLD) in November 2024 at Boson University in Boston, Massachusetts.

The first experiment in this project, demonstrating that 6-month-olds are sensitive to the [-z] but not [-s] allomorph, has recently been published in Child Language. You can read the paper here. You can also read my MA thesis here which was based on an earlier version of this work and discusses some of the other experiments in this project.

A participation certificate from the UCLA Language Lab

Collaborators: Ekaterina (Katya) Khlystova and Megha Sundara

Infants are exposed to substantial phonetic variability in speech and must determine which variants should be encoded in their lexical representations. Word-final /t/ in American English provides a particularly informative case, as it is frequently realized as a glottal stop ([ʔ]) in infant-directed speech.

Across two Headturn Preference Procedure experiments with 14-month-old English-learning infants, we tested whether infants encode the canonical [t], the frequent [ʔ] variant, or an underspecified coda. Infants recognized familiar /t/-final words produced with both [ʔ] and canonical [t], but not minimal coda mispronunciations. These results show that early lexical representations tolerate phonetic variability while retaining sufficient phonological detail to support word recognition.

The manuscript for this project is still in progress.

Child-directed speech from the Manchester Corpus

Collaborator: Charles Yang

An important task faced by children during language acquisition is the formation of syntactic categories from individual words. Distributional cues such as frequent frames have been proposed but their utility under realistic learning situations remain unclear. Since frequent frames are lexical, they result in a proliferation of word classes that do not correspond to recognized syntactic categories. Thus, we propose the Category Frame Learner (CFL) model which uses the Tolerance Principle, a mathematically precise method of language learning, to combine lexical frames into category frames defined over syntactic categories.

This work was presented at a poster session of the Boston University Conference on Language Development (BUCLD) in November 2021 and as a talk at the Linguistics Society of America Annual Meeting in January 2022. You can view the BUCLD video overview here and the poster here.

This project appeared in the proceedings of the BUCLD and can be accessed here.