27th October 2025 12:20 PM

PhonDAWG - Phonetics Lab Data Analysis Working Group

Sam will present a preliminary analysis of his ambisyllabicity experiment.

28th October 2025 04:30 PM

ASL Linguistics Lecture Series: Dr. Naomi Caselli

The Cornell Linguistics Department proudly presents Dr. Naomi Caselli, Associate Professor, Boston University.   Dr. Caselli will lecture on:  "ASL Vocabulary Acquisition!"

ASL/English Interpretation provided, light refreshments to follow. 

Abstract:

This talk addresses four central questions in research on deaf children’s language acquisition:

1. Can children with hearing parents develop age-expected sign-language vocabularies?

2. How do parent sign language skills relate to child vocabulary outcomes?

3. Does learning sign language interfere with spoken language development?

4. Does early access to sign language support long-term academic achievement?

These questions are at the center of longstanding debates in deaf education, with significant implications for families, educators, and policymakers.

Drawing on recent population-level studies and emerging tools such as the ASL CDI 2.0, this presentation situates these issues within the broader evidence base on language deprivation, access, and development.

Speaker Bio:

Dr. Naomi Caselli is an associate professor of Deaf education, director of the Deaf Center, and the director of the AI and Education initiative at Boston University.

She is hearing, and her first languages are American Sign Language (ASL) and English. She leads a research team that works to make research on language—across education, computer science, linguistics, psychology, and medicine—inclusive of sign languages, and to ensure all deaf children have access to language.

Her research is centered on three questions:

• How does early language experience shape how deaf children learn language?
• How is the sign language lexicon structured, learned, and processed?
• How can we responsibly use AI to make the world more accessible to sign language users?

She is internationally recognized for creating open-access sign language databases that support research and education.

29th October 2025 12:20 PM

Phonetics Lab Meeting

Annabelle will lead a discussion on this paper on Whispered speech.  

"Auditory speaker discrimination by forensic phoneticians and naive listeners in voiced and whispered speech" October 2015, International Journal of Speech Language and the Law 22(2):229-248, DOI:10.1558/ijsll.v22i2.23101

Abstract:

In whispered speech some important cues to a speaker’s identity (e.g. fundamental frequency, intonation) are inevitably absent.

In the present study we investigated listeners’ ability to discriminate between speakers in short utterances in voiced and whispered speech.

The performances of a group of 11 forensic phoneticians and a group of 22 naive listeners were compared in a binary forced-choice speaker discrimination task, with 48 same-speaker and 60 different-speaker pairs of short speech samples (≤ 3 s) in each test. Listeners were asked to say whether the two voice samples in each pair were produced by the same or different speakers, and to give a certainty rating.

The results showed that speaker discrimination is more difficult in whispered than in voiced speech, and that while the phoneticians were only slightly better than the naive listeners in voiced speech, the gap widened in whispered speech.

Phoneticians were more cautious in their responses, but also more accurate than naive listeners. When unsure, the phoneticians tended to say two utterances came from different speakers, whereas naive listeners tended to say two utterances came from the same speaker.

Results support the view that trained phoneticians may have an advantage over naive listeners in auditory speaker discrimination when the signal is degraded.

30th October 2025 04:30 PM

Colloquium Talk Series - Marten van Schijndel

The Cornell Linguistics Department is proud to present Dr. Marten van Schijndel, Associate Professor of Linguistics at Cornell University, who will give a talk titled:  "Largely Language Modeling".

Abstract:

Language modeling is an old task which has gained popularity as the backbone of the current AI boom. This is, in part, because the task is simple to implement at scale, and in part because the method has achieved good success at a number of language tasks.

However, the learned representations of neural language models are opaque, making it difficult to assess what they encode. Many current discussions about language models center around their emergent properties; their apparent ability to transcend the language modeling task and instead achieve human-like language understanding.

While language models work well for natural language processing tasks, I argue that language model representations differ in key ways from human linguistic representations.

I discuss multiple studies from my group that identify weaknesses in these models and in the ways they are used which produce mismatches with human linguistic response patterns.

While I provide methods for mitigating some of these weaknesses, caution should be employed when using these models since they are largely language models.