Phorum

The Berkeley Phonetics, Phonology and Psycholinguistics Forum ("Phorum") is a weekly talk and discussion series featuring presentations on all aspects of phonetics, phonology, and psycholinguistics. We meet on Fridays from 4(:10)-5pm (unless specified otherwise below), in Dwinelle 1229 (Zoom link shared upon request). Phorum is organized by Kai Schenck. My email is "kai_schenck" @berkeley.edu.

Schedules from previous semesters can be found here.

Spring 2026

February 13

Kai Schenck (UC Berkeley): Unsupervised learning at the limits of phonology: unbounded circumambient patterns

Formal language theory (FLT) is able to quantify the computational complexity of phonological processes by treating them as string-based input-output mappings that are accepted by some formal grammar. Although non-optional phonological mappings have long been known to be a subset of the regular input-output mappings (Heinz 2011a), the attested phonological processes which require the most complexity to model are known as unbounded circumambient (UCA) – requiring information from both sides of a undergoer that may, in principle, be unboundedly far away (Jardine 2016, McCollum et al. 2020, Meinhardt et al. 2024). However, the unattested Sour Grapes harmony pattern (Wilson 2003a, 2006) also falls into this class (Lamont 2019, Meinhardt et al. 2021, Yolyan 2025). Input-output mappings based in FLT thus are unable to cleanly separate attested from unattested processes – various explanations have been proposed to account for this discrepancy such as phonotactic complexity (Lamont 2019) or diachronic stability (Prickett 2025). While several studies investigate the learning of Sour Grapes compared to a simpler attested process (Finley 2008, Lin & Myers 2010, Lai 2022, Prickett 2022, 2025), no studies compare either to attested UCA processes.

I investigate whether Sour Grapes harmony is more difficult to learn compared to attested UCA processes by training Generative Adversarial Networks (GANs) on artificial language input. GANs have been used extensively in work on the learning of phonological alternations (Beguš 2020, 2021b, Barman et al. 2024) and
are general learners with no language-specific substantive bias (Beguš, Zhou, et al. 2023), making them ideal to test whether Sour Grapes harmony is inherently more difficult to learn despite its similar classification to attested processes in FLT. I train three GANs, each on an artificial language with the same input forms that had a different output harmony system: one language with a simpler bidirectional harmony system, one that resembles attested UCA processes, and one that implements Sour Grapes. I find that, although there are small indicators that the bidirectional system was learned more easily, there is no statistically significant difference in learning of all three systems when only tokens with active spreaders are examined. This is in line with some previous some results for humans (Lin & Myers 2010, Prickett 2022), suggesting both that the formal complexity of Sour Grapes is not responsible for its unattestedness and that a general learner such as a GAN learns both simpler harmony processes and UCA processes to approximately the same level regardless of their attestedness.

February 20

Noah Macey (UC Berkeley): What makes an accent 'foreign'? Preliminary evidence from self-supervised speech models.

This preliminary work investigates whether and how L1 and L2 accents differ systematically. As a motivating example, imagine that a linguist meets several people with accents she has never heard before. Without asking for any biographical information, would she be able to tell which speakers learned English as a first language? 

There are three possibilities for how the linguist would perform, corresponding to three hypotheses about L1 and L2 accents. First, if our linguist makes the L1 vs. L2 determination with near 100% accuracy, then we can say that L1 and L2 accents are categorically different. Second, if our linguist is no better than chance at distinguishing L1 vs. L2, then we can say that there is no acoustic difference between L1 and L2 accents (though obviously there are social differences). Equivalently, we could say that every L2 accent is a potential L1 accent---there is no way to distinguish between them without resorting to biographical information. Third, if our linguist makes a better-than-chance but worse-than-certain judgement about each speaker, then we can say that L1 and L2 accents cluster together in accent-space, with significant overlap, but that there is no hard-and-fast acoustic difference between L1 and L2 speech.

The research presented here contributes to this question by comparing speech-model representations of different accents. The data comprise standardized phrases from the George Mason University Speech Accent Archive (Weinberger, n.d.), representing 1,204 English speakers of various backgrounds. Following Chernyak et al. (2024), an algorithm from the HuBERT family (Hsu et al. 2021) is used to embed the accent recordings in an acoustic space. Once embedded, the recordings are interpretable as trajectories through a three-dimensional space. These trajectories are clustered to test if L1 and L2 accents group together, and a k-nearest neighbors classifier tests whether L1/L2 category is predictable. Finally, inter- and intra-accent variance is tested as well, in an attempt to replicate the findings of Chernyak et al. (2024). The clustering, classifier, and variance analyses fail to distinguish L1 and L2 accents, providing weak evidence against systematic acoustic differences between L1 and L2 speech.

February 27

Larry M. Hyman (UC Berkeley): A Gap in OCP Tonal Effects

In this presentation I have two goals. First, I want to present an unattested gap in how the Obligatory Contour Principle (OCP) potentially interacts with H(igh) Tone Spreading (HTS), adding to the inventory presented by Myers (1997). Second, I will present some of the facts of HTS in the Thɔnkɔ dialect of Limba, a Niger-Congo isolate spoken in Sierra Leone (Hyman & Kamara 2025a,b). The relation between the two is that the Limba facts, independently intriguing, seem to involve the gap that I have in mind [no spoiler here]. However, I will show that the facts are better treated without invoking the apparently unattested repair in question. I will begin by giving a brief refresher of how the OCP affects tone, especially in systems where H is a privative tone contrasting with Ø. I briefly ask whether the reason that the gap has not been attested has to do with the substance of tone or the formal relation (ranking) between the general OT constraints which apply to other aspects of phonology as well.

March 13

Lev Michael (UC Berkeley) and Samuel Piahuantze (Ashéninka community linguist): Stress and (non-)cyclicity in Pichis Ashéninka

In this talk we describe stress in the Pichis dialect of Ashéninka, based on a collaboration between the two authors, one an Ashéninka community linguist, and the other an academic linguist. Ashéninka is an Arawakan language of Peruvian Amazonia, and previous work on Ashéninka prosody has played an important role in the development of phonological theory (e.g., Itô 1989, Spring 1990, Black 1991, Prince and Smolensky 2004). We extend previous work on the Ashéninka stress system (D. Payne 1980, J. Payne 1990) by focusing on the interaction of cyclic and non-cyclic morphology in stress assignment, showing that non-cyclic morphology has a profound, but previously undocumented, impact on Ashéninka stress patterns. We find that to a significant degree, interactions between cyclic and non-cyclic morphology conform with the predictions of the Level Ordering Hypothesis and the Strict Cycle Condition (Kiparsky 1982), but that the language also exhibits some surprising phenomena, including primary stress shift out of the cyclic/Level 1 domain, and in some circumstances, apparent rebracketing (Pesetsky 1985), involving conversion of non-cyclic morphemes to cyclic ones.

March 20

Gašper Beguš (UC Berkeley): Orca vowels and consonants: convergent spectral structures across cetacean and human speech

The vocal communication system of orcas (Orcinus orca) has so far been analyzed primarily in terms of the fundamental frequency (F0) modulations, i.e. the frequency of their phonic lips vibration. The calls have been divided into clicks, pulsed calls, whistles and types thereof. By analyzing 61 hours of on-orca acoustic recordings and controlling for the effect of high-frequency components (HFC) and F0, we report structured formant patterns in orca vocalizations including diphthongal trajectories. Broadband spectrogram analysis reveals previously unreported formant patterns that appear independent of F0 and HFC and are hypothesized to result from air sac resonances. This study builds on the recent report of formant structure in vowel- and diphthong-like calls in another cetacean, sperm whales (Physeter macrocephalus). Using linguistic techniques, we further demonstrate that some calls are reminiscent of human consonant-vowel sequences, featuring bursts or abrupt decreases in amplitude. We also show that individual sparsely distributed clicks gradually transition into high frequency tonal calls, which aligns with analysis of sperm whale codas as vocalic pulses. The paper makes methodological contributions to the cetacean communication research by analyzing orca vocalizations with both narrowband and broadband spectrograms. The reported patterns are hypothesized to be actively controlled by whales and may carry communicative information. The spectral patterns shown in this study provide an added dimension to the orca communication system that merits further analysis and demonstrates convergent evolutions of similar phonological features in cetaceans (orca and sperm whale) and human communication systems.

April 3

Jinyoung Jo (Stanford): Can Large Language Models Predict English Phrasal Stress?

Abstract: We investigate whether the contextual predictability of a word derived from a large language model (LLM) can predict its perceived prominence, and how its contribution compares to more traditional linguistic predictors such as syntactic structure and context-invariant informativity. We examine their effects using word-level prominence ratings from presidential speeches. The results showed that contextual probability does predict prominence, with less predictable words perceived as more prominent, as expected. However, its effect is smaller than that of syntactically derived stress and a context-invariant informativity measure. We also report variability across annotators. Overall, the findings suggest that while LLM-based contextual predictability explains prominence to some extent, sentential stress remains primarily grounded in syntactic structure and relatively fixed lexical properties. We highlight the complementary roles of linguistic theory and language technologies in phonological research.

April 10

Zach Wellstood (UC Berkeley): The phonologization of tonal depression and the emergence of a fourth tone height in Cua (Kalahari Khoe)

Cua is a highly endangered Kalahari Khoe language of the Tshwa subgroup, spoken by the Cua people indigenous to the Eastern Kalahari region in and around Diphuduhudu, Botswana. In this talk, I present phonetic and morphophonological evidence for an analysis of Cua's tonal system in which tonal depression (by a typologically unnatural class of depressor consonants: voiced/aspirated obstruents, delayed aspirated clicks, and /h/) has been phonologized as the insertion of a /L/ tone, and has triggered 'anti-depressant raising' of a historical *L tone, leading to the genesis of a fourth  contrastive tone height. I argue that this process has resulted in a timing shift in the alignment of tones relative to their tone-bearing units, and account for this by associating tones to subsegments (Shih & Inkelas 2019) rather than morae. This research demonstrates (i) how an already-dense tonal system can undergo 'broad tonogenesis' (Hyslop 2022) from an original three-tone system into a synchronic four-tone system, (ii) how extra-complex tritonal contours can evolve from bitonal contours, and (iii) how consonant-tone interactions can affect not only pitch, but also timing and alignment.

April 24

Larry Lyu (UC Santa Cruz): 

May 1

Michelle Kamigaki-Baron (University of British Columbia):

Topics