Poko (Skou, Papua New Guinea) displays a complex tone system, with three contrastive levels, toneless syllables and floating tones. Lexical tone is characterised by robust patterns of anti-alignment, wherein high (H) tones may not be initial and low (L) tones may not be final (McPherson & Dryer 2021; McPherson 2022). This article analyses the postlexical realisation of tone, especially the behaviour of floating tones, rising tones and toneless syllables. The Poko tone system shows unique twists on cross-linguistic patterns, such as the OCP, tone raising before L and the avoidance of non-final rising tones. We demonstrate that the behaviour of floating tones cannot be accounted for in a constraint-based theory with global evaluation, as in traditional Optimality Theory or Harmonic Grammar. Instead, the data motivate the use of directional Harmonic Serialism (Lamont 2022b), wherein changes are made incrementally and directionally, thus avoiding the creation of ties among otherwise similar candidates.

This article investigates phonological complexity by using artificial neural network and Bayesian structural equation models to derive representations of phonological complexity from counts of the segments associated with particular features in languages’ phonemic inventories. These latent representations can then be used alongside principal component analysis to further analyse how interactions between phonological features affect overall complexity, and what phonological complexity patterns a model can detect in a phonological feature data set. The results indicate that the per-feature segment counts investigated tend to contribute positively to a language’s complexity, and that the latent complexity variables approximate a log-normal distribution. This implies that phonological complexifications co-occur with other complexifications diachronically while tending to be more constrained at the upper and lower ends of the complexity range.

This article presents a case of superadditive ganging-up cumulativity in the metrical phonology of Bosnian/Croatian/Montenegrin/Serbian (BCMS). BCMS individually permits monomoraic feet and feet with a toneless head mora, but prohibits toneless monomoraic feet. Across BCMS dialects, several prosodic processes conspire against this doubly marked structure. Because of the superadditive character of this interaction, both Optimality Theory and, importantly, Harmonic Grammar require local constraint conjunction to capture the ban on toneless monomoraic feet in BCMS. This demonstration constitutes evidence for conjoined constraints in weighted constraint grammar. The study contributes to the typology of cumulativity effects by documenting superadditive ganging up in a categorical prosodic pattern, whereas virtually all previously reported cases of superadditivity have been observed in variable phonological patterns.

This article examines the default stress pattern of Sentani (Papuan), situates it within the typology of iambic stress patterns, and provides an analysis within the Weak Bracketing framework for metrical stress theory. Sentani is unique in sometimes employing clash (adjacent stressed syllables) and other times employing lapse (adjacent stressless syllables) in an effort to avoid final stress in even-parity forms. Clash is employed in four-syllable forms, but lapse is employed in longer even-parity forms. Key to the analysis are constraints insisting that both the initial foot and the final foot carry a stress. When the initial and final foot are adjacent (i.e., in four-syllable forms), insisting that both be stressed results in a clash. When the initial and final foot are not adjacent (i.e., in longer even-parity forms), a medial foot emerges without stress in order to avoid clash.

This article tests theories of verb stress in the Northwest Caucasian language Abkhaz using a new database corpus of 3,115 inflected forms of 445 verbs. I describe the creation of the corpus and show how it can be used to gain new insights into principles of Abkhaz stress assignment, which depend on complex interactions between phonology and polysynthetic verbal morphology. I implement a previous theory of Abkhaz stress assignment (Dybo 1977) in a computer program and use the corpus to assess empirically how well it accounts for stress patterns across the lexicon of eventive verbs in Abkhaz. I show how this empirical evaluation identifies both strengths and weaknesses of the theory, and use these to propose a revised theory of Abkhaz stress assignment. The revised theory ties with or outperforms the original in all verb categories, accounting for the stress alternations in 40 additional verbs, which comprise almost 10% of the corpus. This shows that corpora, combined with computational implementations of phonological theories, can be used to further our understanding of highly complex phonological data sets.

This article explores the role of token frequency and global prosodic length in conditioning optional paradigm uniformity in Japanese voiced velar nasalisation, with data from two wug-tests carried out with speakers of Tōhoku Japanese. Experiment 1 demonstrates that frequency-conditioning observed in corpus data is reproduced in existing and novel compounds and holds at the level of the speaker. Experiment 2 focuses on a typologically unusual pattern where overall compound length in mora seems to influence nasalisation, a candidate for a ‘counting’ pattern in phonology. We find that instead of overall length, speakers are sensitive to the length of the second member of the compound, undercutting the viability of the mora-counting analysis. We discuss the importance of the results in adjudicating between existing models of how token frequency impacts the phonological grammar and suggest that only theories that allow individual morphemes to exhibit frequency-sensitive behaviour are sufficiently expressive to model the finding.

The subregular hypothesis posits that all phonological markedness constraints can be described by some principled, learnable subclass of the regular languages. In this article, I classify a wide range of attested long-distance phonological markedness constraints, covering stress, harmony and tone, with focus on interactions of constraints over multiple tiers. To this end, connections are established between propositional logic over multiple tiers and algebraic properties of formal languages. These techniques allow for mechanical verification or refutation of membership in a class. Modelling the constraints and their induced patterns as formal languages, I demonstrate that the entire range lies within the propositional level, including Uyghur backness harmony and Karanga Shona tone, which have been presented as challenges to aspects of the subregular hypothesis.

We discuss three patterns of palatalisation in Czech, each of which is associated with certain suffixes. The data suggest that the trigger of palatalisation is not the initial vowel of these suffixes, but different sets of floating melodic features. We provide a formal analysis of the palatalisation patterns, as well as of the internal structure of the Czech phoneme inventory, in terms of Element Theory and strict CV. This allows us to straightforwardly model their lateral (leftward) effect, as well as some peculiar behaviour in the context of labials and the lateral . Besides Element Theory and strict CV, we argue that this analysis provides further support to a substance-free view of phonology, in which phonological representations do not necessarily have a universal, fixed phonetic implementation.

This article discusses cases of phonological abstractness and opacity and shows how they are eminently learnable, given certain assumptions about the innate cognitive endowment that learners bring to acquisition. I argue that opacity is not a learning problem but its solution. I propose that a distinction in patterning between two types of i in Inuit dialects is best explained by positing that surface [i] represents a merger of two underlying vowels. The apparently similar case of /i/ in Uyghur is shown to require a different type of solution. A review of contrasting approaches to prefix selection in Esimbi shows that opacity plays no role in evaluating their relative learnability. Though a celebrated case of opacity in Polish appears to have been misanalysed, an abstract analysis can still be motivated to account for alternations in certain lexemes. Uniting these cases is a preference for phonological analyses over diacritics or suppletion.

Recent quantitative work on the variable [g]~[ŋ] alternation in compounds of certain dialects of Japanese has revealed token frequency of the compound as a whole, and of the compound’s second member in its freestanding form, to be important predictors of the alternation. We propose a formal phonological analysis that integrates usage-based factors like frequency with the action of the phonological grammar, extending mechanisms of lexicon–grammar interaction previously proposed in the context of Lexical Conservatism. We demonstrate that our model fits the experimental data better than – or at least comparably to – a theoretically naïve statistical model proposed in previous work. Based on the success of our modelling, we discuss the role of token frequency in phonological patterning more broadly, and how the mechanism that we propose might be extended to unify a range of contradictory frequency-dependent processes that have been observed in the literature.

Some researchers have recently argued that sound-symbolic requirements can cause phonological alternations, suggesting that sound-symbolic patterns and phonological patterns may be governed by similar – or perhaps the same – mechanisms. Against this theoretical development, this article further addresses the question of how similar phonological systems and sound-symbolic systems are, by focusing on their counting capability. It has been established that phonological constraints can count only up to two segments. To examine whether a similar restriction holds in sound-symbolic patterns, we experimentally addressed the question of whether three segments of the same sort can cause stronger sound-symbolic images than two segments. The results of three experiments using Pokémon names demonstrate that three segments do indeed cause stronger sound-symbolic meanings than two segments. The overall results suggest that phonological systems and sound-symbolic systems differ in that only the latter have a certain type of counting capability.

Metaphony provides a unique window into the nature of the capacities of phonological grammars. The representational issues metaphonic processes highlight are exactly those core to computational analysis. Prominent examples of such processes, under various systems of phonological and morphological representation, are examined using algebraic methods, which describe the structure that arises when certain sequences exhibit the same behaviours. This analysis clarifies how representational choices interact with computational complexity. In particular, the simplest characterisations are obtained when the patterns are analysed at the syllabic level, as opposed to the segmental level, and when distinct morphemes are not demarcated by morpheme boundaries but by morphologically marked phonological material. Furthermore, these analyses speak directly to issues in learning, acquisition, production and perception.

This article investigates whether high vowel deletion (HVD) in Québec French (QF) is conditioned by prosodic structure. In Prosodic Phonology, prosodic constituents are typically assumed to be universal. However, the universality of certain constituents, such as the foot, has been questioned for languages like French, where prominence is obligatory only at the phrase level. We probe this issue by analysing HVD, a lenition process that is sensitive to prominence. We conducted two experiments. Experiment 1 examined the role of phrasal tone profile in HVD, while Experiment 2 examined whether word-internal constituency affects HVD. The results show that, while tone profile does not affect HVD, the process is preferred in dependent positions of iambic feet in even-parity forms. Odd-parity forms show no preference regarding HVD, suggesting that even- and odd-parity forms are parsed differently in QF.

This article focusses on the identity effect found in reduplication and how it can be accounted for in a formal theory. Specifically, this article demonstrates that the diminutive reduplication in Huozhou Chinese offers new empirical evidence supporting a correspondence-based analysis of reduplication. The reduplicative patterns can be understood as an effect of Agreement by Correspondence (ABC), thereby confirming a previously overlooked prediction of ABC. The case study demonstrates that ABC offers a more straightforward and reasonable analysis of Huozhou reduplication than Base–Reduplicant Correspondence Theory. Given the utility of ABC demonstrated in various other phenomena, including assimilation and dissimilation, the additional utility in reduplication establishes ABC as a more general and adaptable theory.

Metaphony and umlaut are well-studied assimilations, and metaphony, in particular, has attracted a range of different analyses. This article casts both in the framework of Element Theory, but as quite distinct processes. We then study the complex assimilation pattern of the Khoisan language ǃXóõ, between the first vowel of its bimoraic words and the second vowel, also affected by the consonants in the word. The most recent analysis, by Lionnet, argues that probabilistic and cumulative gradient effects are required to account for this process. We show that the process can be seen as formally a combination of umlaut and metaphony. Moreover, our analysis explains other assimilations and predicts a new assimilation, which is confirmed by recent data.

Discreteness of representations is a core commitment of classical generative phonology. However, one of the biggest arguments levelled against discreteness has come from the phenomenon of incomplete neutralisation – where the output of some phonological process on a feature [F] is phonetically distinct from underived [±F]. A sizeable body of work has contended that incomplete neutralisation crucially undermines the sufficiency of discrete representations, arguing instead for the inclusion of gradient representations. In this article, we investigate backness harmony in Uyghur, evaluating recent claims of incomplete neutralisation. Results from two production studies replicate findings from previous work on the language, but we show that, with proper experimental controls, the pattern is derivable without recourse to gradient phonological representations. We then discuss the conceptual and theoretical challenges for gradient representations along with criteria for true phonologically relevant gradience.

This article addresses the effect of phrasal rhythm on synthetic adjectival comparative forms in Russian that exhibit optional truncation of the final vowel, such as ${sil'n\acute{e}e}$ ˜ ${sil'n\acute{e}j}\ {slon\acute{a}}$ ‘stronger than an elephant’. In a corpus study of such comparative forms, using a generalised mixed-effects model to analyse these data, we show that phrasal rhythm affects the distribution of truncation. It is dispreferred in contexts like ${sil'n\acute{e}e\ z\acute{e}bry}$ ‘stronger than a zebra’, where it would yield two adjacent stressed syllables (${sil'n\acute{e}j\ z\acute{e}bry}$). This effect is categorical, and the study does not identify any gradient effect of distance on truncation. Stress placement in the comparative form itself also affects the distribution of -ee vs. -ej: comparative forms with stress on the suffix exhibit a lower proportion of truncation. These findings show that Russian exhibits rhythmic optimisation at the phrase level.