This thematic issue almost did not happen. One of us (JH) was almost killed two days after the deadline for article submissions. As a pedestrian on a sidewalk minding his own business, he was struck by a car that ran a red light and lost control after a collision. So when we write that we are delighted to be writing this introduction, over one year later, we both really mean it.

This paper investigates the phonetic specification of contour tones through a case study of the Mandarin rising tone. The patterns of variation in the realisation of the rising tone as a function of speech rate indicate that its specifications include targets pertaining to both the pitch movement and its end points: the slope of the F0 rise, the magnitude of the rise, and the alignment of the onset and offset of the rise. This analysis implies that the rising tone is overspecified, in that any one of the target properties can be derived from the other three (e.g. slope is predictable from the magnitude and timing of the rise). As a result, the targets conflict, and cannot all be realised. The conflict between tone targets is resolved by a compromise between them, a pattern that is analysed quantitatively by formulating the targets as weighted, violable constraints.

This paper accounts for stress-pattern diversity in languages such as English, where words that are otherwise equivalent in terms of phonotactic structure and morphosyntactic category can take both initial and final stress, as seen in ˈlentil – hoˈtel, ˈenvoy – deˈgree, ˈresearchN – reˈsearchN and ˈaccessV – acˈcessV. Addressing the problem in general and abstract terms, we identify systematic conditions under which stress-pattern diversity becomes stable. We hypothesise that words adopt stress patterns that produce, on average, the best possible phrase-level rhythm. We model this hypothesis in evolutionary game theory, predict that stress-pattern diversity among polysyllabic word forms depends on the frequency of monosyllables and demonstrate how that prediction is met both in Present-Day English and in its history.

Following the Articulatory Model of Handshape (Keane 2014), which mathematically defines handshapes on the basis of joint angles, we propose two methods for calculating phonetic similarity: a contour difference method, which assesses the amount of change between handshapes within a fingerspelled word, and a positional similarity method, which compares similarity between pairs of letters in the same position across two fingerspelled words. Both methods are validated with psycholinguistic evidence based on similarity ratings by deaf signers. The results indicate that the positional similarity method more reliably predicts native signer intuition judgements about handshape similarity. This new similarity metric fills a gap in the literature (the lack of a theory-driven similarity metric) that has been empty since effectively the beginning of sign-language linguistics.

This paper examines the role of weight in stress assignment in the Portuguese lexicon, and proposes a probabilistic approach to stress. I show that weight effects are gradient, and weaken monotonically as we move away from the right edge of the word. Such effects depend on the position of a syllable in the word, as well as on the number of segments the syllable contains. The probabilistic model proposed in this paper is based on a single predictor, namely weight, and yields more accurate results than a categorical analysis, where weight is treated as binary. Finally, I discuss implications for the grammar of Portuguese.

This paper reports on a pattern of voicelessness dissimilation in the Kordofanian language Moro. Voiceless stops and affricates become voiced before a voiceless obstruent in a transvocalic configuration. The dissimilation is robust, and productive across morphological contexts. Phonetically, voicing in Moro is realised as a difference between prevoicing and short-lag voice onset time. This makes [voice] the most realistic featural characterisation; using another feature like [spread glottis] in lieu of [–voice] doesn't explain the contrast. Consequently, dissimilation of voicelessness in Moro is strong evidence that [voice] is a binary feature, and that […voice] may be phonologically active despite being ‘unmarked’. We show that when [–voice] is admitted, the Moro pattern is straightforwardly analysed on a par with other cases of dissimilation. Our analysis uses the theory of surface correspondence, which carries no crucial assumptions about markedness; other theories of dissimilation are considered in an online supplement.

This paper examines a key difference between constraint conjunction and constraint weight additivity, arguing that the two do not have the same empirical coverage. In particular, constraint conjunction in weighted probabilistic grammar allows for superadditive constraint interaction, where the effect of violating two constraints goes beyond the additive combination of the two constraints’ weights alone. A case study from parasitic tone harmony in Dioula d'Odienné demonstrates superadditive local and long-distance segmental feature similarities that increase the likelihood of tone harmony. Superadditivity in Dioula d'Odienné is formally captured in Maximum Entropy Harmonic Grammar by weighted constraint conjunction. Counter to previous approaches that supplant constraint conjunction with weight additivity in Harmonic Grammar, information-theoretic model comparison reveals that weighted constraint conjunction improves the grammar's explanatory power when modelling quantitative natural language patterns.

The apparent loss of initial obstruents in Basque borrowings from Romance (e.g. laru ≪ Lat. claru) is striking. While Proto-Basque is generally reconstructed as lacking initial clusters, the expected repair in loans, based on typology, phonology and phonetics, is copy-vowel epenthesis, not obstruent loss. Indeed, there is evidence for a vowel-copy process in Basque in other loans with obstruent–sonorant clusters (e.g. gurutze ≪ Lat. cruce). We suggest that initial obstruent loss before /l/ but not /r/ is related to Romance developments. In the Romance varieties in contact with Basque, /fl pl bl kl gl/ all show evidence of neutralisation to /ʎ/ word-initially. We hypothesise that obstruent loss in words like Basque laru reflects influence from local Romance languages at a time when Basque lacked /ʎ/. In contrast, vowel copy conforming to Basque syllable structure was the norm in Romance loanwords with clusters not affected by this process.

Templatic morphology involves the appearance of a fixed shape on a morpheme in a specific morphological context. This paper makes two claims: the morphological context of a template is syntactically cyclic, resulting in recursive prosodic word structure, and the shape of a template results from prosodic well-formedness conditions on the internal prosodic word. Templatic morphology in Chukchansi Yokuts illustrates these claims: affixes that trigger templates transfer the root to the phonology before other material is transferred, so that the root forms a prosodic word which is internal to the whole word. Roots with one underlying vowel are augmented to meet a disyllabic minimality requirement on prosodic words; the resulting disyllable forms a light–heavy iamb, to optimally satisfy Chukchansi parsing requirements. Templatic morphology falls out from the predictable interaction of the syntax–phonology interface and general phonological properties of a language, and needs no special apparatus or diacritics.

‘Phonologisation’ is a process whereby a phonetic phenomenon enters the phonological grammar and becomes conceptualised as the result of categorical manipulation of phonological symbols. I analyse the phonologisation of a predictable phonological pattern in Welsh, with particular attention to identifying criteria for whether phonologisation has occurred. I argue for a model where phonologisation experiences bottom-up and top-down biases. From the bottom up, there is pressure to phonologise phenomena with a categorical distribution; from the top down, there exist formal constraints on featural specification. I focus on the requirement for featural specifications to obey the Contrastivist Hypothesis, which denies that redundant features can be involved in phonological computation, in the context of a framework with emergent features. I suggest that the Contrastivist Hypothesis provides a useful boundary condition for emergent-feature theories, whilst independent phonologisation criteria provide contrastivist approaches with a more solid conceptual underpinning.

This paper introduces subfeatural representations to capture subphonemic distinctions at work in ‘subphonemic teamwork’. The unusual case of Laal is presented, in which rounding harmony requires two triggers: a round vowel and a labial consonant. The coarticulatory effect of the labial consonant is shown, on the basis of instrumental evidence, to incur a distinctive, but non-contrastive, intermediate level of rounding on the target vowel, analysed as being featurally [−round], but subfeaturally 〚xround〛 (0 < x < 1). 〚xround〛 vowels are shown to form a separate natural class, which can be independently targeted by phonological processes. Subfeatural representations are argued to constitute an advantageous reification of phonetic knowledge, forming a more solid basis for phonetically driven models of phonology. This proposal builds on the insights of previous literature that perceptual representations are needed in phonology, while eschewing the need for direct reference to phonetics.

Behavioural findings indicate that English, Mandarin and Korean speakers exhibit gradient sonority sequencing preferences among unattested initial clusters. While some have argued these results support an innate principle, recent modelling studies have questioned this conclusion, showing that computational models capable of inducing generalisations using abstract phonological features can detect these preferences from lexical statistics in the three languages. This paper presents a computational analysis of the development of initial clusters in Polish, which arguably presents a stronger test of these models. We show that (i) the statistics of Polish contradict the Sonority Sequencing Principle (SSP), favouring sonority plateaus, (ii) models that succeed in the other languages do not predict SSP preferences for Polish and (iii) children nonetheless exhibit sensitivity to the SSP, favouring onset clusters with larger sonority rises.

The functional load hypothesis of Berinstein (1979) put forward the idea that languages which use a suprasegmental property (duration, F0) contrastively will not use it to realise stress. The functional load hypothesis is often cited when stress correlates are discussed, both when it is observed that the language under discussion follows the hypothesis and when it fails to follow it. In the absence of a more wide-ranging assessment of how frequently languages do or do not conform to the functional load hypothesis, it is unknown whether it is an absolute, a strong tendency, a weak tendency or unsupported. The results from a database of reported stress correlates and use of contrastive duration for 140 languages are presented and discussed. No support for the functional load hypothesis is found.

Opaque phonological patterns are sometimes claimed to be difficult to learn; specific hypotheses have been advanced about the relative difficulty of particular kinds of opaque processes (Kiparsky 1971, 1973), and the kind of data that is helpful in learning an opaque pattern (Kiparsky 2000). In this paper, we present a computationally implemented learning theory for one grammatical theory of opacity, a Maximum Entropy version of Stratal OT (Bermúdez-Otero 1999, Kiparsky 2000), and test it on simplified versions of opaque French tense–lax vowel alternations and the opaque interaction of diphthong raising and flapping in Canadian English. We find that the difficulty of opacity can be influenced by evidence for stratal affiliation: the Canadian English case is easier if the learner encounters application of raising outside the flapping context, or non-application of raising between words (e.g. life with [ʌɪ]; lie for with [aɪ]).

This paper describes and analyses the vowel-harmony system of the Kordofanian language Moro. Moro has a cross-height dominant-recessive raising harmony system in which high vowels and a central mid vowel trigger harmony, while peripheral mid vowels and a central low vowel are harmony targets. Schwas can co-occur with any of the vowels, appearing inert to harmony. Yet when schwas occur alone in a morpheme, some trigger harmony and some do not. We suggest that an original ATR-harmony system shifted to a height system via merger and centralisation, producing two distinct central vowels, rather than a single schwa. One vowel patterns with the higher vowels in triggering harmony, and the other patterns with the lower vowels. We also propose that a particle-based representation offers the best characterisation of the groupings of target and trigger vowels in the language.

Metres are typically classified as being accentual (mapping stress, as in English) or quantitative (mapping weight, as in Sanskrit). This article treats the less well-studied typology of hybrid accentual-quantitative metres, which fall into two classes. In the first, stress and weight map independently onto the same metre, as attested in Latin and Old Norse. In the second, stress and weight interact, such that weight is regulated more strictly for stressed than unstressed syllables, as illustrated here by new analyses of Dravidian and Finno-Ugric metres. In both of these latter cases (as well as in Serbo-Croatian), strictness of weight-mapping is modulated gradiently by stress level.

This paper presents a Maximum Entropy learner of grammars and lexicons (MaxLex), and demonstrates that MaxLex has an emergent preference for minimally abstract underlying representations. In order to keep the weight of faithfulness constraints low, the learner attempts to fill gaps in the lexical distribution of segments, making the underlying segment inventory more feature-economic. Even when the learner only has access to individual forms, properties of the entire system are implicitly available through the relative weighting of constraints. These properties lead to a preference for some abstract underlying representations over others, mitigating the computational difficulty of searching a large set of abstract forms. MaxLex is shown to be successful in learning certain abstract underlying forms through simulations based on the [i]~[Ø] alternation in Klamath verbs. The Klamath pattern cannot be represented or learned using concrete underlying representations, but MaxLex successfully learns both the phonotactic patterns and minimally abstract underlying representations.

The Australian language Arrernte has been argued by Breen & Pensalfini (1999) and Evans & Levinson (2009) to present a case of VC syllabification with coda maximisation, rather than CV syllabification with onset maximisation. In this paper we demonstrate that greater insights into a number of phenomena are achieved when they are analysed with CV syllabification and onset consonants that are moraic, a possibility independently proposed for a wide range of languages by Topintzi (2010). We review a range of evidence from phonetic studies, acquisition and musicology that points towards CV syllabification in Arrernte, and analyse allomorphy, stress assignment, reduplication and the transpositional language game ‘Rabbit Talk’ in terms of reference to moraic structure. The results lend themselves to new directions in the analysis of Arrernte, and provide further evidence for moraic onsets in prosodic morphology.