1. Overview of the volume
Primitives of Phonological Structure is a volume comprising 12 chapters that arose from a workshop held at GLOW 37 in 2014. At the heart of this volume are pointed questions about the nature of phonological representations, especially at the subsegmental level. The editors’ motivations and goals are laid out both in the preface and in the first chapter of the book: ‘The primitives of phonological theory – whether we call them features, elements, gestures, or by some other name – represent the smallest units in phonology and stand in some relation to phonetic reality. At least this much most phonologists would agree on; but once we begin to ask further questions concerning the precise nature of the primitives, we find a significant divergence of views’ (p. 1). Not all of these views are taken up here, but the volume is admirable in terms of the depth at which a few key questions are addressed. In particular, the editors highlight several that the volume’s contributors aim to address:
-
• What are phonological forms ‘made of’, particularly at what we usually refer to as the subsegmental level? If segments are the atoms of phonological representation, what are the subatomic units?
-
• What is the phonetic content of phonological units? What is the relationship between phonetic reality and phonological primitives?
-
• Where do phonological primitives come from? Are they innate/universal or acquired/emergent from phonological data?
Additional questions arise to a somewhat lesser degree in the service of looking at the ‘bigger’ questions listed above. These include whether segments are actually important representational/organisational units, whether phonological primitives are binary or privative and whether/how primitives are subsegmentally organised. Taken as given or well-known to readers are topics such as the autosegmental nature of features (Goldsmith Reference Goldsmith1976, Reference Goldsmith1990) and the more ‘standard’ proposals surrounding feature geometry (Clements Reference Clements1985). These are perhaps topics that phonologists might expect to anchor a volume on phonological primitives, but in largely setting them aside, the editors leave room for the authors to explore the questions listed above in greater depth.
2. Major themes of the volume
In this section, I provide a description of the chapters of the book and how they speak to one or more of the major themes listed in §1.
2.1. What are phonological forms made of?
Many, if not most, of the chapters in this volume centre on the question of what (subsegmental) units serve as the primitives of phonological representation. Chapter 2 provides a clear, concise summary of the most influential proposals, including distinctive features (Jakobson et al. Reference Jakobson, Fant and Halle1952; Chomsky & Halle Reference Chomsky and Halle1968), gestures (Browman & Goldstein Reference Browman and Goldstein1986, Reference Browman and Goldstein1989) and elements (Kaye et al. Reference Kaye, Lowenstamm and Vergnaud1985).
2.1.1. Features vs. gestures
Chapters 2 (by Vaux & Miller), 6 (by Kim) and 10 (by Gafos & Sorensen) concern whether features or gestures should be adopted as phonological primitives. Interestingly, the volume includes two chapters (Chapters 2 and 6) that explicitly argue against gestures and in favour of features, before Chapter 10, in which gestural representations are actually explicated. Chapter 2 lays out a number of arguments against the adoption of gestural representations. Ultimately, though, all of these can be boiled down to a single criticism of the tenet held in early versions of Articulatory Phonology that within a gestural score (a gestural representation of a phonological form), gestures cannot be added, deleted or altered. This is seen by many researchers as a hurdle to the application of gestural phonology to phenomena that are typically associated with generative phonology.
However, in the nearly 40 years since Articulatory Phonology was first proposed, many researchers working with gestures have set aside this tenet of the theory. In doing so, they have successfully embedded gesture-based representations within generative frameworks. For instance, Hall (Reference Hall2003) claims that while ‘intrusive’ vowels result from the relative timing of consonantal gestures, truly epenthetic vowels include gestures that are present in the phonological output but not the input. More recently, many authors have used Optimality Theoretic (Prince & Smolensky, [1993] Reference Prince and Smolensky2004) constraints from the Max, Dep and Ident families in analyses built around gestural addition, deletion and parameter alteration. These include Tejada’s (Reference Tejada2012) analysis of tonal processes in Sierra Juarez Zapotec and Weinan Chinese, Smith’s (Reference Smith2018) analyses of vowel and vowel–consonant harmony, Walker & Proctor’s (Reference Walker and Proctor2019) analysis of liquid distribution in General American English and Bradley’s (Reference Bradley, Colina and Martínez-Gil2020) analysis of rhotic distribution in Spanish. Taken together, these works indicate that the assumption that gestures are neither added nor deleted from phonological forms is not ubiquitous among theoretical phonologists who adopt gestural representations.
Chapter 6 similarly makes a case against the adoption of gestural representations (and in favour of features) by appealing to assumptions held within early versions of the theory of Articulatory Phonology. The chapter presents results from several phonetic studies of Korean and claims that the data cannot be accounted for by gestural representations. One such criticism about the lack of a necessary gesture for laryngeal height is easily addressed: the inclusion of a gesture for laryngeal height can be proposed based on phonetic evidence that such a gesture is necessary. The studies cited in this chapter present just such evidence, as does work by Oh & Lee (Reference Oh and Lee2022), who propose that a laryngeal height gesture is active in Korean.
It is also stated throughout Chapter 6 that a further shortcoming of a phonological theory employing gestures is an inability to appeal to word positions and morpheme boundaries despite their import in analysing phonological phenomena. While it may be the case that morphological information is not implemented in the original Articulatory Phonology proposed by Browman & Goldstein (Reference Browman and Goldstein1986), more recent work, both experimental and theoretical, has made strides in this area. Cho (Reference Cho2001) presents the results of a phonetic study that suggest that intergestural timing in Korean is sensitive to morphological structure, while Smith (Reference Smith, Jesney, O’Hara, Smith and Walker2017) presents a gestural analysis of idiosyncratic vowel harmony in Classical Manchu that relies on reference to a gesture's association with a word-initial syllable.
While recent work employing gestural representations speak to the theoretical issues brought up by Chapters 2 and 6, Chapter 10 focuses on the kinematic and mathematical underpinnings of gestures as units adapted from a dynamically defined theory of motor control. Relatively, less attention is paid to the role of gestures as primitives of use within a system of theoretical phonology. While some of the discussion in this chapter is a welcome introduction to the workings of gestures and intergestural relations, a greater focus on phonological implications might have made a greater contribution to this volume.
2.1.2. Features vs. elements
Chapters 3 (by Baroni) and 11 (by Prince) focus on elements, the phonological primitives assumed in Element Theory (Kaye et al. Reference Kaye, Lowenstamm and Vergnaud1985). Both provide comprehensive and useful overviews of the development of the various sets of elements that have been proposed throughout various iterations of the theory. In addition, Chapter 11 provides an introduction to the workings of the theory that will be welcome to non-practitioners.
Comparatively less attention is given in these chapters to arguments in favour of adopting elements as phonological primitives over either features or gestures. One exception to this is Chapter 11’s discussion of Element Theory’s ability to account for patterns in both child acquisition and adult aphasia in terms of interphonemic distance. With the adoption of elements rather than features as the primitives of phonological representations, interphonemic distance measurements emerge as an appealing explanation for observed rates of different types of simplifications and substitutions in phonological forms in French.
2.1.3. Subsegmental organisation of features
Several chapters in the volume focus on arguing for specific forms and/or organisations of features rather than on arguments in favour of adopting features over either gestures or elements as phonological primitives. In addition to arguing for a specific geometry that subsegmentally organises features, Chapter 8 (by Lahiri & Kotzor) addresses one of the secondary questions put forth by the editors: whether features are binary or privative. The experimental results presented in this chapter serve as a fairly effective basis for the argument that speakers mentally represent phonological properties of segments as being either present or absent.
Chapter 9 (by Schwartz) introduces a compelling model of subsegmental feature geometry in which different levels of hierarchical structural positions lead to differences in how articulatorily defined features are phonetically interpreted and realised. The theory has parallels with Clements’s (Reference Clements1991) theory of registers for the height feature [open], as well as Parkinson’s (Reference Parkinson1996) Incremental Constriction Model, which relies on stacked instances of the height feature [closed]. Both are models in which the interpretation of a single, articulatorily based feature relies on its location within a subsegmental, hierarchical structure. However, these parallels are not explored in the chapter.
Chapter 12 (by Puggaard-Rode, Botma, and Grijzenhout) focuses on a novel variation of Q Theory (Shih & Inkelas Reference Shih, Inkelas, Kingston, Moore-Cantwell, Pater and Staubs2014) referred to as Q-CV. Like Q Theory, Q-CV assumes that segments (Qs) are made up of multiple subsegmental units called qs, each of which is a feature bundle. Rather than arguing in favour of the Q-CV framework or the adoption of Qs and qs as phonological primitives, the chapter instead examines the viability of utilising Q Theoretic representational assumptions for stop consonant contrasts. While Q Theory has typically been deployed to represent typologically uncommon segments with internal temporal variation, the chapter effectively demonstrates that the theory can represent both typologically common and uncommon types of stop consonants.
2.2. What role do phonetic facts play?
Another core question this volume aims to address has to do with the debate between phonological frameworks that do admit a role for phonetic substance versus those that are so-called ‘substance-free’. A highlight of the volume is the juxtaposition of Chapters 5 (by Hall) and 7 (by Iosad). While Chapter 5 advocates for a model of phonology in which phonetic reality informs both the definition of phonological primitives and the application of phonological processes, Chapter 7 advocates for a phonological framework in which features are substance-free.
Chapter 5 provides a clear and convincing overview of the debate surrounding substantive versus substance-free phonology, ultimately coming out in favour of what the author refers to as ‘substance use in moderation’. In contrast, Chapter 7 presents an analysis of a single phenomenon, Welsh provection, as evidence for the strengths of a substance-free approach to phonology. While its scope is narrower, it makes an effective case for the utility of adopting features that are not necessarily phonetically transparent. The account rests on the status of the feature [(±)tense], whose spread from a stressed vowel to a following consonant triggers consonant devoicing. By assuming that a single feature is responsible for the phenomenon, this analysis receives a simple treatment as a case of simple local feature spreading. However, it necessitates that this feature must receive differing phonetic implementation when it is associated with vowels (tongue root position) versus consonants (voicing). The chapter does not take the firm position that Welsh provection provides strong evidence for the substance-free nature of all features. Rather, it reasonably proposes that Welsh provection provides evidence that at least some features may be based on phonological behaviour and not only on transparently defined phonetic implementation.
2.3. Are features innate/universal or acquired/emergent?
The final question the editors of this volume raise has to do with whether features are innate or acquired, and relatedly whether they are universal across languages or emergent from the data a language learner is exposed to during acquisition.
Chapter 4 (by van 't Veer) presents a model of phonological acquisition and applies it to data from several Dutch-acquiring children. The model progresses by child learners’ ‘activating’ features as they gain evidence for their active status in the grammar that shapes their phonological inventories. Whether these features are selected from a predetermined, language-universal set is left unspecified, as is whether ‘activation’ is equivalent to a feature being created in the mind of a learner versus being selected from a premade set.
Similarly, Chapter 5 (discussed in §2.2) includes a brief discussion of featural activation in the context of the contrastive hierarchy (Dresher Reference Dresher2009). While learners of different languages may follow different trajectories of feature activation and as a result end up with different subsegmental organisations of features, assumptions about the universality of the feature set itself are not completely clear. Chapter 11 (discussed in §2.1.2) likewise proposes that the precise set of elements that are active within a language’s phonological system is a language-specific property. However, the chapter does not explicitly state whether the set of possible elements is universal/innate and simply activated in the mind of a speaker of a certain language, and if so, by what sort of mechanism.
Of the main questions introduced at the outset of the volume, the question of the innateness of phonological primitives is the one that gets the least direct attention. However, many chapters address it in some way. In particular, the innate versus emergent feature debate comes up in the context of the debate around phonologically substantive versus substance-free features (discussed in §2.2). Substance-free phonology is often associated with the assumption that features are acquired and/or emergent from language-specific data rather than being innate and universally defined.
3. Evaluation of the volume
Primitives of Phonological Structure is an ambitious volume, asking fundamental questions about the nature of phonological forms. They are questions any phonologist should take seriously, as the assumptions we make about phonological primitives and their organisation ultimately make huge differences in the successes and failures of our analyses once we apply a grammar to them. The nature of phonological representations is a huge topic, and ultimately one that no single volume could ever conquer. Instead, the editors focus on three somewhat smaller themes/questions. While others might have narrowed down this fundamental question differently and addressed quite different questions about the nature of phonological representation, this volume is ultimately an informative and thought-provoking attempt at answering the questions it sets forth in Chapter 1.
One weakness of this volume is that its chapters are somewhat disjoint, each speaking to the main themes of the volume but ultimately doing so mostly independently of one another. The volume might have been more effective if chapters had been organised thematically, either by ordering them differently or by explicitly breaking them up into sections.
In addition, this volume could have been improved by the authors’ including more references to the contents of each other’s chapters, especially in the numerous instances in which they disagree. Chapters 5 and 7 are a notable exception, with their authors referencing each other’s proposals. It is no surprise, then, that this pair of chapters presents perhaps the most compelling debate in the volume. Similar debates, or at least theoretical comparisons, would have been welcome between Chapters 7, 9 and 12, in which features receive different phonetic interpretations based on their structural positions; and between Chapters 4, 5 and 11, in which features/elements become phonologically active during acquisition.
While the chapters in this volume operate too independently from one another for it to be used as a textbook, many (if not most) individual chapters could be included as readings in a topic-specific course or seminar. Chapters 5 and 7 serve as effective counterparts to one another in the discussion of the role of phonetics within phonology, and Chapter 4 would be a welcome inclusion in a course or seminar on phonological learning/acquisition.
Throughout the volume, complex figures, including rule terminology, autosegmental structure, mathematical notation and data visualisation, are all rendered effectively. Typos are minimal and do not impact readability. Clarity of writing varies from chapter to chapter, but is overall high.
Any volume that explores the nature of phonological representations is a welcome one, and Primitives of Phonological Structure is ambitious in asking several critical and fundamental questions about the topic. This volume will ideally spur additional work exploring phonological primitives in light of the important contributions that have been made in recent years. One fruitful avenue is the application of gestural representations to a greater range of phonological phenomena than the theory of Articulatory Phonology was originally intended to model (discussed in §2.1.1). Another is the increased role of gradience, both phonetic, as in work by Lionnet (Reference Lionnet2017) and McCollum (Reference McCollum2019), and cognitive, as in work in frameworks such as Gradient Symbolic Computation (Smolensky & Goldrick Reference Smolensky and Goldrick2016; Hsu Reference Hsu2019).
