2.1. Part 1: Melody and segmental representation

The introduction for this part is written by Florian Breit, and is followed by a further four original research chapters. Despite somewhat baroque language here and there, the introduction provides a brief yet solid overview of the basic tenets of ET: rather than (usually) binary features, mostly but not always defined in articulatory terms, the primitives of ET are privative, autonomous (they can be directly and independently interpreted phonetically) and acoustically defined.

From the principle of autonomous interpretation, combined with the understanding that elements are not inherently ‘consonantal’ or ‘vocalic’, it follows that processes such as vowel–glide interactions are successfully accounted for in ET. The ubiquitous vowel–glide pairs [i $\sim $ j] and [u $\sim $ w] are easily accounted for, since what is needed to turn one into the other is simply a matter of reassigning the element involved from a vocalic to a non-vocalic syllable position or vice versa, no subsegmental reorganisation needed. Liquids, in ET, are very similar to vowels, in that they consist only of non-resonance elements. In fact, Backley (Reference Backley2011) goes so far as to propose that a third vowel–glide pair exists: [a $\sim $ R], where R stands for any rhotic. Still, intricate patterns that are not prima facie straightforward do persist, and Bendjaballah’s contribution about l-vocalisation in Upper Austrian German provides not only relatively unfamiliar data, but also a solid account of the seemingly divergent influence of the lateral in different phonological contexts as a case of licensing interactions between like and different elements in different segments. The lateral approximant l, in ET, is represented as $|$ I, U $|$ (Backley Reference Backley2011; see also de Kok et al. Reference de Kok, Botma and Veer2019). No non-resonance elements are involved, and hence it cannot be a surprise that l is prone to vocalisation, especially in postvocalic position. By far most instances of l-vocalisation result in a vowel (or vocoid) that is closer to the lateral’s original $|$ U $|$ nature, at the expense of its $|$ I $|$ inclinations. In other words, some kind of back, high and often round vowel is the outcome. Upper Austrian German provides a counterpoint, however, where l-vocalisation may result in [ɪ] in some circumstances. Bendjaballah shows that previous research, which focused on interaction with the preceding nucleus, was too narrow in focus, but that rather, the immediately preceding segment determines the outcome, in what is akin to a kind of asymmetric OCP situation: if the preceding segment contains $|$ U $|$ but not $|$ I $|$ , the outcome is [ɪ], while that same $|$ U $|$ -containing segment may trigger rounding in an immediately preceding nucleus. This chapter provides a solid addition to the body of literature that highlights the interesting and sometimes seemingly unexpected behaviour of liquids, and the Upper Austrian German data show interesting parallels and differences with published patterns in first language acquisition (Marshall & Chiat Reference Marshall and Chiat2003; Inkelas & Rose Reference Inkelas and Rose2008), and it would be interesting to see a comparison.

Cyran’s chapter on final devoicing in Polish seeks to find the boundaries of what can be ascribed to phonology in a narrow sense, and in a modular model of grammar, where the different modules (phonetics, phonology, syntax, all they way up to pragmatics) have no direct access to the internal workings of any other module. The phenomena at hand are exceptions to final obstruent devoicing (FOD) in Polish – exceptions that appear to be motivated not by phonological considerations but rather by factors of pragmatics (contrast retention, for example). Polish has traditionally been seen as an FOD language, but what if, under certain circumstances, devoicing is incomplete or even absent? Under the strictly modular view, where phonetics and phonology are black boxes to each other, there should be no difference in phonetic implementation between an underlyingly voiceless final obstruent and one that has been rendered voiceless by FOD. Yet the data speak otherwise. Must we then abandon the modular perspective? Or is FOD part of the language-specific phonetics rather than the phonology? These are not easy questions, but Cyran’s contribution certainly goes beyond the standard laryngeal realism that is often associated with ET, and proposes a further development dubbed ‘Laryngeal Relativism’. Here, the typological asymmetry between voicing languages and aspiration languages is retained, but whereas in laryngeal realism, the two types differ both on which member of the opposition is marked, and which label ([Aspirated] or $|$ H $|$ vs. [Voice] or $|$ L $|$ ) is applied, this latter difference is replaced by a single, substance-free label, whereas the series to which the label is ascribed is still a parametric choice that languages must make. Cyran argues that Polish has both an $|$ H $|$ variant and an $|$ L $|$ variant, and that the fact that both display voicing retention provides evidence for this substance-free approach. While this particular analysis appears to remain agnostic about the nature of the phonological primitives, the question of substance in ET is one that deserves much more attention, and the inclusion of Cyran’s chapter in this book is a good sign.

Both Breit’s and Rennison’s chapters tackle more theory-internal questions, such as how strictly elements are bound to their own unique autosegmental tier, or whether different elements can live on one and the same tier. Rennison argues that the latter is, or should be, a viable representation. If two different types of elements (let’s say, $|$ U $|$ and $|$ A $|$ ), occupy the same autosegmental tier, they must necessarily be ordered sequentially. From this ordering, Rennison derives an account of asymmetry in Turkish vowel harmony. Breit explores the homology between syntax and phonology, which flows from a foundational quest for GP: the government in that term is modelled after the government in Government and Binding theory, the dominant paradigm in syntactic circles in the 1980s. With contemporary generative syntactic thinking centred mostly on Merge and the properties of recursion, and the traditional view of phonology being that it is devoid of the latter, Breit shows that there is a place for Merge in phonology after all, and that a more rigorous definition of subsegmental head-ness might well follow from that. An interesting counterpoint arises between Breit’s and Rennison’s chapters, in that the authors appear to have diametrically opposed positions on the relevance of chirality in phonological representations. Clearly, this is one area of ongoing development in the field.

2.2. Part 2: Prosody and constituent structure

The amount of hierarchical structure in phonological representations is a point of much contention in GP. Two more or less opposite positions exist on this matter: on the one hand, we have the ‘arborealist’ approach, self-branded as GP 2.0, that seeks to find explanation for phonological phenomena in hierarchical relations between constituents (Kaye & Pöchtrager Reference Kaye and Pöchtrager2013). At the other extreme, the ‘deforestation’ approach takes the aforementioned inability of phonology to perform recursive operations to mean that constituenthood is to be absent from phonological representations as much as possible (Lowenstamm Reference Lowenstamm, Durand and Laks1996; Scheer Reference Scheer2004).

The latter approach is represented by Balogné Bérces & Ulfsbjorninn on Standard Southern British English (SSBE) and Yoshida on word-level prominence in Japanese and English, whereas GP 2.0 is represented by Pöchtrager, also on English. The other three chapters here are less explicit in their affiliation with one or the other framework.

It is conspicuous that out of the six chapters in this part, four discuss properties of English phonology, and three of those concern themselves, in some way, with vowels. Balogné Bérces & Ulfsbjorninn, for example, discuss the ongoing process of monophthongisation of diphthongs in SSBE. Lowenstamm provides an intriguing analysis of so-called irregular weak verbs, such as keep $\sim $ kept, which are short to begin with, and involve vowel shortening in the past tense. The real surprise here is that Lowenstamm proposes an analogy to certain aspects of Arabic, placing English alongside templatic languages. Finally, Pöchtrager provides an analysis of light diphthongs such as [ju] in English, with an interesting typological survey for context.

Fathi’s chapter also discusses issues of vowel length, albeit in Cairene Arabic, where an apparent alternation between long and short vowels has puzzled phonologists for decades. Fathi’s solution is as simple as it is elegant: rather than opting for either pre-suffix vowel lengthening (Watson Reference Watson2002) or shortening of underlyingly long vowels (McCarthy Reference McCarthy, Alhawary and Benmamoun2005), Fathi argues that the contrast is one of stress-induced prominence, not length. While perhaps the least theoretically bounded chapter in this part, the solution is intriguing and merits rigorous follow-up.

One could be forgiven for thinking, then, that all of current-day GP research regarding prosody and constituent structure is focused on vocalic problems, but Harris & Chiu pick up the gauntlet of a problem that has haunted syllable theoreticians of any persuasion from the beginnings of syllable theory itself: the status of #sC clusters. Do not think, however, that this is yet another regurgitation of an old problem. In contrast, this chapter is only one of two in the volume that reports on an experimental study, providing new and original data on the syllabification of word-initial #sC clusters in Cypriot Greek – confirming that a naïve syllabification of such sequences as tautosyllabic onset clusters does not capture their true nature.

Yoshida’s contribution concludes this part, and in some ways, is most prototypically concerned with constituenthood: it aims to ground the relatively uncontroversial observation that lexical words can have at most one site for primary stress or pitch accent to manifest (Hyman Reference Hyman2006) in a development of Charette’s (Reference Charette1990) proposal of Licensing Projection, by introducing a Licensor tier, where only ungoverned, non-empty positions may project. Interestingly, this approach also promises to shed light on headedness in the subsegmental domain: the difference between a headed $|$ I $|$ and non-headed $|$ I $|$ is that the former projects onto the Licensor Tier, whereas the latter does not.

The chapters in this part are, largely, perhaps not the most ground-breaking contributions to the volume, with Harris & Chiu providing an original and innovative counterpoint. Instrumental or experimental data are by no means the be-all and end-all of phonological argumentation, but it is certainly refreshing to see aspects of ET/GP’s strong points, namely its commitment to theoretical rigour, being tested in arenas that are traditionally more foreign to the paradigm.

2.3. Part 3: Emptiness, schwa, and epenthesis

One of the more salient characteristics of ET/GP, as viewed from the outside, is its apparent level of comfort with so-called ‘empty’ categories. Under the right circumstances, a syllable’s nucleus may remain unoccupied by melodic material, and traditionally a large part of ET/GP research has been devoted to uncovering (a) what these circumstances are, precisely, and (b) how such ‘empty’ positions are interpreted phonetically. The introduction to this section is provided by Youngberg.

We have already, by now, encountered empty nuclei in this volume, most explicitly perhaps in Yoshida’s chapter. Empty positions – that is, roughly speaking, X-slots on the timing tier that do not contain any subsegmental material – have been successfully invoked in the analysis a wide range of phenomena. One example is vowel–zero alternations, where, rather than framing alternations in terms of shifting syllable structure, some morphophonological operations are said to cause a certain nucleus to remain uninterpreted. Apparently, unexpected consonantal sequences (unexpected from a sonority hierarchy-informed view of syllabification) appear to be perfectly well-behaved when allowing for a rigorously restricted distribution of empty nuclei (Harris Reference Harris1997), and the typology of word-final consonants suddenly becomes more descriptively, and explanatorily, adequate (Kaye et al. Reference Kaye, Lowenstamm and Vergnaud1990; Piggott Reference Piggott1999).

This part offers a bit more typological variation compared to Part 2, and features the second of the two experimental studies. In it, Kenstowicz presents an acoustic analysis of a single-speaker corpus of Connemara Irish, where the consonant sequence in CVRC words (where R=sonorant) is regularly broken up by a schwa, leading to potential neutralisation with a class of words of the form CVCəC.

For more typological variation, Kula tackles sequential Bantu NC-clusters, and the restrictions on voicing of the C-part of such clusters as described in Meinhof’s Law (ML). Paraphrasing Kula, ML can be summarised as *NC _vce vNC _vce . A general repair strategy is the simplification of either of the two NC clusters by removing the occlusion, creating a nasal geminate. Similar to, but not exactly like rendaku in Japanese (which is unrestricted by sequentiality), this kind of problem is exactly where GP shines, and Kula duly demonstrates the power of general licensing statements when treated right – in this case, by constructing a so-called ‘licensing path’, where the domain of any kind of licensing statement is what determines its relative prominence.

Szigetvári provides an analysis of intrusive plosives in English, and although this chapter does not report an experiment, like Harris & Chiu, or an acoustic corpus study, like Kenstowicz, the inclusion of a survey gives some additional empirical weight to the chapter. Scheer provides an account of CəCj sequences in French by means of a Strict CV-analysis (Scheer Reference Scheer2004). In such sequences, the schwa may sometimes be left silent, but in other cases it must be pronounced. In earlier work, Charette analysed the difference as stemming from the syllabic affiliation of either the Cj part of the sequence (Charette Reference Charette1991) or the C(ə)C-part (Charette Reference Charette1998, Reference Charette and Ploch2003). Scheer updates the empirical foundation of this phenomenon by means of a grammaticality judgment survey, and concludes that the earlier Cj-based analysis was closest to an adequate description of the data, but that for the full picture, the status of that C must also be considered: In French C₁əC₂j sequences, schwa may be dropped, except if C₂ is a lateral.Footnote ¹

Cristófaro Silva shows, through a classic GP analysis of final consonants in Brazilian Portuguese, that the language is moving away from preferring non-empty final nuclei, to a state where empty final nuclei are preferred, thus rendering it in the ‘final onset’ class in the typology of Piggott (Reference Piggott1999).

Finally, although this is in actuality the first chapter in Part 3, Cavirani colours outside the traditional lines of GP by incorporating insights from Optimality Theory (Prince & Smolensky [1993] Reference Prince and Smolensky2004), more specifically Turbidity Theory (Goldrick Reference Goldrick2001). As noted by the author, GP has, as a matter of tradition or convention, never paid much attention to problems of phonological derivation or computation, but rather has focused on building a rigorous understanding of phonological representations. It has been very successful in this, but a phonology without operations is no phonology at all. In this sense, Cavirani is pushing important boundaries, and the placement of this chapter at the beginning of Part 3 is well-deserved.

All in all, whereas Part 1 contains contributions that fit comfortably in the realm of subsegmental phonology and ET, Part 3 is perhaps most squarely the home of GP. This part contains a balanced mix, both in terms of typological breadth and in mixing more straightforward implementations of GP principles with work that aims to extend the boundaries of the framework.

2.4. Part 4: Prosodic structure and recursion

That brings us to the last part of the book, with an introduction by Youngberg & Breit. Throughout this review, we have already come across the question of recursion in phonology a few times, and so it is interesting to see an entire book part dedicated to this matter. The excitement is slightly dampened, however, by the realisation that there are only two original chapters in this part (apart from the excellent introduction).

At face value, it would make sense to assume that phonology does not do recursion; naïvely, what would it even mean for, say, a syllable to be embedded in another syllable? In fact, this is exactly the type of recursion that we are presented with in the two chapters here. Both chapters assume only the phonological primitives to have ontological value, with all of phonology built up recursively from these primes. In Precedence-free Phonology (PfP, presented here in the contribution by Backley & Nasukawa), the set of self-projecting units is made up of the resonance elements $|$ I $|$ , $|$ U $|$ , and $|$ A $|$ . Each language must have at least and at most one of these parametrically designated as its base element; this base element is the one that, when in a bare structure, is phonetically interpreted as the neutral or epenthetic vowel. This is its recursor, and by projecting upwards it provides scaffolding for other elements to express their acoustic qualities. In this sense, the approach is quite reminiscent of work on the Modulated Carrier Signal (Harris & Urua Reference Harris and Urua2001; Harris Reference Harris2006), in which a baseline acoustic carrier signal (usually assumed to be schwa-like in timbre) is modulated by linguistically relevant modulators (the elements). In PfP, however, not only is the notion of the carrier signal operationalised by means of a recursive element, this recursion builds upward and upward, fulfilling the role of traditional constituents such as the syllable or foot, rendering these epiphenomenal at best. The current chapter demonstrates the value of such ideas in accounting for the distribution of nasal vowels in French and moraic nasals in Japanese.

Radical CV Phonology, presented in the final chapter by van der Hulst, takes the drive towards minimalism, that is present in so much of ET/GP (either explicitly or implicitly), and turns it up to eleven. Here, constituenthood is fully abandoned, and instead, all of the prosodic hierarchy is represented in terms of recursive layers of C and V units. Whereas traditionally the coda of a syllable would be plausibly posited to be C, it is perfectly admissible for it to be C dominating a further CV syllable; in fact, this is precisely the manner in which trochaic feet are represented. Naturally, as the descriptive adequacy of the theory must be at least comparable to that of others, giving up this much syntagmatic richness comes at the price of an increase of paradigmatic complexity: the representation of a Dutch word such as dominee [domine] ‘preacher’ comes out as a single tree with a head V dominating a dependent C, dominating a V that dominates a left-branching C and a right-branching C, which in turn dominates a right-branching V… et cetera. It is quite deep. As an interesting corollary, while the abandonment of constituents might seem radical to some, the upshot is the elimination of the need for empty positions, which might make others more at ease. There is a trade-off in there somewhere.