It will be very useful to review the fable of the blind people and the elephant to start this review.Footnote ¹ If it is not familiar to you, it goes like this: a group of blind people are put in proximity to an elephant, so that each of them ends up touching the elephant. Each person then describes their experience and what they think they encountered. One reports a wall (they touched the side of the elephant). One reports a snake (they touched the trunk). One reports a brush (they touched the end of the tail). And so on. The moral of the fable likely varies as much as the actual story, and ranges from the elephant touchers killing each other due to disagreements in their reports to sublime understanding being reached by all after listening to one another.

The excellent volume edited by S. J. Hannahs & Anna R. K. Bosch makes phonology the elephant in the room, and provides the opportunity to see how well our field is doing in describing and understanding our elephant. I will not presume to provide a critical evaluation of all the individual chapters, but will go through the highlights of each one, in anticipation of the discussion at the end.

Hannahs & Bosch provide a general overview of the volume in Chapter 1, and make two important points. The first is that almost the entirety of phonology at this point in time is a reaction to Chomsky & Halle (Reference Chomsky and Halle1968; SPE), and the second is the importance of the question: ‘What counts as a phonological fact?’. Given that almost every chapter provides distinct takes on both of these points, this introductory chapter paints the present phonological landscape as fairly fractured. Consequently, the broad representative approach taken in the handbook is a necessity; maybe if people read the entire handbook, some of the intellectual infrastructure in phonology will be improved.

The first four chapters following the introduction are all focused on Optimality Theory (OT). Pavel Iosad (‘Optimality Theory: motivations and perspectives’) provides a very successful introduction to the theory. The presence of this type of chapter indicates that OT may not be the fully dominant theory or lingua franca of phonology that many people think it is. In the next chapter, ‘Current issues and directions in Optimality Theory: constraints and their interaction’, Martin Krämer discusses theory-internal questions in OT. This is very interesting, because parallels to the post-SPE debates can be seen here in the discussion on the nature of constraints. OT may or may not have matured enough to require a period of pruning.

The next two chapters focus on how interfaces are dealt with in OT. Michael Ramsammy (‘The phonology–phonetics interface in constraint-based grammar’) works through a case study of Gorgia Toscana to demonstrate how phonology, phonetics and sociophonetics interact in OT. The spirantisation data presented here is exquisite. Ramsammy's analysis, which parcels different explanations into different modules, does the work of helping us understand Gorgia Toscana, but, once the modular organisation is set up, it seems to me that it doesn't really matter if the analysis is implemented in OT, or in any other framework. To me, this is the hallmark of excellent work. Finally, Ricardo Bermúdez-Otero (‘Stratal Phonology’) walks us through Stratal Phonology, which is a descendant of Lexical Phonology. This is an excellent chapter, presenting the issues involved in the interaction between morphology and phonology and how the cycle has played a fundamental role in understanding them. That Bermúdez-Otero makes a connection with non-OT literature in this area is welcome, because it is unfortunately rare. The (morpho-)syntax–phonology interface is a very dense and complicated area of research, and Bermúdez-Otero does a good job in providing an introduction to contemporary issues and results.

The next three chapters mirror the initial structure of this volume, but instead focus on rule-based phonology (RBP). Thomas Purnell (‘Rule-based phonology: background, principles and assumptions’) takes responsibility for the introductory chapter, and does a good job of reminding the readers about the genesis of generative phonology and the driving assumptions behind rules. The key part of Purnell's chapter is the importance placed on representations in many RBP research programmes. Interest in representations (on the part of any phonologist) makes our field a subcomponent of cognitive science. This is important, because it connects our field to a wide array of approaches and sources of data that help restrict our hypotheses (in a positive way).

Bert Vaux & Neil Myler's following chapter (‘Issues and prospects in Rule-Based Phonology’) is meant to present issues and prospects related to RBP, but is concerned primarily with comparisons of OT and RBP; conspiracies favour OT, while ‘crazy’, arbitrary or unnatural processes favour RBP. For a chapter on RBP, there is perhaps too much discussion of OT. The section on Universal Grammar is a good survey of current work on the acquisition of phonology, not necessarily in RBP or any particular theoretical model.

Heather Newell's chapter (‘The syntax–phonology interface in Rule-Based Phonology’) provides a Distributed Morphology approach to how (morpho-)syntax determines the scope of computation in phonological chunks. Although this chapter also compares RBP and OT approaches, these phenomena appear to be orthogonal to the contribution that Newell makes. Newell represents the contemporary version of the break from Lexical Phonology started by Halle & Vergnaud (Reference Halle and Vergnaud1987), where affixes were arbitrarily marked as either cyclic or non-cyclic. Progress on this approach has clearly been made, because Newell presents evidence where the cyclic vs. non-cyclic attribute can, at worst, be correlated with many morphosyntactic and semantic aspects. At best, Newell predicts the phonological behaviour of affixes that cause bracketing paradoxes and the illusion of morphemes that have different phonological behaviour in different blocks of rules.

Newell's and Bermúdez-Otero's chapters make a perfect pairing, because they mirror the discussion between Lexical Phonology and Halle & Vergnaud that was abandoned in the early 1990s, with the advent of OT. Both chapters demonstrate that we have made great progress since that era; we do not need to be concerned that the field is recreating the wheel. One thing that many might not want to admit is that, as the (morpho-)syntax–phonology interface is better understood, the results become less probative in comparing OT with RBP. The positive results are available to both models of phonology.

The volume turns to another approach to phonology for the next three chapters (Chapters 9–11). Tobias Scheer & Nancy C. Kula (‘Government Phonology: Element Theory, conceptual issues and introduction’) provide an introduction to Government Phonology (GP) that pursues reactions to SPE that are distinct from OT and RBP. GP differs from OT in focusing almost completely on representational issues. It also remains distinct from RBP, in that the theory of elements replaces the role of distinctive features in creating a minimal set of primitives that does not overgenerate. Scheer & Kula do an excellent job of setting the historical stage for the origins of GP in relation to phonology in the 1980s.

Tobias Scheer & Eugeniusz Cyran's chapter, ‘Syllable structure in Government Phonology’, deviates from the pattern of chapters set up by OT and RBP in not being a ‘current issues’ type of chapter. Instead, it reads more like a survey of work on syllable structure in GP for GP practitioners. Consequently, this chapter is a bit more difficult to read than the previous ones. It does offer an extensive discussion of how syllable structure is analysed within GP in a lateral approach, where government and licensing relations are defined in a strictly linear fashion (both can be non-local). There is some connection to earlier non-GP work on syllable structure to help the reader appreciate the analysis and argumentation, but not as much as in the previous chapter.

The final chapter discussing GP, ‘Interfaces in Government Phonology’, has Scheer & Cyran presenting how phonology interfaces with both the morphosyntax and phonetics modules. Covering both interfaces is a very nice feature of this chapter, since they are separate in the previous sections of this handbook. The understanding of interfaces within GP is one of strict modularity, where the interface does the translation between two modules with different vocabularies. This position is not unique to GP. For the morphosyntax–phonology interface, GP restricts the phonological structure that can be manipulated by the morphosyntax to only syllable structure. On the phonology–phonetics interface, GP aligns itself with the substance-free phonology movement, but does so in its own unique manner.

In Chapter 12, ‘Dependency Phonology’, the theory is presented by Harry van der Hulst & Jeroen van de Weijer as an approach focused mainly on representations. Dependency Phonology (DP) is based on the work of John Anderson, and is yet another distinct reaction to SPE. The focus on hierarchical representations in phonology makes DP appear very similar to GP to any reader who is unfamiliar with both approaches. However, they are actually diametrically opposed on the fundamental nature of representations. DP strives for acoustic grounding as the basis for the primitives of its representational system, and downright rejects ‘substance-free’ entities, eliminating the empty V and C positions that form the basis of syllable structure in GP. There is also no latent reference to distinctive features at all in the representational system. It should also be obvious from the name of the theory that the head–dependency relationship carries much of the explanation for the combinatorics involved within the representations.

The next two chapters present work that links phonology and connectionist models. John Alderete & Paul Tupper's chapter (‘Connectionist approaches to generative phonology’) provides a straightforward presentation and discussion of existing connectionist-based phonological works on syllabification, vowel harmony and assimilation, among others. As part of this, some claimed advantages of a connectionist implementation of phonology are discussed. Serendipitously, there is an extended published discussion of connectionism (neural nets) and phonology in the online perspectives section of the March 2019 issue of Language (Pater Reference Pater2019, together with six responses), which considers the pros and cons of incorporating connectionism into phonology. In his chapter, ‘Interfaces in connectionist phonology’, Joseph Paul Stemberger takes a different direction in the interpretation of interfaces from other authors in this volume. Stemberger discusses how connectionist models interface with phonology, and provides a more in-depth introduction to different types of connectionist models than the Alderete & Tupper chapter. Another difference from the preceding chapter is that Stemberger's main focus is on how connectionist models provide explanation for well-documented phenomena in morphology, language production, phonological development and adult speech errors, among others. Stemberger is also refreshingly honest in pointing out that ‘hand tailoring’ and symbolic representations play a larger role in connectionist work than most have thought.

This first set of chapters is grouped as Part I: ‘Theoretical frameworks’, with the following chapters constituting Part II: ‘Approaches’. I didn't fully understand this distinction, other than noticing that the chapters in Part II are single stand-alone ones.

Charles Reiss begins the second part with his chapter, ‘Substance Free phonology’. The bulk of this chapter is a review and reiteration of rationalist approaches to phonology, linguistics and cognitive science. This is useful and important information, but Reiss's presentation, which aims to dissuade phonologists from pursuing particular research programmes, will likely not have the effect that he wants. The very end of the chapter, where Reiss discusses current research that is in line with Substance Free Phonology, should have been the focus of the chapter, and would likely spur the type of work that the author thinks is important.

Jordon Fenlon, Kearsy Cormier & Diane Brentari's chapter, ‘The phonology of sign languages’, does an excellent job of presenting the problem of analysing sign languages in an accessible way that highlights the clear parallels in representational issues in spoken and manual languages. The discussion about mode-specific aspects of language is especially valuable, since it should help address questions about substance raised by other contributors in this volume. Given the general agreement in multiple previous chapters, future handbooks should include more space for signed languages. This single chapter covers the content of multiple chapters in Part I (e.g. introduction, contemporary issues, interfaces).

Chapter 17, ‘Phonology as an emergent system’, is by Diana Archangeli & Douglas Pulleyblank, who present a view in which phonology is the result of the interaction of storing multiple morphemes and then using phonotactic knowledge and relational statements to determine the form that occurs. The claimed benefit of this approach is that phonology can be built out of non-linguistic processes of memory, similarity, frequency, symbolic systems and entropy (page 478). Working in this way, phonology will not need innateness (or at least the need will be drastically reduced), and does not need to worry about abstractness, autosegmental representations, unnatural classes or opacity.

‘Laboratory phonology’, by Abigail C. Cohn, Cécile Fougeron & Marie K. Huffman, presents the history and philosophy of this approach. The exposition argues for a broader phonology, where specific phenomena ruled out by other contributors in this volume should not be ignored. Instead, these phenomena play a fundamental aspect in the understanding of sound patterns in speech. This difference is directly the result of Laboratory Phonology's focus on an experimental approach, in contrast to the formal, logical and (discrete) mathematical methods preferred by other contributors. (The authors also show excellent taste in referring to the blind folk and elephant story that began this review.)

‘Articulatory Phonology’, presented by Nancy Hall in Chapter 19, represents another view on how to analyse phonological data. In contrast to the approaches discussed in some of the earlier chapters, Articulatory Phonology focuses on the physical articulations involved in speech, as opposed to its acoustic characteristics. This creates a different representational system, which provides different insights and observations about phonology. Hall's presentation of the approach to sequencing gestures and how this illuminates aspects of syllable structure is excellent.

Stefan A. Frisch's presentation of ‘Exemplar theories in phonology’ indirectly echoes arguments made in the Laboratory Phonology chapter in favour of expanding the purview of phonology. By focusing on frequency effects, Frisch argues for the necessity of exemplar-type representations in phonology. Exemplar representations are distinct from the representations usually assumed by phonologists, in that they are more numerous and contain both more phonetically specific information and non-phonological contextual information, such as speaker, gender and location. Storing these extra pieces of information along with phonological information yields more numerous representations, and this additional information provides for new connections to phonetics and sociophonetics. The end of this chapter clearly notes that these proposals on exemplar representations are additive to the phonology, and do not replace the role that symbolic representations play in some phenomena.

The following chapter, ‘Algebraic phonology’, by Iris Berent, is the perfect companion for Frisch's, because it argues that phonology requires the exact opposite of exemplars. Berent presents arguments in favour of a symbolic phonology (see many of the earlier chapters in this volume), based on whether connectionist models can learn an algebraic rule. The main thrust of this argument is that phonological processes demonstrate equivalence classes and across-the-board behaviour, which must be reflected in any model of phonology. The case study that Berent focuses on is repetition of segments in an AA, AAB, ABB or non-repeating ABC pattern. Connectionist models are able to learn the AA pattern, but only for the segments that are in the training set. Given novel segments, a connectionist learner cannot recognise the AA pattern unless the model is specifically built to encode symbolic rules (see Stemberger's comments about hand-built connectionist models). Berent concludes that a symbolic component is necessary for phonology. Mirroring Frisch's chapter, Berent admits at the end of the chapter that an exemplar-type system is also necessary to fully account for phonological phenomena.

The penultimate chapter is Michael Hammond's ‘Statistical phonology’. It is the most level-headed and straightforward chapter in the volume, and connects the role of mathematics and statistics in phonology, starting with Labov's work on variable rules and continuing to contemporary MaxEnt grammars. For linguists who are not mathematically inclined, the chapter may appear intimidating, but it actually contains fairly basic mathematics. The best feature of this chapter is that Hammond balances proper accolades for advances in statistics in phonology with the question of whether this added level of sophistication actually models humans. The final chapter in the handbook is Bart de Boer's ‘Phonology and evolution’, which discusses the possible roles of biological and cultural evolution in phonology. Evolution is an extremely complex topic, and it appears to be very difficult to derive conclusions about its relevance for phonology. Determining the role of biological evolution in phonology is problematic, because the bulk of the vocal tract, including the larynx, consists of soft tissue and cartilage, which do not fare well in fossil records. Cultural evolution in phonology is also limited, due to the availability of written records only since circa 3500 BCE and the creation of acoustic recording devices as recently as the mid 19th century, time scales that are far too short for evolutionary work. De Boer implicitly suggests that the most exciting work on the evolution of phonology is modelling agent behaviour in computer simulations.

I now turn to a discussion of the value of this handbook. First, I will dispense with some minor points that must be made, but do not take away from the overall contribution of this book. I don't like endnotes, and am not sure that anyone does. There are minor inconsistencies and errors in the proof-reading and editing that I am fairly confident are the result of contemporary publishing houses eliminating copy-editing support. Surveying the names of the contributors, there appears to be a two-to-one male–female bias. Other questions about a broader representation of linguists in the academy are difficult to ascertain from names, but there appears to be work that needs to be done on this front.

I can now turn to the positives of this volume. Overall, I have benefited greatly from reading this entire volume, and believe that other phonologists would too. The editors have done an excellent job in covering most of the elephant that we know of as phonology. It is clear that phonology as a field has grown, and knowledge has been added since the publication of SPE. If this volume represents reactions to SPE, then our field is pretty lucky to have such a founding document. I think this conclusion is only strengthened by the fact that the bulk of work in the field can be read as identifying and remedying where SPE was wrong.

There is a surprising amount of agreement across the volume, if one reads past what is said about other approaches. One theme that is agreed upon is that there are different types of phonological ‘facts’. The reaction to this admission across the field is diverse, though. Some phonologists argue that phonology should be made ‘smaller’ by eliminating responsibility for certain phenomena. Others want to make phonology ‘bigger’ by including more phenomena. One might just say that the demarcation of what a phonological fact is is a personal choice for the researcher, and defines particular research programmes. However, this is clearly not the perspective presented in the volume. Instead, we can see an aspect of SPE that has not been rethought or rejected by contemporary phonologists. This is the disjunctive argument style, where decisions about analyses are strictly binary, and broad concepts such as innate vs. learned, gradient vs. discrete, abstract vs. concrete, simplicity vs. complexity, etc. are used as definitive judgements. This is not good, and is the main source of disagreement about what a ‘phonological fact’ is. I believe that the retention of this aspect of SPE is due to phonology as a field not updating its knowledge about the philosophy of science and other relevant disciplines.

My impression of the general phonologist's assumptions about how to do science is that three things are highly valued: unification, simplicity and experimentation/quantification. These mantras can be seen lurking behind almost all disagreements in this volume. Unification, put simply, is the desire to produce a single source of explanation (see Cat Reference Cat2017 for an overview). A desire for unification in phonology drives both the argument for a smaller phonology, with only symbolic/discrete resources, and for a larger phonology, with only exemplar/gradient resources. Notice that it is the source of the explanation that drives which phenomena count as facts for phonology. The question that we phonologists should ask ourselves is whether we can get a better grasp on our field if we pursue a pluralist position (see Cat Reference Cat2017: §5.2), where we accept multiple sources of explanation. This would not force us to choose between categorical and gradient representations. If we assume that categorical explanations in phonology are higher-level (e.g. more abstract, less connected to physics) and gradient explanations lower-level (e.g. less abstract, closer to physics), then Sober (Reference Sober1999) provides a relevant argument to show that unification would not be as useful for phonology as people think.

Adopting a pluralist position allows us to reconsider simplicity in phonology. The SPE evaluation metric trained phonologists to prefer a single answer, but this may not accurately reflect phonology as a cognitive system. We see glimpses of this through a direct pairing of the Frisch and Berent chapters, which argue strongly for diametrically opposed representations, exemplars vs. symbols, yet in the end both authors admit the need for the other type of representation. Is this a problem? Not if there are actually multiple cognitive systems supporting phonology. Multiple memory systems which are equated with learning appear to be the emerging (possibly established) consensus in psychology (Ashby & Maddox Reference Ashby and Todd Maddox2005, Murray et al. Reference Murray, Wise and Graham2017, Chechille Reference Chechile2018). Extending pluralist thinking towards the tension between experimentation and observation would also be helpful (Bogen Reference Bogen2017 provides an overview). This would allow phonology to properly value the full range of work being conducted, from documentation (purely observational) to psycholinguistics (purely experimental). The emerging popularity of computational phonology also fits here as an example of where modelling serves as a proxy for experimentation (see Shen Reference Shen2019 for how this works in astrophysics).

To wrap up and return to our elephant, the previous critical comments in this review promote certain endings to the fable over others. I'd prefer the universal transcendental knowledge ending by recognising the validity of the reports from different elephant touchers. The handbook edited by Hannahs & Bosch provides a step along this path, especially if it is combined with some cross-disciplinary grounding. Phonology should not be afraid of cross-disciplinary interactions, because, as evidenced in this volume, the field has made great strides since SPE, and has produced some very sophisticated results. These results provide unique insights from human language into the topics we draw on from other disciplines. Consequently, we have much to offer to non-linguists, as long as we can start a conversation. A key part of conversation is listening to the other party and, according to the elephant fable, this leads to deeper understanding. Listening to the work contained in this volume has enriched my understanding of phonology and I suspect it will have the same effect on other readers.