Phonological typology (not to be confused with the identically titled 2016 monograph by Matthew K. Gordon) is a collection of eleven papers which grew out of a 2013 Oxford workshop organised by Frans Plank, Larry Hyman and Aditi Lahiri.Footnote * In the preface, the editors lament the low profile of phonology in modern linguistic typology. The challenge to the contributors was to each present an example of how typology ‘can and ought to be done in state-of-the-art phonology!’ (p. vii). The formulation of this challenge, as well as the title of the original workshop (‘What is phonological typology – and why does it matter?’), imply that phonological typology has perhaps not gotten very far because no one really agrees on what it is supposed to be doing and where it is supposed to go. In large-scale comparisons, how can we get past superficial observations which, for phonologists, may not correspond to cross-linguistically comparable or psychologically valid entities? In this regard, the editors mention segment inventories, basic phonotactics and the largely discredited notion of syllable vs. stress timing (see Nolan & Jeon Reference Nolan and Jeon2014). Of course, the theoretical literature contains many outstanding cross-linguistic surveys of individual phonological phenomena; yet, apart from notable exceptions like Gordon (Reference Gordon2016) and the recent Joseph Greenberg Award winner, Easterday (Reference Easterday2017), one still frequently perceives a gap in connecting these dots into a larger vision that would define phonological typology as either its own subfield or a unified set of analytical methods.
The volume does not directly prescribe a solution to the phonological malaise in typology, and with so many different angles and topics gathered in one place, readers are left to draw their own conclusions about the larger enterprise of ‘discovering and accounting for order in diversity’ (p. vii) in the sound systems of the world's languages. Taken together, however, the high-quality contributions connect many higher-level dots, and (to echo Plank's chapter) convincingly demonstrate that there need be no typological malaise in phonology. In what follows, I will discuss the individual chapters, and attempt to disentangle some of the interwoven threads running through them. Perhaps the primary one is the issue of which level(s) of representation deserve to be typologised over; see also de Lacy's (Reference de Lacy2017) and Round's (Reference Round2017) reviews of Gordon (Reference Gordon2016). And, given the unanimous agreement that abstraction is involved, how can the typologist best assess evidence to establish robust generalisations? In a sense, these are just the basic problems of linguistic analysis. Nevertheless, they take on a new urgency in the typological context, where the validity of the whole may depend on the soundness as well as comparability of orders-of-magnitude more parts.
In the opening paper, ‘What is phonological typology?’, Hyman takes up two main points. One is to argue in favour of property-driven typology, i.e. the study of individual phenomena and how they vary across languages, as opposed to an alternative style of typology in which languages are categorised into holistic types (e.g. ‘tone language’). He argues that the label-based classification of whole languages oversimplifies cross-linguistic diversity, rather than examining or attempting to reach a new understanding of it. The other main point is that even though modern phonology has been concerned with cross-linguistic comparison since its structuralist inception, particularly with regard to mappings between levels of representation and the ways in which superficially similar sounds can be underlain by different structures, there is still a gap between the concept of typology inherent to phonology, and modern typology as more frequently practised in morphosyntactic circles. Whereas the latter investigates ‘what's where why’ (Bickel Reference Bickel2007), i.e. not only attested phenomena but also their geographical distribution and history, phonology tends to be more squarely focused on the ‘what’, secondarily on the ‘why’ and hardly at all on the ‘where’. Furthermore, in phonology we do not see much use of quantitative typological methods, nor of deductive approaches; most theoretically driven typological investigations of specific phenomena are inductive in nature, seeking out attested instances and generalising over them.
It is interesting to consider the rest of the volume in light of Hyman's discussion. Most of the papers are still primarily concerned with ‘what’, but there is a mix of deductive (e.g. Heinz, Gussenhoven) and inductive (e.g. Brohan & Mielke, Broselow, Riad) methods. One is also struck by the relative lack of Optimality Theory, a framework often thought of as incorporating a typological perspective. Aside from mentions of constraints here and there, OT is limited to a couple of isolated tableaux in Kiparsky's and Broselow's papers. This scarcity may reflect the fact that most of the papers are more concerned with representations than with computation. There are two main exceptions, although neither focuses on the type of phenomenon-specific microtypology for which OT is well suited: Heinz, who explores computational aspects of phonological generalisations in a way that is not compatible with additional theoretical hypotheses about constraint interaction, and Brohan & Mielke, who examine various recurring alternation types, again without going into the synchronic whys behind individual cases.
In the second paper, ‘An implicational universal to defy: typology ⊃ ¬phonology ≡ phonology ⊃ ¬typology ≡ ¬(typology ∧ phonology) ≡ ¬typology ∨ ¬phonology’, Frans Plank presents a historical overview of how phonology has been treated in linguistic typology from medieval times to the present, including recent textbooks and conference, journal and database statistics. Unsurprisingly, these results confirm the impressionistic minority status of phonology. He then confronts the question of whether phonology is different from morphosyntax in some substantive inherent way that explains its relative absence from the typological scene. Does phonology vary more wildly across languages, such that there are fewer universals and recurring co-variations? Or another possibility: is the phonological component of languages less complex or contentful than their morphosyntax? Plank observes that, unlike syntax, phonology never underwent a sociohistorical bifurcation into theoretical and typological wings. This has resulted in extensive use of cross-linguistic examples and comparison in phonology, but also a lack of focus and sophistication in typological methods. The specific expertise required to analyse sound systems will be a challenge for raising the profile of phonology within the community of typologists who practise and develop such methods.
After the editors’ overviews, Paul Kiparsky (‘Formal and empirical issues in phonological typology’) explains why a theory-neutral typology is impossible. Taking up the theme of representational levels, he argues that the locus of phonological universals and relevant typological generalisations is the output of the ‘l-phonemic’ level (corresponding to the word level of Lexical Phonology), rather than postlexical or surface levels, generative morphophonemic URs or even the traditional structuralist phoneme. Kiparsky incisively reanalyses a classic parade of purportedly exceptional phenomena (VC syllables in Arrernte, minimal vowel inventory in Kabardian, etc.), showing that all looks typologically normal at the l-phonemic level, whose status as perhaps the principal level of psychologically significant abstraction is also motivated by various independent lines of argument.
Moving to the other end of the representational spectrum, Ian Maddieson (‘Is phonological typology possible without (universal) categories?’) addresses the tantalising question of whether phonological typology can be done in a completely bottom-up way, avoiding any preconceived labels – an idea that, among linguistic subfields, is uniquely conceivable for phonology, due to the possibility of generalisation directly over the phonetic signal. However, Maddieson examines various examples of typologies along continuous dimensions (rhythm, sonority), and shows that even these do not escape an initial stage of categorisation where some discontinuity or distinction (e.g. acoustic profiles characteristic of consonants vs. vowels) is selected as the basis for classification. He quotes Harnad (Reference Harnad, Cohen and Lefebvre2005) – ‘to cognize is to categorize’ – and suggests that the apparent inescapability of categorisation is not necessarily a bad thing, since it is unclear whether we can say anything interesting about languages without delineating properties of one kind or another, even if the properties themselves can be gradient.
Next, we move away from representations and squarely into the phonological grammar. In ‘The computational nature of phonological generalizations’, Jeffrey Heinz surveys his and colleagues’ recent research on the mathematical properties of markedness constraints and input–output mappings, which appear to occupy quite a restricted area of the Chomsky hierarchy – namely, the regular region. Interestingly, the computational typology does not mirror the computational possibilities of mainstream phonological theories, strongly suggesting that these theories are missing something. Heinz outlines a framework for constructing rigorous intensional descriptions of what are or are not possible phonological generalisations. For me, this suggests a promising method of evaluating Evans & Levinson's (Reference Evans and Levinson2009: 445) claim that ‘once the full diversity [of linguistic systems] is taken into account, the UG approach becomes quite implausible’. The overall gist of the chapter, if not the details, can be understood even by readers unfamiliar with the analytical principles involved. Heinz also makes clear that current results are based on representations of phonological objects as strings, which invites the following question: if most of the other papers in the volume are concerned with representations, which are still controversial and very much still under development, how will advances in representational theory affect future results in computational typology? One hopes that this will be an active ongoing dialogue.
Anthony Brohan & Jeff Mielke (‘Frequent segmental alternations in P-base 3’) offer a different perspective on typologising over input–output mappings. Using ‘P-base’ (Mielke Reference Mielke2008), an extensive phonological database, newly reorganised to enable a focus on alternations and distributions, they conduct a large-scale investigation into the typology of phonological processes. Full-colour charts illustrate the relative incidence of alternations between different pairs of featural properties. There is a top-down, label-based analysis of basic properties of alternations, in order to discover recurring rule types via the co-occurrence of properties (e.g. ‘regressive’ and ‘assimilation’), as well as a more bottom-up, category-discovering exploration of recurring feature groupings. This paper stands out in the collection for its methodology: discussions of things like sampling issues and sources of uncertainty answer the editors’ call for more quantitative sophistication in phonological typology. Anecdotally well-accepted facts such as the prevalence of nasal place assimilation cannot be confirmed in the absence of such databases, and the quantitative approach furthermore enables the provision of evidence for disproportionately attested co-occurrences, e.g. the frequency of nasal place assimilation ‘not [being] accounted for by the frequency of its parts’. Such asymmetries show how typological investigations can open new questions for both phonetics and phonology, both synchronic and diachronic.
The two papers on computation are followed by three papers on contrast and featural representations. In ‘Predicting universal phonological contrasts’, Aditi Lahiri examines the cross-linguistic behaviour of coronals, in order to show that the right representational analysis can be precisely what brings order to apparent diversity. There could be a theoretically shallower typology, but this would largely miss the point. Lahiri compares various feature theories, and points out some of their problems in accounting for the attested typology of processes like coronal harmony and palatalisation. In contrast, the Featurally Underspecified Lexicon (FUL) is able to account for these coherently, while also resolving apparent exceptions. It achieves this typological result by using monovalent features, having neither separate C and V tiers nor any dependencies, and assuming universal underlying underspecification of Coronal and Plosive.
Next, in ‘Contrastive feature hierarchies as a new lens on typology’, B. Elan Dresher, Christopher Harvey & Will Oxford return to the classic theme introduced by Hyman, of how superficially similar sound systems can be profoundly different in their representations and structure. In the debate on what phonological typologists should typologise over, they propose that inventories should be analysed not in terms of segments as such, but rather in terms of the feature hierarchies that generate them. They illustrate and motivate the approach with a variety of case studies (e.g. Manchu, Yupik-Inuit, Algonquian and Ob-Ugric) and, as Kiparsky does for the l-phonemic level, provide diachronic evidence for feature hierarchies as a psychologically salient aspect of language structure. It is worth noting that, while feature hierarchies are generally constructed over ‘underlying’ segments (see Hall Reference Hall2007 on the Contrastivist Hypothesis), this is not incompatible with Kiparsky's proposal. Thus the study of feature hierarchies at the l-phonemic level emerges as a new way to do inventory-based typology.
Taking a different angle, Ellen Broselow (‘Laryngeal contrasts in second language phonology’) uses L2 phonology as a possible source of evidence for universal markedness patterns. In a detailed survey of the literature on obstruent voicing and devoicing, she finds that L2 phonological processes to a large extent mirror typological tendencies across attested L1s, even in the absence of any obvious motivation for those processes from a particular L1. However, there are also L2 patterns that do not reflect attested systems or typological tendencies. Broselow suggests that many of these fail to phonologise in, for example, nativised contact varieties of English, giving us a way to disentangle analytic and channel bias (Moreton Reference Moreton2008) as sources of the observed patterns. Because of gaps in the phonetic detail around production and perception of obstruent voicing in many cases of L1/L2 pairings, fine-grained hypotheses cannot always be tested on the basis of existing studies, but this is a fruitful area for future research.
The volume is rounded out by two papers on prosody. In ‘The phonological typology of North Germanic accent’, Tomas Riad zooms in on Norwegian and Swedish, presenting the volume's only microtypology of a specific, yet richly varying, phenomenon across a group of related language varieties. The typology demonstrates that the principles of the system, and their representation in speakers’ minds, are phonological in nature. He argues that only abstract categorical parameters can accurately describe the space of dialectal variation, which appears phonetically diverse but boils down structurally to just three primary, binary variables. Phonetic differences can be pinpointed as diachronic links, but the logic of the typology shows that they do not belong to the fundamental underlying system. This paper is the most up-to-date overview of North Germanic accent, which has received many treatments over the years.
Finally, in ‘Prosodic typology meets phonological representations’, Carlos Gussenhoven elaborates a basic universal representational model which consists of segments, a prosodic structure that contains and organises the segments, and two ways in which segments can be linked with the prosodic structure (alignment and association). A particularly intriguing consequence is that tone is essentially segmental. He discusses each of the three components in depth, arguing that imprecise conceptualisations and blurring of key boundaries (such as that between association and alignment) have led to confusion in phonological analysis, and by extension, typology. The metrical grid, which does not fit into the model, overgenerates; and tone, stress and accent each belong to a different component, so typologies which assume that they are structurally parallel run into problems.
The volume is organised by empirical phenomenon. This makes sense, since many papers treat multiple and/or overlapping themes in the relationship between phonology and typology, so grouping along those lines would be difficult. Most of the papers centre around demonstrations of how specific theoretical ideas can resolve typological confusion (including the establishment of universals), and/or how typology can give evidence for either the correct phonological analysis of specific phenomena or general properties of UG. As mentioned, the debate over the proper levels or entities to typologise over is one that figures prominently. Brohan & Mielke's paper is arguably the most reminiscent of typology as conventionally understood, in that it establishes relative frequencies of phenomena and recurring combinations of properties with a large database, while also demonstrating that this can be done in a phonologically sophisticated way that is flexible enough to accommodate new parameters such as those developed in the other papers. Despite the point raised by Hyman (citing Gouskova Reference Gouskova2013) about the fragmentation of phonology as a field, the papers in this volume do appear to share a common purpose.
The overall theoretical bent left me wondering whether the lack of transparent unity between representations, processes and surface generalisations is somehow more problematic in phonology than in morphosyntax. That is, however abstract and complex morphosyntax is, is the futility of generalisation without deeper analysis somehow more drastic in phonology, such that relevant properties are less readily looked up in a descriptive grammar? To what extent do we need to get the theory and analyses right before getting more typologically ambitious? Ultimately, this volume shows that a productive way forward – rather than pondering such questions from an armchair – is to keep doing theory and typology in tandem and in dialogue, so that they can inform and build on each other as ways of finding ‘order in diversity’.
