Current research in phonological typology

1 Aims and background

Despite the fact that phonology has historically been integral to the development of modern approaches to linguistic theory and typology (Greenberg et al. 1978; Hockett 1955; Jakobson 1941; Martinet 1955; Trubetzkoy 1939), its study is still underrepresented in many important venues (Hyman 2007; Hyman and Plank 2018). Therefore, we (Moran and Grossman) organized two conference workshops in 2019 aimed specifically at highlighting current research in phonological typology. The first, entitled ‘Phonological (in)stability and language evolution,’ took place at the annual meeting of the Societas Linguistica Europaea (SLE), from August 21–24, in Leipzig, Germany. The second, which shares the title with this special volume, was held at the meeting of the Association for Linguistic Typology (ALT), from September 4–6, in Pavia, Italy. The present special issue emerged from these two workshops. Its aims are twofold.

First, like those before us (e.g., Hyman and Plank 2018), we aim to raise the profile of phonological typology within typology and beyond. Phonological typology is particularly important for a number of reasons. First, phonology is one of the few domains of language in which we have substantial cross-linguistic data. Second, phonology has traditionally been a testing ground for linguistic theories, including a range of structuralist, generativist, and functionalist or usage-based theories; many concepts first developed for phonology were later expanded to other domains of language, and phonological data have been the basis for diverse perspectives on the relationship between synchrony and diachrony. Third, there is broad agreement among linguists of every persuasion that the sound patterns of languages are ultimately phonetically-based, which means that typological studies can make reference to the observable and measurable facts of production and perception, which are themselves grounded in human physiology and cognition. Fourth, phonological typology provides materials for hypotheses linking conventional aspects of language with extra-linguistic variables, such as geography and topography, demographic aspects of speech communities, genes, and more. Despite these advantages, some truly foundational methodological and theoretical issues in phonological typology are still unresolved. For example, the fundamental units of cross-linguistic comparison are still debated, and the reliability of existing cross-linguistic databases has been questioned. Moreover, phonological typology is often marginalized within typology, where morphosyntax and to a lesser extent lexical semantics dominate the discussion. In light of the disparity between the empirical, methodological, and theoretical importance (and achievements) of phonological typology, on the one hand, and its prominence within typology on the other, renewed attention to its study seems warranted.

Second, we are interested in contributing to the ongoing search for the factors that shape the phonological systems of the world’s languages, specifically by exploring the question of ‘what’s where why?’ (Bickel 2007; Hyman 2014, 2018). This question represents a recent paradigm shift in linguistic typology, which moves away from the goals of classical typology, such as the classification of languages into types and the identification of gaps between logically possible and attested patterns of language (Comrie 1989; Dryer 1989, 1992; Greenberg et al. 1978). Instead, due to factors inherent to the nature of the population of languages accessible to linguists today, typologists have increasingly come to formulate explicit causal theories about cross-linguistic diversity and its sources (Bickel 2014, 2015, 2017; Nichols 1992; Nettle 1999).

After we issued calls for papers, we received submissions on a wide range of topics within phonological typology – reflecting its diverse and engaged practitioners. The authors of the papers in this special volume demonstrate the challenges involved in doing phonological typology today – both in theory and in practice. Complementing these discussions are papers that explicitly model phonological patterns and processes. These contributions bring together some of the state-of-the-art research underway. They also highlight the many ongoing and lively debates regarding both phonological theory and linguistic typology.

We begin with a brief historical overview of phonological typology and its key questions, achievements, and challenges. While the amount of new high-profile research due to increased access to comprehensive reference materials, detailed datasets, and computational tools, methods and models, is impressive – it may be even more impressive that some key issues regarding cross-linguistic comparability in phonological typology remain largely unresolved today. We discuss these issues in the hope that researchers will find constructive ways to move forward, turning critical aporias into theoretical, methodological, and empirical opportunities and challenges.

We then summarize the contributions in this volume, highlighting how the authors tackle some of the big questions in phonological typology today.

2 Phonological typology: key issues

Phonology has always been typological in nature because its aim is to determine the possible range of phonological systems through cross-linguistic comparison (Hyman 2014). Its inseparability from typology is also reflected in their shared historical development. Typology, at least in its structuralist phase, began with the cross-linguistic study of phonological systems (Hockett 1955; Jakobson 1941; Trubetzkoy 1939) and later the classification of languages into types in search of language universals (Greenberg et al. 1978).^[1] Recent approaches shift away from language classification and towards explanations for the distribution of linguistic properties and processes across space and time (Nichols 1992).

Maddieson (this volume) writes, “Why languages are the way they are today is the most intriguing, but the most difficult part of the question, because the typologist must first correctly answer ‘what?’ and ‘where?’”. At first glance, getting the answers to ‘what and where?’ is now easy because of increased access to data, together with tools for quantitative analysis and data visualization. For example, with a few lines of code, a researcher nowadays can download and combine numerous freely available datasets (with linguistic and non-linguistic information) and then interactively explore patterns of interest. And due to language reference catalogs such as Glottolog (Hammarström et al. 2022), results about these patterns can also easily be plotted as points or polygons onto world maps.

This increased ease of access to data, quantitative approaches, and visualization techniques has given rise to a powerful paradigm for exploratory analysis in linguistic typology. However, this new data science approach has not solved, but rather exacerbated the interpretability of research results in light of fundamental and decades-old questions in phonology and typology. Two of the most fundamental questions both deal with cross-linguistic comparability: (i) what are the units of comparison (e.g., speech sounds, phonemes, syllables, morae, phonological processes)? And (ii) how can they be compared?

Today, as in all data-driven science, typologists need the qualitative and quantitative skills to undertake research in typology and also to confidently call bullshit (Bergstrom and West 2021) on research presented in quantitative form – whether through statistics or data visualizations (cf. Bickel and Nichols 2020; McNew et al. 2018). How did we reach our present situation, in which larger quantities of data, increasingly sophisticated statistical tools, and more powerful computational resources, permit large-scale analyses that present challenges in terms of reliability of data, appropriateness of methods, and interpretability and evaluability of results?

The trend towards “big” and open data in typology started nearly 20 years ago, most prominently because of the publication of the World Atlas of Linguistic Structures (WALS; Dryer and Haspelmath 2013). First published as a coffee table book and CD-ROM application, it has become an interactive web application with downloadable raw data. The success of open typological databases – first WALS and now many others published by MPI-EVA, MPI-SHH, and by other institutions – has profoundly changed the kinds of questions and the amount of research being published in typology. Nevertheless, the comparability problem remains, and is more prevalent now that researchers combine typological databases curated at different levels of linguistic analysis.

The design of the chapters on phonology in WALS and many other typological databases builds on the landmark work in synchronic phonological typology undertaken by Maddieson (1984) in his publication of the UCLA Phonological Segment Inventory Database (UPSID). By selecting a genealogically balanced sample of the world’s languages with carefully analyzed segment inventories and explicit phonological features, Maddieson aimed to address the fundamental questions of comparability and inference of language universals that had been raised in the decades before during the US National Science Foundation funded Typology and Universals project at Stanford from 1969 to 1976 (PIs Joseph Greenberg and Charles Ferguson).^[2]

Greenberg and his colleagues are best known for their research on the classification of languages into types and the search for language universals.^[3] Much of the resulting research during these years was published in the narrowly distributed and now difficult to access 20-volume series titled, Working Papers on Language Universals.^[4] Some of the highlights from that series were published in 1978 by Stanford University Press in a four-volume collection titled Universals of Human Language. Volume Two in this series is devoted specifically to phonological typology and includes the chapters:

1. Correlations of stops and fricatives (Gamkrelidze)

2. Palatalization (Bhat)

3. Universals of vowel systems (Crothers)

4. Syllabic consonants (Bell)

5. Nasal vowels (Ruhlen)

6. Consonant clusters (Greenberg)

7. Consonant harmony (Vihman)

8. Metathesis (Ultan)

9. Tone (Maddieson)

10. Phonological processes (Ferguson)

11. Word demarcation (Hyman)

12. Intonation (Bolinger)

13. Size-sound symbolism (Ultan)

Most of the research projects in the Typology and Universals project were based on convenience samples of languages in grammars available in the personal libraries of Ferguson and Greenberg or in the Stanford libraries.

Their work on phonological typology involved the development and analysis of the Stanford Phonology Archive (SPA), the first computer-based typological database (Crothers et al. 1979; Sherman and Vihman 1972). SPA is a convenience sample of 200 languages that includes information about phonological inventories, including the description of phonemes, allophones and their conditioning environments, and marginal sounds. Work on and with SPA raised two important questions regarding characterizing the distribution of cross-linguistic variation; both are still relevant and unresolved today:

1. What constitutes adequate descriptive categories for cross-linguistic comparison (Sherman and Vihman 1972: 163)?

2. How do we construct a cross-linguistic sample of languages that captures genealogical and areal diversity and thus allows for proper characterization of the distribution of linguistic phenomena (Sherman 1975)?

At the heart of the question of “what constitutes the basis for cross-linguistic comparison?” is whether “formal” descriptive (i.e., grammatical) categories can be equated and made comparable across languages (cf. Heath 2016). What has ensued in the decades since SPA is the debate about equatable concepts in phonology (e.g., Simpson 1999; Vaux 2009; Anderson 2017; Kiparsky 2018; Maddieson 2018), but also broadly in typology (Haspelmath 2010; Newmeyer 2010). The question about what the “right” descriptive categories (or comparative concepts) are, is highlighted by the perennial problem of phonemes – and “emic” categories generally. Here again one sees that phonological typology inherits the conceptual problems of both phonology and typology.

The second question is how does one create a cross-linguistic sample of languages that captures genealogical and areal diversity, and therefore provides for proper characterization and inference about the synchronic and diachronic distributions of linguistic phenomena (Sherman 1975)? What confounds must be taken into account when applying quantitative methods on cross-linguistic datasets in which we know there are biases? Languages are not statistically independent data points. Languages may share features because they are genealogically related or because they have been in contact. This statistical sampling problem was raised by Sherman (1975) and Bell (1978) and in the decades since – due particularly to increased data access – has become more pertinent, resulting in many debates in the typological literature (e.g., Rijkhoff and Bakker 1998; Janssen et al. 2006; Bakker 2011; Moran et al. 2012). While balanced samples are still used, we note a recent trend towards denser databases that allow factors such as genealogical affiliation and geographical location to be taken into account.

Work on and with SPA was presumably informed by Hockett’s (1955) A Manual of Phonology. Much of the manual is devoted to the development of a typology of sound systems based on a sample of 55 languages which were analyzed in terms of their consonant and vowel inventories, syllable structures, tones, and phonotactics.^[5] One example is shown in Figure 1 (Hockett 1955: 139). Hockett provides a hand-calculated and hand-drawn plot of the vowel-consonant balance from his language sample. The plot contains several layers of information about the distribution of vowels in the languages in the sample.

Figure 1:

Vowel-consonant balance from Hockett (1955: 139).

Also insightful is Hockett’s discussions in A Manual of Phonology and elsewhere on information theory, entropy, and functional load (e.g., Hockett 1955, 1967), in part because at the time these measures were conceivable for cross-linguistic analysis, but were impractical to calculate by hand.^[6] Furthermore, principles such as functional load require a notion of phonological contrast on which language comparison can be made. Thus, early work in phonological typology by Hockett and others built on research in previous decades on defining the notion of phonological contrast.

Trubetzkoy (1939) proposed some general principles of the organization of sound systems and their limitations. He provided a full exposition of the theory of phonemes as bunches of binary oppositions (adhering to the principle that ideally, when describing a particular language, one should strive to use as few oppositions as possible). However, distinctive features as the most basic phonological unit may differ in how they are modeled, e.g., do they describe articulatory or acoustic features and how do they map to the physiological mechanisms involved in production and perception? This key issue in phonological analysis is directly relevant to the general problem of comparability in linguistic typology.

So when did the comparability problem emerge in phonology and typology? The much cited quote by Sapir – nearly 100 years ago now – shows the roots of the relationship between the two:

… it almost goes without saying that two languages, A and B, may have identical sounds but utterly distinct phone[mic] patterns; or they may have mutually incompatible phonetic systems, from the articulatory and acoustic standpoint, but identical or similar [phonemic] patterns (Sapir 1925: 43).

Despite the fundamental problems inherent to phonological typology, research on the cross-linguistic distribution of sound patterns has continued in the 21st century, with a recent monograph presenting the field (Gordon 2016); studies of particular sound patterns (e.g., Bateman 2007, 2011 on palatalization, Nikolaev and Grossman 2018 on affricates, Easterday 2019 on highly complex syllable structure) and of phonological universals (Hyman 2008; Pericliev 2008); investigations of the structure of phonological segment inventories (Clements 2003, 2009; Coupe et al. 2009; Dunbar and Dupoux 2016; Nikolaev and Grossman 2020); typological reconsiderations of fundamental assumptions in phonological theory (e.g., Mielke 2008; Mackie and Mielke 2011); analyses of areal aspects of phonology (e.g., Stolz and Levkovych 2021; Nikolaev and Grossman 2018; Blevins 2017), studies of the impact of the physical environment on sound patterns (e.g., Everett 2013, 2017; Everett et al. 2015, 2016; Roberts 2018). The 21st century has also seen a number of new databases with global scope, such as LAPSyD (Maddieson et al. 2013), P-Base (Mielke 2013), PHOIBLE Online (Moran and McCloy 2019), as well as databases of particular areas, families, or phenomena, e.g., SAPHON for South America (Michael et al. 2015), EURPHON for Eurasia (Nikolaev 2018), Australian phonemic inventories (Round 2019), BDPROTO for ancient and reconstructed languages (Marsico 1999; Marsico et al. 2018; Moran et al. 2020), and SEGBO for borrowed phonological segments (Grossman et al. 2020), just to name a few. Data from phonological typology have been used in studies of human history and prehistory, as well as language evolution, e.g., Atkinson (2011; see also responses, including Cysouw et al. 2012; Maddieson et al. 2011; Sproat 2011; Wang et al. 2012), Creanza et al. (2015), Dediu et al. (2017), Blasi et al. (2019). And more recently, computational approaches to phonological typology have flourished (e.g., Cotterell and Eisner 2017; Pimentel et al. 2020; Salesky et al. 2020) and include expansive cross-linguistic corpora for phonetic-acoustic analysis (e.g., Black 2019; Jackson et al. 2020; Cohen Priva et al. 2021; Seifart et al. 2022; Ylonen 2022).

3 Contributions to this special collection

This special issue’s title conveys a broad theme and of course we cannot cover all of the vibrant current research going on in phonological typology. We solicited contributions from participants at both conference workshops and we also circulated widely a general call for papers. We asked for submissions related to these questions, but we were also open to other questions related to phonological typology:

1. What should be the big questions in phonological typology? Are any new ones on the horizon?

2. What is the current theoretical and empirical status of phonological universals?

3. To what extent can we do typology at the macro versus micro levels?

4. What phonological properties have been under-studied typologically?

5. Why are phonemic categories (and descriptive categories in general) inadequate for typology? What are the alternatives?

6. How do different approaches to typology, e.g., Canonical Typology or Distributional Typology, apply to phonology?

7. What can sound changes and synchronic alternations in typological databases tell us about diachrony?

8. How can we model the impact of language contact on phonological distributions?

9. What can we discern from the landscape of particular geographic areas from a typological perspective?

10. How does language acquisition affect worldwide phonological diversity?

We received a wide range of submissions from leading scholars in the field – and to our delight – from both the new and old guards. The views of these researchers provide not only breadth, but also depth, on current issues in phonology and typology.

We submit these three categories, based on themes in the articles, as our typology of the papers (however, we admit many other typologies are possible):

1. Diachrony & synchrony

2. Challenges

3. Models

4 Concluding remarks

The articles in this special issue are a good representation of the current work being done in phonological typology today. Segment inventories, prosodic systems, and phonotactics are the focus of most of the papers (with the study of tone largely absent in this volume, despite its importance in phonology). The perspectives adopted include synchronic, diachronic, areal, phylogenetic, acquisitional, and theoretical-methodological. Almost all include a quantitative dimension. Some exploit existing databases and others are based on newly-curated data; some leverage sophisticated analytical techniques. All have findings that increase our knowledge. In fact, phonological typologists theorize, observe, generalize and even experiment within a roughly agreed-upon paradigm. As a result, our knowledge of the sound patterns of the world’s languages and the factors that shape them is steadily and incrementally increasing, case study by case study.

But linguists thrive on new questions, of the sort that can lead to received ideas and methods being questioned or even jettisoned. Despite a narrative of progress from classical typology to present-day typology, with a possible paradigm shift from Greenberg’s Question (what is a possible or impossible language?) to Nichols-Bickel’s Question (what’s where why?),^[8] we see substantial conceptual continuity from the early days of phonological typology until today. It is not clear whether the way forward will emerge from new methods or combinations thereof, from new sources of data, or from a rethinking of foundational concepts like ‘phonological segment’ or ‘syllable.’ Will our problems be solved once we can induce phonological generalizations at multiple levels of abstraction directly from speech recordings? Will an ideal multidisciplinary constellation of traditional typology, experimental phonetics, computational modeling, and other sciences, tell us what’s where why? Perhaps, but it is possible that as we hone our existing tools and create new ones, truly new thoughts are out there waiting to surface and to be put into practice. It may even be, as one of Linguistic Typology’s editors suggests, that what’s where why has been the wrong question all along; this kind of suggestion is exactly the kind of stimulus that we need, pushing us to frame new questions that orient and contextualize our research.

So is phonological typology a fish swimming incessantly forward, sometimes awake and sometimes asleep, or is it a hitherto unknown amphibian, waiting to take on a new body-plan? We do not know. But we take this opportunity to call upon practitioners of phonological typology to continue improving our data, our methods, and to develop new theories, while being awake to the possibility of asking new and unsettling questions that can break our established patterns of thought and action.