[Lingtyp] papers on non-uniqueness in tone and stress

Erich Round erichslists at gmail.com
Sat Feb 6 13:28:08 UTC 2021


Hi all,

Martin, thanks for engaging on this topic.  In a few short sentences you raise multiple issues that deserve thought.  I’ll respond quickly to three and then come back to non-uniqueness and biases.

Quickly:

  *   You mention analysis. The key to non-uniqueness is that languages often give us multiple options for analysing them into their parts: do I group [a,b,c,d] vs [e,f] or I do group them [a,b,c] vs [d,e,f]? (This is what I loosely called “carving at joints”; it’s the stuff of classic structuralist analysis.) Often, good motivations will be available for multiple options, hence the potential for multiple analyses. Multivariate approaches like Canonical Typology and Multivariate Typology (Bickel 2010, 2011) can respond to this predicament by asking: in what ways do [a,b,c,d] group together in opposition to [e,f], and in what ways do [a,b,c] group together in opposition to [d,e,f]. (And note, CT and MT don’t do the carving, but they can enlighten us about how and why the multiple options arise). I find this more satisfying than decreeing that “the comparative concept is [a,b,c,d] so just use it”. However, your point was actually about how we label these subdivisions after we’ve created them. That’s an interesting topic, but I don’t think it’s directly relevant here.
  *   You seem newly concerned that Round (2019) might suffer biases. Indeed it might. But this would be equally true if my aim had been, as you first thought, to assign these languages to comparative concepts. Biases do warrant concern though (see below).
  *   You’re right I want to “have it both ways”, to treat languages as systems and compare them. This is a familiar red line that has separated your views from other typologists for a while now. I appreciate that you had an argument couched in philosophical terms that seemed to lead to your conclusion. Formulating such arguments is hard and valuable work, but in my view Spike (2020) demonstrates that the argument fails, predominantly because it’s based on premises that turn out to be false. This is fine; typology benefits from such debates: you raised some interesting problems; Spike engaged with them and showed them to be apparent, not real. The field advances.

On biases and non-uniqueness:

All scientific work is prone to biases, we all strive to avoid them, and so vigilance is ever necessary. One reason I’m so interested in non-uniqueness is that non-uniqueness is one route by which biases can creep into otherwise good typologising, and it can be hard work to guard against or even detect. Here’s an example.

In a pilot study, Round & Macklin-Cordes (2015) looked at the choices made by linguists, about whether to bundle nasal+stop sequences like [mb] into single ‘complex segment’ phonemes or to keep them as two-phoneme sequences in 312 doculects of Australia and New Guinea. Those choices could be predicted with an accuracy of 87%, based on (1) a set of fine-scale phonological and phonetic variables. But then, we could improve the prediction to 92% (i.e., remove one third of the remaining uncertainty!) if we additionally factored in (2) whether the languages were spoken on the Australian continent or not (irrespective of the facts of the sound system), and (3) whether the Author appeared to be SIL-trained. (For those interested: given the same sound system characteristics, SIL and/or New Guinea both favour the positing of complex segments; {Australia & not-SIL} favours sequences).

What (1), (2) and (3) suggest is that choices do follow the facts of the languages, but those facts at times are insufficient to produce a unique analysis on their own. Once non-uniqueness arises, our choices may then follow our communities of practice. These are sociological biases among linguists, not characteristics of languages. Importantly though, they’re not merely noise (which blurs the picture) but bias (which tilts the answer in one direction). If linguists’ communities of practice align with language families or areas, then what appears to be a genealogical/areal fact about languages may actual be, to some extent, a fact about linguists. An otherwise impeccably designed typological study which asked straight-up “are there complex segments /mb/” would fold these sociological issues in with the language facts.

Why does it matter?  Because we don’t want the results of our hard work to be undermined like this.  What can we do about it?  At the very least, remain mindful that non-uniqueness exists, and consider how it might be playing into our careful cross-linguistic comparisons.

Best,
Erich


Bickel, Balthasar. 2010. “Capturing Particulars and Universals in Clause Linkage.” In Clause Linking and Clause Hierarchy: Syntax and Pragmatics, edited by Isabelle Bril, 51–101. Amsterdam: Benjamins.
Bickel, Balthasar. 2011. “Multivariate Typology and Field Linguistics: A Case Study on Detransitivization in Kiranti (Sino-Tibetan).” In Proceedings of the Conference on Language Documentation and Linguistic Theory 3, edited by Peter K Austin, Oliver Bond, David Nathan, and Lutz Marten, 3–13. London: SOAS.
Round, Erich, and Jayden L. Macklin-Cordes. 2015. “On the Design, in Practice, of Typological Microvariables.” Conference Presentation presented at New Developments in the Quantitative Study of Languages, Helsinki.
Spike, Matthew. 2020. “Fifty Shades of Grue: Indeterminate Categories and Induction in and out of the Language Sciences.” Linguistic Typology 24 (3): 465–88. https://doi.org/10.1515<https://doi.org/10.1515/lingty-2020-2061>
 <https://doi.org/10.1515/lingty-2020-2061>
/lingty-2020-2061<https://doi.org/10.1515/lingty-2020-2061>



From: Lingtyp <lingtyp-bounces at listserv.linguistlist.org> on behalf of Martin Haspelmath <martin_haspelmath at eva.mpg.de>
Date: Saturday, 6 February 2021 at 3:23 am
To: "lingtyp at listserv.linguistlist.org" <lingtyp at listserv.linguistlist.org>
Subject: Re: [Lingtyp] papers on non-uniqueness in tone and stress

Thanks, Erich! Yes, it's true that comparative concepts "do not address the issue of non-unique analysis", because they are not about analysis, but about classification.

Linguists often conflate these two things – they say things like "I analyze phenomenon X as Y", where "Y" is a category known from some other language. But this works only if Y is assumed to be an innate building block.

So if one invokes phenomena that are also found in other languages, one should say "I classify X as Y". And I think that classification can and should be unique (unlike analysis), because the classificatory categories are defined in such a way that they apply equally everywhere. (They are "cookie-cutters" whose shape and size is not dependent on the dough, to use Erich's apt metaphor.)

What Round (2019) apparently attempts to do is to have it both ways: "carve languages at their joints", but in a way that still allows comparison, by "choosing judiciously" among different (equally possible) analyses. I fear that such "judicious choices" may introduce too much subjectiveness (criteria selection bias, as Tallman 2021 calls it: https://ling.auf.net/lingbuzz/005720).

Erich is absolutely right that by choosing "cookie-cutter"-type comparative concepts (e.g. WALS-type category-like concepts, or parallel text passages, or Nijmegen-style visual stimuli), one incurs "the cost of losing sight of the systems being characterised" – except that "losing sight" suggests that one could do both things simultaneously.

I'm not sure whether there is a "growing consensus", though. I replied to Himmelmann's 2019 paper here: https://dlc.hypotheses.org/2447, and I will reply to the critical comments on my "General linguistics" paper (mostly by generative linguists such as David Adger: https://ling.auf.net/lingbuzz/005691) next month. It still seems to me that the main lack of consensus is between advocates of innate building blocks and everyone else.

Martin
Am 05.02.21 um 08:37 schrieb Erich Round:
Hi all,

Martin Haspelmath writes,

different issues here: … (iii) how one links language-particular phenomena to comparative concepts; Erich Round's paper on “Australian Phonemic Inventories Contributed to PHOIBLE 2.0” https://doi.org/10.5281/zenodo.3464333 is a clear example of this last type.

This misconstrues what the study does altogether, but it also raises a point worth delving into, so thanks to Martin in the spirit of zigzagging towards the light:

Round (2019) considers the language-internal analyses of a large sample of Australian languages. Relevant to Adam’s topic, phonemic analysis is famously non-unique, and a given language typically allows multiple possible analyses. To put it another way, there are multiple ways to carve a sound system at its joints. Round (2019) chooses among these multiple, possible, language-internal analyses, endeavouring to ensure that the principles of the analysis are comparable across languages. Thus, when languages in the dataset do differ, those differences are more likely to reflect empirical differences in the languages themselves, rather than artifactual differences due to linguists doing phonemicisation differently. As Larry Hyman (2017:144) puts it so well, “we aim to typologize the linguistic properties, not the linguists”. Discussion of this kind of typological data preparation is relatively prominent within phonology: see Hyman (2017), van der Hulst (2017), Kiparksy (2018), and of course Ian Maddieson’s (1984) classic study which treats the issue with great care.

In contrast, Comparative Concepts (CC’s) do not seem to me to address the issue of non-unique analysis, because they don’t seek to characterise languages in their own terms. CC’s, in Haspelmath’s sense, are like cookie-cutters that slice through languages, picking out a predetermined shape chosen by the analyst (so that we can ask what we find within it); they deliberately don’t carve languages at their joints. Round (2019) does carve languages at their joints, only it admits that there are many ways to do so, and attempts to choose judiciously among them, given the aim of constructing a dataset that aids insightful comparison and typologising.

Why is this important to many typologists?  Because we regard languages as organised systems, and want to understand them as such. A challenge, though, is that complex systems admit of multiple different characterisations.  Typological methods which attend to this challenge of multiple analysis / non-uniqueness seek to ameliorate the distractions and illusions that can be thrown up by different choices of analysis, while still remaining committed to studying the system.  CC’s in Haspelmath’s sense offer the promise of reducing variation in analysis, but at the cost of losing sight of the systems being characterised.  Whether that is a cost worth bearing is evidently still a matter of contention in current-day typology.*

Best,
Erich


* Though the growing consensus is, it’s perfectly possible to do good typology without paying it. Bickel 2021, Round & Corbett 2020, Spike 2020 and Himmelmann 2019 all make this point from separate angles.

New references (other references are in my earlier post, below):
Bickel, Balthasar. 2021. “Beyond Universals and Particulars in Language: A Reply to Haspelmath (2021).” https://ling.auf.net/lingbuzz/005707.
Himmelmann, Nikolaus P. 2019. Against trivializing language description (and comparison). Under review.
https://ifl.phil-fak.uni-koeln.de/prof-himmelmann/publikationen
Maddieson, Ian. 1984. Patterns of Sounds. Cambridge: Cambridge University Press.
Spike, Matthew. 2020. “Fifty Shades of Grue: Indeterminate Categories and Induction in and out of the Language Sciences.” Linguistic Typology 24(3):465-488. https://doi.org/10.1515/lingty-2020-2061


From: Lingtyp <lingtyp-bounces at listserv.linguistlist.org><mailto:lingtyp-bounces at listserv.linguistlist.org> on behalf of Martin Haspelmath <martin_haspelmath at eva.mpg.de><mailto:martin_haspelmath at eva.mpg.de>
Date: Thursday, 4 February 2021 at 11:32 pm
To: "lingtyp at listserv.linguistlist.org"<mailto:lingtyp at listserv.linguistlist.org> <lingtyp at listserv.linguistlist.org><mailto:lingtyp at listserv.linguistlist.org>
Subject: Re: [Lingtyp] papers on non-uniqueness in tone and stress

It seems that there are at least three different issues here:

(i) whether all speakers of a language have the same system even when their conventional behaviour is identical; there happens to be an example of indeterminacy in the latest issue of Phonological Data and Analysis (see Matthew Gordon's earlier message):
Bennett, W. G., & Braver, A. (2020). Different speakers, different grammars: Productivity and representation of Xhosa labial palatalization. Phonological Data and Analysis, 2(6), 1–29. https://doi.org/10.3765/pda.v2art6.9

(ii) on what basis one decides between different analyses of a language-particular system; e.g. Schane's (1968) example of English [spin], which can be phonemicized as /sbin/ (with phonetic devoicing of /b/ after sibilant) or /spʰin/ (with phonetic deaspiration in the same environment).

(iii) how one links language-particular phenomena to comparative concepts; Erich Round's paper on “Australian Phonemic Inventories Contributed to PHOIBLE 2.0” https://doi.org/10.5281/zenodo.3464333 is a clear example of this last type. It seems that the issue in Chácobo that Adam Tallman mentioned ("tone" vs. "stress") also falls in this category.

Phonologists do not always distinguish between (ii) and (iii) (particular description vs. general comparison), as pointed out prominently by Lass (1984) and Simpson (1999) (cited by Erich). But Kiparsky (2018) (also cited by Erich) explicitly rejects the distinction – I have argued against Kiparsky here: https://dlc.hypotheses.org/1817.

Best,
Martin

Am 04.02.21 um 13:28 schrieb Erich Round:
Hi Adam,

I’ve enjoyed the conversations you’ve sparked here on the list recently, please keep them coming!

Thanks for raising an important topic.  I have some paper suggestions below.  I’d start by saying, though, that you might be getting formal phonologists wrong.  Generative theorists from the start were well aware of the non-uniqueness problem, and that’s one reason why they were so keen on metrics to evaluate multiple candidate grammars.  Now, that’s not to say it proved to be plain sailing, but there’s a deep appreciation of the problem buried in the theory, even if for practical purposes much theoretical work (just like much typological work) assumes only one analysis in order to get some other task completed in a finite amount of time.  In optimality theory, the notion of Richness of the base is one new-ish incarnation of attempts to deal with the matter.

Canonical Typology (Corbett 2005, Round and Corbett 2020) provides the conceptual tools for asking not just whether ‘the best analysis’ is A, B or C, but to what extent, in multiple different regards, A, B and C differ and therefore can be considered (dis)advantageous in different ways. This helps us clarify why and how multiple analyses arise in the first place. My forthcoming chapter (2021) on phonotactics in Australian languages discusses this with respect to complex segments; Kwon & Round (2015) discuss it with respect to phonaesthemes; my review (2017) of Gordon’s Phonological Typology (2016) discusses the idea of doing typology over a distribution of possible analyses (which I term ‘factorial analysis’) and points out some places where Gordon’s own work covertly does this when confronted with non-uniqueness. Parncutt (2015) applies the idea to reduplication, and a current PhD student of mine, Ruihua Yin presented some of her fascinating results regarding sonority sequencing at the Australian Linguistics Society conference in December; her thesis should be finished early this year, and will be a major undertaking in this kind of typology. Round (2019) discusses how I addressed the issue of non-uniqueness when compiling a typologically nuanced set of 400 Australia phoneme inventories for Phoible. Natalia Kuznetsova’s work (2019) is relevant to prosody and responds to Hyman’s (2006) classic paper. Other serious discussions of the issue from various angles, typically very thoughtful and some quite in-depth are: Hockett 1963, Lass 1984, Simpson 1999, Hyman 2007, 2008, 2017, Dresher 2009, van der Hulst 2017, Kiparksy 2018.

Best,
Erich

Corbett, Greville G. 2005. “The Canonical Approach in Typology.” In Linguistic Diversity and Language Theories, edited by Zygmunt Frajzyngier, Adam Hodges, and David S Rood, 25–49. Amsterdam: John Benjamins.
Dresher, B. Elan. 2009. The Contrastive Hierarchy in Phonology. Cambridge: Cambridge University Press.
Gordon, Matthew K. 2016. Phonological Typology. Oxford University Press.
Hockett, Charles F. 1963. “The Problem of Universals in Language.” In Universals of Language, edited by Joseph Greenberg, 1–29.
Hyman, Larry. 2006. “Word-Prosodic Typology.” Phonology 23: 225–57.
Hyman, Larry M. 2007. “Where’s Phonology in Typology?” Linguistic Typology 11: 265–71.
Hyman, Larry M. 2008. “Universals in Phonology.” The Linguistic Review 25: 83–137.
Hyman, Larry M. 2017. “What (Else) Depends on Phonology?” In Dependencies in Language, edited by Nicholas Enfield, 141–58.
Kiparsky, Paul. 2018. “Formal and Empirical Issues in Phonological Typology.” In Phonological Typology, edited by Larry M. Hyman and Frans Plank, 54–106. Berlin: De Gruyter Mouton.
Kuznetsova, Natalia. 2019. What Danish and Estonian can show to a modern word-prosodic typology. In Goedemans, R., Heinz, J., & van der Hulst, H. (Eds.). The study of word stress and accent: Theories, methods and data. CUP.
Kwon, Nahyun, and Erich R. Round. 2015. “Phonaesthemes in Morphological Theory.” Morphology 25 (1): 1–27.
Lass, Roger. 1984. “Vowel System Universals and Typology: Prologue to Theory.” Phonology Yearbook 1: 75–111.
Parncutt, Amy. 2015. “Towards a Phonological Typology of Reduplication in Australian Languages.” Honours Thesis, University of Queensland.
Round, Erich R. 2017. “Review of Gordon, Matthew K. Phonological Typology, OUP 2016.” Folia Linguistica 51 (3): 745–55.
Round, Erich R. 2019. “Australian Phonemic Inventories Contributed to PHOIBLE 2.0: Essential Explanatory Notes.” https://doi.org/10.5281/zenodo.3464333.
Round, Erich R. forthcoming 2021. “Phonotactics.” In Oxford Guide to Australian Languages, edited by Claire Bowern. Oxford: Oxford University Press. DOI: 10.13140/RG.2.2.23022.13120
Round, Erich R., and Greville G. Corbett. 2020. “Comparability and Measurement in Typological Science: The Bright Future for Linguistics.” Linguistic Typology 24 (3): 489–525.
Simpson, Adrian P. 1999. “Fundamental Problems in Comparative Phonetics and Phonology: Does UPSID Help to Solve Them.” In Proceedings of the 14th International Congress of Phonetic Sciences, 1:349–52. Berkeley: University of California.
Van der Hulst, Harry. 2017. “Phonological Typology.” In The Cambridge Handbook of Linguistic Typology, edited by Alexandra Y. Aikhenvald and Robert MW Dixon, 39–77. Cambridge: Cambridge University Press.


From: Lingtyp <lingtyp-bounces at listserv.linguistlist.org><mailto:lingtyp-bounces at listserv.linguistlist.org> on behalf of TALLMAN Adam <Adam.TALLMAN at cnrs.fr><mailto:Adam.TALLMAN at cnrs.fr>
Date: Thursday, 4 February 2021 at 9:20 pm
To: VAN DE VELDE Mark <Mark.VANDEVELDE at cnrs.fr><mailto:Mark.VANDEVELDE at cnrs.fr>, "lingtyp at listserv.linguistlist.org"<mailto:lingtyp at listserv.linguistlist.org> <lingtyp at listserv.linguistlist.org><mailto:lingtyp at listserv.linguistlist.org>
Subject: Re: [Lingtyp] papers on non-uniqueness in tone and stress

Thanks, yes, I've read this paper.

Adam





Adam James Ross Tallman (PhD, UT Austin)
ELDP-SOAS -- Postdoctorant
CNRS -- Dynamique Du Langage (UMR 5596)
Bureau 207, 14 av. Berthelot, Lyon (07)
Numero celular en bolivia: +59163116867
________________________________
De : Lingtyp [lingtyp-bounces at listserv.linguistlist.org<mailto:lingtyp-bounces at listserv.linguistlist.org>] de la part de Mark Van de Velde [mark.vandevelde at cnrs.fr<mailto:mark.vandevelde at cnrs.fr>]
Envoyé : jeudi 4 février 2021 11:57
À : lingtyp at listserv.linguistlist.org<mailto:lingtyp at listserv.linguistlist.org>
Objet : Re: [Lingtyp] papers on non-uniqueness in tone and stress

Dear Adam:

I can recommend Hyman (2012).

All the best,

Mark

Hyman, Larry M. 2012. In defense of prosodic typology: A response to Beckman and Venditti. Linguistic Typology. De Gruyter Mouton 16(3). 341–385. https://doi.org/10.1515/lity-2012-0014.




On 04/02/2021 11:12, TALLMAN Adam wrote:
Hello all,

I'm looking for papers on the notion of non-uniqueness in phonology (or morphosyntax if applicable). I have three so far (Chao, Hockett, and Schane).

I'm particularly interesting in non-uniqueness in the domain of the description of suprasegmentals - like when we have a system that seems to mix tone and (other types of) prominence whether the system should be described as tonal with a stress mapped to it or vice versa. Phonologists discuss the issue as if there is an obvious unique best way of describing such relations in all cases. But I think that's probably false and it choosing one over the other just amounts to an expositional decision - some of  the discussion in Tallman and Elias-Ulloa (2020) point in this direction in Chácobo.

There's also the related issue of when the acoustic correlates of some phonological category are organized in such a way as to genuinely merit the designation "tone". Phonologists seem to assume that this issue is trivial or obvious - again, I think this is probably false (the notion is more open ended than is recognized) regardless of the phonological evidence that can be rallied in support of one position or another.

@Article{chao:1934:phonemes,
    title = {The non-uniqueness of phonemic solutions of phonetic systems},
    author = {Yuen Ren Chao},
    journal = {Bulletin of the Institute of History and Philology, Academia Sinica},
    year = {1934},
    volume = {4},
    number = {},
    pages = {363-397},
    %doi = {},
    %urldate = {},
}

@incollection{hockett:1963:universals,
    Author = {Charles F. Hockett},
    Booktitle = {Universals of language (Volume 2)},
    Editor = {Joseph H. Greenberg},
    Pages = {1-29},
    Publisher = {MIT Press},
    Address = {Cambridge, MA},
    Title = {The problem of universals in language},
    Year = {1963},
    Edition = {}}

@Article{schane:1968:nonuniqueness,
    title = {On the non-uniqueness of phonological representations},
    author = {Sanford A. Schane},
    journal = {Language},
    year = {1968},
    volume = {44},
    number = {4},
    pages = {363-397},
    %doi = {},
    %urldate = {},
}

@Article{tallman:eliasulloa:2020:acoustics,
    title = {The acoustic correlates of stress and tone in Chácobo (Pano)},
    author = {Adam J.R. Tallman},
    journal = {The acoustic correlates of stress and tone in Chácobo (Pano): A production study},
    editor = {Adam J.R. Tallman and José Élias-Ulloa},
    year = {2020},
    volume = {147},
    number = {4},
    pages = {3028},
    doi = {https://doi.org/10.1121/10.0001014},
    %urldate = {2019-07-04},
}

Adam





Adam James Ross Tallman (PhD, UT Austin)
ELDP-SOAS -- Postdoctorant
CNRS -- Dynamique Du Langage (UMR 5596)
Bureau 207, 14 av. Berthelot, Lyon (07)
Numero celular en bolivia: +59163116867


_______________________________________________

Lingtyp mailing list

Lingtyp at listserv.linguistlist.org<mailto:Lingtyp at listserv.linguistlist.org>

http://listserv.linguistlist.org/mailman/listinfo/lingtyp
--
[Image removed                                   by sender. LLACAN]
Mark Van de Velde
Directeur du LLACAN (CNRS-INaLCO)
mark.vandevelde.cnrs.fr<https://mark.vandevelde.cnrs.fr>
bantu.cnrs.fr<https://bantu.cnrs.fr>



_______________________________________________

Lingtyp mailing list

Lingtyp at listserv.linguistlist.org<mailto:Lingtyp at listserv.linguistlist.org>

http://listserv.linguistlist.org/mailman/listinfo/lingtyp


--

Martin Haspelmath

Max Planck Institute for Evolutionary Anthropology

Deutscher Platz 6

D-04103 Leipzig

https://www.shh.mpg.de/employees/42385/25522



_______________________________________________

Lingtyp mailing list

Lingtyp at listserv.linguistlist.org<mailto:Lingtyp at listserv.linguistlist.org>

http://listserv.linguistlist.org/mailman/listinfo/lingtyp



--

Martin Haspelmath

Max Planck Institute for Evolutionary Anthropology

Deutscher Platz 6

D-04103 Leipzig

https://www.shh.mpg.de/employees/42385/25522
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.linguistlist.org/pipermail/lingtyp/attachments/20210206/19bf8eaa/attachment.htm>


More information about the Lingtyp mailing list