1 of 19

Sampling tonal languages��Dmitry Gerasimov, INALCO�

​

ThoT

Nov. 16, 2023

�

1st ThoT workshop, Paris, Nov. 30 – Dec. 1, 2023

2 of 19

Sampling tonal languages

  1. What is the general set of all tonal languages of the world?
  2. How can we build a representative sample thereof �for the purposes of the project?

ЗАГОЛОВОК ПРЕЗЕНТАЦИИ

2

02.02.20ГГ

3 of 19

The universe of tonal languages

  • “Modern standard Chinese, like nearly half of the world’s languages, uses pitch patterns to distinguish between one word and another.” [Ashby & Maidment 2005: 163]
  • “Probably all languages have structural intonation. However, only about half have lexical tone,..” [Gussenhoven 2004: 12]

​

​

ЗАГОЛОВОК ПРЕЗЕНТАЦИИ

3

02.02.20ГГ

4 of 19

The universe of tonal languages

  • “Tell someone you are writing a book about ‘Tone’, and they look blank, and yet by some estimates as much as 60–70 per cent of the world’s languages are tonal.” [Yip 2002: 1]
  • “Of the 6520 languages in the families listed in Table 2, the indications are that 3044, or 46.7 %, are tonal. <…> Moreover, given the way that the sampling was conducted this figure is more likely to be an over-estimate…” [Maddieson 2023: 12]

​

​

ЗАГОЛОВОК ПРЕЗЕНТАЦИИ

4

02.02.20ГГ

5 of 19

WALS

No tones: 307

Simple: 132

Complex: 88

​

​

ЗАГОЛОВОК ПРЕЗЕНТАЦИИ

5

02.02.20ГГ

6 of 19

LAPSyD

No tones: 503

Marginal: 8

Simple: 119

Moderately �complex: 53

Complex: 68

​

​

ЗАГОЛОВОК ПРЕЗЕНТАЦИИ

6

02.02.20ГГ

7 of 19

Larry Hyman’s database

665 tonal languoids

​

​

ЗАГОЛОВОК ПРЕЗЕНТАЦИИ

7

02.02.20ГГ

8 of 19

Larry Hyman’s database

665 tonal languoids,

of which I’ve been able�to code 627

99 top-level families�(per glottolog),�inc. 60 represented �by 1 language�

​

​

ЗАГОЛОВОК ПРЕЗЕНТАЦИИ

8

02.02.20ГГ

9 of 19

[Hammarström & Mannby, forthc.]

  • Over 10K digitized sources on 4587 languages.
  • Automated search for “tone”, “tonal”, etc. + their translational equivalents.
  • Number of hits in the source > certain threshold => Tone = TRUE.
  • The majority of sources for each language.
  • Manual check of 600 languages.

​

​

ЗАГОЛОВОК ПРЕЗЕНТАЦИИ

9

02.02.20ГГ

10 of 19

[Hammarström & Mannby, forthc.]

  • 1865 languages identified as “tonal” (ca. 40 % out of 4587).
  • Manual check of 600 languages returned a 9% margin of error.
    • 28 false positives
    • 26 false negatives

​

​

ЗАГОЛОВОК ПРЕЗЕНТАЦИИ

10

02.02.20ГГ

11 of 19

Language sampling

  • Convenience sample: grabbing whatever languages are within your reach;
  • Variety sample: capturing as much linguistic diversity as possible;
  • Probability sample: testing quantitative hypotheses;
    • e.g.: in terms of TDI, do languages cluster around the four loci of omnisyllabic, tonemic, etc., or is it a gradual continuum?

​

​

​

ЗАГОЛОВОК ПРЕЗЕНТАЦИИ

11

02.02.20ГГ

12 of 19

Language sampling

  • We primarily want a variety sample;
  • While still keeping an eye open for a potential probability (sub)sample; �(cf. [Dahl 2008; Miestamo 2016; Guzmán Naranjo & Becker 2022; inter alia].
  • However, lack of sufficient descriptions may push us towards something more like a convenience sample.

​

​

ЗАГОЛОВОК ПРЕЗЕНТАЦИИ

12

02.02.20ГГ

13 of 19

Building our sample

​

​

​

  • Compiling information from WALS, LAPSyD, LH and HH&EM.
  • Special attention to cases when different databases disagree.
  • Removing apparent false positives from HH&EM.

​

​

​

​

ЗАГОЛОВОК ПРЕЗЕНТАЦИИ

13

02.02.20ГГ

True

​

​

Bali (Democratic Republic of Congo)

bcp

Atlantic-Congo

True

Complex

Y

Bench

bcq

Ta-Ne-Omotic

False

​

​

Babine

bcr

Athabaskan-Eyak-Tlingit

True

Moderately

Kohumono

bcs

Atlantic-Congo

True

​

Y

Awad Bing

bcu

Austronesian

True

​

​

Bana

bcw

Afro-Asiatic

14 of 19

Building our sample

  • The databases give us over 160 first-level families (per glottolog) presumably containing tonal languages.
  • We also decided to split major tone-rich families (AtlCongo, AfrAs, SinTib, NucTNG), bringing the total number of taxa well over 220.

​

​

​

​

ЗАГОЛОВОК ПРЕЗЕНТАЦИИ

14

02.02.20ГГ

15 of 19

Building our sample

  • But:
    • Some taxa prove to only contain false positives from HH&EM;
    • Some small taxa prove to be crucially underdescribed;
    • At some point it makes sense to convert to a less-fine grained genetic classification like that of Ethnologue (cf. [Miestamo 2016]).

​

​

​

​

ЗАГОЛОВОК ПРЕЗЕНТАЦИИ

15

02.02.20ГГ

16 of 19

Building our sample

  • We also decided to split major tone-rich families (AtlCongo, AfrAs, SinTib, NucTNG), bringing the total number of taxa well over 220.
  • Of these, about 80 contain a single presumably tonal language.
  • The number of smaller taxa that make it into the sample determines how many “slots” we need to distribute among larger, more diverse taxa.

​

​

​

​

ЗАГОЛОВОК ПРЕЗЕНТАЦИИ

16

02.02.20ГГ

17 of 19

Building our sample

  • A manual check of 25 “one-language taxa”:
    • 7 are “false positives”;
    • 6 we can’t use because of scarcity of description;
    • 7 are suitable;
    • 5 are pending further consideration.

​

​

​

ЗАГОЛОВОК ПРЕЗЕНТАЦИИ

17

02.02.20ГГ

18 of 19

References

  • Ashby, M. & Maidment, J. (2005). Introducing Phonetic Science. CUP.
  • Dahl, Ö. (2008). An exercise in a posteriori language sampling // STUF 61/3: 208-220.
  • Gussenhoven, C. (2004). The Phonology of Tone and Intonation. Cambridge: CUP.
  • Guzmán Naranjo, M. & Becker, L. (2022). Statistical bias control in typology // Linguistic Typology 26/3, 605-670.
  • Hammarström, H. & Mannby, E. (forthc). On the world-wide distribution of tonal languages.
  • Maddieson, I. (2013) Tone. Chapter #13A in WALS.

​

​

ЗАГОЛОВОК ПРЕЗЕНТАЦИИ

18

02.02.20ГГ

19 of 19

References

  • Maddieson, I. (2023) Tone is not predominant; Tone is not primordial // Proceedings of ICPhS 2023.
  • Miestamo, M., Bakker, D. & Arppe, A. (2016) Sampling for variety // Linguistic Typology 20/2, 233-296.
  • Yip, M. (2002). Tone. Cambridge: CUP.

​

ЗАГОЛОВОК ПРЕЗЕНТАЦИИ

19

02.02.20ГГ