1 of 60

Tone adaptation in loanwords:� from Chinese Mandarin to Nuosu Yi

Yao Zhang

Cornell University, Department of Linguistics

04/06/2025

PLC 49

2 of 60

Takeaway

​

  • In Nuosu Yi loanwords, there is an overrepresentation of low tones and an underrepresentation of high tones.
  • Phonetic similarity to the borrowed language is significant in tone adaptation as in segmental adaptation, but is not necessarily preserved even when it’s not prevented by native markedness constraints.
  • Loanwords can amplify the markedness restrictions in native phonology of the borrowing language.

​

2

3 of 60

Why do we care about…

  • Tone adaptation

​

  • Nuosu Yi tone adaptation

3

4 of 60

Why do we care about Tone Adaptation

In loanword phonology,

  • loanwords tolerate more markedness violations than native words.
  • the tolerance often appears in sub-lexicons.

​

​

​

​

​

​

​

​

​

Inkelas et al. 1996; Itô & Mester 2017; Jurgec 2010; Simonovic 2012, 2015

4

5 of 60

Why do we care about Tone Adaptation

In loanword phonology,

  • loanwords tolerate more markedness violations than native words.
  • the tolerance often appears in sub-lexicons.

​

e.g.

English gyoza← Japanese gyooza ‘dumpling’

​

​

​

​

Inkelas et al. 1996; Itô & Mester 2017; Jurgec 2010; Simonovic 2012, 2015

5

6 of 60

Why do we care about Tone Adaptation

In loanword phonology,

  • loanwords tolerate tolerates more markedness violations than native words.
  • the tolerance often appears in sub-lexicons.

​

e.g.

English gyoza← Japanese gyooza ‘dumpling’

No /gy-/ in CMU Pronouncing Dictionary—

​

​

​

Inkelas et al. 1996; Itô & Mester 2017; Jurgec 2010; Simonovic 2012, 2015

6

gy

7 of 60

Why do we care about Tone Adaptation

In loanword phonology,

  • loanwords tolerate tolerates more markedness violations than native words.
  • the tolerance often appears in sub-lexicons.

​

e.g.

English gyoza← Japanese gyooza ‘dumpling’

No /gy-/ in CMU Pronouncing Dictionary—

So marked, so foreign!!!

Yet still retained in English.

Inkelas et al. 1996; Itô & Mester 2017; Jurgec 2010; Simonovic 2012, 2015

7

gy

8 of 60

Why do we care about Tone Adaptation

In loanwords phonology,

  • Segmental adaptation tolerates more markedness violations than native words.
  • The tolerance often appears in sub-lexicon.

​

  • Suprasegmental adaptation may “retreat to the unmarked”: non-discriminate assignment of a default tone.
  • Tone-Tone adaptation is understudied.

​

​

​

​

​

Inkelas et al. 1996; Itô & Mester 2017; Jurgec 2010; Simonovic 2012, 2015; Kang 2010; Kenstowicz, 2005

​

8

9 of 60

Why do we care about Nuosu Yi

  • Uniqueness:

East Asian tonal languages retreat to the unmarked by ignoring phonetic similarity and assigning suprasegmental feature based on default mechanisms (Kang 2010).

​

​

​

​

​

​

​

​

​

​

Kang 2010; Pan 1990 ; Zhu 1997, 1999

9

10 of 60

Why do we care about Nuosu Yi

  • Uniqueness:

East Asian tonal languages retreat to the unmarked by ignoring phonetic similarity and assigning suprasegmental feature based on default mechanisms(Kang 2010).

  • Possibility:

Nuosu Yi contains a large number of Chinese loanwords due to long-term, deep interaction with Chinese Mandarin.

​

​

​

​

​

​

​

​

Kang 2010; Pan 1990 ; Zhu 1997, 1999

10

11 of 60

Introduction to Nuosu Yi

​

A Sino-Tibetan tonal language spoken in China

​

11

12 of 60

Nuosu Yi as a tonal language

12

  • Nuosu Yi:
  • Spoken mainly in Liangshan, Sichuan, China (and in northern Yunnan)
  • Sino-Tibetan

Tibeto-Burman

Lolo–Burmese

Loloish

Yi

13 of 60

Tone system in Nuosu Yi

13

Tone

Native

Loan

55

tʰi55 kɔ 33 ‘here’

ʦʰv̩55 ‘vinegar’

44

ŋɔ21 wɔ44 ‘pl marker’

hɔ44 sə33 ‘peanut’

33

ŋa33 ‘I’

ʧʰə33 ‘car’

21

kʰə21 mv̩ 33‘how’

ta21 ‘(political) party’

14 of 60

Tone adaptation patterns in Nuosu Yi

​

Similarity does not always work

​

14

15 of 60

Source tones and adaptations

  • Nuosu Yi tones: 55, 44, 33, 21
  • Sichuan Mandarin, Chengdu dialect tones: 45, 42, 21, 213

15

16 of 60

Source tones and adaptations: multiple patterns

  • Nuosu Yi tones: 55, 44, 33, 21
  • Chengdu Mandarin tones: 45, 42, 21, 213

Tone adaptation

​

​

​

​

​

​

​

​

​

​

​

​

Yang 1951

​

16

Sichuan Mandarin

Nuosu Yi

Example

Count

45

33

‘ampere’ 安培 ŋan45 pei22 → ŋa33 pʰi21

124

21

‘differential’ 微分 wei45 fən44 → wo21 fi33

10

55

‘tomato’ 番茄 fɑn45 ʨʰiɛ21 → fa55 ʨʰi21

2

44

‘corn’ 包谷 pɑu45 ku22 → pu44 kv̩21

1

17 of 60

Source tones and adaptations: mismatch in freq

  • Nuosu Yi tones: 55, 44, 33, 21
  • Chengdu Mandarin tones: 45, 21, 42, 213

Tone adaptation

​

17

Sichuan Mandarin

Nuosu Yi

Example

Count

45

33

‘ampere’ 安培 ŋan45 pei22 → ŋa33 pʰi21

124

21

‘differential’ 微分 wei45 fən44 → wo21 fi33

10

55

‘tomato’ 番茄 fɑn45 ʨʰiɛ21 → fa55 ʨʰi21

2

44

‘corn’ 包谷 pɑu45 ku22 → pu44 kv̩21

1

18 of 60

Match vs. Mismatch

18

Corpus: 463 disyllabic loanwords extracted from Han-Yi Dictionary (1979/1989)

19 of 60

Match vs. Mismatch

19

Corpus: 463 disyllabic loanwords extracted from Han-Yi Dictionary (1979/1989)

Match in both pitch level and contour pattern

​

  • More mismatch in 45 and 42
  • More match in 21 and 213

20 of 60

Tone Frequency in Nuosu Yi Loanwords

20

Corpus: 463 disyllabic loanwords extracted from Han-Yi Dictionary (1979/1989)

21 of 60

Tone Frequency in Nuosu Yi Loanwords

21

Corpus: 463 disyllabic loanwords extracted from Han-Yi Dictionary (1979/1989)

  • More usage of 21 and 33
  • Less usage of 55 and 44

22 of 60

Why…

22

  • …mismatch?
  • …imbalance between match and mismatch in different source tones?
  • …so many 21?
  • …so few 55 and 44?

23 of 60

Native Tone Restrictions in Nuosu Yi

​

33 > 21 > 55 > 44

​

23

24 of 60

Tone Freq in Native Nuosu Yi

Corpus: 12210 disyllabic native words extracted from The Grand Yi-Chinese Dictionary (2008)

​

24

​

​

​

25 of 60

Tone Freq in Native Nuosu Yi

25

26 of 60

Tone Freq in Native Nuosu Yi

26

  • Strong preference for 33

​

33 results from the merger of two historical non-creaky tones and one high stopped tone.

​

​

​

​

​

​

​

​

​

​

​

​

Bradley 1990, Lama 1991

27 of 60

Tone Freq in Native Nuosu Yi

27

  • Slight dispreference for 44
  • Dispreference for 44 on the 2nd syl

​

​

44 is a secondary tone resulted from syntactic derivation, later fossilized in sublexicon in specific word position.

​

​

Bradley 1990, Lama 1991

​

28 of 60

Tone Freq in Native Nuosu Yi

28

  • Dispreference for 55 and 21

​

55 and 21 in native Nuosu Yi lexicons have tonotactically strict distribution, as 55 only appear on “tense” vowels and 21 on “lax” vowels.

​

​

​

Bradley 1990, Lama 1991

29 of 60

Tone Freq Differs in Native and Loaned words

29

  • 33-preference in native words is retained in loanwords;

​

​

​

30 of 60

Tone Freq Differs in Native and Loaned words

30

  • 33-preference in native words is retained in loanwords;
  • Sliɡht 21-preference in native words is magnified in loanwords;

​

​

31 of 60

Tone Freq Differs in Native and Loaned words

31

  • 33-preference in native words is retained in loanwords;
  • Sliɡht 21-preference in native words is magnified in loanwords;
  • Slight 55/44-dispreference in native words is magnified in loanwords.

​

32 of 60

Factors shaping Nuosu Yi

Tone Adaptation Patterns

​

Markedness + Faithfulness

Markedness outweighs Faithfulness

​

32

33 of 60

Input and Constraints

  • Chengdu Mandarin surface tone: 45~55/21~22/42~44/213~13~21~22

​

  • Constraints:

A. Faithfulness (Ident: The input and output should be identical)

      • Ident(high)
      • Ident(low)
      • Ident(leftedge)
      • Ident(rightedge)

​

B. Markedness (specific restriction on tones)

      • *55, *44, *33, *21: No certain tone shall appear as output.

33

34 of 60

MaxEnt Model

​

  • Weighted constraints are used to assign probabilities to outputs.

​

​

​

​

​

​

​

​

​

​

​

​

Goldwater & Johnson 2003; Hayes & Wilson 2008

34

35 of 60

MaxEnt Model ↔ Classic OT

​

  • Weighted constraints are used to assign probabilities to outputs.
  • Weight of constraints in MaxEnt ≈ Ranking in Classic OT

(higher weight in MaxEnt ≈ higher ranking in Classic OT)

  • Higher Δlog-likelihood = larger effect

​

​

​

​

​

Hayes, Wilso 2008

35

36 of 60

Results of MaxEnt Model

36

Loanwords:

37 of 60

Addressing faithfulness

37

Nuosu Yi:

​

  • prefers to retain low pitch

38 of 60

Addressing faithfulness

38

Nuosu Yi:

​

  • prefers to retain low pitch.
  • prefers to retain the left edge pitch of a contour tone.

​

39 of 60

Addressing faithfulness

39

Nuosu Yi:

​

  • prefers to retain low pitch
  • prefers to retain the left edge pitch of a contour tone

​

  • …due to the existence of low tones in the source language and the lack of rising contour in the recipient language.

40 of 60

Addressing markedness

40

In Nuosu Yi:

​

  • 44 is the least preferred tone.

​

41 of 60

Addressing markedness

41

In Nuosu Yi:

​

  • 44 is the least preferred tone.
  • 55 is the second least preferred tone.

​

​

42 of 60

Addressing markedness

42

In Nuosu Yi:

​

  • 44 is the least preferred tone.
  • 55 is the second least preferred tone.
  • 33 is the most preferred tone while 21 the second.

43 of 60

Comparing Faithfulness and Markedness

43

  • Markedness >> Faithfulness

​

  • *44 >> *55 >> Ident(low) >> Ident (leftedge) >> *21

44 of 60

Comparing with Native Markedness

  • Constraints: markedness only

​

  • Result:

​

44

45 of 60

Comparing with Native Markedness

In the loanword model,

  • 44 remains the most marked tone.

45

46 of 60

Comparing with Native Markedness

In the loanword model,

  • there is a relatively larger difference in the weights of the most marked tone (44) and the least marked tone (33).

46

47 of 60

Comparing with Native Markedness

In the loanword model,

  • markedness of 55 is amplified, while markedness of 21 is diminished.

47

48 of 60

Comparing with Native Markedness

​

​

48

In the loanword model,

  • there is a clear tendency towards the markedness of high tones.

49 of 60

Explaining different distribution

49

  • High tones are generally avoided, but more in loanwords than in native words.

​

50 of 60

Explaining different distribution

50

  • High tones are generally avoided, but more in loanwords than in native words.
  • 33 is preferred in both native and loan words due to its “naturalness”.

​

51 of 60

Explaining different distribution

51

  • High tones are generally avoided, but more in loanwords than in native words.
  • 33 is preferred in both native and loan words due to its “naturalness”.
  • The large number of 21 in Yi loanwords results from the larger number of source tones in Chengdu Mandarin, shaped by both faithfulness to pitch level/pattern and restrictions of high tones.

​

52 of 60

Explaining the mismatch

52

Why…

  • mismatch?
  • …imbalance between match and mismatch in different source tones?
  • …so many 21?
  • …so few 55 and 44?

​

​

  • Markedness of high tones outweighs faithfulness to surface similarities in loanwords.

​

53 of 60

Conclusions

​

2 strategies, 2 mechanisms

​

​

53

54 of 60

Conclusions on Nuosu Yi Tone Adaptation

In loanwords, Nuosu Yi…

  • underrepresents high tones
  • overrepresents low tones

​

54

55 of 60

Conclusions on Nuosu Yi Tone Adaptation

Underrepresentation of high tones results from:

  • amplified markedness of 44 and 55;
  • retreatment to the unmarked.

​

​

55

56 of 60

Conclusions on Nuosu Yi Tone Adaptation

Underrepresentation of high tones results from:

  • amplified markedness of 44 and 55;
  • retreatment to the unmarked.

​

Overrepresentation of low tones results from:

  • faithfulness to low pitch level and left edge pitch;
  • diminished markedness of 21.

​

​

56

57 of 60

Conclusions on Loanword Phonology

  • Loanwords do not necessarily violate more markedness constraints than native words. Instead, they can amplify certain markedness restrictions, especially in suprasegmental adaptation of East Asian languages.

​

​

​

57

58 of 60

Conclusions on Loanword Phonology

  • Loanwords do not necessarily violate more markedness constraints than native words. Instead, they can amplify certain markedness restrictions, especially in suprasegmental adaptation of East Asian languages.

​

  • When native markedness is amplified in loanwords, faithfulness to the phonetic similarity remains influential, but only within the bounds of the unmarked domain.

​

​

58

59 of 60

References

  • Bradley, D. (1990). The Status of the 44 Tone in Nosu. La Trobe. Journal contribution.
  • Edmondson, J. A., Esling, J. H., & Ziwo, L. (2017). Nuosu Yi. Journal of the International Phonetic Association, 47(1), 87-97.
  • Goldwater, S., & Johnson, M. (2003). Learning OT constraint rankings using a maximum entropy model. In Proceedings of the workshop on variation within Optimality Theory (pp. 111-120).
  • Hayes, B., & Wilson, C. (2008). A maximum entropy model of phonotactics and phonotactic learning. Linguistic inquiry, 39(3), 379-440.
  • Inkelas, S., Orgun, C. O., & Zoll, C. (1996). Exceptions and static phonological patterns: cophonologies vs. prespecification. Unpublished Manuscript: UC Berkeley and University of Iowa.
  • Itô, J., & Mester, A. (2017). The phonological lexicon. The handbook of Japanese linguistics, 62-100.
  • Jurgec, P. (2010). Disjunctive lexical stratification. Linguistic Inquiry, 41(1), 149-161.
  • Kang, Y. (2010). Tutorial overview: Suprasegmental adaptation in loanwords. Lingua, 120(9), 2295-2310.
  • Simonović, M. (2012). The emergence of postcyclic prosody in loanword integration: Toneless Latinate adjectives in Serbo-Croatian. Acta Linguistica Hungarica (Since 2017 Acta Linguistica Academica), 59(1-2), 221-243.
  • Simonović, M. (2015). Lexicon immigration service-Prolegomena to a theory of loanword integration. Univerza v Novi Gorici.
  • 阿子莫小英. (2019). 凉山彝语中的汉语借词研究. 西南民族大学硕士论文.
  • 拉玛兹偓. (1991).试论彝语次高调产生的原因. 民族语文, (5), 35-38.
  • 潘正云. (1990). 现代凉山彝语音译词规范初探. 西南民族大学学报: 人文社会科学版, (4), 33-39.
  • 朱文旭. (1997). 凉山彝语中的汉语借词. 民族语文, (4), 39-41.
  • 朱文旭. (1999). 彝语古汉语借词的一种形式. 民族语文, (6), 76-76.

​

59

60 of 60

Thank you!

60