Perceptual attrition of lexical tone among L1 Yoruba-speaking children in Canada Saliu Shittu and Anne-Michelle Tessier * University of Alberta BUCLD39 * Friday Nov 7, 2014

Acknowledgments

• members of the University of Alberta Phonology Lab • our anonymous reviewers • all the families who contributed their time and data, especially Mrs. Adeladan • partial funding provided to the first author: University of Alberta‘s GSA Travel

grant and Boston University’s Paula Menyuk Student Travel Award.

BA

CK

GR

OU

ND

METH

OD

S

Selected References

• Niger-Congo family (Nigeria, Benin Republic, Togo…) • Three lexical tones: High, Mid and Low

SUMMARY This study investigates perceptual attrition of lexical tone in Yoruba (Niger-Congo). Two participant groups -- children and adults -- were all born into Yoruba-speaking homes, and now live in English-dominant Canada. We compare their accuracy in three tonal perception tasks: discrimination, identification and lexical recognition. Overall, child participants show poor tonal perception in all tasks, compared to adults: H tone is most easily perceived; tonal contours can be particularly hard. The best predictors of participant accuracy include age and relative exposure to L1 Yoruba vs. L2 English.

• How does perceptual attrition affect Yoruba lexical tone among children? Which perceptual tasks are most affected?

• Does the relative ‘strength’ of Yoruba tones predict rates of attrition? Are HL and LH contours particularly susceptible?

• Which external factors – age, exposure and environment – most strongly predict attrited children’s successful tonal perception? Do these factors vary in their effects across tasks?

RESEARCH QUESTIONS

• ‘Strength’ of tones: H tone as strongest (see e.g. Akinlabi 1985; Pulleyblank 1986)

• M: most difficult in adult processing and child acquisition (Bakare, 1995; Orie 2006, 2012)

• HL and LH ‘contours’: also difficult?

Lexical Tone in Yoruba

Bkgd: How and when is tone acquired? Here: How and when can tone be lost?

Focus: L1 Yoruba/L2 English children

{sshittu, amtessier}@ualberta.ca

Bkgd: On L1 tonal perception

• L1-specific tonal perception achieved btwn 6-8mos (see e.g. Harrison 2000; Mattock et al 2008; also Tse, 1978)

Final Tone Identification Lexical Minimal Pairs

Participants

Assessing Language Input and Environmental Factors: ALEQ (Paradis, 2011)

Three Perception Tasks AX Discrimination

CHILDREN (N=21) • ages 8-15 (mean: 11.6) • exposed to L1 Yoruba since birth • living in Alberta • hear Yoruba spoken daily • do not (ever?) speak it

ADULTS (N=7) (~ controls) • ages 42-47(mean: ) • native L1 Yoruba speakers • also L2 fluent English speakers • parents of experimental children • speak Yoruba daily at home

Alberta Language Environment Questionnaire (http://www.linguistics.ualberta.ca/en/CHESL_Centre/Questionnaires.aspx#aleq) measures include: place of birth months in L2 context language richness score (below) age of L2 exposure/arrival age at test

language richness score: proportion of Yoruba (L1) vs. English (L2) exposure, speaking vs. hearing

with parents and other family, at school, with peers, in media (TV, radio, songs), etc...

• Phonological attrition driven by language-external and -internal factors (e.g. Bullock and Gerfen, 2004)

• Very few tonal attrition studies • Yeh and Lin (2013); Yeh and Lu (2012): - tonal attrition vs. language contact in Hakka - attrition causes tonal errors across all tasks - tonal perception less attrited than production

This Study: On L1 (tonal) attrition

Akinlabi, A. 1985. Tonal Underspecification and Yoruba Tones, Ph.D. Dissertation, University of Ibadan. * Bakare, C.A. 1995. Discrimination and identification of Yoruba tones: perception experiments and acoustic analysis. In Language in Nigeria (K. Owolabi, ed.) Ibadan: Group Publishers. * Harrison, P. 2000. Acquiring the phonology of lexical tone in infancy. Lingua 110. 581-616. * Mattock, K., M. Molnar, L. Polka and D. Burnham. 2008. The developmental course of lexical tone perception in the first year of life. Cognition 106: 3. 1367-1381 * Orie, O. 2006. L2 Acquisition and Yoruba Tones: Issues and Challenges. InProceedings of ACAL 36 (O. Arasanyin et al eds.) 121-128. Cascadilla Press * Orie, O. 2012. Acquisition reversal. The effects of postlingual deafness in Yoruba. Studies on language acquisition no. 47. De Gruyter Mouton * Pulleyblank, D. 1986. Tone in Lexical Phonology. Dordrecht: Reidel. * Yeh, C-H and Y-H Lin. 2013. The Attrition of Hai-lu Hakka’s Tonal System. Proceedings of ICPLC 2013 * Yeh, C-H and C.J. Lu, 2012. The Effect of Language Attrition on Low Level Tone in Hakka. Proceedings of Speech Prosody 2012.

Relative strength of tones?

Better than Chance Scores

• methods from Yeh and Lu (2012) • nonce words used in

discrimination and identification • item order randomized within task

• (C)V or V ̃ syllables • ≅ numbers of each possible

Yoruba vowel in target syllables • materials pre-tested: identification

by naive native speaker • 4-10 training trials before each

task, with feedback

ALEQ Predictors of Accuracy

Monosyllabic (n=11) Disyllabic (n=11)

sound pic1 pic2 ? sound pic1 pic2 ?

[bo]

H

‘peel’

cover(L)

1 [odo]

MH

river(ML)

mortar

2

[ro]

M

think(L)

plough

2 [akp͡a]

LH

scar

arm(MH)

1

Monosyllabic (n=24) Disyllabic (n=24)

lͻ H dũ H same luwa HM k͡pͻki HM same

wɛ L ga L same are ML jeba ML same

ta M tɛ M same dͻku LH bɛmu LH same

dͻ H ju M diff bani HM wɛtͻ LH diff

g ͡bo M bɛ L diff seda LH kedũ ML diff

ge L bu H diff orĩ LL fĩju HH diff

Disyllabic (n=24) Trisyllabic (n=24) g͡beju HH H ak͡pͻtͻ LLM M

tikͻ LL L fͻsãmu MLL L

diwͻ HM M g͡bĩgure LHH H

k͡peta LM M rokalͻ MLM M wodã HL L bͻlɛmu HLH H ikͻ LH H ifalɛ LHL L

• nonce words used to control for familiarity • easier perceptual task • disyllabic materials harder? tonal memory? • H tone easiest? M tone hardest?

• nonce words used to control for familiarity • more difficult task • trisyllabic materials harder? • do HL and LH contours make it harder still? • is this task easier or harder for attriting children?

[bá] ‘to meet’ F0=908Hz

[bà] ‘to land’ F0=248Hz

[bā] ‘to weave’ F0=637 Hz

CONCLUSIONS

• linear mixed-effect (lmer4) models • fixed effects: item and participant • tests for collinearity of predictors

• older Age of Arrival • shorter Months in Canada • not born in Nigeria (confound?)

• older Age of Arrival • shorter Months in Canada • not born in Nigeria (confound?)

• older Age of Arrival • older Age at Test (slow, on-going lexical development)

RES

ULTS

51%

37%

54%

74%

54%

96%

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

discrim ident lexical

child

adult

discrim ident lexical

child 5/21 (24%) 5/19 (26%) 5/19 (26%)

adult 5/7 (71%) 3*/7 (43%) 7/7 (100%)

error bars: 1SD

45% 56%

66% 79%

0%

20%

40%

60%

80%

100%

same tones diff. tones

child

adult

accuracy for 2σ

same-tone items: best: HH, LH > HL (mean: 55%) worst: LM > LL > MH, HM (mean: 29%)

accuracy for 1σ

diff-tone items best: H vs. L (71%), L vs H (64%)

Advantage of tonal contrast

Overall Accuracy

Χ2 tests on accuracy: different tone pairs > same tones pairs (p < 0.01 for both groups)

No effect of word size

No sig. diff. between 1 and 2-syllable items

(p > 0.1 for both groups)

53% 49%

71% 76%

0%

20%

40%

60%

80%

100%

1σ 2σ

child

adult

45% 33% 33%

52% 54% 57%

0%

20%

40%

60%

80%

100%

H M L

Child

Adult

Advantage of H tone children: only H-tone identified better than chance

Disadvantage of tonal contours

No word size effect

38%

55%

36%

53%

0%

20%

40%

60%

80%

100%

2σ 3σ

No sig. diff. between 1 and 2-syllable items

Largest Child v.s Adult Difference

• children 54% vs. adults 96% • possible greater need for F0 normalization

among attrited listeners?

On the ‘strength’ of Yoruba tones in attrition – H easiest? M hardest? • H tone easiest to identify ; no difference for adults • relative strength of tones may influence lexical recognition • HL and LH contours easy to discriminate, but hard to identify

On tonal perception and children’s L1 attrition in a L2 non-tonal environment • attriting children’s tonal perception is greatly reduced (though tonal short term memory is not?) • exposure to spoken L1 at home not sufficient to support tonal maintenance • difficult tasks even for adults: are they also attrited compared to monolingual Yoruba adults?

Relative strength of tones in 1σ items?

54% 54%

93% 100%

0%

20%

40%

60%

80%

100%

1σ 2σ

No sig. diff. between 1 and 2-syllable items

No word size effect!

59% 55% 46%

0%

20%

40%

60%

80%

100%

H L M

children: H tone 1σ words

just sig. better identified than M tones (p = 0.05)

33% 25%

49% 37%

0%

20%

40%

60%

80%

100%

final L final H

contour non-contour

Χ2 tests on child accuracy: LH worse than MH, HH HL worse than ML, LL (LH: p< 0.01; HL, p = 0.02)