!xóõ click perception by english, isizulu, and sesotho listeners 738
Post on 20-Jan-2016
230 views
TRANSCRIPT
!Xóõ click perception by English, Isizulu, and Sesotho listeners
738
Catherine T. Best1,2
Anthony Traill3
Allyson Carter4
K. David Harrison2,5
Alice Faber2
1 Wesleyan University2 Haskins Laboratories3 University of the Witwatersrand4 University of Arizona5 Swarthmore College (USA)
Supported by NIH grants DC00403 & HD01994
Adult listeners find many but not all nonnative phonological contrasts difficult to discriminate. The Perceptual Assimilation Model (PAM, Best 1995) attributes those discrimination differences to the way listeners assimilate the nonnative phones to their native phonologies, which vary among languages. We tested perception of two !Xóõ click contrasts by speakers of American English vs. speakers of Isizulu and of Sesotho, two African click languages that lack the target contrasts. Both African groups were expected to assimilate the stimuli to native clicks and to discriminate them accordingly. Americans were expected to perceive them more often as non-
ABSTRACT
speech and to discriminate both contrasts well. Because Isizulu has a richer click system than Sesotho, Isizulu listeners should assimilate !Xóõ clicks to native clicks more often, and discriminate at least one contrast better than Sesotho listeners. Americans should excel on any contrast that an African group assimilates to a single click. As predicted, Isizulu listeners most often assimilated !Xóõ clicks to native ones; Americans were most likely to hear them as nonspeech. Sesothos found one contrast more difficult to categorize and discriminate than the other groups. Both African groups assimilated the other contrast to a single native click, with lower discrimination than Americans.
BACKGROUND
Categorization and discrimination of unfamiliar nonnative phonetic contrasts is often difficult for adults
• but more recent research indicates that for some nonnative contrasts, perception may be good or excellent
What possible factor(s) may account for variations in perception of various nonnative contrasts by naïve adults?
Theoretical Models:Speech Learning Model (SLM): Flege
• L2 learners hear individual nonnative phones as identical or similar to native phonemes, or as new
Natural Language Magnet model (NLM): Kuhl• prototypes formed by native speech experience act
as perceptual magnets to same-category neighbors
Fragile-Robust Hypothesis (FRH): Burnham• psychoacoustically robust nonnative contrasts are
more easily maintained and learned than fragile ones
Perceptual Assimilation Model (PAM): Best
• nonnative phones assimilated to articulatorily-similar native contrasts/phonemes or heard as nonspeech
Approach: Cross-linguistic perception tests Research design:
• Listeners of languages that display crucial differences in phonological/phonetic details
• Mutually-nonnative stimulus contrasts that relate to listener languages in critically different ways
Theoretical considerations:• SLM: equivalence classifications should vary across
nonnative stimuli and listener groups
• NLM: stimulus properties should vary re: groups’ acoustic exposure and similarity to native prototypes
• FRH: stimulus contrasts should vary along the fragile-robust psychoacoustic dimension
• PAM: contrasts should differ among listener groups in expected assimilations to native contrasts
Choice of Stimuli & Listener Groups (Table 1) !Xóõ click POA (place of articulation) contrasts (2):
• mutually nonnative: across groups that vary in potentially relevant acoustic/phonetic experience
• acoustic/phonetic similarity varies: re: contrasts and listener groups
• psychoacoustics vary: robust vs. more fragile contrast
• assimilations should vary: groups should differ in varied speech and nonspeech percepts of the clicks
Listener groups(3):• American English: no click phonemes, tend to hear
clicks as nonspeech sounds
• South African languages (2): both have clicks but lack the target POA contrasts and phonetic accompaniment
Isizulu has click POA contrasts, Sesotho does not
Stimuli & Listener Languages
STIMULI LISTENER LANGUAGES
Target Clicks: !Xóõ (Khoisan)
Isizulu (Bantu)
Sesotho (Bantu)
English
bilabial + [x] [áx] √
dental + [x] [ñx] √palatal + [x] [úx] √alveolar + [x] [óx] √
Articulatorily Related Phonemes:bilabial click +/- [h][á(h)] √dental click +/- [h][ñ(h)] √ √palatal click +/- [h][ú(h)] √alveolar click +/- [h][ó(h)] √ √ √lateral click +/- [h][Ñ(h)] √ √velar fricative [x] √ √ √velar affricate [kxH] √ √velar ejective [k'] √ √
Table 1
Predictions:Speech Learning Model (SLM):
• Isizulu: [áx] is new while [ñx] is similar to native [ñ]), thus [áx]-[ñx] is easy to discriminate.
[úx]-[óx] are both similar to native [ó], thus difficult.
• Sesotho: [áx]-[ñx] are both new, thus easy.[úx]-[óx] are both similar to native [ó], thus
difficult.• Americans: All clicks new, thus both contrasts easy.
Fragile-Robust Hypothesis (FRH):• [úx]-[óx] contrast is psychoacoustically extremely
salient, [áx]-[ñx] is notably less salient; the more robust contrast should be easier to discriminate.
• Both contrasts are nonnative to all groups, thus no group differences are expected.
Natural Language Magnet model (NLM):• Isizulu: [áx] and [úx] non-protoypes, [ñx] and [óx]
attracted to native [ñ]) and [ó] prototypes, thus both contrasts moderately, equally easy to discriminate.
• Sesotho: [óx] attracted to native [ó], all others are non-prototypes, thus [áx]-[ñx] easier than [úx]-[óx].
• Americans: All clicks are non-protoypes, thus both contrasts very easy to discriminate.
Perceptual Assimilation Model (PAM):• Isizulu: [ñx] assimilated to native [ñ]) better than
[áx], both [úx]-[óx] assimilated to native [ó], thus [áx]-[ñx] (Category Goodness) > [úx]-[óx] (Single Category).
• Sesotho: all clicks assimilate to native [ó], thus both contrasts difficult; worse than Isizulus on [áx]-[ñx].
• Americans: more likely to hear clicks as nonspeech; should discriminate [úx]-[óx] better than other groups.
METHODS
Participants:
•American English (n = 16) from northeast United States
• Isizulu (n = 13) from Kwazulu-Natal Province, South Africa
• Sesotho (n = 13) from the Southern Sotho area of Free State Province, South Africa
• additional inclusion requirements: native speakers of respective languages
no experience with !Xóõ or similar click systems
tested in native language and country
Stimuli (Figures 1 & 2, Table 2):• !Xóõ speaker: adult female from Lokalane, Botswana
• stimulus tokens: 4 /Ca/ syllables with velar-fricated clicks, 5 tokens each: áx, ñx, úx, óx
• POA contrasts: both nonnative to all Ss bilabial - dental [áx] [ñx] (more fragile) palatal - alveolar [úx] [óx] (robust)
Tasks: (separately for each contrast):
• Discrimination: categorial AXB
• Categorization: labeling (re: native phonemes and/or nonspeech sounds) and goodness ratings of all individual tokens
Bilabial [áx] Dental [ñx]
Time (s)0 0.988545
–0.7129
0.701
0
Time (s)0 0.988545
0
8000
Time (s)0 0.940932
–0.3616
0.4619
0
Time (s)0 0.940932
0
8000
Example tokens: bilabial/dental click contrast Figure 1
Palatal [úx] Alveolar [óx]
Time (s)0 0.978322
–1
1
0
Time (s)0 0.978322
0
8000
Time (s)0 0.96644
–0.9539
0.8775
0
Time (s)0 0.96644
0
8000
Example tokens: palatal/alveolar click contrastFigure 2
Mean acoustic values for !Xóõ stimulus tokensbilabial dental palatal alveolar
[áx] [ñx] [úx] [óx]
SYLLABLE:
duration (msec) 905.44 914.42 897.40 893.14
CLICK:
amplitude (RMS) 57.70 43.70 35.06 110.34
click centroid (Hz) 8516.76 9465.39 4179.62 1210.33
stop centroid (Hz) 1289.11 1994.45 2360.57 1685.67
FRICATIVE:
amplitude (RMS) 32.43 9.02 10.16 3.52
fric. centroid (Hz) 5012.55 3417.17 5280.33 2784.02
VOWEL (peak):
F0 (Hz) 176.40 178.20 182.00 183.00
F1 (Hz) 909.80 911.30 914.28 909.80
F2 (Hz) 1219.02 1214.56 1211.56 1198.12
F3 (Hz) 3174.56 3226.82 3253.72 3223.84
Table 2
RESULTSCategorization (Figures 3 & 4, Table 3):
• Americans perceived nonspeech (fully or partially) most often, especially for dental and alveolar clicks.
• Sesotho listeners perceived more nonspeech than Isizulu listeners, except for palatal clicks.
• Sesothos assimilated both contrasts equally to native [ó], as did Isizulus for [úx]-[óx], but they heard a goodness difference in [áx]-[ñx].
Discrimination (Figure 5):• Sesothos were equally poor on both contrasts.
• Isizulus did better on [áx]-[ñx] than [úx]-[óx].
• Americans did better on [úx]-[óx] than [áx]-[ñx] and better on [úx]-[óx] than both Sesothos and Isizulus.
0
20
40
60
80
100
Speech Nonspeech Mixed
!Xóõ Click Labelingassimilation type x language group
AmericanSesothoZulu
% responses (non-missing)
ASSIMILATION TYPE
Figure 3
0
20
40
60
80
100
Speech-only Nonspeech Mixed
!Xóõ Bilabial Click Labels
AmericanSesothoZulu
% responses (non-missing)
ASSIMILATION TYPE
0
20
40
60
80
100
Speech-only Nonspeech Mixed
!Xóõ Dental Click Labels
AmericanSesothoZulu
% responses (non-missing)
ASSIMILATION TYPE
0
20
40
60
80
100
Speech-only Nonspeech Mixed
!Xóõ Palatal Click Labels
AmericanSesothoZulu
% responses (non-missing)
ASSIMILATION TYPE
0
20
40
60
80
100
Speech-only Nonspeech Mixed
!Xóõ Alveolar Click Labels
AmericanSesothoZulu
% responses (non-missing)
ASSIMILATION TYPE
Figure 4
Most Frequent Assimilations (chosen > 50% by any individual subject)
Assimil.
Type: bilabial dental palatal alveolar
SPEECH: [áx] [ñx] [úx] [óx]
American /f/, /T/ /t/, /T/ /t/ /k/, /p/
Sesotho/óhx/, /xh/,
/t/
/óhx/, /xh/,
/Ñx/, /Ñ/, /óh/
/óx/, /ó/, /óh/,
/xh//óx/, /ó/, /óh /
Isizulu/ñh/, /ñhx/,
/gñ/,/Ñh/,/Ñhx//ñh/ /óh/, /óhx/,
/x/
/óh/, /óhx/,
/gó/NOTE: [Ñ] = lateral click ([Ñ] and [ñ] are not native to Sesotho)
NONSPEECH:
American fizzle tsk, kiss snap, slap, whack, click
pop, whack, click
Sesotho snap, cracklesnap, crackle, knife-closing
snap, crackle, shoe heel hit
snap, crackle, stone hit
Isizulu clap, gun, unclear sound
Table 3
50
60
70
80
90
100
Bilabial-Dental Palatal-Alveolar
!Xóõ Click Contrastsdiscrimination x language group
AmericanSesothoZulu
% correct discrimination
CLICK CONTRAST
chance
Figure 5
[úx]-[óx][áx]-[ñx]
CONCLUSIONSTheoretical implications:
• SLM and NLM and FRH predictions all failed to account for one or more findings
• PAM predictions were most straightforwardly supported by the results.
Counter-intuitive result predicted by PAM:• Americans, who lack linguistic experience with clicks, outperformed the click-experienced Isizulu and Sesotho listeners on one contrast
Further research:• confirm and extend to other groups/contrasts
• assess the types of stimulus information non-native listeners use (e.g., acoustic, articulatory)
REFERENCES
Beach, D.M. (1938). The phonetics of the Hottentot language. Cambridge U.K.: W. Heffer & Sons, Ltd..
Best, C.T. (1995). A direct realist perspective on cross-language speech perception. In W. Strange (ed.) Speech perception and linguistic experience: Issues in cross-linguistic research (p. 171-204). Timonium, MD: York Press.
Best, C.T. & Avery, R.A. (1999). Left hemisphere advantage for click consonants is determined by linguistic significance. Psychological Science, 10, 65-69.
Best, C.T., McRoberts, G.W., & Goodell, E. (2001). American listeners' perception of nonnative consonant contrasts varying in perceptual assimilation to English phonology, Journal of the Acoustical Society of America, 109, 775-794.
Best, C.T., McRoberts, G.W., & Sithole, N.M. (1988). Examination of perceptual reorganization for nonnative speech contrasts: Zulu click discrimination by English-speaking adults and infants. Journal of Experimental Psychology: Human Perception and Performance, 14, 345-360.
Burnham, D.K. (1986). Developmental loss of speech perception: Exposure to and experience with a first language. Applied Psycholinguistics, 7, 207-240.
Flege, J.E. (1995). Second-language speech learning: Theory, findings, and problems. In W. Strange (ed.), Speech perception and linguistic experience: Issues in cross-linguistic research (p. 233-277). Timonium MD: York Press.
Kuhl, P. K. & Iverson, P. (1995). Linguistic experience and the "perceptual magnet effect." In In W. Strange (ed.) Speech perception and linguistic experience: Issues in cross-linguistic research (p. 121-154). Baltimore MD: York Press.
Ladefoged P. & Traill, A. (1994). Clicks and their accompaniments. Journal of Phonetics, 22, 33-64.
Traill, A. (1994). The perception of clicks in !Xoo. Journal of African Languages & Linguistics, 15, 161-174.
Traill, A. (1978). Another click accompaniment in !Xóõ. Khoisan Linguistic Studies, 5, 22-29.
Traill, A. (1977). The phonological status of !Xõo clicks. In A. Traill (ed.) Khoisan Linguistic Studies 3 (p. 107-131). Johannesburg: University of Witwatersrand African Studies Institute.
Bila
bia
l
[áx]
Denta
l
[ñx]
Time (s)0 0.988545
–0.7129
0.701
0
Time (s)0 0.988545
0
8000
Time (s)0 0.940932
–0.3616
0.4619
0
Time (s)0 0.940932
0
8000
Pala
tal
[úx]
Alv
eola
r
[óx]Time (s)
0 0.978322–1
1
0
Time (s)0 0.978322
0
8000
Time (s)0 0.96644
–0.9539
0.8775
0
Time (s)0 0.96644
0
8000