Cross-Language Speech Perception Notes (Werker & Tees)

Experiment 1: Thompson glottalized velar vs. uvular (/ki/ vs /qi/)

• Objective: Test generality of early cross-language speech perception and identify the time course of the decline.
• Subjects:
  • Infants: 12 full-term 6–7 months old (avg ~6 months 29 days)
  • Adults: 10 English-speaking adults; 5 Thompson-speaking adults
• Stimuli: Thompson non-English contrast /ki/ (glottalized velar) vs /qi/ (glottalized uvular). Four exemplars per category; duration ~400\text{ ms} per token with ~1500\text{ ms} silent interval.
• Apparatus & Procedure:
  • Head Turn (HT) paradigm for infants; adults responded with button press.
  • Conditioning phase to associate a sound change with a visual reinforcer (toy activation).
  • Test phase: Change trials from /ki/ to /qi/; infant must anticipate the change within a 4.5\,\text{s} window and head-turn toward the reinforcer.
  • Criterion: 8/10 correct responses with ≤2 errors.
• Key results:
  • Thompson adults: all reached criterion; English adults: significantly poorer performance; English infants: 8/10 reached criterion.
  • Statistical note: overall chi-square significant, \chi^2=8.94, p<0.05; English adults performed worse than Thompson and infant English groups.
• Conclusion: Infants (even without specific experience) discriminate non-native glottalized velar/uvular contrasts; adults show language-specific limitations.

Experiment 2: Developmental time course; two non-English contrasts

• Contrasts tested:
  • Thompson ki/qi (glottalized velar vs uvular)
  • Hindi ta/ta (retroflex vs dental relies on place of articulation)
• Subjects:
  • 6–8 months: English infants
  • 8–10 months: English infants
  • 10–12 months: English infants
  • Additional groups: Hindi- and Thompson-exposed infants (11–12 months) for comparison
• Stimuli & Procedure:
  • English labial/alveolar contrast /ba/−/da/ included to verify discrimination ability before/after non-native testing (two-phase criterion).
  • Stimuli: four exemplars per sound; duration ~500\text{ ms}; interstimulus interval ~1500\text{ ms}.
  • HT paradigm as in Experiment 1.
  • Criterion: 8/10 correct changes; additional requirement that discrimination on native contrast before and after non-native change be evident.
• Results:
  • 6–8 months: most infants reached criterion on both contrasts.
  • 8–10 months: all 6–8 mo infants remained able on Hindi; fewer on Thompson; mixed results on both contrasts.
  • 10–12 months: few or none reached criterion on non-native contrasts; by this age, performance on non-native contrasts in English-learning infants resembled adults.
  • Exposure effect: infants raised in Hindi or Thompson homes retained discrimination at 11–12 months on at least one non-native contrast.
  • Statistical notes: overall chi-square for both contrasts significant, \chi^2=21.67 (ki/qi) and \chi^2=24.59 (ta/ta), with p-values p<0.001 for the age-group differences and further pairwise contrasts (e.g., between 6–8 and 10–12 months, and between 8–10 and 10–12 months).
• Conclusion: The decline in non-native phonetic discrimination occurs within the first year and is strongly influenced by native-language experience.

Experiment 3: Longitudinal replication

• Rationale: Address drop-out bias and track within-subject change across ages.
• Subjects: 6 infants tested at three ages:
  • Time 1: 6–8 months (avg ~7m15d)
  • Time 2: 8–10 months (avg ~9m2d)
  • Time 3: 10–12 months (avg ~10m22d)
• Stimuli & Procedure:
  • Non-native contrasts: Hindi ta/ta and Salish ki/gi; Native contrast: English /ba/−/da/
  • HT procedure as before
• Results:
  • Time 1 (6–8 mos): all 6 reached criterion on both non-native contrasts.
  • Time 2 (8–10 mos): all six reached criterion on Hindi ta/ta; only 3 reached criterion on Salish ki/gi.
  • Time 3 (10–12 mos): none reached criterion on non-native contrasts; all could discriminate native English /ba/−/da/.
  • Pattern mirrors cross-sectional results; Hindi contrast showed an abrupt decline around 10–12 months, while Thompson (Salish) showed a more gradual loss.
• Table reference: summarizes “Infant Discrimination Performance on Two Non-English Speech Contrasts” (Reached Criterion vs. Not). Data illustrate converging cross-sectional and longitudinal patterns.

General Discussion and Implications

• Core finding: Young infants can discriminate many phonetic distinctions across languages without targeted experience; a decline emerges with native-language experience within the first year.
• Perceptual reorganization: infants’ initial broad sensitivity to phonetic categories becomes tuned to phonemic distinctions used in the learner’s native language.
• Functional implications: Tuning may facilitate language learning by focusing processing on linguistically relevant contrasts; might align with phonological development and language production/ comprehension trajectories.
• Open questions:
  • Is the decline due to general cognitive maturation or to language-specific perceptual learning?
  • How do changes in perception relate to production and comprehension development?
  • Are there cross-language patterns that extend to other phonetic domains beyond stop consonants, sibilants, vowels, and liquids?
• Future directions:
  • Examine links between cross-language perception, vocal output, and language comprehension.
  • Investigate whether perceptual changes reflect broader cognitive changes or are specifically tuned to phonology.

Key Concepts and Terms

• Phonetic category vs. phonemic distinction
• Universal phonetic contrasts in infancy and language-specific tuning with experience
• VOT (voice onset time) differences as cross-language contrasts
• Back stops: velars vs. uvulars; glottalization effects
• HT (head-turn) paradigm for infant speech perception
• Stimulus exemplars and control of acoustic variability
• Criteria for discrimination: 8/10 correct responses; pre/post native discrimination checks
• Perceptual reorganization and its relation to language acquisition

Notation and Statistics (Selected)

• Criterion for discrimination: 8/10 correct responses with ≤2 errors: 8/10
• Time window for response: 4.5\text{ s}
• Stimulus durations: 400\text{ ms} (Experiment 1) and 500\text{ ms} (Experiment 2) per exemplar
• Key p-values and tests:
  • Overall group differences: p<0.05 in Experiment 1 (\chi^2 = 8.94)
  • Experiment 2: \chi^2 = 21.67 (ki/qi) and \chi^2 = 24.59 (ta/ta); both with p<0.001 for age-group contrasts
• Formant and spectral cues cited: third formant transitions; amplitude and duration of bursts (e.g., /k/ vs /q/ bursts)
• Longitudinal vs cross-sectional patterns show consistent decline timing across contrasts, with language exposure mitigating the decline