Cross-language analysis of phonetic units in language addressed to infants

Science

Volumn 277, Issue 5326, 1997, Pages 684-686

  Kuhl, Patricia K ^a   Andruski, Jean E ^a   Chistovich, Inna A ^b   Chistovich, Ludmilla A ^b   Kozhevnikova, Elena V ^b   Ryskina, Viktoria L ^b   Stolyarova, Elvira I ^b   Sundberg, Ulla ^c   Lacerda, Francisco ^c  

^a UNIVERSITY OF WASHINGTON   (United States)

^b Early Intervention Institute   (Russian Federation)

^c STOCKHOLM UNIVERSITY   (Sweden)

Author keywords

[No Author keywords available]

Indexed keywords

ARTICLE; CHILD DEVELOPMENT; HUMAN; LANGUAGE DEVELOPMENT; PHONETICS; PRIORITY JOURNAL; RUSSIAN FEDERATION; SWEDEN; UNITED STATES;

ADULT; FEMALE; HUMANS; INFANT; LANGUAGE DEVELOPMENT; MOTHERS; PHONETICS; RUSSIA; SPEECH ACOUSTICS; SPEECH PERCEPTION; SWEDEN; UNITED STATES;

EID: 0030868299     PISSN: 00368075     EISSN: None     Source Type: Journal    
DOI: 10.1126/science.277.5326.684     Document Type: Article

References (44)

1. V. Fromkin, S. Krashen, S. Curtis, D. Rigler, M. Rigler, Brain Lang. 1, 81 (1974); H. L. Lane, The Wild Boy of Aveyron (Harvard Univ. Press, Cambridge, MA, 1976).
2. V. Fromkin, S. Krashen, S. Curtis, D. Rigler, M. Rigler, Brain Lang. 1, 81 (1974); H. L. Lane, The Wild Boy of Aveyron (Harvard Univ. Press, Cambridge, MA, 1976).
3. L. A. Petitto, in Developmental Neurocognition: Speech and Face Processing in the First Year of Life, B. de Boysson-Bardies, S. de Schonen, P. Jusczyk, P. McNeilage, J. Morton, Eds. (Kluwer, Dordrecht, Netherlands, 1993), pp. 365-383.
4. P. W. Jusczyk, A. Cutler, N. J. Redanz, Child Dev. 64, 675 (1993); P. W. Jusczyk, A. D. Friederici, J. M. I. Wessels, V. Y. Svenkerud, A. M. Jusczyk, J. Mem. Lang. 32, 402 (1993).
5. P. W. Jusczyk, A. Cutler, N. J. Redanz, Child Dev. 64, 675 (1993); P. W. Jusczyk, A. D. Friederici, J. M. I. Wessels, V. Y. Svenkerud, A. M. Jusczyk, J. Mem. Lang. 32, 402 (1993).
6. P. K. Kuhl, K. A. Williams, F. Lacerda, K. N. Stevens, B. Lindblom, Science 255, 606 (1992).
7. P. K. Kuhl, Curr. Opin. Neurobiol. 4, 812 (1994).
8. _ and A. N. Meltzoff, J. Acoust. Soc. Am. 100, 2425 (1996).
9. C. E. Snow, in Talking to Children: Language Input and Acquisition, C. E. Snow and C. A. Ferguson, Eds. (Cambridge Univ. Press, Cambridge, 1977), pp. 31-49.
10. A. Fernald and T. Simon, Dev. Psychol. 20, 104 (1984); D. L. Grieser and P. K. Kuhl, ibid. 24, 14 (1988).
11. A. Fernald and T. Simon, Dev. Psychol. 20, 104 (1984); D. L. Grieser and P. K. Kuhl, ibid. 24, 14 (1988).
12. A. Fernald, Infant Behav. Dev. 8, 181 (1985); _ and P. Kuhl, ibid. 10, 279 (1987 ) .
13. A. Fernald, Infant Behav. Dev. 8, 181 (1985); _ and P. Kuhl, ibid. 10, 279 (1987 ) .
14. An earlier report [N. B. Ratner, J. Child Lang. 11, 557 (1984)] noted phonetic modifications in linguistic input to nine older children (specific ages were not given; children ranged from "preverbal," described as older than 9 months of age, up to those with mean length of utterances = 4.0, that is, children assumed to be 2 to 4 years of age). Although the direction of the data was similar to that shown here, no statistical analyses were reported.
15. B. Lindblom, J. Acoust. Soc. Am. 35, 1773 (1963).
16. G. E. Peterson and H. L. Barney, ibid. 24, 175 (1952).
17. J. Hillenbrand, L. A. Getty, M. J. Clark, ibid. 97, 3099 (1995).
18. N. Chomsky, Rules and Representations (Columbia Univ. Press, New York, 1981).
19. The mean ages of infants in the three countries were as follows; United States, 14.3 weeks (range, 9.1 to 18.1 weeks); Russia, 15.7 weeks (range, 9 to 22 weeks); Sweden, 13.3 weeks (range, 11 to 17.7 weeks).
20. In English, the words were "bead" for /i/, "pot" for /a/, and "boot" for /u/. In Russian, the words were "knizhka" (book) and "vilka" (fork) for /i/, "lapa" (paw) and "palets" (finger) for /a/, and "busy" (beads) and "ruchka" (pen) for /u/. To make it easier to incorporate preselected words in conversation, we provided mothers with small toys representing the objects. In Swedish, all content words containing the vowels /i/, /a/, or /u/ were analyzed. In all languages, women were recorded on one occasion; the conversation in each condition lasted about 20 min. Women were instructed to speak naturally and (when applicable) to use the preselected words at least three times during each conversation. All analyzed words contained stressed vowels.
21. The three languages were chosen because they represent the substantially different vowel systems that occur across languages: Russian is a 5-vowel system, American English is a 9-vowel system, and Swedish is a 16-vowel system (with the allophonic length variation) [J. Liljencrants and B. Lindblom, Language 48, 839 (1972)].
22. Formants are frequency regions in which the amplitude of acoustic energy is high, reflecting natural resonances created in the vocal tract. Formants are numbered (F1, F2, and so forth, from the lowest frequency upward); the frequency value cited is the center of the band of energy. F1 and F2 are the most important formants for vowel identification and are reported here. See K. N. Stevens, in Introduction to Communication Sciences and Disorders, F. D. Minifie, Ed. (Singular, San Diego, CA, 1994), pp. 399-437.
23. All words were low-pass filtered at 7.5 kHz and sampled at 15 to 16 kHz with specially designed software implemented on either a 12-or 16-bit 486 computer. Formant measures on the English and Russian words were made by a trained acoustician (J.E.A.) using narrow-band spectrograms, fast Fourier transform spectra, and autocorrelation linear prediction coefficient spectra. An automatic formant tracker (16 kHz, 18 poles, six formants) was used for the Swedish words. Reliability was assessed in two ways, each using a randomly selected 10% of the words in the data set: (i) Human reliability was assessed by having English and Russian words remeasured by J.E.A. and by a second trained analyst, and (ii) human-machine reliability was assessed by having American English words remeasured by machine and by having Swedish words remeasured by J.E.A. Reliability, expressed as the mean percentage difference between measures of individual formants in each vowel, was uniformly high. For human reliability across all formants and vowels, the mean intrascorer error was 5.95% and the mean interscorer error was 5.43%. For human-machine reliability across all formants and vowels, the mean error was 6.10%.
24. The vowel center formant measure most accurately reflects the speaker's intended frequency and is the one reported here, but all three locations produced the same pattern of results.
25. 1,27 = 110.64, P < 0.0001.
26. Each formant of each vowel in each language was assessed by MANOVA to examine the effects of addressee. All listed effects are significant at the 0.01 level.
27. P. Tallal et al., Science 271, 81 (1996); M. M. Merzenich et al., ibid., p. 77; M. Studdert-Kennedy and M. Mody, Psychon. Bull. Rev. 2, 508 (1995).
28. P. Tallal et al., Science 271, 81 (1996); M. M. Merzenich et al., ibid., p. 77; M. Studdert-Kennedy and M. Mody, Psychon. Bull. Rev. 2, 508 (1995).
29. P. Tallal et al., Science 271, 81 (1996); M. M. Merzenich et al., ibid., p. 77; M. Studdert-Kennedy and M. Mody, Psychon. Bull. Rev. 2, 508 (1995).
30. P. J. Swoboda, J. Kass, P. A. Morse, L. A. Leavitt, Child Dev. 49, 332 (1978).
31. D. Grieser and P. K. Kuhl, Dev. Psychol. 25, 577 (1989); P. K. Kuhl, Percept. Psychophys. 50, 93 (1991).
32. D. Grieser and P. K. Kuhl, Dev. Psychol. 25, 577 (1989); P. K. Kuhl, Percept. Psychophys. 50, 93 (1991).
33. Lindblom argued that speakers intuitively use "hyperarticulated speech" to accommodate the needs of listeners [B. Lindblom, in Speech Production and Speech Modeling, W. J. Hardcastle and A. Marchai, Eds. (Kluwer Academic, Dordrecht, Netherlands, 1990), pp. 403-439].
34. P. Iverson and P. K. Kuhl, J. Acoust. Soc. Am. 97, 553 (1995).
35. K. Johnson, E. Flemming, R. Wright, Language 69, 505 (1993).
36. It is of theoretical interest to compare the values of the American English "prototype" /i/ vowel used in P.K.K.'s laboratory studies (4, 25, 27) with the hyperarticulated vowels produced by mothers in this study. P.K.K.'s prototype was the male speakers' average /i/ in Peterson and Barney's (12) data set. When the formant values for Peterson and Barney's female average /i/ were compared with mothers' productions of /i/ in the present study, (i) the formant values fell very near the mean of American English mothers' infant-addressed /i/ (Fig. 1), and (ii) the feature values were virtually identical to the mean shown for American English mothers' infant-directed /i/ (Fig. 2). Thus, the earlier findings showing infants' superior categorization when listening to a "prototype" are consistent with our results showing that mothers hyperarticulate vowels when addressing their infants. In effect, the comparison suggests that P.K.K.'s previous tests used a vowel that would be described as "hyperarticulated." The comparison indicates that the Peterson and Barney formant values (derived from words spoken in citation speech) are not representative of formant values in adult-directed natural speech [see also (13)]. Also, in previous studies (4, 25, 27, 28), vowels judged to be "best instances" (i) maximize spectral features to a greater extent than do those judged to be poor instances, and (ii) are more extreme than those produced by adults in normal conversation.
37. Greater variety also improves phonetic category learning in foreign speakers [S. E. Lively, J. S. Logan, D. B. Pisoni, J. Acoust. Soc. Am. 94, 1242 (1993)].
38. G. Fant, Speech Sounds and Features (MIT Press, Cambridge, MA, 1973); R. Jakobson, C. G. M. Fant, M. Halle, Preliminaries to Speech Analysis: The Distinctive Features and Their Correlates (MIT Press, Cambridge, MA, 1969).
39. G. Fant, Speech Sounds and Features (MIT Press, Cambridge, MA, 1973); R. Jakobson, C. G. M. Fant, M. Halle, Preliminaries to Speech Analysis: The Distinctive Features and Their Correlates (MIT Press, Cambridge, MA, 1969).
40. S. S. Stevens, J. Volkmann, E. B. Newman, J. Acoust. Soc. Am. 8, 185 (1937). The mel scale equates the magnitude of perceived differences in pitch at different frequencies.
41. R. D. Kent and A. D. Murray, ibid. 72, 353 (1982).
42. J. S. Perkell and D. H. Klatt, Eds., Invariance and Variability in Speech Processes (Erlbaum, Hillsdate, NJ, 1986).
43. J. R. Saffran, R. N. Aslin, E. L. Newport, Science 274, 1926 (1996).
44. Supported by the William P. and Ruth Gerberding Professorship and NIH grant DC 00520 (P.K.K.), a grant from the Social Sciences and Humanities Research Council of Canada (J.E.A.), and Bank of Sweden Tercentenary Foundation grant 94-0435 (F.L.). We thank E. Stevens for help on all aspects of the data analysis and A. N. Meltzoff for helpful comments on an earlier draft of the manuscript.