6.3 What is the nature of the input?
6.3.2.1 Is BTR a language-teaching mechanism?
Newport, Gleitman and Gleitman 1977 provided scientific test of the motherese hypothesis to address this question. Recognizing that mothers vary in the degree to which they use BTR, they reasoned that if a mother shows a greater
29See Ingram 1995 for discussion, Schieffelin 1979 for study of the Kaluli in Papua New Guinea.
amount of motherese, then, following (1), the child should show faster language acquisition. They also tested a “Fine Tuning Hypothesis”: if a mother is acting as teacher, then as the mother’s speech grows in complexity, then so should the child’s (Gleitman, Newport and Gleitman 1984; Valian 1999 for discussion).
Fifteen American mother–daughter pairs in three age groups were studied (12–15, 18–21, and 24–27 months), with children’s MLUs ranging from 1.00 to 3.46 (mean 1.65). The pairs were interviewed twice, six months apart. Both times, adult speech was recorded, analyzed and coded for various specific mea- sures: e.g., for specific measures of length, complexity, utterance type, and repetition, as was child speech (Newport, Gleitman and Gleitman 1977, 116–
118, tables 5.1 and 5.2). Correlations were then computed between every prop- erty of maternal speech and growth in child language on relevant measures (132, table 5.3).30
Results showed that the vast majority of properties of maternal speech did not correlate positively with developing complexity in child speech. The length or complexity of a mother’s utterances did not correlate with the same features in the child’s language; nor did amount of repetition by the mother correlate with any form of growth measured. Growth of complex sentence structures in the child’s speech did not correlate with any property of maternal speech.
Only two correlations were significant: (a) the number of yes/no questions in the mother’s speech (e.g., “Do you want to take a bath now?”) correlated with the development of overt auxiliaries in the verb phrases of child speech (e.g., “I will jump”), although absolute amount of auxiliaries used in mother’s speech did not;
and (b), noun phrase inflections (e.g., plurals) in the child developed in correlation with amount of deixis (e.g., “That’s a dog”) in the mother’s speech. A Fine Tuning Hypothesis was not supported. Mothers’ MLU was found to correlate with age, but not with language development in children.
These findings disconfirm a strong form of the Motherese Hypothesis, although they do not suggest that children cannot or do not attend to specific properties of the input. They suggest a “semi autonomous unfolding of language capabilities.”
Effects of maternal input are those which match the biases of the learner, which act “as a filter through which the linguistic environment exerts its influence”
(137). These results begin to factor out which properties of the input infants may select. The input is not the primary determinant of the universal aspects of language knowledge, e.g., those involved in complex sentence formation, but it may affect language specific factors which require induction, e.g., the lexical form of the auxiliary verb used in English, or the morphology involved in English pluralization.
30These researchers realized that finding simple positive correlations would not suffice to confirm either the Motherese or the Fine Tuning Hypothesis. “Basic Motherese may be used more when the child is least sophisticated linguistically, but also the child may grow the fastest the less his linguistic sophistication, i.e., the more he has left to learn” (Newport, Gleitman and Gleitman 1977, 133), thus providing the spurious result that more growth correlated with more Motherese.
To correct this, the researchers partialled out variance due to the child’s age and language level.
114 c h i l d l a n g ua g e
Do mothers need to provide “super vowels”?
In another study of BTR, mothers’ speech to their two- to five-month-olds in three countries (USA, Russia and Sweden) was subjected to spectrographic analy- sis and showed that “mothers addressing their infants produced acoustically more extreme vowels than they did when addressing adults” (Kuhl et al. 1997, 684).
The researchers concluded that “language input to infants provides exceptionally well-specified information about the linguistic units that form the building blocks for words” (1997, 684). In all countries, when the mothers’ vowels, /a/, /i/ and /u/, were measured in terms of vowel formant frequency, fundamental frequency (pitch) and duration, not only was there an increase in duration and fundamental frequency, but the vowel triangle was found to be “stretched.”31
We do not know if such phenomena are influential in language acquisition, nor how universal they are (e.g., Quich´e Mayan). Acoustic exaggeration of a vowel space would not alone provide the information to determine the phonemic contrasts of a language.32In Swedish, children must distinguish at least sixteen variations within this vowel space corresponding to different vowels, while in English they must distinguish nine, and in Russian only five. As we have seen, linguistic units do not exist in the acoustic information.33Yet children must dis- cover these, including the contrasts which create them, and the system which relates them.34
Kuhl et al. suggest several ways in which vowel space expansion might
“enhance learning” (1997, 684). (a) The three vowels may be more easily per- ceived because they are “more distinct” from each other. However, as we see in chapter 8, infants make very fine distinctions in the speech stream, even more so the younger they are (see chapter 8). Thus infants appear not to need acoustic expansion, or hyperarticulation, for initial speech perception. (b) “[E]xpanding the vowel triangle allows mothers to produce a greater variety of instances rep- resenting each vowel category without creating acoustic overlap between vowel categories. Greater variety may cause infants to attend to non-frequency-specific spectral features that characterize a vowel category, rather than to any particular set of frequencies the mother uses to produce a vowel” (Kuhl et al. 1997, 685).
If so, some form of linguistic unit must guide analysis so that variety is not a hindrance to categorization.35
31The “vowel triangle” is a schematic representation of the range of tongue positions possible within the human oral tract. This BTR effect appears to be linked to content rather than function words (van de Weijer 2001).
32For treatment of “vowel systems” across languages, see Hockett 1955, 83f, and Ladefoged 2001.
33Even phonetic units do not “exist” in an expanded vowel space without further analysis on this space.
34With an expanded triangle, there are more sound possibilities, each of which may or may not be significant.
35Kuhl et al. 1997 provide an analysis of their acoustic data in a form that corresponds to compact–
diffuse and grave–acute distinctions, linguistic features which were proposed by Jakobson to innately guide language acquisition (chapter 8).