Origin of Serial-Output Complexity in Speech

Abstract

During the babbling and early speech stages (7–18 months), infants strongly prefer to repeat the same consonant and, less often, the same vowel in multisyllabic utterances (e.g., /baba/), even when the target word has two different consonants. These utterances apparently have a simple structure, typically consisting only of repeated cycles of mandibular oscillation (called frames), with other speech articulators (lips, tongue, soft palate) in either a resting or nonresting position, but not moving independently (MacNeilage, Davis, & Matyear, l997). How, then, does an infant increase serial-output complexity by reducing this consonant harmony to the statistically low levels required by adult languages? The most well-known relatively discrete step toward this goal during the first-50-word stage (12–18 months) is to begin the first syllable with a labial consonant (involving the lips) but then follow it with a coronal consonant (involving the tongue front: e.g., “man,” “bad,” “pat”). In a review of seven reports involving five different language communities (MacNeilage & Davis, 1998), this pattern was observed in 21 of 22 infants, including 2 who produced this pattern when attempting words with the opposite sequence (e.g., “soup” → “pooch”). Table 1 is a summary of results from a study of the words of 10 infants during the 50-word stage (MacNeilage et al., 1997). Consonant-vowel-consonant (CVC) and consonant-vowel-consonant-vowel (CVCV) words were studied. The labial consonants were /b/, /p/, and /m/, and the coronal consonants were /d/, /t/, and /n/. Nine of the infants showed the trend toward more labial-coronal (LC) than coronal-labial (CL) sequences, and the 10th infant showed no trend. The overall ratio of LC to CL sequences was 2.55. This preference is not only characteristic of infants, but also present in languages. Table 2 shows ratios of LC to CL sequences in the first and second consonants (with an intervening vowel) in words in 10 diverse languages representing several different major language families.1 The preferences in all these languages are statistically significant beyond the .001 level in chi-square tests, except for Swahili and Japanese, which alone shows a countertrend. The results suggest that languages perpetuate a strong infant preference. The common occurrence of this preference in infants and adults may be of fundamental significance. Infants are unlikely to be simply copying adults, as their preferences tend to be stronger than adult preferences (especially in the 2 infants who reversed adult CL sequences), even though they have as yet learned only a few words. Some considerations suggest that a syllable beginning with a labial consonant is simpler to produce than one beginning with a coronal consonant. First, although the former may be made simply with a cycle of mandibular oscillation, the latter requires an additional tongue-positioning movement (MacNeilage & Davis, 1998). Second, there is an increase in labial consonants relative to coronal consonants when infants go from the babbling stage to the firstword stage (Boysson-Bardies et al., 1992), a trend that can be interpreted as a regression toward easier production forms in the face of the new functional demand to interface the motor system with the lexicon. Steger and Werker (1997) have recently reported an analogous example of apparent simplification of operation at the signalprocessing level in the presence of demands associated with concurrent building of a mental lexicon for speech perception. Infants show less discrimination of fine phonetic detail when required to pair words with objects than they show in syllable discrimination tasks. The usual excess of coronals over labials in babbling seems to be counterevidence to this labial-ease hypothesis. However, it may be due in part to the fact that there are a great deal more coronals than labials in the typical adult language being assimilated. It has been shown that babbling infants have some sensitivity to major differences in consonant frequencies between languages (Boysson-Bardies & Vihman, 1991). The third consideration favoring the labial-ease hypothesis is that infants who have been prevented from vocalization during the babbling and early speech periods by early tracheostomies show a very strong preference for labials over coronals in their initial posttracheostomy speech efforts (e.g., Locke & Pearson, 1992). The achievement of increased serial-output complexity by beginning a word with a simple cycle of mandibular oscillation and adding a tongue movement to the next cycle is not presently explicable in terms of the phonological component of a genetically specified universal grammar (Chomsky, 1986). The pattern is more likely to be a result of self-organization. Just as infants may more easily simulate serial complexity of a word by beginning with a mandibular cycle not including the tongue and then adding tongue action, earlier hominids, under pressures for increases in the size of their linguistic message set,