The role of the input on the development of the LC bias: A crosslinguistic comparison

Highlights

• Cross-linguistic differences reflecting the properties of the lexicons contrasted.

• Specific perceptual LC/CL biases acquired on the basis of lexicons’ properties.

• Emergence of a perceptual LC bias at 10 months in French-learning infants.

• Emergence of a perceptual CL bias at 13 months in Japanese-learning infants.

Abstract

Previous studies have described the existence of a phonotactic bias called the Labial–Coronal (LC) bias, corresponding to a tendency to produce more words beginning with a labial consonant followed by a coronal consonant (i.e. “bat”) than the opposite CL pattern (i.e. “tap”). This bias has initially been interpreted in terms of articulatory constraints of the human speech production system. However, more recently, it has been suggested that this presumably language-general LC bias in production might be accompanied by LC and CL biases in perception, acquired in infancy on the basis of the properties of the linguistic input. The present study investigates the origins of these perceptual biases, testing infants learning Japanese, a language that has been claimed to possess more CL than LC sequences, and comparing them with infants learning French, a language showing a clear LC bias in its lexicon. First, a corpus analysis of Japanese IDS and ADS revealed the existence of an overall LC bias, except for plosive sequences in ADS, which show a CL bias across counts. Second, speech preference experiments showed a perceptual preference for CL over LC plosive sequences (all recorded by a Japanese speaker) in 13- but not in 7- and 10-month-old Japanese-learning infants (Experiment 1), while revealing the emergence of an LC preference between 7 and 10 months in French-learning infants, using the exact same stimuli. These crosslinguistic behavioral differences, obtained with the same stimuli, thus reflect differences in processing in two populations of infants, which can be linked to differences in the properties of the lexicons of their respective native languages. These findings establish that the emergence of a CL/LC bias is related to exposure to a linguistic input.

Introduction

Studies focusing on the analysis of the lexicons of various natural languages have revealed the existence of different phonotactic tendencies which occur consistently and crosslinguistically. The study of the origins of these consistent patterns across languages is a recurrent theme in the study of human cognition, as they are taken to reflect properties and constraints of the mechanisms underlying language processing and acquisition. For example, at the syllabic level, languages privilege open syllables (Consonant–Vowel, e.g. /ma/) over closed syllables (Vowel–Consonant, e.g. /am/; Kawasaki-Fukumori, 1992, Rousset, 2003) and they privilege Consonant–Vowel sequences sharing the same place of articulation (e,g. /de/ or /ko/ rather than /ke/ or /do/; MacNeilage & Davis, 2000a). At the intersyllabic level, languages have been shown to favor syllable sequences where consonants are articulatorily different (e.g. /baga/) over reduplications (e.g. /baba/; Rochet-Capellan & Schwartz, 2005). In addition, among these variegated forms, sequences starting with a labial consonant followed by a coronal consonant (e.g. /bat/) are privileged over the opposite pattern (e.g. /tap/; MacNeilage et al., 1999, MacNeilage and Davis, 2000b, Vallée et al., 2001, Gonzalez-Gomez and Nazzi, 2012a). This tendency is known as the Labial–Coronal (LC) bias. The present study will explore one factor that determines the emergence of the LC bias during development, frequency in the input, therefore focusing on the role of perceptual constraints by comparing infants learning either French, a language with an LC bias in its lexicon, or Japanese, a language that had been proposed to have a lexicon that does not have a clear LC bias (an issue we come back to in more details in the following). This kind of study is crucial to better understand influences of the perceptual and production systems as they support a robust phonotactic phenomenon observed cross linguistically.

The Labial–Coronal bias was first found in early production studies. During the 50-word-stage (12–18 months), infants tend to produce more Labial–Coronal (LC) than Coronal–Labial (CL) structures (Ingram, 1974, Locke, 1983, MacNeilage and Davis, 2000a, MacNeilage et al., 1999). This tendency was found in 9 out of the 10 English-learning infants tested by MacNeilage et al. (1999) and in 21 out of the 22 infants learning English, German, Dutch, French or Czech (MacNeilage & Davis, 2000a). Furthermore, the LC bias has not only been found in infant production studies but also in two typological studies examining 17 different languages (English, Estonian, French, German, Hebrew, Maori, Quechua and Spanish: MacNeilage et al., 1999; Afar, Finnish, French, Kannada, Kwalkw’ala, Navaho, Ngizim, Quechua, Sora and Yup’ik: Vallée et al., 2001).

The first interpretations for the origins of this bias were articulatory. One interpretation was offered within the frame-content theory (MacNeilage and Davis, 1990, MacNeilage, 1998, Davis and MacNeilage, 2004). It was proposed that infants tend to begin an utterance with a lip closure and then follow with a tongue contact closure (MacNeilage and Davis, 2000a, MacNeilage et al., 2000). Since Labial-Vowel sequences are proposed to be pure frames (sequences that are produced by lip closure enabled by rhythmic mandibular oscillations without independent movement of the tongue within the sequence), while Coronal–Vowel sequences are fronted frames (sequences based on rhythmic mandibular oscillations with pre-positioning of the tongue at the alveolar ridge; no independent movement of the tongue from this pre-set position within the CV sequence), infants would tend to start with a labial consonant and then add a coronal one, rather than the other way round, resulting in the LC bias.

A second articulatory explanation known as the “Labial–Coronal Chunking Hypothesis” was proposed by Sato, Vallee, Schwartz, and Rousset (2007). Their results from adult speeded articulation tasks showed that when French adults produce Consonant–Vowel–Consonant–Vowel sequences containing a labial (L) and a coronal (C) consonant at a fast articulatory rate, their productions tend to shift to Consonant-Consonant–Vowel sequences. When this happened, French adults produce LC-vowel sequences more frequently than CL-vowel sequences (e.g. both /bete/ and /tebe/ shifted to /b’te/). Based on these results (extended to Japanese adults by Tsuji, Gonzalez-Gomez, Medina, Nazzi, & Mazuka, 2012), Sato, Vallee, Schwartz, and Rousset (2007) suggested that the LC bias might be explained by the higher articulatory stability of LC sequences compared with CL ones. It is important to emphasize that both articulatory explanations assume the existence of a language-general (that is, non-language specific) LC bias, predicting similar patterns across languages.

More recently, the importance of considering perceptual explanations for the LC bias has been stressed (Nazzi et al., 2009, Gonzalez-Gomez and Nazzi, 2012a). This perspective hypothesized that the LC bias results from infants’ exposure to the linguistic input. This hypothesis was based on the previous observation of links existing between infants’ preferences for specific sound sequences and their frequencies in the language. Indeed, prior studies have shown that by 9 months of age, infants have become sensitive to the phonotactic properties of their native language, preferring legal over illegal sequences (Jusczyk et al., 1993a, Friederici and Wessels, 1993, Sebastián-Gallés and Bosch, 2002), and also more frequent over less frequent phonotactically legal sequences (Jusczyk, Luce, & Charles-Luce, 1994). With regard to the LC bias, it has been shown that LC sequences are significantly more frequent than CL sequences in the lexicons of 17 different languages (MacNeilage et al., 1999, Vallée et al., 2001). According to the perceptual-based perspective, the LC bias might thus be a result of infants’ exposure to a linguistic input containing more LC than CL sequences.

The results of three recent perceptual studies seem to bring support to this perceptual hypothesis. The head-turn preference procedure (HPP) was used to explore French-learning infants’ preference for lists of LC or CL sequences (words or pseudo-words in French pronounced by a native female speaker). Results showed that between 7 and 10 months of age, French-learning infants start preferring the lists corresponding to the LC sequences, the significantly more frequent phonotactic structure in French (Nazzi et al., 2009, Gonzalez-Gomez and Nazzi, 2012a, Gonzalez-Gomez and Nazzi, 2012b). This was found even before infants start producing LC or CL structures, in both full term and preterm infants of the same chronological age, disfavoring the production interpretation; moreover, the data from the preterm infants further rules out the possibility of an emergence solely based on maturational level (Gonzalez-Gomez and Nazzi, 2012a, Gonzalez-Gomez and Nazzi, 2012b).

Importantly, the predictions of the perceptual-based account change depending on the properties of the input: thus, if CL sequences are favored over LC sequences in the input, this account predicts a preference for CL over LC sequences. This contrasts with the above-mentioned articulatory explanations, which predict a general preference for LC patterns. In order to critically assess these perceptual-based predictions, Gonzalez-Gomez and Nazzi (2012c) conducted a more detailed analysis of the French lexicon, finding that the LC bias was not homogeneously present across consonantal classes in French: while the LC bias is clearly present for plosive (plosive consonants are produced by stopping airflow through the mouth, i.e., labials /p/ and /b/ vs. coronals /t/ and /d/) and nasal sequences (nasals are produced by airflow through the nose, i.e., labial /m/ vs. coronal /n/), this is not the case for fricative sequences (fricatives are produced by forcing airflow through a narrow opening in the mouth i.e., labials /f/ and /v/ vs. coronals /s/ and /z/). Accordingly, Gonzalez-Gomez and Nazzi (2012c) tested the level of generalization at which these phonotactic acquisitions operate. In a series of experiments, 10-month-old French-learning infants’ preferences for LC or CL structures in plosive, nasal and fricative sequences were evaluated. The results indicated an LC preference for plosive and nasal sequences, but a CL preference for fricative sequences, suggesting that the LC bias reflects the properties of the input and is acquired at the level of classes of consonants defined by their manner of articulation.

The results of Nazzi, Bertoncini, and Bijeljac-Babic (2009) and Gonzalez-Gomez and Nazzi, 2012a, Gonzalez-Gomez and Nazzi, 2012b, Gonzalez-Gomez and Nazzi, 2012c thus seem to suggest that the early LC/CL biases are perceptually-based, reflect infants’ learning about the structural regularities of the French lexicon, and result from exposure to the input rather than from maturation. Importantly, the results of Gonzalez-Gomez and Nazzi (2012c) on French further suggest that the perceptual LC bias (found for plosives and nasals) can be reversed in case the input shows a CL bias (for fricatives). Therefore, the contribution of Gonzalez-Gomez and Nazzi (2012c) is to establish that different types of consonants lead to different biases in the same infant population, as predicted by input frequencies in their native language. However, it cannot be completely excluded that for some yet-unidentified (linguistically-relevant or characteristic of the stimuli used) acoustic properties, plosives and nasals are processed differently than fricatives, leading to opposite LC/CL biases, independently of input frequencies¹.

Given this, another crucial step in showing the importance of early perceptual acquisition of input statistics in the emergence of LC/CL biases is to show that the same stimuli are processed differently by infants learning languages that differ on this dimension at the input level. Therefore, as further motivated below, the present study used the same stimuli (plosive LC and CL words produced in Japanese) to investigate the emergence of different LC/CL preferences in Japanese- and French-learning infants reflecting input differences in these two languages. This will also allow us to generalize to a language other than French the study of the emergence of LC/CL biases.

The choice for Japanese was originally motivated by the need to test a population learning a language having a lexicon that does not have a clear LC bias (contrary to French). According to MacNeilage et al. (1999) this is the case for Japanese. Their results showed not only that the Japanese lexicon does not have an LC bias, but that it tends to have the opposite pattern, that is a CL advantage. However, their results were based on a very small sample of words (68 words extracted from a travel dictionary), calling for more thorough analyses. Employing a larger database, Tsuji et al. (2012) found that the adult Japanese lexicon in fact has a general LC bias. However, a more fine-grained analysis based on the findings of Gonzalez-Gomez and Nazzi (2012c) revealed that this bias is not homogenously distributed, but changes across consonant classes defined by manner of articulation: the overall LC bias extended to sequences of nasals, while a CL bias was found for plosive sequences.

Therefore, exploring the processing of plosive sequences in Japanese appeared as a good basis for further investigation of the perceptual factors determining the LC/CL biases. In this context, Tsuji et al. (2012) explored Japanese adults’ production and perception of plosive sequences containing a labial consonant (/p/ or /b/) and a coronal consonant (/t/ or /d/). The results revealed that Japanese adults have an LC bias in production, supporting the explanations in terms of articulatory constraints (MacNeilage and Davis, 2000a, MacNeilage et al., 2000, Sato et al., 2007). However, Japanese adults did show a perceptual CL bias for these plosive sequences, showing the influence of language exposure on perceptual biases as had been previously suggested (Nazzi et al., 2009, Gonzalez-Gomez and Nazzi, 2012a, Gonzalez-Gomez and Nazzi, 2012b, Gonzalez-Gomez and Nazzi, 2012c). Based on these results, Tsuji et al. (2012) concluded that in adulthood there are two different biases, a productive LC bias due to constraints of the articulatory system, and a perceptual CL bias based on distributional frequencies in the lexicon. Moreover, these findings suggest that there might be a default LC bias, defined by articulatory constraints, which will partly (but not entirely) determine the structure of the lexicon across languages, and will lead to both production and perceptual LC biases. When exceptions to the LC bias are found in the input (for example, on fricative-fricative sequences in French, or on plosive-plosive sequences in Japanese), then a perceptual bias for the opposite CL bias is observed (perceptual CL bias for fricative sequences in French-learning infants, and plosive sequences in Japanese adults). The present study will explore whether a perceptual CL bias for plosive sequences can be found in Japanese-learning infants.

The role that input properties play on the development of the perceptual LC bias was explored using exactly the same LC and CL sequences, recorded by a native Japanese speaker in two different infant populations. Given the properties of the Japanese and French lexicons, the input-based explanation will predict opposite preference patterns for plosive sequences: a CL preference for Japanese-learning infants and an LC preference for French-learning infants. Given previous data on French-learning infants, no biases should be found at 7 months of age, but language-specific biases should appear from 10 months onward (Nazzi et al., 2009, Gonzalez-Gomez and Nazzi, 2012a, Gonzalez-Gomez and Nazzi, 2012b, Gonzalez-Gomez and Nazzi, 2012c).

Before conducting the perceptual studies, different frequency analyses were conducted in the Japanese lexicon, both in an infant-direct speech (IDS) corpus and in an adult-direct speech (ADS) corpus. This is important given that MacNeilage et al. (1999) used a corpus having a very small number of words, and that Tsuji et al. (2012) used only adult corpora. Thus the present analyses will first allow the verification of these phonotactic properties in the Japanese lexicon. Second, they will establish whether IDS shows a similar or a different pattern compared to ADS.