ICP / INPG

This study presents analysis results of 4 French children's early sound production, beginning with the onset of canonical babbling and comprising the following first seven weeks. The data is part of a more ambitious project, which includes 6 children recorded over a period of 9 months. The infants have been recorded in their family environment using audio-visual equipment. Recording sessions took place once every fortnight, and lasted for approximately 45 minutes each.

Numerous studies on developing productive phonology have chosen the onset of canonical babbling as an appropriate starting point to set in observation (Smith et al., 1989; Robb & Bleile, 1994, Davis & MacNeilage, 1995; Oller et al., 1997...). From this stage onwards, infants produce the characteristic rhythmically repetitive series of open-closed alternations resulting in our perception of strings of "syllable-likes". Canonical babbling seems to be hard-wired and develops independently of cognitive capacities (Down-syndrome babies start canonical babbling at the same age as their typically-developing peers, Stoel-Gammon, 1998). Determination of the acquisition of phonetic and phonological versatility can be limited to the following factors that govern peripheral behavior: – the acquisition of motor skills which includes the integration of self-generated movements and the consequent proprio- and exteroceptive feedback, – audio-visual perception of exterior stimuli, – the auditive perception of the self-generated sounds. Progress towards mastery of articulation, and phonology, by the child depends essentially on the interplay between these factors and on the acquisition of the corresponding perceptuo-motor experience.

The role of motor skills and proprio- and exteroceptive feedback have not been discussed to the same extent as the role of audio-visual perception in phonological development, which is mostly due to methodological/ethical constraints on producing data. Notwithstanding, intriguing data has been collected, permitting to infer about respective states of motor skill acquisition: – acoustical data pertaining to the locus equation paradigm (Sussman, Duder, Dalston & Duder, 1999) which allows for the assessment of coarticulation in CV sequences, basically lip and tongue movements; – even more recently, articulatory data on lip and jaw movements in 1-, 2-, 6-year olds and young adults confirmed the necessity of parsimony in crediting infants with speech motor control capacities: the dominance of jaw movement over little or no active lip movement in /baba/-type utterances in young children has been measured instrumentally (Green, Moore, Higashikawa & Steve, 2000).

Possibly the best-known framework for motor skill acquisition, the "Frame/Content Theory" (MacNeilage, 1998), posits that motor skills relevant to infant speech are limited to control of the mandible only. Babbling sounds, typically repeated in sequences of what we perceive as consonantal and vocalic elements, are considered to be mere consequences of the opening/closing phases of mandibular oscillations: the closing phase yields a consonantal sound, the opening phase a vocalic sound. A typical /baba/ sequence would then be the emergent byproduct of mandibular oscillations with accompanying phonation.

In our study, we aimed at tracing back motor constraints from canonical babbling output. Any study dealing with preverbal utterances is confronted with the question of how to describe infant sound production adequately. Especially if one wants to obtain information about the movements and their restrictions at the basis of babbling sound patterns, one sees little interest in symbol-based transcription systems developed for adult sound production.

Hence, another perspective one can adopt is to describe (as opposed to "transcribe") what infants actually do within their articulatory and phonatory space. For this purpose, Koopmans-van Beinum & van der Stelt (1986, 1998) developed the "Amsterdam System for Transcription of Infant Vocalizations" (AMSTIVOC), based on the description of universal principles of movement proper to the human speech production instrument. Within this system, the notion of "utterance" is defined and delimited by the respiratory cycle; then, each utterance is described according to phonatory and articulatory characteristics. Examples of those characteristics include occurrence, direction and continuity of phonation, voice onset and quality, intonation, number, type and location of articulatory movements within one utterance, place and manner of closures, aperture and front-backness of vowel-likes. This system – applied several times in normalization, diagnostic and mother-infant interaction research (van der Stelt & Koopmans-van Beinum, 1986; Clement et al., 1995; van der Stelt, 1993) – has been further refined to meet the special needs of a precise description of the sound patterns of canonical babbling. Within this framework, features of phonation and articulation can be described independently from one another. Thereby, this sound production description accounts of actual articulation events or phases in a close-up fashion that tries to be more faithful than the segmental transcription of IPA

The Grenoble version of AMSTIVOC will be presented as a tool; moreover, the analysis results of its application to the first 7 weeks of canonical babbling of 4 infants will be discussed.

One striking result is the massive occurrence of so-called silent jaw wags (SJW, e.g. babbling utterances without phonation, cf. Meier, McGarvin, Zakia & Willerman, 1997), which were produced reliably in all of the 13 analyzed recording sessions. The ratio of the number of SJW over the number of phonated jaw wags varies between a minimum of 10% and up to 60% per session (mean: 35%). SJW do not only occur separately, but also at the beginning, in the middle and at the end of phonated babbles. This, the great majority of voiced over voiceless Cs, and the circumstance that phonated babbles also occur in an ingressive mode (during the inspiratory phase as opposed to "normal" expiratory speech) all indicate that the coordination of the three speech subsystems, namely respiration, phonation, supraglottal articulation, has yet to be learned.

The issue of CV cooccurrences (MacNeilage & Davis, 1995) will be addressed and confirmed. The basic idea of the Frame/Content Theory, mandibular oscillations carrying the otherwise passive articulators tongue and lower lip, signifies a major motor constraint. Consequently, the tongue is predicted not to move actively within CV sequences (within the spatio-temporal frame of mandibular oscillations). This leads to the prediction of biomechanically motivated cooccurrences: all sessions taken together, the infants produced the labial C - central V cooccurrence with a majority of 63% over the two other possible combinations (labial C - front V or labial C - back V). The coronal C - front V cooccurrence was observed in 70% of all CV occurrences where C was coronal. Given the high number of analyzed CV sequences (2123), these results are significant at the p < .001 level. The back C - back V cooccurrence has not been validated.

Overall frequencies of labial vs. coronal articulation in CV sequences yield a majority of coronals, although proportions change from the 1^st to the 2^nd, and again from the 2^nd to the 3^rd recording sessions. Upon analysis of each single infant's production, we can draw clearly different profiles for each of them. With regard to their production during the very first recording session, child L produced 72% of labial CVs, child C 55% of palatal CVs, child A 77% of coronal CVs and child N produced distributed proportions of 33% labial, 35% coronal, 10% palatal and 10% velar CVs (the missing 12% are double place articulations, which occurred in all children in variable proportions). Those profiles will be confronted with the notion of a "pure frame" (MacNeilage & Davis, 1995) and results from articulatory modelling (Vilain, Abry, Badin & Brosda, 1999).

The observed proportions of reduplicated vs. variegated babbling will be discussed with regard to their implications. Although we applied severe criteria (motivated by an articulatory distinction: front-back movement of the tongue qualifies for the label "variegated", all other changes like manner and minor front-back movements result in the label "reduplicated"; exact criteria will be given), variegation occurred in 45% of all multicyclic utterances.

Even though many questions remain unanswered, we can already state that articulatory motor control can be seen as a necessary, yet surely not sufficient, factor to be taken into consideration when trying to explain phenomena of productive phonological development.

Stefanie Brosda1, Rafael Laboissière¹ & Christian Abry²

<brosda,rafael,abry>@icp.inpg.fr

1) ICP / INPG

UPRESA CNRS 5009

46, av. Félix Viallet

38031 Grenoble CEDEX 1

2) ICP - Stendhal

UMR CNRS

Université Grenoble III

BP 25

38040 Grenoble CEDEX 9