Fillers as signs of associative learning

Young children often insert 'fillers' in their first multiwordutterances: vocalizations that do not correspond to conventional words. For instance, it is hard to determine the meaning of the syllables [m] and [∂] in utterance (1). So far, fillers have been investigated in French (i.a. Veneziano & Sinclair, 2000; Kilani-Schoch & Dressler, 2000), English (i.a. Peters & Menn, 1993; Feldman & Menn, 2003), Spanish (i.a. Lopez-Ornat, 2001), but barely in Dutch. We will try to fill this gap in this paper by analyzing the longitudinal data of one Dutch child, Cato. (1) [m] pick [∂] flowers (English speaking boy, age1;6, from Peters & Menn, 1993) Methodology The data of Cato are part of the CLPF database (Fikkert, 1994, Levelt, 1994, available through CHILDES). They span a period of 10 months from age 1;10 (MLU=1.1, SD=0.4) until age 2;7 (MLU=3.9, SD=2.3). Bimonthly observation sessions were transcribed orthographically and phonemically. Fillers were not annotated consistently in the original database. First of all, most targetless vowels were attached to words. We decided to detach these vowels, except when they could not be interpreted as an affix (Taelman, 2004). Secondly, there was a confound between articles and fillers in the original database, due to similarities in distribution and phonological form. We recoded the fillers and articles based on the phonological form of the child’s rendition: if the form was a close approximation to the phonological form of an article, we interpreted it as an attempt at an article. All other instances were annotated as fillers, independently of the location in which they appeared. Based on this definition, we found 621 fillers. Most of them were schwas or schwa-like vowels (sometimes preceded by a consonant). Some were nasals: n, m, N. Nasal fillers At the start of observations, articulatory factors determined the insertion of nasal fillers to a high extent. Until age 2;0, words that started with a stop were often preceded by an homorganic nasal filler: /n/ before /d/ or /t/, /m/ before /b/ or /p/. Probably, this homorganic nasal consonant originated from the child’s inadequate control of nasality during mouth closure. After age 2;0, the percentage of nasal fillers before stops diminished, signalling a better control of nasality. Now, nasal fillers frequently appeared in a new context: before nouns starting with an /h/. These nasals are reminiscent of the nasal in the Dutch indefinite article /∂n/ that often occurs before nouns in the input. Our analysis revealed that the child had inferred a statistical rule from the occurrence of a nasal before /h/: she also produced nasal fillers before verbs starting with an h, and before attempts at targets that did not begin with /h/ in the adult model, but that were produced with an initial /h/ by the child.