Markedness vs . Frequency Effects in Coda Acquisition

The problem of individual variation has long plagued the field of language acquisition (e.g., Bloom, 1970; Nelson, 1981). This has been especially true in the area of phonology (e.g., Stoel-Gammon, 1985; Vihman, 1993; Vihman & Greenlee, 1987). Individual preferences for certain consonants begin at the babbling stage of development and persist into the formation of children’s first words (Vihman, Macken, Miller, Simmons & Miller, 1985). Other individual differences emerge later, with variation reported in patterns of syllable structure (e.g., Barlow, 1997; Ingram, 1978; Rose, 2000), word structure (e.g., Fikkert, 1994) and filler syllables (e.g., Peters & Menn, 1993) This high level of individual variation in children’s early phonologies led Fergusson and Farewell (1975) to propose that each child had an individual ‘learning style’. This suggests that learning language is a random process, with few constraints on the learning paths a child might take. Rulebased approaches to the acquisition of phonology (e.g. Smith, 1973), where many unmotivated phonological rules were invoked to capture a child’s developing phonological system, only served to reinforce this notion of randomness. However, as Pater & Barlow (2002) demonstrate, the types of grammars attested are actually a restricted set of possible grammars. That is, the types of possible individual variation appear to be constrained. Several decades ago Jakobson (1941/68) proposed that children would acquire the unmarked sounds of language first, and only later acquire the more marked segments (i.e., those less frequently attested in the world’s languages). Jakobson thought that those segments most commonly found crosslinguistically (e.g., /p/, /n/, and /a/) would be the easiest to perceive and produce, and would therefore be the first to be acquired. Although Jakobson’s claims have never been substantiated at the segmental level, markedness proposals for explaining developmental learning paths have