The recent burgeoning of cross-linguistic research in the ﬁeld of reading and spelling development has exposed how central the English language has been to theorizing about literacy acquisition. Early work by Heinz Wimmer, in particular, brought this to the forefront of discussion with his documentation of differences in the expression of developmental dyslexia in German compared to English (Wimmer & Hummer, 1990; Wimmer, 1993). A process of change initiated by this work has led to more ﬂexible models of reading development in which the concept of orthographic depth (Frost et al., 1987) and related linguistic differences now shape the formulation of the acquisition process (e.g. Ziegler & Goswami, 2005). As horizons have broadened, one further consequence of a research literature
deeply rooted in the English language is the customary restriction of interest to the case of monosyllabic words. The connectionist models which appeared in the 1990s are typical but certainly not isolated examples of this (Seidenberg & McClelland, 1989; Plaut et al., 1996; Harm & Seidenberg, 1999). Traditionally, syllabic or morphemic units, the components of polysyllabic words, featured only in later or higher level phases of developmental models, for example, Frith’s (1985) orthographic stage or Ehri’s (1992) consolidated phase. Whilst this line of research has been productive for English, which contains an unusually high proportion of monosyllables (approximately 12 per cent of the English words in the CELEX database are monosyllables), many languages contain few monosyllables and beginning readers encounter mostly polysyllabic words. Even within English early reading texts, the number of polysyllabic words is not negligible (see Table 3.1). This raises the issue of how young children learn to read polysyllabic words. In the case of monosyllables, there is a longstanding expectation that the lexicon is
organized in terms of onset-rime segments which deﬁne families or neighbourhoods of rhyming words (Glushko, 1979). One advantage of this organization in English is that the rime unit offers a level of orthographic consistency, which leads to more accurate pronunciation of words than correspondences based on vowel or ‘initial consonant plus vowel’ segments (Treiman et al., 1995). Accordingly, both dual-route (Patterson & Morton, 1985) and connectionist models of word recognition (Seidenberg & McClelland, 1989) have made reference to this onset-rime level of structure. One view is that rhyming skills promote an organization in the orthographic lex-
icon based on the rime unit (Goswami & Bryant, 1990; Goswami, 1993). According to Goswami (1993) ‘orthographic analysis is founded in phonological skills’ with rhyming skill playing an early and formative role. Recently, Duncan et al. (2007)
examined how this view might adapt to the polysyllabic domain by investigating children’s conception of rhyme for disyllabic stimuli. Children were asked to play a game in which they produced a rhyme for target monosyllables and then subsequently for disyllables. When the disyllabic stimuli had initial stress, the data showed a strong commitment to a phonological structure composed of an onset plus a unit termed the superrime. This is a higher order unit introduced by Berg (1989), which, together with the onset, heads the internal structure of initial-stress disyllables (see Figure 3.1). The superrime itself can be split into a rime unit and a following syllable which retains its internal onset-rime structure. Awareness of monosyllabic rhyme has been thought to lead to the use of ortho-
graphic rime analogies in reading (Goswami & Bryant, 1990; Goswami, 1993). If this link between rhyme and reading holds for disyllabic words, the results of Duncan et al. (2007) suggest that orthographic analogies may not always involve the rime unit. For example, it is the superrime, and not the rime unit, which appears to be the source of rhyming similarity for disyllabic words with initial stress. This suggests that if children derive the common spelling patterns from rhyming categories of disyllabic words, it is orthographic units at the level of the superrime which should be represented. Supportive evidence comes from Goswami et al. (1998), who investigated analogy use when reading monosyllabic and disyllabic nonwords. Orthographically familiar nonwords shared either a rime (monosyllables) or a superrime (disyllables) with a real word (e.g. cake ! dake; comic ! bomic). The orthography of these nonwords was then altered to produce structures which maintained the original phonological pattern but were orthographically unfamiliar (e.g. daik; bommick). Goswami et al. found speed and accuracy advantages for the orthographically familiar nonwords amongst children with reading ages of 7 years and above. Goswami et al. (1997) replicated this ﬁnding and extended it to trisyllabic stimuli constructed in a similar manner (e.g. daffodil ! taffodil, tafoddyl).