The nature and use of sub-lexical inferences in early reading
This thesis considers the nature and use of orthographic sublexical inferences by 6 year old children. Previous research by Goswami (1986) appears to demonstrate that even young children are able to use sublexical inferences. Typically, Goswami has shown children a clue word such as 'beak' and then shown previously unknown analogous target words such as 'peak' sharing orthographic rimes (medial vowel and terminal consonant(s)) with clue words, 'bean' which shares the head (initial consonant(s) and medial vowel), and control words such as 'bank'. Children typically read more target words which share rimes with taught clue words than other targets, suggesting that rime inferences are privileged in early reading. One problem with Goswami's task is that both clue and target word are presented concurrently, possibly supporting the strategic use of inferences. Experiment 1 therefore contrasted inference use when clue words were either pretaught or concurrently presented. Inference use was evident in the presence of concurrent reminders of clue word pronunciation, but was not evident when a clue word was pretaught. Subsequent experiments investigated inference use when children were given greater prior exposure to clue words sharing orthographic and phonological patterns with targets, but where concurrent prompts were avoided. Rime inferences (e.g. 'leak' - 'peak') and vowel inferences (e.g. 'meat' - 'peak') were contrasted. Results revealed equivalent improvements for both sets of words, suggesting a) that children can make inferences in the absence of concurrent clue words as long as they have had substantial exposure to other words sharing analogous letter-sound patterns, and b) that there is no advantage for words sharing rimes over words sharing other orthographic units such as vowel digraphs when tested under such conditions. The results of these studies and parallel correlational studies are interpreted in terms of models of vowel digraph inferences.