Semantic mediation in word recognition in beginning readers

Substantial controversy has surrounded the premise that semantic information may make an important contribution to the development and function of fast and efficient word recognition. This debate has been fuelled by the alternative explanations of semantic processing offered by the two major models of word recognition (Parallel Distributed Processing, PDP and Dual Route Cascade, DRC). Whilst some data has indicated facilitative effects of semantic information for all reading abilities and word types, other data suggests that these are restricted to people with poor decoding skills and to conditions where word recognition is more difficult, such as for low frequency words. Large variation in the age groups used to explore semantic effects has substantially contributed to the mixed findings in the literature, as have results from studies examining reading acquisition when compared to studies examining word recognition processes. The current series of experiments aimed to reconcile these discrepancies by a) investigating the contribution of specific forms of semantic information to the word learning of beginning readers and b) examining whether semantic information affects orthographic and phonological decisions in beginning readers. In Experiment 1, Grade 1 children (n = 77, Mage= 6 years, 6 months) of varying decoding abilities were trained to read non-words with support from different forms of semantic information (high meaning, low meaning, high imageability, low imageability). The results showed that high meaning stimuli produced an accuracy advantage over all other semantic stimuli for decoders of good and average ability. However, poor decoders received equal benefit from high and low meaning stimuli, with both producing an accuracy advantage over the image stimuli. Although facilitation effects were evident for reaction time performance overall, these did not differ as a function of decoding ability. Experiment 2 investigated whether the facilitative effects of semantics for word learning were also present when this information was used to prime lexical decisions. New high and low meaning and high imageability stimuli were developed and the low imageability condition was replaced with a non-semantic control condition. A new group of Grade 1 readers (n = 61, Mage = 6 years, 9 months) responded to word and pseudo homophone pairs, preceded by the presentation of the semantic primes. Semantic facilitation effects were consistent with those found in Experiment 1, however, these varied between accuracy and reaction time measures. For good and average decoders, semantic facilitation effects were restricted to reaction time advantages, whereas poor decoders received a performance benefit for accuracy only. Experiment 3 aimed to extend these results by investigating whether semantic priming effects could also be found for a phonological decision task and whether performance effects for both tasks differed as a function of word frequency. A new group of Grade 1 readers (n = 45, Mage= 6 years, 9 months) responded to the lexical decision task and to a phonological decision task, created by replacing previous words with pseudoword and pseudohomophone pairs. Sets of high and low frequency words and associated semantic primes were developed for each task, maintaining meaning and control conditions only. The results for the lexical task were consistent with those of Experiment 2 and showed that the contribution of high and low levels of semantic information differed depending on the frequency of the words presented. Findings for the phonological task indicated that although performance was not affected by the frequency manipulation, semantic priming effects for good and average decoders were similar to those in Experiments 1 and 2. In contrast, performance data failed to show semantic priming effects for poor decoders. Overall, the results of these studies suggest that the presentation of meaning related information can facilitate both word learning and children's ability to recognise words based on orthographic and phonological knowledge. The semantic facilitation effects support previous research findings, however, they indicate that the influence of semantic information can extend to readers of good decoding ability and the processing of high as well as low frequency words. The findings do not fit neatly with current models of word recognition, however, they show predominant support for the PDP model in which semantic processing is integral. The results indicate that beginning readers could benefit from reading instruction methods that provide specific semantic information and that optimal results will be achieved when this information is meaning-based rather than image-based and of a level suitable to the decoding ability of the reader.