A connectionist approach to making the predictability of English orthography explicit to at-risk beginning readers: evidence for alternative, effective strategies
V W Berninger,R D Abbott,R Brooksher,Z Lemos,S Ogier,D Zook,E Mostafapour
DOI: https://doi.org/10.1207/S15326942DN1702_06
Abstract:A case is made (and illustrated with empirical data with children) for connectionist models that are not only computationally explicit but also instructionally explicit. First-graders (N = 128) at the bottom of their classes in reading (average 11.5 percentile on nationally normed tests) participated in a 3-layer intervention. In the first layer, kept constant for all treatment groups, the alphabet principle was taught, making functional spelling units and alternations explicit. In the second layer, which varied systematically across treatment groups, children received different kinds of tutor modeling in learning a set of words of varying spelling-sound predictability, using different connections between printed and spoken words, singly or in combination. In the third layer, also kept constant, children read and discussed illustrated books. Over the 4-month, 24-lesson intervention, all 7 treatment groups in the second layer improved more in word-specific learning than a contact control group that received phonological and orthographic awareness training without explicit instruction on orthographic-phonological connections. Of these 7, only 3 kinds of explicit modeling (whole word, letter-phoneme, and combined whole word and letter-phoneme) resulted in greater transfer to untrained words than the contact control or the other 4 kinds of explicit modeling. Results are discussed in reference to the controversy over whether dual route or connectionist models best account for the acquisition of reading.