[TS-1.1] Generalisation of untrained irregular words in spelling

Kohnen, S. , Nickels, L. , Barisic, K. , Moore, K. , Kezilas, Y. , Krajenbrink, T. & Colenbrander, D.

ARC Centre of Excellence in Cognition and its Disorders (CCD); Macquarie University, Sydney, Australia

Previous spelling intervention studies have found improvements for untrained irregular words (i.e., generalisation). However it is unclear which words are most likely to improve and which mechanisms drive this generalisation. We conducted a case-series intervention study with children who have developmental spelling difficulties (surface dysgraphia) to examine whether: (1) generalisation occurs for orthographic neighbours of trained words; (2) generalisation is facilitated by large neighbourhood size; or by (3) overall high frequency; and whether (4) generalisation is more likely to occur for words whose spelling, while imperfect, is somewhat known prior to training. The training (direct and delayed copying) was completed by 11 children and led to significant improvements in spelling for all participants. Most children (N=7) showed generalisation for untrained words. On average, the group of children improved significantly in spelling untrained neighbours (e.g., rough) of trained words (e.g., dough), while untrained non-neighbours matched for frequency and length did not improve. This pattern was also shown by three individual children. Group analysis revealed significant improvements for untrained words high in both neighbourhood size and frequency, but words that were high in either frequency or neighbourhood size did not yield significant improvements. This pattern was not significant for any individual. For all children, the closer that the pre-training misspelling was to the correct spelling (e.g., rouf vs rufe for rough) the greater the likelihood of generalisation for that word. This case-series provides further evidence that generalisation of improvement from treated to untreated irregular words in spelling is a relatively common, albeit small, effect across participants. The results show that neighbourhood plays an important role in generalisation supporting Sage and Ellis? theory that generalisation occurs due to co-activation of orthographically similar words and that this only occurs for words that already have a (rather than no) representation in the orthographic lexicon.