Abstract
We tested the predictions of a dual-route model of complex word reading according to which morpho-orthographic segmentation is hypothesised to require a fine-grained orthographic code that would be particularly sensitive to letter order, whereas morpho-semantic representations are hypothesised to be most rapidly accessed via a coarse-grained orthographic code that is less sensitive to letter order. We predicted that letter transpositions would disrupt morpho-orthographic processing more than morpho-semantic processing. In line with these predictions, Experiment 1 showed no priming from opaque pseudo-derived primes containing a letter transposition at the morpheme boundary (masetr-mast) relative to replaced letter controls (masicr-mast) in the presence of significant priming from transposed-letter transparent derived primes (banekr-bank). Similarly, Experiment 2 showed that although complex nonword primes (bankity-bank) generate significant priming effects relative to unrelated primes (farmity-bank), the same primes with letter transpositions (banikty-bank) do not prime relative to transposed unrelated primes (farimty-bank).
Acknowledgements
Jonathan Grainger and Daisy Bertrand were supported by ERC grant 230313.
Notes
1The model was estimated using treatment coding for the factors. Hence, the individual interaction terms provide tests of the comparisons between individual levels in the 3×3 design. In order to evaluate all comparisons, we refitted the model six times, each time choosing a different reference level in the treatment coding of Item Type and Prime Type. p-values were corrected for multiple testing following Benjamini and Hochberg (Citation1995).