During a visual rhyming task, deaf participants traditionally perform more poorly than hearing participants in making rhyme judgements for written words in which the rhyme and the spelling pattern are incongruent (e.g. "hair/bear"). It has been suggested that deaf participants' low accuracy results from their tendency to rely on orthographic similarity. To test this interpretation more directly, we compared profoundly and prelingually deaf, orally educated participants and hearing participants' accuracy during a visual rhyming judgement task in which the two words of a pair share the orthographic rime, in order to discourage usage of a purely orthographic strategy. Accuracy was lower in deaf than in hearing participants. The gradient of difficulty between items, together with the finding of a significant correlation between accuracy and the consistency of the grapheme to rhyme, suggest that difference in accuracy between groups might be explained by an over regularization in deaf people, which is probably linked to less diversified phonological representations.