The current study investigated the modulation by orthographic knowledge of the final overlap phonological priming effect, contrasting spoken prime-target pairs with congruent spellings (e.g., "carreau-bourreau", /karo/-/buro/) to pairs with incongruent spellings (e.g., "zero-bourreau", /zero/-/buro/). Using materials and designs aimed at reducing the impact of response biases or strategies, no orthographic congruency effect was found in shadowing, a speech recognition task that can be performed prelexically. In lexical decision, an orthographic effect occurred only when the processing environment reduced the prominence of phonological overlap and thus induced participants to rely on word spelling. Overall, the data do not support the assumption of early, automatic activation of orthographic representations during spoken word recognition. (Contains 3 tables and 4 footnotes.)