The present cross-sectional study investigated the development of phonological recoding in beginning readers of Dutch, using a proofreading task with pseudohomophones and control misspellings. In Experiment 1, children in grades 1 to 3 rejected fewer pseudohomophones (e.g., wein, sounding like wijn 'wine') as spelling errors than control misspellings (e.g., wijg). The size of this pseudohomophone effect was larger in grade 1 than in grade 2 and did not differ between grades 2 and 3. In Experiment 2, we replicated the pseudohomophone effect in beginning readers and we tested how orthographic knowledge may modulate this effect. Children in grades 2 to 4 again detected fewer pseudohomophones than control misspellings and this effect decreased between grades 2 and 3 and between grades 3 and 4. The magnitude of the pseudohomophone effect was modulated by the development of orthographic knowledge: its magnitude decreased much more between grades 2 and 3 for more advanced spellers, than for less advanced spellers. The persistence of the pseudohomophone effect across all grades illustrates the importance of phonological recoding in Dutch readers. At the same time, the decreasing pseudohomophone effect across grades indicates the increasing influence of orthographic knowledge as reading develops.