AbstractIn this paper, we report the case of a patient who produced many errors in productive tasks that are analysable as a function of morphological properties of the target and/or the response. We propose a quantitative and a qualitative analysis of the derivational errors produced in a picture-naming task, in order to determine whether these errors result from a morphological impairment. The analysis is conducted with reference to minimal predictions that can be inferred from current models of morphological decomposition. It is shown that the morphological errors made by the patient are explainable without appealing to a morphological organisation of the lexicon. More precisely, we account for these errors in the context of a general two-stage retrieval mechanism that applies both to affixed and unaffixed words.