In the last years, it has become a matter of intense debate whether regular and irregular aspects of language are processed within the same cognitive module as postulated by a number of connectionist models, or whether separate modules have to be assumed, with a word frequency-insensitive symbolic processing module for regular morphology (cook, cook-ed), and word frequency-sensitive lexical storage for irregular morphology (drink, drank). While single process models predict overlap of the neuronal activation associated with regular and irregular morphological processing, a high degree of non-overlapping activations would favor dual process models. During four PET scans, we presented subjects with infinitive verb forms, which had to be inserted with appropriate inflection in a neutral sentence frame, hi a crossed design, verbs were varied between scans along the dimensions 'regular vs. irregular' and Tugh vs. low spoken frequency of past tense/participle forms (HF vs. LF)'. Sentence frames requiring past tense (He ... something.) or participle formation (He has ... something.) were randomized within scans to avoid response strategies. In two further scans (baseline conditions), HF and LF verbs were presented in meir inflected form, so that they could be inserted in die sentence frame without morphological production. Subjects and Methods