Within linguistics, words with a complex internal structure are commonly assumed to be decomposed into their constituent morphemes (e.g., un-help-ful). Nevertheless, an ongoing debate concerns the brain structures that subserve this process. Using functional magnetic resonance imaging, the present study varied the internal complexity of derived words while keeping the external surface structure constant as well as controlling relevant parameters that could affect word recognition. This allowed us to tease apart brain activations specifically related to morphological processing from those related to possible confounds of perceptual cues like word length or affix type. Increased task-related activity in left inferior frontal, bilateral temporo-occipital and right parietal areas was specifically related to the processing of derivations with high complex internal structure relative to those with low complex internal structure. Our results show, that morphologically complex words are decomposed and that the brain processes the degree of internal complexity of word derivations. (Contains 2 figures and 2 tables.)