The exact mechanisms that govern the visual identification of complex words (e.g., build-er) are not entirely clear. In particular, it is not obvious why the system would identify (pseudo–)morphemes in words that only have an appearance of morphological complexity (e.g., corn–er, iron–y). One hypothesis is that this phenomenon is driven by letter co–occurrence regularity - morphemes are also recurrent clusters of letters. To test this hypothesis, we assessed masked priming as induced by morphologically structured (e.g., bulb-er) and non-morphologically structured nonwords whose endings were either as frequent (e.g., bulb-le) or less frequent (e.g., bulb-og) than suffixes. We observed equivalent facilitation by the three prime types, suggesting that neither the frequency nor the morphological status of word endings affects masked morphological priming with nonwords. Since this challenges the pattern of results typically observed with word primes, we ran a second experiment where morphological (bulb-er) or non-morphological (bulb-og) nonword primes were compared with transparent derivations (deal-er), pseudo-derived words (corn-er) and orthographic controls (dial-og). The data show a graded pattern of results, with maximal facilitation for transparent primes and no priming for orthographic controls; nonword primes elicited comparable facilitation again, regardless of their morphological status.