A visual lexical decision task was used to examine the interaction between base frequency (i.e., the cumulative frequencies of morphologically related forms) and semantic transparency for a list of derived words. Linear mixed effects models revealed that high base frequency facilitates the recognition of the complex word (i.e., a "base frequency effect"), but the magnitude of this declines with semantic transparency. These results suggest that the extent to which the constituents of a complex word contribute to its recognition is dictated by semantic transparency. The implications of these findings are discussed in the context of localist-connectionist models of morphological processing.