Pinker’s influential presentation of the distinction between the combinatoric units of language (the “words”) and the mechanisms that organize the units into linguistic constituents (the “rules”) rested on a strong, but ultimately incorrect, theory about the connection between a speaker’s internalized grammar and his/her use of language: that what is linguistically complex, and thus constructed by the grammar, is not memorized; thus experience with complex constituents (as measured in corpus frequency, for example) would have no effect on processing such complex constituents. I argue that recent results within linguistics and within psycho- and neuro-linguistics show instead that memory and frequency effects are irrelevant to the linguistic analysis of language but always influence processing, across simple and complex constituents. Phrases and words can be shown always to decompose down to the level of morphemes both in representations and in processing, and, contrary to Pinker’s claim, the “memorized” status of a complex structure holds no import for its linguistic analysis. On the other hand, speakers’ experience with language is always reflected in their use of language, so frequency effects are always relevant to processing, even for completely regular combinations of words and morphemes. I will present neurolinguistic evidence for full decomposition of irregular forms (such as English irregular verbs), as well as evidence for frequency effects for regular combinations of morphemes and words.