There is mounting evidence that language users are sensitive to distributional information at many grain-sizes. Much of this research has focused on the distributional properties of words, the units they consist of (morphemes, phonemes), and the syntactic structures they appear in (verb-categorization frames, syntactic constructions). In a series of studies we show that comprehenders are also sensitive to the frequencies of compositional four-word phrases (e.g. don't have to worry): more frequent phrases are processed faster. The effect is not reducible to the frequency of the individual words or substrings and is observed across the entire frequency range (for low, mid and high frequency phrases). Comprehenders seem to learn and store frequency information about multi-word phrases. These findings call for processing models that can capture and predict phrase-frequency effects and support accounts where linguistic knowledge consists of patterns of varying sizes and levels of abstraction. (Contains 1 figure and 2 tables.)