Language comprehension is fast and seemingly effortless. However, in spite of long knowing what brain regions enable this feat, our knowledge of the precise neural computations that these frontotemporal regions implement remains limited. One highly controversial question is whether there exist functional differences among the neural populations that comprise the language network. Leveraging the high spatiotemporal resolution of intracranial recordings, we clustered the timecourses of responses to sentences and linguistically degraded conditions and discovered three response profiles that robustly differ in their temporal dynamics. Computational modeling of these profiles suggested that they reflect different temporal receptive windows (TRWs), with average TRWs of 1, 4, and 6 words. The electrodes exhibiting these profiles were interleaved across the language network, further suggesting that all language regions have direct access to distinct, multi-scale representations of linguistic input—a property that may be critical for the efficiency and robustness of language processing.