Speech processing, with its temporal cadence and multi-scale syntactic invariants (syllables, words), is a paradigmatic example where rhythms have been proposed to play a key role [1], with a speech-modulated hierarchical structure of intercoupled cortical oscillations correlating with successful comprehension. Experiments show that speech recognition remains largely intact when compressed up to a certain temporal factor [2] and the re-spacing of chunks of incomprehensible compressed speech with silences recovers comprehension. We hypothesize that rhythm-based top-down semantic context inference is a key mechanism for such re-spaced recovery of compressed speech. We propose a computational model that incorporates a wide range of brain-rhythm data mechanistically and accounts for time-invariant word recognition. In this model, the hierarchically arranged interacting rhythms actively maintain top-down and bottom-up information flow during the inference process: theta-gamma interactions predict and parse phonemes/syllable sequences, while the delta-rhythm adaptively generates the inferred context. We show that word recognition degrades when the speed of words and syllables is compressed beyond the delta and theta rhythms, respectively. The top-down contextual delta-implemented context allows us to explain why re-spacing compressed speech recovers comprehension. Our model further predicts that delta-implemented word context allows for syllable parsing without a precise locking of the theta-rhythmic activity.(A) Schema of BRyBI (B-C) rhythm decoupling and errors depend on uncertainty (D) Surprise as ERP in BRyBI. (E) Intelligibility for interrupted, (F) temporally segmented, and (H) temporally segmented compressed speech. (G) different speech rates;[1] Giraud AL, Poeppel D. Nat Neurosci (2012). [2] Ghitza O, Greenberg S. Phonetica (2009).