Investigation, K.M. and K.H.; 10 Resources, K.M., K.M.T., T.A.V., and C.J.H.; Writing -Original Draft, K.M., K.H., K.M.T., T.A.V., 11 and C.J.H.; Visualization, K.M., K.M.T., and C.J.H.; Supervision, T.A.V. and C.J.H.; Funding 12 Acquisition, K.M., K.M.T., T.A.V. and C.J.H. Summary 1After we listen to a series of words, we can silently replay them in our mind. Does this mental 2 replay involve a re-activation of our original perceptual representations? We recorded 3 electrocorticographic (ECoG) activity across the lateral cerebral cortex as people heard and then 4 mentally rehearsed spoken sentences. For each region, we tested whether silent rehearsal of 5 sentences involved reactivation of sentence-specific representations established during 6 perception or transformation to a distinct representation. In sensorimotor and premotor 7 cortex, we observed reliable and temporally precise responses to speech; these patterns 8 transformed to distinct sentence-specific representations during mental rehearsal. In contrast, 9we observed slower and less reliable responses in prefrontal and temporoparietal cortex; these 10 higher-order representations, which were sensitive to sentence semantics, were shared across 11 perception and rehearsal. The mental rehearsal of natural speech involves the transformation 12 of time-resolved speech representations in sensorimotor and premotor cortex, combined with 13 diffuse reactivation of higher-order semantic representations. 14 Keywords: ECoG, sentence repetition, verbal short-term memory, subvocal rehearsal 15 premotor cortex (dPMC) of the left hemisphere. Furthermore, increased activation in these 1 areas during silent rehearsal predicted more accurate behavioral recall of the sentence content. 2 Consistent with prior literature (e.g., Cheung et al., 2016;Glanz et al., 2018) the SMC and dPMC 3 responded rapidly during sentence perception, encoding sub-second properties of the input. 4The fidelity of sensory responses in SMC and dPMC was exceeded only by the superior 5 temporal gyrus (STG) and middle temporal gyrus (MTG). When sentences were silently 6 rehearsed, SMC and dPMC again exhibited sentence-specific activity patterns, but the activity 7 patterns were distinct from those observed during perception of the same sentences. 8Altogether, the data support a model in which "motor" circuitry (SMC and PMC) supports 9 verbal short-term memory via a sensorimotor transformation (Cogan et al., 2014). 10We also observed sentence-specific activity in anterior prefrontal cortex (aPFC) and 11 temporoparietal cortex (TPJ). Sentence-specific activity in these areas was less temporally 12 precise and less reliable than in sensory or motor areas. However, patterns in prefrontal areas 13 were sensitive to the contextual meaning of the sentence being rehearsed. Moreover, the 14 representations in these high-level areas were not transformed, but were instead shared across 15 the perception and rehearsal of specific sentences. Activation in these higher order areas is 16 therefore consistent with a...