Studies of silent sentence reading have shown indirect evidence for the impact of a default projected "implicit prosody" on sentence processing, such as longer processing times for sentences with final syntactic interpretations inconsistent with their assumed default prosody (e.g. Fernández, 2003). While explanations of such effects associate them with the presence of an auditory image, or "inner speech", there is no direct evidence tying such an image directly to the measured effects--that is, there is little information about how implicit prosody "sounds" in the mind of the reader. In two visual-to-auditory cross-modal priming experiments, we look for evidence of a link between the implicit prosodic contour generated during silent reading and the explicit prosodic contour pronounced on a subsequentlypresented auditory probe word. Pairs of text sentences that contained corrective contrasts (e.g. Jacquelyn didn't pass the test. Belinda passed the test.) were immediately followed by probes pronounced with pitch accent patterns consistent (BELINDA) or inconsistent (BELINDA) with the corrective contrast in the read text. Results in an initial experiment showed no priming from implicit to overt prosody, but instead a processing advantage for contrasts in verb position over those in subject position, as well as an advantage for no-accent probes over contrastively accented ones. The experiment was repeated with participants separated into two groups on the basis of the overt prosody they produced when reading a passage aloud. Results showed that individual differences in the read-aloud production correlate with the response pattern in the cross-modal task. The results provide some initial evidence about the nature of the auditory image produced as inner speech during silent reading.