Evidence of expert's evidence is evidence

John Hardwig has championed the thesis (NE) that evidence that an expert EXP has evidence for a proposition P, constituted by EXP's testimony that P, is not evidence for P itself, where evidence for P is generally characterized as anything that counts towards establishing the truth of P. In this paper, I first show that (NE) yields tensions within Hardwig's overall view of epistemic reliance on experts and makes it imply unpalatable consequences. Then, I use Shogenji-Roche's theorem of transitivity of incremental confirmation to show that (NE) is false if a natural Bayesian formalization of the above notion of evidence is implemented. I concede that Hardwig could resist my Bayesian objection if he interpreted (NE) as a more precise thesis that only applies to community-focused evidence. I argue, however, that this precisification, while diminishing the philosophical relevance of (NE), wouldn't settle tensions internal to Hardwig's views.