This paper focuses on a specific kind of (epistemic) normative conflict, collateral normative conflict -viz., where cognition's working badly at the global level of general dispositions to believe is the price to be paid for its working well locally. I argue that such normative conflicts are much rarer than Williamson (2021) and Lasonen-Aarnio ( 2010) take them to be, even though, and contra proponents of revisionary defeat (e.g., Brown 2018), knowers can, as Williamson and Lasonen-Aarnio rightly maintain, at least sometimes disregard misleading evidence from reliable sources. My rationale for the rarity of collateral epistemic conflicts draws from recent insights by Sosa (2021) on the appropriateness of aiming, in certain domains of inquiry, not just at knowledge, but at knowledge firsthand. A consequence of the rationale offered, however, is that an entirely different kind of normative conflict -what I call cross-modal normative conflictturns out to be much more common than appreciated.