Learning nothing from everything

I analyze how a Sender can use verifiable evidence to influence a Receiver about a binary state when the relevance of information is ex ante uncertain and asymmetrically known by the Sender. The Sender has access to two pieces of binary evidence, one they know to be perfectly informative of the state and one that is completely uninformative. Full disclosure of evidence is always possible in equilibrium, but the Receiver generically cannot fully unravel which piece of evidence is relevant. Consequently, the Receiver may learn very little – or even nothing – about the state even when all evidence is disclosed.