The most promising concrete suggestion I see here is to adopt verbal conventions for distinguishing direct and indirect evidence. I'm not sure the word "impression" really connotes direct evidence, though with enough consistent usage in that mode we might carve out a common meaning to that effect. But we actually have a whole range of indirection; where would the cutoff in that range be? If I actually looked something up recently in an encyclopedia, while someone else just vaguely remembers looking it up sometime long ago, is that my impression or my belief?