Perceptions are true and not describable by propositions, while beliefs are testable propositions and therefore not a criterion of truth.
Now I do think that is a very clear and useful statement.
An example he uses, with reservations, is that of a photograph
And that's an interesting example and gives something concrete to talk about. The photograph becomes something to which we can compare future observations ,but in and of itself the photograph is a limited representation and doesn't really contain any opinion about what is on the film. Sounds like a good article.