If people differ in an absolutely general way in their estimations of epistemic relevance, and if we follow the holism of meaning and individuate int… - Jerry Fodor

From the point of view of semantics, errors must be accidents: if in the extension of "horse" there are no cows, then it cannot be required for the meaning of "horse" that cows be called horses. On the other hand, if "horse" did not mean that which it means, and if it were an error for horses, it would never be possible for a cow to be called "horse." Putting the two things together, it can be seen that the possibility of falsely saying "this is a horse" presupposes the existence of a semantic basis for saying it truly, but not vice versa. If we put this in terms of the crude causal theory, the fact that cows cause one to say "horse" depends on the fact that horses cause one to say "horse"; but the fact that horses cause one to say "horse" does not depend on the fact that cows cause one to say "horse"...

Suppose that the organism is given the problem of determining the analysis of a stimulus at a certain level of representation- e.g., the problem of determining which sequence of words a given utterance encodes. Since, in the general case, transducer outputs underdetermine perceptual analyses, we can think of the solution of such problems as involving processes of non-demonstrative inference. In particular, we can think of each input system as a computational mechanism which projects and confirms a certain class of hypotheses on the basis of a certain body of data.

[T]he degree of confirmation assigned to any given hypothesis is sensitive to properties of the entire belief system … simplicity, plausibility, and conservatism are properties that theories have in virtue of their relation to the whole structure of scientific beliefs taken collectively. A measure of conservatism or simplicity would be a metric over global properties of belief systems.