A possible context
Think of the Agrippan trilemma. Suppose – with the traditional definition – that knowledge has to be justified. For any claim, there seem to be three ways to justify it:
1. One can continue to think up fresh justifications for each preceding justification.
2. One can make a dogmatic assumption.
3. One can repeat something already said
The first leads to an impossible – because vicious – infinite regress. No claim would be justified until every one of the claims justifying it was also justified. So no claim would ever be justified.
The second and third seem to be versions of foundationalism and coherence.
Coherence: The worry about the third horn is this: if our justifications of claims only ever circle back through other claims, how does that whole system ever connect to, and thus say something about, the empirical world? Why isn’t it, in McDowell’s phrase, merely a ‘frictionless spinning in the void’?
So if not coherence, can we block the regress of justifications by finding a suitable place to stop it? If we are thinking about empirical claims (by contrast with logical or mathematical claims) then we will want an observation claim or a perceptual report. So let’s try a form of foundationalism based on that.
Foundationalism: But Davidson comments that only a belief can serve as a reason for a belief. If justification is a matter of giving reasons – a move in the space of reasons – then, goes the thought, the only thing with the right rational properties is a belief. (My cat Sootica, goes the thought, is a very nice cat but she is not, herself, a reason for anything.) But if this Davidsonian thought is correct, then it is hard to know how we can terminate justifications in any appeal to how the world is.
Take the final claim that one makes or belief one holds ‘closest’ to the impact of the world: a perceptual report or perceptual belief. How is it justified? Not, ex hypothesi, by another claim / belief. But if it is a bit of the world, a fact, say, how does it justify the belief / claim?
Davidson himself attempts to escape the ‘frictionless spinning in the void’ of coherence whilst still holding that only a belief can serve as a reason for a belief by arguing that the world can have a causal connection to belief via experiences. Experiences – a bit like sensations – are caused by the world and cause beliefs.
But the problem with that is that that isn’t a rational connection. How can the fact that one thing causes another justify it? Most causal relations do not justify things. They are not in that business.
Sellars’ argument
Sellars attempts to find a middle position between coherence and foundationalism.
If something is to count as a perceptual report it must meet 2 conditions:
1) It must be reliable.
2) The subject who makes it must know it is reliable. Otherwise, trained parrots would count as making perceptual reports. But parrots are not justified in their ‘utterances’.
But answering 2) means that a subject must think of her reports as reliable. So she must have a conception of how the world works and how her reports work, under suitable conditions. But if that is the case, this is not a form of the Myth of Given because perceptual reports are now fallible. If her world view is wrong, that will infect her perceptual reports.
So neither mere coherence (because perceptual beliefs are not inferred from anything else) but not foundationalism (since perceptual reports are fallible and can be undermined by failures of the world-view).