Overview- Sellars’ project is to attack the framework of givenness – the seductive idea that "there are epistemic primitives – beliefs or other mental states that have some positive epistemic status but that are non-inferential, conceptually simple, and epistemically independent and efficacious." (D&T, p.7; c.f. Brandom, p. 122)

Classical sense-data theories are a prime example of this framework of givenness.

1. The argument from illusion (Ayer):

Sense-data theorists were interested in what it is that one sees when one’s perceptions do not line up with reality. Their conclusion: we directly experience "sense data."

1. When I hallucinate, say, a pink elephant, the pinkness that I see cannot be the pinkness of any physical object
2. But the pinkness that I see when I hallucinate a pink elephant is indistinguishable from the pinkness that I see when I really do see a pink elephant.
3. So, we never directly experience the physical properties of objects in the material world. Rather, we actually perceive non-physical, subjective "sense data," from which we must infer the existence and nature of the external world.

Phenomenalism (a kind of idealism): What we normally think of as material objects really are complexes built up out of sense data.

Linguistic phenomenalism: The meanings of our ordinary material-object talk are to be analyzed in terms of sense data.

2. The postulation of sense data (or sense impressions) seems to fit well with a foundationalist account of empirical knowledge (or justification).

Regress argument for foundationalism (Brandom, p. 124): To count as knowledge, a belief must be justified. Beliefs can be used to justify other beliefs. But one cannot have a circle of justification (some proposition ultimately being used in its own justification). Nor can one have a non-terminating chain of justification (or a regress). So for there to be any knowledge at all, there must be basic beliefs, which are not justified by other beliefs, and which can be used to justify others. Such "foundational" beliefs are justified all by themselves; they are "self-authenticating."

3. What then could serve as such foundational beliefs?

A hum-drum observation: Consider the justification of any ordinary belief about how things are in the external world. Typically such justifications simply terminate in claims about how things look or appear to us.

So it is tempting to suppose that statements involving that which you see most directly – sense data or impressions – could serve as this foundation of knowledge. (Voila- empiricism!)

Some Cartesian observations:

1. Statements involving sense-data or impressions seem more secure (or certain) than statements involving the external world. <remember the cogito? One knows oneself and the contents of one’s mind better than anything else.>

-How can one be questioned about how things look or appear to you?

2. This picture lends itself well to dualism. That which comprises the foundation of knowledge (the mental) is of a wholly different ontological kind than that which doesn’t. Indeed, the mental can be characterized as that which is known more directly and with greater certainty.

4. Sellars’ response in Part I: (The ambiguity in sense-datum theories)

What it is that we see when we hallucinate (or have any experience) is just not of the correct sort to serve the justificatory role played by something in the foundation of knowledge.

1. That which serves in the foundation of knowledge must be truth-evaluable (or believable) and thus conceptually articulated: a seeing that something is thus-and-so. <It must be "epistemic">
2. Sense-data (or impressions) are typically thought of as particulars, and thus not truth-evaluable. As such, they are not conceptually or propositionally articulated, or appropriately epistemic (many ways of saying the same thing).
3. Thus sense data cannot justify any beliefs, and so couldn’t belong to any putative foundation of knowledge.

The overall point is that (1) S senses x. [S has a red sense-datum or impression.]

Does not imply (2) S senses that x is f. Having sensory impressions of, say, pinkness, is different from knowing that something is pink.

Note: that does not mean that sense-impressions could not figure in a causal explanation of perception (see the Brandom diagram, p. 126). (1) above might be causally necessary for (2), just not sufficient.

6. A further way to see this distinction, which should embarrass sense datum theories (section 6): Classical empiricism appears committed to inconsistent commitments:

1. The ability to apply concepts in experience (the ability to see that there is a pink elephant) and so come up with epistemically efficacious mental states requires learning and experience.
2. But the ability to have sense impressions (to see a pink elephant) does not require learning and experience.

5. Sellars’ diagnosis of this strong temptation to classify sense-data or impressions as "mental states" like thoughts or beliefs:

But Sellars warns us that assimilating sensations to thoughts is a trap. In due course, he will show us how to properly conceive of the notions of thoughts and of sense impressions, and will even show how we can explain their seeming subjectivity in a way that disarms Rylean suspicions.