Foundationalism

Ask why you believe something and you get another belief. Ask why that one, and the question repeats. There are only three ways it can end: the chain runs forever, it loops back on itself, or it stops. Foundationalism says it stops — there are basic beliefs that are justified without being justified by other beliefs, and every other justified belief traces back to them.^[1] The looping option it rejects outright as circular reasoning, and the infinite chain as something no finite mind actually has — though infinitism bites that bullet.

The disagreements are about what gets to be basic. Classical foundationalism wanted the foundation to be indubitable, which is why Descartes ended up with I think, therefore I am and very little else — a base that secure is too small to rebuild anything on. Modest foundationalism, the version most people hold now, asks only that basic beliefs be justified non-inferentially and lets them be defeasible: your experience of a red wall justifies believing the wall is red, defeasibly, until you learn about the stage lighting. That is enough to stop the regress without pretending to certainty — the concession skepticism was always going to extract.

The serious objection is Sellars' myth of the given. Whatever sits at the bottom faces a dilemma: if it is a raw sensation with no propositional content, it is not the kind of thing that can be a reason for anything; if it does have propositional content, it is the kind of thing that can be right or wrong, and then we can ask what justifies it. Either the foundation can't do the job, or it isn't a foundation. Coherentism takes that argument as its starting point.

Deep dive · The regress trilemma

Strip the argument to its bones. Let $J(p)$ mean "the belief that $p$ is justified," and suppose justification is inferential: $J(p)$ holds only because $J(q)$ holds for some $q$ that supports $p$ . Iterate. The chain $p_1 \leftarrow p_2 \leftarrow p_3 \leftarrow \cdots$ admits exactly three shapes, the Agrippan or Münchhausen trilemma:

Shape	Position	Cost
Infinite, non-repeating	Infinitism	No finite believer completes it
Returns to an earlier link	Coherentism	Looks question-begging
Terminates	Foundationalism	Owes an account of the terminus

Foundationalism takes the third horn by denying the supposition: some $p$ has $J(p)$ non-inferentially, so the chain has somewhere to stop. Everything then turns on the support relation running upward from the base. Classical versions wanted it to be deductive, so certainty at the bottom would transmit undiminished to the top. It doesn't work — almost nothing about the external world follows deductively from facts about your own experience. Modest versions let the relation be inductive instead, which reaches ordinary beliefs about tables and galaxies at the price of transmitting something weaker than certainty. That is the trade the field has largely accepted: a foundation you can actually build on, holding up a structure that could still be wrong.