The liar paradox, attributed to the Greek philosopher Eubulides of Miletus who lived in the fourth century B.C, is the paradoxical statement

I am lying now.

or

This statement is false.

As opposed to the Epimenides paradox, this statement is indeed paradoxical: assuming that the statement is true, then it must be false; assuming it is false, then it is not false. No truth value can be consistently assigned to the statement.

Even the conclusion that the statement is neither true nor false leads to a contradiction: the statement claims to be false, but isn't, so it claims a falsehood and is therefore false.

To avoid having a sentence refer to its own truth value, one can also construct the paradox

The following sentence is true.

The preceding sentence is false.

The proof of Gödel's incompleteness theorem essentially consists of a formally correct formulation of a variation of this paradox in the context of a sufficiently strong axiomatic system A:

A proof exists in A that this sentence is false.

If a proof exists using only the axioms in A that the statement is true, then this implies that there is also a proof that the statement is false. Conversely, if a proof exists in A that the statement is false, then this proof is an example showing that the statement is true. Thus, if a proof exists either way, the system is inconsistent, in that a single statement can be proven to be both true and false.

On the other hand, if there exists no proof in A of the statement either way, then no contradiction arises. The system A is called incomplete in this case: there exists a statement which can neither be proven nor disproven in A.

Similarly, by using the statement "No proof exists in A that this statement is true", we can see that in a consistent system there are statements that are "clearly" true, which cannot be proven to be so in A.

That A can be shown to be true if it is false and false if it is true has led some to conclude that it is neither true nor false. This has given rise to the following, strengthened version of the paradox:

This statement is not true.

If it is neither true nor false, then it is not true, which is what it says, hence it's true, etc. This has led some, notably Graham Priest, to posit that the statement is both true and false. Joachim Bromond (2002) has confuted this third truth value by means of a re-strengthened liar which says:

This statement is only false.

(Priest disagrees. See Priest, forthcoming)

Then there's Yablo's version of the paradox. Consider a list of sentences which is infinitely long in both directions. The sentences all say the same thing: All of the subsequent statements are false. Pick one statement at random. So it's true if all of the subsequent statements are false. But if all of the subsequent statements are false, then what they say is indeed the case: they say that all of the statements subsequent to them are false, and ex hyposthesi they are false. That contradiction means that the picked statement should be false, but its selection was arbitrary, implying all the statements must be false; again this leads to their description of subsequent statements being true. So like the liar, they're true if they're false and false if they're true, yet no propositions predicate falsity of themselves. This is sufficent to suggest that the liar does not depend upon self reference.

Consider for a moment the opposite of the liar:

This statement is true.

It's true if it's true and false if it's false, but which is it? There seems to be nothing intrinsic to the proposition which determines whether it is true or false. Its truth value seems radically underdetermined. It can be argued that this implicity leads to a contradiction: It can consistently be treated as true and it can be consistently treated as false. If that is so, then one person can mark it down as true, another as false: but if that is the case, then one and the same proposition is both true and false at the same time, a contradiction. Either that, or it is neither true nor false. But it seems to be perfectly meaningful. So either we must conclude that it's meaningless after all or reject the principle of bivalence, and conclude that indeed some meaningful statements can be neither true nor false. Any adequate solution to the liar will have to resolve its twin sister as well. But furthermore, if it is neither true nor false, then it claims to be true and is not, hence, it's false. The view that it is neither true nor false is contradictory.