Self-Reference: What Is This Topic About?

Core idea

A name can refer to something it is itself a part of. "This sentence contains five words" talks about the very sentence it occurs in — harmless self-reference. But push it further and trouble erupts. Consider the sentence:

Priest's argument: "This very sentence is false" — if it is true, then what it says holds, so it is false; if it is false, then it says so, so it is true.

This is the liar paradox, traced to the ancient Greek Eubulides. The sentence appears to be both true and false. Its cousin, "this very sentence is true", appears to be neither: nothing settles its value either way. Self-reference, then, seems to break the comfortable assumption from earlier topics that every sentence is true or false, but never both and never neither.

Why it matters

Not just a parlour trick

The same structure surfaces in serious mathematics. A set is a collection of objects, and sets can be members of other sets — even of themselves. Russell's paradox asks about R, the set of all sets that are not members of themselves. If R is a member of itself, then it qualifies as one of the non-self-members, so it is not. If it is not, then it qualifies, so it is. Like the liar, R both is and is not a member of itself. These paradoxes sit at the foundations of set theory; they cannot be waved away.

Expanding the truth values

Priest's response is to take the extra possibilities seriously. Assume that in any situation a sentence may be true only, false only, both, or neither. The truth conditions for negation, conjunction, and disjunction stay exactly as before — ¬a is true when a is false, and so on — but they are now applied to a wider range of cases. A sentence that is both true and false has a negation that is also both; the machinery still computes, it just has more inputs to handle.

Validity survives, but verdicts change

Validity keeps its definition: no situation makes the premises true and the conclusion not true. But situations are richer now, and that changes outcomes. The troubling chain from Truth Functions — or Not? — q, ¬q, therefore p — now comes out invalid. Take a situation where q is both true and false but p is only false: both premises count as true, yet the conclusion is not true. This is a second, deeper diagnosis of why the inference always felt wrong. One of its two sub-inferences must therefore also be invalid — the step that drops a disjunct when its partner is denied — because the truth of ¬q no longer rules out the truth of q.

The crack: contradiction returns

Even this generous account is not safe. Take the cousin, "this sentence is true", supposed to be neither true nor false. If it is neither, then in particular it is not true — but it says it is true, so it is false, contradicting "neither". Or tweak the liar to "this sentence is not true": if true, then not true; if not true, then true. We no longer have a sentence that is merely both T and F — we have a sentence that both is and is not true. Self-reference has been a tangled, contentious issue ever since Eubulides, and it remains one.

Key takeaways

Mental model

Copy sourceZoom

Practical application

When a statement, rule, or specification talks about itself, check whether the self-reference is benign or vicious. "This document supersedes all earlier versions" is harmless. "This rule does not apply to itself" — or a rule asserting its own invalidity — can generate a liar-style loop where no consistent reading exists. Spotting the loop early prevents systems, contracts, and definitions from quietly contradicting themselves.

Example

A software config flag reads: "Ignore this configuration line." If the system obeys it, it ignores the line — including the instruction to ignore — so it should not ignore it; but then it reads the line and ignores it again. The flag has no stable behaviour. It is the liar paradox in a config file: a self-referential instruction whose every interpretation undermines itself. The fix is not cleverer parsing but forbidding the self-reference outright.

Related lessons