A quantifier tells you how many values a claim is about. The universal quantifier, written \(\forall\), means for all, for every, or for each. The existential quantifier, written \(\exists\), means there exists a or there is a. A sentence that begins with \(\forall x \in X, P(x)\) is true only when \(P(x)\) holds for every element of \(X\). A sentence that begins with \(\exists x \in X, P(x)\) is true when \(P(x)\) holds for at least one element of \(X\). To negate such a sentence you flip the kind of quantifier: not "for all" becomes "there exists" a value where the property fails, and not "there exists" becomes "for all" values the property fails.

Without the fog

Ordinary "and" and "or" are not enough once a set may be infinite. Richard Hammack's Book of Proof introduces \(\forall\) and \(\exists\) for exactly that reason. "Every element of \(X\) is odd" would otherwise become an endless chain of "and". "There is at least one odd element of \(X\)" would become an endless chain of "or". The two new symbols package those ideas.

MIT's Mathematics for Computer Science notes say the same thing in the language of predicates. A predicate is a claim whose truth depends on a variable. You then make one of two common assertions: the predicate is always true, or the predicate is sometimes true. Those are the universally quantified statement and the existentially quantified statement. In symbols, \(\forall x \in D, P(x)\) is read "for all \(x\) in \(D\), \(P(x)\) is true." \(\exists x \in D, P(x)\) is read "there exists an \(x\) in \(D\) such that \(P(x)\) is true."

English can hide which one you meant. The MIT notes point out that "if you can solve any problem we come up with, then you get an A" might mean every problem, or at least one problem. In a proof you do not get to leave that choice hanging. Write the quantifier you intend.

Worked example: why order matters

When a statement has two quantifiers, their order is part of the meaning. Hammack's example is the cleanest small case.

Claim A: \[ \forall x \in \mathbb{R},\; \exists y \in \mathbb{R},\; y^3 = x. \] In words: for every real number \(x\), there is a real number \(y\) whose cube is \(x\). That is true, because every real number has a real cube root. The choice of \(y\) is allowed to depend on \(x\).

Now swap the quantifiers to get claim B: \[ \exists y \in \mathbb{R},\; \forall x \in \mathbb{R},\; y^3 = x. \] In words: there is one fixed real number \(y\) such that \(y^3 = x\) for every real number \(x\). No single \(y\) can have that property, so claim B is false. Hammack stresses that the two sentences have entirely different meanings. The MIT notes give the same warning: swapping a universal quantifier with an existential quantifier usually changes the proposition. Their Goldbach example makes the same point. "For every even integer greater than 2 there exist primes that add to it" is not "there exist two primes that add to every even integer greater than 2."

How to negate a quantified statement

Hammack records the two rules you will use constantly: \[ \sim(\forall x \in X,\, P(x)) \;=\; \exists x \in X,\, \sim P(x), \] \[ \sim(\exists x \in X,\, P(x)) \;=\; \forall x \in X,\, \sim P(x). \] The MIT notes say this as a general principle: moving a "not" across a quantifier changes the kind of quantifier. "It is not the case that everyone likes to snowboard" means "there exists someone who does not like to snowboard." "There does not exist anyone who likes skiing over magma" means "everyone dislikes skiing over magma."

If several quantifiers are stacked, you apply the rule from the outside in. Hammack starts with the true cube root claim and computes \[ \sim(\forall x \in \mathbb{R},\; \exists y \in \mathbb{R},\; y^3 = x) = \exists x \in \mathbb{R},\; \forall y \in \mathbb{R},\; y^3 \neq x. \] In words: there is a real number \(x\) such that no real \(y\) cubes to it. That negated sentence is false, as it should be, because the original claim was true. The pattern is mechanical. Flip each quantifier, then negate the innermost statement.

A useful everyday warning from Hammack: do not say "all integers are not even" when you mean "not all integers are even." The first sentence says there are no even integers. The second says there is at least one odd integer, which is the correct negation of "every integer is even."

A fill in template

  1. Write the claim with the quantifiers in reading order. Start with "for every" or "there exists," name the set each variable ranges over, then state the property.
  2. Decide how you would test it. A universal claim dies with one counterexample. An existential claim lives if you can exhibit one example, or otherwise prove that at least one value works.
  3. If the claim is nested, keep the dependence honest. In \(\forall x, \exists y\), the value of \(y\) may depend on \(x\). In \(\exists y, \forall x\), the same \(y\) must work for every \(x\).
  4. To negate, walk from the left. Change every "for all" into "there exists" and every "there exists" into "for all." Then negate the innermost property. Read the new sentence aloud and check that it is exactly the refusal of the original claim.
  5. Translate back into English before you prove anything. If you cannot say the negated sentence in words, you are not ready to use it inside a larger argument.

The link to for every epsilon there exists a delta

Once the two quantifiers and their order are clear, the opening line of a limit definition is no longer a spell. Hammack even uses that sentence as a translation exercise: for every positive number \(\varepsilon\) there is a positive number \(\delta\) with a stated implication. The order is the meaning. You are given the tolerance first. Then you are allowed to choose a window, and that window may depend on the tolerance.

The swapped sentence, "there exists a \(\delta\) such that for every \(\varepsilon\) the implication holds," would demand one window that works for every tolerance at once. That is a different, much stronger claim, and it is not the definition of a limit. The epsilon delta intuition page is where you write the actual limit argument, with linear and other examples. This page only unpacks the two quantifiers that page is negotiating. The intro to proofs pathway and the real analysis and proofs topic map both treat that order as a habit you name before any algebra.

Checkpoints

  1. Translate "every integer is even" into symbols, then write its negation in symbols and in English. Give one counterexample that shows the original claim is false.
  2. In your own words, why is \(\forall x \in \mathbb{R},\; \exists y \in \mathbb{R},\; y^3 = x\) true while \(\exists y \in \mathbb{R},\; \forall x \in \mathbb{R},\; y^3 = x\) is false?
  3. Negate \(\forall x \in \mathbb{R},\; \exists y \in \mathbb{R},\; y^3 = x\) by flipping each quantifier. Read the result in English.
  4. Why is "for every \(\varepsilon > 0\) there exists a \(\delta > 0\)" not the same as "there exists a \(\delta > 0\) that works for every \(\varepsilon > 0\)"?

Common mistakes

  • Negating "for all" as "for all, not." The negation of "every integer is even" is not "every integer is not even." It is "there exists an integer that is not even."
  • Swapping the order of mixed quantifiers. \(\forall x, \exists y\) lets \(y\) depend on \(x\). \(\exists y, \forall x\) demands one \(y\) that works for every \(x\). Those are different claims.
  • Flipping the quantifiers and forgetting to negate the inner property. After you change \(\forall\) to \(\exists\), you still have to replace \(P(x)\) with its negation.
  • Saying "all ... are not" when you mean "not all." Hammack flags this as a sentence that may pass in casual talk and must never pass in a proof.

Related reading

Related tools

The epsilon delta finder searches for a window at one tolerance you type, and the numerical limit estimator samples nearby values. Both are scratch pictures. Neither proves a quantified claim, because a proof has to handle every allowed value, not only the ones a table happened to show.

Sources

Last reviewed: 8 November 2026.

Common questions

What do the two main quantifiers mean?

The universal quantifier means for all, for every, or for each. The existential quantifier means there exists or there is at least one. A universally quantified claim is true only if the property holds for every allowed value. An existentially quantified claim is true if the property holds for at least one allowed value.

Why does the order of quantifiers matter?

When you mix a for all with a there exists, swapping their order usually changes the meaning. For every real \(x\) there is a real \(y\) with \(y\) cubed equal to \(x\) says every real number has a cube root. There is a real \(y\) such that for every real \(x\), \(y\) cubed equals \(x\) says one fixed number cubes to every real number, which is false.

How do you negate a quantified statement?

Move the word not across each quantifier and change its kind. The negation of for all \(x\), \(P(x)\) is there exists an \(x\) such that \(P(x)\) fails. The negation of there exists an \(x\) such that \(P(x)\) is for all \(x\), \(P(x)\) fails. If several quantifiers are stacked, flip each one from left to right and then negate the innermost claim.

How does this help with for every epsilon there exists a delta?

That sentence is a universal claim followed by an existential claim. The window may depend on the tolerance you were given. It is not the claim that one window works for every tolerance. The epsilon delta explainer is where you write the limit argument itself.