Antisymmetric Is Not the Opposite of Symmetric

Language is a terrible teacher of mathematics. The word antisymmetric sounds like the opposite of symmetric — "if a relation is symmetric it is not antisymmetric, and vice versa." That folklore is wrong in three different ways, and it produces some of the most resilient JEE mistakes. The correct intuition is almost never taught cleanly: antisymmetric is about distinct elements, and it lives on a totally different axis from symmetric. A relation can be both symmetric and antisymmetric at the same time. It can also be neither.

The two definitions, side by side

Symmetric: for all a, b \in A, if (a, b) \in R then (b, a) \in R.
Antisymmetric: for all a, b \in A, if (a, b) \in R and (b, a) \in R, then a = b.

Symmetric demands every directed edge have its reverse. Antisymmetric allows reverses only between an element and itself — self-loops. The moment two different elements are linked both ways, antisymmetry is broken.

Symmetric forces both arrows between distinct elements. Antisymmetric forbids both arrows between distinct elements — but allows self-loops and allows one-way edges freely.

The three folklore errors

Error 1. "Antisymmetric means never symmetric." False. Antisymmetric says: if both (a, b) and (b, a) are in R, it must be because a = b. It does not forbid one-way edges; it does not forbid any particular asymmetry — it only forbids mutual edges between different elements.

Error 2. "If R is symmetric, it can't be antisymmetric." False. Consider R = \{(1, 1), (2, 2), (3, 3)\} on \{1, 2, 3\}. Every pair (a, b) in R has a = b. Is it symmetric? Yes — for each (a, a) \in R, (a, a) \in R. Is it antisymmetric? Yes — the only (a, b), (b, a) pairs that are both in R are the self-loops, and they satisfy a = b. Symmetric AND antisymmetric simultaneously. In fact, equality is the canonical such relation.

Error 3. "Not symmetric" and "antisymmetric" mean the same thing. False. "Not symmetric" means there exists some (a, b) \in R with (b, a) \notin R — a witnessed asymmetry. "Antisymmetric" is a universal claim about every mutual pair. A relation can be not symmetric and not antisymmetric, or symmetric and antisymmetric. The two labels are on different axes.

Four worlds table

Because the two properties are independent, every relation lands in one of four boxes.

	Antisymmetric	Not antisymmetric
Symmetric	Only self-loops (subset of diagonal)	Full mutual pairs
Not symmetric	Directed edges, no mutual pairs off-diagonal	Mixed: some one-way, some mutual off-diagonal

Example per box on A = \{1, 2\}:

Symmetric and antisymmetric: \{(1,1)\} or \{(1,1), (2,2)\}.
Symmetric not antisymmetric: \{(1,2), (2,1)\} — mutual pair between distinct 1 and 2.
Not symmetric and antisymmetric: \{(1,2)\} — one-way edge, no reverse.
Not symmetric, not antisymmetric: \{(1,2), (2,1), (1,1)\} — mutual pair between distinct elements breaks antisymmetric; symmetric would demand (2,2) which is missing, so not symmetric either.

Why the four-world table is the right mental model: symmetric and antisymmetric are two separate facts about R's behaviour on distinct pairs. Each fact can be true or false independently. Most students collapse the axes into one and treat symmetric and antisymmetric as negations — the table rips that mistake out cleanly.

The useful mental picture — the main diagonal

Draw R on a grid as you would for any subset of A \times A. Cells on the main diagonal are the self-loop cells (a, a). Cells off the diagonal come in mirrored pairs: (a, b) and (b, a) are reflections across the diagonal.

Symmetric: for every shaded cell off the diagonal, its mirror is also shaded.
Antisymmetric: for every shaded cell off the diagonal, its mirror is not shaded.

The diagonal cells themselves are their own mirrors — they do not participate in the off-diagonal rule, which is why they can be freely shaded in either property. A relation is both symmetric and antisymmetric iff it is entirely on the diagonal.

Why the "anti-" prefix misleads

In English, anti- usually negates. In mathematics, the prefix has drifted — "antiderivative" is a near-inverse, "antisymmetric" describes a constraint on how pairs interact but not the logical negation of symmetric. The property is better read as "off-diagonal asymmetry": off the diagonal, mutual pairs are not allowed. The word does not fight symmetric; it imposes its own rule, which happens to coincide with symmetric on the diagonal and disagree with it off the diagonal.

Classic examples in the right boxes

Equality (=) on any set. Symmetric ✓, antisymmetric ✓, reflexive ✓. Only self-loops.
"Less than or equal to" (\leq) on \mathbb{R}. Not symmetric (you don't have 3 \leq 5 and 5 \leq 3), antisymmetric ✓ (if a \leq b and b \leq a, then a = b). The classic partial-order example.
"Has the same age as" on people. Symmetric ✓, antisymmetric ✗ (two distinct people can share an age).
"Is a sibling of" on a family. Symmetric ✓, antisymmetric ✗.
"Is a proper divisor of" on \mathbb{N} (strict, a \neq b). Not symmetric, antisymmetric ✓ — but not reflexive.

The one-shot antisymmetry test

To test if R is antisymmetric, scan for pairs of distinct elements with mutual edges. If you find any pair a \neq b with both (a, b) \in R and (b, a) \in R, antisymmetric fails. If you don't find any such pair (after exhausting the list or arguing abstractly), antisymmetric holds.

Concrete procedure on a finite relation:

List all pairs (a, b) \in R with a \neq b.
For each such pair, check if (b, a) is also in R.
If any such pair has its reverse also in R, antisymmetry fails — report the pair (a, b) as counter-example.
If no such pair does, antisymmetric holds.

Self-loops are irrelevant to antisymmetry. You can skip them entirely.

checking the four-box independence

Problem: On A = \{1, 2, 3\}, classify each relation by (symmetric?) and (antisymmetric?).

R_1 = \{(1,1), (2,2), (3,3)\}. Only self-loops. Symmetric ✓ (the reverse of a self-loop is itself). Antisymmetric ✓ (no two distinct elements are mutually related — in fact, no two distinct elements are related at all). Both.

R_2 = \{(1,2), (2,1)\}. Mutual pair (1, 2), (2, 1) with 1 \neq 2. Symmetric ✓. Antisymmetric ✗ — the pair 1 \neq 2 with both edges is a counter-example.

R_3 = \{(1,2), (2,3), (1,3)\}. All one-way. Symmetric ✗ — (1, 2) \in R but (2, 1) \notin R. Antisymmetric ✓ — no mutual pair between distinct elements.

R_4 = \{(1,2), (2,1), (1,3)\}. Symmetric ✗ — (1, 3) without (3, 1). Antisymmetric ✗ — (1, 2), (2, 1) with 1 \neq 2.

Four relations, four different cells in the symmetric/antisymmetric table. Confirms they are independent.

The takeaway rule

Stop reading "antisymmetric" as "opposite of symmetric." Read it as "no mutual pairs between distinct elements." Then it stops colliding with symmetric and starts being its own precise, independent property. The two labels can be true together (equality), true alone (\leq or \{(1, 2), (2, 1)\}), or false together. Four boxes — put every relation in one of them and your confusion ends.