Students often read "antisymmetric" and assume it means "the opposite of symmetric," the way "irrational" means "not rational." It does not. Antisymmetric is a completely separate property with its own rule. "Not symmetric" is the negation of symmetry; antisymmetric is a stronger condition about what happens when pairs do reverse.
Both can fail. Both can hold. A relation can be neither symmetric nor antisymmetric. That alone proves the two ideas are different — if antisymmetric were just "not symmetric," every relation would satisfy one or the other, never both-fail.
The two definitions, side by side
Symmetric
R on A is symmetric if for every a, b \in A:
Every arrow has its reverse.
Antisymmetric
R on A is antisymmetric if for every a, b \in A:
The only pairs allowed to reverse are the diagonal pairs (a, a).
Not symmetric is the logical negation of symmetric: there exists some pair (a, b) \in R with (b, a) \notin R. One witness is enough.
Antisymmetric is a for-every statement: for every pair of distinct elements a \neq b, you never find both (a, b) and (b, a) together in R.
Why: the two properties are in different logical moods. "Symmetric" is "for every pair, reverse is also in." "Antisymmetric" is "for every pair, if both directions are in, then the two elements were equal." "Not symmetric" just negates the first. Antisymmetric adds new structure — it restricts which pairs R is allowed to have — it does not simply drop the symmetric requirement.
The four-quadrant table
Every relation on A lands in exactly one of four boxes.
Why is the top-right cell (symmetric and not antisymmetric) almost impossible? If R is symmetric and contains any pair (a, b) with a \neq b, then (b, a) \in R too — and that already violates antisymmetry. So the only way to be both symmetric and antisymmetric is to have no off-diagonal pairs at all: the equality-style relations where R \subseteq \{(a, a) \mid a \in A\}.
A relation that is neither
This is the clearest proof that "not symmetric" \neq "antisymmetric." Take A = \{1, 2, 3\} and
- Symmetric? The pair (2, 3) \in R but (3, 2) \notin R. So R is not symmetric.
- Antisymmetric? Check: both (1, 2) and (2, 1) are in R, and 1 \neq 2. That already violates antisymmetry. So R is not antisymmetric either.
Both fail. Neither property holds. This is forbidden if antisymmetric were the negation of symmetric, so antisymmetric cannot be the negation of symmetric.
Interactive: flip through the four cases
Quick recognition rules for exams
- Symmetric holds iff the grid of R is mirror-symmetric across the main diagonal.
- Antisymmetric holds iff, for every off-diagonal cell (a, b) with a \neq b that is in R, the mirror cell (b, a) is not in R.
- Not symmetric holds iff you can find a single off-diagonal pair (a, b) \in R whose mirror (b, a) \notin R.
Symmetric and antisymmetric are two independent yes/no questions. "Not symmetric" answers only one of them and leaves the other open. To check antisymmetry, look at every off-diagonal pair in R and confirm its mirror is missing; do not assume it follows from symmetry failing somewhere.
Is "divides" antisymmetric on $\mathbb{Z}^+$?
Suppose a \mid b and b \mid a with a, b positive integers. Then b = ka and a = lb for positive integers k, l, giving a = l(ka) = (lk)a, so lk = 1. Since k, l are positive integers, k = l = 1, which forces a = b.
So yes — on \mathbb{Z}^+, the divisibility relation is antisymmetric. And it is definitely not symmetric (2 \mid 4 but 4 \nmid 2). So divisibility lives in the "antisymmetric, not symmetric" cell of the table.
Why: on \mathbb{Z} (allowing negatives), divisibility loses antisymmetry — for instance 3 \mid -3 and -3 \mid 3 but 3 \neq -3. Always check which set the relation lives on before concluding antisymmetry.
One-liner
"Not symmetric" means the mirror rule fails at least once. "Antisymmetric" means the mirror rule can only succeed on the diagonal. They live on different axes, and a single relation can independently satisfy neither, one, the other, or both.
Related: Relations · Symmetry Check Visualisation · A Relation Both Symmetric and Antisymmetric — Possible? · Equivalence Relations