Circular permutations

Line up five friends for a photograph and there are 5!=1205! = 120 distinct orders. Seat those same five friends around a circular table and something changes: rotate everyone one seat to the right and the photograph looks identical — same neighbours, same left-right relationships, just relabelled. A circular arrangement has no fixed starting point, so the usual factorial overcounts every arrangement by the number of ways it can be rotated into itself. This article builds the fix. We derive why circular arrangements of nn distinct objects number (n−1)!(n-1)! rather than n!n!, work out when clockwise and anticlockwise seatings must be counted as different (a table) versus the same (a necklace that can be flipped), and combine this with the "glue" trick for people who must sit together. The reasoning uses only the fundamental principle of counting you already know from linear permutations — applied to a shape with no edges.

The core insight

Circular arrangements of nn distinct objects number (n−1)!(n-1)!, not n!n!, because fixing one object's position removes the rotational duplicates that a row does not have; halve this further to (n−1)!/2(n-1)!/2 only when the arrangement can physically be flipped over.

The core idea

Why a circle is not a row. In a row, position 1 is genuinely different from position 2 — there is a definite "first" seat. Around a circle, there is no such anchor: the arrangement A-B-C-D-E and the arrangement E-A-B-C-D (everyone shifted one seat) describe exactly the same seating, because everyone still has the same left neighbour and right neighbour. Any one circular arrangement of nn people can be rotated into nn different-looking linear listings, all representing the same physical seating.

The fixing trick. Since ordinary linear counting gives n!n! arrangements of nn distinct objects, and each real circular seating has been counted nn times over (once per rotation), the number of genuinely distinct circular arrangements is

n!n=(n−1)!\frac{n!}{n} = (n-1)!

where: nn is the number of distinct objects being arranged, and (n−1)!(n-1)! is the count of distinct circular permutations once rotations are treated as identical. Equivalently, fix any one object's seat as a reference point — this uses up the rotational freedom in one stroke — and arrange the remaining (n−1)(n-1) objects relative to it, an ordinary linear permutation of (n−1)!(n-1)! ways.

Clockwise versus anticlockwise. When "left" and "right" are physically distinguishable — people at a table, cars at a roundabout — a clockwise seating and its mirror-image anticlockwise seating are genuinely different arrangements, and (n−1)!(n-1)! is the final answer. But when the whole arrangement can be picked up and flipped over — beads on a necklace, flowers in a garland, a keychain — a clockwise arrangement and its flipped anticlockwise twin look identical, so every pair collapses into one:

Distinct arrangements (flippable)=(n−1)!2\text{Distinct arrangements (flippable)} = \frac{(n-1)!}{2}

Restricted circular arrangements. When specific objects must sit together, glue them into a single block first (as in linear permutations), then apply the circular formula to the reduced set of units, and finally multiply by the internal arrangements within the block.

Figure
Five people in a row give five factorial orders while the same five people around a circle collapse to four factorial because every rotation looks identical
A row cares about the first seat. A circle does not. in a row: 5! = 120 orders A B C D E seat 1 is a fixed, distinguishable starting point around a circle: (5-1)! = 24 arrangements A B C D E rotating every seat one place gives the identical seating, so n! is divided by n
Derivation ACircular permutations of n distinct objects

The goal is to show, using only the fundamental principle of counting and the ordinary linear-permutation formula a student already owns, that the number of genuinely distinct ways to arrange nn distinct objects around a circle — where a rotated copy of an arrangement is not counted as a new arrangement — is exactly (n−1)!(n-1)!, and to make explicit why the fixing device is a valid counting trick rather than a physical assumption.

Step 1. Count all linear listings first. If the nn objects were instead written out in a row, the fundamental principle of counting gives n!n! distinct orderings — nn choices for the first slot, n−1n-1 for the second, and so on down to 11.

Step 2. Identify the overcounting caused by rotation. Take any one circular seating and list it starting from each of the nn positions in turn, going the same direction each time. This produces nn different linear listings, but all nn of them describe the very same circular arrangement — same neighbours all round, just written down starting from a different seat. So every genuine circular arrangement has been counted exactly nn times among the n!n! linear listings of Step 1.

Step 3. Divide out the overcount. Since each of the true circular arrangements accounts for exactly nn of the n!n! linear listings, the number of distinct circular arrangements is n!/n=(n−1)!n!/n = (n-1)!.

Step 4. Confirm with the fixing argument. An equivalent way to see this: fix one specific object's seat as the reference (say, always seat A first, wherever that happens to be around the table). This single choice absorbs all nn rotational copies into one, because "A's seat" is no longer a free choice — only the other (n−1)(n-1) objects' positions relative to A vary, and arranging them is an ordinary linear permutation of (n−1)(n-1) objects, giving (n−1)!(n-1)!. Both routes agree, confirming the formula.

Worked example

In how many ways can 6 people be seated around a circular table (a) if clockwise and anticlockwise seatings are considered different, and (b) if they are considered the same? Extend this: in how many ways can the same 6 people be seated around the table if 2 particular people must always sit together?

Step 1. Directly apply the circular formula for part (a), with clockwise and anticlockwise treated as distinct (true for seating at a physical table): with n=6n = 6, (n−1)!=5!=120(n-1)! = 5! = 120.

Step 2. For part (b), if clockwise and anticlockwise are to be treated as the same arrangement (as if the table itself could be flipped), halve the Step 1 result: 5!/2=120/2=605!/2 = 120/2 = 60.

Step 3. For the "2 people together" extension, glue the pair into a single unit, leaving 6−2+1=56 - 2 + 1 = 5 units to seat around the circle. Apply the circular formula to these 5 units: (5−1)!=4!=24(5-1)! = 4! = 24.

Step 4. Within the glued pair, the 2 people can swap places in 2!=22! = 2 ways, and by the fundamental principle of counting the two choices combine by multiplication: 4!×2!=24×2=484! \times 2! = 24 \times 2 = \mathbf{48}.

Final answers: (a) 120 seatings, (b) 60 seatings, and 48 seatings when 2 particular people must sit together.

Common traps

  • Using n!n! instead of (n−1)!(n-1)! for a plain circular arrangement — forgetting to remove the rotational duplicates is the single most common error.
  • Halving every circular answer out of habit — the /2/2 correction applies only when the object can physically be flipped over (necklace, garland, bracelet), never for seating at a table where left and right are distinguishable.
  • Forgetting the internal k!k! factor when a "together" constraint is glued into a circular arrangement — the pair (or block) can still be internally reordered.
  • Using the wrong base count after gluing — once kk objects are glued into one unit, the circular formula applies to (n−k+1)(n - k + 1) units, giving (n−k)!(n-k)!, not the original (n−1)!(n-1)!.

Ask the tutor: If it is unclear whether a specific circular arrangement problem needs the divide-by-2 correction, ask the AI tutor whether the object in the question — table, necklace, keychain — can physically be flipped over, since that single check settles it.

Why do we divide by 2 for a necklace but not for people at a table?

The /2/2 correction accounts for arrangements that become identical when the whole shape is flipped over. A necklace or garland can physically be turned over, so a clockwise arrangement and its mirror-image anticlockwise arrangement look exactly the same object and must be counted once. A table cannot be flipped in that way — the chair to your left stays to your left regardless of which direction you look — so clockwise and anticlockwise seatings around a table remain genuinely different arrangements, and only (n−1)!(n-1)! applies, with no halving.

Is fixing one person's seat cheating, since seats around a real table are not actually identical?

No — fixing a seat is purely a counting device, not a claim about physical chairs. What matters in a circular arrangement is only each object's position relative to the others, since the entire circle can be rotated freely without changing who sits next to whom. By declaring one object's seat as the reference point, you remove that rotational freedom from the counting problem entirely, and every other arrangement of the remaining (n−1)(n-1) objects around it is now genuinely distinct. It works no matter which object you choose to fix, and the real seats at the real table are never assumed to differ.

How do I handle a problem where some people must NOT sit together, in a circle?

Use the same complement strategy as linear permutations: first count the total circular arrangements with no restriction, (n−1)!(n-1)!, then subtract the circular arrangements where the given people ARE seated together (found by gluing them into a block, applying (n−k)!(n-k)!, and multiplying by the internal k!k! arrangements of the block). The difference gives the arrangements where they are never adjacent. The circular formula only changes how the "total" and "together" counts are computed; the subtraction logic itself is identical to the linear case.

Before this

Where this leads