Roman Numerals Rules: Seven Letters, Three Rules, and How to Spot an Invalid Numeral

Strip away the history and the marble engravings, and the Roman numerals rules fit in three sentences. Symbols add when they're equal or shrinking in value. A smaller symbol in front of a bigger one subtracts — but only six specific pairs are allowed to do this. And no symbol repeats more than three times, with V, L and D never repeating at all. That's the whole rulebook. Everything else on this page — the tables, the error list, the mystery of the clockmaker's IIII — is those three rules spelled out, stress-tested, and applied.

If you arrived here uncertain whether IC means 99 (it doesn't), you'll leave with a way to settle that question for any numeral you meet. And if you just want the big picture of where these numbers came from and why we still use them, that's a different question — I covered it in what Roman numerals are; this page is the rulebook itself.

Seven Letters, and Only Seven

The entire system is built from seven Latin letters:

Two mnemonics are in circulation: "MeDiCaL XaVIer" (the values in descending order) and "My Dear Cat Loves Xtra Vitamins Intensely". Pick whichever sticks. The word origins are honest helpers too — C really does trace to centum and M to mille — though scholars caution the letter-shape match came after the symbols already existed, so treat the etymology as a memory aid, not an origin story.

Now the part people keep googling: there is no other letter. No A, no B, no K, nothing beyond these seven. Search queries like "what does K stand for in Roman numerals" have a short answer — it doesn't stand for anything, because K isn't in the alphabet. If you see a K, someone is writing knife-edge code or a product name, not a Roman numeral.

Rule 1: Equal or Descending Symbols Add

When symbols repeat, or shrink, as you read left to right, you add. II is 1+1. XXX is 10+10+10. VI is 5+1, XVII is 10+5+1+1. Reading a well-formed numeral is just tallying values as you go — the value of each letter never changes, because the system has no place value. The 2 in "20" and "200" means different things; the X in XVI and XXX means ten, always.

This is also your first validity check: values should never grow as you move right. Once they do, you'd better be looking at one of the six sanctioned pairs — which brings us to the rule everyone gets wrong.

Rule 2: Subtraction, Six Pairs Only

A smaller symbol placed before a bigger one subtracts. IV is 5−1, XL is 50−10, CM is 1000−100. Simple so far. The catch is that only six of these pairs are legal:

Why exactly these six? Two constraints do the trimming. First, only I, X and C may act as the subtracting letter — the powers of ten, if you like. Second, each of them may only subtract from the next two values above it: I can subtract from V or X, X from L or C, C from D or M. Three letters times two targets each: six pairs, and Wikipedia confirms these are the only subtractive forms in standard use.

Every "clever shortcut" you're tempted by falls outside that whitelist. IC for 99? Invalid — I may only subtract from V and X. The correct form is XCIX, ninety plus nine. The same logic kills IL for 49 (it's XLIX), XM for 990 (CMXC), VC for 95 (XCV), and LM for 950, which is CML. The Government of Canada's style guide states the constraint the cleanest way I've found: only I, X and C can be subtracted, "and only from the next two numerals that are larger than it" — hence XCIX, not IC; CML, not LM (Government of Canada).

One subtlety when a small symbol sits between two large ones: it binds to its right-hand neighbor. XIV isn't (10−1)+5; it's 10+(5−1)=14. The subtraction always happens with the symbol immediately after.

Rule 3: No Runaway Repeats

A symbol can appear at most three times in a row. That's why 4 is IV rather than IIII, why 40 is XL not XXXX. Alongside this, V, L and D — the "five" symbols — are never repeated at all. You'll never see VV for 10 or DD for 1000, and for once there's a tidy reason: repeating a five gets you nothing new. VV is just X, DD is just M. The rule kills redundancy, not possibility.

Together the three rules produce a strict format: values descend from left to right, the only interruptions are those six pairs, and nothing runs four deep. Which makes invalid numerals surprisingly easy to catch — once you have a checklist.

Building Any Number: Split, Translate, Glue

To write a number, split it into thousands, hundreds, tens and units, translate each chunk independently, then glue the pieces together largest-first:

39 breaks into 30 + 9: XXX + IX = XXXIX. 944 is 900 + 40 + 4: CM + XL + IV = CMXLIV. 1984 is 1000 + 900 + 80 + 4: M + CM + LXXX + IV = MCMLXXXIV. And 2026 is simply MMXXVI. Missing chunks drop out — 1066 is M + LX + VI = MLXVI, with no hundreds symbol because there are no hundreds to write.

Reading works in reverse: scan left to right, add each value, and subtract whenever you hit one of the six pairs. When I'm not sure of myself, I do it both ways — write the numeral by the table above, then run it through the Roman numerals converter and compare. Takes five seconds, catches transpositions, and it's a habit worth keeping whenever a numeral is headed somewhere permanent.

This table is also the chart to pin up if you're converting numbers regularly; it generates every standard numeral from 1 to 3999 — the ceiling, because the thousands column runs out after MMM and there's no fourth M. (For an entry-by-entry 1-to-100 list, the Roman numerals chart spells out every one.) Larger numbers exist, but they need the overline (a bar multiplying a symbol by 1000), which is an extension rather than core Roman numerals.

Invalid Numerals: Catch Them Before Anyone Else Does

Here's the payoff of knowing the rules rather than guessing. Each of these shows up constantly — on clock towers, in Instagram captions, and, painfully often, in tattoo shops:

To audit any numeral, ask four questions. Are the values generally non-increasing left to right? Is every small-before-large pair one of the six? Do V, L or D appear twice? Does anything repeat four times? Ask them of anything you meet; four yeses and the numeral is standard.

The check matters double for roman numeral tattoos, which is where a slipped digit stops being a typo and becomes a lifelong souvenir. Ink doesn't get patched like a web page. If a numeral is headed somewhere permanent and fails the four-question audit, fix it first, then confirm against the converter.

A note on "wrong" forms you'll meet in the wild: many aren't modern errors at all, just older conventions. Medieval scribes wrote iij for 3 and viij for 8, with a final j to stop tampering; additive forms like IIII, XXXX and CCCC coexisted with subtractive ones for centuries; a French document might render 99 as IIIIXXXIX, mirroring the spoken quatre-vingt-dix-neuf. The National Archives documents these variants for people reading historical papers (The National Archives) — worth knowing before you "correct" a 400-year-old inscription that was never wrong.

Why Clocks Show IIII

And so we arrive at the four that launched a thousand forum threads. By the modern rules, 4 is IV. Look at a clock or watch with Roman numerals, though, and the 4 position almost always reads IIII — while the 9 position, obligingly, reads IX. The watch industry is unanimous about this; Citizen's own support pages field the question as a formal FAQ.

Several explanations coexist, and honestly none has been crowned the winner:

• IIII came first. The additive form is the older way to write 4; subtractive notation spread through Europe only after the printing press (per Wolfram MathWorld, citing the historian of mathematics F. Cajori). Early clockmakers inherited a world where IIII was routine, and the earliest famous example — the Wells Cathedral clock, late 1300s — already used it. Seiko's museum notes IIII remained the general European form for 4 until roughly the 17th century (Seiko Museum).
• Visual balance. The most popular theory in the trade. With IIII at 4 and VIII at 8, the dial splits into three balanced zones — I/II/III/IIII, V/VI/VII/VIII, IX/X/XI/XII — and the four strokes of IIII counterweight the four of VIII across the axis. On radially printed dials, where the lower numerals read upside down, IIII is also easier to distinguish than IV.
• Legibility for the barely literate. Addition is easier than subtraction, and IV mirrors VI in a way IIII never can. This mattered in an era when clocks were the only public timekeeper and most of the congregation couldn't read.
• The legends. Romans supposedly avoided chiseling IV because it opens the name IVPPITER — Jupiter, king of the gods. A later French version blames a king: Charles V (or Louis XIV, depending on the teller) objecting to seeing I subtracted from his own numeral V. Fun stories, both flagged as folklore by the very watch sites that repeat them — the molding-economy theory (IIII and VIIII would need only three casting molds) gets similar self-deprecating treatment.

Whatever mix of tradition and aesthetics holds it in place, IIII is not a mistake, and it isn't universal either: Big Ben's dial uses a crisp IV, and a few modern watchmakers do the same. Even the Romans mixed systems — on the Colosseum's surviving gates, 40 is XL but 4 is IIII, so gate 44 reads XLIIII. The standard rules tell you what's correct; the clock face tells you what's customary. When the two disagree on a dial, custom wins.

Quick Answers

Is IIII a valid Roman numeral? On paper and in formal modern use, no — 4 is IV. On clock and watch dials, yes, by an unbroken tradition dating to the 14th century. Context decides.

Why is 99 XCIX and not IC? Because I is only allowed to subtract from V and X. To build 99 you assemble it by place: 90 (XC) + 9 (IX) = XCIX. The system forbids reaching two columns over — that's the "next two numerals larger" rule doing its job.

What's the largest standard Roman numeral? 3999, written MMMCMXCIX. The thousands column tops out at MMM, and going higher requires the overline extension that multiplies a symbol by 1000.

How do I check a Roman numeral quickly? Run the four-question audit — order non-increasing, subtractive pairs within the six, no doubled V/L/D, no four-in-a-row — or paste it into the converter and see if the round trip holds.

Whichever way you verify, do verify. The rules are short; the errors are long-lived.