The Coleman–Liau Index: the readability formula built for computers

Almost every readability formula has the same weak link: counting syllables. Flesch, Flesch–Kincaid, Gunning Fog, SMOG — all of them need to know how many syllables are in each word, and a computer is genuinely bad at that. English spelling barely encodes syllable boundaries, so syllable counters lean on dictionaries and heuristics that quietly fail on names, brand words, technical terms, and anything OCR mangled. Every one of those errors flows straight into the score.

The Coleman–Liau Index was built in 1975 by Meri Coleman and T. L. Liau to dodge that problem entirely. Their insight was simple: you don't need syllables to predict difficulty — you can use letters, which a machine counts perfectly every time. That one substitution makes it one of the most reliable readability formulas to automate, and it's why it shows up so often inside software.

The formula, and what L and S actually mean

The index returns a U.S. grade level from two ratios:

CLI = 0.0588 × L − 0.296 × S − 15.8

• L is the average number of letters per 100 words. Longer words push L up, which pushes the grade up.
• S is the average number of sentences per 100 words. More sentences per 100 words means shorter sentences — so a higher S lowers the grade. That minus sign in front of S is doing exactly what you'd expect: short sentences are easier.

A result of 9 means the text suits a ninth-grade reader; 12 is high-school senior; 16 is roughly college graduate. The two coefficients and the constant were fit by Coleman and Liau against cloze-test results — a comprehension test where readers fill in deleted words — which is still treated as the gold standard for measuring whether prose was actually understood.

Letters, not characters — a distinction that matters

It's tempting to lump Coleman–Liau in with the Automated Readability Index as "the character-based formulas," and the family resemblance is real. But they don't count the same thing. Coleman–Liau counts letters — the alphabetic characters that make up words. ARI counts characters, which in most implementations includes digits and punctuation attached to words.

On clean prose the gap is tiny. On text full of numbers, code, currency symbols, or URLs it isn't, and the two formulas can drift apart. If your content is heavy on those, Coleman–Liau's letters-only approach tends to be the steadier of the two, because it ignores the punctuation noise rather than feeding it into the score.

When Coleman–Liau is the right pick

Reach for it whenever syllable counting is the thing most likely to break:

• OCR'd or scanned text. Optical character recognition garbles word shapes in ways that wreck syllable estimates but barely move a letter count. Coleman–Liau degrades gracefully where Flesch-family formulas produce nonsense.
• Content thick with proper nouns. Product names, place names, and people's names are exactly where syllable dictionaries miss. Letters don't care whether a word is in a dictionary.
• Social-media and user-generated text. Slang, hashtags, and invented words have no reliable syllable count. They still have letters.
• Anywhere you need determinism. Because the inputs are unambiguous, the same text always yields the same score, with no dependence on which syllable library happens to be installed.

Why it disagrees with Flesch–Kincaid

Run the same paragraph through Coleman–Liau and Flesch–Kincaid and you'll usually land within a grade or two. When they split further apart, it's almost always because of how each one judges word difficulty.

Flesch–Kincaid uses syllables. Coleman–Liau uses letter length. Most of the time those agree — long words tend to have more syllables. But not always. A word like "available" is four syllables in only nine letters; "strengths" is one syllable in nine letters. Text packed with short-but-polysyllabic words reads as harder to Flesch–Kincaid and easier to Coleman–Liau. Text packed with long-but-simple words flips it. When the two disagree by more than a grade, that mismatch is the tell: read the passage yourself and decide whether it's the syllables or the raw word length that better reflects the difficulty your reader will feel.

Where it quietly misleads

Coleman–Liau inherits the blind spots of every mechanical formula, plus one of its own.

• It equates long with hard. Familiar long words ("understanding", "information") inflate the grade even though readers breeze past them, while a short, rare word ("writ", "tort") sails through unflagged. Vocabulary-sensitive work is better served by Dale–Chall, which checks words against a familiarity list rather than measuring their length.
• Short samples are noisy. A single sentence or a stray heading can swing the score wildly. Like all these formulas, it wants a few hundred words before the number means much.
• It measures mechanics, not meaning. Letter and sentence statistics say nothing about logical flow, argument quality, or tone. A confusing paragraph of short, simple words can score as easy while being genuinely hard to follow.

How to use it in practice

Treat Coleman–Liau as your robust readability check — the one you trust when the text is messy and syllable-based scores look unreliable. Pair it with one syllable-based formula for a general audience and a vocabulary-aware one if your readers are children or English learners. If all three point the same direction, act on it. If Coleman–Liau and a syllable formula disagree by more than a grade, that's your signal to read the passage and use judgment rather than trusting either number outright.

When you want the score without doing the letter-counting by hand, the Coleman–Liau Index calculator computes L, S, and the resulting grade level instantly — and because it needs no syllable dictionary, it works the same way on the awkward, proper-noun-heavy, OCR'd text where other formulas start to wobble.