How Readability Scores Work

A readability score is two measurements wearing a trench coat: how long your sentences are, and how heavy your words are. The Flesch Reading Ease and Flesch-Kincaid Grade formulas have turned those two numbers into a verdict since the 1940s, and both run on the same tiny inputs — words per sentence and syllables per word. Once you have seen the arithmetic, the scores stop being magic and start being an editing tool.

Two formulas, the same two ingredients

Rudolf Flesch published Reading Ease in 1948. It subtracts penalties from a starting value of 206.835 and reports how easy a text is — higher is simpler:

Flesch Reading Ease = 206.835 - 1.015 x ASL - 84.6 x ASW

  ASL = words / sentences   (average sentence length)
  ASW = syllables / words   (average word length, in syllables)

Years later the formula was flipped around for the U.S. Navy, which wanted to file its technical manuals by school grade. Working with J. Peter Kincaid in the 1970s, the same two ingredients produced the Flesch-Kincaid Grade Level, where higher is harder and the output is meant to be read as a U.S. school grade:

Flesch-Kincaid Grade = 0.39 x ASL + 11.8 x ASW - 15.59

That is the entire trick. There is no vocabulary list, no grammar analysis, no dictionary — just two ratios. This is why the scores are fast, consistent, and a little crude: the formulas cannot tell a well-chosen long word from a badly chosen one.

How syllables get counted

Everything hinges on the syllable count, and syllables are genuinely hard to define — dictionaries disagree. Tools use a heuristic, and ours (in the Text Statistics tool) works like this: keep only the letters a-z; if three or fewer letters remain, call it one syllable; drop a silent trailing e unless it is part of a consonant-plus-le ending like "able" or "turtle"; then count each run of vowels — a, e, i, o, u, and y — as one beat.

cat       -> 1        (short word, one run of vowels)
syllable  -> 3        (y, a, and the kept -le ending)
beautiful -> 3        (eau is one vowel run, not three)
queue     -> 1        (ueue is one run — right by luck)

The heuristic misses too. rhythm scores 1 (really 2 — the y is doing two jobs), science scores 1 (the ie is one run), and every scores 3 (really 2). The saving grace is that the errors point in all directions, so across a whole passage they mostly wash out. You can check any single word in the Syllable Counter, which shows its per-word verdicts so you can see the heuristic's reasoning.

A worked example, with the real numbers

Run this sentence through the tool:

Many readers skim long sentences before they lose interest and click away.

The tool counts 12 words, 1 sentence, and 20 syllables (Many 2, readers 2, skim 1, long 1, sentences 3, before 2, they 1, lose 1, interest 3, and 1, click 1, away 2). That gives ASL = 12/1 = 12 and ASW = 20/12 ≈ 1.67. Now the formulas:

Reading Ease = 206.835 - 1.015 x 12 - 84.6 x (20/12)
             = 206.835 - 12.18 - 141.0
             = 53.7

Grade level  = 0.39 x 12 + 11.8 x (20/12) - 15.59
             = 4.68 + 19.67 - 15.59
             = 8.8

The tool reports exactly this: Reading Ease 53.7, grade 8.8, labeled "Fairly Difficult (10th–12th grade)". (It computes with full precision and rounds for display, so 20/12 enters the math as 1.6667…, not 1.67.)

Reading the bands — and their quirks

The Reading Ease number maps to a level band:

• 90 and up — Very Easy (5th grade)
• 80–89 — Easy (6th grade)
• 70–79 — Fairly Easy (7th grade)
• 60–69 — Standard (8th–9th grade)
• 50–59 — Fairly Difficult (10th–12th grade)
• 30–49 — Difficult (College)
• below 30 — Very Difficult (College graduate)

Two quirks follow straight from the formula. Very short words can push the score above 100 — "The cat sat on the mat. It ran to the sun." scores 116.7, and its grade level comes out at -1.6, meaning "below first grade". At the other end, jargon-heavy text can go below 0. Most web writing aims for the 60–70 band: Standard, readable by a wide audience without sounding like a children's book. A physics paper at 35 is doing its job; a landing page at 35 has a problem.

What editing actually moves the needle

Since the score is only two ingredients, only two levers exist. Splitting sentences attacks ASL; swapping latinate words for short ones attacks ASW. Both formulas punish sentence length gently (coefficients 1.015 and 0.39) and word heaviness savagely (84.6 and 11.8) — one 4-syllable word in a short passage costs more than a whole extra sentence. Note the two scores can disagree with each other: a text can land in one band on Reading Ease and a different grade on Flesch-Kincaid, because they weight the ingredients differently. Treat the numbers as a compass heading, not a grade for your intelligence — and remember the constants are tuned for English only.

Reading time, the cousin metric

The same word count powers a friendlier number: how long a text takes to get through. The Reading Time Estimator divides words by a speed and rounds up to the next second — 200 words per minute for silent reading and 130 for speaking aloud by default. A 900-word post is 4m 30s to read and 6m 56s to read aloud, which is why a script that "reads in five minutes" is really a six-minute talk. Under the hood the counting is simple: words are just runs of non-whitespace, the same splitting a regex does (see How Regular Expressions Work).

Try it yourself

Paste anything you are writing into the Text Statistics tool to see word, sentence, and syllable counts plus both readability scores with their level labels. Drill into a suspicious word with the Syllable Counter, which shows the per-word syllable verdicts the score is built on. And before you publish, run the piece through the Reading Time Estimator to get the honest read-aloud length, not just the silent-reading estimate.