How to build a cloze test that actually measures comprehension

A cloze test looks almost too simple to be useful: take a passage, delete some words, and ask the reader to fill them back in. Wilson Taylor introduced the technique in 1953 as a way to measure how well a reader could reconstruct meaning from context — not just recognize words, but actively predict what belongs, using grammar, vocabulary, and the surrounding sentence as evidence. Done well, it's one of the few exercises that tests comprehension and production at the same time. Done carelessly — blanks placed at random with no thought to what they're testing — it produces a worksheet that measures almost nothing.

Here's what actually goes into building a cloze test that means something, and how to read the result once students take it.

Two ways to choose what gets blanked

There are two standard approaches to picking which words disappear, and they test different things.

Interval (fixed-ratio) deletion removes every Nth word — classically every fifth word, which is the original Taylor method. This is the more rigorous, more "standardized-test" version: because the blanks fall wherever the count lands, you can't cherry-pick easy or hard deletions, and results are comparable across different passages and different test-takers. The tradeoff is that a fifth-word interval will sometimes blank a word that's genuinely unguessable from context (a proper noun, a number, a highly specific term) and sometimes blank a trivial function word (like "the" or "and") that any reader fills in without thinking.

Random deletion at a set difficulty percentage blanks a chosen proportion of words — say 15% or 25% — chosen at random rather than at fixed intervals. This gives you direct control over difficulty: a lower percentage produces an easier worksheet with more context intact around each blank; a higher percentage produces a harder one. It's less standardized than the interval method, but far more practical for a teacher building graded exercises for a specific class.

Why excluding proper nouns matters

Both deletion methods can land on a capitalized proper noun — a character's name, a place, a brand. Proper nouns are usually a bad test of comprehension: either the reader already knows the name from earlier in the passage and simply recalls it (testing memory, not context inference), or it's a name they have no way to derive from context at all (testing nothing but guessing). Most cloze generators, including this one, offer an option to skip capitalized words when choosing blanks, which keeps every blank testing the same thing — contextual and grammatical inference — instead of mixing in name-recall as a separate, unrelated skill.

This matters more than it looks. A test with three blanked character names in a row tells you nothing about reading level; it tells you whether the student was paying attention to who's who. Turning that option on before generating the test removes an entire category of noise.

How to score the result

Traditional cloze scoring uses the reader's percentage of exact-word correct answers (not just "a word that fits the meaning") against three bands, developed by researchers including John Bormuth in the 1960s and widely adopted in reading-level research since:

• Above 60% correct — independent reading level. The text is comfortable; the reader could handle it unsupported.
• 40–60% correct — instructional level. The text is appropriately challenging with guidance — the sweet spot for classroom material.
• Below 40% correct — frustration level. The text is too difficult for this reader without significant support.

These thresholds are deliberately lower than they'd be for a multiple-choice quiz, because exact-word cloze scoring is a much harder standard — a synonym that preserves meaning perfectly still counts as wrong under strict scoring. Some teachers relax this to "any word that fits grammatically and semantically," which shifts the bands upward but sacrifices some of the comparability that made the original scale useful.

What a cloze test is good for beyond the classroom

Language teachers use cloze exercises to build comprehension worksheets, but the same technique doubles as a readability check: if a passage produces a very low fill-in rate across a class, that's a signal the vocabulary or sentence structure is above the group's level, independent of what a formula like Flesch-Kincaid reports on the same text. It's also a natural fit for ESL vocabulary review (blank the target vocabulary specifically rather than at random), self-study reading practice, and even light editing self-checks — if you can't predict your own next word in a passage you wrote, a reader with less context definitely can't either.

Building your own

The Cloze Test Generator takes any passage and produces a fill-in-the-blank version using either method: fixed-interval blanking for a standardized, comparable test, or random blanking at a difficulty percentage you choose for a graded worksheet. Turn on proper-noun exclusion to keep every blank testing context rather than memory, then paste in the answer key it generates alongside the exercise. Start with an interval of five and a moderate difficulty percentage, try it on a passage you know well, and adjust from there once you see which blanks your test-takers find fair and which feel like guessing games.