A fill in the blank question takes the options away. There is nothing to recognise and nothing to eliminate: the student either produces the word or does not. That is what makes the shape worth the extra care it needs, because a gap that three different words could fill is a worse question than the multiple choice it replaced.
A cloze passage is the same idea at the length of a paragraph: several gaps, each depending on the sense of the whole. EdProcto carries both, as the fill in the blank and cloze passage templates, and marks each gap by rule against the answers you say you will accept.
This guide covers which word to remove, how many gaps a passage can hold, how to mark what students type, and what the research says the shape measures.
What is the difference between fill in the blank and cloze?
A fill in the blank question is one sentence with one missing word or phrase. A cloze passage is a connected text with several words missing, where filling one gap often depends on having understood the sentences around it.
The cloze procedure was introduced by Taylor (1953) as a way to measure how readable a text was: remove words, ask readers to restore them, and count how many they get. It moved from there into language testing and then into every subject, because it is quick to write and quick to mark.
| Fill in the blank | Cloze passage | |
|---|---|---|
| Text | One sentence | A paragraph or more |
| Gaps | One | Several |
| Measures best | Recall of a term in context | Vocabulary, grammar and comprehension together |
| Main risk | More than one word fits | Gaps that can be filled without reading the rest |
Which word should you blank?
Blank the word that carries the idea you taught, and only a word that the rest of the sentence pins down. The question to ask of every gap is: could a student who understood the lesson put anything else here and be right?
Compare three versions of one sentence.
- "The ____ is the powerhouse of the cell." One word fits, and it is the term the lesson was about. Good.
- "The mitochondrion is the ____ of the cell." This tests a slogan. A student who writes "energy source" knows more than one who writes "powerhouse" and would be marked wrong.
- "The mitochondrion ____ the powerhouse of the cell." This tests the verb "is". Nothing was measured.
Rules that follow from this:
- Blank content words, not function words, unless grammar is the point of the lesson. In a language class, blanking the article or the preposition is exactly right.
- One gap in a single sentence. Two gaps in a short sentence leave too little context for either.
- Put the gap late in the sentence where you can. The student reads the context first and then recalls, instead of holding an empty slot in mind while reading.
- Do not lift the sentence from the textbook. A copied sentence rewards memory of the page. Say the idea in your own words.
- Make every gap the same length on the screen, so the size of the box is not a clue.
- Avoid "a" or "an" straight before the gap when it would give away whether the answer starts with a vowel.
Every fifth word, or chosen by hand?
Choose by hand. The original procedure removed words mechanically, at a fixed interval, and that is still common in reading tests. For a classroom quiz it wastes most of the gaps on words that test nothing you taught.
The research backs the choice. Alderson (1979) varied the text, the deletion frequency and the way answers were scored, and found that the results shifted enough with those choices that a cloze passage could not be assumed to be a valid test just because it was a cloze. He concluded that mechanical cloze tests measured lower-order language skill better than higher-order skill. Gellert and Elbro (2013) took the opposite approach and selected each gap so that filling it needed an understanding of the ideas in the text. In a study of 204 adults, their ten-minute cloze test correlated strongly (r = .84) with a standard thirty-minute comprehension test built from questions, and only a small part of that link was explained by decoding or vocabulary.
So a cloze passage measures whatever its gaps are chosen to measure. Picked by a counter, the gaps test local grammar. Picked by a teacher, they can test the argument.
How many gaps should a cloze passage have?
Leave enough text between gaps that each one can be worked out, and leave the first and last sentences whole. There is no settled number, and Alderson's results are a warning against trusting a fixed ratio. As a working rule for a quiz on taught content, one gap in every sentence or two is plenty, and a passage of 100 to 150 words carries six to ten gaps comfortably.
Whether students use the wider passage at all has been tested directly. Trace (2020) analysed 449 cloze gaps from 15 passages, taken by 675 first-language and 2,246 second-language readers, and found that gaps can draw on information at the level of the sentence and at the level of the whole passage. Some gaps need only the words either side. Others cannot be filled without the paragraph. If comprehension is what you want, write more of the second kind: gaps whose answer is settled by something said two sentences earlier.
A quick check before you publish: cover the rest of the passage and read only the sentence with the gap. If the answer is still obvious, that gap tests the sentence, not the passage.
How should you mark typed answers?
Decide what counts before the quiz runs, and write the list down. Marking typed answers has two traditions, and they give different scores for the same student.
- Exact word. Only the word that was removed counts. It is quick and consistent, and it marks a correct synonym wrong.
- Acceptable word. Any word that fits the meaning and the grammar counts. It is fairer, and it needs you to list the acceptable words in advance or to read the answers.
The scoring procedure was one of the three things Alderson (1979) varied, and it changed what the test appeared to measure. For a quiz on content you taught, the acceptable-word approach is the honest one: list the term, its common synonyms and its plural or verb forms. For spelling tests and for technical terms where the exact word is the learning, hold to the exact word and say so in the instructions.
Then settle the small things: whether capital letters matter, whether a hyphen matters, and whether a one-letter slip in a long word is a wrong answer or a typing error. In most subjects it is a typing error.
Should you give a word bank?
Only if recognition is what you want to test. A word bank turns the question back into a matching exercise: the student picks from a list instead of recalling. That is easier and it is sometimes right, for young learners or for a first pass over new vocabulary. If you do give one, include more words than gaps, so that the last gap is not filled by elimination.
Do fill in the blank questions teach better than multiple choice?
With feedback, questions that make the student produce the answer tend to teach more than questions that let them recognise it. Kang et al. (2007) had students read short journal papers and then take a multiple-choice test, take a short-answer test, or read the same facts as statements. Three days later, when no feedback had been given on the first test, the multiple-choice group did best. When feedback was given, the short-answer group did best. The authors' reading was that the harder the retrieval, the greater the benefit, as long as the student finds out what the right answer was.
The wider evidence points the same way. Rowland (2014), pooling the studies that compare testing with restudying, found larger benefits from initial tests that required recall than from tests that required recognition.
Two cautions. Those studies used short-answer and recall tests, not cloze passages as such, so a gap fill is a relative of what was tested, not the same thing. And the advantage depends on feedback: a student who types a wrong word and never sees the right one has practised the error.
How does EdProcto build and mark gap questions?
EdProcto is the best tool for gap questions when they are written from your own text, because the two hard parts of the shape, choosing the gaps and listing the accepted answers, are both in front of you before a student sees anything.
Paste the chapter or the handout into the quiz maker on the home page and choose the cloze shape. Procto AI writes the passage and its gaps from your material, and every gap is yours to move, remove or reword. For a single sentence, the fill in the blank template takes a sentence, the missing word and the list of answers you accept. For an answer that is one word with no sentence around it, there is one word.
Marking is by rule and immediate. EdProcto compares what the student typed with your accepted list, ignoring capital letters, spacing and punctuation, and allows a one-letter slip when you switch that on. A cloze passage is marked gap by gap, so a student who fills seven of ten is told so. The accepted answers stay on the server: the page the student sees has the boxes and nothing behind them.
For listening, the dictation shape plays a clip and marks the transcription the same way, which is the spoken form of the same test.
After the quiz, the feedback settings show the correct answer and your explanation at once, after submitting, or not at all, and the results list each question's difficulty; the full analysis, on the Educator plan, adds how well it separated stronger students from weaker ones. A gap that most of the class filled with the same wrong word is either a misconception to teach or an acceptable answer you forgot to list, and the results are where you find out which.
The Free plan carries the core templates, marking by rule and results question by question. Tutor opens all 111 templates and Procto AI's writing from a PDF or a Word file. The pricing page has the detail.
When should you use another shape?
Use another shape when a gap cannot be made unambiguous, or when the answer is longer than a phrase.
- Several plausible answers and you want to see which one students pick: a multiple-choice question, with the wrong answers as options.
- A fact that is simply right or wrong: a true or false statement, in numbers large enough to beat guessing.
- An explanation in the student's own words: a short answer, marked against a rubric.
- A calculation: an estimate, scored by how close the number is.
Quick questions
How many blanks should one sentence have?
One. A second gap in a short sentence removes the context the first one depends on. Use a cloze passage when you want several gaps, and leave a sentence or two of whole text between them.
Should spelling count in a fill in the blank question?
Only when spelling is what you are teaching. Otherwise accept a one-letter slip and the common alternative spellings. EdProcto ignores capitals, spacing and punctuation and can allow a single-letter slip on each question.
What if a student writes a correct answer I did not list?
Add it to the accepted list, then press "Re-mark this quiz" on the results page: every attempt handed in is marked again from the list as it stands, and the next sitting is marked rightly too. This is the commonest fault in gap questions, so read the wrong answers after the first run of any new question. In EdProcto each attempt in the results shows what the student typed.
Is a cloze test a reading test or a knowledge test?
It depends on the gaps. Gellert and Elbro (2013) showed that gaps chosen to need the ideas of the text track comprehension closely, and Alderson (1979) showed that gaps removed mechanically mostly track local language skill. Choose the gaps for what you want to know.
Can students look the answers up?
A typed gap in your own paraphrase is harder to search for than a copied sentence, which is one more reason to rewrite. For a quiz that counts, EdProcto's integrity settings add a time limit, notice a student leaving the tab and log it for you to judge.
References
- Alderson, J. C. (1979). The cloze procedure and proficiency in English as a foreign language. TESOL Quarterly, 13(2), 219–227. https://doi.org/10.2307/3586211
- Gellert, A. S., & Elbro, C. (2013). Cloze tests may be quick, but are they dirty? Development and preliminary validation of a cloze test of reading comprehension. Journal of Psychoeducational Assessment, 31(1), 16–28. https://doi.org/10.1177/0734282912451971
- Kang, S. H. K., McDermott, K. B., & Roediger, H. L. (2007). Test format and corrective feedback modify the effect of testing on long-term retention. European Journal of Cognitive Psychology, 19(4–5), 528–558. https://doi.org/10.1080/09541440601056620
- Rowland, C. A. (2014). The effect of testing versus restudy on retention: A meta-analytic review of the testing effect. Psychological Bulletin, 140(6), 1432–1463. https://doi.org/10.1037/a0037559
- Taylor, W. L. (1953). "Cloze procedure": A new tool for measuring readability. Journalism Quarterly, 30(4), 415–433. https://doi.org/10.1177/107769905303000401
- Trace, J. (2020). Clozing the gap: How far do cloze items measure? Language Testing, 37(2), 235–253. https://doi.org/10.1177/0265532219888617