How to Create Good Active Recall Questions

A good active recall question names one retrieval target and makes you produce it before you look. A weak one rereads the page, dumps the whole topic, or puts the answer in the wording.

If you already retrieve from notes — cue, try, then check — this page is about the cue: how to write active recall questions that are worth answering. The retrieve-before-looking rule lives in how to use active recall with notes. Here the work is the prompt.

Keep the note as the source. Add a few focused prompts that match the kind of knowledge on the page: a mechanism, a comparison, a procedure, a term. You do not need a question for every sentence. You do not need a question at all if a heading already tells you what to reconstruct.

The test is simple. Hide the note. Read only the prompt. Do you know what to produce, and do you have to produce it? If the wording hints the answer, if the target is “everything on this page,” or if you could pass by recognizing a familiar sentence, rewrite or delete.

What a good retrieval question is for

A retrieval question is a cue. It specifies what to produce before you reopen the note.

It is not a scored exam. It is not a study-guide dump. It is a handle: when you see it later, you attempt one idea, then check.

A usable prompt has a few properties. The same demands show up in practitioner writing on prompts for spaced repetition, even when the unit is a note rather than a card:

• Specific target. One idea, distinction, or step — not “explain the page.”
• Answerable from the note. If the answer is not on the page, the prompt is guessing, or the page is incomplete.
• Effortful. You produce the idea: say it, write it, or sketch it. Yes/no and multiple-choice stems mostly train recognition.
• Checkable. After the attempt, compare with the note. Feedback after a retrieval attempt helps, especially when you were wrong or unsure; skipping the check can leave errors in place.
• Does not give the answer away. If the important words are already in the question, you are rereading.

Practice tests and other retrieval attempts often produce better long-term retention than rereading the same material. The advantage is clearer after a delay than immediately after study. That is why the prompt matters: it is what you retrieve from. The loop around the prompt is the parent article.

If you cannot answer a well-aimed prompt, the note may be the problem. Fix the page — add the missing relation, the example, the condition — rather than only rewording the question. Turning a heading into a question while the notes are still messy is repair work after class, not this craft by itself.

A heading can be enough. Osmosis and tonicity is already a cue if you can reconstruct the mechanism from the title. Write a question when the heading is too broad, or when you keep retrieving the wrong slice of the page.

Weak questions vs strong questions

The fastest way to learn the craft is to rewrite. The two running notes below are short pages you might actually keep: one on osmosis and tonicity, one on price ceilings. Same pages, different knowledge types.

Conceptual mechanism

Weak

What is osmosis?

You can recite a textbook sentence and still fail to predict which way water moves.

Strong

What determines the direction of net water movement across a semipermeable membrane?

That names the mechanism. You have to produce the gradient and the constraint (the membrane lets water through, not the solute), then check the page.

Comparison

Weak

What are hypotonic and hypertonic solutions?

Two definitions. The comparison that matters — what happens to the cell — can stay untested.

Strong

How does an animal cell respond in a hypotonic solution versus a hypertonic one, and why is the water movement different?

You have to distinguish the two conditions and use the mechanism, not list vocabulary.

Procedure and application

Weak

Explain price ceilings.

Too broad. You will reread the section and feel done.

Weak (yes/no)

Do price ceilings cause shortages?

A correct “yes” can hide that you mean only a binding ceiling, or that you cannot say what happens to quantity demanded versus supplied.

Strong

A city sets a rent ceiling below the current equilibrium rent. What happens to quantity demanded, quantity supplied, and the resulting gap?

Apply / what-if

Demand for rental units then rises. The ceiling does not change. What happens to the shortage?

That is the kind of thinking you need if the later use is a problem set, not a glossary.

When it fails

When does a price ceiling fail to bind, and what does the market look like then?

A term you actually need

Weak

What is tonicity?

If the page’s job is using tonicity to predict water movement, a definition prompt trains the wrong thing.

Honest term prompt

What does tonicity compare between a solution and a cell?

Use that when you truly need the word. If you only need the prediction, skip the term card and keep the hypotonic/hypertonic comparison.

Recognition traps

Weak (answer in the stem)

Tonicity is the relative concentration of a solution compared with a cell. What is tonicity?

You have already been given the sentence. There is nothing left to retrieve.

Weak (multiple-choice giveaway)

A binding rent ceiling causes a shortage, a surplus, or neither?

You can pick a familiar word. You never have to produce the quantities.

Strong

Predict what happens to an animal cell in a hypotonic solution, and name what drives the water movement.

Or, for the ceiling: produce quantity demanded, quantity supplied, and the gap — without listing the options in the prompt.

Question types that match the knowledge

Practice the kind of thinking you need later. Definition questions mainly train definitions. If you will need to compare, apply, or explain, write those prompts too.

Why-questions that make you explain a fact in your own words can help, especially when you already know enough to generate a sensible explanation. They are not a substitute for later closed-book retrieval. Reviews of study techniques treat elaborative interrogation as a separate move from practice testing. Do not replace “try to retrieve, then check” with an open-notes why-session and call it done.

Do not force a template. Skip the why-question if the honest target is a term or a value. Skip the comparison if the page has one idea. Skip the apply prompt if you have never encoded an example — encode first, then retrieve.

Retrieval practice has also been shown to help on later questions that require comprehension and inference, not only word-for-word recall. That is an argument for including why / how / compare / apply prompts when that is what you will need — not an argument for never writing a definition.

How many questions — and how atomic they should be

For a notes page, a few prompts that target what you care about are usually more usable than a question for every sentence. That is a workflow claim for page review, not a finding that more retrieval is harmful.

Three grains:

Topic cue. You see Price ceilings and reconstruct. Almost no setup. Easy to retrieve the wrong slice, or to skip the part you always skip.

A few page questions. The note stays intact. A handful of prompts can cover the mechanism, the comparison, and the application. This is the usual middle for a conceptual page.

Atomic cards. Each card asks for one small, easily checked piece. SuperMemo’s minimum-information principle — widely followed by Anki users — is a strong design for large atomic decks. It is not the only way to cue recall from a note, and it is not Yuhimo’s product rule.

Anki-style atomic cards are the better tool when exact recall of many small items is the job: vocabulary, isolated facts, formulas you must emit, language forms. Drill those in a deck. Keep the conceptual note for relations that fall apart when split too far.

When to return to the page is spaced repetition for notes, not a question-count problem. You can also space reviews without converting the page into cards.

Cloze, short answer, and open prompts

Fill-in-the-blank cues a specific missing piece inside leftover context. Open questions ask you to produce more of the idea yourself.

Use cloze when the target is a precise term, value, or wording you truly need.

Cloze-shaped (honest term)

Water moves by osmosis toward the side with ______ solute concentration, assuming a membrane permeable only to water.

Use open why / how / compare / apply prompts when you need to explain, distinguish, or use the idea.

Open

Why does net water movement stop when two solutions reach equal water potential?

Cloze is real retrieval of the missing piece. It is a different shape of cue, not a fake form of active recall. Anki provides a special cloze note type. Yuhimo does not provide a cloze card type. If you need mass cloze drilling, use a deck. If you are cueing a notes page, write the open prompt, or treat a short exact-answer question as a term prompt you check against the page.

Do not smuggle the answer into leftover context (“Osmosis is the movement of _____ from high to low water potential”) if those leftovers already tell you the sentence. Then it is a reread with a hole.

Writing your own questions vs using good ones

Writing questions can be useful when you have no good prompts yet, but the learning comes mainly from later retrieval of the right ideas. Untrained self-generated questions are not reliably better than good provided questions, and they can miss what you actually need to remember.

In experiments with study passages, generating questions and then answering them beat rereading and looked similar to answering questions the researchers provided. Generation is not automatically deeper processing.

A later comparison of self-written and provided questions found that answering your own questions was not more beneficial than answering provided ones, and student questions often missed material that later tests asked.

Write prompts when the page has none, or when only you know which slice matters. Prefer answering a sharp prompt over spending the session authoring a bank. If a textbook, lecturer, or friend already wrote a good question, use it.

Editing questions, including AI drafts

Bad questions are cheap to delete. Keep a noisy set and review becomes rereading a quiz.

Edit when the prompt is too broad, too obvious, too many ideas at once, or aimed at a detail you do not care about. Split “explain osmosis, tonicity, and plasmolysis” into the prompts you actually need. Drop the rest.

If a prompt is unanswerable, decide whether the page is missing the idea. Add it to the note, or delete the question. Do not keep a prompt that can only be answered from a lecture you no longer have.

AI drafts are starting prompts, not a finished deck. Generated questions can be too generic, overlapping, or aimed at a sentence that was only context. Review, rewrite, or delete them before they become habit. The draft that asks “What is a price?” on a ceiling page is not retrieval practice. It is clutter.

You do not need a prompt for every paragraph a generator notices. Keep the ones that match your later use.

Common mistakes

Giveaways. The question contains the key term, the direction, or the conclusion.

Yes/no and multiple-choice stems. “Does a hypotonic solution make an animal cell swell?” trains a nod. Ask for the prediction and the why.

Topic dumps. “Explain supply and demand.” Name the slice.

Definition-only pages. If you will have to compare or apply, those prompts have to exist too.

One question per sentence. Usable review dies under maintenance. Choose what is worth retrieving; leave the rest as reference.

Never checking. Produce, then open the note. Unchecked fluency is a guess.

Authoring instead of retrieving. A session spent writing a pile of prompts and answering none is not retrieval practice.

Treating questions as mandatory. If the heading is enough, leave it.

How Yuhimo fits this workflow

Yuhimo keeps the page as the learning unit. Questions are optional retrieval prompts on that page. You can use them in review as cues, or reconstruct from the title.

After you try to recall, you rate how well you remembered the page — Forgot, Barely, Remembered, or Easily. The app does not mark each answer right or wrong.

Write prompts by hand when you know the target. Or generate AI drafts and edit them. The page stays the source; the prompts stay optional.