What the Pretesting Effect Actually Is

A pretest is not a cold final exam.

It is a low-stakes attempt made before you learn the answer. Researchers also call it prequestioning, pre-instruction testing, or errorful generation.

The basic sequence is simple:

• 1
  See a question about material you have not studied yet.
• 2
  Try to answer without looking.
• 3
  Study the correct answer or the lesson that contains it.
• 4
  Test the answer again later.

The first answer can be wrong. In many experiments, it usually is.

What matters is that a genuine learning opportunity follows the attempt. A wrong guess with no correction is only a wrong guess. A wrong guess followed by a clear answer can become a marker: this is the thing I missed; this is how the new information differs.

That makes pretesting different from active recall.

Active recall asks you to retrieve something after you have learned it. Pretesting asks you to generate an answer before you know it.

One strengthens access to an existing memory. The other may improve how you encode the lesson that comes next.

You can use both.

The Experiment Where 95% Wrong Still Helped

The cleanest demonstration came from Lindsey Richland, Nate Kornell, and Liche Sean Kao in 2009.

Across five experiments, participants read a scientific passage about vision. Some received extra study time. Others spent part of the same total time trying to answer questions before reading.

In the first experiment, the pretest group answered only 5% of its initial questions correctly.

The researchers removed those successful answers from the later analysis. They wanted to know what happened after failure, not after lucky prior knowledge.

The failed questions still helped.

Participants later remembered more of the targeted information after the pretest-and-study sequence than after extended study. The pattern survived several attempts to explain it away. In later experiments, key sentences were italicized or keywords were bolded for everyone. In another, participants memorized the questions without trying to answer them.

Trying to answer still produced the stronger targeted result.

One experiment delayed the final test for a week. The benefit remained.

The conclusion was not that errors are magic. The act of searching for an answer appears to make the answer easier to learn when it arrives.

That is an important difference.

The Newer Evidence Shows Exactly Where the Benefit Stops

The strongest correction comes from two recent quantitative reviews.

A 2025 multilevel meta-analysis separated two outcomes:

• →repeated questions about information the pretest had targeted; and
• →new questions about other information in the same learning material.

The targeted benefit was Hedges’ g = 0.66.

For new questions about information that had not been prequestioned, the effect was g = 0.01.

A separate 2024 preregistered meta-analysis found almost the same shape: g = 0.54 across 97 targeted effects and g = 0.04 across 91 general effects.

That is not a small footnote.

It means prequestions aim attention. They do not guarantee broad understanding.

Ask about the wrong things and you may get better at remembering the wrong slice of the lesson. Ask too many narrow questions and you can turn a rich topic into a scavenger hunt for isolated answers.

The lesson is not “always pretest.”

It is “choose the target.”

Why a Failed Attempt Can Improve the Next Pass

Researchers have proposed several mechanisms, and no single explanation owns every result.

A prequestion may create a search set. If I ask, “Why does this formula use a logarithm?” my mind activates related ideas before the answer appears.

It may create a prediction error. I expect one answer, encounter another, and the difference becomes informative.

It may increase curiosity. An unanswered question creates tension that makes the answer easier to notice.

It may simply direct attention. Once I know what I am looking for, a paragraph that would have blended into the page now contains a destination.

The 2023 comprehensive review found more than 60 modern peer-reviewed articles on prequestioning and pretesting. The results span word pairs, trivia, text, video, lectures, laboratory tasks, online studies, and classroom work.

The effect is real enough to use.

The mechanism is unsettled enough that I would not build mythology around it.

Practically, you do not need to know which process carried the benefit today. You need to preserve the parts the evidence keeps returning to:

• →a specific question;
• →an honest attempt;
• →access to the correct information;
• →attention to the correction; and
• →another retrieval attempt later.

Remove the learning opportunity and you remove the point.

The Question-First Loop

I use a five-step loop before material that matters.

It works with a chapter, technical document, course lesson, software tutorial, research paper, or instructional video.

1. Write three answerable questions

Do not ask, “What is this chapter about?”

Ask for decisions, causes, comparisons, or procedures.

For a coding lesson:

• →Why is this function asynchronous?
• →What would break if the call were synchronous?
• →Which error should this branch catch?

For a finance lesson:

• →What changes free cash flow but not net income?
• →Why does depreciation get added back?
• →When would a growing company still have negative free cash flow?

For a design lesson:

• →Which spacing change creates grouping?
• →Why does this hierarchy fail on mobile?
• →What should remain visible when the card is cropped?

Three questions are enough to focus the pass without pretending you already know the entire lesson.

2. Commit to a guess

Write one sentence for each answer.

Do not research yet. Do not soften the answer with a cloud of possibilities. Put down the best explanation you can produce from current knowledge.

“I do not know” is allowed, but make one prediction first if the question is safe to guess about.

The point is not confidence.

The point is contrast.

3. Study for the correction

Now open the lesson.

When you reach relevant information, stop. Compare it with the guess.

Mark each answer:

• →right for the right reason;
• →right for the wrong reason;
• →partly right;
• →wrong; or
• →still unanswered.

This is where passive exposure becomes a correction event.

The newest meta-analysis found stronger targeted learning when feedback accompanied the prequestions. In independent study, the lesson, documentation, answer key, or primary source has to supply that correction.

If the source never answers the question, do not invent closure. Mark the gap and find a better source.

4. Explain the difference

A red X is not enough.

Write:

This step prevents you from replacing one sentence with another without changing the model underneath.

It also connects naturally to the self-explanation effect. The pretest exposes the gap. The explanation makes the correction usable.

If you cannot explain why your answer changed, you may have memorized the correction without understanding it.

5. Retrieve it later

Close the source.

Answer the three questions again the next day. Then ask one changed question that requires the same principle in a new situation.

The pretest improved the entrance to the material. It did not prove that the memory will survive.

This last step turns the sequence into a learning loop:

Pretesting and Retrieval Practice Are Not Opponents

A 2021 series of five experiments, with 1,573 participants in total, compared pretesting and posttesting around expository passages.

Both test types improved performance over a no-test control. In that particular series, pretesting produced higher overall scores and appeared to improve how participants processed the passage that followed.

That does not settle a contest between “before” and “after.”

The authors used specific passage-and-test sequences, short delays of five minutes or 48 hours, and a single practice test per passage. They also cautioned against generalizing the result to every use of testing. Other arrangements have favored posttesting.

The better question is: what job needs doing?

Before study, a question creates a target.

After study, retrieval checks whether the answer is available without the lesson.

Use pretesting to sharpen encoding. Use retrieval practice to strengthen later access. Use feedback in both places.

When Pretesting Is the Wrong Tool

Do not guess first when an incorrect attempt could be dangerous.

Medical procedures, legal filings, electrical work, safety protocols, security settings, and irreversible production changes should begin with authoritative guidance. Low-stakes learning does not mean low-consequence action.

Do not pretest every sentence.

If you are brand new to a dense domain, you may not yet know which questions matter. Use headings, objectives, an expert outline, or a short overview to build the first set.

Do not confuse difficulty with quality.

A question can be hard because it is vague, obscure, or disconnected from the goal. That kind of failure is only noise.

And do not let the prequestions narrow the entire lesson.

The quantitative reviews found little general benefit for unasked information. After answering your targets, add one broad prompt:

That single question protects the edges.

What the Evidence Does Not Prove

The research does not identify one ideal number of prequestions for every adult learner or subject.

It does not show that any wrong answer improves memory.

It does not guarantee transfer to material you never asked about.

It does not establish that self-written questions always match expert-written ones.

It does not prove that pretesting alone will produce durable skill months later.

And it does not turn a poor source into good feedback.

The honest claim is smaller and more useful:

Attempting a well-chosen question before study can improve later learning of the information that question targets, especially when a clear correction follows.

That is enough to change how you begin.

Start with the Question, Not the Lesson

The next time you open a course or tutorial, pause before the first explanation.

Write three questions.

Guess.

Then study with something to find.

When the answer appears, do not celebrate being right or punish yourself for being wrong. Compare the two. Explain the change. Return tomorrow without the source.

A pretest is not a verdict on what you know.

It is a way to tell your attention where the learning has to happen.

Today, choose one lesson you were already planning to study and run the Question-First Loop.

Do not ask whether the material looked clear.

Ask whether the correction changed what you can produce.

This article is educational and is not medical advice. Study methods are not a substitute for individualized professional support.