Single-blind vs double-blind experiments
By Jude Wallis · Published
In a single-blind experiment, one side does not know who received which treatment: either the subjects, or the people who interact with them and measure the response. In a double-blind experiment, neither side knows. Double-blinding keeps expectations on both sides out of the results.
AP Statistics: Unit 1 (topics 1.13 Experimental Design). Single-blind and double-blind experiments belong to topic 1.13, Experimental Design, the last topic of Unit 1 in the Fall 2026 AP Statistics course. A typical task is to identify which side is blinded from a description of a study and to explain what the blinding prevents.
Single-blind vs double-blind: the short answer
Blinding means keeping the treatment assignment secret from someone in an experiment. Two kinds of people can be kept in the dark. The first is the subjects, the people receiving the treatments. The second is the people who interact with the subjects or measure the response, such as the nurse who hands out pills or the doctor who rates each patient's symptoms.
A single-blind experiment blinds exactly one of those two sides. A double-blind experiment blinds both. The count in the name is a count of sides, so a study with two treatment groups that both do not know their treatment is still single-blind if the evaluators know who got what.
What single-blind means
A single-blind experiment hides the assignment from one side only. There are two versions.
In the common version, the subjects do not know which treatment they received, but the researchers who work with them do. Subjects who know they got the real drug may expect to improve and report feeling better for that reason alone. Hiding the assignment from them removes that expectation from the comparison. This usually requires a placebo, an inactive treatment that looks the same as the real one.
In the other version, the subjects know their treatment but the person measuring the response does not. This is the usual choice when the treatment cannot be hidden. A subject knows whether she did eight weeks of yoga or eight weeks of stretching, but the physical therapist who scores her flexibility at the end does not need to know. That is still single-blind, because one side is blinded and the other is not.
What double-blind means
A double-blind experiment hides the assignment from the subjects and from everyone who interacts with them or measures the response. Somebody still holds the key, usually a person who never meets a subject, and the key is opened after the data are recorded.
The second blind matters because evaluators have expectations too. A doctor who knows a patient got the new drug may read a borderline result as an improvement, ask follow-up questions differently, or treat that patient with more attention. None of that has to be deliberate. When the evaluator cannot tell the groups apart, any lean in how responses are scored falls on both groups equally and drops out of the comparison.
Blinding is separate from random assignment. Random assignment decides who gets which treatment. Blinding decides who knows about it afterward. A well-designed experiment uses both when it can.
The differences side by side
| Feature | Single-blind | Double-blind |
|---|---|---|
| Who does not know the assignment | One side: subjects, or those who interact with them and measure the response | Both sides |
| Guards against subjects' expectations | Only if the subjects are the blinded side | Yes |
| Guards against evaluators' expectations | Only if the evaluators are the blinded side | Yes |
| Needs a placebo or lookalike treatment | Only when the subjects are blinded | Yes |
| Typical use | Treatment is obvious to the subject, such as exercise or a diet | Drug trials and taste tests where treatments can be made to look alike |
The table reduces to one question asked twice: do the subjects know, and do the people measuring them know. One no is single-blind. Two is double-blind.
When to use which
Use a double-blind design whenever both sides can be kept from knowing. Pills that look identical, drinks served in unmarked cups, and coded bottles make this cheap, and it protects the response and the measurement at once.
Use a single-blind design when one side cannot be blinded. Subjects cannot be unaware of a diet, a training program, or a teaching method they are living through. In that case blind the person who measures the response. The reverse case also occurs: when two versions of an operation are compared, the surgeon must know which one to perform, but the patient can be kept unaware of which was done.
Blinding matters most when the response is subjective, such as a pain rating or a judged score, because expectation has room to act. It matters less when the response is read off an instrument that leaves no judgment to the reader. The full list of design choices is in how to design an experiment.
The classic mix-up and how to avoid it
The most common mistake is counting groups and calling that a count of blinds. "Neither the treatment group nor the placebo group knew which pill they had, so the study was double-blind" describes one blind. Both groups are subjects. The second blind has to cover the people who deal with the subjects or record the response.
A second mistake is treating a placebo as proof of double-blinding. A placebo is something a group receives. Blinding is about who knows. A placebo-controlled study in which the nurse reads the labels is single-blind. The placebo vs control group page separates those two ideas.
To get it right on a free-response question, name each side and say what it knows. Write that the subjects did not know which treatment they received, and that the people who measured the response did not know either, then say what each blind prevents in the context of the study.
Classifying three studies
Classify each study as not blinded, single-blind, or double-blind. (a) Volunteers take either a caffeine pill or an identical sugar pill. The researcher who hands out the pills and times each volunteer's reaction test has a list showing who got caffeine. (b) Students are randomly assigned to a new tutoring program or to regular study hall. A teacher who does not know the assignments grades all the final essays. (c) Tasters are given cola in cups marked only with a code. The server who pours and records each rating does not know which code is which brand.
List the two sides for every study: the subjects, and the people who interact with them or measure the response.
Study (a): the volunteers cannot tell the pills apart, so the subjects are blinded. The researcher who gives the pills and records the times has the list, so that side is not blinded. One side blinded means single-blind.
Study (b): the students know whether they are in tutoring or study hall, so the subjects are not blinded. The teacher grading the essays does not know the assignments, so the measuring side is blinded. One side blinded means single-blind.
Study (c): the tasters see only a code, so the subjects are blinded. The server who pours and records also does not know the brands, so the measuring side is blinded. Both sides blinded means double-blind.
Check study (a) against the common error: the placebo does not make it double-blind, because the person recording the response knows the assignment.
Study (a) is single-blind (subjects blinded, researcher not). Study (b) is single-blind (grader blinded, subjects not). Study (c) is double-blind. Studies (a) and (b) show the two different ways a study can be single-blind.
How an unblinded evaluator changes a result
In a pain trial, 100 patients are randomly assigned, 50 to a new drug and 50 to a placebo. When a blinded evaluator scores pain relief on a 0 to 10 scale, the drug group averages 3.2 and the placebo group averages 2.4. Suppose instead the evaluator knows the assignments and, without intending to, scores patients in the drug group 0.3 points higher on average. Find the difference the study would report and how much of it comes from the evaluator.
Find the difference with a blinded evaluator: points.
With the unblinded evaluator, the drug group's recorded mean rises by 0.3: . The placebo group's mean stays at 2.4.
Find the reported difference: points.
Find the overstatement: points.
Express the overstatement as a share of the reported difference: , about 27%.
Express it as a share of the blinded difference: , so the reported effect is 37.5% larger than the effect measured under blinding.
The unblinded study reports a difference of 1.1 points instead of 0.8. The extra 0.3 points, about 27% of the reported difference, comes from the evaluator's scoring. Blinding the evaluator removes it, because the evaluator can no longer score one group differently from the other.
Frequently asked questions
Is a study double-blind if both the treatment group and the control group do not know their treatment?
No. Both of those groups are subjects, so that is one blind. A study is double-blind only when the people who interact with the subjects or measure the response also do not know who received which treatment.
Can every experiment be double-blind?
No. Some treatments are obvious to the person receiving them, such as an exercise program or a diet. Those experiments can still be single-blind by keeping the assignment from whoever measures the response.
Does blinding replace random assignment?
No. Random assignment determines which treatment each subject gets and is what allows a cause-and-effect conclusion. Blinding controls who knows the assignment afterward, so that expectations do not change the responses or how they are recorded. A good experiment uses both.