01 · The Question
What Does “Double-Blind” Actually Tell You?
A paper describes a study as double-blind. Who did not know the treatment assignments?
The participants and clinicians? Participants and outcome assessors? Investigators and data analysts?
You might expect a standard answer. Unfortunately, researchers have not always used single-, double-, and triple-blind consistently. Cochrane specifically cautions that broad labels such as “double blind” make it difficult to determine who was actually blinded, while current SPIRIT and CONSORT guidance favors explicitly identifying the people who were unaware of intervention assignments.
The familiar labels can still provide shorthand, but shorthand becomes a problem when readers have to guess what it means.
03 · What You Need to Know
Counting the Number of Blinded Parties Hides the Information That Matters
Blinding, sometimes called masking, is intended to prevent knowledge of intervention assignment from influencing aspects of a study that could bias its results.
The methodological question is therefore not simply how many parties were blinded. It is which parties were blinded and what their knowledge could otherwise influence.
What does single-blind usually mean?
Single-blind commonly indicates that one relevant category of people is unaware of treatment assignment. Frequently this means the participants, but that interpretation should not be assumed universally.
For example, participants might know which intervention they received while outcome assessors remain unaware of condition membership. Describing such a study merely as “single-blind” does not tell the reader which of these arrangements applies.
What does double-blind usually mean?
Double-blind traditionally indicates that two relevant parties are unaware of assignment. In many clinical contexts, this has often referred to participants and investigators or care providers.
Yet “investigator” is itself ambiguous. It might refer to the person administering treatment, recruiting participants, collecting data, assessing outcomes, supervising the study, or analyzing results.
Cochrane therefore cautions that the term “double blind” makes it difficult to know who was blinded and notes that the term has been used inconsistently.
Current SPIRIT guidance is more explicit: trial protocols should state who will be blinded to intervention groups, including at minimum the blinding status of participants, care providers, and outcome assessors. Data analysts may also be relevant.
What does triple-blind mean?
Triple-blind is often used when a third category, sometimes the data analyst, is also unaware of group identity. Older methodological glossaries, for example, have described triple blinding as extending blinding to the statistician in addition to participants and investigators.
But there is no reason to assume every author uses the term in precisely that way.
If a paper says “triple-blind” but never identifies the three blinded parties, the reader still lacks the information needed to evaluate what potential biases were addressed.
| Label |
Common interpretation |
Main problem |
| Single-blind |
One relevant party is unaware of assignment, often participants |
The label may not identify which party was blinded. |
| Double-blind |
Two relevant parties are unaware, often participants and providers or investigators |
Definitions vary, and “investigator” can refer to several different study roles. |
| Triple-blind |
A third party, sometimes the data analyst, is also unaware |
The identity of the third party is not standardized across all usage. |
| Explicit reporting |
Each relevant party's blinding status is stated separately |
Requires slightly more words but communicates the design much more clearly. |
Who can potentially be blinded?
A study involves more roles than the traditional labels imply.
Depending on the design, relevant parties can include:
- participants;
- care providers or people delivering the intervention;
- data collectors;
- outcome assessors;
- adjudicators;
- data analysts.
Not every role exists separately in every study. The same person may occupy several roles. A participant completing a self-reported outcome, for example, is effectively also involved in outcome assessment for that measure.
This is why the more useful question is who needs to be blinded and what their knowledge could influence.
Different people are blinded for different reasons
Participant blinding may reduce changes in behavior, expectations, adherence, additional treatment seeking, or self-report that arise from knowing the assigned condition.
Blinding intervention providers may reduce differential care or co-interventions beyond those required by the protocol.
Blinding outcome assessors can reduce the possibility that knowledge of assignment influences judgment, classification, or measurement.
Blinding data analysts may limit some opportunities for discretionary analytical choices to be influenced by knowing which coded group received the experimental intervention.
Calling all of these simply “blinding” obscures the different bias mechanisms involved.
Blinding should be described outcome by outcome when necessary
A person may be blinded for one outcome but not another.
Suppose a trial evaluates both self-reported pain and an imaging outcome. Participants know their treatment assignment, so the pain report is necessarily provided by an unblinded participant. The imaging scans, however, could be assessed centrally by specialists who do not know the assigned treatment.
It would be misleading to reduce this situation to one global label.
Cochrane's risk-of-bias framework emphasizes that the consequences of awareness can differ across outcomes. The relevant question is whether knowledge of intervention assignment could influence the particular outcome being assessed.
How blinding was achieved matters as much as saying that it occurred
CONSORT 2025 asks authors to describe how blinding was achieved and, where relevant, how similar the interventions were.
This is important because nominally blinded conditions may still be distinguishable. Placebo tablets might differ in taste, shape, smell, packaging, or side effects. Documents sent to supposedly blinded assessors may contain identifying information. Participants may reveal their condition during an assessment.
A useful methods section therefore explains the mechanism used to achieve and maintain blinding rather than merely attaching a label to the study.
Blinding can be compromised during the study
A participant may discover which intervention they received. An assessor may accidentally see the allocation. An emergency may require treatment assignment to be revealed for safety reasons.
CONSORT recommends reporting known compromises to blinding and procedures for emergency unblinding where appropriate.
The important distinction is between intending to blind and actually maintaining blinding during study conduct.
Watch Out
Do not assume that a study remained blinded merely because its protocol described it as blinded. Evaluate what procedures were used, whether assignments could be inferred, and whether known unblinding occurred.
Testing whether participants guessed their assignment is not a simple solution
Researchers sometimes ask participants or providers at the end of a study which condition they think they received and treat correct guesses as evidence that blinding failed.
That interpretation can be difficult.
If an intervention genuinely works or produces recognizable effects, participants may infer their assignment from their outcomes rather than because the blinding procedure was defective. CONSORT 2025 therefore does not require a formal test of blinding, while recommending transparent reporting if investigators choose to conduct one.
Allocation concealment should not be counted as another level of blinding
Allocation concealment prevents people involved in enrollment or assignment from knowing upcoming allocations before participants are assigned. Blinding generally concerns awareness after assignment.
They protect different stages of the study.
A trial could use excellent allocation concealment and then be completely open after treatment is assigned. Conversely, participants and assessors could be blinded after assignment even though the original allocation procedure was poorly concealed.
Calling a study “triple-blind” does not tell you whether allocation concealment was adequate.
An open-label study is not automatically a poor study
Some interventions cannot realistically be concealed. Teachers know which curriculum they deliver. Participants know whether they attended a workshop. Surgeons know which operation they performed.
In such circumstances, transparent reporting is preferable to creative labeling.
Researchers can still blind outcome assessors where feasible, standardize procedures, use outcomes less susceptible to discretionary assessment, monitor protocol deviations, and address other plausible bias pathways.
The absence of a double-blind label is therefore not a verdict on study quality.
