Which methodological feature best reduces observer bias in a clinical trial?

Double blinding ensures that neither the participants nor the investigators (researchers assessing outcomes) know which treatment group each participant is in. This eliminates performance bias (participants not changing behavior based on group assignment) and detection bias (investigators not unconsciously recording outcomes differently based on group). Single blinding (only participants blinded) still allows investigator bias. Randomization addresses selection bias. Placebo controls prevent participants from knowing if they received the active treatment but do not blind the investigator. Double blinding is the strongest methodological protection against bias in a clinical trial.