Intoxicated witnesses are common, making it important to understand alcohol's impact on witness accuracy and suggestibility. Participants assigned to an immediate retrieval condition encoded and recalled in one of the three intoxication conditions: sober control, placebo, or intoxicated. Participants in the delayed retrieval condition were assigned to encode in one of the three intoxication conditions, returned a week later, and were assigned to retrieve in one of the three intoxication conditions. Intoxication condition at encoding was fully crossed with intoxication condition at retrieval in the delayed condition. Participants encoded a mock crime video and retrieved via a forced-choice test, with answers already circled (purportedly by a prior participant); half of the precircled responses were incorrect. When recalling after a delay only, intoxication at encoding increased agreement with incorrect suggested answers and decreased accuracy. Results suggest intoxicated witnesses may benefit from being interviewed immediately rather than after a sobering delay.