Résumé : Alcohol-dependent patients have difficulty recognizing social cues such as emotional facial expressions, prosody, and postures. However, most researchers describing these difficulties rely on labeling tasks. It therefore remains difficult to disentangle genuine emotion-decoding problems from emotion-labeling impairments. In the present study, 35 recently detoxified alcohol-dependent patients were compared with 35 matched controls on four emotion-pairing tasks to explore the distinction between labeling and perceptual abilities. First, 2 tasks were used to assess emotion-labeling ability (labeling task) and working memory (necessary to process emotional stimuli; control matching task). Next, 2 experimental pairing tasks were used to explore unimodal, Face-face or voice-voice) and cross-modal, Face-voice or voice-face) matching abilities in the absence of any labeling requirement. Patients had difficulty accurately processing voices in unimodal tasks and correctly matching emotional stimuli in cross-modal tasks. Specifically, they did not correctly identify neutral stimuli in unimodal or cross-modal tasks and did not correctly identify fear in cross-modal tasks. Reaction times were also slower in these patients. However, accuracy and reaction time (RT) differences between patients and controls were accounted for by including anxiety and depression scores as covariates in the model. These results suggest that emotion-decoding difficulties observed in recently detoxified alcohol-dependent patients are not due to a specific emotion-labeling impairment, but rather involve perceptual difficulties or later integrative processing steps in the brain. Future studies should directly compare depressed or nondepressed alcohol-dependent patients with depressive patients to disentangle the influences of these highly comorbid disorders on nonverbal language perception.