Detection of acute 3,4-methylenedioxymethamphetamine (MDMA) effects across protocols using automated natural language processing

This study used automated speech analysis to detect objective markers of mental states after MDMA and oxytocin administration. The analysis of speech features (acoustic, semantic, and psycholinguistic) from 31 healthy adults across four sessions (MDMA, oxytocin, placebo) showed promising classification results, achieving cross-validated accuracies up to 87% in training/validation and 92% in independent datasets. Oxytocin effects were mainly reflected in acoustic features related to emotion and prosody, while MDMA effects manifested across multiple speech domains. The experimental task influenced speech responses. The findings suggest speech analysis may offer an objective measurement of intoxication-related mental states.

Carla Agurto, Guillermo A. Cecchi, Raquel Norel, Rachel Ostrand, Matthew Kirkpatrick, Matthew J. Baggott, Margaret C. Wardle, Harriet de Wit, Gillinder Bedi