Q&A: Understanding the composition of behavior

Abstract

Understanding the brain requires understanding behavior. New machine vision and learning techniques are poised to revolutionize our ability to analyze behaviors exhibited by animals in the laboratory. Here we describe one such method, Motion Sequencing (MoSeq), which combines three-dimensional (3D) imaging with unsupervised machine learning techniques to identify the syllables and grammar that comprise mouse body language. This Q&A situates MoSeq within the array of novel methods currently being developed for behavioral analysis, enumerates its relative strengths and weaknesses, and describes its future trajectory.