A new general-purpose technique has been developed for making sense of neural networks that are trained to perform natural-language-processing tasks, in which computers attempt to interpret freeform texts written in ordinary, or ‘natural,’ language (as opposed to a structured language, such as a database-query language).