The rise of sensor-derived data, from air quality and traffic to motion and health information, provide new contextual opportunities. Just as meaning and context have supported improvements in speech recognition, so too will these new data aid in predictions and decision-making. Sensor data, we argue, should be viewed as part of the discourse and pragmatic context for natural language. Advances in chemical sensing add new capabilities, but sensor data must not be interpreted in isolation. Instead, it forms part of the knowledge environment for an artificial intelligence approach that incorporates natural language understanding. In human discourse, context influences interpretation. The ability to collect contextual information is skyrocketing due to advances in sensor capability. Understanding what people say without knowing that they said it in the context of chemical attack, or while running, or in 110° heat, risks missing critical data. It makes sense to exploit all of that information to support human language technology.