Contemporary scholars describe mathematical modeling as a transformation of a real-world problem to a mathematical problem and back again. This paper treats a critical issue in the modeling process: how modelers determine if the transformation from the real world to mathematics was carried out well. I present an empirically derived typology of validating activities explaining how validating functions to ensure a mathematical model will yield a reasonably accurate prediction. The typology arose from analysis of four engineering undergraduates' production of 276 instances of validating. The nuances of validating suggest that creating and maintaining relationships between reality and mathematics is more complex than a transformation and that we should afford a more prominent role to validation in the modeling process.