Driver attention is often assessed via glance behaviour, typically by measuring glances away from the forward roadway or by directly measuring glances to non-driving related targets. This approach can be used to detect distracting events, but it does not check whether all situationally relevant targets are sampled. Here, we evaluate the usefulness of the MiRA-theory as basis for attention assessment. A field study was conducted with 23 participants driving an instrumented vehicle on an urban route. The participants wore a head-mounted eye tracker. Data reduction included the identification of target areas that needed to be sampled, whether they were sampled or not, and whether relevant or irrelevant other traffic was present. Additionally, a gaze-by-gaze analysis identified gaze direction, purpose, and target. As predicted, drivers sampled all required target areas that necessitated a glance away from forward. Target areas roughly in the forward direction, like zebra crossings, were probably sampled with peripheral vision, but this could not be reliably confirmed with the equipment used. The glance direction distribution was found to correspond well to the a- priori-defined requirements. A higher number of parallel requirements induced a larger share of glances with the purpose to check for traffic. Relevant traffic was monitored more than irrelevant traffic. A higher number of parallel requirements was associated with reduced spare visual capacity. Nominal glance target identification was less linked to the requirements. We therefore recommend that “traditional” glance-based attention assessment should be complemented with a purpose-based glance assessment protocol coupled with situation dependent pre-defined requirements.