Student success in a physics degree has been shown to depend on more than just performance in course assessment: important additional factors include student attitudes and beliefs about their subject. We have used an instrument (CLASS) that measures how student epistemologies evolve over the course of their undergraduate degree. Our previous work has sampled a cross-section of students at all levels across the physics undergraduate programme at Edinburgh in a single academic year, and found that student attitudes and beliefs remain essentially static. Here, we present fully longitudinal data collected over the past three years, where we track the evolution of the attitudes and beliefs of one group of students. We find broadly similar results: attitudes and beliefs remain surprisingly consistent over time. This suggests that a 'cross-sectional' or 'pseudo-longitudinal' study (collecting snapshot data in one year) is a valid methodology, rather than necessarily having to wait several years to accumulate truly longitudinal data.