Insight into the stability of dietary behavior over time is important, because only a single measurement of diet is often available to study the association between eating behavior and the occurrence of chronic diseases many years after baseline data collection. Little is known about changes in dietary patterns over time. The current study examined the (internal) stability and reproducibility of dietary patterns and the transition of individuals between patterns over time from 3 surveys within one study population by using cluster analysis. The dietary intake of participants in the Doetinchem Cohort Study in 6113, 4916, and 4520 adults in 1993-1997, 1998-2002, and 2003-2007, respectively, was measured using a validated food-frequency questionnaire. Stability and reproducibility of dietary patterns were studied by examining the optimal number of clusters per survey by comparing the contribution of food groups to total energy intake within the clusters over time and by studying transitions of individuals between clusters over time. A low-fiber bread pattern and a high-fiber bread pattern were identified in all 3 surveys. Over time, dietary patterns were comparable in terms of foods contributing most to total energy intake, suggesting good reproducibility. Nevertheless, only 41.8% of the participants were consistently assigned to the same dietary pattern for all 3 surveys. This implies that, over time, similar dietary patterns were found at the group level, but that ignoring individual transitions between dietary patterns during follow-up may lead to misclassification of a large proportion of the study population.