Age–period–cohort analysis refers to the interpretation of temporal trends in disease incidence or mortality rates in terms of the three time scales, age, calendar date (period), and year of birth (cohort). We present graphical and analytic methods for understanding trends in disease rates. A fundamental issue of identifiability arises in some models because of a linear constraint on these time scales: year of birth plus age equals the current calendar year. We review parametric and nonparametric approaches to estimate identifiable aspects of the data, such as the curvature of temporal trends.