Accuracy assessment of land-use and land-cover classifications obtained from remotely-sensed data is recognized as a critical step in evaluating thematic maps. In this Letter, thematic accuracy assessment is formulated in terms of a finite population sampling problem. Several examples are given to illustrate how this sampling perspective is applied to problems in accuracy assessment. Finite population sampling methods have the flexibility to provide estimators and variance estimators for a wide variety of sampling designs and parameters of interest in accuracy assessment.