Simple algorithmic methods for integrating remotely sensed data with existing cartographic databases do not provide satisfactory results. One of the main difficulties is in reconciling spatial differences between the two data sources. The differences are due to such diverse factors as temporal variations, spatial errors in the map data, and topographic effects in the remotely sensed data. To investigate the extent of this data-integration problem, a map--image congruency evaluation (MICE) knowledge-based system was developed which performs three distinct operations: (i) pre-processing for uniform representation of both the image and cartographic datasets; (ii) spatial reasoning on the data with the MICE system; (iii) presentation of a congruency evaluation map. Results are presented for a forested area in British Columbia, where Landsat multispectral scanner data are integrated with a provincial forest-cover map.