This study evaluated the potential of an autumn LANDSAT MSS diazo-enhanced composite which consisted of one yellow copy of band4, three magenta copies of band 5 and one cyan copy of band 7, to map an area of small scattered forestlands. The standard colour composite of the same scene was also used for comparative purposes. The best diazo composite was made by high-contrast films. The diazo technique used is discussed briefly. Interpretation errors were characterized as commission errors and omission errors which were further classified by size and into errors of boundary placement and errors of identification. Furthermore, they were classified according to land-cover/use and stocking-level stand-size classification systems. The three types of accuracy calculated (classification, interpretation and mapping) indicated, first, the potential of mapping forestlands on a LANDSAT image and, second, that the diazo composite improved the interpretation performance over the standard composite. On both images most of the interpretation errors were classified as boundary errors. The most frequently misclassified areas were agriculture, treed bog and brushland and also lowland hardwoods, mixed hardwoods and oak-hickory stands. The stocking level of the stands affected the interpretation more than the stand size, Over 50 per cent of the interpretation errors on both images were less than 2-0 ha in size. However, the distribution of errors skewed more towards smaller size classes in the case of the diazo composite.