In this article the consistency of forest damage assessments on conifer trees is analysed, by using different methods in estimating the accuracy in assessments of defoliation and discolouration. The data originate from control surveys in the Swedish national forest damage survey, as well as from national and international training courses. Standard deviation of differences in the assessment of defoliation on single trees is found to be about 10% units for Norway spruce and 8.5% for Scots pine. Problems in correctly assessing damaged Scots pine trees and discoloured Norway spruce trees are revealed by measures of agreement (Kappa statistic). Results from several years of national training courses indicate that, on an average, the observer teams do not significantly differ from a national standard, but significant differences between observer teams are found. The presented estimates indicate a substantial within observer error compared to the between observer error. The results indicate that the long-term development of forest damage, rather than short-term fluctuations, is the most important information from these kinds of inventories.