The consistency of visual assessment of tree defoliation, which represents the most widely used indicator for tree condition, has frequently been in the focus of scientific criticism. Thus, the objective of the present study was to examine the consistency of the defoliation data from the annual national training courses for the forest condition survey in Germany from 1992 to 2012. Defoliation assessments were carried out in stands of beech (Fagus sylvatica), oak (Quercus robur and Quercus petraea), Norway spruce (Picea abies), and pine (Pinus sylvestris). Among the observer teams, the absolute deviation from the observer mean of all years was ±4.4 % defoliation and the standard deviation of defoliation was ±5.5 %. On average, 94 % of the assessments were located within the ±10 % interval of deviation from the mean. Tree species-specific differences did not occur when all years were considered. A trend towards increasing consistency was observed from 1992 to 2012, in particular for oak and spruce. The deviation of defoliation assessments depended non-linearly on the level of defoliation with highest deviations at intermediate defoliations. In spite of high correlations and agreements among observers, systematic errors were determined in nearly every year. However, within-observer variances were higher than between-observer variances. The present study applied a three-way evaluation approach for the assessment of consistency and demonstrated that the visual defoliation assessment at the national training courses in general produced consistent data within Germany from 1992 to 2012.