Differences in vegetation cover estimation by field biologists of the 8th National Forest Inventory were tested. Eleven observers estimated the canopy coverages of six forest taxa in 25 sample plots, located in one stand. The experiment was arranged after the field work. The coverage of Vaccinium vitis-idaea and the ground layer appeared to be the most difficult to estimate. the mean of the highest estimator was about double that of the lowest one. The least abundant species and the sample plots with the smallest coverages had the largest estimation errors. The most important compositional gradient of the data was natural, even though the test was made in a homogeneous area. However, the effect of the observer could be recognized. The differences between observers could be caused by the differences both in visual estimation level and in placing the sampling frame. The results suggest that tests should always be made when several observers are used in vegetation surveys. If calibration is used, it should be made separately for each species.