The integration of artificial movable landmarks (tetrahedral tents of 3.46 m height) into the longdistance course control of bees was investigated in an environment that provided a minimum of natural orientation cues. Bees were trained along a row of four landmarks which were equally spaced in a setup of 300 m in length. The animals were then tested in situations where the position of the landmark line indicated a direction of the feeding site other than the (originally trained) sun-compass direction. This discrepancy was systematically varied. We show that the sun compass is dominant over the landmarks. This preference is negligible if the deviation between the two indicated directions is small, but it increases as the angle of deviation becomes larger. In contrast, if the entire sky is overcast by clouds, more bees will follow the direction indicated by the artificial landmarks than under sunny conditions. However, as the deviation between the trained direction and the landmark direction increases, more individuals will land at the training site.