1. The directions of waggling dances of bees, trained for a feeding place in a given direction, are recorded on a horizontal comb. The comb is covered by a white tent with a central circular opening towards the zenith. 2. The orientation of the dances is seriously disturbed under spectral filters, absorb- ing UV. The orientation of the dances is much less disturbed under greyfilters, which decrease the level of stimulation of all types of receptors. These results show that the polarization perception is based on the integration of the contributions of at least two different types of receptors. 3. By selective blocking of polarized light in certain areas of the sky by means of polaroidfilter configurations (Fig. 3c), the importance of spatial differences in the stimulation ratio between the blue and UV receptors is shown. The blue receptor contribution appeared to be predominant in the solar vertical, whereas the con- tribution of the UV receptors is predominant in a celestial band perpendicular to the solar vertical in the zenith. 4. The colour pattern, which according to the colourpattern hypothesis (VAN DER GLAS, 1975), bees would perceive as a result of the polarization of the skylight, is simulated by an artificial colourpattern, i.e. a configuration of colour filter segments (Fig. 8, 10 and 12) in the colours bluegreen and purple, combined with a depolarizer for the incident skylight. The "stable" dances (see methods) under this colourpattern are orientated in the predicted directions. Control experiments with greyfilter con- trasts in a similar configuration, in which the stable dances of the bees are mainly desorientated, proved that the colours in the pattern are essential. It seems that the direction of the bluegreen segments is used by the bees as a beacon for the menotactic orientation in the dances. 5. If half of the field of view around the zenith of the normal free sky is screened, the orientation is better when the dividing-line of the free half-moon shaped field of view is perpendicular to the solar vertical than when this dividing-line is directed along the solar vertical. With similar apertures for artificial colourpatterns (Fig. 16), the same effects are observed. Consequently the differences in orientation behaviour in both cases is due to differences in screening of essential features of the colour pattern. Suggestions for possible senso-neural integration of the information of fields of ommatidia are given.