Fig. 4. Haze experiment. (A) 180° fisheye view of the celestial hemisphere on a hazy morning. Compared with clear atmospheric conditions, the degree of polarization across the whole sky is reduced. (B,C) Polarotactic response as a function of the degree of polarization for a uniform stimulus. The effective degree of linear polarization (d) of a medium-sized (25°) zenithal stimulus was reduced from d=100% to 0% by changing the ellipticity of light (see Materials and methods). Test data (d=1% to 53%) are indicated by black, motivation controls (d=100%) by gray, and zero controls (d=0%) by white (24 series of 17 individuals). (B) Survey of results. Relative strength of the polarotactic response (S/S_mot; mean ± s.d.) plotted against the effective degree of linear polarization. (C) Distribution of S-values and (D) modulation of walking direction with stimulus orientation for some of the polarization levels tested (see polarization ellipses to the left). Data in D are normalized and plotted as described in Fig. 3D. Note: a reduction in polarization to d=53% did not impair the polarotactic response. With lower d-levels, the response strength decreased. However, there was a statistically significant orientation to polarized light at least down to a d-level of 7% (P<0.01).