During each rotation the plane of polarization was rotated by 360° (0° defined as the E-vector parallel to the longitudinal body axis of the animal). The LEDs for unpolarized stimulation (ultraviolet [365 nm], green [520 nm],

blue [460 nm]) were attached to the rotation stage via radial arms extending from the zenith, so that each LED pointed toward the animal (angular size: 3°). With every rotation, each LED passed through all possible azimuth directions at constant elevation. Photon flux rate was equal for all unpolarized stimuli. Over the course of experiments, rotation velocity was either set to 30°/s or 60°/s, and both clockwise as well as counterclockwise rotations were applied in direct sequence. For blocking light to the dorsal region of the compound eye, a small Hydroxychloroquine concentration piece of black tape was positioned directly in front of the eye. Identical stimuli were applied before, during, and after

the shielding. For eliminating polarization during control experiments with zenithal unpolarized light, a diffuser was inserted into the light path. Residual polarization during 360° rotations of the polarizer/diffuser was found to be below 5%. Intensity of polarized light was adjusted to match the unpolarized light intensity resulting from insertion of the diffuser. Neuronal responses to rotations of the polarizer as well as to azimuthal rotations of unpolarized light spots were analyzed with custom designed scripts in Spike2 software. Each spike occurring during a rotation was assigned its corresponding angle (either E-Vector or azimuth). These angles were tested for significant CB-839 mouse difference from randomness using the Rayleigh test for axial (E-vector angles) or circular data (azimuth angles). If activity during rotations was significantly different from randomness, the resulting mean angle was defined as the preferred E-vector secondly or azimuth angle of the examined neuron. For circular plots,

spiking activity during rotations was calculated for 10° bins, averaged over all rotations within each neuron, and plotted against E-vector orientation or azimuth angle, respectively. The response amplitude (R) was calculated as described in Heinze et al. (2009). In brief, R is a measure for the summed absolute deviation from mean activity during stimulus application. Thus, the higher the value of R, the stronger is the response to the stimulus. R values for periods without stimulation were obtained to calculate background variability. Statistical comparison between shielded and unshielded stimulus conditions were performed by analyzing R values for each of the conditions. After R values were normalized to the unshielded response value, a paired t test was then used to compare the shielded response to background variability while a one-sample t test against a hypothetical mean of 1 was performed to compare the shielded response against the normalized unshielded response.