TIMOTHY H. GOLDSMITH 789 



serving color vision. In ilic vertebrate eye, the shift in wavelength 

 ol maximum sensitivity to longer wavelengths with increasing in- 

 tensity of illumination — the Purkinje shift — is one manifestation of 

 this duplicity. 



Hanstrom (25) described retinula cells with long and short axons 

 and suggested, with no supporting physiological evidence, that these 

 two types of cells are analogous to rods and cones. Fingerman and 

 Brown (14, 15) reported a Purkinje shift in the "spectral response 

 curve" of DrosopJiila. These authors used narrow^ bands of wave- 

 lengths as test lights, ^vitli, however, no regard for their relative in- 

 tensities. The "spectral sensitivity" curves of Fingerman and Brown 

 are therefore in reality some composite function of the emission 

 spectrum of their lamp and the action spectrum for phototaxis. More 

 to the point, the consistent decrease in the wavelength of maximum 

 "sensitivity" with decreasing intensity of illumination claimed by these 

 authors is not supported h\ their data. With relative intensities of 

 1, 1/10, 1/100, and 1/1000 their curves peaked at 520, 540, 478, and 

 490-502 m/i,, respectively. 



There are several other observations which bear on this problem. 

 W^eiss (62) reported that although at low intensities 365 mfx was 

 more effective for phototaxis than the broad band of wavelengths be- 

 tween 470 and 528 m/x, at high intensities the order of attractiveness 

 was reversed. This report stands without confirmation and warrants 

 further investigation. 



Analysis of electroretinograms recorded from the compound eye 

 of the worker honeybee provides no evidence of a shift in spectral 

 sensitivity analogous to the Purkinje shift in the first three logarith- 

 mic units above the threshold of the retinal action potential (20) . 

 That is, curves of magnitude of response vs. log intensity are virtually 

 parallel at different wavelengths. Similar results have been reported 

 for the cockroach (57) . On the other hand, Antrum (1) reported 

 that in CalUphora similar response-energy curves are not parallel and 

 that the slopes are shallower at shorter wavelengths; therefore, the 

 wavelength of maximum sensitivity should depend on the magnitude 

 of the ERG's on which the spectral sensitivity curve is based. As was 

 discussed above, the maximum was found by Autrum at 540 m^. 

 at high intensities, and the spectral sensitivity curves appeared to 

 broaden and the peaks to shift to shorter wavelengths as the threshold 

 of the ERG was approached. Prompted by these data, Schneider (49, 

 50) used a behavioral response to a striped drum illuminated with 

 intensities which were presumably below the threshold of the retinal 



