Cellular Mechanisms of the Color Constancy in the Fish Retina

Abstract. We supposed that color-coding double-opponent ganglion cells can display color constancy, for their receptive field (RF) surround might supply the center with the information about the ambient illumination, and its opponent influence on the center might well make necessary scaling for the central signals.
Responses of a certain type of color-coding ganglion cells to flat colored stimuli moving across the RF center were recorded from their axonal terminals in tectum opticum. All scene projected onto the RF was illuminated with light whose spectral composition was changed in a wide range. Under white illumination blue-green stimuli elicited OFF central response while the red ones elicited ON-OFF central response. When the RF surround was colored white, these types of responses remained unchanged while the incident light was varied from red to blue altering the spectral content of the light reflected from the stimuli. When the RF surround was covered with black velvet (i.e., in lack of information about the illumination), one could change the pattern of the central response by varying spectral composition of the illumination. Moreover, covering the surround with colored papers we transformed the pattern of the central response in a predictable manner even under white illumination. This means that the constantly illuminated surround signals the cell about the character of illumination.
Horizontal cells are known to organize the surround of the ganglion cells. Thus, owing to their prolonged polarization during illumination, horizontal cells can realize some color constancy mechanisms just at the very beginning of the visual pathway.
Preliminary reports of these findings have been made elsewhere in Russian¹ and in English².

• Introduction
• Materials and Methods
• Stratification of Tectal Activity
• General Properties of Color-Coding Units
• Color Constancy
• Conclusion

One of the possible mechanisms of the color constancy consists in discounting the illumination. It requires to determine color of the prevailing illumination what visual system can do even if there is no illuminant itself in the visual field. In this case visual system uses some indirect features of the illumination. For example, it is sufficient to have some white object in the visual field to estimate the illumination color.

Mechanisms of the color constancy may be of different levels of complexity, the simplest ones, that do not need memory or learning, may be of retinal origin.

Color-coding opponent ganglion cells of the fish retina were chosen as objects for the following reasons:

• fishes are known to have good color vision,
• color constancy was found in behavioral experiments with fishes³,
• opponent code permits to judge what color the ganglion cell perceives,
• surround of the receptive field might supply the center with the information about the ambient light,
• its opponent influence on the center might well make necessary scaling for the central signals.

The receptive fields (RF) of the unit was mapped by means of a flashing light-point and moving black or white spots, stripes or edges. Then, a special device to project moving stimuli into the RF was positioned at the visual field and the right projection was adjusted with a mirror.
All the scene was illuminated with two independent sources. Their intensities and spectral contents were varied in a wide range.
While investigating the color constancy stimuli (sheets of colored papers) were moved with a constant velocity across the RF center on a velvet-black background, the RF surround being covered with a stationary white or black screen. The spike activity of the unit was recorded in response to introduction (ON response) and withdrawal (OFF response) of the stimulus.

The relative efficiencies (quantum catches) of the papers under white incandescent lamp illumination were calculated for each cone type from
  • spectral sensitivities of the cones,
  • spectral reflectances of the papers and
  • the spectrum of the illuminant.
The efficiency of perfect white diffuser was taken as 100%.

Positions of the papers relative efficiency in the cyprinid color space are shown in the RG-plane projection to the right.

The same types and the same sequences of retinal afferents were observed in the tectum
of the pike⁴ and some marine⁵ fishes:

golden gray mullet	Liza aurata (Risso) / Mugil auratus (Risso)
wrasses	Symphodus sp. / Crenilabrus sp.
red mullet	Mullus barbatus ponticus Essipov
high-body pickarel	Spicara smaris (Linnaeus)
Black Sea scad	Trachurus mediterraneus ponticus Aleev

Properties of the detectors of vertical and horizontal lines were thoroughly investigated later and published elsewhere^6,7.

Color-coding units receive inputs from red and green cones as revealed in color-matching experiments.

Fig.G3 (to the right). Two color-matching distribution coefficients (shown in black) obtained for the color-coding unit O-103.
The points indicate relative intensities of long-wave and short-wave lights (dashed spectral bands) in color mixtures, indistinguishable from certain monochromatic light (abscissa), the absence of reaction in response to the color substitution being a criterion of color match.

This type of retinal afferents demonstrates clear color-coding properties, their ON response being an indicator of the relative amount of a long-wave radiation in the light, reflected from the stimulus.

Fig.G4 (to the left). The correspondence between the space of responses (ON-OFF) of the color-coding unit O-43 and the RG-plane of the cyprinid color space.
Responses of the unit were recorded to presentations of various colored sheets of paper in black surround under white illumination. Dimensions of the rectangles at the figure correspond to dispersions of the experimental points. The curvilinear grid represents the distortion of the color space (see the RG-plane projection in section Materials and Methods for comparison).

Fig.C2 (to the right). Cell responses to the green stimulus under white and red illumination in black and white surround.
One can see that responses change (arrow) after switching on an additional red light source, when the surround is black, and do not change, when it is white.

Fig.C3 (to the left). An increase in number of spikes in the ON response of the unit O-25 to introduction of the green stimulus into its RF center with reddening the illumination when the RF surround is black (curve 1) and the relative constancy of the response to the same stimulus when the RF surround is white (curve 2).

During these experiments, human observer also perceives the stimulus as red at large intensities of the additional red illumination, if the surround is covered with a black screen and there are no information about the illumination, but sees the stimulus as green, if the surround is white.

Fig.C4 (to the right). Experiments on simultaneous color contrast.
Under white illumination when the RF surround is covered with a green screen, ganglion cell considers the illumination to be green and responds to the gray stimulus as to the red one - its OFF-response increases. This is a well known phenomenon of simultaneous color contrast.
Opposite illusion is observed when the RF surround is covered with a red screen.

Note, that neither surrounds nor illumination change in the course of stimulation in these experiments. Only gray paper moves across the RF center on a velvet-black background.

Horizontal cells have been shown to organize the surround of the ganglion cell receptive field⁸. Artificial sustained polarization of the horizontal cells, as well as sustained illumination of the ganglion cell RF surround, modifies the response of the ganglion cells to stimuli presented in their RF center⁹. So, it is reasonable to suggest horizontal cells to be responsible for realization of some color constancy mechanisms just at the very beginning of the visual pathway.

Literature

9. E.M.Maximova, V.V.Maximov. The role of horizontal cells in organization of ganglion cells concentric receptive fields in the fish retina. In: Visual Information Processing and Control of Motor Activity - Proceedings of the International Symposium (Sofia,1969), Bulgarian Acad. of Sci., Sofia, 1971, pp. 31-39 (in Russian)

Questions and comments to

Back to IITP Laboratory 8 Activity Page

Back to IITP Electronic Posters Page

Back to IITP Projects Page ( IITP russian Projects Page )

Last Update: 14 February, 2000