Why Sensory Ecology Needs to Become More Evolutionary
21
reasons for optimism. First, Menzel and his coworkers had established a database
that included the color receptor sensitivity functions of a large variety of
hymenopteran species (Peitsch et a!. 1992) and insects in general (Menzel and
Backhaus 1991). This database suggested that insects could, in principle, produce
pigments with values of maximum sensitivity (Amax) anywhere from 320 to 630
nm, and that the number of color receptor types varied widely between species.
This, in combination with the fact that insects occupy a very wide range of visual
environments, made studying the sensory ecology of their color vision look
promising. Second, Backhaus (1991) had just developed a physiological model of
bee color vision, which allowed quantitative predictions of the similarity of flower
colors, and of flowers and their background. Such a model is an essential tool
to measure the quality of a color vision system and, to date, such models are still
not available for any animal species besides bees and humans. Third, unlike
many other animals, bees seemed an ideal study subject because the relevant
visual tasks are comparatively easily identified: bees obtain their food from
flowers, and so selection should favor color vision systems that allow for swift
detection of flowers and reliable identification of the most rewarding species
(Chittka 1997).
Spectral sensitivity functions of color receptors have roughly Gaussian
characteristics, and the exact shape of the curve can be easily predicted if the Amax
is known (Stavenga et a!. 1993). Our evolutionary model calculations consisted of
moving three color receptor sensitivity curves along the wavelength scale. For
each theoretical combination of receptors thus generated, the quality of the color
vision system for flower color coding was determined. The result was striking: the
optimal color receptors generated by the evolutionary model invariably occurred
at Amax = 330, 430, and 550 nm, which is very close the most common Amax really
found in flower-visiting bees (Chittka and Menzel 1992). This result was
independent of whether we varied one, two, or three photoreceptors. It was also
independent of the particular set of flowers used (Chittka 1996a).
Since the optimal set of color receptors might also depend on the particular kind
of opponent coding in the brain, the mode of this processing, too, was varied - the
result remained unchanged (Chittka 1996a). An engineer could hardly design a
better receiver for flower colors than the color vision system of bees. Does this,
however, mean that flower colors indeed drove the evolution of bee color vision?
This is an attractive notion for sensory ecologists. It is joined by other studies in
which correlations between the results of model calculations and reality were
taken as evidence for adaptation (Lythgoe and Partridge 1989). Thus, many
colleagues took our finding to mean that bee color vision indeed adapted to flower
colors, although we explicitly stated in the discussion of the original paper that
this is not necessarily the case (Chittka and Menzel 1992).
Indeed, there are several complications. Models are useful to generate
hypotheses of optimality, but a correlation between a model and biological traits
does not resolve how these traits evolved. Using models to reject a hypothesis of
evolutionary causality is much more straightforward. Had the optimal color
receptors derived from the model calculations been different from the ones found
in "real animals", then this would have indicated that evolution has not optimized
21
reasons for optimism. First, Menzel and his coworkers had established a database
that included the color receptor sensitivity functions of a large variety of
hymenopteran species (Peitsch et a!. 1992) and insects in general (Menzel and
Backhaus 1991). This database suggested that insects could, in principle, produce
pigments with values of maximum sensitivity (Amax) anywhere from 320 to 630
nm, and that the number of color receptor types varied widely between species.
This, in combination with the fact that insects occupy a very wide range of visual
environments, made studying the sensory ecology of their color vision look
promising. Second, Backhaus (1991) had just developed a physiological model of
bee color vision, which allowed quantitative predictions of the similarity of flower
colors, and of flowers and their background. Such a model is an essential tool
to measure the quality of a color vision system and, to date, such models are still
not available for any animal species besides bees and humans. Third, unlike
many other animals, bees seemed an ideal study subject because the relevant
visual tasks are comparatively easily identified: bees obtain their food from
flowers, and so selection should favor color vision systems that allow for swift
detection of flowers and reliable identification of the most rewarding species
(Chittka 1997).
Spectral sensitivity functions of color receptors have roughly Gaussian
characteristics, and the exact shape of the curve can be easily predicted if the Amax
is known (Stavenga et a!. 1993). Our evolutionary model calculations consisted of
moving three color receptor sensitivity curves along the wavelength scale. For
each theoretical combination of receptors thus generated, the quality of the color
vision system for flower color coding was determined. The result was striking: the
optimal color receptors generated by the evolutionary model invariably occurred
at Amax = 330, 430, and 550 nm, which is very close the most common Amax really
found in flower-visiting bees (Chittka and Menzel 1992). This result was
independent of whether we varied one, two, or three photoreceptors. It was also
independent of the particular set of flowers used (Chittka 1996a).
Since the optimal set of color receptors might also depend on the particular kind
of opponent coding in the brain, the mode of this processing, too, was varied - the
result remained unchanged (Chittka 1996a). An engineer could hardly design a
better receiver for flower colors than the color vision system of bees. Does this,
however, mean that flower colors indeed drove the evolution of bee color vision?
This is an attractive notion for sensory ecologists. It is joined by other studies in
which correlations between the results of model calculations and reality were
taken as evidence for adaptation (Lythgoe and Partridge 1989). Thus, many
colleagues took our finding to mean that bee color vision indeed adapted to flower
colors, although we explicitly stated in the discussion of the original paper that
this is not necessarily the case (Chittka and Menzel 1992).
Indeed, there are several complications. Models are useful to generate
hypotheses of optimality, but a correlation between a model and biological traits
does not resolve how these traits evolved. Using models to reject a hypothesis of
evolutionary causality is much more straightforward. Had the optimal color
receptors derived from the model calculations been different from the ones found
in "real animals", then this would have indicated that evolution has not optimized