20
Lars Chittka and Adriana Briscoe
ambient light in deep water (Lythgoe 1972; Hunt eta!. 1996; Douglas eta!. 1998).
Other presumed cases of adaptive spectral tuning include humans, whose
trichromatic systems might have evolved as a response to frugivory (Mollon
1989), and bees, whose receptors were suggested to be evolutionarily tuned for
flower color coding (Chittka and Menzel 1992). Such intuitively appealing stories
of adaptation easily find their way into textbooks. Studies which do not find an
obvious match between sensory traits and the environment do exist (Crandall and
Cronin 1997; Fleishman et al. 1997), but they often achieve much less fame. Even
in deep-sea fish, which have long been held as classical examples of adaptive
visual pigment tuning, the situation is far from resolved. Fish which live in similar
photic environments, but belong to different taxa, often have different visual
pigments, whereas closely related fish species sometimes have similar visual
pigments, even if they inhabit different light habitats (Douglas et al. 1998).
Turning to bees, the notion that pollinator color vision is tuned to floral colors is
compromised by the finding that arthropods living under entirely different visual
conditions, such as the beach isopod Ligia (Hariyama eta!. 1993), the freshwaterdwelling bug Notonecta (Bmckmoser 1968), nocturnal hawkmoths (White et a!.
1994 ), and the larval ocelli of some Lepidoptera (Ichikawa and Tateda 1982) have
similar sets of color receptors. On the basis of such difficulties, Goldsmith ( 1990)
concluded that phylogenetic and molecular constraints might play more important
a role in determining the wavelength positioning of color receptors than is good
for any pan-adaptionist scenario. We wish to reiterate this warning, and to add
several additional ones.
Our most important caveat is that to show that a trait is adaptive, we must
demonstrate that it has an impact on fitness (Endler 1986; Reeve and Sherman
1993). What is the evolutionary significance of a model, for example, which
shows that for a given visual task, one set of color receptors is 5% better than
another? If this really translates into 5% more lifetime reproductive success, the
effect of selection will probably be significant over evolutionary time. On the
other hand, it is just as possible that 5% improved performance in some criterion
will be absolutely irrelevant to fitness. To our knowledge, there is not a single
study in sensory ecology that resolves this problem. Such fitness tests are
challenging, but should be possible. In what follows, we will lay out a research
agenda that include several steps towards such tests. We hope this treatise will
stimulate a more evolutionary approach to sensory ecology, and a better
understanding of why many animals see the world in colors so differently from
ourselves.
2 Uses and Limitations of Model Calculations
Ten years ago, one of us (L.C.), in collaboration with R. Menzel, set out to
identify the adaptive significance of bee color vision. The idea was to generate a
theoretically optimal color vision system for the task of flower color coding, and
to compare this with the system really implemented in bees. We had considerable
Lars Chittka and Adriana Briscoe
ambient light in deep water (Lythgoe 1972; Hunt eta!. 1996; Douglas eta!. 1998).
Other presumed cases of adaptive spectral tuning include humans, whose
trichromatic systems might have evolved as a response to frugivory (Mollon
1989), and bees, whose receptors were suggested to be evolutionarily tuned for
flower color coding (Chittka and Menzel 1992). Such intuitively appealing stories
of adaptation easily find their way into textbooks. Studies which do not find an
obvious match between sensory traits and the environment do exist (Crandall and
Cronin 1997; Fleishman et al. 1997), but they often achieve much less fame. Even
in deep-sea fish, which have long been held as classical examples of adaptive
visual pigment tuning, the situation is far from resolved. Fish which live in similar
photic environments, but belong to different taxa, often have different visual
pigments, whereas closely related fish species sometimes have similar visual
pigments, even if they inhabit different light habitats (Douglas et al. 1998).
Turning to bees, the notion that pollinator color vision is tuned to floral colors is
compromised by the finding that arthropods living under entirely different visual
conditions, such as the beach isopod Ligia (Hariyama eta!. 1993), the freshwaterdwelling bug Notonecta (Bmckmoser 1968), nocturnal hawkmoths (White et a!.
1994 ), and the larval ocelli of some Lepidoptera (Ichikawa and Tateda 1982) have
similar sets of color receptors. On the basis of such difficulties, Goldsmith ( 1990)
concluded that phylogenetic and molecular constraints might play more important
a role in determining the wavelength positioning of color receptors than is good
for any pan-adaptionist scenario. We wish to reiterate this warning, and to add
several additional ones.
Our most important caveat is that to show that a trait is adaptive, we must
demonstrate that it has an impact on fitness (Endler 1986; Reeve and Sherman
1993). What is the evolutionary significance of a model, for example, which
shows that for a given visual task, one set of color receptors is 5% better than
another? If this really translates into 5% more lifetime reproductive success, the
effect of selection will probably be significant over evolutionary time. On the
other hand, it is just as possible that 5% improved performance in some criterion
will be absolutely irrelevant to fitness. To our knowledge, there is not a single
study in sensory ecology that resolves this problem. Such fitness tests are
challenging, but should be possible. In what follows, we will lay out a research
agenda that include several steps towards such tests. We hope this treatise will
stimulate a more evolutionary approach to sensory ecology, and a better
understanding of why many animals see the world in colors so differently from
ourselves.
2 Uses and Limitations of Model Calculations
Ten years ago, one of us (L.C.), in collaboration with R. Menzel, set out to
identify the adaptive significance of bee color vision. The idea was to generate a
theoretically optimal color vision system for the task of flower color coding, and
to compare this with the system really implemented in bees. We had considerable