Microspectrophotometry of visual pigments and oil droplets in a marine bird, the wedge-tailed shearwater Puffinus pacificus: topographic variations in photoreceptor spectral characteristics

The nature of colour vision varies throughout the animal kingdom, as does the morphological structure of the eye, and this great diversity in ocular design is thought to represent adaptations to different life styles(Walls, 1942; Rochon-Duvigneaud, 1943; Meyer, 1977; Lythgoe, 1979). Our understanding of colour vision in birds is derived from only a handful of species, but it is evident that there is considerable interspecific variation in retinal design (e.g. Slonaker,1897; Wood, 1917; Goldsmith et al., 1984; Martin, 1985; Partridge, 1989; Bowmaker et al., 1997; Hart, 2001a,b; Ödeen and Håstad,2003). Nevertheless, in order to establish which features of the avian retina are adaptive for specific visual tasks it is necessary to obtain data from more species.

Only 3% of the world's birds are classified as seabirds(Nelson, 1980), but their highly specialized life styles may provide some insights into avian retinal design. However, while the optical structure of the eyes of marine species has received considerable attention (Sivak,1976; Sivak et al., 1977, 1987; Martin and Brooke, 1991;Martin, 1998, 1999; Martin and Prince, 2001), only limited and partial data are available on their photoreceptor spectral sensitivities (Liebman, 1972; Bowmaker and Martin, 1985; Bowmaker et al., 1997; Ödeen and Håstad,2003). Described here are new data on the spectral absorption characteristics of the visual pigments and cone oil droplets in the retinal photoreceptors of the wedge-tailed shearwater Puffinus pacificus,measured using microspectrophotometry. The wedge-tailed shearwater is a marine pelagic species that spends most of its life on the open ocean, approaching land only long enough to breed (Schodde and Tidemann, 1997). During the day, shearwaters forage low over the sea's surface looking for fish, crustaceans and cephalopods. Although predominantly diurnal in habit, they may come ashore after dark during the breeding season (Warham,1990).

Three fledgling wedge-tailed shearwaters Puffinus pacificus(Gmelin 1789) were caught under Queensland Parks and Wildlife Service Permit(WITK00494802) on Heron Island (Queensland, Australia) during May 2003. They had become entangled in Pisonia grandis seeds but were otherwise undamaged. Birds were held in darkness for at least 1 h prior to killing by approved humane methods (overdose of barbiturate anaesthetic followed immediately by decapitation). All procedures were conducted on Heron Island and were approved by the University of Queensland Animal Ethics Committee.

Following enucleation, dark-adapted eyes were dissected in 340 mOsmol kg^–1 phosphate-buffered saline (PBS; Oxoid, Basingstoke, UK)and retinal tissue prepared for analysis with a microspectrophotometer as described elsewhere (Hart et al., 1998, 1999, 2000a,b; Hart, 2002). Photoreceptors were mounted in a solution of 340 mOsmol kg^–1 PBS containing 10% dextran (MW 282,000; Sigma D-7265). All preparations were conducted under infrared (IR) illumination provided by a bank of IR light-emitting diodes(LEDs) and visualized using an IR image converter (FJW Optical Systems Inc.,Palatine, IL, USA).

The microspectrophotometer (MSP) used was a new field-portable single-beam wavelength-scanning instrument of simple design constructed by the author. The filament of a 12 V 50 watt tungsten halogen lamp was focused by a biconvex quartz lens (focal length f=50 mm) onto the entrance slit (1 mm width) of a Jobin-Yvon H10-61 UV-VIS monochromator (JY Horiba, France). The monochromator contained a directly driven, concave, holographic grating that dispersed the incident light and focused the diffracted wavefront onto the exit slit (1 mm width). An image of the exit slit was projected by another f=50 mm biconvex quartz lens onto a variable rectangular field aperture that controlled the dimensions of the measuring beam (minimum 1μm×1 μm). Light passing through the aperture was linearly polarized to take advantage of the inherent dichroism of photoreceptor outer segments when illuminated side-on. An image of the aperture was demagnified and focused into the plane of a specimen on a micrometer-manipulated microscope stage using a Zeiss (Germany) ×40 0.75 numerical aperture(NA) water immersion objective. Above the stage, a Leitz (Germany) ×100 1.32 NA oil immersion objective imaged the measuring beam onto the photocathode of a photomultiplier tube (PMT; Model R928, Hamamatsu, Japan) or, via a sliding mirror that could be introduced to the beam path, onto the image plane of a black and white charge-coupled device (CCD) video camera. To view the specimen using the CCD camera, light (>900 nm) from an IR LED was directed into the beam path using a fixed 45° beamsplitter (19 mm diameter No. 0 coverglass; ProSciTech, Australia) positioned below the condenser objective.

To compensate for longitudinal chromatic aberration in the condenser objective, which would otherwise cause defocus of the measuring beam in the specimen plane when scanning through the spectrum (between 370 nm and 800 nm),a MIPOS 3 SG piezoelectric translator and 12 V 40 SG piezo amplifier(Piezosystem, Jena, Germany) were used to move the condenser automatically during the scan. The wavelength calibration of the MSP was checked with a calibrated Ocean Optics USB2000 spectroradiometer (Ocean Optics Inc., Dunedin,FL, USA) and was accurate to better than ±1 nm.

Sample and baseline scans were made from cellular and tissue-free regions of the preparation, respectively. Dimensions of the measuring beam varied from approximately 1 μm×1 μm for oil droplets and small cone outer segments to 2 μm×10 μm for rods. The photocurrent induced by light from the measuring beam reaching the photomultiplier was converted to a voltage by a transimpedance headstage amplifier (Hamamatsu C6271, Schizuoka,Japan) and differentially amplified to reject common-mode voltages. The voltage was further amplified and scaled to suit the input range of a 12-bit successive-approximation sample and hold analogue-to-digital converter that was subsequently interrogated by a Gateway laptop microcomputer viathe parallel port. Each scan consisted of a `downward' long- to short-wavelength pass and an `upward' short- to long-wavelength spectral pass. Corresponding wavelength data from the downward and upward spectral passes were averaged together. To reduce the effects of in-scan bleaching, only one sample scan was made of a given outer segment and this was combined with a single baseline scan. Following these `pre-bleach' scans, outer segments were bleached with full spectrum `white' light from the monochromator for 3 min and an identical number of sample and baseline scans made subsequently. The`post-bleach' spectrum thus created was deducted from the pre-bleach spectrum to create a bleaching difference spectrum for each outer segment. To establish visual pigment-oil droplet pairings, the spectral absorptance of the oil droplet associated with the outer segment (if present) was also measured in the same way.

Baseline and sample data were converted to absorbance values at each wavelength and spectra normalized to the peak and long-wavelength offset absorbances determined by fitting a variable-point unweighted running average to the data (Hart, 1998). A regression line was fitted to the normalized absorbance data between 30% and 70% of the normalized maximum on the long-wavelength limb and the regression equation used to predict the wavelength of maximum absorbance(λ_max) following the methods of MacNichol(1986) and Govardovskii et al.(2000). Spectra from each photoreceptor type that satisfied established selection criteria(Levine and MacNichol, 1985; Hart et al., 1999) were averaged and reanalysed. For display, averaged spectra were overlaid with a rhodopsin (vitamin A₁-based) visual pigment template of the sameλ _max generated using the equations of Govardovskii et al.(2000).

Sample and baseline data were converted to absorptance and normalized to the maximum and long-wavelength offset absorptances obtained by fitting an 11-point unweighted running average to the data. Oil droplet absorptance spectra were described by their cut-off wavelength (λ_cut),which is the wavelength of the intercept at the value of maximum measured absorptance by the line tangent to the oil droplet absorptance curve at half maximum measured absorptance (Lipetz,1984a). For comparison with other studies (e.g. Partridge, 1989) the wavelength corresponding to half maximum measured absorptance(λ_mid) was also calculated(Lipetz, 1984a).

The spectral transmittance (250–800 nm) of the ocular media in the anterior segment of the eye (cornea, aqueous humour and lens) was measured along the optical axis using an Ocean Optics S2000 spectroradiometer. Broad spectrum white light from an Ocean Optics PX-2 pulsed xenon lamp was delivered to either the corneal or lenticular aspects of the anterior segment of a hemisected eye at normal incidence using a 1 mm diameter fibre optic. The anterior segment was supported on a hollow metal tube (10 mm in length and 5 mm internal diameter) screwed onto the sub-miniature type A (SMA) connector of a 0.4 mm diameter fibre optic that relayed transmitted light to the S2000. Data collection was controlled via a Toshiba laptop microcomputer. After correcting the S2000 for electrical dark current, sample and baseline readings were made – with and without the anterior segment of the eye in the beam path, respectively – and converted to transmittance. In total,13 spectra were obtained from the anterior segment of a single eye; of these,six were illuminated from the corneal aspect and seven from the lenticular aspect. All spectra were interpolated at 1 nm wavelength intervals, smoothed using an 11-point unweighted running average and averaged together.

Macroscopic examination of the fundus of the shearwater eye revealed an obvious visual streak approximately 1.5 mm dorsal to the top of the pecten that ran horizontally across the retina. Microscopic examination of the dissected retina showed that the photoreceptors in the visual streak(Fig. 1C) were smaller and more densely packed than in the peripheral retina(Fig. 1A). Moreover, the pigmentation of the oil droplets found in the cone photoreceptors in the visual streak was greatly reduced compared to those found in the peripheral retina. There were, for instance, no bright yellow or red oil droplets in the centre of the visual streak. The transition from heavily pigmented to pale-coloured oil droplets occurred on the edge of the visual streak(Fig. 1B).

Microspectrophotometric data for visual pigments (Figs 2, 3) and oil droplets(Fig. 4) measured outside the visual streak are summarized in Table 1. The retina of the wedge-tailed shearwater contained five different types of visual pigment in seven different classes of photoreceptor. On the basis of goodness-of-fit to visual pigment templates(Govardovskii et al., 2000),all of the visual pigments were considered to be rhodospins, where the chromophore is 11-cis retinal. Rods contained a medium-wavelength sensitive (MWS) visual pigment with a λ_max at 502 nm. There were four spectrally distinct types of single cone. Firstly, a violet sensitive (VS) type with a 406 nm λ_max visual pigment and a`transparent' T-type oil droplet that showed no significant absorptance down to at least 370 nm. Secondly, a short-wavelength sensitive (SWS) type with a 450 nm λ_max visual pigment and a pale greenish-yellow C-type oil droplet with a λ_cut at 445 nm. Thirdly, a MWS type with a 503 nm λ_max visual pigment and a golden yellow Y-type oil droplet with a λ_cut at 506 nm. Lastly, a long-wavelength sensitive (LWS) type with a 566 nm λ_max visual pigment and a red R-type oil droplet with a λ_cut at 562 nm. Both the principal and accessory members of the double cone pair contained the same 566 nm λ_max visual pigment that was also found in the LWS single cones. While the accessory member did not contain an oil droplet, the principal member had a colourless oil droplet with a λ_cut at 413 nm.

In the central retina, cone oil droplets appeared either transparent/colourless or pale green (Fig. 1C). The transition from `normally' pigmented oil droplets located towards the periphery (Fig. 1A)to these less pigmented types in the centre of the visual streak occurred over a short distance (approximately 50 μm when measured on a flat-mounted retina) on the edges of the visual streak (Figs 1B, 5A). Unlike in the periphery,definitive visual pigment-oil droplet pairings in the central retina/visual streak could not be made because of the smaller size of the cone outer segments (<1 μm diameter where present) and the fact that the retinal tissue was not dispersed in order to preserve orientation. Nevertheless, it is likely that there were several different types of cone in the central retina,each of which contained a different visual pigment, because there were oil droplets of slightly different diameters(Fig. 1C) and with different spectral absorptance characteristics (Fig. 5B).The absorptance spectra of oil droplets measured in the central retina fell into five distinct groups (labeled t, c, y, r and p in Fig. 5B). The most common droplet type (p) had a spectral absorptance (mean λ_cut=414 nm, N=12) that closely resembled the P-type oil droplets of the principal member of the LWS double cones in the peripheral retina. Transparent oil droplets (t) showing no significant absorption across the spectrum were also present, and were identical to the T-type oil droplets associated with the VS visual pigment in the periphery. The remaining three types of oil droplet were classified as c, y and r and were considered to be less pigmented versions of the C-, Y- and R-type oil droplets found in the SWS, MWS and LWS single cones at the periphery. This was deduced by taking the measured absorptance spectra for the central c, y, and r oil droplets (Fig. 5B) and modeling the effect of increasing the density of the carotenoid pigment(s) they contain to levels estimated for the C, Y and R-type oil droplets found in the periphery. The results of this modeling are shown in Fig. 5C. Peak absorbances for the peripheral Y- and R-type oil droplets (16.6 and 2.4, respectively) were estimated from their λ_cut values using the equations given by Lipetz (1984b) for turtle oil droplets. The relationship between C-type oil dropletλ _cut and peak absorbance has not been quantified and was taken to be 1.0 for peripheral C-type oil droplets; this value is the lower limit for pigmented droplets in the turtle retina(Lipetz, 1984b). (The peak absorbance of densely pigmented oil droplets cannot be measured accurately with a microspectrophotometer due to the effects of bypassing light(Lipetz, 1984b), which in the current machine limit the upper absorbance values that can be measured to about 0.9 for a 3 μm diameter droplet, as indicated by the flat-topped absorptance spectra for the C-, Y- and R-type oil droplets in Fig. 4A.) Smoothed absorptance spectra for the central c, y, and r oil droplets were converted to absorbance, scaled by a factor equivalent to the ratio of central(measured) oil droplet absorbance to peripheral (estimated) oil droplet absorbance (3.2, 8.5 and 60.3 for the C:c, Y:y and R:rtypes, respectively) and converted back to absorptance for display. The modeled oil droplets, labeled c′, y′ and r′ in Fig. 5C,resemble the peripheral C-, Y- and R-type droplets, havingλ _cut at around 445, 490 and 540 nm, respectively. From Figs 5A,C and 1A–C it seems more than likely that, rather than losing particular cone types, the oil droplets associated with the SWS, MWS and LWS visual pigments gradually reduce their carotenoid pigmentation the nearer they are to the centre of the visual streak.

The method used to measure the spectral transmittance of the ocular media in the anterior segment of the eye was not ideal. More reliable results can be obtained using a spectrophotometer fitted with an integrating sphere (Hart et al., 1998, 1999; Hart, 2002) because light scattered by the tissue is collected more efficiently. Nevertheless, the method used in this instance was a successful improvisation in the field, the only drawback being that the absolute maximum transmittance of the ocular media could not be determined with sufficient accuracy. For this reason, the averaged transmittance spectrum was normalized to the measured maximum transmittance at long wavelengths. As can be seen from Fig. 6, the ocular media of P. pacificus transmits wavelengths down to approximately 300 nm, with a wavelength at 0.5 transmittance (λ_T0.5) of 335 nm.

With the exception of the transparent T-type oil droplet associated with the VS (or UVS) single cones, avian cone oil droplets generally act as long-pass cut-off filters. Consequently, the spectral sensitivity of a given cone is the product of the spectral absorptance of the visual pigment in the outer segment and the spectral transmittance (1-absorptance) of the oil droplet with which it is paired. Relative quantal spectral sensitivities were calculated for single cone photoreceptors in different regions of the wedge-tailed shearwater retina (Fig. 7A,B), as follows. Visual pigment spectral absorptance was modeled using mathematical templates (Govardovskii et al., 2000) of the appropriate λ_max(Table 1). Outer segments were assumed to be 16 μm long (Morris and Shorey, 1967) and contain a visual pigment with a specific(decadic) absorbance of 0.014 μm^–1(Bowmaker and Knowles, 1977). Predicted photon catches for each cone type were also adjusted according to the cross-sectional area of the relevant oil droplet they contained(Table 1) and the transmittance of the ocular media (Fig. 6). For cones in the peripheral retina (i.e. outside the visual streak; Fig. 1A), the VS, SWS, MWS and LWS cone visual pigments were combined with their corresponding oil droplets,the absorptance spectra of which are labeled T, C, Y and R, respectively, in Fig. 4A. Because the T-type oil droplets do not contain short-wavelength-absorbing carotenoid pigments, the peak sensitivity of the VS cone (407 nm) is very similar to theλ _max of the visual pigment it contains. However, spectral filtering by the C-, Y- and R-type oil droplets shift the peak spectral sensitivities of the SWS, MWS and LWS single cones (472, 538 and 600 nm,respectively; Fig. 7A) to wavelengths longer than the λ_max of their respective visual pigments.

In the central retina, the VS, SWS, MWS and LWS single cones were assumed to contain the t, c, y and r droplets shown in Fig. 5B (see above). Calculated spectral sensitivities are also displayed in Fig. 7A. The spectral location of the peak sensitivity of the VS single cone was the same as in the periphery. However, the reduced pigmentation of the presumptive C-, Y- and R-type oil droplets in the centre of the visual streak results in smaller shifts in the peak sensitivities of the SWS, MWS and LWS single cones towards long wavelengths (466, 512 and 573 nm, respectively). The length of the cone outer segments was assumed to be identical between central and peripheral regions, as appears to be the case for other bird species (e.g. Rojas et al., 1999). Nevertheless, the total photon catch of all single cone types in the centre of the visual streak was less than that at the periphery because of the reduced cross sectional area of the oil droplets (mean diameters 1.4, 2, 2 and 2.1μm for t, c, y and r oil droplets,respectively), despite the reduction in oil droplet pigmentation (which at the periphery reduces the overall photon catch of the outer segment by 30, 47 and 39%, respectively, for the SWS, MWS and LWS single cones). To further illustrate the transition in cone spectral properties from peripheral to central retina, the spectral sensitivities of LWS single cones on the edge of the visual streak were also calculated(Fig. 7B, traces 1–4),using the absorptance spectra for oil droplets that changed from red to orange in this retinal area (Fig. 5A,spectra 1–4).

In general, the retina of the wedge-tailed shearwater is typical of other diurnal birds. It contains rods, double cones and four types of single cone. Rod, MWS single cone and LWS single/double cone visual pigmentλ _max values show relatively little variation between bird species. The biggest interspecific difference is in the λ_maxof the visual pigment found in the single cones that contain a T-type oil droplet; values measured microspectrophotometrically range from 359 to 424 nm(for a review, see Hart,2001b). All of these UVS/VS type visual pigments are thought to belong to the same `SWS1' opsin class(Yokoyama, 2000) and are conjugated with the same chromophore (11-cis retinal). Differences in their λ_max values are, therefore, due only to differences in the amino acid sequence of their opsin protein (Wilkie et al., 1998, 2000; Yokoyama et al., 2000a,b; Hunt et al., 2001). Interestingly, the λ_max of the SWS visual pigment found in the single cones with C-type oil droplets covaries with theλ _max of the SWS1 opsin-based visual pigment found in the UVS/VS single cone, as does the λ_cut of the C-type oil droplet and, accordingly, the spectral sensitivity of the SWS single cones(Bowmaker et al., 1997; Hart, 2001b). The strong correlation between UVS/VS single cone and SWS single cone spectral sensitivities suggests that avian photoreceptors are tuned to optimize the extraction of chromatic information over a given spectral range, the short wavelength limit of which varies between species.

With a λ_max at 406 nm, the VS visual pigment of the wedge-tailed shearwater is similar to the only other marine birds for which data are available, the Manx shearwater Puffinus puffinus (402 nmλ _max; Bowmaker et al.,1997) and the Humboldt penguin Spheniscus humboldti (403 nm λ_max; Bowmaker and Martin, 1985). This type of VS visual pigment, which differs subtly from those found in anseriform and galliform species(λ_max 415–426 nm), is also found in the ostrich Struthio camelus (405 nm λ_max; Wright and Bowmaker, 2001) and the feral pigeon Columba livia (409 nm λ_max; Bowmaker et al., 1997) and so is not exclusive to marine bird species. On the basis of SWS1 opsin DNA sequences, Ödeen and Håstad(2003) have suggested that the distribution of UVS/VS visual pigment λ_max values among bird species is complex and probably reflects ecological adaptations rather than phylogenetic relatedness, as seems to be the case for cone photoreceptor abundance (Goldsmith et al.,1984; Partridge,1989; Hart,2001a). There are likely to be many different adaptive reasons for a particular complement of photoreceptor spectral sensitivities, depending,for example, on the physical habitat and foraging or predator avoidance behaviours of a given species.

As an oceanic marine bird, the visual system of the wedge-tailed shearwater appears to be well adapted to its photic environment. Shearwaters use a variety of strategies to capture their prey, which consists largely of fish,squid and crustaceans (Nelson,1980), but feed mainly while floating on the water or during shallow dives made either from the surface or a short distance above it(Warham, 1990). They must,therefore, be able to detect their prey through the air–water interface,although this is a complex visual task: plunge-diving birds most cope not only with the effects of refraction on the apparent position of the prey item underwater (Katzir, 1993), but also with specular reflection of skylight from the surface that interposes a contrast-reducing glare (Lythgoe,1979). Reflections from the surface of the water have the same spectral distribution as skylight, but the spectral radiance of the upwelling light from the ocean – against which prey must be detected – is dominated by the absorption and scatter of the water(Austin, 1974). Considering this phenomenon, Lythgoe(1979) proposed that a visual system most sensitive to wavelengths in which the upwelling light is rich, and the surface reflectance relatively poor, would be adaptive for through-surface vision.

The spectral distribution of upwelling light from a body of water depends on the amount of particulate matter, dissolved organic compounds and chlorophyll it contains. Different types of water have different levels of these substances and, accordingly, different upwelling spectral radiances or`colours' (Austin, 1974; Jerlov, 1976). Fig. 8 shows the calculated difference between upwelling and surface-reflected radiances for three different water types: `blue', `blue-green' and `green', calculated from data in Austin (1974). The geographical range of the wedge-tailed shearwater extends throughout the tropical and subtropical Indian and Pacific Oceans(del Hoyo et al., 1992). Consequently, they forage mainly over oceanic `blue' water types (Jerlov types I, IA and IB; Jerlov, 1976)whose upwelling spectral radiance is relatively richer in short wavelengths between about 400 and 500 nm (Fig. 8). Below 400 nm, submerged objects would become increasingly difficult to distinguish against the surface glare and the 406 nmλ _max VS visual pigment of the wedge-tailed shearwater would,therefore, be more useful than a UVS-type visual pigment(λ_max<400 nm) for through surface vision. It is also interesting to note that the proportion of VS cones in the wedge-tailed shearwater retina (approximately 16% of all cone types) is at least twice as high as in many terrestrial bird species (e.g. Goldsmith et al., 1984; Hart et al., 1998, 2000b; Wright and Bowmaker, 2001) and marine species (e.g. silver gull Larus novaehollandiae, noddy tern Anous minutus) that surface seize but do not plunge dive(Hart, 2001a). Moreover,despite their suggested role in movement detection(Campenhausen and Kirschfeld,1998), the retinae of blue-water marine birds generally contain a lower proportion of double cones (wedge-tailed shearwater 34%, silver gull 30%, noddy tern 29%) compared to terrestrial species (range 35–56%; Hart, 2001a), possibly because their long-wavelength sensitivity is less useful for through-water vision.

The LWS visual pigment of the wedge-tailed shearwater has a similarλ _max (566 nm) to other terrestrial bird species (see Hart, 2001b) and is not shifted towards shorter wavelengths as in the Humboldt penguin Spheniscus humboldti (Bowmaker and Martin,1985). The 543 nm λ_max LWS visual pigment of the penguin is presumably an adaptation to the restricted spectral bandwidth it encounters when foraging in the ocean at depths of 30 m or more(Williams, 1995), where longer wavelengths are attenuated more rapidly than shorter wavelengths with increasing depth (Jerlov,1976). Plunge-diving shearwaters, on the other hand, take most of their prey within 2 m of the surface(Warham, 1990) where the spectral distribution of light available for vision is less restricted.

The wedge-tailed shearwater retina has a distinct area centralis horizontalis or visual streak running horizontally across the retina. A similar band of increased cell density – with or without a central fovea– has been shown, either anatomically or ophthalmoscopically, in several other procellariiform species: the sooty albatross Phoebetria fusca,shy albatross Diomedea cauta, Manx shearwater Puffinus puffinus, sooty shearwater Puffinus griseus, soft-plumaged petrel Pterodroma mollis, Fulmar petrel Fulmaris glacialisand giant petrel Macronectes giganteus(Wood, 1917; O'Day, 1940; Lockie, 1952; Hayes and Brooke, 1990). Some shearwaters, for example the little shearwater Puffinus assimilis,lack a visual streak and instead have a simple area centralis;differences in the retinal topography of petrel species are undoubtedly related to visual ecology, most likely feeding behaviour(Hayes and Brooke, 1990).

Several specialized functions have been attributed to linear areas in birds and other animals, including movement detection, spatial orientation and fixation of the horizon (reviewed in Meyer, 1977). Measurements of the visual fields of the Manx shearwater eye reveal that the long axis of the visual streak is aligned parallel to the physical horizon(Martin and Brooke, 1991), as is the case for other animals that inhabit open, relatively featureless environments (Hughes, 1977; Meyer, 1977). The region of visual space from just above the horizon to just below it will be of great significance to birds such as shearwaters in finding food. Shearwaters usually forage within 10 m of the water's surface(Haney et al., 1992) and search for prey solitarily, only converging to form flocks when a source of food is located and an individual is observed dropping to the surface to feed(Warham, 1990). Haney et al.(1992) calculated that the mean horizontal distance over which procellariiform seabirds were recruited visually to a feeding flock was around 4.5 km and even proposed a theoretical limit of 20–30 km. It is evident, therefore, that a visual streak sampling the physical horizon with a high spatial resolving power would be adaptive for the detection of potential food sources on the open ocean.

The most intriguing feature of the wedge-tailed shearwater retina is that the pigmented cone oil droplets lose their coloration in the central area of the visual streak. These pale-coloured oil droplets in the visual streak are of marginally different sizes (Fig. 1C) and have slightly different absorptance spectra(Fig. 5B). Moreover, the transition from brightly coloured to almost colourless droplets can be seen at the edge of the visual streak (Figs 1B, 5A). It seems likely,therefore, that rather than only one cone type being present at the centre of the visual streak, all different types of cone are present, but contain oil droplets with absorptance spectra that are different from those in the peripheral retina. This has also been observed in other species that have very small, densely packed photoreceptors in their central retina, e.g. sacred kingfisher Todiramphus sanctus(Hart, 2001a) and laughing kookaburra Dacelo novaeguineae (N. S. Hart, unpublished). Pigmented oil droplets act as long-pass cut-off filters, blocking almost all light below a critical wavelength (λ_cut). Consequently, the effect of a coloured oil droplet on a given cone type is to narrow its spectral sensitivity function, shift the peak sensitivity to a wavelength longer than the λ_max of the visual pigment it contains and reduce the overall photon catch (Bowmaker,1977). From the predicted spectral sensitivities of the cone photoreceptors in the wedge-tailed shearwater retina(Fig. 7A), it is evident that the reduction of spectral filtering by oil droplets in the central retina results in greater overlap between adjacent spectral classes and a shift in the peak sensitivity of the SWS, MWS and LWS single cone types towards shorter wavelengths compared to those at the periphery. Reduced overlap of adjacent cone spectral sensitivities is thought to improve the discrimination of broadband (`natural') reflectance spectra and enhance colour constancy(Govardovskii, 1983; Vorobyev, 1997, 2003; Vorobyev et al., 1998). However, the benefits of spectral filtering by oil droplets are strongly dependent on light intensity because they reduce the overall quantum catch of the cone and, accordingly, increase photoreceptor signal noise. In the wedge-tailed shearwater retina, the cones become narrower and more densely packed with decreasing eccentricity (Fig. 1), presumably to enhance spatial acuity in the visual streak. The associated decrease in photon capture area results in a lower quantal sensitivity compared to peripheral cones(Fig. 7A), despite the reduction in oil droplet pigmentation. Moreover, ganglion cells in retinal regions of high spatial acuity, such as the visual streak, tend to have smaller receptive fields (Rodieck,1973), receive inputs from fewer cones (e.g. Manx shearwater; Lockie, 1952) and,consequently, have a lower signal-to-noise ratio than ganglion cells in the periphery. Theoretical models suggest that the benefit of coloured oil droplets for colour discrimination is marginal at lower light intensities(Vorobyev, 2003). This trade-off between spatial acuity (small cones, low summation) and contrast sensitivity (big cones, high summation) may preclude the presence of highly pigmented oil droplets in centrally located cones because they would further reduce quantum catch below a particular noise threshold. This is probably also the reason why nocturnal and crepuscular species of bird do not have highly pigmented droplets (Muntz,1972; Bowmaker and Martin,1978).

If, as in chickens Gallus gallus(Osorio et al., 1999), outputs from the four single cone types in the wedge-tailed shearwater retina are compared using opponent mechanisms in a tetrachromatic colour vision system, a coloured object whose image falls on the photoreceptors in the visual streak will produce different cone opponent signals than if it fell on the peripheral retina, especially if its reflectance is rich in short wavelengths of light. Whether or not the neural circuitry of the opponent mechanisms compensates for this topographic variation in colour perception is unknown and this is clearly an area for future study.

 
• CCD

  charge-coupled device

 
• f

  focal length

 
• IR

  infrared

 
• LED

  light emitting diode

 
• LWS

  long-wavelength sensitive

 
• MSP

  microspectrophotometer

 
• MWS

  medium-wavelength sensitive

 
• NA

  numerical aperture

 
• PBS

  phosphate-buffered saline

 
• PMT

  photomultiplier

 
• SMA

  sub-miniature type A connector

 
• SWS

  short-wavelength sensitive

 
• UVS

  ultraviolet-sensitive

 
• VS

  violet sensitive

 
• λcut

  cut-off wavelength (oil droplet)

 
• λmax

  wavelength of maximum absorbance (visual pigment)

 
• λmid

  wavelength of half maximum-measured absorptance (oil droplet)

 
• λT0.5

  wavelength of 0.5 transmittance (ocular media)

The author thanks Prof. David Vaney, Prof. Justin Marshall and Dr Julia Shand (who all donated equipment used in the construction of the microspectrophotometer), Dr Misha Vorobyev, the staff of Heron Island Research Station and the Queensland Parks and Wildlife Service. The author was funded by a University of Queensland (UQ) Postdoctoral Research Fellowship and a UQ New Staff Start-up Grant. All experiments complied with the `Principals of Animal Care', publication No. 86-23, revised 1985, of the National Institute of Health.