We have used behavioural tests to determine the intensity thresholds of colour vision in Bourke's parrots (Neopsephotus bourkii) and budgerigars (Melopsittacus undulatus). We have also examined the relationship between these thresholds and the optical sensitivities of single photoreceptors using morphological methods. Bourke's parrots lose colour vision in brighter light (0.4 cd m–2) than budgerigars (0.1 cd m–2) and both birds lose colour vision in brighter light(`end of civil twilight') than humans (0.02 cd m–2,`moonlight'). The optical sensitivities of single cones are similar in both birds (budgerigar 0.27 μm2 sr, Bourke's parrot 0.25μm2 sr) but Bourke's parrots have more (cone to rod ratio,1.2:1.0), thinner (2.8 μm) and longer rods (18.5 μm) than budgerigars(2.1:1.0, 3.4 μm, 13.3 μm). Bourke's parrots thus have an eye type that,with a flexible pooling mechanism, allows for high resolution or high absolute sensitivity depending on the light conditions. The results nicely agree with the activity patterns of the birds, Bourke's parrots being active during the day and in twilight while budgerigars are not normally active before sunrise and after sunset. However, Bourke's parrots have fewer cones than budgerigars,which implies that a smaller number of cones are pooled within each retinal integration area. That could explain why Bourke's parrots have a higher intensity threshold of colour vision than budgerigars. Furthermore, the study emphasises the need to expand the sensitivity measure so that photoreceptor integration units are used rather than single receptors.
Colour vision allows for the discrimination between stimuli with different spectral compositions (colours) even if they have the same brightness. Colour vision is also more reliable than achromatic vision in variable light conditions because of colour constancy(Campenhausen, 1986; Vorobyev et al., 2001; Johnsen et al., 2006). However, for most vertebrates, colour vision functions only in bright light(photopic conditions) since cones are less sensitive than rods, which mediate spatial vision in dim light (scotopic conditions).
In mesopic conditions both rods and cones operate(Kelber et al., 2003), and the lower radiance limit of the mesopic range is the threshold of colour vision. To our knowledge, this threshold is known in only two vertebrate species;human, Homo sapiens (Wyszecki and Stiles, 1982) and horse, Equus ferus caballus(Roth et al., 2008), which both lose colour vision around 0.02 cd m–2.
Nothing is known about the mesopic intensity range in birds. This is unfortunate since birds are highly dependent on their visual sense and use colour vision in vital behaviours such as foraging and mate choice [e.g. Cuthill et al. (Cuthill et al.,1999) and references therein]. The limits of colour vision are therefore valuable information for biologists who seek to understand how birds detect and respond to important cues in changing light environments.
The discrimination of colours is possible if the incident light is absorbed by at least two spectrally different cones generating signals that are compared by opponent mechanisms (Kelber et al., 2003). Colour vision is lost when the number of photons reaching the cones is insufficient to generate strong enough signals relative to the noise of the system (Vorobyev,1997). The intensity threshold of colour vision thus depends on the photon capture efficiency of the eye, its optical sensitivity. The optical sensitivity of the eye for an extended light source can be estimated by the f-number of the eye (Martin,1983), which is the focal length divided by the diameter of the entrance pupil (the pupil size as viewed through the cornea). The f-number describes the brightness of the retinal image; a small f-number indicates a smaller but brighter retinal image compared with eyes of the same size but with higher f-numbers. However, a more informative measure is the optical sensitivity of single photoreceptors that combines the eye's f-number with the size of the photoreceptors and the density of the visual pigments(Warrant and Nilsson, 1998). Finally, to understand sensitivity on the behavioural level, the retinal organisation is of importance since the signals can be strengthened if the outputs from several photoreceptors are pooled, i.e. if the effective receptor diameter is enlarged (Land and Nilsson,2002; Warrant,2008).
In this study we used behavioural tests to determine the absolute intensity thresholds of colour vision in budgerigars (Melopsittacus undulatusShaw 1805) and Bourke's parrots (Neopsephotus bourkii Gould 1841). These are both Australian parrot species with similar body sizes and lifestyles. However, budgerigars become active at sunrise and end their activity around sunset (Wyndham,1980), whereas Bourke's parrots are active both during the day and in twilight (MacGillivray,1927; Davies, 1972; Fisher et al., 1972; Collar, 1997). Davies(Davies, 1972) reported that Bourke's parrots start to arrive at waterholes to drink more than 1 h before sunrise and that maximum numbers of Bourke's parrots were observed at the waterholes 35 min before sunrise, which is before the beginning of civil twilight and corresponds to a luminance level of about 0.03 cd m–2. Fisher et al. (Fisher et al., 1972) reported very similar observations. Other sources claim that Bourke's parrots can be seen flying well after sunset, sometimes as late as 21:00 h (Collar, 1997),which is well after the end of astronomical twilight.
This might suggest that Bourke's parrots have evolved more sensitive eyes and if so, possibly also colour vision in dimmer light. Such an adaptation could benefit Bourke's parrots considering that the spectral composition of light changes strongly during twilight, which makes object brightness an unreliable cue for animals without colour or lightness constancy(Johnsen et al., 2006).
Bourke's parrots have larger eyes and larger pupils than budgerigars but,as discussed, these characters reveal little about sensitivity(Martin, 1983; Land and Nilsson, 2002). Therefore, we also performed morphological investigations so that the behaviourally determined thresholds can be related to the optical sensitivities at the single photoreceptor level.
MATERIALS AND METHODS
Healthy wild-type coloured birds were acquired from local breeders and fed with appropriate seed mixes supplemented with vitamins, minerals, fruits, and vegetables. During experimental days, the birds got all their food in the behavioural experiments except for a small food portion at the end of each day. The animals were kept in conditions that followed the ethical guidelines from the Swedish Board of Agriculture and all experiments were approved by the Board (M18-07, M206-07).
The birds were tested in a rectangular cage measuring 500 mm×960 mm×740 mm (width, length, height) placed in a completely dark room. Four adjustable fluorescent tubes (Biolux L18W/965; Osram, München, Germany)and 30 adjustable light emitting diodes (white, 5 mm, 20 cd; Kjell &Company, Malmö, Sweden) placed above the cage produced a homogenous light of 252–11.5 cd m–2 (fluorescent tubes and LEDs) and 3.12–0.0156 cd m–2 (LEDs; Fig. 1), measured as the light reflected from a white standard positioned at the floor of the cage. One of the short end walls of the cage consisted of a black board with two horizontally aligned windows 55 mm×55 mm (313 mm apart) for stimulus presentation. These windows could be closed by a removable black cover. The birds observed the stimuli from a start perch at 827 mm distance, from which each stimulus covered 3.8° of the birds' visual fields. Boxes with hatches and feeder perches were placed below each stimulus window and these were filled with seeds from the birds' feeding mixes.
Behavioural test procedure
The birds were trained to choose between positive green and negative blue stimuli presented simultaneously. The tests started with a two-tone auditory signal after which the black cover in front of the stimuli was removed. Initially, the feeding box of the positive stimulus was opened directly after the stimuli became visible and the bird was free to eat. After a few days the birds had to approach the positive stimulus in order to get the reward and finally, only contact with the correct feeder perch released the reward. The reward consisted of access to the food for 4–6 s after which the feeding box was closed and the stimuli covered. The bird had to return to the start perch in order to initiate a new presentation. Incorrect choices were not punished but only followed by the return of the stimuli cover. When the birds had learned the procedure they were trained in bright light with test sets of 20 presentations each until they reached or exceeded 75% correct choices in two subsequent sets.
The tests started at the highest luminance level (252 cd m–2). Before each session, the birds were allowed to adapt to the luminance level of that test for 10–15 min. After the completion of a test set, the light was dimmed, and a new session was initiated after the adapting interval. Each bird completed 60 choices at each of six to nine luminance levels and the intensity threshold of colour vision was determined by calculating (by interpolation of the choice frequency curves) the luminance level at which the birds made 65% correct choices (binominal test, N=60, P<0.05).
The stimuli were printed on white Munken Pure Copy paper (Artic Paper Munkedals AB, Uddevalla, Sweden) with an ink-jet printer (PIXMA pro9000;Canon, Tokyo, Japan). The radiance of the stimuli was measured with a spectroradiometer (RSP900-R; International Light, Peabody, MA, USA). We assume that the parrots use four types of single cone (the ultraviolet-sensitive, the short wavelength-sensitive, the middle wavelength-sensitive, and the long wavelength-sensitive cone) for colour vision, and double cones and rods mediate spatial vision (Maier and Bowmaker, 1993) (for a review, see Osorio and Vorobyev, 2005). The spectral sensitivities of the budgerigar photoreceptors were modelled using absorbance data of the visual pigments, the oil droplets and the ocular media (Bowmaker et al., 1997; Lind and Kelber, 2009)together with a visual pigment template(Govardovskii et al., 2000)and an oil droplet model (Hart and Vorobyev, 2005). The photoreceptors of Bourke's parrots are probably tuned like those of the budgerigar since photoreceptor sensitivities of birds are very similar within families(Hart, 2001). The quantum catch of the photoreceptors and the colour loci within receptor space(Fig. 1) were calculated as described elsewhere (Balkenius and Kelber,2004; Roth et al.,2008).
Using the quantum catch of double cones and rods, we constructed four brightness versions of the green and five versions of the blue stimulus(Fig. 1). These stimuli were arranged in 10 pairs and the green colour was brighter for the birds in five of the pairs. Each stimulus pair was used twice in a pseudorandom order resulting in a 20 choices test set. This set was used with random alternation in presentation order (stimuli pairs 1,2...20 or 20,19...1) and presentation side (left, right). By this procedure we ensured that all stimuli could be discriminated, that the positive stimuli were presented equally often to the left and to the right and that choices based on stimuli brightness would result in 50% correct choices.
The steady state pupil dynamics of three budgerigars and three Bourke's parrots were measured in the same trial cage that was used in the behavioural experiments. Two fluorescent lights (Repti Glo 2.0 Compact, 26W; Exo Terra,Holm, Germany) were added in order to expand the luminance range to 455 cd m–2. The birds were placed on a plastic perch with scale markings and filmed 30 s at each light level with a video camera (DCR-TRV 730E; Sony, Tokyo, Japan) equipped with infrared light emitting diodes (LEDs). The birds were adapted to each light level for 15 min before the recordings. Three frames, in which the birds were looking directly into the camera, were extracted using Adobe Premier Pro software (Adobe systems; Mountain View, CA,USA).
Two Bourke's parrots and two budgerigars were anaesthetized with carbon dioxide and decapitated. One eye from each of the birds was excised and used for the investigations of the retina. Small circular portions of the retina from the posterior pole of the eyes were cut out and fixed in 2.5%glutaraldehyde and 0.15 mol l–1 cacodylate buffer overnight. The samples were then dehydrated stepwise with ethanol and embedded in Epon. Three sections from each sample (1–2 μm thickness, 50 μm intervals) were mounted on glass slides, dyed with Azur II–Methylene Blue and photographed together with a scale using a light microscope. The photoreceptor dimensions in the micrographs were measured with a free software program, ImageJ v.1.41 (Rasband,2008).
Double cones, single cones and rods were identified in the retinas. The dimensions of the single cones were used for the optical sensitivities of the colour vision systems. The rod dimensions were used to determine the sensitivities of the eyes in scotopic conditions. The double cones were only included in the estimation of the cone to rod ratios and each double cone was counted as one.
One eye from each of two Bourke's parrots and both eyes from one budgerigar were, while still intact inside the skulls, quickly frozen at –80°C directly after decapitation. The eyes were then placed in a cryostat and sectioned horizontally. Photographs of the eyes and a scale were taken at 100μm intervals in order to obtain the largest eye diameters. The eye dimensions in the photographs were measured with ImageJ(Rasband, 2008). Schematic eyes were constructed using mean parameters and Gullstrand's schematic eye model as described by Land and Nilsson(Land and Nilsson, 2002). For these calculations we set the refractive indices of the aqueous and vitreous humours to 1.337, the value found in other birds(Martin, 1982; Martin, 1986; Martin et al., 2001). We assume emmetropic schematic eyes, which means that the refractive indices of the lenses can be approximated as being the only adjustable parameter.
An examination of seven pigeon (Columba livia) eyes with the same method yielded a mean focal length of 7.96 mm and an axial length of 11.96(O.L., unpublished data). These values are very close to those obtained in earlier studies (focal length=7.91, axial length=11.62) in which other methods, including refractive index measurements, were used(Marshall et al., 1973; Martin and Brooke, 1991).
The intensity threshold of colour vision
All birds learned to associate the green stimuli with the food reward and mean choice frequencies indicate that both species readily discriminate the colours at higher light levels (Fig. 2). Mean intensity threshold of colour vision was higher in Bourke's parrots (0.4 cd m–2) than in budgerigars (0.1 cd m–2; Fig. 2). The variation in the data of the Bourke's parrots was large compared with that of the budgerigars; one Bourke's parrot lost colour vision at approximately the same luminance level as the budgerigars whereas another Bourke's parrot performed much worse than all the other birds(Fig. 2). If the worst performing individual is excluded from the analysis, mean intensity threshold of colour vision decreases to 0.2 cd m–2 for Bourke's parrots, which is still at a higher light intensity than the threshold of the budgerigars.
In bright light (455 cd m–2), budgerigars and Bourke's parrots had similar entrance pupil diameters (2.3 mm; Fig. 3). In dimmer light, the pupil of Bourke's parrot widened more and at 0.0156 cd m–2,it was 1 mm larger than that of the budgerigar(Fig. 3). There also appears to be a difference in the mechanisms controlling the pupil dynamics. Although there is a log-linear relationship between the budgerigar pupil diameter and intensity within the tested luminance range, this relationship follows a sigmoid curve in Bourke's parrots with a very steep slope between 11.5 and 3.12 cd m–2 (Fig. 3).
The lens refractive indices of the schematic eyes were adjusted assuming emmetropic eyes. This yielded lens refractive indices of 1.439 and 1.458 for budgerigars and Bourke's parrots respectively, and focal lengths of 4.0 mm for budgerigars and 5.0 mm for Bourke's parrots(Fig. 4C,D, Table 1). The minimum f-number of budgerigars was 1.33 and 1.25 for Bourke's parrots(Table 1).
Retina morphology and optical sensitivity
The intraspecific variation of the photoreceptor dimensions is quite high but still negligible for the calculation of the optical sensitivities since this is a measure interesting only in orders of magnitude(Warrant and Nilsson, 1998). The data from each species were therefore pooled(Table 1). We could only find small differences in cone dimensions between the species but the rods in Bourke's parrots were thinner (2.8±0.4 μm; mean ± s.d.) and longer (18.5±2.2 μm) than in the budgerigar (3.4±0.5 and 13.3±1.8 μm). Bourke's parrots also have a larger proportion of rods in their retinas (cone to rod ratio 1.2:1 compared with budgerigars 2.1:1; Table 1, Fig. 4E,F).
Because the f-numbers and the cone properties are so similar in both species it is not surprising to find that the optical sensitivities for single cones are almost identical (Table 1). This is true also for the single rods because even though the absolute sensitivity of Bourke's parrots is increased by the long rods, it is decreased as much by their small diameters(Table 1, Fig. 4E,F). All calculated optical sensitivities refer to single photoreceptors; the effect of receptor pooling remains unknown.
The intensity limit of colour vision
Our initial hypothesis, that Bourke's parrots might see colours in dimmer light conditions than budgerigars, is not supported by our findings. On the contrary, the absolute intensity threshold of colour vision is slightly higher in Bourke's parrots than in budgerigars(Fig. 2). This is true even if the worst performing Bourke's parrot is excluded from the analysis (see Results). Moreover, both parrots lose colour vision around the end of civil twilight, whereas mammals still see colours at moonlight intensities(Fig. 2)(Wyszecki and Stiles, 1982; Roth et al., 2008). The optical sensitivities of single cones cannot explain these differences since these values are very similar for both birds, and higher for the birds than for the mammals (Table 1). However, a more complete description of the sensitivity of the avian eye would also include the filtering of light through the retinal oil droplets, the integration time of the receptors and finally, the rate of receptor pooling.
Oil droplet filtering and integration time
When comparing birds with mammals it is important to recognize that birds have coloured oil droplets through which incident light is filtered before it reaches the cone outer segments (reviewed by Hart, 2001). The absorption by these oil droplets decreases the amount of light reaching the cones by approximately 50% (Vorobyev et al.,1998). Moreover, budgerigar photoreceptors have shorter integration times (14 ms) than mammals (25 ms)(Ginsburg and Nilsson, 1971; Jarvis et al., 2002) at comparable light intensities, which further decreases the relative sensitivity of the cones. Calculations including these effects result in budgerigar single cone sensitivities that are similar to those of humans, ∼0.08μm2 sr per integration time(Roth et al., 2008). Humans still see colour in dimmer light (Roth et al., 2008), which implies that yet other factors are involved.
For the comparison of the sensitivities between the parrots, the filtering properties of oil droplets and the receptor integration times are likely to be similar and thus less important. Instead, we conclude that the retinal organization and the pooling of photoreceptors are of more influence. Bourke's parrots have a retina with a cone to rod ratio of 1.2:1(Table 1) that is comparable to values found in earlier studies of birds that are active both during twilight and in daylight (Rojas et al.,1999). The cone to rod ratio of budgerigars is higher, 2.1:1(Table 1), which appears to be typical for diurnal birds (Rojas et al.,1999; McNeil et al.,2005). Furthermore, the rods of Bourke's parrots are longer and of smaller diameters than those of budgerigars.
These findings agree well with the activity patterns of the birds, the rod-rich retina of Bourke's parrots most probably enables vision in dimmer light than is possible for budgerigars. The small diameters of the rods probably gives them a high flexibility so that they can use a lower level of pooling at higher light intensities to increase spatial resolution and a higher level of pooling in dimmer light to increase sensitivity. This is different in budgerigars, which are normally not active at light intensities lower than their intensity threshold of colour vision(Fig. 2)(Wyndham, 1980).
However, the higher rod density comes at a cost, a relatively low density of cones. As a result, Bourke's parrots cannot pool as many cones as budgerigars within each retinal integration area without loosing spatial resolution, which would explain the differences in the intensity thresholds of colour vision between the species.
Dimensionality of colour vision
So far we have discussed sensitivity at the receptor level, however, colour vision in dim light may also be limited by post-receptor mechanisms. Colour opponent mechanisms are less tolerant to noise than the additive systems of monochromats (Vorobyev, 1997). Noise increases with decreasing light intensity(Warrant, 2008) and it has been suggested that `the number of distinguishable colours is proportional to the light intensity raised to the power of n/2'(Vorobyev, 1997), where n is the dimensionality of the system (monochromats n=1,dichromats n=2 etc.) Thus, tetrachromatic parrots should have higher intensity thresholds of colour vision than dichromatic horses and trichromatic humans if everything else is similar. It is, however, not likely that all other visual properties are constant in as different animals as birds and mammals. Therefore, we are left to conclude that further investigations within this field require more experimental data before general trends and concepts can be disclosed.
Very few measurements of the steady state pupil dynamics in birds have been performed. The pupil dynamic of Bourke's parrots(Fig. 3) is similar to those of blue-fronted parrots (Amazona aestiva) and grey parrots(Psittacus erithacus), studied by Lind and colleagues(Lind et al., 2008), whereas the less active pupil of budgerigars (Fig. 3) is more similar to that of Ural owls (Strix uralensis)(Lind et al., 2008). This difference might reflect the light conditions to which the birds must adapt. Animals with activities that span over a great range of luminance levels, e.g. Bourke's parrots (Davies,1972), or activity in quickly changing light conditions such as flights in and out of dense vegetation, e.g. blue-fronted parrots and grey parrots (Juste, 1996; Seixas and Mourão,2002), probably require more responsive pupils and larger pupil dynamic ranges than strictly diurnal, e.g. budgerigars(Wyndham, 1980), or strictly nocturnal activities, e.g. Ural owls(Martin, 1990) (cf. Arrese, 2002).
Bourke's parrots are active at luminance levels at which they cannot be considered to use colour vision. Instead, they use achromatic vision and their eyes do indeed seem to be well adapted for this purpose. However, the visual guidance for the behaviours that Bourke's parrots express in dim twilight is probably of a rather simple nature since the variable twilight light spectrum makes brightness differences a less reliable cue for achromatic vision(Johnsen et al., 2006). By contrast, budgerigars are mostly roosting at light conditions in which colour vision is not functioning.
The use of the optical sensitivity for single photoreceptors to predict the intensity threshold of colour vision is a rather crude method. The sensitivity measure should be expanded to also include receptor transduction mechanisms,receptor pooling rates, and possibly also post-receptor processing in order to make it a more reliable predictor of visual thresholds.
LIST OF ABBREVIATIONS
We thank Eva Landgren and Rita Wallén for technical assistance and members in the Lund Vision Group for helpful discussions. We also thank three anonymous reviewers for helpful comments on the manuscript. Financial support from the Swedish Research Council is gratefully acknowledged.