ABSTRACT
Vision is an important sensory modality in birds, which can outperform other vertebrates in some visual abilities. However, sensitivity to achromatic contrasts – the ability to discern luminance difference between two objects or an object and its background – has been shown to be lower in birds compared with other vertebrates. We conducted a comparative study to evaluate the achromatic contrast sensitivity of 32 bird species from 12 orders using the optocollic reflex technique. We then performed an analysis to test for potential variability in contrast sensitivity depending on the corneal diameter to the axial length ratio, a proxy of the retinal image brightness. To account for potential influences of evolutionary relatedness, we included phylogeny in our analyses. We found a low achromatic contrast sensitivity for all avian species studied compared with other vertebrates (except small mammals), with high variability between species. This variability is partly related to phylogeny but appears to be independent of image brightness.
INTRODUCTION
Vision is an important sensory modality in many bird species, as evidenced by their large eye relative to body mass ratio (Brooke et al., 1999). Indeed, some birds outperform all vertebrates in the ability to resolve fine details in a visual scene (hereafter called visual spatial resolution), with some diurnal raptors having the highest visual spatial resolution in the animal kingdom (see Mitkus et al., 2018; Potier et al., 2020 for reviews). However, high contrast in a visual scene is also needed to resolve fine details. Therefore, not only the visual spatial resolution, but also the contrast sensitivity (as well as other visual capabilities) of a species has to be studied to fully characterize and understand a visual system.
Visual pathways can be distinguished into chromatic and achromatic pathways. A distinction therefore is also made between chromatic and achromatic contrast sensitivity (see Lind and Kelber, 2011 for details on chromatic contrast sensitivity). Achromatic contrast sensitivity is the ability of an organism to distinguish luminance differences between two objects, or an object/surface and its background/environment. As with spatial resolution, achromatic contrast sensitivity decreases when ambient light levels decrease (Reymond and Wolfe, 1981). High achromatic contrast sensitivity enables animals to detect and discriminate subtle differences in luminance, which can help them identify important visual cues in their environment. In particular, animals need high contrast sensitivity to detect objects in fog and to see shadows (Land and Nilsson, 2012).
In spite of its importance, contrast sensitivity remains understudied in vertebrates, including birds. To our knowledge, achromatic contrast sensitivity has been studied in only 12 bird species (see Table 1 for information about species; Reymond and Wolfe, 1981; Porciatti et al., 1989; Türke et al., 1996; Lee et al., 1997; Schmid and Wildsoet, 1998; Hodos et al., 2002; Ghim and Hodos, 2006; Harmening et al., 2009; Jarvis et al., 2009; Lind and Kelber, 2011; Lind et al., 2012; Shi and Stell, 2013; Haller et al., 2014; Potier et al., 2018). These studies reveal that achromatic contrast sensitivity varies between avian species, but generally is lower than in medium-sized and large mammals (see Table 1 for information about avian species; Bisti and Maffei, 1974; De Valois et al., 1974; Merigan, 1976; Jacobs, 1977; Langston et al., 1986). However, accounting for differences in the results of these studies on the comparison of birds' achromatic contrast sensitivity can be difficult because various methodologies are used to evaluate this ability. In particular, Hodos et al. (2002) have revealed that the pattern electroretinogram (PERG) method underestimates the contrast sensitivity of birds by 53% compared with a behaviour-based method. The phylogenetic diversity of the investigated species in these 14 studies is limited to only six orders. There is a considerable need for further research examining achromatic contrast sensitivity in birds across a broader range of species and phylogenetic groups, with more species not classically studied in the laboratory. This would allow a comparative analysis among different avian species and would make it possible to study whether properties of contrast sensitivity are derived from a common evolutionary ancestor or whether they are a result of specific visual adaptations to the environment (Hart, 2001; Ryan et al., 2017).
We obtained a contrast sensitivity function (CSF) for the wide-field moving gratings using the optocollic reflex (OCR) of 32 species from 12 phylogenetic orders, and with different ecologies. From these CSFs we extracted the peak contrast sensitivity along with the corresponding spatial frequency and the high frequency cut-off. We also intended to explore the origin of the variation in peak contrast sensitivity. As low contrasts require a high level of light to be seen (Land and Nilsson, 2012), we aimed to investigate a potential positive relationship between contrast sensitivity and retinal image brightness. To determine the extent to which phylogenetic relatedness influences achromatic contrast sensitivity, we performed the analyses with and without control for phylogeny.
MATERIALS AND METHODS
Experimental subjects
We tried to estimate the contrast sensitivity of 137 captive individuals from 35 species and 12 orders belonging to different bird-keeping facilities in France, including two research institutes, five zoological parks and one wildlife hospital (Table S1). All individual birds had reached their adult size and were fed daily by their usual keeper. Birds of 33 species were housed in outdoor aviaries, subjected to natural photoperiod and light conditions. Two other species, the blue tits (Cyanistes caeruleus) and zebra finches (Taeniopygia guttata), were kept in indoor aviaries. Blue tits were subjected to natural photoperiod, while the zebra finches' room was artificially lit for 12 h a day (see supplementary Materials and Methods 1 for more details). Species and number of individuals studied varied according to bird availability at the time of the experiment.
Experiments were carried out in accordance with the European Union directive on the protection of animals used for scientific purposes (2010/63/EU) and French legislation. The protocol and species studied were approved by a local French ethics committee in animal experimentation and authorized by the French general directorate of research and innovation (APAFIS# 31775-2021041311196892 v5). In agreement with French law, birds were handled by their usual keepers under the permits issued to different parks (Réserve Africaine de Sigean, certificate to maintain birds 11-264 delivered 23 February 2018 to Antoine Joris, the director of the zoological park; Espace Rambouillet; national certificate to maintain birds delivered to the director of the falconry, Romuald de Romans, 3 July 1997; Le Grand Parc du Puy du Fou; national certificate to maintain birds 17-85-154; Zoodyssée, national certificate to maintain birds 78/2022/D102; Les Ailes de l'Urga, national certificate to maintain birds DDPP-19-108).
For reasons of bird welfare, the duration of the experiment was limited to a maximum of 3 h per individual. If necessary, the experiment could be stopped at any time and these 3 h could be achieved in several sessions over several days.
Assessing the contrast sensitivity of birds
Experimental set-up
Birds were placed either in a glass cylinder with a wooden perch or in a transparent U-shaped box (Fig. 1A,B) according to their body size and morphology.
The glass cylinder of 10 cm in diameter, 25.3 cm in height and with 2 mm thick wall was used for small birds (Table S1). The top of the cylinder was locked by a removable 0.3 mm thick PVC sheet, which was perforated to allow airflow.
Larger species were placed in a large U-shaped box (Table S1), 100×50×150 cm (length×width×height), with a black wooden floor and ceiling, and a 2 mm thick polycarbonate wall. A removable wooden door locked the box at the rear (Fig. 1). According to the bird's body size, the door could be moved forward/backward to adapt the size of the box. Different wooden boards could be added on the floor to optimize height of the box for each species (Fig. 1). The door and the ceiling were made of two wooden boards spaced 1 cm apart: a perforated panel to let the air pass through and a solid and smaller panel so that the bird could not see through the holes of the other panel. A wooden perch was added in the box if needed, based on ecology of species (Fig. 1). All wooden surfaces were painted matte black (Liberty black, matte, GoodHome).
Species of intermediate size were placed in a medium size U-shaped box [60×25×70 cm (l×w×h); Table S1] with all possible adjustments as described above for the large box.
Both experimental boxes and cylinder were placed at the centre of an octagon, or an arc of seven facets following the shape of a dodecagon, made with computer monitors (Acer KG281KAbmiipx, 28″, 62×37 cm display area, 3840×2160 pixels, 60 Hz, Acer) in portrait orientation. The height of the boxes/cylinder was adjusted so that the head of the bird was levelled with the centre of the screens (Fig. 1). The glass cylinder was placed in the centre of an octagon of 8 computer monitors (Fig. 1D) so the stimulus covered 360×69.5 deg of the bird's visual field (horizontal×vertical extent). The U-shaped boxes were placed in the centre of an arc formed by seven computer monitors, following the shape of a dodecagon (Fig. 1C) so the stimulus covered 210×48.4 deg of the bird's visual field (horizontal×vertical extent). Stimuli of such size are large enough to elicit a bird's OCR (Schmid and Wildsoet, 1998).
To assess the possible impact of the polycarbonate wall of the U-shaped box, we repeated the measurements on falconry-trained American kestrels (Falco sparverius), with individuals able of staying on a perch in front of the monitors without any box. The bird was placed on a black perch in the centre of an arc of seven computer monitors following the shape of a dodecagon, with tarsus loosely fastened to the perch with leather jesses. Contrast sensitivity functions based on two American kestrels, tested in the setup with and without the polycarbonate wall, were similar (Fig. S1), although no statistical test could be performed.
The experimental setup was placed in an opaque black cloth tent [260×250×190 cm (l×w×h)]. The base of the monitors and the floor were also covered by an opaque black cloth to minimize possible disturbing visual cues. The experimenter stayed outside the tent during the entire experiment.
Two video cameras (25 fps; Sony FDR-AX43 4K) were placed above the monitors with an inclination towards the ground of 30–40 deg and approximately separated by 90 deg (both at 45 deg from the bird's body axis). These cameras allowed to observe and record the bird's behaviour during the experiment. One of these cameras was connected to a screen located outside the tent so that the experimenter could monitor the bird during the entire procedure. Because small passerines can move their heads rapidly, the head movement characteristic of the OCR was difficult to see with the two cameras. Therefore, a third camera (25 fps; Mobius 1080 Action Camera, Lens A2, Mobius) was placed on the PVC sheet to record the passerines from above.
Stimuli
A vertical sinusoidal achromatic grating, at variable angular velocities, spatial frequency and contrast, was generated in MATLAB (R2020b) using Psychtoolbox (Brainard, 1997). This grating was mirrored on each screen of the experimental setup using two HDMI splitters (8-port, 4 K, 60 Hz, ST128HD20, StarTech). A bird's OCR was elicited with an angular stimulus velocity ranging from 5 to 30 deg s−1, depending on species (Table S1), a spatial frequency ranging from 0.08 to 2.83 cycles deg−1 and a Michelson contrast ranging from 0.02 to 0.99 (Table S2). The contrast was evaluated by measuring the luminance of darkest and lightest area of the grating (converted into a square wave signal to facilitate measurement) at the centre of the monitor with a lightmeter (Hagner ScreenMaster, B. Hagner, Sweden) (Table S2). As far as we know, there is no documentation allowing us to define the most appropriate angular velocity for each species. Thus, the angular velocity used for each species was determined in preliminary tests by the experimenter; it corresponded to the angular velocity at which the OCR was most pronounced using maximum contrast at different spatial frequencies used for the species. The angular velocity used was the same within a species, except for Eurasian stone curlew (Burhinus oedicnemus) (Table S1).
Spatial frequencies were determined by the experimenter for each species at the beginning of the experiment in order to be sure to frame the spatial frequency range corresponding to the contrast sensitivity curve.
The stimulus was presented for 20 or 30 s, depending on species (Table S1). A duration of 20 s was generally sufficient to observe an OCR (Schmid and Wildsoet, 1998; Shi and Stell, 2013; Wagner et al., 2021). The duration of the signal was therefore determined by the experimenter beforehand at maximum contrast for different spatial frequencies. If 20 s were not enough to observe an OCR, the duration of the stimulus presentation was set to 30 s.
The direction of the stimulus (clockwise or anticlockwise) was changed in a pseudo-random order, i.e. the number of consecutive trials in the same direction was limited to three.
The illuminance at the bird's head position was 120 lx inside the U-shaped box, 321 lx inside the cylinder, and the average stimulus luminance was 101±2 cd m−2 [mean±s.e.m.; measured with a lightmeter (Hagner ScreenMaster, B. Hagner, Sweden)].
Experimental procedure
The experimental session started 5–20 min after the bird was placed in the setup depending on the individual's habituation to its new environment. The presence of OCR was assessed ‘by eye’ by the experimenter (C. B.), similarly as in Schmid and Wildsoet (1998) or Shi and Stell (2013). At the end of the experiment, the two or three views were combined into one video with multiple views (using the Multi View function, PlayMemories Home, Sony software), which was then divided into 10 min long segments. The experimenter (C.B.) was thus able to view the trials post hoc and validate or not the presence of OCR blindly, i.e. being not aware of the angular velocity, spatial frequency, contrast of the stimulus and individual. Only the results from the blind double-checked trials were used in this study. The experimenter (C.B) was the same in the whole experiment to avoid observer bias.
Each trial consisted of a specific combination of angular velocity, spatial frequency and contrast, separated by 5 s pause (or longer if the bird did not look at the monitors), during which a homogeneous grey (isoluminant to the sinusoidal gratings) was displayed on the monitors.
Each session began with a stimulus grating having maximum contrast (Michelson contrast of 0.99) to ensure that the bird was expressing an OCR detectable by the experimenter. The contrast was then decreased and increased in a stepwise manner to find a contrast threshold where birds no longer exhibited the reflex. The typical sequence started with 0.99 contrast, and then half decreased to 0.50 then to 0.25, etc., until the bird no longer expressed an OCR. The contrast was then increased halfway between the last two steps. If an OCR was expressed it was then lowered again to the midpoint between the last two steps, until finding the lowest contrast at which the OCR was visible. Following Shi and Stell (2013) method, we used a ‘four of five’ criterion. Each specific combination was presented 5 times and the response was considered reliable if the OCR was elicited at least 4 times out of the 5 trials. Thus, the lowest contrast with a validated 4 out of 5 criterion was defined as the threshold contrast. Once this threshold was reached, the experiment was repeated with the same angular velocity, but with a different spatial frequency. To obtain a contrast sensitivity function, the experiment was repeated at 4–10 different SFs for each species.
Retinal image brightness variable
Low contrasts require a high level of light to be seen, making the contrast sensitivity dependent on the eye's sensitivity to light (Land and Nilsson, 2012). As in Lisney et al. (2012a,b, 2013) we used the corneal diameter (CD) to the axial length (AL) ratio as a proxy of the retinal image brightness. A large CD allows for a larger pupil size, which raises the number of photons that reach the retina, improving visual sensitivity (Hall and Ross, 2007). The AL is a proxy of focal length and is therefore inversely proportional to retinal image brightness (Kirk, 2006). Thus, the CD:AL ratio is positively linked to the retinal image brightness. High values of the CD:AL ratio are consistently found in animals that primarily are active and live in dim light conditions compared with animals that live in higher light levels (Kirk, 2004, 2006; Hall and Ross, 2007; Veilleux and Lewis, 2011; Lisney et al., 2012a,b).
When available, CD and AL reported by Ritland (1982) were used for calculating the CD:AL ratio for each species. When these data were not available for a species, CD and AL measured by Ritland (1982) on related species were used (see supplementary Materials and Methods 2 for details), except for the white-tailed eagle (Haliaeetus albicilla) whose data were provided by M.M. from an unpublished study.
Data analysis
All analyses were performed with R v. 4.1.2. (https://www.r-project.org/). Contrast sensitivities were expressed as the inverse of the Michelson contrast at a threshold. A double exponential function was fitted to the contrast sensitivity data with the method of least squares (Uhlrich et al., 1981) to produce a CSF for each species (see Fig. S1 for an example of the CSF compared with measurement points used to create the CSF). For each species, a single CSF was calculated using all the contrast thresholds at each SF, for all the individuals (i.e. contrast thresholds were not averaged per SF; Ghim and Hodos, 2006).
We performed a linear regression to test for a potential positive relationship between the peak contrast sensitivity and the CD:AL ratio. Phylogenetically related species often share similar characteristics or abilities. We tested the relationship between peak contrast sensitivity and CD:AL ratio with a phylogenetic linear regression (R package phylolm; https://CRAN.R-project.org/package=phylolm; Ho and Ane, 2014) to separate ecological processes resulting from a common ancestry (Veilleux and Kirk, 2014). Taking phylogeny into account allows to control for the potential influence of evolutionary history on the traits studied. According to our data, we used the most probable tree (consensus.edges function from the phylolm R package) between 1000 phylogenetic trees extracted from Birdtree with the Hackett backbone method (Jetz et al., 2012). The variation in the peak sensitivity explained by phylogeny was extracted using the R function R2_pred (Ives and Li, 2018; Ives, 2019) on a null phylogenetical regression.
Additional exploratory linear relationships have been tested between the peak contrast sensitivity and (1) the high frequency cut-off, (2) the position of peak sensitivity, (3) the normalised bandwidth of CSF, (4) the AL or (5) the body mass of species (Table S3, Fig. S2). Similarly, we explored the linear relationships between the CD:AL ratio and (1) the high frequency cut-off, (2) the position of peak sensitivity or (3) the normalised bandwidth of CSF (Table S3, Fig. S2). To determine the bandwidths of each CSF, curves were normalized and SFs were converted to octaves (when converted into octaves, the abscissa value of each CSF data point corresponds to the logarithm base 2 of the SF of the point divided by the SF of the contrast sensitivity peak). Bandwidths of CSFs were measured at half the maximum of the peak (i.e. the middle in the ordinate between the value of the contrast sensitivity peak and Michelson contrast−1=1).
RESULTS
For the 137 birds investigated in this study, we were able to determine a contrast threshold for at least one SF of 103 individuals representing 32 species (see supplementary Materials and Methods 3 for details about individuals that did not respond). The contrast threshold was estimated for 4–10 SFs per species, with 1–10 individuals per SF (all data are available at https://doi.org/10.7910/DVN/05M6Z1).
Contrast sensitivity function
A contrast sensitivity function was calculated for 32 species (Fig. 2). From the CSFs, we extracted the peak contrast sensitivity (expressed as Michelson contrast−1), the position of peak sensitivity in the spatial frequency domain (hereafter position of sensitivity peak), as well as the high frequency cut-off (Table 1).
Contrast sensitivity in Accipitriformes was variable, with the peak sensitivity ranging from 11.9 to 25.6, with position of sensitivity peak from 0.7 to 1.1 cyc deg−1 (Table 1; Fig. 2A). The high frequency cut-off in Accipitriformes varied from 2.3 cyc deg−1 for the bald eagle (Haliaeetus leucocephalus) to 17.8 cyc deg−1 for the griffon vulture (Gyps fulvus) (Table 1; Fig. 2A).
The four Falconiformes species studied had similar contrast sensitivity, with the peak sensitivity ranging from 20.2 to 28.6, with position of sensitivity peak from 0.7 to 1.2 cyc deg−1 (Table 1; Fig. 2B), and a high frequency cut-off from 3.6 cyc deg−1 for the Eurasian hobby (Falco subbuteo) to 5.5 cyc deg−1 for the American kestrel (Falco sparverius) (Table 1; Fig. 2B).
We recorded a greater diversity of contrast sensitivity in Passeriformes, ranging from 12.8 for blue tits, with position of sensitivity peak at 0.4 cyc deg−1, to 29.9 for northern raven (Corvus corax), with position of sensitivity peak at 1.1 cyc deg−1 (Table 1; Fig. 2C). The high frequency cut-off also varied greatly, ranging from 1.7 cyc deg−1 for the zebra finch to 5.0 cyc deg−1 for the northern raven (Table 1; Fig. 2C).
In Anseriformes, the northern shoveler (Spatula clypeata) and greylag goose (Anser anser) had similar CSFs with low peak contrast sensitivity compared with other bird species (5.7 and 5.8, respectively), whereas northern pintail (Anas acuta) had a peak sensitivity of 16.5 (Table 1; Fig. 2D). However, positions of the sensitivity peak (from 0.3 to 0.5 cyc deg−1) and cut-off frequencies (from 1.1 to 1.9 cyc deg−1) of these 3 species were similar (Table 1).
In Pelecaniformes, the northern bald ibis (Geronticus eremita) differed from the African spoonbill (Platalea alba) and the western cattle egret (Bubulcus ibis) with a peak sensitivity of 11.7 compared with 5.3 and 6.2, respectively, and a position of sensitivity peak at 0.6, 0.5 and 1.6 cyc deg−1, respectively (Table 1; Fig. 2E). The high frequency cut-off of these species was 2.2 for the northern bald ibis, 1.5 for the African spoonbill and 6.3 cyc deg−1 for the western cattle egret (Table 1; Fig. 2E).
For other species studied – common swift (Apus apus), Eurasian stone curlew (Burhinus oedicnemus), white stork (Ciconia ciconia), yellow-necked spurfowl (Pternistis leucoscepus), demoiselle crane (Grus virgo), little bustard (Tetrax tetrax) and Chilean flamingo (Phoenicopterus chilensis) – we observed high variability in contrast sensitivity, with a peak sensitivity ranging from 4.9 for the Chilean flamingo, with position of sensitivity peak at 0.3 cyc deg−1, to 23.0 for the common swift, with position of sensitivity peak at 0.5 cyc deg−1 (Table 1; Fig. 2F). The high frequency cut-off varied from 1.8 cyc deg−1 for the yellow-necked spurfowl to 4.0 cyc deg−1 for the white stork (Table 1; Fig. 2F).
Normalisation of CSFs shows that despite differences in high frequency cut-off and position of peak sensitivity, the general shape of the CSFs is very similar, but the bandwidth varies greatly depending on species (Fig. 3).
Effects of phylogeny and CD:AL ratio on the peak contrast sensitivity variability
Two linear regressions were performed to estimate the effect of the CD:AL ratio (a proxy of the retinal image brightness) on the peak contrast sensitivity, one controlled for phylogeny and one without phylogeny. We did not find any significant effect of the CD:AL ratio whether phylogeny was considered or not (Table 2; Figs 4 and 5). Peak contrast sensitivity was highly conserved across our subset of avian phylogeny (Pagel's lambda=0.79). Phylogeny explained 25% of the variation in peak contrast sensitivity (Fig. 5).
DISCUSSION
This study provides the first phylogenetically controlled analysis of variation in the peak contrast sensitivity of birds. Although, as far as we know, the contrast sensitivity of 12 bird species had been previously assessed, these studies did not allow for easy comparison between them as values come from both operant conditioning and pattern electroretinogram (PERG) methods, and none of them used anatomy or phylogeny to understand variation among species.
The relatively low peak contrast sensitivity is shared among birds
PERG and behavioural methods provide complementary insights and yield similar results concerning sensitivity peak positioning. However, it appears that for Columba livia, PERG tends to underestimate the peak contrast sensitivity and the high frequency cut-off by 53% and 37%, respectively, compared with the operant conditioning measures (Hodos et al., 2002). If we assume that this level of underestimation holds for other bird species studied here, then the peak contrast sensitivity values for Passeriformes fall within the range of values reported by Ghim and Hodos (2006) for the common starling (Sturnus vulgaris), another Passeriforme (see Table 1). Similarly, the peak contrast sensitivity of the American kestrel (Falco sparverius) was close to the value found by Ghim and Hodos (2006) for this species with PERG (Table 1). However, the positions of the sensitivity peak we found for the American kestrel and Passeriformes species were lower than the values found using PERG for the American kestrel and the Common starling, respectively (Ghim and Hodos, 2006), unlike the Northern raven (Corvus corax), which had a similar value to the reported positions of the sensitivity peak (Table 1).
Although contrast sensitivity tested with stationary stimuli underestimates peak contrast sensitivity in budgerigars (Haller et al., 2014), our results were in the range of previously measured values with an operant conditioning method with stationary or moving stimuli (Table 1). Indeed, our measures of peak contrast sensitivity ranged from 4.9 for the Chilean flamingo to 29.8 for the northern raven, falling in the range of previously reported values: from 4.8 for the Bourke's parrot (Neopsephotus bourkii) (Lind et al., 2012) to 31 for the American kestrel (Hirsch, 1982) (Table 1). Positions of sensitivity peak varied from 0.3 cyc deg−1 for the Chilean flamingo to 1.6 cyc deg−1 for the western cattle egret, and are in the lower part of previously reported values, which ranged from 0.5 cyc deg−1 for domestic chickens (Gallus gallus) (Shi and Stell, 2013) to 10 cyc deg−1 for the American kestrel (Hirsch, 1982). Interestingly, the peak contrast sensitivity found for the American kestrel in our study (28.61) is similar to that reported by Hirsch (1982), but the position of sensitivity peak (1.12 cyc deg−1) is 10 times lower. Similarly, the peak contrast sensitivity of the common buzzard (Buteo buteo) (17.1) is close to the 11.8 measured by Potier et al. (2018) for the closely related species, the Harris's hawk (Parabuteo unicinctus), but the position of sensitivity peak (0.8 cyc deg−1) is eight times lower (Table 1). This highlights that values of bird peak contrast sensitivity obtained by operant conditioning and optocollic reflex are comparable, but their associated spatial frequencies are not. The stimulus velocity used during the experiment may affect the CSF. However, Haller et al. (2014) showed that for low or high SFs, the contrast sensitivity varies with the velocity of the stimulus, while this is not the case with intermediate SFs. Hence, stimulus velocity should not have an impact on the peak contrast sensitivity, but it may have an impact on the position of the peak, and the high frequency cut-off calculated from the CSFs may be underestimated.
Overall, our study highlights low achromatic contrast sensitivity of birds compared with large mammals, which are two to six times more sensitive [peak contrast sensitivity of 116 for cats (Bisti and Maffei, 1974) and 77–175 for primates (De Valois et al., 1974; Merigan, 1976; Jacobs, 1977; Langston et al., 1986)]. Studies of contrast sensitivity in these large mammals have been conducted using operant conditioning. Similarly to birds tested in operant conditioning, the position of sensitivity peak of these mammals (2 to 4.5 cyc deg−1) is higher than those found in our study, except for the cat (0.2 cyc deg−1) and the greater galago (Galago crassicaudatus; 0.8 cyc deg−1) (Bisti and Maffei, 1974; Langston et al., 1986). The contrast sensitivity of birds is comparable to that of small mammals such as rats (peak contrast sensitivity ranges from 7 to 20) (Birch and Jacobs, 1979; Keller et al., 2000) and squirrels (from 12 to 32) (Jacobs et al., 1980, 1982). Interestingly, although studies on small mammals were performed with operant conditioning, the range of position of sensitivity peak is similar to that of birds (0.1 to 0.7 cyc deg−1). Aquatic species such as fishes, sharks and octopuses seem to be more sensitive to low contrast than birds since the peak sensitivity ranges from 19 to 125 (Northmore and Dvorak, 1979; Bilotta and Powers, 1991; Northmore et al., 2007; Ryan et al., 2016; Santon et al., 2019; Nahmad-Rohen and Vorobyev, 2020). Although different experimental methods have been used for these studies (including operant conditioning and OKR), the position of sensitivity peak (0.1 to 0.4 cyc deg−1) remains low compared with birds.
The low achromatic contrast sensitivity observed in avian species may be attributed to a trade-off between achromatic contrast sensitivity and ultraviolet sensitivity, as proposed by Ghim and Hodos (2006). The addition of a visual component, such as a fourth type of cone, is potentially costly from an energy point of view and it can be expected that it will only be implemented if it is an advantage for the species in question (Niven and Laughlin, 2008). In addition, the contrast increment thresholds (difference thresholds) of UV-sensitive cones of goldfish were found to be five times larger than those of other cones (Hawryshyn, 1991). This trade-off hypothesis is supported by the fact that large mammals such as primates do not appear to be sensitive to ultraviolet light (see Hunt et al., 2001 for review) and have a greater sensitivity to achromatic contrasts than smaller mammals and birds, some of which are sensitive to ultraviolet (Hunt et al., 2001). However, as not all bird species are UV sensitive, it would be interesting to compare the achromatic contrast sensitivity of UV-sensitive birds with those that are not.
Relationship between CD:AL ratio and peak contrast sensitivity variability
Unexpectedly, the peak contrast sensitivity of birds was not significantly related to the eye's CD:AL ratio. A high level of photons reaching the retina is required to distinguish between random and patterned photon distribution, especially when the contrast is low because of the reduced difference in photon numbers between adjacent objects, or between an object and its background. Therefore, a high level of light is required to detect low contrasts, and by extension, the contrast sensitivity of an eye should be related to its light sensitivity (Land and Nilsson, 2012). The eye's sensitivity to light partly depends on the retinal image brightness, which is determined by the f-number [the ratio of the eye's focal length (f) to the aperture (pupil) diameter (D); the smaller the f-number, the brighter the retinal image]. The aperture varies with the size of the pupil but cannot be greater than the corneal diameter and thus the eye width itself. Wider eyes should therefore allow higher retinal image brightness. As AL and CD are often used as proxies of the eye's focal length and pupil diameter respectively, the CD:AL ratio may be used as a proxy of the f-number to represent the retinal image brightness and sensitivity of an eye (Martin, 1982; Hall and Ross, 2007; Hall et al., 2012).
Parameters other than just the retinal image brightness affect eye sensitivity, such as the ratio of the retinal sampling unit size (d) to the focal length f (Land, 1981). If the eye size is scaled up proportionally to increase pupil diameter D, the focal length f increases too, causing a decrease in the d to f ratio and thus light sensitivity. Moreover, spatial summation – signal pooling from many photoreceptors to the same retinal ganglion cell – improves an eye's sensitivity by increasing the effective sampling unit size d (Warrant, 1999; Land and Nilsson, 2012). Finally, an eye's sensitivity may also be affected by the post-retinal processing of the information (Martin, 1982; Land and Nilsson, 2012). Therefore, the variation of contrast sensitivity observed in birds may be dependent on higher information processing levels as well as on brightness of the retinal image.
Relationship between phylogeny and peak contrast sensitivity
Our study reveals a relatively low contrast sensitivity shared by many avian phylogenetic orders. There is great variability between species, of which 25% is explained by phylogeny. Thus, a part of the contrast sensitivity variation among birds is linked to common ancestry.
The part of the variation not explained by phylogeny may be correlated with the birds' ecology. When moving through an environment, close obstacles appear to move faster than distant ones (motion parallax effect; Rogers and Graham, 1985). An increase in the optic flow speed may lead to blurry retinal images if temporal resolution and contrast sensitivity are not high. A high contrast sensitivity should help the birds to see and avoid obstacles better. Thus, we speculate that the contrast sensitivity may be positively related to the density of objects around a flying bird. Species foraging or migrating in flocks, like starlings, or flying in dense habitats, like the booted eagle, might have a higher peak contrast sensitivity than those flying alone or in open habitats. It would be interesting to investigate such a relationship to better understand variation in bird contrast sensitivity. However, given the diversity of environments in which birds may fly, data on more bird species are needed to test this hypothesis.
Acknowledgements
We thank all the teams, and particularly Thierry Bouchet, Alexis Sahnoune, Oriane Chevasson, Patrice Potier, Romuald de Romans, Antoine Joris, Marie-Pierre Puech and Frédéric Angelier, of Le Grand Parc du Puy du Fou, Zoodyssée, Les Ailes de l'Urga, Espace Rambouillet, the Réserve Africaine de Sigean, Goupil Connexion and the Centre d'Etudes Biologiques de Chizé for allowing us to conduct experiments with their birds and for their help during experiments. We also thank David Degueldre for helping to design and build the equipment needed for these experiments.
Footnotes
Author contributions
Conceptualization: C.L.M.B., O.D., F.B., S.P.; Methodology: C.L.M.B., O.D., F.B., M.M., S.P.; Validation: C.L.M.B.; Formal analysis: C.L.M.B., O.D., F.B., M.M., S.P.; Investigation: C.L.M.B.; Resources: S.P.C.; Writing - original draft: C.L.M.B.; Writing - review & editing: O.D., F.B., M.M., S.P.C., A.B., S.P.; Visualization: C.L.M.B.; Supervision: O.D., F.B., S.P.; Project administration: C.L.M.B., O.D., F.B., S.P.; Funding acquisition: O.D., F.B., A.B.
Funding
C.L.M.B. was supported by a PhD fellowship from various members of the French MAPE programme (Reduction of Bird Mortality at Operating Wind farms), including ADEME (French Environment and Energy Management Agency), OFB (French Office for Biodiversity), DREAL Occitanie (Regional Directorate for the Environment, Planning and Housing of Occitanie), LABEX CEMEB (UM, Mediterranean Centre for the Environment and Biodiversity), FEE (France Wind Energy Association) and the SER (Union of renewable energies), the Occitanie region, the Ministry of Ecological Transition and Solidarity (DGEC, Directorate-General for Energy and Climate) and 25 wind energy operators.
Data availability
Data used for analyses are available at https://doi.org/10.7910/DVN/05M6Z1.
ECR Spotlight
References
Competing interests
The authors declare no competing or financial interests.