The presence of UV wavelengths improves the temporal resolution of the avian visual system

Terrestrial vertebrates, not least birds, are dependent on the ability to quickly and accurately assess their visual environment. Good visual function is essential for orientation, foraging, detecting and escaping predators, as well as identifying conspecifics and evaluating potential mates. Birds are very visual animals – the active and airborne lifestyles of most avian species require rapid detection, interpretation and response to visual stimuli. The ability to resolve rapid movement is determined by the temporal resolution of the visual system (Jones et al., 2007).

Perception of movement involves responding to stimuli that change both over time and in space (Fahle and Bach, 2006). Characteristics of the temporal response can be studied by estimating the temporal luminance contrast – for example, modulation threshold and critical flicker frequency (Arden, 2006). Critical flicker frequency (CFF) is the frequency at which a flickering light stimulus can no longer be resolved and appears continuous to the observer. Measuring CFF is a convenient method for behaviourally assessing temporal properties of visual systems of animals other than humans. Even though the exact relationship between the ability to perceive movement and measures such as CFF in birds is not known, CFF is probably a direct or indirect indication of relative capability to detect movement.

Temporal contrast sensitivity and CFF have been studied in a number of taxa in scotopic (dark adapted), as well as photopic (light adapted) conditions (Loop and Berkley, 1975; Toyoda and Coles, 1975; Bernholz and Matthews, 1975; Bilotta et al., 1998; Woo et al., 2009) and has also been used as a diagnostic method in medicine (e.g. Chang et al., 2007; Rubin and Kraft, 2007). In birds, CFF and its properties have been studied in only a few species: the chicken Gallus gallus (Jarvis et al., 2002; Nuboer et al., 1992; Railton et al., 2009), the pigeon Columba livia (Dodt and Wirth, 1953) and the budgerigar Melopsittacus undulatus (Ginsburg and Nilsson, 1971). Most findings show quite simply that flicker detection improves with increasing light intensity – that is, higher CFF values are obtained with brighter stimuli (e.g. Jarvis et al., 2002; Nuboer et al., 1992). More interestingly, CFF and its rate of increase with light intensity also appear to vary with the spectral properties of light (Nuboer et al., 1992).

Spectral sensitivity and colour vision in birds has been studied more extensively than temporal resolution (reviewed in Hart, 2001; Kelber et al., 2003; Bennett and Théry, 2007; Osorio and Vorobyev, 2008). In particular, UV vision, and its function, has been of great interest as it gives birds an additional dimension in the colour space, but is lacking in most mammals, including humans (Goldsmith, 1990; Maier, 1992; Bennett and Cuthill, 1994; Cuthill et al., 2000; Maddocks et al., 2002; Håstad et al., 2005; Schaefer et al., 2007). Several studies have shown significant effects of UV on avian mate choice (Bennet et al., 1996; Bennett et al., 1997; Hunt et al., 1997; Johnsen et al., 1998; Pearn et al., 2001; Alvarez et al., 2004) (but see Banks, 2001) and foraging (Viitala et al., 1995; Church et al., 1998; Honkavaara et al., 2004). So far, no one has explored the potential importance of UV for visual tasks that are not based on perception of colour, such as detection of motion and flicker.

In birds, motion detection, as well as achromatic or luminance vision, are thought to be possible functions of the double cone photoreceptor (Campenhausen and Kirschfield, 1998; Osorio et al., 1999; Goldsmith and Butler, 2005; Hart and Hunt, 2007). Despite being highly abundant in the avian retina and comprising ca. 50% of all cones (Bowmaker et al., 1997; Hart, 2001), the function of the double cones has not been determined with confidence (Goldsmith and Butler, 2005; Hart and Hunt, 2007).

The purpose of this study was to test how the spectral properties of light affect the temporal resolution of the avian visual system by estimating CFF in the chicken (Gallus gallus) in a behavioural trial. The CFF values were determined for light of different wavelengths and a range of light intensities. We tested two broad-spectrum light stimuli: white UV-free and full-spectrum (white with addition of UV), as well as two spectrally narrow stimuli: UV and yellow. The only difference between the broad-spectrum stimuli was the presence of an additional peak in the UV part of the spectrum (400 nm) for the full-spectrum light. The aim was to produce light that is more similar to natural daylight than the artificial lighting often used when conducting behavioural studies indoors and to compare the two. The narrow-spectrum UV stimulus was tested to explore further the possible importance of UV vision for flicker detection. The yellow light stimulus (590 nm) was chosen to match as closely as possible the peak sensitivity of the double cone and test whether the estimated CFF for this wavelength would indicate higher temporal resolution, and thereby greater importance of the double cone, for the perception of flickering light.

We designed an operant conditioning chamber (size 550 mm × 650 mm × 550 mm), similar to the one described and used by Jarvis and colleagues (Jarvis et al., 2002) and Prescott and Wathes (Prescott and Wathes, 1999); see Fig. 1. Two circular light stimuli were located in the wall of the chamber. Stimuli were produced by two light sources, each of which consisted of three light-emitting diodes (LEDs) (5 mm, 30DEG) placed in a triangular formation at the distal end of an aluminium tube. The stimuli were projected through two UV-transparent Perspex panels with neutral density and diffusion filters placed between the panels and the light sources. Light from outside was prevented from entering and affecting the stimuli. A feeder was located between the panels at the floor level. The light flicker functions were produced by two function generators (2 MHz, GW Instek, Suzhou, China) connected by means of a switch so that the signal from each generator could be interchanged between the two light sources, from one side to the other.

We created four different light stimuli by using three types of LEDs: white (Avago Technologies, Malaysia), UV (single peak at 400 nm, Hero, South Korea) and ‘amber’ (single peak at 590 nm, Avago Technologies, Malaysia). The stimuli were white UV-free (three white LEDs), full-spectrum (two white, one UV LEDs), UV (three UV LEDs) and yellow (three ‘amber’ LEDs) (Figs 2 and 3). Flickering or continuous light of various frequencies and intensities could then be produced by using the function generators in combination with neutral density (50% transmittance and 25% transmittance; Lee Filters, Andover, UK) and diffusion (75% transmittance: Lee Filters) filters. The UV-transparency of the filters was confirmed through spectrophotometer measurements. The light intensity could also be adjusted by changing the distance between the light sources and Perspex panels. Two wave functions were chosen to create a flickering stimulus: a sine wave (100% modulation) and a square wave (60% modulation).

We also measured the chosen light intensity levels (units: cd m^–2) directly on the surface of the Perspex panels and as lux from a distance of 10 mm in front of the stimulus window to estimate the light environment that the birds were in when pecking the panels. These measurements were performed only for the white UV-free stimulus, using a Hagner ScreenMaster instrument (B. Hagner AB, Solna, Sweden). The four light levels are given in Table 2.

The full-spectrum light was adjusted to match, as closely as possible, the spectral distribution of natural daylight. The ratio of relative quantum catch between the UV-part of the spectrum, up to 425 nm (emitted by the UV-LED), and the rest of wavelengths (emitted by the white LEDs) was calculated and compared with the CIE standard data on the spectral power distribution of sunlight (D65). The light was then adjusted to yield an approximate ratio of 1:10 (UV: white), the same as D65.

White leghorn chicks (genotype Bovans) were used in the study. The birds were reared indoors from day-one of age under a combination of shaded natural daylight and high-frequency fluorescent light (on average, 150 lx). Chickens were randomly assigned into groups: all light stimuli were combined with both wave functions. Half of the birds were trained to peck at a flickering stimulus (20 Hz) inside the operant conditioning chamber and the other half at a continuous stimulus (2000 Hz) to receive a food reward from the feeder. The chickens were first trained to peck at the light on one side of the chamber and then the other, while the alternative stimulus (one that birds were not trained to peck) was lit on the opposite side. When the birds had learned this, the position of the test stimulus was alternated between sides. The light intensity used for training was approximately 200 lx (320 cd m^–2). The birds were both trained and tested in pairs to avoid distress induced by isolation. Each pair was tested for all four light-intensity levels and the testing order of intensities was randomized.

Only one bird per pair learned the task successfully. Once the more active chicken in a pair started to associate a light stimulus with food, the other individual was no longer motivated to peck the panels as it could receive some food while the active bird was pecking. On a few occasions, both birds started to learn the task. These pairs had to be separated and provided with a more passive ‘partner’ each as it was not possible to reward correct pecks consistently if the two birds pecked at both panels simultaneously. Also in cases where birds showed aggression towards each other, especially if the passive individual disturbed the procedure by, for example, monopolising access to the feeder, pairs were separated.

Chickens were trained between one and 10 weeks of age and tested between five and 12 weeks of age. Periods of training and testing overlapped because individual birds learned the procedure at a different pace. The birds were starved for a period of 3 h before each training or testing session. There was no additional lighting inside the chamber, and the ambient light levels there were 2–4 lx, when the light stimuli were switched off. The experiments were approved by the ethical council for animal testing in Uppsala, Sweden (Uppsala djurförsöksetiska nämnd).

Testing commenced once birds successfully identified the test stimulus at least 80% of the time. Immediately before testing, the birds were placed inside the experimental arena and were given 5 min to adapt to the light conditions, while the alternative stimulus (20 or 2000 Hz) was lit in both lamps. Next, the test stimulus was lit in one of the lamps until the bird responded by pecking at the panels. Five pecks at the panel with the test stimulus were required for successful discrimination, and this granted three-second access to the feeder. During the time when this reward was presented, the alternative stimulus was lit in both lamps; then, the test stimulus was switched on again. Pecking the alternative stimulus or showing disinterest for more than five minutes was regarded as incorrect discrimination and was not rewarded. During the testing procedure, the flickering frequency of the stimulus was increased in 10 Hz increments, starting from 20 Hz, until correct discrimination was no longer recorded. Frequency was then decreased in steps of 5 Hz until discrimination was correct; thereafter, it was again increased by 1 Hz at a time to obtain a CFF value. CFF was defined as the lowest frequency at which the flickering stimulus appeared continuous – that is, 1 Hz above the highest frequency that birds could discriminate correctly. After a CFF value had been estimated, the test ended with switching off both lamps and rewarding birds by opening the feeder. If a CFF could not be estimated owing to passivity and disinterest of the birds, they were simply removed from the experimental chamber.

The results were analyzed using Statistica 8.0 software, as a factorial repeated measures ANOVA design, with light stimulus and wave function as between-subject factors and light intensity level as a within-subjects factor (repeated measures).

Chicken CFF values determined in this study ranged from 44 Hz (white UV-free light, intensity 1) to 83 Hz (UV light, intensity 4). The highest values were observed for yellow and UV light, followed by the full-spectrum light (see Fig. 4A,B for details). CFF values for the white UV-free light appear to be lower compared with the rest of the stimuli. The statistical analysis could only be performed for full-spectrum light and white light (eight subjects per group) as an insufficient number of birds were able to learn the testing procedure for the yellow (three subjects) and UV (two subjects) groups (see Discussion).

The effects of both light stimulus (F_1,12=104.52; P<0.0001) and intensity level (F_3,36=274,23; P<0.0001) on CFF were highly significant. There was also a significant interaction effect between the two factors (F_3,36=9.18; P=0.00012), meaning that the slope of the CFF as a function of light intensity differs between full-spectrum light and white UV-free light. No effect of the wave function on CFF could be observed in this study (F_1,12=0.00; P=n.s.).

Our results indicate that, if some UV light is added to the stimulus, birds can perceive higher frequencies of flickering light, compared with only white light – that is, the temporal resolution is improved by the addition of UV wavelengths. Even though the full-spectrum light used here is not identical to natural daylight, it gives a good representation of wavelengths that are perceived by birds and stimulates all types of cone photoreceptor cell in the retina. The presence of UV wavelengths in the light environment has already been shown to have important biological effects on mate choice (e.g. Bennett et al., 1996; Hunt et al., 1997; Johnsen et al., 1998) and foraging (Viitala et al., 1995; Church et al., 1998; Honkavaara et al., 2004). We propose that it has significant effects also on other visual functions, such as flicker perception, at least over the range of light intensities tested in this study. The difference in CFF between white and full-spectrum light also suggests that, despite the processes of artificial selection imposed on domestic poultry, the chicken visual system responds better to the light environment that it had originally been adapted to. Even the relatively high CFF values obtained with the UV light show that many generations of selection and rearing in poor lighting conditions have apparently not led to lowered sensitivity for UV light. The high spectral sensitivity in UV part of the spectrum, observed in behavioural studies (Burkhardt and Maier, 1989; Maier, 1992), has been attributed to differences in receptor adaptation to the background light and lack of near-UV in indoor lighting (Vorobyev and Osorio, 1998; Goldsmith and Butler, 2003). This is unlikely to be the case in our study as the background lighting included natural daylight and the birds had time to adapt to the light stimuli before testing.

The yellow light, similarly to the UV, yielded apparently higher CFF values than full-spectrum and white light. The peak wavelength of this stimulus was chosen to match the peak sensitivity of the double cone. As a suggested function of the double cone is motion detection (Campenhausen and Kirschfeld, 1998), it might also be involved in flicker detection. If the double cone were, to a greater extent than other cone types, involved in flicker detection, we would expect higher CFF values for this light. The results from this study do not suggest such a function of the double cone. On the contrary – the UV stimulus, which almost entirely excludes the double cone, yielded similar or higher values, indicating that the UV-sensitive cone (SWS1) is just as important for the ability of birds to detect flicker. The subjects tested with both yellow and UV light showed little variation between individuals, which suggests that the CFF values were estimated correctly. Overall, the individual variation seems to be low among the birds, perhaps indicating low genetic variation due to the artificial selection.

Few subjects could be tested with UV as well as yellow light because the spectral properties of the light seemed to affect bird behaviour and the learning process. Individuals that were trained using white and full-spectrum light learned to peck the panels without apparent difficulties, whereas those trained with UV or yellow stimuli appeared to be distressed by the light and hesitated to approach the panels. Some of these birds failed to associate the light with food, and others were so cautious that they never pecked a panel more than once and could not maintain motivation throughout a testing session. As a result, these birds were not ready for testing within the time-frame available for conducting the experiments. A bird was excluded if it, after 10 training sessions (15 min each), still could not associate light stimulus with food. All pairs got equal training time each day, and they were trained in different order/time of the day, hence differences in behaviour seem to have been caused by the properties of the light.

Several studies suggest that birds perceive novel colours as aversive and show avoidance towards, for example, black-and-yellow patterns, a behaviour that might be associated with aposematic coloration of insect prey (Lindström et al., 1999; Kelly and Marples, 2004; Ham et al., 2006; Johnston and Burne, 2008). This might to some extent account for the cautiousness of our birds towards yellow and UV stimuli. We would have expected, however, that several weeks of daily exposure should have been sufficient for the chickens to become accustomed to the colours. It has also been suggested that, as a consequence of domestication and artificial selection, white leghorn hens differ from red jungle fowl in their social behaviour and have impaired learning ability (Lindqvist and Jensen, 2009; Kirkden et al., 2008) even though they do not seem to suffer major visual impairment (Karlsson et al., 2009). We observed that some behavioural aspects considerably complicated the process of learning the task in this study: the white leghorn chicks lacked a strong motivation to obtain food, were cautious and showed avoidance behaviours towards the light stimuli and were passive in general.

An increase in CFF was observed with increasing light intensity, as has been well documented for human vision. In human photopic vision, CFF is known to increase with increasing retinal illuminance and number of retinal ganglion cells stimulated (Rovamo and Raninen, 1988). Furthermore, the shape of the human flicker sensitivity function has been measured and modelled by Rovamo and colleagues (Rovamo et al., 1999) and shown to depend on luminance and temporal noise (Rovamo et al., 1999). Previous studies on chicken similarly show that CFF increases with mean light intensity (Jarvis et al., 2002; Nuboer et al., 1992). The CFF values estimated for white light in this study are quite similar to those measured by Jarvis and colleagues (Jarvis et al., 2002) (39–71 Hz for intensities of 10–1000 cd m^–2). It is rather difficult to make a direct comparison of our work to that of Nuboer and colleagues (Nuboer et al., 1992) as those authors quantified intensity in a different way, although the range of CFF values is similar. When comparing different studies, it should be considered that factors such as background light level, intensity estimation method, criteria for correct discrimination and, not least, breed and age of the birds might have an affect on the results.

The light levels used in our study were probably not sufficient to estimate the absolute maximum values of CFF in birds, domesticated or wild. We would expect that a further increase in light intensity could yield higher values. The shape of the CFF curve as a function of light intensity could be better understood by increasing the light to such levels that allow estimation of the absolute maximum CFF values, to the point at which the CFF could be expected to decrease owing to photobleaching effects. Such a decrease was not observed in this study, indicating that maximum values have not been reached.

The present study aimed to investigate flicker perception in photopic conditions, and we have therefore only included cone sensitivity in the intensity estimation. However, some effect from the rods cannot be excluded entirely, especially at the lowest light intensities, where the light conditions might have been within the mesopic range (when both cones and rods operate) (Lind and Kelber, 2009a). The effect from rods at these intensities might contribute to the difference in slopes of the curves. It is difficult to estimate such an effect as the mesopic range in birds is unknown and the boundary for colour vision (lower limit of the mesopic range) has so far only been determined in two species of parrot (Lind and Kelber, 2009b). The spectral composition of the light used in this study is such that the perceived intensities of the full-spectrum and white stimuli would have been affected similarly (Table 1). Rods have a slower response to flicker than cones (e.g. Dodt and Wirth, 1953). Nevertheless, CFFs were higher for full-spectrum than for white light, indicating quite clearly that the rods did not contribute to the recorded values. Our results from UV and yellow stimuli would have hardly been affected at all by a possible involvement of rods (Table 1).

The different wave functions did not show any effects on CFF. In order to obtain the same light intensity for both functions, the square wave had to have a slightly lower modulation (60%) than the sine wave (100%). Nevertheless, this is not likely to have had an important effect on the results as modulation down to even 35% is only expected to have a slight effect on CFF values (van de Grind et al., 1973). The aim of using the two different functions was to confirm that flickering stimuli produced by slightly different methods result in similar estimates of CFF. Sine and square wave functions have been used previously (Tyler, 1987) to estimate CFF for selectively stimulated long-wavelength-sensitive (LWS) cones in humans. The two waveforms produced a similar response in terms of CFF values (Tyler, 1987). From the conclusions of our study, it appears that also in birds the waveform of the flicker does not affect the results.

Temporal resolution in birds is far from being completely understood. For example, the individual effect of all five cone types in flicker sensitivity needs to be explored further. As light of different wavelengths yields CFF functions with different heights and slopes (Nuboer et al., 1992) (this study), the contribution of individual cone types to flicker detection might differ as well. Furthermore, most CFF research has so far been performed on domestic birds, such as chickens, and is insufficient to extrapolate to the visual functions in wild species or to possible differences between species in general. The temporal resolution abilities in a species are likely to be affected by the spectral composition and intensity of the ambient light in its natural habitat or its evolutionary history (Land and Nilsson, 2002). On the one hand, chickens might show lower temporal resolution in bright daylight than small passerines, for example, which fly more often and faster in a brighter environment and for which it therefore would be more profitable to update the retinal image at a faster rate (Evans et al., 2006). In dim and spectrally skewed light regimes similar to the habitat of origin of the red jungle fowl, on the other hand, the chicken might have comparatively high temporal resolution.

Our conclusion that the ability to detect motion might be affected by the wavelength of light is perhaps not so unexpected. When studying visual perception in birds and other animals, properties such as spectral sensitivity, spatial and temporal resolution are often treated as independent concepts with separate functions. We should remember, however, that these are just different components of one integrated visual system that has been shaped by adaptation to a visual environment and lifestyle where colours, shapes and movements interact.

We thank Göran Arnquist and James K. Bowmaker for valuable advice on experimental design and cone sensitivity calculations, and Tom Lisney for useful comments on the manuscript. Also, we thank Heidi Lindfeldt and Jenny Nielsen at Funbo-Lövsta for all their practical help with the chickens and Almut Kelber for kindly lending her light meter.

 
• CFF

  critical flicker frequency

 
• LEDs

  light-emitting diodes

 
• LWS

  long-wavelength sensitive

 
• MWS

  medium-wavelength sensitive

 
• SWS1

  short-wavelength-sensitive type 1

 
• SWS2

  short-wavelength-sensitive type 2

 
• UV

  ultraviolet