Greater flamingos Phoenicopterus roseus use uropygial secretions as make-up
- 582 Downloads
It was long thought that the colour of bird feathers does not change after plumage moult. However, there is increasing evidence that the colour of feathers may change due to abrasion, photochemical change and staining, either accidental or deliberate. The coloration of plumage due to deliberate staining, i.e. with cosmetic purposes, may help individuals to communicate their quality to conspecifics. The presence of carotenoids in preen oils has been previously only suggested, and here we confirm for the first time its presence in such oils. Moreover, the carotenoids in the uropygial secretions were the same specific pigments found in feathers. We show not only that the colour of feathers of greater flamingos Phoenicopterus roseus became more colourful due to the application of carotenoids from uropygial secretions over the plumage but also that the feathers became more colourful with the quantity of pigments applied over them, thus providing evidence of cosmetic coloration. Flamingos used uropygial secretions as cosmetic much more frequently during periods when they were displaying in groups than during the rest of the year, suggesting that the primary function of cosmetic coloration is mate choice. Individuals with more colourful plumage initiated nesting earlier. There was a correlation between plumage coloration before and after removal of uropygial secretions from feathers’ surfaces, suggesting that the use of these pigmented secretions may function as a signal amplifier by increasing the perceptibility of plumage colour, and hence of individual quality. As the cosmetic coloration strengthens signal intensity by reinforcing base-plumage colour, its use may help to the understanding of selection for signal efficacy by making interindividual differences more apparent.
KeywordsCarotenoids Cosmetic coloration Plumage colour Plumage maintenance Signals Uropygial secretions
Carotenoids are the basis of some plumage colours, such as red, yellow and orange (Hill 2002). These pigments are incorporated into feathers during moult, and as they cannot be synthesised by animals, but have to be ingested with food, plumage colour may be used to communicate the ability to obtain resources (Hill 2002; Searcy and Nowicki 2005). Plumage colour has been traditionally considered as a static trait, with little opportunity for individuals to alter their appearance if conditions change between moulting periods. However, there is increasing evidence that the colour of feathers may change due to abrasion, photochemical change and staining, either accidental or deliberate (Uchida 1970; Negro et al. 1999; Piersma et al. 1999; McGraw and Hill 2004; Blanco et al. 2005; Figuerola and Senar 2005; Montgomerie 2006; Delhey et al. 2007; Surmacki 2008). Indeed, some bird species are known to modify the colour of their feathers or other body parts by means of the deliberate application of substances, which may be either external to the birds (Negro et al. 1999; Montgomerie 2006; Delhey et al. 2007; Surmacki and Nowakowski 2007) or produced by the birds themselves (Uchida 1970; Montgomerie 2006; Delhey et al. 2007; Surmacki and Nowakowski 2007; Piault et al. 2008; López-Rull et al. 2010). Among the substances produced by the birds are the secretions of the uropygial gland, which may be pigmented orange, red or yellow (Stegmann 1956; Vevers 1985). It has been suggested that carotenoid pigments are present in preen oils, and that these pigments tinge the plumage when artificially transferred onto it (Stegmann 1956). It has also been suggested that the secretions of the uropygial gland may act as cosmetics by modifying the spectral shape of the reflected light (Delhey et al. 2007). However, to our knowledge, neither the presence of carotenoids in the uropygial secretions, neither their potential role as cosmetics because of the changes that they may produce in feather coloration, have been demonstrated.
The use of cosmetics has been reported in some vertebrate taxa, such as fishes, birds and mammals, and has received little attention by behavioural ecologists (Delhey et al. 2007). Although it has been suggested that this type of coloration has a signalling function, by providing a reinforcing mechanism linking body coloration and fitness-related traits (Negro et al. 1999; Piersma et al. 1999), the mechanisms by which this is achieved have not been addressed. To act as an honest signal, the cosmetic coloration should have some cost (Zahavi 1975; Grafen 1990). In the case of birds, time and energy costs may be associated with the maintenance of plumage coloration (Delhey et al. 2007; Griggio et al. 2010). First, if the cosmetics are carotenoids, frequent reapplication would be required to maintain plumage coloration since these pigments bleach rather quickly when exposed to ambient conditions, thus causing the fading of colours (Vevers 1985; Delhey et al. 2007). Second, to apply the cosmetics on feathers, birds may use specific, time-demanding behaviour (Delhey et al. 2007; Griggio et al. 2010) so that time devoted to other activities may be limited. In addition, there may also be physiological costs since carotenoids are used in some biological functions, and because they are potentially a limiting resource, trade-offs among such functions are likely (see Discussion).
In most cases, the proposed signalling function of cosmetic coloration has been sexual, as this type of coloration mainly develops during breeding or displaying (Delhey et al. 2007). Accordingly, we predicted that (1) the use of cosmetic coloration should be more prevalent when individuals are acquiring mates than during other parts of the year, and that if the pigments used for cosmetic coloration act as a reinforcing mechanism of honest signalling, (2) the pigments found in the uropygial secretions should be the same as those found inside feathers, and then (3) there should be a correlation between the use of cosmetic coloration and base-plumage colour (i.e. the quantity of pigments on feathers’ surfaces should be related to feather colour once external pigments are removed). Finally, if the cosmetic coloration functions as a kind of honest signal, there should be a correlation between its use and some measure of individual quality.
The reliability of honest signals is ensured because they are costly to produce (Zahavi 1975; Grafen 1990). But signals should not only communicate reliably the quality of individuals—they also need to be efficient in reaching target destination and elicit a response (Guilford and Dawkins 1991; Maynard Smith and Harper 2003). Because plumage colour changes at moult, several months before mate choice, the information conveyed by colour may be out of date when individuals make assessments (Searcy and Nowicki 2005; Montgomerie 2006). Under these circumstances, the ability of individuals to physically manipulate the perception of their own signals may make the signals (i.e. plumage colour) more easily to be perceived at times when such signals must be functional. In this context, our findings may have important implications to the understanding of selection for signal efficacy, as the cosmetic coloration may reinforce base-plumage coloration, and thus may make interindividual differences more apparent, which is consistent with the idea that redundancy increases efficacy (Maynard Smith and Harper 2003).
We tested the above predictions using the greater flamingo Phoenicopterus roseus as a model. This species is ideal to look for evidence of cosmetic coloration, as the adult plumage exhibits temporal variation in colour that does not seem to be related to moult (Cramp and Simmons 1977; Shannon 2000). Furthermore, carotenoids (mainly canthaxanthin) are responsible for their pink colour (Fox 1975). Greater flamingos are monogamous and both males and females participate in incubation and chick rearing (Johnson and Cézilly 2007). The rate of mate splitting between consecutive breeding seasons is very high (98%, Cézilly and Johnson 1995). Flamingos display in mixed groups of males and females several months before breeding (Cramp and Simmons 1977), suggesting that assessment of potential mates is very important. Plumage colour may play a function during such assessment, as during the displays the birds exhibit the most coloured plumage patches (Cramp and Simmons 1977). Hence, we first studied seasonal variations in plumage colour in relation to courtship activity. Next, we looked for the pigments that may tinge the plumage both in the secretions of the uropygial gland and on feathers’ surface, i.e. external to the plumage. We then studied whether there was specific maintenance behaviour of plumage related to the cosmetic acquisition of coloration. Finally, we asked whether the cosmetic coloration of plumage was correlated with a reliable predictor of annual reproductive success: date of egg-laying.
Materials and methods
Allocation of colour to neck plumage
To record seasonal variations in plumage colour of greater flamingos, we visited three wetlands in southern Spain (Guadalquivir marshes—36°55′63″ N, 6°15′89″ W; Odiel marshes—37°15′56″ N, 6°59′29″ W; Fuente de Piedra lake—37°06′86″ N, 4°46′17″ W), during 2004–2006, and recorded the colour of neck feathers of adult birds by scanning flocks using a spotting scope. The two marshes are the main feeding areas during the chick provisioning period of greater flamingos breeding at Fuente de Piedra (Amat et al. 2005). We allocated the colour to three scores: (1) very pale pink that at distance looks white, (2) pale pink and (3) pink (see Electronic supplementary material). We performed the observations to allocate plumage colours in the morning (06:00–11:00 h, GMT). Although we used a coarse colour-scoring method, human vision may provide a valid proxy for avian perception of interindividual differences in plumage coloration (Hill et al. 1999; Seddon et al. 2010). There was a close agreement among three observers (JAA, AG, MAR) in the assignment of neck colour scores to 50 individual flamingos (Kendall coefficient of concordance, W = 0.89, P < 0.001).
To relate neck plumage colour scores to laying dates, we conducted observations at the Fuente de Piedra breeding colony during 42 days in 2004, from late February until late June. The observations were conducted during the morning (06:00–11:00 h, GMT) on individually marked birds with PVC rings that could be read from a distance of up to 300 m using a telescope. We recorded the date and neck colour the first time that individually marked birds were observed in the breeding island. We used the date of first sighting in the breeding colony as a surrogate of laying date because there was a significant relationship between the date of first observation and actual laying date (Spearman’s rank correlation, rs = 0.72, n = 34, P < 0.001), and this allowed us to increase sample size. To check whether the colour fades just after the flamingos start to breed, i.e. when individuals stop participating in group displays, neck colour was recorded once more when the individuals were observed again in the colony at least 21 days later (mean ± SD = 51.6 ± 22.9 days, n = 193) after they were observed the first time. As incubation lasts about 30 days (Cramp and Simmons 1977; Johnson and Cézilly 2007), the second time that we recorded the individuals they were usually attending chicks.
It is well known that in many bird species laying date advances with the age of the female (e.g. Perdeck and Cavé 1992). Because of this, age may be a confounding factor in the relationship between plumage coloration and laying date so that we controlled for age effect (see below). Age in relation to date of first observation in the colony was known for 219 greater flamingos that were marked as chicks with PVC rings with individual codes.
Maintenance behaviour of plumage
We recorded the maintenance behaviour of plumage by flamingos during periods in which group displays were observed, which in southern Spain usually takes place between October and April, and periods in which no such displays were observed (May–September). During group displays, the plumage is ruffled, the necks are stretched upwards and the heads are jerked from side to side in fixed rhythm, and a function of group displays may be mate choice (Cramp and Simmons 1977; Johnson and Cézilly 2007). We distinguished between two types of maintenance behaviour: ‘rubbing’ (defined also as ‘daubing’ [Uchida 1970]), in which the head is thrown back, the crown rests on the upper back and is then rotated from side to side (see Plate 1c in Kahl 1972), usually after rubbing their cheeks directly on their uropygial glands; and ‘preening’, in which the individual uses lateral strokes of its bill to preen the upper breast and lower neck feathers (see Plate 1a in Kahl 1972). While the aim of ‘rubbing’ is cosmetic (see below), that of preening is mainly directed to rearrangement of feathers and removal of ectoparasites. To record the maintenance behaviour, we scan-sampled (Altmann 1974) flocks of flamingos at Odiel and Guadalquivir marshes, as well as at Fuente de Piedra lake, between 06:00 and 11:00 h (GMT), during both the displaying period (October–April) and outside that period (May–September). We recorded the maintenance behaviour of a mean number of 108.6 ± (SD) 56.4 individuals during each recording day (n = 23), allocating the behaviour to either rubbing or preening, and assigning neck colour scores to those individuals. Although not all birds were marked, the probability of resampling the same individuals was small. Out of 333 observations on individually marked birds, we resampled 19 individuals twice (38 observations, 11.4%), and three individuals were resampled three times (2.7%). However, in all these cases, resampling was done in different months, or even years. For such reasons, the observations on the same individuals may be considered as independent.
Pigments in the uropygial secretions and on neck feathers
We obtained samples of uropygial secretions from captive greater flamingos by softly massaging the nipple of the preen gland (Reneerkens et al. 2002), in October (19 individuals), February (15 individuals) and July (12 individuals), with the aim of studying seasonal variations in the concentration of carotenoids in the uropygial secretions. We sampled 35 individuals, of which seven were sampled twice and two were sampled three times. The captive birds came from the same population where behavioural data were recorded. We collected up to 50 mg of uropygial secretions per sample using micropipettes and transferred the samples into 2-mL Eppendorf round-capped tubes for subsequent extraction of pigments. The samples were kept at 4°C until transportation to a laboratory, within the next 5 h after collection, where they were frozen at −20°C until analyses within the next week. Before freezing, we recorded to the nearest 0.1 mg the mass of secretions of every sample using a Mettler Toledo electronic balance. For carotenoid extraction, we added 60–150 μL of acetone (the quantity varied depending on the intensity of the apparent colour) to each sample of uropygial secretions. This was shaken for 2 min, and subsequently it was subject to sonication for 2 min. The sample was then centrifuged at 12,000×g for 5 min and the upper layer stored at −30°C until analysis using high-performance liquid chromatography (HPLC).
HPLC analyses were carried out using a Waters 600E separation module and a Waters 996 PDA detector, controlled by the Empower Pro software. Separation was performed on a Kromasil C18 column (250 × 4.6 mm ID, particle size 5 μm) by using the chromatographic method (Mínguez-Mosquera and Hornero-Méndez 1993), which consists of a binary solvent gradient acetone–water at a flow rate of 1.5 mL min−1. The diode array detector wavelength was set at 450 nm and the UV–visible spectra of each peak were recorded and stored online in the 350–600-nm wavelength range. Identification of carotenoids present in the uropygial secretions, as well as on feathers’ surface (see below), was performed by separation and isolation of the pigments by thin layer chromatographic and cochromatography with standards, acquisition of UV–visible spectra in different solvents, as well as chemical derivatisation microscale tests for the examination of functional groups (Eugster 1995). The different properties (chromatographic, spectroscopic and chemical) of the pigments were compared with standards and data in the literature (Foppen 1971; Davies and Köst 1988; Britton 1995). Quantification was performed using external standard calibration curves from injection of progressive concentrations of the reference pigment.
To determine whether pigments in the uropygial secretions were transferred onto the plumage, we collected neck feathers, as well as uropygial secretions as indicated above, from 16 captive individuals. Carotenoid concentrations in the uropygial secretions of these 16 birds were determined as indicated above. The feathers were introduced in 15-mL Falcon tubes filled with acetone for pigment extraction. In the laboratory, these tubes were shaken during 2 min, using Vortex, after which they were subject to sonication during 2 min. Pigment extracts were transferred to a rotatory flask and the solvent was evaporated until it was dry. The residue was dissolved in 100–300 μL of N,N-dimethylformamide (the quantity varied depending on the intensity of the apparent colour), filtered through a nylon net (0.45-μm mesh size) into an Eppendorf tube and stored at −30°C until analysis. In a pilot study, we did not obtain any pigment from feathers from which external uropygial secretions had been previously removed, indicating that our method was adequate. That is, with our procedure we did not remove any pigment internal to feathers.
Effect of pigment removal on feathers’ colour
The effect of pigment removal on colour was studied by comparing feather colour before and after removal of external pigments. To measure colour, we scanned neck feathers <4 h after collection and again after removal of external pigments. We used an Epson Perfection 1250 scanner, with a resolution of 1,200 pixels. Scanned images of neck feathers were imported at maximal resolution and saved as JPEG files, from which hue, saturation and brightness were recorded using Photoshop Elements (Adobe, San Jose, CA, USA). The eyedropper was set at 5 × 5 pixels and placed over five points along the feather images (see Electronic supplementary material) where hue, saturation and brightness were measured and then averaged for each feather. Hue defines the colour itself, for example, red in distinction to blue or yellow. The values for the hue axis are expressed in degrees and run from 0 to 360°, beginning and ending with red and running through green, blue and all intermediary colours; in the particular case of flamingo feathers, increasing hue values refers to less red (paler) colours. Saturation indicates the degree to which the hue differs from a neutral grey; the values run from 0%, which is no colour saturation, to 100%, which is the fullest saturation of a given hue at a given percentage of illumination. Brightness indicates the level of illumination. The values run as percentages; 0% appears black (no light) while 100% is full illumination, which washes out the colour (it appears white).
We quantified colour as it is perceived by the human eye. Birds, however, are also sensitive to ultraviolet (UV) light (Bennett and Cuthill 1994), and if uropygial secretions affect UV reflectance, this may have important effects on plumage colour as perceived by birds. However, the uropygial secretions are unlikely to play a major role in modifying plumage UV reflectance (Delhey et al. 2008; but see Griggio et al. 2010).
We used Wilcoxon matched pairs test to compare plumage colour scores before and after chick hatching as well as feathers’ colour before and after pigment removal. Spearman rank correlations were used to examine the relationship between plumage colour and the concentration of pigments on feathers’ surfaces, as well as the relationship between feathers’ colour before and after pigment removal, and the relationship between birds’ age and laying dates. We used Kendall partial-rank correlations to control for base-plumage colour on the relationship between plumage colour and the concentration of carotenoids on feathers’ surfaces, and also to control for the effect of birds’ age on the relationship between plumage coloration and laying dates. Kruskal–Wallis ANOVA was used to analyse monthly variations in plumage colour scores, seasonal variations in the concentrations of pigments in the uropygial secretions and differences in laying dates according to plumage colour scores. Finally, we used χ2 tests to analyse differences in maintenance behaviour of plumage between displaying and non-displaying periods as well as in maintenance behaviour according to neck plumage colour scores. Analyses were performed using STATISTICA (StatSoft 2001). The significance level was alpha <0.05.
Seasonal variations in plumage colour
Pigments in uropygial secretions and feathers
Plumage coloration and laying dates
There are very few previous studies on the effects of uropygial secretions on plumage coloration, and the results are contradictory. While some studies found that preen waxes have no effect on coloration (Reneerkens and Korsten 2004; Delhey et al. 2008; Surmacki 2008), another study found that after application of preen waxes the plumage was more colourful (López-Rull et al. 2010). These differences among studies may lie in variations of the optical properties of uropygial secretions across species. Here, we have shown for the first time that the pigments in the uropygial secretions are the same found in feathers, and that greater flamingos may use these pigments with cosmetic purposes.
It has been suggested that the cosmetic behaviour could have a non-signalling origin, being a side effect of using coloured substances for feather maintenance and without signalling value (Delhey et al. 2007). Here we have shown that the colour of plumage changes with the application of carotenoid-rich secretions over it by the birds themselves, that the intensity of coloration increases with the quantity of pigments applied onto the plumage, that the concentration of carotenoids in the uropygial secretions changes seasonally in accordance with plumage colour, that for the application of such carotenoids the birds use specific behaviour and that the more colourful birds start breeding earlier. All this is consistent with the notion that pigments in the uropygial secretions are used as cosmetic, and that this has a signalling function.
As suggested by López-Rull et al. (2010), cosmetic colours may play an important role in courtship and mate choice, and would update the signal value of plumage colour, mainly by providing a more recent snapshot of the bearer’s quality than colours acquired by moult some months before (Montgomerie 2006). We found a strong correlation between plumage coloration before and after removal of uropygial secretions from feathers’ surfaces. This may indicate that those secretions would add relatively little variation to base colour, i.e. to the colour produced by carotenoids deposited into feathers, and therefore that the use of cosmetic coloration would not have a signalling value. Nevertheless, after the application of uropygial secretions the plumage of greater flamingos was more colourful so that the use of these secretions may function as a signal amplifier by making the signal more detectable by conspecifics (Hasson 1991; López-Rull et al. 2010), and in turn increasing the perceptibility of individual quality.
That the primary function of cosmetic coloration in flamingos may be related to mate choice is supported by the fact that rubbing behaviour was much more frequent during periods when the birds were displaying in groups than outside those periods, and the colour of feathers due to cosmetic coloration faded after egg hatching, likely because the concentration of carotenoids in the uropygial secretions was lower than during displaying periods. In other waterbirds with carotenoid-based plumage colour, it has also been found that plumage colour fades just after the birds start to breed (Shannon 2000; Hays et al. 2006). In the case of flamingos, there may be three reasons for this. First, it may be that breeding birds face a trade-off between using carotenoids for cosmetic purposes or in ‘fighting’ free radicals produced by oxidative stress (see Wiki 1991 and Møller et al. 2000 for physiological functions of carotenoids) during frequent commuting between breeding and distant foraging sites during chick provisioning (Amat et al. 2005). Second, and especially for females just before laying, carotenoids may be re-allocated from uropygial secretions to yolk, as the latter is rich in carotenoids (Fox 1975). And third, carotenoids may be re-allocated to chicks’ food, as adults feed their chicks with a secretion containing carotenoids (Fox 1975).
The flamingos with more coloured plumages started laying earlier, which may have a survival value for their offspring because the first pairs gained control to the best breeding sites (Rendón et al. 2001). This suggests that a secondary function of cosmetic coloration is to signal status during competition for nesting sites. In addition, the fact that rubbing behaviour is also recorded outside the displaying period, though to a much lesser extent than during the displaying period, may indicate that this secondary function of cosmetic coloration to signal individual status may also operate at foraging areas to gain access to the best sites.
In addition to production costs, honest signalling may also have maintenance costs. Because the carotenoids bleach rather quickly when exposed to ambient conditions, frequent reapplication of uropygial secretions rich in these pigments should be necessary to keep the plumage colourful (Delhey et al. 2007). Therefore, the more colourful individuals should invest more in plumage maintenance than the less colourful ones, and as time spent on maintenance cannot be devoted to other activities, the use of cosmetic coloration may be considered as a high-maintenance handicap leading to honest signalling (Walter and Clayton 2005; Griggio et al. 2010). Indeed, we found that rubbing behaviour, which is related to the spreading of uropygial secretions over the plumage with cosmetic purposes, was much more frequently performed by the more colourful birds than by the paler ones.
Given that cosmetic coloration may be related to individual quality, our findings may have important implications for the theories of sexual selection and signalling, highlighting the key role of the manipulation of plumage colour by the birds themselves to improve signal efficacy, as in flamingos. It remains to be determined whether a similar mechanism is at work in other bird species with carotenoid-based plumage coloration.
We thank M. Adrián, O. González, P. Rodríguez, N. Varo and M. Vázquez for helping to capture flamingos and taking samples, and F. Cézilly, K. Delhey, K.J. McGraw, J.J. Negro, J.C. Senar and two referees for commenting on the manuscript. The Consejería de Medio Ambiente of the Junta de Andalucía and ‘Cañada de los Pájaros’ provided facilities. Funding was provided by Ministerio de Educación y Ciencia of Spain with EU–ERDF support (research grants BOS2002-04695 and CGL2005-01136/BOS).
- Britton G (1995) UV/visible spectroscopy. In: Britton G, Liaan-Jensen S, Pfander H (eds) Carotenoids: spectroscopy, vol 1B. Birkäuser, Basel, pp 13–63Google Scholar
- Cramp S, Simmons KEL (1977) The birds of the Western Palearctic, vol 1. Oxford University Press, OxfordGoogle Scholar
- Davies BH, Köst HP (1988) Carotenoids. In: Köst HP (ed) Handbook of chromatography, vol 1. CRC, Boca Raton, pp 3–188Google Scholar
- Eugster CH (1995) Chemical derivatization: microscale tests for the presence of common functional groups in carotenoids. In: Britton G, Liaan-Jensen S, Pfander H (eds) Carotenoids: isolation and analysis, vol. 1A. Birkäuser, Basel, pp 71–80Google Scholar
- Fox DL (1975) Carotenoids in pigmentation. In: Kear J, Duplaix-Hall N (eds) Flamingos. Poyser, Berkhamsted, pp 162–182Google Scholar
- Hill GE (2002) A red bird in a brown bag: the function and evolution of ornamental plumage coloration in the house finch. Oxford University Press, OxfordGoogle Scholar
- Johnson A, Cézilly F (2007) The greater flamingo. Poyser, LondonGoogle Scholar
- Maynard Smith J, Harper D (2003) Animal signals. Oxford University Press, OxfordGoogle Scholar
- Møller AP, Biard C, Blount JD, Houston DC, Ninni P, Saino N, Surai PF (2000) Carotenoid-dependent signals: indicators of foraging efficiency, immunocompetence or detoxification ability? Avian Poult Biol Rev 11:137–159Google Scholar
- Montgomerie R (2006) Cosmetic and adventitious colors. In: Hill GE, McGraw KJ (eds) Bird coloration, vol. I. Harvard University Press, Cambridge, pp 399–427Google Scholar
- Searcy WA, Nowicki S (2005) The evolution of animal communication: reliability and deception in signaling systems. Princeton University Press, PrincetonGoogle Scholar
- Shannon PW (2000) Plumages and molt patterns in captive Caribbean flamingos. Waterbirds 23(Spec Publ 1):160–172Google Scholar
- StatSoft Inc (2001) STATISTICA (data analysis software system), version 6. StatSoft, TulsaGoogle Scholar
- Uchida Y (1970) On the color change in Japanese crested ibis. A new type of cosmetic coloration in birds. Misc Rep Yamashina Inst Ornithol 6:54–72, In Japanese, with English summaryGoogle Scholar
- Vevers HG (1985) Colour. In: Campbell BC, Lack E (eds) A dictionary of birds. Poyser, Berkhamsted, pp 99–100Google Scholar