Feathers

Within-feather variation was an important source of diversity of nuptial plumages. Feathers differed in colour and pattern. All colours in the feather samples that were re-examined (Revisited Dataset) could easily be classified (examples in Figure 4) according to the scoring system (see Figure 2). Each feather had only one or two colours. Black and all hues were found in plain feathers, but more commonly in patterned ones, but pure white was never combined with any other colour. Thus, coloration of patterned feathers was always a combination of black and only one hue. No feathers with two hues were detected (although the intensity of pigmentation could gradually change from base to tip), nor with black and white, nor with black, white and a hue. Many types of patterns were found. Figure 5 shows a number of examples, not a complete series.

Most feathers sampled from each individual were very similar, both in colour and pattern, but some could diverge. Some samples had a combination of feathers with black with one hue plus either plain black feathers, or plain feathers of the same hue. In addition, plain white feathers were found in some samples. No samples were found that had more than one hue and had — besides plain feathers in some instances — highly differing patterns.

Ruff

In the Primary Dataset, 12% of the birds were scored to have only one ruff colour, 73% had two ruff colours, and 15% three. Ruffs were plain white in 4%, plain hue in 1%, and plain black in 8% of the birds. All other birds had combinations of colours. Table 1A gives percentages of colours (primary, secondary and tertiary) and Table 1B patterns (first and second regular, and first and second irregular). An example of how primary and secondary colours associated is shown in Table 2. As colours should be different, all values in the cells of the descending diagonal (from combinations of colour codes ‘11’-‘11’ to ‘51’-‘51’) are zero. The distribution in the non-diagonal cells differed from random (χ²₈₉ = 566, P < 0.001). Combinations between black and hue, black and white, and white and hue were relatively common. In contrast, combinations between different hues (the values within the inner rectangle) were rare (zero or very low). If hues combined at random, the expected sum of the non-diagonal values in the inner rectangle is 105.5, but the observed sum was only 31. The difference is highly significant (χ²₅₅ = 239, P < 0.001). Thus, primary and secondary colour rarely represented two hues.

Figure 4.

Samples of the colours of the feathers. The fifth panel (31) represents a sample of the second generation (alternate) plumage, as this colour is rare in nuptial plumages. All other samples are from third generation plumages.

Figure 5.

Samples of different patterns of feathers. From left to right: wide bars, broken bars, two variants of fine bars, black threads, black spots, four different variants of fairly large patches at the tip, and black and hue separated by the shaft.

Table 1.

Percentages in the Primary Dataset of (A) the primary, secondary and tertiary ruff colours, the primary and secondary head tuft colours, and (B) the first and second regular and irregular patterns.

Table 2.

Frequencies (counts) in the Primary Dataset of combinations of primary and secondary ruff colours (χ²₈₉ = 566, P < 0.001, according to Goodman 1968). Bold numbers in the dark grey cells are higher than expected by chance, numbers in the white cells are lower than expected. The inner rectangle represents the combinations of hues.

Table 3.

Frequencies (counts) in the Primary Dataset of combinations of primary ruff colours and primary colours of head tufts (χ²100 = 3094, P < 0.001). Bold numbers in the dark grey cells are higher than expected by chance, numbers in the white cells are lower than expected by chance. The inner rectangle represents the combinations of hues.

Considering all ruff colours (primary, secondary and tertiary), combinations of two different hues were scored for only 3% of the birds (Tables S1–S3) and combinations of three different hues for 0.2%. Closer inspection (Revisited Dataset) revealed that almost all combinations of different hues could be classified in one of three categories: (1) including hue 21 (straw: almost white) rather than being scored as pure white, (2) including hue 41 (brown: almost black) rather than being scored as pure black, or (3) including two almost similar hues. All these could be due to ambiguity in the scoring system. In addition, small deviations of dominating colours might be caused by environmental influences during plumage development. In summary, ruff coloration consisted of one (white, hue or black), two (white + hue, white + black or hue + black) or three colours (white + hue + black), but we never observed two distinct hues in the same ruff.

Regular patterns of the ruff emerged with patterned feathers. Feather patterning in the ruff was rather uniform in these birds. Irregular patterns of the ruff appeared when (some) plain (often white) feathers diverged from the dominant feather colour. Regular and irregular patterns associated in non-random ways (χ²₃₅ = 501, P < 0.001) (Table S4); generally, non-plain regular patterns excluded irregular patterns. Regular patterns were highly associated with the first and second colours of the ruff (χ²₇₀ = 738 and 530, both cases P < 0.001), but not with the tertiary colours (Tables S5–S7). Irregular patterns were clearly associated with primary, secondary and tertiary colours of the ruff (χ²₅₀ = 342, 250 and 92, all cases P < 0.01) (Tables S8–S10).

Head tufts and association with the ruff

Percentages of primary and secondary colours of head tufts are presented in Table 1A. As in the ruffs, associations between primary and secondary colours of head tufts were non-random (χ²₈₉ = 5 73, P < 0.001, congruent with Goodman 1968), and combinations between different hues were rare (26 observed; 80.8 expected) (Table S11).

In many, but certainly not all, individuals, the colours and patterns of the head tufts and the ruff were similar. Most obvious were the relations between the colours. An example (primary colour of ruff compared with primary colour of head tufts) is shown in Table 3. It is striking that ruff and head tufts tended to have similar colours (values in the dark grey cells of the descending diagonal). In addition, various hues (code ‘21’, ‘22’, ‘23’, ‘32’) were associated with black (‘51’). Almost the same associations were found in most other colour comparisons (primary and secondary colour ruff with primary and secondary colour head tufts) (Tables S12 – S15). The distributions were different from random, all cases χ²₁₀₀ > 1000, P < 0.001. In most cases the values of the cells in the descending diagonal were much higher than expected by chance. The association between black and the hues was clear for the comparison between secondary colour ruff and secondary colour head tufts, but not for the other combinations. Comparisons between the tertiary colour of the ruff and the two colours of head tufts also gave significant non-random distributions (χ²₈₁ = 228 and χ²₉₀ = 142, both cases P < 0.001), but no obvious trends (Tables S16–S17). Both primary and secondary colours of the ruff often reappeared as primary or secondary colours in the head tufts. Primary and secondary colour of the ruff might reappear in the same way in the head tufts, but they might also interchange. No associations could be established between the patterns of ruffs and head tufts, as the latter had not been scored in the Primary Dataset. The regular and irregular pattern of the ruff, however, appeared to be associated with the primary and the secondary colour of the head tufts (Tables S18–S21). In all four comparisons the distributions were different from random (χ²₇₀ = 579 and 382, χ²₅₀ = 290 and 246, all cases P < 0.001).

General ‘plumage category’ as scored in the Revisited Dataset was significantly associated with the properties ‘plain parts’, ‘general pattern for ruff and head tufts’ and ‘presence of strongly contrasting parts’ (χ²₂₁ = 555, 566 and 251, all cases P < 0.001) (Tables S22–S24). The major associations are summarized in Table 4. Three of the nine possible plumage categories appeared to be rare or absent. Portraits and feather samples of typical representatives of the six remaining categories are displayed in Figure 6.

Table 4.

Major associations (combinations that occur remarkably more often than expected by chance) of plumage categories (ruff + head tufts) with ‘parts that remain plain’, ‘general pattern’ and ‘presence of contrasting parts’ (Revisited Dataset).

Figure 6.

Portraits and feather samples of six males, typical representatives of the general plumage categories: 1461959 (white—white), 1337634 (white—hue), 1337636 (hue—hue), 1353498 (hue—black), 1353460 (black—hue) and 1409083 (black—black).

Combined nuptial plumage coloration and diversity

For birds in complete nuptial plumage, we rarely scored more than one distinct hue (ruff and head tufts together, see Table 3). In the central part of the table, within the inner rectangle, the values in cells are zero or very low, except for the diagonal. The sum of these non-diagonal values (combinations of different hues) is 18, which is below the expected value of 253.0 if the hues associated at random (χ²₅₅ = 1737, P < 0.001). Similar trends were found in the other colour comparisons between ruff and head tufts (Tables S12–S17). In all cases the observed values were considerably lower than the expected ones. This indicates that, as with the finding for ruffs alone (see above), in the large majority of cases only one hue is found besides black and/or white. This can also be concluded from the Lek Dataset (15 males had no hue — only black and/or white, 78 males had one hue, and 5 were scored with two hues that were almost identical in all cases). The Revisited Dataset revealed that the original descriptions (Primary Dataset) in which two or — in very rare cases — three hues had been distinguished, quite often referred to hues that were very similar and could fade into each other within one and the same feather. Almost all cases with more than one hue could be due to measurement error (see Methods). In fact, only one out of 781 photos seems to represent a case of a male with two distinct hues in its nuptial feathers (Figure 7). All other plumages could easily be classified as pure white, white with one hue, white with black, white with one hue and black, one hue with black, or pure black. In a few cases ‘white’ seemed to be close to a very light hue, and in a few other cases ‘black’ seemed to be close to a very dark hue. Plumages with only one pure hue (without any white or black patterning) were not seen, and thus seem to be very rare among Ruffs, but do exist (D.B. Lank, pers. comm.).

Figure 7.

Head and feather sample of male 149101, apparently with two different hues in its nuptial plumage.

Although the nine plumage characteristics presented in Table 1 can combine to an impressive number of different phenotypes (about 5 × 10⁶), most of these possibilities did not actually occur. As we have seen, combinations of different hues are excluded. Nevertheless, in our sample of 1814 males we observed 801 different combinations: 565 combinations were scored only once, but various combinations more often, some of them even more than a dozen of times. Three examples: (1) 31 males plain white ruff and head tufts, (2) 24 males gold straw ruff with black moderate flecking, head tufts black with gold straw, and (3) 11 males black ruff and head tufts with light rust broken bars.

Wattles in association with feather coloration

Almost all males that were caught on leks in late April and May had well developed wattles (Lek Dataset). In contrast, males caught on migration in early and mid-April did not have visible wattles. Thus, wattles develop after most nuptial feather growth is completed. All wattles of an individual had the same colour, but there was variation between individuals. Colours ranged from bright yellow (lower male in Figure 6), to orange (second male in Figure 6), and bright red (third male in Figure 6). This might suggest that wattle colour is associated with the yellowish and reddish hues of the feathers. In the Revisited Dataset, wattle colour could be scored in 375 cases and thus be related to plumage colour. Three categories of wattle colours (yellow, orange, red) were tabulated against five categories of plumage colours: white (code: ‘11’), brown (‘21’ + ‘31’ + ‘41’), red (‘22’ + ‘32’ + ‘42’), yellow (‘23’ + ‘33’ + ‘43’), and black (‘51’). An example (wattle colour compared with primary colour of ruff) is shown in Table 5. The distribution of the frequencies was different from random (χ²₈= 17.3, P = 0.027 with primary colour of ruff, and χ²₈ = 22.5, P = 0.004 with primary colour of head tufts) (Tables S25–S26). However, there were no indications of an association between wattle colour and plumage hue. Instead, white plumages co-occurred with red wattles about twice as often as expected.

Table 5.

Frequencies (counts) in the Revisited Dataset of combinations of wattle and primary ruff colours (χ²₈= 17.3, P = 0.027). The bold number in the dark grey cell is higher than expected by chance, numbers in the white cells are lower than expected.

Table 6.

Average (A) body mass (g) and (B) wing length (mm) for birds with the same primary colours of ruff, primary colours of head tufts, regular pattern of ruff, and irregular pattern of ruff (Primary Dataset). Bold values in the dark grey cells are significantly higher than at least one of the values in the same series, values in the white cells are significantly lower (P < 0.05, t-tests for comparisons between the first and second column, Games-Howell post-hoc tests for comparisons among all colours or patterns. ANOVAs are based on the complete range of colours (including white) or patterns (including pattern 1).

Associations between plumage type and body size

Body mass and wing length in the Primary Dataset varied considerably among individuals (resp. 120–335 g, mean = 233; 163–207 mm, mean = 192). Average body mass and wing length also differed between groups of individuals with similar colour or pattern (mass: 222–244 g, SD = 20-40; length: 190–193 mm, SD = 2-5; Table 6). However, differences were small, in the order of 10 g for body mass (less than 5% of the total), and about 1 mm for wing length (less than 1% of the total). Body mass was lower and wing length shorter in birds that had white as the primary ruff colour (code: ‘11’) than in the non-white birds (t-tests, P = 0.001 and 0.017, resp.). For birds that had white as the primary tuft colour this trend was not significant (t-test, P = 0.066 and P = 0.052, resp.). Birds with regular patterns had the same body mass as those without (t-test, P = 0.995), but had longer wings (t-test, P = 0.002). Birds with irregular patterns had a lower body mass and shorter wings than the birds without (t-tests, P < 0.001 both cases, Table 6). Variation in body mass and wing length in most cases differed significantly among the colours and patterns (ANOVAs body mass: between primary ruff colour F_10,1762 = 2.10, P = 0.022; primary tuft colour F _10,1757 = 2.51, P = 0.005; regular ruff pattern F_7,1763 = 1.52, P = 0.156; irregular ruff pattern F _5,1762 = 5.40, P < 0.001; ANOVAs wing length: between primary ruff colour F_10,1800 = 1.33, P = 0.207; primary tuft colour F_10,1795 = 3.24, P < 0.001; regular ruff pattern F_7,1801 = 2.18, P = 0.033; irregular ruff pattern F _5,1800 = 5.01, P < 0.001), but the separate colours did not statistically differ from each other and only a few patterns appeared to differ from other patterns (Games-Howell post-hoc tests, Table 6).

Timing of the moult

In the Primary Dataset, primary colour of the ruff was the only characteristic that varied significantly with moult phase (χ²₃₀ = 48.6, P = 0.017, Table 7) (Tables S27–S33), because the sample of final phases had more white males than expected by chance. Yet, there were no differences between catching date distributions of the different primary colours (ANOVA, F _10,1803 = 1.24, P = 0.26). We examined the association between moult phase and the presence of secondary and tertiary colours, and regular and irregular patterns (Tables S34–S38), and found that only the presence of secondary ruff and head tuft colours varied with phase (χ²₃ = 24.5 and 28.0, P < 0.001 in both cases). In particular, secondary colours tended to occur more often than expected by chance in the samples of the early phases.

In the Revisited Dataset, the distributions of moult phase against plumage category, plain parts of plumage, general pattern, and presence of contrasting parts differed significantly from random (respectively χ²₂₁ = 37.8, P = 0.014; χ²₉ = 30.6, P < 0.001; χ²₉ = 24.4, P = 0.004; χ²₉= 17.0, P = 0.049) (Tables S39–S42). Similarly to the Primary dataset, the white—white and white—hue plumage categories, and the plain patterns were most frequent amongst birds with almost completed ruff feather growth (Table 8). The coarse patterns and contrasting patches were most common in birds in the early stages of moult, whereas plain plumages were most common in birds in final moult stages. This could mean that plain birds arrived later, or, that the plain feathers grow later and cover the patterned ones.

Table 7.

Frequencies (counts) in the Primary Dataset of combinations of moult phase and primary ruff colours (χ²₃₀ = 48.6, P = 0.017). Bold numbers in the dark grey cells are higher than expected by chance, numbers in the white cells are lower than expected.

Table 8.

Major associations of moult phase with ‘plumage categories’ (ruff + head tufts), ‘parts that remain plain’, ‘general pattern’ and ‘presence of contrasting parts’ (Revisited Dataset).

Development

The third feather generation (supplemental plumage) of a Ruff appears in the course of April (e.g. Drenckhahn 1968, Jukema & Piersma 2000), starting with feathers of breast and shoulders (ruff) and followed by feathers of the head (head tufts). By the end of the moult, after the small feathers near the bill are lost, the temporarily enlarged facial wattles become visible. Before that time, between January and April, the large wing coverts, mantle, scapulars, tertials and, commonly, half of the number of rectrices are moulted, partly in the wintering area (e.g. Schmitt & Whitehouse 1976, Pearson 1981). These feathers most probably represent the second plumage generation (alternate plumage), directly replacing the basic plumage (Jukema & Piersma 2000). Nevertheless they are quite colourful, more so than the second generation feathers of breast, ruff and tufts. Bar-tailed Godwits (Limosa lapponica baueri) also grow an alternate plumage of barred feathers, which after just 3–5 weeks is replaced with supplemental red feathers (Conklin & Battley 2011).

During pre-supplemental moult plumages gradually change. Coarse patterns and contrasting patches tend to become rare and plain, and white plumages become more common (see Timing of the moult). This seems to reflect the process of plumage perfection during moult. Initially, the male's plumage is ‘sloppy’ because of the mixing of feathers of both the alternate and supplemental plumage. This sloppy appearance is strengthened because the growth schedules of various feathers differ. By the end of moult the plumage looks wellcared for. It is possible that plain feathers, especially the black ones, develop somewhat later or slower tha nfeathers that are patterned. Observations on males that were kept in captivity suggest that the process of plumage perfection also occurs from year-to-year during their first years of life (JvR unpubl. data). Moults seem to proceed more smoothly with increasing age, and nuptial plumages seem to become less patterned and more brilliant.

New feathers are created by the same feather follicles (e.g. Prum 1999, Yu et al. 2004) at the same feather tracts as the feathers of the previous plumage. Yet, the new feathers differ in size, shape, pattern and colour from second generation feathers. The old were short and striped, the new feathers are longer and more colourful (Jukema & Piersma 2000). These differences are the result of changing physiological factors, in particular the amount of circulating testosterone (e.g. Kimball & Ligon 1999, Bókony et al. 2008, McGraw 2008, Roulin & Ducrest 2011). For instance, the production of melanin pigments is enhanced by testosterone. Although the relationship between testosterone and plumage in the Ruff has been demonstrated by castrating males (van Oordt & Junge 1936) and by implanting testosterone in females (Lank et al. 1999), in this species the female mimics are strikingly deviant. These males have very large testes in comparison to other males, but do not develop large and showy feathers and do not have a dark (melanised) appearance (Jukema & Piersma 2006). It is possible that they use a mechanism that neutralizes the effects of testosterone on growing feathers, and maybe, also on their behaviour. This happens, for instance, in male ducks, that aromatize testosterone into oestrogen at the time they moult into eclipse plumage (e.g. Kimball & Ligon 1999).

In shorebirds melanins are the major (possibly the only) feather pigment (e.g. Toral et al. 2008, Stoddard & Prum 2011). There are two groups of melanin: eumelanin (black and dark brown) and pheomelanin (reddish and yellowish hues). They are produced in cells called melanocytes, stored in small cell organelles (melanophores), that can be transferred to other cells, for instance to the cells of the developing feather (keratinocytes). Melanocytes are distributed over the skin, quite often in particular patterns. As older satellite males keep an increasing area of white head feathers during the whole year (van Rhijn 1983, 1991, D.B. Lank pers. comm.), some regions appear to lose their melanocytes in the course of years. It is plausible that pheomelanin and eumelanin are formed in succession at the same site and are subsequently packed as a core consisting of several particles of pheomelanin surrounded by a layer of eumelanin (Bush et al. 2006, Ito & Wakamatsu 2008). Pigments have to be distributed in a particular way within a growing feather to create the feather pattern. With rather simple reaction-diffusion models the complex patterns that have been found in feathers can be simulated (Prum & Williamson 2002).

The variety of ruff and head tuft colours seems to be sufficiently explained by the existence of feathers without pigments (white feathers) and feathers with eumelanins and/or pheomelanins (black and various hues). Although the mechanistic details are unknown, it is possible that different types of pheomelanins are involved in the different hues, as has been established for human hair (e.g. Piletic et al. 2010). It is also possible that they arise by different blends between pheomelanin and eumelanin (e.g. McGray et al. 2005), as has been shown for human hair too (e.g. Ito & Wakamatsu 2011). Most probably an individual male produces only one type of pheomelanin (or perhaps one blend of pheomelanin with eumelanin) that causes its characteristic hue.

In addition, some males with plain black feathers stand out on leks and in the staging areas because of a bluish or greenish gloss over their feathers (iridescent structural coloration). These glosses have not been scored because they cannot reliably be detected on photos. Whereas colours from melanins are hardly affected by health and food availability at the time the feather grows (but see McGraw 2007, 2008), structural coloration seems to be strongly influenced by these factors (McGraw et al. 2002). Similarly, the colours from some other pigments, such as carotenoids, are related to the condition of the animal (e.g. Hill 1999). These pigments have not yet been found in the plumage of shorebirds, but the bright coloration of the wattles near the bill base (yellow, orange and red) suggest that carotenoids are involved. Condition in male Ruffs was not found to be related to wattle coloration, but instead to the extent of the area with wattles (Lank & Dale 2001).

Genetics

Breeding experiments have shown that variation in mating strategies in Ruffs is based on genetic factors (Lank et al. 1995, 2013), but the genetic factors underlying nuptial plumage variation remains unknown. Although our analysis was restricted to phenotypic variation, our data may help to generate hypotheses on genetic control of nuptial plumage. In fact, we tried to derive a number of general properties from our results that may together define an individual-specific nuptial plumage: (1) whiteness (presence of skin areas without melanocytes), (2) blackness (proportion of eumelanin among the melanins produced in melanocytes), (3) hue (type of pheomelanin or blend of pheomelanin with eumelanin), (4) feather pattern (diffusion characteristics of pigments in growing feathers), (5) plumage category (combination of ruff and head tufts), and (6) wattle colour. The observation that individual nuptial plumage patterns do not change with age suggests that the molecular processes of feather and wattle growth are fixed within an individual, but it remains unclear whether and how the underlying genes or gene expression varies between individuals. Below we review what is known so far, but bear in mind that nuptial plumage development involves an unknown number of coding and regulatory genes.

The first property, whiteness, is the most likely candidate for separating satellite males from independent males. Most satellite males have areas of skin, possibly without (active) melanocytes, that produce white feathers. This is not necessarily restricted to the supplemental plumage, but may also happen in the other plumages. Areas of skin that produce white feathers, for the supplemental plumage are either absent or relatively small in independent males. Satellite males have been shown to possess one or two dominant satellite alleles (Lank et al. 1995). The finding that all pure white males (see Figure 1) observed on leks were satellite males, suggests that ‘a large area of frontal skin without (active) melanocytes’ is a pleiotropic effect of the satellite allele. Alternatively, the ‘whiteness’ factor is closely linked to the satellite allele. However, reproductive strategy is not fully associated with white plumages (Figure 8). Apparently, in addition to the ‘satellite allele’, size and position of the areas without (active) melanocytes are also under control of other genes.

The second and third properties, blackness and hue, are regulated by the proportion of eumelanin among the melanins and by a type of pheomelanin or an unknown number of blends of pheomelanin with eumelanin, and might be quantitative traits under influence of a number of genes of which the effects are added. The potentiality of these properties could appear from the highly variable melanocortin-1 receptor (MC1R) gene (Mundy 2005). Differences between the hues may also relate to variations in the intensity of pigmentation (McGraw et al. 2005).

The fourth, patterning of feathers, has been simulated with reaction-diffusion models (Prum & Williamson 2002). These simulations suggest that the number of (molecular) instructions is not necessarily high for creating a complicated colour pattern in a feather. The way that pheomelanin pigments are packed together in granules with eumelanin (Ito & Wakamatsu 2008) could facilitate this process.

Figure 8.

Reproductive strategy is not fully associated with white plumages. The two males at the right side are both independent males. Their ruffs are (almost) white and their head tufts black. The male at the left side of the picture is a satellite male with a white ruff and light brown head tufts. The male on the background is probably a satellite male.

The fifth, plumage category, is constrained by the factors discussed above. Possibly only for the non-white plumages extra information is required to control the relationship between ruff and head tufts. Ruff and head tufts may have almost similar feathers or somewhat reversed characteristics, especially in blackness. A possible cause is variation in the density of (active) melanocytes in the skin.

The sixth, colour of wattles, has been suggested to be based on one gene with two non-dominant alleles (van Rhijn 1991). Heterozygotes would develop orange wattles, one homozygote red wattles and the other yellow wattles. Their frequencies agreed with the Hardy-Weinberg law (van Rhijn 1991, p. 62). Wattle colour remains largely constant over years, but seems to be independent of other plumage characteristics. For that reason it is likely to be under control of independent genes. Considering the brightness of these colours, pigmentation by carotenoids is more likely than pigmentation by melanins (e.g. Grether et al. 2004); also, carotenoids are the major pigments in the skin of birds (e.g. Olson & Owens 2005). Carotenoids have to be obtained from food resources, implying that not all individuals possess these pigments at any time (e.g. Olson & Owens 2005). As in melanins, pigmentation by carotenoids may be under control of testosterone (e.g. Blas et al. 2006). In addition, the intensity of the carotenoid pigmentation is expected to depend on the condition of the male (e.g. Hill 1999, Simons et al. 2012). In some cases the variation in carotenoid pigmentation could be ascribed to genetic differences (e.g. Eriksson et al. 2008, Walsh et al. 2011).

We suggest that the molecular regulation of wattle colour operates independently of plumage, but consider it plausible that the regulatory factors for plumage are interdependent, except those for whiteness.

Evolution

The diversity in the plumage of Ruffs continues to raise evolutionary questions (e.g. Roulin 2004). Generally breeding plumages are thought to be an effect of mate choice for quality, or attractiveness (e.g. Endler & Basolo 1998, Ryan 1998, Arnqvist 2006). A choice for quality has direct benefits for the fitness of the offspring (health and status) and may be regarded as a case of natural selection. A choice for attractiveness has only indirect effects on the fitness of the offspring (attractive sons) and should be regarded as a pure case of sexual selection. In Ruffs, only very few plumage characteristics may indicate quality (e.g. Lank & Dale 2001). So far it has proved impossible to correlate plumage characteristics with female interest or mating success (Shepard 1975, Hill 1991).

Two other explanations have been given for plumage diversity in Ruffs: (1) to signal status (satellite or independent male), and (2) to signal individual identity (van Rhijn 1983, Dale et al. 2001, Lank & Dale 2001, Dale 2006). Clearly, satellite males will benefit from signalling their mating strategy. Their white plumage provides a “passport” for admission to mating sites and in combination with particular postures, restrains the territory holders to attack them (Hogan-Warburg 1966, van Rhijn 1991, Lank & Dale 2001). It is less clear why an individual male benefits from displaying his identity in such a conspicuous way. It may be helpful for a satellite male to recognize its tolerant host, or for a territory owner to recognize its neighbour (and vice versa). However, the central question is whether male reproductive success is improved by being identifiable. Females may make the decision to mate with a particular male after several visits to various males. If these visits are scattered over days and over locations, easy identification of males is advantageous for females. But is it also advantageous for males? It may be advantageous for the one that was finally chosen, but certainly not for the others that were not chosen for mating.

Being identifiable is beneficial for territorial males that have settled on a lek. Their neighbours have tested their fighting abilities and will avoid their territory. This will minimize violence to ritual attacks or very short fights. The extreme plumage diversity among Ruffs may be useful for individual identification because they do not sing (Lank & Dale 2001). After all, individual recognition is also widespread in largely monomorphic birds and other animals that live in groups (e.g. van Rhijn & Vodegel 1980). However, identification may not be beneficial for a territorial male that is newly settling on a lek. A new male is vigorously attacked by all neighbouring males and has to fight heavily for a long time before it is tolerated by the others. Only a very small number of males get settled without such fighting; in these cases the settling male closely resembled the previous owner of the territory or another territorial male that had recently left the lek (JvR unpubl. data). Thus, it is questionable whether a male benefits from being identifiable.

Thus, and perhaps hardly surprisingly, even after our detailed description and analysis of nuptial plumage variation, the evolutionary drivers of the extreme diversification in Ruffs remain beyond our grasp. Yet, we suggest that sexual selection is involved in major ways. If, as seems to be the case for Ruffs, females do not select males with particular colours or patterns, they might still select males with odd plumages as these are most stimulating for their sensory system (e.g. Knoppien 1985, Eakley & Houde 2004, Hughes et al. 2013, van Rhijn 2013). Also, odd males are conspicuous, and for that reason alone they could be attractive mating partners. A preference for odd mates creates diversity within a species without direct fitness advantages of ‘being different from most others’, but it could initiate a self-propelling, ‘run-away’ process. But why would such selection for diversity be so rare?