The data presented in this report are based on observations of the tail structures of museum specimens kept at the Royal Ontario Museum, Toronto, Canada, and the American Museum of Natural History, New York, USA. Only specimens in good condition were included in this study. A total of 21 samples of blue fowl ( P. cristatus), including two albinos and two hybrids (between strains originating from Bangkok and Cameroon), were examined. The wild blue fowl samples originated from India (N = 3), Sri Lanka (N = 2), Kenya (N = 1), or were captive samples originating from the United States (N = 4 blue, N = 2 albino). The remaining 11 samples were of unknown origin and included two hybrids. We also had access to seven green fowl ( P. muticus) samples at the Natural History Museum, which originated from Malaysia (N = 4), Bangkok (N = 1), or were of unknown origin (N = 2). Wherever dates were noted, most of the field samples were collected during the early 1900s. Conversely, the captive samples were collected as recently as 1942.

We counted the number of eyespots, eyespot feathers, and fishtail feathers displayed by the included specimens ( Figure 3). In the cases where an eyespot was missing due to damage, we counted it as if it were present to estimate the total number of eyespots. Very few of the samples had a uropygium that was in good condition. Of the museum specimens six had intact uropygia in P. cristatus and one in P. muticus. Wherever we were certain that the integrity of the sample had not been compromised, we counted the rows of feather imprints on the uropygium ( Figures 2B, 4). Dryad: A long tail of truth and beauty: The developmental basis of complexity, symmetry, and beauty in the evolution of the peacock’s tail. Dryad My Dataset: doi: 10.5061/dryad.1g1jwstwg.

To illustrate the eyespot symmetry exhibited by trains in displays ( Figure 5), we made use of Adobe Illustrator’s built-in features (Adobe Photoshop and Adobe InDesign) provided by Graphic Designers’ services at Media Production Services, McMaster University. Briefly, Adobe Illustrator was used to map concentric circles following eyespot distribution on the train in the display by using Figure 1 as a template. These sets of concentric circles were used as the basis for text paths to determine how eyespots may distribute themselves upon the unfolding of the train for display.

To simulate and illustrate the train’s eyespot distribution, we used the following observations and/or assumptions: (1) Based on our overall observations (7 P. cristatus, and 1 P. muticus), we concluded that each cell on the uropygium ( Figure 1B) represents the base of a corresponding feather (attempting to determine direct one-to-one correspondence would have required damaging the specimens); (2) we used a row of 10 or 11 dots in the shape of the (oval) uropygium to represent train feathers but we used a flat and not a convex surface as the latter was not possible. In the result section we discuss why using a convex surface in the simulation would not have changed the outcome; (3) In line with our observation ( Figure 1), we used an arbitrary constant distance between spots within a row and we progressively increased the size of the dot spots between rows, from the bottom (newest) to the top (oldest) feathers, to represent the size of the eyespots as seen on the train; (4) We assumed that each dot represents an eyespot at the end of the feather which is supported by our observation; and (5) we only simulated 2-dimensional position of eyespot distribution and not their three dimensional distribution on the train surface. We simulated four spatial feather arrangements: 10/11 zigzag, 10/11 parallel, 10/10 zigzag, and 11/11 zigzag ( Figure 5). The feather grid was expanded bilaterally symmetrically to mimic the animal’s train expansion. It is important to point out that the simulated surface shown in Figure 5 represents the front, flat face of the train.

To uncover the anatomical development plan that determines the bilateral symmetry of eyespots on the train, we examined the uropygium (N = 6) of museum specimens. The uropygium is a fleshy and bony structure at the posterior extremity of a bird’s body that supports the tail feathers. Tail feathers are attached to the dorsal (convex) side of the oval-shaped uropygium, which serves as an anchor plate ( Figure 3). The ventral (concave) side of the anchor plate exhibits several interesting features: 1) It has basal impressions of tightly-packed, parallel rows of feather follicle insertion points; 2) The feather follicles grow progressively larger from the anterior (bottom, younger) to the posterior (top, older) end and are laid out in sequential but alternating rows of 10/11, corroborating the alternate arrangement of 10/11 feathers first reported by Dakin and Montgomerie (2011) except that the alternate rows have a zigzag pattern; 3) The posterior feather follicles radiate in an arch, mimicking the fan formation shape of the train in the display. The convexity of the anchor plate makes the train a three-dimensional structure (i.e., an oval trapezoid-shaped dish), such that the feathers are projected outwards at different angles and lengths. Notably, the anchor plate itself appears to be a product of continuous growth and the sequential addition of new rows of feathers each year. The one-to-one correspondence between the pattern of the feather follicles (size, alternate arrangement, and progression from anterior to posterior) on the ventral side of the anchor plate and the symmetrical positions and size of the feathers on the dorsal side is unmistakable and is supported by our observations ( Figure 2).

To determine whether any variation exists in the anatomical development plan found on the uropygium that would affect the number of eyespots, we examined several specimens from the Royal Ontario Museum and the American Museum of Natural History. The results of this analysis, based on a limited number of specimens in good shape, are summarized in Table 1 and Figure 4. First, the number of feather rows, as determined by the number of follicle rows on the uropygium, varied from 17 to 19 (N = 7), whilst the anterior two to three rows had minor eyespots. Notably, this result is consistent with the findings of Dakin and Montgomerie (2011). The anterior-most rows of eyespots appear minor due to their slow growth and maturity. If one or a few rows of feathers are added each year, then based on the peacock’s longevity of about 20 years in nature ( Lindstedt and Calder, 1976) the number of eyespots can reach over 200; however, the late-aged, developmentally immature eyespots (small size and lack of full coloration) will remain insignificant in their effect on female choice.

Second, the zigzag arrangement of feathers appeared as an invariant trait in our sample ( Table 1). While this one-to-one correspondence between clearly visible follicle insertion patterns and train morphology may be expected, our data suggest that the developmental plan does not permit random (one feather at a time) variation in eyespot number ( Table 1).

Third, the number of fishtail feathers varied around the modal number of 30, 40, or 50 (N = 21), which is what we would expect if fishtail feathers were produced by three, four, or five rows of feather follicles during development, respectively ( Figure 4). In our museum samples, the highest number of fishtail feathers observed was around 40. However, Figure 1 indicates that the number of fishtail feathers can exceed 50.

Fourth, the number of eyespot feathers varied from 101 to 161 (N = 21), not counting the minor feathers. This variation is attributable to individual variation in terms of age, time of year at capture, and any damage from prior handling at the museum.

Fifth, we speculate that number of eyespot feathers and feather length are independent traits that vary asymptotically as a function of age. This is consistent with Manning’s (1989) finding that eyespot number increases during the first 4 to 7 years of the animal’s life and thereafter increases more slowly or remains effectively constant.

We also had access to a small number of P. muticus (green peacock) specimens ( N = 7) from Southeast Asia ( Table 1; Figure 4) for comparative analyses, which could provide clues as to whether peacock train morphology reflects a similar developmental pattern across taxa or whether sexual selection and other environmental factors have resulted in different trajectories of train morphology. In the green peacock, the number of fishtails ranged from 38 to 39 (around the expected mode of 42), while the number of eyespots ranged from 115 to 154.

Data from both species appeared to be homogenous ( Figure 4). One green peacock specimen was in pristine condition, which enabled us to assess the anatomical development pattern of feather follicle rows in the uropygium. As per the blue peacock, feather follicle rows in the green fowl followed the same alternating 10/11 arrangement, suggesting that this developmental arrangement may be an invariant feature across peafowl species. Interestingly, similar alternating patterns of train feather arrangements are also found in the wild turkey, Meleagris ocellata, a member of the same family (Phasianidae). This observation entertains further research on whether the alternating pattern of feather arrangement might be common across other members of the family and thus an invariant developmental trait.

We investigated the significance of the arrangement of feathers in rows of 10/11 and their “zigzag” alignment (see Figure 2). To achieve this, we used graphic design software to reconstruct the unfolding of the peacock’s open train based on the developmental layout of the feather follicle insertion points observed in the anchor plate ( Figure 5, left). For simplicity, we assumed that the feather follicle insertion points were eyespots without the feather stalks. Using Adobe Illustrator, we modeled the fan formation in “display.” We subjected the eyespot distribution on the anchor plate to a uniform, bilaterally symmetrical force to spread the eyespots to the left, right, and upward, thereby pushing them radially and uniformly outward in a curved semi-circular space to mimic the fanning pattern of the train during display. Remarkably, our reconstruction yielded an eyespot distribution that is very similar to that observed on the peacock train (compare Figure 5A to Figure 2), with five feathers on either side of the mid-feather in each row. As expected, we found a one-to-one correspondence between the condensed, developmental–anatomical layout of the feather follicle insertion points on the uropygium and the symmetrical distribution of eyespots on the train. It is important to point out that the zigzag arrangement of cells, when spread out uniformly, gives rise to the well-known hexagon cellular packing (compare Figure 5A and Figure 1). We did not simulate a convex surface as this was not possible, however doing so would not have changed the distribution pattern (hexagonal) of eyespot as we were simulating 2-D image of eyespots distribution as seen from front and not their 3-D distribution patterns. In other words, the simulation results presented in Figure 5 show what the imprint/picture would look like if an expanded train were compressed on a flat surface. This is what we would see from a distance.

We further explored what the eyespots distribution pattern would be like if the feathers were arranged in 10/11 parallel rows instead of the zigzag arrangement that we observed in this study. Notably, the train eyespot pattern that we obtained is remarkably different from the 10/11 zigzag pattern. Our reconstruction of a parallel arrangement yielded a palm-leaf-like pattern that fanned out in parallel rows of eyespots rather than the pattern observed in a peacock’s train on display ( Figure 5B). While both patterns are striking, the 10/11 zigzag arrangement yields a denser and uniformly symmetrical arrangement of eyespots, as seen in the animal.

Because the above results raised the question of why 10 or 11 rows were observed, we simulated a 10/10 parallel and 11/11 parallel pattern; the results are shown in Figures 5C and D. Although the 10/10 and 11/11 patterns were like the 10/11 parallel pattern, they showed certain interesting differences. While the feathers lined up in equally spaced, straight lines over the entire span in the 10/11 parallel pattern, the 10/10 and 11/11 parallel patterns resulted in the equal spacing of individual eyespots towards the end of the train, with wavy lines in the center. It is obvious that it was not the zigzag pattern itself but the zigzag pattern of 10/11 that gave rise to the uniform distribution of eyespots on the train. The 10/11 parallel pattern may not be developmentally possible.

Feather is an important non-embryonic model of animal development (for a recent review see Chen et al., 2015), and much research has been done both anatomical and molecular involving the role of genetic and epigenetic mechanisms ( Jiang et al., 1999, 2004; Maini et al., 2006; Chen et al., 2015). Feathers develop from feather follicle buds on the surface of the epidermis ( Prum, 1999; Chen et al., 2015). Pigmentation based colors in plants and animals can be explained by local variation in gene expression. In birds, movement of melanocytes up the hollow core of the feather is said to control the variation in plumage color ( Prum and Williamson, 2002; Chen et al. 2015). In “structural color,” such as in peacock feathers, the colors are the result of light reflection from the modification and anatomical diversity of the barbules’ photonic structures ( Zi et al., 2003; Freyer et al. 2018).

We were motivated by the superficial but interesting structural similarity between the iridescence patterning of the eyespots and that of the body plan of the whole animal with a fully expanded train. The entire peacock resembles one giant eyespot in the sense that it boasts a deep-blue body surrounded by a dense zone of bluish-green eyespots, which corresponds to an individual eyespot with its deep-blue inner circle surrounded by radiating oval rings of mixed colors. The same pattern of blue can be observed in the peacock’s crest crown also (data not shown) which have been shown to be involved in communication ( Kane et al., 2018).

We hypothesized that the rings of eyespots may be determined by the same cellular plans as the color patterning of the whole animal. We took advantage of the similarity between the rings on the individual eyespots and the spatial eyespot rings traced on the train and modeled backward from the eyespot to feather follicles and back to the eyespot ( Figure 6). Starting from each eyespot ( Figure 6A) we inferred colored rings on the train following the pattern seen in one eyespot ( Figure 6B) and from there on the follicle rows on the uropygium ( Figure 6C). We inferred that the five feather follicles on each side of the central feather make an inverted palindrome and correspond to the five different structural colors seen in the eyespot ( Figure 6D). From this we inferred follicles growth and epidermal invagination giving rise to a concentric ring of structural material (stem cells) ( Figure 6E) inside the feathers which expand and create the ring pattern on the eyespot. ( Figure 6F) Cellular invagination is a well-known process of embryology ( Rauzi et al., 2013; Pearl et al., 2017) but we stress that we are basically connecting dots from eyespot distribution on the train to the concentric rings of individual eyespots and it must remain speculation until further investigated.

The zigzag pattern of feather formation observed from the peacock’s uropygium was able to explain both the hexagonal eyespot symmetry on the train and as well as the unique eyespot rings pattern within eyespots.

The primordial feather follicles arising on the uropygium in rows of 10/11 are arranged in a zigzag manner. It is this zigzag arrangement of 10/11 that creates the complex hexagonal arrangement of eyespots in the peacock’s tail, and thus its complexity and beauty ( Figure 5, upper left). Hexagonal arrangement is a well-known feature of cellular arrangement in developmental biology ( Sengel, 1976; Classen et al., 2005). The same rule of zigzag arrangement between feather follicles that leads to the complex distribution of eyespots on the train also applies to cellular arrangement of cell lineages within feathers, giving rise to rings on the eyespot, which, in combination with the different structural -iridescence colors, imparts the eyespot its beauty. The qualitative annual growth of eyespot number in rows of 10/11 and the quantitative growth of train length and, essentially, of the former’s dependence on the latter for its presentation and display to females may make the two traits inseparable in their effects on female choice.

There are two aspects of development that constrain eyespot variation: bilateral symmetry and asymptotic tail growth. The anatomical arrangement and bilateral symmetry of eyespot feathers preclude random genetic variation by addition of single feather at a time and the only apparent source of variation in eyespot number is the annual addition of new feather rows as the animal grows. Thus, as Takahashi et al. (2008) and Dakin and Montgomerie (2011) found, most males produce approximately 165–170 eyespots, and the small amount of eyespot number variation that exists within and between populations is due to extrinsic factors such as breakage or damage from predation. The actual number of eyespots can exceed 200, as suggested by the number of feather rows on the uropygium shown in Figure 3, but many of them would be too small to be visible through photography or to be effective to illicit female response. The idea that the maximum eyespot number in adult animals is invariant within and across populations is consistent with the annual addition of new rows of feathers, which occurs rapidly at a younger age (~4–7 years), while slowed growth or complete cessation of new feather development occurs after a certain age ( Manning, 1989). It is important to point out that the lack of variation in eyespot number applies to the lifetime total number for individuals and not to the population, which may contain eyespot variation arising from different age classes.

Asymptotic tail growth is the result of aging, loss of vigor and diminishing return in fitness accruing from tail growth. After a male has reached his peak reproductive age and achieved the threshold train size to begin mating, his mating success will increase as a function of both the size of his train and his vigor in mounting a display. But with declining vigor with age and competition from younger males, the individual fitness function will plateau and decline despite large train size. At this point, the law of diminishing returns will commence, and each additional row of feathers will grow slowly, making them look minor, and exert a minimal effect on the feather length and fitness function. Moreover, limitations with respect to maximum eyespot number and the lack of correlation between train length and mating success may—as suggested by Takahashi et al. (2008)—indicate that these traits have reached a threshold. And since males do vary in mating success ( Petrie et al., 1991; Yasmin and Yahya, 1996; Loyau et al., 2008), this variation cannot be due to variation in eyespot numbers and other factors may be involved.

Female choice theory, in the case of the peacock’s tail, relies on the assumption that eyespot number and train length are correlated, and that female choice based on the former leads to a correlated increase in the latter. However, the results of this study, and others ( Takahashi et al., 2008; Dakin and Montgomerie, 2011), suggest that eyespot number is an internally-determined, invariable trait, and thus cannot be the sole factor driving the evolution of the peacock’s long tail. Given that train feathers are rapidly added in rows of 10 or 11 during the first 4–7 years of a peacock’s life ( Manning, 1989) to eventually reach 165–170 at adulthood, any variation in feather growth rate could be the basis of female choice if the size of the train were directly or indirectly a criterion of female mate attraction—or attention. As such, rapidly growing males may develop more elaborate trains with bigger size and more eyespots and thus become preferential targets of female choice during the peak period in the first 4–7 years of life, after which feather number may increase at a slower rate ( Manning, 1989).

Our study brings out three important results that have implications for how females decide on mates. First, the invariant nature of the total number of eyespots in adult males does not allow for variation in fully adult males for selection; the only available variation is that between age classes in a population. Second, eyespots and train length are developmentally correlated; it is the tail growth that pulls the eyespots up, spreads them out, and makes them visible to females. While eyespot brightness has been shown to affect male mating success ( Loyau et al., 2007; Dakin and Montgomerie, 2013), we hypothesize that in deciding on mates, especially from distance, females may be using the size of the “eyespot-studded train” rather than the number of eyespots. This may explain why it has been difficult to separate the effect of eyespots from the effect of train size ( Petrie et al., 1991).

Finally, an important and hitherto neglected trait that is likely to mediate male–female interaction is male vigor. Variation in vigor means that a younger youthful male with fewer eyespots may outperform an older and bigger male with more eyespots ( Petrie, 1993). The new trait, eyespot–train size, may either influence female choice directly through sensory capture ( Kirkpatrick and Ryan, 1991) or indirectly through male vigor and beauty, but not solely based on eyespots ( Mitoyen et al., 2019).

We can consider a two-stage model of female choice, in which females are attracted by male size or train height and sounds from a distance, and exercise mate choice based on eyespot beauty, vigor, and behavior from close proximity. Such a model would be able to explain some of the contradictory results between different studies. It may explain, for example, why females were insensitive to removal of as much as 20% of the eyespots before showing any effect in their behavior ( Dakin and Montgomerie, 2011).

The small sample size of museum specimens used in this study may raise concerns about the validity of our results. The most important results of this study, shown in Figure 5 and Figure 6, are qualitative and/or theoretical modelling, and are unaffected by sample size. Simulation is a form of hypothesis testing, but it lacks direct anatomical testing which is outside the scope of our lab. The zigzag arrangement of feathers and its effect on the train’s symmetry and complexity ( Figure 5) were simulated and shown to conform to observations from this work and at least the number and the symmetry of feathers is supported by previous work ( Dakin and Montgomerie, 2011). The rudimentary model of eyespot development ( Figure 6) would also need to be investigated by developmental work. Future tests can involve (1) anatomical dissection of live animals to test the number and the arrangement of feathers as well as the correspondence between feathers and feather buds, (2) developmental work on young peacocks to test the addition of feather rows as well as feather growth as a function of age, (3) further work on the line of Dakin and Montgomerie (2011) to test the effect of varying eyespot numbers on female choice, and (4) work to test the effect of variation in male vigor as a function of age on female choice.