Plumage color degradation indicates reproductive effort: an experiment

Plumage color has traditionally been regarded as a static ornamental trait, but evidence is accumulating for significant color changes without molt that typically reduce the conspicuousness of ornamentation. In some species, the social partner seems to increase its reproductive investment if the color trait is experimentally enhanced, suggesting that color change could act as a signal. However, the information content of this signal is so far unclear. For example, birds in poor condition or making greater effort may deteriorate more severely. We used brood size manipulations to alter the reproductive effort of male and female collared flycatchers Ficedula albicollis. Both sexes showed less severe decline in some reflectance attribute of their white breast when their brood was experimentally reduced. In each sex, greater deterioration of the reflectance trait affected by the manipulation was accompanied by increased feeding rate by the partner. These feeding patterns do not prove, but are consistent with, a compensatory response by the partner to induced degradation. The manipulation effects on color change we detected confirm for the first time that plumage color deterioration can indicate current reproductive effort, thereby providing a potential fitness advantage to social partners that react to such deterioration.

A central question concerning sexual ornaments is the source of their reliability that provides sufficient benefits for the chooser to counterbalance the costs of considering the signal 1 .Plumage coloration is a well known target of sexual selection, and it is typically categorized based on the proximate background of individual differences.Pigment-based color gets its information content from pigment quality or quantity 2,3 .Chromatic structural coloration (typically UV-blue or iridescent) involves the generation or self-assembly of ordered nanostructures and it may therefore gain its information content from the precision of such structures 4 .The most contentious type of coloration is white plumage patches.These are widespread in many taxa [5][6][7] , and have been repeatedly shown to indicate aspects of individual quality such as condition, parasite load and reproductive output [8][9][10] .These patches contain neither pigment nor ordered nanostructure 11 , so they must gain their reliability from the quantity or quality of the reflective surfaces 12 .
However, information content and reliability are more complicated than this because plumage color does not seem to be a static trait.Even though typically one or a maximum of two molts are performed each year 13 , coloration has been shown to change without feather replacement.Its conspicuousness may increase 14,15 , but it typically decreases 16,17 .Plumage color change between molts may confound analyses of signal information content or function 18 or delimit the most informative period of the trait 15 .Moreover, partners seem to respond to color changes.Within-season manipulations of the coloration of a breeding bird seem to induce changes in parental investment by the partner 19,20 .Color change may constitute a signal of condition or quality for the social mate or even the neighbor.However, this point has not been confirmed yet, as no study to our knowledge has experimentally demonstrated the information content of within-season plumage color shifts, that is, the reason why the observed changes in partner investment would be adaptive.
In white plumage areas, the lack of feather-strengthening melanin 21 disproportionately exposes feather structure to mechanical abrasion and therefore potentially to color degradation 22,23 .Therefore, white area reflectance is an ideal candidate for dynamic signaling.Here we report on experimental manipulations of brood size where we assessed the change of reproductive effort and the immediate degradation of a sexually selected white ornamentation system in response to the manipulation.We conducted the study in collared flycatchers (Ficedula albicollis).In the study population, the reflectance of white plumage areas is sexually selected as part of a plumage-level integrated ornamentation system, where mating patterns follow the overall brightness and relative ultraviolet reflectance of the plumage in both sexes 24 .Differential degradation of color across the plumage has been suggested to occur, frequenting the areas more exposed to damage, particularly the breast 18 .This also influences the integration and functioning of the ornamentation system 25 .Here we applied experimental manipulations of brood size 26,27 to clarify the causal effect of reproductive effort, and studied the degradation of color during nestling rearing across three major white areas of the plumage: wing patch and breast in both sexes, and forehead patch in males.Previous studies suggested area-dependent degradation that presumably reflects different structural background and differential exposure to damage in different plumage areas 18 , so we conducted this study at the level of individual white areas.We predicted that experimentally increased reproductive effort would produce increased degradation and reduced reproductive effort would lead to reduced degradation, especially in the breast which is the most vulnerable to damage.We also observed own and partner feeding patterns in relation to ornament degradation in the late nestling stage to see whether these patterns are consistent with the assumption that social partners adjust their apparent feeding activity to the effort-related degradation of their mate's reflectance.This latter analysis is tentative but necessary.If we find no relationship between the degradation of the trait affected by the manipulation and partner feeding in this scenario where both trait change and partner feeding have increased variance due to the experiment, then this falsifies the assumption that the trait is used as a signal by the other sex.In such a case, any conclusion concerning "signal information content" would have to be treated with caution.If, on the other hand, the analysis detects a significant relationship between trait change and partner feeding, then we cannot exclude the possibility that trait change is a signal.In this case, it is valid to talk about the information content of this putative signal trait, but further experiments, involving direct color manipulation, will be necessary to confirm that the trait is indeed used as a signal.

Manipulation, reproductive effort and output
The feeding rate of parents reacted to manipulation in both sexes, but in a year-dependent manner (Table 1).In 2021, the enlarged group showed greater feeding rate from both others (Scheffé tests, p < 0.001) while reduced broods were fed less often than controls but the difference was not statistically significant (p = 0.120).In 2022, only the reduced group (lower feeding rate) differed from controls (p = 0.021) and tended to differ from enlarged ones (p = 0.096), while control and enlarged broods were very similar (p = 0.880).The total mass of all nestlings before fledging (nestling biomass production) also depended on manipulation in a year-dependent manner (year × manipulation F 2,40 = 3.71, p = 0.033).In 2021, all manipulation categories differed in the expected direction (reduced > control > enlarged; manipulation F 2.18 = 35.12,p < 0.001; all Scheffé tests p < 0.013).In 2022 (F 2,22 = 7.52, p = 0.003), reduced broods produced less biomass (reduced vs. control p = 0.007, reduced vs. enlarged p = 0.020), but control and enlarged broods were very similar (p = 0.999).The initial body mass of parents depended on sex but not any other predictor (Table 1).The body mass changes of both sexes were negative overall during nestling rearing, but they were far smaller in males than in females (manipulation groups pooled, paired t tests, Table 2).When corrected for sex-standardized initial body mass, body mass change showed significant effects of sex, year and manipulation, but no interaction whatsoever between these (Table 1).The effect of manipulation was caused by differences between the reduced and the other groups (reduced > control p = 0.015, reduced > enlarged p < 0.001), with little difference between enlarged and control groups (p = 0.288).In sum, patterns in nestling biomass production, feeding rate and the body mass changes of the parents all indicate that brood size reduction successfully altered reproductive effort but we cannot state this with similar certainty for brood size enlargement.

Color degradation and reproductive effort
During nestling rearing, plumage color decline was prevalent across the white areas (manipulation groups pooled, paired t tests, Table 2).Breast brightness did not systematically decline in either sex, while wing patch brightness declined in males but not in females.Breast and wing patch UV chroma declined in both males (breast only marginal) and females.Finally, the forehead patch of males systematically degraded in both brightness and UV chroma.Before the manipulation, there was no difference between the manipulation categories in any of the white color traits of either sex irrespective of year (Table 3).Changes of reflectance after manipulation (corrected for original value, all details in Table 3) yielded manipulation effects for breast UV chroma in males (Fig. 1).The reduced group differed from the control (p = 0.039) and enlarged groups (p = 0.017) but control and enlarged groups did not differ (p = 0.619).There was also a manipulation effect on breast brightness in females (Fig. 2).The reduced group differed from controls (p = 0.003) but other comparisons were not significant because the enlarged group was intermediate (reduced vs. enlarged p = 0.285, control vs. enlarged p = 0.210).No significant manipulation effect was found for other white color traits, and the manipulation × year interaction was not significant for any trait.Finally, if we pool the control and the enlarged groups for the analysis because their reproductive effort has not been shown to differ (i.e.comparing reduced broods to all other broods), the same results are obtained, with male breast UV chroma and female breast brightness but no other trait significantly affected by the manipulation.This alternative analysis is shown in Supplementary Table S1.

Color degradation and feeding patterns
Partner feeding rate was significantly negatively related to changes in the color trait that responded to the manipulation in the given sex (Fig. 3), but there was no effect of sex, the non-responsive trait of the signaler or the sex × trait change interactions (Table 4, first block).When we repeated the same analysis with change values corrected for pre-manipulation value, no parameter was significant (Table 4, second block).Own feeding rates showed no significant pattern with raw color change irrespective of sex (Table 4, third block), but after correcting change for original value, a sex-dependent relationship with the manipulation-responsive trait emerged (interaction of sex and responsive trait change in Table 4, fourth block).Corrected change in the trait that had responded to manipulation was significantly negatively related to feeding rate in males (F 1,34 = 15.57,p < 0.001) but there was no significant relationship in females (F 1,43 = 0.31, p = 0.583).

Discussion
Our manipulations successfully changed reproductive effort and output, as shown by significant differences in offspring biomass production and the feeding rates and body mass changes in both parents in the brood size reduction group compared to the control and enlarged groups.Similar comparisons indicated that some white plumage color traits responded to the manipulation in the expected direction (less degradation in reduced than in other groups, manifested as higher brightness and UV chroma).Therefore, our data provide experimental evidence that white color degradation may reflect reproductive effort during nestling rearing.
Although both sexes reacted to the manipulation in both feeding and color deterioration, different aspects of color were affected in the two sexes, allowing us to analyze partner feeding rate in relation to the affected trait of the given sex as a possible indicator, although not a proof, of a receiver reaction.This analysis would not have been possible if the same trait had been affected by the manipulation in males and females.Moreover, the lack of correlation between male and female feeding rates within the manipulation groups indicates that negotiation between partners over care also does not confound our results.The results show that the feeding rate of the partner was significantly related to change of the color trait affected by the experimental manipulation of brood size, but not when this change was corrected for original color value.This is precisely what we would expect if the partner directly reacted to degradation in a compensatory manner.The observed degradation strongly depends on original value due to the dependence of abrasion proneness on original macrostructural intactness 28 and also due to random measurement (observation) error.Nevertheless, the partner evidently cannot "correct" degradation for original color and it can only adjust its behavior to raw degradation.Own feeding rates after degradation, on the other hand, were related to change in the experimentally degraded color trait only in males, and the relationship appeared only when degradation was corrected for original value.This is what we expect if the structural mechanism (or other mechanism such as dirt accumulation) causing the degradation is related to male feeding patterns in the late nestling stage.Finally, no relationship with either own or partner feeding rate appeared for the "sister" color trait that did not respond to experimentally modified reproductive effort.Our inference is obviously indirect.The only way to demonstrate a causal link between color degradation and partner feeding is to directly manipulate color and we did not do such manipulations.Most of the variation in our data is experimental so we cannot statistically rule out the possibility that the correlation of color change and partner feeding is due to independent manipulation effects on both traits, although in this case we would expect the strongest relationship between color change and own feeding, and that relationship was not even present in one sex.Moreover, the change in both color and feeding is obviously gradual and our data give only a snapshot of this process that does not even involve a control stage (because the female still incubated the nestlings at the point of manipulation).Nevertheless, the feeding patterns we present are closely consistent with the hypothesis that partners adjust their feeding rates to changes in the color trait of each sex that reflects reproductive effort, but not to a similar trait that does not reflect reproductive effort.
Returning to the color degradation data, at least two particular points are worth mentioning.First, our previous studies of white reflectance degradation paid particular attention to the white wing patch.Both the short-term degradation 28 and the actual expression 29 of white wing patch reflectance were shown to correlate with feather macrostructural state.Moreover, reflectance was suggested to gain part of its quality-indicator value from macrostructural state, as macrostructural state reflected previous year reproductive effort and predicted current year reproductive success 29 .Our current study shows no manipulation effect on either the macrostructure (results not shown here) or the reflectance of the wing patch.As the wing patch is on the flight feathers, it is expected to be more resistant to wear and degrade over longer time scales than contour feathers 30 .In the context of our present results, previous findings indicate that the information content of the wing patch concerning current reproductive potential 29 is due to long-term aspects of body condition (and possibly individual quality), and not short-term condition fluctuations.This is an interesting addition to the signal value of an ornament in which patch size is also very informative and functionally important in both sexes (reviewed in ref 31 ).
The second interesting point concerning our results is that the only significantly affected area was the breast (UV chroma in males and brightness in females).We previously showed in females that, compared to other plumage areas, breast brightness and UV chroma have very low within-individual repeatability and great timeproportional change in expression between incubation and late nestling rearing (but particularly during incubation).We suggested that this might be because of the extreme exposure of this area to dirt and abrasion during parental duties 18 .Moreover, UV chroma has been implicated in greater area-specific deterioration than Table 4. Effects of sex, color change and their interactions on partner and own feeding rates at 10d of nestling age.Results are given both without (raw) and with (residual) correction of color change for original color value.*, p < 0.05; **, p < 0.01; SCRT, standardized change in the manipulation-responsive trait; SCNRT, standardized change in the not manipulation-responsive trait.brightness in our population, with consequences for mating patterns 18,25 .Therefore, deterioration of the breast UV chroma of males by manipulated reproductive effort is not surprising.Female breast UV chroma also shows a non-significant tendency in the same direction (p = 0.133).
On the other hand, female breast brightness significantly responded to the manipulation while the homologous male trait did not (p = 0.685).Female breast feathers are much less reflective in the late nestling period than male feathers 24 , which could be because feather macrostructure differs between the sexes, with male feathers also being more resistant to wear, but note that the female breast also contains some melanin and this confounds both reflectance and degradation.A less complicated explanation is that females incubate the offspring up to 6 days of age, thereby exposing themselves to far greater amounts of dirt and wear than males.Therefore, if brightness is a less degradation-prone trait than UV chroma 32,33 , females are the sex in which we predominately expect such degradation.Finally, it is interesting that breast reflectance traits showed the smallest overall decline among the white plumage areas during nestling rearing, yet the effects of reproductive effort were the most visible on the breast.This may be because birds can easily reach their breast plumage for cleaning and ordering, but at the same time the parental care duties may leave the most visible traces on the breast.
Despite abundant correlative evidence for short-term degradation of plumage color [16][17][18]32,34 , little attention has been paid to the possible role of this trait type as a dynamic signal, although this is important for multiple reasons. First manipulations of reproductive effort to induce changes in coloration at the next molt are a standard though rarely used test of condition dependence.Of the three such studies we are aware of, two measured the current-year color (or patch size) of the birds after nestling rearing and therefore after any related abrasion 35,36 , whereas one study used a generalized across-season patch size measure 37 .Any manipulation aimed to test condition-dependence of color by molt may inadvertently alter current color by the time of measurement due to degradation and therefore confound the baseline to which future ornamentation is compared.This could blur or distort the effects of the manipulation on apparent color change, so measurements of current color need to be very carefully timed at future applications of this classical method, especially if abrasion-prone traits are involved.
Second, coloration of breeding birds is most often measured when much of the investment in the breeding bout is already over (second half of nestling rearing), so color degradation may influence our assessment of the mating and success consequences and therefore fitness relevance of color 38,39 .Such an effect can indeed be detected in the UV coloration of our population 18,25 .Therefore, it would be important to widely quantify the degree of plumage color degradation during nestling rearing, and its effect on mating patterns and success correlations.
The third consequence of the effort-dependent rapid degradation we suggest here is the substantial complication of causal relations between coloration and success.Reproductive output may be determined by the dynamic balance of parental investment by both partners where the investment may depend on both own and partner ornamentation 27,[40][41][42] .Partners have been suggested to react to changes in the provisioning of their mate 43,44 .Analogously, the assessment of within-individual color changes by the partner may transform the problem of ornament-dependent parental care to a dynamic feedback process.The relationship between ornamentation and effort is confounded by wear and it therefore changes during the reproductive bout.As a consequence of changing ornamentation and investment, partner investment and its relation to ornamentation may also change, possibly also changing partner ornamentation.Dynamic skin color traits in birds have been shown to elicit immediate responses from the receiver 45,46 , but ornamental skin colors are relatively rare.However, if within-season plumage color degradation is widespread, then the utility and interpretation of the "good parent" 41 and differential/ compensatory investment 47 hypotheses of sexual selection depends on testing, in as many species as possible, whether reproductive investment changes with rapid shifts in partner ornamentation.Such responses in reproductive investment have been shown e.g. in both male and female blue tits Cyanistes caeruleus 48,49 .
In conclusion, three conditions need to be met for plumage color to qualify as a dynamic, continuously assessed signal.First, it needs to change in the relevant period (e.g. a given phase of breeding), which has been widely demonstrated.Second, its changes need to be considered by the partner, for which there is also evidence already.Finally, it needs to convey relevant information to the receiver, which our present study experimentally demonstrated.We hope that these results contribute to the refocusing of attention from static to dynamic aspects of plumage color, and to the improvement of empirical approaches and therefore better understanding of the functioning and evolution of this sexual trait type.

Ethics statement
We conducted this study with ethical approval by the institutional animal welfare committee of Eötvös Loránd University (permit numbers T-012/2015, T-020/2017) and with research permit from the regional nature conservation authority (permit number PE-06/KTF/920-7/2018).We conform to relevant national guidelines and regulations.The reporting of our work conforms to the ARRIVE guidelines.

Field methods
We conducted the experiment in the breeding seasons of 2021 and 2022.We conducted a brood size manipulation in pairs of synchronously hatched broods with the same clutch size (6 or 7) and a maximum brood size difference of one.We used a similar but non-manipulated third brood as control 26,27 .We did the manipulations within five days in both years, so date variation was minimal.At the nestling age of 2 days, we transferred two nestlings from one brood to the other, and four nestlings in the reverse direction, thereby enlarging the first brood by two nestlings and reducing the second by two nestlings while keeping the ratio of own and foreign young the same in the two broods.We measured nestling masses at 2, 8 and 12 days of nestling age.At two and ten days of nestling age, we captured the parents, identified or ringed them, measured body mass (spring balance) and morphology

Figure 1 .
Figure 1.Change in the breast UV chroma of males in relation to brood size manipulation.Change is corrected for original value by calculating linear model residuals.The number of data is 11 for reduced, 14 for control and 13 for enlarged broods.

Figure 2 .
Figure 2. Change in the breast brightness of females in relation to brood size manipulation.Change is corrected for original value by calculating linear model residuals.The number of data is 17 for reduced, 18 for control and 13 for enlarged broods.

Figure 3 .
Figure 3. Relationship between raw color change of the manipulation-responsive trait of the given sex (standardized for sex) and the feeding rate of the partner at 10d of nestling age.Black, reduced broods; dark grey, control broods; light grey, enlarged broods.

Table 1 .
Original body mass, body mass change and feeding rate of parents in relation to sex, year, manipulation and their interactions.Body mass change is corrected for original body mass by including it as a covariate.

Table 3 .
Original color variables and their changes in relation to year, manipulation and their interaction.Color change is corrected for original color by including it as a covariate.*, p < 0.05; **, p < 0.01; ***, p < 0.001.