Ultraviolet crown colouration affects contest outcomes among male blue tits, but only in the absence of prior encounters


  • Oscar Vedder,

    Corresponding author
    1. Animal Ecology Group, Centre for Ecological and Evolutionary Studies, University of Groningen, P.O. Box 14, 9750 AA Haren, The Netherlands
      Correspondence author. E-mail: o.h.vedder@rug.nl
    Search for more papers by this author
  • Elske Schut,

    1. Animal Ecology Group, Centre for Ecological and Evolutionary Studies, University of Groningen, P.O. Box 14, 9750 AA Haren, The Netherlands
    Search for more papers by this author
  • Michael J. L. Magrath,

    1. Animal Ecology Group, Centre for Ecological and Evolutionary Studies, University of Groningen, P.O. Box 14, 9750 AA Haren, The Netherlands
    2. Department of Zoology, University of Melbourne, Vic. 3010, Australia
    Search for more papers by this author
  • Jan Komdeur

    1. Animal Ecology Group, Centre for Ecological and Evolutionary Studies, University of Groningen, P.O. Box 14, 9750 AA Haren, The Netherlands
    Search for more papers by this author

Correspondence author. E-mail: o.h.vedder@rug.nl


1. Static animal colour patches may function among competitors to minimize conflict escalation over resources, by serving as a signal of resource holding potential or aggressiveness. Empirical evidence for the use of colour patches in conflict resolution is largely restricted to pigment-based colours (melanins and carotenoids) and rarely defines the context in which the signals are used.

2. Here we test whether structural-based ultraviolet (UV) crown colouration functions in conflict resolution among dyads of first-year male blue tits Cyanistes caeruleus in captivity, in a context that is known to favour aggressive individuals.

3. We found that on first encounter in a pairwise context, experimentally UV-reduced males were significantly more likely to lose to control-treated opponents than expected by chance. However, this disparity was less pronounced when conflicts were settled with physical fighting or when the opponent was considerably smaller in size.

4. When the same dyads were tested again several weeks later, but with the UV treatment reversed among opponents, none of the effects remained significant, but instead the winner was most likely to be the individual that won at their first encounter.

5. Our results suggest that structural-based UV colouration can affect the outcome of an interaction, but that size differences and the outcome of initial interactions between opponents can override the influence of this signal in conflict resolution. Whether there is a functional basis to maintain a link between aggressiveness and colouration may thus be highly dependent on the general context under which individuals compete in the wild.


Successful reproduction in animals requires the acquisition of sufficient resources, like food, nest sites, territories and mates. As these are often limited in availability, natural selection will strongly favour traits that enhance an individual’s ability to obtain such resources, often at the expense of others. When two individuals compete over a resource at least three factors come into play. (i) The subjective resource value (Maynard Smith 1982). If there is an asymmetry in the perceived value of the contested resource between competitors, the individual that has more to gain (or lose) will be more prone to escalate the conflict towards actual fighting and consequently more likely to win, as its opponent may favour avoiding the potential costs of fighting, like the risk of injury, and thus retreat. Such a scenario applies, for example, when there are differences in hunger level or prior residency between opponents (e.g. Sandell & Smith 1991; Lemel & Wallin 1993). (ii) The resource holding potential (RHP) of both competitors, defined as the ability to win an all out physical contest (Parker 1974; Maynard Smith 1982). When both competitors are highly motivated, conflicts are more likely to escalate to fighting, thus favouring the individual with the higher RHP. (iii) The level of aggressiveness of both competitors. Apart from contest-specific differences in motivation, individuals may also inherently differ in their tendency to resort to costly fighting. Herein aggressive individuals are defined as individuals that possess a higher propensity to escalate contests, independent of their RHP (Barlow, Rogers & Fraley 1986; Maynard Smith & Harper 1988). This strategy may be particularly favourable when both competitors judge a resource to be low in value (Maynard Smith & Harper 1988). With the recent renewed interest in animal ‘personality’ traits, it has become apparent that aggressiveness is an important component of a wider behavioural syndrome, variously characterized as risk-prone, unresponsive, pro-active, bold or fast exploring that has an underlying genetic basis (rev. by Sih et al. 2004).

Originally Rohwer (1975, 1982) hypothesized that it would be adaptive for both individuals in a contest to exhibit their status, so as to assess each other’s competitive abilities without wasting time, energy or risking injury in physical fights. Empirical studies have since attributed such status signalling functions to a wide array of colourful patches exhibited by a diverse range of animals, but especially in birds (Senar 1999) and lizards (Whiting, Nagy & Bateman 2003). However, it has become less clear which factor in determining social status (from the three discussed above) is actually signalled. Maynard Smith & Harper (1988) argued that apparently arbitrary or conventional signals (i.e. signals that are not necessarily physically or physiologically connected to RHP) can signal aggressiveness. These signals would only predict the outcome of conflicts over low-value resources, as conflicts over high-value resources will always be settled by differences in RHP, either via corresponding signals or directly by fighting. The authors support this prediction by presenting some previously unpublished empirical results on three bird species, showing that status signals only predict contest outcome at feeding sites with a low food intake rate. Recently, empirical evidence from paper wasps Polistes dominules also suggests that signal use may be limited to conflicts in which the benefit of winning is low (Tibbetts 2008).

In birds, melanin-based signals, like the black throat patch in male house sparrows (Passer domesticus; Moller 1987; Gonzalez et al. 2002; Nakagawa et al. 2007) and the black ventral breast stripe in great tits (Parus major; Wilson 1992; Lemel & Wallin 1993), may best fit the pattern expected for a signal of aggressiveness, because these generally only correlate with status when resource values are low (see Lemel & Wallin 1993 for a detailed example). The correlation with social status may be mediated by testosterone, promoting both melanin expression (Evans, Goldsmith & Norris 2000; Buchanan et al. 2001; Gonzalez et al. 2001) and aggressive behaviour (rev. by Wingfield et al. 1987). In contrast, colour patches based on carotenoid pigments, like those in widowbirds (Euplectes; Pryke et al. 2002; Pryke & Andersson 2003), red-winged blackbirds (Agelaius phoeniceus; Smith 1972; Peek 1972) and scarlet-tufted malachite sunbirds (Nectarinia johnstoni; Evans & Hatchwell 1992), appear to signal social status in contests over high-value resources like territories. As carotenoids can not be synthesized de novo by animals (Olson & Owens 1998), the reliability of these signals to reveal genuine RHP may be maintained by the difficulty of acquiring sufficient carotenoids in their diet (Pryke et al. 2002). Hence, carotenoid signals are most likely to reveal RHP, instead of aggressiveness, and individuals should respect these signals also when the risk of fighting is high.

In addition to pigmentation, colours can also arise from the selective reflectance of light from surface structures (Land 1972). Such structures mainly reflect the green, blue, violet and the ultraviolet (UV) part of the colour spectrum (Andersson 1999; Shawkey et al. 2003; Prum 2006). There is some indication that UV-blue structural colouration may be condition dependent (Keyser & Hill 1999; McGraw et al. 2002; Siefferman & Hill 2005a) and a recent experiment revealed a positive effect of testosterone on UV reflectance in first-year male blue tits (Fig. 1; Cyanistes caeruleus; Roberts, Ras & Peters 2009). Regarding a role in conflict resolution, there is experimental evidence that UV reflectance signals status in Augrabies flat lizards (Platysaurus broadleyi; Stapley & Whiting 2006), and positive correlations between measures of social dominance and UV reflectance have been reported in birds (Keyser & Hill 2000; Siefferman & Hill 2005b; Pryke & Griffith 2006). The only experimental support for a function of UV reflectance in competition in birds comes from a study in which territorial male blue tits with fertile females acted less aggressively towards taxidermic mounts with reduced crown UV-reflectance, compared to control males (Alonso-Alvarez, Doutrelant & Sorci 2004). However, these results were not repeatable in a different population when UV-manipulated mounts were presented simultaneously instead of sequentially (Korsten, Dijkstra & Komdeur 2007). Furthermore, they can also be interpreted in a mate choice context if, for example, a UV-reduced male poses less threat to a focal male’s paternity (see Delhey et al. 2007), especially as status signalling studies generally report greater aggression towards individuals with reduced signal expression (e.g. Jarvi & Bakken 1984; Pryke et al. 2002; Pryke & Andersson 2003).

Figure 1.

 Blue tit male (photo by Sandra Bouwhuis).

Nevertheless, blue tits are likely candidates for UV status signalling as they are known to display their UV reflective crown feathers in competitive interactions (Stokes 1962; Scott & Deag 1998). We tested this possibility in a previous study where the crown UV reflectance of free-living blue tits was reduced experimentally, but our findings failed to support a status signalling function for UV colouration, because neither status nor conflict escalation rate was affected by the manipulation (Vedder et al. 2008). However, such effects may be very difficult to detect in wild populations because of the strong site-dependent dominance structure that resident blue tits exhibit when moving around in the vicinity of their territories in winter (Colquhoun 1942; Hansen & Slagsvold 2004; Korsten et al. 2007). Consequently, the outcome of interactions may be much more significant than suggested by the value of the contested resource, as all interactions reaffirm the local territorial structure. Hence, although these previous findings can exclude UV reflectance signalling RHP, it may still act as a signal of aggressiveness. By definition, a static signal like an inconcealable plumage patch can not signal contest-specific motivation.

Here we specifically test whether the UV crown plumage of blue tits serves as a signal of aggressiveness in first-year male blue tits. We experimentally manipulate the UV crown plumage to test whether UV reflectance is used as a cue to estimate an opponent’s aggressiveness, and not merely a correlate of social dominance. We let the birds compete in pairwise trials over food, as such a context has been shown to favour aggressive, fast exploring, individuals in the closely related great tit (Verbeek, Boon & Drent 1996) and signal manipulations in great tits only appear to have an effect in this context (Lemel & Wallin 1993). To explore the consistency of signal use in dominance trials, we repeated all trials with reversed colour manipulations. We predict that if UV reflectance is used as a signal of aggressiveness, opponents of UV-reduced individuals would behave more aggressively towards these individuals, such that conflicts not resolved by fighting will be more often lost by the UV-reduced individuals.


Origin and housing of birds

In spring 2007 and 2008, wild nestling blue tits were taken into captivity and their sex determined from a blood sample using molecular sexing (following Griffiths et al. 1998). For the experiment, we used 22 male nestlings in 2007 and an additional 10 in 2008. In 2007, these nestlings were collected from 12 different 10-day-old broods from a wild nestbox population at the Hoge Veluwe National Park in the Netherlands (52°02′N, 05°51′E) and hand reared in captivity (following Verbeek, Drent & Wiepkema 1994). In brief, the birds were placed in artificial nests in incubators and fed on a mixed diet including wax moth larvae and minced beef heart. Nestlings remained in incubators until fledged and were then placed in small cages (0·6 × 0·4 × 0·4 m) in groups of three to four individuals. Hand-feeding continued until they were observed to eat independently. Approximately 1 month after hatching, they were transferred to outdoor aviaries (3 × 2 × 2 m) at the Biological Centre, University of Groningen. The males were kept in groups of up to eight individuals, until we started the dominance trials. Between trials, males were housed in the same aviary with the same group of individuals.

In 2008, the 10 male nestlings were taken into captivity from a wild nestbox population on the grounds of the Biological Centre, University of Groningen (53°10′N, 06°36′E). Nestlings were swapped between four different nests, to create two broods that were taken into captivity. Just before fledging the two broods were placed in new nest boxes in outdoor aviaries (3 × 2 × 2 m) together with one of the parents (that was caught at the nestbox). The parents were ad libitum provided with mealworms and wax moth larvae, and supplemented with caterpillars from the wild. In both cases, the parent (one male and one female) continued feeding the nestlings in captivity. From 4 days after fledging, the parents were separated from their brood for part of the day, to stimulate the fledglings to forage independently. Initially separation lasted for 2 h but this period was increased progressively until the fledglings were completely independent (24 days after fledging). At this time, the parents were released and the male fledglings kept under the same conditions as described above for the 2007 cohort.

Pairwise dominance trials

The first rounds of pairwise dominance trials took place in early November (in both 2007 and 2008). During each round of trials, half of all first-year males were randomly assigned to a UV-reduction treatment, whereas the other half was control treated (see below). The UV treatments of males were evenly distributed over the outdoor aviaries where they were kept before and between the rounds of trials, and randomly distributed with regard to their nest of origin in the wild. Pairwise dominance trials were staged such that, during each round, every individual competed with at least two individuals from a different aviary (i.e. unfamiliar) and two individuals from the same aviary (i.e. familiar), one each day, and always against an opponent of the opposite treatment. Hence, in one round all first-year males competed at least four times, against four different opponents of the opposite treatment. In November 2007, this was achieved by performing 47 trials involving the 22 males that hatched in spring 2007. In November 2008, we performed 25 trials involving the 10 males that hatched in spring 2008. In both winters, the trials were repeated in early January (in 2007/2008 after 51 days, in 2008/2009 after 49 days) with the exact same combinations of opponents (hereafter referred to as ‘dyads’), but with the UV-reduction and control treatments reversed for all individuals. This allowed sufficient time for the original treatments in November to wear off.

One day prior to each round of trials, all males were removed from their aviaries, measured and treated as described below, and placed in one compartment of a cage that consisted of two compartments equally divided by a sliding wooden panel. Compartments (0·4 × 0·4 × 0·4 m, following Verbeek, Boon & Drent 1996) were fitted with a perch and birds had ad libitum access to food and water, and were visually isolated from each other. These cages were placed outside near their original aviaries, such that the light–dark cycle was unchanged and the birds could observe each other under natural light conditions during the trials. The following morning dominance was assessed by removing the slide between the compartments for 20 min. During this period all food was removed, except for a small bowl of commercially available dried insect food (Witte Molen, Meeuwen, the Netherlands), placed in the centre of the cage. Hence, the contested resource was of only low value to both individuals, which should favour the individuals that perceive their opponent as risk-averse or non-aggressive. In the 20 min. period, we scored the outcome of all competitive interactions, either over food or space, and noted whether they involved physical fighting. It was impossible to clearly separate interactions over food from interactions over personal space, since they generally did not tolerate each other in close proximity (<10 cm, O. Vedder & E. Schut, personal observations). In most cases at least one individual acted aggressively (displaying, chasing), after which the opponent responded submissively (fleeing, crouching). Interactions occasionally escalated to a few brief physical fights (on average 1·38 ± 0·31 SE fights, in trials where fighting occurred) after which one individual retreated and behaved submissively. Individuals were recognizable by colourbands and a single interaction was won if the opponent fled or crouched. All observations were made from a distance of about 3 m. After 20 min, the slide was replaced and in the afternoon the individuals were relocated to compartments adjacent to their next opponent in preparation for the following trial.

Morphological and colour measurements

In both winters, the males were measured 1 day before the start of each round of trials (November and January). Body mass was measured to the nearest 0·1 g with a spring balance (Pesola, Baar, Switzerland). Tarsus-length was measured to the nearest 0·1 mm with a sliding calliper, and length of the third (outermost) primary feather (hereafter referred to as ‘P-3’) to the nearest 0·5 mm with a ruler.

The spectral reflectance of the crown feathers was measured with an USB-2000 spectrophotometer with illumination from a DH-2000 deuterium-halogen light source (both Avantes, Eerbeek, the Netherlands). The measuring probe was held at a right angle against the plumage, i.e. both illumination and recording were at an angle of 90° to the feathers. Five replicate readings were taken from the centre of the crown for each individual, by lifting and replacing the probe between each reading. These five readings were smoothed by calculating the running mean over 10 nm, and the average of the most commonly used colour index (i.e. ‘UV chroma’) in UV colour signalling studies calculated, by dividing the sum of reflectance between 320 and 400 nm by the sum of reflectance between 320 and 700 nm (R320–400/R320–700). All second measurements (in January) of body mass, tarsus-length, P-3-length and UV chroma were highly repeatable, within individuals, with the initial measurements in November (Table 1).

Table 1.   Estimates and test statistics for within-individual repeatability of trait measurements before the first (November) and second (January) experimental round
Dependent variableRepeatability (±SE)n (measured twice)FP
  1. Repeatability estimates were derived according to Lessells & Boag (1987) and the standard errors according to Becker (1984).

Tarsus-length0·889 (±0·04)3119·98<0·001
P-3-length0·968 (±0·01)3161·82<0·001
Body mass0·676 (±0·10)315·17<0·001
UV chroma0·659 (±0·10)304·87<0·001

The UV-reduction treatment (following Johnsen et al. 2005) consisted of painting the UV-blue crown feathers with a dark blue (colour code 003) Edding 4500 T-shirt marker pen (Ahrensburg, Germany). The control treatment consisted of the same procedure, but with a light blue pen (colour code 010). To confirm the direct effect of both treatments, three replicate crown reflectance readings were taken directly after the treatment, following the protocol described above. The UV-reduction treatment decreased the UV chroma index by an average ± SE of 10·4 ± 0·94% (Fig. 2), which was highly significant (tpaired = 10·82, = 30, < 0·001). The control treatment increased UV chroma by an average ± SE of only 0·3 ± 0·59% (Fig. 2), which was not significant (tpaired = −0·58, = 31, = 0·568). The treatments were re-applied to all individuals every afternoon before the next dominance trial. Random sampling of some individuals some weeks later suggested that the treatment lasted a few weeks, confirming that UV reflectance was reduced during the trials (O. Vedder, personal observations). There were no significant initial differences between the group of males that was UV reduced and the group that was control treated in any of the traits (tarsus-length, P-3-length, body mass, original UV chroma) measured before the first round of dominance trials (Table 2). As suggested by the high repeatability of these traits, there were also no significant differences in these traits between treatment groups before the second round of trials (all > 0·40).

Figure 2.

 Reflectance plotted against wavelength of the original crown colouration of 32 males (30 measured twice and two measured once) and their colouration after control treatment (32 males measured once) and after the UV-reduction treatment (31 males measured once). Standard errors are presented at 20-nm intervals. The dotted reference line at 400 nm represents the distinction between the human (400–700 nm) and bird (320–700 nm) visible range in colour spectrum.

Table 2.   Summary of independent t-tests for pre-manipulation differences in morphological traits between the group of males that were UV reduced and the group of males that were control treated in the first round
Dependent variableMeanntP
ControlUV reducedControlUV reduced
Tarsus-length (mm)17·3817·3016160·480·636
P-3-length (mm)51·6351·4416160·350·727
Body mass (g)11·4911·621616−0·660·512
UV chroma (%)29·7730·131615−0·380·709

Data and analyses

In total, we performed 72 trials in the first round (both years combined), of which 68 dyads were repeated with reversed treatments in the second round (one male from 2007 died between rounds so we could not repeat four dyads). An individual was assigned as the winner of the trial if it won significantly more interactions than its opponent, based on a binomial probability test of all interactions within the 20 min trial period. In 21 trials (11 in the first round and 10 in the second round) no significant winner could be assigned, and these trials were scored as draws and omitted from the analyses. Draws resulted mainly from the lack of aggression from both opponents, as an average of only 12·7 (range: 0–46) interactions were observed in these trials, compared with an average of 24·5 (range: 5–87) interactions in trials with a clear winner. In trials with a clear outcome, on average 23·9 ± 1·56 SE of interactions were won by the winner and 0·6 ± 0·15 SE by the loser.

We tested if the probability of UV-reduced individuals winning trials deviated from random (0·50) using generalized linear mixed models (GLMM) with a binomial error distribution and a logit link function. As the probability of an individual winning an interaction is not independent of the chance that the opponent loses, we only analysed interactions from the perspective of the UV-reduced individuals. In this way we avoided pseudo-replication, as each interaction was only counted as a single datapoint, and the outcome of an interaction can be analysed as a binary response (i.e. win or lose). To account for repeated interactions of UV-reduced individuals and opponents, we included both the identity of all UV-reduced males and their identity when they acted as opponents (control treated) as cross-classified random effects. Categorical fixed effects added to the initial model were ‘year’ (the winter in which the trials were performed), ‘familiarity’ (yes, if they came from the same aviary), ‘round’ (first or second occasion the dyad was tested) and ‘fighting’ (yes, if physical fighting took place within the trial period). We also included the interaction between ‘familiarity’ and ‘round’ as the effect of familiarity may be different in the second round because all males had already competed with their opponents in the first round. Covariates added to the initial model were the difference in tarsus-length, the difference in P-3-length, the difference in body mass and the difference in original UV chroma between the UV-reduced individual and its control-treated opponent. In all cases, positive values indicate a greater absolute value for the UV-reduced individual. We chose to use differences instead of absolute values as the difference between opponents is more biologically meaningful and this way the model constant reflects the probability of winning for an UV-reduced individual when there are no other differences with its opponent. We failed to measure the spectral reflectance of one male’s crown before treatment in the first round, so models including this term are based on four fewer trials. To account for intercorrelatedness between the size-variables (i.e. differences in tarsus-length, P-3-length and bodymass), we included only one of these covariates at a time in three initial models. The final model that included the most significant of these size variables was subsequently expanded with the other two size variables, following the same procedure, to test for possible independent effects of these variables.

Final models were obtained by backward stepwise removal of non-significant terms. Significance (< 0·05, two-tailed) was assessed using the Wald statistic. Non-significance was confirmed by re-entering the term in the final model. The GLMMs were fitted using the Markov chain Monte Carlo (MCMC) procedure in MlwiN 2.02 (Rasbash et al. 2004). All other analyses were performed in spss 14.0 (SPSS Inc., Chicago, Illinois, USA).


Analyses of both rounds combined showed that UV-reduced individuals in the first round were significantly less likely to win than the UV-reduced individuals in the second round (see the effect of ‘round’ in Table 3). As the constant represents the first round probability of winning for the UV-reduced individuals, this probability was significantly less than the random expectation of 0·50 (Table 3). However, this probability significantly increased in cases (= 12) where physical fighting occurred during conflict resolution, and when the UV-reduced individual was considerably larger (as represented by a greater tarsus-length) than its opponent (Table 3). Year and familiarity did not affect the probability of an UV-reduced individual winning, nor did the difference in P-3-length, body mass or original UV chroma (Table 3). When entered as the only size-variable in the model (i.e. without difference in tarsus-length), neither difference in P-3-length or difference in bodymass was significant (> 0·15 for both).

Table 3.   Model summary of generalized linear mixed model testing for effects on the probability of an UV-reduced individual winning a 20 min dominance trial against a control treated opponent
Explanatory variableCoefficient (±SE)χ2Δd.f.P
 Constant−1·982 (±0·73)7·3510·007
 Round (first = ref)3·047 (±1·04)8·6110·003
 Fighting (no = ref)2·224 (±1·06)4·3910·036
 Diff. in tarsus-length1·646 (±0·73)5·0910·024
 Year−0·506 (±0·99)0·2610·609
 Familiarity (no = ref)0·751 (±0·56)1·7810·182
 Familiarity × round0·207 (±1·12)0·0310·854
 Diff. in P-3-length0·293 (±0·25)1·3810·240
 Diff. in body mass0·190 (±0·48)0·1610·689
 Diff. in UV chroma0·021 (±0·12)0·0310·862

For closer examination, we subsequently analysed each round separately in initial models that included only the significant terms from the overall model presented above (i.e. difference in tarsus-length and occurrence of physical fighting). The final model for the first round confirmed that UV-reduced individuals had a lower than random chance of winning (24·6% of conflicts were won by the UV-reduced individual (= 61); constant ± SE = −1·550 ± 0·77, χ2 = 4·10, d.f. = 1, = 0·043) and the positive effect of a greater tarsus-length (coefficient ± SE = 2·395 ± 1·15, χ2 = 4·33, d.f. = 1, = 0·037) (see Fig. 3 for visualization of these logistic effects). Although the occurrence of physical fighting was rejected from the model, inclusion of this term in the final model did suggest that the probability of UV-reduced individuals winning tended to increase if physical fighting was involved in conflict resolution (coefficient ± SE = 3·213 ± 1·85, χ2 = 3·02, d.f. = 1, = 0·082). The final model for the second round yielded no significant effects (all > 0·10). However, the outcomes in the second round were very much predicted by the outcomes in the first round, as the same male won in 46 of 52 cases (88·5%) where there was a clear winner in the same dyad in both rounds.

Figure 3.

 Probability of UV-reduced males winning dominance trials against control-treated opponents in relation to the difference in their tarsus-length (positive values indicate a greater tarsus for the UV-reduced male) in the first round of trials. Logistic model prediction (solid line) is based on real conflict outcomes (binary datapoints). Overlapping datapoints are placed on the top of each other. Dotted reference lines indicate the random winning probability (0·5 on y-axis) and the estimated probability when there is no difference in tarsus-length (0·0 on x-axis).

The significantly greater probability of winning for control-treated individuals in the first round, combined with the high repeatability of winner identity between rounds, would suggest that UV-reduced males should have been more likely to win contests in the second round, as the second round consisted of the same dyads as in the fist round, with only the UV treatment reversed among opponents. However, because this was not the case [67·2% of conflicts were won by the UV-reduced individual (= 58); constant ± SE = 1·175 ± 0·78, χ2 = 2·28, d.f. = 1, = 0·131], most of the few changes in winner identity between rounds must have been in favour of the individual that was control treated in the second round. Indeed, in 83·3% (= 6) of dyads where the loser of the first round became the winner in the second round, the individual lost when UV reduced, but won when control treated. In 53·8% (= 13) of dyads in which there was a draw in one round, the winner of the other round was control treated. Three dyads ended in a draw in both rounds, and four dyads were only performed once (two wins of control and two draws), due to the death of one male between rounds.


Here we show that on first encounter in a pairwise setting, first-year male blue tits with reduced crown UV reflectance lose significantly more interactions to control treated males than predicted by chance. However, when the UV-reduced male was of larger size (as indicated by a greater tarsus-length) than its opponent, or when conflicts were settled by physical fighting, its probability of winning increased significantly. When the same combinations of opponents interacted again in the same pairwise setting c. 50 days later, but with the UV treatment reversed, none of the above mentioned effects remained significant. Instead, the outcome in the first round appeared to be a strong predictor of the outcome in the second round. Together, these results suggest that male blue tits use multiple cues and strategies in conflict resolution.

Across various taxa, body size is generally an important determinant of RHP (Kokko, Lopez-Sepulcre & Morrell 2006). A difference in size that can be easily observed by an opponent may override all other cues in conflict settlement, because actual fighting against a considerably larger opponent would risk injury with only a low probability of winning. That the effect of our signal manipulation was weaker in conflicts settled with fighting is consistent with prediction, as we did not manipulate true RHP. Thus, the role of UV crown reflectance in conflict resolution may be limited to situations where the opponents have not previously interacted, there is little difference in true RHP, and conflicts have a low probability to escalate to actual fighting. This situation would seem uncommon among territorial blue tits in the wild and thus may account for the absence of an effect of UV manipulation on social status in a previous study of free-living blue tits (Vedder et al. 2008).

Instead, our results suggest that crown UV reflectance may be used as a signal of aggressiveness. Although there is debate over whether such signals are evolutionary stable (Owens & Hartley 1991; Johnstone & Norris 1993; Hurd 2006), a signal that is only respected when the value of the contested resource is low conforms to expectations of signals for aggressiveness (Maynard Smith & Harper 1988). Respecting such signals will reduce the risk of injury in conflicts over low value resources, whereas misleading signalling of aggressiveness will only increase the probability of conflict escalation for risk-averse individuals. Aggressive or risk-prone behaviour may yield benefits in the short term, but may compromise future returns, leading to the stable coexistence of different behavioural types in a population (Wolf et al. 2007). This would explain the occurrence of functional variation in signals that lack clear production costs, like those resulting from the presence/absence of melanin pigments (e.g. Lemel & Wallin 1993; Qvarnstrom 1997; Gonzalez et al. 2002; Tibbetts & Dale 2004). Empirically this is nicely demonstrated by the observation that male house sparrows with initially large black throat patches showed increased winter survival after signal enlargement, whereas the reverse occurred among initially small-patched males after signal enlargement (Nakagawa et al. 2008). Hence, the lack of any benefit of increased signal expression for risk-averse individuals, instead of high production costs, can functionally explain the stability of signalling aggressiveness (see also Folstad & Karter 1992 for additional testosterone-mediated trade-offs in life-history strategy).

Mechanistically, for melanin-based signals this link with aggressive, risk-prone behaviour is thought to be mediated by testosterone (Evans, Goldsmith & Norris 2000; Buchanan et al. 2001; Gonzalez et al. 2001). Interestingly, it was recently reported that testosterone stimulates the expression of crown UV reflectance in first year male blue tits, via preening behaviour (Roberts, Ras & Peters 2009), adding to previous research that showed UV reflectance to correlate positively with testosterone in first-year, but not so in older males (Peters et al. 2006). Perhaps, more UV reflective individuals receive more social stimulation, at least among recently established first-year males, because they are signalling an aggressive, risk-prone strategy. As such, the signal may also provide positive feedback for the stimulation of testosterone and thus aggressiveness (cf. McGraw, Dale & Mackillop 2003; Safran et al. 2008), further contributing to within individual behavioural consistency across contexts and time (Wolf, van Doorn & Weissing 2008). The stable site-dependent dominance structure among territorial males, as an alternative way of resolving conflicts between individuals, and the concomitant lack of function of UV reflectance in conflict settlement among wintering territorial males (Korsten et al. 2007; Vedder et al. 2008), may break down the cycle of positive feedback that signal expression can have on testosterone levels via social stimulation (Safran et al. 2008) and explain the absence of a positive association between testosterone and UV signalling in older more established males.

This leads us to the absence of an effect of our UV manipulation during the second round of trials. Experimental studies of status signalling using a reciprocal design are rare, as we know of only two previous studies that applied reversed colour treatments in the same dyads of individuals (Pryke et al. 2002; Pryke & Andersson 2003). In contrast to our findings, these studies, performed in red-collared widowbirds Euplectes ardens and red-shouldered widowbirds Euplectes axillaris, did show that conflict outcomes could be reversed by reversing the treatment. However, as these signals are carotenoid-based and likely to signal true RHP, instead of aggressiveness, the cost of disregarding these signals may be too high to risk using alternative cues in conflict resolution. Capability of long-term social memory has been suggested for species in a range of taxa varying from paper wasps (Sheehan & Tibbetts 2008) to African elephants (Loxodonta africana; Hart, Hart & Pinter-Wollman 2008), and also in birds (Godard 1991). Possibly the information gained from the first encounter about the opponents aggressiveness (based on the UV signal) is retained but in subsequent encounters associated with the opponent’s identity and no longer the signal per se, potentially explaining the high repeatability of conflict outcome within dyads between rounds. This would also explain the observation in wild, territorial blue tits that conflict outcomes are almost solely determined by the difference between opponents in distance from territory to the site where the conflict takes place, without obvious signs of aggression (Colquhoun 1942; Korsten et al. 2007; Vedder et al. 2008). Only individual recognition combined with sense of place would allow for such conventions to be stable, after initial settlement, without signalling of contest-specific motivation.

Social information must be context-specific for this line of reasoning to hold, as otherwise the absence of an effect of familiarity (i.e. coming from the same aviary) cannot be explained. In great tits, the effect of ‘personality’ on the probability of winning differs between contexts. Whereas fast-exploring, more aggressive, males are favoured in a pairwise setting (Verbeek, Boon & Drent 1996), slow but thorough explorers perform better in stable hierarchies in aviaries (Verbeek et al. 1999). Similarly, the size of the black ventral breast stripe determines conflict outcome in great tit males in pairwise settings, but not in aviaries (Lemel & Wallin 1993), which was also shown for the black bib in coal tits (Periparus ater; Brotons 1998). Apparently, the kind of information that is used to settle conflicts varies depending on the social context. As such, it is possible that individuals disregard information gathered in a different context when confronted with the same opponent.

Alternatively, the relatively slow recovery to original UV reflectance after the reduction treatment may have led to loser-effects (Hsu, Earley & Wolf 2006) occurring within the aviaries where the males were housed in between rounds. If such loser-effects carried through to the second round of trials, it would also explain the high repeatability in outcomes and absence of a treatment effect in the second round. However, this scenario appears unlikely because in the second winter, we did not observe any effect of the first round UV-reduction treatment on dominance status within the aviaries, in between rounds (O. Vedder and Y. Roelofs, unpublished data).

In conclusion, we provide evidence in birds (for lizards see: Stapley & Whiting 2006) that structural-based UV plumage is used in the settlement of conflicts, but only under specific social conditions. As the effect of our manipulation overruled differences in UV reflectance prior to manipulation, our results imply that male blue tits exhibit the behavioural plasticity to adjust their aggressive response to an opponents signal. The mechanistic link between UV reflectance and testosterone in first-year male blue tits (Peters et al. 2006; Roberts, Ras & Peters 2009) would provide a basis for some degree of within individual consistency in aggressiveness linked to signal expression, necessary for a static signal to provide information on an individuals inherent aggressiveness. However, in the wild, as individuals become locally established, individual recognition may replace the function of UV reflectance as a means to resolve conflicts, which may cause both the functional and mechanistic basis for the link between aggressiveness and colouration to deteriorate with age. Future work should aim to more accurately determine the social context in which colour signals are displayed and received.


We would like to thank Peter Santema and Yvonne Roelofs for practical assistance with the performance of the dominance trials. Piet Drent and Kees van Oers assisted in hand-rearing the nestlings in 2007. Caroline Isaksson gave advice on how to take the nestlings into captivity in 2008. Sandra Bouwhuis and two anonymous referees provided valuable comments on earlier versions of the manuscript. Permission for all procedures involved in the experiment was granted by the Animal Experiments Committee (DEC) of the University of Groningen. The research was financially supported by Grant No. 028696 from the European community’s sixth framework programme (FP6/2002–2006) to J.K.