As previously described in Demeyrier et al. (2016), four features of the urban matrix were measured within a 50 m radius around each nest-box: vegetation cover, car traffic, pedestrian count, and intensity of artificial light. Demeyrier et al. (2016) provide all the details on how these measures were performed, but also how they were analyzed in a principal component analysis (PCA, see Figure 4 in Demeyrier et al., 2016). From this PCA, we extracted the two major axes: PC1 reflecting habitat naturalness and hereafter abbreviated “naturalness” (positively related to vegetation cover and negatively correlated with car traffic and artificial lights, explaining 56.6% of the variance), and PC2 reflecting the pedestrian frequency, hereafter abbreviated “pedestrian” (explaining 16.7% of the variance in the urban environmental features).

The behavioral trials were performed in both La Rouvière forest and Montpellier in 2014, 2015, and 2016. Great tit parents were caught in their breeding nest-boxes while feeding their 9–15 days old nestlings. A captured bird was kept for around 20 min in total, during which time it was submitted to the following sequence of events: capture in the nest-box, handling aggression scoring, resting phase in cloth bag for 5 min, measure of breath rate during handling, resting phase of 2 min in small acclimatization compartment adjacent but separated from the main open-field cage (Figure 1), exploratory scoring in open-field trial during 4 min, ringing (if not already ringed), morphological measurements (body mass, tarsus, tail and wing lengths), and blood sampling.

First, immediately after its capture, a bird was tested for handling aggression by the capturer. The handler held the bird high up in the legs with one hand and nagged the bird with one finger from his/her other hand following a standardized procedure where the finger approaches the bird but does not touch it. After ca. 15 s of test and observation, the handler attributed a handling aggression score (HA, following the standardized criteria list in Table S4 of Dubuc-Messier et al., 2016, see also Brommer and Kluen, 2012; Class et al., 2014), ranging from 0 (completely unreactive bird) to 3 (bird aiming for the handler's finger all the time, spreading its wings and tail), with 0.5 increments. Following this first behavioral assay, the bird was placed in a bag for 5 min and brought to the open-field apparatus which was less than 5 min walking distance from the nest-box. Second, right after it was removed from the cloth bag, the bird was held still on its back by the handler who counted twice in a row the number of breaths during 30 s (breaths were visually expressed via fast movements of the chest). Handling breath rate (HBR) was the number of breaths per minute, averaged over the two measures (Carere and van Oers, 2004; Fucikova et al., 2009; Brommer and Kluen, 2012; Kluen et al., 2014). HBR has been shown in other tit populations to be positively related to stress response (Carere and van Oers, 2004). Third, a “novel environment test” was performed in a unique experimental open-field cage built following the exact set up and dimensions described in Stuber et al. (2013) (Figure 1). This kind of set-up has been classically used to measure exploration behavior in great tits (e.g., Stuber et al., 2013) and blue tits (e.g., Mutzel et al., 2013; Dubuc-Messier et al., 2016), and provides exploratory scores that are repeatable and well correlated to those performed in the laboratory (e.g., in Dutch and German great tits, Dingemanse et al., 2002; Stuber et al., 2013). Right after the HBR measure, the bird was placed in a double door small compartment on the side of the open-field cage for acclimatization during 2 min with no visual cue other than the small cage of dimensions 11 × 12 × 11 cm. Following this resting time, the bird was released through a sliding door into the exploration cage of 61 × 39 × 40 cm dimensions, where it was filmed during 4 min by a camcorder (CH4 Sony Handycam) fixed 110 cm away from the wired side of the open-field cage. The bird was free to move around the cage and/or to settle on three perches. The light was standardized since the open-field cage was placed in a wider cage of dimension 61 × 150 × 40 cm, lighted by two led lights. Following previous studies on great tits, the total number of flights and hops during the open-field trial was used as a proxy of exploratory behavior (e.g. Dingemanse et al., 2002). This number was square root transformed (to minimize deviation from normality and following Nicolaus et al., 2016) and hereafter called exploration score. VD analyzed all videos using the software JWatcher (Blumstein et al., 2006) and scored this behavior.

All statistical analyses were run with the software R (version 3.3.2, R Core Team, 2016). We used several datasets to restrict the limitations of missing data in the various variables. In a first step of the analyses, we compared breeding, personality and physiological traits between the forest and the urban habitats using two-sided t-tests on the entire dataset available (considering the data from both known and unknown parents), hence not restricting to observations where the identity of one parent was known. For this and all other analyses, we removed estimations of number of hatchlings and number of fledglings for all broods that had been (experimentally) manipulated after the egg-laying stage, which explains smaller sample sizes for these two variables compared to lay date and clutch size.

The first step in the statistical analysis compared population means using a dataset where many female identities were unknown (since parents were not captured during the first 2 years of monitoring) providing an idea of results that can be obtained without comprehensive monitoring and individual ringing.

In a second step, we more formally tested for differences between forest and urban birds with univariate linear mixed models (LMM, package lme4, Bates et al., 2015) which included habitat type (forest versus city) as a fixed effect and female identity as a random effect. All LMMs initially included as fixed effects habitat (forest versus urban), year, a habitat × year interaction and other trait-specific fixed effects (see below). The fixed effect model was selected in a backward stepwise procedure starting with a model including all variables. The significance of each fixed effect was tested using likelihood-ratio tests (Bates et al., 2015). While all behavioral and physiological traits showed no significant correlations (Tables 1C,D), the three reproductive performance traits were highly correlated (Tables 1A,B and Table S1). Hence clutch size was included as a fixed effect in brood size models (hereafter called hatching success), and brood size was included as a fixed effect in number of fledglings models (hereafter called fledgling success). For handling aggression and breath rate, the initial full model included fixed effects for sex (and sex × habitat interaction), age, individual life-long capture rank (from 1 to 3, for first to third capture of the same individual) and ambient temperature, and observer/manipulator identity as a random effect. For exploration score, we included sex (and sex × habitat interaction), age, test rank (from 1 to 3), and temperature as fixed effects. As our factor of interest, habitat was kept in all models, even if not significant, to provide the effect size and statistics associated with its fit in the model. Since these LMMs included a female random effect, the dataset used in this second step included breeding parameters that had been recorded for broods where the female was captured only, hence the number of observations was reduced compared to the first dataset.

Based on the best models retained in this step 2, we estimated an adjusted repeatability with the package rptR (Nakagawa and Schielzeth, 2010) for all focal traits. We estimated overall repeatabilities across the whole dataset but also sex-specific and habitat-specific repeatabilities.

In a third and final step, we used only data collected in the city of Montpellier to contrast the measured traits against the two urbanization gradients revealed by the PCA of environmental features in the city: habitat naturalness, abbreviated “naturalness” (PC1) and pedestrian frequency, abbreviated “pedestrian” (PC2). Model selection was similar to that in step 2 (with similar fixed and random effects structure), apart from the use of naturalness and pedestrian instead of the habitat variable which previously was used to compare forest and city. As previously, naturalness was kept in all models since it was our main variable of interest.

Sample sizes are indicated in Table 2 (for step 1), Table 3 (for step 2), and Table 4 (for step 3). While models for phenology and reproductive success were female-specific, models on behavior and physiology pooled data on both sexes, with balanced sampling across sexes since 47–48% of observations were on males (detailed sex-specific sample sizes are in Table S3).

We confirm here a commonly found breeding divergence, which is that city birds breed earlier and produce smaller clutches, and even more so in more urbanized areas of the city (Tables 3, 4, Figures 2A,B and 3A,B). Such urban effects on the timing of breeding and clutch size have been demonstrated before in great tits (e.g., Dhondt et al., 1984; Wawrzyniak et al., 2015) and many other avian species (see review in Marzluff, 2017). Note however that in a recent comparative analysis of forest and urban passerines at a macro-geographic scale (including 138 great tit populations), Vaugoyeau et al. (2016) found that average laying date was not correlated with the degree of urbanization, while average clutch size decreased with urbanization in collared (Ficedula albicollis) and pied flycatchers (Ficedula hypoleuca) but not in great tits. Later laying and smaller clutches in cities have been attributed to the urban heat island effect, as well as the artificial lights and the lack of adequate resources (i.e., arthropods) during the breeding season (Chamberlain et al., 2009; Kempenaers et al., 2010).

Hatching success was lower in urban broods, yet this did not lead to lower fledging success (Table 3, Figure 2D) nor to lower fledgling numbers (Table S2). Within the city, fledging success varied significantly across years but again was not significantly lower in more urbanized areas. It is possible that the slightly lower fledgling productivity in urban broods (Table 2, Table S2) was not significant in our 6 years study because of insufficient power. However, it remains that while the other breeding traits show strong significant differences between forest and city, it did not translate in a similarly strong divergence for fledgling success. This result contrasts with previous findings in great tits (Bailly et al., 2016) and with the common assessment that the urban environment induces strong constraints on avian reproduction (Gil and Brumm, 2014). It raises the very interesting question of whether urban birds might display adaptive breeding strategies that differ from birds in the forest, yet do not lead to lower reproductive success (see further discussion on this possibility, below). Reduced clutch size would hence be the result of an adaptive process rather than the expression of energetic constraints. Overall in terms of POLS, the breeding data suggested that city birds displayed one element pointing to slower life histories (smaller clutches) yet did not differ consistently from forest birds in their reproductive output.

Our results on personality traits and breath rate were more consistent with the prediction of faster behavioral syndromes in urban compared to forest habitats. Handling aggression was not significantly influenced by the urban habitat, yet city birds were faster explorers with generally faster breath rate under constraint (Table 3). Breath rate also varied across the urban gradient, yet with important annual variation (Table 4, Figure 3C). Previous rural vs. urban avian comparisons for these traits found: lower short-term stress response in urban European blackbirds (Partecke et al., 2006), reduced physiological stress and faster exploratory behavior in urban dark-eyed juncos Junco hyemalis (Atwell et al., 2012; Abolins-Abols et al., 2016), higher territorial aggression in urban song sparrows Melospiza melodia (Evans et al., 2010), urban Australian magpies Gymnorhina tibicen (Cilento and Jones, 1999), urban Eurasian coot Fulica atra (Minias, 2015), and urban noisy miners Manorina melanocephala (Lowry et al., 2011), and finally higher breath rates in urban great tits during handling (Torne-Noguera et al., 2014, but see Senar et al. from this issue).

Mean breath rate during handling for Montpellier great tits was similar to that found in the study by Torne-Noguera et al. (2014, i.e., around 130 breaths/min) yet it was only around 2% higher than breath rate in forest birds, compared to a 14% difference in Barcelona (Torne-Noguera et al., 2014). Our singular analysis along the urban gradient revealed strong annual variation in the breath rate response to urbanization level (Figure 3C), which could partly explain the discrepancies between previous studies (Partecke et al., 2006; Torne-Noguera et al., 2014; Senar et al. this special issue). While heart rate and breath rate are highly related (see e.g., Dubuc-Messier et al., 2016), their interpretation is complicated by the fact that they reflect both sympathetic and parasympathetic reactivity (e.g., Koolhaas et al., 1999), and that responses to long term or acute stresses differ (Abolins-Abols et al., 2016). In rodents more aggressive and active individuals have a strong sympathetic response yet a low parasympathetic reactivity to acute stress, which translates into increased heart rate response (Koolhaas et al., 1997). Similarly, domestic hens with active coping styles (high-feather peckers) have faster heart rates during manual restraint (Korte et al., 1999). Hence overall, in the context of POLS, higher breath rate and faster exploration scores in urban great tits are suggestive of a faster pace-of-life.

While we have demonstrated in this study that forest and city great tits display strong phenotypic differences for breeding parameters, behavioral traits and physiological reaction to stress, two fundamental questions remain unanswered. The first question is whether these physiological differences emanate from underlying genetic differences, rather than plasticity. This would suggest that urban birds form a genetically distinct ecotype compared to forest birds, possibly because of a process of local adaptation. The second, related question, is whether the phenotypic differences we have described here result from an adaptive process since both plasticity and genetic processes (such as drift) can result in maladaptive responses.

As regards the first issue of a plastic vs. genetic origin in the phenotypic differences, a genetic difference can occur only if the focal traits are heritable, and in the case of restricted gene flow (Lenormand, 2002). An interesting point to make here is that in our study populations, hatching and fledging success were not repeatable, and there is very little evidence for heritability of reproductive success in the literature (review in Postma, 2014). All other focal traits studied here were repeatable and they were previously shown heritable, in great tits or in closely related avian species (for lay date and clutch size see review in Postma, 2014, for aggression and exploration behavior see review in Dingemanse and Dochtermann, 2014, for breath rate see Class et al., 2014). This overview on heritability suggests that a rapid evolution of the focal traits is possible except for hatching and fledging success. Recent results in population genomics also provide useful knowledge on the neutral and adaptive genomic divergence of Montpellier great tits at both scales studied here (forest vs. city and urban gradient). Using a RAD sequencing analysis of nearly 50,000 SNPs in 140 great tits, Perrier and colleagues (Perrier et al. in review) show that the neutral genetic structure is weakly impacted by urbanization, with an average Fst between La Rouvière forest and Montpellier city of 0.009 and an Fst of 0.012 between the most and least urbanized sites within the city of Montpellier. While all Fst were significant, their low value suggests that gene flow is common along the urban gradient and between forest and city. However, it could well be that non-random gene flow, e.g., via individuals matching their habitat choice to their phenotype (Edelaar et al., 2008; Edelaar and Bolnick, 2012) reinforces local adaptation. In fact it has been suggested that urbanization should favor individuals with high behavioral plasticity (Lowry et al., 2013). Our preliminary results on habitat-specific repeatability (Table S3) show lower repeatability in the behavior and physiology of urban compared to forest birds which could be indicative of higher plasticity, although this is not true for breeding traits. An interesting avenue would hence be to focus on plasticity as a trait per se, evaluating how/whether it is heritable and selected for in the urban landscape and across the range of traits studied here (e.g., following the approach of Nussey et al., 2005 for plasticity of timing of breeding). We also envisage in the future a common garden study to experimentally test whether birds from more or less urbanized areas yet raised in aviaries display genetically based differences in the traits studied here.

Addressing the second question to find out whether the phenotypic variation described here is adaptive or not will require more long term data (including survival and reproductive data) to compare the direction and force of selection acting on the focal traits across the habitats (Isaksson, 2015). It is commonly assumed that the novelty of the urban environmental features induces new and strong selection pressures providing the potential for rapid trait change (Markó et al., 2013; Alberti, 2015; Hendry et al., 2017). However, direct empirical measures of urban natural selection remain scarce (but see e.g., the emblematic peppered moth case, Cook and Saccheri, 2013). New genomic resources offer a powerful way to bypass the data consuming classical selection analyses by testing for genome-wide signatures of selection associated with urbanization, even in non-model species (Harris et al., 2015). While at first hand, breeding parameters show decreased clutch size and hatching success in city great tits compared to their forest conspecifics (Table 3), a very important result from our study is that these city birds do not have lower fledgling rate or success (Table 3 and Table S2). This stability of reproductive performance in the nest despite strong differences in breeding traits, suggests that city birds might have evolved an adapted breeding strategy in response to urbanization with smaller clutches and broods to adjust to the lack of food for nestlings in the urban habitat (Chamberlain et al., 2009). A comparison of morphological traits in our focal populations also points to possibly adaptive changes since urban great tits are smaller and lighter than forest birds yet they do not differ in body condition (Demeyrier et al. in prep.).

Previous studies have shown enhanced annual survival in urban adult birds, in particular because of higher resource availability over winter (Rebolo-Ifran et al., 2015; Marzluff et al., 2016). In the context of the pace-of-life theory, such decreased probability of mortality would be indicative of a slower pace-of-life (Réale et al., 2010). However, new predation pressures in the city are also thought to affect avian survival, especially nestlings', and overall rural/urban survival comparisons are still scarce and show no consistent pattern (Chamberlain et al., 2009). With only 3 years of breeding captures, our study lacked power to conduct capture-mark-recapture analyses, yet it is most definitely our objective to pursue the long term research and compare survival rates across the urbanization gradient in the future. In the same vein, a longer monitoring will enable a comparison of recruitment success across urbanization, providing a complementary perspective to our results on fledgling success. Survival data will also be crucial to understand selection pressures acting on urban great tits, and to compare them with selection in more natural habitats.

While the present analysis focused on first broods following classical approaches in the literature, considering second broods in the study of forest and urban reproductive strategies could reveal further divergences. Between 2011 and 2016, 11.6% of Montpellier urban females laid second broods vs. 4.5% in La Rouvière forest (along with 20.6 and 23.1% of replacement broods). In our study sites in the city of Montpellier, Pinus spp. and the evergreen oak Quercus ilex are dominant tree species. In these breeding habitats, later breeding could be favored because the food required to rear nestlings peaks later in the season (Blondel et al., 2006). Also, in such habitats, second clutches can be typically larger than first broods, while the reverse is true in deciduous habitats such as found in La Rouvière forest (Lambrechts et al., 2008). Hence a complementary analysis involving both first and second broods might confirm that city birds have adopted a different yet adaptive breeding strategy.

Finally, the colonization of an urban landscape by an avian species is likely to be a long process, possibly leading to a succession of adaptive phenotypes. In line with the general finding that individuals that are superior colonizers are inferior competitors (Yu and Wilson, 2001), it has been suggested that while a proactive fast lifestyle would be favored during the colonization stage of urban habitats (Sol et al., 2013), the higher sensitivity of reactive slow individuals to the environmental conditions might favor them on the long run when facing ecological novelty (Robertson et al., 2013). In the case of such a sequence of colonization, findings regarding the pace-of-life of birds adapted to the urban habitat will depend a. on how advanced the colonization process is, and b. on how much the urban population is isolated from surrounding rural populations. Although Montpellier is a millenary city where great tits have a long history of presence, it is still undergoing a rapid urbanization process, which could explain the high prevalence of bold birds in the most urbanized areas. Also, recent genetic analyses based on genome-wide RAD-sequences have revealed that Montpellier and La Rouvière great tits show significant but low genetic differentiation (Perrier et al. in review). Uncovering the direction of gene flow between city and forest, and testing for a source-sink dynamic would reveal whether the city population could be regularly supplied with new, possibly bold, individuals.

Overall, our study shows strong phenotypic divergence between forest and urban great tits, but also across the urbanization gradient. Behavioral and physiological traits suggested that urban birds were on a faster end of the pace-of-life continuum than their forest conspecifics, yet breeding data did not support this conclusion. Further investigations on the plastic vs. genetic origin of the differences revealed here, along with the selection forces on the focal trait, will improve our understanding of how urban life shapes great tit pace-of-life in the future.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The handling Editor declared a past collaboration with one of the authors AG and states that the process nevertheless met the standards of a fair and objective review.