Disentangling drivers of reproductive performance in urban great tits: a food supplementation experiment

Urbanization is a fast-developing process worldwide, currently involving more than 50% of the human population (Marzluff, 2001; Gaston et al., 2015). This urban expansion induces degradation of natural environments, thereby promoting species that are able to adapt to cities (Clergeau et al., 2006; Adams and Lindsey, 2011) while often threatening native species (Aronson et al., 2014). Indeed, urban habitats involve new constraints in terms of breeding sites and anthropic disturbances (Hedblom and Söderström, 2012), but also offer new resources (Clergeau et al., 2006; Mackenzie et al., 2014). For birds, resource abundance and nature are drastically modified in urban habitats (Isaksson and Andersson, 2007; Gladalski et al., 2015). First, urban breeding habitats are notoriously poor in natural avian food such as caterpillars (Perrins, 1965) because of the low density of trees. Second, exotic species of trees are often introduced in urban areas without their associated insect cohorts (Mackenzie et al., 2014), which also contributes to the low abundance of arthropod prey for birds. Third, human waste can represent a new resource type, although its nutritional properties, in particular fat-richness, can be detrimental for avian growth (Meillère et al., 2015). In urban environments, birds could switch to these novel food resources to compensate for the low abundance and poor quality of ‘natural’ food, e.g. arthropods, that is less nutritive than caterpillars (Mackenzie et al., 2014), or to other types of food provided by feeders (Robb et al., 2008; Chamberlain et al., 2009). These constraints on resources could be worsened by a high density of birds in urban areas because of the attractiveness of breeding sites, the availability of food resources in winter and a more favourable microclimate (Payevsky, 2006; Hedblom and Söderström, 2012). Such increased population density is thus likely to lead to higher competition for a poor quantity and quality of food (Shochat, 2004; Chace and Walsh, 2006).

Food resources available in urban areas could thus be expected to be detrimental for birds and to induce a non-adaptive investment in reproduction (Anderies et al., 2007). Resource acquisition is a key factor involved in the trade-off between breeding success and somatic maintenance of parents (Martin, 1987; Karell et al., 2009). During the avian breeding period, the timing and abundance of resources have a critical influence on avian breeding success, as they are strong determinants of nestling body condition and survival (Perrins, 1991; Hõrak et al., 1999; Reed et al., 2013). Hence, it is not surprising that food availability and quality have been identified as major drivers of the decline in nestling mass and lower fledging success for passerines in urbanized landscapes (Solonen, 2001; Chamberlain et al., 2009; Hedblom and Söderström, 2012; Seress et al., 2012; Meillère et al., 2017). In 1965, Perrins' study of great tit nestlings (Parus major) in gardens revealed that they were lighter than nestlings from a nearby forest, and that urban chicks displayed all the signs of dying of starvation (Perrins, 1965). Likewise, one of the drivers of the decline in house sparrow (Passer domesticus) populations in European cities could be the poor food conditions (Meillère et al., 2015). Great tits also face reduced breeding performance and nestling body condition in urban environments (Solonen, 2001; Chamberlain et al., 2009; Bailly et al., 2016; Demeyrier et al., 2016), possibly driven by constraints in food availability and accessibility (Remacha and Delgado, 2009), as well as a high heterogeneity of the urban habitat (Amrhein, 2014).

The detrimental effects of low food resources could be exacerbated in urban areas because of artificial cues. In systems facing human-induced rapid environmental change (HIREC; e.g. Sih, 2013; Hale and Swearer, 2016), cues exploited for habitat selection could be decoupled from the true quality of the habitat, potentially causing maladaptive responses. In a recent study (Demeyrier et al., 2016), we have demonstrated that urban great tits from the city of Montpellier (in the south of France) adjust their breeding investment to artificial breeding cavity size in a maladaptive way. We experimentally showed that great tits preferred settling in large breeding cavities relatively to smaller options. Individuals from these largest cavities also invested higher in egg production, yet had a lower fledging success compared with those from medium-sized cavities. One hypothesis to explain these results is that females produce broods that are too large when choosing large cavities, and are then trapped because of a lack of food to rear these relatively large broods (Anderies et al., 2007; Robertson et al., 2013). Rytkönen and Krams (2003) proposed the same hypothesis to explain the lower breeding success in northern European compared with central European great tits.

The present study aims at testing experimentally the role of food abundance in mediating breeding performance in an urban system where maladaptive breeding investment associated with breeding cavity size has been previously reported (Demeyrier et al., 2016). We performed a feeding experiment over 2 years on wild urban great tits breeding in nest boxes varying in size in the city of Montpellier.

Moreover, we investigated in detail the variation in breeding performance according to the local breeding environment within our urban population. As highlighted by Amrhein (2014), urban habitats need ‘to be characterized by small-scale variation in environmental factors that include differences in vegetation and natural food resources’, which is rarely the case as urban studies focus mainly on urban parks (e.g. Björklund et al., 2009 in Barcelona, Spain; Hedblom and Söderström, 2012 in Uppsala, Sweden; Markowski et al., 2013 in Lodz, Poland). The nest-box distribution in our urban habitat was designed specifically to take into account the wide variety of urban landscapes comprising both parks and residential or commercial streets.

Previous avian studies with food supplementation experiments have found contrasted results on bird reproduction performances. Indeed, some studies showed positive effects of food supplementation (Robb et al., 2008) whereas other studies showed no or even negative effects (Harisson et al., 2010; Sillanpää et al., 2010; Ruffino et al., 2014). Following the suggestion by Robb et al.’s (2008) study that food is a key limiting factor in the urban environment, we predicted that: (1) fledgling body condition and fledging success, as well as parental body condition, would be improved by food supplementation; (2) supplementation would be more beneficial in the most urbanized environments; and (3) supplementation would be more beneficial for broods reared in large nest boxes, where birds are most impacted by urban constraints (see Demeyrier et al., 2016). Overall, this study aims at understanding how food resources influence the reproductive performance of free-ranging urban great tits.

The current work was conducted under an agreement with the host research institute CEFE (Centre d'Ecologie Fonctionnelle et Evolutive, C34172-11) and the CNRS (Centre National de la Recherche Scientifique). In addition, the specific work was conducted following authorizations obtained from the national authorities in France: (1) authorization to work with and capture wild birds from the Centre de Recherches sur la Biologie des Populations d'Oiseaux (CRBPO, Permit number 1904, Museum Paris to A.G.); (2) authorization to work with and capture wild birds in the specific county (authorization number 2012167-003 from Hérault department to A.G.); and (3) personal certificates for animal experimentation (A34-470 & A34-496 from the Direction Départementale de la Protection des Populations de l'Hérault to A.G. and A.C.).

Since 2011, urban great tits (Parus major Linnaeus 1758) have been monitored from March to July in the city of Montpellier, in the south of France (43°36′43″N, 3°52′38″E). This city counted 272,084 residents in 2013, with a density of 4783 residents km^–2. It covered an area of around 57 km² with around 11 km² of green areas. The city was equipped in autumn 2010 with 243 nest boxes placed along an urbanization gradient. At the beginning of the 2011 breeding season, only 168 nest boxes were still there, because of human damage. During the 2012–2015 breeding seasons, the number of nest boxes fluctuated between 163 and 180.

Food supplementation experiments were conducted during the breeding seasons of 2013 and 2014 in two wooden nest-box types that differed in their bottom floor area (medium=121 cm² versus large=210 cm²; see Demeyrier et al., 2016 for details). The distance between neighbouring boxes was at least 100 m, to limit intraspecific competition (Remacha and Delgado, 2009). All nest boxes were visited once per week from the beginning of the breeding season (late March) until the fledgling phase, to collect data on occupation rate, clutch size and fledging success. Adults were trapped inside their nest boxes with a mechanical trap when nestlings were ∼9 days old. They were ringed with unique leg rings provided by the CRBPO (Paris, France); their right tarsus length was measured with a calliper and their body mass measured using a Pesola balance. Nestling body mass was also measured at that time, corresponding to the peak of nestling food requirement (Perrins, 1991). At this age of 9 days, nestlings are still in their growth stage, hence tarsus length measures are not possible because of growth cartilage presence. When nestlings were 14–16 days post-hatching, we re-measured their body mass, as well as their right tarsus length to calculate a body condition index (i.e. body mass controlled for nestling size).

To quantify local habitat characteristics per territory, we defined a disk with a radius of 50 m around each box (Hinsley et al., 2002; Hedblom and Söderström, 2012; following Demeyrier et al., 2016). This allowed us to take six measurements reflecting the level of artificialization inside the city: vegetation cover (oak cover, tree cover and global vegetation cover), motorized and non-motorized traffic, and light pollution. To provide a proxy of the abundance of insect prey for tits in each territory the three vegetation cover variables (Mackenzie et al., 2014; see Demeyrier et al., 2016) were measured from aerial pictures of the city of Montpellier (http://www.montpellier.fr/4053-cartographie-ancienne-de-montpellier.htm), taken in summer 2011 and analysed using the software Delta. We also measured, using 5 min counts, motorized, non-motorized and pedestrian traffic, at each nest box during the nestling stage (Hedblom and Söderström, 2012; Katti and Warren, 2004; see Demeyrier et al., 2016). These counts were cross-validated with information available from measurements taken during longer periods by the Montpellier urban service (http://data.montpellier3m.fr/, R²=0.96, P<0.001; Demeyrier et al., 2016). Finally, light pollution around each breeding cavity was quantified by the spatial surface overlap between the 50 m disk around each nest box and a 50 m disk around each street lamp (Mapinfo data, http://data.montpellier3m.fr/, Quantum GIs 1.7.4, Kempenaers et al., 2010; see Demeyrier et al., 2016 for details). Principal component analyses on these six urban factors (three types of vegetation cover, motorized traffic, pedestrian traffic, artificial lighting; ade4 R package, Dray et al., 2016) revealed two major axes of variation in these anthropic disturbance measures: PC1 was related to the degree of environment artificiality/naturalness (56.6% of the variance, with positive loadings for the three types of vegetation cover and negative ones for light pollution and traffic), called PC1 or ‘habitat naturalness’ thereafter, whereas PC2 (or ‘human presence’) was related to only the pedestrian frequency (16.6% of the variance) (details in Demeyrier et al., 2016).

We performed a supplementation experiment in 2013 and 2014, during the first seasonal breeding attempts, in order to follow a homogeneous sample according to breeding stage. All medium (n₂₀₁₃=55, n₂₀₁₄=52) and large (n₂₀₁₃=50, n₂₀₁₄=51) nest boxes were equipped with a feeder placed inside the box during this 2-year experiment (Sillanpää et al., 2010) (Fig. 1).

We randomly assigned a control status (empty feeder) to half the large and half the medium-sized nest boxes, the other halves being experimental nest boxes where supplementary food was provided in the feeder. Throughout the breeding season after hatching, the status of nest boxes (empty/full) was attributed alternatively within nest boxes of the same size. In line with the literature and with previous feeding experiments on blue tits (Cyanistes caeruleus) and great tits (Eeva et al., 2009; Gienapp and Visser, 2006; Grieco, 2001; Sillanpää et al., 2010), following hatching, feeders were refilled every second day with two species of live mealworm larvae (Alphitobius diaperinus and Tenebrio molitor), and a vitamin and carotenoid complement (Nutribird) (Sillanpää et al., 2010). Mealworm larvae quantities were modulated according to the number of nestlings and their age (Gienapp and Visser, 2006; Grieco, 2001) (see Table S1). From 0 to 8 day of age, nestlings were provided with small Tenebrio larvae (Alphitobius diaperinus); from 8 to 15 days of age, nestlings were provided with miller Tenebrio larvae (Tenebrio molitor). These quantities represent 20-50% of the daily required biomass for nestlings during the feeding period (Eeva et al., 2009; Grieco, 2001). We added a Nutribird complement in feeders, varied according to worm quantity (Sillanpää et al., 2010; Koivula et al., 2011). The Nutribird complement was provided along the proportion of 5 g of Nutribird (a full coffee spoon) for 15 g of worms. The carotenoid supplement was 0.0015 mg per full coffee spoon. The Nutribird complement provided proteins (1.05 g per full coffee spoon), oligo elements (0.225 g per full coffee spoon) and diverse vitamins (1.395 mg per full coffee spoon). Each time we refilled the feeder, we checked whether the larvae were eaten, although a small consumption was difficult to measure. In 21% of refills, at least half of the worms had disappeared and were not found in the nest box. Thus, in the remaining 80% of the refills, fewer than half of the worms disappeared.

All the nests were handled in the same way with the following monitoring methods: (1) weekly visits to control all nest-box contents (also during the experimental stage); (2) daily visits around the hatching date to determine the hatching date of all nests; (3) nestling manipulation at 9 and 15 days post-hatch of all nests; and (4) capture of parents and associated phenotypic measurements of all nests (between 9 and 15 days post-hatch). Only the food-supplemented nest boxes were additionally briefly visited to fill up the feeders (less than 1 min per feeder visit). Thus, we assume that the additional visit to fill up the feeders should have had a minor effect on the outcome of the experiment relatively to the potential impact (if any) of the above-mentioned monitoring procedures.

We used the breeding monitoring data collected in 2013 and 2014 from all nest boxes equipped with an inside feeder to test for a food supplementation effect on nestling body condition and fledging success, along the urbanization gradient within the city. In addition, we evaluated whether the food supplementation affected parental physical condition.

We used general linear mixed models (GLMM) run with R package lme4 to analyse nestling body condition (at 9 and 15 days of age) with a normal error distribution, nestling survival until fledging with a binomial error distribution (0 or 1 for each nestling), and finally adult body condition with a normal error distribution. We followed a stepwise backward regression model selection (keeping fixed effects with P<0.1).

To address our objective to investigate the role of resources on breeding performance using our supplementation experiment, we modelled nestling body condition, individual nestling survival as a proxy of fledging success, and adult body condition within the feeding experiment. Explanatory variables were: nest-box size, feeder status (food supplemented/control), PC1 (habitat naturalness), PC2 (human presence), year and all biologically relevant interactions (nest-box size×feeder status, year×feeder status, PC1×feeder status, PC2×feeder status, nest-box size×feeder status×PC1). All these interactions explore the potential effect of food availability in relation to the nest-box size (see Demeyrier et al., 2016) and the local environment (PC1 and PC2) in an experimental framework. Additional explanatory factors were: number of nestlings in the brood when modelling nestling body condition; number of hatchlings when modelling fledging success; and, finally, number of nestlings, breeders' age (1 year old or more) and breeders' sex when modelling adult body condition. Nest-box number was included as a random effect as 60% of nest boxes were occupied more than once within the 2-year experiment. Maternal identity (i.e. mother ring number) was also included as a random effect, to account for sibling relatedness and also for environmental/genetic maternal effects. When modelling adult physical condition, ring number was included as a random effect as 8% of adults were trapped twice.

Contrary to our predictions, there was no significant statistical interaction between food supplementation, nest-box size and habitat naturalness, on nestling phenotype (Table 1). We found a significant negative effect of supplementary food on 9-day-old nestling body mass (body mass, in grams, in relation to feeder treatment: mean±s.d. food supplemented=13.17±2.17; control=13.64±2.73; Table 1; Fig. 2).

Regarding the body condition of 15-day-old nestlings (i.e. body mass controlled for nestling tarsus length), we found a significant negative main effect of human presence and a marginally significant interaction between supplementary food and human presence (Table 1). Namely, in the food supplemented broods, nestling body condition was more negatively impacted in areas with higher human presence (Fig. 3). There was also a marginal effect of nest-box size, whereby nestlings from medium-size nest boxes were in higher body condition than nestlings from large nest boxes (Table 1). Finally, all three nestling traits (9-day-old body mass, 15-day-old condition and survival until fledging) showed significant annual variation. All three nestling traits were lower in 2014 than in 2013 (Table 1).

There was a significant positive association between individual nestling survival until fledging (fledged=1, died=0) and the level of naturalness of the breeding environment (Table 1). However, there was also a significant interaction between the food supplementation treatment and the naturalness level on individual nestling survival until fledging (Table 1). Nestling survival until fledging decreased overall in food-supplemented broods (Fig. 4A), and this detrimental experimental effect was strongest in the less-natural environments (Fig. 4B).

When exploring adult body condition (i.e. body mass relative to tarsus length), we found a significant interaction between supplementary food and human presence, as illustrated in Fig. 5 (Table 2). Adult body condition was negatively impacted by the supplementary feeding only in areas with high human presence. However, parents of food-supplemented broods were in better body condition when human presence was lower (Fig. 5).

We also found that birds were marginally in worse body condition when they were 1 year old rather than older (body condition in relation to age, in grams: mean±s.d.; 1 year=16.46±0.78, >1 year=16.95±1.02). The classic sexual size dimorphism was also confirmed, with males heavier than females (body condition in relation to sex: mean±s.d.; male=17.25±0.83, female=16.33±0.87) yet the experimental effects were similar for both sexes (Table 2). Finally, adult body condition varied significantly between the 2 experimental years, i.e. adult body condition was overall worse in 2014 than in 2013 (Table 2).

Contrary to our expectations, the experiment resulted in an overall negative effect of food supplementation for traits associated with breeding performance (nestling phenotype and survival). At least three possible explanations could account for this unpredicted experimental outcome, when contrasting our results with the literature.

The first explanation for the observed pattern is that the food supplemented was not of sufficient quality for a nestling diet, particularly in areas where food resources available are already penalizing their growth (e.g. Meillère et al., 2015). Our choice of diet (mealworms and Nutribird complement) was based on the most commonly used food for great tits in aviaries (Eeva et al., 2009; Gienapp and Visser, 2006; Grieco, 2001; Sillanpää et al., 2010). Negative effects of food supplementation have previously been found in at least two other great tit studies (Harisson et al., 2010; Sillanpää et al., 2010; Ruffino et al., 2014) and these negative effects have been attributed to a lack of nutrients other than proteins in the supplementation diet (Eeva et al., 2009). Indeed, mealworms contain mostly protein, fat and water. We anticipated this potential problem by adding a vitamin and carotenoid complement; however, the results obtained support the idea that this may not be sufficient or may not be assimilated by birds when they eat the worms. Moreover, Simons et al. (2014) found that providing supplemental carotenoids to zebra finches (Taeniopygia guttata) could have negative effects on reproductive performances, such as clutch size, in constraining environments. Adding nutrients to mealworms could hence also lead to the negative effect seen in our supplementation experiment.

Moreover, parental body condition, and hence feeding abilities, could also mediate the negative effect of food supplementation. Perhaps the low body condition of parents classically found in urban areas (Bokony et al., 2009; Meillère et al., 2015) could impact their foraging capacities, resulting in a greater use of artificial food instead of natural food (Robb et al., 2008). However, parents involved in our experiment did not display decreased body condition in relation to the naturalness level, whereas they were impacted by the human presence (Fig. 5) and so were their nestlings (Fig. 3). Hence, this hypothesis is unlikely to explain the detrimental effect of supplementation, which is mostly revealed in areas with higher human presence (Fig. 3) and in most urbanized areas (Fig. 4B).

Second, we could postulate that the experiment was conducted during exceptionally good years where supplementation with a low diverse food source might have negatively impacted breeding performance. However, nestling survival until fledging was not significantly different across the 6-year study period in this urban great tit population (P=0.4; 2011-2016), supporting the idea that the 2 years of food supplementation were not exceptionally good. A full understanding of the underlying environmental drivers (e.g. weather fluctuations) associated with the significant year effects for all traits (Tables 1 and 2) will require multiyear studies.

Finally, we cannot rule out, as a third possible explanation, the possibility that experimental broods were negatively affected by the disturbance caused by the added food within the nest box. In our study, we provided supplementary food inside each nest box in order to avoid negative effects of feeding due to increased competition for visible food sources present outside the boxes (Solonen, 2001; Robb et al., 2008). In fact, in great tits, high competition in urban areas is most probably responsible for smaller clutch sizes and fewer fledglings per breeding attempt, because of a combination of food shortage and high bird density (Solonen, 2001; Hedblom and Söderström, 2012). Moreover, Briga et al. (2017) showed that in high-competition areas, low food availability results in shorter lifespan for large-brooded zebra finches. Yet, the choice of placing feeders inside the boxes could induce a stronger behavioural and possibly stress-induced reaction from the breeding parents. Even though we habituated the birds with the presence of feeders inside the nest boxes by adding them before the breeding season, similar to studies by Eeva et al. (2009) and Sillanpää et al. (2010), we cannot exclude the possibility that the novelty of the feeder might be perceived as being much stronger for a full rather than an empty feeder due, for example, to the movement or smell of the worms. However, it is important to notice that feeder status (food supplemented or control) affected body condition of 15-day-old nestlings and adults, as well as nestling survival until fledging, always in interaction with another variable associated with local territory environment. Food supplementation was indeed found to have a positive effect in addition to certain levels of naturalness of the breeding sites. Thus, if perturbation due to items in the feeders was the prevailing effect, such interactions would not have been predicted.

These non-exclusive potential explanations for our main experimental results are probably not exhaustive. Indeed, we are aware that food supplementation effects could be modulated by environmental factors that are not controlled for in such experiments, such as natural resources or latitudes, but also by behavioural variations in birds, such as food catching, or by the experimental design, such as the timing of supplementation or the accessibility of food (Ruffino et al., 2014).

An initial aim of the present work was to evaluate whether food resources were the key parameter mediating the maladaptive breeding investment observed in association with large artificial cavity size (Demeyrier et al., 2016). Thus, we predicted an interaction between cavity size and food supplementation treatment, with a contribution of local conditions (naturalness level): we envisaged that when provided with additional food supply, larger broods in larger nest boxes would no longer display lower fledgling numbers compared with smaller broods in medium-sized boxes. However, no interaction of the two key factors associated with food and cavity size was detected on nestling or parental traits. This absence of significant result did not necessary indicate that the role of food was not present but we were not able to detect it with our protocol.

Nestling survival until fledging was overall decreased in supplemented broods (Fig. 4A), and this detrimental experimental effect was strongest in the less natural environments (Fig. 4B). This result was the exact opposite of the positive effect that was expected from supplementary feeding.

In a highly constrained (i.e. more artificial) environment, as stated above, the addition of poor-quality food could be a penalty for birds rather than a ‘bonus’, i.e. nestling survival until fledging could be negatively impacted in a highly artificialized habitat, whereas in a more natural environment, supplementary food, even of a poor quality, could improve the nestling survival until fledging (Fig. 4B). In fact, in natural environments, most studies have found a positive effect of food supplementation or no effect on nestling survival until fledging, according to the review by Robb et al. (2008). Following a feeding experiment in Spotless Starlings (Sturnus unicolor), Cuervo et al. (2011) suggested that artificial food could cause nestling satiety, which could lead to less begging for higher quality natural food. In this case, artificial food could not compensate completely for the lack of natural food (Cuervo et al., 2011) and it could in fact worsen nestling survival until fledging. One element that is crucially missing in this puzzle is a fine knowledge about great tit diet in cities, and how this diet, and in particular the nestling diet, varies across areas in the heterogeneous urbanized landscape (see, for example, Isaksson and Andersson, 2007; Andersson et al., 2015; Toledo et al., 2016).

Interestingly, the experiment resulted in an interaction between the effects of food supplementation and the local pedestrian frequency on 15-day-old nestling body condition (Fig. 3) as well as on adult body condition (Fig. 5). Nestlings and adults from supplemented broods presented a decreased body condition (relative to controls) only in areas with numerous pedestrians. Human presence could induce a temporary halt in nestling provisioning because of a behavioural reaction of parents, due to a hormonal response to stress (Müller et al., 2006, on blue tits) or due to a personality-linked response (e.g. flight initiation distance and landing distance, Fernández-Juricic et al., 2006). Our procedures might have added too strong a disturbance in areas where human presence is already high, resulting in a detrimental effect on nestling body condition. Dubiec (2011) suggested that the response of birds to human presence depends on parental body condition. Indeed, this author revealed a trade-off between reproduction and survival in great tits: parents with a low body condition deserted the nest more easily than parents with a high body condition (Dubiec, 2011). This hypothesis could be worth considering in our population because in our experimental design, the body condition of the breeders was influenced by human presence in relation to our feeding experiment, even if it was not influenced by urbanization level (Table 2).

In areas highly disturbed by human presence, birds could modulate their foraging strategy in relation to the predation risk and travel longer distances in order to provision their nestlings in safer or less stressful conditions (Sol et al., 2013; Voelkl et al., 2016). When facing a high predation risk, as human presence could be perceived, parents might also display longer latency periods before returning to their nest (Bokony et al., 2012), also affecting negatively the condition of their nestlings. In poor habitats such as urbanized areas, great tit parents might also be oblivious to human presence, as previous studies have shown that great tit females modulate their risk-taking behaviour according to the risk of nestling starvation (Quinn et al., 2012). When we provided food inside the nest box, parents could reduce their foraging distance, and hence be more disturbed by pedestrian frequency, as indicated by the interaction between human presence and feeding treatment. Indeed, although knowledge on foraging strategies in urban environments is very scarce (Amrhein, 2014), it is likely that parents forage further away from their nest box in poor habitat, such as the more urbanized areas. Previous studies in several bird species support the idea that human presence can modify the spatial and temporal foraging patterns, thereby modulating foraging costs (Fernández-Juricic and Tellería, 2000; Bokony et al., 2012; Sol et al., 2013; Voelkl et al., 2016) and that tolerance to disturbance is reduced with an increasing stimulus such as pedestrian presence (Fernández-Juricic and Tellería, 2000).

Our experiment was one of the first to explore experimentally the role of food availability on breeding performance in a city environment. The results we obtained in our Mediterranean study system highlight that the response to the question is potentially complex and that the presence of artificial food during the breeding season might exacerbate a negative effect of urbanization. In order to better understand processes that associate food resources and nestling rearing conditions, an alternative approach that could be conducted is to manipulate brood size (Bańbura et al., 2013; Hõrak, 2003). If breeders had to feed fewer nestlings and if food resources were a major constraint in urbanized areas, a brood size experiment would contribute to the same objective as this study, i.e. understanding how food resources influence the reproductive success of wild urban great tits and whether the ecological trap effect described previously is due to a lack of food and/or to the low-quality resources of urban environments. In light of the results from the present manipulation, such a brood size experiment would solve the issue of controlling for food quality in the alleviation of parental effort.

We thank all the people that assisted in the field work: Pascal Pagano, Stéphane Pouplard, Rémy Spitaliery from the City Council of Montpellier, Matthieu Abouladzé, Frédéric Lidon from the Zoo of Montpellier, Ananda Christophe and Coline Cannone. We thank two anonymous reviewers for useful suggestions on an earlier version of the manuscript.