Abstract

Adaptation to urban habitats presumably requires changes in cognitive, behavioral, and physiological traits enabling individuals to exploit new resources. It is predicted that boldness, reduced neophobia, and enhanced problem-solving and learning skills might characterize urban birds compared with their rural conspecifics, while exposure to novel pathogens might require an enhanced immunity. To test these predictions, we assessed problem solving, color discrimination learning, boldness, neophobia, and immunocompetence in the bullfinch Loxigilla barbadensis , a highly opportunistic and innovative endemic bird in Barbados, wild-caught from a range of differently urbanized sites. Birds from urbanized areas were better at problem solving than their rural counterparts, but did not differ in color discrimination learning. They were also bolder but, surprisingly, more neophobic than rural birds. Urban birds also had an enhanced immunocompetence, measured with the phytohemagglutinin antigen. Our study sheds light on the trade-offs acting on animals exposed to changing environments, particularly in the context of urbanization.

INTRODUCTION

Urbanization is considered one of the most severe threats for biodiversity and has been shown to dramatically alter animal abundance and diversity through the extinction of native species or changes in their distribution ( Case 1996 ; Crooks 2004 ; Sol et al. 2014 ). Characterizing traits that enable successful species to thrive in urban habitats is a key to a better understanding of the evolution of urban ecosystems. Species that are abundant in urban areas are expected to be characterized by behavioral and physiological traits that increase their ability to survive and reproduce in such environments ( McKinney and Lockwood 1999 ; Evans et al. 2009 ). Boldness and plasticity are, according to a review by Lowry et al. (2013) , the 2 most important factors that affect a species’ response to urban environments. Boldness is often measured by how well animals tolerate proximity to humans, neophobia by how animals react to novel environmental features, whereas plasticity can be assessed by readiness to solve a novel food-finding problem. There is overwhelming evidence on dozens of avian species that urban populations tolerate a closer approach by humans than do rural populations of the same species ( Cooke 1980 ; Knight et al. 1987 ; Valcarcel and Fernández-Juricic 2009 ; Evans et al. 2010 ; Lowry et al. 2011 ; Atwell et al. 2012 ; Møller and Tryjanowski 2014 ), but results on flexible problem solving and tolerance of novel environmental features are both much rarer and less clear. Problem-solving differences between conspecifics from urban and rural populations are sometimes weak (e.g., urban house sparrows better on only one of 4 tasks, Papp et al. 2014 ) or confounded by other variables (e.g., body size, Papp et al. 2014 ), while urban birds are often more or equally neophobic toward novel environmental features compared with rural ones ( Echeverría et al. 2006 ; Echeverría and Vassallo 2008 ; Bókony et al. 2012 ; Miranda et al. 2013 ).

In this article, we examine differences between urban and rural Barbados bullfinches ( Loxigilla barbadensis ), an endemic opportunistic and innovative species that is very successful in urban areas, but which is also abundant in less-disturbed areas of Barbados ( Webster and Lefebvre 2000 ; Reader et al. 2002 ; Ducatez et al. 2013 ). The island of Barbados shows a strong gradient of human disturbance, thus providing an excellent environment to study the effects of urbanization. In addition to boldness, neophobia, and innovative problem solving, which Sih and Del Giudice (2012) predict should be associated, we test 2 variables that could be linked with urbanization, enhanced immunocompetence and faster reversal learning. Reversal learning is the classical measure of behavioral flexibility in psychology. Contrary to obstacle removal tasks in innovative problem solving, much is known about the neural ( Lissek et al. 2002 ), genetic ( Krugel et al. 2009 ), and psychological ( Cools et al. 2002 ) mechanisms of reversal learning, an advantage that could ground ecological studies of flexibility in a wider literature. Tebbich and Teschke (2014) have used reversal learning to show that Darwin’s finches from an arid zone that goes through sharp variation in dryness make fewer reversal errors than conspecifics from a less variable cloud forest. As far as immunocompetence is concerned, comparative studies on both birds ( Garamszegi et al. 2007 ; Vas et al. 2011 ; Soler et al. 2012 ) and primates ( McCabe et al. 2015 ) have shown that increased contact with pathogens is one of the costs of innovative foraging. Enhanced immunocompetence might thus be one of the responses that behaviorally flexible animals develop or evolve given the wide and novel array of pathogens they encounter as a result of invasive ( Sol, Duncan, et al. 2005 ) and generalist ( Ducatez, Clavel, et al. 2015 ) lifestyles that go with flexibility.

MATERIALS AND METHODS

Subjects

Fifty-three Barbados bullfinches were captured in mist nets between February and May at 8 different sites throughout the island of Barbados that were selected in order to obtain a wide range of urbanization rates ( Table 1 ). Urbanization rates were calculated using the percentage of anthropogenic structures in a 1 km 2 area around the point of capture (as in Jacquin et al. 2013 ; Table 1 , also see map: Supplementary Figure S1 ).

Table 1

Summary of captured birds and their site of capture

Site name GPS coordinates Anthropization n  
Swans +13°14′10.96″, −59°35′17.16″ 2.1% n“rural” = 27  
Bruce Vale +13°13′18.98″, −59°33′30.74″ 2.4%  
White Hill +13°13′18.24″, −59°34′31.68″ 3.7%  
Jah +13°15′18.80″, −59°35′14.56″ 5.6%  
Bellairs +13°11′31.21″, −59°38′25.20″ 18.0% n“urban” = 26  
Jamestown park +13°11′18.84″, −59°38′7.79″ 21.1%  
Payne’s Bay +13°9′47.83″, −59°38′10.71″ 25.3%  
Bridgetown +13°5′50.98″, −59°37′21.65″ 54.7%  
Total   53  
Site name GPS coordinates Anthropization n  
Swans +13°14′10.96″, −59°35′17.16″ 2.1% n“rural” = 27  
Bruce Vale +13°13′18.98″, −59°33′30.74″ 2.4%  
White Hill +13°13′18.24″, −59°34′31.68″ 3.7%  
Jah +13°15′18.80″, −59°35′14.56″ 5.6%  
Bellairs +13°11′31.21″, −59°38′25.20″ 18.0% n“urban” = 26  
Jamestown park +13°11′18.84″, −59°38′7.79″ 21.1%  
Payne’s Bay +13°9′47.83″, −59°38′10.71″ 25.3%  
Bridgetown +13°5′50.98″, −59°37′21.65″ 54.7%  
Total   53  

Summary of all captured birds by site. Urbanization was measured as the percentage of a satellite map covered by human landmarks (roads, buildings, etc.).

Morphological measurements

Morphological measurements were taken at capture on all 53 birds by the same person (J.N.A.); measurements were taken 3 times in succession on each bird, and the mean value of the 3 measures was used in the analyses below. Individuals were weighed using a digital pocket scale (precision to 0.1g). We measured tail length as the length of the longest straightened rectrix using a metal ruler (precision to 0.5mm). Wing length was taken with a raised-end ruler as the length of the unflattened wing chord (precision to 0.5mm). Calipers were used to measure the metatarsi, bill, and head (precision to 0.05mm). Metatarsal length was measured from the intertarsal joint to the last scale before the toes. Bill length was measured from the tip to the anterior edge of the nostril. Head length was measured from the anterior edge of the nostril to the back of the head following the angle of the bill ( Audet et al. 2014 ). Residuals of body weight against wing length were used as a proxy of body condition.

Captivity conditions

Birds were housed in individual cages (height: 92cm × width: 73.5cm × length: 81cm) that were visually but not acoustically isolated from each other in an indoor aviary. Tests started after a 2-day habituation period during which the birds were fed ad libitum. On the day of the first behavioral test, birds were food-deprived overnight for 14h. Behavioral tests began at 9:00, a 1-h pause was given at noon during which they were given 10min to feed and the tests stopped at 16:00. Birds were then fed ad libitum until the next overnight deprivation (starting at 19:00). Birds were given a commercial mix of finch seeds when fed before and between the tests and also as a reward for behavioral tests. During the tests, the observer (J.N.A.) was hidden behind an opaque curtain and observed through small holes at a distance of 5 m from the cages. Birds lost on average 0.7g of their body weight at capture, which represents a mean of 4.3% of their initial weight. At the end of the captivity period (7 days), birds were released at their initial site of capture. Three out of the 53 birds died from unknown causes during captivity testing (1 female from White Hill, 1 female from Swans, and 1 male from Bellairs); they were excluded from analyses. All experiments were conducted according to Animal Use Protocol 2013-7140, approved by the McGill University Animal Care Committee and permit 8434/56 from the Natural Heritage Department of the Barbados Ministry of Environment and Drainage.

Behavioral tests

Tests were always given in the same order to reduce the potential biases emerging from habituation (see Ducatez, Audet, et al. 2015 for a detailed explanation). Behavioral tests started on day 3 of captivity with boldness assessment. Birds were presented with an open Petri dish full of seeds (same dish and food as during the habituation period), and the experimenter hid behind the curtain until the bird had fed (all birds fed within 12min). The same procedure was repeated on the 3 following days of captivity to assess repeatability of the boldness measure. On day 3, after the boldness test, neophobia was assessed by presenting a novel object beside the Petri dish until the bird fed or reached the maximum 20-min limit of the trial, in which case this 20-min latency was recorded for the individual. Neophobia was measured as the latency to feed, minus the mean latency of pre-control (previous boldness measurement) and post-control (additional boldness measurement). Failure to remove boldness could result in a confounding of an animal’s response to the novel object and to the human presenting it ( Greenberg 1983 ). The first novel object was a 30-cm yellow stake ( Supplementary Figure S2A ). To estimate repeatability of neophobia, we took another measure of neophobia on day 6, after changing the novel object to 2 brightly colored and textured balls (dog toys, 50-mm diameter) placed directly on each side of the dish ( Supplementary Figure S2B ). As our measure of neophobia, we used the mean of the neophobia latencies obtained on days 3 and 6. There was no significant difference between neophobia measured on the first presentation of the 2 objects (days 3 and 6: P = 0.235).

On day 3, after the neophobia trial, we assessed problem-solving ability using the lid-drawer task (see Supplementary Movie 1 ). A 2cm × 3cm × 3cm drawer made of white plastic was constructed with a circular opening (1.5-cm diameter) at the top covered with a lid to which a hook was attached ( Supplementary Figure S2C ). The birds had the opportunity to gain access to food by opening the lid or by pulling the drawer. Birds were given a maximum of 15 trials each lasting 5min. The problem-solving score was defined as the latency to succeed, which was started when the individual touched the apparatus for the first time, thus removing initial boldness or neophobia effects from the problem-solving score. The other problem-solving task, the tunnel task, was given on day 4 of captivity. It consisted of a transparent rectangular box (height: 3cm × width: 3cm × length: 10cm) opened on only one side ( Supplementary Figure S2E : as presented to the birds and Supplementary Figure S2F : opened, see also Supplementary Movie 2 ). A transparent cylindrical tube containing seeds and topped with a loose fitting white lid was inserted at the closed end of the tunnel and a wooden stick was attached to it so that the birds had to pull on the stick to get the tube out of the tunnel. Once the tube was out, the bird had to remove its lid to gain access to seeds. Birds were given a maximum of 15 trials each lasting 5min, and problem-solving latency was measured in the same way it was for the lid-drawer task.

On day 5 of captivity, a color discrimination task was made to first assess acquisition learning ability. The test apparatus consisted of 2 Petri dishes (same as the one used for the boldness assessment), each inserted in a wooden platform (10×10×10cm 3 ) painted either green or yellow and open on one side, placed at each extremity of the cage ( Supplementary Figure S2D ). A “color bias” trial was first made, where the bird was allowed to eat from 1 dish, and the color of the wooden platform chosen by the bird was considered as its preferred color. The other color thus became the rewarded one in order to control for initial color bias. The Petri dish inside the unrewarded color contained seeds glued to the bottom of the dish, so that no difference could be seen from a distance but the seeds were impossible to remove for the birds. This task was designed to measure discrimination learning without a problem-solving or motor skill component because the bird only had to choose a color without performing a novel motor task. Novelty was also reduced because the birds were already habituated to feed from similar (but not color associated) Petri dishes. On each trial, the 2 platforms were introduced simultaneously inside the cage. The bird was given up to 5min to choose a dish. If the bird chose the rewarded color, it was allowed to feed for 15s. If the bird chose the unrewarded color, the 2 platforms were immediately removed by the experimenter. A “choice” is defined as the first peck movement toward the seeds in the Petri dish on either side. Because the seeds were glued on the unrewarded side, the peck yielded no reward on this side. The location of the rewarded platform was switched at each trial to control for spatial preference. The success criterion was reached once the birds chose successively the correct (rewarded) color for 7 consecutive trials ( Boogert et al. 2010 ). On the day after this criterion was reached, we assessed reversal learning. We switched the rewarded color and tested the birds in the same way we did in the acquisition phase. On the first reversal learning trial, all birds initially chose the previously rewarded color (which was incorrect at this stage), indicating that they effectively learned the color stimuli, and not a potential perceptible difference in the Petri dishes. See Supplementary Figure S2 for pictures of all tasks. On completion of all tasks (including boldness and neophobia assessment), birds were allowed to feed for 2min. A new task was started only when every bird had completed the previous one (either success or maximum number of trials reached). This allowed for a relatively constant food intake while keeping the birds hungry enough to be motivated.

Immunocompetence assessment

Immunocompetence was assessed using a phytohemagglutinin (PHA) injection, a measure of the cellular immune response. Measurement of PHA-induced swelling in birds is a well-established immunoecological technique that has the advantage of assessing general innate immunity (and to a lesser extent adaptive immunity), and it is easily performed in the field ( Martin et al. 2006 ). It was performed on the last day of captivity (day 7) by subcutaneously injecting PHA at a concentration of 5 µg/g (e.g., 0.033mL of a 3mg/mL PHA solution for a 20-g bird) in the proximal portion of the wing, as described in Martin et al. (2006) . We measured swelling of the tissue with a micrometer caliper (Mitutoyo, Mississauga, Ontario, Canada) by subtracting the wing thickness before injection from thickness of the same region 21.5±0.6h after the injection.

Sex typing

Loxigilla barbadensis is monochromatic, so molecular sexing of individuals is required. Approximately 50 µL of blood was sampled by puncturing the brachial vein. DNA was extracted from blood, and PCR sexing was performed following Audet et al. (2014) .

Statistical analyses

To test for an effect of urbanization on our different variables, we separated our capture sites into “urban” ( n = 4 sites; mean urbanization score = 30% ± 12%) and “rural” sites ( n = 4 sites; mean urbanization score = 3% ± 1%; see Supplementary Table S1 ) and used this binary variable in our linear models. Normality of the data was assessed using D’Agostino–Pearson tests. The only datasets that did not follow a Gaussian distribution were the results of the 2 problem-solving tasks. Therefore, we computed the P value only for mean differences of the latter variables using a nonparametric t -test (Mann–Whitney) for the data presented in Figure 1b (note however that the computed problem-solving PC1 followed a Gaussian distribution). To test for the effect of urbanization on all other variables along with all potential confounding variables, we performed linear models and then conducted stepwise variable selection until only significant effects remained. A Principal component analysis was performed on latency to solve the lid-drawer and tunnel tasks and the first component, which explained 64% of the variance ( Supplementary Figure S3 ), was used as the general problem-solving score. For all models, tarsus length (which was found to differ between rural and urban environments), sex, body weight, and body condition along with urbanization were used as explanatory variables. For neophobia, we added boldness as a potential confounding variable. For problem-solving and discrimination learning models, we also added neophobia along with boldness as potential confounding variables. Correlations between each variable and percent urbanization as a continuous rather than a binary variable were also tested.

Figure 1

Behavior and immunity in urban versus rural environments. (a) Temperament. Boldness (latency to feed following human disturbance) is higher in birds coming from urban environments than birds from rural environments ( P = 0.0056). Neophobia (average of latency to feed in the presence of a 2 different objects on 2 different days) is higher in urban birds compared with rural birds ( P = 0.0010). (b) Problem solving. In both problem-solving tasks, the latency to succeed is lower in urban individuals compared with their rural counterparts (lid-drawer P = 0.0338; tunnel P = 0.0206). (c) Discrimination learning. The number of trials to succeed in acquisition and reversal learning does not significantly differ between rural and urban birds (acquisition learning P = 0.7970; reversal learning P = 0.9835). (d) Immunocompetence. Intensity of reaction following PHA injection is higher in urban than in rural birds ( P < 0.0001). * P < 0.05, ** P < 0.01, *** P < 0.001.

Figure 1

Behavior and immunity in urban versus rural environments. (a) Temperament. Boldness (latency to feed following human disturbance) is higher in birds coming from urban environments than birds from rural environments ( P = 0.0056). Neophobia (average of latency to feed in the presence of a 2 different objects on 2 different days) is higher in urban birds compared with rural birds ( P = 0.0010). (b) Problem solving. In both problem-solving tasks, the latency to succeed is lower in urban individuals compared with their rural counterparts (lid-drawer P = 0.0338; tunnel P = 0.0206). (c) Discrimination learning. The number of trials to succeed in acquisition and reversal learning does not significantly differ between rural and urban birds (acquisition learning P = 0.7970; reversal learning P = 0.9835). (d) Immunocompetence. Intensity of reaction following PHA injection is higher in urban than in rural birds ( P < 0.0001). * P < 0.05, ** P < 0.01, *** P < 0.001.

Additionally, we tested whether associations between behavioral and immunity variables varied between urban and rural populations. To that aim, we built models with proxies of cognition, problem solving, or immunocompetence as response variables, and urbanization and proxies of behavior or immunocompetence along with their interactions as fixed effects.

Finally, we also tested all models using a linear mixed model (LMM) approach with the capture site as a random variable; the results of the latter models are presented in Supplementary Tables S1–S6 .

Repeatability was calculated using the rpt.adj function from the rptR package ( Nakagawa and Schielzeth 2010 ) in R 3.2.1, which allows a comparison of latencies adjusted for a given parameter. For boldness, we used the latencies measured on 4 consecutive days and added the day of measurement (habituation parameter) in the model, as latencies usually decrease as the birds habituate to this test and to human presence in general. For neophobia, we used the 2 measures of neophobia (days 3 and 6) and added measurement day in the model. The individual ID was used as the random variable.

JMP software (SAS Institute, Cary, NC) was used to compute all linear models, SPSS Amos software (IBM, Armonk, NY) for path analyses, rptR package in R (R Core team, Vienna, Austria) for repeatability calculations, and Prism 5.01 (GraphPad software, La Jolla, CA) to draw graphs.

RESULTS

Morphology

None of the morphological traits, including body condition, differed between rural and urban birds, except tarsus length (rural birds: 0.41mm longer than urban birds, t = 2.22, P = 0.0307), which became nonsignificant after Bonferroni corrections (see Table 2 ). To be conservative, we nevertheless included tarsus length as an explanatory covariable in all subsequent linear models.

Table 2

Final linear models after variable selection from models that included all potential confounding variables

Final models ~ urbanization N r2 t P (α = 0.007)  
Tarsus (length) 52 0.09 −2.22 0.0307 
Boldness (latency) 52 0.15 −2.91 0.0056a 
Neophobia (latency) 52 0.24 3.55 0.0010 
Problem solving (PC1, latency) 52 0.15 −2.94 0.0049 
Acquisition learning (trials) 52 0.00 −0.04 0.7970 
Reversal learning (trials) 52 0.00 0.02 0.9835 
Immunocompetence (thickness) 45 0.33 5.10 <0.0001 
Final models ~ urbanization N r2 t P (α = 0.007)  
Tarsus (length) 52 0.09 −2.22 0.0307 
Boldness (latency) 52 0.15 −2.91 0.0056a 
Neophobia (latency) 52 0.24 3.55 0.0010 
Problem solving (PC1, latency) 52 0.15 −2.94 0.0049 
Acquisition learning (trials) 52 0.00 −0.04 0.7970 
Reversal learning (trials) 52 0.00 0.02 0.9835 
Immunocompetence (thickness) 45 0.33 5.10 <0.0001 

Following stepwise selection of variables, only urbanization remained in all models, except when stated a . Thus, t ratios and P values for only urbanization effects are compiled in this table. Values in bold represent significant values following Bonferroni corrections (specified α).

a Sex also remained significant after variable selection ( P = 0.0366) but not after Bonferroni correction.

Temperament

Boldness and neophobia measurements were both repeatable. The computed repeatability for boldness was 0.427 (standard error [SE] = 0.066, confidence interval [CI] = [0.300, 0.548]) and was significant ( P < 0.0001). The repeatability for neophobia was 0.350 (SE = 0.138, CI = [0.072, 0.629]) and was also significant ( P = 0.0109). Boldness was higher in birds living in urban environments than in birds living in rural environments: Urban birds were faster at eating after human disturbance compared with rural birds ( Figure 1a , left panel). After stepwise selection of potential confounding variables, urbanization remained significant ( t = −2.91, P = 0.0056) as well as sex (mean boldness for rural females = 143s, rural males = 43s, urban females = 23s, urban males = 62s, t = 2.16, P = 0.0366) ( Table 2 , see Supplementary Table S1 for detailed models). However, after Bonferroni corrections, boldness was no longer significantly explained by sex and only urbanization remained as a predictor of boldness (see Table 2 ). In contrast, neophobia was higher in urban birds ( Figure 1a , right panel). Following stepwise selection of variables, urbanization was the only significant explanatory variable for neophobia ( t = 3.55, P = 0.0010) ( Table 2 , see Supplementary Table S2 for detailed models).

Innovative problem solving and discrimination learning

The lid-drawer task was completed by all birds from both environments. Twenty-six percentage of the rural birds and 50% of urban birds succeeded in completing the tunnel task. Latency to succeed at the 2 problem-solving tasks (lid-drawer and tunnel) varied significantly with urbanization (lid-drawer: tnonparametric = −2.18, P = 0.0338; tunnel: tnonparametric = −2.39, P = 0.0206; Figure 1b ) and were correlated with each other ( r = 0.280, P = 0.042). Urban birds performed better than rural birds on the problem-solving PC1 ( Supplementary Figure S3B ). Following variable selection in the linear models, urbanization remained as the only significant variable explaining problem-solving performance (PC1: t = −2.94, P = 0.0049, Table 2 , see Supplementary Table S3 for detailed models). In contrast, discrimination learning scores did not differ between birds living in the 2 environments ( Figure 1c ). For acquisition learning, models yielded no significant effect for all tested predictors (urbanization: t = −0.036, P = 0.7970, Table 2 , see Supplementary Table S4 for detailed models). Similarly, reversal learning did not significantly vary with urbanization nor with any other tested predictor (urbanization: t = 0.020, P = 0. 9835, Table 2 , see Supplementary Table S5 for detailed models).

Immunity

The injection of PHA triggered a significant swelling of the skin (mean difference = 236.5±21.6 µm, Mann–Whitney U = 78, P < 0.0001). When comparing birds from both environments, urban birds had a 2.6-fold stronger PHA reaction than rural birds ( Figure 1d ). Following stepwise selection of variables in a linear model, only urbanization remained as the only significant factor explaining PHA response ( t = 5.10, P < 0.0001, Table 2 , see Supplementary Table S6 for detailed models).

All the variables previously mentioned as significant remained so after Bonferroni corrections ( nVARIABLES = 7, P < 0.007) except tarsus length ( P = 0.0307) (see Table 2 ). Using a mixed model approach that included capture site as a random variable also resulted in the same variables remaining significant following stepwise variable selection (all P < 0.05, see Supplementary Tables S1–S6 ).

Urbanization gradient

In addition to the analyses categorizing urbanization as a binary variable, we repeated our analyses using percentage of urbanization for each of our 8 capture sites, yielding a continuous gradient. We then reran models using the same procedure, and we used with the binary urbanization variable, that is, backward selection with all potential explanatory variables. The urbanization gradient was the only significant predictor of boldness ( t = −2.12, P = 0.0392, Supplementary Figure S4A ) and neophobia ( t = 3.39, P = 0.0015, Supplementary Figure S4B ). Latencies to solve the 2 problem-solving tasks were negatively related to urbanization gradient (lid-drawer: t = −2.38, P = 0.0211; tunnel: t = −2.51, P = 0.0151). Furthermore, when the 2 tasks were combined into 1 principal component expressing problem-solving ability ( Supplementary Figure S3A ), it was significantly correlated with the urbanization gradient ( t = −3.16, P = 0.0026) ( Supplementary Figure S4C ). Scores for acquisition and reversal learning remained nonsignificantly associated with the urbanization gradient (acquisition: urbanization t = −0.54, P = 0.5948; reversal: urbanization t = −0.57, P = 0.5709). Immunocompetence, as measured by skin swelling following PHA injection, varied strongly with the urbanization gradient, and this was also the only variable remaining after stepwise variable selection ( t = 7.97, P < 0.0001) ( Supplementary Figure S4D ).

Interactions between behavioral and immunocompetence variables

Linear models that included interactions between urbanization and all previously measured variables (behavioral and immunological) were tested ( Supplementary Table S7 ). No significant interaction was found in any of the models ( Supplementary Table S7 ). The only significant effect found between all combinations of variables was between acquisition learning and reversal learning when including urbanization in the model ( r2 = 0.195/0.196, P = 0.002, Supplementary Table S7 ), individuals needing fewer trials to succeed at acquisition learning also needed less trials to succeed at reversal learning.

Path analysis

To summarize our results and test for dependencies among variables, we constructed a path analysis. We tested every plausible path between variables, and the model presented in Figure 2 is the only one in which all paths were significant ( P < 0.01, see legend for all model statistics). This suggests that, in accordance with our linear models, there is no correlation and/or interaction among the significant behavioral variables and immunocompetence and that urbanization is the only common driver of variation in the traits we measured.

Figure 2

Path analysis showing dependencies between variables. Every plausible path was tested, and this model is the only one in which all paths are significant. The model suggests that behavioral variables and immunocompetence are not directly affecting each other and that urbanization is the main variable influencing the measured variables. Variables e1 to e4 represent the error terms. Model statistics: chi square = 1, 302; degrees of freedom = 6; model probability level = 0.972; Akaike information criterion = 29.302; Browne–Cudeck criterion = 32.662; all individual paths P < 0.01.

Figure 2

Path analysis showing dependencies between variables. Every plausible path was tested, and this model is the only one in which all paths are significant. The model suggests that behavioral variables and immunocompetence are not directly affecting each other and that urbanization is the main variable influencing the measured variables. Variables e1 to e4 represent the error terms. Model statistics: chi square = 1, 302; degrees of freedom = 6; model probability level = 0.972; Akaike information criterion = 29.302; Browne–Cudeck criterion = 32.662; all individual paths P < 0.01.

DISCUSSION

Animals living in urbanized habitats are likely to be advantaged by traits that allow them to profit from human-derived food sources ( Sol et al. 2011 ). Here, we jointly measured behavioral and immunological traits predicted to affect birds’ success in urban areas. We showed that, as predicted, urban bullfinches were bolder, faster at problem solving, and had a stronger immune response than rural bullfinches ( Figure 1a , b , d ). Contrary to our predictions, however, urban birds were more neophobic and did not differ from rural birds in discrimination learning ( Figure 1a ,c). Our path analysis suggests that urbanization affects each of the traits we measured independently ( Figure 2 ). Differences in problem solving and immunocompetence between rural and urban birds were not explained by morphology, body condition, sex, boldness, or neophobia and were significant both when urbanization was considered as a continuous gradient and as a binary variable ( Figure 1b ,d and Supplementary Figure S4C,D ).

Urbanization is only one of the ecological contexts in which temperament, cognitive abilities, and innovativeness are expected to diverge. The expectation that boldness, low neophobia, and flexibility should all covary with ecology has also been applied to populations that experience different degrees of environmental harshness ( Roth et al. 2010 ; Tebbich and Teschke 2014 ; Kozlovsky et al. 2015 ). Darwin’s finches from a variable arid zone are more neophilic and faster reversal learners than conspecifics from a more stable cloud forest, but they are also more neophobic and not better at an obstacle removal problem ( Tebbich and Teschke 2014 ). Mountain chickadees from a harsher, higher elevation are better problem solvers than conspecifics from a milder lower elevation, but equally neophobic ( Kozlovsky et al. 2015 ). Black-capped chickadees from a harsh seasonal environment (Alaska) are both better problem solvers and less neophobic than conspecifics from a more benign southerly environment (Kansas; Roth et al. 2010 ). The relationship between flexibility and neophobia is thus inconsistent in the 3 studies. The same inconsistency characterizes comparative studies of urbanization, where access to new foods in novel environments (e.g., refuse at dumps with intense truck traffic, leftovers at tables with intense pedestrian traffic) should logically favor positive covariation between problem solving and neophobia. Sol et al. (2011) report that urban mynas are better at solving a technical innovation problem than are suburban ones, but they do not eat a new food faster; they were also less neophobic and more exploratory in pecking more often at the test apparatus. The study of Bókony et al. (2012) on house sparrows shows no effect of urbanization on neophobia, whereas our results shows that urban birds are more neophobic. The contradictory data on urbanization thus support the conclusions of Griffin and Guez (2014) in their review of innovation and problem solving: Neophobia does not generally covary with problem-solving ability. Why this is so is puzzling and warrants a closer look at both the conceptual basis of neophobia/neophilia and at the different ways of assessing it.

Despite their inclusion under the umbrella term “behavioral flexibility,” the fact that problem solving and reversal performance do not covary in our study fits with the results obtained by Tebbich et al. on Darwin’s finches ( Tebbich et al. 2010 ; Teschke et al. 2011 ; Tebbich et al. 2012 ; Tebbich and Teschke 2014 ), Griffin et al. (2013) on Indian mynas, Isden et al. (2013) on spotted bowerbirds, and Ducatez, Audet, et al. (2015) on Carib grackles. In most problem-solving tasks, performance is measured by the speed with which an animal removes an obstacle blocking access to food. Motor acts directed at inappropriate parts of the apparatus entail minimal costs and innovative animals routinely direct a wide diversity of movements to the apparatus at a fast rate (see Griffin and Guez 2014 and Griffin et al. 2014 for discussions on the role of motor diversity in problem solving). In contrast, reversal errors are more costly. They entail a time-out between unsuccessful trials, with the added disturbance of human intervention in tests that are not automated. These cost differences might lead to a speed accuracy trade-off as performance at problem-solving and reversal learning tasks relies on different skills. Problem solving requires fast, diversified pecks at many parts of the obstacle, whereas reversal learning entails accurate inhibition of response to previously rewarded stimuli. It is thus not surprising that performance on the 2 tasks shows either no relationship (our study, Tebbich et al. 2012 ; Isden et al. 2013 ) or a negative one ( Ducatez, Audet, et al. 2015 ; Griffin et al. 2013 ). Griffin and Guez (2014) conclude that obstacle removal is a valid experimental test of feeding innovations in the wild. Whether reversal learning, in particular multiple serial reversals, is also an ecologically valid test for innovativeness is open to question. Repeated, sudden reversals of the ability of stimuli to predict rewards could be a useful test of flexibility in humans, but it might not represent an ecologically relevant situation for animals forced to opportunistically switch to a new food or new technique when their usual foraging behavior does not work. In discussing the fact that innovative, tool using woodpecker finches make more reversal errors than nontool using small tree finches, Teschke et al. (2011) suggest that extractive foraging with tools requires perseverance, but reversal learning depends on the opposite, rapid change.

In line with predictions from the literature on both birds and primates, urban bullfinches showed both better innovative problem solving and increased immunocompetence. Møller (2009) obtained similar results in 39 urban species compared with rural congeners or relatives, with a higher innovation rate and a larger bursa of Fabricius (a key immune organ) in urban species. By definition, intraspecific comparisons like ours include fewer confounding variables than interspecific ones and are a more direct test of urbanization effects. The ability to mount a strong immune response is only one of the physiological adaptations urban birds have been shown to have. Suburban Florida scrub jays have lower plasma corticosterone levels than woodland jays, even when the latter are supplemented with high-fat, high-protein food ( Schoech et al. 2004 ). In conditions of acute stress, urban Eurasian blackbirds show a lower level of corticosterone than rural ones ( Partecke et al. 2006 ), as well as lower levels of oxidative stress ( Costantini et al. 2014 ). Urban and rural blackbirds further show different single-nucleotide polymorphisms for SERT genes ( Mueller et al. 2013 ), which are associated with anxiety-related traits. Finally, urban blackbirds have lower levels of blood parasites than do rural ones ( Geue and Partecke 2008 ).

Ecological conditions that favor differences in behavior, cognition, innovation, and physiology are likely to be sensitive to time and to population isolation before they lead to evolutionary divergence. Common garden experiments on populations that are separated by vast distances and long-term environmental differences provide the best evidence for evolved adaptive responses. This is the case for Alaskan and Kansas populations of black-capped chickadees studied by Roth et al. (2010 , 2012) and Pravosudov et al. (2013) . Eurasian blackbirds have been urbanized since the 1820s, and genetic differences with woodland conspecifics appear to have evolved independently in several areas of Europe ( Mueller et al. 2013 ). Barbados is a very small island, and it would be surprising if urban and rural bullfinches were geographically isolated, even if the island has a high population density and the original vegetation of the island has been destroyed and replaced by sugar cane and other anthropogenic plants for over 350 years. Enhanced boldness, problem solving, and immunocompetence in urbanized bullfinches might all be experience-driven responses to environmental variation in food, human disturbance, and pathogens. Individuals with different phenotypes might also choose habitats based on traits that provide the best context-dependent benefits, in the same way that longer- and shorter-winged Zenaida doves ( Sol, Elie, et al. 2005 ; Monceau et al. 2011 ) in Barbados feed at sites where territorial defense or group feeding is favored by food distribution ( Goldberg et al. 2001 ). At temporal and spatial scales where selection is unlikely, as is probably the case for Barbados bullfinches, or at scales where long-term trends and isolation might lead to genetic divergence, urbanization is one of the key situations that can help us understand how some animals respond positively to anthropogenic change.

SUPPLEMENTARY MATERIAL

Supplementary material can be found at http://www.beheco.oxfordjournals.org/

FUNDING

This work was supported by an Animal Behavior Society student research grant, a Society of Canadian Ornithologists student research award, and a FQRNT doctoral scholarship to J.N.A., a postdoctoral fellowship from the Fondation Fyssen to S.D., and a NSERC Discovery grant to L.L.

All experiments were conducted according to Animal Use Protocol 2014-7140, approved by the McGill University Animal Care Committee and permit 8434/56 from the Natural Heritage Department of the Barbados Ministry of Environment and Drainage. We thank L. Jacquin and M. Couture for helpful discussions and technical help.

J.N.A. and S.D. performed the experiments and analyses; J.N.A., S.D., and L.L. wrote the manuscript. All authors gave final approval for publication.

REFERENCES

Atwell
JW
Cardoso
GC
Whittaker
DJ
Campbell-Nelson
S
Robertson
KW
Ketterson
ED
.
2012
.
Boldness behavior and stress physiology in a novel urban environment suggest rapid correlated evolutionary adaptation
.
Behav Ecol
  .
23
:
960
969
.
Audet
JN
Ducatez
S
Lefebvre
L
.
2014
.
Morphological and molecular sexing of the monochromatic Barbados bullfinch, Loxigilla barbadensis
.
Zool Sci
  .
31
:
687
691
.
Bókony
V
Kulcsár
A
Tóth
Z
Liker
A
.
2012
.
Personality traits and behavioral syndromes in differently urbanized populations of house sparrows ( Passer domesticus )
.
PLoS One
  .
7
:
e36639
.
Boogert
NJ
Monceau
K
Lefebvre
L
.
2010
.
A field test of behavioural flexibility in Zenaida doves ( Zenaida aurita )
.
Behav Processes
  .
85
:
135
141
.
Case
TJ
.
1996
.
Global patterns in the establishment and distribution of exotic birds
.
Biol Conserv
  .
78
:
69
96
.
Cooke
AS
.
1980
.
Observations on how close certain passerine species will tolerate an approaching human in rural and suburban areas
.
Biol Conserv
  .
18
:
85
88
.
Cools
R
Clark
L
Owen
AM
Robbins
TW
.
2002
.
Defining the neural mechanisms of probabilistic reversal learning using event-related functional magnetic resonance imaging
.
J Neurosci
  .
22
:
4563
4567
.
Costantini
D
Greives
TJ
Hau
M
Partecke
J
.
2014
.
Does urban life change blood oxidative status in birds?
J Exp Biol
  .
217
:
2994
2997
.
Crooks
K
.
2004
.
Avian assemblages along a gradient of urbanization in a highly fragmented landscape
.
Biol Conserv
  .
115
:
451
462
.
Ducatez
S
Audet
JN
Lefebvre
L
.
2015
.
Problem-solving and learning in Carib grackles: individuals show a consistent speed–accuracy trade-off
.
Anim Cogn
  .
18
:
485
96
.
Ducatez
S
Audet
JN
Lefebvre
L
.
2013
.
Independent appearance of an innovative feeding behaviour in Antillean bullfinches
.
Anim Cogn
  .
16
:
525
529
.
Ducatez
S
Clavel
J
Lefebvre
L
.
2015
.
Ecological generalism and behavioural innovation in birds: technical intelligence or the simple incorporation of new foods?
J Anim Ecol
  .
84
:
79
89
.
Echeverría
AI
Vassallo
AI
.
2008
.
Novelty responses in a bird assemblage inhabiting an urban area
.
Ethology
  .
114
:
616
624
.
Echeverría
AI
Vassallo
AI
Isacch
JP
.
2006
.
Experimental analysis of novelty responses in a bird assemblage inhabiting a suburban marsh
.
Can J Zool
  .
84
:
974
980
.
Evans
J
Boudreau
K
Hyman
J
.
2010
.
Behavioural syndromes in urban and rural populations of song sparrows
.
Ethology
  .
116
:
588
595
.
Evans
KL
Newson
SE
Gaston
KJ
.
2009
.
Habitat influences on urban avian assemblages
.
Ibis
  .
151
:
19
39
.
Garamszegi
LZ
Erritzøe
J
Møller
AP
.
2007
.
Feeding innovations and parasitism in birds
.
Biol J Linn Soc
  .
90
:
441
455
.
Geue
D
Partecke
J
.
2008
.
Reduced parasite infestation in urban Eurasian blackbirds ( Turdus merula ): a factor favoring urbanization?
Can J Zool
  .
86
:
1419
1425
.
Goldberg
JL
Grant
JWA
Lefebvre
L
.
2001
.
Effects of the temporal predictability and spatial clumping of food on the intensity of competitive aggression in the Zenaida dove
.
Behav Ecol
  .
12
:
490
495
.
Greenberg
R
.
1983
.
The role of neophobia in determining the degree of foraging specialization in some migrant warblers
.
Am Nat
  .
122
:
444
453
.
Griffin
AS
Diquelou
M
Perea
M
.
2014
.
Innovative problem solving in birds: a key role of motor diversity
.
Anim Behav
  .
92
:
221
227
.
Griffin
AS
Guez
D
.
2014
.
Innovation and problem solving: a review of common mechanisms
.
Behav Processes
  .
109
(
Pt B
):
121
134
.
Griffin
AS
Guez
D
Lermite
F
Patience
M
.
2013
.
Tracking changing environments: innovators are fast, but not flexible learners
.
PLoS One
  .
8
:
e84907
.
Isden
J
Panayi
C
Dingle
C
Madden
J
.
2013
.
Performance in cognitive and problem-solving tasks in male spotted bowerbirds does not correlate with mating success
.
Anim Behav
  .
86
:
829
838
.
Jacquin
L
Récapet
C
Prévot-Julliard
AC
Leboucher
G
Lenouvel
P
Erin
N
Corbel
H
Frantz
A
Gasparini
J
.
2013
.
A potential role for parasites in the maintenance of color polymorphism in urban birds
.
Oecologia
  .
173
:
1089
1099
.
Knight
RL
Grout
DJ
Temple
SA
.
1987
.
Nest-defense behavior of the American crow in urban and rural areas
.
Condor
  .
89
:
175
.
Kozlovsky
DY
Branch
CL
Pravosudov
VV
.
2015
.
Elevation-related differences in parental risk-taking behavior are associated with cognitive variation in mountain chickadees
.
Ethology
  .
121
:
383
394
.
Krugel
LK
Biele
G
Mohr
PN
Li
SC
Heekeren
HR
.
2009
.
Genetic variation in dopaminergic neuromodulation influences the ability to rapidly and flexibly adapt decisions
.
Proc Natl Acad Sci USA
  .
106
:
17951
17956
.
Lissek
S
Diekamp
B
Güntürkün
O
.
2002
.
Impaired learning of a color reversal task after NMDA receptor blockade in the pigeon ( Columba livia ) associative forebrain (neostriatum caudolaterale)
.
Behav Neurosci
  .
116
:
523
529
.
Lowry
H
Lill
A
Wong
BBM
.
2011
.
Tolerance of auditory disturbance by an avian urban adapter, the noisy miner: tolerance of auditory disturbance by an avian urban adapter
.
Ethology
  .
117
:
490
497
.
Lowry
H
Lill
A
Wong
BBM
.
2013
.
Behavioural responses of wildlife to urban environments
.
Biol Rev
  .
88
:
537
549
.
Martin
LB
Han
P
Lewittes
J
Kuhlman
JR
Klasing
KC
Wikelski
M
.
2006
.
Phytohemagglutinin-induced skin swelling in birds: histological support for a classic immunoecological technique
.
Funct Ecol
  .
20
:
290
299
.
McCabe
CM
Reader
SM
Nunn
CL
.
2015
.
Infectious disease, behavioural flexibility and the evolution of culture in primates
.
Proc Biol Sci
  .
282
:
20140862
.
McKinney
ML
Lockwood
JL
.
1999
.
Biotic homogenization: a few winners replacing many losers in the next mass extinction
.
Trends Ecol Evol
  .
14
:
450
453
.
Miranda
AC
Schielzeth
H
Sonntag
T
Partecke
J
.
2013
.
Urbanization and its effects on personality traits: a result of microevolution or phenotypic plasticity?
Glob Chang Biol
  .
19
:
2634
2644
.
Møller
AP
.
2009
.
Successful city dwellers: a comparative study of the ecological characteristics of urban birds in the Western Palearctic
.
Oecologia
  .
159
:
849
858
.
Møller
AP
Tryjanowski
P
.
2014
.
Direction of approach by predators and flight initiation distance of urban and rural populations of birds
.
Behav Ecol
  .
25
:
960
966
.
Monceau
K
Wattier
R
Dechaume-Moncharmont
F-X
Motreuil
S
Cezilly
F
.
2011
.
Territoriality versus flocking in the Zenaida dove ( Zenaida aurita ): resource polymorphism revisited using morphological and genetic analyses
.
Auk
  .
128
:
15
25
.
Mueller
JC
Partecke
J
Hatchwell
BJ
Gaston
KJ
Evans
KL
.
2013
.
Candidate gene polymorphisms for behavioural adaptations during urbanization in blackbirds
.
Mol Ecol
  .
22
:
3629
3637
.
Nakagawa
S
Schielzeth
H
.
2010
.
Repeatability for Gaussian and non-Gaussian data: a practical guide for biologists
.
Biol Rev Camb Philos Soc
  .
85
:
935
956
.
Papp
S
Vincze
E
Preiszner
B
Liker
A
Bókony
V
.
2014
.
A comparison of problem-solving success between urban and rural house sparrows
.
Behav Ecol Sociobiol
  .
69
:
471
480
.
Partecke
J
Schwabl
I
Gwinner
E
.
2006
.
Stress and the city: urbanization and its effects on the stress physiology in European blackbirds
.
Ecology
  .
87
:
1945
1952
.
Pravosudov
VV
Roth
TC
2nd
Forister
ML
LaDage
LD
Kramer
R
Schilkey
F
van der Linden
AM
.
2013
.
Differential hippocampal gene expression is associated with climate-related natural variation in memory and the hippocampus in food-caching chickadees
.
Mol Ecol
  .
22
:
397
408
.
Reader
SM
Nover
D
Lefebvre
L
.
2002
.
Locale-specific sugar packet opening by Lesser Antillean bullfinches in Barbados
.
J Field Ornithol
  .
73
:
82
85
.
Roth
TC
2nd
Gallagher
CM
LaDage
LD
Pravosudov
VV
.
2012
.
Variation in brain regions associated with fear and learning in contrasting climates
.
Brain Behav Evol
  .
79
:
181
190
.
Roth
TC
LaDage
LD
Pravosudov
VV
.
2010
.
Learning capabilities enhanced in harsh environments: a common garden approach
.
Proc Biol Sci
  .
277
:
3187
3193
.
Schoech
SJ
Bowman
R
Reynolds
SJ
.
2004
.
Food supplementation and possible mechanisms underlying early breeding in the Florida scrub-jay ( Aphelocoma coerulescens )
.
Horm Behav
  .
46
:
565
573
.
Sih
A
Del Giudice
M
.
2012
.
Linking behavioural syndromes and cognition: a behavioural ecology perspective
.
Philos Trans R Soc Lond B Biol Sci
  .
367
:
2762
2772
.
Sol
D
Duncan
RP
Blackburn
TM
Cassey
P
Lefebvre
L
.
2005
.
Big brains, enhanced cognition, and response of birds to novel environments
.
Proc Natl Acad Sci USA
  .
102
:
5460
5465
.
Sol
D
Elie
M
Marcoux
M
Chrostovsky
E
Porcher
C
Lefebvre
L
.
2005
.
Ecological mechanisms of a resource polymorphism in Zenaida doves of Barbados
.
Ecology
  .
86
:
2397
2407
.
Sol
D
González-Lagos
C
Moreira
D
Maspons
J
Lapiedra
O
.
2014
.
Urbanisation tolerance and the loss of avian diversity
.
Ecol Lett
  .
17
:
942
950
.
Sol
D
Griffin
AS
Bartomeus
I
Boyce
H
.
2011
.
Exploring or avoiding novel food resources? The novelty conflict in an invasive bird
.
PLoS One
  .
6
:
e19535
.
Soler
JJ
Peralta-Sánchez
JM
Martín-Vivaldi
M
Martín-Platero
AM
Flensted-Jensen
E
Møller
AP
.
2012
.
Cognitive skills and bacterial load: comparative evidence of costs of cognitive proficiency in birds
.
Naturwissenschaften
  .
99
:
111
122
.
Svensson
E
Ricklefs
RE
.
2009
.
Low diversity and high intra-island variation in prevalence of avian Haemoproteus parasites on Barbados, Lesser Antilles
.
Parasitology
  .
136
:
1121
.
Tebbich
S
Stankewitz
S
Teschke
I
.
2012
.
The relationship between foraging, learning abilities and neophobia in two species of Darwin’s finches
.
Ethology
  .
118
:
135
146
.
Tebbich
S
Sterelny
K
Teschke
I
.
2010
.
The tale of the finch: adaptive radiation and behavioural flexibility
.
Philos Trans R Soc Lond B Biol Sci
  .
365
:
1099
1109
.
Tebbich
S
Teschke
I
.
2014
.
Coping with uncertainty: woodpecker finches ( Cactospiza pallida ) from an unpredictable habitat are more flexible than birds from a stable habitat
.
PLoS One
  .
9
:
e91718
.
Teschke
I
Cartmill
EA
Stankewitz
S
Tebbich
S
.
2011
.
Sometimes tool use is not the key: no evidence for cognitive adaptive specializations in tool-using woodpecker finches
.
Anim Behav
  .
82
:
945
956
.
Valcarcel
A
Fernández-Juricic
E
.
2009
.
Antipredator strategies of house finches: are urban habitats safe spots from predators even when humans are around?
Behav Ecol Sociobiol
  .
63
:
673
685
.
Vas
Z
Lefebvre
L
Johnson
KP
Reiczigel
J
Rózsa
L
.
2011
.
Clever birds are lousy: co-variation between avian innovation and the taxonomic richness of their Amblyceran lice
.
Int J Parasitol
  .
41
:
1295
1300
.
Webster SJ, Lefebvre L. 2000. Neophobia by the Lesser-Antillean bullfinch, a foraging generalist, and the Bananaquit, a nectar specialist. Wilson Bull. 112:424–427.

Author notes

Handling editor: Louise Barrett