Fear of humans and urban invasiveness

Literature on introduced species suggests that behavioural flexibility, in the form of learning, cognition and/or rapid adjustment to new conditions, allows animals to be successful when invading novel habitats [27]–[28]. However, unlike typical biological invasions, urban areas present birds with all of the novel conditions characteristic of new environments (e.g., food resources, competitors, or breeding sites) as well as an extraordinary selective factor, i.e. humans. In this sense, our study indicates that species that are variable regarding their fear of humans and/or vehicles, which are also those with relatively larger brains, are more likely to invade urban habitats than apparently tame species. Abundant, small and generalist species also tended to show higher urban invasiveness. However, when simultaneously tested, only the effect of variability in FID remains significant, suggesting that the ability to cope with this extreme anthropogenic habitat change can be largely related to inter-individual variability in a specific behavioural trait such as fearfulness [16]. Within-species variability in fear of people thus emerges as a proximate behaviour that could explain urban invasion by a small subsample of tame individuals, while RBS could be the ultimate responsible behind the variability in this behaviour among individuals, as suggested by our confirmatory path analysis. Our results obtained from a contemporary scenario of invasion strongly support the suggestion by Møller [8] of a selection of individuals with reduced FID in urban environments.

While admitting that other unexplored behavioural traits related to brain size or FID could also contribute to explaining urban invasiveness, our results suggest that only tame individuals from variable species would cross the disturbance frontier and thus be able to live in urban environments, hence supporting the disturbance-induced habitat selection hypothesis [15]. One could ask why individuals from apparently tame species do not become urban invaders. The likely answer would be that individual fearfulness (as measured through FID) must be interpreted regarding its within-species variability. Both risk perception and the costs of fleeing from people likely vary among species, and thus just the FID of an individual tells us little about its tolerance of people if it is not compared with the variability shown by its conspecifics. In this sense, variable species seem to include some individuals which perceive human proximity as less risky than their conspecifics, being thus able to coexist with people. However, in species with low variability all individuals would be similarly affected by human disturbance, creating few opportunities for some individuals to invade urban areas.

Looking for alternative explanations to the above hypothesis, the low FIDs of urban individuals compared to their rural conspecifics could partially result from individuals habituating to human disturbance after they settled in cities, thus increasing their tameness with time. Cooke [26] found 30 years ago that birds tend to be more approachable in urban than in rural habitats, suggesting that birds in urban areas come into contact with people more often and with greater proximity, having a greater opportunity to learn within what distance a human can approach before being a danger. Since then, habituation has been often argued to explain differences in FID among areas with different human disturbance [29]. Those studies, however, were based on population means instead of on individual responses to an increased human presence. The only study so far dealing with changes in individual responses did not find evidence for a consistent short-term individual habituation to human disturbance in rural birds, using the same burrowing owl population included in this study [15]. Habituation was neither supported when relating the differences in FID between urban and rural conspecifics to time as species become urban in Old-World cities [7]. Nonetheless, further research on habituation is needed. Despite the strong within-individual consistency in FID found, some burrowing owls slightly habituated to human disturbance while others became more afraid of people [15], and we cannot discard the possibility that the relationship between individual consistency in FID and habituation could change among species.

Brain size, individual behaviour and behavioural flexibility

The brain size-environmental change hypothesis predicts that behavioural flexibility carries fitness benefits to individuals facing novel or altered environmental conditions [17]. The principle underlying this hypothesis is the idea that enlarged brains afford advantages to individuals in dealing with environmental change when the response demands behavioural flexibility in the form of learning and innovation. Relative brain size correlates with measures of behavioural flexibility, linked to, for example, innovation capacities and the ability to deal with new environments, which could satisfactorily explain the success of several species as alien invaders [17]–[18], [28]. From these studies, one could assume that individuals from large brain species are equally flexible. However, recent research shows that species often exhibit inter-individual differences in their responses to a variety of situations such as feeding, mating, or avoiding/escaping predators. These behavioural tendencies, personality traits or behavioural syndromes (such as activity, shyness-boldness, exploration, and aggressiveness [30]–[31]) can greatly determine how these species respond to changes such as those produced by human development [16], [31]–[32].

A growing number of studies show that the majority of a population's niche width is determined by inter-individual variation [33]–[35]. Along this same line of evidence, individuals within populations show different behaviours that are heritable [36]–[37], relatively inflexible [15], [30], and linked to fitness traits, thus being favoured or disfavoured by selection depending on the particular ecological conditions experienced by the population [30], [38]. Within this context, and taking into account results presented here, behavioural flexibility should be regarded as a specific trait encompassing variability among individuals, but not necessarily within individuals. Recent studies on individual variations in FID have shown that this behaviour has a strong individual component in two bird species [15]–[16] and a reptile, Agama planiceps [39]. Moreover, these authors also found links between FID and other individual behaviours [16], [39], suggesting that FID can be considered as a personality trait sensu Réale et al. [30]. Therefore, although much more research is needed, we must not discard that within-individual consistency in FID could be a generalised fact in animal behaviour. No studies have been performed however to test whether FID changes across situations or is the result of early experience that then is fixed throughout life, but studies on personality envisage these possibilities for individual behaviours [31]. However, as FIDs resulted highly repeatable within individuals [15]–[16], there would be a low potential cognitive effect of larger brains on the flexibility of adult individuals (i.e., those studied here) regarding their fear of humans. Although the actual mechanism underlying the positive relationship between RBS and variability in FID found in this study remains unexplained, our results suggest an association between enlarged brains and the evolution of behavioural variability among individuals, not just the capacities of individuals to modify their behaviour in potentially adaptive ways but through increased differences in individual traits. Thus, one prominent contribution of our results is that the behaviours of individuals, but not the average behaviour at the level of species, are important during the invasion process. As previously suggested [30], further research on biological invasion processes would benefit from the perspective of variability in animal personalities.

Ethics Statement

Field work conducted here was not invasive and did not require the manipulation of live animals, and measuring FID from a car did not suppose an additional disturbance to that coming from daily car traffic. Brain masses were obtained from the literature and from birds found recently killed by cars in the roads. Therefore, this work did not require specific permits by the relevant Argentinean nor Spanish authorities.

Study area

We selected the area of Bahía Blanca as a study site, on the Atlantic coast of Argentina, a relatively young city founded by European colonists in 1828. It was a small village until the middle of the twentieth century, reaching ca. 293,000 inhabitants in very recent years. The city is surrounded by natural habitats (mostly grasslands and pasturelands, with small interspersed patches of xerophytic forests and scrublands) where human presence and activities are negligible. Both pedestrian records (0–0.1 pedestrians/h) and traffic volume (0.34–2.4 cars/h) were extremely low in natural (hereafter rural) habitats when compared to typical figures for First World countries (11–325 cars/h; [1], [40]). Sampling in rural habitats was restricted to areas 20–150 km from the city to avoid potential confounding effects of urban-rural ecotones on both behaviour and relative abundance of birds (see below). Field work was conducted during wintering and summering months of 2003–2008.

Urban invasiveness and explanatory variables

We classified species within a gradient of urban invasiveness based on their relative abundance (measured through censuses following road transects; see e.g. [41]) in urban and rural habitats. Road transects were shown to perform as well as other measurement techniques such as foot transects or point counts in estimating relative abundances of a variety of bird species in similar Argentinean open habitats [42]. In summer 2004, transects totalling 59 and 150 km were conducted in urban and rural habitats, respectively, at a constant speed (ca. 20 km/h), avoiding windy and rainy days and the hottest hours of midday. Near-road abundances (hereafter, abundances) were estimated as the number of individuals of each species recorded per km [41]–[42]. As is the case for any census methodology, differences in abundances between species may be biased by their differential detectability. In our case, smaller species could be more frequently missed when conducting road transects. This is not a problem, however, when comparing within-species abundances in rural and urban areas, since any body size bias should be common to both areas. After calculating the difference in abundance in rural minus the abundance in urban areas for each species (i.e., within-species relative abundance), urban invasiveness was scored as null or occasional (species recorded in rural habitats that were never or very rarely seen at urban sites), medium (species commonly recorded at urban sites but with higher abundances in rural habitats), and high (species showing higher abundances in urban compared to rural habitats). We chose these rough scores because urban invasion is a contemporary process in our study area: several species have become urban within the last 5–15 yr, while urban populations of others continue to increase (Authors unpubl. observations). Therefore, relative abundances may not have been established in some species and thus the use of finer scoring scales could imply their incorrect categorization. Nonetheless, the use of five scores for the categorization of urban invasiveness gave similar results to those presented in this paper. For the same reason (contemporary, dynamic process), we did not differentiate between potential urban exploiters and urban adapters (i.e., species living in the city proper and species living in the surrounding, less urbanized areas, respectively) as defined by Kark et al. [43]. Nonetheless, both types of species do not seem to differ in terms of behavioural flexibility as measured through relative brain size and the number of feeding innovations [43].

We recorded FIDs of different birds to car approach as a measure of the ability of individuals to cope with human disturbance. We are confident that we mostly sampled different individuals since 1) we surveyed ca. 750 km of different unpaved roads and streets across a large study area (ca. 5,000 km²), and 2) road surveys covered a number of territories of territorial species, such as the burrowing owl, that we identified in the course of other field-work tasks [15]. Therefore, the likelihood of resampling individuals for FID was negligible. To measure FID, we drove a small grey car at a slow, steady speed (ca. 10 km/h). Birds measured were typically perched close to (usually within 15 m) unpaved roads in rural habitats or streets at urban sites. When we selected a focal bird we did not stop for the identification of the species but approached it driving through the route at the same speed until it flew. Therefore, the approach was nearly directional (i.e., following the straight road or street in direction to the bird) and was done in the same way that usual car traffic unintentionally approaches birds. If the focal bird was close to others, we only measured FID from the focal one. Then, one of the two authors seated in the front of the car identified the species, using binoculars if needed, and measured the FID of the focal bird through the closest open window. Therefore, we cannot differentiate bird responses to the car from those to humans that were clearly visible to birds. Nonetheless, FIDs of one of the species obtained by an approaching human [15] give similar results to those when measured from a car (Authors, unpublished data). FIDs were obtained only from adult birds, discarding fledglings and juveniles (whose flight skills could be compromised by their young ages). Each FID was measured by using a LEICA laser distancemeter (measuring range: 10–800 m; accuracy: ±1 m) or through direct measurement (for distances<10 m), thus obtaining the actual Euclidean distance [9]. We recorded 1,393 FIDs from 61 species in rural habitats and 691 FIDs from 41 species in urban habitats (Figure S1), individual FIDs ranging from 0.5 to 122 m. However, to avoid potential biased results associated with small sample sizes, statistical analyses (see below) were performed following Cooke [26] and Blumstein [9] by using information from species for which we obtained FIDs from at least 10 individuals measured in rural habitats (n = 42 species, see Figure S1). As a measure of within-species variability in FID in rural habitats, we used the coefficient of variation (CV_FID) instead of the variance because, contrary to this measure, CV was not related to body size (Regression of variance against body size: R² = 0.25, n = 42, P = 0.0015; Regression of CV_FID against body size: R² = 0.003, n = 42, P = 0.93). Thus, we can test the relative importance of CV_FID and body mass including them simultaneously in the same models.

Information on overall brain masses (in grams) was available for 27 species (44, D. Sol, unpubl. data, Authors unpubl. data). As is the case in most comparative works on brain size, the number of individuals from which brain size and body mass is available per species was generally low (see Figure S1). However, variance in both brain size and body mass is much higher among than within bird species, which is required for the feasibility of a comparative analysis [44]. Body masses were obtained from the same sources as available brain masses, completed with information from published body mass compilations [45]–[47] and own unpublished data for the rest of species. Following Sol et al. [17], we calculated the residuals of a log-log linear regression of brain mass against body mass. The relationship was positive and linear (linear regression, R² = 0.75; F_1,25 = 75.08, P<0.0001) and the residuals were uncorrelated to body mass (r = 0.003, P = 0.99, n = 27), hence, we used them as a measure of relative brain size [17].

The abundance of each species in rural areas was estimated as the average number of birds/km (see above). As previously mentioned, larger species could be more easily detected from a car than smaller ones. Contrary to when we previously estimated within-species relative densities in urban and rural areas, the between-species relative densities obtained in rural areas could be seriously affected by such a bias. However, abundances in our study species in rural habitats were uncorrelated to our measure of body size (i.e., body mass; r_s = 0.014, P = 0.93, n = 42), despite the fact that sampled species widely differed in size (body masses ranging from 13 to 890 g, see Figure S1).

Dietary and habitat generalisms were obtained as the number of the major food/habitat types recorded in the literature for each species (diet: grasses and herbs, seeds and grains, fruits and berries, pollen and nectar, vegetative material, invertebrates, vertebrates, and carrion; habitat: forest, mixed scrub, grassland, marsh and wetland, and cultivated and farm lands, [17], [48]). Environmental tolerance was calculated as the whole latitudinal distribution of each species (in degrees) including their breeding and non-breeding ranges [25].

Statistical analyses

Generalized Linear Mixed Models (GLMMs, using the cumulative logit link function and the multinomial error distribution for categorical, ordered data; GLIMMIX procedure in SAS 9.1) were performed to investigate the relative influence of our explanatory variables on urban invasiveness (as a categorical, ordered variable with three levels; see above) while controlling for potential phylogenetic effects (Family and Order as nested random factors, following [17]). The same random factors were fitted when testing in a GLMM the relationship between CV_FID and brain size of the species, using a normal distribution of errors and the identity link function. As FIDs did not differ between seasons and years within species and habitats (all P>0.36), data were pooled for analyses. CV_FID and average FID were not significantly related (r = 0.13, P = 0.40, n = 42), so we included both variables in the same models without problems of colinearity. We compared alternative models (Fig. 5) to ascertain the links between relative brain size, CV_FID and urban invasiveness through Confirmatory Path Analysis, using Structural Equation Modelling (SEM) in AMOS 5. The Akaike's information criterion (AIC) was used for model selection, using differences in AIC scores (lower scores indicated greater statistical support). Models with AIC scores differing from that of the lowest score by more than two were considered to be unsupported statistically [49]. Finally, a Paired t-test was used to compare FIDs of urban and rural conspecifics for the 20 species from which we sampled at least 10 individuals in both habitats.