Wild birds recognize individual humans: experiments on magpies, Pica pica
- First Online:
- Cite this article as:
- Lee, W.Y., Lee, S., Choe, J.C. et al. Anim Cogn (2011) 14: 817. doi:10.1007/s10071-011-0415-4
- 603 Views
The ability to distinguish among heterospecific individuals has been reported in only a few animal species. Humans can be viewed as a special type of heterospecifics because individuals differ widely in behavior, ranging from non-threatening to very threatening toward animals. In this study, we asked whether wild magpies can recognize individual humans who had accessed their nests. We compared the behavior of breeding pairs toward individual humans before and after the humans climbed up to the birds’ nests, and also toward climbers and non-climbers. We have evidence for (i) aggressive responses of the magpie pairs toward humans who had repeatedly accessed their nests (climbers) and a lack of response to humans who had not accessed the nest (non-climbers); (ii) a total lack of scolding responses toward climbers by magpie pairs whose nests had not been accessed; (iii) a selective aggressive response to the climber when a climber and a non-climber were presented simultaneously. Taken together, these results suggest that wild magpies can distinguish individual humans that pose a threat to their nests from humans that have not behaved in a threatening way. The magpie is only the third avian species, along with crows and mockingbirds, in which recognition of individual humans has been documented in the wild. Here, we propose a new hypothesis (adopted from psychology) that frequent previous exposure to humans in urban habitats contributes to the ability of birds to discriminate among human individuals. This mechanism, along with high cognitive abilities, may predispose some species to learn to discriminate among human individuals. Experimental tests of these two mechanisms are proposed.
KeywordsCognitionPredationIndividual recognitionPre-exposure effectMagpie
Many animals discriminate among conspecific individuals as well as between conspecifics and heterospecifics (see Tibbets and Dale 2007; Kazial et al. 2008; McLean and Rhodes 1991; Templeton and Greene 2007), but only a few studies have asked whether animals can discriminate among heterospecific individuals (Slobodchikoff et al. 1991; Levey et al. 2009; Marzluff et al. 2010), and in what circumstances this is likely to happen (Ferrari et al. 2008). The ability to recognize individuals of other species may be beneficial for prey when the degree of predation threat varies among individuals of predatory species (Slobodchikoff et al. 1991; Levey et al. 2009), when the prey can view the predator long before an attack is possible (Walters 1990), and when the predatory species does not pose an extreme threat (Ferrari et al. 2008). Considering that some humans may pose a threat to wild animals in anthropogenic habitats, we can predict that animals in these landscapes may develop the ability to distinguish human individuals that pose a threat from individuals perceived as non-threatening.
Recognition of individual humans or their images has been studied in several domesticated and captive animals. Honey bees, pigeons, sheep, and horses are not only able to recognize human individuals (Davis and Balfour 1992; Davis 2002; Dittrich et al. 2010), but they can also recognize individual human faces in photos or still images (Herrnstein and Loveland 1964; Kendrick et al. 2001; Peirce et al. 2001; Dyer et al. 2005; Adachi et al. 2007; Stone 2010). However, only three studies have experimentally tested the ability to recognize human individuals in natural populations (Slobodchikoff et al. 1991; Levey et al. 2009; Marzluff et al. 2010). Two of these studies involve birds. Northern mockingbirds (Mimus polyglottos) on a university campus seemed to recognize human individuals that had touched their nests, and therefore, that were probably perceived as posing a threat (Levey et al. 2009). American crows (Corvus brachyrhynchos), members of a family (Corvidae) known for the high cognitive abilities (Emery 2006), also learned to recognize artificial facial masks that were worn by researchers during trapping on a university campus and its surroundings (Marzluff et al. 2010). The researchers proposed that these two species of birds may be good at rapidly learning to discriminate among individual humans because they have especially high general “cognitive” abilities (Marzluff et al. 2010) or “perceptive ability and rapid learning” (Levey et al. 2009), which makes them well pre-adapted for urban habitats, as well as to interactions with humans and predators associated with human settlements.
In this study, we examined whether another species of Corvidae, the black-billed magpie (Pica pica), can distinguish human individuals who pose a threat to their nests from humans perceived as non-threatening. Wild magpies react aggressively to observers that climb to their nests (Buitron 1983), and increase nest defense behaviors in response to repeated human approaches (Redondo and Carranza 1989). However, previous studies did not show whether they respond selectively to nest climbers or increased their aggressive behavior with repeated access to the nests by the climbers. The main aim of this study was to test whether magpies can distinguish individual humans that threatened the birds by climbing to their nests, from other humans that should be perceived as non-threatening. An additional aim was to further explore hypothetical explanations of these abilities in birds by incorporating a pre-exposure effect in learning.
Materials and methods
We studied a black-billed magpie (Pica pica sericea) population on the campus of Seoul National University (SNU, Seoul, Korea). Magpies on the SNU campus are used to the presence of many people: on SNU campus (surface area of 1.4 km²), there were 28,669 students and faculty registered in October 2008; resulting in an estimated human density of 20,478 people/km². The core observations and experiments were conducted during the breeding season (April–July) in 2009. Additional data from the 2008 breeding season were also included in the analyses (see “Comparison of responses to non-climbers and climbers”).
Responses to the climber
As a part of an annual breeding survey of over 50 pairs, we regularly observed the behavior of breeding pairs from a distance that does not cause disturbance to the birds, moving along the standard pedestrian walkways. Starting from the time of egg laying, the nests were visited once a week. One of five members of the 2009 survey team, WYL: “the climber” repeatedly accessed all of the nests except one (which was accessed by PGJ) using a cargo crane during the breeding season from late March to mid-June in 2009. Each climbing visit took approximately 30 min. The two climbers (WYL and PGJ) wore different clothes during each climbing visit. Three to four days after climbing, each climber re-visited the nest area that he had accessed earlier (moving to within 15 m of the nest once after each climbing event). On the designated day, the climber first made sure that the breeding pair at the accessed nest did not respond to at least five random passersby that approached the magpies to within 6 m. Then, the climber walked toward the magpies on a pedestrian road and recorded the response of the birds. Eleven breeding pairs (N = 11) were accessed and tested in this way four to nine times each (average per nest is seven tests; the interval between consecutive tests on a pair averaged 10.8 days from March to June 2009). The climber wore different clothing for each test. The presence or absence of the aggressive response toward the person (defined as “scolding and following the human”) was recorded. These responses are unmistakably different from other behaviors and therefore, the fact that the climber was not blind with respect to the treatment should not bias the resulting variable. To examine the effect of the number of climbing visits prior to the test on the probability of an aggressive response during the test, we used generalized linear mixed models (PROC GENMOD in SAS 9.1), with the number of climbing events at a nest prior to the test as an explanatory variable and the presence or absence of an aggressive response as the response variable. We used the logit link function and assumed a binomial error distribution. Since the response of each breeding pair was measured several times, we treated each breeding pair as a block in the analysis. PROC GENMOD deals with correlated data from repeated sampling using generalized estimating equations (GEEs; SAS User’s guide), which was also used in generating predicted probabilities of aggressive responses.
To compare responses between birds at the nests that were accessed and those that were not accessed, we performed the following observations (each magpie pair was tested within the immediate vicinity to the pair’s nest after verification of the identity of the pair). Between March 3 and April 18 2009, we recorded the reactions of birds to two approaches toward the vicinity of the nest and compared the reactions of birds at nests that had been previously accessed by the climber with the reactions of birds whose nests had not been previously accessed by the climber. The former group consisted of responses by breeding pairs tested after two and three climbing events involving their nests (once after each climbing event). The latter group consisted of responses by the pairs at nests that had never been accessed. These nests were chosen randomly from among all of the active nests that had not been accessed. We did not notice any clear difference between the two groups with respect to their distribution on the campus, proximity to buildings, or to paths on which humans usually walk. We set the score for a pair to value “1” if at least one of the two tests triggered an aggressive response and to “0” if no scolding response was observed during the two tests. Fisher’s exact test (Zar 1999) was used to test whether there was a difference between pairs at accessed and not-accessed nests in the proportion of aggressive responses.
Comparison of responses to non-climbers and climbers
Finally, to complement the results from the 2009 season, we asked whether there was a difference in magpie responses toward an approaching climber in 2009 (N = 11) and toward non-climbers (N = 14 tests) in the previous breeding season (2008). The data came from different years, different experimental designs, and theoretically we cannot exclude the possibility that any statistical difference between the responses to climbers and non-climbers is due to the effect of year, or the number of approaching experimenters (two non-climbers in 2008 vs. one climber in 2009). However, the tests were performed at the same breeding stage in the same population, and we do not have any indications that the 2 years differed drastically in any ways that should affect the responses of the birds to humans. Hence it is rather unlikely that factors unrelated to the characteristics of the approaching persons affected the outcome. The 2008 data came from a separate study on the responses of magpies to the people who gazed at the birds while approaching them—a situation that resembles both an approach by a natural predator and the approach of the climber in 2009. The breeding survey in 2008 was conducted with the same methods as in 2009. The identity of the climber was not fixed in 2008: various team members climbed to the nests and the responses of magpies to approaching climbers were not tested. The experiment on human gaze in 2008 was conducted by people who never took part in nest climbing. Two people (“non-climbers”) walked toward a magpie foraging on the ground in the vicinity of the sidewalk near the nest. The response of the bird was recorded by a third person who was at least 10 m away. The territories that were used in the experiments were different in 2008 and in 2009. Considering the total study population of about 100 birds, the probability that the birds tested in the 14 tests in 2008 were tested again in the 11 tests in 2009 is relatively small. Therefore, we assumed independence and used Fisher’s exact test for 2 × 3 table (Freeman and Halton 1951; Zar 1999) to examine whether there is a difference in magpie responses toward the approaching non-climbers (N = 14 tests in 2008) and climbers (N = 11 tests in 2009).
Responses to the climber
The territorial magpies showed no visible aggressive responses to the climber or a random passerby before the nest was first accessed by the climber. The breeding pairs aggressively scolded the climber when he was handling the eggs and nestlings during visits to the nests. The birds sometimes attempted to physically attack the climber by pecking his head. In all 11 nests analyzed here (i.e., those accessed in 2009), the nest owners were occasionally joined by 1–2 neighbors, but never by any of the remaining ten pairs used for the analysis. The probability of aggressive responses to the walking (approaching) climber increased with the number of the preceding climbing events at a nest (Fig. 1a; χ12 = 9.15, P = 0.0025), which also corresponds to an increased response when the chicks were in the later stages of development. This contrasts with the total lack of aggressive reaction by the same pairs throughout the whole 2009 breeding season toward random passersby (recorded prior to each observation of responses to the climber).
No aggressive responses toward the climber were detected throughout the 2009 season (4–6 visits per nest) at the nests that were never accessed. Consequently, the pairs whose nests were accessed were more likely to show aggressive responses (at least once during two visits; see Methods) than those whose nests were not accessed (Fisher’s exact test; df = 1, P = 0.0006).
In the experiment where the climber and the non-climber were simultaneously presented, the birds clearly showed aggressive behavior selectively to the climber (Fig. 2b, c; one-sample median test; Z5 = 2.201, P = 0.03; H0: median = 0).
Comparison of responses to non-climbers and climbers
No aggressive reactions were recorded to the approaches of non-climbers. Most often the magpies responded by flying away (Fig. 2d). Hence the distribution of responses to non-climbers was different (Fisher’s exact test; df = 2, P < 0.001) from that to the climbers (Fig. 2d). See “Methods” for more explanations about the two data sets used in this comparison.
How do the birds recognize individual humans?
Our results suggest that wild black-billed magpies, like mockingbirds (Levey et al. 2009) and crows (Marzluff et al. 2010), are able to distinguish an individual human who poses a threat from other humans in an area of high human density. Considering that even arthropods (Hemmi and Merkle 2009) habituate to the continuous non-threatening presence of predators, and magpies are known to habituate to non-threatening models of owls (Buitron 1983), it is not surprising that magpies, as well as mockingbirds and crows (Levey et al. 2009, Marzluff et al. 2010), did not scold numerous humans walking in a standard manner along sidewalks and paths. However, among the individual humans approaching the magpies in a manner that might have attracted the birds’ attention (watching the birds), only those humans who had earlier accessed the birds’ nests were scolded aggressively. This result is similar to previous reports of scolding by mockingbirds and crows of only those humans who had posed a threat to the birds earlier (within several days or weeks in Levey et al. 2009 and Merrit 1983; or within several years in Marzluff et al. 2010).
The rate of scolding by magpies increased after multiple climbing events, which parallels an observed increase in mockingbirds’ reactions to repeated approaches by an observer to their nests and may be caused by a perception of the returning intruder as a greater threat (Levey et al. 2009). These increases in the defensive reactions in mockingbirds and magpies are also consistent with previously reported associations between avian nest defense and the stage of the breeding cycle (Andersson et al. 1980; Greig-Smith 1980; Buitron 1983; Knight and Temple 1986; Redondo and Carranza 1989; Brunton 1990). Although we cannot determine whether magpies learned to properly distinguish the climber from other humans right after the first access to the nest or whether the birds gradually acquired this ability, the discrimination experiment clearly showed that the aggression was directed only at the climber. In American crows, neighboring individuals seemed to learn to recognize dangerous facial masks using information from crowds of birds that mobbed humans with masks (Marzluff et al. 2010). In our study, none of the tested pairs were close neighbors, and the pairs are extremely territorial during the breeding season. Therefore, the possibility that tested individuals learned to identify the climber by visiting their neighbors during climbing events or during reactions of birds to the climber is unlikely.
What kind of physical cues might have been used by the magpies for discrimination between human individuals? Because the humans who took part in the experiments wore random sets of clothing (except during the two experimental trials with identical clothing in the discrimination experiment), magpies’ discrimination between human individuals was not based on clothing. The people who participated in our experiments were all Koreans (except for one participant in one set of tests) with short black hair and similar hair styles, which makes it unlikely that features of hair style could be the cue for individual recognition by magpies. Because, like Levey et al. (2009), we did not use artificial face masks (unlike Marzluff et al. 2010), we are unable to directly disentangle the effect of facial features from other cues such as body posture. However, we chose pairs of humans for this experiment that generally had quite similar body heights and postures but different facial features (Fig. 1). Therefore, we think that non-facial features were not drastically different between the climbers and the non-climbers. We hypothesize that, like crows (Marzluff et al. 2010), the magpies use facial features to distinguish climbers from non-climbers. This hypothesis is consistent with a report by Buitron (1983) who, after accessing magpie nests with an exposed face, was scolded by magpies only when she did not wear a bandana across her face (mouth and nose). Paying attention to frontal or facial areas, and especially the eyes, is well documented in discrimination among conspecific individuals (Guhl and Ortman 1953; Whitfield 1986), and in recognition of heterospecifics Scaife (1976a, b), including detection and recognition of predators by prey (Curio 1975; Coss et al. 2005).
What mechanisms lead to the ability of wild birds to recognize individual humans?
Here, we discuss two hypotheses: the “higher cognitive abilities” hypothesis and the “pre-exposure to stimuli” hypothesis, which are complementary and non-mutually exclusive. Together, they can explain why some wild birds in urban habitats are able to recognize individual humans.
According to the post hoc hypotheses proposed by Levey et al. (2009) and Marzluff et al. (2010), crows and mockingbirds (and magpies if we include our results) may have learned to discriminate among individual humans because they may have especially high cognitive abilities not present in some other birds (such abilities are undoubtedly present in corvids, Emery 2006). Furthermore, the two studies added a suggestion that such abilities predispose (“pre-adapt”, Levey et al. 2009) these species to succeed in novel urban environments, implying that they are pre-existing in relation to the adaptations of these birds to life in proximity to humans. This “higher cognitive abilities” hypothesis remains untested, as no experimental studies on wild birds with varying cognitive abilities are available from a number of species found in both remote natural habitats and urban habitats. However, the demonstration of abilities to recognize individual humans in a range of animal species with clearly different cognitive abilities (honey bees, chickens, pigeons, sheep, dogs, llamas, penguins, seals, rabbits, horses, lizards, and octopuses; Davis and Balfour 1992; Davis 2002; Dittrich et al. 2010) in captivity, where they are exposed to the presence of humans probably as often as are birds on university campuses, suggest that the three species may not be unique with respect to their potential for recognition of individual humans.
Predicted abilities to recognize individual humans in species with different cognitive/perceptual abilities in environments with prominent exposure to human presence (e.g., urban habitats) versus remote habitats with little or no human presence
Major effects on development of abilities to recognize human individuals
Predicted abilities to recognize individual humans
High cognitive abilities
Low cognitive abilities
High cognitive abilities
Low cognitive abilities
High cognitive abilities hypothesis
Effect of general high cognitive abilities (regardless of any pre-exposure effect; species specific cognitive properties)
Pre-exposure to stimuli hypothesis
Effect of pre-exposure that does not depend on cognitive abilities
Interaction between the two mechanisms
Interaction between “Cognition” and “Pre-exposure”: stronger effect of pre-exposure in species with higher cognitive abilities
Very good recognition
Intermediate recognition abilities
Intermediate recognition abilities
In summary, the magpie is only a third avian species in which recognition of individual humans have been experimentally determined in a wild population. In order to explain these abilities in urban birds, Levey et al. (2009) and Marzluff et al. (2010) proposed a post hoc hypothesis based on the possible association between the birds’ ability to recognize human individuals in urban habitats (e.g., human faces) and the bird’s pre-existing high cognitive and perceptual abilities. We think that the effect of the subjects’ repeated pre-exposure to stimuli on the subsequent increase in abilities to discriminate between the stimuli (e.g., different faces) offers an additional explanation. None of the three studies (Levey et al. 2009; Marzluff et al. 2010; and our study) provide appropriate data to determine how these two hypothetical mechanisms, which are complementary and non-mutually exclusive, contribute to the observed discriminatory abilities of the three urban species. Future studies should involve species of clearly different cognitive abilities tested in a standard manner in two types of habitats: heavily human-populated urban areas and wild natural habitats where exposure to human presence is minimal. We provide a framework of predictions for such a research program (Table 1).
We are grateful to Changku Kang, Choongwon Jeong, Changsoo Yang, and Woncheol Song for serving as “non-climbers” and Kyungseon Seo, Jihyun Oh for participating in the work. We thank magpie team members Woohjung Kim, Giran Ko, and Yoonsook Lee for their continuing efforts on the breeding survey. We particularly thank Dr. Susan Lappan for linguistic help and Youngeun Jo, Hyunkyung Cha, and Soyon Hwang for providing data from the 2008 gaze experiment. We also thank Byungsoon Jang, Hongsup Shin, Heeyoon Kim, and Changseok Han for discussions and comments. PGJ thanks the Dean, Faculty, and Administration of the School of Biological Sciences, and especially Prof. Sa-Ouk Kang and Prof. Jung-Hye Roe for help, advice, and patience during his adaptation to life at a Korean University. This work was funded by the Long-term Ecological Monitoring Program on Animal Populations, Research Grant of 2007 from the Korean Ministry of Environment from Ewha Womans University, Korean Research Foundations grant No. KRF-0409-20090137, KRF-0409-20070120, National Research Foundation Grant No. NRF-2010-0029613, KRF-0409-20080118 and NRF-2010-K001149, NRF-2010-0009006, the Brain Korea 21 Project 2009, and Research Grants (3344-20090054, 3344-20080067, 3344-20070018 and 3344-20100051) from the College of Natural Sciences, Seoul National University.
Conflict of interest
The authors declare that they have no conflict of interest.
The experiments described in this report comply with the current laws of the country in which they were performed—South Korea.