Report

Social Learning of Migratory Performance

See allHide authors and affiliations

Science  30 Aug 2013:
Vol. 341, Issue 6149, pp. 999-1002
DOI: 10.1126/science.1237139

Follow the Leader

How birds migrate between wintering and breeding grounds, often over thousands of kilometers through difficult conditions, remains mysterious. The recovery of North American Whooping Cranes by release of captive-reared birds trained to migrate by following aircraft provided an opportunity for Mueller et al. (p. 999; see the cover) to analyze 8 years of data for individual birds. The presence of older birds within a group of migrating cranes significantly decreased the deviations the flock took from a straight line migration path. The lack of evidence for a genetic component indicates that social learning dominates any innate capacity in developing migratory behavior.

Abstract

Successful bird migration can depend on individual learning, social learning, and innate navigation programs. Using 8 years of data on migrating whooping cranes, we were able to partition genetic and socially learned aspects of migration. Specifically, we analyzed data from a reintroduced population wherein all birds were captive bred and artificially trained by ultralight aircraft on their first lifetime migration. For subsequent migrations, in which birds fly individually or in groups but without ultralight escort, we found evidence of long-term social learning, but no effect of genetic relatedness on migratory performance. Social learning from older birds reduced deviations from a straight-line path, with 7 years of experience yielding a 38% improvement in migratory accuracy.

Mechanisms underlying the complex phenomenon of animal migration have been particularly well studied in birds (15). In some taxa, individuals may migrate alone, unaided by conspecifics and relying instead mostly on endogenous, genetically inherited navigation programs (6). In other species, innate programs alone are not sufficient, and experiential learning is critical to successful navigation, as adult animals often have markedly better navigational capabilities than juveniles (79). Information transfer from more experienced individuals to inexperienced ones can be essential to navigational success, especially for species that travel in groups (810). Current hypotheses, richly supported by theoretical studies (1113), posit that social learning, coupled with interindividual coordination of movements, is essential to successful migration and the maintenance of group structure. Despite these advances, disentangling the contributions of experience and social learning (mediated through cultural transmission of knowledge) versus innate navigation programs (inherited genetically) to migratory performance represents a key challenge in understanding animal behavior (2).

To address that challenge, we used databases emerging from long-term investment in the conservation of whooping cranes (Grus americana). Formerly widespread in North America, whooping cranes are now endangered and restricted to a relictual wild population, migrating between northwestern Canada and the Texas coast, and birds that originate from reintroduction efforts. The most successful reintroduction effort to date established the eastern migratory population (EMP). Most birds in the EMP migrate between a summering range centered on Necedah National Wildlife Refuge in Wisconsin and a wintering range in the area of Chassahowitzka National Wildlife Refuge in Florida (Fig. 1).

Fig. 1 Whooping crane location data.

(A) Migration map for the EMP of whooping cranes (2002–2009). We identified each bird’s summer and winter ranges in each year using the mean coordinates of all locations for that individual during summer and winter times when birds are not migratory. We then identified the straight-line path for each migration event linking consecutive summer and winter (or winter and summer) ranges for each bird. We calculated the deviation of each migratory relocation from the straight-line path and used this as a simple proxy for migratory performance. Variation in data availability over the 8 years of the study precluded application of more complex measures of deviation, such as those based on full trajectories that might take into account heterogeneity in wind strength and direction, topography, and the availability of suitable stopover sites. (B) Typical migratory pattern for two 1-year-old individuals migrating in spring 2005 traveling without (red) and with (blue) older birds.

We used 2002–2009 relocation data for the EMP, which have two distinct features that allow us to partition genetic and socially learned aspects of migration. First, the EMP is wholly derived from a captive breeding program, and the pairwise genetic relatedness of all migrants is known from studbook pedigrees (14). Second, new captive-bred, naïve-to-migration birds were transferred each summer to Necedah National Wildlife Refuge and, during their first fall, were trained on their southbound migration route by human-piloted ultralight aircraft (15, 16). Thus, all birds were initially trained to follow the same migration route. Although another release method has been used in this population more recently, we included only ultralight-trained birds in our study. After this first training flight, the cranes migrate freely, flying in groups with other cranes but without ultralight aircraft on all subsequent north- and southbound migrations. Because of their endangered status, the stepwise progression of each bird’s route on each migration is intensively monitored, yielding both detailed spatial information and comprehensive information on group composition on each trip.

The necessity for ultralight training suggests that successful migration in whooping cranes depends on both social learning and innate programs. As in storks (17), southward autumn migrations by naïve-to-migration, captive-reared juveniles flying in the absence of experienced individuals would be unlikely to lead to a successful journey, suggesting that cultural transmission of information is important (15). Innate programs influence initiation of migration in that ultralight-trained birds can initiate the northbound spring migration independently of experienced birds (or ultralight aircraft) (15).

To disentangle the effects of social learning and innate navigation programs on migratory performance, we extracted data on four predictor variables from the crane databases: (i) the age of all individuals on each flight (which we hypothesize as a measure of experiential learning), (ii) the age of the oldest individual(s) in a migrating group of cranes (which we hypothesize provides an upper limit on experiential learning within each group, as one or more individuals may be the oldest in a group), (iii) the group size as a measure for potential group navigation, and (iv) genetic relatedness (which captures interindividual nonindependence in the birds’ innate navigation programs). We used deviations from a straight-line path between summer and winter ranges on the migratory route of individual birds as a proxy for migratory performance and built a hierarchical linear mixed model to examine how much of those deviations at each observed location on the migratory route could be explained by individual age, age of the oldest individual(s) in a migratory social group, group size, and genetic relatedness on both individual and group levels (18). In addition, our model included the effects of sex and season (18).

Social learning facilitated long-term increases in the accuracy of migration. The age of the oldest individual(s) in a group improved migratory performance by ~5.5% per year of age (Fig. 2), decreasing the average deviation from a straight-line path by ~4.2 km per year of age for each relocation event [posterior mode: –4.2 km, 95% highest posterior density interval (HPDI): –1.1 to –7.2 km]. Flight groups in which the oldest individual(s) had a migratory age of 1 year were predicted to deviate ~76.1 km from the straight-line path per relocation, whereas in groups in which the oldest individual(s) was 8 years old, the predicted deviation was only 46.8 km. Thus, 7 years of experience translated into a 38% improvement in migratory performance (Fig. 2). Overall, autumn locations were predicted to deviate 36.3 km more from the straight-line paths than did spring locations (95% HPDI: –21.0 to –54.3 km) (Fig. 2). We found no significant effects of sex (posterior mode: 2.4 km, 95% HPDI: –2.9 to 8.8 km), individual migratory age (as opposed to age of the oldest bird in a group) (posterior mode: –1.4 km, 95% HPDI: –3.8 to 1.2 km), or group size (posterior mode: –0.3 km, 95% HPDI: –4.8 to 5.8 km). The lack of an effect of group size is interesting, as theory suggests that larger group sizes may aid navigation (13). In addition, a fixed effect for mean group genetic variance [i.e., the mean breeding value (19) potentially predicting better migratory accuracy] was indistinguishable from zero (standardized posterior mode: –0.03 km, 95% HPDI: –10.7 to 24.9 km) (Fig. 2). This finding implies that closely related birds did not migrate more similarly to each other (either increased or decreased accuracy) than did less related birds.

Fig. 2 Migratory performance of whooping cranes measured as deviation of locations from straight-line path versus age of oldest individual(s) in each flight group with more than one individual.

Original data [red; means with 95% confidence intervals (CIs)] and model predictions (black; posterior modes, quartiles, and 95% HPDI) accounting for variation in several other factors, including migratory season, bird sex, and the birds’ genetic relatedness, are shown. Sample sizes (N) are relocation events. (Inset) Posterior distributions of overall model terms. Mean additive genetic variance refers to the group effect of genetic variance [i.e., the mean breeding value (18)].

On average, 1 year olds that traveled with older birds deviated by 63.9 km from the straight-line paths, which was 34% less than for 1 year olds that traveled in same-age groups (mean deviation: 97.1 km) (Fig. 3). Accounting for other sources of variation, the modeled difference was 44.7 km (95% HDPI: 6.6 to 85.7 km). Groups of 1-year-old birds migrating without older birds were particularly prone to large deviations from the straight-line route. Fully 25% of locations of 1 year olds flying without older birds exceeded 150 km of deviation, the largest deviation observed for mixed-age groups (Fig. 3).

Fig. 3 Distance from straight-line path for locations of 1-year-old birds that migrated with older bird(s) compared with 1-year-old birds that migrated in groups without older bird(s).

Box plots providing minimum, maximum, medians, and upper and lower quartiles are shown in gray. Means and 95% CIs are shown inside the box plots.

Previous research has contributed to overall understanding of the role of experience, social transmission of knowledge, and innate programs in navigation and route-learning of birds. However, results must be drawn piecemeal from across diverse studies. For example, translocation experiments involving several bird species have demonstrated individual learning (9), social learning (20), and innate programs (21) in isolation. The crane analyses reported here provide an integrated, multiyear portrait of these critical issues within a single species.

We show that learning of migration routes by whooping cranes takes place over many years and that social transmission of knowledge by experienced older birds yields progressive improvements in migratory performance of younger birds. In cranes, social learning may contribute to improved navigation through spatial memory of landscape features. Tracking studies on the relictual wild population of whooping cranes suggest that memory of landmarks across small scales and long-distance responses to large-scale topography may aid navigation (22, 23). Experience may also manifest as improved coping with weather patterns such as wind drift, as has been shown for raptors (24). More than 75% of our relocations were to the east of the straight-line migratory path, which accords with the predominantly westerly winds in the region (fig. S1). The absence of experienced adults may lead to failed or misdirected migration (8, 17). In our study, we found that younger birds lack accuracy only when unaided by older birds (Fig. 3), suggesting the absence of intentional exploration. However, greater deviations in the fall migration may arise from explorations in search of alternative overwintering areas, because birds do not always return to the initially trained overwintering area (Fig. 1A).

In other bird species, the timing and direction of migration are strongly heritable (6). Even though we did not find a significant effect of genetic relatedness on migratory performance, other evidence indicates that innate programs must play a role in some aspects of crane migration. For example, the first independent northward migration can be initiated by flight groups that consist of only juvenile birds, demonstrating that some elements of migration knowledge need not be culturally transmitted in whooping cranes.

Beyond their contributions to understanding social learning of migration, our findings have important implications for conservation and reintroduction efforts of whooping cranes. If experience and learning accrue with time for crane reproduction, as demonstrated here for migration, additional experience may also improve successful reproduction in the wild, especially given the potential links between migratory performance and breeding performance (25). Because the average age of the whooping crane EMP is itself increasing, further improvements in migratory performance are expected. Previous studies have demonstrated that leadership plays an important role in bird homing navigation and that more experienced birds are more likely to become leaders (26); our results show that social learning enhances group navigation performance for long-distance migrants and that the benefits of experience accrue over many years.

Supplementary Materials

www.sciencemag.org/cgi/content/full/341/6149/999/DC1

Materials and Methods

Fig. S1

References (2733)

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
  2. Acknowledgments: T.M. and W.F.F. were supported by NSF Advances in Biological Informatics award 1062411, T.M. and R.B.O. were supported by the Landes-Offensive zur Entwicklung Wissenschaftlich-ökonomischer Exzellenz programme (Hesse, Germany), and T.M. was supported by the Robert Bosch Foundation. Data were provided by the Whooping Crane Eastern Partnership and K. Jones and are deposited in the Dryad Repository: http://doi.org/10.5061/dryad.1r0f7. We thank F. Hailer, S. Servanty, E. Grant, J. Calabrese, R. Reynolds, S. Via, E. L. Neuschulz, C. Rushing, D. Bennu, and three anonymous reviewers for helpful comments and discussions.
View Abstract

Navigate This Article