The ability to learn about the spatial environment plays an important role in navigation, migration, dispersal, and foraging. However, our understanding of both the role of cognition in the development of navigation strategies and the mechanisms underlying these strategies is limited. We tested the hypothesis that complex navigation is facilitated by spatial memory in a population of Chrysemys picta that navigate with extreme precision (±3.5 m) using specific routes that must be learned prior to age three. We used scopolamine, a muscarinic acetylcholine receptor antagonist, to manipulate the cognitive spatial abilities of free-living turtles during naturally occurring overland movements. Experienced adults treated with scopolamine diverted markedly from their precise navigation routes. Naive juveniles lacking experience (and memory) were not affected by scopolamine, and thereby served as controls for perceptual or non-spatial cognitive processes associated with navigation. Further, neither adult nor juvenile movement was affected by methylscopolamine, a form of scopolamine that does not cross the blood–brain barrier, a control for the peripheral effects of scopolamine. Together, these results are consistent with a role of spatial cognition in complex navigation and highlight a cellular mechanism that might underlie spatial cognition. Overall, our findings expand our understanding of the development of complex cognitive abilities of vertebrates and the neurological mechanisms of navigation.
Animals use a variety of cognitive processes to address complex navigation tasks . In many migratory species, learning as well as individual or social experiences are important factors in successfully locating alternative seasonal habitat [2–7]. For example, paths taken by sandhill cranes become more precise with age and experience , and sea turtles imprint on their natal site and return as adults (e.g. [6,8,9]). Similarly, the ability to gather information about the spatial environment plays a critical role for migration, dispersal and foraging [10,11]. Some animals appear to learn to take paths associated with landscape features  and may use complex spatial memory to navigate between the nesting and foraging sites . Although much progress has been made in recent years (e.g. [4–9]), the mechanisms that underlie navigation behaviour and the full role of cognition (herein defined as ‘all processes involved in acquiring, storing and using information from the environment’ ) in navigation remain enigmatic, especially in vertebrates . In particular, the role of spatial memory in navigation has been controversial, especially in some taxa [15–17].
Given the inherent difficulties of controlling, manipulating and tracking wild animals, few studies have manipulated animal cognition in the wild as a means of understanding navigation. Many studies demonstrate that species with behaviours that might rely on spatial memory (e.g. migration) tend to have enhanced spatial memory performance in the laboratory [3,17,18]. Although informative in many respects, this approach provides only correlative evidence of a relationship between cognition and navigation (sensu [19,20]). Experimental manipulations of cognition and its effect on navigation in wild vertebrates, therefore, remain sparse [1,11]. Studies that do attempt cognitive manipulation often either focus inadvertently on orientation (as opposed to navigation itself) or generate results that can be explained by orientation alone. Thus, the role of cognition and memory in navigation per se often remains unclear. For example, Kohler et al.  treated homing pigeons with scopolamine, a muscarinic acetylcholine receptor (mAChR) antagonist, and found these pigeons oriented towards home with less accuracy and were slower to return to the roost relative to control birds. While Kohler et al.  showed an effect of the drug on animal movement, they did not monitor the birds' movements in real time with high precision either before or after pharmacological manipulation. Thus, Kohler et al.  could not demonstrate the degree of spatial deviations or alteration in specific aspects of navigation. Such difficulties are inherent in many study systems, and therefore, common in many studies of spatial memory in wild vertebrates. Thus, our understanding of the development and mechanisms underlying complex spatial cognition and strategies for navigation is limited .
To better address the possible role of spatial memory in navigation, we pharmacologically manipulated the behaviour of semi-aquatic turtles (Chrysemys picta, the Eastern painted turtle) in a system known for extremely precise navigation . In our system, turtles are forced to navigate terrestrial habitat after annual aquatic habitat loss. Animals in our focal population locate far-off, permanent water sources using one of four very precise (±3.5 m), highly predictable routes (animals use the same routes year after year; figure 1) [22,23]. Naive turtles must navigate this site prior to age 4 to be able to successfully navigate as adults [22,23]. The precision of the routes suggests that turtles use a fine-scale, ground-based cue when learning to navigate [22,23]. Although the information used to learn navigation routes is probably multimodal, previous studies in reptiles suggest vision or olfaction as possibilities [24–30], and our own work suggests that aspects of UV vision might facilitate learning the navigational routes .
The extreme precision in navigation, the repeatability of complex movements and the critical period for learning to navigate seen in this system provide a unique opportunity to make very specific predictions about the nature of navigation of these animals. We used this unique system to test the importance of spatial memory for navigation in free-living animals using an established pharmacological technique to manipulate spatial cognition in these turtles. Unlike some other model systems (e.g. migrating birds or sea turtles), our system allows us to monitor animal navigation in real time with exceptionally high spatial and temporal resolution. Thus, we can clearly document the behavioural ramifications of experimental manipulation of cognition in freely navigating animals.
Here, we present a series of controlled field experiments in which we studied the role of cognition in navigating C. picta by manipulating mAChRs, receptors known to play a role in spatial memory (e.g. [31,32]). First, we blocked mAChRs in navigating adult turtles using scopolamine. We predicted that if adult turtles use spatial memory to navigate, they should deviate from their paths when injected with scopolamine. If they use some non-spatial-memory-based technique, then they should continue to navigate with high precision. Second, to control for possible effects of scopolamine in the peripheral nervous system (PNS), we treated a separate group of adults with methylscopolamine, which cannot enter the central nervous system (CNS) [31,32]. If adult turtles use some aspect of the CNS to navigate these routes, then methlyscopolamine should have no effect on navigation. If, on the other hand, turtle navigation is a function of mACh-related processes in the PNS or if scopolamine interferes with peripheral navigation processes, then methylscopolamine should cause adult turtles to deviate from the predicted routes. Finally, we controlled for the possibility that scopolamine might affect cognitive or perceptual processes in the CNS unrelated to spatial memory by pharmacologically manipulating naive juveniles (1–3 years old) with either scopolamine or methylscopolamine. If either drug disrupts any navigation-relevant process other than spatial memory, then these naive juveniles, who have not yet formed memories of the site, should not be able to navigate with high precision. If, however, scopolamine only disrupts spatial memory and methylscopolamine acts as a functional PNS control, then naive turtles will retain the ability to navigate routes that they had never previously encountered.
2. Material and methods
(a) Model species
This study focused on C. picta, a long-lived (approx. 25 years) pond turtle inhabiting a variety of fresh water bodies across the northeastern United States . This species has been studied extensively (e.g. [22,31–35]), particularly in reference to movements, orientation and navigation (e.g. [23,29,34,36–39]), learning (e.g. [22,31,35]), and spatial memory (e.g. ).
Semi-aquatic animals like C. picta are uniquely susceptible to stressors associated with habitat loss. Navigating to new habitats requires that they traverse a physiologically hostile terrestrial matrix , where they are exposed to elevated risks of hyperthermia, desiccation and predation. Rapid terrestrial navigation would minimize these risks, and selection would plainly favour navigation efficiency. It is against this selective backdrop that we address the role of spatial memory in navigation.
(b) Study system
We conducted our study at Chesapeake Farms, a 3300 acre wildlife management and agriculture research area in Kent Co., MD, USA (39.194° N, 76.187° W; figure 1; ). The focal portion of the site is composed of five wetland impoundments (three permanent and two temporary). The temporary ponds (each approx. 2.5 ha in area) have experienced a rapid draining (the entire pond is drained in approx. 24 h) each summer for nearly 100 years for the purpose of wetland management. After draining, resident turtles leave the pond and navigate to alternative aquatic habitats using one of four very precise, complex and highly predictable routes described above (; see also figure 1).
(c) Assessing turtle movements in search of alternative habitat
We used radiotelemetry to investigate the means by which turtles navigate to alternative aquatic sites after the draining as per our previous work [22,23]. We captured turtles using baited hoop traps and basking traps, and fitted them with radiotransmitters (models RI-2B or BD-2, Holohil Systems, Ltd, Ontario, Canada, less than 4.5% body mass). We tracked turtles via remote triangulation (e.g. ), taking high precision (spatial resolution of 2.5 m) locations on all animals more than three times per hour for the entirety of their terrestrial journey.
(d) Pharmacological manipulation
To investigate the role of spatial memory in adult turtle navigation, we performed pharmacological manipulations of spatial memory using scopolamine, a mAChR antagonist known to disrupt spatial memory in this species [31,32]. We allowed turtles to navigate their routes naturally prior to manipulation, thereby allowing each individual to act as its own control. Once adult turtles (n = 9) navigated approximately half of their paths, they were injected with scopolamine hydrobromide in physiological saline (6.4 mg kg−1, IP; Sigma-Aldrich, S0929) .
To control for side effects of scopolamine on the PNS  and thus, to demonstrate that the observed effects of scopolamine are cognitive in nature (i.e. the behaviours observed are the result of CNS manipulations), we treated an independent group of adults (n = 9) with scopolamine methylbromide (hereafter methylscopolamine) in physiological saline (6.8 mg kg−1, IP; Sigma-Aldrich, S8502) , a quaternary form of scopolamine that binds to mAChRs with the same affinity as scopolamine, but does not cross the blood–brain barrier . Methylscopolamine is widely used as a control for scopolamine's potential PNS side effects, effects that might otherwise mask its manipulation of cognition (e.g. thirst, pupil dilation, thermal sensitivity, heart rate) [42,43]. Prior experimental work in this species showed no reductions in physiological drive or motivation associated with methylscopolamine, thereby confirming its validity as a PNS control (e.g. [31,32]). As before, we allowed turtles to navigate approximately half of their routes prior to manipulation.
Scopolamine is known to disrupt spatial memory in this species [31,32], but scopolamine might affect other cognitive or perceptual processes in the CNS unrelated to spatial memory or cognition (e.g. anxiety, motivation). To control for these factors, and to decouple them from aspects of spatial cognition, we also manipulated animals that were naive to the system, and thus, had no experience with or memory of navigation. Naive juvenile turtles within the critical period (1–3 years old) are able to successfully navigate routes despite their lack of experience or memory . We were able to exploit this ability of naive animals as a control of a variety of experimental factors relating to the effects of these drugs on the brain. We are assuming that the main differences between the behavioural abilities of naive juveniles and experienced adults are rooted in a critical learning period in juveniles, whereby juveniles can learn novel routes, but adults cannot . Consequently, naive adults outside of the critical period have already been shown to be unable to learn/navigate traditional routes ([22,23]; Krochmal and Roth 2015, unpublished data). Thus, they could not be used as controls in our experiment.
We translocated naive juvenile turtles to our study site from a distant location (Chester River Field Research Station, Queen Anne's Co., MD; 39.222° N, 76.986° W; ). We determined the age of juveniles using growth rings (a technique with high accuracy in young turtles [22,23,44,45]). These two sites were separated by 18.5 km, a distance that greatly exceeds even the largest overland movements in C. picta (e.g. ), and were not connected by water. Thus, we were assured that the navigation ability of translocated turtles did not reflect experiences gained at the site prior to translocation (i.e. animals were indeed naive ).
Naive juveniles were translocated to the focal site and, after the draining, were allowed to choose and navigate approximately half of a route, at which time we injected them with either scopolamine (n = 7) or methylscopolamine (n = 6) at the same concentrations used in adults. Given that naive juveniles lack memory of the paths, yet can still navigate the paths based on local sensory cues , they can act as a control for many of the CNS effects of scopolamine. If either drug disrupts navigation-relevant behaviours or processes other than spatial memory, then we predict that naive juveniles should deviate from the traditional paths. Turtles hatched on-site exhibit identical navigation precision to naive translocated animals, and all naive turtles—irrespective of home population—seem to navigate by relying on the same phenomenon to learn and remember routes [22,23].
(e) Statistical analyses
We compared the navigation patterns among treatment groups as per our previous analyses . Briefly, we compared path specificity and precision by calculating the spatial variability of paths taken by each individual as the distance of each individual from the traditional route. Using LOAS (Ecological Software Solutions) and ArcGIS v. 10.2.1 (Esri Industries), we first documented the traditional path taken by resident adults in previous years . Given that resident paths are historically accurate to 3.5 m, we statistically compared all turtle movements to this template by calculating the mean distance of each location for each individual (using the individual turtle as sampling unit) from that of the historical paths. We analysed these distances across treatment groups with a general linear model with Fisher's LSD post hoc comparison . We also produced a measure of the deviation of movements relative to the residents' paths by examining the proportion of points that overlapped the traditional 3.5 m path (see the electronic supplementary material).
To further characterize movements, we quantified movement rate, time spent stopped, and latency to resume movement after stopping across all treatment groups and levels. Movement rate was calculated as distance moved (metre) per time (second), for observations during which the animals were moving (all time stopped was removed from rate calculations). Latency to resume movement was calculated as the duration of each individual cessation of movement while navigating. All measures were calculated as individual animal means and then pooled within age class and treatment. By addressing these aspects of movement, we aimed to assess changes in ability, motivation or strategy to reach alternative water across different phases of navigation, age classes and pharmacological treatments. Post-injection measurements were initiated 15 min after the injection to remove any potential impact of handling. We compared these aspects of movement before and after injection using a within-subjects design. The turtles in the adult scopolamine treatment deviated significantly from their paths upon injection, but then re-established their paths (see below). Thus, we also analysed movement for this group after they recovered their navigation ability and returned to the path.
(a) Navigation of adult turtles was disrupted with scopolamine, but not methylscopolamine
Upon treatment with scopolamine, adults were immediately disoriented and moved off of their paths, roaming significantly away from their predicted, historical route (general linear model, F3,27 = 35.091, p < 0.001; figures 1a and 2; electronic supplementary material, maps and figure S1). We note that this response was robust, occurring in all habitat types and at all locations where adults received scopolamine. Compared with their movement before injection, turtles became significantly slower (F2,26 = 432.872, p < 0.001), spent more time stationary (F2,26 = 190.790, p < 0.001) and were slower to restart movement (F2,26 = 252.559; p < 0.001) after treatment with scopolamine (see table 1 for descriptive statistics and electronic supplementary material, table S2 for raw data).
This striking effect of scopolamine was temporary, with all scopolamine-treated adults regaining precise navigational abilities, returning to their paths at varying points across the specific routes, and navigating to their alternative aquatic habitats with the high precision typical of this population (p > 0.999; figure 1a; electronic supplementary material, maps). Our field observations of the timing of the turtles' ability to regain navigation precision are consistent with laboratory estimates of the duration of the effect of scopolamine used in this study (approx. 6–8 h [31,32]).
Nevertheless, because we could not monitor the concentration of drug metabolites in our animals in the field, we cannot conclusively say when the drug lost its effect. Thus, we categorized recovery movements conservatively in light of the loss and subsequent gain of navigational ability inferred from behaviour (the turtles returned to the traditional path). Once the turtles returned to their path, movement rates (p = 0.742), time stopped (p = 0.636) and latency to move (p = 0.868) were not significantly different from pre-injection movements (table 1; electronic supplementary material, tables S1 and S2). Furthermore, a subset of these animals (n = 7 of 9) were monitored navigating these same paths in the previous years. The pre-injection and recovery movements of these turtles were not significantly different across treatments (F3,27 = 1.834, p = 0.165) or from movements of the same individuals on the same paths the previous year (no manipulation; all ps > 0.268; see electronic supplementary material, tables S1 and S2).
The ability of adult turtles to navigate was completely unaffected by treatment with methlyscopolamine, with animals moving with equal precision both before and after injection (p = 0.997; figures 1a and 2; electronic supplementary material, maps and figure S1). No turtle positions were detected beyond the 3.5 m buffer of the historic path for any turtle in this group, and all animals navigated with the expected precision to their traditional alternate aquatic habitats. Adults treated with the methylscopolamine control showed no difference in their movement rate (t8 = 0.837, p = 0.542), time stopped (t8 = 0.492, p = 0.636) or latency to move after stopping (t8 = 0.171, p = 0.868) when compared with their pre-injection movements (table 1). Likewise, adult movement in this group was not different from untreated adults tracked in previous years (n = 60; movement rate: F3,78 = 0.180, p = 0.909; LSD, p = 0.498; time stopped: F3,78 = 0.186, p = 0.906; LSD, p = 0.502; latency to move: F3,78 = 1.430, p = 0.240; LSD, p = 0.495; electronic supplementary material, tables S1 and S2).
(b) Navigation of naive juveniles was not disrupted by scopolamine or methylscopolamine
Neither scopolamine nor methylscopolamine altered the spatial abilities of naive juveniles. Navigational precision was not altered by injection of either drug, with pre-injection precision not differing from post-injection precision in either group (p = 0.997; figures 1b and 2; electronic supplemental material, maps and figure S1). No turtle positions were detected beyond the 3.5 m buffer of the path for any juvenile turtle treated with either scopolamine or our PNS control and all naive juvenile turtles were successful in navigating to alternative aquatic habitat. Furthermore, neither scopolamine (movement rate: t6 = 1.241, p = 0.561; time stopped: t6 = 0.714, p = 0.502; latency to move: t6 = 0.238, p = 0.820) nor methylscopolamine (movement rate: t6 = 1.009, p = 0.352; time stopped: t6 = 0.592, p = 0.575; latency to move: t6 = 0.119, p = 0.909) affected movement before or after treatment (table 1). Finally, we note that juvenile movement patterns were not different from untreated adults tracked in previous years (n = 60; movement rate: F3,78 = 0.180, p = 0.909; LSD, all ps > 0.723; time stopped: F3,78 = 0.186, p = 0.906; LSP, all ps > 0.725; latency to move: F3,78 = 1.430, p = 0.240, LSP, all ps > 0.117; see electronic supplementary material, tables S1 and S2).
We tested the importance of cognition and cognitive processes in navigation by pharmacologically blocking the mAChRs of free-ranging turtles during a period of highly predictable, precise and quantifiable movement. Scopolamine dramatically disrupted the adults' ability to navigate. We controlled for physiological side effects of this drug by both using methylscopolamine and by observing the effect of both drugs on naive animals that successfully navigate without memory. Given the lack of effect of methylscopolamine on adult navigation, the lack of effect of both methylscopolamine and scopolamine on naive juveniles, and the strong predictive nature of the study system, we can rule out strictly sensory and perceptual means as the mechanism for adult navigation. Thus, our study suggests a role for cognitive processes in complex navigation, is consistent with the hypothesis that these cognitive processes involve spatial memory, and highlights a cellular mechanism that might underlie memory formation and recall navigating animals.
In the light of the complex nature of cognition and of our pharmacological manipulations, we must interpret our results with some caution. First, mAChRs are extremely widespread both in the PNS and CNS, and are involved in numerous behaviours and physiological processes (e.g. ). Additionally, scopolamine is a general antagonist of mAChRs, affecting all muscarinic receptor types. Thus, scopolamine can have numerous physiological, behavioural and neurological effects, and despite its common use to manipulate cognitive function [42,43], including in this species (e.g. [31,32]), some authors question its validity as a means of doing so (reviewed by ).
We have, however, taken great steps to account for these potential confounding factors by (i) using a robust experimental design backed by strong inference within a highly predictable system, (ii) including in our experiment both experienced and naive animals, and (iii) using a control drug that does not cross the blood–brain barrier . We observed no effect of methylscopolamine on navigation precision on any animal, and therefore, conclude that the effects of scopolamine in the adult turtles were not associated with the PNS. For example, scopolamine is known to dilate pupils ; however, methylscopolamine induces approximately twice this effect . Had the marked decreased in navigation ability of the adult scopolamine treatment been due to pupil dilation (or any other PNS effect), we would have expected lower performance in the methylscopolamine group. We also note that some studies in mammals have found an effect of methylscopolamine on cognition , although studies in turtles (e.g. [31,32]) did not. In our study, neither of these issues was apparent, as we did not see any changes in movement precision or rate in adults treated with methylscopolamine. Thus, we conclude that methylscopolamine is a reasonable control for our experiment, a conclusion that is consistent with previous studies (e.g. [41,42]) including those conducted in this species (e.g. [31,32]).
A major strength of our system is that by using naive juveniles as a control, we were able to dissociate cognitive processes involving memory of the site from non-cognitive processes (those not involving information obtained from prior experiences with the site, i.e. learning and memory). Although naive juveniles certainly use cognitive processes while navigating (sensu , i.e. they are using sensory cues to learn paths, thereby creating spatial memory), they cannot (by the fact that they are translocated to the site) have memory of the site. While navigation via direct social cues is common in animals (e.g. ), the timing of movements our system precludes juveniles from learning by direct social observation of adults . Instead, these juveniles, as well as resident juveniles during their first draining year, must be identifying paths based on other processes using fine-scale, ground-based cue(s) . Our previous work failed to find support for the use of olfaction in navigation, but did find support for the idea that naive juveniles may use indirect social cues such as UV reflected light from excretia or secretia , an idea consistent with prior studies in turtles [26–30]. Irrespective of the cue(s) by which naive juveniles navigate and learn their paths, their navigation was not disrupted when treated with either scopolamine or methylscopolamine, demonstrating that the marked deviations in the use of space observed in scopolamine-treated adults were probably not a function of disrupting sensory capabilities (e.g. vision, olfaction) or motor coordination. Moreover, these results were not probably the effect of an age-related sensory/perceptual loss. The vast majority of sensory studies are conducted on adults (e.g. [24,25,29,31,32,35,37]) including studies of UV vision [27,28,47], suggesting that adults can indeed detect the cues probably important for juvenile navigation.
Nevertheless, it is conceivable that observed behavioural patterns could, in principle, be the result of an interaction between scopolamine and age. Based on our results, we can conclude that the cognitive processes used by adult turtles to navigate that were disrupted by scopolamine are not the ones used by juveniles. However, consistent with any study of the behavioural expression of complex cognitive processes, our study has some limitations. We do not think that our results are rooted in age-specific gains in the importance or the sensitivity of mAChR, whereby adults were more susceptible to the mAChR antagonist scopolamine than juveniles. The literature suggests the opposite—the number and sensitivity of mAChR decrease with age (e.g. [48,49]). Furthermore, we observed no such change in behaviour—every juvenile performed the task perfectly with precision consistent with untreated animals in our previous work  (see electronic supplementary material, maps). However, although we have eliminated peripheral effects and many aspects of sensory perception, we cannot eliminate a complex interaction between scopolamine and levels of motivation, anxiety or other aspects of perception (e.g. differential perception of risk) with age. Indeed, there are certain aspects of motivation and perception that are difficult, if not impossible, to control, particularly under natural conditions in the field. Nevertheless, given the similarity in behaviour between juveniles and adults in this study and our previous work [22,23], the body of work on sensory reception in this species [22–30], and the previous literature demonstrating the effects of mAChR-blocking drugs on cognition in this and other species [31,32,41–43], we find it unlikely that an age-by-scopolamine interaction underlies the observed pattern of behaviour, yet remain open to the possibility that such a complex interaction exists.
The complexity of cognitive abilities of vertebrates emphasizes the need for future studies to examine the mechanisms and development of cognitive processes of free-living animals. Future studies should consider the impact of individual experiences on behavioural plasticity and cognition, the role of the environment in setting this duration of plasticity from an evolutionary perspective, and the neurological mechanisms that underlie these phenomena. For example, two key qualities of our system are the severity of the environmental change for a semi-aquatic animal coupled with the predictability of that change. The selection pressure for efficient navigation in our system is high, as both the cost of terrestrial movements (e.g. thermal stress, predation) and the consequences of not finding water (e.g. desiccation) are severe. Yet the draining of ponds has occurred annually at the site for approximately100 years, perhaps enhancing the efficacy of learning or other cognitive strategies for navigation. Although cognitive processes which underlie navigation are often difficult to dissociate from more simplistic explanations, especially in free-living animals , our study highlights the importance of cognition in understanding how animals respond to their environments (e.g. [50–52]). Thus, we advocate for an integrative and naturalistic approach to cognitive research that both observes and manipulates animals in situ to test hypotheses related to animal cognition. Using experimental manipulation and strong inference to integrate multiple fields of animal behaviour will be pivotal to unravelling the evolution of cognition, the relationship between the brain and the environment and their influence on complex behaviours such as navigation [51,52].
This work was reviewed and endorsed by the Maryland Department of Natural Resources (SCO # 51936) and the Institutional Animal Care and Use Committees of Washington College (protocol #F11-004 and SU11-001) and Franklin and Marshall College (2013-48; 2014-73). Our field study was conducted in accordance with all local, state and federal laws and by permission of the land owners.
Spatial data (maps) and movement timing data are included in the electronic supplementary material.
Both authors contributed equally to this project.
We declare we have no competing interests.
This research was funded by Washington College's Provost's Office, Middendorf Fund, Hodson Trust, and Franklin and Marshall's Hackman Fund and College of Grants.
We thank E. Counihan, W. Gerwig, S. Giordano, F. Rauh, A. Roth, S. Rush, N. Simmons, J. Sullivan and K. Wachter for assistance in the field. We thank M. Conner, R. Fleegle and D. Startt at Chesapeake Farms and Chino Farms for permission and access. The Washington College GIS Program helped with the preparation of the maps. We thank M. Bashaw, J. Lesku, A. Owens, A. Powers, N. Rattenborg and two anonymous reviewers for helpful comments and discussion.
- Received October 22, 2015.
- Accepted January 20, 2016.
- © 2016 The Author(s)
Published by the Royal Society. All rights reserved.