The Croatan National Forest, which is located on the coast of North Carolina (N34° 51.624′ W77° 03.165′), was established in 1936 as a multi-use US National Forest. It is boarded on three sides by the Neuse River, Bogue Sound, and White Oak River, respectively, and is surrounded by a mosaic of farming, commercial, and housing developments. The forest covers approximately 647 km2 of land and is characterised by at least seven dominant ecosystems, including long-leaf pine (Pinus palustris) forests, cypress (Taxodium distichum) swamps, pocosins, salt estuaries, open savannahs, sand beaches, and mixed pine/hardwood forests. There are four ‘wilderness areas’ that are off-limits to the public, some of which include pristine longleaf pine stands. The forest is managed by the US Forest Service of the US Department of Agriculture, which allows controlled timber extraction within designated areas. Fire is part of the natural ecology of this region and the US Forest Service conducts prescribed burns within the Croatan to reduce the hazard of uncontrollable wildfires. Other human activities allowed within the Croatan include hiking, camping, boating, biking, all-terrain vehicle driving, horse riding and seasonal hunting and fishing.
The study spanned one summer and two winter seasons between 2019 and 2020 (see “Task design and testing locations”). Total precipitation was 119.4 mm in summer and 180.1 mm in winters. Average minimum temperature was 21.7 ± 2.4 °C in summer and 5.2 ± 5.3 °C in winters. Average maximum temperature was 33.2 ± 2 °C in summer and 17 ± 5.7 °C in winters. Weather data came from the National Oceanic and Atmosphere Administration (https://ncdc.noaa.gov/cdo-web/search, retrieved 9 March 2020).
Aside from the current study, no formal studies have been published on the raccoon population within the Croatan. Subjects were wild, unmarked, and free ranging. Thus, the identities, sexes, and ages of raccoons that participated in cognitive testing could not be determined. Raccoons were classified as fully weaned if they consumed the hard food rewards provided at testing platforms and were larger than the length of the pipe task (30 cm) (Gehrt 2003; Okuyama et al. 2013).
The task, hereafter the “stick task”, required raccoons to use a stick to push or rake food from a pipe (Fig. 1). Many “tool kits” among wild animals include sticks for extracting out-of-reach food (van Schaik et al. 1999; Tebbich et al. 2007; Moura and Lee 2004; Rutz and Clair 2012). In the wild, animals might acquire knowledge about the properties and possible tool-related functions of sticks if, for example, they inadvertently dislodged food after displacing a stick, leading to positive reinforcement and repetition of the behaviour in the future (Alcock 1971). Stick tasks have been used to assess tool-using abilities in chimpanzees (Pan troglodytes), corvids, and capuchins (Sapajus apella) (Visalberghi et al. 1995; Bagotskaya et al. 2013), all of which could solve even more complex versions of the task used in the present study.
Approximately six pieces of commercial dog or cat food, i.e. high value bait for raccoons (Taulman and Williamson 1994; Andelt and Woolley 1996; Gehrt 2003; Schlexer 2008), were placed in the middle of a PVC pipe (length: 30.2 cm, diameter: 2 cm). The PVC material was too thick for subjects to break open and both ends of the pipe were too small for subjects to insert their hands to reach the food. The pipe was fixed to a wooden platform (length: 122 cm, width: 23.4 cm, height: 90.5 cm) using metal clamps to prevent animals from removing or tilting them (Fig. 1a, b). Raccoons are natural climbers, spend much of their time arboreally (Gehrt 2003), and were therefore capable of climbing onto the platforms. Two sticks made from smooth processed wood (length: 45.2 cm, diameter: 0.9 cm) were placed next to the pipe (Fig. 1a, b); raccoons could either use these novel sticks or natural sticks that they found themselves from their surroundings.
High value bait (e.g. dog/cat food, fish oil, and/or hot dog water) (Taulman and Williamson 1994; Andelt and Woolley 1996; Gehrt 2003; Schlexer 2008) was placed freely on each platform, outside the pipe, to motivate raccoons to approach platforms and engage in testing. Bait was either an olfactory sign of food presence, or in quantities too small to satiate the visiting racoons (e.g. 8 pieces of dried dog food). Previous research has reported no negative impact of human scent or trail cameras on wild raccoons’ willingness to approach materials manufactured and handled by humans (Munoz et al. 2014; Edmunds et al. 2018).
Raccoons should be physically and perceptually capable of operating the task. Although they are not entirely “primate-like” in terms of manual dexterity, raccoons are well-known for being able to use their hands to lift, hold, push, pull, and/or carry a wide variety of objects (e.g. locks, latches, levers, lids, plugs, rocks, strings, cups, and drawers) that vary in complexity, length, diameter, rigidity, and weight (e.g. Davis 1907; McDougall and McDougall 1931; Michels et al. 1961; Iwaniuk and Whishaw 1999; Snow et al. 2017; Stanton et al. 2017; Daniels et al. 2019). This includes being able to grab, pull, and vertically lift smooth wooden sticks of a similar size (30.5 cm) to those used in the current study (McDougall and McDougall 1931). Although indeed raccoons frequently use both hands to grasp objects (Iwaniuk and Whishaw 1999), handling and manipulating the sticks using their hands as “tongs” is all that would be necessary to solve the stick task. Davis (1907) also notes that through practice, raccoons can acquire the ability to use each forepaw independently with greater quickness and accuracy than they formerly could using both hands together. Similarly, Stanton et al. (2017) note that raccoons can grip and manoeuvre the handle of a metal scoop with one hand to insert it into a pipe.
The height of the task should also be appropriate for raccoons, since the skeletal morphology of this species includes a humerous, ulna, radius, and a wide, fan-like scapula with a subscapular fossa (Iwaniuk and Whishaw 1999). These bones enable raccoons to have extensive freedom of movement with their forelimbs (e.g. 180° vertical movement and rotation) without having to perform manual actions from a set posture or orientation (Iwaniuk and Whishaw 1999). Given that raccoons also have mouths to aid manoeuvring objects around in their hands (Davis 1907; McDougall and McDougall 1931), this gives them even more dexterity to aid manipulations/positioning of objects in ways that “handless” stick-using species cannot (e.g. birds).
In terms of visuo-motor skill, experimental studies show raccoons have adequate abilities for attending to fine-scale features of objects (Michels et al. 1961) and directing their hands towards relatively small targets (e.g. coins, peanuts, and buttons; Davies 1907; Breland and Breland 1961; Iwaniuk and Whishaw 1999). Their binocular vision also allows them to perceive depth and pick objects up and place them into containers, including pipes (Stanton et al. 2017). Thus, raccoons in the current study could—at the very least—grab one end of the stick and insert it into the pipe, then use their hand(s) and/or mouth to push the remaining segment of the stick forward until the stick is fully inside the pipe.
Tasks were administered once in 70 locations throughout the Croatan, which included a range of habitats (e.g. mixed hardwood forests, pine forests, swamps, and savannahs) suitable for raccoons (Gehrt 2003). Locations were spaced at least 1 km apart. None of the locations contained anthropogenic sources of food (e.g. garbage bins) accessible to raccoons.
Testing took place in one summer season (17 June–29 July 2019) and two winter seasons (13 February–1 March 2019 and 4–25 February 2020). For logistical reasons, tasks were available at 9 or 10 locations at any given time for 8.9 ± 7.4 days per location. Any animal could voluntarily participate in testing during these times, but only data from the first animal to visit a location were used in analyses.
Recording raccoon behaviour on testing platforms
An infrared motion-sensor camera (Enkeeo PH760) was placed horizontally on a tree away from each testing platform to record subjects’ behaviour (Fig. 1a). Cameras had a 120° sensing angle, a triggering distance of 20 m, and were set to their maximum sensitivity to ensure they would detect any movement at or around the platforms. Video lengths were set to their maximum coverage, i.e. a five second trigger delay, 10-min recording bouts and 5 s in between each bout. Camera lenses were sprayed with defogger and, where necessary, understory vegetation was removed to ensure optimal visibility between the camera and platform. Raccoons that operated the task were considered to be participants; the amount of time participants spent operating the task was based on the total amount of time they spent using their hands, mouth, and/or sticks to extract the food from the pipe while they were on the platform (Fig. 1c). An independent observer randomly scored 50% of videos to perform an interobserver reliability test with the original coder (F.B.M.).
Natural stick availability
The Croatan is characterised by a diverse range of ecosystems (see “Study Site”), but it was unclear whether or to what extent natural sticks would be readily available in some of them (e.g. treeless savannahs and fire-degraded habitats). Moreover, even if a given location contained woody debris, it was uncertain whether the right kind of debris (e.g. sticks without branches and of a certain length, diameter and straightness) would be available for raccoons to have opportunities to interact with and learn how to use them as tools. Thus, data collection on the availability of natural tools within the Croatan were evaluated.
Photos were taken in February 2019 at each of the 70 locations where the task was administered and later coded for the presence or absence of at least one natural stick that was suitable for solving the task. In February 2020, an “in-person” search for sticks was conducted at 35 (50%) of these locations. For both the photos and in-person searches, the sticks needed to be a suitable length and width for extracting food from the pipe and found with relatively little effort, which was defined in this study as any viable stick that could be found in under 10 s within a 2 m radius of the platform (Fig. 1 in ESM 1). An independent observer randomly scored 50% of photos to perform an interobserver reliability test with the original coder (FBM).
Novel stick exploration
Trail camera videos were used to code whether raccoons explored the novel sticks by sniffing them with their noses and/or touching them. An independent observer randomly scored 50% of videos to perform an interobserver reliability test with the original coder (FBM).
Reducing the risk of pseudoreplication
Raccoons that participated in this study could not be marked for identification and their home ranges may have overlapped (Gehrt 2003). Most home range estimates for the species fall between 0.5 and 3 km2 despite wide variation in geographic location, sample size, and methodology (e.g. Johnson 1970; Gehrt and Fritzell 1997; Walker and Sunquist 1997; Gehrt 2003). In the south-eastern United States where the current study took place, the largest average home range on record for a population is 0.7 km2 in females and 2.6 km2 in males (Walker and Sunquist 1997). Thus, after collecting data across all 70 locations, the risk of resampling the same individuals was reduced in the current study by only analysing videos from locations that were > 3.4 km2 apart, i.e., larger than the typical home range of the species plus an additional 10% “buffer”. The risk of resampling was further reduced by only using videos of the first raccoon to visit a given location from the time they first climbed onto the platform until they climbed down to the ground and moved out of the camera’s range of view. Finally, in some cases, it was possible to distinguish between raccoons based on their physical markings (see “Video Analysis”).
Raccoons were recorded at 23 of the 70 locations and were all fully weaned. Two locations were < 3.4 km2 apart and so the video from one of these locations was randomly selected. On four other occasions, raccoons visited locations that were < 3.4 km2 but were included in analyses because it was possible to distinguish between individuals based on physical markings (e.g. tail band length and colouration) (Figs. 2–6 in ESM 1). Due to camera malfunctions, videos from two locations only depicted raccoons as they climbed down from the platforms. Thus, videos from 20 locations (10 from summer and winter, respectively) were retained for analyses of raccoons’ participation and performance on the task. Lastly, video from one location depicted a raccoon whose body blocked their responses to the sticks. Thus, videos from 19 locations (10 from summer and 9 from winter) were retained for the stick exploration analysis.
Two types of intraclass correlation coefficients (Shrout and Fleiss 1979) were calculated to determine interobserver agreement between raccoons’ operation time on the task (i.e. time spent trying to gain access to food in pipes). The first intraclass correlation coefficient, ICC (3, 1), indicates the reliability of single ratings. The second, ICC (3, k), indicates the reliability of the mean scores across k raters (two raters in the present study). Cohen’s kappa coefficients were calculated to determine interobserver agreement on raccoons’ methods of stick exploration (sniff, handle, or both), tool availability judged from photos (i.e. presence or absence of sticks), and the degree of overlap between the tool availability scores derived from the original photos and the in-person searches.
Chi-squared tests were used to test whether there were more locations with natural sticks available than locations where they were absent. Chi-squared tests were also used to test whether raccoons at each location were more likely to acknowledge versus ignore trail cameras, and whether they were more likely to use tactile versus alternative means (e.g. sniff) to explore the novel sticks.
A Mann–Whitney U test was used to compare seasonal differences (winter versus summer) in the average amount of time raccoons spent operating the task, and whether operating time differed according to whether or not the subject looked in the direction of the trail camera during their session.
The data analysed in this study are provided in Tables 1–5 in ESM 1. Statistical analyses were conducted using SPSS Version 26.