Intrinsic Ball Retrieving in Wolf Puppies Suggests Standing Ancestral Variation for Human-Directed Play Behavior.

Domestication dramatically alters phenotypes across animal species. Standing variation among ancestral populations often drives phenotypic change during domestication, but some changes are caused by novel mutations. In dogs (Canis familiaris) specifically, it has been suggested that the ability to interpret social-communicative behavior expressed by humans originated post-domestication and this behavior is thus not expected to occur in wolves (Canis lupus). Here we report the observation of three 8-week-old wolf puppies spontaneously responding to social-communicative behaviors from an unfamiliar person by retrieving a ball. This behavioral expression in wolves has significant implications for our understanding and expectations of the genetic foundations of dog behavior. Importantly, our observations indicate that behavioral responses to human social-communicative cues are not unique to dogs. This suggests that, although probably rare, standing variation in the expression of human-directed behavior in ancestral populations could have been an important target for early selective pressures exerted during dog domestication.

Introduction

Domesticated animals express dramatic phenotypic alterations compared with their ancestral species (Darwin, 1868, Driscoll et al., 2009). Although phenotypic change can be attributed to novel mutations, a growing body of evidence suggests that evolutionary change relies heavily upon standing genetic variation (Barrett and Schluter, 2008, Larson et al., 2014). Indeed, although few novel mutations with large effects account for some phenotypic differences between domestic and ancestral populations (Parker et al., 2009, Larson et al., 2014), animal domestication was likely initiated by selection on standing genetic variation within ancestral populations (Larson et al., 2014). The potential for domestic phenotypes to derive from existing variation has been well demonstrated in the farm fox project (Belyaev et al., 1985, Trut et al., 2009), where strong selection regimes based on observed variation in the behavioral trait tameness (i.e., reduced aggression and increased docility) among pre-selection foxes brought about rapid occurrence of classic morphological phenotypes associated with domestication. Clarifying whether the basis for traits selected upon during early domestication are variants from ancestral populations is central to developing our knowledge of the domestication process. For instance, wild species expressing variation for the trait tameness are arguably more likely to be successfully domesticated compared with species that do not (Dobney and Larson, 2006). Therefore, disentangling whether phenotypic change in domesticates is caused by novel mutations or selection on standing ancestral variation is important if we are to advance our understanding of which traits had a fundamental role during initial stages of animal domestication.

The dog (Canis familiaris), which was domesticated from the gray wolf (Canis lupus) at least 15,000 years ago (Driscoll et al., 2009), shows extreme phenotypic variation as a species. Present-day dogs are bred for highly breed-specific requirements for behavior and morphology (Svartberg, 2006, Mehrkam and Wynne, 2014), and although a large amount of the resulting variation is believed to originate from standing genetic variation in ancestral populations (Ostrander and Wayne, 2005), novel mutations have had a significant impact during breed formation (Larson et al., 2014). For instance, black coat color (Candille et al., 2007, Anderson et al., 2009), chondrodysplasia (foreshortened limbs, Parker et al., 2009), and brachycephaly (pathologically short muzzle, Schoenebeck et al., 2012) are traits that have occurred in modern dogs through novel mutations. An additional example comes from a genome-wide analysis of genetic difference between dogs and wolves, identifying dogs as having an increased copy number of the amylase locus (AMY2B), which was argued to be a novel adaptation to a starch-rich diet in early-domesticated dogs (Axelsson et al., 2014). However, investigations of a wider range of individuals revealed standing variation in amylase copy numbers in wolves, thereby shifting the AMY2B example from being a novel mutation important in domestication to yet another example of selection upon standing variation as an essential substrate for domestication (Freedman et al., 2014). This critical distinction has important implications for hypothesizing how dog domestication could have taken place. Thus, the AMY2B example illustrates the importance of continued research on ancestral species to better describe existing variation among wolves, thereby avoiding misclassification of traits expressed in dogs as novel.

Although much progress has been made in studying the morphological and physiological differences between wolves and dogs, understanding the basis and origins of behavioral variation has proven more elusive (Larson and Fuller, 2014). One behavioral skill that has been suggested to be novel in dogs compared with wolves is their interspecific social competence (Topál et al., 2009, Miklósi and Topal, 2013). Specifically, it has been posited that, unlike wolves, dogs possess unique skills to interpret human cues (Hare et al., 2002, Topál et al., 2009) and that these skills might have arisen after the domestication process from the gray wolf had been initiated (Hare et al., 2002, Hare and Tomasello, 2005, Miklósi and Topal, 2013). The ability to interpret human social cues has received considerable interest from researchers comparing behavior in dogs and wolves. However, owing to substantial differences in testing procedures, environmental factors, and interpretation of results, consensus among these studies is lacking (Hare et al., 2002, Miklósi et al., 2003, Gácsi et al., 2009, Udell et al., 2008, Udell et al., 2012). Consequently, whether wolves have the ability to interpret human social cues, or whether this is a novel trait in dogs, remains unresolved.

The ability to follow human gestures to access a food reward has been demonstrated in a range of both domesticated and non-domesticated species (Maros et al., 2008, von Bayern and Emery, 2009, Hall et al., 2011, Nawroth et al., 2013). However, these studies all include some degree of prior training and/or familiarity with the communicative person. We remove all these factors, including the food reward, by focusing upon human-directed play with an unfamiliar person as a behavior exemplifying human-animal cooperation and animals' ability to interpret human social cues. Human-directed play behavior has been reported in some domesticated species (Melotti et al., 2014, Mertens and Turner, 2015), including dogs (Horváth et al., 2008, Rooney and Bradshaw, 2002, Rooney et al., 2001, Tóth et al., 2008). Dogs can interpret human play cues and adjust their behavioral repertoire when playing with a human instead of a conspecific (Rooney et al., 2000, Rooney et al., 2001). Within a domestication context, wherein animals have been selected for greater tolerance of and interactions with humans, interspecific human-directed play behavior represents a highly relevant behavior to address. However, to date only one study exists comparing human-directed playfulness in a domesticated species and its ancestral proxy species (Hansen Wheat et al., 2018), and studies on human-directed play behavior in wolves have never been attempted.

Here we report on the spontaneous expression of human-directed play behavior, in the form of ball retrieving for an unfamiliar person, in 8-week-old, hand-raised wolves. Our observations occurred during a standardized test in which wolves, with no prior training, are vocally encouraged to retrieve a ball and thus respond to social-communicative behaviors from a human they had never met before. Based on the existing literature (Hare et al., 2002, Topál et al., 2009, Hare and Tomasello, 2005, Miklósi and Topal, 2013), we expected that human-directed play behavior as a trait occurred after the initial domestication of dogs and that wolves therefore would not respond to interspecific social-communicative behaviors or engage in human-directed play with a stranger.

Results

Our observations occurred during a subtest in a standardized test battery aimed at describing the behavior of puppies at the age of 7–9 weeks. Specifically, the subtest quantifies social play and cooperation with an unfamiliar person, the puppy assessor, who throws a tennis ball across the test room. The puppy is given the opportunity to chase after and play with the ball, before the puppy assessor encourages the puppy to cooperate with her and retrieve the ball to her. The subtest is repeated three consecutive times and cooperation is scored on a scale from 1 to 5, where 1 is no expressed interest in the ball and 5 is full cooperation/retrieving (Table 1). Thirteen hand-raised wolf puppies were subjected to this test at the age of 8 weeks. Wolves had not been trained and had only spent time with their caregivers prior to testing, i.e. at the time of testing it was the first time they interacted with an unfamiliar person.

Table 1

Behavioral Scoring

Score	Behavior
1	The puppy shows no interest in the ball
2	The puppy plays with the ball on its own, but aborts
3	The puppy plays with the ball on its own, but ignores the puppy assessor's call
4	The puppy responds to the puppy assessor's call, initiates retrieving but releases the ball
5	The puppy responds to the puppy assessor's call and retrieves the ball to her

Cooperation in the three consecutive retrieving tests is measured on a scale from 1 to 5, where 1 is no cooperation and 5 is full cooperation.

Three wolves, all from the 2016 litter, fully retrieved the ball at least two times, and one of those wolves fully retrieved the ball all three times (Score: 5, Figure 1, Video S1). One of the wolves fully retrieving the ball two times also played with the ball in one of the trials, but ignored the puppy assessor's call (score: 3, Video S2). One wolf from the 2014 litter and one from the 2016 litter showed some interest in playing with the ball on at least one trial but aborted (Score: 2). Eight wolves (four from the 2014 litter, both from the 2015 litter and two from the 2016 litter) showed no interest in the ball in any of the three trials (Score: 1, Video S3).

Figure 1

Behavioral Scores

Cooperation scores in the retrieving test for 13 wolves on three consecutive trials (shading from light to dark with the first trial being light, second medium, and third dark). Behavior is scored on a scale from 1 to 5. Only scores 4 and 5 include partial or full retrieving, respectively. Photo credit: Christina Hansen Wheat.

Video S1. Score 5, Related to Figure 1

Eight-week-old wolf puppy (Sting) fully retrieving the ball upon encouragement from the puppy assessor. This behavior is scored as 5, full cooperation.

Video S2. Score 3, Related to Figure 1

Eight week-old wolf puppy (Elvis) playing with the ball on its own, but ignoring the puppy assessor's call. This behavior is scored as 3.

Video S3. Score 1, Related to Figure 1

Eight-week-old wolf puppy (Hendrix) showing no interest in the ball. This behavior is scored as 1, no cooperation.

Discussion

Here we provide the first empirical evidence that wolves, and not only dogs, express interspecific play with a human based on social-communicative cues. Our finding is surprising given that dogs' ability to interpret social-communicative behavior expressed by humans has been suggested to be a novel trait occurring after domestication had been initiated (Hare et al., 2002, Hare and Tomasello, 2005, Topál et al., 2009, Miklósi and Topal, 2013). Importantly, our results suggest that, although probably rare, standing variation in the expression of human-directed behavior, including play, in ancestral populations could have been an important target for early selective pressures exerted during dog domestication.

Our observations of three wolf puppies retrieving a ball are highly relevant for the on-going discussion on how domestication affects behavior and further have significant implications for our understanding and expectations of the genetic foundations of behavior in modern-day dogs. Specifically, in relation to current attempts to reveal the genomic basis of behavioral changes during domestication (Pendleton et al., 2018, Freedman et al., 2014, Kukekova et al., 2018), our observations indicate that signatures of selection for human-directed behavior in dogs are likely to be weak and prone to false positives (sensu lato Messer and Petrov, 2013, Pritchard et al., 2010). This is because (1) we must now consider that selection likely acted upon standing variation in interspecific social-communicative behavior in wolves, (2) this behavior almost certainly has a polygenic genetic architecture, and (3) samples sizes in recent genomic studies are small and therefore lacking sufficient power to detect the expected selection dynamics.

In sum, we argue that, to answer questions about the evolutionary foundation of dog behavior, research attention should refocus away from solely conducting direct species comparisons and include studies upon whether or not specific behavioral variation inherently exists among wolves. Identifying such instances has important ramifications upon expectations of how dog domestication proceeded.

Limitations of the Study

We present results based on a limited number of wolves. However, because our results provide proof of concept by demonstrating the presence of a specific behavioral trait in wolves, the number of tested individuals is not crucial for the interpretation of our results.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.