How wolves were first domesticated is unknown. One hypothesis suggests that wolves underwent a process of self-domestication by tolerating human presence and taking advantage of scavenging possibilities. The puppy-like physical and behavioural traits seen in dogs are thought to have evolved later, as a byproduct of selection against aggression. Using speed of selection from rehoming shelters as a proxy for artificial selection, we tested whether paedomorphic features give dogs a selective advantage in their current environment. Dogs who exhibited facial expressions that enhance their neonatal appearance were preferentially selected by humans. Thus, early domestication of wolves may have occurred not only as wolf populations became tamer, but also as they exploited human preferences for paedomorphic characteristics. These findings, therefore, add to our understanding of early dog domestication as a complex co-evolutionary process.
Citation: Waller BM, Peirce K, Caeiro CC, Scheider L, Burrows AM, et al. (2013) Paedomorphic Facial Expressions Give Dogs a Selective Advantage. PLoS ONE 8(12): e82686. doi:10.1371/journal.pone.0082686
Editor: Claire Wade, University of Sydney, Australia
Received: July 8, 2013; Accepted: October 27, 2013; Published: December 26, 2013
Copyright: © 2013 Waller et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: The study was funded by a WALTHAM Foundation Research Grant to BMW, JK and AB. The funder had some input during study design, but did not influence study findings, interpretation of results or writing of the manuscript. The funder had some input during study design, but did not influence study findings, interpretation of results or writing of the manuscript.
Competing interests: The authors declare that author SM is an employee (Research Manager) of WALTHAM, a division of Mars Inc. All other authors have declared that no competing interests exist. This does not alter their adherence to all the PLOS ONE policies on sharing data and materials.
Wolves were domesticated early in the history of human civilization , and have since evolved into dogs whose lives are now inextricably linked to those of humans. The initial steps that led to wolves becoming domesticated, however, is unknown. One hypothesis suggests that wolves underwent a process of self-domestication as tamer individuals took advantage of opportunities to scavenge from human settlements during the agricultural revolution . In support of this theory is recent evidence that domestic dogs exhibit genetic mutations to a starch-rich diet . During domestication, dogs have departed from wolves on various other behavioral and physical dimensions , , , one of the most striking being paedomorphism. In many ways dogs appear more like wolf puppies than wolf adults. These features are thought to have evolved as a byproduct of the domestication process, and arose accidently when aggression was actively selected against , , for a detailed review see .
Paedomorphic features, however, could have evolved much earlier in response to human preferences. Domestic cats have developed modified purr vocalizations that appear to solicit increased care from human hosts by mimicking human infant cries , and which may have increased tolerance of cats in human environments during domestication. Likewise, the shorter snout and wider cranium of the dog give the dog face a more puppy like appearance (although there is variation between breeds) which may have evolved as the well documented human preference for paedomorphic facial characteristics  was exploited. Paedomorphic facial features can be further enhanced through use of upper face facial muscle contractions that lift the brow to increase the apparent height and overall size of the orbital cavity (i.e. the apparent size of the eyes: figure 1). Large eyes relative to the rest of the face are a prominent feature in human infants and are associated with perceived cuteness of and motivation to invest in human infants by human adults , . Toys (teddy bears) that display this trait are also preferred , . Infantile facial features are similarly preferred in pet dogs and cats , and manipulation of infant-like facial traits increases perceived cuteness . However, in all of these studies humans are making forced choices in experimental conditions. In addition, demonstrating visual preference does not necessarily mean that these animals are (or have been) selected preferentially. To demonstrate whether these human preferences translate into differential investment we need to examine which dog characteristics incur a current selective advantage. Current fitness is not necessarily indicative of past selection of course, but it is a common assumption in behavioural ecology and evolutionary anthropology.
Figure 1. Example of facial movement AU101 (inner brow raiser) in a domestic dog (Rhodesian Ridgeback, not a subject in the study), increasing the height and overall size of the orbital cavity (eye): A) neutral on right side of face, B) AU101 on right side of face.doi:10.1371/journal.pone.0082686.g001
Juvenile traits other than face may have also been subject to selection, of course. Tail wagging and other submissive behaviours are more common in wolf puppies than adult wolves but persist in the adult dog , and are more often human directed . Such behaviours, however, are not human-like or even universally mammalian, so it is unlikely that they would be as salient as the face (to humans), which is widely understood to be an attention grabbing stimulus in both humans and other animals.
Dog facial expressions have been described in classic studies , but as the facial muscles of social mammals (including humans) can exhibit great subtlety of movement, standardised methods for facial movement measurement are needed to make accurate observations within and between species. Scientists must use validated, anatomically based systems for recording facial expression. First, facial expressions are processed as whole units in an automatic, streamlined manner which makes it difficult to see the detail accurately . Second, human observers tend to categorise facial expressions in terms of emotion, which can affect how comparisons between species are made . The Facial Action Coding System (FACS: ) is an anatomically based facial expression coding system used in humans to counter these problems, which identifies observable facial changes associated with underlying muscle movement. Recently, the system has been successfully modified for use with chimpanzees , rhesus macaques , hylobatids  and orangutans . The systems are objective, reliable and standardised, and allow subtle movements to be identified and quantified.
In the current study we used shelter dog rehoming as a proxy for dogs’ selection over time. We tested whether humans (when adopting dogs from a shelter) actively select for dogs, which appear more juvenile in the face as a result of facial muscle contraction. AU101 (inner brow raiser) raises the medial portion of the brow increasing the apparent size of the eyes in relation to the face, and as such enhances one of the features of the face associated with infants. Subtle facial muscle movements were recorded using an anatomically based facial muscle coding system (DogFACS). We examined whether frequent use of these movements (AU101: inner brow raiser) was associated with selection by humans using real world shelter dog adoption speed as a proxy for human selection over evolutionary time.
Materials and Methods
This study was carried out in strict accordance with the recommendations in the ASAB/ABS guidelines for the use of animals in research and was approved by the University of Portsmouth Animal Ethics Committee.
Development of DogFACS
Footage from 28 privately owned dogs of varying breeds (approximately 8–10 hours) from the Max Planck Institute for Evolutionary Anthropology DogLab was the primary source for DogFACS development. In addition, we sourced approximately 100 clips from www.youtube.com (permission granted from the copyright holder of each clip) and used ad hoc footage from 86 dogs at four dog shelters (Portsmouth City Dog Kennels; Wood Green, The Animal’s Charity in Cambridge; The Dog’s Trust, West London, Harefield and RSPCA Southridge Animal Centre, London). Each facial movement was documented by appearance changes, minimal criteria for identification and comparison to other species, in line with FACS terminology (Table 1). The muscular basis of each facial movement was verified in light of dissection of a face from a specimen of a domestic dog (AMB) as well as previously published dissections . The manual is freely available and requires certification to use (www.dogfacs.com).
Table 1. Comparison of action units (AUs) and the underlying facial muscles in humans  and dogs.doi:10.1371/journal.pone.0082686.t001
Shelter Dog Data Collection
The study used a correlational design using data from a one-shot, timed observation. Dogs were observed at the same four re-homing shelters (above). The modal breed group (bull breeds, which includes all breeds derived from the molasser breed, Staffordshire Bull Terriers, Mastiffs and mixed bull breeds: as classified by the shelter staff using criteria from the UK Kennel Club) was chosen for analysis to minimise the variance associated with breed differences, and totalled 29 dogs. Each dog was filmed for a 2 min period (focal sampling) during controlled first contact with the experimenter. The experimenter approached the subject’s kennel room and stood in front of the room with a neutral stance and holding out one hand. Each 2 min video sample (from each dog subject) was coded using DogFACS to record the frequency of facial movements (full DogFACS coding), duration of tail wagging and time spent at the front of the kennel in close proximity to the experimenter. The number of days between becoming available for re-homing and leaving the shelter was recorded. Reliability assessment was conducted on the behavioural coding (DogFACS AUs and other behaviours: Table 2) for 30% of the sample (8 dogs) using Wexler’s Agreement (Ekman et al., 2002):
Table 2. Wexler’s agreement calculations for the behavioural coding.doi:10.1371/journal.pone.0082686.t002
Proximity and tail wagging were treated as categorical variables by using number of bouts instead of overall duration for the reliability and agreement was also assessed using Wexler’s Agreement.
Two dogs were removed as their time before re-homing was greater than the upper quartile by more than 1.5 IQR (82+87 days), and thus were perceived to be outliers (and their long stay most likely due to unusual factors). Our final sample included 27 dogs for analysis (Age range = 7–96 months, M = 29.46 months). Nonparametric correlations (as the dependent variable was not normally distributed) were used to explore the relationships between the behavioural variables and the number of days before re-homing (Table 3). AU101 and time at the front of the kennel were the only variables significantly negatively correlated with days before re-homing, indicating that dogs that produced more of these behaviours were re-homed quicker. Tail wagging was positively correlated indicating that dogs that tail wagged more were re-homed slower. Note, however, that if Bonferroni corrections were applied, no variables would be deemed significant so these exploratory findings should be taken with caution. Visual inspection of scatter plots and curve estimation were used to explore the relationships further. Time spent at the front of the kennel and tail wagging had very weak or no linear or curvilinear relationships with the dependent variable. AU101 had a significant power curve relationship with the dependent variable and the model explained a significant proportion of the variance in re-homing speed (R2 = 0.39, F(1,25) = 15.63, p<0.005), see Figure 2. From the regression equation (y = 114.12x−0.515, see Table 4) we can predict that a dog that produces five AU101 during the 2 min observation will stay in the shelter for 49.83 days on average, but if it produces 10 AU101, this would be reduced to 34.88 days, and if it produces 15, this would be reduced to 28.31 days. As there is a negative power relationship the slope becomes less steep as AU101 increases, and so the benefit (in terms of re-homing) in producing AU101 reduces with increasing AU101.
Figure 2. Relationship between frequency of AU101 and days before re-homing in the dog shelter.
Curved line shows the power estimation.doi:10.1371/journal.pone.0082686.g002
Table 3. Relationship between behaviours exhibited during the 2-homing.doi:10.1371/journal.pone.0082686.t003
Table 4. Regression statistics (power curve fit) between AU101 and the number of days before re-homing, showing unstandardised co-efficients (B) and the associated standard error (SE B), standardised co-efficients (β) and significance values (P).doi:10.1371/journal.pone.0082686.t004
Domestic dogs who produced a high frequency of facial movement to raise the inner brow (AU101) were adopted more quickly from re-homing shelters. As AU101 enhances a key feature of paedomorphism (eye size and height: ) this suggests that dogs have evolved to manipulate the human preference for paedomorphic features using the face. This is the first empirical evidence that paedomorphism plays a key role in humans’ current selection of dogs, and the first time that actual investment has been used as an indicator of preference. If the selection process in the shelter context emulates past selection during domestic dog evolution, this preference may have also been at work during early dog domestication.
Interestingly, tail wagging and close proximity to the human were not strongly associated with speed of selection by adopters, despite being factors that are commonly believed to indicate a friendly temperament. In fact, higher durations of tail wagging resulted in a longer period before re-homing. This finding further supports the growing evidence that indirect manipulation of human sensory preferences (particularly a preference for juvenile facial characteristics) has been a particularly powerful selective force in domestication , , even more so than genuine indicators of temperament. Importantly, it is highly possible that these facial expressions do not correlate with suitability as a pet, but, like superficial morphological traits, are still preferred over more relevant behavioural traits .
In humans, the equivalent facial movement to AU101 is AU1(inner brow raiser), which features heavily in human sadness expressions . It is possible, therefore, that human adopters were responding not to paedomorphism, but instead to perceived sadness in the dogs looking for adoption. However, it is also possible that the human sadness expression is itself derived from paedomorphism, and that sadness is attributed to this specific facial movement because it enhances paedomorphism and thus perceived vulnerability. Another possibility is that humans are responding to the increase in white sclera exposed in the dogs as the orbital cavity is stretched through AU101 action. Visibile sclera is a largely unique human trait  (which likely contributes to our extensive gaze following abilities) and people are more likely to cooperate or behave altruistically when exposed to cues of being watched , . It is unclear, however, whether it is the sclera specifically or simply the presence of eyes per se which has such a powerful affect on human behavior and attention, and so this is more a complimentary hypotheses as opposed to an alternative.
Our real world data show that domestic dogs who exhibit paedomorphic characteristics are preferentially and actively selected by humans as pets from rehoming shelters. This therefore supports the hypothesis that paedomorphic characteristics in domestic dogs arose as a result of indirect selection by humans rather than only being a by-product of selection against aggression. Whether our findings are transferable to other contexts, such as breeding, is unknown, and it is possible that modern breeding practices put emphasis on such specific morphological and behavioural traits that this effect is obscured. However, given that recent evidence leans towards early wolf domestication arising from tolerance of their presence rather than direct selection per se , , adoption from shelters might be a more appropriate proxy than modern breeding. We can therefore speculate that early domestication of wolves may have occurred not only as wolf populations became tamer , , but also as they exploited human preferences for paedomorphic characteristics.
The raw data.
We would like to thank the following for access to their shelter dogs: Portsmouth City Dog Kennels; Wood Green, The Animal’s Charity in Cambridge; The Dog’s Trust, West London, Harefield and RSPCA Southridge Animal Centre, London. We thank Jamie Whitehouse for reliability coding. We also thank Alison Colyer (Statistics Team, WALTHAM®) and Ed Morrison for helpful comments on the statistics, and two anonymous reviewers for constructive help revising the manuscript.
Conceived and designed the experiments: BMW AMB JK. Performed the experiments: KP CCC LS. Analyzed the data: BMW. Contributed reagents/materials/analysis tools: AMB. Wrote the paper: BMW JK. Assisted with study design: SM.
- 1. Druzhkova AS, Thalmann O, Trifonov VA, Leonard JA, Vorobieva NV, et al.. (2013) Ancient DNA analysis affirms the canid from Altai as a primitive dog. PloS ONE 8.
- 2. Coppinger R, Coppinger L (2001) Dogs: A startling New Understanding of Canine origin, Behavior and Evolution. New York: Scribner.
- 3. Axelsson E, Ratnakumar A, Arendt M-L, Maqbool K, Webster MT, et al. (2013) The genomic signature of dog domestication reveals adaptation to a starch-rich diet. Nature 495: 360–364.
- 4. Zeuner F (1967) Geschichte der Haustiere. München: Bayerischer Landwirtschaftsverlag.
- 5. Clutton-Brock J (1995) Origins of the dog: domestication and early history. In: Serpell J, editor. The Domestic Dog: its evolution, behaviour and interactions with people. Cambridge: Cambridge University Press. 7–20.
- 6. Hare B, Plyusnina I, Ignacio N, Schepina O, Stepika A, et al. (2005) Social Cognitive Evolution in Captive Foxes Is a Correlated By-Product of Experimental Domestication. Current Biology 15: 226–230.
- 7. Trut L (2001) Experimental studies of early canid domestication. In: Ruvinsky A, Sampson J, editors. The Genetics of the Dog. New York: CABI Publishing.
- 8. Hare B, Wobber V, Wrangham R (2012) The self-domestication hypothesis: evolution of bonobo psychology is due to selection against aggression. Animal Behaviour 83: 573–585.
- 9. McComb K, Taylor AM, Wilson C, Charlton BD (2009) The cry embedded within the purr. Current Biology 19: R507–R508.
- 10. Sternglanz SH, Gray JL, Murakami M (1977) Adult preferences for infantile facial features - ethological approach. Animal Behaviour 25: 108–115.
- 11. Glocker ML, Langleben DD, Ruparel K, Loughead JW, Gur RC, et al. (2009) Baby schema in infant faces induces cuteness perception and motivation for caretaking in adults. Ethology 115: 257–263.
- 12. Hinde RA, Barden LA (1985) The evolution of the teddy bear. Animal Behaviour 33: 1371–1373.
- 13. Morris P, Reddy V, Bunting R (1995) The survival of the cutest: who’s responsible for the evolution of the teddy bear? Animal Behaviour 50: 1697–1700.
- 14. Archer J, Monton S (2011) Preferences for infant facial features in pet dogs and cats. Ethology 117: 217–226.
- 15. Little AC (2012) Manipulation of infant-like traits affects perceived cuteness of infant, adult and cat faces. Ethology 118: 775–782.
- 16. Gacsi M, Gyori B, Miklosi A, Viranyi Z, Kubinyi E, et al. (2005) Species-specific differences and similarities in the behavior of hand-raised dog and wolf pups in social situations with humans. Developmental Psychobiology 47: 111–122.
- 17. Feddersen-Petersen DU, Ohl F (1995) Ausdrucksverhalten beim Hund. Jena, Stuttgart: Gustav Fischer Verlag.
- 18. Calder A, J., Young AW, Keane J, Dean M (2000) Configural information in facial expression perception. Journal of Experimental Psychology: Human Perception and Performance 26: 527–551.
- 19. Waller BM, Bard KA, Vick SJ, Pasqualini MCS (2007) Perceived differences between chimpanzee (Pan troglodytes) and human (Homo sapiens) facial expressions are related to emotional interpretation. Journal of Comparative Psychology 121: 398–404.
- 20. Ekman P, Friesen WV, Hager JC (2002) The facial action coding system. Salt Lake City: Research Nexus.
- 21. Vick SJ, Waller BM, Parr LA, Pasqualini MCS, Bard KA (2007) A cross-species comparison of facial morphology and movement in humans and chimpanzees using the Facial Action Coding System (FACS). Journal of Nonverbal Behavior 31: 1–20.
- 22. Parr LA, Waller BM, Burrows AM, Gothard KM, Vick SJ (2010) Brief communication: MaqFACS: A muscle-based facial movement coding system for the rhesus macaque. American Journal of Physical Anthropology 143: 625–630.
- 23. Waller BM, Lembeck M, Kuchenbuch P, Burrows AM, Liebal K (2012) GibbonFACS: A muscle-based facial movement coding system for hylobatids. International Journal of Primatology 33: 809–821.
- 24. Caeiro C, Waller BM, Zimmermann E, Burrows AM, Davila-Ross M (2013) OrangFACS: A muscle-based facial movement coding system for orangutans (Pongo spp.). International Journal of Primatology 34: 115–129.
- 25. Evans HE (1993) Miller’s Anatomy of the Dog. Philadelphia: W.B. Saunders Company.
- 26. King T, Marston LC, Bennett PC (2012) Breeding dogs for beauty and behaviour: Why scientists need to do more to develop valid and reliable behaviour assessments for dogs kept as companions. Applied Animal Behaviour Science 137: 1–12.
- 27. Kobayashi H, Kohshima S (2001) Unique morphology of the human eye and its adaptive meaning: Comparative studies on external morphology of the primate eye. Journal of Human Evolution [print] 40: 419–435.
- 28. Bateson M, Nettle D, Roberts G (2006) Cues of being watched enhance cooperation in a real-world setting. Biology Letters 2: 412–414.
- 29. Francey D, Bergmueller R (2012) Images of eyes enhance investments in a real-life public good. PLOS ONE 7.