Welcome to Scribd, the world's digital library. Read, publish, and share books and documents. See more
Download
Standard view
Full view
of .
Look up keyword
Like this
2Activity
0 of .
Results for:
No results containing your search query
P. 1
Domestication Hypothesis for Dogs' Skills with Human Communication

Domestication Hypothesis for Dogs' Skills with Human Communication

Ratings: (0)|Views: 15|Likes:
Published by TexunNYC

More info:

Categories:Types, Research, Science
Published by: TexunNYC on Jun 07, 2011
Copyright:Attribution Non-commercial

Availability:

Read on Scribd mobile: iPhone, iPad and Android.
download as PDF, TXT or read online from Scribd
See more
See less

11/26/2012

pdf

text

original

 
Forum Articles
The domestication hypothesis for dogs’ skills with human communication:a response toUdell et al. (2008)andWynne et al. (2008)
Brian Hare
a
,
*
, Alexandra Rosati
a
, Juliane Kaminski
b
, Juliane Bra¨ uer
b
, Josep Call
b
, Michael Tomasello
b
a
Department of Evolutionary Anthropology & Center for Cognitive Neuroscience, Duke University, Durham, NC, U.S.A.
b
Max Planck Institute for Evolutionary Anthropology, Leipzig, Germany
a r t i c l e i n f o
 Article history:
Received 7 February 2009Initial acceptance 15 June 2009Final acceptance 25 June 2009Available online 29 August 2009MS. number: AF-09-00115R 
Keywords:
cognitive evolutiondog cognitiondomesticationsocial cognition
Domestic dogs have special skills in comprehending humancommunicativebehaviours(Hare&Tomasello2005;Miklosi2008).Dogs across a rangeof breeds use human communicative cues suchas pointing or physical markers to find food that is hidden in one of two hiding places (controls rule out the use of olfactory cues;Cooper et al. 2003; Hare & Tomasello 2005; Miklosi & Soproni2006). In direct comparisons, dogs are even more skilled thanchimpanzees at using human communicative cues when searchingfor food (Hare et al. 2002; Bra¨ uer et al. 2006). Moreover, a numberof studies suggest that dogs understand human gestures commu-nicatively, as a number of possible low-level explanations havebeen ruled out (e.g. only responding to movement; reflexively co-orienting; using only familiar cues, etc.;Hare et al. 1998; Agnettaet al. 2000; Soproni et al. 2001; Miklosi 2008).The sophistication that dogs show in using human communi-cative cues led researchers to investigate the origins of these abil-itiesandtoconcludethatthesesocialskillsarenotsimplyinheritedfrom wolves nor are they simply learned as a result of exposure tohumans in ontogeny, but rather they have evolved as a result of domestication (Hare & Tomasello 2005). Although wolves canpotentially learn to use human communicative cues (much likenonhuman primates), they do not show these skills as youngpuppies and must be explicitly trained to express dog-like skills inresponse to human communicative cues (Agnetta et al. 2000; Hareet al. 2002; Viranyi et al. 2008). This conclusion holds even forwolves reared in identical conditions with a group of dogs for thepurpose of comparing their social skills with humans. In summary,it is unlikely that dogs simply inherited their unusual skills fromtheir last common ancestor with wolves (Viranyi et al. 2008).Additionally, dogs develop their ability to use human commu-nicative cues, such as pointing cues or gaze cues, as young puppiesregardless of rearing history. Even puppies as young as 6–9 weeksofagecanuseahuman’scommunicativecues,includingunadoptedpuppies still living with their littermates and having little exposuretohumansbeyondroutinecare(Hareetal.2002;Riedeletal.2008;Viranyi et al. 2008). This suggests that dogsuse of humancommunicative cues does not require extensive exposure tohumans (e.g.Hare et al. 2002).These two major findings led researchers to explore the possi-bility that dogs evolved their unusual ability to use humancommunicative cues not only during, but as a direct result of,domestication (Miklosi et al. 2003; Hare & Tomasello 2005).Comparisons between experimentally domesticated foxes anda matched control line of foxes support the hypothesis thatdomestication, or selection against aggression towards humans,can, as a by-product, lead to enhanced skills in comprehendinghuman gestures. Experimental foxes bred over 40 generations toapproach humans without fear were more skilled at using humangestures than were control foxes that were not bred over the same
*
Correspondence: B. Hare, Department of Evolutionary Anthropology & Centerfor Cognitive Neuroscience, Duke University, Durham, NC 27708, U.S.A.
E-mail address:
b.hare@duke.edu(B. Hare).
Contents lists available atScienceDirect
Animal Behaviour
0003-3472/$38.00
Ó
2009 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.doi:10.1016/j.anbehav.2009.06.031
Animal Behaviour 79 (2010) e1–e6
 
period for their reaction to humans. Crucially, the control andexperimentalfoxesparticipatedintheexperimentsatsimilarlevels(i.e.madeequalnumbersofchoiceswithintestsessions)and,whentested with a nonsocial cue, the control foxes were actually moreskilled than the experimental foxes, effectively ruling out thepossibility that the control foxes weretoofearfulor unmotivated toparticipate in all human-led tasks (Hare et al. 2005). The findingsfromtheseexperimentswithfoxesprovidefurthersupportthatthedomestication of dogs actually enhanced their ability to usecommunicative cues provided by humans.Recently, two papers challenge the ideas (1) that dogs outper-form wolves in using human communicative gestures (Udell et al.2008) and (2) that dogs require very limited human exposure toshow initial skill in using such communicative cues (Wynne et al.2008). In the first of these studies,Udell et al. (2008)present findings suggesting that wolves are more skilled than dogs in usinghuman communicative cues in a food retrieval task. Specificallytheyfoundthatwolves weremoreskilledthanshelterdogsandpetdogs tested outdoors, and equally as skilled as pet dogs testedindoors. In a second study,Wynne et al. (2008)critiqued anexperiment byRiedel et al. (2008)that examined the use of humancommunicative cues by dog puppies. Riedel and colleagues repor-ted a series of three studies showing that domestic dog puppiescomprehendedahumanpointinggestureasearlyas6weeksofage.Wynne and colleagues suggest that this study did not rule out thepossibility that human experience plays a decisive role. In bothpapers this team of authors claims that their evidence and rean-alysis refute the domestication hypothesis, and they suggest thatdifferences among canid species in social skills is largely due toenvironmental factors during rearing rather than being a result of dog domestication.To evaluate the evidence presented in these studies, we firstdiscuss several methodological concerns that we have about theapproach of Udell et al. (2008), then we reanalyse their data basedon these methodological concerns. We also present a test of shelterdogs naı¨ ve to cognitive testing to examine whether it is the casethat shelter dogs are less skilled at using human communicativecues than other groups of dogs. Finally, we directly rebut thecritique of Wynne et al. (2008)and argue that there remains noevidence of significant differences in performance between dogs of different ages in their use of human communicative cues. Weconclude that the domestication hypothesis remains the bestexplanation for dogs’ special skills for communicating withhumans.
REANALYSIS OFUDELL ET AL. (2008): ARE WOLVES MORESKILLED THAN DOGS?
In this section we describe the method of Udell et al. (2008),highlight important methodological differences with previousstudies and re-examine their data. The authors kindly provided uswith a trial-by-trial data set of both studies so that we could usea more conventional data analysis to re-examine the performanceof their subjects.Udell et al. (2008)examined the use of a pointing cue by fiveexperimental groups of canids (one group of wolves, four groups of dogs), each with eight subjects of adult age. UnlikeHare et al.(2005)andViranyi et al. (2008), none of the subjects were reared forthepurposesoftheexperiment,sotheauthorscouldnotbesureof the subjects’ previous relevant experiences. This raises thequestion of how much previous experience the wolves had withperforming human-directed tasks; this is particularly importantgiven that these wolves had been used in public education showsand had been intensively exposed to humans. The wolves inUdelletal.’s(2008)study,likethoseinHareetal.(2002)andViranyietal. (2008),were tested by caregivers in familiar outdoor enclosures,but unlike inHare etal.’s (2002)study, the caregiverinUdellet al.’s (2008)studytested the wolves while standing inside the subjects’enclosure. Three of the dog groups consisted of family-reared petdogs that were tested by different combinations of familiar orunfamiliar experimenters and testing locations: dogs in the ‘homeunfamiliar’ group were tested at their homes by an experimenter;dogs in the ‘outdoor familiar’ group were tested in a less familiaroutdoor location by a familiar human; and dogs in the ‘outdoorunfamiliar’ group were tested in a less familiar outdoor location byan experimenter. The final group consisted of dogs living at a localshelter that were tested in a less familiar room by an unfamiliarhuman experimenter.Unlike all previous studies,Udell et al. (2008)gave subjects fourwarm-up trials in which they placed food on top of one of twohiding locations in view of subjects so that, once placed, the foodwas visible to the subject (usually in a warm-up phase withprimates or dogs, food is hidden under different hiding locations asthe subjects watch). During this warm-up, food was alwayscompletely visible and all subjects correctly retrieved the visiblefood in all four trials. After the warm-up, all subjects were thengiven 10 experimental test trials. However, the methodology usedin the test trials was very different from that of previous studies. InUdell et al.’s (2008)study, no food was hidden, but rather theexperimenter pointed and ‘when a subject indicated a correctchoice, the experimenter clicked [a clicker device] and then drop-ped a piece of food on the chosen container’ (page 3). This taskdiffers greatly from the traditional method in which a humanindicates the location of hidden food to a dog (reviewed in:Hare &Tomasello 2005; Miklosi 2008).The performance of each individual subject was compared tochance using binomial probabilities (authors assigned chance to50%), one-sample
tests, one-way ANOVAs and post hoc compar-isons (Bonferroni tests). In addition tothe experimental trials, afterevery second experimental trial, a control trial was run that wasidentical to the experimental trial with the exception that nopointing cue was provided (these trials acted as control for olfac-tory cues that the subjects might be using). Unlike other studies, if subjects made incorrect choices for more than three trials consec-utively they were given another warm-up trial in which food wasplaced on one of the two hiding locations as the animal watched;this was done to rule out the possibility that the subjects’ failureswere due to lack of food motivation.Based on their analysis of the data,Udell et al. (2008)concludedthat the performance of the five groups of subjects differedsignificantly from one another. Wolves and dogs tested indoorswere the only groups to use the pointing gesture at above-chancelevels,andwolves evensignificantlyoutperformedtheshelterdogsintheiruseofthehumanpointinggesture.Afollow-upstudy(Udellet al. 2008) compared two groups of dogs for their ability to followa human point; one group was tested while they were behinda fence while the other group was tested while both the experi-menter and subject were on the same side of the fence.Udell et al.(2008)reported that the dogs tested behind fences made moreincorrect choices than did the group with no fence.In describing the methods of Udell et al. (2008)we have high-lighted severalmethodological differences between their work andprevious studies (e.g. reviewed inMiklosi 2008) regarding testprocedures. However, the most important difference concerns thecoding of trials and subsequent data analysis, which we argue arestatistically invalid. Although not stated explicitly in the paper, wenow know that subjects were coded in both studies as making anincorrect choice not only if they chose the cup that the experi-menter did not point at, but also in trials in which they made nochoice at all (C. Wynne, personal communication). This is
B. Hare et al. / Animal Behaviour 79 (2010) e1–e6 
e2
 
a particularly problematic method (and different from all previousstudies; reviewed inMiklosi 2008) given that their statisticalanalysis relies on using 50% as random choice. In this system of scoring,‘incorrect’ choices havea greaterexpectedprobability thancorrect choices because they include both the choice of the incor-rect cup and no choice (no participation). To use an analogy, thisscoring procedure would be like flipping a coin and consideringheads as correct, tails as incorrect, and trials without flipping asincorrect also (e.g. it is a truism that there is 0% chance that a coinwill land heads if it is never flipped). In previous studies theexclusion or repetition of no-choice trials has been the standard,because in no-choice trials, subjects often become distracted oreven leave the testing area (i.e. temporarily losing motivation tosolve the problem).We now present the results of the same data using an almostidentical analysis to that originally presented inUdell et al. (2008)but following the more conventional method of examining sepa-rately (1) participation (making a choice or not) and (2) the level of correctchoices(choosingthecupindicatedbyapointinsteadoftheone ignored).Figure 1presents the mean percentage of trials fromUdell et al.’s (2008)firstexperiment inwhich subjects from each of the five groups actually participated by approaching and makinga choice in the 10 experimental trials. Our one-way ANOVArevealed a significant difference in the level of participationbetween the five groups (
4,35
¼
9.06,
<
0.001). Post hoccomparisons (Bonferroni tests) revealed that the dog home unfa-miliar group participated significantly more often than the otherthree dog groups, but did not differ from the wolf group (
<
0.05for all significant tests). In addition, the wolf group participatedsignificantly more often than the dog outdoor familiar and shelterdog groups (
<
0.05). Wolves also participated marginally moreoften than the dog outdoor unfamiliar group (
¼
0.054).Figure 2presents the mean percentage of trials in whichsubjects that participated chose the container indicated by theexperimenter over the incorrect container (the one that was notindicated). A one-way ANOVA revealed no significant differencesbetween the five groups’ performance in choosing the correctcontainer in trials in which they participated (
4,35
¼
0.308,
¼
0.871). One-sample
tests revealed that all groups but theshelter dog group chose the container indicated by a point atabove-chance levels in trials when they made a choice (wolf:
7
¼
2.992,
<
0.02; dog home unfamiliar:
7
¼
4.32,
<
0.003; dogoutside familiar:
7
¼
4.344,
<
0.03; dog outside unfamiliar:
7
¼
3.422,
<
0.011; shelter dog:
7
¼
1.46,
<
0.19).Figure 3presents the mean percentage of trials fromUdell et al.’s (2008)second experiment, in which the seven dogs from both groups(
¼
14) participated in and made correct choices. As in the firstexperiment,whilethetwogroupsdifferedinthenumberoftrialsinwhich they participated (
12
¼
2.43,
<
0.05), their actual perfor-mance was identical in trials where a choice was made (
11
¼
0.11,
¼
0.92). One dog was excluded from our new analysis because itnever made a single choice; surprisingly, it was still counted asa subject in the original publication.In direct opposition to the conclusions of Udell et al. (2008),wolves did not outperform dogs in using human communicative
10080
   %    C  o  r  r  e  c   t
6040200
   W  o   l  f   o  u  t  d  o  o  r  f  a  m  i   l  i  a  r   D  o  g     h  o  m  e  u  n  f  a  m  i   l  i  a  r   D  o  g    o  u  t  d  o  o  r  f  a  m  i   l  i  a  r   D  o  g    o  u  t  d  o  o  r  u  n  f  a  m  i   l  i  a  r   D  o  g    s   h  e   l  t  e  r
******
Figure 1.
Mean
Æ
SE percentage of trials (out of 10 trials) in which subjects from thefive experimental groups inUdell et al. (2008)made a choice by touching one of twochoice options within 3 min. **
<
0.05 (post hoc comparisons between all groups,Bonferroni correction).
10080** **NS6040200
   W  o   l  f   o  u  t  d  o  o  r  f  a  m  i   l  i  a  r   D  o  g    o  u  t  d  o  o  r  f  a  m  i   l  i  a  r   D  o  g    o  u  t  d  o  o  r  u  n  f  a  m  i   l  i  a  r   D  o  g     h  o  m  e  u  n  f  a  m  i   l  i  a  r   D  o  g    s   h  e   l  t  e  r
   %    C  o  r  r  e  c   t
Figure 2.
Mean
Æ
SE percentage of correct choices for one of two hiding locations bysubjects in the five experimental groups inUdell et al. (2008). The line represents thelevel of chance performance. *
<
0.05 (one-sample
tests comparing each group tochance or 50%).
Behind fenceFenceNS*1008060
   %    C  o  r  r  e  c   t
40200Participation Correct choice
Figure 3.
Mean
Æ
SE percentage of trials (out of 10 trials) in which dogs that weretested behind a fence and dogs that were not tested behind a fence participated inchoice tests (by choosing one of two locations) and chose the correct location inUdellet al.’s (2008)study. *
<
0.05 (independent
tests).
B. Hare et al. / Animal Behaviour 79 (2010) e1–e6 
e3

You're Reading a Free Preview

Download
scribd
/*********** DO NOT ALTER ANYTHING BELOW THIS LINE ! ************/ var s_code=s.t();if(s_code)document.write(s_code)//-->