A behaviour test on German Shepherd dogs: heritability of seven different traits

Page created by Franklin Webb
 
CONTINUE READING
Applied Animal Behaviour Science 79 (2002) 113±132

       A behaviour test on German Shepherd dogs:
           heritability of seven different traits
               Silvia Ruefenachta,*, Sabine Gebhardt-Henrichb,
                      Takeshi Miyakea, Claude Gaillarda
           a
            Division of Genetic Epidemiology, Institute of Animal Genetics, Nutrition and Housing,
                  University of Berne, Bremgartenstrasse 109a, CH-3012 Berne, Switzerland
       b
         Division of Animal Housing and Welfare, Institute of Animal Genetics, Nutrition and Housing,
                  University of Berne, Bremgartenstrasse 109a, CH-3012 Berne, Switzerland
                                          Accepted 11 July 2002

Abstract

   In this study, genetic and non-genetic effects on behavioural traits were estimated, based on records
of the ®eld behaviour test of the Swiss German Shepherd Dog breeding club. This standardized test
has been applied since 1949 and comprised the following seven traits: self-con®dence, nerve stability,
temperament, hardness, sharpness, defence drive and ®ghting drive. The analyses were based on the
test results of 3497 German Shepherds between 1978 and 2000. Gender, age, judge and kennel had
signi®cant effects on all behaviour traits. The heritabilities were calculated using three different
methods and ranged between 0.09 and 0.24, with a standard error varying between 0.04 and 0.06.
Phenotypic correlations among the traits lay between 0.28 and 0.94, the genetic correlations between
0.34 and 1.0. No signi®cant correlations between hip dysplasia scores and the behavioural traits were
found ( 0.04 to 0.01). The modest genetic improvement over the last 25 years in the studbook
population of the German Shepherd dog (GSD) was due to the low heritabilities of the behaviour
traits, but mainly because of the low selection intensities after the test (only 8% failed). Some
recommendations were made to improve the test and selection response.
# 2002 Elsevier Science B.V. All rights reserved.

Keywords: Behaviour test; Breeding; Dog; Genetic; German Shepherd; Heritability; Selection response

  *
    Corresponding author. Present address: Interdisciplinary Dermatology Unit, Department of Clinical
Veterinary Medicine, Laenggassstrasse 128, CH-3012 Berne, Switzerland. Tel.: ‡41-31-631-26-18/22-71;
fax: ‡41-31-631-25-41.
E-mail address: silvia.ruefenacht@itz.unibe.ch (S. Ruefenacht).

0168-1591/02/$ ± see front matter # 2002 Elsevier Science B.V. All rights reserved.
PII: S 0 1 6 8 - 1 5 9 1 ( 0 2 ) 0 0 1 3 4 - X
114           S. Ruefenacht et al. / Applied Animal Behaviour Science 79 (2002) 113±132

1. Introduction

   The behaviour of dogs is of growing public concern. In today's hectic, technical and loud
world, dogs are nevertheless expected to behave without endangering or annoying people.
Behaviour tests are used in breeding programs, as well as for the prediction of the potential
danger of particular dogs, for the selection of service dogs, for the prediction of behaviour
of dogs from shelters, as performance testing, etc. Standardized behaviour tests for the
purpose of evaluating the character or temperament of dogs in regard to people and
environment have been applied in Switzerland in various breeds for breeding purposes.
Every breeding club uses its own test with a different request pro®le. Frightened dogs with
a disposition to aggression and with a tendency to run away are generally excluded from
breeding. Special behaviour tests have been designed and used to predict the potential
danger of a particular dog (Planta and Netto, 1997, behaviour test used in Germany; Lower
Saxony, 2002, http://www.ml.niedersachsen.de/wesenstest.htm). Other performance tests,
such as dog racing, hunting tests, ``Schutzhund'' tests (guard dogs), etc. are applied for
selection purposes too. Besides their use as selection criteria, they are also used as aptitude
tests. The objectives of aptitude tests are to predict the suitability and future performance of
the dogs, e.g. as police dogs (Slabbert and Odendaal, 1999), guide dogs for the blind
(Wilsson and Sundgren, 1997a,b, 1998), family dogs (Venzl et al., 1989), and shelter dogs
(van der Borg et al., 1991; Weiss and Greenberg, 1997; Ledger and Baxter, 1997).
   Behavioural traits can only be integrated in breeding programs, if they can be measured
accurately, as objectively as possible, and if they demonstrate signi®cant genetic variation.
In one of the ®rst comprehensive studies, Scott and Fuller (1965) looked for develop-
mental, behavioural, and genetic differences in ®ve different dog breeds and their crosses
and backcrosses in different age groups. They found signi®cant genetic variation and
important maternal effects in different behavioural traits. Since Scott and Fuller (1965),
heritabilities of a multitude of behavioural traits have been estimated in various dog breeds.
In Table 1, heritability estimates of different studies are presented. The heritability values
concerning the character of dogs were heterogeneous ranging from 0 to 0.58 with an
average of 0.20. In the other two groups of traits (achievements in training/working and
sensitivity) heritabilities were lower averaging 0.12 in both groups. Comparisons between
these values have to be done with care because the de®nitions of the traits, the sampling
procedures, the methods of evaluations, the breeds, etc. varied between the authors.
   The aim of this study was to estimate the importance of genetic and non-genetic effects
on behavioural traits measured by the behaviour test of the Swiss German Shepherd Dog
breeding club. Genetic parameters such as heritabilities and genetic correlations should
point out which traits could be suitable to be included in a breeding program.

2. Materials and methods

2.1. Data

  Since 1949 the Swiss German Shepherd breeding club has applied a standardized
behaviour test (Seiferle and Leonhardt, 1984). Results of these tests have been stored in a
S. Ruefenacht et al. / Applied Animal Behaviour Science 79 (2002) 113±132             115

Table 1
Heritabilities of behaviour traits in dogs

Traitsa                         Heritability    Breed                    References

Character
  Activity                        0.53          German Shepherd          Wilsson and Sundgren (1998)
  Adaptiveness                    0.00/0.04b    German Shepherd          Reuterwall and Ryman (1973)
  Affability                      0.37          German Shepherd          Wilsson and Sundgren (1997b)
  Affability                      0.15          Labrador                 Wilsson and Sundgren (1997b)
  Affability                      0.17/0.09b    German Shepherd          Reuterwall and Ryman (1973)
  Character                       0.12          German Shepherd          Pfleiderer-HoÈgner (1991)
  Closeness                       0.04          California guide dogs    Scott and Bielfelt (1976) (cited in
                                                                         Mackenzie et al., 1986)
  Concentration                   0.28          Labrador and others      Goddard and Beilharz (1982)
  Confidence                      0.16          American guide dogs      Bartlett (1976) (cited in Mackenzie
                                                                         et al., 1986)
  Courage                         0.05/0.13b    German Shepherd          Reuterwall and Ryman (1973)
  Courage                         0.26          German Shepherd          Wilsson and Sundgren (1997b)
  Courage                         0.28          Labrador                 Wilsson and Sundgren (1997b)
  Contact 1                       0.21          German Shepherd          Wilsson and Sundgren (1998)
  Contact 2                       0.42          German Shepherd          Wilsson and Sundgren (1998)
  Excitability                    0.09          Labrador and others      Goddard and Beilharz (1982
  Energy                          0.05          American guide dogs      Bartlett (1976) (cited in Mackenzie
                                                                         et al., 1986)
  Fear                            0.46          Labrador and others      Goddard and Beilharz (1982)
  Forgetting unpleasant           0.1/0.17b     German Shepherd          Reuterwall and Ryman (1973)
    incidents
  Hardness                        0.15          German Shepherd          Wilsson and Sundgren (1997b)
  Hardness                        0.2           Labrador                 Wilsson and Sundgren (1997b)
  Heel                            0.10          California guide dogs    Scott and Bielfelt (1976) (cited in
                                                                         Mackenzie et al., 1986)
  Intelligence                    0.06          American guide dogs      Bartlett (1976) (cited in Mackenzie
                                                                         et al., 1986)
  Large ball                      0.27          German Shepherd          Wilsson and Sundgren (1998)
  Motivation                      0.34          Div. purebred dogs       Scott and Fuller (1965)
  New-experience response         0.06          California guide dogs    Scott and Bielfelt (1976) (cited in
                                                                         Mackenzie et al., 1986)
  Nervousness                     0.58          Labrador and others      Goddard and Beilharz (1982)
  Nerve stability                 0.25          German Shepherd          Wilsson and Sundgren (1997b)
  Nerve stability                 0.17          Labrador                 Wilsson and Sundgren (1997b)
  Self-right                      0.22          American guide dogs      Bartlett (1976) (cited in Mackenzie
                                                                         et al., 1986)
  Shriek                          0.24          German Shepherd          Wilsson and Sundgren (1998)
  Success                         0.44          Labrador and others      Goddard and Beilharz (1982)
  Suspicion                       0.10          Labrador and others      Goddard and Beilharz (1982)
  Temperament                     0.51          German Shepherd          Mackenzie et al. (1985)
  Temperament                     0.15          German Shepherd          Wilsson and Sundgren (1997b)
  Temperament                     0.10          Labrador                 Wilsson and Sundgren (1997b)
  Tug of war                      0.48          German Shepherd          Wilsson and Sundgren (1998)
  Willing in training             0.12          California guide dogs    Scott and Bielfelt (1976) (cited in
                                                                         Mackenzie et al., 1986)
  Willingness                     0.03          American guide dogs      Bartlett (1976) (cited in Mackenzie
                                                                         et al., 1986)
116               S. Ruefenacht et al. / Applied Animal Behaviour Science 79 (2002) 113±132

Table 1 (Continued )

Traitsa                        Heritability     Breed                    References

  Willingness                    0.22           Labrador and others      Goddard and Beilharz (1982)
  Willing new experience         0.24           California guide dogs    Scott and Bielfelt (1976) (cited in
                                                                         Mackenzie et al., 1986)
  Yelp                           0.22           German Shepherd          Wilsson and Sundgren (1998)
Achievement in training and working characteristics
  Ability to cooperate         0.28            German Shepherd           Wilsson and Sundgren (1997b)
  Ability to cooperate         0.35            Labrador                  Wilsson and Sundgren (1997b)
  Come                         0.14            California guide dogs     Scott and Bielfelt (1976) (cited in
                                                                         Mackenzie et al., 1986)
  Defence drive                  0.20           German Shepherd          Wilsson and Sundgren (1997b)
  Defence drive                  0.22           Labrador                 Wilsson and Sundgren (1997b)
  Fetch                          0.24           California guide dogs    Scott and Bielfelt (1976) (cited in
                                                                         Mackenzie et al., 1986)
  Fetch                          0.21           German Shepherd          Wilsson and Sundgren (1998)
  Fighting instinct              0.04           American guide dogs      Bartlett (1976) (cited in Mackenzie
                                                                         et al., 1986)
  Fighting the leash             0.44           Div. purebred dogs       Scott and Fuller (1965)
  Footing-crossing               0.06           California guide dogs    Scott and Bielfelt (1976) (cited in
                                                                         Mackenzie et al., 1986)
  Hare tracking                  0.03           German Hunting dogs      Geiger (1973) (cited in
                                                                         Pfleiderer-HoÈgner, 1979)
  Hare tracking                  0.02           German Wirehaired        KraÈmer (1981) (cited in Hruby,
                                                Pointer                  1991)
  Man-work                       0.06           German Shepherd          Pfleiderer-HoÈgner (1979)
  Nose                           0.01           German Hunting dogs      Geiger (1973) (cited in
                                                                         Pfleiderer-HoÈgner, 1979)
  Nose                           0.03           German Wirehaired        KraÈmer (1981) (cited in Hruby,
                                                Pointer                  1991)
  Obedience                      0.16           Div. purebred dogs       Scott and Fuller (1965)
  Obedience                      0.01           German Hunting dogs      Geiger (1973) (cited in
                                                                         Pfleiderer-HoÈgner, 1979)
  Obedience                      0.01           German Wirehaired        KraÈmer (1981) (cited in Hruby,
                                                Pointer                  1991)
  Obedience                      0.09           German Shepherd          Pfleiderer-HoÈgner (1991)
  Playful fighting               0.16/0.21b     German Shepherd          Reuterwall and Ryman (1973)
  Pointing                       0.01           German Wirehaired        KraÈmer (1981) (cited in Hruby,
                                                Pointer                  1991)
  Prey drive                     0.05           Labrador                 Wilsson and Sundgren (1997b)
  Prey drive                     0.31           German Shepherd          Wilsson and Sundgren (1997b)
  Protective instinct            0.12           American guide dogs      Bartlett (1976) (cited in Mackenzie
                                                                         et al., 1986)
  Retrieve                       0.20           German Shepherd          Wilsson and Sundgren (1998)
  Search                         0.00           German Hunting dogs      Geiger (1973) (cited in
                                                                         Pfleiderer-HoÈgner, 1979)
  Search                         0.03           German    Wirehaired     KraÈmer (1981) (cited in Hruby,
                                                Pointer                  1991)
  Self-defence                   0.11/0.26b     German    Shepherd       Reuterwall and Ryman (1973)
  Self- and handler-defence      0.04/0.16b     German    Shepherd       Reuterwall and Ryman (1973)
  Sharpness                      0.13           German    Shepherd       Wilsson and Sundgren (1997b)
S. Ruefenacht et al. / Applied Animal Behaviour Science 79 (2002) 113±132                117

Table 1 (Continued )

Traitsa                           Heritability       Breed                   References

  Sharpness                         0.11             Labrador                Wilsson and Sundgren (1997b)
  Sit                               0.06             California guide dogs   Scott and Bielfelt (1976) (cited in
                                                                             Mackenzie et al., 1986)
  Tracking                          0.10             German Shepherd         Pfleiderer-HoÈgner (1991)
  Traffic                           0.12             California guide dogs   Scott and Bielfelt (1976) (cited in
                                                                             Mackenzie et al., 1986)
  Trained response                  0.08             California guide dogs   Scott and Bielfelt (1976) (cited in
                                                                             Mackenzie et al., 1986)
Sensitivity
  Body sensitivity                  0.16             California guide dogs   Scott and Bielfelt (1976) (cited in
                                                                             Mackenzie et al., 1986)
  Body sensitivity                  0.10             American guide dogs     Bartlett (1976) (cited in Mackenzie
                                                                             et al., 1986)
  Body sensitivity                  0.33             Labrador and others     Goddard and Beilharz (1982)
  Ear sensitivity                   0.00             California guide dogs   Scott and Bielfelt (1976) (cited in
                                                                             Mackenzie et al., 1986)
  Ear sensitivity                   0.25             American guide dogs     Bartlett (1976) (cited in Mackenzie
                                                                             et al., 1986)
  Hearing sensitivity               0.00             Labrador and others     Goddard and Beilharz (1982)
  Sound shy                         0.14             Labrador and others     Goddard and Beilharz (1982)
  Auditory disturbance              0.04/0.15b       German Shepherd         Reuterwall and Ryman (1973)
  Nose distraction                  0.00             Labrador and others     Goddard and Beilharz (1982)
  Nose distraction                  0.12             American guide dogs     Bartlett (1976) (cited in Mackenzie
                                                                             et al., 1986)
  Distraction                       0.08             Labrador and others     Goddard and Beilharz (1982)
  Dog distraction                   0.09             Labrador and others     Goddard and Beilharz (1982)
  Dog distraction                   0.27             Labrador and others     Goddard and Beilharz (1982)
   a
       Original terms used by the authors.
   b
       Heritabilities of males/heritabilities of females.

database starting in 1978. Data of the present study consisted of behaviour test results of
3497 dogs from 1978 until 2000. Fifty-two percent were females and 48% males. The
average age at test was 21.5 months (9.1) where 71% were between 12 and 23 months,
22% between 24 and 35 months and 7% were 3 years or older. The average number of
progeny per dam was 2.2 (ranging from 1 to 18) and per sire 3.3 (ranging from 1 to 52). All
the dogs were born in 729 kennels. Among the 3497 dogs, 122 were inbred. The average
inbreeding coef®cient of all dogs was very low (0.1%). The inbreeding coef®cient of the
inbred dogs ranged from 0.7 to 14% (83 dogs were between 0.7 and 5%, 27 dogs between 5
and 10% and 12 dogs between 10 and 14%).

2.2. The standardized behaviour test

   The test is standardized and has remained unchanged for years (Seiferle and Leonhardt,
1984), except for the traits self-defence and ®ghting drive, which were not tested after 1989
(see later). In most cases, the owner handles the dog (handler). During the test the handler
118           S. Ruefenacht et al. / Applied Animal Behaviour Science 79 (2002) 113±132

should not force the dog into submission. Only a few commands should be given to allow
the dog to react independently in the different situations. Except at the beginning and
during the last part (handler-defence) the dog is not leashed. The dog must be at least
12 months of age and should not be older than 2 years. It can be tested only once.
   The behaviour test is performed in different locations in Switzerland during the year. It
takes place in an open area, which can be as large as half a football ®eld. The test lasts 30±
40 min per dog.
   Judges have to take a monthly 3-year training course including a ®nal examination
before they are licensed as a judge for the behaviour test. They also have to attend a yearly
brush up meeting. Between 1978 and 2000, 38 judges had rated on average 92 dogs
(ranging from 6 to 419). In order to obtain a good estimate of the effect of the judges in the
statistical analyses, we pooled the 16 judges with fewer than 30 evaluated dogs.
   The behaviour test is divided into eight different parts.

2.2.1. Part 1: Approaching the handler
   The judge approaches the handler, who keeps the dog on the leash. The judge veri®es the
identity of the dog by controlling its ear tattoo. The dog is let off the leash and the judge
asks the handler several questions about the dog: age, living space, contact with the
environment, history of ownership, relationship of dog to handler, previous working dog
training, etc.

2.2.2. Part 2: Behaviour in friendly situations
  Several different situations are tested: 10±20 people, the judge, handler and the dog
move around freely. Then the people form a line and walk towards the dog and the handler,
®rst at a normal pace, then faster. While they are walking towards each other, the handler
with the dog crosses the line formed by the people. Next, dog and handler walk through a
narrow lane of standing people. Then the people form a wide circle, while dog and handler
are in the middle of it. The people walk towards the dog and the handler, ®rst at a normal
pace, and then at running speed while clapping their hands until they almost touch the dog.
All these situations have to be done in a friendly and not threatening way for the dog. The
sequences can be changed.

2.2.3. Part 3: Reactions to different environmental stimuli
  Several optical and acoustical objects are presented to the dog in a playful manner:
bicycle bell, empty cans in a bag, large steel plate, plastic cans ®lled with stones and an
umbrella. A blanket or large piece of plastic is held high while handler and dog pass
underneath it, then it is placed on the ¯oor, and dog and handler tread on it. There should be
no provocation or intimidation, it should be more like a game for the dog. The methods and
objects can vary and are changed frequently.

2.2.4. Part 4: Reaction to gunfire
  The reaction to gun®re shot with a big bore blank pistol is tested at a distance of about
20 m. The ®rst shot is ®red when the dog and handler are moving away from the shooter, a
second one when they are moving parallel to him. The movements of the shooter should not
be visible for the dog.
S. Ruefenacht et al. / Applied Animal Behaviour Science 79 (2002) 113±132     119

2.2.5. Part 5: Play with a toy (dummy made of French linen)
   First the handler plays with the dog while holding the toy in his hands, inciting the dog to
bite, gripping it and dragging it. Then the judge takes the toy and continues the animation
while increasing the distance to the handler.

2.2.6. Part 6: Handler-defence
   The handler holds the dog on a 1 m leash. The judge ®rst approaches the handler in a
friendly way and shakes hands with him. Then the judge takes a few steps away from the
handler and the dog and suddenly attacks the handler in a way that looks physically and
verbally serious.

2.2.7. Part 7: Self-defence (not tested since 1990)
   The dog is tied to a stable object and left alone for 3 min. He cannot see the handler during
this test. The judge approaches the dog in a friendly way, seeks contact with him, speaks
some words and pets him. Then, as in part 6 of the test, the judge takes a few steps away from
the dog and then suddenly attacks the dog in a way that looks physically and verbally serious
by waving at the dog and threatening it with a stick without actually beating it.

2.2.8. Part 8: Fighting drive (not tested since 1990)
   The handler holds the dog. The judge threatens the dog (in the same way as described
before in part 7). The dog can be gently struck with a ¯exible, padded stick. Two blows can
be given on the ¯anks, thighs, or withers. The handler encourages the dog to attack the
``enemy''. Now the judge runs 20±30 m away. The handler lets the dog go, still encoura-
ging it vocally to attack the escaping judge. The judge turns around before the dog reaches
him. Although using aggressive and threatening motions, words, and gestures with the
stick, the judge does not beat the dog anymore. The handler stays in the same place and is
allowed to encourage the dog to ``®ght''.

2.3. Evaluation of the behaviour test

  The judge watches the dog attentively during the whole test and then assesses the
behaviour in eight different traits (Seiferle and Leonhardt, 1984). The subjective grading is
recorded verbally:

   Self-confidence (ability to react to new situations): A calm, interested, self-confident,
    fearless, uninhibited, dauntless, attentive, and friendly behaviour is desired without any
    signs of aggression, distrust or tendency to run away in the presence of optical or
    acoustical stimuli or sudden approach of strangers.
   Nerve stability (the way the dog reacts to the different test parts): The dog should react
    neither nervously nor hypersensitively nor jumpy.
   Reaction to gunfire: The behaviour should be dauntless, maybe interested but not
    fearful.
   Temperament (physical flexibility and intensity of reaction to different environmental
    stimuli): The dog should be temperamental and interested in the environment with a big
    radius of action.
120                   S. Ruefenacht et al. / Applied Animal Behaviour Science 79 (2002) 113±132

         Hardness (severity or ability to accept unpleasant perceptions without being deeply
          impressed afterwards): A forceful play with the toy is desired as is untiring interest in
          the different visual and acoustic objects.
         Sharpness (ability to react in an aggressive way towards a serious or serious looking
          attack): The sharpness is desired only when the dog is threatened. After the end of the
          threatening period, the dog has to calm down immediately and has to be friendly
          towards the participating people and the judge. This kind of sharpness is defined as
          desired sharpness.
         Defence drive (the ability and desire to guard and protect the threatened handler): The
          dog should show a very well-developed defence drive.
         Fighting drive (ability and desire of unimpressed and self-confident dogs to attack an
          ``enemy''): The dog should tolerate the gentle stick beats, persecute and seriously
          attack the threatening judge.
   One single person of the breeding club transformed the verbal grades into numerical
scores. The lowest scores re¯ect the desired behaviour pattern. The higher the scores are,
the increasingly undesired the behaviour is (Tables 2 and 3).
   In order to obtain a measure of the overall behaviour, the grades of the seven particular
characteristics are added to an overall score. Because only some dogs were tested for
®ghting drive, this score was not included in the overall score. Dogs with the best grades
in every trait reached the overall score of 7. When dogs passed the behaviour test
successfully, then the trait outcome was coded with a 1 and those that failed got a 2. One
of the following features was suf®cient for failing: nervousness, insecurity, anxiety,
distrust, gun shyness, missing defence drive combined with anxiety, fear-related sharp-
ness, over-sharpness. If dogs showed none of these features, they passed the test,
regardless of the grades in the other behaviour traits. In case of failure, they are excluded
from breeding.

Table 2
Transformation of the verbal gradings into scores for self-confidence and nerve stability

Behaviour pattern                                                    Scorea for

                                                                     Self-confidence              Nerve stability

Self-confident, stable nerves                                        1                            1
Self-confident, still fearless, good-natured, attentive              1                            2
Still self-confident, still fearless, good-natured, attentive        2                            1
Fearless, good-natured, attentive                                    2                            2
Fearless, indifferent, reserved                                      2                            3
Reserved, distrustful, combative                                     3                            2
Reserved, distrustful, frightened                                    3                            3
Fearful, frightened                                                  4                            3
Fearful, frightened, aggressive                                      4                            4
Self-confident, fearless, combative, aggressive                      1                            4
Aggressive, over-sharpness, termination of the test                  5                            5
      a
          Score 1: most desirable; score 5: most undesirable.
S. Ruefenacht et al. / Applied Animal Behaviour Science 79 (2002) 113±132                  121

Table 3
Transformation of the verbal grades into scores for reaction to gunfire, temperament, hardness, fighting drive,
defence drive, and sharpness

Score     Reaction to          Temperament        Hardness           Fighting     Defence      Sharpness
          gunfire                                                    drive        drive

1         No fear              Temperamental      Middle to good     Pronounced   Pronounced   Desired
2         Slightly impressed   Slightly reduced   Slightly reduced   Good         Good         Slightly reduced
3         Impressed            Reduced            Reduced            Latent       Latent       Missing
4         Fearful              No temperament     Dauntfull          Absent       Absent       Slightly increased
          a                    a                  a                  a            a
5                                                                                              Over-sharpness
          a                    a                  a                  a            a
6                                                                                              Fear-related
    a
        In reaction to gunfire, temperament, hardness and fighting drive the grades range from 1 to 4.

2.4. Statistical methods

  Descriptive statistics were implemented by use of the SAS1 software (Release 8.0, SAS
Institute Inc., NC, USA, 1999). Linear models were applied to analyse the in¯uence of
®xed non-genetic effects on the different behavioural traits. These results were used to
determine the ®xed effects for the genetic analysis. The applied model could explain 32±
40% of the total variation in the analysed traits (Table 4). Signi®cant differences were often
found between the gender, ages, judges, evaluation system before and after 1990, and the
kennels but rarely for the locations of the test (Table 4).
  The different behaviour patterns were rated in four to six categories; therefore, the
phenotypic distribution of these traits may not follow a normal distribution. Evaluations of
categorical traits are tricky because many statistical methods assume normality. In order to
feel more con®dent about the estimated heritabilities variance components were calculated
by a restricted maximum likelihood (REML) and a Bayesian analysis.

2.4.1. REML approach
   Estimates of genetic variance and covariance components were obtained by the
derivative-free REML approach (Boldman et al., 1995). In this analysis, the behavioural
traits were treated as quantitative traits. Heritabilities and genetic correlations were
estimated based on these (co)variance components. The following linear animal model
was ®tted to the data:
          yijklm ˆ gi ‡ sj ‡ pk ‡ kl ‡ b1 ageijklm ‡ b2 age2ijklm ‡ am ‡ eijklm
where yijklm is the behaviour trait of dog m, gi the fixed effect of gender (i ˆ 1, 2), sj the
fixed effect of evaluation system (j ˆ 1 (up to 1989), 2 (from 1990)), pk the fixed effect of
the judge (k ˆ 1±23), kl the fixed effect of the kennel (l ˆ 1±729), age of the dog at
examination as a covariant (linear and quadratic), am is the additive genetic effect of the
dog m, and eijklm is the random residual. The random effects were distributed independently
as multivariate normal with mean zero and covariance matrix:
             2                
            a         Asa 0
      Var       ˆ
            e          0    Is2e
122
                                                                                                                                                                 S. Ruefenacht et al. / Applied Animal Behaviour Science 79 (2002) 113±132
Table 4
Significance levels of fixed effects and coefficient of determination (R2) for eight behavioural traits, the outcome and the overall score

Effect          Self-          Nerve           Reaction        Temperament Hardness            Defence         Fighting        Sharpness     Outcome   Overall
                confidence     stability       to gunfire                                      drive           drive                                   score

Gender
S. Ruefenacht et al. / Applied Animal Behaviour Science 79 (2002) 113±132     123

where A is the numerator relationship matrix, I is the identity matrix of relevant size and
scalars are variance components to be estimated. The REML method based on the animal
model includes the effect of selection by means of the relationship matrix (Mrode, 1996),
which results in robust and unbiased estimates of heritabilities.

2.4.2. Bayesian analysis with Gibbs sampling
   Theory of polygenic model and genetic analysis of quantitative traits by Bayesian
analysis with Gibbs sampling is based on the ®ndings of Wang et al. (1994). Conditional
posterior density functions needed to construct the polygenic model were obtained with the
program package MAGGIC (Janss, 1998). The same model was used as in the REML
approach. In this analysis, the burn-in period and the spacing were 20,000 and 100,
respectively. A total of 270,000 Gibbs samplings were performed to obtain 2500 samples
for each parameter.
   The basic theory of genetic analysis of categorical traits was based on Wright's
liability concept (Wright, 1934) in which the underlying continuous variable, named
liability, with thresholds imposes a discontinuity on the visible categorical traits. In order
to construct conditional posterior density functions needed for the liability concept, the
method of threshold model by Bayesian analysis with Gibbs sampling introduced by
Sorensen et al. (1995) was adapted for the animal model. As suggested by these authors
s2e ˆ 1:0 was used as a standardization of the scale. In addition, the maximum and
minimum liability values were ®xed at ‡10.0 and 10.0 on the standardized liability
scale. In this analysis, the dependent variable yijklm was replaced by the liability lijklm. The
®xed effects pk and kl of the above model were treated in this analysis as random effects,
as suggested by Hoeschelle and Tier (1995) and discussed by Moreno et al. (1997). To
treat them as random effects is useful because their mean is zero and therefore does not
change the normal shape of liability distribution. The ®xed effects gi and sj that have only
two classes were treated as covariates. In categorical trait analysis, it is generally rather
dif®cult to obtain convergence compared to the quantitative trait analysis because the
number of unknown parameters was increased (e.g. number of liability values for each
individual and threshold value). That was the reason why a larger number of samplings
was applied for the categorical trait analysis. A total of 540,000 Gibbs samplings were
performed to obtain 5000 samples for each parameter. The burn-in period and spacing
were 40,000 and 100, respectively.
   Mode and mean of the 2500 or 5000 samples were taken to estimate the heritability.
Together with the means, standard deviations were also calculated, which correspond to the
standard error of the estimates.
   The heritability (h2) is de®ned as the ratio of additive genetic variance to the total
phenotypic variance. The phenotypic correlation between two characters is de®ned as
phenotypic covariance over the product of the two standard deviations.
   The dogs' breeding values (BVs) for the different behaviour traits were estimated
according to the best linear unbiased prediction (BLUP) method (Henderson, 1973) using
the MTDFREML software. The value of an individual (related to a trait), judged by
the mean value of its progeny, is called the breeding value of an individual and is twice
the mean deviation of the progeny from the population mean (Falconer, 1983). To
calculate the genetic trend between 1975 and 1999, averages of breeding values of each
124                    S. Ruefenacht et al. / Applied Animal Behaviour Science 79 (2002) 113±132

year of birth were estimated. The trend was then characterized by ®tting a regression
curve to these averages.

3. Results

   The distributions of the grading scores of all behavioural traits were skewed (Fig. 1).
The most frequent score was 1 (48±87%). The overall score showed a skewed distribu-
tion too, since 37% of the dogs reached the top score of 7, 12% a score of 8 and 6% a score
of 9. The dog with the poorest grading obtained an overall score of 29 (distribution not
shown).
   The ®xed effects based on the REML analysis with the animal model are presented in
Table 5. In all analysed traits, the males reached a better score on average than the females.
The differences expressed in percent of the phenotypic standard deviation between gender
varied between 39% (overall score) and 18% (outcome). In all traits but reaction to gun®re,
temperament and hardness the average scores before 1990 were signi®cantly higher than
afterwards, the range of the differences lay between 37% (outcome) and 15% (overall
score). Highly signi®cant differences could be observed between judges. The largest
difference between two judges was observed for sharpness which amounted to 125% of the
phenotypic standard deviation; the largest difference for reaction to gun®re was only 45%.
All other largest differences lay between 68 and 95%. The effect of kennel was highly
signi®cant too (Table 4). As dogs got older the grading scores increased, i.e. they did worse
on average than younger dogs. This relationship followed a quadratic regression curve; for
an example, see Fig. 2. The average increase of the scores from 18 to 30 months of age
ranged between 15% (reaction to gun®re) and 38% (sharpness) of the phenotypic standard
deviation.
   The heritability estimates calculated by REML and with Gibbs sampling, by treating
the traits as quantitative ones, were the same or very similar (Table 6). They varied

Table 5
Fixed effects (%) of the phenotypic standard deviation of the corresponding trait

Traits                               Difference between

                                     Genders              Evaluation        The two                18 and 30
                                                          systema           extreme judges         months of age

Self-confidence                      34                   28                 93                    24
Nerve stability                      37                   29                 92                    28
Reaction to gunfire                  20                   ns                 45                    15
Temperament                          28                   ns                 68                    36
Hardness                             28                   ns                 71                    21
Defence drive                        31                   20                 89                    21
Fighting drive                       35                   ns                 77                    26
Sharpness desirability               26                   34                125                    38
Outcome                              18                   37                 82                    14
Overall score                        39                   15                 95                    31
      a
          ns: not significant.
S. Ruefenacht et al. / Applied Animal Behaviour Science 79 (2002) 113±132
Fig. 1. Distribution of the grading scores in the different behavioural traits: SC, self-confidence; NS, nerve stability; RG, reaction to gunfire; T, temperament; H,
hardness; DD, defence drive; FD, fighting drive; Sh, sharpness.

                                                                                                                                                                        125
126              S. Ruefenacht et al. / Applied Animal Behaviour Science 79 (2002) 113±132

Fig. 2. Effect of age on the score of self-confidence. The score increased from the age of 18 months ( 0.06) to
the age of 30 months (0.11) by 0.17, which corresponded to 24% of the phenotypic standard deviation.

between 0.09 (sharpness) and 0.24 (reaction to gun®re). The values for the outcome
and the overall score were 0.19/0.20 and 0.18/0.18, respectively. The heritability for
®ghting drive was estimated with records up to 1990 (N ˆ 1671) and reached values of
0.11/0.13. The standard error of the estimates ranged from 0.04 to 0.06. When traits
were treated as categorical ones, the heritabilities were generally the same or somewhat
larger but only in one case (reaction to gun®re) it was much larger than in the other
two methods.
   All correlations between the behavioural traits were positive and moderate to high
(0.31±1.0; Table 6). Genetic correlations were always larger than the corresponding
phenotypic correlations. The highest genetic correlation of 1.0 was observed between
self-con®dence and nerve stability. The others ranged between 0.66 and 0.90 except the
genetic correlations between sharpness and reaction to gun®re, temperament, as well as
defence drive (0.34±0.47). Since the grading of the behavioural traits was part of the
outcome and overall score, these correlations were not estimated (autocorrelation).
   Scores for hip dysplasia (FCI standard) were available from a part of the dogs
(N ˆ 1166). The phenotypic correlations between hip dysplasia and behavioural traits
were quite small ranging from 0.04 (temperament) to 0.01 (outcome).
   The effectiveness of the practised selection on behaviour during the last 25 years (1975±
1999) is presented in Fig. 3. All trends were signi®cantly negative, i.e. during this time,
behaviour patterns have been improved, although only moderately. The curvilinear
regressions were mainly caused by the fact that since the late 1980s the averages of
the breeding values remained stable or increased slightly. The regressions of self-con-
®dence, nerve stability and overall score ®tted the data best (R2: 0.74±0.81), followed by
reaction to gun®re and defence drive (R2: 0.54 and 0.47) and were low for temperament,
hardness and sharpness (R2: 0.33±0.28).

4. Discussion

   In this retrospective study of the last 25 years results of ®eld tests of the standardized
behaviour test were evaluated. The available German Shepherd dog (GSD) population
consisted of dogs which were intended to be registered as breeding animals. This material
is therefore not a representative sample of the whole GSD Swiss population since only
Table 6
Heritabilities of various behavioural traits (on the diagonal), phenotypic (above the diagonal) and genetic (below the diagonal) correlations among these traits in
German Shepherd dogs

                       Self-confidence      Nerve stability      Reaction to gunfire       Temperament       Hardness             Defence drive       Sharpness

                                                                                                                                                                      S. Ruefenacht et al. / Applied Animal Behaviour Science 79 (2002) 113±132
                           a
Self-confidence        0.18                 0.94                 0.47                      0.54              0.61                 0.50                0.56
                       0.18  0.05b
                       0.21  0.05c
Nerve stability        1.00                 0.18                 0.47                      0.52              0.59                 0.48                0.57
                                            0.18  0.05
                                            0.22  0.05
Reaction to gunfire    0.69                 0.72                 0.23                      0.32              0.41                 0.34                0.28
                                                                 0.24  0.06
                                                                 0.42  0.09
Temperament            0.77                 0.68                 0.80                      0.17              0.53                 0.53                0.31
                                                                                           0.18  0.05
                                                                                           0.17  0.05
Hardness               0.87                 0.79                 0.75                      0.67              0.14                 0.72                0.44
                                                                                                             0.15  0.05
                                                                                                             0.17  0.05
Defence drive          0.68                 0.66                 0.87                      0.81              0.90                 0.10                0.38
                                                                                                                                  0.11  0.05
                                                                                                                                  0.11  0.04
Sharpness              0.83                 0.83                 0.44                      0.34              0.77                 0.47                0.09
                                                                                                                                                      0.10  0.04
                                                                                                                                                      0.13  0.04
   a
     Heritability estimation with the REML method.
   b
     Heritability estimation with Bayesian analysis, Gibbs sampling as a quantitative trait.
   c
     Heritability estimation with Bayesian analysis, Gibbs sampling as a categorical trait.

                                                                                                                                                                      127
128             S. Ruefenacht et al. / Applied Animal Behaviour Science 79 (2002) 113±132

Fig. 3. Genetic trend of behavioural traits from 1975 until 1999: BVs, breeding values; R2, coefficient of
determination; P, significance of the fitted regression.

about 10% of an annual birth cohort attends the behaviour test. The selection of these 10%
is not random as the future dog owners choose the puppy according to the pedigree, the
kennel, the conformation, colour, behaviour and gender. The education of these young dogs
commonly includes puppies' school and various behaviour training courses. In addition,
most of the dogs attending the behaviour test were especially trained by experienced dogs
handlers. The preselection and the special training were not ideal conditions to estimate
S. Ruefenacht et al. / Applied Animal Behaviour Science 79 (2002) 113±132    129

genetic parameters. These handicaps, however, could be at least partially compensated by
many generations of records and the application of the animal model approach.
   Besides random sampling, it is most important that the recorded traits were accurately
and objectively measured. Even if behavioural traits are well de®ned, the grading of the
performance of a dog will always be subjective. In studies with several judges, an
additional non-genetic source of variation will affect the grading and in¯ate the total
variance. As suggested by Wilsson and Sundgren (1997b) it is important to keep the
number of people evaluating the dog's behaviour down to a minimum and to have routines
that regularly coordinate the standard of these experts. Under ®eld test conditions, as in the
present study, it was not possible to keep the number of judges small considering the time
interval (25 years) and the wide dispersion of the dog owners in Switzerland. In contrast to
P¯eiderer-HoÈgner (1979) who found no signi®cant differences between judges, the effect
of judges was highly signi®cant for all traits in our population even in spite of annual
training and continuing education of the judges.
   In all traits, scores increased signi®cantly when dogs got older. This is in contrast to
Wilsson and Sundgren (1997a) who found no age effect in dogs between 15 and 20 months
of age. P¯eiderer-HoÈgner (1979) was not able to show effects of age on behaviour scores in
11-month- to 4-year-old dogs. One possible explanation why older dogs did more badly in
all the behaviour traits is that judges were more rigorous when rating these dogs. Another
explanation could be that some dogs did not meet the standards to pass the test at a younger
age and their owners decided to lengthen the preparation period in order to reach the
required level. This practice is understandable because, according to the rules of the
breeding organisation, the behaviour test can be performed only once. If the dif®culty to
reach the level of education is an effect of the dog itself, then the model for the genetic
analyses had to be modi®ed by removing the age effect from the model. In that case, the
heritabilities should increase if the mentioned hypothesis is true. In fact, the heritabilities
calculated with the reduced model were equal (sharpness) or 0.01±0.04 larger than the ones
with the full model. Although the differences between the two values were not signi®cant,
it cannot be excluded that at least some dogs needed more time to be ready for the test
because they were (genetically) less quali®ed.
   One reason for the differences between genders may be the sex-related difference in
social behaviour of female and male dogs (Wells and Hepper, 1999; Pal et al., 1999).
Another reason could be that males, which are more often kept for working purposes than
females (P¯eiderer-HoÈgner, 1979), received different training and had other experiences
than dogs kept for breeding or as companion or guide dog.
   Differences between genders may also be affected by the choice of the behavioural traits.
In the applied behaviour test, only general and guard dog traits were rated but no traits for
companion, hunting or guide dogs. In their study with Labradors, Goddard and Beilharz
(1982) found that males were less often rejected for fearfulness and excitability in the
training program than females but more for distraction. Wilsson and Sundgren (1997a)
showed that GSD and Labrador males were better in courage, prey drive and defence drive.
However, they found breed dependent sex differences, e.g. males scored better in nerve
stability in GSD but there was no signi®cant difference in Labradors. In ability to cooperate
GSD males were better than GSD females, but it was opposite in Labradors, where females
scored better.
130           S. Ruefenacht et al. / Applied Animal Behaviour Science 79 (2002) 113±132

   Since the number of progeny per dam was low (a mean of 2.2 progeny), maternal effects
were not investigated. Furthermore, these effects would probably be completely con-
founded with the kennel effects. Wilsson and Sundgren (1998) mentioned that maternal
effects are most likely to be found in behaviour characteristics tested at juvenile age and are
small or negligible in dogs more than 1 year old.
   The heritabilities estimated by the different evaluation methods were very similar,
which indicate that this genetic parameter was reliably estimated. One exception was
observed for reaction to gun®re where the heritability calculated with Gibbs sampling by
treating the trait as a categorical one, was much higher (0.42) than with the two other
methods (0.23 and 0.24). This discrepancy may be due to the nearly binary character of the
trait with 88% of the dogs with the best score of 1 and 10% with a score of 2. The estimated
heritabilities were moderate to low, i.e. one cannot expect spectacular genetic improve-
ment. Although comparisons between heritability values have to be carried out with care
because of differences between de®nitions of the traits, sampling procedures, methods of
evaluations, breeds, etc. most of our estimates were similar to those published in the
literature (Table 1). Our heritability of self-con®dence (0.18) was very close to the one of
Bartlett (1976) (cited in Mackenzie et al., 1986) in American guide dogs (0.16); the same
was valid for nerve stability (0.18) where Wilsson and Sundgren (1997a,b) found values of
0.25 and 0.17 for GSD and Labrador, respectively. The latter authors estimated herit-
abilities for temperament of 0.15 and 0.10 for GSD and Labradors whereas our estimate
was 0.17; Mackenzie et al. (1985), however, found a much higher value of 0.51 in GSD.
Wilsson and Sundgren (1997a,b) obtained similar values for hardness (0.15 and 0.20) and
sharpness (0.13 and 0.11) in GSD and Labradors as in our study (0.15 and 0.09,
respectively) but their heritability estimate for defence drive (0.20 and 0.22) is larger
than our value (0.10).
   All correlations among traits were positive and showed a moderate to high relationship,
i.e. all these traits could have, at least to some extent, a common genetic background. This
situation facilitates the breeding work because no antagonistic correlations were found.
The genetic correlation of 1 between self-con®dence and nerve stability means that the
same trait is recorded twice. By looking at the de®nition and the scoring system of self-
con®dence and nerve stability, it becomes clear why both traits are so closely related. One
of them could easily be removed from the behaviour test without affecting the ef®ciency of
the breeding program.
   Goddard and Beilharz (1982) found genetic correlations of 0.89 between nervousness
and sound shy compared with 0.72 between nerve stability and reaction to gun®re in our
study. Wilsson and Sundgren (1997a) obtained 0.07±0.31 smaller phenotypic correlations
among comparable traits, except the relationship between defence drive and sharpness,
which was larger (0.46 and 0.38, respectively). The largest observed difference was
between nerve stability and temperament where they found a correlation of 0.03 against
0.52 in our study. These substantial differences between both studies can arise from the
different de®nitions of the traits, from the different sampling procedures and from the
different methods of evaluation.
   In contrast to Mackenzie et al. (1985), who estimated a phenotypic correlation of 0.31
between hip dysplasia and temperament, no relationship was found between hip dysplasia
and any of the tested traits in our study ( 0.04 to 0.01).
S. Ruefenacht et al. / Applied Animal Behaviour Science 79 (2002) 113±132                   131

   The modest genetic improvement over the last 25 years in the studbook population of the
GSD was partly due to the low heritabilities of the traits, but mainly because of the low
selection intensities after the behaviour test (only 8% failed). To be more rigorous at
selection is not an evident task, if we keep in mind that other traits had to be considered in
the breeding program such as hip dysplasia, body conformation and working tests.
Selection could be more rigorous by increasing the number of dogs that attend the
behaviour test. A further improvement would be to select the dogs not only on their
personal achieved results, but on their breeding values where their own performances and
those of all relatives are combined. Modi®cations of the grading system could contribute to
better results too. The actual grading is at the same time a rating by giving a score of 1 for
the most desirable behaviour. This leads to a quite skewed distribution of the traits, which
make genetic evaluation more dif®cult. Instead of classifying the dog's performance on a
desirability scale, it could be done on a descriptive scale, e.g. for sharpness: reduced,
slightly reduced, optimal, slightly too much, too much. The distribution would become
more symmetrical and easier to evaluate.
   An other open question is, if it would not be more ef®cient to split the actual behaviour
test in two distinct parts: one for the general behaviour where dogs' performances are
focused on their ability to behave in today's world without endangering or annoying people
or other dogs, and another one for working purposes as defence drive and sharpness.

References

Boldman, K.G., Kriese, L.A., van Vleck, L.D., Kachman, S.D., 1995. A Manual for Use of MTDFREMLÐA
   Set of Programs to Obtain Estimates of Variances and Covariances. US Department of Agriculture,
   Agricultural Research Service, University of Nebraska, Lincoln.
Falconer, D.S., 1983. Introduction to Quantitative Genetics, 2nd ed. Essex, England.
Goddard, M.E., Beilharz, R.G., 1982. Genetics of traits which determine the suitability of dogs as guide-dogs for
   the blind. Appl. Anim. Ethol. 9, 299±315.
Henderson, C.R., 1973. Sire evaluation and genetic trend. In: Proceedings of the Anim. Breed. Genet. in honour
   of J.L. Lush. ASAS and ADSA, Champaign, IL, pp. 10±41.
Hoeschelle, I., Tier, B., 1995. Estimation of variance components of threshold characters by marginal posterior
   modes and means via Gibbs sampling. Genet. Sel. Evol. 27, 519±540.
Hruby, A., 1991. Populationsgenetische Untersuchungen von Leistungs- und Wesensmerkmalen bei
   Gebrauchshunden. Vet. Med. Diss., University of Veterinary Medicine, Vienna.
Janss, L.L.G., 1998. MAGGIC: a package of subroutines for genetic analysis with Gibbs sampling. In:
   Proceedings of the Sixth World Congress on Genet. Appl. Livest. Prod., vol. 27. UNE, Armidale, NSW,
   Australia, pp. 459±460.
Ledger, R.A., Baxter, M.R., 1997. The development of a validated test to assess the temperament of dogs in a
   rescue shelter. In: Proceedings of the First International Conference on Veterinary Behavourial Medicine,
   Birmingham, UK, pp. 87±92.
Lower Saxony, 2002. Federal State of Germany, Law/Regulation About Dangerous Animals (Gefahrtier-
   VerordnungÐGefTVO), http://www.ml.niedersachsen.de/wesenstest.htm (accessed25 February 2002).
Mackenzie, S.A., Oltenacu, E.A.B., Leighton, E., 1985. Heritability estimate for temperament scores in German
   Shepherd dogs and its genetic correlation with hip dysplasia. Behav. Genet. 15, 475±482.
Mackenzie, S.A., Oltenacu, E.A.B., Houpt, K.A., 1986. Canine behavioural geneticsÐa review. Appl. Anim.
   Behav. Sci. 15, 365±395.
Moreno, C., Sorensen, D., GarcõÂa-CorteÂs, L.A., Varona, L., Altarriba, J., 1997. On biased inferences about
   variance components in the binary threshold model. Genet. Sel. Evol. 29, 145±160.
132              S. Ruefenacht et al. / Applied Animal Behaviour Science 79 (2002) 113±132

Mrode, R.A., 1996. Linear Models for the Prediction of Animal Breeding Values. CAB International, Oxon, UK,
    pp. 38±40.
Pal, S.K., Ghosh, B., Roy, S., 1999. Inter- and intra-sexual behaviour of free-ranging dogs (Canis familiaris).
    Appl. Anim. Behav. Sci. 62, 267±278.
Pfleiderer-HoÈgner, M., 1979. MoÈglichkeiten der ZuchtwertschaÈtzung beim Deutschen SchaÈferhund anhand der
    SchutzhundepruÈfung I. Vet. Med. Diss., Ludwig-Maximilian University of Munich.
Planta, D.J.U., Netto, W.J., 1997. Behavioural testing for aggression in the domestic dog. Appl. Anim. Behav.
    Sci. 52, 243±263.
Reuterwall, C., Ryman, N., 1973. An estimate of the magnitude of additive genetic variation of some mental
    characters in Alsatian dogs. Hereditas 73, 277±284.
Scott, J.P., Fuller, J.L., 1965. Genetics and the Social Behavior of Dog. The University of Chicago Press,
    Chicago.
Seiferle, E., Leonhardt, E., 1984. In: Erni, S., Druck, A.G. (Eds.), Wesensgrundlagen und WesenspruÈfung des
    Hundes, 2nd ed. Utznacherstr. 3, CH-8722 Kaltbrunn, Switzerland.
Slabbert, J.M., Odendaal, J.S.J., 1999. Early prediction of adult police dog efficiencyÐa longitudinal study.
    Appl. Anim. Behav. Sci. 64, 269±288.
Sorensen, D.A., Andersen, S., Gianola, D., Korsgaard, I., 1995. Bayesian inference in threshold models using
    Gibbs sampling. Genet. Sel. Evol. 27, 229±249.
van der Borg, J.A.M., Netto, W.J., Planta, D.J.U., 1991. Behavioural testing of dogs in animal shelters to predict
    problem behaviour. Appl. Anim. Behav. Sci. 32, 237±251.
Venzl, E., Unshelm, J., Oldigs, B., 1989. Der Campell-test: eine WesenspruÈfung beim Hund, durchgefuÈhrt an der
    Rasse Beagle. Kuratorium fuÈr Technik und Bauwesen in der Landwirtschaft, Deutsche VeterinaÈrmedizi-
    nische Gesellschaft (Eds.), Aktuelle Arbeiten zur artgemaÈssen Tierhaltung 1989. Darmstadt Giessen,
    pp. 134±152.
Wang, C.S., Rutledge, J.J., Gianola, D., 1994. Bayesian analysis of mixed linear models via Gibbs sampling with
    an application to litter size in Iberian pigs. Genet. Sel. Evol. 26, 91±115.
Weiss, E., Greenberg, G., 1997. Service dog selection tests: effectiveness for dogs from animal shelters. Appl.
    Anim. Behav. Sci. 53, 297±308.
Wells, D.L., Hepper, P.G., 1999. Male and female dogs respond differently to men and women. Appl. Anim.
    Behav. Sci. 61, 341±349.
Wilsson, E., Sundgren, P.-E., 1997a. The use of a behaviour test for the selection of dogs for service and
    breeding. I. Method of testing and evaluating test results in the adult dog, demands on different kinds of
    service dogs, sex and breed differences. Appl. Anim. Behav. Sci. 53, 279±295.
Wilsson, E., Sundgren, P.-E., 1997b. The use of a behaviour test for the selection of dogs for service and
    breeding. II. Heritability for tested parameters and effect of selection based on service dog characteristics.
    Appl. Anim. Behav. Sci. 54, 233±239.
Wilsson, E., Sundgren, P.-E., 1998. Behaviour test for eight-week-old puppiesÐheritabilities of tested behaviour
    traits and its correspondence to later behaviour. Appl. Anim. Behav. Sci. 58, 151±162.
Wright, S., 1934. An analysis of variability in number of digits in an inbred strain of guinea pigs. Genetics 19,
    506±536.
You can also read