Measuring herding behavior in Border colliedeffect of protocol structure on usefulness for selection

Transcription

1 Journal of Veterinary Behavior (2013) 8, 9-18 RESEARCH Measuring herding behavior in Border colliedeffect of protocol structure on usefulness for selection Per Arvelius a, Sofia Malm a, Kenth Svartberg b, Erling Strandberg a a Department of Animal Breeding and Genetics, Swedish University of Agricultural Sciences, Uppsala, Sweden; and b Department of Anatomy and Physiology, Swedish University of Agricultural Sciences, Uppsala, Sweden. KEYWORDS: Border collie; herding trait; heritability; selection Abstract From 1989 until 2003, the Swedish Sheepdog Society used a standardized method, Herding Trait Characterization (HTC), to describe herding behavior of individual Border collies for breeding purposes. In the HTC, which existed in 2 consecutive versions, the dog s herding behavior was described using predefined scales typically consisting of 6 classes within each trait. In total, 1,663 dogs participated in the first version and 951 in the second. The main difference between the versions was the structure of the protocols used to record the traits. In the first version, the scales were designed to describe the intensity in the expression of the measured traits. In the second version, the most desirable behavioral expression was placed in the middle of the scale. Another difference was that in the second version, the judges were given more freedom for their own interpretations, that is, the scales were more subjective. The objective of this study was to examine the quality of each version from a breeding perspective, and to analyze the reasons for possible differences in this respect between them. Heritability estimates for the 17 traits of the earlier version of the HTC ranged from 0.14 to 0.50 (weighted average: 0.30), all significantly different from 0. Corresponding heritabilities for the 19 traits in the later version were substantially lower ( , weighted average: 0.16), 3 of them not being significantly different from 0. Owing to the moderate-to-high heritabilities for the traits measured in the earlier version of the HTC, it would be possible to accomplish effective selection of breeding animals for most of the measured traits. It is plausible that the less neutral and more subjective protocol of the later version is the main cause for the lower heritabilities. Ó 2013 Elsevier Inc. All rights reserved. Introduction Herding Trait Characterizationddesigned to be a tool for Border collie breeders Present address: Kenth Svartberg, Svartbergs Hundkunskap, Alunda, Sweden. Address for reprint requests and correspondence: Per Arvelius, MSc, Department of Animal Breeding and Genetics, Swedish University of Agricultural Sciences, PO Box 7023, Uppsala, Sweden; Tel: ; Fax: per.arvelius@slu.se The Swedish dog population has been estimated to contain just more than 800,000 animals, but only very few of them are used for sheep herding (Egenvall et al., 1999). The skills of the sheepdogs are of importance not only for the herder but also (from an animal welfare perspective) for the sheep. These skills are expected to have some genetic /$ - see front matter Ó 2013 Elsevier Inc. All rights reserved. doi: /j.jveb

2 10 Journal of Veterinary Behavior, Vol 8, No 1, January/February 2013 background, that is, they can be modified by breeding. Successful breeding requires that the genetically most suited individuals can be identified and selected. In Sweden, sheepdog trials have played an important role when selecting sheepdogs for breeding. However, it is not self-evident that dogs being successful in trials also are the ones best suited for farm work, neither from a phenotypic nor from a genetic point of view. Among other things, this led the Swedish Sheepdog Society (SSS; Svenska Vallhundsklubben in Swedish), the breed club in Sweden for Border collie, to initiate the development of the so-called Herding Trait Characterization (HTC; Arbetsbeskrivning in Swedish) in the late 1980s (Frick, personal communication; Pyk, 2004, personal communication). In contrast to traditional sheepdog trials, the HTC was a noncompetitive method to describe how individual dogs typically expressed a number of traits considered important for herding ability. Hence, no winner was nominated among the participating dogs. The HTC existed in 2 consecutive versions. Version 1 was in use from 1989 to 1995, and version 2 was in use from 1996 to In the first version, 17 traits were measured, of which 12 can be regarded as herding traits. The second version contained 19 traits, 14 being herding traits. Most traits in version 1 are also in version 2, but some have been excluded and new ones included. Both versions had the same purpose: selection for better sheepdogs. The main difference between them was the structure of the protocols used to record the traits; the predefined grading alternatives are almost always different between the 2 versions. In the first version, large effort was put into creating scales according to intensity, that is, the intent was to measure the intensity or the strength of the expression of a trait. The descriptions used for the various grades were written not to be interpreted as good or bad (Frick, 2004). When changing the protocol in 1996, another philosophy was followed in which what was considered the most desirable expression of a trait was placed in the middle of the scale. The intention behind this was that the assessed dogs should be normally distributed around the center of the scale (Holgersson, 2001). Another difference was that in the second version, the judges were given more freedom for their own interpretations, that is, the scales were more subjective. Effective breeding for temperament in dogs requires good measurement methods The heritability of a trait is a significant parameter to describe the quality of the measurement method, in terms of how well it captures genetic differences between individuals. The higher the heritability, the greater part of the differences between individuals has genetic background. Consequently, heritability is an important measure of the potential for, say, a temperament test to function as a basis for selection. Measurements of temperament typically show low heritabilities compared with, for example, some morphological measurements such as osteochondrosis, hip dysplasia, withers height, or body weight (Nicholas, 2010; Viklund et al., 2008). One reason for this is that temperament is difficult to measure; the greater the measurement error, the smaller part of total variance will be due to genetic differences, resulting in a less useful test. Therefore, all else being equal, it is of importance to choose a measurement method that maximizes heritability when constructing a test for breeding purposes. Effective breeding for temperament in dogs requires that the measurements are utilized It is obvious, and has been stated in several studies, that breeding for temperament in dogs would be much more effective if modern methods for genetic evaluation were used more extensively (Correau and Langlois, 2005; Karjalainen et al. 1996; Liinamo et al., 1997; Ruefenacht et al., 2002; van der Waaij et al., 2008; Willis, 1995). This would enable selection based on estimated breeding values (EBVs), rather than on phenotypes as is the most common method today. An EBV is a statistical prediction of an individual s genetic contribution, for the measurements included in the EBV, to a given population if used for breeding. Among production animals, the most commonly used method today to estimate breeding values is mixed linear models resulting in best linear unbiased prediction (BLUP). When a so-called BLUP animal model is applied, all available information on relative s results contributes to an individual s EBV, which also will be adjusted for systematic environmental effects. EBVs are beneficial not only because they provide good measurements of an individual s potential genetic contribution to its offspring, but also because they are simple and straightforward and thereby easy to understand for both breeders and puppy buyers. As a consequence, introduction of EBVs can decrease the risk that breeders do not see the value of a test, and/or have difficulties interpreting the results and use them in a proper way. Prediction of EBVs also enables calculation of the genetic trend, yielding a more accurate reflection of genetic change compared with the phenotypic trend. The purpose with the HTC was to be used as a basis for selection, yet no breeding values for individual dogs were estimated and made available to the breeders. The situation is similar for many of the different temperament tests for dogs used around the world; genetic parameters are quite rarely estimated, and only in exceptional cases are breeding values put to use in practical breeding. Aims The aim of this study was to estimate heritabilities for the different traits included in both versions of the HTC, with 2 purposes: (1) to examine the quality of the HTC

3 Arvelius et al Measuring herding behavior in Border collie 11 from a breeding perspective, and (2) to enable a comparison of heritabilities between versions, including a discussion about possible differences, thereby gaining general insight into how temperament traits are best measured for breeding purposes in dogs. Material and methods Description of the HTC The HTC was carried out in connection with sheepdog training, most commonly performed at the introductory courses for sheepdogs that were organized by the local clubs of SSS. The person(s) executing the HTC described how the dog typically expressed various traits, and this expression was graded on a predefined scale. Hence, the HTC was not a set of standardized test situations, as is the case with, for example, the Dog Mentality Assessment (DMA) used by the Swedish Working Dogs Association (for a description of the DMA, see Svartberg and Forkman, 2002). The SSS recommended the characterization to be done by the course instructor together with the club s local breeding consultant and the dog owner. However, also characterizations done by either the instructor or the consultant (together with the owner) were accepted. Because the characterization usually involved the course instructor, he/she had the opportunity to observe and assess the dog over a long time before the actual recording was done (an introductory course usually consists of 7-10 sessions). Neither the instructor nor the consultant needed to have any education regarding the HTC to be allowed to perform it. However, to become an instructor, one had to participate in an education program for instructors as well as should have trained several sheepdogs. There were no limits as to the age of the dog being characterized; however, it was recommended that the age be less than 3 years. It was also emphasized that the dog should have received basic training in herding and should be able to cooperate with the handler; these were considered as prerequisites to make it possible to correctly assess the traits of interest (Holgersson, 2001). Unfortunately, there is no cohesive documentation about the HTC where the included traits are clearly defined. Using course materials, protocols for the HTC, and interviews with SSS breeding consultants (Frick, personal communication; Holgersson, 2004, personal communication; Pyk, 2004, personal communication; Svensson, 2004, personal communication), we have tried to describe the traits as good as possible (Tables 1 and 2). In both protocols, the scale in most cases consists of 6 classes within each trait. Typical behaviors of dogs at each end of the scale are included in Tables 1 and 2. When translating the trait labels from Swedish to English, we have strived to find the equivalents to the Swedish terms used within the English speaking herding community. However, such equivalents do not always exist. Data The analyses were based on all HTC results of Border collies, and pedigree data for these dogs. Most of the HTC data were contributed by the SSS but was complemented by some characterizations that were done before the first version was officially adopted. These unofficial characterizations, contributed by Dr. Lennart Swenson (Department of Animal Breeding and Genetics, Swedish University of Agricultural Sciences), were carried out in the same general way as the later official ones (Frick, 2004, personal communication). Pedigree data for the dogs were received from the Swedish Kennel Club. The total material consisted of 2,614 dogs, with a relatively even sex ratio and approximately 2 / 3 assessed using version 1 (Table 3). The proportion of intact versus neutered animals is not known; however, among Swedish dogs, in general, almost all dogs were intact during the period in question (Egenvall et al., 1999). In addition to the results for each dog, the HTC data also contained information about sex and year of assessment. A relatively small number of dogs were not assessed for all traitsd6 dogs in version 1 and 28 dogs in version 2. To use as many dogs as possible for each analysis, the number of dogs actually analyzed in a particular analysis may differ somewhat. Statistical analysis Results from the 2 protocols were analyzed separately. Initially, a descriptive analysis was done, where mean and standard deviation for each of the 17 traits in version 1 and 19 traits in version 2 were calculated. The only systematic environmental effects available for inclusion in the statistical model were sex and assessment year. These and their interaction were tested for significance in preliminary analyses (variance and regression analysis) using the procedure Proc GLM in SAS (SAS, 1999). Proc GLM uses the method of least squares to fit general linear models. In years 2002 and 2003, few dogs were assessed, and these were combined into one class. Based on the results of these preliminary analyses, the following mixed linear animal model was used for the estimation of variance components for each trait: y ijk 5m1sex i 1year j 1a k 1e ijk ð1þ where y ijk is the observation for the trait in question, m is overall mean, sex i is the fixed effect of sex, year j is the fixed effect of assessment year (j ,., 1995 or 1996,., 2002), a k is the random effect of breeding value of the animal k wnd(0, As 2 A ) (where A is the additive relationship matrix and s 2 A is the additive genetic variance), and e ijk is the random residual wind(0, s 2 E ) (where s2 E is the residual variance). The heritability for a trait was defined as h 2 5s 2 A =ðs2 A 1s2 E Þ.

4 Table 1 The 17 traits in version 1 of the HTC Trait Description of the trait Typical behaviors shown by dogs scored to either end of the scale 1. Affability (toward humans) How interested the dog is to interact with unknown persons. 0: Avoids contact, removes, and/or growls or bites. 5: Actively seeks contact with humans and, for example, jumps at them and/or licks them in the face. 2. Social behavior (toward dogs) The dog s tendency to seek confrontation with other dogs. 0: Goes away or lies on back. 5: Often seeks confrontation, that is, strives to fight. 3. Trainability without livestock How easily the dog learns while not in a herding situation. 0: Needs many rehearsals to learn and soon forgets what it has learned. 5: Learns new skills quickly. 4. Trainability with livestock How well the dog cooperates with the handler in a herding situation, that is, how keen it is to adjust its behavior according to the handler s desires. 0: Very difficult for the handler to make the dog listen/react when herding. 5: Easy to make contact with. 5. Ability to relax The dog s ability to relax while another dog is working. 0: Lies down and rests. 5: Whines, barks, and pulls the leash. 6. Balance The dog s ability to work in balance with the handler, that is, to take the position on the opposite side of the flock that affects the livestock to move toward the handler. 7. Natural working distance The distance to the livestock that the dog itself chooses to work at if not influenced by the handler. 8. Effective working distance The distance where the livestock become affected by the dog and start to move away. 9. Pace The gait and speed that the dog prefers when moving the livestock and how this changes depending on the flock behavior. 10. Natural ability The dog s ability to foresee and counteract the livestock s movements and thereby keep the flock together. 0: Must always be directed by the handler to find the right position. 5: Does always strive to place itself in the balance position. The dog is also meticulous and attentive to the handler. 0: 0-1 m. 5:.10 m. 0: 0-1 m. 5:.10 m. 0: Always gallop when moving livestock. 5: Never uses higher speed than walking pace when moving the livestock. 0: Often loses individual animals or the whole flock even if reminded when it is about to happen. 3: Never loses any animals, but places itself in the correct position in time to prevent this. 11. Focus How focused on its task the dog is in the herding situation. 0: Not interested, easily distracted, for example, by birds or other dogs. 5: Uninterruptedly focuses the livestock, regardless of what goes on in the surroundings. 12. Power The dog s ability to move livestock. 0: Cannot move livestock. 5: Moves threatening animals without support and from a long distance. 13. Oscillating movements How the dog moves from side to side in relation to the flock while working. 0: Powerful and heated movements. 5: Stalks without oscillating movements. 14. Eye How intensively the dog visually focuses the livestock. 0: Never fixates using eye. 5: Fixates, locks up until livestock move. 12 Journal of Veterinary Behavior, Vol 8, No 1, January/February 2013

5 Arvelius et al Measuring herding behavior in Border collie 13 0: No circle, straight through flock. 5: Makes a very wide circle, of the magnitude 35 m. 15. Out-run How wide circle the dog takes when moving from the handler to the balance point. Assessed at a position mid-between handler and balance point. 0: Increases speed, goes through the flock. 5: Stops, thereafter moves slowly toward the flock, alternatively 16. Lift How the dog behaves when the livestock start to get affected by the dog, that is, when at effective working distance. has troubles moving again. 0: Never bites, not even when provoked. 5: Often bites, also without having been provoked, and holds the grip. 17. Grip The dog s propensity for biting the livestock, where on the animals it bites and how it bites. For all genetic analyses the computer program DMU (Madsen and Jensen, 2000) was used, applying an AI-REML method (average information-restricted/residual maximum likelihood) for the estimation of variance components. Results and discussion Descriptive analysis Descriptive analysis of the traits showed that for most traits, the whole scale has been used in both versions of the HTC. Generally, the scales have been more extensively utilized in version 1; for all 14 comparable traits except 2 ( Lift vs. Lift and Oscillating movements vs. Oscillating movements [softness] ), the standard deviation was higher for the ones belonging to version 1 (Table 4). One reason for the higher variation in version 1 might be the reluctance to put dogs in the extreme categories in version 2 owing to the less neutral and more subjective descriptors of the scales in version 2. For instance, for effective working distance, a person judging might be less inclined to mark a dog in version 2 as 0: Fails to move the livestock regardless of distance or 5: Can move any animals, also from a long distance, than as having a distance of 0: 0-1 m or 5:.10 m in version 1. Similarly, some of the new traits in version 2 have words with very negative connotation, for example, for Courage, 0: fearful, cowardly or 5: lacks self preservation. Effect of sex and year For both sex and year, the preliminary analyses of these systematic environmental effects showed significant effects (P % 0.05) for roughly half of the traits in both versions of the HTC. Approximately 50% of the significant results had P % Therefore, both these effects were included in the mixed linear animal model (1) used for genetic analyses. The interaction was significant in only 1 case ( Eye in version 2). A plot of the combined main and interaction effects did not reveal any systematic trend; therefore, the interaction was not included in the genetic analyses (model [1]). Regarding sex differences, the general results from both versions indicated that male dogs were more confrontational with other dogs. Female dogs were more easily trained and cooperative with respect to their handlers. Male dogs could work at a longer distance from the livestock and still affect them. They also were less fearful of the livestock and used Oscillating movements more often, whereas female dogs showed more Eye. Looking at trends over time, in both versions of the HTC, both sexes were judged as showing more and more Eye and wider Out-runs. Male dogs also showed calmer Lifts and more Natural ability as the years passed by. Looking at trends over time for version 1 of the

6 Table 2 The 19 traits in version 2 of the HTC Trait Description of the trait Typical behaviors shown by dogs scored to either end of the scale 1. Affability (toward humans) Same as trait 1 in version 1. 0: Avoids contact, removes, and/or growls or bites. 5: Actively seeks contact with humans and for example jumps at them and/or licks them in the face, officious. 2. Social behavior (toward dogs) Same as trait 2 in version 1. 0: Bites in anxiety, very subordinate. 5: Aggressive toward both sexes. 3. Handler relation Similar to trait 3 in version 1, but more emphasis on how easy it is to make the dog subordinate (while not in a herding situation). 0: Very unwilling to absorb the handler s directives. 5: Always attentive and obedient toward the handler. 4. Cooperation Similar to trait 4 in version 1, but more emphasis on how easy it is to make the dog subordinate (while herding). 0: Seldom any reaction when the handler requests attention. 5: Always attentive and obedient toward the handler. 5. Oscillating movements (speed) Similar to trait 13 in version 1. 0: Lots of oscillating movements. 5: Moves (too) slow, lies down if the livestock stop. 6. Balance Same as trait 6 in version 1. 0: Moves the livestock away from the handler. 5: Difficult to make the dog work out of balance. 7. Natural working distance Same as trait 7 in version 1. 0: Works so close to the livestock that the speed increases until the herd is shattered. 5: Works too far away to make the livestock move. 8. Effective working distance Same as trait 8 in version 1, but in combination with the dog s ability to move livestock. 0: Fails to move the livestock regardless of distance. 5: Can move any animal, also from a long distance. 9. Oscillating movements (softness) Similar to trait 13 in version 1. 0: Heated and arrhythmic movements. 5: No oscillating movements. 10. Natural ability Same trait as trait 10 in version 1. 0: No anticipation ability, loses animals or the whole flock even if adverted when it is about to happen. 3: Extremely anticipating, becomes blocked and fails to move the livestock owing to the eagerness to counteract any tendency to change in direction. 11. Work ethic Describes the dog s will to work. 0: Often quits working, constantly has to be encouraged. 5: Excessive will to work, seeks tasks even when there is nothing to do. 12. Courage Describes how fearless the dog is when acting toward the livestock. 0: Fearful and cowardly, stops working if challenged or attacked, might bite nonthreatful animals in the rear end. 5: Lacks self-preservation, never backs down. 13. Style Describes the dog s body posture while working or watching livestock. 0: Unchanged posture. 5: Lowers the whole body to the extent that it crawls. 14. Eye Same as trait 14 in version 1. 0: Never fixates using eye. 5: Magnetic, locks up. 15. Out-run Same as trait 15 in version 1. 0: No circle, straight through flock. 5: Makes a too wide circle, looses contact with the livestock. 16. Lift Same as trait 16 in version 1. 0: Increases speed, goes through the flock. 5: Stops, thereafter moves slowly toward the flock, alternatively has troubles moving again. 14 Journal of Veterinary Behavior, Vol 8, No 1, January/February 2013

7 Arvelius et al Measuring herding behavior in Border collie Grip Similar to trait 17 in version 1. 0: Cannot be forced to bite. 5: Bites unprovoked, often and/or nasty. 18. Activity Describes the dog s ability to relax in everyday situations. 0: Difficulties to relax, nervous. 5: Always takes it easy, indolent. 19. Will to drive Describes the dog s will to move livestock. 0: Shows no interest to move the livestock. 5: Too much will to move livestock, hard to control the direction of drive. Table 3 Number of male and female dogs per year that participated in the HTC Year Male Female Total Version 1 of the HTC Sum ,663 Version 2 of the HTC Sum HTC alone, both sexes were judged as being able to move livestock from an increasing distance. Heritability estimates for the observed traits Estimated heritabilities for the traits in the 2 versions of the HTC are shown in Table 5. There is a clear tendency to higher heritabilities in version 1 than in version 2; for almost all comparisons of similar traits, the heritability was higher in version 1. The weighted averages of heritabilities across all traits were 0.30 and 0.16, respectively, and for the herding traits 0.34 and 0.20, respectively (the heritabilities were weighted by the inverse of their squared standard errors). Of the 17 traits in version 1, 5 had high (.0.4), 10 had medium high ( ), and 2 had low heritability, whereas of the 19 traits in version 2, only 1 had high heritability and 11 had low heritability. Three of these 11 were not significantly different from 0 (defined as.1,645! standard error, i.e., a 1-sided t test), whereas all estimates in version 1 were. The differences in heritabilities between the 2 versions of the HTC have implications for applied breeding. This can be illustrated by how much the genetic gain differs if selecting for a trait measured in version 1 versus the equivalent trait in version 2. For example, the first trait ( Affability ) showed a heritability of 0.40 when measured using version 1 of the HTC. When using version 2, the heritability was If selecting only on this trait and on phenotype alone, the accuracy of the breeding values becomes 0.63 and 0.41 (the square root of the heritability), respectively. All else being

8 16 Journal of Veterinary Behavior, Vol 8, No 1, January/February 2013 Table 4 Means and standard deviations (SD) for all traits in both versions of the HTC and for averages of subsamples of traits Version 1 Version 2 Trait Mean SD Trait Mean SD Affability (toward humans) Affability (toward humans) Social behavior (toward dogs) Social behavior (toward dogs) Trainability without livestock Handler relation Trainability with livestock Cooperation Ability to relax Balance Balance Natural working distance Natural working distance Effective working distance Effective working distance Pace Natural ability Natural ability Focus Power Oscillating movements Oscillating movements (speed) Oscillating movements (softness) Work ethic Courage Style Eye Eye Out-run Out-run Lift Lift Grip Grip Activity Will to drive Average for all traits Average for all traits Average for herding traits Average for herding traits Average for herding traits similarly defined between versions Traits with similar definitions between versions are placed side-by-side Average for herding traits similarly defined between versions equal, this means that the genetic gain for Affability will become 53% higher if the breeding animals are chosen by using the results from version 1. Expressed in another way, the progress in the breeding will be 53% faster. Similarly, for Social behavior (heritability: 0.16 and 0.03, respectively), the accuracy becomes 0.40 and 0.17, meaning a 135% higher genetic gain if the selection is based on version 1. On average, if making all 2 by 2 comparisons of traits that can be regarded as equivalent between versions, the genetic gain is 50% higher if selecting for a trait using version 1. A major reason for the higher heritabilities in version 1 of the HTC is most likely the differences in how the scales in the test protocols are designed. First, the main criterion when constructing the scales in version 1 was to arrange the classes according to intensity in reaction. In version 2, also desirable/undesirable behavior was included as criteria. This change can be expected to have made the scales less easily defined and therefore more difficult to use, resulting in decreased reliability. Second, the more objective scales in version 1 are likely to contribute to the simplicity of that protocol. For instance, the measures of Natural and Effective working distance in version 1 were based on objective length measured in meters, whereas in version 2, the scales were formulated as, for example, too far away and long distance. The resulting 2 heritabilities from version 1 were percentage units higher than those from version 2. Third, the less neutral ratings in version 2 might have had the effect that the judges became more reluctant to give dogs the worst grades. It is probably more difficult to remain objective if the protocol forces you to evaluate and tell the owner how good or bad a dog is, rather than just in a neutral manner describe its temperament traits or how prone it is to express different behaviors. This can be expected to have influenced the heritabilities in 2 ways. As discussed in the section Descriptive analysis, this is likely an important reason for the more extensive use of the whole scales in version 1, resulting in better prerequisites for high heritabilities. We did not do any analysis on interrater reliability, but it is plausible that this reliability decreases with a less neutral protocol. The reasoning behind this is that judges can be expected to differ in making the assessment based only on what the dogs actually do, without being influenced by irrelevant circumstances (such as the feelings of the dog owner). As a consequence, heritabilities would decrease.

9 Arvelius et al Measuring herding behavior in Border collie 17 Table 5 Estimates of heritability (h 2 ) and its standard error (SE) for the traits measured in version 1 and 2 of the HTC Version 1 Version 2 Trait h 2 SE Trait h 2 SE Affability (toward humans) Affability (toward humans) Social behavior (toward dogs) Social behavior (toward dogs) Trainability without livestock Handler relation Trainability with livestock Cooperation Ability to relax Balance Balance Natural working distance Natural working distance Effective working distance Effective working distance Pace Natural ability Natural ability Focus Power Oscillating movements Oscillating movements (speed) Oscillating movements (softness) Work ethic Courage Style Eye Eye Out-run Out-run Lift Lift Grip Grip Activity Will to drive Traits with similar definitions between versions are placed side-by-side. It is not possible to be certain that the difference between protocols is the only reason for the higher heritabilities for version 1. Two traits, Lift and Eye, were measured in almost the same way across the 2 versions. Nevertheless, the heritability in version 2 was lower for both of these traits. This might indicate that the general conscientiousness in doing the HTC declined over time, whichdif truedcan be expected to have caused increased random variation in the ratings. There is no observable evidence of this except that the interest for testing has deteriorated, as judged by the number of tested dogs per year (Table 3), and that the proportion of dogs with missing values for one or several traits has increased. The lower number of annually tested dogs for version 2 might also be an indication that a more selected portion of the population was observed at HTC, which could cause lower genetic variation. If valid, these circumstances would have lower heritabilities as a consequence. There are very few genetic studies on sheepdogs and herding traits. Based on 2,745 results of 337 Border collies, Hoffmann et al. (2003) estimated heritabilities for the various herding aspects from approximately 0 to Swenson (1983) used data from the predecessor of HTC (an examination after having completed a SSS herding course) and estimated a heritability for Obedience at 0.29, whereas all herding traits had heritabilities,0.2. Functionally, herding behavior is homologous to hunting behavior (Coppinger et al., 1987), which makes comparisons interesting. Most studies on hunting traits in dogs have reported lower estimates than those found for the HTC herding traits (Vangen and Klemetsdal, 1988; Karjalainen et al., 1996; Liinamo et al., 1997; Brenøe et al., 2002). The exception is one study on hunting traits for flat-coated retrievers (Lindberg et al., 2004). Schmutz and Schmutz (1998) also reported high heritabilities for different hunting traits, but most of them were not significantly different from 0. Because they used parent offspring regression, their results might not be directly comparable with ours. One reason for the comparatively high heritabilities in our study could be that the scores given to a dog was based on an evaluation over a long period. It is more common to have a test that takes place once, and therefore may be influenced much more by purely temporary factors. The HTC score is more to be considered as an average of several tests, although the averaging takes place inside the head of the person doing the scoring. Conclusions Our results suggest that when measuring behavioral traits for breeding purposes, one should consider objectivity of

10 18 Journal of Veterinary Behavior, Vol 8, No 1, January/February 2013 the scales. Furthermore, it is wise to be careful with passing value judgments in the test situation. Owing to the moderate-to-high heritabilities for the traits measured in the earlier version of the HTC, it would be possible to accomplish effective selection of breeding animals for most of the measured traits. However, this does not automatically mean that it is possible to conclude that the first version is a better tool to select for more talented sheepdogs. For example, validity remains to be analyzed; there might be a stronger correlation between measured traits and breeding goal traits in version 2 than in version 1. For several of the measurements, the heritabilities are high enough for selection based on phenotypic values to be reasonably successful. However, the accuracy of the breeding value would be higher if a linear mixed model was used to predict BLUP breeding values, including information from all relatives. This would require a centralized genetic evaluation, which is more costly than using the phenotypic values. Even so, we argue that the considerably better prerequisites for genetic progress (due to the higher accuracy) speak in favor of applying routine estimation of EBVs. Acknowledgments The authors thank Dr. Lennart Swenson for his suggestion to make genetic analyses on the HTC data. They also thank the members of the Swedish Sheepdog Society who answered our questions about the HTC, and assisted in finding appropriate translations to English of the Swedish terms in the test protocols: Pierre Frick, Susanne Gustafsson, Ylva Holgersson, Larry Jones, Mosse Magnusson, Ulf Pyk and Joel Svensson. The Swedish Sheepdog Society, Dr. Lennart Swenson, and The Swedish Kennel Club are acknowledged for the permission to use the data. References Brenøe, U.T., Larsgard, A.G., Johannessen, K.R., Uldal, S.H., Estimates of genetic parameters for hunting performance traits in three breeds of gun hunting dogs in Norway. Appl. Anim. Behav. Sci. 77, Correau, J.F., Langlois, B., Genetic parameters and environmental effects which characterise the defence ability of the Belgian shepherd dog. Appl. Anim. Behav. Sci. 91, Coppinger, R., Glendenning, J., Torop, E., Matthay, C., Sutherland, M., Smith, C., Degree of behavioral neoteny differentiates canid polymorphs. Ethology 75, Egenvall, A., Hedhammar, A., Bonnett, B.N., Olson, P., Survey of the Swedish dog population: age, gender, breed, location and enrolment in animal insurance. Acta. Vet. Scand. 40, Frick, P., Lokalprov (LP) och arbetsbeskrivning som avelsinstrument lite historik. Swedish Sheepdog Society webpage. Available at: Accessed February 17, Hoffmann, U., Hamann, H., Distl, O., Genetische Analyse von Merkmalen der Leistungsprüfung für Koppelgebrauchshunde. 1. Mitteilung: Leistungsmerkmale. Berl. Münch. Tierärzl. Wochenschr 116, Holgersson, Y., Arbetsbeskrivningen. Omkring 2, Karjalainen, L., Ojala, M., Vilva, V., Environmental effects and genetic parameters for measurements of hunting performance in the Finnish spitz. J. Anim. Breed. Genet. 113, Liinamo, A.E., Karjalainen, L., Ojala, M., Vilva, V., Estimates of genetic parameters and environmental effects for measures of hunting performance in Finnish hounds. J. Anim. Sci. 75, Lindberg, S., Strandberg, E., Swenson, L., Genetic analysis of hunting behavior in Swedish Flatcoated Retrievers. Appl. Anim. Behav. Sci. 88, Madsen, P., Jensen, J., A User s Guide to DMU A Package for Analysing Multivariate Mixed Models. Department of Animal Breeding and Genetics. Research Centre Foulum, Danish Institute of Agricultural Sciences (DIAS), Tjele, Denmark. Nicholas, F.W., Introduction to Veterinary Genetics, 3rd Ed. Wiley- Blackwell, West Sussex, UK, pp. 146, 225. Ruefenacht, S., Gebhardt-Henrich, S., Miyake, T., Gaillard, C., A behaviour test on German Shepherd dogs: heritability of seven different traits. Appl. Anim. Behav. Sci. 79, SAS, Release 8.2. SAS Institute Inc, Cary, NC. Schmutz, S.M., Schmutz, J.K., Heritability estimates of behaviors associated with hunting in dogs. J. Hered. 89, Svartberg, K., Forkman, B., Personality traits in the domestic dog (Canis familiaris). Appl. Anim. Behav. Sci. 79, Swenson, L., Vallprover som urvalsinstrument vid avel för bättre vallhundar. Department of Animal Breeding and Genetics, Swedish University of Agricultural Sciences, Uppsala. Van der Waaij, E.H., Wilsson, E., Strandberg, E., Genetic analysis of results of a Swedish behavior test on German Shepherd dogs and Labrador retrievers. J. Anim. Sci. 86, Vangen, O., Klemetsdal, G., Genetic studies of finnish and norwegian test results in 2 breeds of hunting dogs. In: Proceedings of the Sixth World Conference on Animal Production, Helsingfors, Finland, pp Viklund, A., Thorén Hellsten, E., Näsholm, A., Strandberg, E., Philipsson, J., Genetic parameters for traits evaluated at field tests of 3- and 4-year-old Swedish warmblood horses. Animal 2(12), Willis, M.B., Genetic aspects of dog behaviour with particular reference to working ability. In: Serpell, J. (Ed.), The Domestic Dog-its Evolution, Behaviour and Interactions with People. Cambridge University Press, Cambridge, pp