Evaluation of the Campbell test and the influence of age, sex, breed, and coat color on puppy behavioral responses

Abstract

The aim of this study was to evaluate the Campbell test and discover if there is a link between a puppy’s scores and factors such as age, breed, sex, sex-breed interaction, size, Fédération Cynologique Internationale (FCI) groups, and coat color. The Campbell test was performed on 342 puppies (191 males and 151 females) of different breeds. The results show that the criteria used by Campbell to classify puppies are incomplete, and that it is more appropriate to use numerical values for each type of answer. In general, the mean value obtained, regardless of sex and breed, corresponded to the Campbell’s submissive stable category. The mean value was higher in male dogs than in females.

Introduction

Potential dog owners are not only concerned about the physical appearance of a dog, but also about its adult personality. Numerous experts in veterinary clinical ethology, therefore, have developed complicated tests designed to predict a puppy’s character; Wilsson and Sundgren’s test is an example (1). Of all the various tests available, the Campbell test is the most popular and commonly used method of temperament testing in puppies. It is easy to administer, provides quick results for developing conclusions, and has been used in this study. Behavior tests, like the Campbell test, conducted on puppies, however, could be poor predictors of future behavior. Unlike Beaudet et al (2), the level of prediction of the Campbell test (3) is not determined here, but an attempt is made to assess the influence of many factors on puppies’ scores without Campbell’s interpretations of those behaviors. For this reason, these results should only be interpreted for puppies.

The most important objectives of our study are justified by the following statements:

1. Puppies evaluated using Campbell’s test tend to display high levels of dominance or high final scores (2).

2. Adult male dogs are more dominant than females (4–6), and certain breeds are more dominant than others (7,8).

3. Socialization in relation to a stimulus is adequate when the dog’s response is well-balanced. Enough interaction with that stimulus during the socialization process is the key (9). A lack of human socialization will cause a puppy to fear humans (10).

4. The color of a dog’s coat could be linked to the animal’s behavior (11–13) since English cocker spaniels having a single coat color display higher levels of aggression than those that are parti-colored (14,15,17); golden spaniels are the most dominant aggressive (14,17) displaying the highest scores in the Campbell’s test, followed by black-coated, and then parti-colored dogs (16).

Unlike Campbell, we cannot state that behavioral tendencies studied using the test are associated with dominance-submission tendencies in puppies (3). The final mean score obtained by a puppy for the various subtests, therefore, has a positive correlation to puppies’ dominance levels. For that reason, the terms “dominance/submission” have been used in the text strictly relying on Campbell’s personal description and interpretation of behaviors (3).

The Campbell test (3) was applied to various breeds of puppies between 6- and 8-wk old, to evaluate if:

1. Campbell’s groups are well-defined;

2. the puppies evaluated by this test tend to display high mean scores;

3. all 5 subtests are different or if they measure different characteristics, even though they are linked (correlated) to behavioral tendencies in the puppies. Behavioral tendencies probably result from a number of factors; therefore, it is impossible to measure these tendencies using a single subtest or a group of subtests that measure the same characteristics;

4. there is a link between a puppy’s mean score and factors such as age, breed, sex, and sex-breed interaction;

5. the extent to which inadequate human socialization or a total lack thereof might influence the test results and which of the subtests are most likely to detect puppies that have received inadequate human socialization. This group, puppies with inadequate human socialization, is defined using data collected from puppy owners. The litters studied, had had no human interaction;

6. there is link between the size of a puppy and its mean score;

7. there is any connection between the different groups of breeds established by the FCI (World Canine Association) and the puppy’s mean score; and

8. golden English cocker spaniels display the highest mean scores in the Campbell’s test, followed by black-coated and then parti-colored dogs.

Material and methods

Study sample

The Campbell test was carried out on a total of 342 puppies (191 male and 151 females), both purebred animals (American pitbull terrier, American Staffordshire terrier, Andalusian hound, bobtail, boxer, chihuahua, English bulldog, English cocker spaniel, Doberman, Argentine Dogo, fox terrier, German shepherd, golden retriever, Irish setter, standard poodle, pekingese, pointer, Presa Canario, rottweiler, Siberian husky, Spanish mastiff, Spanish water dog, and Yorkshire terrier) and mixed breeds, between 6- and 8-wk old. A minimum of 30 dogs from each breed studied were used in order to have a sufficient number of dogs to make breed comparisons and upon which to draw useful conclusions. The dogs were assessed individually and in accordance with the principles defined by Campbell (3). Puppies were randomly evaluated individually, during the day (8 am–6 pm) and in the same place (an 8-m × 4-m room), at the Department of Veterinary Medicine and Surgery (University of Cordoba). The length of the test was the same for each puppy (2 to 3 min). There was no other person, animal, or object in the room that could distract them. The test leader (TL) was always the same person (male) and had had no previous contact with the puppy. He remained impassive and showed no signs of emotion throughout the test. The TL observed how the puppies behaved in the various different sections of the test and classified them according to the Campbell test responses. A handler, who was the owner of the puppy, placed the puppy in test room.

The study was carried out over 4 years. All the litters were from different parents and from private homes. The recruitment list was selected from newspaper advertisements that listed puppies for sale. Owners were contacted by telephone and asked to participate in the study; 46% of the owners agreed to participate.

The English cocker spaniels were subdivided into 3 different sub-breeds (golden, black, and parti-colored) since several studies have observed a link between the color of an English cocker spaniel’s coat and its level of aggression.

The only litter studied that had not been in contact with humans and, therefore, lacked human socialization, was a litter of Andalusian hounds that came from a house in the country. This litter was studied independently of the other Andalusian hounds.

Description of the test

The Campbell test was performed in order to assess behavioral tendencies, according to Campbell associated to dominance-submission tendencies in puppies (3). The 5 parts of the test that must be conducted when puppies are between 6- and 8-wk old are as follows:

1. Social Attraction — the puppy is placed at one end of the room (or in the center of a particularly large room) facing the wall. The TL quickly moves in the opposite direction away from the puppy, kneels down and claps his hands to get the puppy’s attention. Possible responses include: a) the puppy comes readily, tail up, seeking contact with the TL; 2) the puppy comes readily, tail down, makes no contact with the TL; c) the puppy comes readily, tail down; d) the puppy comes hesitantly, tail down; e) the puppy does not come or runs away.

2. Following — The puppy is placed at one end of the room at the TL’s feet. The TL walks away in the opposite direction, making sure he gets the puppy’s attention. Possible responses include: a) the puppy follows the TL readily by his feet, tail up, trying to play; b) the puppy follows the TL readily by his feet, tail up; c) the puppy follows the TL readily, tail down; d) the puppy follows hesitantly; e) the puppy does not follow or runs away.

3. Restraint — The TL places the puppy on its back on the floor, holding the puppy down with one hand on its chest. Possible responses include: a) the puppy struggles vigorously, biting or growling, tail wagging; b) the puppy struggles vigorously, no biting or growling, tail wagging; c) the puppy struggles, then calms down; d) the puppy does not struggle and may lick the TL’s hands.

4. Social Dominance — The TL holds the puppy gently around its neck with one hand while stroking backwards along its neck and back for 30 s. Possible responses include: a) the puppy resists, growls and/or tries to bite; b) the puppy resists but does not demonstrate aggressive behavior; c) the puppy resists for only a short period of time; d) the puppy adopts a supine position; e) the puppy walks away and does not return.

5. Elevation Dominance — The TL picks up the puppy, holding it around its chest (placing his hands between its hind legs) a short distance above the floor. Possible responses include: a) the puppy struggles vigorously, growls and/or tries to bite; b) the puppy struggles vigorously but does not exhibit aggressive behavior; c) the puppy struggles, calms down and/or licks the TL’s hands; d) the puppy does not struggle and may lick the TL’s hands.

According to Campbell (3), the scores for the different parts of the test indicate: a) excessive dominance; b) dominance; c) balanced submission; d) excessive submission; e) independence or inadequate socialization. Responses during the test indicate the level of dominance displayed by the puppy: maximum dominance corresponds to (a); dominance to (b); balanced submission to (c); excessive submission to (d); and independence or excessive fear to (e). An independent dog may be defined as a dog with no need of human interaction.

The potential relationship between sex and breed were evaluated using Campbell’s classification criteria and the 5 categories he established (a — Excessively dominant; b — Dominant; c — Stable submissive, d — Excessively submissive; e — Independent, or with a lack of human socialization) depending on the number of answers given for each level of dominance (a, b, c, d, e). The test was carried out in the following sequence: 1) Social Attraction, 2) Following, 3) Restraint, 4) Social Dominance, and 5) Elevation Dominance.

Statistical analysis

Evaluation of each sex following Campbell’s criteria

The total number of answers of each type scored by each sex was calculated and divided by the total number of individuals of that sex. The average values for each type of answer scored for each sex was obtained to observe if there were any significant differences between the sexes and if these differences fit into any of Campbell’s groups. The average number of answers for each sex was divided by that of the other sex in order to calculate how often a sex obtains 1 type of answer over another.

Evaluation of each breed following Campbell’s criteria

The same procedure was followed for the breed factor. The number of answers that each breed scored for each type of response (a,b,c,d,e) was calculated. This figure was then divided by the total number of individuals in that breed. The number produced indicated the likelihood that a given breed would display certain behaviors, allowing them to be classified into 1 of the Campbell groups.

Study of the different factors by assigning numerical values to the different types of answers

For the statistical study, a score of between 4 and 0 was assigned to each Campbell subtest: 4 corresponded to maximum dominance (a); 3 to dominance (b); 2 to stable submission (c); 1 to excessive submission (d); and 0 to independence or excessive fear (e). Hence, each puppy obtained a final mean score for the various subtests. The higher the mean score, the more dominant the puppy, according to Campbell, although it was felt that these results should be depicted as the mean score for “behavioral tendencies” (without specifications) and not for “dominant-submission tendencies” according to Campbell’s interpretations.

Averaging the scores washes out relevant information recorded from the Campbell test categories, which are meant to be looked at in the context of consistency of individual scores. Nevertheless, averaging is necessary for a complete statistical study when many factors like this are involved.

Having previously applied the square root data transformation for normalization of variables, SAS (18) software was used to carry out the following statistical analyses:

1. Test factor — a one-way analysis of variance (ANOVA) was used to compare the results from the 5 subtests (5 levels per subtest factor), to find out if there were any significant differences among them. If significant differences were observed, Duncan’s test was applied to determine the number of statistically significant different groups.

2. To analyze breed, sex, sex-breed interaction, and litter factors, an analysis of covariance (ANCOVA) was carried out on adjusted factor means for the variable “age of the puppy”. If significant differences were observed, Student’s t-test and the least mean squares (LMS) algorithm were used in order to make inferences about factor means.

3. The dog’s size — an ANCOVA was performed in order to determine the influence of the adult animal’s size on the final mean score exhibited by the puppy.

4. The link between the score and the breed groups established by the World Canine Association (FCI = Fédération Cynologique Internationale) — an ANCOVA was performed to determine statistically significant differences for each factor and Student’s t-test and the LMS algorithm were applied to determine the number of statistically significant groups.

5. To detect puppies with inadequate human socialization, an ANCOVA, Student’s t-test, and the LMS algorithm were used; however, for this factor, the subtests were grouped into 2 blocks: one with subtests 1, 2, and 4, and another with subtests 3 and 5.

6. The influence of coat color in the English cocker spaniel — a one-way ANOVA was applied. If significant differences were observed, Student’s t-test and the LMS algorithm were applied to determine the number of statistically significant different groups. Since a total of 52 golden cocker spaniels and 42 nongolden-coated dogs (28 black and 14 parti-colored) were to be assessed, the study would first compare the Golden cocker spaniel with the rest of cockers and subsequently compare all the types of cockers.

Results

Evaluation of each sex in accordance with Campbell’s criteria

Table I shows the average number of answers of each type (a,b, c,d,e) obtained for each sex and breed. Each sex was compared and analyzed for the possibility of applying Campbell’s classification. The following results provide important data:

Table I.

Average number of answers of each type (a, b, c, d, e) obtained for each sex and breed

Sex	ā	b̄	c̄	d̄	ē	N
Male (M)	0.9666	1.2124	1.7038	0.8559	0.1359	460
Female (F)	0.1788	1.0529	2.0530	1.5761	0.1390	439
Sex Ratio	M/F = 5.4060	M/F = 1.1515	F/M = 1.2050	F/M = 1.8415	F/M = 1.0228
Breeds	ā	b̄	c̄	d̄	ē	N
A Pitbullte	0.7619	1.3810	2.0476	0.8095	0.0000	42
A.Staffor	0.7838	1.3243	1.8649	1.0270	0.0000	37
Bobtail	0.2000	1.2000	2.2000	1.4000	0.0000	30
Boxer	0.6667	1.0667	2.0000	1.0000	0.2667	30
Bulldog	0.25	1.0000	2.0000	1.7500	0.0000	32
Poodle	0.1667	0.5000	1.8333	2.0000	0.5000	30
Chihuahua	0.1429	0.7143	2.4286	1.7143	0.0000	35
Cocker	0.3333	1.0476	2.0000	1.6190	0.0000	42
Gold Cock	0.7308	1.4230	1.5000	1.2693	0.0769	52
Doberman	0.4286	1.7143	1.7143	1.1429	0.0000	35
Do Argen	0.8667	1.3333	1.8000	0.9333	0.0667	30
Foxterrier	0.5556	1.5556	2.0000	0.7778	0.1111	36
Golden	0.1429	0.5714	2.2857	2.0000	0.0000	35
Husky	0.0000	0.5000	2.1667	1.8333	0.5000	30
Span. mast	0.6400	1.0400	1.7600	1.2400	0.3200	50
Mixed	0.4000	1.5500	2.0000	1.0000	0.0500	40
Germ she	0.5455	1.1818	1.8636	1.3182	0.0909	44
Pekingese	0.7500	0.7500	2.2500	1.2500	0.0000	36
Span wate	0.2000	1.0000	2.0000	1.8000	0.0000	30
And houn	0.3125	1.4375	1.8750	1.3750	0.0000	32
And houn*	0.0000	0.3750	1.5000	0.5000	2.6250	8
Pointer	0.0000	1.0000	2.2000	1.8000	0.0000	30
Presa Can	0.7500	1.4375	1.7500	1.0625	0.0000	32
Rottweiler	0.8974	1.3077	1.6410	1.1282	0.0256	39
Irish setter	0.2500	1.3750	1.8750	1.5000	0.0000	32
Yorkshire	0.6667	1.1667	2.3333	0.8333	0.0000	30
Average	0.5819	1.1345	1.8743	1.2076	0.1374	899

Symbols ā–ē are the average number of answers of each type obtained for each sex and breed.

N = total number of puppies.

A Pitbullte — American Pit Bull Terrier, A Staffor — American Staffordshire Terrier, Cocker — English cocker spaniel, Do Argen — Dogo Argentino, Gold Cock — golden English cocker spaniel, Golden — golden retriever, Span mast — Spanish mastiff, Germ she — German shepherd, Span wate — Spanish water dog, And houn — Andalusian hound, And houn* — Andalusian hound not socialized, Presa Can — Presa Canario.

1. For excessive dominance (a) — males tend to obtain an average of 0.9666 subtests and females 0.1788. This means that the probability of displaying excessively dominant behavior is 5.4060 times higher in male dogs than in females.

2. For dominance (b) — males tend to obtain an average of 1.2124 subtests with this value and females 1.0529. This means that the probability of displaying dominant behavior is 1.1515 times higher in males than in females.

3. For stable submission (c) — males tend to obtain 1.7038 subtests with this value and females 2.0530. The probability of displaying this type of behavior, therefore, is 1.2050 times higher in females than in males.

4. For excessive submission (d) — males obtain an average of 0.8559 subtests with this value and females 1.5761. The probability of displaying this type of behavior is, therefore, 1.8415 times higher in females than in males.

5. For independence or inadequate human socialization (e), males display independent behavior in 0.1359 subtests and females in 0.1390. These values are very low, and statistically equal according to the ANOVA. It can, therefore, be concluded that there are no significant differences between males and females in terms of independent behavior and that dogs are generally social animals.

6. It does not seem possible to classify males and females into any of the groups Campbell defines in his classification.

The analysis indicates that males have a more marked dominant character than females and that, in general, neither males nor females tend to display independent behavior according to Campell’s categorization.

Evaluation of each breed in accordance with Campbell’s criteria

Table I summarizes the average number of answers for each type (a,b,c,d,e) obtained by each breed. The number of breeds that could be classified according to Campbell’s criteria were analyzed, and the following conclusions were developed:

1. There are no excessively dominant breeds, because no breed recorded 2 or more (a) answers along with several (b) answers, which is Campbell’s requirement for this classification (although certain dogs did obtain this score individually).

2. No dominant breeds were revealed in this study; none had 3 or more (b) answers.

3. No breeds were classified into the stable submissive group; none had 3 or more (c) answers.

4. No breeds fell into the excessive submission category; none had a majority of (d) answers, nor into the independence group [with 2 or more (e) answers]. The only exception was the Andalusian hound, which would be included in group (e), because of inadequate human socialization. It was the only litter studied, however, that had not been socialized with humans.

Although no breeds could be classified into Campbell’s groups, there are breeds in which the probability of displaying excessively dominant or dominant behavior is higher than other breeds.

Test results for the different factors

Subtest factor

When an ANOVA was performed for the different subtests, very different significant values (P < 0.0001) were observed. Table II shows the results of Duncan’s multiple-range test for comparisons of means, in which the 5 Campbell subtests were significantly different.

Table II.

Duncan’s test to subtest factor

Subtest	Average	N	Duncan group
1	2.96199	899	A
2	2.77778	899	B
3	2.11404	899	C
4	1.79532	899	D
5	1.23099	899	E

A, B, C, D, E are significantly different Duncan groups.

Influence of age, sex, breed, sex-breed interaction, dog size, and FCI groups on the results of the 5 subtests

A positive and linear relationship was observed between the age of the puppy and the scores obtained in the ANOVA and Duncan’s test: the older the puppy the higher the score (Table III).

Table III.

Duncan’s test to age factor

Age	Average	Duncan group
8 weeks	2.2832	A
7 weeks	2.0678	B
6 weeks	1.8314	C

A, B, C are significantly different Duncan’s groups.

The ANCOVA showed that all the factors studied are highly significant and that sex-breed interaction is also highly significant (P < 0.0001).

In relation to the dog’s sex, male puppies obtained a higher significant average than did females.

Table IV shows the clusters delineated by Student’s t-test to compare least mean squares. The significantly different groups of breeds were placed in decreasing order of score. The 1st classification of Table IV shows the order for all groups; the 2nd classification shows the results for males only; and the 3rd classification shows the results for females. An interaction between sex and breed is observed for the mean value of the results, because the breeds with the highest scores in males do not correspond to the breeds with the highest scores in females. The breeds with the highest scores, regardless of sex are: the American pitbull terrier, the Dogo Argentino, the rottweiler, the Presa Canario, the American Staffordshire terrier, the fox terrier, the Yorkshire terrier, and the Doberman. In the male group, the most dominant breeds are the rottweiler, the Doberman, the Yorkshire terrier, and the Presa Canario. In the female group, the most dominant breeds are the American pitbull terrier, the mixed breeds, the American Staffordshire terrier, and the fox terrier.

Table IV.

Student’s t-test of the least mean squares in breeds

Group	Breeds	Average	Group	Breeds — Males	Average	Group	Breeds — Females	Average
A	A. pitbull terrier	2.4190	A	Rottweiler	2.7455	A	A. pitbull terrier	2.2222
A	Dogo Argentino	2.4000	A	Doberman	2.6667	A	Mixed	2.1636
A	Rottweiler	2.3846	A	Yorkshire	2.6500	A	A. Staffordshire	2.1500
A	Presa Canario	2.3750	A	Presa Canario	2.6222	A	Foxterrier	2.0667
A	A. Staffordshire	2.3730	B	Dogo Argentino	2.5800	B	Presa Canario	2.0571
A	Foxterrier	2.3333	B	A. pitbull terrier	2.5667	B	Dogo Argentino	2.0400
A	Yorkshire	2.3333	B	Boxer	2.5500	B	Doberman	2.0000
A	Doberman	2.2857	B	A. Staffordshire	2.5429	B	Pekingese	2.0000
B	Mixed	2.2500	B	German shepherd	2.4909	B	Cocker	1.9798
B	Pekingese	2.2000	B	Foxterrier	2.4667	B	Rottweiler	1.9176
B	Boxer	2.1733	B	Pekingese	2.4000	B	Irish setter	1.9000
B	Cocker	2.1702	B	And. hound	2.3778	B	Spanish mastiff	1.8833
B	German shepherd	2.1545	B	Mixed	2.3556	B	And. hound	1.8286
B	And. hound	2.1375	B	Cocker	2.2848	B	German shepherd	1.8182
B	Spanish mastiff	2.0880	B	Spanish mastiff	2.2769	B	Spanish water dog	1.8000
B	Irish setter	2.0750	B	Bobtail	2.2667	B	Boxer	1.7429
B	Bobtail	2.0400	B	Irish setter	2.2500	B	Pointer	1.7333
B	Bulldog	1.9500	B	Bulldog	2.2000	B	Husky	1.7333
B	Spanish water dog	1.9200	C	Chihuahua	2.1333	B	Bulldog	1.7000
C	Chihuahua	1.8571	C	Pointer	2.1000	B	Bobtail	1.7000
C	Pointer	1.8400	C	Spanish water dog	2.0000	C	Yorkshire	1.7000
C	Golden	1.7714	C	Golden	1.9000	C	Chihuahua	1.6500
C	Poodle	1.5667	C	Caniche	1.8500	C	Golden	1.6000
C	Husky	1.5333	C	Husky	1.3333	C	Poodle	1.0000
D	And. hound*	0.9250	D	And. hound*	0.9600	D	And. hound*	0.8667

A,B,C,D are significantly different Student’s t-test groups.

^*And. hound — not socialized.

The ANCOVA performed for dog size indicated that there are no significant differences between small and large breeds of dogs. The adult size of the breed does not influence the final score observed in the puppy.

The ANCOVA revealed that the puppy’s FCI group was highly significant (P < 0.0001). Table VI shows the groups defined by applying Student’s t-test and the LMS algorithm to the different FCI groups; calculations were performed for the total groups as a whole, as well as males and females separately. Males obtained higher values, and the groups that displayed the highest values were the terriers (group 3), the molossers (group 2), and the mixed breeds.

Table VI.

Student’s t-test of the least mean squares for the FCI groups in all the puppies, males, and females

Group	FCI	Average	Group	FCI	Males average	Group	FCI	Females average
A	3	2.3781	A	2	2.5701	A	Mixed	2.16360
A	2	2.2777	A	3	2.5488	A	3	2.1333
A	Mixed	2.2500	A	1	2.4429	A	2	1.9148
B	1	2.1333	A	Mixed	2.3556	B	8	1.8870
B	8	2.1017	B	8	2.2389	B	1	1.8000
B	5	1.9273	B	5	2.1159	B	5	1.8000
C	7	1.8727	B	7	2.0800	C	7	1.7000
C	9	1.8353	B	9	2.0667	C	9	1.5750

A,B,C are Student’s t-test groups significantly different.

Mixed dogs have been included in the statistical analysis. FCI groups of breeds studied are:

1 — German shepherd

2 — Rottweiler, Spanish mastiff, Boxer, Doberman, Presa Canario, and Dogo Argentino

3 — American Staffordshire terrier, American pit bull terrier, Yorkshire terrier, and fox terrier

5 — Siberian husky and Andalusian hound

7 — Ireland setter and Pointer

8 — English cocker spaniel

9 — Poodle and Pekingese

Efficiency of grouping subtests 1–2–4 and 3–5 in order to group puppies that display inadequate human socialization

A litter of Andalusian hounds was the only litter studied that had not been in contact with humans and, therefore, lacked human socialization. This litter was studied independently of the other Andalusian hounds, because of this significant difference. Lack of human socialization in the Campbell test is revealed by a very low mean score, and this litter of Andalusian hounds obtained the lowest mean score (0.9250). The dogs tended to run away whenever they had that option (subtests 1, 2, and 4), thereby obtaining the minimum score (0 points) for most subtests. In subtests 3 and 5, where the option of running away was not available to them, their response was aggressive; therefore, this breed obtained higher scores than other dogs for these subtests. The aggressive response was caused by fear, since they displayed typical fearful motivated postures: lowering the head, flattening the ears back against the head, tucking the tail between the legs, and avoiding eye contact.

Table V shows the breed clusters delineated by Student’s t-test and the LMS algorithm for subtests 1–2–4 and 3–5. In relation to subtests 1–2–4, there is a great difference between the score of the Andalusian hounds that had not been in contact with humans (the lowest) and the other dogs assessed, to such an extent that this is the only “breed or group” that can be included in group (C). In subtests 3–5, the Andalusian hounds that had not been in contact with humans, once again form a group that is totally separate from the other dogs. In this case, their mean score is higher than the other dogs, although the differences are not as marked as those for subtests 1–2–4, since highly dominant puppies that have been socialized will also tend to obtain high scores in subtests 3–5.

Table V.

Student’s t-test of the least square means in breeds for the subtests 1–2–4 and 3–5

Group	Subtests 1–2–4 Breed	Average	Group	Subtests 3–5 Breed	Average
A	A. pitbull terrier	2.9048	A	And. hound	2.1250
A	Dogo Argentino	2.8889	B	Yorkshire	2.0000
A	Presa Canario	2.8750	B	Rottweiler	1.8205
A	A. Staffordshire	2.8288	B	Mixed	1.7750
A	Rottweiler	2.7607	B	Foxterrier	1.7222
A	Foxterrier	2.7407	B	A. pitbull terrier	1.6905
A	Doberman	2.7143	C	A. Staffordshire	1.6892
A	Pekingese	2.6667	C	German shepherd	1.6818
A	Boxer	2.5778	C	Cocker	1.6703
B	Mixed	2.5667	C	Dogo Argentino	1.6667
B	Yorkshire	2.5556	C	Irish setter	1.6563
B	Cocker	2.5035	C	Doberman	1.6429
B	And. hound	2.5000	C	Presa Canario	1.6250
B	German shepherd	2.4697	C	And. hound	1.5938
B	Spanish mastiff	2.4400	C	Boxer	1.5667
B	Bobtail	2.4000	C	Spanish mastiff	1.5600
B	Irish setter	2.3542	C	Bobtail	1.5000
B	Bulldog	2.2500	C	Husky	1.5000
B	Spanish water dog	2.2000	C	Bulldog	1.5000
B	Chihuahua	2.0952	C	Pekingese	1.5000
B	Pointer	2.0667	C	Spanish water dog	1.5000
B	Golden	1.9524	C	Golden	1.5000
B	Poodle	1.6111	C	Poodle	1.5000
B	Husky	1.5556	C	Chihuahua	1.5000
C	And. hound*	0.1250	C	Pointer	1.5000

A,B,C are significantly different Student’s t-test groups.

^*And. hound — not socialized.

Influence of coat color: golden, black, and parti-colored

The ANOVA performed using a square root transformation, revealed a significant difference between golden-coated and nongolden-coated dogs, and also between the 3 different types of coats (P < 0.0001).

Student’s t-test and the LMS algorithm revealed significant differences: animals with golden coats displayed higher scores than those with other coat colors (Table VII); and, when comparing the 3 types of coats, the highest mean scores, in decreasing order, were golden, black, and parti-colored (Table VII).

Table VII.

Student’s t -test of the least mean squares in English cocker spaniel

Duncan group	Coat color	Average
A	Golden	2.2923
B	No golden	2.0190
B	Black	2.0967
C	Parti-colored	1.8636

A,B,C are significantly different Student’s t-test groups.

Discussion

Evaluation of each sex and breed in accordance with Campbell’s criteria

Although the results clearly reveal that, according to Campbell’s interpretation, males are more dominant than females and that certain breeds are more dominant than others, none of these groups could be classified into Campbell’s groups (3). Campbell’s categories seem to be too rigid and not adequately defined. Given this preliminary conclusion, it seems that a more accurate picture could be obtained if numerical values were assigned to each type of answer. It would then be possible to obtain a final score that would quantify character, and allow for better grouping. It would also be possible to study other factors and statistical variables.

Different factors

The 5 Campbell subtests might be significantly different because they measure different behavioral tendencies, even though they are linked (correlated) to a dominance/submission factor.

The 1st subtest obtained the highest values, followed by the 2nd subtest, and then the 3rd. This might be because the puppy becomes familiar with the test situations and is, therefore, able respond to the tests giving higher values. This may also be possible because the puppy assumes that these test situations are less dangerous and is therefore not as frightened (fear can modify or even block the way in which the puppy responds, possibly making the animal more cautious). The puppy, therefore, may behave in a more natural way compared to the last 2 subtests. In particular, this statement is supported by the last subtest (Elevation Dominance), which resulted in the lowest score (1.2300). Picking the puppy up might make it feel powerless and when faced with this new circumstance, it may respond by keeping still (value 1 = d). This was the most frequently observed response, in contrast to subtest number 4, which had the 2nd lowest average (1.7953).

Age is an important factor since within the range of 6- to 8-wk, it was clear that the older dogs were more alert and therefore tended to respond with higher scores. For this reason the scores of other factors for the “Age” variable were adjusted using an ANCOVA.

The results indicate that regardless of sex and breed, the score obtained by most dogs corresponds to the Submissive Stable group. This differs from the findings recorded by Beaudet et al (2) who reported that the puppies in their study obtained high levels of dominance at 7-wk of age. However, Beaudet et al (2) found that at 16-wk of age, puppies displayed a tendency towards regression (lower levels of dominance).

Other studies with adult dogs confirmed that males are more dominant than females (10–12). In this study, male puppies were observed to have a higher mean than females. This suggests that “behavioral tendencies” analyzed using Campbell’s test, associated or not associated with dominance, already have higher scores when the dog is a puppy.

Efficiency of grouping subtests 1–2–4 and 3–5 in order to group puppies that display inadequate human socialization

Although Campbell has never validated his test and interpretations associated to dominance-submission, an agreement was found with the (e) interpretation (independence or lack of human socialization: “the puppy ran away or did not follow the test leader”) in this study since the puppies with inadequate human socialization ran away. Similarly, an independent dog was defined as a dog with no need of human interaction; therefore, these dogs have no interest in following the test leader.

These results might indicate that grouping subtests 1–2–4 and 3–5 could be a very useful way of detecting puppies with inadequate human socialization, although data collected from puppy owners could be decisive. According to Campbell, the option (e) can indicate either inadequate human socialization or independence, and grouping the subtests may allow clear differentiation between independent puppies and puppies with inadequate human socialization. It was observed that independent puppies obtained low scores for subtests 1–2–4, but unlike under-socialized puppies that ran away, independent animals showed no interest in the examiner. Independent puppies will not necessarily score highly in subtests 3–5, whereas puppies with inadequate human socialization will. This was the case, for example, with huskies, which had been socialized but obtained the 2nd lowest score for subtests 1–2–4 (the lowest was that of the Andalusian hounds that had not been in contact with humans).

An independent puppy will obtain a low score due to the influence of subtests 1–2–4; because of this, a better evaluation of “behavioral tendencies” would be obtained using subtests 3–5. Out of all the breeds of dogs evaluated, the most independent were the huskies; the only breed of spitz dog studied. This could be because this breed has only recently been domesticated and has a marked phylogenetic similarity to its relative the wolf.

Influence of coat color: golden, black, and parti-colored

Several studies on English cocker spaniels have observed a link between the animals’ coat color and levels of aggression (14,15,17). The coat colors listed in decreasing order for displays of aggression are: golden, black, and parti-colored (14,17). This order has been shown to be the same for the scores in the Campbell’s test (16), and the results herein.

Conclusions

According to Campbell’s classification

1. The procedure based on assigning numerical values to the different types of answers is a quick, easy, and accurate way of classifying traits of dominance-submission in puppies although relevant information recorded from the test categories could be washed out by averaging.

2. Regardless of sex and breed, the score obtained by most dogs corresponds to Campbell’s Submissive Stable group.

3. Higher levels of dominance in males compared with females are already evident when the dog is a puppy, as shown in all groups of subtests. Males displayed traits that were between stable submission and dominance, whereas females mainly fell into the stable submission category with a few traits of excessive submission.

4. Very few puppies displayed independence (if they had received adequate human socialization) and scores were statistically equal between males and females. This reinforces the statement that the dog is a social animal and therefore establishes clear relationships of dependence (need of interaction) with people.

According to the puppies’ scores without Campbell’s interpretations

1. The 5 Campbell subtests are significantly different. The first 3 subtests provide the highest scores and subtest 5 yields the lowest scores, possibly because the puppy becomes accustomed to each situation and behaves more naturally compared to subtest 5.

2. Age is highly influential in the final score obtained by the puppies. There is a positive correlation, between age and final scores, in that the same dog tends to obtain higher scores at an older age.

3. Final mean scores have a clear relationship with breed type and may also vary depending on whether the dogs are male or female.

4. Independence or a lack of human socialization in puppies is revealed in the Campbell test by very low scores (0 points) in subtests 1–2–4. Unsocialized puppies obtain a very high mean score in subtests 3–5 and independent dogs obtain different scores depending on their behavioral tendencies assessed using subtests 3–5. All subtests might be very useful in detecting puppies that display a lack of human socialization, but only subtests 1–2–4 will help detect independent puppies.

5. Out of all the breeds studied, the husky (the only spitz dog studied) displayed the highest levels of independence. Subtests 3–5 are the only subtests that assess behavioral tendencies in dogs with high levels of independence.

6. The size of the breed does not influence the final mean score in the puppy.

7. According to the groups established by the FCI, Terriers (group 3) and Molossers (group 2) display the highest scores. Mixed breeds would also be included in this group.

8. Behavioral tendencies in English cocker spaniels vary depending on the animal’s coat color. Final mean score associated with agression in decreasing order are as follows: golden, black, and parti-colored dogs.