Author Names: Carly Hill (BSc (Hons Animal Behaviour and Welfare) and Lucy Bearman-Brown
The aim of the current study was to investigate the efficacy of three positive reinforcement methods in training dogs, as measured by the number of trials taken to learn a task. The dogs (n=24) were randomly placed into one of three positive reinforcement conditions; food, play or stroking. They were trained to walk around a cone with a pole in the top using one of these methods. In the play condition, when the dogs were successful they were invited to play with a toy in the shape of a rugby ball on a rope. In the food condition, the dogs were given a piece of chicken from a pouch round the trainer’s waist. In the stroking condition, the dogs were stroked either down their flank or across their back depending on the dog’s preference. The dogs were considered trained when the trainer could stand three metres away from the pole, cue the behaviour and the dog respond correctly for 75% of 10 cues. No significant differences were found between the three conditions in the percentage of successfully performed trials, the number of trials attempted or the time taken to learn the new cue. However, there are many factors that could have influenced the results such as the trainability of the breeds, the neuter status of the dogs, or the age of the dog. All these factors can affect the acquisition of new cues. Further study with consideration of these factors and a larger sample size would be beneficial.
The prevalence of problem behaviours reported by dog owners such as aggression, separation anxiety, chewing and chasing has increased over recent years, in part, due to greater owner knowledge and awareness (Blackwell et al, 2008). The welfare of the dogs displaying these behaviours can be compromised, both as these behaviours can be a result of anxiety and also as there is an increased risk of the dogs being relinquished into rescue centres or euthanized (Blackwell et al, 2008). Many owners would keep their dog if the problems could be resolved (Blackwell et al, 2008). Due to these welfare concerns, the need to train dogs effectively has become increasingly important (Hiby, Rooney and Bradshaw, 2004). In animal training, there are many different techniques that can be used which have varying degrees of recommendation and success (Heindenreich, 2007). Positive reinforcement involves the application of a stimulus after the presentation of a desired behaviour to increase the likelihood of the behaviour being repeated, whereas negative reinforcement is the removal of a stimulus when the desired behaviour is displayed to increase the frequency of the behaviour (Heindenreich, 2007).
There is a large volume of research dedicated to comparing the use of positive and negative reinforcement methods and evaluating the effectiveness and the welfare impacts of each method (Deldalle and Gaunet, 2014; Herron, Shofer and Reisner, 2009; Blackwell et al, 2008; Haverbeke et al, 2008; Hiby, Rooney and Bradshaw, 2004; Schilder and van der Borg, 2004). Many studies have found that using negative reinforcement and positive punishment techniques are associated with the animals showing behavioural signs of stress and fear towards their handlers (Deldalle and Gaunet, 2014; Haverbeke et al, 2008; Schilder and van der Borg, 2004). Deldalle and Gaunet (2014) compared the different training methods. Although their sample size was small as observations were only carried out in two schools from the same area, dogs trained using negative reinforcement showed more stress related behaviours when given commands than those taught with positive reinforcement. The dogs taught using positive reinforcement were significantly more attentive during the sessions. In Schilder and van der Borg’s (2004) study looking at the short and long term behavioural effects of using a shock collar in training, they found that stress remained even after the shock collar used to train them had been removed. The dogs associated their handler’s presence with an aversive stimulus and displayed lower body positioning and aggressive behaviours, especially when given commands where they were usually shocked (Schilder and van der Borg, 2004). However, similar to Deldalle and Gaunet (2014) the sample size was limited to 16 dogs in the experimental group and 15 dogs in the control group. Haverbeke et al (2008) found that not only did dogs trained via negative reinforcement show more stressed behaviour, the use of negative reinforcement techniques also did not equate to high performance scores. It was found that the four trainers that used play as a reward when the dogs performed correctly had better performance scores than the military handlers. The use of negative reinforcement did not stimulate the dogs to pay attention, resulting in more punishment, and the dogs were slower to learn the command (Haverbeke et al, 2008).
Questionnaires have also been used to survey dog owners and a number of studies have found that the use of negative reinforcement and punishment methods to train dogs were associated with more undesired behaviours being reported by owners compared to owners who only used positive reinforcement techniques (Herron, Shofer and Reisner, 2009; Blackwell et al, 2008; Hiby, Roney and Bradshaw, 2004). Hiby, Rooney and Bradshaw (2004) conducted a study looking at training methods and the number of problems behaviours reported by dog owners. Dogs trained via negative reinforcement and punishment demonstrated more problem behaviours. Herron, Shofer and Reisner (2009) found that dogs being trained using negative reinforcement and punishment also displayed more aggressive responses. This indicates a welfare concern as the aggressive responses were defensive, therefore potentially showing fear and stress as a result of the methods being used in training. There is also the danger that the resultant aggressive behaviour could lead to dogs being abandoned or euthanized. It also has safety concerns for the owners, as if the dog responds aggressively there is a chance they could be bitten (Herron, Shofer and Reisner, 2009). Similarly, Blackwell et al (2008) found that training using punishment was linked to a higher level of undesirable behaviour when compared to dogs only trained using positive reinforcement. Those dogs taught with both positive reinforcement and punishment were associated with higher aggressive scores (Blackwell et al, 2008). Overall, it has been found in a range of studies that positive reinforcement training is better for the welfare of the dogs, safer for the owners and more effective when trying to train new behaviours (Herron, Shofer and Reisner, 2009; Blackwell et al, 2008; Hiby, Roney and Bradshaw, 2004).
Training can have many benefits for both owner and dog. In a study conducted by Mariti et al (2013) comparing the level of attachment between pet and working dogs and their owners, whilst no significant results were obtained there was a trend towards dogs trained in search and rescue having a higher attachment level than pet dogs. The sample size of this study was small and further consideration with a larger sample size would be beneficial to explore potential relationships between training and attachment.
Dog respond to signals from humans such as following human gaze, pointing, body posture and touch to aid them in solving tasks such as finding hidden food (Bentosela et al, 2008). Dogs use their gaze and vocalisations to communicate in these tasks, and training has been found to influence these communicative responses (Marshall-Pescini et al, 2009; Bentosela et al, 2008). Bentosela et al (2008) investigated how a dog could be trained to look up at the owner’s face through a series of trials involving reinforcement, reinforcer omission and extinction. The results showed that reinforcement increased the communicative responses of the dog whilst omission and extinction reduced the communicative responses. Bentosela et al (2008) also showed that trained dogs are more likely to look up at their owners without prompting whilst walking than those who have received no training, suggesting that the communication between dogs and humans is influenced by the level of training that the dog receives. The communication response of dog gaze to the human face is beneficial as it aids social cohesion and cooperation between dog and owner (Elgier et al, 2009a) and can help in solving cognitive tasks (Elgier et al, 2009b). Marshall-Pescini et al (2009) found that communicative behaviours directed towards humans were significantly influenced by the type of training a dog has received. When comparing agility dogs, search and rescue dogs and untrained dogs in a simple socio-cognitive task involving food hidden in a container, agility dogs looked between their owners and the apparatus more frequently than untrained dogs and search and rescue dogs. This could be due to the way that agility dogs are trained to pick detect subtle changes in body posture in order to guide them around a course (Marshall-Pescini et al, 2009). They also found that during the “unsolvable task”, where the container was stuck down preventing the dog from obtaining the food, search and rescue dogs would bark more than the agility and untrained dogs. Again, this could be due to the way they are trained to bark when they have found the missing person at the end of their search. This study showed that the training experience of the dog has a significant effect on the way that they communicate with their owners (Marshall-Pescini et al, 2009).
In a study by Marshall-Pescini et al (2008) comparing trained to untrained dogs, the dogs had to press a paw pad or lift the lid of the apparatus to obtain the food inside. Trained dogs were found to be more confident and independent in problem solving tasks as they spent more time focusing on the apparatus rather than looking back towards their owner (Marshall-Pescini et al, 2008). It was possible through their training that the dogs had learnt to display a variety of different behaviours until the correct one resulted in the food reward which may explain the greater time spent interacting with the apparatus (Marshall-Pescini et al, 2008).
Training can also be used to correct problem behaviours such as rushing and barking at the door when visitors arrive (Yin et al, 2008). Using a remote controlled positive reinforcement technique, experimenters and then eventually owners were able to train their dog to run to a target mat, lie down and stay there when visitors appeared at the door (Yin et al, 2008). This technique was found to be more effective than punishment based methods commonly used such as squirting water into the mouth of the dog when barking, kneeing the dog in the chest when dashing at the door, or jerking the dog back on a choke chain lead (Yin et al, 2008).
Within positive reinforcement training, there are many stimuli that can act as reinforcers that result in the behavioural response being repeated. It has been found that dogs can have different preferences for the reinforcer used, such as stroking or verbal praise (Feuerbacher and Wynne, 2015; Feuberbacher and Wynne, 2014) or different types of food reward (Vicars, Miguel and Sobie, 2014), and that these reinforcers may vary in their effectiveness (Fukuzawa and Hayashi, 2013). Fueberbacher and Wynne (2014) investigated whether dogs preferred food as a reward rather than stroking. The dogs were found to prefer food over stroking, and there was not a satiation effect of receiving food. Later, Feuerbacher and Wynne (2015) compared whether shelter and owned dogs preferred being stroked or receiving verbal praise. They found that both shelter dogs and owned dogs preferred being stroked to receiving verbal praise, even to the point that owned dogs preferred being stroked even when it was their owner giving verbal praise and the experimenter giving strokes. Again, with this pairing they found the dogs did not become satiated in the amount of petting they received (Feuerbacher and Wynne, 2015).
Fukuzawa and Hayashi (2013) conducted a study to compare the effectiveness of three different types of reinforcer; food, stroking and verbal praise to reinforce a behaviour that the dogs already knew (“come”). The results showed that food was the most effective reinforcer as it took the least response time during training. They did however, find that the other reinforcers, whilst not as effective as the food, did work as reinforcers (Fukuzawa and Hayashi, 2013).
Vicars, Miguel and Sobie (2014) investigated whether the dog’s preference could be tested using a paired stimulus preference assessment whereby six food items, rated by the dog’s owner for perceived preference, were paired up and presented against each other for the dogs to choose in 15 paired stimulus assessments. They found that their test was an effective way of determining a dog’s preference in terms of food as a reinforcer. Vicars, Miguel and Sobie (2014) conducted a second follow up study that investigated how their subjects’ preference had changed over time. They discovered that the preference test should be repeated after a period of time as the preference of the dog can change (Vicars, Miguel and Sobie, 2014). Whilst this test used food to find out the preference of the dog it could be argued that it could be adapted to find out which method of reinforcement (such as food, stroking, verbal praise, play) is preferred by the dog in order to determine the most effective reinforce to use when training.
Play behaviour is displayed and continued throughout a dog’s life, even when they become adults, and there are many different types of play (Bradshaw, Pullen and Rooney, 2014). Rooney, Bradshaw and Robinson (2001) found that humans performing signals such as the “Bow” and the “Lunge” can increase the frequency and the duration of play displayed by dogs with humans, showing that dogs can read the body language of humans to engage in play as they do with other dogs. However, Bradshaw, Pullen and Rooney (2014) also identified that play behaviour differs between dog-dog play and dog-human play. Dog-dog play can be competitive, such as winning possession of a toy and is considered a show of dominance (Bradshaw, Pullen and Rooney, 2014). Whereas, with dog-human play, the winning possession of a toy is rewarding to the dog but is not a display of dominance (Bradshaw, Pullen and Rooney, 2014). In fact, play has been found to be important in the formation of emotional bonds between dogs and their owners (Bradshaw, Pullen and Rooney, 2014) which has a positive impact on the quality of life of the dog (Marinelli et al, 2007).
Although play is mentioned in literature as being rewarding for dogs (Bradshaw, Pullen and Rooney, 2014) and reported as being used as a reinforcer during training with success (Haverbeke et al, 2008) there is no literature to date on how effective play is as a form of reinforcement in comparison to other reinforcers. The aim of this current study was to adapt the methodology performed by Fukuzawa and Hayashi (2013) to investigate which positive reinforcement method (food, stroking and play) is more effective. This was measured by the number of trials taken, in 10-minute training sessions one to three times a week, to learn a new cue for a sample of 24 rescue dogs. The null hypothesis was that there will be no significant difference in the effectiveness of the different positive reinforcement methods, whilst the alternate hypothesis is that one of the forms of reinforcement (play, stroking, food) will be more effective than the others, with the dogs taking fewer trials to learn the novel cue
An opportunistic sample of 24 rescue dogs were used within this study. The dogs were housed at a local rescue centre and comprised fourteen males and ten females. The dogs’ age ranged between 10 months and 10 years (mean: 4.39 years). A variety of different breeds were used including Jack Russell terriers (3), Labradors (1), Staffordshire bull terriers (5), American bull dogs (4), Chihuahuas (1), Lurchers (1), Rottweilers (1), and a variety of cross breeds including Husky cross (1), Staffordshire bull terrier cross (3), Jack Russell terrier cross (1), American bull dog cross (1), and German Shepherd cross (2). The rescue centre was provided with information about the study and its potential benefits and risks and signed a consent form to allow the use of the dogs and premises.
2.2 Experimental apparatus
Training sessions took place in the rescue shelter’s training room which is used for testing how dogs behave within a home environment to different stimuli. The training room consisted of a desk and office chair to the far end of the room, two armchairs opposite the office desk, a bookshelf next to the armchairs, and a crate for smaller dogs adjacent to the bookshelf. For the purpose of this study the door to the crate was kept shut. Water was available throughout the study. All dogs were trained three times per week and once per day. Previous studies have found that dogs have better acquisition of commands if trained weekly and with one session per day (Demant et al, 2011; Meyer and Ladewig, 2008). However, time constraints necessitated training the dogs three times a week rather than the optimum once a week (Demant et al, 2011). A novel cue was used which involved the dog walking around a cone with a garden weave pole stuck through the top of the cone. The cone had rice inside to act as a weight and duct tape was used to secure the pole into the cone as seen in figure 1.
Figure 1 –photograph of the construction of the equipment used for training (Author, 2017a)
2.3 Experimental conditions
The dogs were randomly placed into the three conditions and were thus trained using one of three positive reinforcement types; food, stroking or play. In the food condition the dogs were given a piece of chicken, from a training pouch around the trainer’s waist, as a reward for displaying the correct response to the cue. In the stroking condition the dogs were either stroked down their flank, or across their back depending on their preference. The dogs were not stroked on their head as many of the dogs did not like their heads being touched. In the play condition, the dogs were invited to play tug of war with a rugby ball shaped mesh toy on a rope as seen in figure 2. If the dogs did not initiate play in this way, the trainer would throw the toy for the dog to chase. The timing of the reinforcement was practised by the experimenter to optimise the acquisition of the cue. Yamamoto, Kikusui and Ohta (2009) found that a delay in the reinforcement can affect the dog’s efficiency in both learning new cues and also their retention of old cues.
Figure 2 – Photograph of the rugby ball on a rope toy used in the play condition. (Author, 2017b)
The dogs received one 10-minute training session per day, three times a week to optimize the dogs’ learning capability (Demant et al, 2011; Meyer and Ladewig, 2008). The training was broken down into four stages. Training was considered to be concluded if the dogs could correctly walk around the pole on cue with the trainer stood three metres away from the pole for 75% of 10 cues given in a final test. If the dog failed the test (i.e. got less than 75% of the cues correct) the dog resumed training and was then retested when the trainer felt that the dog was confident in completing the cue. The dogs would be submitted into the test when they performed the cue correctly for the previous ten cues. The dog’s age, breed, the number of cues attempted, the number of cues correctly completed and the number of training sessions taken to learn the cue were all recorded. The testing phase was counted as a training session.
2.4.1 Stage One
In the first stage the dogs were either taken for a walk by the trainer, or were taken to the location of the training sessions in order for the dogs to become accustomed to the trainer. If taken for a walk the trainer would often call the dog back to them and would pet the dog before setting off walking again. If taken to the training room, the dogs would be invited to play with toys that were available, or they would receive attention from the trainer. This was not restricted to one session as for some dogs that had experienced a difficult background it was harder for them to trust new people. The subject moved on to the next stage when they did not display any behavioural indicators of stress such as lip licking, yawning, paw lifting and panting (Rooney, Gaines and Bradshaw, 2007). Any dogs showing signs of extreme levels of stress such as lowering of body posture, including their ear position, tail position and general body position (Beerda et al, 1999), would have been removed from the study however, this did not occur.
2.4.2 Stage Two
Once the dog was accustomed to the trainer, the training sessions began. The first stage was to lure the dogs round the pole either using chicken (food condition), the toy (play condition), or the trainer’s hand or occasionally their body (stroking condition). The dog was able to move on to the next step once they had got the previous ten cues correct without hesitation. Once the dog completed being lured round correctly for the previous ten cues the trainer combined a verbal cue with the action of the dog walking round the pole.
2.4.3 Stage Three
Once the dog was confident walking round the pole on cue the trainer increased the distance between the trainer and the pole. This was done gradually by adding an extra step away from the pole each time the dog was completing the trials without hesitation correctly for the previous ten cues. If the dog became confused the trainer would go back a step until the dog was confident in the correct response before trying to increase the distance again. This continued until the trainer was three metres away. Once the dog was confident in completing the cue at this distance the trainer moved on to the next stage.
2.4.4 Stage Four
In the fourth stage the trainer tested the dog, by giving the dog ten attempts to correctly perform the cue. If the dog performed the cue correctly 75% of the time the dog was considered trained and training sessions would cease with this dog. An incorrect cue was considered as the dog failing to perform the correct behaviour after 5 seconds. If the dog failed to perform the cue correctly 75% of the time the dog resumed training sessions until the dog was deemed confident with the cue by the trainer and retested.
2.5 Statistical analysis
Once data was collected, the percentage for the number of trials the dogs performed successfully was calculated. The percentage for the number of trials successfully performed was calculated using Microsoft Excel by dividing the number of trials successfully completed by the number of trials attempted and multiplying by 100. A Shapiro Wilk normality test was then performed to determine if the data was normally distributed. The data was found to be non-parametric so a Kruskal Wallis test was performed to determine if there was a difference in the percentage of the number of trials successfully performed between reinforcement conditions. This statistical test was appropriate due to the study testing for differences in continuous dependent data between multiple unrelated independent conditions (Field, 2009). A Shapiro Wilk normality test was then performed on the data for the number of trials the dogs attempted. This data was found to be normally distributed so a one-way ANOVA was performed to investigate if there were any differences in this measure between the reinforcement methods. The amount of time taken to learn the cue was calculated by multiplying the number of training sessions by 10 (the length of each training session). This was then tested for normality using a Shapiro Wilk test. The data was found to be non-parametric so a Kruskal Wallis test was performed to determine whether there was an effect of the reinforcement conditions on the time taken to learn the cue.
3.1 Percentage of the number of trials successfully performed
Analysis via a Kruskal Wallis test found that there was no significant difference in the percentage of the number of trials successfully performed between the three conditions (H (2) = 3.885, P=0.143) so the null hypothesis that there would be no difference between the three forms of reinforcement (play, stroking, food) in their effectiveness in training a novel cue was retained (Figure 3).
Figure 3 – Boxplot showing the median of the data and the interquartile ranges of the percentage trials achieved for each condition.
3.2 Number of trials attempted
Analysis via a one-way ANOVA test found that there was no significant difference in the number of trials attempted between the different conditions (F (2,21) =0.320, P=0.729) so the null hypothesis that there would be no difference between the three forms of reinforcement (play, stroking, food) in their effectiveness in training a novel cue was retained (Figure 4).
Figure 4 – Boxplot showing the median of the data and the interquartile ranges of the trials attempted for each condition.
3.3 Time taken to learn the cue
Analysis via a Kruskal Wallis test found that there was no significant difference in the time taken to learn the cue between the three groups (H (2) = 0.350, P=0.839) so the null hypothesis that there would be no difference between the three methods of reinforcement (play, stroking, food) in their effectiveness in training a novel cue was retained (Figure 5).
Figure 5 – Boxplot showing the median of the data and the interquartile ranges of the time take to learn for each condition.
The aim of the current study was to compare three positive reinforcement methods (play, stroking, food) to determine if one was more effective than the others, as measured by the number of trials taken by dogs to learn a new cue. The results of the current study indicate that the different positive reinforcement techniques did not differ in their effectiveness as there was no significant difference between the reinforcement conditions in the percentage of the number of trials that were successfully achieved, the number of trials that were attempted, or the time it took to train the new cue. Since there were no significant differences in the training measures found between the three conditions it could be suggested that all methods are effective as the dogs all succeeded in learning the new cue.
4.2 General discussion
4.2.1 Individual differences
Individuals within a species can display different behavioural responses that are consistent across time and context, meaning they have different personalities (Dingemanse et al, 2009). These different personalities can result in individuals behaving differently from others within the same context (Dingemanse et al, 2009; Wolf et al, 2007). Dog personality, and individual differences, were not considered within this study and could have affected the results. The personality of the dog may have affected their preference for the reinforcement methods and therefore, could have affected the acquisition of the new cue. Future research considering the effects of personality on trainability and preference for training methods would be of value.
4.2.2 Breed differences
A variety of breeds of dog were used in the study. Much research has explored breed differences in trainability (Mehrkam and Wynne, 2014; Udell et al, 2014; Starling et al, 2013; Turcsan, Kubinyi and Miklosi, 2011; Helton, 2010; Gacsi et al, 2009; Wobber et al, 2009; Serpell and Hsu, 2005). Breed differences can be important both in terms of the selection for different traits within breeds (Udell et al, 2014; Turcsan, Kubinyi and Miklosi, 2011; Helton, 2010), and also in terms of how the morphological differences of breeds can affect their cognitive (Gasci et al, 2009) and physical ability (Helton, 2010).
Different dog breeds have been found to vary in their trainability which is important when considering effective ways of training dogs, as some breeds will be easier to train that others (Udell et al, 2014; Turcsan, Kubinyi and Miklosi, 2011; Helton, 2010; Gacsi et al, 2009; Serpell and Hsu, 2005). Although boldness and trainability are distinct behaviour traits in domestic dogs (Turcsan, Kubinyi and Miklosi, 2011) it has been found that boldness can influence the trainability of dogs, especially those who perform in sports (Starling et al, 2013). Boldness and trainability can be linked to a genetic predisposition for different breeds, which can be related back to their original historic function (Turcsan, Kubinyi and Miklosi, 2011). Using four different behavioural traits; boldness, calmness, trainability and dog sociability to characterise the different dog breeds, the authors concluded that there was a large difference in trainability and boldness between different dog breeds (Turcsan, Kubinyi and Miklosi, 2011). Similarly, Mehrkam and Wynne (2014) proposed that being able to predict the typical behavioural tendencies of different breeds is invaluable in providing training for the different breeds which suits them for the roles they can fulfil.
Serpell and Hsu (2005) ranked 11 different breeds of dog using trainability scores collected using a C-BARQ questionnaire completed by owners. Certain breeds such as Labradors, English springer spaniels, Golden retrievers, Poodles, Rottweilers and Shetland sheepdogs were rated highly for trainability. Other breeds such as the Dachshund, West highland terrier, and Yorkshire terriers formed the middle scoring dogs and the Siberian Husky and the Bassett hound scored very low for trainability (Serpell and Hsu, 2005). There was evidence that the dogs from working lines in both Labradors and English springer spaniels scored significantly higher trainability scores than those from show lines of the same breed (Serpell and Hsu, 2005).
Helton (2010) conducted four different experiments to test whether Coren’s (1994) list of the most highly trainable dog breeds was accurate. Helton (2010) argued that the trainability of a dog is different from the intelligence of a dog. Some breeds are viewed as having low trainability, such as terriers, as they are more independent workers however, this does not reflect their intelligence. Coren’s highly trainable dogs were more likely to appear in agility competitions than the low trainability dogs and due to this were more likely to win medals (Helton, 2010). The highly trainable dogs were significantly faster than the low trainability dogs but despite this, there was no significant difference in the precision of the two groups of dogs measured by the number of faults made during an agility course. Helton (2010) discovered that the dogs did not differ in the number of hours put into training the different groups, suggesting that the limiting factor for the non-elite, low trainability group was speed and morphological structure (Helton, 2010).
Gasci et al (2009) classified the breeds into three different groups, cooperative workers, independent workers and mongrels and tested them in a two-object pointing choice test. Their hypothesis was that dogs from the cooperative group who were selected to perform tasks that involved visual guidance from human partners would perform better in following directional pointing than the other groups. Their results showed that the dogs within the cooperative group performed better than the other groups (Gasci et al, 2009). However, a limitation of the study is the subjective nature of classifying the breeds into different groups. Gasci et al (2009) classified the Siberian husky within the independent workers group, however, in Wobber et al’s (2009) study the Siberian husky was placed within the cooperative worker group and was found to be sensitive to the social signals given by humans. The different expectations of different breeds in terms of classification as cooperative or independent may have influenced the results (Udell et al, 2014).
Considering the differences in trainability found between dog breeds in the literature, and the range of different breeds used in this study due to the opportunistic sampling method, it is important to consider that breed differences could have affected the results of this study. The low sample size for each breed prevents consideration of this in the current study, however future research using a larger sample size for each breed would be beneficial.
22.214.171.124 Selection for different traits
The different breeds used in this study may also have affected the results due to other breed differences, beyond just trainability. Udell et al (2014) investigated how selection in the domestication process had affected the predatory sequence of three different breeds of dog; Border collie, Airedale terrier, and Anatolian Shepherds. Through selection humans have weakened the wolf’s predatory sequence in domestic dogs to either remove the final steps of killing their prey or inhibiting the sequence completely (Udell et al, 2014). Through domestication the “orient – eye – stalk – chase” component of the original predatory sequence in Border collies has become exaggerated. Whereas, in Airedale terriers, domestication has maintained this predatory sequence and in Anatolian shepherds domestication has inhibited it (Udell et al, 2014). Border collies were found to be most successful in a two-choice-point following test, where the experimenter would point to one of two containers containing food (Udell et al, 2014). Both the Border collie and the Airedale terrier performed above chance in their trials. The Anatolian shepherds showed more restraint which could be due to a genetic predisposition for this response (Udell et al, 2014).
Udell et al (2014) completed a follow up study on the Anatolian shepherd dogs alone to determine whether, by giving the Anatolian shepherd dogs more training trials, the response was due to a socio-cognitive deficit or due to their lack of experience and known breed inhibition. Giving the Anatolian shepherds more training trials improved their performance as they overcame their behavioural inhibition (Udell et al, 2014). This finding could suggest that the selection for different behavioural traits in different breeds of dogs (Turcasan, Kubinyi and Miklosi, 2011) could affect the length of time needed to train, but not affect their overall trainability or reflect their intelligence (Helton, 2010). This could have affected the current study as, due to the range of breeds used, different behavioural traits of the breeds could have affected the dogs’ rate of acquisition of the task.
126.96.36.199 Morphological differences
The morphological differences seen in different dog breeds can affect their success in cognitive tasks (Gasci et al, 2009), the methods used to achieve success (Mehrkam and Wynne, 2014) and also their ability to perform in sports (Helton, 2010).
Gasci et al (2009) conducted a follow up study from their cooperative worker, independent worker and mongrel two-way object choice test, to test whether brachycephalic dogs performed better than dolichocephalic breeds due to the positioning of their eyes. They found that brachycephalic dogs were more successful in the two-way choice test than the dolichocephalic dogs (Gasci et al, 2009). Mehrkam and Wynne (2014) discovered that brachycephalic dogs were more likely to use their paws due to their short snout in cognitive tasks, such as trying to get food out of Kong toys, compared to dolichocephalic breeds, who were more likely to use their snout to obtain the food.
Helton (2010) recognized that the morphological attributes of a breed of dog can influence the perceived trainability of the breed. In studying agility dogs Helton (2010) found that the height of the elite dogs was close to the mean average height of all breeds of dog. The non-elite dogs tended to be more heterogenous and lay at either end of the height spectrum (Helton, 2010). It was proposed that this average height was perceived as more trainable due to large dogs being harder to control and manoeuvre whereas very small dogs required more effort on the part of the trainer to get down to the dog’s level to interact with the dog, as is needed for training. The elite dog appeared to be a compromise between speed, with larger dogs having a longer stride, and the flexibility and manoeuvrability of smaller dogs (Helton, 2010). Whilst height is not an accurate perception of the trainability of dogs it can influence trainers breed choice (Helton, 2010).
In the current study, morphological differences in the breeds used could have impacted upon their cognitive processes in understanding the task (Mehrkam and Wynne, 2014). The nature of the task (walking/running around a pole) could have affected the dogs’ ability to complete the task as the more “elite” shaped dogs may have been able to perform better than the “non-elite” shaped dogs (Helton, 2010).
Another factor that may have affected the results of this study is the dogs’ age. Age can affect boldness and therefore, the trainability of dogs (Starling et al, 2013). Starling et al (2013) investigated factors that affect boldness in dogs through data collected via a C-BARQ questionnaire completed by owners. A negative correlation was found between the age of dogs and boldness (Starling et al, 2013). This could be due to age related discomfort, such as the dog experiencing pain, or the dog being less likely to take risks such as engaging with other dogs or humans (Starling et al, 2013). It could also be a cause of a loss of interest in exploring their surroundings (Starling et al, 2013). This could also be applied to training as the dog may lose interest in learning new cues or may be suffering discomfort and may not want to engage with training (Starling et al, 2013). This could have been the case in the current study as there was a range of ages of dogs and the older dogs may have been less inclined to engage with the training.
4.2.4 Gender and neuter status
Mixed findings have been found regarding the effect of neutering on trainability (Starling et al, 2013) and whether there are breed specific differences (Serpell and Hsu, 2005). In Starling et al’s (2013) study, they found that male dogs were bolder than female dogs and the act of neutering male dogs decreased the boldness of the male. In contrast, Serpell and Hsu (2005) found that gender made no difference to the trainability of the dog. However, they did find some breed specific examples where gender did have an effect, such as that male Dachshund and West highland terriers were significantly more trainable than their female counterparts (Serpell and Hsu, 2005). Higher trainability scores were found in neutered Shetland sheepdogs and Rottweilers, and whilst neutering had no effect on the trainability of female Shetland sheepdogs, it resulted in higher trainability in male Shetland sheepdogs (Serpell and Hsu, 2005). More research is required to confirm whether neutering has a definitive effect on the trainability of dogs as many owners choose to have their dogs neutered to increase their dog’s trainability (Serpell and Hsu, 2005). Sex and neuter status could have impacted the result of the current study. Whilst both male and female dogs were used in the study, all the dogs were neutered and so this could have affected their trainability.
The current study had a number of limitations. The study had a limited sample size (n=24), and although the sample size was larger than that used in Fukuzawa and Hayashi’s (2013) study (n=15) it was still relatively small. A smaller sample size can negatively influence the validity of this study as the sample is unlikely to represent the whole population (Field, 2009). The likelihood that the effects of the study are true and not a result of chance are limited (Button et al, 2013). It is also possible that by having a smaller sample size this could have resulted in a false negative or an exaggerated effect if an effect had been found (Button et al, 2013).
Another limitation, similar to Deldalle and Gaunet’s (2014) study, is the use of only one observer when observing dogs’ behaviour during training. The current study only used the experimenter as both the trainer and the observer. There was not a second observer to validate whether the cue had been achieved and so some bias and subjectivity was introduced and could have affected the results.
As the sample was an opportunistic sample, it restricted the selection of breeds as the sample consisted of dogs that the rescue centre had in at the time. Many studies have shown that there are specific breed differences that affect the trainability of dogs and these could have affected the results (Mehrkam and Wynne, 2014; Udell et al, 2014; Button et al, 2013; Starling et al, 2013; Turcsan, Kubinyi and Miklosi, 2011; Helton, 2010; Gacsi et al, 2009; Wobber et al, 2009; Serpell and Hsu, 2005). This limitation, like the sample size, could affect the validity of the study as it does not represent the entire population (Field, 2009).
4.4 Future research
Future research could improve the current study by increasing the sample size and use a greater sample for each breed allowing consideration of breed differences. Further research could investigate the difference between pet dogs and rescue dogs. Udell, Dorey and Wynne (2010) found that shelter dogs do not perform as well as pet dogs in certain human guided tasks, suggesting that these responses are not innate as previously thought and requires some learning which could affect the training process. Future research could also investigate which form of positive reinforcement owners report using for different activities such as heelwork to music, agility, search and rescue and obedience to see if there is a difference in the reported effectiveness of each method for the activity.
Although this study found no significant differences in training between the positive reinforcement techniques, further study with a larger sample size may be warranted. The results may have been affected by a range of factors that can influence the trainability of dogs such as breed traits, morphological differences, age, gender and neuter status. Dog breeds can differ in their trainability and certain dogs excel at different tasks, such as dogs that work more independently versus dogs who work cooperatively with humans. Age can also affect the trainability of dogs, as the older the dog becomes the greater the reduction in the boldness of the dog, making the dog less likely to learn new cues. Gender and neuter status require more research into their effects on the trainability of the dogs. Evidence shows that there are breed specific examples where having the dog neutered affects the trainability but this is not widespread across the breeds. Finding an effective method of training is vital for dogs as the prevalence of problem behaviour is increasing alongside the associated concerns of relinquishment and euthanasia. Since, if the problem behaviour could be retrained, many owners would keep their dog, finding effective methods of training is becoming increasingly important. Further study in this important area of research is warranted.
Author (2017a) photograph of the construction of the equipment used for training [photograph]
Author (2017b) photograph of the rugby ball on a rope toy used in the play condition [photograph]
Beerda, B., Schilder, M.B., Van Hooff, J.A., De Vries, H.W. and Mol, J.A. (1999) Chronic Stress in Dogs Subjected to Social and Spatial Restriction 1 Behavioural Responses. Physiology and Behaviour [online]. 66 (2), pp. 233-242. [Accessed 28 January 2016].
Bentosela, M., Barrera, G., Jakovcevic, A., Elgier, A.M. and Mustaca, A.E. (2008) Effect of Reinforcement, Reinforcer Omission and Extinction on a Communicative Response in Domestic Dogs (Canis familiaris). Behavioural Processes [online]. 78 (3), pp. 464-469. [Accessed 16 February 2016].
Blackwell, E.J., Twells, C., Seawright, A. and Casey, R.A. (2008) The Relationship Between Training Methods and the Occurrence of Behaviour Problems, as Reported by Owners, in a Population of Domestic Dogs. Journal of Veterinary Behavior [online]. 3 (5), pp. 207-217. [Accessed 28 January 2016].
Bradshaw, J.W.S., Pullen, A.J. and Rooney, N.J. (2015) Why Do Adult Dogs Play?. Behavioural Processes [online]. 110, pp. 82-87. [Accessed 19 January 2017].
Button, K.S., Ioannidis, J.P.A., Mokrysz, C., Nosek, B.A., Flint, J., Robinson, E.S.J. and Munafo, M.R. (2013) Power Failure: Why Small Sample Size Undermines the Reliability of Neuroscience. Nature reviews Neuroscience [online]. 14 (5), pp. 365-376. [Accessed 22 February 2017].
Deldalle, S. and Gaunet, F. (2014) Effects of 2 Training Methods on Stress-related Behaviors of the Dog (Canis familiaris) and on the Dog-owner Relationship. Journal of Veterinary Behavior [online]. 9 (2), pp. 58-65. [Accessed 16 February 2016].
Demant, H., Ladewig, J., Balsby, T.J.S. and Dabelsteen, T. (2011) The Effect of Frequency and Duration of Training Sessions on Acquisition and Long Term Memory in Dogs. Applied Animal Behaviour Sciences [online]. 133 (3), pp. 228-234. [Accessed 28 January 2016].
Dingemanse, N.J., Kazem, A.J.N., Reale, D. and Wright, J. (2009) Behavioural Reaction Norms: Animal Personality Meets Individual Plasticity. Trends in Ecology and Evolution [online]. 25 (2), pp. 81-89. [Accessed 06 March 2017].
Elgier, A.M., Jakovcevic, A., Barrera, G., Mustaca, A.E. and Bentosela, M. (2009a) Communication between domestic dogs (Canis familiaris) and humans: Dogs are good learners. Behavioural Processes [online]. 81 (3), pp. 402-408. [Accessed 06 March 2017].
Elgier, A.M., Jakovcevic, A., Mustaca, A.E., Bentosela, M. and , (2009b) Learning and owner-stranger effects of interspecific communication in domestic dogs (Canis familiaris). Behavioural Processes [online]. 81 (1), pp. 44-49. [Accessed 06 March 2017].
Feuerbacher, E.N. and Wynne, C.D.L. (2014) Most Domestic Dogs (Canis lupus familiaris) Prefer Food to Petting: Population, Context and Schedule Effect in Concurrent Choice. Journal of the Experimental Anlaysis Od Behaviour [online]. 101 (3), pp. 385-405. [Accessed 06 March 2016].
Feuerbacher, E.N. and Wynne, C.D.L. (2015) Shut Up and Pet Me! Domestic Dogs (Canis lupus familiaris) Prefer Petting to Vocal Praise in Concurrent and Single-alternative Choice Procedures. Behavioural Processes [online]. 110, pp. 47-59. [Accessed 16 February 2016].
Field, A. (2009) Discovering Statistics Using Ibm Spss Statistics. 4th ed. London: Sage Publications Ltd.
Fukuzawa, M. and Hayashi, N. (2013) Comparison of 3 Different Reinforcements of Learning in Dogs (Canis familiaris). Journal of Veterinary Behavior [online]. 8 (4), pp. 221-224. [Accessed 16 February 2016].
Gacsi, M., McGreevy, P., Kara, E. and Miklosi, A. (2009) Effects of Selection For Cooperation and Attention in Dogs. Behavioural and brain functions [online]. 5 (1), p. 31. [Accessed 22 February 2017].
Haverbeke, A., Laporte, B., Depiereux, E., Giffroy, J.M. and Diederich, C. (2008) Training Methods of Military Dog Handlers and Their Effects on the Team’s Performances. Applied Animal Behaviour Science [online]. 113 (1), pp. 110-122. [Accessed 16 February 2016].
Heidenreich, B. (2007) An Introduction to Positive Reinforcement Training and Its Benefits. Journal of Exotic Pet Medicine [online]. 16 (1), pp. 19-23. [Accessed 16 February 2016].
Helton, W.S. (2010) Does Perceived Trainability of Dog (Canis lupus familiaris) Breeds Reflect Differences in Learning Or Differences in Physical Ability. Behavioural Processes [online]. 83 (3), pp. 315-323. [Accessed 22 February 2017].
Herron, M.E., Shofer, F.S. and Reisner, I.R. (2009) Survey of the Use and Outcome of Confrontational and Non-confrontational Training Methods in Client-owned Dogs Showing Undesired Behaviors. Applied Animal Behaviour Science [online]. 117 (1), pp. 47-54. [Accessed 16 February 2016].
Hiby, E.F., Rooney, N.J. and Bradshaw, J.W.S. (2004) Dog Training Methods: Their Use, Effectiveness and Interaction with Behaviour and Welfare. Animal Welfare [online]. 13 (1), pp. 63-69. [Accessed 08 February 2016].
Marinelli, L., Adamelli, S., Normando, S. and Bono, G. (2007) Quality of life of the pet dog: influence of owner and dog’s characteristics. Applied Animal Behaviour Science [online]. 108, pp. 143-156. [Accessed 17 February 2017].
Mariti, C., Ricci, E., Carlone, B., Moore, J.L., Sighieri, C. and Gazzano, A. (2013) Dog Attachment to Man: A Comparison Between Pet and Working Dogs. Journal of Veterinary Behavior [online]. 8 (3), pp. 135-145. [Accessed 24 May 2016].
Marshall-pescini, S., Passalacqua, C., Barnard, S., Valsecchi, P. and Prato-Previde, E. (2009) Agility and Search and Rescue Training Differently Affects Pet Dogs’ Behaviour in Socio-cognitive Tasks. Behavioural Processes [online]. 81 (3), pp. 416-422. [Accessed 16 February 2016].
Marshall-pescini, S., Valsecchi, P., Petak, I., Accorsi, P.A. and Previde, E.P. (2008) Does Training Make You Smarter? The Effects of Training on Dogs’ Performance (Canis familiaris) in a Problem Solving Task. Behavioural Processes [online]. 78 (3), pp. 449-454. [Accessed 16 February 2016].
Mehrkam, L. and Wynne, C.D.L. (2014) Behavioural Differences Among Breeds of Domestic Dogs (Canis lupus familiaris): Current Status of the Science. Applied Animal Behaviour Science [online]. 155, pp. 12-27. [Accessed 22 February 2017].
Meyer, I. and Ladewig, J. (2008) The Relationship Between Number of Training Sessions Per Week and Learning in Dogs. Applied Animal Behaviour Science [online]. 111 (3), pp. 311-320. [Accessed 28 January 2016].
Rooney, N.J, Gaines, S.A. and Bradshaw, J.W. (2007) Behavioural and Glucocorticoid Responses of Dogs (Canis familiaris) to Kennelling: investigating mitigation of stress by prior habituation. Physiology and Behaviour [online]. 92 (5), pp. 847-854. [Accessed 06 March 2017].
Rooney, N.J., Bradshaw, J.W.S. and Robinson, I.H. (2001) Do Dogs Respond to Play Signals Given by Humans?. Animal Behaviour [online]. 61 (4), pp. 715-722. [Accessed 19 January 2017].
Schilder, M.B.H., van der Borg, J.A.M. and , (2004) Training Dogs with Help of the Shock Collar: Short and Long Term Behavioural Effects. Applied Animal Behaviour Science [online]. 85 (3), pp. 319-334. [Accessed 16 February 2016].
Serpell, J.A. and Hsu, Y.A. (2005) Effects of Breed, Sex, and Neuter Status on Trainability in Dogs. Anthrozoos [online]. 18 (3), pp. 196-207. [Accessed 22 February 2017].
Starling, M.J., Branson, N., Thomson, P.C. and McGreevy, P.D. (2013) Age, Sex and Reproductive Status Affect Boldness in Dogs. The Veterinary Journal [online]. 197 (3), pp. 868-872. [Accessed 22 February 2017].
Turcsan, B., Kubinyi, E. and Miklosi, A. (2011) Trainability and Boldness Traits Differ Between Dog Breed Clusters Based on Conventional Breed Categories and Genetic Relatedness. Applied Animal Behaviour Science [online]. 132 (1), pp. 61-70. [Accessed 22 February 2017].
Udell, M.A.R., Dorey, N.R. and Wynne, C.D.L. (2010) The Performance of Stray Dogs (Canis familiaris) Living in a Shelter on Human-guided Object-Choice tasks. Animal Behaviour [online]. 79 (3), pp. 717-725. [Accessed 28 February 2017].
Udell, M.A.R., Ewald, M., Dorey, N.R. and Wynne, C.D.L. (2014) Exploring Breed Differences in Dogs (Canis familiaris): Does Exaggeration Or Inhibition of Predatory Response Predict Performance on Human-guided Tasks?. Animal Behaviour [online]. 89, pp. 99-105. [Accessed 27 February 2017].
Vicars, S.M., Miguel, C.F. and Sobie, J.L. (2014) Assessing Preference and Reinforcer Effectiveness in Dogs. Behavioural Processes [online]. 103, pp. 75-83. [Accessed 16 February 2016].
Wobber, V., Hare, B., Koler-matznick, J., Wrangham, R and Tomasello, M. (2009) Breed Differences in Domestic Dogs (Canis familiaris) Comprehension of Human Communicative Signals. Interaction Studies [online]. 10 (2), pp. 206-224. [Accessed 28 February 2017].
Wolf, M., Van Doorn, G.S., Leimar, O. and Weissing, F.J. (2007) Life-history Trade-offs Favour the Evolution of Animal Personalities. Nature [online]. 447 (7144), pp. 581-584. [Accessed 06 March 2017].
Yamamoto, M., Kikusui, T. and Ohta, M. (2009) Influence of Delayed Timing of Owners’ Actions on the Behaviors of Their Dogs, Canis familiaris. Journal of Veterinary Behavior [online]. 4 (1), pp. 11-18. [Accessed 16 February 2016].
Yin, S., Fernandez, E.J., Pagan, S., Richardson, S.L. and Snyder, G. (2008) Efficacy of a Remote-controlled, positive-reinforcement, dog-training system for modifying problem behaviors exhibited when people arrive at the door. Applied Animal Behaviour Science [online]. 113 (1), pp. 123-138. [Accessed 16 February 2016].