Abstract

This project assessed the welfare of dogs trained with pet training aids, specifically remote static pulse collar systems (e-collars). Previous work has focused on a very limited number of devices in a very limited range of contexts and the evidence of the impact of such devices on dog's overall quality of life is inconclusive. Project AW1402 aimed to assess the physical characteristics of the e-collars and the physiological, behavioural and psychological consequences of their use in dog training in four objectives.

1. Investigate the resistance in the neck skin of a range of dogs

2. Measure the physical output properties of the devices under investigation

3. Evaluate methods for recording behavioural/psychological measures of emotional state in the context of dog training.

4. Investigate the long term behavioural, physiological and psychological effects of using training devices in the domestic dog

A representative selection of e-collars was purchased to allow the assessment of electrical properties in laboratory tests and the evaluation of manuals (Objective 2). As the electrical output of the e-collars depended on the impedance presented by the dogs' necks, this was measured first on a sample of dogs of a number of breed and cross-breeds under dry and wet conditions (Objective 1). This was done under supervision of an animal welfare specialist and did not cause pain or distress as indicative from the dogs' behaviours.

The impedance of dogs can be modelled as a passive resistance with a value of about 10kΩ (10th -90th percentile range 4 – 150kΩ) for wet dogs and 600kΩ (22 – 950kΩ) for dry dogs. The momentary stimulus generated by the e-collars comprised a sequence of identical short voltage pulses. The continuous stimulus comprised a much longer sequence of the same voltage pulses. There were considerable differences between tested e-collar models in the voltages, the number of pulses in, and length of each stimulus, but little variation within individual models of e-collars. The peak voltage delivered by e-collars varied significantly with the resistance of the dog, from as much as 6000V at 500kΩ to 100V at 5kΩ. The highest voltages were generated for only a few millionths of a second. To allow meaningful comparisons between e-collars (taking into account the differences in electrical characteristics), a stimulus strength ranking indicator (SSRI) was developed. This showed differences between the selected e-collars, as well as differences in the relationship between momentary and continuous stimuli.

Manuals were clear on operation, but gave varying levels of information on using the e-collar in training. Generally they did not adequately explain their full potential, for instance with respect to using the tone or vibrate functions. Advice in manuals was not always taken up by end-users as evident from responses in owner questionnaire collected as part of objectives 3 and 4.

A pilot study involving 10 dogs with prior experience of e-collars and 10 control dogs (matched by age, sex, breed and where possible behavioural problem) was conducted to develop and evaluate protocols for assessing dog welfare in home and training environments (Objective 3). This was followed by a larger field study (Objective 4) involving 65 dogs with prior experience of e-collar training and 65 matched controls. Cases and matched controls for Objective 4 were initially recruited from a separate training methods survey distributed to dog owners to reduce sampling bias, but this was later supplemented by other recruitment methods.

Data collection in Objective 4 included:-
1. An owner questionnaire to collect demographic data on dogs and owners; and owner-reports of behaviour during training and efficacy of training methods.
2. First passage urine to measure cortisol, creatinine, and metabolites of the neuro-transmitters serotonin (5-HIAA) and dopamine (HVA).
3. Saliva for assay of cortisol prior to and during training.
4. Observations of dog behaviour during fitting of inactivated e-collar
5. Observation of dog behaviours during a series of standard training tasks (“stay”, “leave” and “recall” and the situation for which the focal device was used) given by both owner and a researcher and conducted in the context where the focal device had been originally used for training. Each set of tests were repeated both without (Test 1) and with (Test 2) the wearing of a dummy or inactivated e-collar to enable comparisons to be made between measures for the same dogs when wearing an e-collar (which may predict the application of stimulus for the dog) and not.
6. A spatial discrimination task designed to use judgement bias to assess underlying affective state.

Questionnaire data included type of device used, time since use, owner perceptions of the success of training, and owner reports of behavioural responses to use. Training methods used by owners in the control group could be sub-divided into those mainly using positive reinforcement (reward based) training, and those using methods based largely on punishment or negative reinforcement. Most owners (68%) purchased e-collars new, mainly from the internet, though some owners borrowed or purchased second hand collars. Problems with recall (40%) and livestock worrying (33%) accounted for the majority of reasons for e-collars use, although some manuals included information on use for basic obedience. Owner reports on operation of devices suggested they were often unclear as to how best use e-collars in training and some appeared not to have followed manual advice (if available). 36% of owners reported vocalisations on first use, and 26% on subsequent use of e-collars. This suggested that operating levels may not have been set in accordance with manufacturer’s instruction (where available), though due to owners often being unable to recall how they used the device this could not always be verified. Owners reported the addressed behaviours to be more severe in e-collar trained dogs than the controls. Owners showed a high degree of satisfaction with the effectiveness of all the training approaches used, though owners from the e-collar group were more likely to state they would prefer to try other forms of training in the future.

No significant differences between groups were identified for behaviours shown during collar fitting, although a wide range of behavioural responses among dogs were noted. These differences were considered likely to reflect response to novelty in the control group, and the specific events that usually followed collar fitting in the e-collar group such no consequence, going for a walk or stimulus application. Because of high variability between dogs, it was considered that differences in measures between the first series of training tasks (Test 1; conducted with no collar) and the second series of training tasks (Test 2; conducted with dogs wearing a dummy collar) would be more reliable than absolute differences between groups. There was a significant increase in salivary cortisol between tests in the e-collar group compared to the sub-group of dogs trained using positive reinforcement. A behavioural scale incorporating proportion of training period tense, an inverse of proportion of training time relaxed, and proportion of time with attention directed at owner (whoever was training) significantly increased in the e-collar group, as compared to both the whole control group and the sub-set of dogs predominantly trained using positive reinforcement. These differences may reflect increased emotional arousal in e-collar dogs as a result of previous learned associations with the collar. Data was collected for a further 11 control dogs who experienced both sets of standard training tasks but wearing no collar to test for potential order effects. Their behavioural and physiological responses were consistent with control dogs who wore the e-collar for the second set of tasks.

There was some evidence of higher baseline cortisol in control dogs compared with e-collar dogs in both the urinary cortisol: creatinine (reflecting cortisol production overnight before researcher arrival) and baseline salivary cortisol (taken after the arrival of the researcher and likely to be influenced by the events associated with visitor arrival and greeting) particularly when considering just the positive reinforcement sub-group. However these differences were small and found not to be significant when a multiple comparison Bonferroni correction was applied. There were no significant differences in neurotransmitter metabolites between the two groups. Neither were there significant differences between control and e-collar dogs with respect to speed to ambiguous probes in the judgement bias task. However, in the latter case, group effects were confounded by strong effects of arena size where different test spaces had been used.

Overall, this project has highlighted the very variable outcomes between individual dogs when trained using e-collars. The combination of differences in individual dog’s perception of stimuli, different stimulus strength and characteristics from collars of different brands, differences between momentary and continuous stimuli, differences between training advice in manuals, differences in owner understanding of training approaches and how owners use the devices in a range of different circumstances are likely to lead to a wide range of training experiences for pet dogs. This variability in experience is evidenced in the data from trained dogs such as owner reports of their dogs’ response to e-collar use.

Significant differences were, however, found in data collected from e-collar and control dogs undergoing standard training tests with and without dummy e-collars. These included a difference in the change in salivary cortisol between tests with e-collar dogs showing an increase and positive reinforcement dogs showing a lowering of salivary cortisol between the tests. There were also behavioural changes that were consistent with changes in emotional state, with e-collar dogs showing an increase in a behavioural scale incorporating time spent tense and the inverse of time relaxed between the two situations. These training tasks were designed as far as possible to replicate the context where e-collar training had occurred in the past, and indicate a shift towards higher levels of physiological and behavioural arousal in the e-collar dogs as well as a tendency to focus more on the owner than when they had not been wearing a collar.

Thus it seems reasonable to conclude that the previous use of e-collars in training is associated with behavioural and physiological responses that are consistent with negative emotional states. It is therefore suggested that the use of e-collars in training pet dogs leads to a negative impact on welfare, at least in a proportion of animals trained using this technique.