Accepted Manuscript

The Effects of Using Aversive Training Methods in Dogs – A Review

Gal Ziv

PII: S1558-7878(17)30035-7
DOI: 10.1016/j.jveb.2017.02.004
Reference: JVEB 1041

To appear in: Journal of Veterinary Behavior

Received Date: 18 March 2016

Revised Date: 13 December 2016
Accepted Date: 11 February 2017

Please cite this article as: Ziv, G., The Effects of Using Aversive Training Methods in Dogs – A Review,
Journal of Veterinary Behavior (2017), doi: 10.1016/j.jveb.2017.02.004.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.
 ACCEPTED MANUSCRIPT
1

1 Running Head: AVERSIVE TRAINING IN DOGS

4 The Effects of Using Aversive Training Methods in Dogs – A Review

PT
5

RI
6 Gal Ziva

SC
a
8 The Zinman College of Physical Education and Sport Sciences, Wingate Institute, Netanya,

9 Israel

10
U
AN
11

12
M

13
D

14
TE

15 Corresponding Author:

16 Gal Ziv, PhD

EP

17 Tel Aviv 69360

18 Israel
C

19 +972-54-4412010
AC

20 [email protected]

21
 ACCEPTED MANUSCRIPT
2

22 ABSTRACT

23 The purpose of this paper is to review a series of studies (N= 17) regarding the effects of using

24 various methods when training dogs. The reviewed studies examined the differences between

25 training methods (e.g., methods based on positive reinforcement, positive punishment,

PT
26 escape/avoidance, etc.) on a dog’s physiology, welfare, and behavior towards humans and other

RI
27 dogs. The reviewed studies included surveys, observational studies, and interventions. The

28 results show that using aversive training methods (e.g., positive punishment and negative

SC
29 reinforcement) can jeopardize both the physical and mental health of dogs. In addition, while

30 positive punishment can be effective, there is no evidence that it is more effective than positive

31
U
reinforcement-based training. In fact, there is some evidence that the opposite is true. A few
AN
32 methodological concerns arose from the reviewed studies. Among them are small sample sizes,

33 missing data on effect size, possible bias when coding behavior in observational studies, and the
M

34 need to publish case reports of bodily damage caused by aversive training methods. In
D

35 conclusion, those working with or handling dogs should rely on positive reinforcement methods,
TE

36 and avoid using positive punishment and negative reinforcement as much as possible.

37
EP

38

39
C

40 Keywords: dog training, reinforcement, punishment, training methods, electronic collar

AC

41
 ACCEPTED MANUSCRIPT
3

42 Introduction

43 Domestic dogs are an integral part of human culture, and their welfare is an important

44 concern for owners, caretakers, veterinarians, behavior specialists, and all those working or

45 handling them. Much controversy exists in the veterinary and the dog training community

PT
46 regarding the efficacy and possible negative unintended outcomes of various training methods of

RI
47 dogs. These training methods can range from reward-based to aversive, and individuals who

48 work with dogs choose training methods based on several factors such as their level of education,

SC
49 their previous success with different methods, and their individual set of morals.

50 Both classical and operant conditioning processes are usually involved in any dog

51
U
training method. These processes have been researched extensively, and information about them
AN
52 can be found in both academic (e.g., Chance, 2003) and professional (e.g., Reid, 1996) books.

53 For the purpose of this review it is important to briefly define classical and operant conditioning.
M

54 Classical conditioning is the process of pairing a neutral stimulus (e.g., the conditioned stimulus)
D

55 with an unconditioned stimulus (e.g., food) (Chance, 2003). This process allows an animal to
TE

56 make an association between the two stimuli. In contrast, operant conditioning is a procedure in

57 which a behavior becomes stronger or weaker depending on its consequences (Chance, 2003). In
EP

58 general, there are four possible consequences in operant conditioning: 1) positive reinforcement

59 – a behavior is strengthened by the presentation of a stimulus (that the animal wants), 2) negative
C

60 reinforcement – a behavior is strengthened by the removal of an unpleasant stimulus that the

AC

61 animal wants to avoid , 3) positive punishment – presenting an unpleasant stimulus that causes a

62 reduction in the strength of a behavior, and 4) negative punishment – the removal of a stimulus

63 that the animal seeks out, which causes a reduction in the strength of a behavior (Chance, 2003).
 ACCEPTED MANUSCRIPT
4

64 Traditional training methods tend to rely on positive punishment and negative

65 reinforcement. Unfortunately, using these operant principles can have negative effects on dogs’

66 health and behavior (Beerda et al., 1998). In his book “Coercion and its Fallout”, Murray

67 Sidman suggests that “…what makes the noncoercive alternatives necessary....is the vast

PT
68 catalogue of punishment’s side effects – consequences of punishment that cancel out its

RI
69 benefits…” (Sidman, 2000, p. 80). Indeed, using punishment can be accompanied by a number

70 of possible undesirable, negative, and potentially injurious (to the subject) effects, such as

SC
71 escape behavior, aggression, and apathy (Chance, 2003). Importantly, while negative

72 reinforcement uses a removal of an unpleasant stimulus, this stimulus must first be presented.

73
U
Thus, the presentation of the unpleasant stimulus can be considered positive punishment for the
AN
74 behavior that occurred just prior to its appearance. As Sidman (2000) suggested, negative

75 reinforcement and punishment involve the same events, but which function differently. It is often
M

76 unappreciated that negative reinforcement involves coercion, which is accompanied by negative

D

77 unintended outcomes as well (e.g., prevents an animal from relaxing its vigilance, causes
TE

78 fearfulness of novelty, and causes reluctance to explore) (Sidman, 2000). In contrast to training

79 methods that use positive punishment and negative reinforcement, other methods rely mainly on
EP

80 positive reinforcement which, according to Sidman (2000), is not coercive. Positive

81 reinforcement carries less risk of negative unintended outcomes.

C

82 The debate among trainers who tend to use positive punishment and negative
AC

83 reinforcement and those who prefer methods that rely on positive reinforcement is ongoing. For

84 example, in one editorial, Overall (2007) explained why electronic collars (which deliver

85 electronic shocks to the dogs and are usually used as positive punishers or negative reinforcers)

86 are not and should not be used for behavior modification in dogs, due to their aversive nature and
 ACCEPTED MANUSCRIPT
5

87 to the lack of scientific data on their effectiveness. However, others suggest that such collars can

88 be an effective training tool (e.g., Christiansen et al., 2001). While using positive punishment and

89 negative reinforcement can be effective, the question of whether using them is ethical or not is

90 open to debate. Friedman (2009) suggests that the relative intrusiveness of behavior modification

PT
91 techniques should be examined, and that minimally intrusive (but still effective) methods should

RI
92 be used. Friedman (2009) suggests that behavior interventions should not be chosen solely

93 because they are convenient or effective, since they may produce detrimental unintended

SC
94 outcomes in the learner.

95 Since there are contrasting opinions regarding the use of different dog training methods,

96
U
it is important to provide as much data as possible on this topic. Such data will allow
AN
97 practitioners to choose training methods wisely and thus provide effective and minimally

98 aversive behavior modification tools to dog owners and to those working with or caring for dogs.
M

99 Hence, the purpose of this review is (1) to review the literature regarding the effects of different
D

100 training techniques (e.g., positive punishment and/or negative reinforcement versus positive
TE

101 reinforcement) on dogs’ behavior and welfare, and (2) to suggest possible implications of the

102 research findings to other researchers and to those working with or caring for domestic dogs.
EP

103

104 Methods
C

105 A search for articles written in the English language was conducted using three
AC

106 computerized databases: Scopus, Google Scholar, and PubMed. A combination of the following

107 terms was used: punishment in dog training, aversion, punishment, shock collars, electronic

108 collars, choke collars, prong collars, dog training methods. A manual search of the reference

109 lists from the relevant articles was performed as well. The search was completed in October,
 ACCEPTED MANUSCRIPT
6

110 2016. Only articles that directly compared the effects of two or more training methods on dogs’

111 behavior and welfare were included. Articles that were published in non-academic journals were

112 excluded from this review. The search yielded 17 studies, which are reviewed in the current

113 article and are summarized in Table 1.

PT
114

RI
115 Results

116 The results are divided into four sections. The first section reviews articles that compared

SC
117 different training methods; the second section examines dog-to-dog aggression; the third section

118 reviews studies on the use of electronic collars or electronic pet containment systems; and the

119
U
fourth section examines the effects of aversive training techniques on the physical health of dogs.
AN
120

121 Comparison between training methods

M

122 One survey of 326 dog owners examined whether the use of different training methods
D

123 was related to the level of obedience and to the occurrence of behavior problems in dogs over
TE

124 one year of age (Hiby et al., 2004). The results revealed that punishment-based training methods

125 were related to a larger number of reported behavior problems compared to reward-based
EP

126 training. In addition, the highest obedience scores were reported by owners who used reward-

127 based training only, followed by those who used a combination of reward and punishment-based
C

128 methods, and lastly by those using punishment only. As Hiby et al. (2004) indicated, the results
AC

129 of this survey suggest that reward-based training methods are associated with both higher levels

130 of obedience and fewer behavior problems in dogs owned by a population of average dog

131 owners.
 ACCEPTED MANUSCRIPT
7

132 Another survey of 192 dog owners was conducted in the United Kingdom (Blackwell et

133 al., 2008). The training techniques were listed in three categories: (1) positive reinforcement, (2)

134 negative reinforcement, and (3) positive punishment. No techniques were categorized as negative

135 punishment. However, some of the training techniques reported as negative reinforcement should

PT
136 have been categorized as negative punishment (e.g., withdrawing of attention/time out, and

RI
137 withdrawing of rewards). Hence, the relationship between negative reinforcement and dogs’

138 behavior is difficult to interpret. Regardless, the use of punishment when training dogs was

SC
139 related to an increase in both fear and aggression. In contrast, using positive reinforcement only

140 was associated with the lowest scores on fear, aggression, and attention-seeking behaviors. Using

141
U
a combination of positive reinforcement and positive punishment was related to the highest
AN
142 aggression scores (Blackwell et al., 2008).

143 A third survey of 140 dog owners who scheduled an appointment for treating their dog’s
M

144 behavior problems in a veterinary hospital (Herron et al., 2009), revealed that direct and indirect
D

145 confrontational training methods were related to aggressive behavior. For example, yelling “no”
TE

146 at the dog was related to aggression in 15% of the cases (18 of 122 dogs); performing an “alpha

147 roll” (i.e., forcefully putting the dog on its back and holding it down) was related to aggression in
EP

148 31% of the cases (11 of 36 dogs); hitting or kicking the dog was related to aggression in 41% of

149 the cases (12 of 28 dogs); forcefully releasing an item from the dog’s mouth was related to
C

150 aggression in 38% of the cases(15 of 39 dogs); using a spray bottle was related to aggression in
AC

151 20% of the cases (10 of 51 dogs); and grabbing the dog by the jowls or scruff was related to

152 aggression in 26% of the cases (7 of 27 dogs). In contrast, using neutral or reward-based methods

153 was rarely related to aggressive behaviors.

 ACCEPTED MANUSCRIPT
8

154 A fourth survey of 3,897 dog owners found that compared to positive reinforcement and

155 negative punishment, the use of positive punishment and negative reinforcement was related to

156 an increased risk for aggression towards family members (odds ratio 2.9) and towards unfamiliar

157 people outside of the house (odds ratio 2.2) (Casey et al., 2014).

PT
158 Lastly, it is possible that factors such as a dog’s size, breed, age, and sex can affect

RI
159 owners’ behavior and choice of training methods. A survey of 1,276 dog owners compared the

160 relationship between training methods and behavior in small (<20 kg) and large dogs (>20 kg)

SC
161 (Arhant et al., 2010). As in previous surveys, positive relationships between the use of

162 punishment and excitability, and between punishment and aggression, were found in both small

163
U
and large dogs. A relationship between punishment and fearfulness was found in small dogs
AN
164 only. In addition, reward-based responses to behavior problems (e.g., calming or distracting the

165 dog) were related to aggression as well. Correlation values were low to moderate (usually <.3).
M

166 In addition, the cutoff weight for small and large dogs (i.e., 20 kg) may not have been specific
D

167 enough to reveal differences in dogs of different sizes.

TE

168 In summary, it appears that the use of training methods that are based on positive

169 punishment and negative reinforcement are related to higher incidences of behavior problems,
EP

170 aggression, and fear. Reward-based responses to behavior problems were related to aggression

171 only in one study (Arhant et al., 2010). It is important to note, however, that the five reviewed
C

172 studies (i.e., Arhant et al., 2010; Blackwell et al., 2008; Casey et al., 2014; Herron et al., 2009;
AC

173 Hiby et al., 2004) were based on the owners’ subjective answers to questionnaires, and are prone

174 to methodological difficulties such as recall bias and the misunderstanding of terms or questions

175 that were presented in the questionnaires. In addition, causality cannot be established from the

176 data in these studies, as it is not known whether the dogs’ aggression or the presence of behavior
 ACCEPTED MANUSCRIPT
9

177 problems led to the use of aversive training methods, or if the use of aversive training methods

178 caused aggression and other behavior problems. Hence, the results of such studies should be

179 interpreted appropriately and serve as a basis for future studies that are based on direct

180 observation or specific interventions, rather than as a basis for causal inference. In future studies,

PT
181 researchers should also take into account the dogs’ size as a covariate in their statistical analysis,

RI
182 as it can have an effect on dog owners’ choice of training methods.

183 Finally, at least according to one study (Blackwell et al., 2008), inconsistency in training

SC
184 methods was related to the highest aggression scores. As the authors of this study suggested, it is

185 possible that the inconsistency in training methods led to uncertainty or anxiety in the dogs,

186
U
which in turn led to aggressive behavior. It is also possible that this finding merely suggests that
AN
187 owners tried a variety of training methods to modify aggressive behavior. As this is the only

188 study that reported this finding, more research on this topic is warranted.
M

189 Studies with direct observations of dogs

D

190 Researchers in three studies directly observed the training techniques and the behavior of
TE

191 owners and their dogs (Deldalle and Gaunet, 2014; Haverbeke et al., 2008; Rooney and Cowan,

192 2011).
EP

193 In one study, one of the researchers visited the home of each of 53 dogs and their owners

194 and conducted several tests of behavior and obedience (Rooney and Cowan, 2011). All sessions
C

195 were filmed and a history of training methods used by owners was obtained. The dogs of owners
AC

196 who tended to use punishments showed less interaction with the experimenter. In addition, dogs

197 that were trained with physical punishment were less likely to approach a stranger and played

198 less with their owners. It was also found that dogs that were mostly trained with reward-based

199 methods scored higher in their ability to learn a novel task.

 ACCEPTED MANUSCRIPT
10

200 In another exploratory study (Deldalle and Gaunet, 2014), one of the researchers

201 observed training classes at two dog training schools. In one school, the dogs were trained using

202 positive reinforcement and in the other school they were trained with negative reinforcement.

203 The researcher observed and recorded the dogs’ behavior as they were walking on a leash and

PT
204 responding to the “sit” command from their owners. More dogs showed stress-related behaviors

RI
205 and low body postures in the negative reinforcement group compared to the positive

206 reinforcement group. In addition, fewer dogs gazed at their owners in the negative reinforcement

SC
207 group, compared to the positive reinforcement group [during walking on leash, 4% (1 of 26)

208 dogs) vs. 63% (15 of 24 dogs), respectively; during the “sit” command, 38% (10 of 26 dogs) vs.

209
U
88% (21 of 24 dogs), respectively]. These results suggest that dogs’ welfare may be threatened
AN
210 by the use of negative reinforcement. Moreover, the data on gazes towards the owner may

211 suggest that the relationship between owner and dog can be compromised when training with
M

212 negative reinforcement. However, it is also possible that the dogs from the positive
D

213 reinforcement school looked at their owners because they were waiting for a reward, or that they
TE

214 were reinforced more often for this behavior compared to the dogs from the negative

215 reinforcement school. It is important to note that positive and negative punishment were not
EP

216 evaluated in this study. Both the history of punishment and the acute punishments delivered

217 throughout the training sessions could have affected the results of this study. For example, when
C

218 using positive reinforcement, the withdrawal of food can be considered negative punishment. In
AC

219 contrast, in order for a stimulus to serve as a negative reinforcement, one has to apply positive

220 punishment first so that there will be an aversive stimulus to remove.

221 Lastly, a third study carried out evaluations of the performance of 33 military dog-

222 handler dyads when performing standardized obedience exercises and protection work
 ACCEPTED MANUSCRIPT
11

223 (Haverbeke et al., 2008). In general, the teams’ performance (i.e., score based on correct and

224 incorrect exercises) was relatively low, with ~66% success in obedience exercises and ~39%

225 success in protection work exercises. Dogs that received more aversive stimuli (either positive

226 punishment or negative reinforcement – e.g., pulling on the leash, hanging dog by collar, verbal

PT
227 scolding, hitting) were more distracted and showed poorer performance compared to dogs that

RI
228 received less aversive stimuli. In addition, the dogs showed a lower posture after the infliction of

229 aversive stimuli by their handlers. The authors of this study suggested that the welfare of dogs

SC
230 that received aversive stimuli during training was threatened, although this could not be directly

231 proven by the data.

232
U
In summary, the reviewed studies suggest that aversive training methods (e.g., positive
AN
233 punishment and negative reinforcement) may negatively affect the behavior and welfare of dogs.

234 Moreover, none of the studies showed any evidence that aversive training methods are more
M

235 effective than reward-based training. In fact, according to Haverbeke et al. (2008), the opposite
D

236 appears to be true. Hence, those working with dogs are encouraged to rely on reward-based
TE

237 training methods.

238 However, two limitations are noteworthy. First, while observational studies provide more
EP

239 robust data than surveys, they still do not necessarily provide support for causation. Second, in

240 the first study (Rooney and Cowan, 2011), only the researcher who visited the dogs rated their
C

241 behaviors. This researcher also interviewed the owners. Hence, as the authors mentioned, there is
AC

242 a risk for unconscious bias. Similarly, in the second study (Deldalle and Gaunet, 2014) only one

243 researcher recorded the behaviors of dogs while visiting the training centers. In the third study

244 (Haverbeke et al., 2008), the authors did not report whether more than one researcher rated the
 ACCEPTED MANUSCRIPT
12

245 behaviors. In future studies, two researchers – preferably blind to the hypotheses or study groups

246 – should code behaviors, and inter-rater reliability should be reported.

247 Lastly, since only two training schools were compared by Deldalle and Gaunet (2014),

248 individual variations could have led to some of the reported differences. Hence, as the authors of

PT
249 this study suggested, future studies should include a larger sample of training schools to

RI
250 compensate for these individual variations.

251

SC
252 Dog-to-dog aggression

253 Only one study examined the relationship between training methods and dog-to-dog

254
U
aggression (Roll and Unshelm, 1997). Questionnaires were given to dog owners with a dog that
AN
255 was injured by another dog (n=151) and dog owners with a dog that had injured another dog

256 (n=55) while they were in a veterinary clinic. The results showed that dogs that were trained by
M

257 hitting or “shaking”, and dogs owned by individuals who believed that without training a dog
D

258 will be out of control, tended to be on the aggressor side of the dog-to-dog aggression. In
TE

259 contrast, dogs of owners who believed that training should be fun and that it would be

260 advantageous to have a trained dog were found to be more often on the victim side. In addition,
EP

261 dogs of owners who shouted and gave clear commands were also found more often on the victim

262 side. The explanation for the last finding is not readily apparent. It is possible that the dogs that
C

263 were on the receiving end of shouting were more fearful in general and had been punished for
AC

264 previous displays of aggression.

265

266 Electronic collars and electronic containment systems

 ACCEPTED MANUSCRIPT
13

267 The use of electronic collars is highly controversial. Some trainers suggest that such

268 collars are effective in modifying behavior, while other trainers find them inhumane and avoid

269 using them. When possible, practitioners should base their choices of training methods on

270 scientific data. Hence, data on the relationship between the use of electronic collars and dogs’

PT
271 behavior are described in this section.

RI
272 Observational studies and surveys

273 In one study (Polsky, 2000), descriptive data about dog aggression that may have been

SC
274 elicited by electronic containment systems were collected. Five cases in which dogs inflicted

275 multiple uninhibited bites on humans in the presence of an active electronic containment system

276
U
were found. In all cases, the dog was an intact male with no prior displays of aggression. During
AN
277 the attack, all dogs failed to show gross warning signs (e.g., snarling, growling) before biting. In

278 four out of the five attacks, the dogs received a shock before the attack. These results should be
M

279 read with caution, as the data were collected from legal reports and the behavior was not directly
D

280 observed by trained professionals. Still, since all of the dogs had not shown serious aggression
TE

281 before the incident, it is plausible that the aggressive behavior was elicited by the shock, by the

282 classically conditioned response to a threat tone that precedes the shock, or by a response to
EP

283 other environmental stimuli that preceded the shock.

284 In contrast to the previous study, a survey of 3,897 dog owners did not find any
C

285 difference in undesirable behaviors among those using various training techniques (Blackwell et
AC

286 al., 2012). Results showed that only a small proportion of dog owners used electronic collars

287 (3.3%, N=133), electronic bark collars (1.4%, N=54), or electronic containment systems (.9%,

288 N=36). It is possible that these small numbers prevented a meaningful statistical comparison

289 between methods. As for training effectiveness, the reported training success for teaching a dog
 ACCEPTED MANUSCRIPT
14

290 to come when called or to prevent the dog from chasing was higher in the reward-based group

291 compared to the electronic collars group. However, this could be due to a number of

292 confounding factors, such as the seriousness of the problem behavior and owners’ perception of

293 success.

PT
294 Lastly, the effects of using electronic collars were examined in a study that directly

RI
295 observed 32 dogs that were trained as police service dogs or watchdogs (Schilder and van der

296 Borg, 2004). The researchers also observed the behavior of dogs that were shocked in past

SC
297 training with dogs that had never received shock before. Dogs trained with electronic collars

298 vocalized and presented body postures associated with stress or fear (e.g., tongue flicking,

299
U
lowering ear positions, lowering of the body/tail) for a fraction of a second after receiving an
AN
300 electric shock. Both during free walking and during training sessions without shocks, dogs that

301 were trained with shocks in the past showed more stress-related behaviors than dogs that were
M

302 not trained with shocks. In addition, shocked dogs were more stressed than control dogs on the
D

303 training grounds and also in a park unrelated to training. The fact that stress-related behaviors
TE

304 were seen outside the training grounds, but in the presence of the handlers, suggests that dogs

305 associate the possibility of getting shocked with the presence of their handlers. The authors of
EP

306 this study concluded that using electronic shocks for training is not only unpleasant, but is also

307 painful and frightening for the dogs. Hence, it appears that even dogs that make it through
C

308 demanding training programs suffer from the aversive training methods.
AC

309 Interventional studies

310 One study compared the application of shock to three groups of laboratory-bred Beagles

311 (Schalke et al., 2007). The first group of dogs received the shock precisely when they grabbed a

312 prey dummy. This led to an increase of ~22% and ~31% in absolute and relative salivary
 ACCEPTED MANUSCRIPT
15

313 cortisol, respectively. The second group received the shock if they failed to respond to a recall

314 while hunting the prey dummy (~114% and ~160% increase in absolute and relative salivary

315 cortisol, respectively). The third group received arbitrary and unpredictable shocks (~336% and

316 ~328% increase in absolute and relative salivary cortisol, respectively). The fact that the Beagles

PT
317 that were shocked unpredictably had extremely high cortisol levels is not surprising, as they

RI
318 could not predict and had no control over the coming shocks. In the second group, it is possible

319 that the elevated levels of cortisol were due to the fact that the recall was trained without a prey

SC
320 dummy but was tested with it. Hence, it was difficult for the dogs to control their first reaction to

321 chasing the prey. In the first group, the predictability of the shock could have led to the relatively

322
U
small increase in cortisol levels, but another explanation is possible. While an increase in the
AN
323 concentration of cortisol can represent an increase in stress (Dreschel and Granger, 2005), it can

324 also represent the physical activity level of the dog. Indeed, elevation in cortisol concentration
M

325 can occur as a result of both low-intensity and high-intensity exercise (Radosevich et al., 1989).
D

326 However, the dogs in this study (Radosevich et al., 1989), exercised for 90 minutes on a
TE

327 treadmill, and while plasma cortisol gradually increased with the duration of exercise, large

328 elevations were seen only after 15-30 minutes of exercise. Since the dogs in Schalke et al.’s
EP

329 (2007) study ran after prey for less than two minutes a day, and since plasma cortisol samples

330 were taken 10 minutes after the administration of the shock, it is unlikely that the short exercise
C

331 contributed significantly to the elevation in cortisol levels.

AC

332 A second study (Cooper et al., 2014) compared the behavior and cortisol levels of dogs

333 trained to come when called, in three training groups: (1) using an electronic collar, (2) training

334 without an electronic collar but by the same trainers from group 1, and (3) training without an

335 electronic collar by trainers who believe in reward-based training. The study found no
 ACCEPTED MANUSCRIPT
16

336 differences in training effectiveness between groups. The dogs that were trained with electronic

337 collars tended to spend more time in a tense state, carried their tails lower, interacted less with

338 the environment, and yawned more compared to the dogs that were trained without the electronic

339 collars by reward-based trainers. No differences in urinary cortisol levels were found between

PT
340 groups, but salivary cortisol levels were higher in the group trained by the reward-based trainers

RI
341 compared to the two other groups throughout the experiment. However, comparing the

342 difference between post-training cortisol to pre-training cortisol revealed no differences between

SC
343 the three groups, suggesting that these values were not due to training methods.

344 While the previous studies compared aversive and non-aversive training methods, the

345
U
effects of three aversive training methods on learning as well as on possible unintended
AN
346 outcomes were examined in a study of forty-two police dogs (Salgirli et al., 2012). The dogs

347 were required to heel while a person serving as a decoy tried to distract them and cause them to
M

348 leave the handler’s side. In a counterbalanced design, the dogs received either a pull on a pinch
D

349 collar, a shock from an electronic collar, or a quitting signal that was conditioned to signify the
TE

350 withdrawal of a reward. A similar number of dogs learned to disregard the distraction with the

351 use of the electronic collar (n=39) and the pinch collar (n=32), compared to only three dogs with
EP

352 the use of a quitting signal. A plausible explanation for these results is that the dogs receiving the

353 quitting signal did not understand what was expected of them in this specific setting. Indeed, the
C

354 training of the quitting signal was done with a toy and not with a provoking person. Expecting
AC

355 the dogs to generalize the quitting signal with a toy to a different scenario seems unrealistic.

356 Hence, it is not surprising that the quitting signal failed to elicit the required behavior.

357 Although not statistically significant, 64.3% (N=27) of the dogs showed an extreme

358 backwards ear position after being punished with the pinch collar, compared to 38.1% (N=16)
 ACCEPTED MANUSCRIPT
17

359 that were punished using the electronic collar. In addition, approximately 43% (N~18) of the

360 dogs being trained with a pinch collar or an electronic collar showed a lowering of their back,

361 and approximately 31% (N~13) of the dogs crouched. Vocalization was seen in approximately

362 60% of the dogs with the electronic collar (N~25) compared to approximately 23% of the dogs

PT
363 with the pinch collar (N~10) (exact values were not reported in the original study; values

RI
364 extracted from figures). Lastly, out of the four dogs that responded to the quitting signal, two

365 showed a backward ear position and one showed an extreme lowering of body posture and

SC
366 crouching.

367 Unfortunately, cortisol levels are difficult to interpret in this study. This is because actual

368
U
cortisol values and effect sizes were not reported, and since there are contradictions in the
AN
369 reporting of results. For example, the authors state that no significant differences in relative

370 cortisol levels were found between groups, but at the same time they report that the relative
M

371 cortisol level after the quitting signal was significantly higher than after the use of a pinch collar
D

372 (Salgirli et al., 2012, p. 534). Still, 17 dogs showed maximal cortisol values with the quitting
TE

373 signal, 15 dogs showed maximal cortisol values with the electronic collar, and 10 with the pinch

374 collar. Statistical significance was not reported for these values.
EP

375 The results of this study suggest that the use of positive punishment in the form of a

376 pinch collar or an electronic collar can have detrimental effects on dogs’ physical and mental
C

377 welfare. In addition, using negative punishment without clear or consistent instructions of what
AC

378 is expected of the dogs can lead to fear and stress.

379 Bark collars

380 Bark collars are a different type of electronic collars. Unlike regular electronic collars

381 that are operated manually by the handler, these collars are designed to automatically deliver a
 ACCEPTED MANUSCRIPT
18

382 shock every time a dog barks. One study (Steiss et al., 2007) compared the use of electronic bark

383 collars, lemon spray collars (i.e., instead of an electronic shock, a spray of an unfavorable odor is

384 sprayed during barking), and control (inactivated) collars. Dogs in both the electronic and spray

385 collars groups barked less than dogs in the control group. While no statistically significant

PT
386 elevation in plasma cortisol or in Adrenocorticotropic hormone (ACTH) was reported in any of

RI
387 the groups (perhaps due to the small sample sizes), calculations of effect sizes portray a different

388 picture. These calculations revealed an increase in plasma cortisol between the first day of

SC
389 wearing the activated collars and the acclimation stage to the collars (Cohen’s d = 1.3 and 1.7 for

390 the electronic collar and the spray collar, respectively). Similar effect sizes were calculated for

391
U
ACTH (Cohen’s d = 0.82 and 1.9 for the electronic collar and the spray collar, respectively).
AN
392 These effect sizes are considered large, and represent significant elevations in plasma cortisol

393 and ACTH as a result of wearing the activated collars.

M

394 The fact that bark collars can in some cases reduce barking was shown previously, in a
D

395 study that compared the satisfaction of owners when using either an anti-bark citronella collar
TE

396 (88.9% satisfaction; N=8) or an anti-bark electronic collar (44.4% satisfaction; N=4) in a sample

397 of nine dogs (Juarbe-Diaz and Houpt, 1995). However, the more relevant question is not whether
EP

398 these bark collars work, but rather whether there are other effective training methods that can

399 alleviate barking without the added stress that is associated with them.
C

400 In summary, except for one study (Blackwell et al., 2012), all of the observational and
AC

401 interventional studies reviewed suggest that the various types of electronic collars may pose risks

402 to dogs’ welfare. Indeed, Schilder and van der Borg (2004) showed the risks to dogs’ welfare

403 even when the collars are operated by experienced trainers. The fact that dogs associate shocks

404 with the presence of handlers (probably due to classical conditioning) is not surprising, and is
 ACCEPTED MANUSCRIPT
19

405 troubling. It could be argued that the association these dogs make between their handlers and the

406 painful shocks can make them less reliable in situations when faultless performance is most

407 needed. This putative association should be examined in future studies.

408 The smallest elevations in cortisol levels when using electronic collars were reported

PT
409 when well-timed shocks were delivered in a controlled environment (Schalke et al., 2007). While

RI
410 no data are available on this topic, it is unlikely that dog owners would have the necessary skills

411 or experience to use such collars, nor would they operate the collars in a controlled environment.

SC
412 Lastly, shock collars, even in the hands of the most experienced trainers, can only

413 provide information regarding what behavior not to perform. These devices do not give the dog a

414
U
choice of an alternative behavior to perform. Hence, given the available data and in order to
AN
415 avoid risking the dogs’ welfare, trainers should avoid using electronic collars when training

416 dogs.
M

417
D

418 The effects of aversive training techniques on physical health

TE

419 The psychological unintended outcomes of aversive training methods have been

420 examined, but the effect of aversive training methods on the physical health of dogs should also
EP

421 be examined. It appears that stress can be associated with aversive training methods. Increased

422 cortisol levels followed shocks from electronic collars (Schalke et al., 2007), and were found in
C

423 dogs with activated electronic and citronella bark collars (Steiss et al., 2007). Training
AC

424 inconsistency and the use of electronic or pinch collars were related to maximal cortisol levels

425 (Salgirli et al., 2012). Importantly, Beerda et al. (1998) reported that unanticipated stimuli such

426 as short electric shocks and sound blasts led to increased salivary cortisol in dogs. Low body

427 posture, body shaking, crouching, yawning, and restlessness were also indicators of acute stress
 ACCEPTED MANUSCRIPT
20

428 (Beerda et al., 1998). Lastly, Dess et al. (1983) showed a marked elevation in mean cortisol with

429 (258% increase) or without (400% increase) control over electronic shocks. Control over the

430 situation was assessed by allowing or preventing the dogs from pushing a lever to stop the shock.

431 Elevations in mean cortisol were seen both in both conditions: if the dogs could have predicted

PT
432 (291% increase) or could not have predicted (374% increase) the coming shock (Dess et al.,

RI
433 1983). Predictability was introduced by presenting an auditory tone prior to the shock or

434 shocking the dogs without signaling it beforehand.

SC
435 Such stress can affect dogs’ physical health. Indeed, a recent review of the effects of

436 stress on animals’ health suggests that stress is associated with various damaging changes to

437
U
physical health in dogs, including suppression of the immune system, gastrointenstinal problems
AN
438 (e.g., diarrhea, vomiting, decreased appetite), delayed puberty, and decreased sperm quality

439 (Mills et al., 2014).The studies reviewed here examined acute stress responses, but it is mostly
M

440 chronic stress that can negatively affect physical health in the long term (Beerda et al., 1997).
D

441 One study found a relationship between dogs’ fear of strangers – a possible powerful stressor,
TE

442 and a shortened lifespan (Dreschel, 2010). The chronic effect of stress due to aversive training on

443 dogs’ physical health should be a relevant topic for future research. More research is needed to
EP

444 clarify the relationship between dogs’ behavior and stress, since behavioral responses can vary

445 between individual dogs and between various stimuli, and can be misinterpreted.
C

446 It is also possible that using certain punitive techniques presents more direct health risks.
AC

447 One study showed increased intra-ocular pressure in dogs while pulling against a collar (Pauli et

448 al., 2006). Importantly, one extreme case report of the effects of a specific punitive technique on

449 the physical health of a one-year-old German Shepherd dog was found (Grohmann et al., 2013).

450 The dog was hung several feet in the air with a choke collar for approximately 60 seconds, and
 ACCEPTED MANUSCRIPT
21

451 subsequently lost consciousness. A few hours after this incident, the dog developed several

452 neurological symptoms (ataxia in all four limbs, circling to the left, disorientation). Magnetic

453 resonance imaging revealed severe cerebral edema due to ischemia. Tragically, the owner chose

454 to euthanize the dog. While this is an extreme case in the published literature, the author

PT
455 suggested that the punitive technique of choking a dog while hanging it in the air is not

RI
456 uncommon, and more cases such as this one may have gone unreported. Veterinarians and

457 trainers should be made aware that hanging a dog in the air or “helicoptering” it (e.g., lifting up

SC
458 by the choke chain and circling the dog in the air) presents a severe threat to dogs’ health and

459 obviously should be avoided. Legislators would do well to make such practices illegal.

460
U
AN
461 Discussion

462 The discussion is divided into two sections: (1) Methodological concerns, and (2)
M

463 Implications for practitioners and researchers.

D

464
TE

465 Methodological concerns

466 This section discusses four methodological concerns regarding the reviewed literature.
EP

467 Sample size and lack of the reporting of effect size

468 The reporting of effect sizes allows readers to assess the practical significance of group
C

469 differences and is not related to sample sizes. This information is of importance since the lack of
AC

470 statistical significance does not necessarily mean that differences between groups are not of

471 consequence. For example, in one of the reviewed studies (Steiss et al., 2007), calculation of

472 effect sizes revealed relatively large differences in plasma cortisol, although those were not

473 statistically significant. The lack of statistical significance was probably due to the small sample
 ACCEPTED MANUSCRIPT
22

474 sizes (i.e., 6-8 dogs in each of three groups). Statistical analyses in future studies should make

475 sure to report effect sizes in addition to null-hypothesis testing.

476 The reliance on surveys and observational data, and the ethics of randomized interventions

477 Many of the reviewed studies were based on surveys and observations. While these

PT
478 studies are valuable, they do not allow researchers to assess causal relationships. Hence,

RI
479 performing randomized controlled interventions in which dogs are randomly assigned to

480 punishment-based training groups versus reward-based training groups may be warranted.

SC
481 However, the results of this review, and the vast literature on punishment in general (e.g.,

482 Durrant and Ensom, 2012; Sidman, 2000), suggest that punishment comes with negative

483
U
unintended outcomes that can be detrimental to an animal. From an ethical perspective,
AN
484 researchers should be cautious before performing such interventions, and should ensure that the

485 dogs’ welfare is not threatened. Once these conditions are met, researchers can be encouraged to
M

486 continue learning about the effects of aversive training methods on dogs’ behavior and welfare
D

487 within the realm of observational studies. Direct observations of larger sample sizes, with a
TE

488 robust methodology of coding the observed behaviors (e.g., using two observers who are blinded

489 to the hypotheses, reporting inter-rater reliability), will allow researchers to provide enlightening
EP

490 data on this topic without endangering the dogs’ welfare.

491 Reporting case studies on physical health risks due to aversive training methods
C

492 Only one study reported a severe case of brain damage, due to hanging a dog in the air
AC

493 with a choke chain as a punitive technique (Grohmann et al., 2013). Since the use of such

494 aversive methods may still be prevalent, it is unlikely that this is the only case. In addition, many

495 punitive techniques involve pressure on the neck, which may lead to dangerous increases in

496 intra-ocular pressure (see Pauli et al., 2006). Veterinarians should be encouraged to publish case
 ACCEPTED MANUSCRIPT
23

497 studies in which dogs were injured due to the use of aversive training methods. Publication of

498 this type of information will allow those who work with dogs to understand the negative

499 physical symptoms related to aversive training methods, and perhaps reduce their occurrences.

500 The operation of electronic collars by experienced trainers

PT
501 In all of the observational and interventional studies that reported the use of electronic

RI
502 collars, experienced trainers operated the collars. The results of the studies suggest that even

503 when experienced trainers operate these collars, the welfare of the dogs could be compromised.

SC
504 However, most dogs are not owned by professional trainers, and the effects of regular pet owners

505 using such collars on dogs’ welfare are not known. Indeed, it is likely that the threat to dogs’

506
U
welfare would be even greater in the hands of unskilled dog owners, who might lack the timing
AN
507 and consistency needed for this type of training to be successful. In such cases, due to the

508 aversive nature of these devices and the likelihood of training ineffectiveness, their use can be
M

509 abusive.
D

510 Implications for practitioners and researchers

TE

511 Despite the methodological concerns, it appears that aversive training methods have

512 undesirable unintended outcomes and that using them puts dogs’ welfare at risk. In addition,
EP

513 there is no evidence to suggest that aversive training methods are more effective than reward-

514 based training methods. At least three studies in this review suggest that the opposite might be
C

515 true – in both pets and working dogs (Blackwell et al., 2012; Haverbeke et al., 2008; Hiby et al.,
AC

516 2004). Since this appears to be the case, it is recommended that the dog training community

517 embrace reward-based training and avoid, as much as possible, training methods that include

518 aversion. For this purpose, it is proposed that Friedman’s (2009) hierarchy of intervention

519 strategies may be a good tool for choosing the least intrusive, yet effective, behavior
 ACCEPTED MANUSCRIPT
24

520 modification tools. Friedman (2009) lists six levels of intervention: (1) arranging distant

521 antecedents (least intrusive); (2) arranging immediate antecedents; (3) positive reinforcement; (4)

522 differential reinforcement of alternative behavior; (5) negative punishment, negative

523 reinforcement, extinction; and (6) positive punishment (most intrusive). According to Friedman

PT
524 (2009), levels 1-4 are sufficient for solving the vast majority of behavior problems in animals.

RI
525 Level 5 may occasionally and under certain conditions be the effective and ethical choice. Level

526 6 is rarely needed or suggested when the practitioner has good teaching skills and the required

SC
527 knowledge of behavior. It may be noted that such ethical hierarchies of intervention that begin

528 with the least intrusive and end with the most intrusive are practiced in children’s education as

529
U
well (Carter & Wheeler, 2005). As Friedman (2009) suggests, it may be wise to borrow such
AN
530 guidelines from the field of applied behavior analysis, as both animals and humans who require

531 behavior modification are often vulnerable and frequently cannot protect themselves. If aversive
M

532 or intrusive methods are chosen, the competence of the handlers is critical in order that they may
D

533 achieve the proper timing and consistency required to allow for quick learning and to avoid
TE

534 abusing dogs and threatening their physical and mental well-being. Handlers’ competence should

535 be defined, regulated, and assessed by relevant regulating agencies based on the
EP

536 recommendations of accredited and experienced animal behaviorists.

537 One could rightly suggest that more studies with better methodologies concerning the
C

538 effects of aversive training methods on dogs’ welfare are needed in order to strengthen the
AC

539 evidence on this topic. However, the data emerging from the current review, as well as available

540 data on the negative unintended outcomes of aversive training methods in other species, such as

541 in humans (e.g., Durrant and Ensom, 2012; Sidman, 2000) and rats (e.g., Ulrich and Azrin,

542 1962), suggest that it is perhaps time to pursue a different focus and approach of research. This
 ACCEPTED MANUSCRIPT
25

543 new line of research will examine how humane, reward-based methods can be improved in order

544 to facilitate better communication between humans and dogs. In turn, such outcomes will allow

545 dogs to modulate their stress, and at the same time improve their ability to effectively understand

546 and respond to the behavior displayed towards them.

PT
547

RI
548 Acknowledgements

549 The author would like to thank Dinah Olswang and Daveena Tauber for their editorial

SC
550 assistance.

551 Conflict of interest statement

552
U
The author has no conflicts of interest relevant to the content of this review.
AN
553 Ethical approval

554 No ethical approval was required as this is a review of literature.

M

555 Authorship
D

556 The idea for the paper was conceived by the author. The paper was written solely by the
TE

557 author.
C EP
AC
 ACCEPTED MANUSCRIPT
26

558 References

559 Arhant, C., Bubna-Littitz, H., Bartels, A., Futschik, A., Troxler, J., 2010. Behaviour of smaller

560 and larger dogs: Effects of training methods, inconsistency of owner behaviour and level

561 of engagement in activities with the dog. Appl. Anim. Behav. Sci. 123, 131-142.

PT
562 Beerda, B., Schilder, M.B.H., van Hooff, J.A., de Vries, H.W., 1997. Manifestations of chronic

RI
563 and acute stress in dogs. Appl. Anim. Behav. Sci. 52, 307-319.

564 Beerda, B., Schilder, M.B.H., van Hooff, J.A., de Vries, H.W., Mol, J.A., 1998. Behavioural,

SC
565 saliva cortisol and heart rate responses to different types of stimuli in dogs. Appl. Anim.

566 Behav. Sci. 58, 365-381.

567
U
Blackwell, E.J., Bolster, C., Richards, G., Loftus, B.A., Casey, R.A., 2012. The use of electronic
AN
568 collars for training domestic dogs: Estimated prevalence, reasons and risk factors for use,

569 and owner perceived success as compared to other training methods. BMC Vet. Res. 8,
M

570 93.
D

571 Blackwell, E.J., Twells, C., Seawright, A., Casey, R.A., 2008. The relationship between training
TE

572 methods and the occurrence of behavior problems, as reported by owners, in a population

573 of domestic dogs. J. Vet. Behav.: Clin. Appl. Res. 3, 207-217.

EP

574 Carter, S.L., Wheeler, J.J., 2005. Considering the intrusiveness of interventions. Int. J. Spec.

575 Educ. 20, 136-142.

C

576 Casey, R.A., Loftus, B.A., Bolster, C., Richards, G.J., Blackwell, E.J., 2014. Human directed
AC

577 aggression in domestic dogs (Canis familiaris): Occurrence in different contexts and risk

578 factors. Appl. Anim. Behav. Sci. 152, 52-63.

579 Chance, P., 2003. Learning and Behavior. Wadsworth Publishing, Belmont, CA.
 ACCEPTED MANUSCRIPT
27

580 Christiansen, F.O., Bakken, M., Braastad, B.O., 2001. Behavioural changes and aversive

581 conditioning in hunting dogs by the second-year confrontation with domestic sheep.

582 Appl. Anim. Behav. Sci. 72, 131-143.

583 Cooper, J.J., Cracknell, N., Hardiman, J., Wright, H., Mills, D., 2014. The welfare consequences

PT
584 and efficacy of training pet dogs with remote electronic training collars in comparison to

RI
585 reward based training. PLOS ONE 9, e102722.

586 Deldalle, S., Gaunet, F., 2014. Effects of 2 training methods on stress-related behaviors of the

SC
587 dog (Canis familiaris) and on the dog–owner relationship. J. Vet. Behav. 9, 58-65.

588 Dess, N.K., Linwick, D., Patterson, J., Overmier, J. B., Levine, S., 1983. Immediate and

589
U
proactive effects of controllability and predictability on plasma cortisol responses to
AN
590 shocks in dogs. Behav. Neurosci. 97, 1005-1016.

591 Dreschel, N.A., 2010. The effects of fear and anxiety on health and lifespan in pet dogs. Appl.
M

592 Anim. Behav. Sci. 125, 157-162.

D

593 Dreschel, N.A., Granger, D.A., 2005. Physiological and behavioral reactivity to stress in
TE

594 thunderstorm-phobic dogs and their caregivers. Appl. Anim. Behav. Sci. 95, 153-168.

595 Durrant, J., Ensom, R., 2012. Physical punishment of children: Lessons from 20 years of
EP

596 research. Can. Med. Assoc. J. 184, 1373-1377.

597 Friedman, S., 2009. What's wrong with this picture? Effectiveness is not enough. J. App.
C

598 Compan. Anim. Behav. 3, 41-45.

AC

599 Grohmann, K., Dickomeit, M.J., Schmidt, M.J., Kramer, M., 2013. Severe brain damage after

600 punitive training technique with a choke chain collar in a German shepherd dog. J. Vet.

601 Behav.: Clin. Appl. Res. 8, 180-184.

 ACCEPTED MANUSCRIPT
28

602 Haverbeke, A., Laporte, B., Depiereux, E., Giffroy, J.-M., Diederich, C., 2008. Training methods

603 of military dog handlers and their effects on the team's performances. Appl. Anim.

604 Behav. Sci. 113, 110-122.

605 Herron, M.E., Shofer, F.S., Reisner, I.R., 2009. Survey of the use and outcome of confrontational

PT
606 and non-confrontational training methods in client-owned dogs showing undesired

RI
607 behaviors. Appl. Anim. Behav. Sci. 117, 47-54.

608 Hiby, E.F., Rooney, N.J., Bradshaw, J.W.S., 2004. Dog training methods: their use, effectiveness

SC
609 and interaction with behaviour and welfare. Anim. Welf. 13, 63-69.

610 Juarbe-Diaz, S.V., Houpt, K.A., 1995. Comparison of two antibarking collars for treatment of

611
U
nuisance barking. J. Am. Anim. Hosp. Assoc. 32, 231-235.
AN
612 Mills, D., Karagiannis, C., Zulch, H., 2014. Stress—its effects on health and behavior: A guide

613 for practitioners. Vet. Clin. North Am. Small Anim. Pract. 44, 525-541.
M

614 Overall, K.L., 2007. Why electric shock is not behavior modification. J.Vet. Behav. 2, 1-4.
D

615 Pauli, A.M., Bentley, E., Diehl, K.A., Miller, P.E., 2006. Effects of the application of neck
TE

616 pressure by a collar or harness on intraocular pressure in dogs. J. Am. Anim. Hosp.

617 Assoc. 42, 207-211.

EP

618 Polsky, R., 2000. Can aggression in dogs be elicited through the use of electronic pet

619 containment systems? J. Appl. Anim. Welf. Sci. 3, 345-357.

C

620 Radosevich, P.M., Nash, J.A., Lacy, D.B., O'Donovan, C., Williams, P.E., Abumrad, N.N., 1989.
AC

621 Effects of low-and high-intensity exercise on plasma and cerebrospinal fluid levels of ir-

622 β-endorphin, ACTH, cortisol, norepinephrine and glucose in the conscious dog. Brain

623 Res. 498, 89-98.

624 Reid, P.J., 1996. Excel-erated Learning. James & Kenneth Publishers, Berkeley, CA.
 ACCEPTED MANUSCRIPT
29

625 Roll, A., Unshelm, J., 1997. Aggressive conflicts amongst dogs and factors affecting them. Appl.

626 Anim. Behav. Sci. 52, 229-242.

627 Rooney, N.J., Cowan, S., 2011. Training methods and owner–dog interactions: Links with dog

628 behaviour and learning ability. Appl. Anim. Behav. Sci. 132, 169-177.

PT
629 Salgirli, Y., Schalke, E., Boehm, I., Hackbarth, H., 2012. Comparison of learning effects and

RI
630 stress between 3 different training methods (electronic training collar, pinch collar and

631 quitting signal) in Belgian Malinois police dogs. Rev. Med. Vet. 163, 530-535.

SC
632 Schalke, E., Stichnoth, J., Ott, S., Jones-Baade, R., 2007. Clinical signs caused by the use of

633 electric training collars on dogs in everyday life situations. Appl. Anim. Behav. Sci. 105,

634 369-380.
U
AN
635 Schilder, M.B.H., van der Borg, J.M., 2004. Training dogs with help of the shock collar: Short

636 and long term behavioural effects. Appl. Anim. Behav. Sci.e 85, 319-334.
M

637 Sidman, M., 2000. Coercion and its Fallout. Revised edition. Authors Cooperative Inc.,
D

638 Publishers.
TE

639 Steiss, J.E., Schaffer, C., Ahmad, H.A., Voith, V.L., 2007. Evaluation of plasma cortisol levels

640 and behavior in dogs wearing bark control collars. Appl. Anim. Behav. Sci. 106, 96-106.
EP

641 Ulrich, R.E., Azrin, N.H., 1962. Reflexive fighting in response to aversive stimulation. J. Exp.

642 Anal. Behav. 5, 511-520.

C
AC
 ACCEPTED MANUSCRIPT

Table 1

A Summary of studies of the effects of aversive training methods on dogs (N = 17, studies are presented by year of publication and

PT
name of first author).

Study Animals & Method Measures Results

RI
Participants
Roll and Dog owners with a Questionnaires completed by Sex Only descriptive statistics reported.

SC
Unshelm, dog injured by all owners in order to compare Breed Dogs that were trained by hitting or “shaking” tended
1997 another dog aggressors and victims. Age to have a history of biting other dogs.
(n=151) and dog Training methods A higher % of victim dogs were owned by dog owners

U
owners with a dog Degree of injury who shout and give clear commands and owners who

AN
that injured another Location of fight believe that training should be fun and that it is
dog (n=55) Owners’ advantageous to have a trained dog.
demographics A higher % of aggressor dogs were owned by owners

M
who believe that a dog will be out of control without
training
Polsky, Adult dogs (n=5) Descriptive data collected in Dog sex, age, All five attacks carried out by adult intact males

D
2000 the form of legal documents, reproductive between the ages of 2-3 years.

TE
animal control officers and status, breed, Adult victims were familiar with the dog, child victims
police officers’ reports. location of attack were not.
Case studies of aggression that relative to border In 4 out of 5 cases, the dog received a shock at the time
EP
was elicited by electronic pet of containment of the attack.
containment systems were system, and Not one of the dogs showed threatening behavior prior
examined. victim’s to attack.
C

familiarity with All attacks included repeated biting of victim.

AC

dog No gross warning signals were given before biting

Hiby et al., Dog owners Questionnaires given to dog Demographics. Using punishment positively correlated with problem
2004 (n=326) with dogs owners as they were walking Training methods behaviors.
over 1 year of age their dog or through veterinary Obedience level No correlation between use of reward-based training
clinics. Problem behaviors and problem behaviors.
Questionnaires returned by Number of behavior problems:
 ACCEPTED MANUSCRIPT

mail. Punishment only > combination of punishment and

reward > reward only or miscellaneous methods.
Punishment-based training was never the most
effective for achieving obedience goals.

PT
Overall obedience was related to reward-based but not
to punishment-based methods
Schilder Dogs that were Videotaping 107 shocks Direct effect of Direct reaction to shocks (e.g., lowering body, high-

RI
and van der trained for official delivered to 31 dogs and shocks on body pitched yelps, barks, squeals, redirected aggression,
Borg, 2004 certificate of police comparing body language to language. avoidance) lasted a fraction of a second.

SC
service (n=15). control dogs that did not Body language of Long-term effects: comparison of shocked (S) dogs and
Dogs that trained receive shocks. dogs that received control dogs (C).
with (n=16) and Comparing behavior of dogs shocks in the past During free walking, obedience, and police work:

U
without (n=15) that received shocks in the past during various lower ear postures and stress-related behaviors: S > C.

AN
electric shocks to dogs that didn’t. conditions in Differences were seen even when walking in a park
which no shocks outside the training grounds.
were given. Stress and lower ear postures: Training > free walking.

M
Schalke et Laboratory-bred Training to stop prey behavior. Heart rate absolute and relative cortisol levels:
al., 2007 Beagles between Three groups: (1) receive Salivary cortisol Group (1) – increase by ~22 and 31%, respectively.
the ages of 1.5-2 shock precisely at the moment Group (2) – increase by ~114 and 160%, respectively.

D
years (n=14) they touch a prey dummy, (2) Group (3) – increase by ~336 and 328%, respectively.

TE
receive shock when failing to
obey a recall during hunting, Increase in maximal heart rate in group (3) on days of
(3) receive arbitrary, training with shock.
EP
unpredictable shocks.
Steiss et Dogs from a private Dogs randomly assigned to 3 Barking Reduced barking in S and E groups compared to C.
al., 2007 no-kill shelter groups: control (C), electronic Activity No statistically significant changes in plasma cortisol
C

(n=21) anti bark collar (E), lemon Plasma cortisol and ACTH. However, large effect sizes of increased
AC

spray anti bark collar (S). ACTH plasma cortisol and ACTH in S and E groups on day 1
Baseline measurements and of wearing the active collar compared to baseline.
measurements after collar was
activated.
Blackwell Dog owners Questionnaires given to dog Demographics Training methods: 16% R+, 12% R+ and R-, 32% R+
et al., 2008 (n=192) with dogs owners walking their dogs or Training methods and P+, 40% combination of all categories.
 ACCEPTED MANUSCRIPT

between the ages of in vet hospitals in the United Obedience level 72% used some P+.
1-15 years Kingdom. Problem behaviors R+ only: lowest score on attention seeking behaviors,
fear, and aggression.
R+ and R-: highest score on attention seeking behavior.

PT
R+ and P+: highest score on aggression.
Reactivity to other dog and people: lowest in R+ only.
Fear and aggression problems higher in dogs that

RI
received any type of punishment
Haverbeke 33 dog and handler Standardized evaluation of 8 Team performance Team performance: 66% success in obedience and 39%

SC
et al., 2008 teams of the obedience exercises and 5 Handler’s behavior success in protection work.
Belgian Defense. protection work exercises to Dog behavior Use of R+ (57.12%) > use of aversive stimuli (R- and
Dogs between the assess teams’ performance. P+) (21.88%).

U
ages of 1-5 years Two evaluations with 20 days Use of aversive stimuli:

AN
and have been in between were performed. protection work > obedience
working between 3 Dogs divided into high- high-performance < low-performance dogs
months to 3 years performance and low- distracted dogs > slightly distracted, not distracted

M
performance groups. 2nd evaluation > 1st evaluation.
Dogs showed lower posture after aversive stimuli in the
2nd compared to the 1st evaluation

D
Herron et Dog owners who Survey regarding previous Frequency of % of dogs responding aggressively to confrontational

TE
al., 2009 scheduled an behavior interventions and intervention use training methods:
appointment for outcomes sent by email, fax, or Aggressive “Alpha roll” (31%), forced release of item in dog’s
behavior postal mail to owners. response due to mouth (38%), hit or kick dog (43%), grab jowls/scruff
EP
consultation in a Survey included a list of 30 intervention (26%), “dominance down” (29%).
veterinary hospital possible interventions. Effect of % of dogs responding aggressively to indirect
(n=140) Interventions categorized by intervention on confrontational training methods:
C

researchers to: aversive, behavior problem “Stare down” (30%), water pistol/spray bottle (20%),
AC

indirect confrontation, reward growl at dog (41%), yelling “no” (15%).

training, and neutral. Only between 0-6% of dogs responded aggressively to
neutral and reward based methods.
Arhant et Survey randomly Questionnaire with 237 short Dog behavior Using P+: small = large dogs.
al., 2010 sent to 5,000 dog questions regarding Training Higher frequency of punishments related to higher
owners. 1,405 demographics, and dog and techniques used aggression and excitability scores in small and large
 ACCEPTED MANUSCRIPT

surveys returned, owner behavior. dogs.

1,276 analyzed Comparison between small Relationship between punishment and aggression
(<20 kg) and large (>20 kg) stronger in small dogs.
dogs. Relationship between punishment and fearfulness and

PT
anxiety found only in small dogs.
Use of reward-based responses to unwanted behavior
related to higher frequency of aggressive behavior.

RI
All correlations < 0.3
Rooney Dog owners and Researcher visited and filmed Owner reported None of 54 the owners reported using reward-based or

SC
and their dogs (n=53) the behavior of each owner and training methods punishment-based methods exclusively.
Cowan, dog at their home while Dog behavior Proportion of punishment-based methods negatively
2011 following specific instructions Owner behavior correlated with dog interaction with experimenter.

U
including ignoring the dog and Proportion of reward-based training methods positively

AN
performing obedience correlated with dog performance at a novel task.
exercises. Using physical punishment: Dogs less interactive
Owners also completed a during play and less likely to interact with

M
questionnaire. experimenter (compared to not using physical
punishment at all).
Dog performance in novel task positively correlated

D
with total rewards delivered and owner patience.

TE
Blackwell 14,566 Questionnaires regarding Training method Only 3.3% reported using E-collars, 1.4% bark E-
et al., 2012 questionnaires demographics, choice of Problem behavior collars, and .9% electronic containment system.
distributed to dog training method and Demographics Higher % of owners using reward-based methods
EP
owners, 27% prevalence of undesired Training success reported success for recall / chasing problems (~97%)
returned (n=3,897) behaviors. compared to E-collar use (~83%) or other aversive
methods (94%).
C

Occurrence of undesired behaviors did not differ

AC

between training methods.

Salgirli et Adult police dogs Set-up of training exercise Salivary cortisol. Use of E-collar and pinch collar led to learning to
al., 2012 (n=42) with human decoys meant to Behavioral disregard the distraction while heeling.
distract the dog into breaking a observations % of dogs with behavior reactions to punishment:
heel. Repeated measures No statistically significant differences between pinch
design. Counterbalanced order (64.3%) and electronic (38.1%) for extreme backward
 ACCEPTED MANUSCRIPT

of three aversive stimuli: (1) ears.

quitting signal, (2) pinch Extreme low body posture: 4.8% of dogs for pinch, 0%
collar, (3) E-collar. One week E-collar.
between conditions for each Vocalization: ~60% E-collar > ~23% pinch.

PT
dog. # of dogs with maximal salivary cortisol values: 17
quitting signal, 15 E-collar, 10 pinch.
Highest cortisol concentration after quitting signal.

RI
Grohmann 1-year-old intact A case study of severe brain Description of Dog suspended a few feet in the air by a choke collar
et al., 2013 male German damage after punitive punishment for approximately 60 seconds. Dog panicked and lost

SC
Shepherd technique with a choke collar. Description of consciousness. After a few hours dog became ataxic on
symptoms all 4 limbs and was circling to left. Several
neurological symptoms.

U
MRI revealed legions which led to a diagnosis of

AN
severe cerebral edema due to ischemia.
Owner chose to euthanize the dog.
Casey et Same questionnaire Questionnaires regarding Risk factors for Compared to R+ and P-, Using P+ and/or R- were

M
al., 2014 as Blackwell et al., demographics, choice of aggressive related to increased risk of aggression towards
2012. training method and behavior members of the family and towards unfamiliar people
14,566 prevalence of undesired outside the house (odds ratio 2.8 and 2.2, respectively).

D
questionnaires behaviors.

TE
distributed to dog
owners, 27%
returned (n=3,897)
EP
Cooper et 63 dogs (no Examining effects of E-collar Behavioral and Training success: No differences between groups in
al., 2014 differences between use on dogs’ welfare. physiological owners’ satisfaction and perception of dogs’
groups in age, sex, Three groups trained for 5 days measures before, improvement.
C

and breed) divided (2 sessions per day) for recall during, and after Behavioral measures:
AC

to three groups in the presence of distractions: training Time spent in tense state: 1 > 2, 3
based on dog (1) E-collar use, (2) same Low tail carriage: 1, 2 > 3
characteristics and trainers with no use of E- Yawning: 1>3
past training history collar, (3) trainers who don’t Vocalization increase with E-collar intensity increase.
advocate E-collar use and no # of commands given: 1, 2 > 3 (twice as many)
use of E-collar. Interaction with environment:1, 2 < 3
 ACCEPTED MANUSCRIPT

Salivary cortisol:
3 >1, 2 throughout study.
Post-training minus pre-training cortisol:1=2=3
Deldalle Dogs in advanced Observation of owner and dog Performance While walking on leash:

PT
and training class from performance of a walk on leash Dog behavior % of dogs gazing at owner: R+ (63%) > R- (4%).
Gaunet, two schools: R+ and a sit command in advanced Owner behavior During the “sit” command:
2014 school (n=24), R- group classes. # of dogs showing mouth licking and yawning and

RI
school (n=26) showing at least 1 of 6 stress-related behaviors:
R- > R+.

SC
% of dogs gazing at owner: R+ (88%) > R- (38%).
% of dogs with low body posture: R- (46%) > R+(8%).
ACTH - Adrenocorticotropic hormone. R+ = Positive reinforcement. R- = Negative reinforcement. P+ = Positive punishment. MRI =

U
Magnetic Resonance Imaging. E-collar = Electronic collar.

AN
M
D
TE
C EP
AC
 ACCEPTED MANUSCRIPT

The Effects of Using Aversive Training Methods in Dogs – A Review

Highlights

PT
• Training methods based on positive punishment may have negative side effects

RI
• Training methods based on negative reinforcement may have negative side effects

• Trainers should rely on positive reinforcement based methods when training dogs

SC
• Aversive training methods should be avoided when training dogs

U
AN
M
D
TE
C EP
AC