ABSTRACT
Domestic dogs are trained using a range of different methods, broadly categorised as reward based (positive reinforcement/negative punishment) and aversive based (positive punishment/negative reinforcement). Previous research has suggested associations between use of positive punishment-based techniques and undesired behaviours, but there is little research investigating the relative welfare consequences of these different approaches. This study used a judgement bias task to compare the underlying mood state of dogs whose owners reported using two or more positive punishment/negative reinforcement based techniques, with those trained using only positive reinforcement/negative punishment in a matched pair study design. Dogs were trained to discriminate between rewarded and unrewarded locations equidistant from a start box, and mean latencies recorded. Their subsequent latency to intermediate ‘ambiguous’ locations was recorded as an indication of whether these were perceived as likely to contain food or not. Dogs trained using aversive methods were slower to all ambiguous locations. This difference was significant for latency to the middle (Wilcoxon Z = − 2.380, P = 0.017), and near positive (Wilcoxon Z = − 2.447, P = 0.014) locations, suggesting that dogs trained using coercive methods may have a more negative mood state, and hence that there are welfare implications of training dogs using such methods.
Source
Keep in mind that this is OWNERS using two or more methods based on a survey, not trainers.
Learning the unrewarded vs rewarded task did not have a difference in how quickly it was picked up, but the difference in latency where owners used two of the following methods: Bark activated citronella collar,
Remote activated citronella collar,
Pet corrector,
Physical punishment (e.g. smacking or shaking),
Remote activated electronic collar,
Bark activated electronic collar,
Water pistol,
Check or choke chain,
‘Rattle can’ or other sound based ‘distraction’ method
It did not differentiate between aversive methods, just that two or more of these were used.
I feel that studying laypeople rather than focusing on the methodology of "proper use" may be a good way to get a general understanding of the overall risks involved from day-to-day ownership, as a significant number people really don't go through a qualified trainer.