Animal Training

I have just finished Karolina Westlund Friman’s Advanced Animal Training course.  https://illis.se/en/

Much to digest and organise into usable training strategies, I need to be more disciplined in writing shaping plans and setting up the sessions.

Now I am going to do Stage 2 of the Natural Animal Centre Equine Behaviour Qualification. In preparation I am reading “Physiology of Behaviour” by N. Carlson (2013) which I managed to get as a PDF some time ago. That along with Paul Chances book on “Learning and Behaviour” ought to keep me busy for a while.

 

 

Punishment Callus

How to avoid problems with negative reinforcement.

If we use R- there are certain things we can do to negate possible negative side effects.

1. Use very low level aversive stimulus to form a behaviour. DO NOT trigger a flight response or use escalating aversive stimuli. Horses can habituate to the stimuli and then more is needed for the required reaction.

2. Remove the aversive stimulus as soon as the animal starts the behaviour, DO NOT wait until the behaviour is finished.

3. Use shaping – so reinforce small approximations of the final behaviour

4. Put the behaviour on a command so the animal can avoid the aversive stimulus when they respond. Either verbal or visual, be consistent and predictable.

5. Don’t use the same command for different behaviours.

6. Use a combination of R- and R+ e.g form the behaviour using an aversive stimulus and remove the stimulus and give a appetitive reinforcer at the same time -.e.g food or scratches.
This may counter-condition the aversive stimulus rather than cause poisoned cues.

An interesting phenomenon is Punishment Callus

“Animals habituate to aversives – this is referred to as the punishment callus. So, escalation is typically not the best solution.” Karolina Westlund Friman

Punishment Callus

When the animal habituates to an aversive stimulus and stops responding. So people escalate the aversive stimulus but the animal may habituate to that too. What have you got left for an emergency situation?
We don’t want animals to habituate – aversives need to be kept for emergency (life and death situations) use only.

If we use mild negative reinforcement e.g a light touch until the animal moves and there is no emotional reaction to the stimulus, it is still aversive but not enough to cause flight or avoidance reactions. Just enough for the animal to want it to stop, it then may do no harm and used in combination with a positive reinforcer may counter condition rather than poison the  cue or result in punishment callous. The aversive stimulus must be extremely low level and not escalate.

This is why the LIMA principles and the Humane Hierarchy make more sense than starting with mild aversives and escalating as the animal habituates. It is like the old riding school ponies who had been kicked so much they habituated and switched off.

PS I still prefer to use positive reinforcement but it is good to know that the odd time we use negative reinforcement like this we do very little harm to our relationship.

Don’t Be Callous: How Punishment Can Go Wrong

http://stalecheerios.com/horse-training/clicker-training-clinic-notes-happy-horses/

https://m.iaabc.org/about/position-statements/lima/

7 ways to get behaviour

More on poisoned cues – see previous post too



http://www.equineclickertraining.com/articles/negative_reinforcement_new.html#poisoned%20cues

http://stalecheerios.com/training-concepts/thoughts-poisoned-cues/

http://shoutout.wix.com/so/90c220b7-7aff-4676-b909-493f2ba082f6#/main

Poisoned cues

This is a really important topic for any animal training, whether you use positive or negative reinforcement but especially if you mix the 2.

Mixing can just be using a mild aversive to form a behaviour then clicking and positively reinforcing the behaviour, this can counter condition the aversive stimulus. This has to be done very carefully to avoid poisoned cues and punishment callus.

“Praise the horse or dog when he gets it right. Correct him when he gets it wrong. That describes many styles of standard animal training. It also describes the perfect set-up for poisoned cues. When you link corrections with motivation, you get problems—that’s what the basic research tells us. The bottom line is this: corrections do not work.” Karen Pryor

Back up is an example – do you use pressure and then say back? So eventually “back” is the command.
Or do you use a target stick and allow the horse to follow that back, reward and say “back” for the cue?
The verbal cue would be the same but the emotions of the horse different. If we mix them and use both aversives and appetitives for the same command/cue you can see how the horse may experience conflicting emotions.

We need one cue for one behaviour, but if we mix quadrants the horse can feel unsure what the consequences of the behaviour will be. We need a separate cue for a positively reinforced behaviour – do not use the command you used if the behaviour was previously taught with negative reinforcement. It is the command/cue not the behaviour that becomes poisoned.

There are some links to articles and an excellent podcast at the end of this post.

https://www.clickertraining.com/node/164

http://www.equineclickertraining.com/articles/negative_reinforcement_new.html#poisoned%20cues

http://stalecheerios.com/training-concepts/thoughts-poisoned-cues/

http://shoutout.wix.com/so/90c220b7-7aff-4676-b909-493f2ba082f6#/main

Next post will be about how to avoid some of the pitfalls of negative reinforcement and punishment callus.

Know your animal.

I am doing a course with Karolina Westlund Friman on animal training, a generic course for all animal trainers.  https://illis.se/en/

In one module we learned how important it is to know the species you work with. This brought me to revisit a blog I did a while ago about Fixed Action Patterns and share one by Mary Hunter on how this can be used in training.

From my Positive Horsemanship blog – first published in 2014.

Kondrad Lorenz and Niko Tinbergen received the Nobel Prize in 1973 for their work in developing an interpretive framework that crystallised the data they collected on animal behaviour in the field (ethology) and in the laboratory (neuroethology).
They observed what animals do and how and where the individual animals spent their time. They recognised that the behaviour of animals seemed to be constructed of elementary motor and sensory units. (Reference Animal Physiology – Eckert 4th edition 1997).

Motors units were called Fixed Actions Patterns – now called Modal Action Patterns as they aren’t as fixed as first thought.

The six properties of fixed/modal action patterns:
1. they are complex motor acts, each consisting of a specific temporal sequence of components – they are not simple reflexes.

2. they are typically elicited by specific key stimuli rather than general stimuli.

3. fixed action patterns are normally elicited by an environmental stimulus: but if the experimenter removes the stimulus after the behaviour has begun, the behaviour will usually continue to completion. This all or none property distinguishes them from simple reflexes.

4. the stimulus threshold for fixed action patterns varies with the state of the animal, and the variation can be quite large.

5. when they are presented with the appropriate stimulus, all members of the species (perhaps that are the same age, sex or both) will perform a given fixed action pattern nearly identical.

6. fixed action patterns are typically performed in a recognisable form even by animals that have had no prior experience with the key stimulus. That is these patterns are inherited genetically, although in many species the patterns can change with experience.

The last property has provoked the debate about nature versus nurture and recently epigenetic studies.

Useful blog about using this knowledge in training.
http://stalecheerios.com/other-species/modal-action-patterns-influence-behavior/

Ditch the Dominance

This article is interesting and adds to the debate about dominance based training or even those based on leadership concepts.

Don’t Be So Dominant During Training – The Horse

“The “alpha” concept of showing dominance when training a horse doesn’t coincide with what equitation science research is revealing, scientists say.”
thehorse.com

This is the original research – http://www.j-evs.com/article/S0737-0806(17)30005-9/fulltext

Quote – Highlights

“It is unlikely that horse–horse social status translates to analogues of human–horse interactions.
The concept of leadership as advocated in many training manuals proves to be unreliable in the horse.
Horses’ responses to training are more likely a result of reinforcement rather than a result of humans attaining high social status and a leadership role.
Knowledge of horses’ natural behavior and learning capacities are more reliable in explaining training outcomes than the application of dominance and leadership concepts.”

So it can be seen that we need a good understanding of equine ethology and learning theory as it relates to training horses.

Liberty and why I don’t use a whip.

Liberty and why I don’t use a whip.

I think the danger is that people misuse the whips, sticks or ropes and chase horses or threaten them. This is common in some forms of liberty training that use negative reinforcement.
Whips and ropes and even people can become conditioned aversive stimuli, so just the presence of the tool or person affects how the horse reacts.
I used to carry a dressage whip when riding my mare, I never touched her with it but she knew it was there (previous training had taught her how to avoid the whip by being obedient but it was a fear avoidance response). Do we really want to train using avoidance? Or do we want a partnership with the horse, horses can become excellent puzzle solvers once we give them a choice.

So for me (and it is a personal choice) I only use a stick as a target to form behaviours. Once on a cue we can fade out the target and use a variable schedule of reinforcement, and a variety of reinforcers – these may be scratches, food or even a favourite behaviour.

Horse and handlerhorse being postively reinfrced

Horses do like to play, as long as we can keep them under their emotional thresholds. Too much activation of the SEEKING system can also cause distress, just as activation of the FEAR (flight response) system can.
If we use aversive stimuli to train then horses can become very vigilant as they work out how to avoid the whip, stick or rope. This hyper-vigilance is exhausting and often we see very animated horses with liberty trainers and as soon as the trainer stops the horses seem to go to sleep, people then mistake this for a happy relaxed horse. We must be careful to watch the horses emotional state in all training.

Negative reinforcement isn’t necessarily a bad thing, if used sparingly and correctly. However we must be aware of how and why any reinforcement works and how it affects the horse.

Positive reinforcement stimulates the cognitive area of the brain, whereas negative reinforcement stimulates the flight/fight response especially in situations where the horse is driven away or chased with a whip.

So I start with the horse at liberty, rather than training with aversive stimuli until the horse learns he can’t leave. This is often what happens in many forms of liberty work, the horse is trained to stay because the consequences for leaving are something he wishes to avoid.
No pressure halters, ropes or round pens are needed when we really give the horse a choice using positive reinforcement