Last week I took part in a clinic with Marlitt Wendt
- a German ethologist and clicker trainer. Since it was a really very interesting experience for me, I'd like to share what I learned from her. It was not so much the novelty of things she said, but she kind of filled the blanks I had in some spots of positive horse training. For instance, she connected for me two different applications of reward based training that are very dear to me - default clicker training and Romy's style of interaction with horses.
The following is my interpretation of what I've learned and I would be very interested in your opinions about it...
Since the clinic was on an intermediate level, Marlitt didn't talk about the clicker basics, but rather emphasised on alternatives to operant conditioning, or at least a broader view of it.
Since clicker training was developed by experts, training mostly exotic animals in zoos, their work is characterised by a highly controllable environment and sessions with a clear timeframe. Clicker training as one can learn from most books, is therefore reflecting those ideal conditions and the requirements to train one precise behaviour at a time.
Horse owners though usually find themselves in a totally different position. They interact with their horses in environments that are not under their control (like on a walk, or in a barn together with other people and horses) and their interaction expands far from short training session in the arena. They interact while grooming, leading, walking, riding, feeding, and so on.
Also the emotional involvement is much more intense than when working with an animal in a zoo. Horse owners usually want to have a sound every-day relationship and not just a set of behaviours.
Those concepts that are so important for horse owners - like friendship, respect, harmony - are very fuzzy states of interaction and almost impossible to split up into distinct behaviours that one could mark and reinforce by operant conditioning. So to reinforce those states as a whole, one would need a more fuzzy method of reinforcing.
Marlitt suggested to have different marker signals for instance. Like a sharp clicker sound to mark a precise criterion of a behaviour, and another, longer marker (a word for example) to mark more fuzzily, or even omitting the marker signal completely to get even more fuzzy.
With a horse's growing history of operant conditioning, this mode of interaction is so well trained, that problems may arise. Some horses get very excited (or stressed) by just the expectation of being clicker trained. This kind of interaction is based on a question being raised and expectations to be fulfilled. It can be actually quite demanding for the animal.
Another problem can be that some horses (like our Lily
) have very holistic view on an interaction. It seems they receive so much sensory input which they feel is relevant to the trained behaviour, that there is a high potential of frustration and/or training of superstitious behaviours.
A remedy for those problems can be to shift the emphasis from operant conditioning to classical (Pavlovian) conditioning for certain tasks. The advantage is that not a specific behaviour is reinforced, but more a state. Like "being together", or "moving in synchrony", or "just relaxing". It's possible to clicker train every little sign of relaxation, but that requires a high level of proficiency and timing for the trainer and a high level of abstraction and generalisation for the horse.
If I just feed my horse in a casual way while we are standing together quietly without expectations, that would reinforce the whole situation and probably lead in the long run to a classical conditioned response of relaxation every time I am present and behave as I did in that situation.
To clearly differentiate the classical from the operant conditioning "mode", I think it would be best to try and avoid all kinds of markers, like a snappy move to the treat bag, a meaningful timing in relation to certain behaviours, or even the anticipation of behaviours that could be marked. What would be helpful instead is a fuzzy focus beyond the horse, soft flowing movements, casually giving out rewards and a general appreciation of the situation as a whole.
I've seen lot of things I've mentioned at work when I watched Romy interact with our horses and I've already started to experiment with our horses myself - to great effect I believe. I will write about it in my diary...