Tuesday 30 October 2012

The trouble with pressure-release




'Pressure-release training' and 'negative reinforcement training' are the same thing by different names. 

If I'd been strictly correct in what I'd said in the picture, I would have said, 

"the problem with negative reinforcement training is that aversives must be present in order to be removed".

The character count may not have violated twitter's stipulations but would you have understood what I'd said?

Now my problem with negative reinforcement since we understand that it is the same as 'pressure-release': That the horse learns to do a desired behaviour by the trainer releasing pressure (or other 'aversive' stimulus) when the horse is doing that desired response. In actual fact from the horse's point of view, the desired behaviour enabled the horse to escape something he found aversive. He learned that  the desired behaviour was effective in this matter and that it was worth doing again!

Pressure-release is a common term, but also (I believe) a misnomer. If we are using it in place of the scientifically correct term 'negative reinforcement' then 'aversive-removal' is probably closer to the mark. This is because pressure isn't the only kind of aversive stimulus that can be released just when the horse is doing what the trainer wants. Perhaps this just doesn't sound as comfortable as 'pressure-release'. 


As an aside I consider the term 'pressure-release' to be a 'reinscription of technology' (I read a lot from the social sciences with respect to animal welfare while undertaking my MSc thesis!). Reinscribing technology is a way of re-wording something to make us feel better about it. A more common example might be saying that a bit is 'stronger' instead of more 'severe', as stronger doesn't sound as painful! Buying a stronger bit must be easier on the mind than buying a more severe one.

The above digression aside, my point is that horses who find a particular stimulus aversive are motivated to avoid it. In order to make sure that they can, they pay attention to all that predicts it. When they've done this they will most likely avoid those stimuli too, since they have become classically associated with the original learning experience. Just as Pavlov's dogs learned that a bell predicted the arrival of food, horses (and dogs) can learn what predicts the presence of something they don't like or are afraid of.

So my problem with negative reinforcement is that aversive stimulation (of whatever nature) must be used in order to be removed, and it will be readily linked to anything else the horse noticed about the learning environment at the time. I have experienced this over the years that I've worked with horses, and it inspired my MSc research. In that I demonstrated that even mild aversive pressure (just a little pull on a lead rope) can increase the time it takes for a horse to voluntarily approach the trainer afterwards - just as Sankey et al found (see ISES proceedings from the 2008 Dublin meeting). www.equitationscience.com

Of course while I have this problem I accept that aversives happen and I aim to reduce that as best I can - of course sometimes I don't see them coming and the horse has to alert me to that sorry fact! I try to learn from it and move on.

1 comment:

  1. this came up on my memories newsfeed- great to read it again !

    ReplyDelete