About six months after I decided to start training Kit I ran into a fairly major problem. I am not sure how it happened, but Kit came to associate my presence with being afraid and so every time I went into his cage he would run and hide and refuse to come out until I had gone again. Not exactly ideal for training. It didn’t matter how non-threatening I was when I went into his cage. He had formed a superstitious behavior. Hiding makes him feel safe and makes the scary people go away. How could I change this habit?
I had thought for years about using a clicker to tell Kit he would be safe. But at the time I didn’t have a strong understanding of learning theory and conditioning and I was looking at it backwards. I wanted to jump right into telling him he would be safe, but how would I make sure my signal actually predicted a feeling of safety?
Once I thought about it in those terms it was obvious. Kit wanted me to leave him alone. So I introduced two sounds. When I approached his cage I made a kissy sound so he wasn’t caught unawares and could go and hide. Then as I left the cage I made a clicky sound. I paired my departure with a signal and this signal came to be interpreted as a safety signal. I didn’t know there was such a thing as a safety signal in the literature at this point, but a friend pointed me in the right direction and that really helped crystallise in my mind what I was trying to achieve.
Safety signals, or learned safety, are one of those odd little phenomena in learning theory that haven’t received a huge amount of attention. It’s only just recently got a second look by scientists in relation to depression. Robert Rescorla was the first to extend Pavlov’s work from conditioned reflexes to conditioned inhibition. When the absence of an unpleasant experience is paired with an unconditioned stimulus, such as a sound, then the sound comes to predict a period of safety. This is most easily achieved by pairing the US with the ending of an unpleasant experience. Fearful behavior is inhibited and it appears that intense negative emotions such as despair are lessened.
Care in interpretation is perhaps required here, though, as always when talking about emotions in animals. The important bit is that once a safety signal is learnt, its presence can inhibit fear both in a known situation the animal has experience with and perhaps to a lesser extent, new situations. Don’t forget this is a learnt signal, though. It needs to be maintained, and if it does not clearly signal safety, it will lose its effectiveness. This is a big challenge for its practical use in my opinion. It’s like training a recall. It’s most effective when built up slowly and frequently reinforced.
I really like the safety signal. I believe it helped Kit a lot. My first clue that his fear was fading was seeing his ears swivelling around listening for that signal. He was clearly paying a fair bit of attention to it. I still use it, but I am very careful how and when. Kit pays less attention to it now, but I use it to tell him nothing at all is going to happen for the next 10 seconds, and sometimes I drop it in when he is a bit worked up.
What I don’t do is use it when he’s dashing around in a panic. The moment to use it is either when he’s considering whether he needs to dash around in a panic or when he comes down from his panic enough to slow down a bit and start paying attention to his surroundings again. But it does seem to make him pause when he’s on the brink of a panic. Pausing is 90% of the way to avoiding a panicky dash by my reckoning.
To read more about safety signals, try Pollack et al (2008) An Animal Model of a Behavioral Intervention for Depression in Neuron, volume 60, pages 149-161. It’s currently free online here.
Rescorla’s chief paper on it is: Rescorla, R.A. (1969). Conditioned inhibition of fear resulting from negative CS-US contingencies. J. Comp. Physiol. Psychol. 67, 504–509. Another early article of Rescorla’s is at the time of writing available free online: RA Rescorla (1969) Pavlovian Conditioned Inhibition, Psychological Bulletin, volume 72(2).
Rogan resurrected the safety signal in this paper: Michael T. Rogan, Kam Sam Leon, David L. Perez and Eric R. Kandel (2005) Distinct Neural Signatures for Safety and Danger in the Amygdala and Striatum of the Mouse, Neuron, Volume 46, Issue 2, 309-320.