The Bubble Blog » The Red Green Show
The Red Green Show
Posted on March 07, 2009 by freshwater
No, this isn't about silly, cabin-bound Canadians. It's about silly, tank-bound goldfish.

Actually, Andi isn't particularly silly. She's smart and a fast learner. But, like any student, she sometimes gets into a rut.

We're still laboring with weave pole training. I gave up on the 2x2 method; it just wasn't working for Andi. Instead I connected our 4 weave poles in a straight row, and I've been using a little luring and a lot of shaping to teach Andi. She's making good progress, but she has one Achille's heel (er, fin). Because earlier tricks involved swimming in circles through a hoop and a tunnel, Andi is convinced that she's meant to do the same with weave poles. She has a nearly irresistable urge to start her weave run in the middle pair of poles. This just won't do; the whole point of weaving is to start at one end and weave through until she reaches the other end. Shaping her to start at the end is going very slowwwwly.

Now, operant conditioning relies on a click to mark desired behavior. But that's not the only tool available to trainers. With my dogs I use a combination of clicks ("Yes! That's the right behavior!") and No Reward Markers (NRMs) ("Oops! That's not right.") I do, in fact, sometimes say "Oops!" when the dog does the wrong thing. (Note: this is not a scolding; it's a neutral "Oops!")

One has to be careful, I think, to use NRMs sparingly. A constant chatter of clicks and NRMs can just confuse the animal. I tend to use them when the dog seems to be mentally stuck.

I haven't been using NRMs with my fish, because I thought it might make the system too complicated. But the more I work with fish, the more I believe their little brains are capable of intricate learning. So I decided to start using an NRM with Andi.

Obviously, saying "Oops" won't be of much use to a goldfish. They probably hear my voice, but not clearly. So I am using a red penlight flash as an NRM and a green penlight flash as the click. I taught Andi to understand her NRM thusly:

First I taught her to touch a green target. (A green soda straw through a suction cup stuck to the outside of the tank.) A green target touch earned a green light "click" followed by a treat. Then I added a second, blue target to the side of the tank. Each time she touched that, I flashed the red light and did not give her a treat. It only took a few reps for her to understand that a red flash meant she had not done the right thing. She very quickly stopped trying to touch the blue target and consistently went for the green.

Now I am applying this red light/green light protocol to Andi's weaves. When she begins a run by going in the middle pole pair, I flash the red light. Several times she has stopped cold, avoiding going in the middle. So far, so good! I just have to be careful not to use the No Reward Marker too much. I want rewards to come more frequently than "oopses." And I don't want her to ALWAYS be averse to going through the middle poles, since she needs to do that in the middle of a weave run.

This is an interesting experiment, though. Andi's success rate will tell me whether the NRM is ultimately a help or a hindrance to learning.