One of the most common arguments I often hear against clicker training for horses is that we “can’t use dog training methods for horses.” However, clicker training wasn’t originally invented for dogs (or dolphins either, as many people falsely claim). The principle of reinforcement, on which clicker training is based, was originally described and researched by B.F. Skinner, using lab rats and pigeons. The specifics vary a bit from species to species, but the underlying principles are the same. If positive reinforcement and clicker training work for dogs, horses, dolphins, pigeons and lab rats, then surely we can clicker train a goldfish!
I started experimenting some with fish training last Spring with my goldfish Blaze. We worked on some basic tricks using the R2 fish training kit, such as following a target, swimming through a narrow hoop and swimming through tunnels. I even made a short video of Blaze doing some tricks. All of our early tricks were taught using luring–Blaze was going where I wanted him to because he was interested in following the feeding wand, which was full of food. I got distracted by other projects and then was gone all of the summer, so I have not done much fish training since March.
Now, this past week, fish training has resumed. I am primarily interested in training some tricks using the principles of shaping. Training with shaping creates a different mindset for the animal than training with luring. Animals that are familiar with shaping are more confident in their behavior, more eager to offer new behavior, and aren’t as likely to get stuck if they don’t understand. As well, great shaping skills make a great trainer and this is one area where I want to keep practicing and improving. So, Blaze got volunteered to help me practice my shaping skills.
The first task, for any good clicker training should be the following two goals:
1) The animal clearly understands that the click means “Yes, that was right! Here comes some food!”
2) The animal understands how to get the food and does so quickly and efficiently.
The second goal is the downfall of many an otherwise good trainer. If there are holes in your food delivery process your training will progress slower and you’ll often end up with a frustrated animal. The animal should understand exactly how and where food is delivered and should go get the food and then immediately move on to the next response. Breakdown in the food delivery process is the trainer’s fault–the trainer hasn’t properly taught the animal the rules surrounding the food delivery process or is slow and clumsy with food delivery.
So, our first goal was effective food delivery. I started out getting Blaze use to the feeding wand again and practicing quickly and accurately delivering the food and reloading the wand. I use floating pellets as reinforcers, but I have to soak them to just the right amount before I can start training. Too little soaking and they float immediately to the top, which makes it harder for Blaze to follow them, too much soaking and they begin to fall apart.
Once our food delivery was going well, I added in my fish “clicker.” A clicker, or marker signal, doesn’t have to be a little metal box clicker. It could be a bell, a spoken word, a flashing light, or almost anything else. The key is to pick something that is distinct, short in duration, unique, and easy for the animal to recognize. For Blaze, I use a penlight that has a white light.
Now I am working on getting Blaze to associate the flash of the light with the presentation of the feeding wand. I think that he’s starting to understand at least something about the flashing light. When I was feeding from the wand the first couple of days, I was always doing it in the front right part of the tank. Once Blaze figured this out, he stayed almost exclusively in this area. Now that I have introduced the light, I have been feeding in the same part of the tank, but I have been clicking the light when he has been in the middle and at the other end of the tank. Today, he was swimming all over the tank during the session, rather than hovering around near the feeding spot.
Whether he understand the click well enough for me to be able to shape behavior, I don’t know. So, my next task is to try and shape or capture a simple behavior. I think I’m going to start by training him to go under the middle of the bridge. This is something he occasionally does, so I know he’s capable of doing it and isn’t afraid or hesitant to do it. Since he can do it, the real question is, can I teach him that swimming under the bridge is what earns him his food reward?
As with any good shaping plan, it’s best to start with small approximations and work slowly to your final goal. So, for now, I’m clicking when he is in the middle region of the tank where the bridge is. I’m also only trying to click if he’s more towards the bottom third of the tank. I’m not sure how long this will take, but I’ll keep you updated as to how our shaping goes!
I am glad to see you talk about training as not being species specific. One of the biggest misconceptions I find is that people always want to know how to train “x” (x=a particular species or breed).
If you have the principles of behavior modification–which I would not label as positive training (because both + and – are used in the process beyond the basics) you can train just about anything.
What is new in the current animal training world is not new at all. In fact, I’ve traced it back to the Hagenbeck zoo in the late 1800s when they used food rewards with their captive animals.
However, the scientific world tested and quantified the process which has morphed into “clicker training” aka “positive training.”
I’ll be focusing more on training on my blog in 2010.
You’re right, Ark Lady!
Animal training has been around for ages, but it’s only recently that we’ve begun to describe it in scientific terms.
One of the biggest advantages of clicker training that I see is that it improves the timing of the trainer and makes the trainer more aware of their actions. Leads to quicker learning and less frustration on the part of the animal! A skilled trainer, though, can often get the behavior they want with or without a clicker.
As for positive, training, I think positive reinforcement is the fastest and most humane way to get most of the behaviors we want. Novice and intermediate trainers are often not skilled enough to use punishment and negative reinforcement and often jump to these procedures too quickly.
I think the goal should be positive training, but we should recognize that some animals and some training situations might benefit from carefully applied punishment or negative reinforcement procedures.
For instance, negative reinforcement can be a great way to shape friendly behavior in fearful or aggressive animals, such as approaching and retreating from a fearful horse and retreating whenever the horse offers calm or friendly behavior.
Mary
Very cool to see a goldfish being trained Mary 🙂 I was interested to see you write that the “spoken word” can be a marker signal when my understanding and experience is that the spoken word doesn’t really work that well as a marker signal because we vary how we say even a simple word like “good” so much.
Good point, Jane!
I think in many instances, a click is much clearer and distinct as a marker signal. However, I do know some people who have been very successful with a verbal marker.
The key, I think, is that the marker must be easy for the animal to see/hear and be an accurate predictor of food delivery. For some dogs (and horses), I think they totally ignore the click and instead are attending to the hand reaching for the treat. So, reaching for a treat becomes a visual marker, rather than the click.
Mary