Home » Library » Learn » Training Theory

Training a Fish: Goldfish-Click

Filed in - Training Theory

Note from Karen Pryor: Ogden Lindsley was one of B.F. Skinner's first graduate students, a past president of the Association for Behavior analysis, and one of the first behavior analysts to grasp the power of shaping with a conditioned reinforcer. As a professor at the University of Kansas, he required his own students to shape behavior; many of them used goldfish. The instructions here for clicker training a goldfish are easy to follow and make a good science project.

As a professor, I required my students to buy their own animal to shape or use a household pet. When their animal was successfully performing on cue, in changed environments with observers present, the student brought his/her animal to class to perform and receive credit for the required TEBRO project. TEBRO meant "Teaching By Reward Only." I renamed "shaping" TEBRO in an attempt to stop students from using the clicker to stimulate or lead behavior as in the "warmer/colder" game. The click should always follow the shaped response class.

The most popular purchased animals were goldfish, which learned rapidly but often died. Some deaths were caused by buying the fish at Walmart or Kresge's at night, when the regular fish guy was at home. The night-time clerks often netted the fish too roughly, scraping the protective slime from their bodies. This slime protects fish from infection. The scraped fish almost always developed scale patches and died. Other deaths were caused by putting the fish in water that was not at room temperature, or had not been let sit overnight to permit the chlorine to escape.

A second problem was contaminating the tank or bowl with too many or too large reinforcements. It is best to lay a glass plate over the tank top, and spread the gold fish food over the plate surface. Then it is easy, using a table knife, to slide just one particle off the plate and into a floating glass food triangle on the water surface.

The most common goldfish performances were:

  1. Swimming under and through a ceramic bridge, castle, or loop made from bent coat-hanger wire.
  2. Swimming in a circle in open water.
  3. Swimming backwards.
  4. Swimming to a finger target placed against the bowl.
  5. Following a laser pointer spot on the pebble floor of the tank.

You must be very, very careful to keep the laser in front of the fish and not let it hit their eyes. The fish are primarily visual and their eyes are very sensitive. We eventually dissuaded the use of lasers because it was so hard to keep them off the fishes' heads and out of their eyes.

The most common "do it" cues were:

  1. Flashing a table lamp nearby.
  2. Flashing a flashlight aimed at the tank.
  3. Placing a finger against the bowl in the same spot.

The most common "did it" cues (clicks) were:

  1. Flashing a flashlight beam at the tank.
  2. Tapping the tank with the edge of the Veeder root hand counter, or table knife to be used to knock a food pellet off the glass plate.

Some students used a green table lamp bulb for the "do it" cue and a white flashlight flash for the "did it" cue (click).

It should take fewer than two weeks of 15-minute daily training sessions to reach your fish performance goal. A short morning session and a short evening session each day gets faster learning than one long single session. Stop a session when your fish does not instantly eat the food pellet dropped into the floating triangle. If the fish modifies your goal by doing something spectacular like swimming a somersault loop the loop on his/her own, CLICK IT! My Sampson was in training to high jump his meter stick, but one day he pulled it down: CLICK! Then off into our championship weightlifting we went!

The students turned in a daily Standard Celeration Chart with counting time (record floor), number of fish attempts, and number of rewards (clicks) administered. From these charts we could see the regularity and duration of daily training sessions, and both the frequency and acceleration of fish performance, and student shaping (clicks).

I am now convinced that "click" is a much better name for the "did it" signal than Skinner's "conditioned reinforcer" or my "response definer." It is shorter, clearer, and both a noun and verb. I am also convinced that "clicker training" is a much better name for the training process than Skinner's "shaping" or my "TEBRO." Skinner's "shaping" even got misused for stimulus fading by some well-known behavior analysts. My "TEBRO" was just too arcane!

Don't let them scrape your fish!