Spring 2021 Goldfish Experiment Lessons Learned / Design Improvements Needed


what was learned from preliminary study

(in rough order of priority)

  1. LACK OF SYMMETRY (and Third Target placement): When fish is unsure whether to go right or left because color is intermediate between red and blue, fish may go back and forth and therefore spend more time in center or may “hesitate” between right and left and spend more time in middle. Spending more time in middle will result in more hits on middle target, not necessarily because fish was aware they may hit wrong large reward target and so hits middle target to at least get lesser reward, but just by increased chance due to location.

Improvement: Tank should have star shape of at least three symmetrical branches (or a division within it of at least three or four paths that are symmetrical)



  • Increase trials and measure amounts of time spent at different locations for groups of fish.

  • Train multiple fish for individual testing.

  • Compare field direction detection behavior at different field strengths to color response behavior

  1. TIME NEAR TARGET CAN BE MORE MEANINGFUL THAN HIT COUNTS (at least for group fish tests): fishes may hover around target without pecking it, and a fish may peck a target but not hover around it. In addition to hitting target, whether or not fishes swam toward and stayed near correct target is relevant evidence. Target hits are more relevant for individual fish trained by reward for hitting target.

Improvement: Video record and gain more accurate measurements of time distribution of fishes throughout tank.

Challenge: How should volumes of tank categorizing fish locations for time measurements be divided? For example, selecting a rectangular cuboid of right ¼ is arbitrary because the target is in front bottom. Tradeoff between more meaningful but harder to measure accurately spheres around targets vs. less meaningful but easier to measure pyramid or other shape.

  1. FISH AVOIDANCE OF LIGHT: Placing light at end of tank produced most color, but fish avoided light, which would affect data collection, so light was placed beneath tank in the middle to minimize this. This was still not ideal, however, because that may decrease time fish spend in middle, lowering hits on middle target.


1. Lighting should be ambient (with no detectable direction)

2. If fish response to magnetic field direction is confirmed, test metacognition responses to magnetic field direction instead of color. Ways to make correct field target increasingly indeterminate:

  • Alternate fields with one longer duration, but increasingly less so (we have buttons for this so we are not switching wires)

  • Alternate fields with one dominant, but increasingly less so (by adjusting field strength)

  • Do not alternate fields but make one increasingly stronger than the other (requires two sets of loops)

  1. TIMING: Fish were hungrier during tests earliest in the day and so possibly more motivated. Hunger motivation probably varied. Times of day of tests not recorded.

Improvement: Tank or tank division should have at least three or four paths that are symmetrical

  1. INCREASED TRAINING THROUGH TESTING: Fish are trained (or untrained) by the testing itself, which needs to be factored into analysis of increases and decreases of number of hits.

Improvement: Perhaps unavoidable, but should be noted in data analysis.