[possibility #] 5. B. F. Skinner developed ways to control all those external variables by enclosing the experiment in a sealed, soundproof, lightproof, etc., box. The results were reliably reproducible, and a great deal was learned. The boxes were soon names “Skinner Boxes” and became the new paradigm for studying animal learning. Skinner and many others switched to pigeons, for various reasons, but others continued to use rats.
When I was undergraduate in the late ’40s, I hung around that lab and helped with some switching and sequencing stuff to make the experiments more convenient. I don’t remember the name of Young, but it was folklore that the change was because someone had found that rats appeared to be able to navigate by distant cues, e.g., the appearance of the ceiling, so that the traditional open-topped maze experiments might be flawed.
2.1 Shortcut learning in Comparative Psychology: unintended cue learning
Rats learned to navigate a complex maze apparently based on subtle colour differences—very surprising given that the rat retina has only rudimentary machinery to support at best somewhat crude colour vision. Intensive investigation into this curious finding revealed that the rats had tricked the researchers: They did not use their visual system at all in the experiment and instead simply discriminated the colours by the odour of the colour paint used on the walls of the maze. Once smell was controlled for, the remarkable colour discrimination ability disappeared… [Nicholas Rawlins, personal communication with F.A.W. some time in the early 1990s, confirmed via email on 12.11.2019]
I asked on Twitter if Rawlins had seen this first hand or if it was secondhand, and the second author stated:
yes, the anecdote happened as described in Nicholas Rawlins’ laboratory at Oxford, confirmed in personal communication with Felix Wichmann in Nov ’19
One interesting lead showed up on Twitter: Marvin Minsky on Usenet 10 April 1993 (
sci.bio“Puling Habits out of Rats”) in response to someone asking ‘who was Mr Young and whatever happened with the mouse studies’:It is not Feynman, but we seem to have a confirmed anecdote of a similar problem in rat experiments. From “Shortcut Learning in Deep Neural Networks”, Geirhos et al 2020:
I asked on Twitter if Rawlins had seen this first hand or if it was secondhand, and the second author stated: