What the work claims
Learning in animals occurs incrementally through trial and error, not via insight, imitation, or observation.
How it was done
Thorndike placed cats in puzzle boxes that required a specific operant response—pulling a lever or pushing a button—to escape. He measured escape time across repeated trials.
The boxes were approximately 20 inches long, 15 inches wide and 12 inches tall. A door was pulled open by a weight on a string that ran over a pulley to a lever or button inside, so pressing the bar or pulling the lever lifted the weight and opened the door. Once the animal made the response it was let out and given a reward, usually food. Thorndike primarily used cats. Put in the box, they wandered restlessly and meowed but did not know how to escape, and eventually stepped on the switch by chance.
To test whether they could learn by observation, he had cats watch other animals escape and compared their times with those of cats that had not watched. He found no difference in their rate of learning. He saw the same with other animals, and observed no improvement even when he placed an animal’s paws on the correct lever, button or bar.
What holds up
The learning curves—plots of escape time across trials—are empirically documented. Cats showed gradual, not sudden, improvement. The method isolated operant response as necessary and sufficient for escape.
The design rested on a prediction. If animals were showing insight, their escape time would suddenly drop to a negligible period, an abrupt drop on the learning curve; animals using ordinary trial and error would show gradual curves. After stepping on the switch by accident once, they pressed it faster in each succeeding trial, until the times levelled off, which gives the curve an s-shape. The curve also suggested that different species learned in the same way but at different speeds.
The experiments were motivated in part by his dislike of claims that animals use extraordinary faculties such as insight. Of the books of his day he wrote: “They have all been about animal intelligence, never about animal stupidity.”
What does not
The work does not support claims about insight, imitation, or observational learning in cats. Physical guidance and conspecific observation produced no measurable improvement in escape latency.
Thorndike’s own conclusions did not stay fixed. In 1929 he said his early theory had been wrong. He denounced his law of exercise completely, having found that practice alone did not strengthen an association and that time alone did not weaken one. He also dropped half of the law of effect, after finding that a satisfying state of affairs strengthens an association but punishment is not effective in modifying behaviour.
His connectionism has also been criticised as too simplistic and reductionistic, and his law of effect and puzzle box method were subjected to detailed criticism by behaviorists and other psychologists.
Why it matters beyond the lab
It established the first controlled, quantitative animal model of learning—shifting psychology from introspection to measurement—and directly enabled later operant conditioning frameworks.
Is it worth your time
Yes, if you need to understand the experimental origin of trial-and-error learning theory in psychology. It is foundational, not current.
