This 1898 monograph by Edward L. Thorndike is a foundational pillar of modern psychology. It marks the decisive shift from "anecdotal" animal psychology—which often attributed human-like reasoning to pets—to a rigorous, experimental science. Thorndike’s work introduced the concept of trial-and-error learning and provided the empirical basis for what would later be known as the "Law of Effect."
TL;DR
Thorndike replaced anthropomorphic stories with controlled "puzzle box" experiments to show that animals learn through the gradual strengthening of associations between stimuli and impulses, not through reasoning or imitation. His quantitative "time-curves" demonstrated that learning is a process of "stamping in" successful behaviors through pleasure, a discovery that revolutionized our understanding of both animal and human intelligence.
What Problem Does This Paper Address?
At the end of the 19th century, animal psychology was dominated by the "anecdote school." Researchers like George Romanes collected stories of "intelligent" pets to argue that animals possessed high-level reasoning, inference, and imitation. Thorndike critiqued this approach as being a "eulogy" rather than a psychology, noting that it ignored "animal stupidity" and failed to explain the actual mental processes involved.
The core question Thorndike tackled was: What is the "mental fact" in association? Does an animal reason that "if I pull this string, the door opens," or is something more primitive happening? He sought to move beyond vague descriptions of "intelligence" to a quantitative measure of how associations are formed, retained, and inhibited.
Methods
Thorndike designed a series of "puzzle boxes"—small enclosures from which a hungry animal could escape to reach food by performing a specific mechanical act, such as pulling a loop of cord, pressing a lever, or turning a wooden button.
The methodology was elegantly simple:
- Hunger as Motive: Animals were kept in a state of "utter hunger" to ensure a standard strength of motive.
- Quantitative Tracking: Thorndike recorded the time taken to escape in successive trials.
- Controlled Learning: He tested whether animals could learn by observation (imitation) or by being manually "put through" the required movements by the experimenter.
Datasets
While not a "dataset" in the modern computational sense, Thorndike’s "subjects" included:
- 13 Cats: Mostly young (3–11 months), tested across various box types (A through L).
- 3 Dogs: Small, adult or near-adult dogs.
- Chicks: Used in pen-based experiments to study simpler associations and social influence.
Key Insights and Ideas
- Learning is Gradual, Not Sudden: If animals reasoned, we would expect a sudden "aha!" moment where the time to escape drops to zero and stays there. Instead, Thorndike’s time-curves show a gradual slope, indicating a slow "wearing smooth of a path in the brain."
- The Association of Impulse: Thorndike argued that animals do not associate ideas (e.g., the thought of the loop and the thought of food). Instead, they associate a sense-impression with an impulse to act. The pleasure of escape "stamps in" the successful impulse.
- Failure of Imitation: In one of his most controversial findings, Thorndike observed that cats and dogs did not learn more quickly by watching a "trained" animal perform the task. He concluded that they lack the "transferred association" required for true imitation.
- Learning by Doing: He found that manually moving an animal's paw to open a latch (instruction by "putting through") did not help them learn. The animal must provide the innervation or impulse themselves for the association to take hold.
Key Results
Thorndike’s primary results were expressed through "time-curves," where the x-axis represents successive trials and the y-axis represents time.
- Quantitative Mastery: For simple tasks (like Box A), cats typically moved from escape times of 100+ seconds to a constant 6–10 seconds within 20–30 trials.
- Complexity Limits: As tasks became more complex (e.g., Box K, which required three separate acts), the curves became much more irregular, and many animals failed to form a perfect association.
- Permanence: Associations were remarkably durable. Animals returned to boxes after weeks or months of no practice often escaped nearly as quickly as they had at the end of their initial training.
Limitations and Future Work
Thorndike was careful to note that his findings applied specifically to the domestic animals he studied. He acknowledged that primates (monkeys) might possess a different order of intelligence, including the ability to form associations from "free ideas" or through imitation—a hypothesis he was unable to test fully due to a "sluggish" subject.
He suggested that future work should focus on:
- Comparative Psychology: Tracing the "origin of human faculty" by studying the transition from practical associations to "free ideas" in children and primates.
- Pedagogy: Applying the "learning by doing" principle to human education, suggesting that actual performance is often superior to mere explanation or imitation.
Summary updated