What Does an Open Door Show? Reading Thorndike's Learning Curves

The final movement can reveal how a door opens: a latch moves and the door becomes free. That result does not by itself reveal how the movement was learned. Edward L. Thorndike's study of animal learning centered on this distinction. A successful action could look remarkable without explaining the process that produced it. Recording what happened from early attempts through later trials could generate a different kind of evidence from telling the story of one impressive performance.

This article follows selected sections of the experimental study republished in Thorndike's 1911 Animal Intelligence. The preface identifies the 1898 investigation among the early studies assembled in the volume. Thorndike explained that the texts were largely retained, with editorial changes and some added notes. The account here therefore examines the earlier research through the 1911 access text. It is not a contemporary survey of the intelligence of every animal species. [1][2]

The selection made by extraordinary stories

Thorndike asks what accounts of animal success leave out. A failure to find a route may attract little attention, while an exceptional return journey becomes a widely repeated story. Collecting reported successes consequently does not supply the distribution formed by successes and failures together. An owner's attachment and a story's purpose can also influence the details selected. His criticism is not simply that every narrator knowingly lies. It concerns which events are considered worth reporting. [1]

Accepting that an event happened is different from finding it sufficient for a general explanation. Unknown previous experience, conditions that are not repeated, and an uncontrolled setting make it difficult to reconstruct learning from one achievement. Thorndike's laboratory method sought to address these gaps. He wanted repetition and recorded earlier encounters to enter the explanation. Stories need not be worthless for repeated observations to answer questions that a story alone cannot settle.

An escape time needs a defined situation

Animals were observed in enclosures whose doors could be released by specified movements. Pulling a loop, depressing a lever, or activating a platform could open different mechanisms. Visible food outside and the tendency to leave confinement formed part of the motivational situation. This describes a historical method, not a procedure recommended for replication with household animals. [1]

Thorndike recorded more than whether an escape occurred. He observed general behavior, how the successful act happened, and the interval spent in the enclosure. Across repetitions, a shorter route to the successful movement became one indication of learning. Time is interpretable here only in relation to the task, its mechanism, and the behavior recorded. An identical duration can correspond to different difficulties in different apparatus.

The initial actions described for the cats did not all directly target the release. They approached openings, reached toward different parts, and acted on movable objects. The successful movement could arise within this activity. Thorndike interpreted repeated success as strengthening the relevant connection while unsuccessful movements declined. That is an explanation of the observed pattern, not a transcript of sentences occurring in the animal's mind. [1]

The curve does not descend at every step

The book gives a verbal sequence of twenty-four trial times for cat number twelve in apparatus A. The first three were 160, thirty, and ninety seconds; the final three were six, six, and seven. There is a substantial overall change, but every later attempt is not faster than the preceding attempt. A very short second escape was followed by a longer third one. The sequence belongs to a specified animal and apparatus, not to an average curve for all cats. [1]

This comparison prevents equating learning with one favorable result. An effective movement may happen accidentally and early on one attempt, then take longer to appear on the next. Thorndike recognized that such contingencies could influence short portions of a curve. The overall direction and each individual point require different readings. Getting out quickly once is not the same observation as responding rapidly and consistently on subsequent encounters.

He considered time to combine the elimination of ineffective activity with more precise performance of the effective act. Some interval also concerned orientation after the animal was placed in the enclosure. A time record can summarize changing behavior without becoming identical to every component of the learning process. When comparing mechanisms or animals, it remains important to understand which activities the endpoint gathers together.

Similar movements need not operate every door alike

Differences between the mechanisms make that issue concrete. In one apparatus, depressing a lever releases the door. In another, lifting a latch also requires outward pressure; otherwise the latch can fall back into place. More complex enclosures require several acts together. “The animal learned to open a door” can conceal these differences. Which combination was performed and in what order belongs in the account. [1]

Previous experience changes the starting situation too. Thorndike described a cat that had learned to direct its paws toward loose objects in an earlier box as bringing a modified tendency to a new one. Previously acquired movements could help without completely solving the new problem. Each series should not be read as detached from the animal's history. Lack of familiarity with a new mechanism is not equivalent to lack of every relevant movement experience.

This matters for the comparison between anecdote and experiment. The experimental record makes more of the history available, but it cannot make history cease to exist. A known previous action can influence exploration without constituting understanding of every new mechanical relation. The account becomes more informative when it specifies that distinction rather than choosing between two extremes: an animal with no relevant experience and an animal assumed to understand the entire device.

The measured interval and the inferred mental process

Thorndike interpreted gradual formation of the connection against the idea that the animal suddenly understood the mechanism and made a considered decision. If a complex solution had been fully grasped at once, he expected subsequent times to fall abruptly and remain low. He nevertheless acknowledged that a simple movement learned quickly could also produce an abrupt descent. A rapid success was not, in his account, sufficient evidence of inference. [1]

Turning this interpretation into the conclusion that animals cannot think under any circumstances would exceed the measurement's scope. The text contains both a qualification about the particular animals studied and very ambitious generalizations. The reliable distinction retained here is between a record from specified tasks and an interpretation of that record. A time curve is not a direct recording of inward thought. Agreement between an explanation and the curve does not mean that the same data have eliminated every possible alternative explanation.

Nor should a historical account conceal confinement and the use of food motivation. Those conditions shaped the observed behavior and limit its identification with freely occurring activity. Describing their role does not supply instructions for personal replication. Understanding a measurement idea and approving the historical procedure are separate judgments. The material here concerns how the evidence was constructed, not a training program.

The source-based conclusion is to give a successful movement a history rather than treat success as its own explanation. Thorndike used sequences of trials to investigate the difference between a fortunate outcome and a more consistent response. The paper remains useful when it moves the description from “the animal succeeded” to “under what conditions, through which repetitions, and with what changes did it succeed?” At the same time, it requires separating what the investigator measured from the mental meaning he attributed to it. An open door is an observed result; an account of learning must be constructed from the history leading to that result.

Sources

1. Thorndike, Edward L. (1911). Animal Intelligence, Chapter II, republishing the experimental investigation of 1898. University-hosted primary text. Criticism of anecdotes and method: pp. 20–29; mechanisms and time curves: pp. 30–42; interpretation concerning inference: pp. 66–75.
2. The same volume. 1911 preface and scope of republication.

Sources reviewed on 14 October 2026. The public-domain historical text is interpreted in original prose. No source figures or direct quotations are reproduced. No animal-training protocol, contemporary intelligence diagnosis, or universal learning-duration recommendation is offered.