A psychologist built a machine that graded and taught in 1926

The classroom note

A teacher grading tests by hand in 1926 could not tell a student, mid-exam, whether an answer was right. Sidney Pressey, a psychologist then at Ohio State University, built a machine to close that gap. In the journal School and Society, Pressey described a box with a window showing a multiple-choice question and keys a student pressed for an answer; if correct the machine advanced, and if wrong it tallied the error and made the student choose again. A later technical account, written by B. F. Skinner and citing Pressey's paper directly, records that the machine could both test and, in Pressey's own claim, teach, because a self-scoring device gives immediate consequences a delayed paper exam cannot.

What the evidence says

Skinner's 1958 account is a competing inventor's retrospective, not a neutral history, written three decades later to introduce his own teaching machine. It is nonetheless specific about what Pressey claimed. Pressey believed immediate feedback would let each student proceed at his own rate, rather than the whole class moving together. By 1932, in a follow-up piece Skinner's paper also cites, Pressey reported disappointment: 'one person alone can accomplish relatively little' against the cost of distributing a physical device, and he was 'regretfully dropping further work on these problems.' A 1984 ERIC-indexed review of the lineage from Pressey to computer-based instruction credits him with 'provided the necessary technological implement' and 'anticipated the present cognitive perspective,' but that review is a secondary retrospective making a case for continuity, not a study of outcomes.

The implementation question

Pressey's own explanation was economic, not pedagogical: manufacturing, distributing and maintaining a physical device required money and coordination one inventor could not supply. Skinner's account adds a second limit: Pressey's machines were testing devices, layered onto learning that had already happened elsewhere, not the primary teaching method. The lesson for any later automated-feedback tool, mechanical or generative, is that 'immediate feedback' is a claim about a mechanism, not about who pays to build, distribute and support it, or where it sits in a lesson.

What holds and what fails

What holds, on this evidence, is narrow: a self-scoring device can change a student's next action before an error repeats, which a delayed grade cannot. What fails is the leap from that mechanism to a promise of automatic, low-cost transformation. Pressey's account describes an invention that worked in isolated tests and failed to scale under the institutions and financing then available. This is an editorial reading of a century-old dispute, offered because the same promise of an 'industrial revolution in education,' in Skinner's phrase, keeps returning.

• Is the claimed benefit the feedback mechanism, or the business model needed to sustain it?
• Who pays to maintain the tool after the inventor or vendor moves on?
• Is the tool testing what was learned elsewhere, or carrying the primary teaching?

Pressey's box did not fail because immediate feedback was a poor idea; the same mechanism reappears in every tutoring system this batch covers. It failed as an institution-scale product, for reasons unrelated to whether the underlying idea was sound.