# Rosenblatt's perceptron treated recognition as something a machine could learn > Frank Rosenblatt was asking how past experience could change what a machine saw in a new pattern. Published 2026-11-08 Canonical: https://machines-on-machines.anth.us/posts/rosenblatt-perceptron-perception/ ::image{src="assets/rosenblatt-perceptron-perception/organization_of_a_biological_brain_and_a_perceptron.png" alt="Rosenblatt’s historical diagrams compare stages of a biological brain and a perceptron as each responds to a letter" credit="Image: Frank Rosenblatt, Public domain, via Wikimedia Commons."} In a 1958 Office of Naval Research article, Rosenblatt drew the perceptron alongside a proposed organization of the brain. [Image source](https://commons.wikimedia.org/wiki/File:Organization_of_a_biological_brain_and_a_perceptron.png). [Licence](https://creativecommons.org/publicdomain/mark/1.0/). Show a machine one example of a pattern and it can store a copy. Recognizing another example that differs from the first takes something more. Frank Rosenblatt's 1958 paper introduced the perceptron as a probabilistic model for how a system could learn to organize and respond to sensory information. Rosenblatt, a psychologist working at the Cornell Aeronautical Laboratory, described a system with sensory, association and response units. Activity could travel from the sensory inputs through associations to a response. Experience changed how readily the system gave particular responses to patterns. ## The paper asked a question about perception The work took inspiration from nervous systems and put forward a model researchers could analyze. Its architecture and training assumptions mattered. A machine that could learn one kind of separation between patterns could still have trouble with a different arrangement. The value of looking back at the original is seeing the question Rosenblatt was trying to answer. It was about how a pattern recognition system might acquire a useful response from experience, a problem later work tackled with different architectures and learning methods. [Read the source](https://pubmed.ncbi.nlm.nih.gov/13602029/).