How did one book nearly destroy neural network research for an entire decade?

The answer

In 1969, Marvin Minsky and Seymour Papert published 'Perceptrons: An Introduction to Computational Geometry,' a rigorous mathematical analysis proving that single-layer perceptrons could not compute certain functions, most notably the XOR (exclusive or) operation. They showed that a perceptron could only learn linearly separable patterns, a severe limitation. While they briefly acknowledged that multi-layer networks might overcome these limitations, the book's overall pessimistic tone convinced many researchers and funding agencies that neural networks were a dead end. Government funding for connectionist research dried up almost immediately, and neural network work entered a 'dark age' that lasted roughly from 1969 to 1982, until Hopfield networks and backpropagation revived interest.