Number 214 (Story #2), February 16, 1995 by Phillip F. Schewe and Ben Stein CLASSIFYING GALAXIES can now be done automatically with the help of neural-network scanners. Normally requiring the discerning eye of a human, the designation of a galaxy's morphology on a sliding scale from elliptical to spiral is a tedious but important job. Automating the process would greatly help an enterprise like the Sloan Digital Sky Survey, which plans to chart more than a million galaxies. A team of astronomers has now tested an artificial scanner based on neural network circuits which "learn" to recognize patterns at they go along. Given a sample of 800 digitized galaxy images, the scanner produced a set of classifications which differed from those of six humans by no more than the difference between the classifications of any two humans. (O. Lahav et al., Science 10 February 1995.)