Social learning models of investment provide an interesting explanation for sudden changes in investment behaviour. Caplin and Leahy (1994) develop a model of social learning in which agents learn about the true state of demand from the investment suspension decisions of other agents. The author tests the main predictions of Caplin and Leahy's model using a unique database of investment projects undertaken by semiconductor plants. She finds that firms that are installing a significant new technology appear to be influenced by social learning, because they are more likely to suspend their investment project when other suspensions occur. A 1 per cent increase in the number of other suspensions increases by 3.6 per cent the probability that an average new technology plant will suspend their investment project. Suspensions by other agents also significantly affect plants that use conventional technology, but that effect is negative. The conventional technology plants are less likely to suspend their investment project when other firms suspend, which suggests that their payoffs are strategic substitutes, as in a "war-of-attrition" game.