Over the past decade, generative grammarians have viewed language acquisition as a process of fixing option points or parameters defined in Universal Grammar. Here David Lightfoot addresses the crucial question of what it takes to set a parameter - of what kind of experience is needed to trigger the emergence of a natural kind of grammar. Lightfoot asserts that parameter-setting is not sensitive to embedded material, and that it is triggered only by robust elements that are structurally simple. He observes that morphological properties play a significant role in setting parameters which have widespread syntactic effects. Using evidence from data on diachronic changes and from current work in syntactic theory, Lightfoot makes precise claims about the triggering experience that can explain a number of historical puzzles. He argues that the changes could have taken place in the way they did only if language acquisition proceeds on the basis of simple, unembedded experiences.