This talk examines the relationship between knowledge and accuracy at both the individual and group level. For individuals, it seems reasonable to expect that the more a person knows about a domain, the more accurate their decisions will be. This expectation is challenged by the so-called "less-is-more" effect, which claims that there are often situations in which a person who knows less about a domain will make more accurate decisions than a person who knows more. We present new data that provide evidence both for and against the less-is-more effect, and a new analysis that reconciles the apparently contradictory findings. For groups, it seems reasonable to expect that the greater agreement there is within a group, the more accurate its decisions will be. We consider data from a number of different domains, including gambling, trivia questions, medical diagnoses, and sports predictions. An analysis of these data shows that the relationship between group agreement and accuracy is complicated, and depends in interpretable ways on the nature of the domain.