We investigated German and Nichols' finding that 3-year-olds could answer counterfactual conditional questions about short causal chains of events, but not long. In four experiments (N =192), we compared 3- and 4-year-olds' performance on short and long causal chain questions, manipulating whether the child could draw on general knowledge to answer. We failed to replicate German and Nichols' result, finding instead that in two experiments (Experiments 1 and 3) there was no difference in performance on short and long causal chain questions and in two experiments (Experiments 2 and 4) children showed the opposite pattern: short causal chain questions were more difficult than long. These two unexpected patterns of results were replicated in a fifth study (N =97). Children with lower language ability found short causal chains more difficult than long. Performance by children with higher language ability was unaffected by the length of the causal chain they had to consider. We found no evidence that children showed precocious counterfactual thinking when asked about recent events in a causal chain and conclude that counterfactual thinking develops after 4 years of age.