Vosniadou and Brewer (1992) claim that children's drawings and answers to questions show that they have naive, theory-like 'mental models' of the earth; for example, they believe it to be flat, or hollow with people inside. However, recent studies that have used different methods have found little or no evidence of these misconceptions. The contrasting accounts, and possible reasons for the inconsistent findings, were tested by giving adults (N = 484) either the original task (designed for 5-year olds) or a new version in which the same drawing instructions and questions were rephrased and clarified. Many adults' responses to the original version were identical to children's 'naïve' drawings and answers. The new version elicited substantially fewer non-scientific responses. These findings indicate that even adults find the original instructions and questions ambiguous and confusing, and that this is the principal reason for their non-scientific drawings and answers. Since children must find the task even more confusing than adults, this explanation very probably applies to many of their non-scientific responses, too, and therefore accounts for the discrepant findings of previous research. 'Naïve' responses result largely from misinterpretation of Vosniadou and Brewer's apparently simple task, rather than from mental models of the earth.