An experiment investigated how methods of questioning affect children's responses regarding the shape of the earth and the day/night cycle. Seventy-two children from Grade 1 and Grade 3 in a middle-class elementary school in Athens, Greece, were tested either by an open method of questioning or by a forced-choice method of questioning. The two methods of questioning produced different results, suggesting that they tap different forms of knowing and encourage different ways of reasoning in elementary astronomy. The open method replicated previous findings by Vosniadou and Brewer [Cogn. Psychol. 24 (1992) 535], showing that the majority of the children gave responses consistent with a small number of internally consistent mental models of the earth. The forced-choice method of questioning produced an increase in scientifically correct responses, but also a decrease in internal consistency. It appears that the forced-choice method of questioning, together with the presentation of the spherical model of the earth, can inhibit the generation of internal models.