The pragmatic tolerance hypothesis (Katsos & Smith, 2010) was originated to explain differences between children and adults concerning scalar implicatures. They introduced Likert-scales to test this hypothesis. We conducted a study with a within subjects design in which we compare children’s binary and scalar responses to the same underinformative sentences. We also used two separate tasks to look at the effects of task difficulty on performance. The results show that the more difficult task, Euler circles, lead to less pragmatic responses compared to the easier task, drawings. Confirming the study by Katsos and Smith (2010; see also Katsos & Bishop, 2011) children choose the middle options on the scale more when they are confronted with underinformative sentences and they choose more extreme options for the control sentences. The comparison with the binary responses however, reveal that the link between the two measuring methods is not as straight forward as we would think.