An analysis of the discourse in these experiments revealed that the average number of arguments per group in experiments in which unexpected results were obtained was significantly higher than the number in the experiments in which the anticipated results were obtained (χ2=6.7 p=0.017).