Complexities in Multi-representational Contradiction Detection

Both of the current studies investigated the role of complexities in sensitivity to consistency between text and graph in the domain of science. Participants received eight experimental materials consisting of a single sentence claim and a line graph that was either consistent with, or contradictory to the claim. Contradictory graphs had a reversed y-axis that resulted in traditional interpretations of the data being incorrect. Study 1 manipulated the complexity of the graphs. Study 2 included the same manipulation, but also manipulated the complexity of the claims that accompanied the more or less complex graphs. Results of both studies demonstrate that participants showed a consistency bias, that is a tendency to say that the text and graph matched regardless of their consistency. Results also showed that participants were less sensitive to consistency when materials were more complex (both graphs and claims) even when prior knowledge, graph literacy, and numeracy understanding was controlled for. When asked what part of the materials were used to justify consistency decisions participants rarely mentioned the y-axis, although frequency of mention was correlated with better sensitivity. Finally, tasks were consistently rated as being not difficult and participants consistently demonstrated high levels of confidence in their decisions regardless of condition.