Note: Asterisks represent significant difference (*p < .05,**p < .01,***p < .001) and are based on the revised analyses (binomial mixed effects models) for the replications. Effect sizes are differences in percent correct. Data here reflect results after excluding subjects with poor catch trial accuracy, except for SNA96-1c where there were no catch trials. In all but one case, we calculated main effects by investigating the mean of subject means. For Frank et al. (2010), we never calculated these means, so we estimated the main effects by applying the reverse logit function on the intercept for the continuous-sentence-length model.