Finally, we used linear regression to examine the predictive power of a model comprising a combination of the variables associated with cross-examination accuracy-change scores. Analysis of Cook’s D values indicated that the data of one female child exerted an undue influence on the overall results (Cook’s D = .13).The data from this participant were therefore excluded. There was good tolerance and low collinearity (>.35). Using the simultaneousmethod, we entered age at second interview, WASI Vocabulary raw score, and amount of free recall information reported at the first interview. A significant model emerged: F(3, 122) = 4.73, p = .004,explaining 8% of the variance (Adjusted R2= .08). The amount of free recall information that participants reported at the first interview uniquely predicted 3.3% of the variance in the effect of cross-examination on accuracy (R2change = .033; standardisedbeta weight = .214).