A key cross-examination tactic in trials of child sexual abuse (CSA) is to highlight inconsistencies between sources of information to discredit the complainant's account. The present study examined the prevalence, origin and nature of inconsistencies arising in the cross-examination of complainants in CSA trials. Further, we examined the association between these inconsistencies and the types of question that elicited them in the earlier police interview of the child witness (i.e. open-ended, specific, or leading). Transcripts of videorecorded interviews (evidence-in-chief) of 73 complainants (15 males, 58 females) and subsequent cross-examinations at trial were coded. Results showed that inconsistencies were raised in the cross-examination of 94.5% of complainants; most between what the children said in their police interview versus their cross-examination. A greater proportion of inconsistencies was associated with specific than open-ended questions asked in the police interview. However, open-ended questions were associated with some inconsistencies, perhaps due to the longer answers they elicited. Shorter police interviews relying mainly on open-ended questions may minimise the opportunity for inconsistencies to arise in cross-examinations. Judges and juries require education about inconsistencies that arise from memory's reconstructive nature and lay people's tendency to use these inconsistencies to make inferences about the unreliability of witnesses.