Age differences in rates of forced confabulation and memory consequences thereof were assessed using a recall task similar to real forensic interview procedures. Children viewed a target video and were tested with the same 18 questions immediately afterward and 1 week later. Of the 18 questions, 12 were answerable; the 6 unanswerable questions referred to information not in the video. Participants in the voluntary confabulation condition had a "don't know" response option; those in the forced confabulation condition did not. Although 6-year-olds and 9-year-olds were equally likely to provide a response to an unanswerable question initially, 1 week later 9-year-olds were significantly more likely than 6-year-olds to repeat their initial confabulated responses. These findings suggest that pressing child witnesses to answer questions they are initially reluctant to answer is not an effective practice, and the consistency of children's responses over time is not necessarily an indication of the accuracy of their eyewitness memory. (Contains 1 figure and 4 tables.)