An experiment was conducted to investigate both the eect of the order of presentation of defence and prosecution evidence and the prior availability of background information on assessment of guilt. Subjects were required to judge the defendant's probability of guilt either after each witness statement (step-by-step) or after having read all witness statements (end-of-sequence). In the step-by-step mode, an order eect was observed with later evidence exerting a greater impact on the subjects' judgment. This recency eect probably occurred because subjects used an anchoring-and-adjustment process: each new piece of evidence was averaged with an anchor judgment re¯ecting the overall assessment of previous items. In the end-of-sequence mode, on the other hand, the order eect depended on the background condition: if background information was provided a recency eect occurred, but if no background information was available a primacy eect was evident. This result might be explained by assuming that subjects tried to integrate witness information into a coherent cognitive pattern. As the judgment is memory-based in the end-of-sequence condition, recent information will be more available than earlier items. However, when no background information was presented, the ®rst evidence items had to be processed at a more semantic, deeper level, resulting in a primacy eect that apparently outweighed the recency eect.