An experiment was conducted to investigate the effects of accent, telephone and a relatively long retention interval (3 or 8 weeks) on speaker identification. Three-hundred and sixty participants heard the target's voice and were asked to identify the target by means of a line-up consisting of 6 voices. Half of the participants were given a target-present line-up and the other half a target-absent line-up. The results showed that 24% of participants correctly identified the target in the target-present condition (hits), whereas 50% of participants incorrectly identified a person as the target in the target-absent condition (false alarms). The speaker with the standard-accented voice was more often correctly recognized than the speaker with the non-standard-accented voice. No difference was found between identification accuracy after one, three or eight weeks and between the telephone and non-telephone conditions. It can be concluded that there is a relatively high probability that an innocent defendant is identified as the perpetrator, even in a procedurally correct voice line-up (in this experiment 8%). Furthermore, reliability may be drastically reduced when the perpetrator has a strong accent, unfamiliar to the listener. On the other hand, reliability of a voice line-up seems not to be affected by a presentation over the telephone, as well as by a retention interval of at least 8 weeks. Copyright © 2006 John Wiley & Sons, Ltd.