ABSTRACT
Eyewitness testimony is central to criminal investigations, yet facial identification accuracy varies as a function of retention interval. This study examined eyewitness identification performance across delays ranging from 1 min to 90 days in two experiments. Participants viewed a target face and later completed lineup identification tasks following varying intervals. Results revealed a non-monotonic, three-phase forgetting trajectory: a rapid decline within the first hour, a period of relative stability between 1 h and 1 day and a more gradual decline extending to 90 days. The early and late phases were well described by logarithmic functions, whereas the intermediate phase showed a near-zero gradient, indicating a temporary plateau in performance. The primary contribution of this work is a detailed characterisation of the shape of the forgetting function across forensically relevant timescales, with practical implications for the timing and evaluation of eyewitness identification evidence.