It has been repeatedly shown that in police lineups, older eyewitnesses are less likely to correctly identify a culprit than younger eyewitnesses. Instead, they are more likely to wrongfully select an innocent suspect or a filler. Here, we used the two-high threshold eyewitness identification model to assess age-related differences in the cognitive processes that underlie these observable responses. First, a large dataset (N = 10.324 responses to lineups) originally collected for a different purpose was reanalyzed. The cognitive processes of culprit-presence detection, culprit-absence detection and guessing-based selection were compared across young to middle aged adults, young-old adults and old-old adults. Then, a novel experiment (N = 6.052 responses to lineups) was conducted, which corroborated the findings of the reanalysis: The probability of culprit-presence detection and, descriptively at least, the probability of culprit-absence detection decreased with age, whereas the probability of guessing-based selection increased. These results highlight possible challenges to the validity of eyewitness testimony from older adults and offer insights on possible strategies to mitigate them.