We conducted five pupil old/new experiments to examine whether pupil old/new effects can be linked to familiarity and/or recollection processes of recognition memory. In Experiments 1–3, we elicited robust pupil old/new effects for legal words and pseudowords (Experiment 1), positive and negative words (Experiment 2), and low‐frequency and high‐frequency words (Experiment 3). Importantly, unlike for old/new effects in ERPs, we failed to find any effects of long‐term memory representations on pupil old/new effects. In Experiment 4, using the words and pseudowords from Experiment 1, participants made lexical decisions instead of old/new decisions. Pupil old/new effects were restricted to legal words. Additionally requiring participants to make speeded responses (Experiment 5) led to a complete absence of old/new effects. Taken together, these data suggest that pupil old/new effects do not map onto familiarity and recollection processes of recognition memory. They rather seem to reflect strength of memory traces in short‐term memory, with little influence of long‐term memory representations. Crucially, weakening the memory trace through manipulations in the experimental task significantly reduces pupil/old new effects.