ABSTRACT Background Persons with aphasia (PWAs) are often impaired in time reference/tense production. It has been suggested that this impairment is due to encoding or/and retrieval deficits. However, to the best of our knowledge, no experimental design that enables teasing apart selective encoding and retrieval deficits has been proposed thus far. Aims This study aims at disentangling time reference-related encoding deficits from time reference-related retrieval deficits in PWAs. Methods & procedures Two sentence completion tasks tapping production of time reference and subject-verb agreement (control condition) were administered to eight Greek-speaking PWAs, eight Russian-speaking PWAs, six Italian-speaking PWAs, seven English-speaking PWAs and four groups of language-, age- and education-matched healthy controls. Task 1 tapped encoding and retrieval processes to a similar extent. Task 2 predominantly tapped retrieval processes. Comparisons between each PWA and the corresponding control group, as well as within-participant comparisons were performed. Outcomes & results All four control groups performed at ceiling. Twenty-eight out of 29 PWAs were impaired in time reference in at least one of the two completion tasks, and all but three PWAs were impaired in production of subject-verb agreement in at least one of the two tasks. In all language groups, there were PWAs exhibiting between-task dissociations. A double dissociation emerged in the time reference condition, as some Greek-, Russian- and English-speaking PWAs performed better on Task 1 than on Task 2, whereas other Greek- and Italian-speaking PWAs performed worse on Task 1 than on Task 2. In the agreement condition, in each language group, there were PWAs performing better on Task 1 than on Task 2. However, none PWA exhibited the opposite pattern. Based on the results, we identified both PWAs with selective time reference-related encoding deficits and PWAs with selective time reference-related retrieval deficits. Conclusions The present experimental design provides a sound basis for teasing apart selective time reference-related encoding deficits and time reference-related retrieval deficits.