The purpose of the study was to disentangle the effects of simultaneous interpreting experience and training on working memory, to examine the effect of language, modality and recall on working memory scores, and to associate memory scores of trainees with interpreting quality. Working memory scores were compared in the L2 reading span task (performed by professional conference interpreters, bilingual controls and interpreter trainees tested before and after training) and in the L1 reading span task and L1 listening span task (performed by interpreters and controls). Data was collected from 68 participants in experiment 1 and from 51 participants in experiment 2. It was analysed by means of linear models, regressions and t-tests. Professional interpreters consistently outperformed controls on all working memory tasks. They performed better in L1 than L2, and their scores were not affected due to modality (visual vs. auditory presentation) and recall mode (serial vs. free). Interpreter training improved working memory scores. Trainees’ higher scores predicted better interpreting performance. This was the first study to adopt a longitudinal design to examine the effect of training on memory of conference interpreter trainees. The study shows that interpreter training (but not experience) improves working memory capacity and predicts interpreting performance.