The increasing digitalization of nuclear power plants’ main control rooms questions the validity of Human Reliability Analysis (HRA) estimates for tasks that were formerly performed by acquiring information on analog panels. This paper presents a re-analysis of two experiments conducted at the Halden Reactor Project that employed the micro-task method for comparing different types of digital and analogue displays. The experiments were conducted with licensed operators at a U.S. power plant training simulator and at the HAMMLAB research simulator in Norway. In the re-analysis the tasks are classified according to two HRA reference sources for human error probabilities, THERP and KAERIs’ HuREX, in order to assess (a) the operators’ reliability on identification tasks, and (b) the impact of the HSI type (digital displays vs. analogue panels). The results show that error-rates for decontextualized micro-tasks are much higher than HRA estimates for the corresponding tasks in average industrial conditions. The advantage of the analog panels over the digital displays was small and not statistically significant, despite the experimental set-up that benefitted the familiar panels. Error rate differences between task types are larger than the differences attributable to the HSI type. Finally, the observed error rates relative differences across tasks are fairly consistent with THERP reference values but not with KAERIs. The results stress the importance of task modelling above HSI concerns, and the risk of overestimating the reliability of apparently easy tasks.