Introduction The present study investigated the strategies used by experienced researchers to find articles in an online bibliographic database. Objective The goal was to shed light on researchers’ actual search behaviour by empirically observing users during the search process. Method We asked 16 neuroscience researchers and 16 researchers in other disciplines of life sciences to perform five search tasks on neuroscience topics using the PubMed database. Think-aloud protocols were recorded while the participants performed the search tasks. Results All researchers managed to find adequate references for the tasks in the limited amount of time allocated. Most participants used similar, very basic moves and tactics to perform the tasks. However, they behaved in many different ways when they had to decide whether to read the retrieved results or to modify their query (according to the number of references they obtained), or when selecting the one or two references they finally gave as answers among all possible answers. Discussion Variability in participants’ behaviour may result from inter-individual differences in either the participants’ mental models of PubMed or cognitive abilities and styles, but the question requires further investigation. The data further suggest that developers of online search engines and databases should concentrate their efforts on improving simple-search interfaces and helping users to correct search errors in real time.