Enormous amounts of empirical facts have been gathered regarding the content and structure of mental models in astronomy, biology, psychology, and so on. Mental models, or their individual parts (beliefs), are usually understood as stable domain-specific representations of a specific group of phenomena. A typical study using response time analysis includes demonstrating a sequence of statements that must be evaluated by the respondents as true or false. A stable experimental result is an increase in response time to statements that conflict with scientific and naive theories, in comparison to coherent statements. Relevance: the study was conducted during the COVID-19 pandemic (2020-2021), so we used a set of semantically related statements about the mechanisms of coronavirus infection, protective measures, etc. Scientific novelty: in this study, we considered the stability factor of the naive representations studied, presenting each statement multiple times. In predecessor studies, statements were presented only once. Research goal: To determine whether reaction time can serve as an indicator of the consistency of strong beliefs about a particular set of phenomena. This study tested the hypothesis that response time would be shorter for coherent answers to semantically related statements in comparison to incoherent answers. The study was conducted online in spring-summer 2020 and winter 2021. Methods: the participants evaluated pairs of statements regarding COVID-19 and their behavior during pandemic conditions, in randomized order. The results from two series of an online survey confirmed this hypothesis and a simple theoretical model behind it. The delta reaction times for coherent statements are lower than for incoherent ones. Findings: no differences were found in the RT of “yes” and “no” responses; we assume that the delta reaction times have a semantic origin.