Abstract: The amount of annually published scholarly articles is growing steadily, as is the number of indicators through which impact of publications is measured. Little is known about how the increasing variety of available metrics affects researchers’ processes of selecting literature to read. We conducted ranking experiments embedded into an online survey with 247 participating researchers, most from social sciences. Participants completed series of tasks in which they were asked to rank fictitious publications regarding their expected relevance, based on their scores regarding six prototypical metrics. Through applying logistic regression, cluster analysis, and manual coding of survey answers, we obtained detailed data on how prominent metrics for research impact influence our participants in decisions about which scientific articles to read. Survey answers revealed a combination of qualitative and quantitative characteristics that researchers consult when selecting literature, while regression analysis showed that among quantitative metrics, citation counts tend to be of highest concern, followed by Journal Impact Factors. Our results suggest a comparatively favorable view of many researchers on bibliometrics and widespread skepticism toward altmetrics. The findings underline the importance of equipping researchers with solid knowledge about specific metrics’ limitations, as they seem to play significant roles in researchers’ everyday relevance assessments.