The number of researchers' publications is a widely used proxy measure for scientific output, individual achievement, and performance. Despite well-known criticism from the bibliometric community, the use of bibliometric databases as a basis for measuring publication output is widespread. At the same time, there are established survey instruments that also measure the publication output per researcher. We use survey-bibliometric matching with Scopus publication records to compare the alternative publication counts. A Scopus author ID match could be found for 70 % of the respondent researchers. The number of publications per researcher varies greatly between these data sources. The correlation is only 0.41 and the average individual Scopus coverage is 55 %. Importantly, we find a very high variance of individual-level coverage within disciplines, something that other approaches fail to detect.