This meta-analysis aims to analyse how the activities of rail passengers have changed in Japan as a result of rapid technological developments. To be eligible for inclusion in this analysis, source studies must have reported the number of passengers performing specific activities, and the number must have been directly counted by surveyors who actually ride on trains. Databases searched included CiNii, J-STAGE, Web of Science, and Google Scholar. References in selected studies were trialled using a snowballing method. In addition, past onboard activities were retrospectively identified by content analysis of YouTube videos in which the surveyors hypothetically travelled on a train and observed the passengers. The use of YouTube videos for meta-analysis of rail passengers’ activities is a novel contribution of this study. The search for the YouTube video was entirely manual. In total, 23 independent studies with 332,355 passengers were included in the analysis. Data were collected from 1983 to 2019. The effect sizes were the proportion of each of the following activities: ‘(a) mobile phones’, ‘(b) sleeping’, ‘(c) reading’, ‘(d) music’, and ‘(e) other’. Meta-regressions were performed with the year of data collection as a moderator. Demonstrating historical changes in activities through statistical analysis is another novel contribution: ‘(a) mobile phones’ and ‘(d) music’ had a significantly increasing trend, ‘(c) reading’ had a significantly decreasing trend, and ‘(b) sleeping’ and ‘(e) other’ did not change. Studies with and without YouTube videos did not affect the conclusions, which supports the use of YouTube videos for the purposes of this study. Ideas are presented for research methods that use directly observed data to explain the possible social factors behind longitudinal variation in travel-based multitasking.