Knowledge of the carbon footprint (CF) of a scientific publication can help to guide changes in behavior for mitigating global warming. A knowledge gap, however, still exists in academic circles. We quantified the CF of a publication by parameterizing searches, downloads, reading, and writing in the processes of publication with both direct and indirect emissions covered. We proposed a time-loaded conversion coefficient to transfer indirect emissions to final consumers. A questionnaire survey, certification database of Energy Star, fixed-asset databases specific to our campus, and reviewed life-cycle-assessment studies on both print media and electronic products were integrated with Monte Carlo simulations to quantify uncertainties. The average CF [(CI: 95%), SD] of a scientific publication was 5.44 kg CO2-equiv. [(1.65, 14.78), 4.97], with 37.65 MJ [(0.00, 71.32), 30.40] of energy consumed. Reading the literature contributed the most, followed by writing and searching. A sensitivity analysis indicated that reading efficiency, the proportion of e-reading, and reference quantity were the most dominant of 52 parameters. Durable media generated a higher CF (4.24 kg CO2-equiv.) than consumable media (1.35 kg CO2-equiv.) due to both direct and indirect reasons. Campus policy makers should thus not promote the substitution of e-reading for print reading at the present stage, because their environmental advantages are highly dependent on time-loaded and behavioral factors. By comparison, replacing desktops with laptops is more attractive, by potentially reducing CFs by 50% and the disproportionate consumption of energy.