The so-called ‘Melatonin Hypothesis’ proposed that decreased nocturnal production of melatonin (MLT) might explain the increased risk of breast cancer that has been formerly attributed to extremely low-frequency (ELF) magnetic fields (MF) of weak intensity. Although the risk of ELF MF upon breast cancer was later dismissed, repeated reports were published of partial inhibition of MLT secretion in rats under long-term (≥ 4 weeks) exposure to weak ELF MF. Since 2004, however, this topic has not been experimentally studied any more. In the present study, we propose to go back to the MLT hypothesis and apply it to childhood leukemia, for which an increased risk has been robustly associated with residential exposure to ELF MF. Contrary to the original hypothesis, however, we do not consider decreased MLT levels, but disruption of circadian rhythmicity per se as the effector mechanism. Indeed, the role of the circadian timing system in the development of childhood leukemia has been well established. Motivation for going back to the MLT hypothesis comes from recent data that suggest magnetosensory disruption by ELF MF in mammals, and magnetosensitivity in humans, together with current evidence for an influence on circadian rhythmicity from disruption of non-photic sensory stimuli of various natures. We thus suggest further study on circadian rhythmicity in humans (children if possible) under long-term exposure to weak ELF MF. Copyright © 2012 John Wiley & Sons, Ltd.