While quasi-isochronous rhythms have been frequently reported in non-human animal communication, they have mostly been considered absent in human communication. In this paper, we argue that human speech shows quasi-isochrony when the rhythm is analyzed based on Inter-Onset Intervals rather than words or syllables, akin to how rhythm is commonly defined for non-human animals. Our findings are grounded on a worldwide sample of 49 languages. Rhythms are relatively stable in the range between 0.2-0.6Hz across languages and independent of linguistic and extra-linguistic factors. This quasi-isochrony may rest on neural oscillations and/or respiratory rhythms.