Language existed before human populations became separated (all descendant populations have language) but language did not emerge until long after these population divergences occurred (behavioral modernity only showed then). Distinguishing capacity for language from externalized language resolves the apparent paradox, eliminates the need of proto-language, and rules out monogenesis. Speech emerged only after the capacity for language became (sufficiently) fixated in the species. This accords well with a fundamental property of human language. Rules mapping to meaning rely on structural properties only, while rules mapping to sound are (also) sensitive to linear order, reflecting properties of sensorimotor modalities. The asymmetry suggests (i) primacy of internal language over speech/sign, and (ii) evolution of capacity of language preceding externalized language. Click phonemes with their unique geneological, genetic and geographical distribution may be relevant here. All biologically Khoisan groups speak click languages, which are spoken by biologically Khoisan groups only. Separation followed possession of internal language but preceded externalized language. Clicks were recruited for externalization in San populations only after deepest separation.