The central niche for language use is social interaction: this is the context in which language is
learned, most heavily used, and doubtless evolved. Interactive language use has well defined
properties which look strongly universal. Amongst these is turn-taking or the rapid alternation
of speakers. Investigations of turn-taking reveal rather stable temporal parameters, with
alternating short bursts of speech (averaging c. 2 secs), separated by modal gaps of only 200 ms
or less. Given the latencies involved in language production (c. 600 ms for a single word, 1500
ms for a simple clause) this implies an overlap in comprehension and production by the
addressee towards the end of the incoming turn, an implication confirmed by neuroimaging
and other measures. Such multitasking must involve a high cognitive load. Looking at the
development of turn-taking in infancy and childhood, one can see relatively quick responses in
the early months slowing down as ever more complex language has to be crammed into short
turns, with children struggling to meet adult norms even in middle childhood. The intensive
processing required by turn-taking suggests it might be a kind of “fossil” with temporal
properties inherited from our primate ancestors before complex vocal language gradually
developed, filling short turns with increasingly complex structures. A glance across our primate
cousins gives some reasons to think this is a plausible scenario.