The idea that ‘gossip’ or vocal exchange of social information was a vital mechanism for bonding early human groups appears plausible and concretely testable (Dunbar 1996, 1998; Dunbar, Duncan and Nettle 1995). The relatively rapid encephalisation seen in archaic-grade Homo sapiens is presumed to reflect the increasing size and complexity of these hominids' social groups (Aiello and Dunbar 1993). Vocal grooming in the first place and ultimately gossip offered alternative mechanisms for servicing such extensive social networks, because they saved valuable time compared with the traditional primate means of manual grooming.

This ‘time-saving’ argument leads to a serious problem for the gossip hypothesis of language origins, however. As our ancestors maximised brain size in response to the pressure for larger groups, they maximised their ‘Machiavellian intelligence’ (Byrne and Whiten 1988; Whiten and Byrne 1997). Humans appear to be selected for a capacity involving both social cooperation and alliance formation, but also manipulation and exploitation of their relationships. We cannot consider gossip as a mechanism of social bonding without factoring in this Machiavellian aspect of manipulating information for selfish purposes (cf. Kemmerer 1997). In the case of primate grooming, time becomes a currency (Byrne 1995: 200–202). The time an individual spends grooming an ally reliably quantifies its commitment to their relationship. Correspondingly, if vocal grooming and gossip mechanisms led to a reduction in time spent grooming per individual groomed, this implies a reduction in the level of commitment signalled to each individual (not necessarily equally distributed).