Evidence for a bimodal distribution in human communication

By ye Wu et all, PNAS, 2010. Sorry, lost the direct link.

Human actions (disease spread, resource alloc, ) are often modelled using a Poission distribution (expected number of events during a time interval T, given exponential distribution for waiting times). But this assumption is not fit by the data for i.e. email. Rather, we see bursts of activity followed by long periods of inaction.

Is this because most models do not incorporate both individuals and interactions??

The authors distinguish between initiated and responsive actions, with Poission distribution for the initiation of communication bursts and power-law distribution of messages within a burst (determined in part by priority queuing) They analyze SMS communication.