@BillIves @tetradian and @gagan_s point to some recently published research suggesting that "collaboration can exact heavy time costs if done inefficiently" [MIT Sloan November 2010]. Yes indeed, that is what the word "inefficiently" usually means. Specious arguments and poor journalism says @flowchainsensei; @tetradian agrees about the poor journalism but thinks there is a valid point about whole-of-context cost of collaboration.
The researchers measured the time cost of collaboration in terms of the amount of time people spend preparing for and engaging in collaboration with others. Undoubtedly there is considerable variation in the amount of time people devote to these activities, and also considerable variation in the benefits gained from these activities. In many situations there will be a trade-off between the amount of preparation and the effectiveness of the collaboration - poorly prepared people may not be able to collaborate effectively - and obviously the greater the number of people involved in a collaboration, the greater the cost of wasting people's time as a result of poor preparation.
If we want to know the efficiency of a particular activity, then we need to have a measure of output as well as input. It may be an interesting observation that some project managers devote more time to collaboration, and consume more collaboration time from their team members, but in order to draw useful conclusions about collaboration efficiency, either we must assume that all project managers ought to need exactly the same quantity of communication, or we must have some way of linking collaboration with the overall performance of the team. After all, there are some working practices requiring high quantities of collaboration, such as pair programming, that are thought to be associated with high levels of overall productivity, so we shouldn't always regard collaboration time as an unnecessary cost. I don't need researchers to tell me too much is too much; what I want to know is how much is too much, in other words some notion of requisite collaboration.
The researchers think "it is important to reduce network connectivity at points where collaboration fails to produce sufficient value". Well yes obviously, but the critical question is how to assess the value of collaboration in the first place. See my post on Collaboration Impact Zones.