The researchers used a computer warfare simulation that examines behaviour in tight, demanding circumstances, where teams of four protect their territories by correctly identifying enemy intruders and then quickly destroying them. Team-mates used separate monitors, but shared a room and could freely converse. They recruited 304 management undergraduates, half of whom were given straightforward group incentives: $10 each if their group outperformed a specified rival group.
The other teams were given mixed incentives: group performance could lead to $5 each , and individually outdoing a specified member of another team garnered another $5. Participants who were individually incentivised were hungrier for scores, being significantly faster at destroying intruders. However, heavily penalised illegitimate attacks ('friendly fire') were more common in these teams. This slump in quality suggests a drop-off in the flow of information typical in close teams, making it harder to detect and ward off errors as attention was turned towards delivering immediate personal objectives.
The study also examined direct helping behaviour, in terms of the efforts made to destroy intruders in team-mate territory rather than your own. This mattered, as each team had a high workload member who was constantly swarmed with as many radar blips as the others had put together. Participants with pure group-level incentives showed more helping behaviours than their mixed incentive counterparts.
Barnes and his colleagues suggest that mixed incentives present a conflict between maximising individual interests and that of the collective, and the temptation is to focus on your own priorities, letting others hold the fort for you. Moreover, if you doubt that they will, you'd be even more of a sucker to vainly do so yourself. This amounts to a social dilemma akin to the prisoner's dilemma, which pressurises players towards self-serving behaviours.
I felt - and the authors do note - that the experimental paradigm relates best to 'task forces' whose urgent tasks necessitate trade-offs between different behaviours. I'm skeptical about generalising to workplaces which are more elastic: I may forgo reading my book over lunch in order to help you out, feel rewarded by this, and spend the afternoon contributing just as much or more to my own goals. Nevertheless, by plugging social dilemmas in to the research on incentives, this article highlights that tweaking incentives can result in tradeoffs, not simply the best of both worlds.