This first commentary will be on the paper by Edwards, Hassall, Sutherland and Y, Selection for protection in an ant-plant mutualism: host sanctions, host modularity, and the principal agent game, published in Proceedings of the Royal Society of Biology (2006) with some additional comments on the arguments Douglas Yu put forward at the TECT meeting.

The central arguments of the paper revolve around the observed facts that:

1. mutualisms evolve from antagonistic interactions,

2. and persist in the face of invasion by specializes parasites, and selection for cheating by the mutualists themselves.

I assess specifically the economic basis for the arguments presented by Edwards et al. and ask the reader to refer to the paper for the specific details offered on the experimental design, results, etc.

The essential points are the following, there are species of plants that have ant symbionts which populate the domatia that are produced by the plant. These domatia produce relevant benefits to the symbiont ant. Symbiont ants protect the plant from herbivory by investing in protection (a sunk cost), and thus maintain access to the domatia. Ants that seem not to protect the plant from herbivory (approx 10% of the ants that inhabit this specific plant species) that are found not to patrol (through some signal) will lose the benefit of the domatia. The argument is that this interaction is one that is customary to the principal-agent literature: there are two types of agents cooperators (protectors) and defectors (cheaters/non-patrollers). The idea is that the plant (the principal) will design a contract so that it can read the signals of the agents in order to know whether the agent is abiding by the contract. If the agent does not abide by the contract it is punished by the removal of the domatia on the branch that it did not protect.

Crucially, we must note several factors that affect how the ant-agents behave and how the plant-principals act. The plant-principals signal their success by the number of domatia that they produce for the ants. The ant-agents signal their successful defence by ensuring that herbivory does not occur. In the context of the experiment, herbivory was simulated by cutting off leaves in the area where the ants patrolled the branches. These branches would then die off if a certain number of leaves were cut off (leaves cut = 4), but not for a certain lower number of leaves removed (leaves cut = 2). This would indicate that only 'gross misconduct/laxity' by the ants (i.e. something similar to the complete lack of protection offered by the one ant species) would result in domatia mortality.

The study found that a certain pertinent level of domatia (d=23) was the level at which selection between cooperators and defectors would be ambivalent. However, if the number of domatia exceeded this level then it is likely that protectors would be selected. This seems to be coherent with the data as the average number of domatia on the plants was 28 in the site under inspection. Consider this as a wage problem in the principal-agent setting. The principal needs to set some optimal wage in order for the agent to maximise their output. Moreover, the wage can act as a selection tool, reading the signals that are sent by the respective agents.

I believe that the paper offers a really high quality and interesting application of economic insights to the world of biology. Edwards et al. correctly assert that modelling mutualism as a market structure would be erroneous on account of the information asymmetries that would be present in such a market and the inability for the mutualists to have a 'complete contract' (me imputing economics to what they are saying). Thus the use of the principal-agent problem as a descriptive technique is almost certainly more accurate than that progressed by Noë and Hammerstein (1994) for markets in mutualisms. The economics that Edwards et al. use seems intuitively applied and theoretically well-grounded. Douglas Yu is currently moving this work forward by looking at other plant-ant mutualisms, as well as at fig wasp-fig tree mutualisms. The paper on that topic is currently a work in progress, but I look forward to its results, especially because of the fact that 'abscising' (cutting off) of figs in which excess oviposition has occurred does not seem to be a general pattern (if I recall his talk accurately).