Figure 1.

Trade-off between tasks leads to phenotypes arranged along polygons in trait space (a) Standard approaches assume a fitness landscape in the space of phenotypic traits. Selection tends to favour phenotypes near the maximum. (b) When the system needs to perform several tasks, each task has a performance function in trait space, whose maximum is called the archetype. Each coloured hill represent a performance function P_i(T) for task i, where T is the vector of traits, and fitness at locale q is an increasing function of all performance functions (c) In an environment where task 1 is more important, fitness is maximized at a phenotype close to archetype 1; In an environment where task 3 is more important, fitness is maximized at a phenotype closer to archetype 3. Fitness maxima in the two environments are different. (d) Fitness maxima in all possible environments in which fitness is an increasing function of performance in the three tasks lie in the full triangle whose vertices are the three archetypes—known as the Pareto front—when the conditions of Shoval et al. [4] are fulfiled. (e) Possible polymorphism structures that keep a phenotype on the front: either mutation effects are parallel to the front (left), or mutations whose components perpendicular to the front cancel out (right). (f) Schematic of the selection process described in the Results section, for the case of three tasks that define a triangle whose vertices are the three archetypes. Locales and individuals best at the locales are sequentially removed into a survivor list, and remaining individuals are removed. Survivors make up the next generation. Here, we consider 7 locales (N = 7); phenotypes that maximize fitness in these locales are connected to the relevant locale with a black line.