Relationships between traits and fitness are complex
The relationships between traits and fitness in a given environment are defined by multiple selective coefficients that shape the evolutionary course of populations. Natural Selection explains much of evolutionary processes that link variation in life expression with success in a given environmental context.
The total selection pressure on a given trait comprises the effects of direct, indirect and correlated selection on that trait. Selection of trait values may be direct, as when the environmental pressures affect the distribution of the values of a trait in the population. Or selection may be indirect, as when distribution of trait values is modified due to correlations with other traits under selection. Indirect selection occurs because phenotypes are an integration of realized multiple trait values within a single individual.
Also, selection may be correlated, as when it acts on particular trait combinations (Lande and Arnold 1983). For example, the combination of fruit size and seed size may be a cue for dispersers about pulp quantity. Thus, a negative correlation between these two traits may be related to higher dispersal success (Sobral et al. 2010). Traits may be pleiotropically correlated because they are related to the same genes and biosynthetic pathways. Other reasons for trait correlations include mechanistic relationships. In these relationships, mechanistic reasons preclude the existence of infinite combinations. An example of a mechanistic relationship would occur when the length of a given organ is correlated with the width. Although these traits could be codified by independent genes, and modified by independent selective pressures, the length and width are mechanistically related. The length of a fruit, for instance, must cohere with the width.
All these kinds of selective pressure can be linear or non-linear—that is, quadratic. (Phillips and Arnold, 1989). Quadratic selection implies a trait has different relationships with fitness. These relationships depend on the trait’s own value, which causes non-linear selective curves. These curves may be concave or convex, where bimodal or unimodal distributions of the traits would be selected (Stinchcombe et al 2008). This analysis implies that the fitness-trait relationship is so complex that it may vary depending on the values of the trait itself.