Phenotypic diversity, also known as phenotypic heterogeneity, occurs when individuals within a species have different phenotypes. Phenotypic diversity plays a crucial role in adaptation, and understanding how it is maintained is fundamental for comprehending the complexity of life. In this thesis, I use theoretical approaches to investigate different mechanisms that support the selection for variation. Specifically: (1) I model the optimal degree of coordination in the division of labour in a non-clonal population, and predict that communication between individuals would be less frequent in a population with low relatedness; (2) I show that the temporal scale of abiotic environmental fluctuations can play a crucial role in favouring a monomorphic or polymorphic population, that short-term variation selects for diversity, but long-term variation selects against it; (3) I develop and test a new form of cheating strategy that manipulates other social group members and find that it can lead to arms-race-like dynamics, which can help explain the genetic complication found in nature; (4) I construct a general model to incorporate different properties in experimental setups and provide general principles for when the cheater-cooperator ratio remains stable and when oscillations occur; (5) I examine how comparative studies, an important tool for validating theoretical predictions, have quantified abiotic environments and highlight the potential problem with correlating climatic quantities and using environmental predictability. Taken together, this body of work extends our knowledge on phenotypic diversity and identifies several promising research directions related to eco-evolutionary feedbacks.