The introduction of different methodologies for the construction and screening of complex protein libraries has provided powerful means in protein engineering for the development of molecules with desired traits. A challenge faced in many situations is to adapt a given methodology for efficient and rapid identification of the most interesting variants present in a library. In the present study, the concept of Darwinian selection based on a growth advantage for clones having the desired trait has been investigated. Using a beta-lactamase-based PCA (protein fragment complementation assay), affinity maturation of a TNFalpha (tumour necrosis factor alpha)-binding Affibody molecule with an initial 2 nM affinity for the target has been performed. Initial characterization of the PCA system, based on the affinity-driven reconstitution of beta-lactamase activity in the periplasm of cells harbouring a library member showing affinity for a co-expressed target protein, showed that the system was responsive to promoter induction level, interaction affinity and applied selection pressure. Using combinatorial protein engineering principles, a 107 library of second-generation Affibody molecules was constructed and subjected to selection of improved variants by library growth in liquid culture. The results show that, after a pre-selection step on semi-solid medium to eliminate non-binding variants, present in the majority, two rounds of selection in liquid culture resulted in an enrichment for binders showing up to 8-fold higher affinity for the TNFalpha target than the ancestral variant. Biosensor analyses showed that the major factor for the improved affinity could be attributed to reduced off-rate constants.