I describe several assays designed to examine how costs and benefits interact in the development of mutualisms between species. A mutualism occurs between alpheid shrimp and gobiid fishes. These shrimp have poor vision but good burrowing ability. Individual shrimp share their burrows with a goby that, with good vision but no burrowing ability, acts as a watch-out warning shrimp when predators approach. In the Caribbean, a single species, Nes longus, which has been described as a mutualist, follows these behaviors. Others, such as Ctenogobius saepepallens, casually use shrimp burrows, rarely warn shrimp of danger, and are better described as commensalists. I found that N. longus more effectively avoids predators while using shrimp burrows than C. saepepallens. Thus, tight mutualism with shrimp is advantageous, especially in areas where shrimp burrows are abundant. I have quantified several behaviors that likely allow N. longus to use burrows more effectively. Why then would C. saepepallens not evolve such behaviors and become a strict mutualist if strict mutualism is advantageous? For gobies, there is likely a cost associated with mutualism with shrimp. To warn shrimp, gobies must remain at a burrow entrances and restrict foraging to that small area. I found that on the same restricted diet, C. saepepallens lost more weight than N. longus. Thus, C. saepepallens may be constrained to a casual association with shrimp due to foraging requirements. This story indicates that strict mutualism may evolve infrequently because few species can overcome the inherent costs of mutualism.