This paper investigates the role that load-limiters play with respect to the performance of occupant protection systems, with focus on performance in frontal crashes. Modern occupant protection systems consist of not just the seat belt, but also airbags, interior vehicle surfaces and vehicle structure. Modern seat belts very often incorporate load-limiters as well as pretensioners. Published research has established that load-limiters and pretensioners increase the effectiveness of occupant protection systems. Some have argued that load-limiters with higher deployment thresholds are always better than load-limiters with lower deployment thresholds. Through testing, modeling and analysis, we have investigated this hypothesis, and in this paper we present test and modeling data as well as a discussion to this data and engineering mechanics to explain why this hypothesis is incorrect. Research presented in this paper shows that because load-limiters are just one component of a multi-component occupant protection system, the performance of the overall occupant protection system cannot be predicted from the load-limiter performance or specifications alone. The overall occupant protection system is designed such that the load-limiter works in conjunction with the webbing stiffness, airbag, vehicle interior surfaces and vehicle structure to provide effective occupant protection. This paper shows that effective occupant protection has been achieved using different combinations of these parameters. The appropriate method of evaluating the effectiveness of an occupant protection system is to evaluate the overall performance of that occupant protection system in sled and crash tests and/or calibrated modeling. Tests of load-limiters alone or reliance on the load-limiter specification alone is not a good indicator of the overall effectiveness of an occupant protection system. Tests were conducted to establish the level at which certain retractor load-limiters deployed webbing. This data was cross-referenced with publically available test data, and shoulder belt loads, chest deflection and chest compression, among other injury metrics. Computer modeling was conducted to assess the effect of varying initial load-limiter deployment loads.