Traditional electromyography-assisted optimization (TEMG) models are commonly employed to compute trunk muscle forces and spinal loads for the design of clinical/treatment and ergonomics/prevention programs. These models calculate muscle forces solely based on moment equilibrium requirements at spinal joints. Due to simplifications/assumptions in the measurement/processing of surface EMG activities and in the presumed muscle EMG-force relationship, these models fail to satisfy stability requirements. Hence, the present study aimed to develop a novel stability-based EMG-assisted optimization (SEMG) method applied to a musculoskeletal spine model in which trunk muscle forces were estimated by enforcing equilibrium conditions constrained to stability requirements. That is, second-order partial derivatives of the potential energy of the musculoskeletal model with respect to its generalized coordinates were enforced to be positive semi-definite. Fifteen static tasks in upright and flexed postures with and without a hand load at different heights were simulated. The SEMG model predicted different muscle recruitments/forces (generally larger global and local muscle forces) and spinal loads (slightly larger) compared to the TEMG model. Such task-specific differences were dependant on the assumed magnitude of the muscle stiffness coefficient in the SEMG model. The SEMG model-predicted and measured L4-L5 intradiscal pressures were in satisfactory agreement during simulated activities. © 2018 IPEM