In this paper, we present a new perspective to design an unpowered exoskeleton for metabolic rate reduction in running. According to our studies on human biomechanics, it was observed that having a torsional spring that applies torque as a linear function of the difference between two hips angles ( -angle), compared with a local spring which applies torque as a function of hip angle ( -angle), provides a better condition for hip moment compensation and, consequently, metabolic rate reduction. Accordingly, a new type of unpowered exoskeleton device for realization of this idea was designed, and a prototype of this exoskeleton was constructed. This exoskeleton was tested on 10 healthy active subjects for running at 2.5 m s-1. In this experiment, 8.0 ± 1.5% (mean ± s.e.m.) metabolic rate reduction (compared with the no-exoskeleton case) was achieved.