I will present results of new paradigms that address the neural computations for human spatial orientation and closed-loop navigation. These computations may rely on sensory signals (visual, vestibular), motor feedback and internal predictions, as well as cognitive signals, which all have differences in noise properties, internal dynamics and intrinsic reference frames. How these signals are integrated in the underlying perception-action loops has remained difficult to quantify since the quality of these signals can typically not be measured in isolation. I will present our recent work combining behavioral modeling and experimental psychophysics to characterize signal (re)weighting in ego- and allocentric reference frames in the perception-action loops for spatial orientation and navigation in health and disease.