Heart rate variability (HRV) is a fundamental basis for evaluating the residual autonomic nerve function in patients with spinal cord injury (SCI). The commonly used methods for analyzing HRV involve calculation of time and frequencydomain or nonlinear indices of RR interval series derived from ECG signals, and the detrended fluctuation analysis (DFA) is one of the most commonly used nonlinear methods. Typically, DFA of HRV can yield two scaling exponents: α1 (<11 beats) and α2 (>11 beats). The present study was aimed to investigate whether the local scaling exponent α(t) (t is the time scale) could reveal the dynamics of HRV that cannot be characterized by spectral analysis and scaling exponents. ECG signals were collected from 12 SCI patients and 15 healthy subjects during upright sitting and prone posture. A new method was used to calculate α(t) involving correcting the α(t) values at small scales. The results showed that, for both sitting and prone postures, α(t) in healthy subjects monotonically decreased in the range of 4 s