A number of reports have advocated the use of Heart Rate Variability (HRV) as a non invasive method of monitoring the Autonomic Nervous System (ANS). In the anesthesia and critical care monitoring settings, the development of an instrument able to provide real-time information about the ANS state at different stages of any procedure would provide improved safety for patients undergoing diagnostic or therapeutic interventions. However, real-time analysis of HRV can be particularly challenging since larger effective lengths of observation provide better spectral resolution. Our study explores a probabilistic approach that analyzes changes in HRV parameters obtained from an autoregressive (AR) model technique using Burg's methods to evaluate very short observation windows while preserving appropriate frequency resolution. These HRV parameters are continuosly compared to a baseline state, and a probability trend is updated during provocative maneuvers. Preliminary results show that trends from classical parameters such as RMSSD and LFn are consistent and reliable instruments capable of providing significant information about ANS fluctuations in a timely fashion.