The objective of this study was to determine whether or not the underlying physiological systems that generates spontaneous arterial blood pressure (ABP), cerebral blood flow velocity (CBFV), and intracranial pressure signals could be adequately approximated as a linear stochastic process. Furthermore, a new measure (C) capable of capturing the degree of nonlinear dependency between two ABP and CBFV signals (including a time-varying situation) was proposed for quantifying the degree of cerebral blood flow autoregulation. A surrogate data test of fifteen ABP, CBFV, and intracranial pressure (ICP) segments was conducted for detecting whether there exists a statistically significant deviation from the null hypothesis of linear signals. The extension of the established block computation method of C measure to an adaptive one was achieved. This new algorithm was then applied to study the C evolution using brain injury patients data from a hyperventilation study and two propofol studies. Nonlinearity has not been detected for all the fifteen recordings, neither has nonlinear dependency between CBFV and ABP. However, their presences in some of the signal segments justified the adoption of a nonlinear measure of dependency capable of characterizing both linear and nonlinear correlations for inferring autoregulation status. C measure started to decrease with the introduction of hypocapnia state indicating that hyperventilation may reduce the dependency of CBFV on ABP fluctuations. On the other hand, complex patterns of C measure evolution were observed among 14 cases of propofol data indicating a nontrivial effect of propofol on the dependency of CBFV on ABP.