The critical closing pressure (CrCP) of cerebral circulation is normally estimated by extrapolation of instantaneous velocity-pressure curves. Different methods of estimation were analysed to assess their robustness and reproducibility in both static and dynamic applications. In ten healthy subjects (mean ± SD age 37.5 ± 9.2 years) continuous recordings of arterial blood pressure (BP, Finapres) and bilateral cerebral blood flow velocity (transcranial Doppler ultrasound, middle cerebral arteries) were obtained at rest. Each session consisted of three separate 5 min recordings. A total of four recording sessions for each subject took place over a 2 week period. A total of 117 recordings contained 34 014 cardiac cycles. For each cardiac cycle, CrCP and resistance-area product (RAP) were estimated using linear regression (LR), principal component analysis (PCA), first harmonic fitting (H1), 2-point systolic/diastolic values (2Ps) and 2-point mean/diastolic values (2Pm). LR and PCA were also applied using only the diastolic phase (LRd, PCAd). The mean values of CrCP and RAP for the entire 5 min recording ('static' condition) were not significantly different for LRd, PCAd, H1 and 2Pm, as opposed to the other methods. The same four methods provided the best results regarding the absence of negative values of CrCP and the coefficient of variation (CV) of the intra-subject standard error of the mean (SEM). On the other hand, 'dynamic' applications, such as the transfer function between mean BP and RAP (coherence and RAP step response) led to a different ranking of methods, but without significant differences in CV SEM coherence. For the CV of the RAP step response though, LRd and PCAd performed badly. These results suggest that H1 or 2Pm perform better than LR analysis and should be used for the estimation of CrCP and RAP for both static and dynamic applications.