The baroreflex (BR) is an important physiological regulatory mechanism which reacts to blood pressure perturbations with reflex changes of target variables such as the heart period (electrocardiogram derived RR interval) or the peripheral vascular resistance (PVR). Evaluation of cardiac chronotropic (RR as a target variable) and vascular resistance (target PVR) BR arms was in previous studies mainly based on the use of the spontaneous variability of the systolic or diastolic blood pressure (SBP, DBP), respectively, as the input signals. The use of other blood pressure measures such as the mean blood pressure (MBP) as an input signal for BR analysis is still under investigation. Making the assumption that the strength of coupling along the BR indicates the more appropriate input signal for baroreflex analysis, we employ partial spectral decomposition to assess in the frequency domain the causal coupling from SBP, MBP or DBP to RR or PVR. Noninvasive beat-to-beat recording of RR, SBP, MBP and DBP and PVR was performed in 39 and 36 volunteers in whom orthostatic and cognitive loads were evoked respectively through head-up tilt and mental arithmetic task. At rest, the MBP was most tightly coupled with RR, in contrast to the analysis of the vascular resistance BR arm where the results showed similar importance of all blood pressure input signals. During orthostasis, the increased importance of SBP as the input signal for BR analysis along the cardiac chronotropic arm was demonstrated. In addition, the gain from MBP to RR was more sensitive to physiological state changes compared to gains with SBP or DBP signal as inputs. We conclude that the coupling strength depends not only on the analysed baroreflex arm but also on the selection of the input blood pressure signal and the physiological state. The MBP signal should be more frequently used for the cardiac baroreflex analysis.