In this paper cross time-frequency (TF) analysis is used to estimate the phase differences and the phase locking between cardiovascular signals. Phase differences give a measure of the changes in the synchronization between two oscillations, while phase locking measures the degree of similarity of these changes across subjects. The methodology is based on the smoothed pseudo Wigner-Ville distribution and includes coherence analysis. In a simulation study involving R-R variability (RRV) signals, this methodology provided accurate estimates of phase differences, with an error characterized by interquartile ranges lower than 2% and 10% for SNR of 20 dB and 0 dB, respectevely. A comparative study showed that the proposed estimator outperformed an estimator based on the integration of the difference between the instantaneous frequencies of the signal spectral component. The presented methodology was used to characterize the interactions between RRV and systolic arterial pressure variability during tilt table test. Head-up tilt caused the phase differences (time delay) to change about 0.48 rad (361 ms) in HF range [0.15, 0.5 Hz]. The phase locking, which decreased immediately after the head-up tilt, was restored in about 2 minutes.