Nuclear magnetic relaxation rates are measured for whole blood, blood plasma, whole blood clots, and plasma clots in vitro. Relaxation rates are linear in the hematocrit and transverse relaxation rates are significantly greater than longitudinal relaxation rates. Longitudinal relaxation rates measured from 0.01 to 42 MHz for proton Larmor frequencies are found to decline monotonically with increasing magnetic field strength; however, the dispersion curves do not follow a simple Lorentzian behavior, which is anticipated in a suspension of particles in a solution of proteins having a distribution of molecular weights. The transverse relaxation rate is a function of the acquisition parameters, in particular, the choice of TE in either Hahn echo experiments or in echo-train experiments. The origin of this dependence of T2 on TE or the interpulse spacing in an echo train is identified with the exchange of water from inside the red blood cell to the outside and is only an important relaxation mechanism in the case where the blood cell membrane is intact and the cell contains deoxygenated hemoglobin. The dependence of the apparent transverse relaxation rate on the interpulse spacing in a Meiboom-Gill-Carr-Purcell pulse sequence provides the estimate that the mean residence time of water inside the blood cell is about 10 ms. These data provide a sound basis for understanding the dependence of magnetic images on magnetic field strength and the choices of the image acquisition parameters, TE and TR.