Knowledge of the 1 H magnetic properties of blood is important for developing models of the MR signal behavior in, for instance, the BOLD effect (1). Such models of free induction decay (FID) or spin echo (SE) experiments require accurate knowledge of the blood susceptibility and relaxation parameters as a function of blood oxygenation level. The blood signal is especially important in situations where the volume of interest contains large vessels and a major signal component arises directly from the MR behavior of bulk blood.It is widely assumed, and has been used in several modeling reports, that blood relaxation in the FID experiment can be described as pure Lorentzian behavior of the blood signal magnitude: S(t) ϰ exp(-R* 2 ⅐ t). It is shown herein that a substantial oxygenation-level-dependent non-Lorentzian component is an attribute of the bulk blood signal relaxation.As a note on nomenclature, the reader is reminded that exponential signal decay in the time domain (S(t) ϰ exp(-R* 2 ⅐ t)) leads to a Lorentzian lineshape (S() ϰ R* 2 / R* 2 2 ϩ 2 ) in the frequency domain. Herein we use the term "Lorentzian behavior" to describe frequency domain and time domain signal characteristics interchangeably. Likewise, Gaussian signal decay in the time domain (S(t) ϰ exp(-AR* ⅐ t 2 )) leads to a Gaussian lineshape (S() ϰ exp(-2 /4AR*)) in the frequency domain. As will be shown in this work, a Gaussian component arises in empirical modeling of the first 100 msec of the blood time domain signal magnitude. However, this is only a useful approximation. Time domain data demonstrate small (less than 2%) but significant systematic deviations from our empirical model, a simple product of Lorentzian and Gaussian components. To emphasize the fact that our empirical model does not fully account for the blood signal's relaxation characteristics (and that the model is not based on a biophysical model of blood magnetic properties) we use the generic phrase "non-Lorentzian signal behavior" to indicate as yet unexplained deviations from pure exponential/Lorentzian characteristics.Studies of the magnetic properties of human blood have been undertaken previously, and the results were insightful (2-13). However, as BOLD-related functional MRI (fMRI) procedures increase in precision, there is need to ensure the quantitative nature of the blood MR parameters upon which interpretation of experimental results may depend. Previous blood MR literature describes varying degrees of mismatch between the experimental conditions used for in vitro studies and human blood in situ. These differences include temperature (6,13,14), pH (13), and the settling of erythrocytes from plasma in the case of static samples (3,5-7,10 -12,14). Further, the relaxation rate constants in whole blood depend on magnetic field strength (3,6); hence it is important to determine these relaxation parameters at magnetic field strengths typical of human imaging systems.Of the in vitro studies cited herein, only two ensured a continuous mixing of the blood sample in order t...