The blood oxygenation level dependent signal of cerebral tissue can be theoretically derived using a network model formed by randomly oriented infinitely long cylinders. The validation of this model by phantom and in vivo experiments is still an object of research. A network phantom was constructed of solid polypropylene strings immersed in silicone oil, which essentially eliminated the effect of spin diffusion. The volume fraction and magnetic property of the string network was predetermined by independent methods. Ten healthy volunteers were measured for in vivo demonstration. The gradient echo sampled spin echo signal was evaluated with the cylinder network model. We found a strong interdependency between the two network characterizing parameters deoxygenated blood volume and oxygen extraction fraction. Here, different sets of deoxygenated blood volume/oxygen extraction fraction values were able to describe the measured signal equally well. However, by setting one parameter constant to a predetermined value, reasonable estimates of the other parameter were obtained. The same behavior was found for the in vivo demonstration. The signal theory of the cylinder network was validated by a well-characterized phantom. However, the found interdependency that was found between deoxygenated blood volume and oxygen extraction fraction requires an independent estimation of one variable to determine reliable values of the other parameter.