When an electromagnetic field of one specific frequency equal to the Larmor frequency of hydrogen protons is sent to the underground by a transmitter on the surface part of its energy is absorbed exclusively by the water molecules. When the excitation field is removed, the absorbed energy is released in the form of a new electromagnetic field which can be detected by a receiver at the surface. This response can only be produced by water, and has some identity characteristics: the released energy has the Larmor frequency for hydro-gen protons and produces a voltage e(t) which amplitude decays exponentially with time until it vanishes, at a rate or decay time that depends on the mean size of the pores. The maximum voltage amplitude is directly proportional to the amount of water. When the exci-tation field is increased the signal comes from deeper parts of the subsurface, allowing making a sounding (Magnetic Resonance Sounding, MRS). The principal factors that affect the measured water signal are the depth and thickness of the water bearing layer, the electrical conductivity of the subsurface, the magnitude and inclination of the geomagnetic field, the type of water containing rocks, the size of the antenna and the electromagnetic noise. The general rule is that the maximum amplitude of the signal decreases when increas-ing the depth of the aquifer. The existence of rocks with high conductivity has a screening effect that lowers the amplitude of the water signal and causes the depth of investigation to be decreased. When magnetic rocks are present, the intensity and gradient of the geo-magnetic field is different in different parts of the subsurface what causes the Larmor frequency be also different and hence MRS signal can not be detected. The best conditions are met in resistive-non magnetic environments, high latitudes, and coarse grained rocks. A mathematical model of the physical phenomenon and its measurement allows getting the theoretical signal produced by a defined con-figuration of the ground water and rocks distribution in the subsurface (modelling or direct problem), as well as deducing the aquifers configuration (distribution of the amount of water and time decay with depth) from the MRS signal actually measured in the field (inver-sion or inverse problem). Accuracy of the results depends on the signal quality and the hypothesis and assumptions made in the model.