LiNi2O4 spinel-type phases were prepared by thermal treatment of electrochemically deintercalated layered Li similar to 0.5NiO2. The phase transformation was followed by Li-7 NMR, showing a gradual change of the signal from the layered compound. The characteristic signal of the latter (related to local Li/vacancy and Ni3+/Ni4+ ordering) vanishes after heating to 150 degrees C and is replaced by a new signal showing faster exchange kinetics (originating from Ni3+/Ni4+ hopping around Li), which progressively transforms into a broad distribution of signals. Around 200 degrees C, a set of three positively shifted signals is observed, corresponding to the appearance of the spinel phase as seen from XRD; these signals disappear after heating to 240 degrees C, corresponding to the beginning of decomposition of the spinel into a disordered R (3) over barm type phase with oxygen evolution as previously shown by Guilmard et al. (Chem. Mater. 2003, 15, 4476 and 4484). In an ideal LiNi2O4 spinel, only one Li-7 NMR signal is expected. DFT (GGA) calculations were carried out and show that the mechanism for the electron spin density transfer from NiO6 octahedra to corner-sharing LiO4 tetrahedra with close to 120 degrees Ni-O-Li configuration is a delocalization one, although the p orbitals on oxygen do not present ideal orientation, leading to a much weaker transfer compared to cases where both Ni and Li are in octahedral coordination with 180 degrees Ni-O-Li configuration. The complex but well-defined experimental NMR signals consistently observed show that the material is far from the ideal spinel structure. However, it could not be correlated to the actual stoichiometry of the compound. It was therefore tentatively assigned to structural defects resulting from incomplete migration of Ni ions from their site to the Li layer in the pristine compound, such as partial occupation of tetrahedral sites.