We calculate the electron paramagnetic resonance (EPR) spectra of the antiferromagnetic spin-1 chain compound Y2BaNi1−xMgxO5 for different values of x and temperature T much lower than the Haldane gap (∼ 100K). The low-energy spectrum of an anisotropic Heisenberg Hamiltonian, with all parameters determined from experiment, has been solved using DMRG. The observed EPR spectra are quantitatively reproduced by this model. The presence of end-chain S = 1/2 states is clearly observed as the main peak in the spectrum and the remaining structure is completely understood. This picture was supported by EPR measurements of [Ni(C 2 H 8 N 2 ) 2 (NO 2 )]ClO 4 (NENP) doped with nonmagnetic ions [7], where resonances corresponding to the fractional spin S = 1/2 states at the "open" ends of the S = 1 Ni chains were observed. Similar measurements for doping with magnetic ions are also consistent with S = 1/2 end states [8]. This effect, if robust, would correspond to the only instance in magnetism where a low energy collective excitation has no classical analog.However, Ramirez et al.[9] also tested the presence of free S = 1/2 states by studying the specific heat of non-magnetic defects in Y 2 BaNiO 5 , with magnetic fields up to 9T and temperatures down to 0.2K. They found that the shape and magnitude of the Schottky anomaly associated with the defects in Y 2 BaNi 1−x Zn x O 5 are better described by a simple model involving spin-1 excitations, instead of the S = 1/2 excitations of the VBS. In order to eliminate the apparent discrepancy with the EPR measurements in NENP, Ramirez et al. [9] pointed out the possibility that a small fraction of ethylene diamine complexes in NENP acquire charge at structural defects induced by Zn doping. As it is not uncommon that structural defects induce a paramagnetic behavior in organic compounds, they propose that similar EPR measurements should be performed on Y 2 BaNi 1−x A x O 5 (A a nonmagmetic ion) where x is more easily calibrated [9]. EPR is an appropriate technique to determine the existence of S=1/2 end states because it allows to distinguish between S=1/2 and S=1 spins in the presence of spatial anisotropy.In a recent paper [6], a precise fit of the above mentioned specific heat measurements was done, solving the low-energy spectrum of an anisotropic Heisenberg Hamiltonian for Y 2 BaNi 1−x Zn x O 5 . It was shown that there is no contradiction between EPR in NENP [7] and specific heat measurements in Y 2 BaNi 1−x Zn x O 5 [9]. The results supported the existence of S = 1/2 excitations for sufficiently long chains and clearly indicated that the anisotropy plays a very important role in the low temperature properties. However, EPR experiments would unambiguously detect the presence of such excitations and definitely confirm the validity of the theory.Recently, Saylor et al [13] have measured the EPR spectra of Y 2 BaNi 1−x Mg x O 5 at temperature T = 2K for different values of Mg concentration. The spectra show a prominent main peak, which can be associated with free S=1/2 spins, surr...