The FRS-ESR facility at GSI provides unique conditions for precision measurements of large areas on the nuclear mass surface in a single experiment. Values for masses of 604 neutron-deficient nuclides (30≤Z≤92) were obtained with a typical uncertainty of 30 µu. The masses of 114 nuclides were determined for the first time. The odd-even staggering (OES) of nuclear masses was systematically investigated for isotopic chains between the proton shell closures at Z=50 and Z=82. The results were compared with predictions of modern nuclear models. The comparison revealed that the measured trend of OES is not reproduced by the theories fitted to masses only. The spectral pairing gaps extracted from models adjusted to both masses and density related observables of nuclei agree better with the experimental data.Significant progress has been achieved over the last years in constructing self-consistent mass models [1,2]. These models aim to reliably describe the properties of nuclei far off the valley of β-stability, where the experimental information is scarce or even not available yet. For instance, in modelling the astrophysical r-process of nuclear synthesis one needs precise knowledge of masses and half-lives of very exotic nuclei and one has to rely on theoretical predictions since most of the nuclides involved have not even been produced in the laboratory yet. The predictions for these nuclides dramatically deviate for the different models [2]. Thus new experimental data on exotic nuclei and consequently better understanding of nuclear structure away from the valley of β-stability is essential for further theoretical development.Odd-even staggering of nuclear binding energies (OES) was detected in the early days of nuclear physics [3] and was explained by the presence of pairing correlations between nucleons in the nucleus [4]. Pairing contributes only little to the total nuclear binding energy but its influence on the nuclear structure is significant.The common way to extract experimental information about the pairing correlations is to measure the value of the OES which approximates the pairing-gap energy (∆) in the standard Bardeen-Cooper-Schrieffer (BCS) theory [5]. The latter quantity is connected with the * Physical Review Letters 95 (2005) 042501, http://prl.aps.org/ strength of the pairing interaction (G):where ε ν is the single-particle energy and λ is the chemical potential. The summation goes over all single-particle levels ν below and above the Fermi energy. In order to evaluate this sum in local pairing functionals a (smooth) cut-off in energy is usually implemented. Neutron (∆ n ) and proton (∆ p ) pairing gaps are usually determined from finite-difference equations of measured masses [6], e.g. by the five-point formulae: