The merger 1 of close binary systems containing two neutron stars should produce a burst of gravitational waves, as predicted by the theory of general relativity 2 . A reliable estimate of the double-neutron-star merger rate in the Galaxy is crucial in order to predict whether current gravity wave detectors will be successful in detecting such bursts. Present estimates of this rate are rather low 3−7 , because we know of only a few doubleneutron-star binaries with merger times less than the age of the Universe. Here we report the discovery of a 22-ms pulsar, PSR J0737−3039, which is a member of a highly relativistic double-neutron-star binary with an orbital period of 2.4 hours. This system will merge in about 85 Myr, a time much shorter than for any other known neutron-star binary. Together with the relatively low radio luminosity of PSR J0737−3039, this timescale implies an order-of-magnitude increase in the predicted merger rate for double-neutron-star systems in our Galaxy (and in the rest of the Universe). PSR J0737−3039 was discovered during a pulsar search carried out using a multibeam receiver 8 on the Parkes 64-m radio telescope in New South Whales, Australia. The original detection showed a large change in apparent pulsar period during the 4-min observation time, suggesting that the pulsar is a member of a tight binary system. Follow-up observations undertaken at Parkes consisting of continuous ∼ 5-hour observations showed that the orbit has a very short period (2.4 hrs) and a significant eccentricity (0.088). The derived orbital parameters implied that the system is relatively massive, probably consisting of two neutron stars, and predicted a huge rate of periastron advanceω due to effects of general relativity. Indeed, after only a few days of pulse-timing observations we were able to detect a significant value ofω.Interferometric observations made using the Australia Telescope Compact Array (ATCA) in the 20-cm band gave an improved position and flux density for the pulsar. Knowledge of the pulsar position with subarcsecond precision allowed determination of the rotational period derivative,Ṗ , and other parameters from the available data span. Table 1 reports results derived from a coherent phase fit to data taken over about five months. The measured value ofω = 16.88 • yr −1 is about four times that of PSR B1913+16 (ref. 9), previously the highestknown. If the observedω is entirely due to general relativity, it indicates a total system mass M = 2.58 ± 0.02 M , where M is the mass of the Sun. Figure 1 shows the constraints on the masses of the pulsar and its companion resulting from the observations so far and the mean pulse profile as an inset. The shaded region indicates values that are ruled out by the mass function M f and the observeḋ ω constrains the system to lie between the two diagonal lines. Together, these constraints imply that the pulsar mass m p is less than 1.35 M and that the companion mass m c is greater than 1.24 M . The derived upper limit on m p is consistent with the