Gravitational waves are expected to be radiated by supermassive black hole binaries formed during galaxy mergers. A stochastic superposition of gravitational waves from all such binary systems will modulate the arrival times of pulses from radio pulsars. Using observations of millisecond pulsars obtained with the Parkes radio telescope, we constrain the characteristic amplitude of this background, A c,yr , to be < 1.0×10-15 with 95% confidence. This limit excludes predicted ranges for A c,yr from current models with 91-99.7% probability. We conclude that binary evolution is either stalled or dramatically accelerated by galactic-center environments, and that higher-cadence and shorter-wavelength observations would result in an increased sensitivity to gravitational waves.Studies of the dynamics of stars and gas in nearby galaxies provide strong evidence for the ubiquity of supermassive (> 10 6 solar mass) black holes (SMBHs) (1). Observations of luminous quasars indicate that SMBHs are hosted by galaxies throughout the history of the universe (2) and affect global properties of the host galaxies (3). The prevailing dark energycold dark matter cosmological paradigm predicts that large galaxies are assembled through the hierarchical merging of smaller galaxies. The remnants of mergers can host gravitationally bound binary SMBHs with orbits decaying through the emission of gravitational waves (GWs) (4).Gravitational waves from binary SMBHs, with periods between ~ 0.1 and 30 yr (5), can be detected or constrained by monitoring, for years to decades, a set of rapidly rotating millisecond pulsars (MSPs) distributed throughout our galaxy. Radio emission beams from MSPs are observed as pulses that can be time-tagged with as small as 20 ns precision (6). When traveling across the pulsar-Earth line of sight, GWs induce variations in the arrival times of the pulses (7).The superposition of GWs from the binary SMBH population is a stochastic background (GWB), which is typically characterized by the strain-amplitude spectrum h c (f)=A c,yr [f/(1 yr -1 )] -2/3 , where f is the GW frequency, A c,yr is the characteristic amplitude of the GWB measured at f = 1 yr -1 , predicted to be A c,yr > 10 -15 (5,(8)(9)(10)(11)(12), and -2/3 is the predicted spectral index (5,(8)(9)(10)(11)(12). The GWB will add low-frequency perturbations to pulse arrival times. While the detection of the GWB would confirm the presence of a cosmological population of binary SMBHs, limits on its amplitude constrain models of galaxy and SMBH evolution (8).As part of the Parkes Pulsar Timing Array project to detect GWs (6), we have been monitoring 24 pulsars with the 64-m Parkes radio telescope. We have produced a new data set, using observations taken at a central wavelength of 10 cm and previously reported methods (6,8), that spans 11 yr, which is 3 yr longer than previous data sets analyzed at this wavelength. In addition to having greater sensitivity to the GWB because of the longer duration, the data set was improved by identifying and correc...