The wide luminosity dispersion seen for stars at a given effective temperature in the Hertzsprung-Russell diagrams of young clusters and star forming regions is often interpreted as due to significant ( \sim 10 Myr ) spreads in stellar contraction age . In the scenario where most stars are born with circumstellar discs , and that disc signatures decay monotonically ( on average ) over timescales of only a few Myr , then any such age spread should lead to clear differences in the age distributions of stars with and without discs . We have investigated large samples of stars in the Orion Nebula Cluster ( ONC ) using three methods to diagnose disc presence from infrared measurements . We find no significant difference in the mean ages or age distributions of stars with and without discs , consistent with expectations for a coeval population . Using a simple quantitative model we show that any real age spread must be smaller than the median disc lifetime . For a log-normal age distribution , there is an upper limit of < 0.14 dex ( at 99 per cent confidence ) to any real age dispersion , compared to the \simeq 0.4 dex implied by the Hertzsprung-Russell diagram . If the mean age of the ONC is 2.5 Myr , this would mean at least 95 % of its low-mass stellar population has ages between 1.3–4.8 Myr . We suggest that the observed luminosity dispersion is caused by a combination of observational uncertainties and physical mechanisms that disorder the conventional relationship between luminosity and age for pre main-sequence stars . This means that individual stellar ages from the Hertzsprung-Russell diagram are unreliable and can not be used to directly infer a star formation history . Irrespective of what causes the wide luminosity dispersion , the finding that any real age dispersion is less than the median disc lifetime argues strongly against star formation scenarios for the ONC lasting longer than a few Myr .