We present a method of directly testing whether time continues to have its usual meaning on scales of <= t_P = sqrt(hbar G/c^5) ~ 5.4E-44 s, the Planck time. According to quantum gravity, the time t of an event cannot be determined more accurately than a standard deviation of the form sigma_t/t = a_o (t_P/t)^a, where a_o and a are positive constants ~1; likewise distances are subject to an ultimate uncertainty c \sigma_t, where c is the speed of light. As a consequence, the period and wavelength of light cannot be specified precisely; rather, they are independently subject to the same intrinsic limitations in our knowledge of time and space, so that even the most monochromatic plane wave must in reality be a superposition of waves with varying omega and {\bf k}, each having a different phase velcocity omega/k. For the entire accessible range of the electromagnetic spectrum this effect is extremely small, but can cumulatively lead to a complete loss of phase information if the emitted radiation propagated a sufficiently large distance. Since, at optical frequencies, the phase coherence of light from a distant point source is a necessary condition for the presence of diffraction patterns when the source is viewed through a telescope, such observations offer by far the most sensitive and uncontroversial test. We show that the HST detection of Airy rings from the active galaxy PKS1413+135, located at a distance of 1.2 Gpc, secures the exclusion of all first order (a=1) quantum gravity fluctuations with an amplitude a_o > 0.003. The same result may be used to deduce that the speed of light in vacuo is exact to a few parts in 10^32.

We present a method of directly testing whether time is `grainy' on scales of <= t_P = (hbar G/c^5)^0.5 ~ 5.4E-44 s, the Planck time. If the phenomenon exists, the energy and momentum of a photon (quantities which are independently measured) will be subject to ultimate uncertainties of the form dE/E ~ dp/p ~ (E/E_P)^a, where E=h/t_P and a~1, because surpassing these will lead to `super-clocks' and `super-rulers'. A well-known consequence is random perturbation of the photon dispersion equation by a correction term which to lowest order modifies the equation to the form p^2 = E^2[1 +/- a_o(E/E_P)^a], where a_o ~ 1 and c=1. As a result, the two wave velocities will fluctuate by different amounts, so that after propagating a sufficiently large distance the phase of the radiation will no longer be well-defined. Now, at optical frequencies, the technique of stellar interferometry readily ascertains whether light from an astronomical object retains is phase information upon arrival. Currently the furthest star from which interference fringes were seen, S Ser, is at ~ 1 kpc, implying that all models with a <= 13/15 are to be rejected. The decisive step, of course, is to test the presence or not of the first order term E/E_P itself (i.e. a=1); for this one must await the imminent observations of extra-galactic sources at > 1 Mpc by the Very Large Telescope Interferometer (the VLTI). Based upon the S Ser result, nonetheless, it is already possible to conclude that there is no first order departure from exactness in time over intervals ~ 4.2E-40 s (~10^4 times longer than the Planck time).