Why is there so little dust in globular clusters? Using ISO to observe raster maps of the central 10x10 arcmin of four clusters, with a detection limit of 1E-4 solar masses of dust, will tell us what the typical level of dust in clusters is, and by looking at clusters of differing masses and metallicities and at different heights away from the Galactic plane, we may be able to discriminate between different stripping mechanisms. The interstellar medium in a globular cluster might be expected to contain about 0.1 solar masses of dust. In the 1E7 years since a typical cluster last crossed the galactic plane (when all the gas and dust would be tidally and ram stripped out) enough stars in a cluster will have gone through the post-AGB stage to inject this amount of dust into the cluster. However dust at this level has not yet been detected in any cluster. As the dust should be at about 50K, observations at 50 to 100 microns are needed. (Observations at other wavelengths only give much higher upper limits.) Pointed IRAS observations of 47 Tuc found 4E-4 solar masses and the IRAS surveys only found higher upper limits for all the other clusters. Some mechanism is removing dust from clusters, on a timescale of about 1E4 - 1E5 years. Possible mechanisms include; (i) ram-pressure stripping by halo gas, (ii) removal by a cluster wind, (iii) ejection by the integrated radiation pressure of Galactic stars, (iv) erosion of grains by the hot x-ray emitting Galactic halo gas. With only the 47 Tuc detection, it is impossible to say either if its 3E-4 solar masses of dust is a typical level, or to discriminate between the different proposed stripping mechanisms.