There is much evidence that nucleation of liquid droplets from vapour as well as nucleation of crystals from both solution and vapour occurs preferentially in surface defects such as pits and grooves. In the case of nucleation of solid from liquid (freezing) the situation is much less clear-cut. We have therefore carried out a study of the freezing of 50 µm diameter water drops on silicon, glass and mica substrates, and made quantitative comparisons for smooth substrates and those roughened by scratching with three diamond powders of different size distributions. In all cases, freezing occurred close to the expected homogeneous freezing temperature, and the nucleation rates were within the range of literature data. Surface roughening had no experimentally significant effect on any of the substrates studied. In particular, surface roughening of mica -which has been shown to cause dramatic differences in crystal nucleation from organic vapours -has an insignificant effect on ice nucleation from supercooled water. The results also show that glass, silicon and mica have at best only a marginal ice-nucleating capability which does not differ appreciably between the substrates.The lack of effect of roughness on freezing can be rationalised in terms of the relative magnitudes of interfacial free energies and the lack of a viable two-step mechanism, which allows vapour nucleation to proceed via a liquid intermediate.