The origin of spurious and irreproducible signals in the measurement of the hyper-Rayleigh scattering (HRS) of solutions is investigated. A new cell design, coupled with continuous solution filtration through a 20 nm filter, is shown to improve the reproducibility of the measurement to +/-5%. A transverse cell dimension of similar to 2 mm reduces absorption of the second harmonic. A simple monochromaticity test is used to identify luminescence, and we recommend that it is made routinely. Scattering from the [Ru(bipy)(3)](2+) cation, attributed to HRS by Zyss et al. [Chem. Phys. Lett. 206, 409 (1993)], appears to be largely due to two-photon excited luminescence. The reliability of calibration procedures is examined, and some revised hyperpolarizabilities are reported. (C) 1996 American Institute of Physics.