The results of the first interlaboratory comparison, carried out in 1991-92 in the framework of the European collaborative Action "Indoor Air Quality and Its Impact on Man", showed unacceptable interlaboratory discrepancies in the case of a thin layer fast decreasing source. A second interlaboratory comparison was subsequently organized and, in order to improve the agreement, the design included: (1) control of the chamber air velocity; (2) control of the source layer thickness; (3) adoption of both dilution and sink mathematical models. The concentrations of 2-(2-butoxyethoxy)-ethanol and of the two Texanol isomers emitted from a water-based paint had to be determined over 13 days to derive the initial emission factor. The results of the 18 participating laboratories from 10 countries can be summarized as follows. The preparation of the paint sample contributed markedly to the variance, because of differences in the paint film thickness. The impact of the chamber itself on the results, if any, was less evident: in fact, a satisfactory agreement of the results has been obtained with chambers of widely different features (capacity range 35 cm³ to 1.5 m³). Despite the use of calibration solutions prepared from the same batch of pure compounds, the analysis of the compound concentrations contributed markedly to the variance, as confirmed by the results of an analytical comparison carried out in parallel with the main comparison. Model fitting has produced a reasonably good description of the data sets, apparently accounting also for the sink due to chamber wall adsorption. The maximum ranges of the estimated emission factors, expressed as a ratio of the highest to the smallest reported value, are 52, 9 and 9 respectively for 2-(2-butoxyethoxy)-ethanol, Texanol-1 and Texanol-2; however, the scattering is markedly reduced if only results obtained with the same GC detector (FID) and from paint samples within a narrow thickness range (50.8-70.6 pm) are considered.