The reliability of the Limulus assay for determining endotoxin levels in settled house dust was examined. Settled dust samples were collected from several rooms in a number of houses using a commercial vacuum cleaner modified to collect samples in 19 by 90 millimeter cellulose extraction thimbles. Samples were desiccated at 4 degrees-C, divided into eight 100 milligram (mg) aliquots, baked 30 minutes at 270 degrees, and placed in a phosphate/triethylamine buffer, pH 7.5. Sufficient 5% saponin or sodium-dodecyl-sulfate (SDS) was added to make a final detergent concentration of 0.1% or 1%. The samples were extracted by sonication for 60 minutes and analyzed for endotoxin by the kinetic Limulus assay with resistant/parallel line estimation and utilizing dilution factors of 1:2 to 1:2,592. Other extracts or untreated dust samples were stored for periods up to 7 months at 4 or -20 degrees before analysis to assess storage stability. Interferences were seen in samples extracted with buffer only that appeared to depend primarily on the dilution factor. For example, at a dilution of 1:2, the estimated endotoxin concentration was 0.30 endotoxin units per milligram (EU/mg). At dilution factors of 1:12, 1:72, 1:432, and 1:2,592, the apparent estimated endotoxin concentrations were 1400, 330, 38, and 30EU/mg, respectively. If the first three dilutions were excluded, the extraction curve was parallel to the standard curve, producing an estimated endotoxin concentration of 41EU/mg. This indicated that if a single dilution had been used in the assay, interferences would have produced a 13.6 fold underestimate or a 34 fold overestimate of the endotoxin concentration depending on which dilution was used. Endotoxin activity was stable for up to 8 to 10 weeks (wk) in untreated dust samples stored at either 4 or -20 degrees-C. Endotoxin activity in extracts stored at 4 or -20 degrees-C, however, decreased by 65% and 86% respectively after 4 to 6wk. Adding SDS or saponin to the buffer before sonication reduced the amount of endotoxin measured. Seven replicate determinations of a control standard using resistant/parallel line estimation yielded within and between run relative coefficients of variation of 3.9% and 7.5%, respectively. The authors conclude that endotoxin is present in dust in homes and other nonindustrial environments, but measuring it accurately presents technical problems. Dust samples frequently contain substances that interfere with the Limulus assay. Using a resistant/parallel line technique and correcting for dilution effects can avoid these problems.