The influence of three commonly used extraction media on the recovery of endotoxin from different filter types was studied using the Limulus amebocyte lysate (LAL) test. Particulate and aqueous endotoxin preparations were used to spike PVC, Teflon coated glass fiber and glass fiber filters. For each filter endotoxin combination, air was sampled at 2 liter/minute for 50 minutes after which the filters were extracted with either pyrogen-free water (PFW), 0.05% Tween 20 in PFW, or a 0.05 M Potassium phosphate/0.01% triethylamine pyrogen free buffer solution. Sample extracts were assayed by the LAL. To determine the interaction between endotoxin, filter type, and extraction medium without air sampling (interaction controls), each extraction media was spiked with aqueous or particulate associated endotoxin, an unspiked filter added to the media, and the sample extracted and assayed for endotoxin. Positive controls consisted of each extraction medium spiked with the aqueous or particulate endotoxin. For particulate spiked samples, there was no difference in recovery of endotoxin from the air sampled PVC and glass fiber filters extracted with either PFW or Tween 20; however, for these extraction media, recovery efficiency from air sampled Teflon coated glass fiber was significantly less than the other two filter media. No difference was observed in particulate spiked air sample PVC and glass fiber filters and the respective positive controls and interaction controls. There was no difference in the recovery of endotoxin from any of the positive controls, including the phosphate buffer control. Regardless of the filter media, all particulate spiked samples (both air sampled and interaction controls) extracted with the phosphate buffer showed significantly lower recoveries than for the other two extraction media. An inhibition study demonstrated that in undiluted samples the phosphate buffer inhibited the recovery of endotoxin. The inhibition could be eliminated by sample dilution, however, filters extracted with the phosphate buffer, then diluted with PFW yielded lower endotoxin recoveries. These data indicate that the phosphate buffer interacts with either the filter media, the endotoxin, or both to reduce the availability of the endotoxin for the LAL analysis. These data confirm previous studies which have shown that filter type and extraction media are important variables in determining the recovery of endotoxin from environmental samples.