The toxicological significance of paraquat concentrations in paraquat poisonings was evaluated by means of multivariate analysis methods. Paraquat could be determined by a newly developed procedure, which involved thin-layer chromatography with flame ionization detector (TLC-FID) and solid-phase extraction with a disposable octadecylsilane cartridge. This new method proved to be simple, rapid and reliable for the analysis of paraquat in our seven cases of suicidal poisoning. The relationship between plasma paraquat concentration (C) and time from ingestion (T) could be best described by the following functions. The regression equation of fatal cases was ln[ln(C X 1000)] = 2.5453 - 0.2114 lnT. The regression equation of survivors was ln[ln(C X 1000)] = 2.1041 - 0.2826 lnT. The discriminant function (D) to separate the fatal and survival cases was D = 1.3114 - 0.1617 lnT - 0.5408 [ln(C X 1000)] (fatal cases: D less than 0, survivors: D greater than 0). The discriminant function was demonstrated to have a high reliability for the toxicological significance in our seven poisoned patients. The significant correlation between plasma paraquat concentration and urine paraquat concentration (C') in our cases was obtained. The regression equation was lnC' = 0.953 lnC + 1.409. This also indicated that urinary concentrations are 3.3 - 4.5 times greater than plasma concentrations. The multiple regression equation among plasma paraquat concentration, time from ingestion, and the ingested volume (V) of Gramoxone (trade name of paraquat), was lnC = 0.009V - 0.232T + 3.612. It is suggested that the determination of paraquat is of great value, and that these data are useful in assessing the severity and predicting the outcome of poisoning for forensic and clinical purposes.