Earlier views indicated that globulin (corticosteroid-binding globulin (CBG) or sex hormone-binding globulin (SBG)) but not albumin binding in plasma, protects steroids from splanchnic metabolism in man. Also, the splanchnic extraction (HE) of a steroid seemed to be highly dependent on the rate of disassociation of the steroid–protein complex. However, the faster rate of disassociation (τ½ = 0·9 s) of cortisol–CBG, as determined by later accurate fluorescence methods, intuitively meant that this complex must disassociate completely in a single 9 s passage through the liver. The low HE of total cortisol was then a puzzling anomaly. Using a differential equation solver (TUTSIM) and a model with unbound, albumin- and globulin-bound pools of steroid (with metabolism of unbound and also possibly albumin-bound steroid), the mechanism of splanchnic metabolism has been studied. The 'complex', probably most realistic, model includes 13 steroids, which can simultaneously bind to plasma albumin, CBG and SBG. The steroid concentration and numbers of occupied binding sites of the globulins decrease during the time of metabolism. The experimental data used are the in-vitro binding characteristics of the steroid–protein complexes, including the equilibrium constants and rates of disassociation and the in-vivo HE of nine steroids, usually measured by direct analysis of hepatic venous blood. However, the HE of cortisol had to be calculated from the metabolic clearance rate/splanchnic blood flow, giving a maximum value of 12%. The fractional metabolic rate of unbound steroid is generally represented by e. A certain value of e (RE) is required to give a remaining steroid concentration after 9 s of metabolism, which is made equal to (1–HE) in the model to simulate splanchnic extraction. If the fractional rate of metabolism of albumin-bound steroid is h (f = h/e), then RE will depend on the value of f. The maximum RE for cortisol is RE0 = 0·42 and RE1 = 0·16 for f = 0 and 1 respectively. For either value of RE, there will be the appreciable reassociation of cortisol to CBG after disassociation of the cortisol–CBG complex. With such reassociation, the total cortisol remaining after 9 s metabolism is fairly independent of the rate of disassociation of the cortisol–CBG complex. This explains the low total HE of cortisol in spite of the high rate of disassociation of cortisol–CBG. Generally, for all nine steroids studied, HE estimations in vivo in humans indicate that the steroid–globulin disassociation rate will only be markedly rate-limiting for dihydrotestosterone–SBG and, to a lesser extent, testosterone–SBG. These results are examples of the principle that the divergence in hormone concentration with the disassociation rate of the hormone-protein complex after metabolism depends on the value of e. The RE0 and RE1 values of the nine steroids: aldosterone, progesterone, testosterone, dihydrotestosterone, androstenedione, androstanediol, oestradiol, oestrone and cortisol have also been calculated from all in-vivo HE estimates. If it is assumed that the steroid in the albumin-bound pool is not metabolized (f = 0), then the RE0 values of the nine steroids considered have a large variation (s.d./mean = 115%). There is a correlation between the RE0 values and their albumin-binding index (BIA). However, if f = 1 then the variation of RE is smaller (s.d./mean = 36%) and there is no correlation between RE1 and BIA. The ratio of the RE0 and RE1 values for a particular steroid is nearly equal to (1 + BIA), which is determined solely by the dynamics of the situation. However, the appropriate value of f depends on the mechanisms involved. With the F1 hypothesis (f = 1), albumin binding does not protect steroids from splanchnic extraction because of metabolism of albumin-bound steroid. With the FO hypothesis (f = 0), this seeming lack of protection of albumin binding is essentially artefactual. Albumin-bound steroid would not be metabolized directly but the fractional rate of metabolism of unbound steroid would be increased with the decreasing polarity of the steroid (i.e. with increasing BIA). With the available data, the choice between the hypotheses can only be on an intuitive basis. Journal of Endocrinology (1991) 131, 339–357