W hen Factor VIII/von Willebrand factor (FVIII/vWF) protein is rechromatographed on 4% agarose in 0.25 M CaCl₂, the protein and vWF activity appear in the void volume, but most of the FVIII procoagulant activity elutes later. Recent evidence suggests that the delayed FVIII procoagulant activity is a proteolytically modified form of FVIII/vWF protein that filters anomalously from agarose in 0.25 M CaCl₂. To test whether or not thrombin is the protease involved, the effect of 0.25 M CaCl₂ on FVIII/vWF and its reaction with thrombin was examined. About 30% of the FVIII procoagulant activity was lost immediately when solutions of FVIII/vWF protein were made 0.25 M in CaCl₂. When FVIII in 0.15 M NaCl was activated with 0.04 U thrombin/ml and then made 0.25 M in CaCl₂, the procoagulant activity of a broad range of FVIII/vWF protein concentrations remained activated for at least 6 h. But, in 0.25 M CaCl₂, the increase in FVIII procoagulant activity in response to thrombin was much more gradual and once activated, the procoagulant activity was stabilized by 0.25 M CaCl₂. When thrombin-activated FVIII/vWF protein was filtered on 4% agarose in 0.15 M NaCl, there was considerable inactivation of FVIII procoagulant activity; however, the procoagulant activity that did remain eluted in the void volume. In contrast, when thrombin-activated FVIII/vWF protein was filtered in 0.25 M CaCl₂, the FVIII procoagulant activity eluted well after the void volume and remained activated for 6 h. The procoagulant peak isolated by filtering nonthrombin-activated FVIII/vWF protein on agarose in 0.25 M CaCl₂ was compared to that isolated from thrombin-activated FVIII/vWF protein. Both procoagulant activity peak proteins had about the same specific vWF activity as the corresponding void volume protein. Before reduction, the sodium dodecyl sulfate gel patterns for the two procoagulant activity peak proteins were the same. After reduction, the gel pattern for the nonthrombin-activated procoagulant activity peak protein contained bands of 195,000, 148,000-120,000, 79,000, 61,000, 51,000, and 18,000 daltons whereas the pattern for the reduced thrombin-activated procoagulant activity peak protein always lacked the higher molecular weight bands, but consistently showed the four lower molecular weight bands to be well resolved. Taken together, these results imply that thrombin generates the FVIII procoagulant activity that is stabilized by 0.25 M CaCl₂ and elutes aberrantly from 4% agarose in that solvent.