A modified torsiometer (Gervais and Vermeire 1983) was used to determine the coagulation kinetics of raw milk and of milk heated for 14.7 s at temperatures ranging from 105 to 125°C with a direct U.H.T. unit. Two coagulation kinetics models (Scott‐Blair and Burnett 1963; Douillard 1973) were equally satisfactory. From the strict point of view of mathematical adjustment, the two models can be taken as equivalent. In all cases, the residual standard deviation was equal to experimental error.The Scott‐Blair equation is not suitable since it may give a negative gel time for some heated milks. The authors demonstrate that this is a mathematical artifact, arising from the form of this equation. The Douillard model is best adapted for determining the time at which curd rigidity will reach a given value. The results also show that heating milk modifies its coagulation rate constant, rather than gelling time and maximum rigi