This paper examines the influence of residual pressures in the range from 10-~to 5 torr on the course of thermal analysis. With the help of examples concerning in particular the thermolysis of gibbsite, AI(OH)3 , it is shown that a) the control of residual pressure is of virtually no use unless the rate of decomposition is also controlled (otherwise, the TG curves represent a composite phenomenon, which is practically unintelligible); b) the influence of residual pressure may be unexpectedly high both on the shape of the TG curves (and therefore on the apparent kinetic parameters) and on the nature (porosity, structure) of the products.The influence of the gas atmosphere on a thermal decomposition of the type solid A ~ solid B + gas is widely accepted. In 1951, Rowland and Lewis [1] in this way explained the usual DTA trace for calcite thermolysis, where the onset of decomposition may be observed as soon as 650 ~ (low partial pressure of CO~) and where "the reaction progressively becomes more vigourous until a peak is reached at about 925 ~ the temperature at which the reaction would begin if the furnace had been filled with COz at one atmosphere pressure. In 1960, Garn suggested simplification of the situation by operating under self-generated atmospheres [2], and several devices were proposed and used in order to control the overall pressure, e.g. the "cold sink", especially convenient in the case of water evolution [3], or the discontinuous and automatic gas extractor [4]. Nevertheless, "vacuum" was often considered as one pressure at which it may be worthwhile to carry out one experiment in order to complement a set of data obtained under higher pressures. Indeed, various authors, such as Eyraud et al. at the time of the first vacuum TG experiments, pointed out the "simplifying" part played by vacuum-operation, which was supposed to eliminate disturbing phenomena such as adsorption or diffusion processes [5] or secondary reactions [6] and to give rise to the simplest mechanism with higher decomposition rate [7]. Now pressures lower than I torr, which are usually referred to as "vacuum", actually cover a wide experimental range, down to 10 -~ torr, whereas the range of medium and *