A method is described for estimating, in man, the degree of deposition of fine particles at various levels of the lungs. It avoids many of the artifacts observed when dealing with animals. Submicronic particles are selected by scrubbing the original cloud through successive, turbulent, liquid barriers, and are then dispersed for 1–2 min into a 0.5 m3 chamber from where well-trained subjects inhale the aerosol for 2 min. During the second minute of breathing, the expired air is collected into a graduated spirometer. At the end of the second minute, a forced expiration is collected in a long tube; the last part of this maximal expiration represents the Haldane-Priestley alveolar air. Using Walkenhorst thermal precipitators with electron microscope screens, the aerosol is sampled simultaneously in the dispersion chamber for inspired air, in the spirometer for mixed expired air, in the last portion of last expiration for end of tidal air and in the last portion of the forced maximal expiration for alveolar air. In this way, the air collected from different levels of the lungs can be compared with the inspired air in order to determine, by particle counting and sizing, the deposition characteristics of the inhaled particles. It was observed that the deposition (in per cent of the number of particles per ml of inspired air) in the mixed expired air and in the last portion of the tidal volume was, as classically known since C. BROWN, inversely proportional to the breathing frequency. Conversely, regardless of the breathing pattern, the per cent deposition was found to average more than 90 per cent in the deep alveolar air where the air exchanges are governed by diffusion. It is only when voluntarily over-ventilating the alveolar spaces that the degree of deposition in the deep parts of the lungs can be reduced. Although, for every breathing rate, the end tidal air is characterized by a higher deposition rate than that of mixed expired air, it cannot be identified with deep alveolar air. When made exclusively of microsized particles, aerosols from hygroscopic salt solutions offer the same deposition curves as insoluble particles. The mean size of the particles collected in the successive portions of the expired air was observed to decrease regularly towards the deep alveolar air. The larger the original granulometric spectrum of the inhaled particles, the greater the difference in the mean size and in the size distribution curves between inspired air, mixed expired air, end tidal air and alveolar air. However, the size distribution curves of the 0.5-micro maximum size particles do not exhibit appreciable changes from inspired air toward alveolar air, suggesting that the particles do not settle before reaching the deep alveolar spaces. Under electronmicroscopic examination the size most frequently observed for particles sampled from deep alveolar air lies between 0.05 and 0.4 micro; their maximum size has never been found to exceed 1 micro. None of these findings appears to be influenced by the number of particles present in the inspired air (from 4000 to 900,000 per ml).