Snow avalanche case reports have documented the survival of skiers apparently without permanent hypoxic sequelae, after prolonged complete burial despite there being only a small air pocket on extrication. We investigated the underlying pathophysiological changes in a prospective, randomised 2 x 2 crossover study in 12 volunteers (28 tests) breathing into an artificial air pocket (1- or 2-l volume) in snow. Peripheral SpO(2), ETCO(2), arterialised capillary blood variables, air pocket O(2) and CO(2), snow density, and snow conditions at the inner surface of the air pocket were determined. SpO(2) decreased from a median of 99% (93-100%) to 88% (71-94%; P<0.001) within 4 min of breathing into the air pocket; the reduction was greater at 1 l, than 2 l, volume air pocket (P=0.013, intention to treat P=0.003) and correlated to snow density (r=0.50, P=0.021, partial correlation coefficient). ETCO(2) rose simultaneously from median 5.07 kPa (3.47-6.93 kPa) to 6.8 kPa (5.87-8.27 kPa; P<0.001), with consequent respiratory acidosis. Despite premature interruption due to hypoxia (SpO(2)</=75%) in 17 of 28 tests (61%), a respiratory steady state prevailed in five tests until protocol completion (30 min). We conclude that the degree of hypoxia following avalanche burial is dependent on air pocket volume, snow density and unknown individual personal characteristics, yet long-term survival is possible with only a small air pocket. Hence, the definition of an air pocket, "any space surrounding mouth and nose with the proviso of free air passages" is validated as the main criterion for triage and management of avalanche victims. Our experimental model will facilitate evaluating the interrelation between volume and inner surface area of an air pocket for survival of avalanche victims, whilst the present findings have laid the basis for future investigation of possible interactions between hypoxia, hypercapnia, and hypothermia (triple H syndrome) in snow burial.