Each subject was fitted with a modified U.K. Royal Air Force Aircrew Respirator Mk 5 (AR5), comprising an appropriately sized oxygen mask (over mouth and nose) with an attached polycarbonate visor and butyl rubber cowl, covering the head, neck, and shoulders. The neck seal was removed for comfort, and the helmet suspension harness was not needed. The mask antisuffocation valve was sealed to exclude light, but the microphone was retained for the same reason. The bulk of the visor was cut away to leave a residual scaffold on which matching polycarbonate scaffolds could be mounted. These supported neutral density (ND) gelatin filters (Wratten; Eastman Kodak, Rochester, NY) of either 1 or 2 optical density units. A port to the mask cavity allowed breathing gas composition to be analyzed continuously with a mass spectrometer (MGA 2000; Airspec, Kent, UK). The AR5 allowed ambient air to be blown gently and independently of the breathing gas supply, across the facial aspect of the ND filters, ensuring that they remained free from misting. Hence, normal corneal oxygenation was preserved, regardless of respiratory condition, and subjects could wear normal corrective spectacles and lenses, as required. Thus, the AR5 enabled total, contemporaneous control of visual and respiratory adaptation state.
Each of the three breathing gases (air, oxygen, and the hypoxic mixture) was supplied to its own dedicated, pressure-demand, breathing gas regulator at a nominal supply pressure. The regulators had identical pressure–flow characteristics and imposed minimal breathing resistance, making them indistinguishable to the user, and their breathing gases were delivered, via a selection tap, to a common mask hose. Mask valves prevented rebreathing of expired gas, and the mass spectrometer traces were monitored to ensure that good face–oxygen mask seals were maintained. Mass spectrometer calibrations were conducted immediately before and after each respiratory condition by using various gas mixtures of known composition, giving a measurement error of less than 1% for physiological partial pressures of oxygen and carbon dioxide. Noninvasive monitoring of blood pressure, heart rate, and oxygen saturation were undertaken with a blood pressure monitor (Finapres 2300; Ohmeda, Englewood, NJ) and a pulse oximeter (7840; Kontron Instruments, Ltd.; Watford, UK) with finger probe. Analog outputs from both devices were calibrated and recorded, together with the mass spectrometer data, using a PC-based data recording and analysis system with software (Powerlab/Chart software; ADInstruments, Castle Hill, NSW, Australia).