This experiment observed the evolution of metabolite plumes from a human trapped in a simulation of a collapsed building. Ten participants took it in turns over five days to lie in a simulation of a collapsed building and eight of them completed the 6 h protocol while their breath, sweat and skin metabolites were passed through a simulation of a collapsed glass-clad reinforced-concrete building. Safety, welfare and environmental parameters were monitored continuously, and active adsorbent sampling for thermal desorption GC-MS, on-line and embedded CO, CO(2) and O(2) monitoring, aspirating ion mobility spectrometry with integrated semiconductor gas sensors, direct injection GC-ion mobility spectrometry, active sampling thermal desorption GC-differential mobility spectrometry and a prototype remote early detection system for survivor location were used to monitor the evolution of the metabolite plumes that were generated. Oxygen levels within the void simulator were allowed to fall no lower than 19.1% (v). Concurrent levels of carbon dioxide built up to an average level of 1.6% (v) in the breathing zone of the participants. Temperature, humidity, carbon dioxide levels and the physiological measurements were consistent with a reproducible methodology that enabled the metabolite plumes to be sampled and characterized from the different parts of the experiment. Welfare and safety data were satisfactory with pulse rates, blood pressures and oxygenation, all within levels consistent with healthy adults. Up to 12 in-test welfare assessments per participant and a six-week follow-up Stanford Acute Stress Response Questionnaire indicated that the researchers and participants did not experience any adverse effects from their involvement in the study. Preliminary observations confirmed that CO(2), NH(3) and acetone were effective markers for trapped humans, although interactions with water absorbed in building debris needed further study. An unexpected observation from the NH(3) channel was the suppression of NH(3) during those periods when the participants slept, and this will be the subject of further study, as will be the detailed analysis of the casualty detection data obtained from the seven instruments used.