Fire statistics from the UK and the USA attribute 60% to 70% of fire fatalities in dwellings to the inhalation of fire toxic smoke. The objective of this project was to provide more toxic yield data from typical compartment fires and in the process develop a methodology for faster generation of such data on bench scale apparatus. The models for overall toxicity assessment (for irritants and asphyxiants) were reviewed and the reported threshold limits for typical smoke toxicants, were collected, categorised and compared for increasing levels of harm. An extensive database was created of yields of toxic species from different materials and under different fire conditions. This highlighted the need for more yield data for under-ventilated fires in compartments. Eight full scale tests were carried out in a room enclosure with ventilation through a corridor to a front access door. Fire loads were wood pallets, cotton linen and towels, typical living room furniture and diesel. The fires were allowed to become fully developed before extinguishment by the local FRS team. Toxic concentrations were monitored in the hot layer and the corridor (through a heated sampling line) using a heated FTIR analyser, calibrated for 65 species. An emissions based model, developed as part of this work, was used to quantify the equivalence ratio and also the toxic species yields, even for the cases where the fuel mass loss rate was unknown. An important finding was the overwhelming contribution of Acrolein and Formaldehyde in most tests, in exceeding the impairment of escape threshold. The modified controlled atmosphere cone calorimeter showed comparable results to the full scale tests for lean burning combustion however it proved difficult at this stage to produce combustion in the rich burning regime and further development of the methodology is needed.