A major cause of death or permanent injury in fires is inhalation of toxic gases. Moreover, every fire is unique, and the range of products, highly dependant on fire conditions, produces a wide variety of toxic and irritant species responsible for the most fire fatalities. Therefore, to fully understand each contribution to the toxicity it is necessary to quantify the decomposition products of the material under the test. Fires can be divided into a number of stages from smouldering combustion to early wellventilated flaming, through to fully developed under-ventilated flaming. These stages can be replicated by certain bench-scale physical fire models using different fuel-to-oxygen ratios, controlled by the primary air flow, and expressed in terms of the equivalence ratio (the actual fuel/air ratio divided by the stoichiometric fuel/air ratio). This work presents combustion product yields generated using a small-scale fire model. The Purser Furnace apparatus (BS7990 and ISO TS 19700) enables different fire stages to be created. Identification and quantification of combustion gases and particularly their toxic components from different fire scenarios was undertaken by continuous Fourier Transform Infrared Spectroscopy. The relationship between type of the fire particularly the temperature and ventilation conditions, and the toxic product yields for four bulk polymers, LDPE, PS, Nylon 6.6 and PVC are reported. For all the polymers tested, except PVC, there is a dramatic increase in the yield of products of incomplete combustion (CO and hydrocarbons) with increase in equivalence ratio, as might be expected. For PVC there is a consistently high level of products of incomplete combustion arising both from flame inhibition by HCl, and oxygen depletion. There is a low sensitivity to furnace temperature over the range 650-850°C, except that at 650°C PS shows an unexpectedly high yield of CO under well-ventilated conditions and PVC shows a slightly higher hydrocarbon yield. This demonstrates the dependence of toxic product yields on the equivalence ratio, and the lack of dependence on furnace temperature, within this range.