External flaming from buildings occurs as a result of a large amount of fuel being available in the room of fire origin in comparison to the amount of ventilation provided. The size of the openings in a compartment affects the amount of oxygen available within the fire room, and hence the amount of combustion that can take place inside. Excess fuel that is not burnt within the room flows out of the opening and combusts upon reaching the oxygen in the air outside. It is in this situation that flames are seen projecting out of the window. Flames projecting from openings pose the threat of fire spread from the room of fire origin. This threat increases with the size of the flames. Thus a dependable method for predicting the size of flames projecting from openings is required. This research addresses the issue of predicting flame heights projecting from openings, based on the heat release rate of a fire. The results are based on laboratory experiments and are presented in non-dimensional form, allowing application to scenarios that have not been specifically tested. This work supports the findings of other researchers, with appropriate adjustments made to compensate for differing experimental approaches. This indicates that the relationships established between the non-dimensional heat release rate and flame height are formed from a sound underlying principle. An empirical relationship between the non-dimensional flame height and heat release rate of a fire is presented in a simplified format to enable ease of use. The temperatures attained, and other observations from the compartment fire experiments are also presented and discussed. These were found to have dependence on a number of factors, with relationships varying between the individual experiments. The widely-used computational fluid dynamics model Fire Dynamics Simulator, Version 4 (FDS), was found to currently be unreliable in modelling the experimental scenarios. The results obtained were unrealistic and bore minimal resemblance to the experimental results, with extensive computational simulation time. The ability of the programme to model the compartment fire scenario requires further investigation to determine whether a finer grid resolution may improve results, or whether it is simply not able to model combustion processes involved at this stage.