Increasingly there is a need for data on "real" fires, whether it be as an aid for fire engineering assessments or for testing mathematical models. This paper describes four large-scale fire experiments that were performed in a simulated two-bed hospital ward. Both smoldering and small flaming ignition sources were employed in the experiments. The smoldering stages of the fires did not produce a buoyant hot layer. The flaming stages of the fires produced similar fire growth curves irrespective of whether there had been a prolonged smoldering stage. A number of criteria for untenability have been compared.