Automatic fire sprinklers use a heat sensitive element such as a glass bulb or fusible link to respond to the heat from a fire. The response of commercial fire sprinkler glass bulbs has been extensively characterised in convection-dominated dry gas flows but in real fires there may be more factors that influence the heat transfer to the bulbs such as radiation from the fire or cooling from adjacent sprinkler sprays. The time of activation is the only indication of the thermal response of typical commercial fire sprinklers using glass bulbs to a fire, but direct temperature measurement using a modified proxy may provide a better understanding of how sprinklers respond in a complex environment. Modified glass bulbs have been created that allow a thermocouple to be inserted in the bulb for direct temperature measurement. In this paper, the thermal response of sprinklers with these modified bulbs has been observed in hot-air wind tunnel plunge experiments and full scale room fire experiments. At the time of activation the measured temperature of the modified sprinklers was found to be higher than the nominal activation temperature specification for the unmodified sprinklers. For the compartment fires, a thermal response model generally predicted longer sprinkler activation times based on ceiling jet temperature and velocity measurements than was observed experimentally.