This paper investigates the physical environmental (indoor and outdoor temperature and relative humidity, wind speed, solar radiation and rainfall) and contextual drivers (time of day) affecting occupants' manual space heating override behaviour during the heating season based on measurements collected in ten UK dwellings. Logistic regression modelling is used to understand the probability of occupants manually overriding their scheduled heating periods. To the authors' knowledge, these are the first stochastic models of manual heating override behaviour developed for residential buildings. The work reported in this paper suggests that occupants' manual overrides are influenced by indoor and outdoor temperature, indoor relative humidity, and solar radiation. In addition, the effects of the physical environmental variables varied in relation to the time of day. At night, none of the physical environmental variables influenced manual overrides. In the morning, afternoon and evening, manual overrides were governed by a mix of indoor air temperature, indoor relative humidity and solar radiation. The models presented can be used in building performance simulation applications to improve the inputs for space heating behaviour in residential buildings and thus the predictions of energy use and indoor environmental conditions. (C) 2018 Elsevier B.V. All rights reserved.