Captive welfare is of great and ever increasing concern, but assessing welfare has largely relied on behavioural and physiological measures. Attention bias is the predisposition to attend to particular stimuli over others and is known to be influenced by emotional state, making it an important indicator of psychological well-being. For example, humans have a bias to attend to threatening facial expressions over positive ones, but this is further pronounced in anxious individuals until a threshold is reached and extreme avoidance occurs. Rhesus macaques (Macaca mulatta) are used as research models in psychopathology, and therefore understanding how they focus on or avoid particular stimuli is important not only in research for human conditions, but also in understanding their psychological well-being within the captive environment. We tested 35 captive-bred, socially-housed female macaques on their attention to two simultaneously-presented images of facial expressions (one aggressive, one neutral), of unknown male conspecifics before and after their yearly health check. As with humans, female rhesus macaques were found to also have a bias to attend to threatening facial expressions over neutral ones. This bias was significantly affected by the testing condition, as well as the social position and breeding status of the individual macaque, and the structure of the social group. While the main focus of the attention bias measure was eye gaze, we also observed behavioural reactions to the threatening stimulus, including strong appeasement facial expressions and exiting the testing area. Our results indicate that macaques do have an attention bias for threatening social stimuli, and stressful events may alter this bias. The degree to which an individual animal is influenced by stressors, resulting in increased vigilance or avoidance, is a measure of psychological well-being and can be tested using attention bias.