It has been highlighted that shame may be an important dynamic risk factor for prevention of violence and recidivism in forensic populations. However, past research investigating the relationship between shame and violence, or recidivism has been inconsistent. Different conceptualisations and measurements of shame used in the literature may explain these inconsistencies. Therefore, a systematic review was conducted to explore how shame was conceptualised in forensic populations and these measures were then evaluated. Findings revealed that most studies did not clearly define shame, and when they did, the same theoretical underpinnings were used in different ways. By assessing the validity and reliability of shame measures, it was revealed that different measures focused on different aspects of shame. This could explain the current confusion in the conceptualisation and measurement of shame in forensic populations, and shed light on inconsistent findings between shame and other constructs. Shame in violent female offenders is an unexplored phenomenon and therefore may involve various complex and unexpected factors. A social constructivist grounded theory approach was applied to the narratives of eight violent female offenders, focusing on thoughts, feelings and life experiences in relation to shame and violence. A model was constructed suggesting that childhood victimisation, in the absence of available, compassionate, secure relationships, may lead to difficulties with emotion regulation. The experience of negative emotions, including shame, may lead to self-harm, substance misuse and violence. It was however demonstrated that this vicious cycle could be broken through the development of secure, positive and compassionate relationships. These findings suggest that shame and attachment may be important factors for treatment and service planning, to meet the unique needs of female offenders.