The Experiences in Close Relationships-Revised (ECR-R) scale, and its predecessor, the Experiences in Close Relationships (ECR) scale have been argued to be the most robust measures of adult attachment to date. Despite this, to date, there has been little research which has thoroughly examined the psychometric properties of the ECR-R. The studies presented in this thesis attempted to address this gap in the literature. When examinations have taken place, authors propose that ECR-R contains a 2-factor model of attachment with the factors representing anxious and avoidant attachment. However, none have considered a bifactor model of attachment which, arguably, would better reflect current theories of attachment. A bifactor model would include an overarching factor of attachment security, together with additional latent variables representing anxious and avoidant attachment. Furthermore, in addition to several other limitations in the previous literature, unnecessary data manipulation has often been carried out which may have undermined findings. The overarching aim of this thesis was threefold: first, to investigate the factor structure of the ECR-R including the exploration of a bifactor model of attachment. Second, to explore the various short versions of the ECR-R including their factor structure, model fit, and differential predictive validity. Finally, to explore the possibility of a new, superior, version of the ECR-R. To examine this (N = 914) participants completed an online survey which included the ECR-R together with measures of relationship satisfaction, perceived partner compatibility, self-esteem, and rejection sensitivity, and were then asked to complete the survey again three months later. In the first study, the ECR-R was found to have good concurrent /convergent validity when compared to the Relationships Questionnaire (RQ) and it was found to reflect a bifactor model of attachment. In the second study, the ECR-R-18 was found to be superior to the full ECR-R in terms of model fit but the model fit obtained could be improved. In the final study, an Exploratory Factor Analysis (EFA) suggested that a bifactor model of attachment with four latent factors (anxious attachment: model of the self, anxious attachment: model of others, avoidant attachment: model of self, and avoidant attachment, model of others) was a best fit to the data. This was a novel finding as these latent factors had not been proposed in previous literature regarding the ECR-R. However, the following Confirmatory Factor Analysis (CFA) suggested that this was not the best fitting model to the data and that a bifactor model with two latent factors (anxious and avoidant attachment) would provide a better fit. As a result, a new, arguably superior, version of the ECR- R is proposed; the ECR-R-28. These findings have important implications for evaluating previous research which has used the ECR-R. The practical implications of these and recommendations for future research are discussed.