Decades of turnover research have identified sets of factors that lead to voluntary employee separation. However, the predictive power of existing turnover models is extremely limited, and does not go far beyond 30%. In contrast, marital research has identified a method for predicting dissolution which has an accuracy of over 90%, based on a couples’ reflections on their past. This paper presents preliminary findings from a pilot study using this method. The study identifies some indicators of distressed and non-distressed marriage, that are transferrable to employment context, and indicators that were only identified in organisational context. The paper concludes with an expectation that this method can extend methods of enquiry used in turnover research.