The disinvestment decision is of crucial importance in many contexts: if funds are tied up for too long in an increasingly poorly-performing project, then opportunities for re-investment may be missed. Optimal disinvestment theory is an important component of real options theory; in some senses it is simply the converse of optimal investment theory, but it seems to have been relatively ignored by experimentalists. Two important papers that do consider it are those of Sandri et al (2010) and Musshoff et al (2013), who conclude that decision-makers stay in projects longer than that prescribed by the optimal behaviour of a risk-neutral agent. They explain this departure through risk-aversion, but without a formal hypothesis under test. We report here on a similar experiment, but we explain the behaviour of the subjects with a formal model and with the estimation of the level of risk-aversion. In addition we explore an alternative hypothesis to explain observed behaviour – namely that subjects are unable to fully backwardly induct and instead work with a shorter, and rolling, horizon, rather than the correct one. Our results show that few subjects appear to be risk-neutral, many seem to be very risk-averse and very few are myopic.