This paper provides a resolution to the contradictory accounts of whether or not the Darrieus turbine can self-start. The paper builds on previous work proposing an analogy between the aerofoil in Darrieus motion and flapping-wing flow mechanisms. This analogy suggests that unsteadiness could be exploited to generate additional thrust and that this unsteady thrust generation is governed by rotor geometry. Rotors which do not exploit this unsteadiness will not self-start. To confirm the hypothesis, unsteady effects were measured and then incorporated into a time-stepping rotor analysis and compared to experimental data for self starting wind turbines. When unsteady effects were included the model was able to predict the correct starting behaviour. The fundamental physics of starting were also studied and parameters that govern the generation of unsteady thrust were explored: namely chord-to-diameter and blade aspect ratios. Further simulation showed that the Darrieus rotor is prone to be locked in a deadband where the thrust is not continuous around a blade rotation. This discrete thrust is caused by the large variation of incidence angle during start-up making the Darrieus blade ineffective during part of the rotation. The results show that unsteady thrust can be promoted through an appropriate selection of blade aspect and chord-to-diameter ratios, therefore self-starting rotors may be designed. A new definition of self-starting is also proposed.