Vertical axis wind turbines boast several advantages over their horizontal counterparts. Their development and large-scale installation has been limited due to their inherent aerodynamic complexity and in particular due to the occurrence of dynamic stall. Vortices associated with dynamic stall induce aerodynamic load transients on the wind turbine blades that can lead to structural fatigue and premature mechanical failure. Despite these inherent risks, vertical axis wind turbines are often operated under kinematic conditions that lead to dynamic stall because this maximizes their power output. The ability to quantify the trade-off between these undesirable load transients and the benefit of increased power generation is of utmost importance for future optimization studies. Here, we introduce a novel experimental setup to study the aerodynamic performance of a blade in a model-scale H-type Darrieus vertical axis wind turbine and present the results of direct blade load measurements using strain gauges for various tip speed ratios. We compare the measured blade load responses with predictions based on Greenberg's potential flow theory for different tip-speed ratios. The magnitude of the load responses and their fluctuations increase with decreasing tip-speed ratio in both the upwind and the downwind halves of the rotation. In the upwind part of the cycle, flow stall is more severe than in the downwind half across all tip-speed ratios. The relative stall magnitude decreases exponentially with increasing tip-speed ratio. The measured blade loads deviate more strongly from the predictions by Greenberg's analytical potential flow model in the downwind half of the cycle than in the upwind half.