We report results from an experiment based on a simplified version of Hart's (1995) property rights theory of the firm. Only one manager invests and this investment is completely specific. In that case the theory predicts that the level of investment is not affected by the level of no-trade payoffs if these payoffs are modeled as threat points (as Hart does). If instead no-trade payoffs are modeled as outside options (as in De Meza and Lockwood, 1998), subgame perfectness predicts that investments are reduced when the no-trade payoff increases from a non-binding (low) level to a binding (high) level. Our findings contradict these predictions. With threat points, investment levels increase (rather than remain constant) when the investor's no-trade payoff goes up. With outside options, investment levels remain constant (instead of decrease) when the value of the no-trade payoff increases. In all cases considered, the average investment level exceeds the predicted level. Actual investment behavior is consistent with the outcomes of the bargaining stage; investors typically receive a higher return on their investment than predicted. Given this, actual investment behavior is close to optimal.