This paper reports the results from one experiment and one quasi-experiment used to investigate the potential problem of "hypothetical bias" in surveys involving an individuals valuation of time. The experiment compares hypothetical and real choices regarding an offer to participate in a survey in exchange for money. The quasi-experiment compares hypothetical and real choices regarding two bus journeys, one fast and expensive and the other slow and cheap. In both of these experiments, real choices differ significantly from hypothetical ones. The paper estimates parametric distributions of the value of time by applying the general method of moments (GMM) estimator. Since the samples are relatively small a parametric bootstrap is used to obtain asymptotic refinement of statistical tests. The results in the experiment as well as in the quasi-experiment suggest a value of time which is higher when the choice is for real than when the choice is hypothetical. Assuming that the value of time distribution is exponential, real choices produce a mean value of time twice as large as the corresponding hypothetical value.