Optimal stopping problems require people to choose from a sequence of values, under the constraint that they cannot return to an earlier option once it is rejected. We study how people solve optimal stopping problems when the distribution of values they must choose from is not uniform, but is constructed to contain many high values or many low values. We present empirical evidence that people adapt to both sorts of environments, and make decisions consistent with using threshold-based models. We then fit a threshold model to our data, inferring the threshold people use, and finding they usually decrease their thresholds faster than is optimal as the sequence progresses. We also present empirical and modelbased evidence that people generally do not adjust their thresholds on the basis of the values they see.