In 2003, the U.S. Geological Survey, in cooperation with Delta County, initiated a study to characterize streamflow gainloss in a reach of Terror Creek, in the vicinity of a mine-permit area planned for future coal mining. This report describes the methods of the study and includes results from a comparison of two sets of streamflow measurements using tracer techniques following the constant-rate injection method. Two measurement sets were used to characterize the streamflow gain-loss associated with reservoir-supplemented streamflow conditions and with natural base-flow conditions. A comparison of the measurement sets indicates that the streamflow gain-loss characteristics of the Terror Creek study reach are consistent between the two hydrologic conditions evaluated. A substantial streamflow gain occurs between measurement locations 4 and 5 in both measurement sets, and streamflow is lost between measurement locations 5 and 7 (measurement set 1, measurement location 6 not visited) and 5 and 6 (measurement set 2). A comparison of the measurement sets above and below the mine-permit area (measurement locations 3 and 7) shows a consistent loss of 0.37 and 0.31 cubic foot per second (representing 5- and 12-percent streamflow losses normalized to measurement location 3) for measurement sets 1 and 2, respectively. This indicates that similar streamflow losses occur both during reservoir-supplemented and natural base-flow conditions, with a mean streamflow loss of 0.34 cubic foot per second for measurement sets 1 and 2. Findings from a previous investigation support the observed streamflow loss between measurement locations 3 and 7 in this study. The findings from the previous investigation indicate a streamflow loss of 0.59 cubic foot per second occurs between these measurement locations. Statistical testing of the differences in streamflow between measurement locations 3 and 7 indicates that there is a discernible streamflow loss. The p-value of 0.0236 for the parametric paired t-test indicates that there is a 2.36-percent probability of observing a sample mean difference of 0.34 cubic foot per second if the population mean is zero. The p-value of 0.125 for the nonparametric exact Wilcoxon signed rank test indicates that there is a 12.5-percent probability of observing a sample mean difference this large if the population mean is zero. The similarity in streamflow gain-loss between measurement sets indicates that the process controlling streamflow may be the same between the two hydrologic conditions evaluated. Gains between measurement locations 4 and 5 may be related to hyporheic flow from tributaries that were dry during the study. No other obvious sources of surface water were identified during the investigation. The cause for the observed streamflow loss between measurement locations 5 and 6 is unknown but may be related to mapped local faulting, 100 years of coal mining in the area, and aquifer recharge.