Many factors other than a species' actual abundance can affect mist-net capture rates. We used ANCOVA models to quantify some potential biases and control their effects, producing adjusted estimates of capture rates that are more directly comparable among mist-net stations. Data came from 46 two-day mist-net sessions from September 1990 to May 1992 at six subtropical forest stations in the Rio Bravo Conservation and Management Area, northwest Belize. Factors evaluated included canopy height at net sites, long-term net shyness (days elapsed between first and last netting day of the entire study period), season (wet vs. dry), total rainfall during a netting session, and temperature. Number of individuals and species captured/10 net-h declined at each net with increasing canopy height above the net. Capture rates differed significantly among some of the stations. Elapsed days and rainfall caused significant bias in capture rates, which were statistically controlled within the ANCOVA, whereas season and temperature did not. Capture rates varied among sessions, but there was a slight and significant decline over the entire study period for all stations combined. Rainfall significantly depressed capture rates somewhat on a daily basis, but capture rates did not differ between wet and dry seasons. When we replaced the station variable in the ANCOVA with mean canopy height, the model was still highly significant, but did not explain as much of the variation in capture rates. Statistical analysis provides an objective means of interpreting data and estimating reliability, but only if statistical assumptions of the analyses are met. We discuss the need for including randomization in the experimental design, standardizing netting protocol, and quantifying sources of bias in the field, before ANCOVA or other parametric statistical techniques can be used to partition effects of biases.