Annual counts of egg masses have been promoted as an appropriate state variable for monitoring populations of some amphibian species. However, if some egg masses are not detected and detectability changes over time, the use of egg mass counts is unreliable. Variation in counts of egg masses may be indicative of variation in actual abundance or variation in detectability. I used closed capture–recapture models to estimate detection probability and evaluate potential sources of variation in the detectability of Wood Frog (Rana sylvatica) egg masses in a pond in Rocky Mountain National Park in 2003 and 2004. Model selection results and model-averaged estimates provided evidence that detection probability varied between years. However, I found no evidence of variation between observers within each year. The results of this study provide additional evidence that detection probability needs to be accounted for if egg mass counts are to be used to infer abundance and population trends in amphibian populations.