Surveys are essential for conservation of populations, because they provide estimates of density that may reveal a change in status. Managers therefore depend on analysis of density estimates to evaluate effectiveness of their actions. However, statistical problems associated with trend analysis make declines caused by environmental change difficult to detect. Here we apply piecewise regression to analysis of trends in eight species of Hawaiian passerines to determine if slopes following a year of documented change differ from slopes estimated before the change. Using estimates of densities of forest birds at Hakalau Forest National Wildlife Refuge, we show that estimates of trends based on 21 years of survey data and expressed as a single slope missed an environmental change in year 14. Analysis of residuals indicates that analysis of trends as a single slope violated the model's assumptions, giving a false impression that populations were either stable or increasing and at variance with independent indicators of population decline based on banded birds. In contrast, piecewise regression shows that the slopes of all eight species were positive before the change and negative after it and meets the model's assumptions. The Hawai'i ‘Ākepa (Loxops coccineus coccineus) and the ‘I’iwi (Vestiaria coccinea) are declining significantly, with several other species likely declining. On the basis of the structure of the data, the positive slopes were anomalous. We suggest that piecewise regression may be the more appropriate technique for dealing with environmental changes toward the end of long-term data sets.