US national and state forest insect and disease surveys provide plentiful information on forest dieback. These data, however, have not been quantified and analyzed systematically to address outstanding questions on the etiology of dieback. This study quantified long-term (1950-1995) trends in the severity of dieback on Acer saccharum Marsh., Fraxinus spp., Betula spp., and Picea rubens Sarg. in US northern hardwoods. A numeric index (0-10 scale) of the severity and extent of dieback was applied using key words frequently found in the surveys. The 18 episodes identified showed considerable variability among species at the local scale, yet systematic, repetitive patterns of dieback at the scale of the region and multidecadal time frame. Six dieback characteristics were evident: episodes showed abrupt onset and subsidence, endured 13.6 years on average, were cyclical, with a frequency of 22.3 years between recurrence, and averaged about two-thirds of maximum possible severity. In contrast to the perception that dieback is happenstance and chaotic, this study supports the hypothesis that, by addressing issues of spatial scale and long-term population dynamics, coherent, generic patterns emerge that are cyclic and predictable. Limitations and advantages of the approaches were discussed in terms of meeting needs of the US Forest Health Monitoring Program for innovative approaches to the analysis of the voluminous field data being assembled nationwide. By developing a quantitative database, environmental correlation and modeling of dieback now become possible.