In summary, numerous factors may contribute to observed differences between longitudinally and cross-sectionally derived measures of annual decline in lung function. The direction and magnitude of these differences appear hard to predict. Furthermore, although these differences can be minimized by careful modeling of the data, they cannot, in general, be completely avoided. It seems plausible, however, that both types of studies should give similar qualitative comparisons of risk factor effects if appropriately modeled. Longitudinal studies are likely to provide the most accurate and reliable estimates of lung function decline for both individuals and populations. Such data may be especially useful in identifying individuals with accelerated declines in lung function but who still have "normal" lung function as measured cross-sectionally. However, such studies require careful attention to quality control and typically require at least 4 years of follow-up before the noise in the data settles down. Multiple measurements, preferably four or more, are also necessary to reliably detect and adjust for survey effects. Cross-sectional studies, on the other hand, are simpler, cheaper, and quicker to conduct than are longitudinal studies. They may be particularly useful as a screening tool for identifying potentially affected or high-risk subjects (e.g., those with low levels of lung function) who may require further medical follow-up and/or ongoing monitoring. Both types of studies have a role in population-based occupational health hazard assessments.