come or cancer mortality, but all three documented significant differences in symptoms recording. The symptoms varied widely, with headache, irritability, and fatigue being common in all three studies. However, all of these symptoms may also increase due to other factors, such as perceived risk.
The authors of these reports drew somewhat different conclusions from their studies, based on the degree to which they believed recall bias accounted for differences in reported symptoms. Recall bias is difficult to avoid if a community is episodically exposed to a noxious agent (such as hydrogen sulfide) with a very powerful odor. Recent studies of such communities in Alberta report that symptoms were more common in exposed groups than in controls, but that there were no difference in objective data of morbidity or mortality (Dales et al., 1989). This paper includes considerable discussion of the problem of recall bias. Hopwood and Guidotti (1988) also show that recall bias operates in the recalling of incidents of acute exposure.
Although it might be concluded that recall bias explains the symptom differences in all of these studies, the real possibility nevertheless exists that the symptoms complained of are more sensitive as indicators of significant exposure than are more severe outcomes. In air pollution studies, recall bias cannot account for daily diary records of respiratory symptoms that relate to variations in air pollution. Longitudinal studies of recorded symptoms and exposure around hazardous-waste sites could similarly avoid much of the problem, although this has not so far, to our knowledge, been attempted.
A different example of the use of cross-sectional community studies is the use of data on birth defects from different localities separated by their proximity to known hazardous-waste sites or contaminated water supplies. The Lipari Landfill study (NJDOH, 1989) found low average birth weight in children born or conceived while the Gloucester County, New Jersey, landfill was operating. After the landfill closed, the differences disappeared, adding plausibility to the association between exposure to airborne pollutants and low birth weight. Follow-up studies of residents close to Love Canal yielded similar results with this end point, with some evidence of an effect that subsided in later years (Goldman et al., 1985).
In community studies, the same exposure is assigned or assumed for all individuals. Individual cross-sectional studies use measures of individual rather than community levels of exposure and correlate them with differences in outcome variables across individuals rather