I suppose if one looked at the history of any well researched variables in organizational behavior one would ®nd that the early studies reported strong eects. As Spector et al. (this issue) point out, the two studies reported in 1988 (Brief et al., 1988;Payne, 1988) which partialled out for negative aectivity/neuroticism in examining the relationship between perceptions of environmental stressors and the reporting of psychological strains, showed very large eects indeed. A scatter plot of studies today would show a few strong eects like those, but many more small eects and some where the partial correlation is larger than the zero-order correlation as the authors themselves show. It should not be forgotten perhaps that the same scatterplot pattern would emerge if the correlations between perceived stressors and reported strains were made available. A re-look at Ghiselli's (1966) The validity of occupational selection tests' would reveal that this scatter from large positive to medium negative is the norm. Why? Accepting the problems of reliability, validity and the unrepresentativeness of samples we are still faced with the question, Why do results vary so widely?'
One of the most impressive parts of this impressive paper gives six reasons why results with NA vary as much as they do. Since I have pointed out elsewhere (Payne, 1988) that NA, locus of control and Type A all correlate with each other, their reasons may also apply to the other two constructs. The six reasons are: the perception mechanism which assumes NA leads people to see more stress in their environment; the hyper-responsivity mechanism which proposes that high NA leads to an exaggerated level of strain; the selection mechanism which argues NA people selfselect into more stressful jobs; the stressor creation mechanism which suggests high NA people create more stress because they are not easy to relate to; the mood mechanism which may in¯uence reports of stressors, strains and NA itself, ®nally the causality mechanism where high stress jobs may lead to the person feeling themselves to be higher on NA. Evidence is martialled to support all the mechanisms. The paper is an admirable and important achievement.
Nevertheless, I would like to share some thoughts which might clarify my thinking if nobody else's! The argument for not partialling out for NA revolves around whether NA is construed to be a bias or to have a substantive eect. Watson, Pennebaker and Folger (1987) argued for partialling out NA because they believed high NA people perceived more stress and, because they are high on NA, report higher strain. This is certainly what I had in mind when I called it a