It is interesting that the previous discussant, Robert Kaplan, and I were each struck by the same quanta in Professor Clark's paper, namely, the correlations given in table 8 and the results of Clark's factor analysis. Kaplan said there is no dependent variable in the analysis; I would merely observe that such is the nature of a factor analysis, to find some coherent structure among a set of variables. Typically, though, the set of variables is provided by observation, not constructed using alternative arithmetic rules concocted by the researcher. What Clark has done is (a) concocted twenty-three measures of fiscal strain, many highly correlated because they are almost arithmetically redundant (i.e., linearly dependent), and (b) factor analyzed these measures to see what he hath wrought. That the factor analysis produced four meaningful factors from these twenty-three variables is interesting. But they reflect no more than was input, namely, the variables created by Clark, and should not be interpreted as the totality of fiscal strain indicators. If indeed these indicators are to be useful to municipalities, one needs the wherewithal to calculate factor scores for each municipality. These factor scores are standardized scores on what would be the magnitude of the indicator if it could be directly measured. Moreover, as time and circumstances changed, so that the values of the twentythree variables changed, it would be useful to be able to calculate the reflection of these changes in the indicators. To do this, one must assume that the factor structure and factor loadings, as developed by Clark's study, are universal constants and invariant with changes of circumstances. This may be straining one's belief in results of factor