The Treatment of Reciprocal Interaction, with or without Lag, in Path Analysis

Turner and Stevens in a recent paper [1959] and Tukey a few years ago [1954] have proposed a modification of the method of path analysis, introduced by the present author [1921]. Their central thesis, the replacement of the standardized path coefficients by the concrete path regressions, has been discussed recently (Wright [1960]). In addition, both of the above papers discussed the application of the method to reciprocal interaction between variables (feedback). A somewhat extended discussion of this important topic seems desirable. The above paper (Wright [1960]) or earlier ones (Wright [1934, 1954]) should be consulted for reviews of the general method. The direct application to instantaneous reciprocal interaction was avoided in these earlier papers because of a complication in method that seemed likely to create confusion. It seemed preferable to deal with two such interacting variables as both completely determined by the same factors rather than as acting directly on each other. On consideration of the papers by Tukey and by Turner and Stevens, I agree that there is more merit in the direct treatment of circular paths by path analysis than I had previously concluded.