We present a new approach to the construction of chemical reaction mechanisms by methods derived, in part, from electronic circuit and systems theoretical techniques. The approach, correlation metric construction (CMC), is based on the calculation and analysis of a time-lagged multivariate correlation function of a set of time-series of chemical concentrations. The time-series are composed of the observed responses of species composing a chemical reaction network to random changes in the concentration of a set of input species. The four-dimensional correlation-time lag function is subsequently transformed into a metric distance function and is analyzed by multidimensional scaling and cluster analysis in order to (1) determine a measure of effective dimensionality of the system; (2) construct a correlation diagram of the reaction mechanism that graphically recapitulates, in large part, the reaction steps in the network by a technique that emphasizes the strengths of coupling among the constituent species; and (3) determine the hierarchy of control in the network and identify possible weakly coupled or uncoupled subsystems. In order to demonstrate the technique, we analyze three different models of common types of chemical reactions. The analysis of these examples, which include enzymatic substrate cycles, mass action kinetics, networks with rate-determining steps, and networks satisfying the steady-state hypothesis, demonstrates that CMC is able to construct informative diagrams which construct, in large part, the underlying chemical reactions and strengths of interactions among the measured species in the network.