Steady state sensitivity analysis, as exemplified by work in metabolic control analysis and biochemical systems theory, has proved a potent tool in the analysis of biochemical networks. A limitation of this method is that it treats the response of a system only under step perturbations. Extension of these results to arbitrary disturbances, as addressed in this paper, can be carried out directly using the frequency domain tools of signal processing and control theory. This analysis, while it remains linear (and hence local), allows improved theoretical understanding as well as new ways of describing the behavior of systems of interest. The theory can be interpreted as a natural extension of metabolic control analysis and as such provides generalizations of the summation and connectivity theorems. The results are illustrated by application to some simple biochemical feedback networks.