In this paper we present a technique for the automatic design for testability of digital circuits based upon the analysis of controllability and observability measures. The new concept of sensitivity is introduced, which is a measure for the degree to which the testability of a circuit improves as increased controllability and observability is achieved over a set of nodes in a circuit. In order to improve the testability of a circuit, three simple transformations are used, namely, the addition of a new primary input and possibly an AND (OR) gate so that a logic 0(1) can be injected into the interior of the circuit, and test points so that internal signal values can be observed. We then introduce the global r-modification problem, which deals with making r (integer) transformations to a circuit in order to improve its testability. This resynthesis problem has been formulated as a mixed integer linear programming problem. A program called Testability Improvement Program (TIP) has been developed for implementing this approach, and experimental results are presented. The work presented is applicable to problems of test generation, the design of fixtures for ATE, and determining the location of test pads on integrated circuit chips when employing electron beam testing.