A review of the evolution of the field of damage identification is presented. Trends of research and the present state of the field are discussed and directions of future research are postulated. Ideally, an automated damage identification method incorporated into the smart structure scheme would be able to detect damage as it is incurred by the structure, determine the location and extent of the damage, predict when and if catastrophic failure of the structure will occur, and alert the operator as to how the performance of the structure is affected in order for appropriate steps to be made to remedy the situation. Obviously, this is no easy task but it is essential that it is clearly defined how the research fits into the ultimate goal of developing an automated, noninvasive damage identification method. In attempting to quantify changes in response characteristics due to damage on a structure, it is very important to be aware of the inherent variabilities one might encounter in acquiring these response characteristics. These variabilities may come from computational algorithms, sensor error, or environmental effects. A method which assumes a priori ignorance to the manifestations of damage in the response characteristics of the structure is presented. This method uses inductive learning to statistically isolate changes in response characteristics due to damage from those due to the inherent variabilities. In order to validate the method, an example is presented which identifies the existence and location of a small test mass on an aluminum plate via the measurement of the structural impedance-response of the plate.