The performance limits of the best types of electrical instrument are considered. The errors from unknown causes have, in general, been reduced to negligible values in specially adjusted instruments, and in this sense, the design of electrical instruments may be said to have achieved finality. However, there is still room for improvement in the reduction of errors from known causes and in the raising of the level of the commercial product, as distinct from the specially adjusted laboratory one.The defects found in instruments submitted for test by the National Physical Laboratory are described in some detail; some of the most important of these are scale errors, level errors and sticking of the movement. There is, however, a very large number of other defects, ranging from irritating trifles to serious faults, many of which could be eliminated by more careful inspection of the instruments by the manufacturers.The design of instruments might be improved with a view to reducing errors from known causes. One error, in particular, which is becoming of increasing importance with modern applications is the frequency error, which is usually small in the best instruments at power frequencies but may become so large at moderately high audio frequencies as to render the instrument useless. The frequency errors of a number of types of instrument are considered, and the improvements made possible by simple types of compensation are demonstrated.