To understand Kant's transcendental deduction of categories we must distinguish between Kant's strategy foe constructing such a deduction and the manner in which this strategy is executed. I argue that both versions of the deduction contain similar strategies in which categories are identified with transcendental conditions of experience. Where the versions differ substantially is in the manner Kant executes the various stages of this strategy. It is pointed out, for instance, that in the objective deduction in A Kant introduces 'understanding' (...) as a defined term (A119), whereas in B Kant postulates understanding as the fundamental activity of synthesis in terms of which he formulates the arguments of each stage of the deduction. Once the distinction between strategy and execution is accepted, much of Vaihinger's evidence for the "patchwork thesis" dissolves. But I also argue that in neither version of the deduction does Kant execute the identificatory strategy with convincing success. (shrink)