It is a humiliating consideration for human reason that it is incompetent to discover truth by means of pure speculation, but, on the contrary, stands in need of discipline to check its deviations from the straight path and to expose the illusions which it originates. But, on the other hand, this consideration ought to elevate and to give it confidence, for this discipline is exercised by itself alone, and it is subject to the censure of no other power. The bounds, moreover, which it is forced to set to its speculative exercise, form likewise a check upon the fallacious pretensions of opponents; and thus what remains of its possessions, after these exaggerated claims have been disallowed, is secure from attack or usurpation. The greatest, and perhaps the only, use of all philosophy of pure reason is, accordingly, of a purely negative character. It is not an organon for the extension, but a discipline for the determination, of the limits of its exercise; and without laying claim to the discovery of new truth, it has the modest merit of guarding against error.
At the same time, there must be some source of positive cognitions which belong to the domain of pure reason and which become the causes of error only from our mistaking their true character, while they form the goal towards which reason continually strives. How else can we account for the inextinguishable desire in the human mind to find a firm footing in some region beyond the limits of the world of experience? It hopes to attain to the possession of a knowledge in which it has the deepest interest. It enters upon the path of pure speculation; but in vain. We have some reason, however, to expect that, in the only other way that lies open to it- the path of practical reason- it may meet with better success.
I understand by a canon a list of the a priori principles of the proper employment of certain faculties of cognition. Thus general logic, in its analytical department, is a formal canon for the faculties of understanding and reason. In the same way, Transcendental Analytic was seen to be a canon of the pure understanding; for it alone is competent to enounce true a priori synthetical cognitions. But, when no proper employment of a faculty of cognition is possible, no canon can exist. But the synthetical cognition of pure speculative reason is, as has been shown, completely impossible. There cannot, therefore, exist any canon for the speculative exercise of this faculty- for its speculative exercise is entirely dialectical; and, consequently, transcendental logic, in this respect, is merely a discipline, and not a canon. If, then, there is any proper mode of employing the faculty of pure reason- in which case there must be a canon for this faculty- this canon will relate, not to the speculative, but to the practical use of reason. This canon we now proceed to investigate.