Pure vs. non-pure a priori propositions in Kant

I'm reading the Norman Kemp Smith translation of Critique of Pure Reason. Kant is famous for his analytic/synthetic distinction, but he also subdivides a priori propositions into pure and non-pure. I am trying to understand this distinction. My difficulty is that Kant seems to contradict himself regarding his paradigmatic example of a non-pure a priori proposition.

In the Introduction (B 3, end of Section I), Kant defines a pure a priori proposition to be one which has "no admixture of anything empirical". As an example of a non-pure proposition, he gives "every alteration has its cause". Alteration, he says, is a concept that one derives from experience. Evidently, involving alteration constitutes an admixture of something empirical.

Okay, fine. But then, a couple of paragraphs later (B 4–5), he gives the proposition "every alteration must have a cause" as an example of a pure a priori proposition! Here is the relevant quote:


(Emphasis added.)

How can these claims be reconciled? How can "every alteration must have a cause" be pure when it involves alteration, which, a page earlier, meant that a proposition was not pure?

I've read a 150 or so pages beyond this point, and I'm enjoying the book, but this contradiction right at the beginning just won't stop bugging me . I'm hoping that someone here can straighten me out .