There exists a Chevalley system (Xα)α∈R in (g,h)(§2, no. 4, Def. 3) such that Xα=xα and X−α=yα for α∈B. For every Chevalley system (Xα′)α∈R having these properties, and for all α∈R,Xα′ is a basis of G∩gα.
For α∈B, put Xα=xα,X−α=yα. For α∈R+ -- B, choose a w∈W such that w(α)∈B and an automorphism φ of g such that θφ=φθ,φ(G)=G and φ(h)=w−1(h) for h∈h (Cor. 2); put Xα=φ(xw(α)),X−α=φ(yw(α)). Then
[X−α,Xα]=φ([yw(α),xw(α)])=φ(Hw(α))=w−1(Hw(α))=Hα
θ(Xα)=θφ(xw(α))=φθ(xw(α))=φ(yw(α))=X−α
so (Xα)α∈R is a Chevalley system. Moreover,
G∩gα=φ(G∩gw(α))=φ(Zxw(α))=ZXα(32)
G∩g−α=φ(G∩g−w(α))=φ(Zyw(α))=ZX−α(33)
Let (Xα′)α∈R be a Chevalley system such that Xα′=xα,X−α′=yα for α∈B. Let S be the set of α∈R such that Xα′=±Xα. By §2, no. 4, Prop. 7, S is a closed set of roots. Since S⊃B∪(−B), we have S = R (Chap. VI, §1, no. 6, Prop. 19). Thus, by (32) and (33), we have G∩gα=ZXα′ for all α∈R.
Bourbaki sets the proof directly after the statement, and the corpus does not separate them, so what is above is the statement and its proof together.