* The expression that's spliced in is done so as a single syntactic unit. In other words, the lambda expression `<< w x y z >>` is parsed via usual conventions as `<< (((w x) y) z) >>`. Here `<< x y >>` is not any single syntactic constituent. But if you do instead `let a = << x y >>;; let b = << w $a$ z >>`, then what you get *will* have `<< x y >>` as a constituent, and will be parsed as `<< ((w (x y)) z) >>`.
* The expression that's spliced in is done so as a single syntactic unit. In other words, the lambda expression `<< w x y z >>` is parsed via usual conventions as `<< (((w x) y) z) >>`. Here `<< x y >>` is not any single syntactic constituent. But if you do instead `let a = << x y >>;; let b = << w $a$ z >>`, then what you get *will* have `<< x y >>` as a constituent, and will be parsed as `<< ((w (x y)) z) >>`.