Against the division of labor in scope and binding Alternative sets and choice functions figure in many linguistic analyses as semantic scope mechanisms, separate from syntactic scope (like quantifier raising) for "genuine" quantifiers (like "nobody"). This division of labor correctly predicts that scope constraints differ between genuine and non-genuine quantifiers, but I argue against it. First, variables are essential for quantifier raising but incompatible with Hamblin's compositional computation of alternatives; I fix this failure to treat "who saw nobody" with a variable-free semantics. Second, choice functions do not predict binding constraints like weak crossover. These problems suggest dealing with scope and binding entirely in semantics, or entirely in syntax.