In this article I argue against Bošković and Takahashi's (1998) analysis of scrambling as base generation (with lowering for Ā-cases). I present evidence from Russian of scope and antireconstruction effects and scrambling/wh-movement parallels, all implicating a “classical” overt movement account of Ā-scrambling. I then discuss theoretical issues unresolved by the base generation/lowering account. Having shown that Ā-scrambling is (upward) movement, I argue that the account of A-scrambled arguments as base-generated also loses its force. In conclusion I suggest an alternative way to eliminate the apparent optionality associated with scrambling, while maintaining the classical analysis of scrambling as upward movement.