This article explores the capacity of the concept of "multi-level governance" to respond to the analytical, empirical and normative challenges posed to political science by EU decision-making processes and structures. After exploring the relationship between "governance" and "multi-level goverance" (MLG), the article tackles the question of the empirical fruitfulness of this latter concept. It argues that the concept is best understood as describing those decision-making processes in which territorial jurisdictions with different comptetencies end up collaborating "on a par", thus upsetting and recasting traditional territorial hierarchies. It also discusses wether MLG, by involving different groups of citizens and fostering the diffusion of shared norms and convictions, contributes to more legitimate decision-making. The conclusions, on all counts, are somewhat mixed. Although MLG appears to describe effectively a new class of phenomena and to suggest intereasing novel hypotheses as to the transformation of the strluctures of democratic rule, it has not yet succeeded in generating new testable propositions nor in suggesting new convincing criteria of legitimacy.