There is a mixed report card for autonomy as it has operated across a range of services, including schooling. Coughlan & Desurmont (2007) noted two decades of top-down imposition of autonomy on schools along with an absence of bottom-up support. This suggests a possible legitimacy deficit for autonomy. However, some economists of education (e.g., Woessmann) and the OECD, point to evidence of autonomy being associated with learning improvement. Four difficulties are encountered when endeavouring to arbitrate between these conflicting perspectives: one concerns the limited number of informants consulted for evidence about autonomy; the second is about the restricted range of measures and indicators from which autonomy judgements are derived; the third is an inconsistency in the analysis of autonomy results (the association between some measures and learning appears to be positive, but in other cases it is negative); the fourth is a complete disconnect between these indicators, measures and judgments, and the dynamics of school level practice. Collectively, these difficulties raise the question of whether autonomy may be more apparent than real. The purpose of this paper is twofold: to analyse critically the significance of autonomy as part of New Public Management and to review the evidence of autonomy and its impact assembled by proponents and critics alike.