Title: Background independence: What's special about general relativity? Author: Oliver Pooley In formulating his general theory of relativity (GR), one of Einstein's goals was the generalization of the restricted relativity principle of special relativity to a principle that upheld the physical equivalence of reference frames in arbitrary states of relative motion. He believed that a generally covariant theory---a theory the equations of which held true with respect to every one of a class of coordinate systems related by smooth but otherwise arbitrary coordinate transformations---would satisfy this requirement. Einstein was quick to drop this interpretation of the significance of general covariance in the face of Kretschmann's criticisms, and the physics community, by and large, has followed him. Two views are surely widespread: (i) that the privileged coordinate systems, and associated absolute states of motion of pre-GR theories live on as the \emph{local} inertial frames and motions accelerated with respect to these [pace Dieks]; (ii) general covariance cannot have any physical content for \emph{any} theory can be expressed in a generally covariant form [but NB qualifications needed here -- Norton 2003; 1995]. I will argue that these claims are essentially correct, but they raise an obvious question: if not its general covariance, what is the conceptual novelty of GR in contrast to pre-GR theories? One possible answer again surely commands widespread acceptance (at least as part of the answer). It is that: (iii) GR lacks \emph{absolute objects}---non-dynamical fields that in pre-GR theories define a fixed spacetime background against which the dynamics of these theories unfolds [ref. to Pitts' recent work, 2005]. It is a commonplace amongst physicists working on the canonical quantization of GR that what is special about the theory is that it is \emph{background independent}. If being background independent is simply a matter of lacking background fields, and if background fields are simply absolute objects, then this view is just an expression of that stated in the previous paragraph. Unfortunately, things are not so simple. Background independence is routinely linked to the \emph{diffeomorphism invariance} of GR. But in what way does a theory's being diffeomorphism invariant differ from its being generally covariant (even when the former is qualified by the adjective `active')? This paper has two aims. (1) I seek to defend claims (i), (ii) and (iii) above against the recent trend that holds that there is a sense of general covariance, or of diffeomorphism invariance, according to which (ii) is false. My principle claim will be that there is no reason to deny that the diffeomorphism group is a gauge group of the appropriately formulated pre-GR theory [and here `appropriately formulated' does not (just) mean in the manner of Sorkin's 2002 example, but simply refers to standard generally covariant formulations, the proper interpretation of which will be discussed]. It will follow that there is no difference between GR and pre-GR theories with regards to the `individuation' of the theories' fundamental entities (pace Stachel), or with regards to which kind of magnitudes the theories take to be genuinely observable (pace Earman). (2) I draw distinctions between the notions of (a) absolute objects, (b) non-dynamical fields and (c) background fields. While GR may be distinguished from certain pre-GR theories in terms of (a) and (b), I argue it is really the notion (c) that counts (GR lacks fields of the type (a), (b) and (c)). One might hope that identifying what is special about GR will help in identifying what makes it so hard to quantize, and, relatedly, what leads to the Problem of Time. I will relate recent discussion of these issues (Belot 2005, Belot and Earman) to the claims of this paper, especially the denial that a substantive notion of general covariance is key to understanding this knot of problems.