I'd like to mention first, to put this discussion into philosophical context, the notion of "positive science" which originates with Comte and has variants in many kinds of positivism. In David Hume we see first the praise of science and the castigation of metaphysics. In Comte we see an awareness that not all that passes for science is good (some entire disciplines are regarded by him as worthless), and that even in domains which are productive, the manner of doing science and the conclusions which scientists draw may be inappropriate.

In positive science the scientist is engaged in undertaking experiments and in reporting the results of his experiments. His theories should not go beyond the experimental data, they should merely summarise it.

What I want to do here is keep this idea in mind, but moderate it somewhat so that we end up with some sense of how a theory of physics may go beyond its experimental mandate, discover some way to separate out the deginite from the speculative, and then consider ways in which the issues which then remain uncertain can be progressed.

I don't know exactly what Comte had in mind when he talked about theories not going beyond the experimental data on which they are based, but there is one very basic way in which we do normally expect them to do so. We expect a scientific theory to be a generalisation, of which the experimental observations provide particular instances. The merit of such a theory lies precisely in its going beyond the experimental observations to tell us what the results would be if many other similar experiments which fall within its scope.

A second way in which positivists seek to limit science is phenomenalistically. Thus, one approach to formalisation of physics attempted by Carnap and others in the most recent major phase of positivism was to do so using a language which spoke only of the phenomena or sense data. This is not what I have in mind here. We need to think and talk about the things which cause these phenomena.

The Logical Positivists had another approach to separating science from metaphysics. This was to be liberal about the language of science but insist that a theory be empirically "verifiable". They ultimately failed to define this notion in a satisfactory way, but even if they had suceeded it would not suffice for my present purposes. For this tests the theory as a whole and I am interested in taking a single theory, say for example General Relativity, and distinguishing physical content from verbal or metaphysical content.

What remains may not qualify as positivist, it is so only in the limited sense of recognising that there may be differences in the character of different aspects of a physical theory which allow further discussion of certain aspects of a physical theory which has solid observational support. The distinctions include something which might reasonably be described as the distinction between physics and metaphysics, though in this case the attitude towards the metaphysics is not so negative as is typical for postivists. I cannot offer any kind of definition of the distinctions which are at stake here, it is part of the enterprise to make the distinctions clearer, but I have no expectation that they will become sharp.

The distinctions of interest here are between physical and verbal content (in which no great weight should be attached to the word "verbal") and that between metaphysical and other content.

Historically it is natural to think of the structure of space and time as metaphysics. In the twentieth century they have come to be viewed as physics, because first special and then general relativity put forward theories which depended upon changes to our ideas about space and time, and turned out to be better models of physical reality. However, though the theories are better models than their predecessors, there may be alternatives which are equally good in accounting for the observational data but which do not involve changes in our conception of space and time. In the case of special relativity, there are equivalent theories which retain Newtonian space-time and are equally consistent with experimental data. I'm not intimate enough with the detail but lets suppose that Newtonian mechanics with absolute motion and a lorentz contraction on bodies according to their absolute velocities is one such theory and call it NFL. Not sure whether there is a stronger relationship between SR and NFL that their joint consistency with the empirical data. Suppose that they are equivalent in a strong way.