All data points are imperfect copies of a Platonic Data Point, an underlying Grand Truth. All data points, provided the experiment that collected them is well-designed, fit into “The grand scheme of things”. It is Not Done in Science to look solely at results that fit with what you’ve recently discovered, and ignore the ones that don’t. You have to take them all together, because they’re all equally approximations to the Truth.

However, based on my interactions with a “real” science department last summer and over the course of this academic year, the way Science (or at least, Experimental Psychology, but I suspect it generalises to some extent) is done is more like this:

All data points collected by me, and my colleagues in this department, are imperfect copies of a Departmental-Aims Data Point, which proves what this department is currently trying to show. All data points collected by Them, and Their Colleagues in Their department, are imperfect copies of a Their-Departmental-Aims Data Point, which proves what Their department is trying to show, which is mutually exclusive to what this department wants to show.

Now, this is not to say that Science is totally failing here: all the scientists referred to above would still, if shown a set of results that prove their theory wrong whilst strongly corroborating the theory of the competing department, back down and accept that they are probably/definitely wrong. So Science still moves forwards as it should. However, in the meantime and while all results are just-a-little-hazy (as tends to be the case) the literature implicitly makes this marked distinction between the stuff from department A and the stuff from department B. Which jars a little with the supposed nature of Science.

This is something that philosophers of science are aware of, I know: my module last year on the Sociology of Scientific Knowledge covered similar ground, as well as other dichotomies along the same general lines. Still, it’s something that can’t hurt to mull over a bit more.