Next Contents


We will discuss scientific realism from the perspective of cosmology, especially primordial cosmology: i.e. the cosmological investigation of the very early universe.

We first (Section 2) state our allegiance to scientific realism, and discuss what insights about it cosmology might yield, as against "just" supplying scientific claims that philosophers can then evaluate. In particular, we set aside (Section 2.2) the traditional methodological worry that cosmology cannot be a science; (because, it is alleged, there being only one universe means that any putative laws of cosmology would have only one instance). We also discuss limitations on ascertaining the global structure of any universe described by general relativity (Section 2.3). Then in Section 3, we review in a realist spirit, some of what is now known about the early universe. Here, "early" will mean, for us, times later than about 10−11 seconds after the Big Bang (!).

The rest of the paper addresses, in more detail, two issues in cosmology that bear on scientific realism: especially the theme, familiar in philosophy of science, of the under-determination of theory by data. But the issues do not (we believe!) threaten scientific realism. Rather, they clarify-and agreed: limit-what a scientific realist should take as definitively established by modern cosmology. They both concern primordial cosmology; where "primordial", and similarly "very early", means for us: "so early that the energies are higher than those which our established theories successfully describe". This will turn out to mean: times earlier than about 10−11 seconds after the Big Bang. The two issues each have a vast literature, written by cosmologists: who have proven - we are happy to report - to be very insightful about the conceptual, indeed philosophical, issues involved.

The first issue (Section 4) concerns the difficulty of observationally probing the very early universe. Thus expressed, this is hardly news: we would expect it to be difficult! But the issue is more specific. In the last thirty years, it has become widely accepted that at times much earlier (logarithmically) than one second after the Big Bang, there was an epoch of accelerating expansion (dubbed inflation) in which the universe grew by many orders of magnitude. The conjectured mechanism for this expansion was a physical field, the inflaton field, ϕ, subject to an appropriate potential V(ϕ); (or maybe by a set of such fields, but most models use a single field). The evidence for this inflationary framework lies principally in (i) its solving three problems that beset previous general relativistic cosmological models (the flatness, horizon and monopole problems), and (ii) its explanation of some main features of the cosmic microwave background radiation (CMB). However, this evidence leaves very undetermined the detailed physics of the inflationary epoch: in particular, it allows many choices for the shape of the potential V(ϕ), and there are nowadays many different models of inflation.

The second issue (Section 5) concerns the difficulty of confirming a cosmological theory that postulates a multiverse, i.e. a set of domains (universes) each of whose inhabitants (if any) cannot directly observe, or otherwise causally interact with, other domains. This issue arises because many models of inflation amount to such a theory. That is: according to many such models, in an epoch even earlier than that addressed in our first issue - and even harder to access - a vast number of different domains came into being, of which just one gave rise to the universe we see around us, while the others "branched off", and are forever inaccessible to us. That is, of course, rough speaking: more details in Section 5. For now, we emphasize that this pictureis very speculative: the physics describing this earlier epoch is not known. The energies and temperatures involved are far above those that are so successfully described by the standard model of elementary particle physics, and far above those probed by the observations described in relation to our first issue (as in Section 4).

Nevertheless, cosmologists have addressed the methodological question how we can possibly confirm a multiverse theory, often by using toy models of such a multiverse. The difficulty is not just that it is hard to get evidence. We also need to allow for the fact that what evidence we get may be greatly influenced, indeed distorted, by our means of observation: that is, by our location in the multiverse. Thus one main, and well-known, aspect concerns the legitimacy of anthropic explanations. That is, in broad terms: we need to ask: How satisfactory is it to explain an observed fact by appealing to its absence being incompatible with the existence of an observer?'

For all the issues we discuss, it will be clear that much remains unsettled, as regards both physics and philosophy. But we will maintain that these remaining controversies do not threaten scientific realism. We can already state the main reason why not. In Section 2, we will take scientific realism to be a claim along the lines "we can know, indeed do know, about the unobservable". But that does not imply that - and it is simply no part of scientific realism to claim that - "all the unobservable is known"; or even, "all the unobservable is knowable". For example, we will maintain that the scientific realist can perfectly well accept - should accept! - that:

  1. the global structure of space-time may be unknowable (indeed: will provably be unknowable, if general relativity is true);

  2. dark matter and dark energy are known to exist, but that their nature is unknown;

  3. even assuming there is an inflaton field ϕ, the potential V(ϕ) governing it may be unknowable; and

  4. a theory that postulates a multiverse faces special difficulties about confirmation.

All these four admissions, corresponding to Sections 2.3, 3, 4 and 5 respectively, are just `epistemic modesty'. They are compatible with the characteristic `epistemic optimism' of scientific realism-that much is already known, and yet more can be: an upbeat note on which our final summary (Section 6) will end.

Issues that we set aside:-
We should also register at the outset several other ways in which our discussion will leave issues unsettled. Partly this is a matter of setting issues aside just because we intend to write a review of mainstream ideas in a limited space-and with limited expertise! And partly it is a matter of the issues being open.

First: there are broad issues directly about scientific realism. For example:

(A): Does the fact that in cosmology, and indeed astronomy, we cannot manipulate the objects and events in question as we do in other sciences, undercut claims of scientific realism; (or of its variants like Hacking's entity realism (Hacking 1983, Miller 2016)?

(B): Does this fact undercut claims about causation, at least when understood in terms of manipulations or interventions (Woodward 2003)?

(C): Does the difficulty of observationally probing the early universe undercut science's usual strategy of confirming new theoretical postulates by finding independent lines of access to the postulated entities? (This of course relates directly to our chosen focus on the under-determination of theory by data.)

Such issues, especially the last, have been discussed very judiciously, and with more detail than here, by Smeenk (2013: Sections 6-8; 2014: Sections 6-8; 2016: Sections 3-4).

Second: there are no less than five specific issues on which this essay might have focused (rather than the under-determination of theory by data) - all of which bear on scientific realism. We list them in a roughly increasing order of specificity vis-à-vis primordial cosmology. Though we will touch on some of them later, we will mostly set them aside; so here we also give some references.

(1): In recent decades, developments in high-energy physics, such as string theory and the recognition that most theories are effective, i.e. limited in their range of validity, has made problematic the confirmation of putatively fundamental theories: and thereby also, the defence of scientific realism. Cf. Dawid (this volume, 2013).

(2): At sufficiently early times after the Big Bang, energies are so high that atoms and even nuclei `melt', so that the proverbial clock and rod with which we measure time and space cannot possibly exist. So we need to scrutinize the limits of application of our temporal and spatial concepts in such regimes. This scrutiny has been undertaken by Rugh and Zinkernagel in several papers (2009, 2016).

(3): Modern cosmology, especially inflationary cosmology, makes much use of probability measures over, for example, some set of possible initial states in a cosmological model. How to define these measures rigorously, and how to then justify the choice of one measure rather than another, are often hard, and disputed, questions. Cf. for example, Schiffrin and Wald (2012); and for philosophers' views, Koperski (2005), Norton (2010) and Curiel (2015).

(4): Inflationary cosmology proposes that quantum fluctuations in the inflaton field became classical and generated slight variations in matter density at early times: variations that led to slight anisotropies in the cosmic microwave background radiation (CMB), which were then magnified by matter clumping together under gravity, leading to stars and galaxies. We give some details of this remarkable mechanism in Section 4.2.2. But we should note at the outset that in the transition from quantum fluctuations to classical fluctuations, one faces quantum theory's notorious measurement problem! After all, `quantum fluctuation' really means `non-zero amplitude for more than one alternative' while `classical fluctuation' means (less puzzlingly!) `jitter in the actual possessed value of a given variable'. Of course, practitioners of inflationary cosmology recognize this; and many appeal to decoherence as a solution. But we believe, along with most aficionados of the measurement problem, that decoherence, though important, is not a complete solution. So the issue remains open; and fortunately, some foundationally-inclined cosmologists pursue it. Cf. for example, Perez et al. (2006), Sudarsky (2011), Cañate et al. (2013), Colin and Valentini (2016).

(5): The hypothesis of a cosmological multiverse raises several issues of scientific method additional to those we will address in Section 5. One is whether the difficulty of confirming the hypothesis, and-or the ensuing need to accept anthropic explanations, prompt a revision in our conception of scientific explanation, or more generally in our conception of science and its method. This has been the subject of considerable debate: not surprisingly, since it borders on general questions about the aim and scope of science, and the perhaps special role of cosmology in humankind's search to understand the universe and our place in it. Surveys of these issues can be found in, for example, Carr (2014, especially Sections 4-6) and Ellis (2014, Sections 6, 8; 2016, especially Sections 2, 3, 6). Ellis and Silk (2014) is a good example of skepticism about the multiverse.

So much by way of a list of issues to be set aside. (The list format is of course not meant to deny that the issues are connected. They obviously are: for example, conceptual advances about fundamental theories, under (1) or (4), might help with (3)'s problems about measures.) We now turn to what we have promised we will address ...

Next Contents