However, a classification is no end in itself. As Ernest Rutherford used to provoke, "science is either physics or collecting stamps". So what are the arguments for presuming the existence of other universes?

4. Is Our Universe Fine-Tuned?

Life exists in this universe. But does it exist in other universes, and if so, are those life-forms similar to those which are typical of Earth? We cannot, as yet, answer these latter questions. However, the fact that life exists at all raises its own fundamental questions, such as, why are the physical and chemical features of this universe compatible with life? Why does this universe appear to be "fine tuned," that is, why are the laws and constants of nature and certain boundary conditions narrowly tailored to those constants which are ideal for life (e.g. Carr 2007; Leslie 1989; Vaas 2004b); for indeed, if these fundamental constants varied by just a few degrees, life as we know it would not be possible.

Life as we know it depends crucially on the laws and constants of nature as well as a variety of boundary conditions which allowed intelligent life to evolve. Nevertheless it is difficult to judge how "fine-tuned" this universe really is, both because it is unclear how modifications of many values together might compensate each other and whether laws, constants and initial conditions really could have been otherwise to begin with. It is also unclear how specific and improbable those values need to be in order for information-processing structures – and, hence, intelligent observers – to develop. If we accept, for the sake of argument, that at least some values are fine-tuned, we must ask, "why" and how this can be explained.

In principle, there are many options for answering this question (Table 2). Fine-tuning might (1) just be an illusion if life could adapt to very different conditions or if modifications of many values of the constants would compensate each other; or (2) it might be a result of (incomprehensible, irreducible) chance, and thus inexplicable (Vaas 1993); or (3) it might be nonexistent because nature could not have been otherwise, and with a fundamental theory we would be able to prove this; or (4) it might be a product of selection: either observational selection within a vast multiverse of (infinitely?) many different realizations of those values (weak anthropic principle), or a kind of cosmological natural selection making the measured values (compared to possible other ones) quite likely within a multiverse of many different values, or even a teleological or intentional selection. Even worse, these alternatives are not mutually exclusive – for example it is logically possible that there is a multiverse, created according to a fundamental theory by a cosmic designer who is not self-sustaining, but ultimately contingent, i.e. an instance of chance.

Table 2: Digging deeper: Laws, constants and boundary conditions are the basic constituents of cosmology and physics from a formal point of view (besides spacetime, energy, matter, fields and forces or more fundamental entities like strings or spin-networks and their properties with regards to content). An ambitious goal and historically at least a successful heuristic attitude is reduction, derivation and unification to achieve more fundamental, far-reaching and simple descriptions and explanations. While uniqueness is much more economical and predictive, multiple realizations – presumably within a multiverse – have recently been proposed as an opposing (but not mutually exclusive) alternative. This table provides a summary of different approaches, possibilities and problems; it is neither complete nor the only conceivable system. (From Vaas 2009b.)

From both a scientific and philosophical perspective the fundamental theory approach and the multiverse scenario are most plausible and heuristically promising (Vaas 2004a,b).

5. Unique-universe versus multiverse accounts

Unique-universe accounts take our universe as the only one (or at least the only one ever relevant for cosmological explanations and theories). It might have had a predecessor or even infinitely many (before the big bang) and/or a successor or infinitely many (after a big crunch), but then the whole series can be taken as one single universe with spatiotemporal phase transitions. From the perspective of simplicity, parsimony and testability it is favorable to try to explain as much as possible with a unique-universe account. A both straightforward and very ambitious approach is the searching for a fundamental theory with just one self-consistent solution that represents (or predicts) our universe. (Of course one could always argue that there are other, strictly causally separated universes too, which do not even share a common generator or a meta-law; but then they do not have any explanatory power at all and the claims for their existence cannot be motivated in a scientifically useful way, only perhaps by philosophical arguments.)

Future theories of physics might reveal the relations between fundamental constants in a similar way as James Clerk Maxwell did by unifying electric and magnetic forces: he showed that three seemingly independent constants – the velocity of light c, the electric constant eo (vacuum permittivity), and the magnetic constant μ_o (vacuum permeability) – are connected with each other: c = (μ_o . e_o)^-0,5. Indeed some candidates for a grand unified theory of the strong, weak and electromagnetic interaction suggest that most of the parameters in the standard model of particle physics are mathematically fixed, except for these three: a coupling constant (the electromagnetic fine-structure constant) and two particle masses (namely that of down and up quarks) (Hogan 2000). A promise of string theory is to get rid of any free parameter – and if this can be accomplished all constants could be calculated from first principles (Kane et al. 2002). Although at present this remains wishful thinking, it is a direction very worth following and, from a theoretical and historical point of view, perhaps the most promising.

So even without an ultimate explanation, fine-tuning might be explained away within a (more) fundamental theory. Most of the values of the physical constants should be derived from it, for example. This would turn the amazement about the anthropic coincidences into insight – like the surprise of a student about the relationship e^iπ = -1 between the numbers e, i and π in mathematics is replaced by understanding once he comprehends the proof. Perhaps the fact that the mass of the proton is 1836 times the mass of the electron could be similarly explained. If so, this number would be part of the rigid formal structure of a physical law which cannot be modified without destroying the theory behind it. An example for such a number is the ratio of any circle's circumference to its diameter. It is the same for all circles in Euclidean space: the circular constant π.

But even if all dimensionless constants of nature could be reduced to only one, a pure number in a theory of everything, its value would still be arbitrary, i.e. unexplained. No doubt, such a universal reduction would be an enormous success. However, the basic questions would remain: Why this constant, why this value? If infinitely many values were possible, then even the multitude of possibilities would stay unrestricted. So, again, why should such a universal constant have the value of, say, 42 and not any other?

If there were just one constant (or even many of them) whose value can be derived from first principles, i.e. from the ultimate theory or a law within this theory, then it would be completely explained or reduced at last. Then there would be no mystery of fine-tuning anymore, because there never was a fine-tuning of the constants in the first place. And then an appreciable amount of contingency would be expelled.

But what would such a spectacular success really mean? First, it could simply shift the problem, i.e. transfer the unexplained contingency either to the laws themselves or to the boundary conditions or both. This would not be a Pyrrhic victory, but not a big deal either. Second, one might interpret it as an analytic solution. Then the values of the constants would represent no empirical information; they would not be property of the physical world, but simply a mathematical result, a property of the structure of the theory. This, however, still could and should have empirical content, although not encoded in the constants.

Otherwise fundamental physics as an empirical science would come to an end. But an exclusively mathematical universe, or at least an entirely complete formal description of everything there is, derivable from and contained within an all-embracing logical system without any free parameter or contingent part, might seem either incredible (and runs into severe logical problems due to Kurt Gödel's incompleteness theorems) or the ultimate promise of the widest and deepest conceivable explanation. Empirical research, then, would only be a temporary expedient like Ludwig Wittgenstein's (1922, 6.54) famous ladder: The physicist, after he has used empirical data as elucidatory steps, would proceed beyond them. "He must so to speak throw away the ladder, after he has climbed up on it." That there is no contingency at all seems very unlikely. So why are some features realized but not others? Or, on the contrary, is every feature realized? Both questions are strong motivations for M.

Multiverse accounts assume the existence of other universes as defined above and characterized with respect to their separation in Table 1. This is no longer viewed as mad metaphysical speculation beyond empirical rationality. There are many different multiverse accounts (see, e.g., Carr 2007; Davies 2004; Deutsch 1997; Linde 2008; Mersini-Houghton 2010; Rees 2001; Smolin 1997; Tegmark 2004, Vaas 2004b, 2005, 2008b, 2010; Vilenkin 2006) and even some attempts to classify them quantitatively (see, e.g., Deutsch 2001 for many-worlds in quantum physics, and Ellis, Kirchner & Stoeger 2004 for physical cosmology). They flourish especially in the context of cosmic inflation (Aguirre 2010; Linde 2005, 2008, Vaas 2008b), string theory (Chalmers 2007, Douglas 2010) and a combination of both as well as in different quantum gravity scenarios that seek to resolve the big bang singularity and, thus, explain the origin of our universe (for recent reviews see Vaas 2010).

If there is a kind of selection, new possibilities emerge within M. One possibility is observational selection (Barrow & Tipler 1986, Leslie 1989, Vaas 2004b). Another is a kind of cosmological natural selection making the measured values (compared to possible other ones) quite likely within a multiverse of many different values (García-Bellido 1995, Smolin 1992, 1997 & 2010, for a critical discussion e.g. Vaas 1998 & 2003). Finally there is teleological or intentional selection (e.g. Ansoldi & Guendelman 2006; Harrison 1995; Leslie 1989; Vidal 2008 & 2010, for a critical discussion e.g. Vaas 2004b, 2009ab).

The strongest version of M is related to the principle of plentitude or principle of fecundity (advocated e.g. by Richard Feynman, Dennis Sciama). According to this principle everything is real, if it is not explicitly forbidden by laws of nature, e.g. symmetry principles. As Terence H. White wrote in his novel The Once and Future King (1958): "everything not forbidden is compulsory". But the question remains: What is forbidden, i.e. physically or nomologically not possible and thus not "allowed" by natural laws? (This is a slippery slope, as Paul Davies 2007 argued: could there even be universes containing magic, a theistic God, or simulations of every weird fantasy? Otherwise some restrictions are necessary!) And is it really true that everything which is physically or nomologically possible is also realized somewhere? Answers to these questions are not known. (It is also unclear whether "metaphysically possible" makes truly sense in contrast to both "logically possible" and "physically possible", cf. Meixner 2008.)

6. Why (Should) We Believe in Other Universes (?)

There are – or could be – at least three main reasons for assuming the existence of other universes: empirical evidence, theoretical explanation, and philosophical arguments (Table 3). These three reasons are independent from each other, but ideally entangled. As far as there is at least some connection with or embedding into a theoretical framework of physics or cosmology, M is part of the scientific endeavour, not only of philosophy. This is also the case in the absence of empirical evidence, if a theoretical embedding exists. And philosophical arguments might at least motivate scientific speculation (Table 4).

Table 3: Arguments for the existence of other universes (cf. Vaas 2010).

Table 4: Controverses about universes (cf. Vaas 2010).

Many cosmologists and philosophers alike argue that M does not belong to science because it cannot be falsified and it can be tested. This criticism is not valid.

7. Falsificationism is Both Too Much and Not Enough

"Scientific statements, refering to reality, must be falsifiable", Karl R. Popper (1930-1933, 1935). Falsifiability has been one of the most important and successful criteria of science. Popper understood it as a demarcation criterium of science (in contrast to metaphysics, logic, pseudoscience...). But what is usually meant here is falsifiability of theoretical systems, or of parts of such a system, not of single statements.

So there is an important distinction: Scientific laws on the one hand must be falsifiable. Take for example Newton’s law – it can be tested, was tested rigorously indeed, and is still under investigation (at very large and very small scales). Hypothetical universal existential statements on the other hand need not and cannot be falsified, but must be verified. Take for instance the discovery of the last two naturally occurring stable elements: Hafnium, atomic number 72, was detected by Dirk Coster and George de Hevesy in 1922 through Xray spectroscopy analysis, after Niels Bohr had predicted it, or its properties. (Actually Dimitri Mendeleev had already predicted it implicitely in 1869, in his report on The Periodic Law of the Chemical Elements). Similarly, Rhenium, atomic number 75, was found by Walter Noddack, Ida Tacke, and Otto Berg in 1925, after Henry Moseley had predicted it in 1914.

In science verification is obviously extremely important. But falsification is not sufficient to disprove an idea in controversial and hypothetical situations even if there are concise predictions. That is, "absence of evidence is not evidence of absence." Otherwise fictive ghosts or unicorns would also be part of science, for we cannot know a priori whether they exist or not, but we can imagine ways to detect them; that is to verify their existence in principle. So what is missing here? It is theoretical embedding. To be reasonably part of science, hypothetical universal existence statements must not only be verifiable, they must also be part of a sufficiently confirmed or established scientific theory or theoretical framework. That was the case in the Hafnium and Rhenium examples, where Mosley and Bohr had their theoretical models about the atomic numbers and the properties of the atoms, explaining regularities in the periodic table of the chemical elements.

Statements about the existence of other universes are not like statements about scientific laws. The latter must be falsifiable, while the former should be taken as universal existential statements which cannot be falsified, but must be verified. So ultimately a multiverse scenario might only be accepted, strictly speaking, if there is empirical evidence for it, i.e. observational data of another universe or its effects. A weaker argument for M would be if a – falsifiable, rigorously tested – theory predicts the existence of other universes, and this theory is well established according to the usual scientific criteria. Still weaker are philosophical reasons; whether they could suffice if they are stronger than alternative statements is a controversial issue and lies at – or beyond – the boundaries of physics and cosmology.

8. Research Programs and Systematicity

From a philosophy of science perspective an exaggerated falsificationism is not useful. As Imre Lakatos (1976 & 1977) pointed out, the scientific endeavour usually takes place in the form of research programs (which include not only the main assumptions and theoretical ideas but also methodological issues) rather than in respect of single hypotheses or theories. And refutations are not a straightforward sign of empirical progress, and this is because research programs grow in a permanent ocean of anomalies.

First, data are theory-dependent, not just empirical; they might simply be false or interpreted wrongly. Thus, theories should not been thrown away too quickly. For example the measured anomalies in Uranus' orbit were not a falsification of Newton’s law of gravity but an indication of Neptune's existence and the starting point for its discovery.

Second, theories are often ahead of data. Thus, theorists should make testable predictions of course, and observers and experimenters should obtain data. Supersymmetry models, for instance, were theories in search of applications initially, and physicists (and even mathematicians) learned a lot from them although there has still not been a single supersymmetric particle discovered (yet). So testability, a falsifiable prediction, is important, but not necessarily at least at the initial stages of theory building. Thus, theories must be given a chance to develop, to be purged of errors, to become more complete.

What Lakatos advocated was, therefore, a sophisticated falsificationism, referring to a (quasidarwinistic) struggle between theories and data interpretation. Given initial theoretical or empirical contradictions, anomalies or data deviances occur. Nevertheless, it is reasonable to keep the theory for a while, especially its "hard core" (at least if there is no alternative – an experimentum crucis is very rarely accomplished). Instead of killing the theory, the "protective belt" of auxiliary hypotheses should be modified first. Often they are taken as responsible for failed predictions too. Of course this effort to save the core could lead to immunization strategies which would in the end expel the theory from science. Furthermore, falsification is often difficult to achieve because core commitments of scientific theories are rarely directly testable and predictive without further assumptions. That is, a theory grows or becomes modified with the discovery of additional evidence.

Over time and with the accumulation of evidence both pro and con, some research programs are more successful than others. Therefore, ad hoc hypotheses to save a hard core could led to program degenerations – especially if they are problem shifts not pointing to other fruitful areas.

A progressive research program is characterized by its growth, along with stunning novel facts being discovered, more precise predictions being made, new experimental techniques being developed. By contrast, a degenerating research program is marked by lack of growth, or by an increase of the protective belt that does not lead to novel facts.

Cosmology provides a lot of examples for these complex interplays between competing theories, data acquisition and interpretation, immunization ("epicycles" are proverbial now), and even paradigm changes. The most prominent were geocentrism versus heliocentrism, metagalaxy versus island universe, and steady state versus big bang. Universe versus multiverse seems to be the next challenge in this direction.

Though the demarcation criteria for science have sometimes been criticized (Laudan 1988) and are somewhat fuzzy indeed, it is usually rationality that rules. And there are many well established criteria for successful research programs, especially the following: many applications; novel predictions; new technologies; answering unsolved questions; consistency; elegance; simplicity; explanatory power/depth; unification of distinct phenomena; and truth (but how can we know the latter – only via the criteria mentioned above?).

These criteria are also useful in analysing modern theoretical cosmology beyond the falsifiability doctrine. Are statements about the "cause" of the big bang and about other universes unscientific if they cannot provide falsifiable predictions (yet)? No. At least not if they are seen as part of a research program which is otherwise well-established and at least partly confirmed by observations or experiments. And indeed (not all, but at least) some of the just-mentioned criteria are fulfilled by the ambitious big bang and multiverse scenarios: many applications? well, no; novel predictions? perhaps yes, partly; new technologies? definitely not (yet); answering unsolved questions? yes; consistency? hopefully; elegance? depends on taste; simplicity? often yes; explanatory power/depth? certainly yes; unification of distinct phenomena? yes indeed; truth? nobody knows, but this is of course the main source of the controversies.

Last but not least there are other approaches to science. For example science can, negatively, characterized by comparing it with pseudoscience (table 5, see Vaas 2008).