Many Worlds in Context∗

Max Tegmark
Dept. of Physics, Massachusetts Institute of Technology, Cambridge, MA 02139; [email protected]
(Dated: Submitted August 3 2008, revised February 14 2010)
Everett’s Many-Worlds Interpretation of quantum mechanics is discussed in the context of other
physics disputes and other proposed kinds of parallel universes. We find that only a small fraction
of the usual objections to Everett’s theory are specific to quantum mechanics, and that all of the
most controversial issues crop up also in settings that have nothing to do with quantum mechanics.

I. INTRODUCTION cus their efforts on those remaining aspects of Everett’s

arXiv:0905.2182v2 [quant-ph] 24 Mar 2010

theory that are uniquely quantum-mechanical. This is

There is now great interest in Everett’s Many-Worlds not to say that the issues in Table 1 with a “No” in
Interpretation of quantum mechanics and the controversy the QM-specific column are necessarily unimportant —
surrounding it.1 A key reason for this is undoubtedly merely that it is unfair to blame Hugh Everett for them
that it connects with some of our deepest questions about or to use them as evidence against his theory alone.
the nature of reality. How large is physical reality? Are Rather than discuss these objections one by one in the
there parallel universes? Is there fundamental random- order they appear in Table 1, this article is structured as
ness in nature? a survey of multiverse theories, addressing the objections
in their natural context. We then return to Table 1 and
The goal of this article is to place both Everett’s theory
summarize our conclusions in Section VI.
and the standard objections to it in context. We will re-
view how Everett’s Many Worlds may constitute merely
one out of four different levels of parallel universes, the
A. The MWI: what it is and what it isn’t
rest of which have little to do with quantum mechanics.
We will also analyze the many objections to Everett’s
theory listed in Table 1, concluding that most of them Let us first spell out what we mean by the Many
are not specific to quantum mechanics. By better under- Worlds Interpretation (MWI) Much of the early criticism
standing this context, quantum physicists can hopefully of the MWI was based on confusion as to what it meant.
avoid reinventing many wheels that have been analyzed Here we grant Everett the final say in how the MWI is
in detail in other areas of physics or philosophy, and fo- defined, since he did after all invent it [1, 2], and take it
to consist of the following postulate alone:

• EVERETT POSTULATE:
∗ To appear in “Many Worlds? Everett, Quantum Theory and Re- All isolated systems evolve according to the
d
ality”, S. Saunders, J. Barrett, A. Kent & D. Wallace (eds), Oxford Schrödinger equation dt |ψi = − ~i H|ψi.
Univ. Press
1 The controversy shows no sign of abating, as evidenced by the More succinctly, “physics is unitary”. Although this pos-
results of the following highly unscientific poll carried out by tulate sounds rather innocent, it has far-reaching impli-
the author at the Perimeter Institute “Everett@50” Conference
cations:
9/22-07:
1. Do you believe that new physics violating the Schrödinger 1. Corollary 1: the entire Universe evolves according
equation will make large quantum computers impossible? (4 to the Schrödinger equation, since it is by definition
Yes/ 29 No/11 Undecided)
an isolated system.
2. Do you believe that all isolated systems obey the Schrödinger
equation (evolve unitarily)? (17 Yes/10 No/20 Undecided) 2. Corollary 2: when a superposition state is ob-
3. Which interpretation of quantum mechanics is closest to served, there can be no definite outcome (wavefunc-
your own? tion collapse), since this would violate the Everett
• 2 Copenhagen or consistent histories (including pos- postulate.
tulate of explicit collapse)
• 5 Modified dynamics (Schrdinger equation modified to Because of corollary 1, “universally valid quantum me-
give explicit collapse) chanics” is often used as a synonym for the MWI. What
• 19 Many worlds/consistent histories (no collapse) is to be considered “classical” is therefore not specified
• 2 Bohm axiomatically (put in by hand) in the MWI — rather, it
• 1.5 Modal can be derived from the Hamiltonian dynamics, by com-
• 22.5 None of the above/undecided puting decoherence rates.
How does corollary 2 follow? Consider a measurement
4. Do you feel comfortable saying that Everettian parallel uni-
verses are as real as our universe? (14 Yes/26 No/8 Unde- of a spin 1/2 system (a silver atom, say) where the states
cided) “up” and “down” along the z axis are denoted |↑i and |↓i.
Assuming that the observer will get happy if she measures
 2

Table 1. Most common worries about Everett’s many-worlds interpretation are not specific to quantum mechanics.
Worry QM-specific? Counterexamples/resolution
1 Popper worry: falsifiable? No General relativity, inflation
2 Occam worry: parallel universes feel wasteful No Level I & 2 multiverses
3 Aristotle worry: math mere approximation No Linked to expernal reality hypothesis
4 Uncertainty worry: How can omniscience allow uncertainty? No Occurs whenever observer ensemble
5 How derive probabilities from deterministic theory? No Occurs whenever observer cloning
6 Unequal probability worry: Why square the amplitudes? Yes Comes from Hilbert space structure
7 ρ worry: describes world or my knowledge of it? Partly Can describe both
8 How judge evidence for/against such a theory? No Classical statistical mechanics
9 Word worry: What do we mean by “exist”, “real”, “is”, etc? No Level I & II multiverses
10 Invisibility worry: Why can’t we detect the parallel worlds? Yes Solved by decoherence
11 Basis worry: What selects preferred basis? Yes Solved by decoherence
12 Weirdness worry No Electric fields, black holes, Levels I & II

spin up, we let |-̈ i, |⌣

¨ i and |⌢
¨ i denote the states of the important to ask whether they are within the purview
observer before the measurement, after perceiving spin of science, or merely silly speculation. They are also
up and after perceiving spin down, respectively. If the a source of confusion, since many forget to distinguish
measurement is to be described by a unitary Schrödinger between different types of parallel universes that have
time evolution operator U = e−iHτ /~ applied to the total been proposed[3, 4].
system, then U must clearly satisfy The farthest you can observe is the distance that light
has been able to travel during the 14 billion years since
U |↑i⊗|-̈ i = |↑i⊗|⌣
¨i and U |↓i⊗|-̈ i = |↓i⊗|⌢
¨ i. (1) the big-bang expansion began. The most distant visible
objects are now about 4×1026 meters away2, and a sphere
Therefore if the atom is originally in a superposition of this radius defines our observable universe, also called
α|↑i + β|↓i, then the Everett postulate implies that the our Hubble volume, our horizon volume or simply our
state resulting after the observer has interacted with the universe. Below I survey physics theories involving par-
atom is allel universes, which form a natural four-level hierarchy
of multiverses (Figure 1) allowing progressively greater
U (α|↑i + β|↓i) ⊗ |-̈ i = α|↑i ⊗ |⌣
¨ i + β|↓i ⊗ |⌢
¨ i. (2)
diversity.
In other words, the outcome is not |↑i ⊗ |⌣¨ i or |↓i ⊗ |⌢
¨i • Level I: A generic prediction of cosmological infla-
with some probabilities, merely these two states in super- tion is an infinite “ergodic” space, which contains
position. Very few physicists have actually read Everett’s Hubble volumes realizing all initial conditions —
29
original 137-page Ph.D. thesis (reprinted in [2]), which including an identical copy of you about 1010 m
has lead to a common misconception that it contains a away.
second postulate along the following lines:
• What Everett does NOT postulate: • Level II: Given the fundamental laws of physics
At certain magic instances, the world undergoes that physicists one day hope to capture with equa-
some sort of metaphysical “split” into two branches tions on a T-shirt, different regions of space can
that subsequently never interact. exhibit different effective laws of physics (physi-
cal constants, dimensionality, particle content, etc.)
This is not only a misrepresentation of the MWI, but corresponding to different local minima in a land-
also inconsistent with the Everett postulate, since the scape of possibilities.
subsequent time evolution could in principle make the
two terms in equation (2) interfere. According to the • Level III: In Everett’s unitary quantum mechan-
MWI, there is, was and always will be only one wavefunc- ics, other branches of the wavefunction add nothing
tion, and only decoherence calculations, not postulates, qualitatively new, which is ironic given that this
can tell us when it is a good approximation to treat two level has historically been the most controversial.
terms as non-interacting.
• Level IV: Other mathematical structures give dif-
ferent fundamental equations of physics for that T-
B. Many worlds galore shirt.

Parallel universes are now all the rage, cropping up in

books, movies and even jokes: “You passed your exam in 2 After emitting the light that is now reaching us, the most distant
many parallel universes — but not this one.” However, things we can see have receded because of the cosmic expansion,
they are as controversial as they are popular, and it is and are now about about 40 billion light years away.
 3

The key question is therefore not whether there is a mul- local conditions for the universal laws that should go on
tiverse (since Level I is the rather uncontroversial cosmo- that T-shirt.
logical concordance model), but rather how many levels
it has.
Below we will discuss at length the issue of evidence
and whether this is science or philosophy. For now, the
key point to remember is that parallel universes are not
a theory, but a prediction of certain theories. The Pop-
per worry listed in Table 1 is the question of whether
Everett’s theory is falsifiable. For a theory to be falsi-
fiable, we need not be able to observe and test all its
predictions, merely at least one of them. Consider the
following analogy:

General Relativity Black hole interiors

Inflation Level I parallel universes
Unitary quantum mechanics Level III parallel universes

Because Einstein’s theory of General Relativity has suc-

cessfully predicted many things that we can observe,
we also take seriously its predictions for things we can-
not observe, e.g., that space continues inside black hole
event horizons and that (contrary to early misconcep-
tions) nothing funny happens right at the horizon. Like-
wise, successful predictions of the theories of cosmologi-
cal inflation and unitary3 quantum mechanics have made
some scientists take more seriously their other predic-
tions, including various types of parallel universes.
Let us conclude with two cautionary remarks before
delving into the details. Hübris and lack of imagina-
tion have repeatedly caused us humans to underestimate
the vastness of the physical world, and dismissing things
merely because we cannot observe them from our van-
tage point is reminiscent of the ostrich with its head in
the sand. Moreover, recent theoretical insights have in-
dicated that Nature may be tricking us. Einstein taught
us that space is not merely a boring static void, but
a dynamic entity that can stretch (the expanding uni-
verse), vibrate (gravitational waves) and curve (gravity).
Searches for a unified theory also suggest that space can
“freeze”, transitioning between different phases in a land-
scape of possibilities just like water can be solid, liquid or
gas. In different phases, effective laws of physics (parti-
cles, symmetries, etc..) could differ. A fish never leaving
the ocean might mistakenly conclude that the properties
of water are universal, not realizing that there is also
ice and steam. We may be smarter than fish, but could
be similarly fooled: cosmological inflation has the de-
ceptive property of stretching a small patch of space in
a particular phase so that it fills our entire observable
universe, potentially tricking us into misinterpreting our

3 As described below, the mathematically simplest version of quan-

tum mechanics is “unitary”, lacking the controversial process
known as wavefunction collapse.
4
 sibility, the lower limit on the size of our universe has kept
growing.
5
against them. Infinite models fit the data, and strong
II. LEVEL I: REGIONS BEYOND OUR COSMIC limits have been placed on the alternatives [7, 8]. In ad-
HORIZON dition, a spatially infinite universe is a generic prediction
of the cosmological theory of inflation [6], so the striking
Let us return to your distant twin. If space is infi- successes of inflation listed below therefore lend further
nite and the distribution of matter is sufficiently uniform support to the idea that space is after all infinite just as
on large scales, then even the most unlikely events must we learned in school.
take place somewhere. In particular, there are infinitely Another loophole is that space is infinite but matter
many other inhabited planets, including not just one but is confined to a finite region around us – the historically
infinitely many with people with the same appearance, popular “island universe” model. In a variant on this
name and memories as you. Indeed, there are infinitely model, matter thins out on large scales in a fractal pat-
many other regions the size of our observable universe, tern. In both cases, almost all universes in the Level I
where every possible cosmic history is played out. This multiverse would be empty and dead. But recent ob-
is the Level I multiverse. servations of the three-dimensional galaxy distribution
and the microwave background have shown that the ar-
rangement of matter gives way to dull uniformity on large
scales, with no coherent structures larger than about 1024
A. Evidence for Level I parallel universes meters. Assuming that this pattern continues, space be-
yond our observable universe teems with galaxies, stars
Although the implications may seem crazy and and planets.
counter-intuitive, this spatially infinite cosmological
model is in fact the simplest and most popular one on
the market today. It is part of the cosmological concor- B. What are Level I parallel universes like?
dance model, which agrees with all current observational
evidence and is used as the basis for most calculations The physics description of the world is traditionally
and simulations presented at cosmology conferences. In split into two parts: initial conditions and laws of physics
contrast, alternatives such as a fractal universe, a closed specifying how the initial conditions evolve. Observers
universe and a multiply connected universe have been se- living in parallel universes at Level I observe the exact
riously challenged by observations. Yet the Level I mul- same laws of physics as we do, but with different initial
tiverse idea has been controversial (indeed, an assertion conditions than those in our Hubble volume. The cur-
along these lines was one of the heresies for which the Vat- rently favored theory is that the initial conditions (the
ican had Giordano Bruno burned at the stake in 16004), densities and motions of different types of matter early
so let us review the status of the two assumptions (infi- on) were created by quantum fluctuations during the in-
nite space and “sufficiently uniform” distribution). flation epoch (see section 3). This quantum mechanism
How large is space? Observationally, the lower bound generates initial conditions that are for all practical pur-
has grown dramatically (Figure 2) with no indication of poses random, producing density fluctuations described
an upper bound. We all accept the existence of things by what mathematicians call an ergodic random field.
that we cannot see but could see if we moved or waited, Ergodic means that if you imagine generating an ensem-
like ships beyond the horizon. Objects beyond cosmic ble of universes, each with its own random initial con-
horizon have similar status, since the observable universe ditions, then the probability distribution of outcomes in
grows by a light-year every year as light from further a given volume is identical to the distribution that you
away has time to reach us5 . If anything, the Level I mul- get by sampling different volumes in a single universe. In
tiverse sounds trivially obvious. How could space not other words, it means that everything that could in prin-
be infinite? Is there a sign somewhere saying ”Space ciple have happened here did in fact happen somewhere
Ends Here–Mind the Gap”? If so, what lies beyond it? else.
In fact, Einstein’s theory of gravity calls this intuition Inflation in fact generates all possible initial conditions
into question. Space could be finite if it has a convex with non-zero probability, the most likely ones being al-
curvature or an unusual topology (that is, interconnect- most uniform with fluctuations at the 10−5 level that
edness). A spherical, doughnut-shaped or pretzel-shaped are amplified by gravitational clustering to form galaxies,
universe would have a limited volume and no edges. The stars, planets and other structures. This means both that
cosmic microwave background radiation allows sensitive pretty much all imaginable matter configurations occur
tests of such scenarios. So far, however, the evidence is in some Hubble volume far away, and also that we should
expect our own Hubble volume to be a fairly typical one
— at least typical among those that contain observers.
A crude estimate suggests that the closest identical copy
4 Bruno’s ideas have since been elaborated by, e.g., [5, 6], all of 29 91
of you is about ∼ 1010 m away. About ∼ 1010 m away,
whom have thus far avoided the stake.
5 If the cosmic expansion continues to accelerate (currently an there should be a sphere of radius 100 light-years identical
open question), the observable universe will eventually stop grow- to the one centered here, so all perceptions that we have
ing. during the next century will be identical to those of our
 6
115
counterparts over there. About ∼ 1010 m away, there their numbers written out in binary. When asked to place
should be an entire Hubble volume identical to ours.6 bets on your room number, you assign an equal probabil-
ity for all of them. However, you can give more interest-
ing odds on what fraction of the ten binary digits on your
C. How derive probabilities from a causal theory? door will be zeros, knowing from the binomial theorem
that it’s 50% for 10 yous, 20% for 10



5 = 254 1 = 45
Let us now turn to worry 4 and worry 5 in Table 1. yous, etc. You can therefore say that you will probably
see a random-looking string of zeroes and ones on your
The Level I multiverse raises an interesting philosophical
point: you would not be able to compute your own future door, with an 89% chance that the fraction of ones will
be between 30% and 70%. This conclusion is exactly the
even if you had complete knowledge of the entire state of
the cosmos! The reason is that there is no way for you same as you would draw if you instead assumed that there
was only one of you, and that you would be placed in a
to determine which of these copies is “you” (they all feel
that they are). Yet their lives will typically begin to differ random room. Or that there was only one one you and
one room, whose 10 digits were each generated randomly
eventually, so the best you can do is predict probabilities
for what you will observe, corresponding to the fractions with 50% probability for both 0 and 1.
of these observers that experience different things. This
kills the traditional notion of determinism even without In Everett’s MWI, probability appears from random-
invoking quantum mechanics. ness in exactly the same way if the branches have equal
However, perhaps it is a uniquely quantum-mechanical amplitude: one you evolves into more than one through
phenomenon that you can end up with subjective indeter- deterministic Hamiltonian dynamics as in Equation (2)
minism even if only a single you exists to start with? No, with α = β = 2−1/2 . The only difference is what the
because we can create the same phenomenon in the fol- physical nature of the cloning process is. In our exam-
lowing simple gedanken experiment involving only classi- ple above, another difference is that the hospital guests
cal physics, without even requiring any sort of multiverse can meet and verify the existence of their clones, whereas
(not even Level I). You are told that you will be sedated, quantum clones cannot because of decoherence — how-
that a perfect clone of you will be constructed (includ- ever, the hospital experiment could easily be modified to
ing your memories), and that the two yous will be woken have this property too, say by keeping the rooms locked
up by a bell at the same time the next morning in two forever or shipping the clones off into deep space with-
identical-looking rooms. The rooms are numbered 0 and out radios. It is therefore observer cloning that is the
1, and these numbers are printed on a sign outside the crux, not what physics is involved in the cloning process.
door. When asked by the anesthesiologist to place a bet You need to end up with more than one post-experiment
on where you will wake up, you realize that you have to you with different recent experiences, but having identi-
give her 50-50 odds, because someone feeling that they cal memories from before the experiment.
are you will wake up in both. When you awaken, you re-
alize that you’d still give 50-50 odds, because even if you In summary, although these classical parallels have not
knew the position of every atom in the universe, you still ended the debates over probability in the Everett picture,
couldn’t know which of the two yous is the one having as evidenced by the continuing controversy over whether
your current subjective experience. When you go out- probability makes sense in the many worlds interpreta-
side, the room number you read will therefore feel just tion (this volume, Part 3, Part 4), they do show that
like a random number to you. worries 4 and 5 appear already in classical physics. That
Now suppose instead that you were told that this ex- is, whenever there are multiple observers with identical
periment would be repeated 10 more times, resulting in memories of what happened before a certain point but
a total of 210 clones in 1024 identical rooms which have differing afterwards, these observers will perceive appar-
ent randomness even if the evolution of their universe
is completely deterministic. Whenever an observer is
cloned, she will perceive something completely indistin-
6 This is an extremely conservative estimate, simply counting all guishable from true randomness. Since both of these phe-
possible quantum states that a Hubble volume can have that are nomena can be realized without quantum mechanics, ap-
no hotter than 108 K. 10115 is roughly the number of protons
that the Pauli exclusion principle would allow you to pack into
parent causality breakdown and randomness are there-
a Hubble volume at this temperature (our own Hubble volume fore not quantum-specific. Unitary quantum mechan-
contains only about 1080 protons). Each of these 10115 slots can ics has these attributes simply because it routinely cre-
115 115
be either occupied or unoccupied, giving N = 210 ∼ 1010 ates observer cloning when an instability rapidly ampli-
possibilities, so the expected distance to the nearest identical
115 115
fies microsuperpositions into macrosuperpositions, while
Hubble volume is N 1/3 ∼ 1010 Hubble radii ∼ 1010 meters. decoherence ensures — effectively — that the doors be-
29
Your nearest copy is likely to be much closer than 1010 meters, tween the clones are kept locked forever. Examples of
since the planet formation and evolutionary processes that have
tipped the odds in your favor are at work everywhere. There such instabilities include most quantum measurements,
are probably at least 1020 habitable planets in our own Hubble Schrödinger’s cat experiment and, probably, certain snap
volume alone. decision processes in the brain.
 7

D. How a multiverse theory can be tested and pioneers of classical statistic mechanics: in the grand en-
falsified semble at the heart of the theory, there would always be
some totally confused observers who repeatedly saw eggs
Is a multiverse theory one of metaphysics rather than unbreak and cups of water spontaneously separate into
physics? This is the concern listed as worry 1 in Table 1. steam and ice.
As emphasized by Karl Popper, the distinction between When they occur in examples not involving quantum
the two is whether the theory is empirically testable and mechanics, these issues are generally considered resolved,
falsifiable. Containing unobservable entities does clearly merely exemplifying what confidence levels are all about.
not per se make a theory non-testable. For instance, a If anybody in any context says that she has ruled some-
theory stating that there are 666 parallel universes, all of thing out at 99.9% confidence, she means that there is a 1
which are devoid of oxygen makes the testable prediction in 1000 chance that she has been fooled. Whenever there
that we should observe no oxygen here, and is therefore is any form of randomness, either ontologically funda-
ruled out by observation. mental as in the Copenhagen interpretation, apparent as
As a more serious example, the Level I multiverse in the MWI or merely epistemological (reflecting our in-
framework is routinely used to rule out theories in mod- ability to predict detector noise, say), there is a risk that
ern cosmology, although this is rarely spelled out explic- we get fooled by fluke data. In most cases, we can reduce
itly. For instance, cosmic microwave background (CMB) this risk as much as we want by performing more mea-
observations have recently shown that space has almost surements, but in some cases we cannot, say when mea-
no curvature. Hot and cold spots in CMB maps have suring the large-scale power spectrum in the cosmic mi-
a characteristic size that depends on the curvature of crowave background, where further measurements would
space, and the observed spots appear too large to be only help if we could perform them in outside of our cos-
consistent with the previously popular “open universe” mic horizon volume, i.e., in Level I parallel universes.
model. However, the average spot size randomly varies The take-home message from this section is that the
slightly from one Hubble volume to another, so it is im- MWI and indeed any multiverse theories can be tested
portant to be statistically rigorous. When cosmologists and falsified, but only if they predict what the ensemble
say that the open universe model is ruled out at 99.9% of parallel universes is and specify a probability distri-
confidence, they really mean that if the open universe bution (or more generally what mathematicians call a
model were true, then fewer than one out of every thou- measure) over it. This measure problem can be quite se-
sand Hubble volumes would show CMB spots as large rious and is still unsolved for some multiverse theories
as those we observe — therefore the entire model with (see [9–14] for recent reviews), but is solved for both sta-
frits entire Level I multiverse of infinitely many Hubble tistical mechanics and for quantum mechanics in a finite
volumes is ruled out, even though we have of course only space.
mapped the CMB in our own particular Hubble volume.
A related issue is worry 8 in Table 1: how does one
judge evidence for/against a multiverse theory, if some III. LEVEL II: OTHER POST-INFLATION
small fraction of the observers get fooled by unusual BUBBLES
data? For example, of a Stern Gerlach apparatus is used
to measure the spin in the z-direction of 10000 particles
If you felt that the Level I multiverse was large and
prepared with their spin in the x-direction, most of the
hard to stomach, try imagining an infinite set of distinct
210000 resulting observers will observe a random looking
ones (each symbolized by a bubble in Figure 1), some per-
sequence with about 50% spin-up, but one of them will
haps with different dimensionality and different physical
be unlucky enough to measure spin up every time and
constants. This is what is predicted by most currently
mistakenly conclude that quantum mechanics is incor-
popular models of inflation, and we will refer to it as
rect.
the Level II multiverse. These other domains are more
This issue clearly has nothing to do with quantum me- than infinitely far away in the sense that you would never
chanics per se, since it also occurs in our last hospital get there even if you traveled at the speed of light for-
example from Section II C. Suppose the 1024 clones are ever. The reason is that the space between our Level I
all considering the hypothesis that what happened to multiverse and its neighbors is still undergoing inflation,
them is indeed the cloning experiment as described in which keeps stretching it out and creating more volume
Section II C, trying to decide whether to believe it or faster than you can travel through it. In contrast, you
not. They all observe their room numbers, and most could travel to an arbitrarily distant Level I universe if
of them find it looking like a random sequence of zeros you were patient and the cosmic expansion decelerates.7
and ones, consistent with the hypothesis. However, one
of the clones observes the room number ”0000000000”,
and declares that the hypothesis has been ruled out at
99.9% confidence, because if the hypothesis were true, 7 Astronomical evidence suggests that the cosmic expansion is
the probability of finding oneself in the very first room is currently accelerating. If this acceleration continues, then even
only 1/1024. Similar issues also tormented some of the the level I parallel universes will remain forever separate, with the
 8

A. Evidence for Level II parallel universes There are models where also such non-integer parame-
ters can vary from one post-inflationary bubble to an-
other.9 In summary, the Level II multiverse is likely to
Inflation is an extension of the big bang theory and
be more diverse than the Level I multiverse, containing
ties up many of its loose ends, such as why the uni-
domains where not only the initial conditions differ, but
verse is so big, so uniform and so flat. An almost expo-
perhaps the dimensionality, the elementary particles and
nentially rapid stretching of space long ago can explain
the physical constants differ as well.
all these and other attributes in one fell swoop (see re-
views [15, 16]. Such stretching is predicted by a wide This is currently a very active research area. The pos-
class of theories of elementary particles, and all avail- sibility of a string theory “landscape” [18, 19], where the
able evidence bears it out. Much of space is stretching above-mentioned potential has perhaps 10500 different
and will continue doing so forever, but some regions of minima, may offer a specific realization of the Level II
space stop inflating and form distinct bubbles, like gas multiverse which would in turn have four sub-levels of
pockets in a loaf of rising bread. Infinitely many such increasing diversity: IId: different ways in which space
bubbles emerge (Figure 1, lower left, with time increas- can be compactified, which can allow both different effec-
ing upwards). Each is an embryonic Level I multiverse: tive dimensionality and different symmetries/elementary
infinite in size8 and filled with matter deposited by the articles (corresponding to different topology of the curled
energy field that drove inflation. Recent cosmological up extra dimensions). IIc: different “fluxes” (generalized
measurements have confirmed two key predictions of in- magnetic fields) that stabilize the extra dimensions (this
flation: that space has negligible curvature and that the sublevel is where the largest number of choices enter, per-
clumpiness in the cosmic matter distribution used to be haps 10500 ). IIb: once these two choices have been made,
approximately scale invariant. there may be a handful of different minima in the effec-
tive supergravity potential. IIa: the same minimum and
The prevailing view is that the physics we observe to-
effective laws of physics can be realized in a many differ-
day is merely a low-energy limit of a more general the-
ent post-inflationary bubbles, each constituting a Level I
ory that manifests itself at extremely high temperatures.
multiverse.
For example, this underlying fundamental theory may be
10-dimensional, supersymmetric and involving a grand Before moving on, let us briefly comment on a few
unification of the four fundamental forces of nature. A closely related multiverse notions. First of all, if one
common feature in such theories is that the potential en- Level II multiverse can exist, eternally self-reproducing in
ergy of the field(s) relevant to inflation has many different a fractal pattern, then there may well be infinitely many
minima (sometimes called “metastable vacuum states”), other Level II multiverses that are completely discon-
and ending up in different minima corresponds to dif- nected. However, this variant appears to be untestable,
ferent effective laws of physics for our low-energy world. since it would neither add any qualitatively different
For instance, all but three spatial dimensions could be worlds nor alter the probability distribution for their
curled up (“compactified”) on a tiny scale, resulting in properties. All possible initial conditions and symmetry
an effectively three-dimensional space like ours, or fewer breakings are already realized within each one.
could curl up leaving a 5-dimensional space. Quantum An idea proposed by Tolman and Wheeler and recently
fluctuations during inflation can therefore cause differ- elaborated by Steinhardt & Turok [20] is that the (Level
ent post-inflation bubbles in the Level II multiverse to I) multiverse is cyclic, going through an infinite series of
end up with different effective laws of physics in differ- Big Bangs. If it exists, the ensemble of such incarnations
ent bubbles — say different dimensionality or different would also form a multiverse, arguably with a diversity
types of elementary particles, like two rather than three similar to that of Level II.
generations of quarks. An idea proposed by Smolin [21] involves an ensem-
In addition to such discrete properties as dimensional- ble similar in diversity to that of Level II, but mutating
ity and particle content, our universe is characterized by and sprouting new universes through black holes rather
a set of dimensionless numbers known as physical con- than during inflation. This predicts a form of a natu-
stants. Examples include the electron/proton mass ra-
tio mp /me ≈ 1836 and the cosmological constant, which
appears to be about 10−123 in so-called Planck units. 9 Although the fundamental equations of physics are the same
throughout the Level II multiverse, the approximate effective
equations governing the low-energy world that we observe will
differ. For instance, moving from a three-dimensional to a four-
intervening space stretching faster than light can travel through dimensional (non-compactified) space changes the observed grav-
it. The jury is still out, however, with popular models predicting itational force equation from an inverse square law to an inverse
that the universe will eventually stop accelerating and perhaps cube law. Likewise, breaking the underlying symmetries of parti-
even recollapse. cle physics differently will change the lineup of elementary parti-
8 Surprisingly, it has been shown that inflation can produce an in- cles and the effective equations that describe them. However, we
finite Level I multiverse even in a bubble of finite spatial volume, will reserve the terms “different equations” and “different laws of
thanks to an effect whereby the spatial directions of spacetime physics” for the Level IV multiverse, where it is the fundamental
curve towards the (infinite) time direction [17]. rather than effective equations that change.
 9

ral selection favoring universes with maximal black hole

production.
In braneworld scenarios, another 3-dimensional world
could be quite literally parallel to ours, merely offset in a
higher dimension. However, it is unclear whether such a
world (“brane”) deserves be be called a parallel universe
separate from our own, since we may be able to interact
with it gravitationally much as we do with dark matter.

B. Fine-tuning and selection effects

Although we cannot interact with other Level II par-

allel universes, cosmologists can infer their presence in-
directly, because their existence can account for unex-
plained coincidences in our universe. To give an analogy,
suppose you check into a hotel, are assigned room 1967
and note that this is the year you were born. What a
coincidence, you say. After a moment of reflection, how-
ever, you conclude that this is not so surprising after all.
The hotel has hundreds of rooms, and you would not have
been having these thoughts in the first place if you had
been assigned one with a number that meant nothing to
you. The lesson is that even if you knew nothing about
hotels, you could infer the existence of other hotel rooms
to explain the coincidence.
As a more pertinent example, consider the mass of the
sun. The mass of a star determines its luminosity, and us- FIG. 3: Difference between Level I and Level III. Whereas
ing basic physics, one can compute that life as we know Level I parallel universes are far away in space, those of Level
it on Earth is possible only if the sun’s mass falls into III are even right here, with quantum events causing classical
the narrow range between 1.6 × 1030kg and 2.4 × 1030kg. reality to split and diverge into parallel storylines. Yet Level
III adds no new storylines beyond levels 1 or 2.
Otherwise Earth’s climate would be colder than that
of present-day Mars or hotter than that of present-day
Venus. The measured solar mass is M ∼ 2.0 × 1030 kg.
At first glance, this apparent coincidence of the habitable
and observed mass values appears to be a wild stroke of
luck. Stellar masses run from 1029 to 1032 kg, so if the sun
acquired its mass at random, it had only a small chance weak interaction were much weaker, hydrogen would not
of falling into the habitable range. But just as in the ho- exist; if it were much stronger, supernovae would fail to
tel example, one can explain this apparent coincidence by seed interstellar space with heavy elements. If the cos-
postulating an ensemble (in this case, a number of plan- mological constant were much larger, the universe would
etary systems) and a selection effect (the fact that we have blown itself apart before galaxies could form. In-
must find ourselves living on a habitable planet). Such deed, most if not all the parameters affecting low-energy
observer-related selection effects are referred to as “an- physics appear fine-tuned at some level, in the sense that
thropic” [22], and although the “A-word” is notorious changing them by modest amounts results in a qualita-
for triggering controversy, physicists broadly agree that tively different universe.
these selection effects cannot be neglected when testing
fundamental theories. In this weak sense, the anthropic
principle is not optional. Although the degree of fine-tuning is still debated (see
What applies to hotel rooms and planetary systems [23–25, 40]) for more technical reviews), these examples
applies to parallel universes. Most, if not all, of the suggest the existence of parallel universes with other val-
attributes set by symmetry breaking appear to be fine- ues of some physical constants. The existence of a Level
tuned. Changing their values by modest amounts would II multiverse implies that physicists will never be able to
have resulted in a qualitatively different universe–one in determine the values of all physical constants from first
which we probably would not exist. If protons were principles. Rather, they will merely compute probability
0.2% heavier, they could decay into neutrons, destabiliz- distributions for what they should expect to find, taking
ing atoms. If the electromagnetic force were 4% weaker, selection effects into account. The result should be as
there would be no hydrogen and no normal stars. If the generic as is consistent with our existence.
 10

IV. LEVEL III: THE MANY WORLDS OF as reviewed in [26–28]. It should be borne in mind that
QUANTUM PHYSICS these two worries remained serious open problems when
Everett first published his work, since decoherence was
If Everett was correct and physics is unitary, then there only discovered in 1970 [29].
is a third type of parallel worlds that are not far away
but in a sense right here. The universe keeps branch-
ing into parallel universes as in the cartoon (Figure 3, A. What are Level III parallel universes like?
bottom): whenever a quantum event appears to have a
random outcome, all outcomes in fact occur, one in each
branch. This is the Level III multiverse. Although more Everett’s many-worlds interpretation has been bog-
debated and controversial than Level I and Level II, we gling minds inside and outside physics for more than
will see that, surprisingly, this level adds no new types of four decades. But the theory becomes easier to grasp
universes. when one distinguishes between two ways of viewing a
Since the volume to which this chapter belongs dis- physical theory: the outside view of a physicist study-
cusses the MWI in great detail, we will summarize the ing its mathematical equations, like a bird surveying a
key points only very briefly. Everett’s MWI is simply landscape from high above it, and the inside view of an
standard quantum mechanics with the collapse postulate observer living in the world described by the equations,
removed, so that the Schrödinger equation holds without like a frog living in the landscape surveyed by the bird.
exception (Section I A). From this, the following conclu- From the bird perspective, the Level III multiverse
sions can be derived: is simple. There is only one wave function. It evolves
smoothly and deterministically over time without any
1. Microsuperpositions (say of an atom going through kind of splitting or parallelism. The abstract quantum
two slits at the same time) are inevitable (the world described by this evolving wave function contains
Heisenberg Uncertainty principle). within it a vast number of parallel classical story lines,
continuously splitting and merging, as well as a number
2. Macrosuperpositions (say of a cat being dead and
of quantum phenomena that lack a classical description.
alive) are also perfectly legitimate quantum states.
From their frog perspective, observers perceive only a
3. Processes occur that amplify microsuperpositions tiny fraction of this full reality. They can view their own
into macrosuperpositions (spontaneous symmetry Level I universe, but the process of decoherence [26, 29]
breaking, Schrödinger’s cat, and quantum measure- — which mimics wave function collapse while preserv-
ments being three examples). ing unitarity–prevents them from seeing Level III parallel
copies of themselves.
4. The superposition of a single macroscopic object Whenever observers are asked a question, make a snap
tends to spread to all other interacting objects, decision and give an answer, quantum effects in their
eventually engulfing our entire universe. brains lead to a superposition of outcomes, such as “Con-
5. Decoherence makes most macrosuperpositions for tinue reading the article” and “Put down the article”.
From the bird perspective, the act of making a decision
all practical purposes unobservable.
causes a person to split into multiple copies: one who
6. Decoherence calculations can determine which
quantities appear approximately classical.
There is consensus in the physics community that both interesting possibilities: either the symmetry is broken early on
double-slit interference and the process of decoherence while the superposition is still microscopic and unobservable (in
have been experimentally observed, showing the pre- which case the collapse process has nothing to do with measure-
ment), or the symmetry is broken later on when the superposi-
dicted behavior. Conclusions 1, 2, 3 and 4 together im- tion is macroscopic (in which case local energy conservation is
ply that astronomically large macrosuperpositions occur. seriously violated by abruptly moving the center-of-mass by a
These are Everett’s parallel universes.10 Worry 10 in Ta- macroscopic amount — even if the mass transfer is not superlu-
ble 1 is addressed by 5, and worry 11 is addressed by 6 minal, it would have to be fast enough to involve kinetic energy
greatly exceeding the natural energy scale of the problem). If this
experiment or Schrödinger’s cat experiment were performed in a
sealed free-falling box, the environment outside the box would
learn how the needle had fallen or whether the cat had died from
10 Note that to avoid creating macrosuperpositions, it is insuffi- the altered gravitational field outside the box (and perhaps also
cient to abandon unitarity. Rather, it is symmetry that must be from recoil motion of the box), causing decoherence. However,
abandoned. For example, any theory where the wavefunction of this complication can in principle be eliminated by keeping the
a system evolves deterministically (even if according to another moving parts spherically symmetric at all times. For example,
rule than the Schrödinger equation) will evolve a perfectly sharp if a metal sphere full of hydrogen contains a smaller sphere at
needle balanced on its tip into a superposition of needles pointing its center full of oxygen at the same pressure which is opened if
in macroscopically different directions unless the threory explic- an atom decays (after which diffusion would mix the gases), the
itly violates rotational symmetry. If the theory does violate this resulting superposition of two macroscopically different density
symmetry and “collapses” the wavefunction, then there are two distributions would leave all external fields unaffected.
 11

keeps on reading and one who doesn’t. From their frog able copies of the same universes–the same old story lines
perspective, however, each of these alter egos is unaware playing out again and again in other quantum branches.
of the others and notices the branching merely as a slight The passionate debate about Everett’s theory therefore
randomness: a certain probability of continuing to read seems to be ending in a grand anticlimax, with the dis-
or not. covery of less controversial multiverses (Levels I and II)
As strange as this may sound, the exact same situation that are equally large.
occurs even in the Level I multiverse. You have evidently
decided to keep on reading the article, but one of your
alter egos in a distant galaxy put down the magazine after C. The unequal probability worry
the first paragraph. The only difference between Level I
and Level III is where your doppelgaängers reside. In Let us now turn to worry 6 in Table 1: how to com-
Level I they live elsewhere in good old three-dimensional pute the apparent probabilities from the wave function
space. In Level III they live on another quantum branch amplitudes when they are not all equal and the wave-
in infinite-dimensional Hilbert space (Figure 3). function collapse postulate has been dropped from the
theory. Since a single approximately classical state often
evolves into a superposition of macroscopically different
B. Level III parallel universes: evidence & states that rapidly decohere as discussed above, it is obvi-
implications ous that observers will experience apparent randomness
just as in our hospital examples from Section II C. How-
The existence of Level III depends on one crucial as- ever, why is it that these probabilities correspond to the
sumption: that the time evolution of the wave function square modulus of the wave function amplitudes (the so-
is unitary. So far experimenters have encountered no de- called Born rule)? For example, in equation (2), why is
partures from unitarity. In the past few decades they the apparent probability for a happy observer equal to
have confirmed unitarity for ever larger systems, includ- |α2 | rather than some other real-valued function of α,
ing carbon 60 buckyball molecules and kilometer-long op- say |α|4 ?
tical fibers. On the theoretical side, the case for unitarity There are a number of arguments that suggest that it
has been bolstered by the discovery of decoherence (see must be this way. For example, one could argue that
[27] for a popular review). Some theorists who work on the sum of the probabilities should be conserved (soR that
quantum gravity have questioned unitarity; one concern it can be normalized to 1 once and for all), and |φ|2
is that evaporating black holes might destroy informa- is the only functional of ψ that is conserved under ar-
tion, which would be a nonunitary process. But a recent bitrary unitary evolution, by the very definition of uni-
breakthrough in string theory known as AdS/CFT corre- tarity. In other words, the business about the squaring
spondence suggests that even quantum gravity is unitary. comes straight from the Hilbert-space structure of quan-
If so, black holes do not destroy information but merely tum mechanics, whereby the inner product defines an L2
transmit it elsewhere. norm but no other norms.
If physics is unitary, then the standard picture of how Other arguments to this end have been proposed,
quantum fluctuations operated early in the big bang must based on information theory [2], decision theory [30] and
change. These fluctuations did not generate initial con- other approaches [28, 31, 32]. But many authors have
ditions at random. Rather they generated a quantum expressed a deeper concern about whether probability in
superposition of all possible initial conditions, which co- the usual sense even makes sense in MWI (often focused
existed simultaneously. Decoherence then caused these around some combination of worries 4 and 5). To this
initial conditions to behave classically in separate quan- end, arguments have been proposed based on Savage’s
tum branches. Here is the crucial point: the distribution approach: whatever intelligent observers actually believe,
of outcomes on different quantum branches in a given they will behave as though ascribing subjective probabil-
Hubble volume (Level III) is identical to the distribution ities to outcomes — probabilities which, as [30, 33, 34]
of outcomes in different Hubble volumes within a single showed, match the Born rule. A rigorous mathematical
quantum branch (Level I). This property of the quantum treatment of this is given by Wallace in Chapter of this
fluctuations is known in statistical mechanics as ergodic- volume.
ity. At the extensive debates about this issue at the “Ev-
The same reasoning applies to Level II. The process erett @ 50” conference at the Perimiter Institute in 2007,
of symmetry breaking did not produce a unique outcome it was clear that these purported Born Rule derivations
but rather a superposition of all outcomes, which rapidly were still controversial. Interestingly, the entire con-
went their separate ways. So if physical constants, space- troversy centered around the equal-probability case (say
time dimensionality and so on can vary among parallel α = β in equation (2)), i.e., getting probabilities in the
quantum branches at Level III, then they will also vary first place (worries 4 and 5 in Table 1). In contrast, the
among parallel universes at Level II. notion that this can be generalized to arbitrary ampli-
In other words, the Level III multiverse adds nothing tudes (worry 6 in Table 1) was fairly uncontroversial. In
new beyond Level I and Level II, just more indistinguish- summary, worry 6 is the first one in Table 1 which is
 12

truly specific to quantum mechanics, but addressing it if E. The weirdness worry

worries 4 and 5 have been settled is arguably a solved
problem. Despite all the elaborate technical and philosophical
worries about the MWI listed in Table 1, many physicists
probably find their strongest objection to the MWI not
in their brain but in their gut: it simply feels too weird,
crazy, counter-intuitive and disturbing.
D. Does the state describe the world or my The complaint about weirdness is aesthetic rather than
knowledge of it?
scientific, and it really makes sense only in the Aris-
totelian world view. Yet what did we expect? When we
A quantum state can be mathematically described by ask a profound question about the nature of reality, do we
a density matrix. But what does this density matrix not expect an answer that sounds strange? I personally
really describe? The state of the universe or your state dismiss this weirdness worry as a failure to appreciate
of knowledge about it? This issue, listed as worry 7 in Darwinian evolution. Evolution endowed us with intu-
Table 1, is as old as quantum mechanics itself and still ition only for those aspects of physics that had survival
divides the physics community. value for our distant ancestors, such as the parabolic tra-
jectories of flying rocks. Darwin’s theory thus makes the
The Everett postulate implies a clear answer to it:
testable prediction that whenever we look beyond the
both! On one hand, the entire universe has a quan-
human scale, our evolved intuition should break down.
tum state which corresponds to a wavefunction, or to
We have repeatedly tested this prediction, and the re-
a density matrix if the state is mixed. Let us call this
sults overwhelmingly support it: our intuition breaks
the ontological quantum state. On the other hand, our
down at high speeds where time slows down, on small
state of knowledge of the universe is described by a lower-
scales where particles can be in two places at once, on
dimensional density matrix for those degrees of freedom
large scales where we encounter black holes, and at high
that we are interested in, both conditioned on what we
temperatures, where colliding particles change identity.
already know (limiting to those branches that we could
To me, an electron colliding with a positron and turning
be on — what Everett termed the “relative state”) and
into a Z-boson feels about as intuitive as two colliding
partial-traced over those degrees of freedom that we know
cars turning into a cruise ship. The point is that if we
nothing about. I will refer to this as the epistemological
dismiss seemingly weird theories out of hand, we risk dis-
quantum state, bearing in mind that it differs from one
missing the correct theory if we stumble across it.
observer to the another — both from a colleague in this
branch of the wavefunction and from yourself in another
branch11 . In other words, the epistemological quantum
state is derivable from the ontological quantum state and F. Two world views
your subjective observations. When quantum textbooks
refer to “the” state, they usually mean the the epistemo- The seemingly endless debate over the interpretation
logical state of a system according to you, after you have of quantum mechanics is in a sense the tip of an ice-
prepared it in a certain way. This is further elaborated berg. In the Sci-Fi spoof “Hitchhiker’s Guide to the
in [36]. Galaxy”, the answer is discovered to be “42”, and the
hard part is finding the real question. Questions about
The density-matrix aspect of this issue is clearly
parallel universes may seem to be just about as deep as
quantum-specific. However, the dichotomy between ob-
queries about reality can get. Yet there is a still deeper
jective and subjective descriptions appears in classical
underlying question: there are two tenable but diamet-
statistical mechanics as well: an ensemble of classical
rically opposed paradigms regarding physical reality and
worlds can be completely described by a probability dis-
the status of mathematics, a dichotomy that arguably
tribution in a high-dimensional phase space, whereas
goes as far back as Plato and Aristotle, and the question
the knowledge of the world by an individual observer
is which one is correct.
is described by a probability distribution in a lower-
dimensional phase space, again computable by condition- • ARISTOTELIAN PARADIGM: The subjec-
ing (the classical equivalent of computing a relative state) tively perceived frog perspective is physically real,
and marginalizing (the classical equivalent of partial trac- and the bird perspective and all its mathematical
ing). language is merely a useful approximation.

• PLATONIC PARADIGM: The bird perspec-

tive (the mathematical structure) is physically real,
11 Whereas the ontological state might be pure and hence describ- and the frog perspective and all the human lan-
ably by a wave function, the epistemological state is generically
mixed and cannot be described by a wavefunction, only by a guage we use to describe it is merely a useful ap-
density matrix. This was pointed out already by Schrödinger proximation for describing our subjective percep-
[35] tions.
 13

What is more basic — the frog perspective or the bird its own universe with different laws. The Level IV mul-
perspective? What is more basic — human language or tiverse is compulsory, since mathematical structures are
mathematical language? Your answer will determine how not “created” and don’t exist “somewhere” — they just
you feel about parallel universes. exist. Stephen Hawking once asked, “What is it that
If you prefer the Aristotelian paradigm, you share breathes fire into the equations and makes a universe for
worry 3 in Table 1. If you prefer the Platonic paradigm, them to describe?” In the case of the mathematical cos-
you should find multiverses natural, since our feeling that mos, there is no fire-breathing required, since the point is
say the Level III multiverse is “weird” merely reflects not that a mathematical structure describes a universe,
that the frog and bird perspectives are extremely differ- but that it is a universe.
ent. We break the symmetry by calling the latter weird In a famous essay, Wigner [41] argued that “the enor-
because we were all indoctrinated with the Aristotelian mous usefulness of mathematics in the natural sciences is
paradigm as children, long before we even heard of math- something bordering on the mysterious”, and that “there
ematics - the Platonic view is an acquired taste! is no rational explanation for it”. This argument can be
In the second (Platonic) case, all of physics is ulti- taken as support for the MUH: here the utility of math-
mately a mathematics problem, since an infinitely intel- ematics for describing the physical world is a natural
ligent mathematician given the fundamental equations of consequence of the fact that the latter is a mathemat-
the cosmos could in principle compute the frog perspec- ical structure, and we are simply uncovering this bit by
tive, i.e., compute what self-aware observers the universe bit. The various approximations that constitute our cur-
would contain, what they would perceive, and what lan- rent physics theories are successful because simple math-
guage they would invent to describe their perceptions ematical structures can provide good approximations of
to one another. In other words, there is a “Theory of how an observer will perceive more complex mathemati-
Everything” (TOE) whose axioms are purely mathemat- cal structures. In other words, our successful theories are
ical, since postulates in English regarding interpretation not mathematics approximating physics, but mathemat-
would be derivable and thus redundant. In the Aris- ics approximating mathematics. Wigner’s observation is
totelian paradigm, on the other hand, there can never unlikely to be based on fluke coincidences, since far more
be a TOE, since one is ultimately just explaining certain mathematical regularity in nature has been discovered
verbal statements by other verbal statements — this is in the decades since he made it, including the standard
known as the infinite regress problem [37]. model of particle physics. Detailed discussions of the
In [38, 39], I have argued that the Platonic paradigm Level IV multiverse, what it means and what it predicts
follows logically from the innocuous-sounding External are given in [38, 40].
Reality Hypothesis (ERH) [38]: “there exists an external
physical reality completely independent of us humans”.
More specifically, [38] argues that the ERH implies the VI. DISCUSSION
Mathematical Universe Hypothesis” (MUH) that our ex-
ternal physical reality is a mathematical structure. The We have discussed Everett’s Many-Worlds Interpreta-
detailed technical definition of a mathematical structure tion of quantum mechanics in the context of other physics
is not important here; just think of it as a set of abstract disputes and the three other levels of parallel universes
entities with relations between them — familiar exam- that have been proposed in the literature. We found that
ples of mathematical structures include the integers, a only a small fraction of the usual objections to Everett’s
Riemannian manifold, and a Hilbert space. theory (summarized in Table 1) are specific to quantum
mechanics, and that all of the most controversial issues
crop up also in settings that have nothing to do with
V. LEVEL IV: OTHER MATHEMATICAL quantum mechanics.
STRUCTURES

Suppose you buy the Mathematical Universe Hypothe- A. The multiverse hierarchy
sis and believe that we simply have not found the correct
equations yet, or more rigorously, the correct mathemat- We have seen that scientific theories of parallel uni-
ical structure? Then an embarrassing question remains, verses form a four-level hierarchy, in which universes be-
as emphasized by John Archibald Wheeler: Why these come progressively more different from ours. They might
particular equations, not others? [38] argues that, when have different initial conditions (Level I), different effec-
pushed to its extreme, the MUH implies that all mathe- tive physical laws, constants and particles (Level II), or
matical structures correspond to physical universes. To- different fundamental physical laws (Level IV). It is ironic
gether, these structures form the Level IV multiverse, that Everett’s Level III is the one that has drawn the
which includes all the other levels within it. If there is most fire in the past decades, because it is the only one
a particular mathematical structure that is our universe, that adds no qualitatively new types of universes.
and its properties correspond to our physical laws, then Whereas the Level I universes join seemlessly, there are
each mathematical structure with different properties is clear demarcations between those within levels II and III
 14

caused by inflating space and decoherence, respectively. postulates: finite space, wave function collapse, ontologi-
The level IV universes are completely disconnected and cal asymmetry, etc. Our judgment therefore comes down
need to be considered together only for predicting your to which we find more wasteful and inelegant: many
future, since “you” may exist in more than one of them. worlds or many words.

B. Are parallel universes wasteful? C. Are parallel universes testable

A common argument about all forms of parallel uni- We have discussed how multiverses are not a theories
verses, including Everett’s Level III ones, is that they but predictions of certain theories, and how such the-
feel wasteful. Specifically, the wastefulness worry (#2 in ories are falsifiable as long as they also predict some-
Table 1) is that multiverse theories are vulnerable to Oc- thing that we can test here in our own universe. There
cam’s razor because they postulate the existence of other are ample future prospects for testing and perhaps rul-
worlds that we can never observe. Why should nature be ing out these multiverse theories. In the coming decade,
so wasteful and indulge in such opulence as an infinity dramatically improved cosmological measurements of the
of different worlds? Yet this argument can be turned microwave background radiation, the large-scale matter
around to argue for a multiverse. What precisely would distribution, etc., will test Level I by further constraining
nature be wasting? Certainly not space, mass or atoms the curvature and topology of space and will test level II
– the uncontroversial Level I multiverse already contains by providing stringent tests of inflation. Progress in both
an infinite amount of all three, so who cares if nature astrophysics and high-energy physics should also clarify
wastes some more? The real issue here is the apparent the extent to which various physical constants are fine-
reduction in simplicity. A skeptic worries about all the tuned, thereby weakening or strengthening the case for
information necessary to specify all those unseen worlds. Level II. If the current world-wide effort to build quantum
But an entire ensemble is often much simpler than one computers succeeds, it will provide further evidence for
of its members. This principle can be stated more for- Level III, since such computers are most easily explained
mally using the notion of algorithmic information con- as, in essence, exploiting the parallelism of the Level III
tent. The algorithmic information content in a number multiverse for parallel computation [43]. Conversely, ex-
is, roughly speaking, the length of the shortest computer perimental evidence of unitarity violation would rule out
program that will produce that number as output. For Level III. unifying general relativity and quantum field
example, consider the set of all integers. Which is sim- theory, will shed more light on Level IV. Either we will
pler, the whole set or just one number? Naively, you eventually find a mathematical structure matching our
might think that a single number is simpler, but the universe, or we will and have to abandon Level IV.
entire set can be generated by quite a trivial computer
program, whereas a single number can be hugely long.
Therefore, the whole set is actually simpler.
Similarly, the set of all solutions to Einstein’s field D. So was Everett right?
equations is simpler than a specific solution. The for-
mer is described by a few equations, whereas the latter Our conclusions regarding Table 1 do not per se argue
requires the specification of vast amounts of initial data either for or against the MWI, merely clarify what as-
on some hypersurface. The lesson is that complexity in- sumptions about physics lead to what conclusions. How-
creases when we restrict our attention to one particular ever, all the controversial issues arguably melt away if
element in an ensemble, thereby losing the symmetry and we accept the External Reality Hypothesis (ERH) [38]:
simplicity that were inherent in the totality of all the el- there exists an external physical reality completely in-
ements taken together. dependent of us humans. Suppose that this hypothesis
In this sense, the higher-level multiverses are simpler. is correct. Then the core MWI critique rests on some
Going from our universe to the Level I multiverse elimi- combination of the following three dubious assumptions.
nates the need to specify initial conditions, upgrading to
Level II eliminates the need to specify physical constants, 1. Omnivision assumption: physical reality must
and the Level IV multiverse eliminates the need to spec- be such that at least one observer can in principle
ify anything at all. The opulence of complexity is all in observe all of it.
the subjective perceptions of observers [42] — the frog
perspective. From the bird perspective, the multiverse 2. Pedagogical reality assumption: physical real-
could hardly be any simpler. ity must be such that all reasonably informed hu-
A common feature of all four multiverse levels is that man observers feel they intuitively understand it.
the simplest and arguably most elegant theory involves
parallel universes by default. To deny the existence of 3. No-copy assumption: no physical process can
those universes, one needs to complicate the theory by copy observers or create subjectively indistinguish-
adding experimentally unsupported processes and ad hoc able observers.
 15

1 and 2 appear to be motivated by little more than hu- about the MWI is whether physics is unitary or not. So
man hübris. The omnivision assumption effectively re- far, experiments have revealed no evidence of unitarity
defines the word “exists” to be synonymous with what is violation, and ongoing and upcoming experiments will
observable to us humans. Of course those who insist on test unitarity for dramatically larger systems.
the pedagogical reality assumption will typically have re- My guess is that the only issues that worried Hugh Ev-
jected comfortingly familiar childhood notions like Santa erett were 10 and 11 from Table 1, which are precisely
Claus, local realism, the Tooth Fairy, and creationism — those which were laid to rest by the subsequent discov-
but have they really worked hard enough to free them- ery of decoherence. Perhaps we will gradually get more
selves from comfortingly familiar notions that are more used to the weird ways of our cosmos, and even find its
deeply rooted? In my personal opinion, our job as sci- strangeness to be part of its charm. In fact, I met Hugh
entists is to try to figure out how the world works, not Everett the other day and he told me that he agrees —
to tell it how to work based on our philosophical precon- but alas not in this particular universe.
ceptions.
If the omnivision assumption is false, then there are
unobservable things that exist and we live in a multi- Acknowledgements: The author wishes to thank
verse. If the pedagogical reality assumption is false, then Anthony Aguirre, David Albert, Bryan Eastin, Peter
the weirdness worry (#12 in Table 1) makes no sense. Byrne, Olaf Dreyer, Mark Everett, Brian Greene, Alan
If the no-copy assumption is false, then worries 4 and 5 Guth, Seth Lloyd, George Musser, David Raub, Martin
from Table 1 are misguided: observers can perceive ap- Rees, Simon Saunders, Harold Shapiro, Lee Smolin, An-
parent randomness even if physical reality is completely ton Zeilinger, Wojciech Zurek, Alex Vilenkin and Frank
deterministic and known. In this case, these fundamen- Wilczek for stimulating discussions, Simon Saunders for
tal conceptual questions raised by the MWI will arise in detailed feedback on this manuscript, and Adrian Kent,
physics anyway, independent of quantum mechanics, and Jonathan Barrett, David Wallace and the Perimeter In-
will need to be solved — indeed, Everett, in providing stitute for hosting a very stimulating meeting on the
a coherent and intelligible account of probability even in MWI.
the face of massive copying, has blazed a trail in showing This work was supported by NSF grants AST-0071213
us how to solve them. & AST-0134999, NASA grants NAG5-9194 & NAG5-
The ERH alone settles worry 9 in Table 1, since what 11099, a grant from the John Templeton Foundation, a
is in the external reality defines what exists. In summary, fellowship from the David and Lucile Packard Foundation
if the ERH is correct, then the only outstanding question and a Cottrell Scholarship from Research Corporation.

[1] H. Everett, Rev. Mod. Phys, 29, 454 (1957) [19] L. Susskind, hep-th/0302219 (2003)
[2] N. Everett 1973, in The Many-Worlds Interpretation [20] P. J. Steinhardt and N. Turok, Science, 296, 1436 (2002)
of Quantum Mechanics, ed. B. S. DeWitt and N. Gra- [21] L. Smolin, The Life of the Cosmos (Oxford Univ. Press:
ham (Princeton: Princeton Univ. Press, available at Oxford, 1997)
[22] B. Carter 1974, in IAU Symposium 63, ed. S. Longair
http://www.pbs.org/wgbh/nova/manyworlds/pdf/dissertation.pdf)
[3] M. Tegmark, Sci. Am., 270 (5/2003), 40 (2003) (Reidel: Dordrecht)
[4] M. Tegmark, astro-ph/0302131 (2003) [23] J. D. Barrow and F. J. Tipler, The Anthropic Cosmolog-
[5] G. B. Brundrit, Q. J. Royal Astr. Soc., 20, 37 (1979) ical Principle (Clarendon Press: Oxford, 1986)
[6] J. Garriga and A. Vilenkin, Phys. Rev. D, 64, 043511. [24] C. J. Hogan, Rev.Mod.Phys., 72, 1149 (2000)
(2001b) [25] M. Tegmark, A. Aguirre, M. J. Rees, and F. Wilczek,
[7] A. de Oliveira-Costa, M. Tegmark, M. Zaldarriaga, Phys. Rev. D, 73, 023505 (2006)
and A. J. S Hamilton, astro-ph/0307282 (2003) [26] D. Giulini, E. Joos, C. Kiefer, J. Kupsch, I. O. Sta-
[8] J. Shapiro, N. J. Cornish, D. N. Spergel, and G. D. Stark- matescu, and H. D. Zeh, Decoherence and the Appear-
man, PRD, 75, 084034 (2007) ance of a Classical World in Quantum Theory (Berlin:
[9] M. Tegmark, JCAP, 2005-4, 1 (2005) Springer, 1996)
[10] J. Garriga, D. Perlov-Schwartz, A. Vilenkin, and S. [27] M. Tegmark and J. A. Wheeler, Sci.Am., 2/2001, 68-75
Winitzki, hep-th/0509184 (2005) (2001)
[11] R. Easther, E. A. Lim, and M. R. Martin, JCAP, 0603, [28] W. H. Zurek, Nature, 5, 181 (2009)
016 (2006) [29] H. D. Zeh, Found. Phys., 1, 69 (1970)
[12] A. Aguirre, S. Gratton, and M. Johnson C, PRL, 98, [30] D. Deutsch, Proc. R. Soc. London A, 455, 3129,
131301 (2007) quant-ph/9906015 (1999)
[13] R. Bousso, PRL, 97, 191302 (2006) [31] C. M. Caves, C. A. Fuchs, and R. Schack, PRA, 65,
[14] D. N. Page, JCAP, 0810, 025 (2008) 022305 (2002)
[15] A. Linde, Sci. Am., 271, 32 (1994) [32] S. Saunders, quant-ph/0211138 (2002)
[16] A. Guth and D. L. Kaiser, Science, 307, 884 (2005) [33] D. Wallace, quant-ph/0211104 (2002)
[17] M. A. Bucher and D. N. Spergel 1999, Sci. Am. 1/1999 [34] D. Wallace, quant-ph/0312157 (2003)
[18] R. Bousso and J. Polchinski, JHEP, 6, 6 (2000) [35] E. Schrödinger, Naturwissenschaften, Volume 23, 844
 16

(1935) [40] M. Tegmark, Ann. Phys., 270, 1 (1998, gr-qc/9704009)

[36] M. Tegmark, PRE, 61, 4194 (2000) [41] E. P. Wigner, Symmetries and Reflections (MIT Press:
[37] R. Nozick, Philosophical Explanations (Harvard Univ. Cambridge, 1967)
Press: Cambridge, 1981) [42] M. Tegmark, Found. Phys. Lett., 9, 25 (1996)
[38] M. Tegmark, Found.Phys., 11/07, 116 (2007) [43] D. Deutsch, The Fabric of Reality (Allen Lane: New York,
[39] M. Tegmark 2007, New Scientist, 9/15/2007, 1997)
arXiv:0709.4024