By Max Tegmark

ata I ,t,-.;'.':i*i11;..
iirriiirffi
''++11p

nlverse S Not just a staple

of science fiction,
other universes are
a direct irnplication
of cosmol ogical observations

Is therc a copy of you

reading this article? A person who is not you but who lives oq You will probably never see your other selves. The farthest
a planet called Earth, with misty mountains, fertile fields and you can observe is the distance that light has been able to trav-
sprawling cities, in a solar system with eight other planets? The el during the 14 billion years since the big bang expansion be-
life of this person has been identical to yours in every respect. gan. The most distant visible obiects are now abour 4 x 1026
But perhaps he or she now decides to put down this article with- meters away-a distance that defines our observable universe.
out finishing it, while you read on.
The idea of such an alter ego seems strange and implausi- ffi...Giffiiliildiffi.,3..
ffi
ble, but it looks as if we will just have to live with it, because it lFhAes oT-ihe same size centered on their planets. They are the
is supported by astronomical observations. The simplest and most sffaightforward example of parallel universes. Each uni-
most popular cosmological model today predicts that you have verse is merely a small part of a larger "multiverse."
= a twin in a galaxy about 10 to the 1028 meters from here. This By this very definition of "universe," one might expect the
= distance is so large that it is beyond astronomical, but that does notion of a multiverse to be forever in the domain of meta-
j
not make your doppelgdnger any less real. The estimate is de- physics. Yet the borderline betrveen physics and metaphysics is
rived from elementary probability and does not even assume defined by whether a theory is experimentally testable, not by
speculative modern physics, merely that space is infinite (or at whether it is weird or involvequnobservable entities. The fron-
least sufficiently large) in size and almost uniformly filled with tiers of physics have gradually expanded to incorporate ever
= matter, as observations indicate. In infinite space, even the most more abstract (and once metaphysical) conc"bpts such as a round
unlikely events must take place somewhere. There are infinite- Earth, invisible electromagnetic fields, time slowdown at high
ly many other inhabited planets, including not just one but in- , speeds, quantum superpositions, curved space, and black holes.
finitely many that have people with the same appearance, name Over the past several years the concept of a multiverse has joined
and memories as you, who play out every possible permutation this list. It is grounded in well-tested theories such as relativity
of your life choices. and quantum mechanics, and it fulfills both of the basic criteria
I

www.sctam.com sctENTtFlc AMERtcAN 3

I

I ,I
of an empirical science: it makes predictions, and it can be fal- all universes in the Level I multiverse would be empty and dead.
sified. Scie_nti":_rfbg-r t But recent observations of the three-dimensional galaxy distri-
of prrallel rniverses. The k.y -@ bution and the microwave background have shown that the
arrangement of matter gives way to dull uniformity on large
LJ::
tVerse?*isis but rather how many levels it has.
scales,with no coherent structures larger than about 1024 me-
Level l: Begond Our Cosmic Horizon ters. Assuming that this pattern continues, spacebeyond our
THE pARALLEL uNIVERSTs of your alter egos constitute the observable universe teems with galaxies, stars and planets.
Level I multiverse. It is the least controversial type. \7e all ac- Observers living in Level I parallel universes experience the
cept the existence of things that we canflot see but could see if same laws of physics as we do but with different initial condi-
we moved to a different vantage point or merely waited, like tions. According to current theories, processes early in the big
people watching for ships to come over the horizon' Objects bang spread matter around with a degree of randomness, gen-
beyond the cosmic horizon have a similar status. The observ- erating all possible arrangements with nonzero probability. Cos-
abie unirre.se grows by a light-year every year ry-UmfA;mt mologists assume that our universe, with an almost uniform dis-
. wair- tribution of matter and initial density fluctuations of one part in
your alter egos 100,000, is a fairly typical one (at least among those that con-
@ die long before
thggllrytg that
come into view, but in principle, and if cosmic expansion co- tain observers ). Ihe-t. a" SUl1pllgggp5|glltgl
operates, your descendants could observe them through a suf- your closest identical copy is 10 to the 10lt *.*rs:-3f . About
f radius
fi ciently powerful teiescope.
If anything, the Level I multiverse sounds trivially obvious' ,.qgtr
How could space notbe infinite? Is there a sign somewhere say- ,i"^r rfr"r 'u*
ing "Space Ends Here-Mind the Gap" ? If so, what lies beyond TEoGiltly"lgg[.rparts over there. About 10 to the 10118 me-
it? In fact, Einstein's theory of gravity calls this intuition into
ouestion. Soace could be finite if it has a convex curvature or Th".-rt. extremely conservative estimates, derived simplv
orrr.oJtopology (that is, interconnectedness). A spherical, by counting all possibie quantum states that a Hubble voiume
", can have if it is no hotter than 108 kelvins. One way to do the
iEd volume and no edges. The cosmic microwave background calculation is to ask how many protons could be packed into
radiation allows sensitive tests of such scenarios [see "Is Space a Hubble volume at that temperature. The answer is 10118 pro-

Finite?" by Jean-Pierre Luminet, Glenn D. Starkman and Jef- tons. Each of those particles may or may not, in fact, be present,
frey R. \7eeks; ScrnNurIc AranRtcAN, Aptil 1,999). So far, which makes for 2 to the 10118 possible arrangements of pro-
however, the evidence is against them. Infinite models fit the tons. A box containing that many Hubble volumes exhausts all
data, and strong limits have been placed on the alternatives' the possibilities. If you round off the numbers, such a box is
Anorher oossibiliry is that space is infinite but matter is con- about 10 to the 10118 meters across. Beyond that box, univers-
es-including ours-must repeat. Roughly the same number
6ned to a finire t.gion iiound-Gjhe historically popular "is-
t could be derived by using thermodynamic or quantum-gravita-
6ut on large scales in a fractal pattern. In both cases, almost tional estimates of the total information content of the universe.
Your nearest doppelgdnger is most likely to be much clos-
er than these numbers suggest, given the processes of planet for-
mation and biological evolution that tip the odds in your favor.
r 0ne of the mang implications of recent cosmological Astronomers that our Hubble volume has at least 1020
observations is that the concept of parallel universes is hi6i66Fplr"etq some mig[t we
no mere metaphor. Space appears to be infinite in size' lf I multfuerse framework is used routinely to eval-
so, then somewhere out there, evergthing that is possible uate theories in modern cosmology, although this procedure is
becomes real, no matter how improbable it is. Begond the rarely spelled out explicitly. For instance, consider how cos-
range of our telescopes are other regions of space that mologists used the microwave background to rule out a finite
are identical to ours. Those regions are a tUPe of pardllel spherical geometry. Hot and cold spots in microwave back-
universe. Scientists can even calculate how distant these ground maps have a characteristic size that depends on the cur-
universes are, on average. vature of space, and the observed spots appear too small to be :
r And that is fairlg solid phgsics. When cosmologists consider consistent with a spherical shape. But it is important to be sta-
theories that are less well established, theg conclude that tistically rigorous. The average spot size varies randomly from
other universes can have entirelu different properties'and one Hubble volume to another, so it is possible that our universe
laws of phgsics. The presence ofthose universes would is fooiingus-it could be spherical but happen to have abnor- =
explain various strange asPects of our own. lt could even mally small spots. Vhen cosmologists say they have ruled out
answer fundamental questions about the nature oftime the spherical model with 99.9 percett confidence, they realiy ;
and the comprehensibilitg of the phgsicalworld. mean that if this model were true, fewer than one in 1,000 Hub-
ble volumes would show spots as small as those we observe.

4 scrENTtFtc AMERtcAN
 Radiui ofSpace Ibillions of light'gearsJ
1 2 4 816o-+ 102s.
80
a
,c ?0
o
o a
a
7023.
^tqt
tso o
E ."""Y
.9
5so g 7021
,"."*
o
q
c
.o
i!
a
40
o
o 101e +\\'
'"u|{'
,-UNIF0RM
fr
g
iso ^$Yr
' DENsITY
o .c lon .w; I
.-\.",1 \i==
E ^i!er i o
.= .r$-r.'l
?zo
o 5-r 1 o
o a
=
o 101s
Fro
o
o
o-tl,'
F It
0 1013+,
20 52 1 0.5 0.2 t07 108 1os 1o1o 1011

Angular Scale (aegreesJ Sphere Radius I light-gears J

CoSMOL0GICAL DATA support the idea that space continues begond the or infinite Icenter). [0ne caveat: some cosmologists speculate that the
confines of our observable universe. The WMAP satellite recentlg discrepant point on the left ofthe graPh is evidence for a finite volume.J ln
measured the fluctuations in the microwave background [/e/t) The addition, WMAP and the 2dF Galaxg Redshift Surveg have found that space
strongest fluctuations are just over half a degree across, which on large scales is filled with matter uniformlU [nght), meaning that other
indicates-after applging the rules of geometru-that sPace is verg large universes should look basicallg like ours.

The lesson is that the multiverse theory can be tested and tween our bubble and its neighbors is expanding faster than you
falsified even though we cannot see the other universes. The key could travel through it. Your descendants will never see their
is to predict what the ensemble of parallel universes is and to doppelgdngers elsewhere in Level II. For the same reason, if cos-
specify a probability distribution, or what mathematicians call mic expansion is accelerating, as observations now suggest)
a "measurer" over that ensemble. Our universe should emerge they might not see their alter egos even in Level I.
as one of the most probable. If not-if, according to the multi- The Level II multiverse is far more diverse than the Level I
verse theory, we live in an improbable universe-then the the- multiverse. The bubbles vary not only in their initial conditions
ory is in trouble. As I will discuss later, this measure problem but also in seemingly immutable aspects of nature. The prevail-
can become quite challenging. ing view in physics today is that the dimensionality of spacetime,
the qualities of elementary particles and many of the so-called
Level ll: 0ther Postinflation Bubbles physical constants are not built into physical laws but are the
IF THE LEVEL I
MULTIVERSE was hard to stomach, try outcome of processes known as symmetry breaking. For in-
imagining an infinite set of distinct Level I multiverses, some stance, theorists think that the space in our universe once had
perhaps with different spacetime dimensionaliry and different nine dimensions, all on an equal footing. Early in cosmic histo-
physical constants. Those other multiverses-which constitute ry, three of them partook in the cosmic expansion and became
a Level II multiverse-are predicted by the currently popular the three dimensions we now observe. The other six are now un-
theory of chaotic eternal inflation. observable, either because they have stayed microscopic with a
Inflation is an extension of the big bang theory and ties up doughnutlike topology or because all matter is confined to a
many of the loose ends of that theory, such as why the universe three-dimensional surface (a membrane, or simply "brane") in
is so big, so uniform and so flat. A rapid stretching of space long the nine-dimensional space.
ago can explain all these and other attributes in one fell swoop Thus, the original symmetry among the dimensions broke.
[see "The Inflationary flniverse," by Alan H. Guth and Paul J.
The quantum fluctuations that drive chaotic inflation could
Steinhard; ScrBvrrrrc Aruremcen, May L984;and "The Self-Re- cause different symmetry breaking in different bubbles. Some
producing Inflationary Universe," by Andrei Linde, November might become four-dimensional, others could contain only two
1 994] . Such stretching is predicted by a wide class of theories rather than three generations of quarks, and still others might
of elementary particles, ,tl available evidence bears it out. have a stronger cosmological constant than our universe does.
""d
The phrase "chaotic eternal" refers to whdt happens on the very Another way to produce a Level II rei*iverse might be
largest scales. Space as a whole is stretching and will continue through a cycle of birth and destruction of universes. In a sci-
doing so forever, but some regions of space stop stretching and entific context, this idea was introduced by physicist Richard C.
form distinct bubbles, like gas pockets in a loaf of rising bread. Tolman in the 1930s and recently elaborated on by PaulJ. Stein-
Infinitely many such bubbles emerge. Each is an embryonic Lev- hardt of Princeton University and Neil Turok of the University
elI multiverse: infinite in size and filled with matter deposited by of Cambridge. The Steinhardt and Turok proposal and related
the energy field that drove inflation. models involve a second three-dimensional brane that is quite
Those bubbles are more than infinitely far away from Earth, literally parallel to ours, merely offset in a higher dimension [see =
in the sense that you would never get there even if you traveled "Been There, Done That," by George Musser; News Scan, Scr-
at the speed of light forever. The reason is that the space be- ENTIFIC AunnICAN, March 20021. This parallel universe is not

6 scre urrrtc AMERtcAN

 FIEI DS
ARE
UN5'lABLE

COMPLEX
STRUCTURES 1:iY
NEXE
.ANNOT EXIST
really a separate universe, because it interacts with ours. But the "Exploring Our Universe and
ues of the physical constants fsee
ensemble of universes-past, present and future-that these Others," by Martin Rees; ScrnNrlrlc AuBnIcaN; December
branes create would form a multiverse, arguably with a diver- 19991. The Level II multiverse theory predicts that physicists
sity similar to that produced by chaotic inflation. An idea pro- will neverbe able to determine the values of these constants
posed by physicist Lee Smolin of the Perimeter Institute in n7a- from first principles. They will merely compute probability dis-
terloo, Ontario, involves yet another multiverse comparable in tributions for what they should expect to find, taking selection
diversity to that of Level II but mutating and sprouting new uni- effects into account. The result should be as generic as is con-
verses through black holes rather than through brane physics. sistent with our existence.
Although we cannot interact with other Level II parallel uni-
verses, cosmologists can infer their presence indirectly, because Level !!l: Ouantum Mang Worlds
their existence can account for unexplained coincidences in our THE LEVEL I AND LEVEL II multiverses involve parallel
universe. To give an analogy, suppose you check into a hotel, worlds that are far away, beyond the domain even of as-
are assigned r oom L967 and note that this is the year you were tronomers. But the next level of multiverse is right around you.
born. What a coincidence, you say. After a moment of reflec- It arises from the famous, and famously controversial, many-
tion, however, you conclude that this is not so surprising after all. worlds interpretation of quantum mechanics-the idea that
The hotel has hundreds of rooms, and you would not have been random quantum processes cause the universe to branch into
having these thoughts in the first place if you had been assigned multiple copies, one for each possible outcome.
one with a number that meant nothing to you. The lesson is that In the early 20th century the theory of quantum mechanics
even if you knew nothing about hotels, you could infer the ex- revolutionized physics by explaining the atomic realm, which
istence of other hotel rooms to explain the coincidence. does not abide by the classical rules of Newtonian mechanics.
As a more pertinent example, consider the mass of the sun. Despite the obvious successes of the theory, a heated debate
The mass of a star determines its luminosity, and using basic rages about what it really means. The theory specifies the state
physics, one can compute that life as we know it on Earth is of the universe not in classical terms, such as the positions and
possible only if the sun's mass falls into the narrow range be- velocities of all particles, but in terms of a mathematical ob-
tween 1.6 x 1030 and 2.4 x 1,030 kilograms. Otherwise Earth's ject called a wave function. According to the Schrodinger equa-
climate would be colder than that of present-day Mars or hot- tion, this state evolves over time in a fashion that mathemati-
ter than that of present-day Venus. The measured solar mass cians term "unitary," meaning that the wave function rotates
is 2.0 x 1030 kilograms. At first glance, this apparent coinci- in an abstract infinite-dimensional space called Hilbert space.
dence of the habitable and observed mass values appears to be Although quantum mechanics is often described as inherently
a wild stroke of luck. Stellar masses run from 1O2e to 1032 kilo- random and uncertain, the wave function evolves in a deter-
grams, so if the sun acquired its mass at random, it had only a ministic way. There is nothing random or uncertain about it.
small chance of falling into the habitable range. But just as in The sticky part is how to connect this wave function with
the hotel example, one can explain this apparent coincidence what we observe. Many legitimate wave functions correspond
by postulating an ensemble (in this case, a number of planetary to counterintuitive situations, such as a cat being dead and alive
systems) and a selection effect (the fact that we must find our- at the same time in a so-called superposition. In the 1920s
selves living on a habitable planet). Such observer-related se- physicists explained away this weirdness by postulating that the
lection effects are referred to as "anthropic," and although the wave function "collapsed" into some definite classical outcome
"A-word" is notorious for triggering controversy, physicists whenever someone made an observation. This add-on had the
broadly agree that these selection effects cannot be neglected virtue of explaining observations, but it turned an elegant, uni-
when testing fundamental theories. tary theory into a kludgy, nonunitary one. The intrinsic ran-
'lfhat domness commonly ascribed to quantum mechanics is the re-
applies to hotel rooms and planetary systems applies
to parallel universes. Most, if not all, of the attributes set by sult of this postulate.
symmetry breaking appear to be fine-tuned. Changing their val- Over the years many physicists have abandoned this view
ues by modest amounts would have resulted in a qualitatively in favor of one develo ped in L9 57 by Princeton graduate stu-
different universe-one in which we probably would not ex- dent Hugh Everett IIL He showed that the collapse postulate
ist. If protons were 0.2 percent heaviep, they could decay into is unnecessary. Unadulterated quantum theory does not, in fact,
neutrons, destabilizing atoms. If the electromagnetic force were pose any contradictions. Although it predicts that one classi-
4 percent weaker, there would be no hydrogen and no normal calreality gradually splits into superpositions of many such re-
stars. If the weak interaction were much weaker, hydrogen alities, observers subjectively experience this splitting merely as
would not exist; if it were much stronger, supernovae would a slight randomness, with probabilities in exact agreement with
fail to seed interstellar space with heavy elements. If the cos- those from the old collapse postulate. This superposition of
=
mological constant were much larger, the universe would have classical worlds is the Level III multiverse. =
blown itself apart before galaxies could form. Everett's many-worlds interpretation has been boggling ;
Although the degree of fine-tuning is still debated, these ex- minds inside and outside physics for more than four decades'
amples suggest the existence of parallel universes with other val- But the theory becomes easier to grasp when one distinguishes

8 scrrurrrrc AMERIcAN
between two ways of viewing a physical theory: the outside Tegmark and John Archibald'Wheeler; ScIrNnntc AratntceN,
view of a physicist studying its mathematical equations, like a February 2001]. Some theorists who work on quantum graviry
bird surveying a landscape from high above it, and the inside have questioned unitarity; one concern is that evaporating black
view of an observer living in the world described by the equa- holes might destroy information, which would be a nonunitary
tions, Iike a frog living in the landscape surveyed by the bird. process. But a recent breakthrough in string theory known as
From the bird perspective, the Level III multiverse is simple. AdS/CFT correspondence suggests that even quantum gravity is
There is only one wave function. It evolves smoothly and de- unitary. If so, black holes do not destroy information but mere-
terministically over time without any kind of splitting or par- ly transmit it elsewhere.
allelism. The abstract quantum wodd described by this evolv- If physics is unitary, then the standard picture of how quan-
ing wave function contains within it a vast number of parallel tum fluctuations operated early in the big bang must change.
classical story lines, continuously splitting and merging, as well These fluctuations did not generate initial conditions at ran-
as a number of quantum phenomena that lack a classical de- dom. Rather they generated a quantum superposition of all
scription. From their frog perspective, observers perceive only possible initial conditions, which coexisted simultaneously. De-
a tiny fraction of this full reality. They can view their own Lev- coherence then caused these initial conditions to behave clas-
el I universe, but a process called decoherence-which mimics sically in separate quantum branches. Here is the crucial point:
wave function collapse while preserving unitarity-prevents the distribution of outcomes on different quantum branches
them from seeing Level III parallel copies of themselves. in a given Hubble volume (Level III) is identical to the distrib-
n7henever observers are asked a question, make a snap deci- ution of outcomes in different Hubble volumes within a single
sion and give an answer, quantum effects in their brains lead to quantum branch (Level I). This property of the quantum fluc-
a superposition of outcomes, such as "Continue reading the ar- tuations is known in statistical mechanics as ergodicity.
ticle" and "Put down the article." From the bird perspective, the The same reasoning applies to Level II. The process of sym-
act of making a decision causes a person to split into multiple metry breaking did not produce a unique outcome but rather
copies: one who keeps on reading and one who doesn't. From a superposition of all outcomes, which rapidly went their sep-
their frog perspective, however, each of these alter egos is un- arate ways. So if physical constants, spacetime dimensionality
aware of the others and notices the branching merely as a slight and so on canvary among parallel quantum branches at Level
randomness: a certain probability of continuing to read or not. III, then they will also vary among parallel universes at Level II'
As strange as this may sound, the exact same situation oc- In other words, the Level III multiverse adds nothing new
curs even in the Level I multiverse. You have evidently decided beyond Level I and Level II, just more indistinguishable copies
to keep on reading the article, but one of your alter egos in a of the same universes-the same old story lines playing out
distant galaxy put down the magazite after the first paragraph. again and again in other quantum branches. The passionate de-
The only difference between Level I and Level III is where your bate about Everett's theory therefore seems to be ending in a
doppelgdngers reside. In Level I they live elsewhere in good old grand anticlimax, with the discovery of less controversial mul-
three-dimensional space. In Level III they live on another quan- tiverses (Levels I and II) that are equally large.
tum branch in infinite-dimensional Hilbert space. Needless to say, the implications are profound, and physi-
The existence of Level III depends on one crucial assump- cists are only beginning to explore them. For instance, consid-
tion: that the time evolution of the wave function is unitary. So er the ramifications of the answer to a long-standing question:
far experimenters have encountered no departures from unitar- Does the number of universes exponentially increase over time?
ity. In the past few decades they have confirmed unitarity for The surprising answer is no. From the bird perspective, there is
ever larger systems, including carbon 60 buckyball molecules of course only one quantum universe. From the frog perspective,
and kilometer-long optical fibers. On the theoretical side, the what matters is the number of universes that are distinguishable
case for unitarity has been bolstered by the discovery of deco- at a given instant-that is, the number of noticeably different
herence [see "100 Years of Quantum Mysteries," by Max Hubble volumes. Imagine moving planets to random new lo-
cations, imagine having married someone else, and so on. Atthe
MAX T€6MARK wrote a four-dimensional version of the comPuter quantum level, there are 10 to the 10118 universes with temper-
game Tetris while in college. ln another universe, he went on to be- atures below 108 kelvins. That is a vast number, but a finite one.
come a highlg paid software developer..ln our universe, however, From the frog perspective, the evolution of the wave func-
he wound up as professor of phgsics and astronomg at the Uni- tion corresponds to a never-ending sliding from one of these 10
versitg of Pennsglvania. Tegmark is an expert in analgzing the to the 10118 states to another. Now you are in universe A, the
cosmic microwave background and galaxg clustering. Much ofhis one in which you are reading this sentence. Now you are in uni-
work bears on the concePt of parallel universes: evaluating evi- verse B, the one in which you are reading this other sentence.
dence for infinite space and cosmological inflation; developing in' Put differently, universe B has an observer identical to one in
sights into quantum decoherence; and studging the possibilitg universe A, except with an extra instant of memories. AII pos-
that the amplitude of microwave background fluctuations, the di- sible states exist at every instant, so the passage of time may be
mensionalitg of spacetime and the fundamental laws of phgsics in the eye of the beholder_afl idea explored in Greg Egan's
can varU from place to place. 1994 science-fiction novel Permutation City anddeveloped by

10 scrENTrFrc AMERtcAN
 The Mgsterg of Probabilitg:

What Are the 0dds?

AS MULTIVERSE THE0RIES gain credence, the sticku issue of howto becomes severe at Level ll, has caused much debate at Level lll,
compute probabilities in phgsics is growing from a minor nuisance and is horrendous at Level lV. At Level ll, for instance, Alexander
into a major embarrassment. lf there are indeed mang identical Vilenkin ofTufts Universitg and others have published predictions
copies of gou, the traditional notion of determinism evaPorates. for the probabilitg distributions of various cosmological
You could not compute Uour own future even if gou had complete parameters. Theg have argued that different parallel universes that
knowledge ofthe entire state ofthe multiverse, because there is no have inflated bg different amounts should be given statistical
wag for gou to determine which of these copies is gou [theg all feel weights proportional to their volume. 0n the other hand, ang
theg areJ. All gou can predict, therefore, are probabilities forwhat mathematician willtell gou that 2 x oo * oo, so there is no objective
gou would observe. lf an outcome has a probabilitg of, sag, 50 sense in which an infinite universe that has expanded bg a factor of
percent, it means that halfthe observers observe that outcome. two has gotten Iarger. Moreover, a finite universe with the topologg
Unfortunatelg, it is not an easg task to compute what fraction of a torus is equivalent to a perfectlg periodic universe with infinite
of the infinitelg mang observers perceive what. The answer volume, both from the mathematical bird perspective and from the
depends on the order in which gou count them. Bg analogg, the frog perspective of an observer within it. So whg should its infinitelg
fraction ofthe integers that are even is 50 percent if gou order smaller volume give it zero statistical weight? After all, even in the
them numericallg [ 1, 2, 3, 4, . .. ] but approaches 100 percent if gou Level I multiverse, Hubble volumes start repeating Ialbeit in a
sort them digit bg digit, the wag gour word processor would [ 1, 10, random order, not periodicallgJ after about 10 to the 10118 meters.
100, 1,000, .. . J. When observers reside in disconnected universes, lf Uou think that is bad, consider the problem of assigning
there is no obviouslg naturalwag in which to order them. lnstead statisticalweights to different mathematical structures at Level lV.
one must sample from the different universes with some statistical The fact that our universe seems relativelg simple has led mang
weights referred to bg mathematicians as a "measure." people to suggest that the correct measure somehow involves
This problem crops up in a mild and treatable manner at Level l, complexitg. -M.T.

physicist David Deutsch of the University of Oxford, indepen- discover mathematical structures rather than create them.
dent physicist Julian Barbour, and others. The multiverse There are two tenable but diarnetrically opposed paradigms
framework may thus prove essential to understanding the na- for understanding the correspondence between mathematics
ture of time. and physics, a dichotomy that arguabl,v goes as far back as Pla-
to and Aristotle. According to the Aristotelian paradigm, phys-
Level IV: Other Mathematical Structures ical reality is fundamental and mathematical language is mere-
THE INITIAL coNDITIoNs and physical constants in the ly a useful approximation. According to the Platonic paradigm,
Level I, Levei II and Level III multiverses can vary, but the the mathematical str:uctr.rre is the true reality and observers per-
fundamental laws that govern nature remain the same. Why ceive it imperfectly. In other words, the two paradigms disagree
stop there? Why not a1low the laws themselves to vary? How on which is more basic, the frog perspective of the observer or
about a universe that obeys the laws of classical physics, with the bird perspective of the physical 1aws. The Aristotelian par-
no quantum effects? How about time that comes in discrete adigm prefers the frog perspective, whereas the Platonic para-
steps, as for computers, instead of being continuous? How digm prefers rhe bird perspective.
about a universe that is simply an empty dodecahedron? In the As children, long before we had even heard of mathemat-
Level IV multiverse, all these alternatir,e realities actually exist. ics, we were all indoctrinated with the Aristotelian paradigm.
A hint that such a multiverse might not be just some beer- The Platonic view is an acquired taste. Modern theor:etical
fueled speculation is the tight correspondence between the physicists tend to be Platonists, suspecting that mathematics de-
worlds of abstract reasoning and of observed reality. Equations scribes the universe so well because the universe is inherently
and, more generally, mathematical structures such as numbers, mathematical. Then a1l of physics is ultimately a mathematics
vectors and geometric objects describe the world with remark- problem: a mathematician with unlimited intelligence and re-
able verisimilitude. In a famous 1959 lecture, physicist Eugene sources could in principle compute the frog perspective-that
P. $Tigner argued that "the enormous usefulness of mathemat- is, compute what self-aware observers the universe contains,
ics in the natural sciences is something bordering on the mys- what they perceive, and w'hat languages they invent to describe
terious." Conversely, mathematical structures have an eerily their perceptions to one another.
real feel to them. They satisfy a central criterion of objective ex- A mathematical structure is an abstract, immutable entity
istence: thel, are the same no matter who studies them. A the- existing outside of space and time. If history were a movie, the
orem is true regardless of whether it is proved by a human, a structure would correspond not to a single frame of it but to the
computer or an intelligent doiphin. Contemplative alien civi- entire videotape. Consider, for example, a world made up of
lizations would find the same mathematical structures as we pointlike particles moving around in three-dimensional space.
have. Accordingiy, mathematicians commonly say that they In four-dimensional spacetime-the bird perspective-these

www-sciam.com SCIENTIFICAIVERICAN TT
particle traiectories resemble a tangle of spaghetd. If the frog time. Most of them are probably devoid of observers. This hy-
sees a particle moving with constant veiocity, the bird sees a pothesis can be viewed as a form of radical Platonism, assert-
straight strand of uncooked spaghetti. If the frog sees a pair of ing that the mathematical structures in Plato's realm of ideas or
orbiting particles, the bird sees two spaghetti strands inter- the "mindscape" of mathematician Rudy Rucker of San Jose
twined like a double helix. To the frog, the world is described State University exist in a physical sense. It is akin to what cos-
by Newton's laws of motion and gravitation. To the bird, it is mologist John D. Barrow of the University of Cambridge refers
described by the geometry of the pasta-a mathematical struc- to as "rt in the sky," what the late Harvard University philoso-
ture. The frog itself is merely a thick bundle of pasta, whose pher Robert Nozick called the principle of fecundity and what
highly complex intertwining corresponds to a cluster of parti- the late Princeton philosopher David K. Lewis called modal re-
cles that store and process information. Our universe is far alism. Level IV brings closure to the hierarchy of multiverses, be-
more complicated than this example, and scientists do not yet cause any self-consistent fundamental physical theory can be
know to what, if any, mathematical structure it corresponds. phrased as some kind of mathematical structure.
The Platonic paradigm raises the question of why the uni- The Level IV multiverse hypothesis makes testable predic-
verse is the way it is. To an Aristotelian, this is a meaningless tions. As with Level II, it involves an ensemble (ir-r this case, the :
question: the universe just is. But a Platonist cannot help'but full range of mathematical structures) and selection effects. As
r,vonder why it could not have been different. If the universe is mathematicians continue to categorize mathematical struc- ;
inherently mathematical, then why was only one of the many tures, they should find that the structure describing our world
mathematical structures singled out to describe a universe? A is the most generic one consistent with our observations. Sim-
:
fundamental asymmetry appears to be built into the very heart ilarly, our future observations should be the mosr generic ones
of reality. that are consistent with our past observations, and our past ob-
As a way out of this conundrum, I have suggested that com- servations should be the most generic ones that are consistent
plete mathematical symmetry holds: that all mathematical struc- with our existence. =
2
tures exist physically as well. Every mathematical structure cor- Quantifying what "generic" means is a severe problem, and
responds to a parallel universe. The elements of this multiverse this investigation is only now beginning. But one striking and
do not reside in the same space but exist outside of space and encouraging feature of mathematical structures is that the sym-

LZ scrENTrFrc AN4ERTcAN
metry and invariance properties that are responsible for the you might think that a single number is simpler, but the entire
simplicity and orderliness of our universe tend to be generic, set can be generated by quite a trivial computei program,
more the rule than the exception. Mathematical structures tend whereas a single number can be hugely long. Therefore, the
to have them by default, and complicated additional axioms whole set is actually simpler.
must be added to make them go away. Similarly, the set of all solutions to Einstein's field equations
is simpler than a specific solution. The former is described by
What Sags 0ccam? a few equations, whereas the latter requires the specification of
TH E s c IENTI F I c rH E o Rr rs of parallel universes, therefore, vast amounts of initial data onsome hypersurface. The lesson
form a four-level hierarchy, in which universes become pro- is that complexity increases when we restrict our attention to
gressively more different from ours. They might have different one particular element in an ensemble, thereby losing the sym-
initial conditions (Level I); different physical constants and par- metry and simplicity that were inherent in the totality of all the
ticles (Level II); or different physical laws (Level [V). It is iron- elements taken together.
ic that Level III is the one that has drawn the most fire in the In this sense, the higher-level multiverses are simpler. Go-
past decades, because it is the only one that adds no qualita- ing from our universe to the Level I multiverse eliminates the
tively new types of universes. need to specify initial conditions, upgrading to Level II elimi-
In the coming decade, dramatically improved cosmological nates the need to specify physical constants, and the Level IV
measurements of the microwave background and the large- multiverse eliminates the need to specify anything at all. The
scale matter distribution will support or refute Level I by fur- opulence of complexity is all in the subjective perceptions of ob-
ther pinning down the curvature and topology of space. These servers-the frog perspective. From the bird perspective, the
measurements will also probe Level II by testing the theory of multiverse could hardly be any simpler.
chaotic eternal inflation. Progress in both astrophysics and The complaint about weirdness is aesthetic rather than sci-
high-energy physics should also clarify the extent to which entific, and it really makes sense only in the Aristotelian world-
physical constants are fine-tuned, thereby weakening or view. Yet what did we expect? Vhen we ask a profound ques-
strengthening the case for Level II. tion about the nature of reality, do we not expect an answer
If current efforts to build quantum computers succeed, they that sounds strange? Evolution provided us with intuition for
will provide further evidence for Level III, as they would, in the everyday physics that had survival value for our distant an-
essence, be exploiting the parallelism of the Level III multiverse cestors, so whenever we venture beyond the everyday world,
for parallel computation. Experimenters are also looking for we should expect it to seem bizarre.
evidence of unitarity violation, which would rule out Level III. A common feature of all four multiverse levels is that the
Finally, success or failure in the grand challenge of modern simplest and arguably most elegant theory involves parallel uni-
physics-unifying general relativiry and quantum field theory- verses b1, default. To deny the existence of those universes, one
will sway opinions on Level IV. Either we will find a mathe- needs to complicate the theory by adding experimentally un-
matical structure that exactly matches our universe, or we will supported processes and ad hoc postulates: finite space, wave
bump up against a limit to the unreasonable effectiveness of function collapse and ontological asymmetry. Our judgment
mathematics and have to abandon that level. therefore comes down to which we find more wasteful and in-
So should you believe in parallel universes? The principal elegant: many worlds or many words. Perhaps we will gradu-
arguments against them are that they are wasteful and that they ally get used to the weird ways of our cosmos and find its
are weird. The first argument is that multiverse theories are vul- strangeness to be part of its charm. m
nerable to Occam's razor because they postulate the existence
of other worlds that we can never observe. \7hy should nature
be so wasteful and indulge in such opulence as an infinity of dif-
ferent worlds? Yet this argument can be turned around to ar- Whg ls the CMB Fluctuation Level 10-s? Max Tegmark and Martin Rees in
gte for a multiverse. Vhat precisely would nature be wasting? Astrophgsicol Journol V01.499, No. 2, pages 526-532; June 1, 1998.
Available online at arxiv.orglabs/astro-ph/9709058
Certainly not space, mass or atoms-the uncontroversial Lev-
ls "The Theorg of Evergthing" MerelU the Ultimate Ensemble Theorg?
el I multiverse already contains an infinite amount of all three,
Max Tegmark in / nnols oJ Phgsics,Vol.27O, N0.1, pages 1-5 1;
so who cares if nature wastes some more? The real issue here November 20, 1998. Available online at arxiv.orglabs/gr-qc/97O4OO9
is the apparent reduction in simplicity. A skeptic worries about Mang Worlds in One. Jaume Garriga and AlexanderVilenkin inPhgsicol
all the information necessary to specify all those unseen worlds. Review,Yol. D64, No. 043511; Julg 26, 2001. Available online at
But an entire ensemble is often much simpler than one of its arXiv.org/abs/gr-qc/01020 10
members. This principle can be stated more formally using the 0ur Cosmic Habitat. Martin Rees. Princeton Universitg Press, 2001.
notion of algorithmic information content. The algorithmic in- lnflation, 0uantum Cosmologg and the Anthropic Principle. Andrei Linde
in Science ond Ultimote Reolitg: From Auontum to Cosmos. Edited bg J. D.
formation content in a number is, roughly speaking, the length
Barrow, P.C.W. Davies and C. L. Harper. Cambridge Universitg Press, 2003.
of the shortest computer program that will produce that num- Available online at arxiv.org/abs/hep-th/02 1 1048
ber as output. For example, consider the set of all integers. information at
The author's Web site has more
\7hich is simpler, the whole set or just one number? Naively, www.hep.upe nn.edu/- m axlmu ltiverse. ht ml

www.sciam.com SCIENTIFIC AMERICAN 13

 \ b
r.\J Nr
l^
\c-.
\\ v q
\"
\ 0(],
tr
a\
N
c\ \i
\l
. c..

\t \D
\i
c,
\)
c.
\) G
\\ xO
N c)
o\
(:,
\\>
U a.

.J $
.N
b t)
\. \.)
s .D
L c..

s \D
C>

c
a\l
\
D
i.'
p,\.t
O
Or

\o
o
o
o

SCIENTIFIC CopUright O 2007, 2003 bg Scientific American, lnc. All rights reserved. Printed.in the U.5.A. No paft ofthis reprint mag be reproduced
cal,
bg ang mechani-
pholographic or electronic process, or in the form of t phonographic recording, nor roU it be stored in a rltrieval igstem, trensmft;d o;othemise
copied for public or private use without written perm issio n ofthe p-ublisher. The traidema rk ind trade name "scl ENIIFIC AI,,|ERl
CAN" and the distinctive lo-

AMERICAN
gotUPe pertaining thereto are the sole PropertU of and are registered underthe name ofscientific American, lnc. page numbers
and internal references
mag varU from those in the original issues.