Godel's Theorem (Part II)

Author(s): Richard Wiebe

Source: The Two-Year College Mathematics Journal, Vol. 6, No. 3 (Sep., 1975), pp. 4-7
Published by: Mathematical Association of America
Stable URL: http://www.jstor.org/stable/3027162 .
Accessed: 12/04/2014 09:11
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

Mathematical Association of America is collaborating with JSTOR to digitize, preserve and extend access to
The Two-Year College Mathematics Journal.

http://www.jstor.org

This content downloaded from 137.219.5.13 on Sat, 12 Apr 2014 09:11:57 AM

All use subject to JSTOR Terms and Conditions

Mathematical

Godel's Theorem (PartII)*

RichardWiebe
P. WiebeisAssistant
ofMathematics
Richard
Professor
at SaintMary'sCollegeofCalifornia.
He has studiedat theUniversity
of Louvain(Belgium)
and theUniversity
of Munster
(West
and hisdegreesincludean M.S. in mathematics
Germany),
fromthe University
of Illinois
(Urbana)and a Ph.D.inphilosophy
fromtheUniversity
of California
(Berkeley).
Besides logicismtherewas anothercontendingview of the nature of mathematics: formalism.
Impressedby what we have called the formalnature of logic,
formalistsmaintainedthat mathematicswas essentiallya matter of syntax-a
questionofdevisingproofswhichadheredto the formalrulesconcerning
the shapes
of the symbols.For the formalists,as forothers,therewere still importantconsiderationsgoverningthe acceptabilityof the syntacticalrules forlogical proofs.
It mustbe impossibleto deduce a statementand also its negaOne was consistency.
tion fromthe axioms of logic. Anotherconsiderationconcernedcompleteness.
The
systemshouldbe completein that everylogicaltruthwhichcould be writtenusing
the symbolsofthe logicwouldbe a theorem.Progresshad alreadybeen made along
theselines.As remarkedearlier,thebasic logicused in axiomaticset theoryhad been
provedconsistent.And then,in a doctoralthesispublishedin 1930,a youngAustrian
logician,Kurt G6del, proved that this same logic was complete.Such proofsare
in the sense that they are proofsabout the language of
termedmetamathematical
mathematics-in this case about the formallanguage of logic.
And so thisis wheremattersstood in 1930. Major tasks wereto showthat PM
and axiomaticset theorywereconsistentand complete.With thisdone, eithersystem would constitutea trulyideal foundationforall of mathematics.As it was,
everyknownmathematicalproofcould be formalizedin eitherof thesesystems.If
they were complete,then, at least as far as the logicistswere concerned,every
mathematicaltruthwhichcould everbe discoveredwas a theoremforwhicha proof
in eitherof the systemsexisted.And if theycould be shownto be consistent,then
theirreliabilitywould be guaranteed.
Ill. AnOversimplification
of theIncompleteness
Theorem
Gddel'sincompleteness
theoremis generallyconcededto be theoutstandingachievementin logic in this century-a centuryin whichlogic has grownphenomenally.
Its proofis also regardedas one ofthe mostbeautifuland remarkablein the history
of mathematics.Unfortunately
it will be possible only to set forththe broadest
outlinesof its proofhere.
Gddel's theoremis about the Principia Mathematicaof Whiteheadand Russell
* PartI

intheMay1975issueofthisJOURNAL.
appeared

This content downloaded from 137.219.5.13 on Sat, 12 Apr 2014 09:11:57 AM

All use subject to JSTOR Terms and Conditions

(PM), and relatedsystemssuch as axiomaticset theoryor,in fact,any formalsystem forarithmetic.It states that in any such systemthereare two sentences-one
the negationof the other,so that one of themis logicallytrue-such that neither
can be derivedfromthe axioms.To understandhow such a theoremis possible,you
must keep in mindthe formalnatureof PM. Exactly what stringsof symbolsare
sentencesand whichsentencesare axiomscan be describedin termsofshapes; there
ofwhichstringsofsentencesare proofs.
is an equallypreciseformalcharacterization
willrecognizethat in order
Anyonewho understandsmathematicaldemonstrations
to demonstratethat no proofof certainsentencesexists,it is absolutelyessentialto
have a precisedefinitionof just what a proofis. Such a definitionexistedin the
strictestformalterms.
Gbdel'sproofcombinedtwoideas, each ofwhichhad been knownforcenturies:
arithmetization
and paradox.The firstimportantexplicitexampleofarithmetization
is analyticgeometry.Numbers,or pairs or triplesof numbers,are associated with
pointsand lines in such a way that certainstatementsabout numbersare true if
and onlyif certaincorresponding
statementsare true of the associated pointsand
between
lines.Thus certainproofsabout numbers,by virtueofthe correspondence
numbersand pointsand lines,also provecertaingeometricaltruths.Godel devised
an ingeniouscorrespondence
betweencertainnumbersand the sentencesand proofs
in PM. Hence,by provingcertainthingsabout certainnumbers,he simultaneously
provedcertainthingsabout proofsin PM.
You will recall that all mathematicalsentencesabout numberscan be translated into the formallanguage of PM, and all the truthsof arithmeticwhichhad
everbeen demonstratedcould be provedas theoremsof PM. Thus certainof these
sentencesofPM about numbersweretrueifand onlyifcertain(other) sentencesof
PM were provable in PM. In effect,then, certainarithmeticsentencesof PM
weresentencesabout othersentencesin PM and theirproofs.
The otheringredientmentionedwas paradox. Here Gbdel deviseda modification of the ancientparadox of the liar. A versionof the originalparadox goes like
this: An individualsays, "The sentenceI am now utteringis false." Now we know
that everysentenceis eithertrueor it isn't. What about our individual'ssentence?
If it is true,thenwhatit assertsis thecase, so it is false.This showsthatthesentence
is false.Consequently,what it assertsis not the case, so it is true.We have shown
boththat the sentenceis trueand that it is not true.The resolutionofthisparadox
is the topic of anotherexposition;we shall onlynoticehow G6del was inspiredby
the paradox. In an ingeniousand by no means trivialfashionhe showedthat there
was a sentenceofthe languageofPM whichsaid ofitselfthat it was not provable.
In fact,he evenshowedhow,theoretically,
thissentencecouldbe writtenout. Let us
call this sentence--B. --B is a sentenceof PM, a sentenceof formalarithmetic
whichsaysthatthereis no numberofa certainkind.By virtueofthearithmetization
ofthesyntaxofPM, --B is trueifand onlyifthereis no proofofa certainsentence,
and in this case that certainsentenceis --B itself.
It is easy to see that --B is true. --B is truebecause it says that thereis no
proofof --B, and that thereis no such proofwe now showby supposingtherewere
such a proof.If therewere such a proof,--B would be falsebecause it says that
5

This content downloaded from 137.219.5.13 on Sat, 12 Apr 2014 09:11:57 AM

All use subject to JSTOR Terms and Conditions

thereis no such proof.Hence, JOBwouldbe a falsebut provablesentence.Thereare

(presumably?)no falseprovablesentencesin PM, so we see that we cannothave a
proofof --B.
Now what about B, a sentencewhichsays that --B is provable?B is false,so
again thereis no proofof B either.Neither --B nor B is provable in PM. B
however,is a truesentenceabout numbers,and so PM is incomplete:not all true
statementsabout numberswhichcan be expressedin PM are provable.
You mighttryto make PM completeby adding --B to the axioms of PM,
but thenthenew enlargedPM would,by a slightmodification
ofGddel's argument,
have a new true and unprovablesentence,and in fact,infinitely
many true,unprovablesentencescould be added as new axiomsand the resultingenlargedsystem
would stillbe incomplete.
If a logical systemsuch as PM is inconsistent,
then everysentenceof the language ofPM is a theorem.So ifPM is inconsistent,
thenit is all too complete:both
--B and B are theorems;and, in fact, Godel in the proofof the incompleteness
theoremassumedthat PM was consistent.(Actuallyhe assumed a bit morethan
that: He assumedthat PM is w-consistent.
A fewyearslater an Americanlogician,
Rosser,showedhow the incompleteness
followsfromthe assumptionof consistency
alone. In what followsit will simplifymattersto call this slightlystrongerresult
Gddel's theorem.)So G6del's theoremestablishesthe following:
(5) If PM is consistent,
thenL-B is notprovable.

Now the sentence"PM is consistent"can be formulated

in thelanguageofPM as a
formalstatementof arithmeticvia Gbdel's arithmetization.
Let us call a sentence
ofPM whichsays that PM is consistentC. It is also possibleto say that --B is not
provablein PM. What is a sentenceof PM whichsays that? --B itself,to be sure.
Gddel's theorem,then,if it were stated in the formalismof PM, would be a very
longsymbolicsentencewhichwe are callingC -* -B. Godel,ofcourse,gave a careful
mathematicalproofof (5), and as you wouldexpect,thisentireproofcan be carried
out in PM. The resultis a proofin PM of C -*
to see
B. Now it is not difficult
that C cannotbe provedin PM, foriftherewerea proofin PM of C, we could immediatelyobtaina proofin PM of --B, but as we have seen, --B is not provablein
PM. This is how Gddelproveda secondtheoremshowingthat no proofwhichcould
in PM could everprovethat PM is consistent.
be formulated
Thus forformalsystemssuchas PM, axiomaticset theory,or any formalization
ofarithmetic,
therewillbe truesentenceswhichcan be writtenin the symbolismof
the systembut which cannot be deduced fromthe axioms. Moreover,there will
never be a satisfactoryproofof the consistencyof such a system.Put more succinctly:arithmeticis incompleteand can neverbe provedconsistent.
IV. TheSignificance
of Godel'sTheorems
In 1931,fewmathematiciansstudiedlogic,and fewphilosophersunderstoodlogic,
so Gddel's theoremsdid not cause the stirin the worldoflogicthat you mighthave
expected.Johnvon Neumann, who had made one of the axiomatizationsof set
the formalist
theoryand had been intentupon furthering
program,saw immediately

This content downloaded from 137.219.5.13 on Sat, 12 Apr 2014 09:11:57 AM

All use subject to JSTOR Terms and Conditions

that formalism'stwo main projectshad been proved impossible.He gave up the

study of foundationsand went on to do importantworkin applied mathematics.
On the otherhand,formanyyearsthetypicalreactionofthosefewmathematicians
who werenot logiciansbut who took the troubleto examineGddel's paper was to
acknowledgethat there were strangearithmeticsentenceswhich were true and
whichcouldn'tbe deduced fromPeano's axiomsbut at the same timeto feelthat
these strangesentenceswould neverbe of any interestor importanceto working
mathematicians.In this connection,we must realize that, while G6del's proofs
called special attentionto only two true but unprovablesentencesof PM, viz.,
--B and C, he mostemphaticallydid not showor even suggestthat therewerenot
others. One might quite naturallywonder if the famous conjectureknown as
Fermat's last theoremmightnot be unprovable.Most mathematicians,however,
wereunconcernedby such eventualities.
Gddel himself,however,in a paper publishedin 1947, conjecturedthat two
famousand importantsentenceswhichhad been the subject of much controversy
wereundecidableon the basis ofset theory.They werethe axiom of choiceand the
continuumhypothesis.The formeris acceptedby mostbut not all mathematicians
and is used to provemanytheoremsimportantin analysis.To citejust one example,
the Tychonoff
theoremof topologythat the productof any collectionof compact
spaces is compactis actually equivalentto the axiom of choice. There are many
other,apparentlyquite different
statements,all equivalentto the axiom of choice.
There is, moreover,a historyofunsuccessfulattemptsto deriveit fromotherprinciplesof set theory.Curiouslyenough,in set theory,even mathematicianswho accepted it were usually carefulto keep track of whichtheoremsthey could prove
withoutit and whichseemedto requireit-as iftheywereperhapsnot so sureafter
ofthe historyofthe parallelpostulate.
all ifit was true.All thisis reminiscent
The axiom of choiceis, as its name suggests,one of the axioms of set theory.
Gddel's theorems,however,apply to axiomaticset theorywithoutthe axiom of
choice (as well as to axiomaticset theorywith the axiom of choice). Moreover,
Godelhad in themeantimeprovedanotherimportantmetamathematical
result,viz.,
that the axiom of choiceis consistentwiththe otheraxiomsofset theory,i.e., that
ifthe otheraxiomsare consistent,thenthe additionofthe axiomofchoicedoes not
rendertheminconsistent.Hence it would neverbe possibleto deduce the negation
ofthe axiomofchoicefromthe otheraxioms,unlesstheywerealreadyinconsistent.
What Gddelconjecturedwas that it is also impossibleto deducethe axiomofchoice
itselffromthe others.If Gddel was right,the axiom of choicewould be one of his
undecidablesentencesand would have the same status logicallyand even historically vis-a-visthe otheraxiomsofset theorythat the parallelpostulatehas vis-a-vis
the otherpostulatesof geometry.Well, as some of you may know,recentlyPaul
Cohen, a youngmathematicianat Stanford,provedGddel's conjectureto be true.
In closing,it would be improperto leave you withthe impressionthat all the
importantrepercussions
ofGddel's theoremhave been mentionedor that theyhave
all been ofthedramaticand negativesort.Gddel'sincompleteness
theoremsand the
methodshe used in provingthemhave had and continueto have a positiveand profoundinfluenceoverthe entirefieldoffoundationsofmathematics.

This content downloaded from 137.219.5.13 on Sat, 12 Apr 2014 09:11:57 AM

All use subject to JSTOR Terms and Conditions