100% found this document useful (1 vote)
532 views12 pagesWhat Numbers Could Not Be (Paul Benacerraff) PDF
Uploaded by
Ignacio Benavides JimenezCopyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
100% found this document useful (1 vote)
532 views12 pagesWhat Numbers Could Not Be (Paul Benacerraff) PDF
Uploaded by
Ignacio Benavides JimenezCopyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
/ 12
What numbers could not be
PAUL BENACERRAF
‘he attention ofthe mathematica fess primar
faces primary upon mathe
matical srt, ands ill dh ares Gn pa) fom
‘in that a ven theory vb so ant sich a same, fom
‘rin how one srs model" inane, or aon
‘ome new struct an shoving Row it elas o previo sted
hes. Bethe athena wate slong o be has ane
“ties oF “objets” (rt or mambo Tanto of
sac" “pois") 0 Work with snd edo ot ingle in
their inner character or ontological status. i.
The phsepica lin, on theo hand, s more serie 0
sates of ontology and wil te ep inersed the tnd or
Wino ents there realy He wil ot Be ied th
‘ein tidmee tht cand ch cls et sch and such
satheaticalsructire. He wil wish tongue move dee io
‘ha theese ow the rele oe ens Ao he
‘ilwishoask wheter te ent Get wih su geno her
icin sone see reat (or continents oa,
ethane more fundaretletes
kat aac, Intention and Decision
Wen bn. ou dea
ony wht ey
th ante opto cnt
soda abet ee aon
be mame st blog ote soe
sneer fey et ef fies
comp ase ube es Crearooonpee et
‘ther ht same flr conga af Gate sae ak.
= ttc, The Foundation of Arithmete
1. The education
Imagine Emie and Johnny, sons of two militant k
nag Sons of two militant logicists ~ children who
‘have not been taueht in the vulgar (old-fashioned) way but for whom the
Mach of be work
MESheAR nah on spews one whi be aur bd «Proce and Gable
at Princeton Unies. Ths pet scksowede, Lam ace
m
What numbers could not be
pedagogical order of things has been the epistemological oder. They did
hot learn straight off how to count. Instead of beginning their mathe-
matical training with arithmetic as ordinary men know it, they first
learned logic ~ in their case, actually set theory. Then they were told
about the numbers. But (0 tell pople in their position about the num
bers was an easy task ~ very much lke the one which faced Monsieur
Jourdain’s tutor (who, oddly enough, was a philosopher) The parents of
‘our imagined children needed oaly to point out what aspect or part of
‘what the children already knew, under other names, was what ordinary
people called “numbers.” Learning the numbers merely involved
learning new names for familiar sets. Old (set-theoretic) truths took on
new (number-theoreti) clothing.
‘The way in which this was done will, however, bear some scrutiny and
reexamination, To facilitate the exposition, I will concentrate on Ernie
fad follow his arithmetical education tots completion. 1 wil then return
to Jobnny.
It might have gone as follows. Ernie was told that there was a set
whose members were what ordinary people refered to asthe (natural)
‘numbers, and that these were what he had Known all along as the ele-
rents of the (infinite) set U. He was further told that there was relation
defined on these “numbers” (henceforth I shall usually omit the shudder
‘quotes, the less-han relation, under which the numbers were well
Ordered. This relation, he learned, was really the one, defined on 3, for
which he had always used the letter “R."” And indeed, speaking incu
tively now, Ernie could verify that every nonempty subset of BE con-
tained a “least element - tha is, ne that bore R to every other member
of the subset, He could also show that nothing bore R to ise, and that
R was transitive, antisymmetric, irrflerive, and connected in 2. In
‘hort, the elements of 2 formed a progression, or series, under R.
‘And then there was 1, the smallest number (for reasons of future con-
venience we are ignoring 0). Ernie learned that what people had been
feferring to as I was realy the element aof SL the first, oF least, clement
fof SU under R. Talk about "successors" (each number is said to have
‘one) was easly translated in terms of the concept of the “next” member
‘of St (under R). At this point, it was no trick to show that the assump-
tions made by ordinary mortals about numbers were in Fact theorems for
Ernie, For on the basis of his theory, he could establish Peano’s axioms ~
‘an advantage which he enjoyed over ordinary mortals, who must more oF
‘o Pal Ziff for hi pl comments onan eater drat of thi paps
Reprtad with tne Kind pom of the eos from the Phiosphia Reiew
mise
m less take them as given, or self-evident, or meaningless-but-useful, of
‘what have you."
‘There are two more things that Erie had to learn before he could truly
be sad to be able to speak with the vulgar. It had to be pointed out 0
him which operations on the members of X were the ones referred 10 a8
“‘addition,”” “multiplication,” “exponentiation,” and so forth. And
here again he was in a position of epistemological superiority. For where
as ordinary folk had to introduce such operations by recursive definition,
4 euphemism for postulation, he was in position to show that these
‘operations could be expiicty defined. So the additional postulates
assumed by the number people were also shown to be derivable in his
theory, once it was seen Which set-theoretic operations addition, multi-
plication, and so forth really are.
‘he last element needed to complete Ernie's education was the expla-
‘nation of the applications of these devices: counting and measurement.
Fr they employ concepts beyond those as yet introduced. But forts-
nately, Emie was ina position to see what it was that he was doing that
corresponded to these activities (we will concentrate on counting, assum-
Ing that measurement can be explained either similarly or in terms of
counting).
There are two kinds of counting, corresponding to transitive and
intransitive uses of the verb “to count.” In one, “counting” admits of 8
direct object, asin “counting the marbles”: in the other it does not. The
‘ase have in mind is that of the preoperative patient being prepared for
the operating room. The ether mask is placed over his face and he i told
{0 count, as far as he can. He has not been instructed to count anything
at all, He has merely been told to count. A likely story is that we norm
ally learn the first few numbers in connection with sets having that
‘number of members ~ that is, in terms of transitive counting (hereby
learning the use of numbers) and then learn how to generate "the rest”
of the numbers. Actually, “the est” always remains relatively vague
matter, Most of us simply learn that we wll never run out, that our nota-
tion will extend as far as we will ever need to count. Learning these
Words, and how fo repeat them in the right order, is learning intransitive
counting. Learning their use as measures of sets is learning transitive
‘counting. Whether we learn one kind of counting before the other is
{immaterial so far asthe initial numbers are concerned. What is certain,
‘and not immaterial, is that we will have to learn some recursive pro:
‘edure for generating the novation in the proper order before we have
learned to count transitively, or to do the latter is, either directly OF
"he det fhe pool ee not dein ws
24
Whar numbers could not be
we sat uy Ie te stp
cnn sa ence ex of ee
Sete tntts ates ees Postal span nae
metaphorically) its elements one by one as we say the numbers one by
see ea tmnt au nan we
eC etS Gon neaenenva me entices come ch
cane cde teeaiecemn tite ene
miata tant etch “pining osm
2h doa te machine sey stow fy
eaten cr wraps as tae
seen aes em cia went ad
St raccoons wh a
ries sol
sc Sten and haa mai cn bead
se eel atandin on ac mene
Maas e scare anlputy en pete
ean tens ray sb We
crete tote peta act ct
sper erin a Sara
Ea oy cee htttin onener rg
esa iste opisen ae
eee ce petra pee tr
Srececchs See towers tamer ee
Sara hose, a emt, fr pepe ied
Sine Decechease a ere crtltrers
pedo Se cenesay ay fran fe ak oreo om te
oe baie Somes cen th they frm tow
‘hos bat requesting spel ao :
Sem tes ise Ma_3t wuld apc, The expand of erally ~ Le. of he
ort amie cmos rae pad pa of
ater ceegete tee ecaty toon anche on one estar aes
‘Shy. proved tnt our appar Ctl enough of e ay off ona & Po.
‘ee Bat wl ecu ol
Ds hhave already seen that Quine denis (by implication) that this constitutes
an additional requirement: “The condition upon all aeceptableexplica-
tions of mumber...can be put...: any progression ~ ie., any infinite
series each of whose members has only finitely many precursors ~ will do
nicely” (See note 3). But suppose, for example, that one chose the pro-
aression A =a),d3,43...-0,--. obtained as follows. Divide the postive
integers into two sequences B and C, within each sequence letting the ele
ents come in order of magnitude. Let B (that 8, Dy, sy...) Be the &-
‘quence of Godel mumbers of valid formulas of quantification theory,
under some suitable numbering, and let C (that is, ¢,,c2-..) be the se-
‘quence of positive integers which are not numbers of valid formulas of
‘quantification theory under that numbering (in order of magnitude in
each ease). Now in the sequence A, for each mletdsy-|=By 81 dy =Cy
Clearly A, though a progression, is not recursive, much less primitive
recursive. Just as clearly, this progression would be unusable as the
humbers ~ and the reason is that we expect that if we know which
hhumbers two expressions designate, we are able to calculate in a finite
‘number of steps which is the “greater” Gin this case, which one comes
later in)? More dramatically, if told that set bhas n members, and that
has m, it should be possible to determine ina finite number of steps
which has more members. Yet itis precisely that whieh is not possible
here. This ability (o tell in a finite number of steps which of two
umber isthe greater) is connected with (both transitive and intransi=
cits possibilty is equivalent co the possibility of gen-
crating (“saying”) the numbers in order of magnitude (that is, in theit
‘order in A). You could not know that you were saying them in order of|
‘magnitude since, no recursive rule existing for generating its member
You could not know what ther order of magnitude should be, This is, of|
‘course, avery strong claim, There are two questions here, both of which
fare interesting and nether of which could conceivably receive discussion
inthis paper. (1) Could « human being bea decision procedure for non
Tue i of cove, fiuy withthe sation of “knowing which numbers 90
‘SurSlow sige eo cme re by te fm ea Abra
{hls of amber, sno dow neal ANahon' tbe an ome eee
than now whi utes Abra’ And say wi that BR!
me no os dei whch the rt, Tac te avd ower, ese
‘tember seta See ie ame hares ee a
a se Si hemor hath the soca pearance is Te et
rogresion of mbes rely testa othe po ate. Tse ee mee
Sold the labora cicumoculos tat would res om doing eve oe
hore. Ove could ge he sane eft By leg the unten be fs of
‘Sion tery, ied oh Geel mumbo and wig ifort So
216
Whar numbers could not be
recursive ets ors the human organism at best a Turing machine (in the
relevant respect)? If the latter, then there could not exst a human being
‘who could generate the sequence A, much less know that this is what he
was doing. Even if the answer to (I), however, is that a human being
‘could be (act oF be used a3) such a decision procedure, the following
‘Question would still arise and need an answer: (2) could he know all
truths of the form
tation of the numbers as some particular set of ses, the two accounts
_Eenerally agree on the relations defined on that set; under both, we have
what is demonstrably a recursive progression and a successor function
Which follows the order ofthat progression. Furthermore, the notions of
cardinality are defined in terms of the progression, insuring that it
‘becomes a theorem for each m that a set has members if and only iit
can be put into one-to-one correspondence with the set of numbers less
than or equal to n. Finally, the ordinary arithmetical operations are
defined for these “numbers.” They do differ in the way in which car-
linalt is defined, for in Erie's account the fact that the number had
in members was exploited to define the notion of having m members. In
all other respects, however, they agree.
fore, if i i not the case that both 3=(6]]] and 3 = (0), (0,{II)
surely is not, then at least one of the corresponding accounts
incorrect as @ result of containing a condition that Is not necessary. It
may be incorrect in other respects aswell, but a least that much is clear.
1 can distinguish two possibilties again: either all the conditions just
ted, which both of these accounts share, are necessary for a correct
and complete account, or some are not. Let us assume thatthe former is
the case, although I reserve the right to discard this assumption if i
‘becomes necessary to question it
all the conditions they share are necessary, then the superfluous cot~
ditions are to be found among those that are not shared. Again there are
‘wo possbilities: ether at least one of the accounts satisfying the cond
tions we are assuming to be necessary but which assigns & definite set (0
cach number, is correct, or none are. Clearly no two different ones can
be, since they are not even extensionally equivalent, much less inten=
sionally. Therefore exacily one is correct or none is. But then the correct
280
Whar numbers could not be
fone must be the one that picks out which set of sets is in fact the
numbers. We are now faced witha crucial problem: if there exists such a
“correct” account, do there also exist arguments which will show it to Be
the correct one? Or does there exist a particular set of sets b, whichis
‘really the numbers, But such that there exists no argument one can give 10
tstablish that it, and not, say, Ernie's set SU, is really the numbers? I
seems altogether too obvious that this latter possibility borders on the
absurd, If the numbers constitute one particular set of sets, and not
another, then there must be arguments to indicate which. In urging this
fam not committing myself Co the decidability by proof of every mathe-
matical question ~ Tor I consider this neither a mathematical question nor
‘one amenable to proof. The answer to the question 1 am raising will
follow from an analysis of questions of the form “Is n=...” It will
‘suffie for now to point to the difference between our question and
Is there a greatest prime p such that p-+2 is also prime?
Does there exist an infinite set of real numbers equivalent with
neither the set of integers nor with the set ofall real numbers?
In awaiting enlightenment on the true identity of 3 we are not awaiting &
proof of some deep theorem. Having gotten as far as we have without
‘seting the identity of 3, we can go no further. We do not know what &
proof of that could look like. The notion of “correct account" is
breaking loose from its moorings if we admit of the possible existence of
‘unjustifiable but correct answers to questions such as this. To take
seriously the question “Is 3 ={{{0]}]?" tour cour (and not eliptically for
“in Etnle’s account?" in the absence of any way of settling it, isto lose
one's bearings completely. No, if such a question has an answer, there
fare arguments supporting it, and if there ate no such arguments, then
there is no “correc” account that discriminates amiong all the accounts
satafying the conditions of which we reminded ourselves a couple of
pages back.
How then might one distinguish the correct account from all the pos-
sible ones? Is there a set of sts that has 2 greater claim tobe the numbers
than any other? Are there reasons one can offer to single out that set?
Frege chose as the number 3 the extension of the concept “equivalent
‘with some 3-membered set”; that is, for Frege & number was an equi
lence class - the clase of all classes equivalent with a given class.
‘Although an appealing notion, there seems litle to recommend it over,
‘say, Ernie's. thas been argued that this is a more fitting account because
‘number words are really class predicates, and that this account reveals
281 that fact. The view is that in saying that there are n F's you are pred-
cating n-hood of F, just as in saying that red is a color you aze pred
cating colorhood of red. do not think thsi true. And neither did Frege
(41950: sex. 57. Its ertanly true that to say
(There are seventeen lions in the 200
is not to predicate seventoen-hood of each individual ion, {suppose that
itis also true that if there are seventeen lions inthe 200 and also sever
{cen tigers in the zoo, the classes of lions-in-the-z00 and tigers-in-the-200
are ina class together, though we shall return to that. It doesnot follow
‘rom this that (1) predicates seventeen-hood of one of those classe. First
of all, the grammatical evidence for this is scanty indeed. The best one
‘can conjure up by way of an example of the occurrence of a number
word in predicative position is a rather artificial one lke
@) The lions in the 200 are seventeen,
If we do not interpre this asa statement about the ages ofthe beasts, we
‘ee that such statements do not predicate anything of any individual lon
‘One might then succumb to the temptation of analyzing (2) as the noun
phrase “The lions in the 200” followed by the verb phrase “are seven-
{een,"* where the analysis is parallel to that of
©) The Cherokees ate vanishing
‘where the noun phrase refers to the class and the verb phrase predicates
something of that class. But the parallel is shortlived. For we #008
‘notice that (2) probably comes inte the language by deletion from
(4) The lions in the z00 are seventeen in number,
which in turn probably derives from something ike
(5). Seventeen lions are inthe 200.
his is no place to explore in detail the grammar of number words. Su
fice it to point out that they differ in many important respects from
words we do no esate to cll predicates, Probably the closes thing «2
4 genuine class predicate involving number words is something on the
‘model of “seventecs-membered” or ‘*has seventeen members.” But the
Step from there to “seventeen” being itself a predicate of classes is along
‘one indeed. Infact, I should think that pointing to the above two predi=
‘ales gives away the show ~ for what isto be the analysis of seventeen”
‘sit occurs in those phrases?
[Not only is there scanty grammatical evidence for this view, there
What numbers could not be
sc to considerable viens ain a ny sera fhe sin
In of uncon mong he mt wer and many” "ew" ly”
“foey" any ands oth wl mesial eva The proper sty
or tse mates wl huve 10 aall_anterconen bu the
Sonpstcatve nna of umber rd be ater sea yo
ow fret hey are rom, ora ates, which do fnstion
{Spreicats We have ready sen hat here re real no occ
or mime wors napa Peat poston (hal yi ae)
hike only pula ese ng along the ns of @ above and
itive ar mpaule The eer aomaly tat munber words
sora outank if sce (oa oer aes oe wan to
hss thea hin having wo apea a he etd an deve wring
find not nde, Tiss vch 4 srong taking tht dean vital
(6) The ve lovely gure tes
fine, tay moifston ofthe postin of" ils an mera
Ia ting te farther othe i he Woe”
Torte son fr denying the pedestvefture of number wor
comes fom the aon store ana of semences sch
with whch we sare. For hat seal ana a
ax clang Nba Lae ban ann A
tetey ONEDo emu IER =A
Tie only predicate in) which remains “ton in the 00"
"vente ging ay to aera Guns, th fn a
‘bic un oturenceof =; anes ofcourse one ies cnser
thes as ob rene o ser. Bt there rs rods indeed
forthe vinta) or pret seetntod of hcl ons
steabe. Numbe words utonomch ke operator sch al”
ote" and so for tha ett make Sis tae fhe
‘hold be accompanied by eadne to make the comeing move
‘iin respen to gual herey proving don pilowophis
“kath cosa eh
SSeS shat arene sod
Be crud ot ee ton
Thee ny hone
‘yt vs tem an le ond sh Te we a
ame tes oa
283 fashion) the existence ofthe one, the many, the few, the all, the some,
the any, the every, the several, and the each,”
But then, what support does this view have? Well, this much: if two
lasses each have seventeen members, there probably exists a class which
contains them both in virtue of that Fact. I say “probably” because this
vaties from set theory to set theory, For example, this is not the case
with type theory, since the two classes have both to be ofthe samme type.
But in no consistent theory is there a class of all clases with seventeen
‘members, at least not alongside the other standard se-theoretical appa:
rats. The existence of the paradoxes i itself a good reason to deny t0
“seventeen” this univocal role of designating the class of all classes with
seventeen members.
‘think, therefore, that we may conclude that “seventeen” need not be
considered a predicate of classes, and there is similarly no necessity 10
of all triples. This s not to deny that "isa clas having
is a predicate of dasses; but that isa different mater
Indeed. For that follows from all of the accounts under consideration."
‘Our present probiem isto see if there is one account which can be estab-
lished to the exclusion ofall others, thereby seitling the issue of which
Sets the numbers really are. And it should be clear by now that there is
not. Any purpose we may have in giving an account of the notion of|
rhumber and ofthe individual numbers other than the question-begaing
‘one of proving ofthe right set of sts that isthe set of numbers, wil be
equally well (or badly) served by any one ofthe infinitely many accounts
satisfying the conditions we set out so tediously. There is little need 10
examine all the possibilities in detail, once the traditionally favored one
of Frege and Russell as been seen not to be uniquely suitable,
Where does that leave us? I have argued that at most one of the in-
rnitely many different accounts satisfying our conditions can be correct,
‘onthe grounds that they are not even extensionally equivalent, and thee-
fore atleast all but one, and possibly al, contain conditions that are not
necessary and that lead to the identification of the numbers with some
Particular set of sets. If numbers are sets, then they must be particular
set, for each set is some particular set. But if the number 3 is really One
set rather than another, it must be possible to give some cogent reason
{or thinking so; for the postion that thisisan unknowable truth is hardly
enable. But there seems to be litle to choose among the accounts, Rela~
tive to our purposes in giving an account of these matters, one wil do a5
aod nee why Hot am he oe wh eve in ating forthe few gant be
hin the bounds poe by consistency.
284
What numbers could not be
wel as another, stylistic preferences aside. There is no way connected
with the reference of number words that will allow us to choose among,
them, for the accounts differ at places where tere i no connection what
‘ever between features af the accounts and our uses ofthe words in ques-
tion, If all the above i cogent, then there site to conclude except that
any feature of an account that identifies 3 with ast isa superfluous one
and that therefore 3, and is fellow numbers, could not be sets a all.
MTL. Way out
In this third and final section, I shall examine and urge some considera-
tions that | hope will lend plausibility to the conclusion of the previous
section, if only by contrast. The issues involved are evidently so
‘numerous and complex, and cover such a broad spectrum of philosophic
problems, that in this paper I ean do no more than indicate what think
they are and how, in general, I think they may be resolved. I hope never-
theless that a more postive account will emerge from these consid
4A: eit. Throughout the festive etn, have wate expressions
Sie Tom
® an
when it anunesxprnin at exrsion 4 ta
yma ure fod ee an tat tas our Jobo sor he ae
fom theses and tight ape tat hd onl ht al ich
Sinemet wee set to damatie he Kn Fane hat 8
Freqnsmigiae tthe requr foram analy ef number = po
thekin gurionFepetook fob: For lary waned he al
Sst deen teh ae for eh sch ent Inf, e wae
Wideominea ae fre el of eli» wth ny mame ove
trpion hance (eile anes onan beled fo name a
Ter cca the pono of) ven the ser an ay
orden tere wee tee Kinds of Wee satising tee