26. H. P. Fairchild, Dictionary of Sociology Scientific Policy 1959-60 (Her Majesty's Sta- Science, Warsaw, 1961). Yugoslavia: S.

Dedi-
(Littlefield, Adams, Paterson, N.J., 1961). tionery Office, London. Sweden: S. Brohult, jer, Nature 187, 468 (1960); estimates from
27. D. J. S. Price, Science since Babylon (Yale Tek. Vetenskaplig Forskning 32, 339 (1961). published official data. China: J. M. H. Lind-
Univ. Press, New Haven, 1961). West Germany: Steifterverband fur die Deutsche beck, In "Sciences in Communist China,"
28. A. M. Weinberg, "Problems of Big Science," Wissenschaft Jahrbuch (1961); private com- Pubi. Am. Assoc. Advan. Sci. No. 68
paper presented at the University of Ten- munication. Canada: Natl. Res. Council Can. (1961). Ghana: private communication from
nessee, Knoxville (1962). Ann. Rept. No. 44 (1960-61). France: Le the Research Council of Ghana. Lebanon and
29. P. Auger, Current Trends in Scientific Re- Progas Scientifique (May 1961). Norway: Egypt: Regional Conference on Scientific Re-
search (UNESCO, Paris, 1962). Tiden (Dec. 1960); R. Major, Science 129, search Facilities and Cooperation (UNESCO,
30. The data on expenditure for research and 694 (1959). Australia: S. Encel, Science 134, Cairo, 1960). Philippines: Government Ex-
260 (1961); , private communication. penditure for R&D: for Manufacturing In-
development are from various sources, as Japan: "Statistical Survey of Researchers in dustries, 1959-60 (National Science Devel-
follows. U.S.: National Science Foundation Japan, 1960" (Bureau of Statistics, Office of opment Board, Manila, 1961). Pakistan: "Re-
Publ. No. 62-9 (1962). U.S.S.R.: K. Meyer, the Prime Minister, Japan); Y. Shizume, port of Scientific Committee of Pakistan"
Das Wissenschaftliche Leben in the USSR Growth-rate of Science in Japan. New Zea- Government of Pakistan, 1960). India: P. S.
(Osteuropa Institut, Berlin, 1959); La Ma- land: S. Encel, private communication. Po- Mahalanobis, "Recent Development in the
tiare Grise en Europe (Brussels, 1960). U.K.: land: Z. Zagadnien Planowania I Koordi- Organization of Science in India" (Indian
Annual Report of the Advisory Council on nacji Badan Naukowich (Polish Academy of Statistical Institute, Calcutta, 1959).

head about 1910. You will recall, then,

the astonishing contention with which
he shocked the mathematical world of
that time-namely, that all of mathe-
Are Logic and Mathematics matics was nothing but logic. Mathe-
maticians were generally puzzled by

Downloaded from http://science.sciencemag.org/ on July 4, 2019

this radical thesis. Really, very few
Identical? understood at all what Russell had in
mind. Nevertheless, they vehemently
opposed the idea.
An old thesis of Russell's is reexamined in the light This is readily understandable when
of subsequent developments in mathematical logic. you recall that a companion thesis of
Russell's was that logic is purely tauto-
logical and has really no content what-
Leon Henkin ever. Mathematicians, being adept at
putting 2 and 2 together, quickly in-
ferred that Russell meant to say that
all mathematical propositions are com-
pletely devoid of content, and from
It was 24 years ago that I entered exalt the latter at the expense of the this it was a simple matter to pass to
Columbia College as a freshman and former, and I determined to read the the supposition that he held all mathe-
discovered the subject of logic. I can essay in order to refute it. But I dis- matics to be entirely without value.
recall well the particular circumstance covered something quite different from Aux armes, citoyens du monde mathe'-
which led to this discovery. what I had imagined. Indeed, contrast- matique!
One day I was browsing in the li- ing aspects of mysticism and logic were Half a century has elapsed since
brary and came across a little volume delineated by Russell, but his thesis this gross misinterpretation of Russell's
by Bertrand Russell entitled Mysticism was that each had a proper and im- provocative enunciation. These 50
and Logic. At that time, barely 16, I portant place in the totality of human years have seen a great acceleration
fancied myself something of a mystic. experience, and his interest was to de- and broadening of logical research.
Like many young people of that age I fine these and to exhibit their inter- And so it seems to me appropriate to
was filled with new emotions strongly dependence rather than to select one seek a reassessment of Russell's thesis
felt. It was natural that any reflective as superior to the other. I was dis- in the light of subsequent development.
attention should be largely occupied armed, I was delighted with Russell's
with these, and that this preoccupation lucent and persuasive style, I began
should give a color and poignancy to avidly to read his other works, and was Definitions and Proofs
experience which found sympathetic soon caught up with logical concepts
reflection in the writings of men of which have continued to occupy at In order to explain how Russell
mystical bent. least a portion of my attention ever came to hold the view that all of
Having heard that Russell was a since. mathematics is nothing but logic, it
logician I inferred from the title of Bertrand Russell was a great pop- is necessary to go back and discuss two
his work that his purpose was to con- ularizer of ideas, abstract as well as important complexes of ideas which
trast mysticism with logic in order to concrete. Probably many of you have had been developed in the decades be-
The author is professor of mathematics at
been afforded an introduction to fore Russell came into the field. The
the University of California, Berkeley, and mathematical logic through his writ- first of these was a systematic reduc-
president of the Association for Symbolic Logic. ings, and perhaps some have even been tion of all the concepts of mathematics
This article is adapted from an address given
5 September 1961 at the 5th Canadian Mathe- led to the point of peeping into the to a small number of them. This
matical Congress, in Montreal. It is reprinted formidable Principia Mathematica process of reduction had indeed been
from the Proceedings of the congress, with
permission. which he wrote with Alfred White- going on for a very long time. As far
788 SCIENCE, VOL. 138
back as the days of Descartes, for ex- aid of algebraic symbols such as the to develop the main theorems con-
ample, we can see at least an imperfect plus sign, the multiplication sign, and cerning these concepts as part of a
reduction of geometric notions to al- the equality sign and of variables. For uniform and systemic development.
gebraic ones. Subsequently, with the example, Boole used the familiar equa- Viewed in retrospect, the contempo-
development of set theory initiated tion P.Q = Q.P. to express the fact that rary logician is struck by the willing-
by Georg Cantor, the reduction of the sentences of the form "P and Q" and ness of Russell and Whitehead to rest
system of real numbers to that of "Q and P" must be both true or both their case on what, for a mathema-
natural numbers marked another great false (whatever the sentences P and Q tician, must be considered such flimsy
step in this process. But perhaps the may be), while the generally unfa- evidence. The world of empirical sci-
most daring of these efforts, the culmi- miliar algebraic equation -(P.Q) = ence, of course, expects to achieve
nating one, was the attempt by a Ger- (-P) + (-Q) indicates that the sen- conviction on the basis of empirical
man mathematician, Gotlob Frege, to tence "Not both P and Q" has the evidence, but the quintessence of the
analyze the notion of natural number same truth value as "Either not P or mathematician's approach, especially
still further and reduce it to a concept not Q." Boole demonstrated that of the mathematical logician's, is the
which he considered to be of a -purely through the use of such algebraic no- demand always for proof before a
logical nature. tation one can effect a great saving in thesis is accepted. Yet you see that
Frege's work was almost entirely the effort needed to collate and apply whereas Russell was interested in estab-
unnoticed in his own time (the last basic laws of logic. Later his work was lishing that in a certain sense all of
three decades of the 19th century), extended and deepened by the Ameri- mathematics could be obtained from
but when Bertrand Russell came upon can C. S. Peirce and the German his logical axioms and concepts, he
Frege's work he realized its great sig- mathematician E. Schroder. And Rus- never really set out to give a proof of

Downloaded from http://science.sciencemag.org/ on July 4, 2019

nificance and gave these ideas very sell himself, working within this tradi- this fact! All he did was to gather the
wide currency through his own bril- tion, found it a convenient basis for basic ideas that had been developed
liant style of exposition. The ultimate a systematic development of all math- in a nonformal and unsystematic way
elements into which the notion of nat- ematics from logic. By combining the by mathematicians before him, and to
ural number was analyzed by Frege symbolic formulation of logical laws say, in effect, "You see that I have
and by Russell were entities which with the reduction of mathematical been able to introduce all this loosely
they called "propositional functions." concepts to a logical core, he was able formulated work within the precise
To this day there persists a controversy to conceive of a unified development framework of my formal system. And
among philosophers as to just what such as was attempted in the Principia it's pretty clear, isn't it, that I have all
ihese objects are, but at any rate they Mathematica. the tools available to formalize such
are connected with certain linguistic further work as mathematicians are
expressions which are like sentences likely to do?"
except for containing variables. Just From Russell to Godel In this respect one is reminded of
as there is a certain proposition asso- the approach of that first great axiom-
ciated with (or expressed by) the sen- What was the Principia like? Well atizer and geometer, Euclid. Euclid,
tence "U Thant is an astronaut," for of course the work is still not com- too, conceived that all propositions of
example, so there is a propositional pleted (only three of four projected geometry-that is, all the true state-
function associated with the expression volumes having appeared); and since ments about triangles, circles, and those
"x is an astronaut." Since propositions Bertrand Russell has most recently other figures in which he was inter-
had long been recognized as consti- seemed to occupy himself with the ested-could be developed from the
tuting one of the most basic portions political effects of certain physical re- simple list of concepts and axioms he
of the domain of investigation of logi- search it may, perhaps, never be com- gave. But in his case, too, there was
cians, and since propositional func- pleted! Nevertheless, one can see never any attempt to prove this fact
tions are very closely related to propo- clearly the intended scope of the work. other than by the empirical process of
sitions, it was natural to consider these, Surprisingly, it reminds one of the deriving a large number of geometric
too, to be a proper part of the subject present massive undertaking by the propositions from the axioms and then
of logic. It is in this sense that Frege Bourbaki group in France. For even appealing to the good will of the
seemed able, by a series of definitions, though the Principia and Bourbaki audience, so to speak. "Well," we may
to arrive at the notion of number, as are very dissimilar in many ways, each imagine him saying, "look how much
well as at the other notions under attempts to present an encyclopedic I have been able to deduce from my
study in various parts of mathematics, account of contemporary mathemati- axioms. Aren't you pretty well con-
starting from purely logical notions. cal research unified by a coherent vinced that all geometric facts follow
The second important line of devel- point of view. from them?"
opment which preceded Russell, and In the Principia, starting from cer- But of course there were mathema-
upon which he drew for his ideas, was tain axioms expressed in symbolic form ticians and logicians who were not
the systematic study by mathema-tical which were intended to express basic convinced. And so the demand for
means of the laws of logic which laws of logic (axioms involving only proof was raised.
entered into mathematical proofs. This what Russell conceived to be logical Actually, the proper formulation of
development was initiated by George notions), the work systematically pro- the problem of whether a system of
Boole, working in England in the mid- ceeds to derive the other laws of logic, axioms is adequate to establish all of
dle of the 19th century. He discovered to introduce by definition such mathe- the true statements in some domain of
that certain of the well-known laws of matical notions as the concept of num- investigation requires a mathematically
logic could be formulated with the ber and of geometric space, and finally precise formulation of the notion of
16 NOVEMBER 1962 789
"true sentence," and it was not until then whenever a sentence is true in pleteness. From this viewpoint it ap-
1935 that Alfred Tarski, in a great every model satisfying these axioms pears that one of the basic elements
pioneering work, made fully evident the there must be a proof of finite length, on which Russell rested his thesis that
form in which semantical notions must leading from the axioms to this sen- all mathematics could be reduced to
be analyzed for mathematical lan- tence, each line of the proof following logic must be withdrawn and recon-
guages. Of course, it is a trivial matter from preceding lines by one of several sidered.
to give the conditions under which any explicitly listed rules of logic. This
particular sentence is true. For exam- result of Godel's is among the most
ple, in the theory of Euclidean geom- basic and useful theorems we have in Consistency and Decision Problems
etry the sentence "All triangles have the whole subject of mathematical
two equal angles" is true if, and only logic. I have been talking about complete-
if, all triangles have two equal angles. But the very next year, in 1931, the ness, which has to do with the ade-
However, Tarski made it clear that hope of further extension of this kind quacy of a formal system of axioms
there is no way to utilize this simple of completeness proof was definitely and rules of inference for proving true
technique in order to describe (in a dashed by Godel himself in what is sentences. But I must mention, also,
finite number of words) conditions for certainly the deepest and most famous a second aim of the Russell-Whitehead
the truth of all the infinitely many sen- of all works in mathematical logic. Principia which also fared ill in the
tences of a language; for this purpose Godel was able to demonstrate that subsequent development of mathemati-
a very different form of definition, the system of Principia Mathematica, cal logic. Russell and Whitehead were
structural and recursive in character, taken as a whole, was incomplete. That very much concerned with the ques-
is needed. is, he showed explicitly how to con- tion of consistency. While they hoped

Downloaded from http://science.sciencemag.org/ on July 4, 2019

Even before Tarski's treatment of struct a certain sentence, about natural to have a complete system, one con-
semantics, indeed as early as 1919, we numbers, which mathematicians could taining proofs for all correct state-
find the first proof of what we call, in recognize as being true under the in- ments, they were also concerned that
logic, "completeness." The mathema- tended interpretation of the symbolism their system should not contain proofs
tician Emil Post (in his doctoral dis- but which could not be proved from of incorrect results. In particular, in
sertation published in that year), limit- the axioms by the rules of inference a consistent system such as they
ing his attention to a very small frag- which were part of that system. sought, it would not be possible to
ment of the system created by White- Now, of course, if Godel had done prove both a sentence and its nega-
head and Russell, was able to show nothing more than this, one might tion.
that for any sentence in that fragment simply conclude that Russell and To understand their concern with
which was "true under the intended Whitehead had been somewhat care- the question of consistency it is neces-
interpretation of the symbols," one less in formulating their axioms, that sary to recall the rude wakening which
could indeed get a proof by means of they had left out this true but un- mathematicians sustained in 1897 in
the axioms and rules of inference provable sentence from among the connection with Cantor's theory of
which had been stated for the system. axioms, and one might hope that by transfinite numbers. For centuries be-
Subsequently, further efforts were adding it as a new axiom a stronger fore the time of Cantor mathematicians
made to extend the type of complete- system which was complete would be simply assumed that anyone who was
ness proof which Post initiated, and achieved. But Godel's proof shows properly educated in their subject could
it was hoped that ultimately the entire that this stronger system, too, would distinguish a correct proof from an in-
system of the Principia could be contain a sentence which is true but correct one. Those who had trouble in
brought within the scope of proofs not provable; that, indeed, if this sys- making this distinction were simply
of this kind. tem were further strengthened, by the "weeded out" in the course of their
In 1930, Kurt G6del contributed addition of this new true but unprov- training and were turned from mathe-
greatly to this development and to able sentence as an axiom, the result- matics to lesser fields of study. And no
this hope when he succeeded in prov- ing system would again be incomplete. one took up seriously the question of
ing the completeness of a deductive And indeed, if a whole infinite se- setting forth, in explicit and mathe-
system based upon a much larger por- quence of sentences were to be ob- matical terms, exactly what was meant
tion of mathematical language than tained by successive applications of by a correct proof.
had been treated by Post. Godel's Godel's method, and added simultane- Now when Cantor began his devel-
proof deals with the so-called "first- ously to the original axioms of Prin- opment of set theory he concerned
order predicate logic," which treats of cipia, the same process could still be himself with both cardinal and ordinal
mathematical sentences containing var- applied to find another true sentence numbers of transfinite type. (These
iables of only one type. When such a still unprovable. numbers can be used for infinite sets
sentence is interpreted as referring to Actually, Godel described a very in very much the same way that we
some mathematical model, its variables wide class of formal deductive systems use ordinary numbers for counting and
are interpreted as ranging over the ele- to which his method applies. And most ordering finite sets.) Many of the
ments of the model; in particular, there students of the subject have been con- properties of transfinite numbers are
are no variables ranging over sets of vinced that any formal system of identical to those of ordinary numbers,
model elements, or over the integers axioms and rules of inference which and in particular Cantor showed that,
(unless these happen to be the ele- it would be reasonable to consider as given any ordinal number b, we can
ments of the particular model). Now a basis for a development of mathe- obtain a larger number, b + 1. How-
G6idel shows that if we have any sys- matics would fall in this class, and ever, in 1897 an Italian mathematician,
tem of axioms of this special kind, hence would suffer a form of incom- C. Burali-Forti, demonstrated that
790 SCIENCE, VOL. 138
there must be a largest ordinal number, has persuaded most, though admittedly of the Principia could some day be
by considering the set of all ordinal not all, logicians that Hilbert's search brought within the scope of such a
numbers in their natural order. Mathe- for a consistency proof must remain procedure.
maticians were unable to find any unfulfilled. Efforts to find decision procedures
point, either in the argument of Can- I should like finally to mention a for various fragments of the Principia
tor or in that of Burali-Forti, which third respect in which the original aim were vigorous and many. The doctoral
they intuitively felt rested on incor- of mathematical logicians was frus- dissertation of Post, for example, con-
rect reasoning. Gradually it was real- trated. The questions of consistency tained some efforts in this direction,
ized that mathematicians had a genuine and completeness clearly concerned and further work was produced during
paradox on their hands, and that they the authors of Principia Mathematica, the succeeding 15 years by logicians of
would have to grapple at last with the but the question of decision procedures many countries. Then in 1936 Alonzo
question of just what was meant by a seems not to have been treated to any Church, making use of the newly de-
correct proof. Later, Russell himself serious extent by Russell and White- veloped notion of recursive function,
produced an even simpler paradox in head. Nevertheless, this is an area of was able to demonstrate that for a cer-
the intuitive theory of sets, based upon study which interested logicians as far tain fragment of mathematical lan-
the set of all those sets which are not back as the time of Leibniz. Indeed, guage, in fact for that very first-order
elements of themselves. Leibniz himself had a great dream: He predicate logic which Godel, in 1930,
This background sketch will make dreamt that it might be possible to had showed to be complete, no decision
clear why it was that Russell and devise a systematic procedure for an- procedure was possible. And so with
Whitehead were concerned that no swering questions-not only mathe- decision procedures, as with proofs of
paradox should be demonstrable in matical questions but even questions completeness and consistency, efforts

Downloaded from http://science.sciencemag.org/ on July 4, 2019

their own system. And yet they them- of empirical science. Such a procedure to establish a close rapport between
selves never attempted a proof that was to obviate the need for inspiration logic and mathematics came to an un-
their system was consistent! The only and replace this with the automatic happy end.
evidence they adduced was that a carrying out of routine procedure. Had Well, I have brought you down to the
large number of theorems had been Leibniz been conversant with today's year 1936. Probably most mathemati-
obtained within their system without high-speed computing machines he cians have heard at least something of
encountering paradox, and that all at- might have formulated his idea by as- the development which I have sketched
tempts to reproduce within the system serting the possibility that one could here. But somehow the education in
of Principia Mathematica the Burali- write a program of such breadth and logic of most mathematicians seems to
Forti paradox, and such other para- inclusiveness that any scientific ques- have been terminated at about that
doxes as were shown, had failed. tion whatever could be placed on tape point. The impression is fairly wide-
As with the question of complete- and, after the machine had been set spread that, with the discoveries of
ness, mathematicians were not satisfied to work on it for some finite length of Godel and Church, the ambitious pro-
with an answer in this form, and there time, a definitive reply would be forth- gram of mathematical logicians in ef-
arose a demand that an actual proof coming. fect ground to a halt, and that since
of the consistency be given for the then further work in logic has been a
system of Principia. (and for other sys- sort of helpless faltering by people,
tems then considered). The great and Logic after 1936 unwilling to accept the cruel facts of
illustrious name of David Hilbert was life, who are still seeking somehow to
associated with these efforts to achieve Leibniz's idea lay dormant for a buttress the advancing frontiers of
consistency proofs for various portions long time, but it was natural to revive mathematical research by finding a
of mathematics, and under his stimulus it in connection with the formal de- nonexistent consistency proof.
and direction important advances were ductive systems which were developed And yet this image is very far indeed
made toward this goal, both by him- by mathematical logicians in the early from reality. For in 1936, just at the
self and by his students. But as with the part of this century. Since these lo- time when, many suppose, the demise
efforts to prove completeness, Hilbert's gicians had been interested in formu- of mathematical logic had been com-
program came to founder upon the lating mathematical ideas within a sym- pleted, an international scholarly soci-
brilliant ideas of Kurt Godel. bolic calculus and then manipulat- ety known as the Association for Sym-
Indeed, in that same 1931 paper to ing the symbols according to prede- bolic Logic was founded and began
which I have previously referred, termined rules in order to obtain publication of the Journal of Symbolic
Godel was able to show that the ques- further information about these mathe- Logic. In the ensuing 25 years this has
tions of consistency and completeness matical concepts, it seemed natural to greatly expanded to accommodate a
were very closely linked to one raise the question of whether one could growing volume of research. And at
another. He was able to show that if a not devise purely automatic rules of present there are four journals devoted
system such as the Principia were computation which would enable one to exclusively to publishing material deal-
truly consistent, then in fact it would reach a decision as to the truth or fal- ing with mathematical logic, while many
not be possible to produce a sound sity of any given sentence of the cal- articles on logic appear in a variety of
proof of this fact! Now this result culus. And while the area of empirical mathematical journals of a less special-
itself sounds paradoxical. Nevertheless, science was pretty well excluded from ized nature.
when expressed with the technical ap- the consideration of 20th-century logi- In the space remaining I should like
paratus which Godel developed, it is in cians seeking such decision procedures, to mention very briefly some of the de-
fact a precisely established and clearly it was perhaps not beyond the hope of velopments in mathematical logic since
meaningful mathematical result which some that a system as inclusive as that 1936.
16 NOVEMBER 1962 791
Sets and Decision Methods not exclude the existence of nondenu- ical connectives and quantifiers. If we
merably many "Urelemente" (objects ask whether any given sentence of this
I have found it convenient for this which are not sets). The independence kind is true for all Abelian groups, it is
exposition to divide research in mathe- of the axiom of choice from systems of possible to answer the question in an
matical logic into seven principal areas. axioms such as that used in Godel's automatic way by using the method of
And first I shall mention the area deal- consistency proof, and the independence Szmielew. But if we are interested in
ing with the foundations of the theory of the continuum hypothesis in any which of these sentences are true for
of sets. known system of set theory, remain all groups, then Tarski's proof shows
To explain the connection of this open questions. that it is impossible to devise a machine
field with logic it should be mentioned More recently the direction of re- method to separate the true from the
that those objects which Russell and search in the area of foundations of set false ones.
Whitehead had called "propositional theory seems to have shifted from that A result closely related to Tarski's
functions" are, in fact, largely indis- of formulating specific axiom systems is that of P. Novikov and W. Boone
tinguishable from what are now called and deriving theorems within them to concerning the nonexistence of a deci-
"'sets" and "relations" by mathemati- consideration of the totality of differ- sion method which would enable one
cians. From a philosophical point of ent realizations of such axiom systems. to solve the word problem for the the-
view there is perhaps still room for It is perhaps J. Shepherdson who should ory of groups, a problem for which a
distinguishing these concepts from one be given credit for the decisive step in solution had long been sought by alge-
another. But since, in fact, the treat- this shift of emphasis, although his braists. In fact it is a simple matter to
ment of propositional functions in Prin- work clearly owes much to Godel's. show that the Novikov-Boone result is
cipia Mathematica is extensional (so Subsequent work by Tarski, R. Vaught, equivalent to the nonexistence of a de-
that two functions which are true of ex- and R. Montague has carried this de- cision method for a certain subset of
actly the same objects are never dis- velopment much further.

Downloaded from http://science.sciencemag.org/ on July 4, 2019

the sentences making up the elementary
tinguished), for mathematical purposes An important tool in their work is theory of groups-namely, all those
this system is identical to one which the concept of the rank of a set, which sentences having a special, very simple,
treats of sets and relations. may be defined inductively as the least form. Hence, this result is stronger than
Among systems of set theory which ordinal number exceeding the rank of Tarski's.
have been put forth by logicians as a all elements of the set. This notion may
basis for the development of mathe- be used to classify models of set theory
matics, the principal ones are the theory according to the least ordinal number Recursive Functions
of types used by Whitehead and Rus- which is not the rank of some set of the
sell themselves, subsequently amplified model. Recently there have been some Now the key concept whose develop-
by L. Chwistek and F. Ramsey, and an very interesting contributions by Azriel ment was needed before negative solu-
alternative line of development initi- L6vy to these studies. His efforts have tions to decision problems could be
ated by E. Zermelo, to which important been directed toward successively achieved was the concept of a recursive
contributions were subsequently made strengthening the axioms of set theory function. Intuitively speaking this is
by A. Fraenkel and T. Skolem. Still so as to penetrate increasingly far into simply a function from natural num-
another system, having certain charac- the realm of the transfinite. bers to natural numbers which has the
teristics in common with each of these A second area that I would delineate property that there is an automatic
two principal forms, was advanced and in contemporary logical research is that method for computing its value for any
has been studied by W. Quine and, to dealing with the decision problem. While given argument. A satisfactory and ex-
some extent, by J. B. Rosser. Of these it is true that the work of Church made plicit mathematical definition of this
systems the Zermelo-type system has it clear that there could be no universal class of functions was first formulated
probably received most attention, along decision procedure for mathematics, by J. Herbrand and Godel. But it re-
with an important variant form sug- there has remained a strong interest in mained for S. C. Kleene to develop the
gested and developed by J. von Neu- findina decision procedures for more concept to such an extent that it now
mann, P. Bernays, and Go5del. modest portions of mathematical the- underlies a very large and- important
Among the significant efforts expend- ory. Of special interest here is Tarski's part of logical research.
ed on these systems were those directed decision method for elementary algebra Much of the work with recursive
toward establishing the status of propo- and geometry, and an important exten- functions has been along the line of
sitions such as the axiom of choice and sion of it which was made by Abraham classifying sets and functions, a clas-
the continuum hypothesis of Cantor. Robinson. Wanda Szmielew has also sification similar to that involving pro-
Here the names of Godel and A. Mo- given an important decision procedure jective and analytic sets in descriptive
stowski are especially prominent. -namely, one for the so-called "ele- set theory. Kleene himself, his students
Godel showed that a strong form of mentary theory" of Abelian groups. By Addison and Spector, and other logi-
the axiom of choice and the generalized contrast, the elementary theory of all cians, including Post, Mostowski, J.
continuum hypothesis are simultaneous- groups was shown by Tarski to admit Shoenfield, and G. Kreisel, have con-
ly consistent with the more elementary of no decision procedure. In fact, tributed largely to this development.
axioms of set theory-under the as- Szmielew and Tarski considered exactly Also to be mentioned are the applica-
sumption that the latter are consistent the same set of sentences-roughly, all tions which initially Kleene, and subse-
by themselves. Mostowski showed that of those sentences which can be built quently others, have attempted to make
the axiom of choice is independent of up by the use of the group operation of the concept of recursive function by
the more elementary axioms of set the- symbol, and variables ranging over the way of explicating the notion of "con-
ory, provided that a form of these ele- group elements, with the aid of the structive" mathematical processes. In
mentary axioms is selected which does equality sign, as well as the usual log- this connection several attempts have
792 SCIENCE, VOL. 138
been made to link the notion of recur- tinct models of these axioms which are only nonnegative values is represented
sive function with the mathematical alike in all respects except for the in- as a sum of squares of rational func-
viewpoint known as intuitionism, a rad- terpretation of the given symbol. (This tions, the number of terms needed for
ical reinterpretation of mathematical proves the completeness of A. Padoa's the representation depends only on the
language which was advanced by L. method of demonstrating nondefinabil- degree and number of variables of the
Brouwer and developed by A. Heyting. ity.) A logical interpolation theorem of given polynomial, and that it is inde-
W. Craig's provides a close link for the pendent of the particular coefficients.
results of Lyndon and Beth.
Algebra, Logic, and Models A sixth area which can be discerned
in recent work on logic concerns the Russell's Thesis in Perspective
A fourth area of logical research theory of proof. This is perhaps the
deals with material which has recently oldest and most basic portion of logic, I hope that this very brief sketch of
been described as algebraic logic. This a search for systematic rules of proof, some of the areas of contemporary log-
is actually a development which can be or deduction, by means of which the ical research will give some idea of the
traced back to the very early work of consequences of any propositions could ways in which logicians have reacted to
Boole and Schroder. However, interest be identified. In recent work, however, the theorems of Godel and Church
in the subject has shifted away from the logicians have begun to depart in rad- which, in the period 1931 to 1936, dealt
formulation and derivation of algebraic ical ways from the type of systems for so harshly with earlier hopes. Speaking
equations which express laws of logic which rules of proof were originally generally, one could describe this reac-
to the consideration of abstract struc- sought. For example, several attempts tion as compounded of an acceptance
tures which are defined by means of have been made to provide rules of of the impossibility of realizing the
such equations. Thus, the theory of proof for languages containing infinitely original hopes for mathematical logic,

Downloaded from http://science.sciencemag.org/ on July 4, 2019

Boolean algebras, of relation algebras, long formulas, such as sentences with a relativization of the original program
of cylindric and polyadric algebras have infinitely many disjunctions, conjunc- of seeking completeness and consistency
all successively received attention; M. tions, and quantified variables. Tarski, proofs and decision methods, an incor-
Stone, Tarski, and P. Halmos are close- Scott, C. Karp, W. Hanf, and others poration of the new methods and con-
ly associated with the central develop- have participated in such efforts. Curi- structs which appeared in the impossi-
ment here. The algebraic structures ously enough, while this direction of bility proofs, and the development of
studied in this domain may be associ- research seems at first very far removed quite new interests suggested by gen-
ated in a natural way with mathemat- from ordinary mathematics, one of the eralization of early results.
ical theories, and this association per- important results was used by Tarski to Now with this background, let us
mits the use of very strong algebraic solve a problem, concerning the exist- return to Russell's thesis that all of
methods in the metamathematical anal- ence of measures on certain very large mathematics can be reduced to logic. I
ysis of these theories. spaces, which had remained unsolved would say that if logic is understood
A fifth area of modern logical re- for many years. clearly to contain the theory of sets
search concerns the so-called theory of The last area of logical research I (and this seems to be a fair account of
models. Here effort is directed toward should like to bring to your attention what Russell had in mind), then most
correlating mathematical properties pos- is a kind of converse study to what we mathematicians would accept without
sessed by a class of structures defined have called algebraic logic. In the latter question the thesis that the basic con-
by means of given mathematical sen- we are interested in applying methods cepts of all mathematics can be ex-
tences with the structural properties of of algebra to a system of logic. But pressed in terms of logic. They would
those sentences themselves. there are also studies in which results agree, too, that the theorems of all
A very early example is Garrett Birk- and methods of logic are used to estab- branches of mathematics can be de-
hoff's result that, for a class of struc- lish theorems of modern algebra. The rived from principles of set theory, al-
tures to be definable by means of a set first to have made such applications though they would recognize that no
of equational identities, it is necessary seems to have been the Russian mathe- fixed system of axioms for set theory
and sufficient that it be closed under matician A. Malcev, who in 1941 in- is adequate to comprehend all of those
formation of substructures, direct prod- dicated how the completeness theorem principles which would be regarded as
ucts, and homomorphic images. Char- for first-order logic could be used to "mathematically correct."
acterizations of a similar nature were obtain a result on groups. Subsequently But perhaps of greater significance
given for classes definable by universal the same technique was used by Tarski is the consensus of mathematicians that
elementary sentences (Tarski) and by to construct various non-Archimedean there is much more to their field than
any elementary sentences (J. Keisler). ordered fields. Perhaps the best-known is indicated by such a reduction of
A related type of result is R. Lyn- name in this area is that of Abraham mathematics to logic and set theory.
don's theorem that any elementary sen- Robinson, who formerly was associated The fact that certain concepts are se-
tence whose truth is preserved under with the University of Toronto in lected for investigation, from among
passage from a model of the sentence Canada. Among his contributions was all logically possible notions definable
to a homomorphic image of that model the application of logical methods and in set theory, is of the essence. A true
must be equivalent to a sentence which results to improve a solution, given in understanding of mathematics must in-
does not contain negation signs. In a 1926 by E. Artin, to Hilbert's 17th volve an explanation of which set-theory
different direction, E. Beth has shown problem (17th of the famous list of notions have "mathematical content,"
that if a given set symbol or relation problems presented in his address to the and this question is manifestly not re-
symbol is not definable in terms of the International Congress of Mathemati- ducible to a problem of logic, however
other symbols of an elementary axiom cians in 1900). Robinson showed that broadly conceived.
system, then there must exist two dis- when a real polynomial which takes Logic, rather than being all of math-
16 NOVEMBER 1962 793
ematics, seems to be but one branch. contained. With reference to the following list, 3 (Princeton Univ. Press, Princeton, N.J.,
for an example of an axiomatic theory of sets 1940).
But it is a vigorous and growing branch, and an important contribution toward its meta- L. Henkin, "Some remarks on infinitely long
and there is reason to hope that it may theory, see Godel. For an account of the de- formulas," in Infinitistic Methods (Pergamon,
cision problem,- see Tarski (1951) for positive New York, 1961).
in time provide an element of unity to solutions and Tarski et al. (1953) for nega- and A. Tarski, "Cylindric algebras," in
tive. The theory of recursive functions and "Lattice Theory" (Proc. Symposium in Pure
oppose the fragmentation which seems their uses is well described in Kleene. Some Mathematics, Providence, 1961) (American
to beset contemporary mathematics- historical remarks on algebraic logic, as well as Mathematical Society, in press), vol. 2.
H. J. Keisler, Indagationes Mathematicae 23,
detailed results, may be found in Henkin and
and indeed every branch of scholarship. Tarski. A very recent and important contribu- 477 (1961).
tion to the theory of models is that of Keisler, S. C. Kleene, Introduction to Metamathematics
Bibliography where reference to earlier works may be found. (Van Nostrand, New York, 1952).
For an account of recent work on infinitely long A. Robinson, Complete Theories (North-Hol-
Some of the early work of Boole, Frege, and formulas, see Henkin; an application of such land, Amsterdam, 1956).
Cantor has been made more available by rela- work to a problem on the existence of certain B. Russell and A. N. Whitehead, Principia
tively recent translations into English. Examples measures appears in Tarski (in press). Ap- Mathematica (Cambridge Univ. Press, Cam-
are the works cited. The Principia Mathematica plications of logic to algebra are described in bridge, England; 1910, 1912, 1913, respec-
of Russell and Whitehead, while largely devoted Robinson. tively), vols. 1-3.
to a formidable formalism, contains many very A. Tarski, A Decision Method for Elementary
readable sections of an introductory or sum- G. Boole, An Investigation of the Laws of Algebra and Geometry (Univ. of California
mary character. An account of the fundamental Thought (Dover, New York, new ed., 1951). Press, Berkeley, 1951).
theorems of Godel and Church may be found G. Cantor, Contributions to the Founding of "Some problems and results relevant to
in Kleene. In describing logical research since the Theory of Transfinite Numbers (Open the foundations of set theory," in "Proceedings
1936 my aim has been not to give a complete his- Court, London, 1915) (P. Jourdain, trans.). of the International Congress for Logic,
tory but only to indicate the range of activity by G. Frege, The Foundations of Arithmetic (Philo- Methodology, and Philosophy of Science,
mentioning some of the most actively cultivated sophical Library, New York, 1950) (J. L. Stanford, 1960" (Stanford Univ. Press, in
areas. Accordingly, I have referred to only a Austin, trans.). press).
very small sample of the literature, selecting K. Godel, "The Consistency of the Axiom of - , A. Mostowski, R. Robinson, Undecidable
(where possible) works which are largely self- Choice," Annals of Mathematics Study No. Theories (North-Holland, Amsterdam, 1953).

Downloaded from http://science.sciencemag.org/ on July 4, 2019

matter with a demonstration in which
the change in wavelength was precisely
measured with a crystal spectrometer
and shown to be h/mc or 0.024 ang-
Arthur Holly Compton, strom units at 90 degrees, as the calcu-
lation had predicted. The audience was
ready, because the apparent conflict be-
Research Physicist tween the corpuscular and the wave
theories of light was in every physicist's
mind. A Nobel-prize discovery had been
made.
But here at Harvard in 1924, in the
I first met Arthur Compton in 1924, happened in physics, the experiment, laboratory of a highly respected in-
in William Duane's x-ray research lab- the interpretation, and the audience vestigator of x-rays, the crystal spec-
oratory at Harvard University. Comp- were not simultaneously ready, and trometer measurements seemed to give
ton had come on a visit to attempt to these prior results attracted little atten- different results. The scattered radiation
discover why Duane and his associates tion. Compton, however, had never showed, as Compton had found, part
could not confirm his discovery of the completely laid aside those gamma-ray of the radiation to be shifted to longer
change of wavelength of x-rays on scat- experiments he performed in Ruther- wavelengths, but Duane interpreted this
tering, now known as the Compton ford's laboratory, and he turned them as "tertiary radiation," of the brems-
effect. I do not know what Compton over and over in his mind, finally reach- strahlung type, caused by the decelera-
had been doing just before he arrived, ing an interpretation based on the trans- tion of photoelectrons ejected from the
but his appearance late that afternoon fer of momentum from light quanta to scatterer by the primary radiation. Ac-
was completely nontypical. He was di- free electrons. Again the "interpreta- tually, the shift at 90 degrees, from
sheveled, unshaven, and obviously over- tion" was not new; the idea of photons carbon, of the K x-rays of molybdenum
tired. He returned to the laboratory or light quanta had long been in the could be quantitatively accounted for
the following morning looking like him- minds of many physicists. Some had by the energy loss in the ejection of
self-a well-groomed, energetic, and even worked out Compton's equations carbon K-electrons. The crucial tests
clear-thinking physicist. for the conservation of energy and of the angular dependence of the shift,
The situation was rather tense, with momentum in the photon-electron col- and of its independence of the atomic
peculiar overtones. Compton was not lision, ending with the wry remark that number of the scatterer, had not been
the first to perform experiments which this would be a beautifully simple the- decisively performed at Harvard.
indicated that scattered x-rays and ory of scattering but was of course un- A peculiar overtone to the situation
gamma rays were more absorbable- tenable because everyone knew that was Duane's great resistance to accept-
that is, of longer wavelength-than scattered light and x-rays were un- ing a photon theory of scattering. It
their primaries. As far back as 1912 changed in wavelength and coherent was Duane and Hunt who, a few years
Sadler and Mesham had observed such with the primary radiation. Compton previously, had quantitatively estab-
an effect in x-rays scattered from car- solved the equations independently, lished the relation between the electron
bon, and Compton himself, in 1921, however, and was the first to publish kinetic energy and the maximum fre-
had followed others in experiments the results. quency of the bremsstrahlung, which,
showing the softening of gamma rays It took Compton to correlate theory in those pre-wavemechanical days, was
on scattering. But, as has several times and experiment and finally to clinch the considered one of the strongest evid-
794 SCIENCE, VOL. 138
Are Logic and Mathematics Identical?: An old thesis of Russell's is reexamined in the
light of subsequent developments in mathematical logic
Leon Henkin

Science 138 (3542), 788-794.

DOI: 10.1126/science.138.3542.788

Downloaded from http://science.sciencemag.org/ on July 4, 2019

ARTICLE TOOLS http://science.sciencemag.org/content/138/3542/788

REFERENCES This article cites 1 articles, 0 of which you can access for free
http://science.sciencemag.org/content/138/3542/788#BIBL

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Use of this article is subject to the Terms of Service

Science (print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for the
Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005. 2017 © The Authors, some rights
reserved; exclusive licensee American Association for the Advancement of Science. No claim to original U.S.
Government Works. The title Science is a registered trademark of AAAS.