The Elements of Mathematical Logic PAUL C. ROSENBLOOM DEPARTMENT 1950 DOVER PUBLICATIONS, INC. ASSOCIATE PROFESSOR, OF MATHEMATICS, SYRACUSE UNIVERSITY THE DOVER SERIES IN MATHEMATICS AND PHYSICS w. prager,ConsultingEditor > FirstEdition 1950 J3V PUBLICATIONS, 1780 BROADWAY, NEW 19, N. Y. Printed& Boundinthe UnitedStatesofAmerica COPYRIGHT DOVER INC. YORK This book is intendedforreaders who, while mature mathe- matically, have no knowledge of mathematical logic. We at- tempttointroducethereader tothemostimportantapproaches tothesubject, and, whereverpossiblewithin thelimitations of space which we have set for ourselves, to give at least afew nontrivialresultsillustrating each of theimportant methodsfor attacking logical problems. Since Lewis' survey of symbolic logic and Jorgensen's treatise on formal logic, both of which are now obsolete, the onlywork of this naturehasbeen theexcellentbook of Church, which is notsuitableforbeginners and whichis not easilyaccessible. Thus thepresentbook differs from those which confine themselves to the detailed develop- ment of oneparticidar system offormal logic. We haveempha- sizedinstead themodern tendencyof analyzing thestructureof asystemasawhole. Wefeel thattoomanyauthorsin thisfield haveoverlookedthenecessityofexhibiting thepowerof logical methodsin non-trivial problems. Otherwise mathematical logic is amereshorthandfor transcribingresults obtainedwithout its aid,notatoolforresearch and discovery. Thusin thechapteron thelogic ofclasses wehavea section on thestructureand representation ofBoolean algebras,which isappliedin thenextchaptertothestudyof deductivesystems. In the thirdchapter wesketch the methods ofRussell, Quine, Zermelo, Curry, and Church for the construction of logics of propositional functions. Finally, we give a brief introduction to the general syntax of language, with applications to unde- cidabilityand incompleteness theorems. Wehaveattemptedtomake theexposition aselementaryas possible throughout. Nevertheless, those who are unfamiliar with modern algebramay find it advisable to skip the proofs in Chapter I,Section 3, on thefirst reading. PREFACE In thelast chapterweuse theprofoundandbeautiful ideasof Post. Wehopethat oneby-product of thisbook will be amore widespread recognition and appreciation of his work, which amountstothecreation of anewbranch of mathematics of the same fundamental importance as algebraand topology. The connoisseur mayfind of some interest (1) theinsistence onthedemonstrablepropertiesofaformal systemasacriterion for its acceptability, (2) thesimpleproof of thecompletenessof thetheoryofcombinators,* (3)thesimpleexplicitexampleof a recursively unsolvable problem in elementary number theory, (4) thefirst connectedexposition of alltheessentialstepsin the proof of Church'stheorem ontherecursive unsolvability of the decisionproblemfor therestricted function calculus. Much of thematerial waspresentedin a coursegivenby the author atLund University, Sweden,in the spring of 1948. It is impossiblefor me to expressadequatelymy debttothe late Professor H. B. Smith forhis constantkindness and gen- erosity. I am grateful to Professors Churchman, Post, Curry, McKinsey, Huntington, and Stone for their friendly encour- agement when I was beginning my mathematical career. I cannotrefrain from also thankingProfessors Cohenand Nagel, since it was a misinterpretation of a footnote in their book which led me to abandon chemistry for mathematics twelve years ago! I thank Dover Publications, Inc. for its unfailing courtesy and helpfulness during the preparation of this book. Finally, I should like to express my gratitude to my beloved wife, Elly, for providing the stimulus and the working condi- tionswithout whichthe bookcould nothavebeenwritten. October 11, 1949 PAUL C. ROSENBLOOM Syracuse,NewYork *Curryhas arrivedindependentlyatessentially thesamesimplification ofthetheoryofcombinators.Thisappearedsincetheabovewaswrittenin Synthase,Vol.VII, 1948^9, No.6-A,p.391-398. IInnttrroodduuccttiioonn 1 CHAPTER I. THELOGIC OFCLASSES Section 1. Informal introduction. Fundamental theorems 1 Section 2. Boolean algebraas adeductivescience 8 Section 3. Thestructure andrepresentationof Boolean algebras 18 11. THE LOGIC Section 1. Fundamentals 28 Section2. Alternativeformulations 31 Section3. Deductivesystems 44 Section 4. Manyvalued logics, modal logics, intuitionism 51 CHAPTER 111. THELOGIC OF PROPOSITIONAL FUNCTIONS Section 1. Informal introduction 69 Section 2. The functional logic ofthe firstorder 74 Section3. Someveryexpressive languages 91 Section4. Combinatorylogics 109 Section5. The developmentof mathematics within an objectlanguage 133 Section 6. Theparadoxes 144 Section 7. Theaxiom ofchoice 146 CONTENTS CHAPTER OF PROPOSITIONS CHAPTERIV. THE GENERAL SYNTAXOFLANGUAGE Section 1. Basic concepts. Simple languages 152 Section 2. Production, canonical languages, extension, and definition 157 Section3. Normal languages. Theorems of Post and Godel 170 appendix 1. Canonical forms ofLx ,L 2, andL 182 z appendix2. Algebraic approach to language. Church's theorem 189 Bibliographical and OtherRemarks 194 Index 209 Inthisbook weshall studythelaws oflogicbymathematical methods. This mayseemunfair,sincelogic isusedinconstruct- ingmathematical proofs, and it might appearthat thestudy of logic shouldcome before thestudy of mathematics. Suchapro- cedureis,however,typical of science.Ouractualknowledge is a narrow band of light flanked on both sides by darkness. We may, on theone hand, goforward and developfurther thecon- sequences ofknown principles. Orelse we maypressbackward theobscurityin which thefoundations of scienceareenveloped. Justbyusingmathematical methods, i.e. by workingwith ideo- grams (symbolsfor ideas) instead of ordinary words (symbols for sounds),wecanthrownewandimportantlight onthelogical principles used in mathematics. This approachhas ledtomore knowledge about logic in one century than had been obtained from the death of Aristotle up to 1847, when Boole's master- piecewas published. Webegin withthesimplest branch ofthesubject, thelogic of classes. After an informal introduction, in whichwederive the propertiesofclassesbyafreeuseofnaiveintuition,weformulate that theoryasadeductive science, thatis, asasciencein which the assumptions are explicitly stated, and in which everything else follows from theassumptions bymeans ofexplicitly stated rules. The assumptions are stated in terms of certain notions which are notanalyzedfurther and are takenasundefined. All otherconcepts of thescience aredefined in termsofthese. Wethenproceedtoastudyof thesystemasawhole. Thatis, instead of developing moreand more consequences of the as- sumptions, we trytofind general characteristics of the science itself. Thisis typical of themodern tendencyto emphasize the structure of a science, to derive theorems about the science, rather thanto concentrateonthe detailedderivation of results i INTRODUCTION II within the science. This study of the structure of the logic of classesculminates in ChapterI, Section3. We then applythesame methodstothelogic ofpropositions. In doing this, we uncover a striking similarity between this scienceand thelogic of classes. It isprecisely throughformulat- ing theselogics as deductivesciencesthatwesee thatboth are special examplesof ageneraltheory. The logic of propositions has beenthe subject of much con- troversy amonglogicians and mathematicians. We discuss the various alternative approaches whichhavebeen proposed. We then try toconstruct general logical theories which are adequatefor at leasta largepartof mathematics. Herewerun into difficulties since theunreineduse of naiveintuitivereason- ingleadstodevastatingparadoxes.Thuswemustseekatheory which admits as much as possible of thereasoning intuitively acceptedasvalid, but includes such restrictions astoevadethe paradoxes. But a profound theorem of Godel shows that no logical theory of avery general type can include methods of reasoning strong enough for the proof of its ownconsistency. Indeed, in any system of logic of this general type, there are propositions which can be proved by an argumentoutside the systembut whichcannotbeprovedwithinthesystem.Thusno formal logical system ofthis type, whichincludesall adequate logicssofar proposed,can contain allvalid modesofreasoning. Allthatwe can hopeforis strongerand strongersystems which are adequatefor more and more powerful arguments, or else some systemradically different from anything sofar proposed. Inordertoarrive atsuchresultsasGodel's,itis necessaryfor us to scrutinize our tools more closely. In a deductive science theundefinedtermsaredenotedbycertain symbols, whichmay be blobs ofprinter's ink, speech sounds, printed marks repre- senting thelatter, etc. Thepropositions of thissciencearecom- municated by means ofthesesigns. These signs, togetherwith the rules governing their use and combination, constitute a language for stating relationships within the science. This is called the object language. In an exposition of the science the assumptions mustbecommunicated in alanguagewhosemean- 111 ingis alreadyassumedtobeknown, sayEnglish. Thisis called thesyntax language. We use the object languageto talk within thescience and the syntaxlanguage to talk about the science. In ordinaryusage the confusion betweenthe two leads to no difficulty,but whenthescienceunderconsideration islogic itself wemust lean overbackwards toavoidunclarity. Theprimitive signs of theobjectlanguagearecalledits alpha- bet. Certaincombinations of thesesigns maybe assignedmean- ings.Suchcombinations areoftencalled wordsorsentences. If a certain combination of signs denotesan object, then this com- bination will be anamefor that object. In speakingabout the objectweuseanameforit. Thus "Deweysmiled" is asentence wherein we mention the man Dewey by using his name, the word "Dewey."Whenwearetalking aboutanameorasymbol, it is convenienttouse a specimenenclosed in quotation marks asanameofthenameorsymbol. Thus," "Dewey""isanameof "Dewey,"which is, in turn, anameof Dewey, whois aman. Again, onp. 2,25thlinefrom thebottom, wearespeakingabout anameof theuniversalclass, whileonthenextlineitisthenull class itself which is mentioned. To avoid the use of names of names of namesand thelike, weshall also use suchphrases as "the letter—" or "the sign —" as names of the symbols of which specimensareexhibited. Itis oftenoverlookedthat while we cannot put a man on the printed page and are thusforced to use a name when writing about him, we do have greater resources when we wish to write about symbols. Inparticular, asentenceis aname of aproposition. We shall say that the sentenceexpresses the proposition, and we shall often use "statement" as a synonymfor "sentence." We shall often use the phrase "the proposition that p" to indicate the proposition expressedby "p."Carefulattentiontothesematters helpsin discussing ticklish questions. We arethusled, inchapter IV,tothemathematical analysisof language. Whereas in the previous chapters our attention is centered on therelationships expressedby the objectlanguage, in thelast chapterwefocus our attentionon thestructureof the languageapartfrom its meaning.The former process is some- IV times called thesemantical study of language,i.e. thestudy of themeaningsexpressedby thelanguage,whilethelatter is often called the syntactical study of the language. The methods we use were developedespecially by Post. Wefind in this chapter that certain classes of languages, which include practically all languages whichhavebeen preciselyformulated, canbe singled out and possessimportant common properties. It is exactlythe mathematical method ofabstractingfrom thespecialfeatures of particular languages which enables us to prove rigorously a numberofprofound generaltruths,wheremetaphysicianswould argueback andforthfor centurieswithout everreaching acon- clusion whichcould be tested. Mathematicallogicis, then,nomereshorthandfor expressing in ideograms what hasalreadybeen discoveredbyreasoning in ordinarylanguage. It is,rather, apowerfuland versatile toolfor solvingproblemswhichare inaccessibleto other methods. In thefollowingwe shall make references thus: 1112 denotessection 2ofchapter III; T2 denotestheorem 2 of thepresentsection; T5.2.3 denotestheorem3of section 2ofchapter V; [27]4 denotes number 4by author 27 in Church's Bibliog- raphy, J.SymbolicLogic, vol. 1,no. 4; [11]35 denotesthearticlebeginning orreviewed onpage 35 of vol. 11, J.SymbolicLogic.