Why resolving PvNP may require a paradigm shift

The philosophical and mathematical significance of Aristotle’s particularisation in the foundations of mathematics, logic and computability
Bhupinder Singh Anand Draft of November 23, 2009.

Abstract I show why efforts which are founded on a classical logic that admits Aristotle’s particularisation cannot resolve the PvNP Problem.

1

Introduction

In the September 2009 issue of the Communications of the Association of Computing Machinery, Vol. 52, #9, Lance Fortnow wrote1 : “. . . we don’t expect the P versus NP problem to be resolved in the near future . . . None of us truly understand the P versus NP problem, we have only begun to peel the layers around this increasingly complex question.” I shall seek to show why—as Fortnow appears to suggest—resolving the PvNP problem may not be the major issue; the harder part may be altering our attitudes and beliefs so that we can see what is obstructing such a resolution. In particular, I shall show, first, why efforts which are founded on a classical logic that admits Aristotle’s particularisation cannot resolve the PvNP problem; and, second, how the belief that Aristotle’s logic of predicates can be validly applied over the structure N 2 of the natural numbers 0, 1, 2, . . . , has obstructed efforts to prove that P=NP.

1.1

Why can we not immediately conclude from G¨del’s o argument that P=NP?

For instance, I note that the standard definition of the class P due to Cook3 admits a number-theoretic function F —viewed formally as defining a unique subset L of the set Σ∗ of finite strings over some non-empty finite alphabet set
1 [Fo09]. 2 See

Section 5 for the notation and definitions of technical terms as used in this paper.

3 [Cook].

1

Σ—in P if, and only if, Σ defines 4 a deterministic Turing machine TMΣ that accepts L and runs in polynomial time. It follows that: Lemma 1 A total arithmetical relation—say F ∗ (x)5 —is in P if, and only if, for any given natural number n as input, the deterministic Turing machine TMF 6 computes F ∗ (n) (when treated as a Boolean function) as either true or false in polynomial time. It further follows from Fortnow’s brief overview of Proof Complexity7 that: Lemma 2 P=NP if we can define an arithmetical relation with a single variable F ∗ (x) over N such that: (i) we can logically8 decide that, for any given natural number n, F ∗ (n) is true;9 (ii) there is no deterministic Turing machine TMF which can decide that, given any natural number n as input, F ∗ (n) is true.10 2 Now, in his seminal 1931 paper on formally ‘undecidable’ propositions of arithmetic, Kurt G¨del showed how to construct a formula—say [R(x)]11 —in o any Peano Arithmetic such that: Lemma 3 If the first-order Peano Arithmetic PA is assumed to be consistent, then:
4 The significance of making this distinction explicit is seen in Section 4, where I show why we should not admit a number-theoretic function in P merely because it is instantiationally equivalent to some member of P. 5 I shall use an asterisk to denote that the number-theoretical relation or function under consideration is arithmetical in G¨del’s sense; see [Go31], p.29. o 6 I shall use this convention to denote the deterministic Turing machine defined by the alphabet set ΣF corresponding to F . Strictly speaking, it follows immediately from Alan Turing’s seminal 1936 paper on computable numbers [Tu36] that the quantifier-free string— say [FQF ]—in the prenex normal form of any PA-formula [F ] corresponds to an alphabet set ΣP A which defines a deterministic Turing machine TMP A for computing the instantiations ∗ ∗ of the arithmetical relation FQF over N , where FQF is the interpretation of [FQF ] under a sound interpretation of PA. 7 See also [Cook]; [Ra02]. 8 The distinction sought to be emphasised here is that a logical argument—such as G¨del’s o reasoning in [Go31]—may metamathematically ([Kl52], §14, p.55) establish the existence of some method for computing an arithmetical value by finitary means (in Hilbert’s sense [Hi25], p.378. See also [Kl52], §15, p.63; [Be59], p.233; [Wa63], p.41; [EC89], p.172.), without requiring that any specific method—such as, say, a Turing machine—be necessary for computing the value. 9 I note that F ∗ (n) is a tautology over N if, and only if, for any given natural number n, F ∗ (n) is true. The definition of a tautology only requires that the truth of F ∗ (n) be decidable instantiationally; it does not require that F ∗ (n) be partial recursive, and therefore decidable algorithmically. 10 This would be the case if, for instance, F ∗ (n) were a ‘Halting’ type of function (such as the one defined by Turing in [Tu36], p.132, §8); which, of course, implies that there is no deterministic Turing machine TMF which can decide for any given natural number n as input that F ∗ (n) is true in polynomial time. 11 In his paper, G¨del refers to this formula only by its G¨del-number r ([Go31], p.25, o o eqn.12).

2

(i) for any numeral [n], the PA-formula [R(n)] is provable in PA12 ; (ii) the PA-formula [R(x)] is not provable in PA13 . 2

If we denote the interpretation of [R(x)] under an interpretation of the Arithmetic over N by R∗ (x)14 , it follows that, if PA has a sound interpretation, then we can logically conclude that, for any given natural number n, the arithmetical proposition R∗ (n) is true. The question arises: Query 1 Assuming that PA has a sound interpretation over N 15 , why can we not immediately conclude from G¨del’s argument that, if the PA-formula [R(x)] o is not provable in PA, then there is no deterministic Turing machine TMR which can decide that, given any natural number n as input, R∗ (n) is true (and therefore P=NP)?

1.2

The significance of ω-consistency for the PvNP problem

I shall show how the answer to this query is obscured by the persisting belief that Aristotle’s 2000-year old logic of predicates remains valid even when applied over an infinite domain such as N 16 . A critical component of this belief is Aristotle’s particularisation, which is the postulation that: Aristotlean Postulate 1 From the assertion: ‘It is (logically17 ) not the case that, for any given natural number n, F ∗ (n) does not hold’ we may always conclude: ‘There is some (unspecified) natural number m such that F ∗ (m) holds’.18 Now, in his 1931 paper, G¨del introduced the concept19 : o Definition 1 PA is ω-consistent if, and only if, there is no PA-formula [F (x)] such that: (i) for any given PA-numeral [n], the PA-formula [F (n)] is PAprovable; (ii) the PA-formula [¬(∀x)F (x)] is provable in PA.
[Go31], p.26(2). follows by Generalisation from G¨del’s actual argument (see [Go31], p.25(1)), which o is that the arithmetical formula [(∀x)R(x)]—to which G¨del refers only by its G¨del-number o o 17Gen r ([Go31], p.25, eqn.13)—is not provable in any Peano Arithmetic. 14 I shall aim to use this notation consistently in this paper. 15 I define such an interpretation finitarily in Section 6; and show that it is sound in Section 7; see also [An09c]. 16 Despite the indisputable objections raised by L. E. J. Brouwer in 1908 [Br08]. 17 I shall presume this qualification to be implicit henceforth. 18 Clearly, the converse is always true. 19 [Go31], p.23.
13 This 12 See

3

It follows that: Lemma 4 If PA is consistent but not ω-consistent, then there is some PAformula [F (x)] such that, under any sound interpretation—say IP A(Sound) —of PA: (i) for any given numeral [n], the PA-formula [F (n)] interprets as true under IP A(Sound) ; (ii) the PA-formula [¬(∀x)F (x)] interprets as true under IP A(Sound) . 2 Now, by Alfred Tarski’s standard definitions of the satisfaction, and truth, of the formulas of a formal system such as PA under an interpretation20 , say IP A(N ) , over N : Lemma 5 If the PA-formula [¬(∀x)F (x)] interprets as true under IP A(N ) , then it is not the case that, for any given PA-numeral [n], the PA-formula [F (n)] interprets as true under IP A(N ) . 2 If the interpretation IP A(N ) admits Aristotlean particularisation over N 21 , it follows that: Aristotlean Postulate 2 If the PA-formula [¬(∀x)F (x)] interprets as true under IP A(N ) , then there is some unspecified PA-numeral [m] such that the PA-formula [F (m)] interprets as false under IP A(N ) . 2 So, the belief that Aristotle’s particularisation holds over N yields the postulate: Aristotlean Postulate 3 If PA is consistent, there can be no PA-formula [F (x)] such that, under any sound interpretation, say IP A(N ,Sound) , of PA over N: (i) for any given numeral [n], the PA-formula [F (n)] interprets as true under IP A(N ,Sound) ; (ii) the PA-formula [¬(∀x)F (x)] interprets as true under IP A(N ,Sound) . Hence, the belief that Aristotle’s particularisation is valid over N implies22 that: Aristotlean Postulate 4 If PA is consistent, then it is ω-consistent. Clearly, if PA is ω-consistent then, since [n = n] is PA-provable for any given PA-numeral [n], we cannot have that [¬(∀x)(x = x)] is PA-provable. Thus, since an inconsistent PA proves [¬(∀x)(x = x)], an ω-consistent PA cannot be inconsistent. It follows that:
20 [Ta33]; see also [Ho01] for an explanatory exposition. However, for standardisation and convenience of expression, we follow the formal exposition of Tarski’s definitions given in [Me64], p.50. 21 As, for instance, in [Me64], pp.51-52 V(ii). 22 The above argument is made explicit in view of Davis’ remark in [Da82], p.129, that such a proof of ω-consistency may be “. . . open to the objection of circularity”.

4

Aristotlean Postulate 5 PA is consistent if, and only if, it is ω-consistent.23

2

The first paradigm shift: G¨del’s reasoning o implies that Aristotle’s particularisation does not hold over the natural numbers

Now, the first component of the paradigm shift24 lies in recognising that— contrary to accepted dogma25 —G¨del’s reasoning in Theorem VI26 of his 1931 o paper on formally undecidable arithmetical propositions admits the conclusion that (see Section 10)27 : Lemma 6 If PA is consistent, then it is not ω-consistent. It follows that: Lemma 7 Aristotle’s particularisation does not hold over N . 2 2

As the classical, ‘standard’, interpretation of PA—say IP A(Standard/T arski) 28 —appeals to Aristotle’s particularisation29 , this paradigm shift also involves recognising—again, contrary to accepted dogma—that the interpretation IP A(St− 30 andard/T arski) is not sound, and does not yield a model of PA . Now, formal quantification in computational theory is currently interpreted— as in classical logic31 —so as to admit Aristotlean particularisation over N as axiomatic32 . However, if Aristotle’s particularisation does not hold over N , it would explain to some extent why efforts to resolve the PvNP problem by arguments that appeal to classical Aristotlean logic cannot prevail.
23 It is pertinent to note that—in order to avoid intuitionistic objections to his reasoning in [Go31]—G¨del explained at some length (in his introduction on p.9 of [Go31]) that his o reasons for introducing the syntactic property of ω-consistency as an explicit assumption in his formal reasoning (see [Go31], p.23 and p.28), was to avoid appealing to the stronger, semantic, concept of classical arithmetical truth—a concept which is implicitly based on an intuitionistically objectionable logic that assumes Aristotlean particularisation is valid over N , and under which—as shown here—PA is consistent if, and only if, it is ω-consistent. Moreover—as I show in [An09b]—Rosser’s ‘extension’ of G¨del’s argument ([Ro36]) succeeds o in avoiding an explicit assumption of ω-consistency only by implicitly appealing to Aristotlean particularisation. 24 In Kuhn’s sense; see [Ku62], p.52. 25 See, for instance, Davis’ remarks in [Da82], p.129(iii) that “. . . there is no equivocation. Either an adequate arithmetical logic is ω-inconsistent (in which case it is possible to prove false statements within it) or it has an unsolvable decision problem and is subject to the limitations of G¨del’s incompleteness theorem”. o 26 [Go31], p.24-26. 27 See also [An09a]. 28 See Section 5. 29 See, for instance, [Me64], p.107 and p.52(V)(ii). 30 I note that finitists of all hues—ranging from Brouwer [Br08] to Alexander YesseninVolpin [He04]—have persistently questioned the soundness of the ‘standard’ interpretation IP A(Standard/T arski) of PA. 31 See [Hi25], p.382; [HA28], p.48; [Be59], pp.178 & 218. 32 In the sense of being intuitively obvious. See, for instance, [Da82], p.xxiv; [Rg87], p.308 (1)-(4); [EC89], p.174 (4); [BBJ03], p.102.

5

3

The second paradigm shift: Turing’s reasoning implies that arithmetical satisfiability can be defined algorithmically

The second component of the paradigm shift lies in recognising that Turing’s 1936 paper on computable numbers admits a sound, finitary, interpretation of PA—say IP A(βrouwer/ T uring) —in which arithmetical satisfiability is definable in terms of Turing-computability.33 . Specifically, this interpretation (see Section 634 ) differs from the standard interpretation IP A(Standard/T arski) of PA only in the requirement that, whereas the latter interpretation defines the satisfaction of atomic PA-formulas in terms of a subjective, and essentially unverifiable, intuitive decidability35 , the former defines the satisfaction of atomic PA-formulas under IP A(βrouwer/T uring) in terms of an objective, and instantiationally verifiable, algorithmic decidability36 . The comparative definitions of the satisfiability of the atomic formulas of PA under the two interpretations are as follows: Definition (Standard/Tarskian satisfiability) If [A(t1 , . . . , tn )] is an atomic wellformed formula of PA, A∗ (t∗ , . . . , t∗ ) the corresponding interpretation over N , n 1 and the sequence of PA-terms [a1 ], . . . , [an ] interprets over N as a∗ . . . , a∗ , n 1 then a∗ . . . , a∗ satisfies [A(t1 , . . . , tn )] under IP A(Standard/T arski) if, and only n 1 if, A∗ (a∗ , . . . , a∗ ) holds over N ; n 1 Definition (Brouwer/Turing satisfiability): If [A(t1 , . . . , tn )] is an atomic wellformed formula of PA, A∗ (t∗ , . . . , t∗ ) the corresponding interpretation over N , n 1 and the sequence of PA-terms [a1 ], . . . , [an ] interprets over N as a∗ . . . , a∗ , n 1 then a∗ . . . , a∗ satisfies [A(t1 , . . . , tn )] under IP A(βrouwer /T uring) if, and only if, n 1 A∗ (a∗ , . . . , a∗ ) is TMA∗ -computable (when treated as a Boolean expression) as n 1 holding over N , where TMA∗ is the Turing machine defined by A∗ (t∗ , . . . , t∗ )37 . n 1 The satisfaction, and truth, of compound formulas of PA under IP A(βrouwer/ follow Tarski’s standard, inductive, definitions (see Section 638 ) as usual . They yield the following consequences (see Section 740 ):
T uring) 39
33 It immediately follows from this that—as suspected, and sought, by David Hilbert [Hi27]— PA is consistent (see Section 7; see also [An09c]). 34 See also [An09c]. 35 Note that Tarski defines the formal sentence P as True if and only if p—where p is the proposition expressed by P (see [Ta33]). In other words, the sentence “Snow is white” is True if, and only if, it is subjectively true in all cases; and it is subjectively true in a particular case if, and only if, it expresses the subjectively verifiable fact that snow is white in that particular case. Thus, for Tarski the commonality of the satisfaction of the atomic formulas of a language under an interpretation (cf. [Me64], p.51(i)) is axiomatic, and need not be objectively verifiable. 36 Moreover, it follows that introduction of an algorithmic method is necessary for the decidability of the satisfaction and truth of PA-formulas under a sound interpretation of PA (see Section 9; see also [An09c]). 37 It follows from Turing’s seminal 1936 paper ([Tu36], pp.138-139) that every quantifier-free arithmetical relation F (x) (when interpreted as a Boolean function) defines a Turing machine TMF such that F (x) is TMF -computable if, and only if, for any given natural number n, TMF will compute F (n) as either true, or as false, over the structure N . 38 See also [An09c]. 39 See, for instance, [Me64], pp.49-51. 40 See also [An09c].

6

Lemma The PA-axioms PA1 to PA8 41 interpret as tautologies under the interpretation IP A(βrouwer/T uring) . 2 Lemma For any given PA-formula [F (x)], the Induction axiom schema PA9 interprets under IP A(βrouwer/T uring) as a tautology. 2 Lemma Generalisation preserves truth under IP A(βrouwer/T uring) . Lemma Modus Ponens preserves truth under IP A(βrouwer/T uring) . 2 2

Hence the interpretation IP A(βrouwer/T uring) of PA is sound, and defines a finitary model of PA.42

3.1

A case for P=NP

Moreover, it follows from the above that (see Section 843 ): Provability Theorem for PA A total arithmetical relation F ∗ (x)—when treated as a Boolean function—defines a deterministic Turing machine TMF which computes F ∗ (x) as always true (i.e., true for any given natural number input n when quantification in F ∗ (x) is interpreted Turing-computably under IP A(βrouwer/T uring) as indicated above) if, and only if, the PA-formula [F (x)] is provable in PA. 2 It follows from the Provability Theorem for PA that: Theorem 1 If an arithmetical relation F is decidable by the deterministic Turing machine TMF as a tautology metamathematically, then the formula [F ] is PA-provable. 2 It further follows that: Lemma 8 For any arithmetical relation F ∗ (x): If the PA-formula [F (x)] is not provable in PA, then there is no deterministic Turing machine TMF such that, given any natural number n as input, TMF decides that F ∗ (n) is true. 2 Hence: Corollary 1 G¨del’s relation R∗ (x) is such that: o (i) for any given natural number n, R∗ (n) is necessarily true; (ii) there is no deterministic Turing machine TMR which can show that, for any given natural number n as input, R∗ (n) is true. 2 This is the assertion that G¨del’s arithmetical relation R∗ (x) is metamatho ematically decidable as always true, but not algorithmically decidable by the deterministic Turing machine TMR as always true. It follows that: Theorem 2 P=NP.
41 See 42 We

2

Section 5 for a formal statement of the PA axioms. thus have a finitary proof that PA is consistent. 43 See also [An09c].

7

4

The third paradigm shift: Church’s Thesis is false

The third component of the paradigm shift—and possibly the most significant one for appreciating why efforts to prove P=NP remain unsuccessful—lies in recognising that—contrary to accepted dogma44 —the term ‘effective computability’ can be precisely defined. For instance, we can define: Definition 2 A computation is effectively computable if, and only if, it halts in finite time using finite resources. Definition 3 A number theoretic function F is effectively computable instantiationally if, and only if, given any sequence of natural number values for its free variables (assumed finite), F is effectively computable. Definition 4 A number-theoretic function is effectively computable if, and only if, it is effectively computable instantiationally. In other words, a number-theoretic function such as F (x) is effectively computable if, and only if, for any given natural number n, F (n) is effectively computable . Prima facie, these definitions adequately capture our intuitive understanding of the term ‘effective computability’. Now, classical theory argues that: Lemma 9 Every Turing-computable function (or relation, treated as a Boolean function) F is partial recursive , and, if F is total , then F is recursive45 . 2 Lemma 10 Every partial recursive function (or relation, treated as a Boolean function) is Turing-computable46 . 2 It follows that the following—essentially unverifiable but refutable—theses are classically equivalent47 : Standard Church’s Thesis48 A number-theoretic function (or relation, treated as a Boolean function) is effectively computable if, and only if, it is partial-recursive49 . Standard Turing’s Thesis50 A number-theoretic function (or relation, treated as a Boolean function) is effectively computable if, and only if, it is Turing-computable51 .
44 See Section 10; also [Kl52], p.300; [Me64], p.227; [Rg87], p.20; [EC89], p.85; [BBJ03], p23. 45 cf. [Me64], p.233, Corollary 5.13. 46 cf. [Me64], p.237, Corollary 5.15. 47 cf. [Me64], p.237. 48 Church’s (original) Thesis The effectively computable number-theoretic functions are the algorithmically computable number-theoretic functions [Ch36]. 49 cf. [Me64], p.227. 50 After describing what he meant by “computable” numbers in the opening sentence of his seminal paper on Computable Numbers [Tu36], Turing immediately expressed this thesis— albeit informally—as: “. . . the computable numbers include all numbers which could naturally be regarded as computable”. 51 cf. [BBJ03], p.33.

8

4.1

The Church and Turing Theses imply P=NP

However, we note that, in its classical formulations, Church’s Thesis ignores the doctrine of Occam’s razor by postulating a strong identity—and not simply a weak equivalence—between an effectively computable number-theoretic function and some algorithmically computable function (see Section 11). Consequently, Church’s Thesis (Turing’s Thesis) does not admit the possibility of an arithmetical function F that is effectively computable instantiationally (i.e., constructively computable), but not partial recursive (since it is not computable by the Turing machine TMF ). Now a consequence of the Provability Theorem for PA is that: Theorem 3 (First Tautology Theorem) G¨del’s tautology R(n) is effectively o computable as true for any given natural number n. Proof G¨del has defined a primitive recursive relation, xBP A y that holds if, o and only if, y is the G¨del-number of a PA-formula, say [Y ], and x the G¨delo o number of a PA-proof of [Y ] ([Go31], p22, dfn. 45). Since every primitive recursive relation is Turing-computable (when treated as a Boolean function), xBP A y defines a deterministic Turing machine TMB that halts on any natural number values of x and y. Now, G¨del has shown, by a meta-mathematical argument52 , that we can o construct a PA-formula [(∀x)R(x)] which is not PA-provable, but is such that [R(n)] is PA-provable for any given numeral [n]. Hence, if g[R(1)] , g[R(2)] , . . . are the G¨del-numbers of the PA-formulas [R(1)], o [R(2)], . . . , it follows that, for any given natural number n, when the natural number value g[R(n)] is input for y, the deterministic Turing machine TMB must halt for some value of x, which is the G¨del-number of some PA-proof of [R(n)]. o Thus, R(n) is effectively computable as true for any given natural number n. 2 It now follows from the First Tautology Theorem that, since there is a number-theoretic relation which—treated as a Boolean function—is effectively computable but not partial recursive53 : Corollary 2 The Church and Turing theses do not hold. 2

The significance of the Church and Turing theses for the PvNP problem is that—as noted in Section 1.1—P=NP if there is an arithmetical tautology R(x) which is not decidable by the deterministic Turing machine TMR as a tautology. It thus follows from the First Tautology Theorem that: Lemma 11 If P=NP, then the Church and Turing theses are false. Corollary 3 If the Church and Turing theses are true, then P=NP. 2 2

Since belief in the validity of the Church and Turing theses seems widespread amongst computationalists, it would explain why the obstacles to proving that P=NP appear unsurmountable.
p25(1). the axioms of a set theory such as ZF do not admit such a distinction, the above also suggests that set-theoretic arguments may not be able to resolve the PvNP problem.
53 Since 52 [Go31],

9

5

Appendix A: Notation, Definitions and Comments

Aristotlean particularisation This holds that an assertion such as, ‘There exists an unspecified x such that F ∗ (x) holds’—usually denoted symbolically by ‘(∃x)F ∗ (x)’—can always be validly inferred in the classical, Aristotlean, logic of predicates54 from the assertion, ‘It is not the case that, for any given x, F ∗ (x) does not hold’—usually denoted symbolically by ‘¬(∀x)¬F ∗ (x)’. Notation In this paper I use square brackets to indicate that the contents represent a symbol or a formula of a formal theory, generally assumed to be well-formed unless otherwise indicated by the context.
In other words, expressions inside the square brackets are to be only viewed syntactically as juxtaposition of symbols that are to be formed and manipulated upon strictly in accordance with specific rules for such formation and manipulation—in the manner of a mechanical or electronic device—without any regards to what the symbolism might represent semantically under an interpretation that gives them meaning. Moreover, even though the formula ‘[F (x)]’ of a formal Arithmetic may interpret as the arithmetical relation expressed by ‘F ∗ (x)’, the formula ‘[(∃x)R(x)]’ need not interpret as the arithmetical proposition denoted by the abbreviation ‘(∃x)R∗ (x).’ The latter denotes the phrase ‘There is some x such that R∗ (x)’. As Brouwer had noted55 , this concept is not always capable of an unambiguous meaning that can be represented in a formal language by the formula ‘[(∃x)R(x)]’. By ‘expressed’ I mean here that the symbolism is simply a short-hand abbreviation for referring to abstract concepts that may, or may not, be capable of a precise ‘meaning’. Amongst these are symbolic abbreviations which are intended to express the abstract concepts—particularly those of ‘existence’—involved in propositions that refer to non-terminating processes and infinite aggregates.

Provability A formula [F ] of a formal system S is provable in S (S-provable) if, and only if, there is a finite sequence of S-formulas [F1 ], [F2 ], ..., [Fn ] such that [Fn ] is [F ] and, for all 1 ≤ i ≤ n, [Fi ] is either an axiom of S or a consequence of the axioms of S, and the formulas preceding it in the sequence, by means of the rules of deduction of S. The structure N The structure of the natural numbers—namely, {N (the set of natural numbers); = (equality); (the successor function); + (the addition function); ∗ (the product function); 0 (the null element)}. The axioms of first-order Peano Arithmetic (PA) PA1 PA2 PA3 PA4 PA5 PA6 PA7 PA8 PA9 [(x1 = x2 ) → ((x1 = x3 ) → (x2 = x3 ))]; [(x1 = x2 ) → (x1 = x2 )]; [0 = x1 ]; [(x1 = x2 ) → (x1 = x2 )]; [(x1 + 0) = x1 ]; [(x1 + x2 ) = (x1 + x2 ) ]; [(x1 0) = 0]; [(x1 x2 ) = ((x1 x2 ) + x1 )]; For any well-formed formula [F (x)] of PA: [(F (0) → (∀x)(F (x) → F (x ))) → (∀x)F (x)].
pp.58-59. see also [An08].

54 [HA28], 55 [Br08];

10

Generalisation in PA If [A] is PA-provable, then so is [(∀x)A]. Modus Ponens in PA If [A] and [A → B] are PA-provable, then so is [B]. Standard interpretation of PA The standard interpretation IP A(Standard/ T arski) of PA over the structure N is the one in which the logical constants have their ‘usual’ interpretations56 in Aristotle’s logic of predicates, and57 : (a) the set of non-negative integers is the domain; (b) the integer 0 is the interpretation of the symbol [0]; (c) the successor operation (addition of 1) is the interpretation of the [ ] function; (d) ordinary addition and multiplication are the interpretations of [+] and [∗]; (e) the interpretation of the predicate letter [=] is the identity relation. Simple consistency A formal system S is simply consistent if, and only if, there is no S-formula [F (x)] for which both [(∀x)F (x)] and [¬(∀x)F (x)] are S-provable. ω-consistency A formal system S is ω-consistent if, and only if, there is no S-formula [F (x)] for which, first, [¬(∀x)F (x)] is S-provable and, second, [F (a)] is S-provable for any given S-term [a]. Soundness (formal system) A formal system S is sound under an interpretation IS if, and only if, every theorem [T ] of S translates as ‘[T ] is true under IS ’. Soundness (interpretation) An interpretation IS of a formal system S is sound if, and only if, S is sound under the interpretation IS .
Soundness in classical logic. In classical logic, a formal system S is sometimes defined as ‘sound’ if, and only if, it has an interpretation; and an interpretation is defined as the assignment of meanings to the symbols, and truth-values to the sentences, of the formal system. Moreover, any such interpretation is a model of the formal system. This definition suffers, however, from an implicit circularity: the formal logic L underlying any interpretation of S is implicitly assumed to be ‘sound’. The above definitions seek to avoid this implicit circularity by delinking the defined ‘soundness’ of a formal system under an interpretation from the implicit ‘soundness’ of the formal logic underlying the interpretation. This admits the case where, even if L1 and L2 are implicitly assumed to be sound, S + L1 is sound, but S + L2 is not. Moreover, an interpretation of S is now a model for S if, and only if, it is sound.58

Categoricity A formal system S is categorical if, and only if, it has a sound interpretation and any two sound interpretations of S are isomorphic.59

6

Appendix B: A finitary interpretation of PA

In the interpretation IP A(βrouwer/T uring) of PA, we interpret the non-logical constants as in the standard interpretation IP A(Standard/T arski) of PA in the ‘usual’ manner, but interpret the logical constants ‘algorithmically’. Specifically, the interpretation IP A(βrouwer/T uring) of PA is obtained if, in Tarski’s inductive definitions of the satisfaction and truth of the formulas of a
[Me64], p.49. [Me64], p.107. 58 My thanks to Professor Rohit Parikh for highlighting the need for making such a distinction explicit. 59 Compare [Me64], p.91.
57 See 56 See

11

formal system such as PA under the standard interpretation IP A(Standard/T arski) of PA—(1a) and (2)-(6) below60 —we apply Occam’s razor and weaken (1a) by replacing it with the algorithmic definition of Turing-satisfiability (1b).

6.1

Defining satisfiability

Thus, the comparative definitions of the satisfiability of the atomic formulas of PA under the two interpretations are as follows: Definition 1a (Standard/Tarskian satisfiability) If [A(t1 , . . . , tn )] is an atomic well-formed formula of PA, A∗ (t∗ , . . . , t∗ ) the corresponding interpretation over n 1 N , and the sequence of PA-terms [a1 ], . . . , [an ] interprets over N as a∗ . . . , a∗ , n 1 then a∗ . . . , a∗ satisfies [A(t1 , . . . , tn )] under IP A(Standard/T arski) if, and only if, n 1 A∗ (a∗ , . . . , a∗ ) holds over N ; n 1 Definition 1b (Brouwer/Turing satisfiability) If [A(t1 , . . . , tn )] is an atomic well-formed formula of PA, A∗ (t∗ , . . . , t∗ ) the corresponding interpretation over n 1 N , and the sequence of PA-terms [a1 ], . . . , [an ] interprets over N as a∗ . . . , a∗ , n 1 then a∗ . . . , a∗ satisfies [A(t1 , . . . , tn )] under IP A(βrouwer /T uring) if, and only if, n 1 A∗ (a∗ , . . . , a∗ ) is TMA∗ -computable (when treated as a Boolean expression) as n 1 holding over N , where TMA∗ is the Turing machine defined by A∗ (t∗ , . . . , t∗ )61 . n 1

6.2

Defining truth under an interpretation

However, both IP A(Standard/T arski) and IP A(βrouwer/T uring) identically define the satisfaction and truth of the compound formulas of PA inductively as usual under the corresponding interpretation as follows: Definition 2 A sequence s satisfies [¬A] under the interpretation if, and only if, s does not satisfy [A]; Definition 3 A sequence s satisfies [A → B] under the interpretation if, and only if, either s does not satisfy [A], or s satisfies [B]; Definition 4 A sequence s satisfies [(∀xi )A] under the interpretation if, and only if, every denumerable sequence over N which differs from s in at most the i’th component satisfies [A]; Definition 5 A well-formed formula [A] of PA is true under the interpretation if, and only if, every denumerable sequence over N satisfies [A]; Definition 6 A well-formed formula [A] of PA is false under the interpretation if, and only if, no sequence over N satisfies [A].
[Me64], pp.50-52. follows from Turing’s seminal 1936 paper ([Tu36], pp.138-139) that every quantifierfree arithmetical relation F (x) (when interpreted as a Boolean function) defines a Turing machine TMF (in the general case, TMF is defined by the quantifier-free expression in the prenex normal form of F (x)) such that F (x) is TMF -computable if, and only if, for any given natural number n, TMF will compute F (n) as either true, or as false, over the structure N .
61 It 60 cf.

12

7

Appendix C: The finitary interpretation IP A(βrouwer/T uring) of PA is sound

Since the PA-axioms PA1 to PA8 do not involve any quantification, it follows straightforwardly from Alan Turing’s 1936 seminal paper on computable numbers62 that: Lemma 12 The PA-axioms PA1 to PA8 interpret as tautologies under the interpretation IP A(βrouwer/T uring) . 2 Further: Lemma 13 For any given PA-formula [F (x)], the Induction axiom schema PA9 interprets under IP A(βrouwer/T uring) as a tautology. Proof The two meta-assertions: ‘[F (0) → (∀x)(F (x) → F (x ))] is true under IP A(βrouwer/T uring) ’ and ‘[(∀x)F (x)] is true under IP A(βrouwer/T uring) ’ both mean63 : ‘For any given natural number n, the Turing machine TMF ∗ computes F ∗ (n) as true’ where F ∗ (x) is the interpretation of [F (x)] under IP A(βrouwer/T uring) . Similarly: Lemma 14 Generalisation interprets under IP A(βrouwer/T uring) as a tautology. Proof The two meta-assertions: ‘[F (x)] is true (i.e., satisfied for any given x) under IP A(βrouwer/T uring) ’ and ‘[(∀x)F (x)] is true under IP A(βrouwer/T uring) ’ both mean: ‘For any given natural number n, the Turing machine TMF ∗ computes F ∗ (n) as true’ where F ∗ (x) is the interpretation of [F (x)] under IP A(βrouwer/T uring) . It is also straightforward to see that:
62 [Tu36]. 63 I note that the interpretation I e P A(βrouwer/T uring) settles the Poincar´-Hilbert debate (see [Hi27], p.472; also [Br13], p.59; [We27], p482; [Pa71], p.502-503.) in the latter’s favour. Poincar´ believed that the Induction Axiom could not be justified finitarily, as any such e argument would necessarily need to appeal to infinite induction. Hilbert believed that a finitary proof of the consistency of PA was possible.

2

2

13

Lemma 15 Modus Ponens preserves truth under IP A(βrouwer/T uring) . We thus have that:

2

Lemma 16 The axioms of PA are constructively satisfied/true under the finitary interpretation IP A(βrouwer/T uring) , and the rules of inference of PA preserve the properties of satisfaction/truth under IP A(βrouwer/T uring) . 2 It follows that the finitary interpretation IP A(βrouwer/T uring) of PA is sound, and so it defines a finitary model of PA.

8

Appendix D: A Provability Theorem for PA

I show that PA is categorical, and can have no non-standard model, since it is ‘algorithmically’ complete in the sense that: Theorem 4 (Provability Theorem for PA) A total arithmetical relation F ∗ (x)— when treated as a Boolean function—defines a deterministic Turing machine TMF which computes F ∗ (x) as always true (i.e., true for any given natural number input n) if, and only if, the corresponding PA-formula [F (x)] is PAprovable. Proof It follows from Turing’s seminal 1936 paper 64 that every quantifier-free arithmetical relation F ∗ (x) (when interpreted as a Boolean function) defines a deterministic Turing machine TMF such that F ∗ (x) is TMF -computable if, and only if, for any given natural number n, TMF will compute F ∗ (n) as either true, or as false, over the structure N . In the general case, TMF is defined by the quantifier-free expression in the prenex normal form of F ∗ (x), and any quantification in F ∗ (x) is to be interpreted Turing-computably under IP A(βrouwer/T uring) as detailed above. Now, we have that: (a) [(∀x)F (x)] is defined as true under the interpretation IP A(βrouwer/T uring) if, and only if, the Turing machine TMF computes F ∗ (n) as always true (i.e., as true for any given natural number n) under IP A(βrouwer/T uring) ; (b) [(∃x)F (x)] is an abbreviation of [¬(∀x)¬F (x)], and is defined as true under IP A(βrouwer/T uring) if, and only if, it is not the case that the Turing-machine TMF computes F ∗ (n) as always false (i.e., as false for any given natural number n) under IP A(βrouwer/T uring) . Moreover, since IP A(βrouwer/T uring) is sound, it defines a finitary model of PA over N —say MP A(β) —such that: If [(∀x)F (x)] is PA-provable, then the Turing machine TMF computes F ∗ (n) as always true (i.e., as true for any given natural number n) in MP A(β) ; If [¬(∀x)F (x)] is PA-provable, then it is not the case that the Turing machine TMF computes F ∗ (n) as always true (i.e., as true for any given natural number n) in MP A(β) . Further, we cannot have that both [(∀x)F (x)] and [¬(∀x)F (x)] are PAunprovable for some PA-formula F ∗ (x), as this would yield the contradiction:
64 [Tu36].

14

(i) There is a finitary model—say M 1β —of PA+[(∀x)F (x)] in which the Turing machine TMF computes F ∗ (n) as always true (i.e., as true for any given natural number n). (ii) There is a finitary model—say M 2β —of PA+[¬(∀x)F (x)] in which it is not the case that the Turing machine TMF computes F ∗ (n) as always true (i.e., as true for any given natural number n). Hence PA is ‘algorithmically’ complete, in the sense that a total arithmetical relation F ∗ (x)—when treated as a Boolean function—defines a Turing machine TMF which computes F ∗ (x) as always true if, and only if, the corresponding PA-formula [F (x)] is PA-provable65 . 2 It follows that PA can have no ‘non-standard’ model, and so we have that: Corollary 4 PA is categorical. 2

9

Appendix E: Introduction of an algorithmic method is necessary for the decidability of the satisfaction and truth of PA-formulas under a sound interpretation of PA

The question arises: Is the introduction of an algorithmic method for the decidability of the satisfaction and truth of PA-formulas under a sound interretation of PA necessary? I give an affirmative answer by showing that, even if the atomic PA-formulas are defined verifiably under an interpretation, say IP A(G¨del) , the interpretation o is not sound 66 if it admits Aristotle’s particularisation. Thus, I define the satisfaction of the atomic PA-formulas under IP A(G¨del) o as follows: Definition 1c (G¨delian satisfiability) If [A(t1 , . . . , tn )] is an atomic well-formed o formula of PA then the sequence of PA-terms [a1 ], . . . , [an ] satisfies [A(t1 , . . . , tn )] under IP A(G¨del) if, and only if, [A(a1 , . . . , an )] is PA-provable. o The satisfaction and truth of the compound formulas of PA are defined inductively under the interpretation IP A(G¨del) as usual by definitions (2)-(6) o in Section 6 above. If we accept that (1c) is a consistent definition of satisfiability, then it is straightforward to show that IP A(G¨del) is sound. o However, it also follows from the definition (1c) above and definitions (2)-(6) in Section 6 that: Lemma 17 (G¨delian universality) A PA-formula such as [(∀x)A(x)] intero prets as true under IP A(G¨del) if, and only if, for any given numeral [n], [A(n)] o is PA-provable. Lemma 18 (G¨delian particularisation) A PA-formula such as [(∃x)A(x)]— o the abbreviation for [¬(∀x)¬A(x)]—interprets as true under IP A(G¨del) if, and o only if, it is not true that for any given numeral [n], [¬A(n)] is PA-provable.
65 Note 66 Compare

that [(∀x)F (x)] is PA-provable if, and only if, [F (x)] is PA-provable. [Br08]; see also [Br23], p.336; [Br27], p.491; [We27], p.483.

15

Further, it folows from the Provability Theorem for PA in Section 8 above that the G¨delian PA-formula [R(x)] is such that: o (i) [(∀x)R(x)] is not PA-provable; (ii) [¬(∀x)R(x)] is PA-provable; (iii) for any given numeral [n], [R(n)] is PA-provable. Now, if IP A(G¨del) is sound, then (ii) implies that it is not the case that, for o any given numeral [n], [R(n)] is PA-provable. It follows that IP A(G¨del) is not sound. o We conclude that: Theorem 5 If an interpretation IP A(S) of PA lacks specification of an algorithmic method for determining the satisfaction of the atomic formulas of PA, then this lack is reflected in the non-constructivity of the interpretations of the universal and existential quantifiers of PA under IP A(S) . Corollary 5 If an interpretation IP A(S) of PA lacks specification of an algorithmic method for determining the satisfaction of the atomic formulas of PA, then IP A(S) cannot distinguish between arithmetical relations that are algorithmically decidable, and those that are instantiationally, but not algorithmically, decidable.

10

Appendix F: A consistent PA is not omegaconsistent

In his 1931 paper67 , G¨del defined a Peano Arithmetic P and showed that: o Lemma 19 If P is consistent and [(∀x)R(x)] is assumed P-provable, then [¬(∀x) R(x)] is P-provable68 . 2 By G¨del’s definition of P-provability69 , it follows that: o Lemma 20 There is a finite sequence [F1 ], . . . , [Fn ] of P-formulas such that [F1 ] is [(∀x)R(x)], [Fn ] is [¬(∀x)R(x)], and, for 2 ≤ i ≤ n, [Fi ] is either a P-axiom or a formal consequence of the preceding formulas in the sequence by the rules of inference of P. 2 Now: Lemma 21 Under every sound interpretation of P, the sequence [F1 ], . . . , [Fn ] ∗ ∗ of P-formulas interprets as a finite sequence F1 , . . . , Fn of arithmetical proposi∗ ∗ tions such that F1 is the interpretation of [(∀x)R(x)], Fn is the interpretation ∗ of [¬(∀x)R(x)], and, for 2 ≤ i ≤ n, Fi is either the interpretation of a Paxiom, or a logical consequence of the preceding formulas in the sequence by the interpretation of the rules of inference of P. 2 We thus have that:
67 [Go31]. 68 This 69 [Go31],

follows from G¨del’s argument in [Go31], p.26(1). o p.22, Definitions #44-46.

16

Lemma 22 There is no sound interpretation of P under which [(∀x)R(x)] interprets as true and [¬(∀x)R(x)] as false. 2 Since both [(∀x)R(x)] and [¬(∀x)R(x)] are closed P-formulas, it, further, follows that: Lemma 23 The formula [(∀x)R(x) → ¬(∀x)R(x)] interprets as true under every sound interpretation of P. 2
∗ ∗ In other words, the implication ‘If F1 then Fn ’ holds in any sound interpretation of P. By G¨del’s completeness theorem, it follows that: o

Lemma 24 ((∀x)R(x) → ¬(∀x)R(x)) is P-provable.
G¨del’s Completeness Theorem In any first-order predicate calculus, the o theorems are precisely the logically valid well-formed formulas (i. e. those that are true in every model of the calculus). 2

2

Since [(A → ¬A) → ¬A] is a theorem of first-order logic70 , we have, by Modus Ponens, that: Lemma 25 (¬(∀x)R(x)) is P-provable. Now, G¨del also showed that: o Lemma 26 If P is consistent, then [R(n)] is P-provable for any given Pnumeral [n]71 . 2 It follows that: Theorem 6 If P is consistent, then it is not ω-consistent. Since G¨del’s argument holds in PA, we further have that: o Theorem 7 If PA is consistent, then it is not ω-consistent.72 2 2 2

11

Appendix G: Church’s Thesis as a weaker equivalence

The reason Church’s thesis does not hold is that the thesis is expressed as a strong, refutable, identity rather than the weaker—and intuitively more plausible —equivalence: Thesis 1 (Weak Church’s Thesis) A number-theoretic function (or relation, treated as a Boolean function) is effectively computable if, and only if, it is instantiationally equivalent to a partial-recursive function. 2
70 See

[Me64], p.32, Ex.1. p.26(2). 72 See also [An08].
71 [Go31],

17

It is significant that G¨del (initially) and Church (subsequently—possibly o under the influence of G¨del’s disquietitude) enunciated Church’s formulation of o ‘effective computability’ as a Thesis because G¨del was instinctively uncomforto able with accepting it as a definition that fully captures the essence of ‘intuitive effective computability’73 . G¨del’s reservations seem vindicated if we accept that a number-theoretic o function can be effectively computable instantiationally, but not algorithmically. The possibility that ‘truth’ may be be effectively computable instantiationally, but not algorithmically, is implicit in G¨dels famous 1951 Gibbs lecture74 , o where he remarks: “I wish to point out that one may conjecture the truth of a universal proposition (for example, that I shall be able to verify a certain property for any integer given to me) and at the same time conjecture that no general proof for this fact exists. It is easy to imagine situations in which both these conjectures would be very well founded. For the first half of it, this would, for example, be the case if the proposition in question were some equation F (n) = G(n) of two number-theoretical functions which could be verified up to very great numbers n.”75 Such a possibility is also implicit in Turing’s remarks76 : “The computable numbers do not include all (in the ordinary sense) definable numbers. Let P be a sequence whose n-th figure is 1 or 0 according as n is or is not satisfactory. It is an immediate consequence of the theorem of §8 that P is not computable. It is (so far as we know at present) possible that any assigned number of figures of P can be calculated, but not by a uniform process. When sufficiently many figures of P have been calculated, an essentially new method is necessary in order to obtain more figures.” The need for placing such a distinction on a formal basis has also been expressed explicitly on occasion77 . Thus, Boolos, Burgess and Jeffrey78 define a diagonal function, d, any value of which can be computed effectively, although there is no single algorithm that can effectively compute d. Now, the straightforward way of expressing this phenomenon should be to say that there are well-defined number-theoretic functions that are instantiationally computable, but not algorithmically computable79 . Yet, following Church and Turing, such functions are labeled as effectively uncomputable80 !
73 See

[Si97].

74 [Go51]. 75 Parikh’s paper [Pa71] can also be viewed as an attempt to investigate the consequences of expressing the essence of G¨del’s remarks formally. o 76 [Tu36], §9, para II. 77 Parikh’s distinction between ‘decidability’ and ‘feasibility’ in [Pa71] also appears to echo the need for such a distinction. 78 [BBJ03], p. 37. 79 Or, preferably, one could borrow the analogous terminology from the theory of functions of real and complex variables and term such functions as computable, but not uniformly computable. 80 The issue here seems to be that, when using language to express the abstract objects of our individual, and common, mental ‘concept spaces’, we use the word ‘exists’ loosely in three senses, without making explicit distinctions between them (see [An07]).

18

“According to Turing’s Thesis, since d is not Turing-computable, d cannot be effectively computable. Why not? After all, although no Turing machine computes the function d, we were able to compute at least its first few values, For since, as we have noted, f1 = f1 = f1 = the empty function we have d(1) = d(2) = d(3) = 1. And it may seem that we can actually compute d(n) for any positive integer n—if we don’t run out of time.”81 The reluctance to treat a function such as d(n)—or the function Ω(n) that computes the nth digit in the decimal expression of a Chaitin constant Ω82 —as computable, on the grounds that the ‘time’ needed to compute it increases monotonically with n, is curious83 ; the same applies to any total Turing-computable function f (n)84 !

References
[Be59] Evert W. Beth. 1959. The Foundations of Mathematics. Studies in Logic and the Foundations of Mathematics. Edited by L. E. J. Brouwer, E. W. Beth, A. Heyting. 1959. North Holland Publishing Company, Amsterdam. George S. Boolos, John P. Burgess, Richard C. Jeffrey. 2003. Computability and Logic (4th ed). Cambridge University Press, Cambridge. L. E. J. Brouwer. 1908. The Unreliability of the Logical Principles. English translation in A. Heyting, Ed. L. E. J. Brouwer: Collected Works 1: Philosophy and Foundations of Mathematics. Amsterdam: North Holland / New York: American Elsevier (1975): pp. 107-111. L. E. J. Brouwer. 1913. Intuitionism and Formalism. Inaugural address at the University of Amsterdam, October 14, 1912. Translated by Professor Arnold Dresden for the Bulletin of the American Mathematical Society, Volume 20 (1913), pp.81-96. 1999. Electronically published in Bulletin (New Series) of the American Mathematical Society, Volume 37, Number 1, pp.55-64. L. E. J. Brouwer. 1923. On the significance of the principle of the excluded middle in mathematics, especially in function theory. Address delivered on 21 September 1923 at the annual convention of the Deutsche Mathematiker-Vereinigung in Marburg an der Lahn. In Jean van Heijenoort. 1967. Ed. From Frege to G¨del: A source o book in Mathematical Logic, 1878 - 1931. Harvard University Press, Cambridge, Massachusetts. L. E. J. Brouwer. 1927. Intuitionistic reflections on formalism. In Jean van Heijenoort. 1967. Ed. From Frege to G¨del: A source book in Mathematical Logic, 1878 o - 1931. Harvard University Press, Cambridge, Massachusetts. Gregory J. Chaitin. 1975. A Theory of Program Size Formally Identical to Information Theory. J. Assoc. Comput. Mach. 22 (1975), pp. 329-340.

[BBJ03] [Br08]

[Br13]

[Br23]

[Br27]

[Ct75]

p.37. Halting Probability is given by 0 < Ω = 2−|p| < 1, where the summation is over all self-delimiting programs p that halt, and |p| is the size in bits of the halting program p; see [Ct75]. 83 The incongruity of this is addressed by Parikh in [Pa71]. 84 The only difference being that, in the latter case, we know there is a common ‘program’ of constant length that will compute f (n) for any given natural number n; in the former, we know we may need distinctly different programs for computing f (n) for different values of n, where the length of the program will, sometime, reference n.
82 Chaitin’s

81 [BBJ03],

19

[Ch36]

Alonzo Church. 1936. An unsolvable problem of elementary number theory. In M. Davis (ed.). 1965. The Undecidable. Raven Press, New York. Reprinted from the Am. J. Math., Vol. 58, pp.345-363. Stephen Cook. The P versus NP Problem. Official description provided for the Clay Mathematical Institute, Cambridge, Massachusetts. Martin Davis. 1958. Computability and Unsolvability. 1982 ed. Dover Publications, Inc., New York. Richard L. Epstein, Walter A. Carnielli. 1989. Computability: Computable Functions, Logic, and the Foundations of Mathematics. Wadsworth & Brooks, California. Lance Fortnow. 2009. The Status of the P Versus NP Problem. Communications of the ACM, Volume 52, Issue 9 (September 2009). Kurt G¨del. 1931. On formally undecidable propositions of Principia Mathematica o and related systems I. Translated by Elliott Mendelson. In M. Davis (ed.). 1965. The Undecidable. Raven Press, New York. Kurt G¨del. 1951. Some basic theorems on the foundations of mathematics and o their implications. Gibbs lecture. In Kurt G¨del, Collected Works III, pp.304-323. o 1995. Unpublished Essays and Lectures. Solomon Feferman et al (ed.). Oxford University Press, New York. Catherine Christer-Hennix. 2004. Some remarks on Finitistic Model Theory, UltraIntuitionism and the main problem of the Foundation of Mathematics. ILLC Seminar, 2nd April 2004, Amsterdam. David Hilbert & Wilhelm Ackermann. 1928. Principles of Mathematical Logic. Translation of the second edition of the Grundz¨ge Der Theoretischen Logik. 1928. u Springer, Berlin. 1950. Chelsea Publishing Company, New York. David Hilbert. 1925. On the Infinite. Text of an address delivered in M¨nster on u 4th June 1925 at a meeting of the Westphalian Mathematical Society. In Jean van Heijenoort. 1967. Ed. From Frege to G¨del: A source book in Mathematical Logic, o 1878 - 1931. Harvard University Press, Cambridge, Massachusetts. David Hilbert. 1927. The Foundations of Mathematics. Text of an address delivered in July 1927 at the Hamburg Mathematical Seminar. In Jean van Heijenoort. 1967. Ed. From Frege to G¨del: A source book in Mathematical Logic, 1878 - 1931. o Harvard University Press, Cambridge, Massachusetts. Wilfrid Hodges. 2001. Tarski’s Truth Definitions. Stanford Encyclopedia of Philosophy. (Winter 2001 Edition). Edward N. Zalta (ed.). (Web essay: http://plato.stanford.edu/entries/tarski-truth/. Stephen Cole Kleene. 1952. Introduction to Metamathematics. North Holland Publishing Company, Amsterdam. Thomas S. Kuhn. 1962. The structure of Scientific Revolutions. 2nd Ed. 1970. University of Chicago Press, Chicago. Elliott Mendelson. 1964. Introduction to Mathematical Logic. Van Norstrand. pp.145-146. Rohit Parikh. 1971. Existence and Feasibility in Arithmetic. The Journal of Symbolic Logic, Vol.36, No. 3 (Sep., 1971), pp. 494-508. Ran Raz. 2002. P=NP, Propositional Proof Complexity, and Resolution Lower Bounds for the Weak Pigeonhole Principle. ICM 2002, Vol. III, 1-3. Hartley Rogers Jr. 1987. Theory of Recursive Functions and Effective Computability. MIT Press, Cambridge, Massachusetts.

[Cook] [Da82] [EC89]

[Fo09] [Go31]

[Go51]

[He04]

[HA28]

[Hi25]

[Hi27]

[Ho01]

[Kl52] [Ku62] [Me64] [Pa71] [Ra02] [Rg87]

20

[Ro36]

J. Barkley Rosser. 1936. Extensions of some Theorems of G¨del and Church. In M. o Davis (ed.). 1965. The Undecidable. Raven Press, New York. Reprinted from The Journal of Symbolic Logic. Vol.1. pp.87-91. Wilfried Sieg. 1997. Step by recursive step: Church’s analysis of effective calculability. Bulletin of Symbolic Logic, Volume 3, Number 2. Alfred Tarski. 1933. The concept of truth in the languages of the deductive sciences. In Logic, Semantics, Metamathematics, papers from 1923 to 1938 (p152-278). ed. John Corcoran. 1983. Hackett Publishing Company, Indianapolis. Alan Turing. 1936. On computable numbers, with an application to the Entscheidungsproblem. In M. Davis (ed.). 1965. The Undecidable. Raven Press, New York. Reprinted from the Proceedings of the London Mathematical Society, ser. 2. vol. 42 (1936-7), pp.230-265; corrections, Ibid, vol 43 (1937) pp. 544-546. Hao Wang. 1963. A survey of Mathematical Logic. North Holland Publishing Company, Amsterdam. Hermann Weyl. 1927. Comments on Hilbert’s second lecture on the foundations of mathematics. In Jean van Heijenoort. 1967. Ed. From Frege to G¨del: A source o book in Mathematical Logic, 1878 - 1931. Harvard University Press, Cambridge, Massachusetts. Bhupinder Singh Anand. 2007. Why we shouldn’t fault Lucas and Penrose for continuing to believe in the G¨delian argument against computationalism - II . The o Reasoner, Vol(1)7 p2-3. . . . . 2008. Why Brouwer was justified in his objection to Hilbert’s unqualified interpretation of quantification. Proceedings of the 2008 International Conference on Foundations of Computer Science, July 14-17, 2008, Las Vegas, USA. . . . . 2009. The significance of Aristotle’s particularisation in the foundations of mathematics, logic and computability II: G¨del and formally undecidable aritho metical propositions. Proceedings of the 2009 International Conference on Theoretical and Mathematical Foundations of Computer Science, July 13-16, Orlando, FL, USA. . . . . 2009. The significance of Aristotle’s particularisation in the foundations of mathematics, logic and computability III: Rosser and formally undecidable arithmetical propositions. Proceedings of the 2009 International Conference on Foundations of Computer Science, July 13-16, Las Vegas, NV, USA. . . . . 2009. The significance of Aristotle’s particularisation in the foundations of mathematics, logic and computability IV: Turing and a sound, finitary, interpretation of PA. Proceedings of the 2009 International Conference on Theoretical and Mathematical Foundations of Computer Science, July 13-16 2009, Orlando, FL, USA.

[Si97] [Ta33]

[Tu36]

[Wa63] [We27]

[An07]

[An08]

[An09a]

[An09b]

[An09c]

Authors postal address: 32 Agarwal House, D Road, Churchgate, Mumbai - 400 020, Maharashtra, India. Email: re@alixcomsi.com, anandb@vsnl.com.

21

Sign up to vote on this title
UsefulNot useful