By a logical measure (for a given language) we mean a syntactically defined function which associates some value with each well-formed formula of the language. Various such logical measures have played a fundamental role in the development of Logic and the Philosophy of Science. The purpose of this paper is to define a logical measure which has much wider applications than measures so far studied.

The new measure has two fundamental advantages, which will be referred to throughout the paper. First of all it can be applied to more (and richer) languages than the older measures. Most of the measures now in use are restricted to the first-order functional calculus, and frequently even to a first-order calculus with one-place predicates only. The measure to be defined will be applicable to richer languages as well, e.g., to functional calculi of all finite orders. But even as far as the first-order functional calculus is concerned, the new measure has a great advantage: We do not have to require that the atomic sentences of the calculus be independent. The previous measures depended in their construction on the requirement that “the basic statements of the language express independent facts.”

By an atomic sentence we mean a well-formed formula formed by applying an h-place primitive predicate to h individuals (a well-formed formula no part of which is well-formed); by a permissible conjunction we mean a conjunction of atomic sentences and negations of other atomic sentences; and the requirement of independence is that an atomic sentence (or its negation) is logically implied by a permissible conjunction only if it (its negation) is one of the components of the conjunction.

Beth's Definability Theorem, and consequently the Interpolation Lemma, fail for the version of quantified S5 that is presented in Kripke's [6]. These failures persist when the constant domain axiom-scheme ∀x□φ ≡ □∀xφ is added to S5 or, indeed, to any weaker extension of quantificational K.

§1 reviews some standard material on quantificational modal logic. This is in contrast to quantified intermediate logics for, as Gabbay [6] has shown, the Interpolation Lemma holds for the logic CD with constant domains and for several of its extensions. §§2—4 establish the negative results for the systems based upon S5. §5 establishes a more general negative result and, finally, §6 considers some positive results and open problems. A basic knowledge of classical and modal quantificational logic is presupposed.

Let me briefly review the relevant model theory for quantified modal logic. Further details can be found in [3] or [7].

The language is obtained from the language for classical first-order logic with identity by adding a unary operator □ for necessity. The atomic formula ‘Ex’ is used as an abbreviation for ‘∃y(y = x)’ and may be read as ‘x exists’.

Let be a relational system whereby D is a nonempty set and P1 is an m1-ary relation on D. With we associate the (weak) monadic second-order theory consisting of the first-order predicate calculus with individual variables ranging over D; monadic predicate variables ranging over (finite) subsets of D; monadic predicate quantifiers; and constants corresponding to P1, P2, …. We will often use ambiguously to mean also the set of true sentences of .

PA is Peano arithmetic. The formula InterpPA(α, β) is a formalization of the assertion that the theory PA + α interprets the theory PA + β (the variables α and β are intended to range over codes of sentences of PA). We extend Solovay's modal analysis of the formalized provability predicate of PA, PrPA(x), to the case of the formalized interpretability relation InterpPA(x, y). The relevant modal logic, in addition to the usual provability operator ‘□’, has a binary operator ‘⊳’ to be interpreted as the formalized interpretability relation. We give an axiomatization and a decision procedure for the class of those modal formulas that express valid interpretability principles (for every assignment of the atomic modal formulas to sentences of PA). Our results continue to hold if we replace the base theory PA with Zermelo-Fraenkel set theory, but not with Gödel-Bernays set theory. This sensitivity to the base theory shows that the language is quite expressive. Our proof uses in an essential way earlier work done by A. Visser, D. de Jongh, and F. Veltman on this problem.

The modern problem of infinity was first raised by Aristotle who held (at least on the popular interpretation) that infinite sets exist potentially (i.e. one more number can always be counted, one more division can always be made in a line segment) but not actually (i.e. the numbers or divisions cannot all exist at one time). In fact, Aristotle not only held that completed infinities never actually exist, but also that they are impossible, that is, that the assumption that they do exist leads to contradictions. To see this, consider Aristotle's view that mathematical entities depend for their existence on the existence of primary substance in which they inhere, coupled with his view that there can be no infinite body. From these it follows that if there were actual completed infinities, infinite bodies would both exist and not exist. The details here are not so important as the idea that if a completed infinite is assumed to exist a contradiction follows.

This negative attitude towards completed infinities flourished for centuries. By the mid-1500's, the German mathematician Stifel (who apparently also invented Pascal's triangle) was moved to condemn irrational numbers simply by their association with the completed infinite:

About a hundred years later, Galileo further discredited the completed infinite by pointing out that line segments of different lengths can be brought into one-to-one correspondence by projection, as can the natural numbers and the perfect squares by assigning each number to its square.

The purpose of the present paper is to give a new, simple proof of the theorem of M. Davis, H. Putnam and J. Robinson [1961], which states that every recursively enumerable relation A(a1, …, an) is exponential diophantine, i.e. can be represented in the form

where a1 …, an, x1, …, xm range over natural numbers and R and S are functions built up from these variables and natural number constants by the operations of addition, A + B, multiplication, AB, and exponentiation, AB. We refer to the variables a1,…,an as parameters and the variables x1 …, xm as unknowns.

Historically, the Davis, Putnam and Robinson theorem was one of the important steps in the eventual solution of Hilbert's tenth problem by the second author [1970], who proved that the exponential relation, a = bc, is diophantine, and hence that the right side of (1) can be replaced by a polynomial equation. But this part will not be reproved here. Readers wishing to read about the proof of that are directed to the papers of Y. Matijasevič [1971a], M. Davis [1973], Y. Matijasevič and J. Robinson [1975] or C. Smoryński [1972]. We concern ourselves here for the most part only with exponential diophantine equations until §5 where we mention a few consequences for the class NP of sets computable in nondeterministic polynomial time.

A boolean algebra is shown to be completely representable if and only if it is atomic, whereas it is shown that neither the class of completely representable relation algebras nor the class of completely representable cylindric algebras of any fixed dimension (at least 3) are elementary.

The infinite-valued statement calculus to which this paper refers is that of Łukasiewicz [10], whose axiomatization was proved complete in [5]. In [9], Rutledge extended this system to include predicates and quantifiers2 and presented a deductively complete set of axioms for the monadic predicate calculus. This paper represents an attempt to axiomatize the full predicate calculus; for the proposed axiomatization, a property akin to but weaker than completeness is proved. An attempt to prove full completeness along similar lines failed; it has since been shown [11] that the set of valid formulas of the infinite-valued predicate calculus is not recursively enumerable. The method of this paper was suggested by Professor J. Barkley Rosser.

We prove that by showing that for any set C not of PA-degree and any set A, there exists an infinite subset G of A or such that G ⊕ C is also not of PA-degree.

Dummett's LC [1] is a system which characterizes all formulas of the propositional calculus which are valid in every chain (for definitions and notation see the first section of [2]). An L-algebra is a Heyting algebra in which (x → y) + (y → x) = 1 for all x, y. L-algebras bear the same relation to LC as Boolean algebras to the classical propositional calculus and Heyting algebras to the intuitionist propositional calculus.

Below are collected some simple results on the theory of free choice sequences [4] or infinitely proceeding sequences (ips) as they are called in [5]. These results are not sufficient to settle the completeness problems for Heyting's predicate calculus [4], as formulated in [12], neither in the strong nor in the weak sense. They are published because they lead to intuitionistically valid versions of completeness proofs which have appeared in the literature, particularly [16], [17], [18], and, with certain reservations, [1].

The problems considered below (except in §8) differ from those of [12] in the following respect: In [12] we were mainly concerned with formulae of Heyting's predicate calculus which were not even classically provable, and showed that the calculus was complete with respect to certain classes of these formulae. The novel feature was that we established this completeness by means of intuitionistically valid methods, in fact methods which can be formalized in Heyting's arithmetic. Here we are primarily concerned with formulae which are classically, but not intuitionistically, provable. As pointed out in [12], the nature of the completeness problem for such formulae is totally different: the class of predicates which provides the required counterexamples must come from a system (with an intuitionistically acceptable interpretation) which, unlike Heyting's arithmetic, is not a subsystem of the corresponding classical system. An example of such a system is an extension FC, given below, of Heyting's formalization of the theory of free choices.

This paper is motivated by the question whether there exists a logic capturing polynomial time computation over unordered structures. We consider several algorithmic problems near the border of the known, logically defined complexity classes contained in polynomial time. We show that fixpoint logic plus counting is stronger than might be expected, in that it can express the existence of a complete matching in a bipartite graph. We revisit the known examples that separate polynomial time from fixpoint plus counting. We show that the examples in a paper of Cai, Fürer, and Immerman, when suitably padded, are in choiceless polynomial time yet not in fixpoint plus counting. Without padding, they remain in polynomial time but appear not to be in choiceless polynomial time plus counting. Similar results hold for the multipede examples of Gurevich and Shelah, except that their final version of multipedes is, in a sense, already suitably padded. Finally, we describe another possible candidate, involving determinants, for the task of separating polynomial time from choiceless polynomial time plus counting.

A. H. Lachlan [2] and C. E. M. Yates [4] independently showed that minimal pairs of recursively enumerable (r.e.) degrees exist. Lachlan and Richard Ladner have shown (unpublished) that there is no uniform method for producing a minimal pair of r.e. degrees below a given nonzero r.e. degree. It is not known whether every nonzero r.e. degree bounds a r.e. minimal pair, but in the present paper it is shown (uniformly) that every high r.e. degree bounds a r.e. minimal pair. (A r.e. degree is said to be high if it contains a high set in the sense of Robert W. Robinson [3].)

Theorem. Let a be a recursively enumerable degree for which a′ = 0″. Then there are recursively enumerable degrees b0 and b1 such that0 < bi < a for each i ≤ 1, and b0 ⋂ b1 = 0.

The proof is based on the Lachlan minimal r.e. pair construction. For notation see Lachlan [2] or S. B. Cooper [1].

By Robinson [3] we can choose a r.e. representative A of the degree a, with uniformly recursive tower {As, ∣ s ≥ 0} of finite approximations to A, such that CA dominates every recursive function where

We define, stage by stage, finite sets Bi,s, i ≤ 1, s ≥ 0, in such a way that Bi, s + 1 ⊇ Bi,s for each i, s, and {Bi,s ∣ i ≤ 1, s ≥ 0} is uniformly recursive.

Si vous ne gardez du calcul des prédicats classique (que j'appellerai Lω) que les formules qui ne font intervenir, tant en occurences libres que liées, que les n premières variables, vous obtenez le langage Ln. Avant de poursuivre, et de façon un peu cavalière, je vous invite à résoudre l'exercice suivant:

Exercice 1. Dans le langage comprenant seulement un symbole de relation binaire, caractériser à isomorphisme près la chaîne à n éléments au moyen d'un seul énoncé de L2.

Vous avez vu que Ln a plus de possibilités d'expression que vous ne l'aviez d'abord pensé, qu'il est ennemi de la forme prénexe, que sa force au contraire réside dans l'utilisation d'une même variable en différentes places d'une formule.

Ces langages ont été étudiés par Léon Henkin dans [1]; il y remarque que L1 est très pauvre: c'est le “calcul des classes”, on y peut seulement exprimer que des combinaisons booléennes de formules atomiques sont vides ou non, et nous n'en parlerons plus; il montre ensuite que tout théorème de L2est prouvable par une suite de déductions (d'un système classique) qui ne fait intervenir que des énoncés de L3. Dans [2], Donald Monk construit pour tout n un théorème de L3 dont toute preuve utilise des énoncés avec au moins n variables. Dans [3], Dana Scott prouve que L2 est décidable; ce résultat est précisé par Michael Mortimer dans [4], qui montre que tout énoncé consistant f de L2 a un modèle fini, de cardinal borné effectivement en fonction de f. Pour n supérieur ou égal à 3, Ln n'est plus décidable car il a même pouvoir d'expression que Lω, comme il est montré dans l'exercice suivant.

In Anderson and Belnap [8] there was developed a semantics for first degree entailments (fde), i.e., entailments A → B between formulas A and B involving only truth-functions (defined in terms of “or” and “not”) and quantifiers. The key ideas were (i) the notion of a frame ⟨P, FIP, I⟩, where P is a set of (intensional) propositions closed under negation and multiple disjunction, I is a domain of individuals, and FIP is the set of functions from I into P; and (ii) the semantic relation of cons (consequence), as obtaining between a set of propositions taken conjunctively, and a set taken disjunctively; and (iii) the notion of an atomic frame, i.e., a frame generated by a set of propositions X closed under negation, such that for any disjoint subclasses Y and Z of X, Y does not bear cons to Z.

In 1947 E. Post [28] and A. A. Markov [18] independently proved the undecidability of the word problem (or the problem of deducibility of relations) for semigroups. In 1968 V. L. Murskil [23] proved the undecidability of the identity problem (or the problem of deducibility of identities) in semigroups.

If we slightly generalize the statement of these results we can state many related results in the literature and state our new results proved here. Let V denote either a (Birkhoff) variety of semigroups or groups or a pseudovariety of finite semigroups. By a very well-known theorem a (Birkhoff) variety is defined by equations or equivalently closed under substructure, surmorphisms and all products; see [7]. It is also well known that V is a pseudovariety of finite semigroups iff V is closed under substructure, surmorphism and finite products, or, equivalently, determined eventually by equations w1 = w1′, w2 = w2′, w3 = w3′,… (where the finite semigroup S eventually satisfies these equations iff there exists an n, depending on S, such that S satisfies Wj = Wj′ for j ≥ n). See [8] and [29]. All semigroups form a variety while all finite semigroups form a pseudovariety.

We now consider a table (see the next page). In it, for example, the box denoting the “word” (identity) problem for the psuedovariety V” means, given a finite set of relations (identities) E and a relation (identity) u = ν, the problem of whether it is decidable that E implies u = ν inside V.

Acquaintance with [11] and [8] will be assumed. Let be the alphabet →, &, ν, ┐, (,), A1, A2, … where →(I), &(C), ν (D) and ┐(N) are called the connectives. The letters appearing in parentheses are the names of the connectives. A1, A2 … are the atoms.

Definition 1.1. A string S is an ICDN word iff there is a finite sequence of strings S1 … Sn = S such that for i, 1 ≤ i ≤ n, either S1 is an atom or S1 is one of the forms (S1→Sk), (S1 & Sk), (S1, ν Sk) or ┐S1,j, k ≤ i.

A standard set of axioms for the intuitionist propositional calculus is the following (we associate to the right, and use the convention that ⊃ binds less strongly than the other connectives).

We study the complexity of the isomorphism relation on classes of computable structures. We use the notion of FF-reducibility introduced in [9] to show completeness of the isomorphism relation on many familiar classes in the context of all equivalence relations on hyperarithmetical subsets of ω.

In A note on the Entscheidungsproblem the author gave a proof of the unsolvability of the general case of the Entscheidungsproblem of the engere Funktionenkalkül. This proof, however, contains an error, in order to correct which it is necessary to modify the “additional axioms” of the system L so that they contain no free variables (either free individual variables or free propositional function variables).

The additional axioms of L other than x=y→[F(x)→F(y)] contain no free propositional function variables, and hence it is sufficient to replace each one by an expression obtained from it by quantifying all the individual variables by means of universal quantifiers initially placed—thus, in particular, x = x is replaced by (x)[x=x]. Moreover the axiom x=y→[F(x)→F(y)] may be replaced by the following set of axioms:

and similar axioms for each of the functions b1, b2, …, bk.