Syntactic translations of classical logic C into intuitionistic logic I are well known (see [Kol25], [Gli29], [Göd32], [Kre58b], [M063], [Cel69] and [Lei71]). Harvey Friedman [Fri78] used a translation of a similar nature, from I into itself, to reprove a theorem of Kreisel [Kre58a] that various theories based on I are closed under Markov's rule: if ¬¬∃x.α is a theorem, where x is a numeric variable and α is a primitive recursive relation, then ∃x.α is a theorem. Composing this with Gödel's translation from classical to intuitionistic theories, it follows that the functions provably recursive in the classical version of the theories considered are provably recursive already in their intuitionistic version. This conservation result is important in that it guarantees that no information about the convergence of recursive functions is lost when proofs are restricted to constructive logic, thus removing a potential objection to the use of constructive logic in reasoning about programs (see [C078] for example). Conversely, no objection can be raised by intuitionists to proofs of formulas that use classical reasoning, because such proofs can be converted to constructive proofs (this has been exploited extensively; see [Smo82]).

Proofs of closure under Markov's rule had required, until Friedman's proof, a relatively sophisticated mathematical apparatus. The chief method used Godel's “Dialectica” interpretation (see [Tro73, §3]). Other proofs used cut-elimination, provable reflection for subsystems [Gir73], and Kripke models [Smo73]. Moreover, adapting these proofs to new theories had required that the underlying meta-mathematical techniques be adapted first, not always a trivial step.

Our purpose is to delineate the extensions (normal and otherwise) of the propositional modal logic K5. We associate with each logic extending K5 a finitary index, in such a way that properties of the logics (for example, inclusion, normality, and tabularity) become effectively decidable properties of the indices. In addition we obtain explicit finite axiomatizations of all the extensions of K5 and an abstract characterization of the lattice of such extensions.

This paper refines and extends the Ph.D. thesis [2] of the first-named author, who wishes to acknowledge his debt to Brian F. Chellas for his considerable efforts in directing the research culminating in [2] and [3]. We also thank W. J. Blok and Gregory Cherlin for observations which greatly simplified the proofs of Theorem 3 and Corollary 10.

By a logic we mean a set of formulas in the countably infinite set Var of propositional variables and the connectives ⊥, →, and □ (other connectives being used abbreviatively) which contains all the classical tautologies and is closed under detachment and substitution. A logic is classical if it is also closed under RE (from A↔B infer □A ↔□B) and normal if it is classical and contains □ ⊤ and □ (P → q) → (□p → □q). A logic is quasi-classical if it contains a classical logic and quasi-normal if it contains a normal logic. Thus a quasi-normal logic is normal if and only if it is classical, and if and only if it is closed under RN (from A infer □A).

In the wake of Silver's breakthrough on the Singular Cardinals Problem (Silver [74]) followed one of the landmark results in set theory, Jensen's Covering Lemma (Devlin-Jensen [74]): If 0# does not exist then for every uncountable x ⊆ ORD there exists a constructible Y ⊇ X, card(Y) = card(X). Thus it is fair to say that in the absence of large cardinals, V is “close to L”.

It is natural to ask, as did Solovay, if we can fairly interpret the phrase “close to L” to mean “generic over L”. For example, if V = L[a], a ⊆ ω and if 0# does not exist then is V-generic over L for some partial ordering ∈ L? Notice that an affirmative answer implies that in the absence of 0#, no real can “code” a proper class of information.

Jensen's Coding Theorem provides a negative answer to Solovay's question, in a striking way: Any class can be “coded” by a real without introducing 0#. More precisely, if A ⊆ ORD then there is a forcing definable over 〈L[A], A〉 such that ⊩ V = L[a], a ⊆ ω, A is definable from a. Moreover if 0# ∉ L[A] then ⊩ 0# does not exist. Now as any M ⊨ ZFC can be generically extended to a model of the form L[A] (without introducing 0#) we obtain: For any 〈M, A〉 ⊨ ZFC (that is, M ⊨ ZFC and M obeys Replacement for formulas mentioning A as a predicate) there is an 〈M, A〉-definable forcing such that ⊩ V = L[a], a ⊆ ω, 〈M, A〉 is definable from a. Moreover if 0# ∉ M then ⊩ 0# does not exist.

Let ℜ = (R, + R,…) be the ordered field of real numbers. It will be shown that the -theory of ℜ is decidable, where denotes the Malitz quantifier of order n in the ℵ1-interpretation.

In this paper we prove that if T is the complete elementary diagram of a countable structure and is a theory as in the title, then Vaught's conjecture holds for T. This result is Theorem 7, below. In the process of establishing this proposition, in Theorem 3 we give a sufficient condition for a superstable theory having only countably many types without parameters to be ω-stable. Familiarity with the rudiments of stability theory, as presented in [3] and [4], will be supposed throughout. The notation used is, by now, standard.

We begin by giving a new proof of a lemma due to J. Saffe in [6]. For T stable, recall that the multiplicity of a type p over a set A ⊆ ℳ ⊨ T is the cardinality of the collection of strong types over A extending p.

Lemma 1 (Saffe). Let T be stable, A ⊆ ℳ ⊨ T. If t(b̄, A) has infinite multiplicity and t(c̄, A) has finite multiplicity, then t(b̄, A ∪ {c̄}) has infinite multiplicity.

Proof. We suppose not and work for a contradiction. Let ‹b̄γ:γ ≤ α›, α ≥ ω, be a list of elements so that t(b̄γ, A) = t(b̄, A) for all γ ≤ α, and st(b̄γ, A) ≠ st(b̄δ, A) for γ ≠ δ. Furthermore, let c̄γ satisfy t(b̄γ∧c̄γ, A) = t(b̄ ∧ c̄, A) for each γ < α.

Since t(c̄, A) has finite multiplicity, we may assume for all γ, δ < α. that st(c̄γ, A) = st(c̄δ, A). For each γ < α there is an automorphism fγ of the so-called “monster model” of T (a sufficiently large, saturated model of T) that preserves strong types over A and is such that f(c̄γ) = c̄0.

In algebra, the properties of having the (strong) amalgamation property and epis being surjective are well investigated; see the survey [10]. In algebraic logic it is shown that to these algebraic properties there correspond interesting logical properties, see e.g. [15], [12], [4], and [8, p. 311, Problem 10 and the remark below it]. In the present paper we show that the varieties Crsα (of cylindric-relativised set algebras) and Boα (of Boolean algebras with operators) have the strong amalgamation property. These contrast to the following result proved in Pigozzi [15]: No class K with Gsα ⊆ K ⊆ CAα has amalgamation property. Note that Gsα ⊆ Crsα ⊆ Boα and CAα ⊆ Boα. For related results see [3], [1], [16], [11]. For more connections with logic and abstract model theory see [14] and §4.3 of [9].

BA denotes the class of all Boolean algebras. Let α be any ordinal. From now on, throughout in the paper, α is an arbitrary but fixed ordinal. Recall from [7, p. 430, Definition 2.7.1] that an α-dimensional BA with operators, a Boα, is an algebra = 〈A, + −, ci, dij〉i, j ∈ α of the same similarity type as CAα's such that , is a BA and the operations ci (i ∈ α) are additive, i.e., ⊨ ci(x + y) = cix + ciy for all i ⊨ α. If ⊨ Boα then I is called the Boolean reduct of . Note that BA = Bo0. A Boα is said to be normal if {ci 0 = 0: i ∈ α} is valid in it, and a Boα is said to be extensive if {x ≤ cix: i ∈ α} is valid in it. Boα's were introduced in [17].

The class Crsα of all cylindric-relativised set algebras is defined in Definition 1.1.1 (iii) of [8, p. 4]. We give a definition in the present paper, too—see Definition 5 below. It is shown in [13] that ICrsα is a variety.

Our main result is (i) of Theorem 1 below, but we obtain (ii)–(vi), too, as a byproduct from the proof.

In this article propositions will be identified with a certain kind of changes of belief. The intended interpretation is that a proposition is characterised by the change it would induce if added to a state of belief. Propositions will thus be defined as functions from states of belief to states of belief. A set of postulates concerning the properties and existence of propositions will be formulated. A proposition will be said to be a tautology iff it is the identity function on states of belief. The main result is that the logic determined by the set of postulates is intuitionistic propositional logic.

The basic epistemic concept is that of a belief model, which is defined as a pair 〈, 〉, where is a nonempty set and is a class of functions from to . The elements in will be called states of belief and they will be denoted K, K′,…. A discussion of the epistemological interpretation of the states of belief can be found in Gärdenfors [2]. Here, no assumptions about the structure of the elements in will be made.

The elements in will be called propositions, and A, B, C, … will be used as variables over . Functions from states of belief to states of belief can be characterised as epistemic inputs. The intended interpretation of the functions in is that they correspond to changes of belief where the new evidence is accepted as “certain” or “known” in the resulting state of belief. This means that not all functions defined on can properly be called propositions.

For any A ⊂ R, the Banach game B(A) is the following infinite game on reals: Players I and II alternately play positive real numbers a1; a2, a3, a4,… such that for n > 1, an < an−1. Player I wins iff ai exists and is in A.

This type of game was introduced by Banach in 1935 in the Scottish Book [15], Problem 43. The (rather vague) problem which Banach posed was to characterize those sets A for which I (II) has a winning strategy in B(A). (There are three parts to Problem 43. In the first, Mazur defined a game G**(A) for every set A ⊂ R and conjectured that II has a winning strategy in G**(A) iff A is meager and I has a winning strategy in G**(A) iff A is comeager in some neighborhood; this conjecture was proved by Banach. Presumably Banach had this result in mind when he asked the question about B(A), and hoped for a similar type of characterization.) Incidentally, Problem 43 of the Scottish Book appears to be the first time infinite games of any sort were studied by mathematicians.

This paper will not provide the reader with any answer to Banach's question. I know of no nontrivial way to characterize when player I (or II) wins, and I suspect there is none. This paper is concerned with a different (also rather vague) question: For which sets A is the Banach game B(A) determined? To say that B(A) is determined means, of course, that one of the players has a winning strategy for B(A).

Throughout this paper, ⊿ will denote a commutative ring with multiplicative identity, 1. The algebras we consider will be associative ⊿-algebras which are not necessarily commutative and do not necessarily contain a multiplicative identity. By standard methods, every ⊿-algebra can be embedded in an existentially closed (e.c.) Δ-algebra—and even in one which is existentially universal (e.u.). (See §0 for more details.)

We shall be studying the ideals of e.c. ⊿-algebras. Since every ideal is a sum of principal ideals, a natural place to begin is with principal ideals. In §1 we show that for an algebraically closed (a.c.) ⊿-algebra A, and elements a, b in A, whether or not b belongs to the principal ideal (a)A generated by a, depends only on the underlying ⊿-module structure of A; more precisely, for b to belong to (a)A it is necessary and sufficient that b satisfies every positive existential formula θ(ν) in the language of ⊿-modules which is satisfied by a (cf. Corollary 1.8). For special classes of rings ⊿ this condition can be simplified (Proposition 1.10): e.g. for Prüfer rings it is enough to consider formulas of the form ∃x(λx = μν); and for regular rings it is enough to consider formulas μν = 0 (where λ, μ ∈ ⊿).

In §2 we use the results of §1 to study e.c. and e.u. algebras over a principal ideal domain (p.i.d.) ⊿ (Note that for ⊿ = Z this includes the case of e.c. rings.) We obtain a necessary and sufficient condition for an a.c. ⊿-algebra to be e.c. (Theorem 2.4). We also show (Theorem 2.2) that in an a.c. ⊿-algebra A every element that is divisible by all nonzero elements of ⊿ belongs to the divisible part D(A) of A. (It should be noted that, while a.c. ⊿-modules are always divisible [ES], an e.c. ⊿-algebra is never divisible: see the end of §0. Moreover, an e.c. ⊿-algebra always contains torsion-free elements: see Remark 2.3.) We prove that every bounded ideal in an a.c. ⊿-algebra is principal (2.7).

The purpose of this paper is not to produce a survey of systems with a universal set: we do not yet understand them well enough. Rather it will concentrate on one particular aspect of them: the curious circumstance that there are half-a-dozen or so distinct proofs of ~ AC available in set theories with a universal set. This began to emerge in 1953 when Specker published in [2] a proof of ~ AC in Quine's system NF. Until recently this was an isolated phenomenon and poorly understood. The proof answered few questions and seemed rather ad hoc, thus inviting an investigation to determine whether this was an artefact caused by the particular axioms for NF, or part of a general conflict between the demands of big sets and AC.

We will start with an informal discussion of the genesis of set theories with a universal set, collecting, en route, a number of desiderata for such theories. A number of new refutations of AC in systems meeting some or all of these conditions will then be presented. The conclusion the reader is invited to draw is that any sensible set theory with a universal set will probably have trouble with AC.

Why do set theory with V e V at all? It contradicts conventional wisdom which teaches us that e is wellfounded. It must be said for this doctrine (let us abbreviate it to WOOF) that it has enabled the rapid execution of the logicist programme, so if that were the sole purpose of set theory we could consider ourselves well served. However most pure mathematicians are platonists and study things not because they are useful but simply because they are there. And the belief that V is not there can, after all, only be the result of early conditioning.

Craig's theorem is a result about the cardinality of a proper basis for the theory of combinators. Its proof given in [3] was shown to be incomplete by André Chauvin [2]. By using a different approach, we give a very short proof of this theorem. We use the notation of [1].

Definition 1. A combinator Q is proper if there exists a natural number n such that for arbitrary variables x1,…,xn we have the following contraction rule:

where C is a pure combination of the variables x1,…,xn. Q is to be understood as an abstract symbol, not as a combination of S and K's. Therefore Q comes with a contraction rule.

Definition 2. A set (Q1,…, Qm} of combinators is a basis for combinatory logic if for every finite set {x1,…, xk} of variables and every pure combination C of these variables, there exists a pure combination Q of Q1,…, Qm such that Qx1 … xk ↠ C.

Craig's Theorem. Every basis for combinatory logic containing only proper combinators contains at least two elements.

Proof. Let {Q} be a singleton basis for combinatory logic, and let us show that we cannot have combinatory completeness. This is an easy consequence of the next two lemmas.

Lemma 1. Q is a projection. That is, Qx1 … xn ↠ xj, for some j.

Proof. Let I be a proper combination of Q such that Ix ↠ x for a variable x, and let M be a term such that Ix ↠ M → x and M → x is a nontrivial contraction.

A series of investigations started by H. Friedman ([5], [6]) and pursued by him, by S. Simpson, and by others (see [7], [16] and the references there) had the objective of determining what are precisely the set existence axioms needed for proving theorems of “ordinary” mathematics. The study centered around -CA0 (for “-Comprehension Axiom”), a fragment of second order arithmetic in which the main body of “ordinary” mathematics can be comfortably developed (of course, after a suitable encoding of the various concepts into numbers and sets of numbers). The main question is always this: given a theorem τ provable in -CA0, do we need all of -CA0, or, maybe, is τ provable in a weaker subsystem? A surprisingly simple pattern emerged: there seem to be just five important systems, denoted (in order of increasing strength) RCA0, WKL0, ACA0, ATR0 and -CA0, such that in most cases, whenever an ordinary mathematical theorem τ is provable in -CA0, it is either provable in RCA0 or it is equivalent to one, call it S, of the other four systems, the proof of the equivalence being done in one of the systems which is weaker than S (most typically, RCA0 or ACA0). This very interesting phenomenon—or “theme” as it was called by Friedman—has been verified for many instances in the realm of analysis and algebra (cf. the references above and the forthcoming book [17]). It is the purpose of this paper to do the same for a particular branch of model theory, namely stability theory.

When dealing with axiomatic theories from a recursion-theoretic point of view, the notion of r.e. preordering naturally arises. We agree that an r.e. preorder is a pair = 〈P, ≤P〉 such that P is an r.e. subset of the set of natural numbers (denoted by ω), ≤P is a preordering on P and the set {〈;x, y〉: x ≤Py} is r.e.. Indeed, if is an axiomatic theory, the provable implication of yields a preordering on the class of (Gödel numbers of) formulas of .

Of course, if ≤P is a preordering on P, then it yields an equivalence relation ~P on P, by simply letting x ~Py iff x ≤Py and y ≤Px. Hence, in the case of P = ω, any preordering yields an equivalence relation on ω and consequently a numeration in the sense of [4]. It is also clear that any equivalence relation on ω (hence any numeration) can be regarded as a preordering on ω. In view of this connection, we sometimes apply to the theory of preorders some of the concepts from the theory of numerations (see also Eršov [6]).

Our main concern will be in applications of these concepts to logic, in particular as regards sufficiently strong axiomatic theories (essentially the ones in which recursive functions are representable). From this point of view it seems to be of some interest to study some remarkable prelattices and Boolean prealgebras which arise from such theories. It turns out that these structures enjoy some rather surprising lattice-theoretic and universal recursion-theoretic properties.

After making our main definitions in §1, we examine universal recursion-theoretic properties of some r.e. prelattices in §2.

In this paper we answer the question of whether all low sets with the splitting property are promptly simple. Further we try to make the role of lowness properties and prompt simplicity in the construction of automorphisms of the lattice of r.e. (recursively enumerable) sets more perspicuous. It turns out that two new properties of r.e. sets, which are dual to each other, are essential in this context: the prompt and the low shrinking property.

In an earlier paper [4] we had shown (using Soare's automorphism construction [10] and [12]) that all r.e. generic sets are automorphic in the lattice ℰ of r.e. sets under inclusion. We called a set A promptly simple if Ā is infinite and there is a recursive enumeration of A and the r.e. sets (We)e∈N such that if We is infinite then there is some element (or equivalently: infinitely many elements) x of We such that x gets into A “promptly” after its appearance in We (i.e. for some fixed total recursive function f we have x ∈ Af(s), where s is the stage at which x entered We). Prompt simplicity in combination with lowness turned out to capture those properties of r.e. generic sets that were used in the mentioned automorphism result. In a following paper with Shore and Stob [7] we studied an ℰ-definable consequence of prompt simplicity: the splitting property.

Let T be superstable. We say a type p is weakly minimal if R(p, L, ∞) = 1. Let M ⊨ T be uncountable and saturated, H = p(M). We say D ⊂ H is locally modular if for all X, Y ⊂ D with X = acl(X) ∩ D, Y = acl(Y) ∩ D and X ∩ Y ≠ ∅,

Theorem 1. Let p ∈ S(A) be weakly minimal and D the realizations of stp(a/A) for some a realizing p. Then D is locally modular or p has Morley rank 1.

Theorem 2. Let H, G be definable over some finite A, weakly minimal, locally modular and nonorthogonal. Then for all a ∈ H∖acl(A), b ∈ G∖acl(A) there area′ ∈ H, b′ ∈ G such that a′ ∈ acl(abb′A)∖acl(aA). Similarly when H and G are the realizations of complete types or strong types over A.

The probability logic is a logic with a natural interpretation on probability spaces (thus, a logic whose model theory is part of probability theory rather than a system for putting probabilities on formulas of first order logic). Its exact definition and basic development are contained in the paper [3] of H. J. Keisler and the papers [1] and [2] of the author. Building on work in [2], we prove in this paper the following probabilistic interpolation theorem for .

Let L be a countable relational language, and let A be a countable admissible set with ω ∈ A (in this paper some probabilistic notation will be used, but ω will always mean the least infinite ordinal). is the admissible fragment of corresponding to A. We will assume that L is a countable set in A, as is usual in practice, though all that is in fact needed for our proof is that L be a set in A which is wellordered in A.

Theorem. Let ϕ(x) and ψ(x) be formulas of LAP such that

where ε ∈ [0, 1) is a real in A (reals may be defined in the usual way as Dedekind cuts in the rationals). Then for any real d > ε¼, there is a formula θ(x) of (L(ϕ) ∩ L(ψ))AP such that

and

The purpose of this work is to present Gentzen-style formulations of S5 and S4 based on sequents of higher levels. Sequents of level 1 are like ordinary sequents, sequents of level 2 have collections of sequents of level 1 on the left and right of the turnstile, etc. Rules for modal constants involve sequents of level 2, whereas rules for customary logical constants of first-order logic with identity involve only sequents of level 1. A restriction on Thinning on the right of level 2, which when applied to Thinning on the right of level 1 produces intuitionistic out of classical logic (without changing anything else), produces S4 out of S5 (without changing anything else).

This characterization of modal constants with sequents of level 2 is unique in the following sense. If constants which differ only graphically are given a formally identical characterization, they can be shown inter-replaceable (not only uniformly) with the original constants salva provability. Customary characterizations of modal constants with sequents of level 1, as well as characterizations in Hilbert-style axiomatizations, are not unique in this sense. This parallels the case with implication, which is not uniquely characterized in Hilbert-style axiomatizations, but can be uniquely characterized with sequents of level 1.

These results bear upon theories of philosophical logic which attempt to characterize logical constants syntactically. They also provide an illustration of how alternative logics differ only in their structural rules, whereas their rules for logical constants are identical.

As is well known, NF is a first-order theory whose language coincides with that of ZF. The nonlogical axioms of the theory are: Extensionality. (x)(y)[(z)(z ∈ x ↔ z ∈ y) → x = y].

Comprehension. (Ex)(y)(y ∈ x ↔ ψ) for every stratified ψ in which x does not occur free (a formula of NF is said to be stratified if it can be turned into a formula of the simple theory of types by adding type indices (natural numbers ≥ 0) to its variables).

Before stating our result, a few preliminaries are in order. Let T be the simple theory of types. If ψ is a formula of T, we denote by ψ+ the formula obtained from ψ by raising all type indices by 1. T* is the result of adding to T every axiom of the form ψ ↔ ψ+. A formula of T is n-stratified (n > 0) if it does not contain any type index ≥ n. A formula of NF is n-stratified if it can be turned into an n-stratified formula of T by adding type indices to its variables. (In practice, we shall allow ourselves to confuse an n-stratified formula of T with the corresponding n-stratified formula of NF). For n > 0, Tn (resp. ) is the subtheory of T (resp. T*) containing only n-stratified formulae. For n > 0, NFn is the subtheory of NF generated by those axioms of NF which are n-stratified. Let = 〈M0, M1,…,=, ∈〉 be a model of T.

Suppose D ⊂ M is a strongly minimal set definable in M with parameters from C. We say D is locally modular if for all X, Y ⊂ D, with X = acl(X ∪ C)∩D, Y = acl(Y ∪ C) ∩ D and X ∩ Y ≠ ∅,

We prove the following theorems.

Theorem 1. Suppose M is stable and D ⊂ M is strongly minimal. If D is not locally modular then inMeqthere is a definable pseudoplane.

(For a discussion of Meq see [M, §A].) This is the main part of Theorem 1 of [Z2] and the trichotomy theorem of [Z3].

Theorem 2. Suppose M is stable and D, D′ ⊂ M are strongly minimal and nonorthogonal. Then D is locally modular if and only if D′ is locally modular.

There are well-known theorems in mathematical logic that indicate rather profound differences between the logic of first-order languages and the logic of second-order languages. In the first-order case, for example, there is Gödel's completeness theorem: every consistent set of sentences (vis-à-vis a standard axiomatization) has a model. As a corollary, first-order logic is compact: if a set of formulas is not satisfiable, then it has a finite subset which also is not satisfiable. The downward Löwenheim-Skolem theorem is that every set of satisfiable first-order sentences has a model whose cardinality is at most countable (or the cardinality of the set of sentences, whichever is greater), and the upward Löwenheim-Skolem theorem is that if a set of first-order sentences has, for each natural number n, a model whose cardinality is at least n, then it has, for each infinite cardinal κ (greater than or equal to the cardinality of the set of sentences), a model of cardinality κ. It follows, of course, that no set of first-order sentences that has an infinite model can be categorical. Second-order logic, on the other hand, is inherently incomplete in the sense that no recursive, sound axiomatization of it is complete. It is not compact, and there are many well-known categorical sets of second-order sentences (with infinite models). Thus, there are no straightforward analogues to the Löwenheim-Skolem theorems for second-order languages and logic.

There has been some controversy in recent years as to whether “second-order logic” should be considered a part of logic, but this boundary issue does not concern me directly, at least not here.