The generic structures introduced by Abraham Robinson are now well established in model theory. Anyone who works with these structures soon begins to feel that, in some sense, the infinite generic structures are large and the finite generic structures are small. Here ‘large’ and ‘small’ do not refer to the cardinality of the structures but are used in the way we describe saturated structures as large and atomic structures as small. In this paper I isolate what I consider to be the large and small existentially closed (e.c.) structures and attempt to determine their role in the class of all e.c. structures.

In the usual context of model theory we are concerned with elementary embeddings and all formulas. E.c. structures are concerned with all embeddings and ∃1-formulas. Thus we need to look at ∃1-analogues of saturated and atomic structures. These ∃1-saturated structures have been around for some time; they are just the existentially universal structures. The corresponding ∃1-atomic structures are not new here (they appear in [8]) but I believe that this paper will add much to the understanding of them.

The bulk of this paper is in §§2 and 4. §2 deals with ∃1-atomic structures, and §4 is concerned with the ∃1-analogues of several results in Vaught's paper [12].

A similar program has been carried out by Pouzet in [6], [7], [8]; however, there the significance of e.c. and e.c. structures is not realized and consequently some of the simplicity of the situation is lost.

Let R be a commutative ring with 1 and R[X 1, …, Xn ] the polynomial ring in n variables over R. Then for any relation f(X) = 0 in R[X] there exists a conjunction of equations φ f such that f(X) = 0 holds in R[X] iff φ f holds in R; φ f is of course the formula saying that all the coefficients of f(X) vanish. Moreover, φ f is independent of R and formed uniformly for all polynomials f up to a given formal degree. In this paper we investigate first order theories T for which a similar phenomenon holds. More precisely, we let TAH be the universal Horn part of a theory T and look at free extensions of models of T in the class of models of TAH . We ask whether an atomic relation t 1(X, a) = t 2(X, a) or R(t 1(X, a), …, tn (X, a)) in can be equivalently expressed by a finite or infinitary formula φ(a) in , such that φ(y) depends only on t i{X, y) and not on or a 1, …, am ∈ A.

We will show that for a wide class of theories T “defining formulas” φ(y) in this sense exist and can be taken as infinite disjunctions of positive existential formulas.

With quantifier elimination and restriction of language to a binary relation symbol and constant symbols it is shown that countable complete theories having three isomorphism types of countable models are “essentially” the Ehrenfeucht example [4, §6].

The main purpose of this paper is to show how partial recursive functions and isols can be used to generalize the following three well-known theorems of combinatorial theory.

(I) For every finite projective plane Π there is a unique number n such that Π has exactly n 2 + n + 1 points and exactly n 2 + n + 1 lines.

(II) Every finite projective plane of order n can be coordinatized by a finite planar ternary ring of order n. Conversely, every finite planar ternary ring of order n coordinatizes a finite projective plane of order n.

(III) There exists a finite projective plane of order n if and only if there exist n − 1 mutually orthogonal Latin squares of order n.

Introduction. The purpose of this paper is to show how results from the theory of inductive definitions can be used to obtain new compactness theorems for uncountable admissible languages. These will include improvements of the compactness theorem by J. Green [9].

In [2] Barwise studies admissible sets satisfying the Σ1-compactness theorem. Our approach is to consider admissible sets satisfying what could be called the abstract extended completeness theorem, that is, sets where the consequence relation of the admissible fragment LA is Σ1-definable over A. We will call such sets Σ1-complete. For countable admissible sets, Σ1-completeness follows from the completeness theorem for LA .

Having restricted our attention to Σ1-complete sets we are led to a stronger notion also true on countable admissible sets, namely what we shall call uniform Σ1-completeness. We will see that this notion can be viewed as extending to uncountable admissible sets, properties related to both the “recursion theory” and “proof theory” of countable admissible sets.

By following Barwise's recent approach to admissible sets allowing “urelements,” we show that there is a natural connection between certain structures arising from the theory of inductive definability, and uniformly Σ1-complete admissible sets . The structures we have in mind are called uniform Kleene structures.

One general program of α-recursion theory is to determine as much as possible of the lattice structure of (α), the lattice of α-r.e. sets under inclusion. It is hoped that structure results will shed some light on whether or not the theory of (α) is decidable with respect to a suitable language for lattice theory. Fix such a language ℒ.

Many of the basic results about the lattice structure involve various sorts of simple α-r.e. sets (we use definitions which are definable in ℒ over (α)). It is easy to see that simple sets exist for all admissible α. Chong and Lerman [1] have found some necessary and some sufficient conditions for the existence of hhsimple α-r.e. sets, although a complete determination of these conditions has not yet been made. Lerman and Simpson [9] have obtained some partial results concerning r-maximal α-r.e. sets. Lerman [6] has shown that maximal α-r.e. sets exist iff a is a certain sort of constructibly countable ordinal. Lerman [5] has also investigated the congruence relations, filters, and ideals of (α). Here various sorts of simple sets have also proved to be vital tools. The importance of simple α-r.e. sets to the study of the lattice structure of (α) is hence obvious.

Lerman [6, Q22] has posed the following problem: Find an admissible α for which all simple α-r.e. sets have the same 1-type with respect to the language ℒ. The structure of (α) for such an α would be much less complicated than that of (ω). Lerman [7] showed that such an α could not be a regular cardinal of L. We show that there is no such admissible α.

In [3], Martin computed the degrees of certain classes of RE sets. To state the results succinctly, some notation is useful.

If A is a set (of natural numbers), dg(A) is the (Turing) degree of A. If A is a class of sets, dg ( A ) = {dg(A): A ∈ A ). Let M be the class of maximal sets, HHS the class of hyperhypersimple sets, and DS the class of dense simple sets. Martin showed that dg ( M ), dg ( HHS ), and dg ( DS ) are all equal to the set H of RE degrees a such that a ′ = 0″.

Let M * be the class of coinfinite RE sets having no superset in M ; and define HHS * and DS * similarly. Martin showed that dg ( DS *) = H. In [2], Lachlan showed (among other things) that dg ( M *)⊆K, where K is the set of RE degrees a such that a ″ > 0″. We will show that K ⊆ dg ( HHS *). Since maximal sets are hyperhypersimple, this gives dg ( M *) = dg ( HHS *) = K.

These results suggest a problem. In each case in which dg(A) has been calculated, the set of nonzero degrees in dg(A) is either H or K or the empty set or the set of all nonzero RE degrees. Is this always the case for natural classes A? Natural here might mean that A is invariant under all automorphisms of the lattice of RE sets; or that A is definable in the first-order theory of that lattice; or anything else which seems reasonable.

Throughout this paper, α will denote an admissible ordinal. Let (α) denote the lattice of α-r.e. sets, i.e., the lattice whose elements are the α-r.e. sets, and whose ordering is given by set inclusion. Call a set A ∈ (α)α*-finite if it is α-finite and has ordertype < α* (the Σ1-projectum of α). The α*-finite sets form an ideal of (α), and factoring (α) by this ideal, we obtain the quotient lattice *(α).

We will fix a language ℒ suitable for lattice theory, and discuss decidability in terms of this language. Two approaches have succeeded in making some progress towards determining the decidability of the elementary theory of (α). Each approach was first used by Lachlan for α = ω. The first approach is to relate the decidability of the elementary theory of (α) to that of a suitable quotient lattice of (α) by a congruence relation definable in ℒ. This technique was used by Lachlan [4, §1] to obtain the equidecidability of the elementary theories of (ω) and *(ω), and was generalized by us [6, Corollary 1.2] to yield the equidecidability of the elementary theories of (α) and *(α) for all α. Lachlan [3] then adopted a different approach.

This paper is devoted to the study of the infinitistic rules of proof i.e. those which admit an infinite number of premises. The best known of these rules is the ω-rule. Some properties of the ω-rule and its connection with the ω-models on the basis of the ω-completeness theorem gave impulse to the development of the theory of models for admissible fragments of the language . On the other hand the study of representability in second order arithmetic with the ω-rule added revealed for the first time an analogy between the notions of re-cursivity and hyperarithmeticity which had an important influence on the further development of generalized recursion theory.

The consideration of the subject of infinitistic rules in complete generality seems to be reasonable for several reasons. It is not completely clear which properties of the ω-rule were essential for the development of the above-mentioned topics. It is also worthwhile to examine the proof power of infinitistic rules of proof and what distinguishes them from finitistic rules of proof.

What seemed to us the appropriate point of view on this problem was the examination of the connection between the semantics and the syntax of the first order language equipped with an additional rule of proof.

We prove the following extension of a result of Keisler and Morley. Suppose is a countable model of ZFC and c is an uncountable regular cardinal in . Then there exists an elementary extension of which fixes all ordinals below c, enlarges c, and either (i) contains or (ii) does not contain a least new ordinal.

Related results are discussed.

Let α be an admissible ordinal, and let (α) denote the lattice of α-r.e. sets, ordered by set inclusion. An α-r.e. set A is α*-finite if it is α-finite and has ordertype less than α* (the Σ1 projectum of α). An a-r.e. set S is simple if (the complement of S) is not α*-finite, but all the α-r.e. subsets of are α*-finite. Fixing a first-order language ℒ suitable for lattice theory (see [2, §1] for such a language), and noting that the α*-finite sets are exactly those elements of (α), all of whose α-r.e. subsets have complements in (α) (see [4, p. 356]), we see that the class of simple α-r.e. sets is definable in ℒ over (α). In [4, §6, (Q22)], we asked whether an admissible ordinal α exists for which all simple α-r.e. sets have the same 1-type. We were particularly interested in this question for α = ℵ1 L (L is Gödel's universe of constructible sets). We will show that for all α which are regular cardinals of L (ℵ1 L is, of course, such an α), there are simple α-r.e. sets with different 1-types.

The sentence exhibited which differentiates between simple α-r.e. sets is not the first one which comes to mind. Using α = ω for intuition, one would expect any of the sentences “S is a maximal α-r.e. set”, “S is an r-maximal α-r.e. set”, or “S is a hyperhypersimple α-r.e. set” to differentiate between simple α-r.e. sets. However, if α > ω is a regular cardinal of L, there are no maximal, r-maximal, or hyperhypersimple α-r.e. sets (see [4, Theorem 4.11], [5, Theorem 5.1] and [1,Theorem 5.21] respectively). But another theorem of (ω) points the way.

A number is a nonnegative integer, and E is the set of numbers. In [3], J. C. E. Dekker introduced the concept of a regressive set of order n as the range of a one-one function f of n arguments such that (i) domain f ⊆ En and, if (x 1, …, xn ) ∈ domain f, and yi ≤ ≤ xi for 1 ≤ i ≤ n, then (y 1, …, yn ) ∈ domain f, and (ii) if 1 ≤ i ≤ n and (x 1, …, xn ) ∈ domain f, then f(x 1 … x i−1, xi ∸ 1, x i+1 … xn ) can be found effectively from f(x 1 … xn ). (0 ∸ 1 = 0 and, for m ≥ 1, m ∸ 1 = m − 1.) Since one can take the view, as Dekker did when first introducing regressive functions in [1], that a regressive set of order one is the range of a function of the above type which is of order one and everywhere defined, it seems natural to study the n-dimensional analogue in which (i) is replaced by “domain f = En .” It is the purpose of this paper to study such sets.

In [4] Kleene gave a definition of recursive functionals of finite type. Later Sacks [5] and Harrington [2] gave definitions of recursion in normal functionals of finite type. These definitions, that Sacks and Harrington assumed equivalent as far as normal objects are concerned, are nevertheless very different: Kleene's definition is given in terms of an inductive definition; Sacks uses simultaneously a hierarchy (the S σ F's) and induction on the ordinals and on the type; Harrington's universe does not use the induction on the type but uses a hierarchy as Shoenfield [6]. In this paper we prove in detail that, as was expected, the three definitions are equivalent.