Mathematical Logic

PREFACE TO THIS VOLUME

Mike YATES,     July 1999

Although the unpublished papers were left to Robin Gandy in Turing’s will, the matter of inquiring into the possibility of their publication seems to have rested in the first instance with Max Newman¹. Newman ran into problems with the assessment of some of the unpublished papers, and then on retiring in 1963 he left the whole matter in Robin’s hands. It was in fact far too big a task for one person but that was not truly realised until 1988 when Professor Furbank, Turing’s executor and overall editor of the Collected Works, stepped in to take a new initiative. There followed an intensive period of activity in which he brought in three additional editors. They did a remarkable job and their volumes were published in 1992.

So why has this, the fourth volume, taken so much longer? As often there is no single reason, but some explanation is due. It was planned at the same time as the other three, becoming the only part of the Collected Works to be edited by Robin. In 1991I was invited to help him and we made progress, though as expected it was slow. I should at this point make two things clear. First, the scope of the volume envisaged at that point covered only what is now in Parts I and II. Secondly, Robin had in fact written very careful first drafts of prefaces for all of the papers except the ‘Ordinal Logics’ paper as far back as 1978! He wished to rewrite them and that is what took time. He particularly wanted to rewrite the preface to the famous 1937 paper and to tackle the ‘Ordinal Logics’ paper. Since Robin had no draft preface for the latter, I did at that time initiate correspondence with Professor Solomon Feferman with a view to possibly using extracts from his excellent paper ‘Turing in the Land of O(z)’ for this purpose.

Then very sadly, Robin died unexpectedly in November 1995, and so I had to take a fresh look at the volume. I have been especially grateful to Professor Feferman for his prefaces, and for overall advice. Turing’s biographer, Dr. Andrew Hodges, has also been particularly helpful. He suggested that this was an opportunity to take some new directions and tidy up some loose ends – hence Part III – and he subsequently provided two prefaces for Part III, including that for the elusive Enigma paper written probably in 1940 but not released until 1996.

I am grateful to Professor Jack Good, who worked with Turing at Bletchley Park, and Dr. Martin Campbell-Kelly for the other two prefaces in Part III.

Finally, I must thank Professor Nick Furbank and Dr. Arjen Sevenster for their help and forbearance in what has been a very protracted task.

¹Professor Maxwell Hermann Alexander Newman, FRS. Newman was involved at certain crucial times in Turing’s life. He introduced him to logic in 1935, was also at Bletchley Park and finally brought him to Manchester in 1948. He wrote Turing’s Biographical Memoir for the Royal Society, which can be found in this volume.

Part I

Computability and Ordinal Logics

Introduction to: Computability and Ordinal Logics

The historical introduction which opens Part I has been adapted from Feferman’s excellent paper [1988]: ‘Turing in the Land of O(z)’. The editor wishes to reiterate his gratitude to both Professor Feferman and Oxford University Press, the original publisher of the volume Herken [1988] which includes that paper.

Historical Introduction

Solomon Feferman

The story of how Turing came to write his papers on computable numbers and ordinal logics is contained in Andrew Hodges’ excellent biography, Alan Turing, The Enigma [1983]. It is retold in the following in condensed form, drawing extensively on Hodges for the relevant biographical details, as well as Kleene [1981] and Feferman [1986] for the development of logic and recursion theory in this period.

Paper 1. On Computable Numbers with an Application to the Entscheidungsproblem

The story begins with Turing’s major achievement, his work on computability, carried out in 1935-6 soon after he became a fellow of King’s College Cambridge at the age of 23. In the Spring of 1935 Turing attended a course on the Foundations of Mathematics given by the topologist M.H.A. Newman. Among other things, Newman explained Hubert’s problems concerning consistency, completeness and decidability of various axiomatic systems, as well as Gödel’s incompleteness results for sufficiently strong such systems. Turing had already been interested in mathematical logic but had been working primarily on other areas of mathematics, especially group theory. Newman’s course served to focus his interests in logic; in particular, Turing became intrigued by the Entscheidungsproblem (decision problem) for the first order predicate (or functional) calculus, and this came to dominate his thought from the summer of 1935 on (Hodges [1983], p. 94). In grappling with this problem he was led to conclude that the solution must be negative; but in order to demonstrate that, he would have to give an exact mathematical analysis of the informal concept of computability by a strictly mechanical process. This Turing achieved by mid-April 1936, when he delivered a draft of Paper 1 to Newman. At first Newman was skeptical of Turing’s analysis, thinking that nothing so straightforward in its basic conception as the Turing machines could be used to answer this outstanding problem. However, he finally satisfied himself that Turing’s notion did indeed provide the most general explanation of finite mechanical process, and he encouraged the paper’s publication.

Neither Newman nor Turing were then aware that the question of analyzing the notion of effective calculability had occupied the attention of Gödel, Herbrand, and especially Church since the early 1930’s. This side of the story is well-told in Kleene [1981], on the origins of recursive function theory.¹ Kleene was a Ph.D. student of Church from 1931 to 1933 (along with Rosser). Church was promoting a universal system for logic and mathematics in the framework of the lambda (λ)-symbolism for defining functions, and set Kleene the problem of developing the theory of positive integers in his formalism, using an identification of the integers with certain λ -terms. The initial steps were rather difficult (even the predecessor function posed a problem), but once the first hurdles were cleared, Kleene was able to show more and more number-theoretic functions definable by the conversion processes of λ -terms. But Church’s original system was shown before long (by Kleene and Rosser in 1934) to be inconsistent, and attention was then narrowed to a demonstrably consistent subsystem, which came to be called the λ -calculus.² The consistency of this subsystem was established by Church and Rosser, with some input by Kleene. As it happened, Kleene had already achieved all of his previous definability results in this restricted calculus. It was clear from the Church-Rosser consistency property that every λ -definable function (in the sense given by convertibility in the λ -calculus) is effectively calculable; moreover, Kleene was able to show each example of an effectively calculable function, which came to mind, to be λ -definable. In view of Kleene’s success in meeting all such challenges, Church was led to make the proposal, which has come to be known as Church’s Thesis, that the λ -definable functions comprise all the effectively definable functions. In the words of Kleene: When Church proposed the thesis, I sat down to disprove it by diagonalising out of the class of the λ -definable functions. But quickly realizing that the diagonalisation cannot be done effectively, I became overnight a supporter of the thesis. (Kleene [1981], p. 59).

Gödel (coming from the University of Vienna) visited the Institute for Advanced Study in Princeton during the year 1933-34, and in the spring of 1934 he gave lectures on his incompleteness results. Notes for these lectures were taken by Kleene and Rosser and, after corrections by Gödel, were circulated at the time. They were subsequently reproduced in Davis’ collection [1965] of basic papers on the undecidable and computable functions, and more recently in Volume I of Gödel’s Collected Works. During the course of these lectures, Gödel presented a definition, based on a suggestion of Herbrand, of general recursiveness as an analysis of the most general concept of computability, using systems of equations. However, at the time Gödel regarded this identification only as a heuristic principle.³ During the same period, Church had conversations with Gödel in which he advanced his own proposal. Gödel regarded this as thoroughly unsatisfactory but Church replied that if [Gödel] would propose any definition of effective calculability which seemed even partially satisfactory [Church] would undertake to prove that it was included in lambda-definability" (Kleene [1981], p. 59). It was only as a result of Turing’s later work on mechanical computability that Gödel would accept this identification, albeit indirectly through a chain of equivalences.

The first step in this chain was the demonstration by Church and Kleene that λ -definability for functions of positive integers is equivalent to general recursiveness. Thus when Church finally announced his Thesis in published form [1936], he formulated the Thesis in terms of general recursiveness but bolstered it by the fact of this equivalence (see Davis [1982], pp. 10-11). Church applied his Thesis in [1936] to demonstrate the effective unsolvability of various mathematical and logical problems, including the decision problem for sufficiently strong formal systems. A year later (15 April 1936, just when Turing was presenting the draft of his paper to Newman), Church submitted his [1936a] with its simple proof of the unsolvability of the decision problem for the first-order functional calculus of logic.

The dismaying news of Church’s work reached Cambridge in May 1936. At first this seemed to pre-empt Turing’s analysis of computability and his result on the Entscheidungsproblem; but Turing’s definition of computability was sufficiently different from that of Church as to warrant separate publication. Thus Turing’s paper was submitted after all, on 28 May 1936; in August 1936 he tacked on an appendix sketching a proof of the equivalence between his notion of computability and λ -definability, thus establishing the second link in the chain of equivalences.

Independently of Turing, but with knowledge of Church’s Thesis, Post also proposed in 1936 a definition of mechanical computability, which was very close to that of Turing. From the conceptual point of view these are now generally regarded as providing the most convincing analyses of the informal notion of effective or mechanical computability. Since 1936 there have been further defined notions of effectiveness, all shown to be equivalent, but there is hardly any other which carries the same immediate conviction that the sought-for general definition is at hand. Indeed, Gödel later championed Turing’s definition in his postscripts to his paper on incompleteness and the 1934 lectures (on the occasion of their reprinting). For example, in the latter he writes that… due to A.M. Turing’s work, a precise and unquestionably adequate definition of the general concept of formal system can now be given. … Turing’s work gives an analysis of the concept of ‘mechanical procedure’ (alias ‘algorithm’ or ‘computation procedure’ or ‘finite combinatorial procedure’). This concept is shown to be equivalent with that of a ‘Turing machine’.⁴ On the other hand Gödel never remarked in print concerning his negative views about Church’s Thesis as initially formulated.

Church himself agreed early on that Turing’s definition was conceptually superior. In his 1937 review⁵ of Turing’s paper, after referring to the equivalence proof in the appendix, Church says: As a matter of fact, there is involved here the equivalence of three different notions: computability by a Turing machine, general recursiveness in the sense of Herbrand–Gödel-Kleene, and λ -definability in the sense of Kleene and the present reviewer. Of these, the first has the advantage of making the identification with effectiveness in the ordinary (not explicitly defined) sense evident immediately – i.e. without the necessity of proving preliminary theorems. The second and third have the advantage of suitability for embodiment in a system of symbolic logic.

For the actual development of the (abstract) theory of computation, where one must build up a stock of particular functions and establish various closure conditions, both Church’s and Turing’s definition are equally awkward and unwieldy. In this respect, general recursiveness is superior, and Kleene’s particular schemata for that (using primitive recursion and the minimum operator μ) are the most tractable. In Kleene’s words: I myself, perhaps unduly influenced by rather chilly receptions from audiences around 1933–35 to disquisitions on λ -definability, chose, after general recursiveness had appeared, to put my work in that format. … I cannot complain about my audiences after 1935…. Still, he defends λ -definability as having… the remarkable feature that it is all contained in a very simple and almost inevitable formulation, arising in a natural connection with no prethought of the result (Kleene [1981], p. 62).

Paper 2. Computability and λ -definability and Paper 3. Systems of logic based on ordinals

Back to Turing in Cambridge, May 1936: he had been understandably let down by Church’s scoop, but gratified that he had independently arrived at the same result concerning the Entscheidungsproblem, and that his own work on computability was recognised (if by no one else but Newman and Church) to have independent merit. What next? Newman recommended that he go study with Church in Princeton, and wrote to Church to see if a visit and any sort of grant could be arranged. Turing also applied for a Procter fellowship, offered by Princeton. There were three of these altogether, one designated for Oxford, one for Cambridge, and one for the Collége de France. Turing failed to win the one for Cambridge, and no other money was forthcoming from Princeton. Still, he thought he could just manage on his fellowship funds from Kings College ($300 p.a.) and decided to spend the year 1936-7 in Princeton; he arrived there at the end of September 1936 (Hodges [1983], pp. 112-117).

Turing was well aware of the growing importance of the Princeton Mathematics Department as a world center for mathematics. It had already been a leader on the American scene when the department was greatly enriched, in the early 1930’s, by the establishment of the Institute for Advanced Study (IAS). The two shared the same facilities (Fine Hall) until 1940, so that the lines between them were blurred and there was a great deal of interaction. The IAS had been built up by Oswald Veblen, with Einstein as one of its first members. It benefited particularly as a result of the exodus of mathematicians and physicists from Germany (and especially from the great Mathematical Institute in Göttingen) following the advent of Nazism. Among the mathematical luminaries that Turing found on his arrival in Princeton were Einstein, von Neumann, Weyl, Courant, Hardy and Lefschetz. (Of these, the first three were members of the IAS, Courant and Hardy were visitors, and Lefschetz was in the Princeton department.) Turing wrote home saying that the mathematics department here comes fully up to expectations. But in logic, he had hoped to find – besides Church - Gödel, Bernays, Kleene and Rosser who were here last year (but) have left (ibid., p. 117). Turing was particularly disappointed to have missed Gödel. As already mentioned, Gödel had visited the IAS in 1933-34; he commenced a second visit in the Fall of 1935 but left after a brief period due to illness. He was to visit the IAS once more, in 1939, before becoming a member in 1940; apparently Gödel and Turing never did meet. Bernays, who had visited in 1935-36, was noted as Hubert’s collaborator on Beweistheorie, the proposed means for carrying out Hilbert’s foundational program, while Kleene and Rosser (who were, previously, Ph.D. students of Church) had left to take up positions elsewhere. Others whom Turing might have looked to were von Neumann and Weyl; both had taken an active interest in logic in earlier years (of which, see below), but were totally engaged in mainstream mathematics at the time of Turing’s arrival on the scene.

Thus Turing was left to draw on Church for direction in the further pursuit of his studies. He dutifully attended Church’s lectures which were rather on the ponderous and laborious side and took notes on Church’s theory of types. Turing also met with him in person from time to time, but Church was… a retiring man himself, not given to a great deal of discussion. Turing was himself on the shy side and, indeed, he was described by Max Newman as a confirmed solitary. The brief characterisations of Church’s style and personality, from Hodges (p. 119), are fair enough, as far as they go. But it should be added that Church was (and is) noted for the great care and precision of his writing and lecturing, and these virtues probably benefited Turing – whose own writing was rough-and-ready and prone to minor