David DeVidi

There are several open problems in the study of the calculi which result from adding either of Hilbert's ϵ‐ or τ‐operators to the first order intuitionistic predicate calculus. This paper provides answers to several of them. In particular, the first complete and sound semantics for these calculi are presented, in both a “quasi‐extensional” version which uses choice functions in a straightforward way to interpret the ϵ‐ or τ‐terms, and in a form which does not require extensionality assumptions. Unlike the classical case, the addition of either operator to intuitionistic logic is non‐conservative. Several interesting consequences of the addition of each operator are proved. Finally, the independence of several other schemes in either calculus are also proved, making use of the semantics supplied earlier in the paper.

_Logical Options_ introduces the extensions and alternatives to classical logic which are most discussed in the philosophical literature: many-sorted logic, second-order logic, modal logics, intuitionistic logic, three-valued logic, fuzzy logic, and free logic. Each logic is introduced with a brief description of some aspect of its philosophical significance, and wherever possible semantic and proof methods are employed to facilitate comparison of the various systems. The book is designed to be useful for philosophy students and professional philosophers who have learned some classical first-order logic and would like to learn about other logics important to their philosophical work.

That choice principles ‘change the logic’ when added to constructive systems is significant for contemporary metaphysics. However, many are surprised by these results, having learned that the Axiom of Choice (AC) is constructively valid. Indeed, even among specialists there were, until recently, reasons for puzzlement-rival versions of Intuitionistic Type Theory, one where (AC) is valid, another where it implies classical logic. This paper accessibly explains the situation, puts the issues in a broader setting by considering other choice principles, and draws philosophical morals for the understanding of quantification, choice principles, and the prospects for constructivism.

Graham Solomon, to whom this collection is dedicated, went into hospital for antibiotic treatment of pneumonia in Oc- ber, 2001. Three days later, on Nov. 1, he died of a massive stroke, at the age of 44. Solomon was well liked by those who got the chance to know him—it was a revelation to?nd out, when helping to sort out his a?airs after his death, how many “friends” he had whom he had actually never met, as his email included correspondence with philosophers around the world running sometimes to hundreds of messages. He was well respected in the philosophical community more broadly. He was for several years a member of the editorial board for the Western Ontario Series in Philosophy of Science. While he was employed at Wilfrid Laurier University in Waterloo, Ontario, several of us at the University of Wat- loo always regarded our own department as a sort of second academic home for him. We therefore decided that it would be appropriate to hold a memorial conference in his honour. Thanks to the generous?nancial support of the Humphrey Conference Fund, we were able to do so in May 2003. Many of the papers in this volume were presented at that conf- ence.

The volume includes twenty-five research papers presented as gifts to John L. Bell to celebrate his 60th birthday by colleagues, former students, friends and admirers. Like Bell’s own work, the contributions cross boundaries into several inter-related fields. The contributions are new work by highly respected figures, several of whom are among the key figures in their fields. Some examples: in foundations of maths and logic ; analytical philosophy, philosophy of science, philosophy of mathematics and decision theory and foundations of economics. Most articles are contributions to current philosophical debates, but contributions also include some new mathematical results, important historical surveys, and a translation by Wilfrid Hodges of a key work of arabic logic.

We discuss the philosophical implications of formal results showing the con- sequences of adding the epsilon operator to intuitionistic predicate logic. These results are related to Diaconescu’s theorem, a result originating in topos theory that, translated to constructive set theory, says that the axiom of choice (an “existence principle”) implies the law of excluded middle (which purports to be a logical principle). As a logical choice principle, epsilon allows us to translate that result to a logical setting, where one can get an analogue of Diaconescu’s result, but also can disentangle the roles of certain other assumptions that are hidden in mathematical presentations. It is our view that these results have not received the attention they deserve: logicians are unlikely to read a discussion because the results considered are “already well known,” while the results are simultaneously unknown to philosophers who do not specialize in what most philosophers will regard as esoteric logics. This is a problem, since these results have important implications for and promise signif i cant illumination of contem- porary debates in metaphysics. The point of this paper is to make the nature of the results clear in a way accessible to philosophers who do not specialize in logic, and in a way that makes clear their implications for contemporary philo- sophical discussions. To make the latter point, we will focus on Dummettian discussions of realism and anti-realism. Keywords: epsilon, axiom of choice, metaphysics, intuitionistic logic, Dummett, realism, antirealism.

The two main accomplishments of this thesis are that it provides the first adequate semantics for Hilbert's epsilon-operator and that it describes a general semantics for term forming operators more flexible than any in the literature. ;The epsilon-operator was introduced by David Hilbert in the 1920s as a term forming operator in first order logic. The semantics so far available for epsilon has been designed for classical two-valued logic, and has required that additional extensionality assumptions be made. This thesis provides complete semantics for epsilon in classical extensional, classical non-extensional, Boolean valued, and intuitionistic first order systems. The natural step to generalizing the technique used in the epsilon case to get a general theory of term forming operators which handles the non-extensional and non-classical cases is then taken. ;The thesis proceeds as follows. Chapter One gives a historical discussion of term forming operators. A brief, self-contained presentation of the untyped lambda-calculus, which illustrates the inevitable differences between lambda and any possible operator in first order logic, follows. A chapter is devoted to solving the syntactical difficulties involved in introducing a variable binding term forming operator to standard languages for first order logic. The semantics for epsilon, and in the intuitionistic case also for another of Hilbert's creatures, tau, takes up the next several chapters. The discussion includes several new completeness and soundness results, and some new results about the extra strength these operators add to intuitionistic logic, including some new independence results. The final chapter includes an argument to the effect that the results earlier in the thesis show that we need a more general theory of term forming operators than any in the literature, and indicates the shape such a theory should take.

RésuméCet article discute diverses confusions, actueles ou potentielles, liées á la bivalence et au tiers exclu. Il s'agit, en particulier, 1) d'examiner divers cas illustrant les rapports entre la bivalence et le tiers exclu ; 2) de discuter la thése selon laquelle le tiers exclu et le schéma-T de Tarskipour la vérité entraînent la bivalence; 3) de proposer quelques remarques sur les rapports entre la bivalence, le tiers exclu et la preuve par l'absurde; 4) de scruter un argument répandu selon lequel le tiers exclu vaut pour les termes «précis» et 5) de commenter l'histoire de la terminologie. L'objectif ici est d'attirer l'attention des philosophes sur une distinction importante qui ne reçoit pas souvent le traitement minutieux qu'elle mérite.

There are several open problems in the study of the calculi which result from adding either of Hilbert's ϵ- or τ-operators to the first order intuitionistic predicate calculus. This paper provides answers to several of them. In particular, the first complete and sound semantics for these calculi are presented, in both a “quasi-extensional” version which uses choice functions in a straightforward way to interpret the ϵ- or τ-terms, and in a form which does not require extensionality assumptions. Unlike the classical case, the addition of either operator to intuitionistic logic is non-conservative. Several interesting consequences of the addition of each operator are proved. Finally, the independence of several other schemes in either calculus are also proved, making use of the semantics supplied earlier in the paper.

Logical Options introduces the extensions and alternatives to classical logic which are most discussed in the philosophical literature: many-sorted logic, second-order logic, modal logics, intuitionistic logic, three-valued logic, fuzzy logic, and free logic. Each logic is introduced with a brief description of some aspect of its philosophical significance, and wherever possible semantic and proof methods are employed to facilitate comparison of the various systems. The book is designed to be useful for philosophy students and professional philosophers who have learned some classical first-order logic and would like to learn about other logics important to their philosophical work.

Michael Friedman has recently argued that Carnap'sLogical Syntax of Language is fundamentally flawed in a way that reveals the ultimate failure of logical positivism. Friedman's argument depends crucially on two claims: (1) that Carnap was committed to the view that there is a universal metalanguage and (2) that given what Carnap wanted from a metalanguage, in particular given that he wanted a definition of analytic for an object language, he was in fact committed to a hierarchy of stronger and stronger metalanguages. We argue that neither of these claims need be accepted. We show that there is no textual evidence for (1) and that if metalanguages are to be used for merely descriptive and not also justificatory purposes, Carnap does not need to define analyticity sufficiently for proving consistency, and so could have given a definition that does not entail a hierarchy of metalanguages.

It is well known that Tarski proved a result which can be stated roughly as: no sufficiently rich, consistent, classical language can contain its own truth definition. Tarski's way around this problem is to deal with two languages at a time, an object language for which we are defining truth and a metalanguage in which the definition occurs. An obvious question then is: under what conditions can we construct a definition of truth for a given object language. Tarski claims that it is necessary and sufficient that the metalanguage be "essentially richer". Our contention, put bluntly, is that this claim deserves more scrutiny from philosophers than it usually gets and in fact is false unless "essentially richer" means nothing else than "sufficient to contain a truth definition for the object language."