Rules in relevant logic. I: Semantic classification (Q1319263)

Relevant logics have been motivated by the idea that a theory of logic is a theory of (correct) inference [see \textit{A. R. Anderson} and \textit{N. D. Belnap jun.}: Entailment. the logic of relevance and necessity. Vol. I (1975; Zbl 0323.02030), \textit{R. Routley} and \textit{R. K. Meyer}, Truth Syntax Modality, Stud. Logic Found. Math. 68, 199-243 (1973; Zbl 0317.02017), \textit{R. Routley, V. Plumwood, R. K. Meyer} and \textit{R. T. Brady}: Relevant logics and their rivals. Vol. I (1982; Zbl 0579.03011), for instance]. Though I know of no one who has put it this way, it would seem on this view that logical truth is a mere by-product of rules of inference. One would accordingly expect that research in relevant logic would emphasize the importance of rules of inference. This has been true to some extent [see Anderson and Belnap, loc. cit., \textit{N. D. Belnap jun.}: ``The formalisation of entailment'' (Yale Univ. Doctoral Diss., New Haven 1959), the reviewer, Notre Dame J. Formal Logic 33, 442-449 (1992; Zbl 0786.03015), for instance], but this work being reviewed and its sequel (reviewed below), are much more what one might have expected. Syntactically, it is traditional to distinguish between derived and admissible rules. This distinction is apparently what also underlies the distinction between logics that are smooth or otherwise in \textit{M. Dummett's} book: Elements of intuitionism (1977; Zbl 0358.02032). And it has been natural to think of this distinction as marking the difference (in an appropriate semantics) between rules that preserve truth and those that merely preserve validity. In this paper, the author uses the Routley-Meyer style semantics of Routley, Plumwood, Meyer and Brady [loc. cit.] to set up five distinct semantic categories for rules of inference, according to what semantic properties they preserve. They are: 1) Preservation of truth in each set up; 2) Preservation of truth in each regular set up; 3) Preservation of validity in each model structure; 4) Preservation of validity in each general model structure (frame); 5) Preservation of validity in the semantics. He then considers a number of syntactically distinguishable classes of rules in Hilbert-style axiomatisations of a number of relevant logics, and proves that they accordingly fall into one or other of the above categories. The logics in question are extensions of the basic affixing system B of Routley, Plumwood, Meyer and Brady [loc. cit.] or extensions of \(\text{B}^{\text{d}}\), which itself comes by adding the following meta-rule to B: If \(A \Rightarrow B\), then \(C \vee A \Rightarrow C \vee B\). Next, the systems of natural deduction for relevant logics presented by \textit{R. T. Brady} in Logique Anal. Nouv. Sér. 27, 355-377 (1984; Zbl 0559.03011) are considered, and classes of rules are syntactically distinguished and shown accordingly to fall into category 1) or 4) of the above. Then the same thing is done for the Gentzen systems of \textit{R. T. Brady} [J. Philos. Logic 19, 35-73 (1990; Zbl 0697.03006), ibid. 20, 97- 117 (1991; Zbl 0725.03007)]. (Again, only categories 1) and 4) are significant.) Finally, the same treatment is given to classical sentential and predicate logic, some normal and non-normal modal logics, and to some many-valued logics. The paper does not discuss the intuitive significance, or lack thereof, of the five categories delineated for rules. One cannot help but wonder whether the additional distinctions mark real logical differences or are mere artefacts of the chosen representational schemes.