\(\mathbf R\) and relevance principle revisited (Q381004)

According to \textit{A. R. Anderson} and \textit{N. D. Belnap jun.} [Entailment. The logic of relevance and necessity. Vol. I. Princeton, NJ: Princeton University Press (1975; Zbl 0323.02030)], the ``variable-sharing property'' (vsp) is a necessary property of any relevant logic. The vsp reads as follows: a propositional logic S has vsp if in each theorem of the form \(A\rightarrow B\), \(A\) and \(B\) share at least a propositional variable. The present paper is a study of vsp in the relevant logic \textbf{R} and its extension \textbf{RM} (\textbf{R} plus the ``mingle axiom'', \(A\rightarrow (A\rightarrow A)\)). The author refers to vsp as ``the relevance principle''. The logic \textbf{R} (and any logic included in it) has the vsp (cf. [loc. cit.]). However, logic \textbf{RM} does not have the vsp but has the following property instead (let us call it vsp\(^{\prime }\) for simplicity reasons): a propositional logic S has vsp\(^{\prime }\) if in each theorem of the form \(A\rightarrow B\), \(A\) and \(B\) share at least a propositional variable or both \(\lnot A\) and \(B\) are theorems (cf. [loc. cit.]).\smallskip Consider now the following propositional constants together with their respective intuitive interpretation (pp. 775--776): df4. \textbf{t} = the conjunction of all true sentences; df5. \textbf{f} = the disjunction of all false sentences; df6. \textbf{T} = the disjunction of all sentences; df7. \textbf{F } = the conjunction of all sentences. Since \textit{W. Ackermann}'s seminal paper [J. Symb. Log. 21, 113--128 (1956; Zbl 0072.00106)], it has been customary to expand the main relevant logics with some or all of these constants. From the vsp perspective, the problem with these expansions is that this property is violated: \((p\wedge \)\textbf{t}\( )\rightarrow (q\vee \)\textbf{t}\()\) is a theorem in the expansion of \textbf{R} with the constant \textbf{t}, for example. The author then discusses some proposals in the literature for solving this problem and, finding them inadequate, proposes the following solution. Firstly, let us set the following definitions: (1) a propositional constant is \textit{strongly metadefinable} in S if its metadefinable but it is not definable in the object language of S (cf. pp. 778--779). Thus, for example, \textbf{t}, \textbf{f}, \textbf{T} and \textbf{F} are strongly metadefinable in \textbf{R}. (2) \(A\) and \(B\) \textit{implicitly} share a propositional variable in \( A\rightarrow B\) if it is possible to identify a shared variable by means of the metadefinitions df4--df7 (cf. Definition 6, p. 777). Thus, for example, \( p\wedge \)\textbf{t} and \(q\vee \)\textbf{t} in the formula recorded above ``implicitly'' share a propositional variable since \textbf{t} is the conjunction of all truth sentences. Then, the proposed modifications of vsp and vsp\(^{\prime }\) are as follows (cf. Definition 8, p. 780):\smallskip -- A logic S has the (modified) vsp if in each theorem of the form \( A\rightarrow B\), (1) \(A\) and \(B\) implicitly share a propositional variable, and (2) the propositional constant(s) and \(A\) and \(B\) are strongly metadefinable.\smallskip -- A logic S has the (modified) vsp\(^{\prime }\) if in each theorem of the form \(A\rightarrow B\), (a) properties (1) and (2) in the definition of the vsp are fulfilled or (b) both \(\lnot A\) and \(B\) are theorems.\smallskip According to these modified definitions, \textbf{R} (\textbf{RM}) and all its extensions by means of the propositional constants \textbf{t}, \textbf{f} , \textbf{T} and \textbf{F} have vsp (vsp\(^{\prime }\)).