Yesterday, @gravbeast asked on Twitter,
Does anyone know why we traditionally use Greek phi and psi for metasyntactic variables representing arbitrary logic formulas? Is it just because ‘formula’ begins with an ‘f’ sound? And chi was being used for other things?
Although Whitehead and Russell already used φ and ψ for propositional functions, the convention of using them specifically as meta-variables for formulas seems to go back to Quine’s 1940 Mathematical Logic. Quine used μ, ν as metavariables for arbitrary expressions, and reserved α, β, γ for variables, ξ, η, σ for terms, and φ, χ, ψ for statement. (ε, ι, λ had special roles.) Why φ for statements? Who knows. Perhaps simply because Whitehead and Russell used it for propositional functions in Principia? Or because “p” for “proposition” was entrenched, and in classic Greek, φ was a p sound, not f?
The most common alternative in use at the time was the use of Fraktur letters, e.g., \(\mathfrak{A}\) as a metavariable for formulas, and A as a formula variable; x as a bound variable and \(\mathfrak{x}\) as a metavariable for bound variables. This was the convention in the Hilbert school, also followed by Carnap. Kleene later used script letters for metavariables and upright roman type for the corresponding symbols of the object language. But indicating the difference by different fonts is perhaps not ideal, and Fraktur may not have been the most appealing choice anyway, both because it was the 1940s and because the type was probably not available in American print shops.