Here's my sketch of a Tarski result about truth not being reflected. (Tarski's Truth Theorem) We're assuming we have the type of terms and a representation relation "reps" between terms. a) We assume that if t reps s then t is closed. Notation and some simple corrolaries (indicated by "Thus"): There are also assumptions about substitution into SUBX(?,?) and Q(?). -x- is a variable (and a term) t/e is substitution of term e for variable -x- in t SUBX(t,r) reps t'/r' if t reps t', and r reps r' SUBX(t,r)/e = SUBX(t/e,r/e)) q(t) reps t Q(t) reps q(r) if t reps r. b) Thus, Q(q(t)) reps q(t). Q(t)/e = Q(t/e) f(t) is SUBX(q(t),SUBX(-x-,Q(-x-))) ) s(t) is f(t)/q(f(t)) Thus, s(t) is SUBX( q(t), SUBX( q(f(t)), Q(q(f(t)))) ) ) by (a) on q(t) Thus, s(t) reps t/(f(t)/q(f(t))) by (b) *) Thus, s(t) reps t/s(t) Not(t) is the term built from term t by the negation-denoting operator c) Thus Not(t)/e = Not(t/e). The Tarskian Argument: Let F(L,T,tr), where L and T are properties of terms and tr is a term, mean 1) forall S:term. L(tr/S) if S reps some term 2) & forall t:term. T(Not(t)) iff L(t) and not T(t) 3) & forall S,t:term. if S reps t then ( T(tr/S) iff T(t)) This is meant to be part of the criterion for T being truth on L, and for tr to denote T (in -x-). L is supposed to be the class of sentences, T the purported truth predicate. Glosses of 1,2,3 are: 1) says that inserting any term-representing term into tr forms a sentence; 2) says that negation is faithfully interpreted by the purported truth predicate (we don't need the faithful interpretation of other connective for this proof); 3) says that (substitution into) term tr represents the purported truth predicate itself under the purported truth interpretation. Then there are no L,T,tr such that F(L,T,tr) because: 4) Assume F(L,T,tr) 5) let S = s(Not(tr)) 6) S reps Not(tr/S) by (5,*,c) 7) L(tr/S) by (4,1,6) 8) T(tr/S) iff T(Not(tr/S)) by (4,3,6) 9) T(Not(tr/S)) iff L(tr/S) & not T(tr/S) by (4,2) 10) T(Not(tr/S)) iff not T(tr/S) by (9,7) T(tr/S) iff not T(tr/S) by (8,10) which is false so (4) is false. qed http://www.cs.cornell.edu/Info/People/sfa/ --------- (modifications from previous versions: - moving the "Not" stuff to after f.p. construction. - and simplifying substitution notation. - changed "FU" to "F" - made L an explicit argument to F - provided glosses of the clauses in the def of F)