DEFINING IN TYPE LOGIC ALL LOGICAL OPERATIONS IN TERMS OF APPLICATION, ABSTRACTION, AND IDENTITY 1. We introduce 1 as a name of a tautology. For this, we need to find a tautology expressed only with λ, ( ( )), and =. Let α 2 VARt. λα.α denotes the identity function ID = {<1,1>,<0,0>}. Since this function is self-identical, (λα.α = λα.α) is a tautology. We define: 1 := (λα.α = λα.α) Three facts: -λα.1 denotes the constant function on 1: C1 = {<1,1>,<0,1>} -λα.α = 1 is equivalent to λα.α This is easy to see: λα.α = 1 (1) is equivalent to (1 = 1) and hence to 1 λα.α = 1 (0) is equivalent to (0 = 1) and hence to 0 Thus λα.α = 1 denotes {<1,1>,<0,0>} -Let x 2 VARa. λx.1 denotes the characteristic function of Da, hence, set-theoretically, it denotes Da. 2. We introduce 0 as a name of a contradiction. For instance, λα.α denotes the identity function, λα.1 denotes the constant function on 1. These functions are not the same function. Expressing that they are the same is a contradiction. We define: 0 := (λα.α = λα.1) Obviously, λα.0 denotes the constant function on 0, C0 = {<1,0>,<0,0>}. 3. Fact: λα.α = 0 denotes NEG = {<1,0>,<0,1>} This is easy to see: λα.α = 0 (1) is equivalent to (1 = 0) and hence to 0 λα.α = 0 (0) is equivalent to (0 = 0) and hence to 1 Thus λα.α = 0 denotes {<1,0>,<0,1>}. We define: := λα.α = 0 4. Let x 2 VARa, φ 2 EXPt. λx.φ denotes the set of all a-entities that have φ (such that φ is true). We saw above that λx.1 denotes the set of all a-entities, Da. Hence, (λx.φ = λx.1) is true iff the set of all a-entities coincides with the set of all aentities that have φ, i.e. iff all a-entities have φ. We define: xφ := (λx.φ = λx.1) 5. Tarski's definition of conjunction. Let f 2 VAR<t,t>, α, β 2 VARt ^ := λαλβ.f[ (f(α) = f(β)) = β ] Correctness proof: We need to prove,for φ,ψ 2 EXPt, that f[ (f(φ) = f(ψ)) = ψ ] is equivalent to (φ ^ ψ) (on the standard semantics of the latter). We prove this by the following equivalences: 1. v f[ (f(φ) = f(ψ)) = ψ ] bM,g = 1 2. for every k 2 D<t,t>: v (f(φ) = f(ψ)) = ψ ] bM,fk = 1 1 and 2 are equivalent by the semantics of . 3. v v v v (f(φ) = f(ψ)) = ψ (f(φ) = f(ψ)) = ψ (f(φ) = f(ψ)) = ψ (f(φ) = f(ψ)) = ψ bM,gfID = 1 and bM,gfNEG = 1 and bM,gfC1 = 1 and bM,gfC0 = 1 2 and 3 are equivalent, because D<t,t> = {ID, C1, C2, NEG}. 4. v v v v (λα.α(φ) = λα.α(ψ)) = ψ bM,g = 1 and (λα.α(φ) = λα.α(ψ)) = ψ bM,g = 1 and (λα.1(φ) = λα.1(ψ)) = ψ bM,g = 1 and (λα.0(φ) = λα.0(ψ)) = ψ bM,g =1 3 and 4 are equivalent by the coinsiderations in the above points 1-4: λα.α denotes ID, λα.α denotes NEG, λα.1 denotes C1, λα.0 denotes C0. 5. v v v v (φ = ψ) = ψ bM,g = 1 and (φ = ψ) = ψ bM,g = 1 and (1 = 1) = ψ bM,g = 1 and (0 = 0) = ψ bM,g =1 4 and 5 are equivalent by λ-conversion. 6. v (φ = ψ) = ψ bM,g = 1 and v ψ bM,g = 1 5 and 6 are equivalent by the following observations: a) (φ = ψ) and (:φ = :ψ) are equivalent, hence also (φ = ψ) = ψ and (φ = ψ) = ψ are equivalent (by extensionality). Thus the second clause in 5 is equivalent to the first, and can be dropped. b) (1 = 1) is equivalent to 1; (0 = 0) is also equivalent to 1. Hence (by extensionality), (1 = 1) = ψ and (0 = 0) = ψ are equivalent to (1 = ψ), and hence to (ψ = 1). But, by λconversion, (ψ = 1) is equivalent to λα.α = 1(ψ), and we saw above that λα.α = 1 has the same denotation as λα.α, hence (by extensionality) λα.α = 1(ψ) is equivalent to λα.α(ψ), and hence (by λ-conversion) to ψ. Hence, both the third and fourth clause in (5) are equivalent to v ψ bM,g = 1. Combining (a) and (b), we see that indeed 5 and 6 are equivalent. 7. v φ bM,g = 1 and v ψ bM,g = 1 6 and 7 are equivalent by the fact that (φ = ψ) = ψ and φ are equivalent. The latter fact is easily shown in a truth table: φ ψ (φ = ψ) (φ = ψ) = ψ 1 1 0 0 1 0 1 0 1 0 0 1 1 1 0 0 Thus, we have shown that 1 and 7 are equivalent, which, of course, indeed means that the expression which defines ^ picks out the correct two-place truth function for conjunction. As usual, with and ^ defined, we can define all other truth functions, and with xφ defined, for any x and any φ, we can define xφ for any x and any φ. This proves that we need only application, abstraction, and identity to define all other logical operations. It should be clear that identity plays a crucial role in these considerations: the power of the theory lies in the combined power of abstraction and identity, and identity itself cannot be defined with just application and abstraction.