One of the fundamental differences in the semantics of functional and logic programming is the way in which we bind variables. In a logic programming language, variables are allowed to be free for as long as the inference engine does not arrive at contradictory bindings during unification.

Such existential values allow us to defer a computation into the future much like lazy evaluation. However, unlike a pure functional language with lazy evaluation, a functional logic language allow functions to return partial data structures (containing unification variables), which may later be unified with complete structures (not containing such).

In this paper, we will present the approach less formally, by comparing an extended Haskell-like functional programming language naively, by the kind of existential variables found in SWI Prolog.