- Up - | Next >> |
-expressions are formed out of ordinary first order formulas using the
-operator. We can prefix the
-operator, followed by a variable, to any first order formula or
-expression. We call expressions with such prefixes
-abstraction s (or, more simply, abstractions). We say that the variable following a
-operator is abstracted over . And we say that the variable abstracted over is (
-)bound by its respective
-operator within an abstraction, just as we say that a quantified variable is bound by its quantifier inside a quantification.
Abstractions
The following two are examples of -abstractions:
In the first example, we have abstracted over . Thus the
in the argument slot of
is bound by the
in the prefix. In the second example, we have abstracted twice: Once over
and once over
. So the
in the first argument slot of
is bound by the first
, and the
is bound by the second one.
Missing Information
We will think of occurrences of variables bound by as placeholders for missing information: They serve us to mark explicitly where we should substitute the various bits and pieces obtained in the course of semantic construction. Let us look at our first example
-expression again. Here the prefix
states that there is information missing in the formula following it (a one-place predication), and it gives this ``information gap'' the name
. The same way in our second example, the two prefixes
and
give us separate handles on each of the two information gaps in the following two-place predication.
- Up - | Next >> |