You are on page 1of 2

Exercise 2 by Thomas Colin and Kim Cruden

TOWARDS A VARIABLE FREE SEMANTICS


PAULINE JACOBSON

In this paper, we argue against the standard account that it is preferable to treat the
semantics of pronouns without making use of variables. We offer an alternative treatment of
pronouns and pronoun binding at the semantic and syntactic levels by making use of new
syntactic and semantic operators, and discuss ample empirical evidence in favor of this
approach.

The standard account


The standard account starts from the intuition that pronouns act like nouns. This intuition
seems to proceed from cases such as (1), (2) and (3), in which pronouns and nouns are subject
to the same syntactic rules (we will argue that this intuition is mistaken).

(1) Every man said that every woman thinks that she should talk to him
(2) Every man said that every woman thinks that Mary should talk to him
(3) Every man said that every woman thinks that Mary should talk to John

The standard account makes use of variables for pronoun binding. For example, in (4), his
dog remains an open expression until it is affected by a value from some assignment
function.

(4) Every man loves his dog

So, in the standard account, pronouns are treated syntactically as nouns. At a semantic level, a
special entity (variables) is created to account for their behavior, and for binding pronouns to
the nouns they represent. This binding is made by applying assignment functions (thus one
has to shift the meaning of a phrase containing a pronoun into a function from assignment
function to properties).

Variable-free semantics
It is possible to describe pronoun binding using syntactic and semantic operators instead of
variables. This is accomplished firstly by treating pronouns as a function from individuals to
individuals (<e, e>; although we introduce the notation ee). Two operations, g and z, are
introduced to keep track of the need to fulfill the argument of this identity function.

For example, in (5), his dog is the function which takes an individual as input and returns
the dog of this individual. Syntactic and semantic shifts are applied to loves to allow it to
take an extra argument, and to "man" to allow it to be taken as input for "his dog" (thus
keeping track of the need to provide an argument for the function "his dog").

(5) Every man loves his dog

The variable-free approach accounts for a number of phenomena detailed in the paper,
including multiple pronouns. It makes use of new types to prevent argument mismatch for
example in cases such as (6) (in which unlike as in (5) we do not want the subject of love
to also bind the pronoun as that would make Mary both lover and lovee).

(6) John thinks Mary loves him.

1
Thus, in the variable-free semantics a pronoun isnt a variable, but denotes a function from
individuals to individuals (the identity function for pronouns like him or her ). Binding
is done by parallel syntactic and semantic shifts using two rules (the g-rule and the z-rule).

Advantages of the variable-free account


Major advantages of the variable-free account are:
The semantic effect is local, unlike in the standard account. Direct compositionality is
possible.
Much of the work is delegated to the syntax.
A layer is removed between surface structure and meaning: assignment functions (and
in some cases, reconstruction steps). There is thus a reduction of complexity.

Empirical motivation for variable-free semantics


In addition, there is ample independent motivation for variable-free semantics, as it allows for
new, intuitively satisfying and efficient explanations for a variety of phenomena in which the
standard account runs into major issues, most notably:
Functional questions. Pronouns are understood as functions in variable-free semantics,
as opposed to nouns requiring binding as in the standard account. The semantics we
propose can thus be effortlessly applied to the analysis of functional questions.
Across the board binding. The locality of pronouns in the variable-free account
permits solving such cases without going through a reconstructive step, as the standard
account does. Besides, we show that reconstruction or use of traces isn't an acceptable
solution for across the board binding, as it adds ambiguities that do not exist in the
original sentence.

Conclusion
The new approach we propose not only allows a gain in simplicity by removing an unneeded
layer of abstraction from semantic analysis. It also gives new, elegant and precise solutions to
sentences that are problematic for the standard approach. It isn't however a general solution to
all problems concerning anaphora. Some of the vexing problems existing of the standard
account remain open: for example, donkey anaphora. The interaction of this theory with
accounts of extraction also remains to be worked out. This is left for future research. We
however believe to have shown good reasons as to why further research on anaphora should
take place within the paradigm of variable-free semantics.

You might also like