208 Kerstin Kunz and Erich Steiner
(1977: 657ff ) distinguishes between ‘forms’ and ‘expressions’, or ‘noun-substitutes’
vs. ‘nominal substitutes’ with respect to pro-forms in a similar way.
In the German literature, the term ‘Substitution’ (often ‘Wiederaufnahme’)
usually covers ‘Ko-Referenz’ and ‘Ko-Denotation’ (Schreiber 1999: 126, 159). More
precisely, it includes all kinds of relations where a ‘proform’ is linked to some other
element in the text. Co-reference may or may not be involved.
While substitution is a process of substituting words by other (semantically
weaker) words usually having the same structural function, and while it does
establish semantic ties, a substitute usually does not co-refer with its ‘antecedent’.
An additional argument in favour of a separate category of substitution for English
is that verbal and clausal substitution are different from co-reference in the classi-
cal sense altogether (see below). Even nominal substitution usually involves type
reference or co-denotation, rather than instantiated co-reference. Finally, English
has grammaticalized some forms of substitution more clearly than German.
Substitution and ellipsis share their semantic properties, yet substitution relies
on semantically bleached lexical material (one, same, do, so etc. for English, ein-,
welch-, etc. for German), whereas ellipsis is realized by zero, or a gap. Substitution
and ellipsis thus belong to one type functionally, but with different constraints on
realization.
Let us compare the relationships of cohesive reference, substitution and lexical
cohesion with each other along several dimensions (see Table 2). These cohesive
relationships all involve lexico-grammatical realization as linguistic forms, and
they also involve some kind of semantic relationship between two or more dis-
course units. Where they are different is in the type of semantic relationship they
prefer: incrementally instantiated co-reference vs. type-reference and co-deno-
tation vs. lexical semantic relation like synonymy, hyponymy etc. They are also
different in that reference, substitution/ellipsis are largely realized in grammar,
whereas lexical cohesion is realized in lexis. The dimensions of our comparisons
are defined in the leftmost column in Table 2.
Starting with row 1, we note Halliday and Hasan’s (1976: 88ff, 303ff ) claim
that “substitution is a purely textual relation, with no other function than that of
cohering one piece of text to another.” Co-reference as a meaning relation is not
the preferred semantic tie here: with a substituting item, reference is to the same
class/type of referents as that of its antecedent, but not to the same instantiated
referent. Cohesive reference, on the other hand, is a semantic relation usually
involving instantiated co-reference. Lexical cohesion, finally, relies on semantic
relationships, usually involving type-reference or conceptual similarity between
types of referents. General nouns, its limiting case, approximate substitution as a
mechanism. Lexical cohesion may trigger co-reference but only when combined
with referential devices in the phrase in question (personal or demonstrative).