228 Kerstin Kunz and Erich Steiner
are functionally and formally different from substitution, whereas ellipsis is func-
tionally similar. Going beyond systemic contrasts, we saw indications of different
patterns of frequencies and different specific meanings for these cohesive devices
between the languages. We also saw that, even though the morphological encod-
ing of case, number and gender is in most areas more explicit in German than in
English, ambiguities and vagueness of scope, especially with ‘generalized’ sub-
stitution, are considerable in German. More textual evidence needs to be found
to substantiate these latter impressions. And we would strongly argue that these
ambiguities and vaguenesses should be psycholinguistically tested as to their
implications for text production and understanding (see e.g. Alves et al. 2010).
Apart from the results discussed above, our corpus-linguistic analysis has also
yielded some findings with respect to register. One interesting observation, which
still requires further testing, is that high frequencies in some registers are not only
a reflection of the written vs. spoken parameter, but are due to variation accord-
ing to contextual functions. Most notably, the experiential domain seems to play
a major role for conceptualizing and linguistically explicitating type-reference or
co-denotation. We hope to obtain more evidence for this assumption with our
analysis of lexical cohesion, i.e. sense relations and lexical chains.
Let us finish with some thoughts on possible explanations and future work. It
is clear that in a diachronic perspective, lexico-grammatical phenomena change
their cohesive functionalities between lexically cohesive, co-referential and sub-
stitutive. Well documented examples are the many cases of formerly fully lexical
words turning through stages as general nouns/verbs into substitutes (e.g. one,
do) into referential pronouns and/or grammatical markers. This falls in line with a
widely assumed direction of language change over time from fully and independ-
ently lexical > cohesive > grammatical > morphological. Some of the possibilities
discussed by us, such as the German general verbs tun/machen, may be undergo-
ing such a development currently, with the associated semantic changes (bleach-
ing), scoping issues, phonological reductions, etc. This calls for the addition of a
diachronic dimension to our architecture in future work.
In considering cohesive mechanisms, we initially identify the formal realiza-
tions (pro-forms, ellipsis, conjuncts, lexical chains, etc.). In a second step, we dis-
tinguish semantics/functions: co-reference vs. type reference vs. logical linking
vs. creating networks of sense relations in chains vs. other types of collocations,
and sometimes even purely grammatical substitutions. In our methodology, we
are treating the semantic/function level as the tertium comparationis. Yet this is
all still systemic. In addition, what our empirical data show is that very differ-
ent frequencies in occurrence of whatever is systemically possible may provide
indications of how categorial the systemic distinctions are felt to be by language