Sometimes you just feel lucky : I was reading the famous article by Charles J. Fillmore, “Corpus linguistics” or “Computer-aided armchair linguistics”, in the proceedings of a Nobel symposium which took place in 1991 (it is known for the introducing descriptions of the armchair and of the corpus linguist who don’t have anything to say to each other) as I decided to read the following article. The title did not seem promising to me, but still, it was written by Halliday :

M.A.K. Halliday, Language as system and language as instance: The corpus as a theoretical construct, pp. 61-77.

The author gives a few insights on the questions which one could ask to a given text to find a language model. One of the points has to do with “text dynamics”. Here is how Halliday defines it :

« It is a form of dynamic in which there is (or seems to be) an increase in complexity over time: namely, the tendency for complexity to increase in the course of the text. » (p. 69)

In fact, Halliday develops a very interesting idea from the textual dimension of complexity, also named the “unfolding of the text” (p. 69), its “individuation” or the “logogenesis” (this word was apparently coined by J. Lemke in Technical discourse and technocratic ideology, 1990).

Then he speaks of “some local increases in complexity”, for instance in scientific and technical discourse :

« An example of a more general increase in complexity is provided by grammatical metaphor. » « Nominal strings like glass fracture growth rate and increasing lung cancer death rate tend to accumulate throughout the text rather than being announced at the beginning. » (p. 70)

The examples above stand for what he calls a “word complex”, which is constructed by the text.

Last but not least, Halliday gives ideas of measures (p. 71) :

  • general ones, with criteria like lexical density, the number of lexical items per ranking clause, perhaps weighted for their overall frequency bands ;
  • more specific ones : the length of nominal chains or adjectival-nominal chains, a selection of verbs which are typically associated with it ;
  • other possible ways to increase complexity : the average number of ranking clauses per sentence, the number of phrases and clauses “rankshifted” inside nominal groups.

There are possibly decreases in the course of a text, but according to him it is harder to see how they could be motivated for themselves.

J. Svartvik (ed.), Directions in Corpus Linguistics, Nobel Symposium 82, Mouton de Gruyter, 1992.