The Greek Verbal Network Viewed from a Probabilistic Standpoint 11 cal features and classes 37. People, according to Halliday, have little difficulty accepting that they use the word go more often than the word walk, and walk more often than stroll to describe motion. However, when the same understanding is applied to grammatical fea- tures—that a person is more likely to use an active construction than a passive, or a positive rather than a negative clause—Halliday has obser ved that «many people object very strongly, and protest that they have a perfect right to choose otherwise if they wish» 38. In order to apply a probabilistic understanding to the grammar of a language, one must adopt a paradigmatic view of grammar. Such a model arranges grammatical categories (aspect, voice, person, gender, etc.) into combinations of either/or choices—for instance a noun is either singular or plural. The language user is most often completely unaware that such choices are being made—and thus the choices are unconscious—yet the language system requires that such choices be made 39. Each grammatical category in a particular system is given meaning by its relationship to the other categories within the system. Nesbitt and Plum state that: «A system is defined as an entry condition together with a set of mutually exclusive options or features, one of which must be selected» 40. So, for instance, plurality has little semantic value unless it is viewed as a choice against (or instead of ) singularity. The use of system in this technical, linguistic sense 37 Halliday, «Corpus Studies and Probabilistic Grammar», p. 31; idem, «Quantitative Studies», p. 2. There has been a considerable amount of investigation into numerical behaviour of lexis in language, resulting in a number of vocabulary measures, such as Zipf ’s Law (G.K. Zipf, The Psychobiology of Language [Boston: Houghton Mifflin, 1935]) and various forms of the Type-Token ratio. These measures have often been uti- lized in authorship attribution studies (e.g. G.U. Yule, The Statistical Study of Literary Vocabulary [Cambridge: Cambridge University Press, 1944]; A. Ellegård, A Statistical Method for Determining Authorship: The Junius Letters, 1769–1772 [Göteborg: Elanders Boktryckeri Aktiebolag, 1962]). See O’Donnell, «Linguistic Fingerprints», pp. 215-16, 230-39. 38 Halliday, «Quantitative Studies», p. 3. Halliday suggests that: «The resistance seems to arise because grammar is buried more deeply below the level of our conscious aware- ness and control; hence it is more threatening to be told that your grammatical choices are governed by overall patterns of probability.» 39 See Porter, Verbal Aspect, p. 9, on conscious and unconscious choice in terms of systemic linguistics. 40 C. Nesbitt and G. Plum, «Probabilities in a Systemic-Functional Grammar: The Clause Complex in English», in R.P. Fawcett and D. Young (eds.), New Developments in Systemic Linguistics. II. Theory and Application (London: Pinter, 1988), pp. 6-38, quota- tion p. 7. They continue: «The entry condition of a system is itself an option in a prior system. So the environment of choice is always that of choices already made. In this way systems form networks of systems organized according to the logical priority of certain options over other options.» See also R.P. Fawcett, Cognitive Linguistics and Social Interaction: Towards an Integrated Model of a Systemic Functional Grammar and the Other Components of a Communicating Mind (Heidelberg: Julius Groos and Exeter University, 1980), pp. 19-25; Porter, Verbal Aspect, pp. 7-16.