Stanley E. Porter and Matthew Brook O’Donnell 14 the complementary ideas of language as system and language as instance. These are not two different entities, but rather two viewpoints upon one phenomenon, that is, the phenomenon of language. Viewing language as a system involves taking a long-term view on multiple instances of language use and making generalizations, whereas viewing language as an instance is an examination of just one occurrence of language in use. Halliday offers the illustration of examining the weather in terms of climate (weather viewed as a system over an extended period) and in terms of day by day weather patterns (weather viewed as an instance). He suggests that: There is only one set of phenomena here: the meteorological processes of precipitation, movement of air masses and the like, which we observe in close-up, as text, or else in depth, as system. But one thing is clear: the more weather we observe, as instance-watchers, the better we shall perform as system-watchers when we turn to explaining the climate 49. Halliday began simple frequency counts for the Chinese language (Mandarin) in the early 1950s (he completed his work on Chinese in 1955, just before the publication of Chomsky’s Syntactic Structures), and assigned probabilities to the terms in the grammatical systems 50. From these frequency counts and probabilities, he wanted to discover «the extent of the association between different systems» 51. One such association investigated by Halliday was negative interrogatives, where there is a com- bination of a choice from the polarity (negative vs. positive) and the indi- cative (declarative vs. interrogative) systems 52. From this early work, and from later work carried out on small samples of English, he developed a «typology of systems». He noticed that when probabilities where applied to the grammatical systems he studied, the systems generally fell into one of two groups: (1) those where the two terms in the system have equal pro- bability of occurring (0.5), and thus there is no marked or unmarked term, and (2) those where there is an unmarked term with a probability of roughly 0.9, and a marked term with a 0.1 probability. These two types of system are referred to as equiprobable and skewed systems, respectively 53. Halliday further suggests that the fact that a language will primarily pos- 49 Halliday, «Language as System and Language as Instance», p. 66. Cf. idem, «A Brief Sketch of Systemic Grammar», in Halliday: System and Function, pp. 3-6, esp. p. 3. 50 M.A.K. Halliday, The Language of the Chinese «Secret History of the Mongols» (Publications of the Philological Society, 17; Oxford: Basil Blackwell, 1959), esp. pp. 207-26. 51 Halliday, «Quantitative Studies», p. 3. 52 He wondered «whether it was possible to predict the number of instances of nega- tive interrogative by intersecting the probabilities of negative (versus positive) with those of interrogative (vs. declarative)» (Halliday, «Quantitative Studies», p. 3). 53 Halliday, «Language as System and Language as Instance», p. 65. He notes that this typology of systems arose initially out of his work on Chinese and was then confirmed by later work on a small corpus of English. He also notes that Svartvik’s work on voice in English (J. Svartvik, On Voice in the English Verb [ Janua Linguarum, Series Practica, 63; The Hague: Mouton, 1966]) provided further support for his theory.