||The adroit use of adversative constructions (Halliday and Hasan 1976) in written text is a key element in building a logical, cohesive and convincing argument. It is the purpose of this paper to examine a number of examples starting from selected connective devices and, firstly, collocating them in their specific syntactic contexts in order to identify any recurrent patterns which come to the fore, as in the following example: In another section, the high-level thinking skills had dropped from a fifth of 14-year-olds to just 5%. However, average intellect had improved in a generation. (Education Guardian online, 27/10/2008) Here the concessive ‘however’ serves as a link between the two sentences and may be schematically represented as follows (S = sentence): S1 CONCESSION (front focus) S2 The connective is classified according to its function and sentence initial position ‘front focussing’ the concession. Note how the position of the connective may also be embedded within the structure of the sentence, diminishing the impact of the concession: Average intellect had, however, improved … Thus: S1 CONCESSION (embedded) S2 Or how it can also be, more rarely, ‘end focussed’ with a significantly marked emphasis (Leech and Short: 223-225) Average intellect had improved in a generation, however. Therefore we find: S1 CONCESSION S2 (end focus) Secondly, it was decided to analyse the examples on a semantic level in order to see how this can be mapped onto the structural level in creating a message which Widdowson (1983: 69-71) defines as “coherent”. Thus our first example becomes: S1(negative -) CONCESSION (front focus) S2 (positive +) An ad hoc corpus was first constructed with material from a widely circulated British national daily newspaper. It was decided to focus on media discourse, since, while newspaper reporting theoretically represents an impartial account of events, “There is always a decision to interpret and represent it in one way rather than another” (Fairclough 1995: 54). The use of adversative constructions plays an important role in representing reality from one perspective, while appearing at the same time to be balanced and unbiased. An initial small-scale manual analysis was carried out on the corpus in order to classify the selected connective devices according to frequency, meaning and function and this was then extended to an analogical comparison which allowed the further extraction of material which was then analysed automatically. The setting up of the corpus, analysis of the data acquired and conclusions will be fully illustrated in the paper. Fairclough, N. (1995) Critical Discourse Analysis, Harlow: Longman. Halliday, M.A.K. and R. Hasan (1976) Cohesion in English, Harlow: Longman. Leech, G.N. and M.H. Short (1981) Style in Fiction, Harlow: Longman. Widdowson, H.G. (1983) Learning Purpose and Language Use, Cambridge: CUP.