LECTURETTE #14: INTRO TO GOVERNMENT THEORY: THE C-COMMAND RELATION (Noam Chomsky is, of course, famous outside of the fields of linguistics, psychology, etc. for his radical political views and his eagerness to express them. So it's perhaps not altogether surprising that some of the terminology he uses in his scholarly work carries some political connota- tions. 'Government', of course, is a term with a long tradition in gram- matical scholarship. But i'm not so sure about 'binding'. I have myself had the experience, and have heard from other linguists of similar ones, of mentioning the syntactic theory known as 'Government & Binding' and having the label understood in a political sense -- binding arbitration, that sort of thing. And one of the ultimate developments of the theory of government, 'binding' (pardon the pun) together also the theories of binding and bounding, is the concept of the 'barrier', which we will be getting to soon. Geoff Pullum has complained about the inconsistency of Chomsky, whose political views are so notoriously leftist, imposing on linguistic terminology such overtones of oppressive totalitarianism. Geoff himself may have been deliberately fighting back at this trend by positing in 1982 a grammatical phenomenon which he called 'Liberation'.) The theory of government in REST is one of the 'heavier' modules of the framework. There is a lot of detail to keep straight here, and it is complicated by the variety of definitions proposed in the literature by different practicioners for some of the relevant relations. At this point i'm going to try to get over this rough ground as lightly as pos- sible without shortchanging anybody. We can come back to some of the gory details later on, if desired. You will remember my statement in lecturette #2 that in REST it is as- sumed that ALL relations of syntactic relevance are to be defined in terms of constituent structure. This is certainly true of government relations. So most of what i'm going to be doing here and in the next two lecturettes is defining those constituent-structure relations that are essential to REST's theory of government. Beyond the basic dominance and precedence relations defined in lecturette #2, the most basic structural relation in REST is c-command. The 'c' stands for 'constituent', and when we get around to talking about other frameworks it will be noted that some of them define parallel relations: LFG has 'f-command', where 'f' stands for 'functional', and HPSG has 'o- command', where 'o' stands for 'obliqueness'. But this is REST where it's all constituent-structure, and we have c-command. C-command is a relation that holds between sisters or, to extend the femi- nine terminology a bit, between aunts and nieces. The simplest and most straightforward (though as we shall see not necessarily the most accurate or useful) definition of c-command is that the c-command relation holds between two nodes A and B if neither dominates the other but every node that dominates one also dominates the other. Since sisters by definition share all dominating nodes they always c-command each other and are hence an example of symmetrical c-command, which is trivial and generally not of much interest. More important is asymmetrical c-command. When they are not sisters, node A c-commands B (but not vice-versa) if every node dominating A also dominates B. In general the definition is restricted so as to ignore intervening non-branching nodes, nodes which themselves have only one daughter apiece. Thus, in (1) the node B c-commands C, D, and E because the first non-branching node dominating B, namely Y, also dominates C, D, and E. Furthermore, because X is a non-branching node not only does it c-command Y, B, C, D, and E but so does A. (1) Z / \ X \ | \ A Y / \ B C / \ D E I mentioned back in lecturette #6 on the SPC that adjunction is regarded as a structure-preserving transformation. Adjunction is also typically regarded as transparent for the purposes of c-command. In an adjunction structure such as (2) not only does X c-command Y, B, and C, but so does A; the node X dominating it is not considered a branching node because it is a copy of another node which it also dominates. (2) Z / \ X \ / \ \ A X Y / \ B C I will mention here, before going on to the more complex government rela- tions whose definition depends on c-command, a variant that has been talked about occasionally in the literature and which is sometimes called 'm-command'. 'M' stands for 'maximal', and 'm-command' is just like 'c- command' except that instead of every branching node only maximal projec- tions count. Thus in a structure like (3) none of the words in the geni- tive NP 'the boy's' c-commands the N1 'blue bike', and none of the words in the N1 'blue bike' c-commands the genitive NP. But the words 'blue' and 'bike' do m-command the genitive NP, asymmetrically, because the first maximal projectin dominating 'blue bike' is the NP at the top, which also dominates the genitive NP. (3) NP /\ / \ / \ / \ NP N1 / \ / \ Det N1 A N | | | | the boy's blue bike Unfortunately, i have to warn you that some researchers use the expres- sion 'c-command' when they mean 'm-command'. It has been suggested that, to the extent that the distinction can be made, it is 'm-command', not 'c-command' that is important, i.e., all structural relations should be defined solely in terms of maximal projections. This may be true, but until it has been established beyond a shadow of a doubt i think we should maintain the teminological distinction between c- and m-command. So much for c-command (and m-command). Coming next: the government rela- tions essential to Theta- and Case-Theory. Best, Steven --------------------- Dr. Steven Schaufele 712 West Washington Urbana, IL 61801 217-344-8240 fcosws@prairienet.org **** O syntagmata linguarum liberemini humanarum! *** *** Nihil vestris privari nisi obicibus potestis! ***