Craig Thiersch (craig@sfs.nphil.uni-tuebingen.de)
Wed, 13 May 1998 16:00:40 +0200
Dear Hyeun Lee, Since Detmar's out preparing his seminars, I thought I'd try to help out by taking over Detmar's Inn for the day, on the question of what "configurational" and "non-configurational" mean. I think it's probably most useful to give an historical answer (i.e., pre-HPSG and pre-G/B). Much (transformational) syntax in the 1960's and 70's defined relationships in terms of tree structures, which determined both linear order and hierarchical structure: to give a very simple example, in an English sentence "the cat bit itself", if the structure were S (1) [ NP_s [_x V NP_o]], i.e., / \ NP_s x / \ V NP_o then NP_s was structurally superior to (commanded) NP_o and the relation between anaphor and antecedent could be defined such that the antecedent had to command the anaphor, ruling out "*Itself bit the cat" (with "itself" the subject). Studies of languages (Japanese, German, and many more) in which, for example, the orders (2) "NP_s NP_o V" and "NP_o NP_s V" both seemed equally acceptable called this into question, the extreme case being Warlpiri, where parts of noun phrases could be "scrambled" as well. The latter inspired the suggestion (due to Ken Hale) that such languages were non-configurational; i.e., unlike "configurartional" languages, which had articulated structures like (1), these languages had "flat" structures, without the node "x". Phenomena like binding of anaphors were to be accounted for by other mechanisms. The alternative was to claim that all languages were configurational and the alternative structures in (2) were to be accounted for by assuming one (e.g., the first) was "basic" with the structure (2a) [ NP_s [_x NP_o V]] and the second derived, with the structure (3) NP_o [NP_s [_x t_o V]] the anaphoric element t_o being phonetically vacuous (SLASH in GPSG/HPSG terms, "trace" in G/B terms). Both approaches have been explored in both frameworks and for both language types; evidence pro and contra is still being gathered. The issue is quite a complicated one, since it involves the interactions of many different components of the grammar (e.g., whether the Binding Theory is based on hierarchical configuration (tree structure) or some level of argument structure (SUBCAT), to name just one example). Furthermore, on the empirical side, languages which look superfically "non-configurational" (e.g., very free constituent order) generally turn out to have various configurational properties which have to be accounted for (e.g., NP_o and V can be "topicalized" but not NP_s and V), and vice versa, languages which look rigidly configurational (e.g., English) turn out to have much more freedom than apparent at first glance. Hope this is helpful. Best, Craig, temporarily in Tuebingen, where it's also 30 C. ------------------------------------------------------------------------ Craig Thiersch (Tilburg University) Lett./Gr.Mod. (Linguistics) e-mail: C.L.Thiersch@kub.nl Kath. Universiteit Brabant tel. +31-13 / 466 31 24 Postbus 90153 FAX: +31-13 / 466 31 10 NL-5000 LE Tilburg, The Netherlands http://cwis.kub.nl/~fdl/general/people/thiersch/index.htm ------------------------------------------------------------------------
This archive was generated by hypermail 2.0b3 on Fri Dec 18 1998 - 20:38:19 PST