Transcript Slide 1
Linguistic Theory Lecture 7 About Nothing Nothing in grammar • Language often contains irregular paradigms where one or more expected forms are absent • E.g. English present tense verb agreement 1st 2nd 3rd Singular Am Are Is plural Are Are Are – We see from the paradigm for be that number and person play a role in determining the form of the verb in the present tense: • But other verbs do not show the same pattern 1st 2nd 3rd Singular Smile Smile Smiles Plural Smile Smile Smile • The only form which shows any agreement is • the 3rd person singular Two choices: – Assume that there is no verbal agreement except for 3rd person singular and for 1st, 2nd and 3rd person singular with the verb be – Assume that there is a complete set of verbal agreements, only most of them are realised by a null morpheme • The second choice is the one usually made as it makes the system more regular Other kinds of nothing • Ellipsis – She wanted to watch the TV, but I didn’t – (want to watch the TV) – * (take any notice) • There is a difference: – null morpheme = absent at the phonological level – elliptical material = present at the semantic level Nothing in the 1960s • One possible way to treat ellipsis is as a deletion: – John drank beer and Bill wine • John drank beer and Bill drank wine • Deletions are recoverable: – * John drank beer and Bill biscuits • John drank beer and Bill ate biscuits • This shows that ‘recoverability’ is a limited notion: – Recoverable from syntactic not pragmatic context • A similar approach can account for the following observations: – The man [who I spoke to] – The man [who I spoke to] – He asked [who I spoke to] – * He asked [who I spoke to] • It is assumed that the same process is • • involved in relative clause and interrogative clause formation But if so, why can the wh-relative delete but not the wh-interrogative? The wh-relative has an antecedent in the noun that it modifies, so is recoverable. The wh-interrogative does not and so is unrecoverable. Equi NP Deletion • John wants [Bill to leave] • Bill wants [to leave] • Bill1 wants [Bill1 to leave] • Equi-NP deletion: In structures: ... NP1 ... NP1 ... Delete the second NP • But due to constraints on transformations, deletion transformations fell out of favour • Chomsky and Lasnik (1977) proposed that instead of a deletion, ‘Equi’ structures involve a phonologically null pronoun (PRO): • Bill1 wants [PRO1 to leave] • PRO has two properties that need to be accounted for: – Its referential behaviour (control) – Its distribution Control • Like most pronouns, PRO can take its reference from an antecedent: – John1 dressed himself1 – John1 thinks Mary likes him1 – John1 wants [PRO1 to be loved] • However, it has special referential properties of its own – Subject/Object control • John1 promised Bill [PRO1 to be good] • John persuaded Bill1 [PRO1 to be good] – Obligatory/Arbitrary control • John1 tried [PRO1/*2 to sing] • [PROarb to sing now] would be inappropriate The distribution of PRO: the PRO theorem • PRO is an NP • But its distribution is not the same as a typical NP: • • • • I saw him I spoke to him He left * I tried [him to sing] • • • • * I saw PRO * I spoke to PRO * PRO left I tried [PRO to sing] • At first sight it seems that PRO cannot appear in a Case position (it is an exception to the Case Filter) • But there are non-Case positions where PRO cannot go either – * John’s picture PRO – * John is very fond PRO • So the restriction on the distribution of PRO is more stringent Government • Government is a relationship between certain elements (governors) and certain positions: – Governors = lexical heads (N, V, P and A) and finite Inflection – Governors govern complement and specifier positions: • XP spec X’ X comp • Case assigners are governors • (but not all governors are Case assignors) • So the set of all Case positions is a subset of the set of all governed positions: Governed positions Case positions • PRO must be ungoverned • Therefore it cannot appear in a Case position Governed positions Case positions PRO Explaining the PRO theorem • Anaphors (reflexive pronouns and NP traces) must have a close by antecedent – John1 admires himself1 – * John1 thinks [Mary admires himself1] – John1 was admired t1 – * John1 was believed [Mary to admire t1] • Pronominals (personal pronouns) cannot have a close by antecedent: – * John1 admires him1 – John1 thinks [Mary admires him1] • Pronominals don’t have to have antecedents at all (anaphors do): – He left – * himself left • There is a part of the structure which contains (at least) the pronoun and a governor = the governing category • Binding theory – A: an anaphor must be bound in its governing category – B: a pronominal must be free in its governing category • Bound = coindexed with an appropriate antecedent • Free = not bound • So pronominals and anaphors are in complementary distribution • (Controled) PRO is like an anaphor – Because it must have an antecedent • (Arbitrary) PRO is like a pronominal – Because it does not need an antecedent • So PRO is a pronominal anaphor • So PRO must be bound and free in its • • • • governing category But this is a contradiction!!! The contradiction can be solved if PRO has no governing category PRO will have no governing category if it is not governed Hence the PRO theorem A typology of empty categories: Overt + pronominal - pronominal + anaphor * himself - anaphor him John Covert + pronominal - pronominal + anaphor PRO NP-trace - anaphor ** Wh-trace * doesn’t exist because all (overt) NPs must have Case and therefore must be governed ** exists, but not in English: missing subject of finite clause in e.g. (most) Romance languages, Slavic languages, Semitic languages, Hungarian, etc.