Practical Information - Generative Linguistics in the Old World
Practical Information - Generative Linguistics in the Old World
Practical Information - Generative Linguistics in the Old World
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
Stefan Ke<strong>in</strong>e(University of Massachusetts, Amherst)Long-Distance Agreement, Improper Movement and <strong>the</strong> Locality of AgreeBackground: As is well-known, H<strong>in</strong>di allows for long-distance agreement (LDA) between amatrix verb and <strong>the</strong> direct object of an embedded <strong>in</strong>f<strong>in</strong>itival verb (see Bhatt 2005 and referencescited <strong>the</strong>re). LDA is generally optional and alternates with m.sg default agreement:(1) Raam-ne roṭii khaa-nii caah-ii / khaa-naa caah-aaRam-erg bread.f eat-<strong>in</strong>f.f.sg want-pfv.f.sg / eat-<strong>in</strong>f.m.sg want-pfv.m.sg‘Ram wanted to eat bread.’The agreement/A-movement correlation: I present novel evidence show<strong>in</strong>g that <strong>the</strong> optionalityof LDA is only apparent. LDA correlates with whe<strong>the</strong>r or not A-subextraction takes place.H<strong>in</strong>di uses both A- and Ā-scrambl<strong>in</strong>g, which are subject to different locality conditions. Moreover,Ā-scrambl<strong>in</strong>g, but not A-scrambl<strong>in</strong>g, is subject to weak crossover (Mahajan 1990). In(2) <strong>the</strong> direct object har billii ‘every cat’ is scrambled above <strong>the</strong> matrix subject us-ke malik-ne‘its owner’, a movement step that could be ei<strong>the</strong>r A- or Ā-scrambl<strong>in</strong>g. LDA is not affected andrema<strong>in</strong>s optional. In (3) <strong>the</strong> object is likewise scrambled but here <strong>the</strong> pronoun embedded <strong>in</strong>side<strong>the</strong> matrix subject is co<strong>in</strong>dexed with it. This movement must be A-movement as Ā-movementwould <strong>in</strong>cur a crossover violation. In contrast to (2), LDA becomes obligatory <strong>in</strong> (3).(2) har billii 1 us-ke 2 malik-ne t 1 ghumaa-nii caah-ii / ghumaa-naa caah-aaevery cat.f its owner-erg walk-<strong>in</strong>f.f.sg want-pfv.f.sg / walk-<strong>in</strong>f.m.sg want-pfv.m.sg‘Its 1 owner wanted to walk every cat 2 .’us-ke 1itsmalik-ne t 1 ghumaa-nii caah-ii / *ghumaa-naaowner-erg walk-<strong>in</strong>f.f.sg want-pfv.f.sg / walk-<strong>in</strong>f.m.sg(3) har billii 1every cat.fcaah-aawant-pfv.m.sg‘For every cat x, x’s owner wanted to walk x.’This pattern generalizes. In (4) and (5) it is <strong>the</strong> <strong>in</strong>direct object that is scrambled above <strong>the</strong> matrixsubject. If it is not co<strong>in</strong>dexed with <strong>the</strong> pronoun <strong>in</strong>side <strong>the</strong> subject DP, as <strong>in</strong> (4), LDA is optional.(5), by contrast, conta<strong>in</strong>s a co<strong>in</strong>dexed pronoun and LDA becomes obligatory.(4) har bacce-ko 1 us-kii 2 mãã-ne t 1 film dikhaa-nii caah-ii /every child-dat his mo<strong>the</strong>r-erg movie.f show-<strong>in</strong>f.f.sg want-pfv.f.sg /dikhaa-naa caah-aashow-<strong>in</strong>f.m.sg want-pfv.m.sg‘His 2 mo<strong>the</strong>r wanted to show a movie to every child 1 .’(5) har bacce-ko 1 us-kii 1 mãã-ne t 1 film dikhaa-nii caah-ii /every child-dat his mo<strong>the</strong>r-erg movie.f show-<strong>in</strong>f.f.sg want-pfv.f.sg /*dikhaa-naa caah-aashow-<strong>in</strong>f.m.sg want-pfv.m.sg‘For every child x, x’s mo<strong>the</strong>r wanted to show a movie to x.’In both examples, <strong>the</strong> LDA controller film ‘movie’ rema<strong>in</strong>s <strong>in</strong> its base position and may receivean <strong>in</strong>terpretation as a weak <strong>in</strong>def<strong>in</strong>ite. This strongly suggests that LDA is not correlated withA-movement of <strong>the</strong> direct object per se. Ra<strong>the</strong>r, LDA is obligatory if any DP A-moves out of <strong>the</strong>embedded clause. Conversely, f<strong>in</strong>ite clauses, which are islands for A-scrambl<strong>in</strong>g but do allowĀ-extraction are also opaque for LDA. This motivates <strong>the</strong> new empirical generalization <strong>in</strong> (6).(6) Generalizationa. If any element is A-moved out of <strong>the</strong> embedded clause, LDA is obligatory.b. Clauses that are opaque for A-extraction are also opaque for LDA.