Notes on computational linguistics.pdf - UCLA Department of ...
Notes on computational linguistics.pdf - UCLA Department of ...
Notes on computational linguistics.pdf - UCLA Department of ...
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
Stabler - Lx 185/209 2003<br />
10.6 Some remaining issues<br />
10.6.1 Locality<br />
When phrasal movement was defined in §9.1 <strong>on</strong> page 170, it will be recalled that we <strong>on</strong>ly allowed the operati<strong>on</strong><br />
to apply to a structure with a +f head and exactly 1 -f c<strong>on</strong>stituent. 48 We menti<strong>on</strong>ed that this restricti<strong>on</strong> is a<br />
simple, str<strong>on</strong>g versi<strong>on</strong> <strong>of</strong> a kind <strong>of</strong> “shortest move c<strong>on</strong>straint” in the sense that each -f c<strong>on</strong>stituent must move<br />
to the first available +f positi<strong>on</strong>. If there are two -f c<strong>on</strong>stituents in any structure, this requirement cannot<br />
be met. This is also a kind <strong>of</strong> “relativized minimality” c<strong>on</strong>diti<strong>on</strong> in the sense that the domains <strong>of</strong> movement<br />
are relativized by the inventory <strong>of</strong> categories (Rizzi, 1990). A -wh c<strong>on</strong>stituent cannot get by any other -wh<br />
c<strong>on</strong>stituent, but it can, for example, get by a -k c<strong>on</strong>stituent.<br />
Notice that while this c<strong>on</strong>straint allows a wh element to move to the fr<strong>on</strong>t <strong>of</strong> a relative clause,<br />
the man whoi you like ti visited us yesterday<br />
it properly blocks moving another wh-element out, e.g. forming a questi<strong>on</strong> by questi<strong>on</strong>ing the subject you:<br />
*whoj did the man whoi tj like ti visited us yesterday * the man whoi I read a statement whichj tj was about<br />
ti is sick<br />
When this impossiblity <strong>of</strong> extracting out <strong>of</strong> a complex phrase like the man who you like was observed by<br />
Ross (1967), he observed that extracti<strong>on</strong> out <strong>of</strong> complex determiner phrases is quite generally blocked, even<br />
when there are (apparently) no other movements <strong>of</strong> the same kind (and not <strong>on</strong>ly in English). For example, the<br />
following should not be accepted:<br />
*whoi did the man with ti visit us<br />
* the hat whichi I believed the claim that Otto was wearing ti is red<br />
How can we block these? The SMC is apparently too weak, and needs to be strengthened. (We will see below<br />
that the SMC is also too str<strong>on</strong>g.)<br />
Freezing<br />
Wexler, Stepanov, et al.<br />
10.6.2 Multiple movements and resumptive pr<strong>on</strong>ouns<br />
In other respects, the SMC restricti<strong>on</strong> is too str<strong>on</strong>g. In English we have cases like (22a), though it is marginal<br />
for many speakers (Fodor, 1978; Pesetsky, 1985):<br />
(22) a. ?? Which violins1 did you say which s<strong>on</strong>atas2 were played t2 <strong>on</strong> t1<br />
b. * Which violins1 did you say which s<strong>on</strong>atas2 were played t1 <strong>on</strong> t2<br />
The example above is particularly troubling, and resembles (23a) famously observed by Huang (1982):<br />
(23) a. ? [Which problem]i do you w<strong>on</strong>der howj to solve ti tj?<br />
b. * Howj do you w<strong>on</strong>der [which problem]i to solve ti tj?<br />
It seems that in English, extracti<strong>on</strong> <strong>of</strong> two wh-elements is possible (at least marginally) if an argument whphrase<br />
moves across an adjunct wh-phrase, but it is notably worse if the adjunct phrase extracts across an<br />
argument phrase. We could allow the first example if wh-Adverbs have a different feature than wh-DPs, but<br />
then we would allow the sec<strong>on</strong>d example too. If there is really an argument-adjunct asymmetry here, it would<br />
apparently require some kind <strong>of</strong> fundamental change in the nature <strong>of</strong> our SMC.<br />
Developing insights from Obenauer (1983), Cinque (1990), Baltin (1992) and others, Rizzi (2000) argues that<br />
what is really happening here is that certain wh-elements, like the wh-DP in a above, can be related to their<br />
traces across another wh-element when they are “referential” in a certain sense. This moves the restricti<strong>on</strong> <strong>on</strong><br />
movement relati<strong>on</strong>s closer to “binding theory,” which will be discussed in §11. (Similarly semantic accounts<br />
have been <strong>of</strong>fered by many linguists.<br />
48 In our implementati<strong>on</strong>, we actually do not even build a representati<strong>on</strong> <strong>of</strong> any c<strong>on</strong>stituent which has two -f parts, for any f.<br />
227