Richard Hudson (r.hudson@linguistics.ucl.ac.uk)
Fri, 8 Sep 1995 08:24:31 -0600 (CST)
Bob Carpenter mentions traces, which he says we have no way of deciding whether # exist. [Pardon the extraction ...] I think the Pickering and Barry evidence *against* should be taken more seriously than they have, and I don't think Ivan and Janet Fodor actually dispose of it. Let me repeat it, in a somewhat different form from theirs. a. If traces exist, they must each have a specific position among the (surface) words. This is the main claim that distinguishes theories that use traces from those that don't. b. Consider (1). (1) What did he give Mary #? If there's a trace, it must be in the position marked #, because that's the only place where a direct object can occur. Now if we replace "Mary" by a very long NP, the effect on processing should be just the same as if the direct object is overt, as in (2). (2) He gave every student capable of understanding his lectures about the syntax of the Mongolian prepositional phrase with special reference to the preposition BA an A. Prttty bad, most people agree. Ok, let's see what happens if we replace "an A" with preposed "what grade" (3) What grade did he give every student capable of understanding his lectures about the syntax of the Mongolian prepositional phrase with special reference to the preposition BA #? Pretty *good*, most people agree. This is odd if indeed the trace has the position marked. Now notice that if it has any position at all, this is the only possible position. In particular, we can be sure that the indirect object phrase "every student .... BA" hasn't been heavy-NP shifted across the direct object, leaving # right after "give", because heavy-NP shift is impossible for bare indirect objects (notice: no "to"). Nor can the postmodifier "capable of .... BA" have been extraposed, leaving # after "student", because delayed adjectives like "capable" can't extrapose: (4) He accepts any students into his class who can ... / *capable of ... Therefore the position of #, the supposed trace, in (3) is correct. c. This is an odd conclusion, because it suggests that overt NPs when delayed do cause processing difficulty, but traces don't. Why? Could it be because the hearer can tell from the verb's subcat list that the direct object position is coming straight after this interminable indirect object NP, and can guess that "what" is the expected direct object? If that were so, the same option should be available for the other subcat option for GIVE, NP + PP: give NP to NP. But if the "to" is present, it's just as bad as (2): (5) Which students did he give his boring lectures on the syntax of the Mongolian prepositional phrase with special reference to the preposition BA and various other very small words to #. The conclusion to which Pickering and Barry come, and which I agree on, is that the link between "what grades" or "which students" and the verb "give" is established as soon as "give" is heard/read - i.e. the link is shown by a dependency, and not by a structural position - and traces are redundant. Maybe they're there, but they do no work, so they may as well not be there. ============================================================================ Prof Richard Hudson Tel: +44 171 387 7050 ext 3152 E-mail: r.hudson@ling.ucl.ac.uk Dept. of Phonetics and Linguistics Tel: +44 171 380 7172 Fax: +44 171 383 4108 UCL Gower Street London WC1E 6BT UK
This archive was generated by hypermail 2.0b3 on Fri Dec 18 1998 - 20:33:04 PST