Re: USAGE: Garden paths

Nik Taylor wrote:

> > But getting to your main point:  I suspect that your requirement that a
> > complementizer or relative pronoun be present in the cases you mention is not a
> > rule of your mental grammar, but rather a 'meta-rule' designed to avoid a potential
> > parsing problem.  The reason I say that is because your requirement is not
> > dependent on the structure of the relative clause per se, but rather on the
> > (accidental) morphological properties of the verb which happens to be contained
> > within the relative clause.
>
> I wouldn't say "accidental morphological properties", it's verbs that
> can take a noun or verb phrase as objects require "that", and verbs that
> may be transitive or intransitive, I'd say that that's a syntactic
> property.  Perhaps that second part could be considered such a
> "meta-rule", but the first part is definitely connected with syntax, and
> not morphology.
Reduced relatives of the type I'm talking about are only possible with transitive verbs
(those which can form passives):

  The poll was conducted by CNN
  The poll [conducted by CNN] was inconclusive

  * The guests were arrived by the airplane
  * The guests [arrived by the airplane] were exhausted

The fact that certain transitive verbs ("break", etc.) can also be used as intransitives
is thus irrelevant in this case:  Since intransitives never appear in reduced relatives
of this type, it shouldn't make any difference whether a given transitive verb can be
used intransitively or not.  The only reason why it would matter is because such a verb
could be misconstrued as intransitive while parsing, leading to problems with
interpretation.

Furthermore, it's not just any old transitive/intransitive verbs which you don't like in
reduced relatives--it's specifically those which lack a separate morphological form for
the past participle.  Again, the reason for this relates to processing:  If a verb has a
separate participial form, there will be no danger of parsing it as a past tense main
verb.


> > Garden path sentences demonstrate that *acceptability*
> > and *grammaticality* are two entirely different things!  A sentence can be
> > unacceptable without being ungrammatical.
>
> Well, maybe, but it seems to be making an artificial distinction, to
> me.  If a structure is unacceptable, and is not produced, how do you
> know that it's NOT simply a grammatical rule?  In other words, if your
> conception of the grammar says that it is grammatical, but it's not
> acceptable, how can you be sure that you haven't misanalyzed the
> grammar, perhaps it really is ungrammatical?  For that matter, why can't
> "meta-rules" as you described earlier be part of the grammar?  What
> justification is there in considering them outside of the grammar, when
> they affect what kind of sentences are produced and what kinds of
> sentences are considered acceptable?
What you call an artificial distinction I call a useful subtlety of the theory.
Linguists who work on processing argue that the parser is a separate component of the
language faculty, which has access to grammatical information but is not itself part of
the grammar.  If a particular construction (here, reduced relatives) is in principal
allowed, but (some or all) speakers reject certain examples of that construction just in
case they're difficult to parse, it would make sense to attribute that fact to a
limitation of the parser rather than a principle of the grammar.  Why build extra rules
into the grammar, when the effects of those rules are entirely duplicated by limitations
of the parser?

It all boils down to the following:  There's nothing per se to keep us from saying that
meta-rules are part of the grammar.  BUT:

(i)  We know that parsing can effect the acceptability of a sentence, as shown by the
centre-embedding examples.
(ii)  As I said before, syntactic rules which act only in the presence of certain
morphological features (which are themselves irrelevant to the construction in question)
are rare at best.
(iii)  Occam's Razor dictates that our model of the grammar should be as simple as
possible.  Therefore, any phenomena which may reasonably be explained by appealing to
extragrammatical factors (such as processing) should not be attributed to the
grammar--especially if doing so would cause us to formulate an unprecedented kind of
rule.

Matt.