Burquest's query/ies

Sun Apr 9 21:27:56 UTC 1995

This is in response to the questions posted by Don Burquest of 
UTexArlington last Thursday, 6/4/95.   

******

Dear Dr. Burquest,

Sorry it's taken me so long to get in touch with you in response to your 
questions, but (1) as Portia said to Figaro, 'urge l'opra, lo sai' and 
(2) the e-mailing system here was been acting peculiar the past couple of 
days.

> Most ... are from a GB background (me too) ... It may be that my GB 
> viewpoint is encumbering me with presuppositions that are hindering my 
> understanding.

I presume you are acquainted with Peter Sells' fine little book on compa- 
rative syntactic theory (my favourite subject!).  Allow me to suggest 
that you also take a look at my 1990 UIUC dissertation (avail. from Uni- 
versity Microfilms), Free Word-Order Syntax: the Challenge from Vedic 
Sanskrit to Contemporary Formal Syntactic Theory.  While the substantive 
issues wrestled with in that work may be tangential to your concerns, it 
does represent an explicit comparison between GB and LFG and might there- 
fore prove of interest.

> Here's how I think f-structure for a given sentence is developed, 
> please correct me if I'm wrong.  A typical verb might be characterized in 
> f-structure along the lines of the following:
>
> (1)	'HAND <(^SUBJ)(^OBJ)(^OBJ2)>'
>
> It seems, however, that there is a representation more basic, namely one 
> in which the entry seems to correspond to the theta grids of GB, along 
> the lines of the following:
>
> (2)	hand (agent) (goal) (theme)
>
> It is principles of grammatical function assignment that then associated 
> these (universal) thematic roles with the (universal) grammatical 
> functions.  I am thinking that because this mapping is well defined, it 
> is rarely referred to, and the representations using grammatical 
> functions is used instead as a sort of shorthand.  Is this correct?

Pretty much.  What is represented in (2) (my numbering) is the verb's 
lexical *semantic structure*, while (1) represents the verb's lexical 
*functional structure*.  The semantic (or 'sigma-') structure in (2) is 
common to the verbs in both the sentences in (3):

(3)	a. Evan handed a toy to the baby.
	b. Evan handed the baby a toy.

The two sentences differ in the mapping of sigma-structure into 
f-structure.  The lexicon of standard English allows 'hand' two options 
in this respect.  One is the one in (1), the other is the one in (4):

(4)	'HAND <(^SUBJ)(^OBJ)(^OBJdat)>'

Beyond that, the way given arguments in the sigma-structure match up with 
given GFs in the f-structure is governed by general principles, as you 
suggest.

> Is it possible to take the position that it is the verb's lexical entry 
> that constrains the remainder of the f-structure of a given sentence?

In general, yes; that's what's meant by the 'PRED' function and the 
(^ = v) notation.  The one caveat i would venture has to do with 
adverbials.  (5) is a perfectly good sentence of English, but 'in the 
playground' doesn't satisfy any subcategorization requirements of 'hand' 
and can't be merged with its lexical semantics in any meaningful way.  
The compositional semantics of LFG allow for it to be *added* to the 
lexical semantics of 'hand', as long as there isn't any semantic conflict 
involved (as there would be, say, if two distinct places were involved, 
or if the lexical semantics of the verb were somehow inconsistent with a 
place-adverbial).  The f-structure of (5) would include an attribute not 
specified in the lexical entry of 'hand', but not inconsistent with it.

(5)	Evan handed the baby a toy in the playground.

> Is it required that SUBJ, OBJ, OBJ2 be NPs?  Where is that information 
> found?

It is not required by the framework as a whole.  It is logically possible 
that the grammars of particular languages may make such a stipulation, 
possibly by means of a global constraint in the lexicon to the effect 
that a certain GF can only be borne by a certain category type.  In 
English, as it happens, many of the adverbials alluded to above happen to 
be encoded in PPs, meaning that the c- and f-structures of sentences line 
(5) have prepositions that have their own subcategorization requirements, 
which the NPs such as 'the playground' satisfy.  In many other languages, 
however, such adverbials are often expressible as bare NPs (with oblique 
case marking, or some such thing).  All this would have to be worked out 
in the details of the syntax/semantics interface, specifically in the 
relation between sentential f- and sigma-structures.  But the important 
thing is that these details would be *language-particular*.

> Do NPs and other constructions have their own c-structure rules?

Sure; and if nobody else has offered you any references in the literature i 
can try to scrounge some up for you.  In the LFG part of my dissertation i 
assumed something like a GPSG-style PS component, where that didn't con- 
flict with the basic axioms and approach of LFG.  But in any case you've 
got to be able to fit determiners and adjectives in, to say nothing of 
embedded relative clauses, etc.

> If I am correct, the development of the f-structure for a given 
> sentence may be characterized as a progression along the lines of the 
> following:
>
>	(a) predicate and its thematic roles selected
>	(b) thematic roles associated with GFs on universal grounds
>	(c) GFs established within f-structure and 'filled in' appropriately

Yes, though there are other equally valid approaches; at least as i 
practice it, LFG is a 'declarative', not a 'procedural' framework.  Like 
other unification-based frameworks, it places constraints on static 
representations but not on how one derives them.

> I don't know what to do; perhaps you have a designated guide?!

Mary Dalrymple, who manages the LFG-list, has a packet of recent papers 
(including an interesting one addressing the theoretical complications of 
the semantic integration of adverbials) on how LFG works which i'm sure she 
would be happy to send you (i myself just got the packet today -- thanks, 
Mary!).

For myself, i would be more than happy to field any questions.  I don't 
promise to always have answers, and some of the answers i have may be 
rather idiosyncratic -- i've been 'doing' LFG analyses for years, pretty 
much on my own.  But wrestling with someone else's questions would be 
great practice towards the day (may it come soon!) when i get the chance 
to actually teach this stuff in a class somewhere.  So by all means get 
in touch with me.

Best,
Steven Schaufele
---------------------
Dr. Steven Schaufele
712 West Washington
Urbana, IL  61801
217-344-8240
fcosws at prairienet.org

**** O syntagmata linguarum liberemini humanarum! ***
*** Nihil vestris privari nisi obicibus potestis! ***