Problems of Lexical Integrity in Monostratal Theories

[Joan Bresnan]

I would like to comment on Dick Hudson's response to Dan Everett's
query about Lexical Integrity:

>>>Dick Hudson said:
 > Dan Everett:
 > At 18:52 31/03/98 -0500, you wrote:
 > >I wonder if there are more recent works 'defending' the Lexical Integrity 
 > >Hypothesis than the Bresnan & Mchombo piece in NLLT 95?
 > 
 > dh: Isn't the LIH the same as the principle of morphology-free syntax
 > (sister of phonology-free syntax) that Zwicky and Pullum have been
 > advocating for many years? E.g. Zwicky 1992, Some choices in the theory of
 > morphology. In R. Levine (ed.) Formal Grammar: Theory and implementation,
 > OUP, 327-71, especially 354. Indeed, isn't it inherent in any monostratal
 > analysis which generates surface structures directly? 
 > 
 > >
 > 
 > ============================================================================

The issue of lexical integrity is not quite this simple.  There are
several issues that cause trouble for monostratal theories:

1. analytic vs. synthetic sequences:

First, even assuming that some sequence of surface elements--say N V
or X NP--is directly generated, how can you tell whether it is
generated as a word or as a phrase?  In many languages there are cases
where complements may or may not be phrasal.  So the question of
criteria for distinguishing wordhood from phrase-hood is relevant even
for monostratal theories.  

2. explaining why and to which processes synthetic sequences are
barriers:

Second, different monostratal theories have many different mechanisms
for expressing grammatical dependencies such as agreement, extraction,
coordination, etc., and the question arises whether the morphological
word is an absolute barrier to all or just some of these processes,
and if so, why or which (and why which).  LFG from its inception was
lucky to have researchers working on typologically diverse languages
involved in its development (e.g. Simpson, Mohanan, Andrews, Neidle,
Zaenen, Maling), and so from the beginning it allowed words to
construct functional structures on a par with phrases.  This
flexibility is not present in all current monostratal theories, and it
leads to a *different* conception of lexical integrity than is found
elsewhere:

words are encapsulated from phrases in their *structural*
formation, but may interact with phrases in specifying functional
structures (complex feature structures).

This formulation, stated in Bresnan and Mchombo 1995 and elsewhere,
originates in Jane Simpson's 1983 doctoral dissertation.  Quite a bit
of recent work has provided interesting exemplification and support
for this "split" conception of lexical integrity (see, for example,
work by Jane Simpson, K.P. Mohanan, Avery Andrews, Farrell Ackerman,
Paul Kroeger, Tara Mohanan, Yo Matsumoto building on Akira Ishikawa
and Mariko Saiki, Miriam Butt, Tracy King, Hye-Won Choi, Peter Sells,
Sells and Young-Mee Cho, Alex Alsina, Sam Mchombo, Ackerman and
Webelhuth, Rachel Nordlinger, Louisa Sadler, Ida Toivonen building on
Jonni Kanerva, Kersti Bo"rjars, Nigel Vincent, Julia Barron, Anette
Frank and others I shouldn't forget).  See Alsina, Bresnan, and Sells
(eds.) 1997 Complex Predicates (CSLI Publications) for some articles
and references from different theoretical perspectives.

3. The problem of mixed categories:

There is still another serious problem for monostratal lexical
integrity, however, even for LFG's flexible "split" formulation: mixed
categories.  Here you have what appear to be phrasal structures of
different categories sharing a single morphologically complex head.  I
briefly review the syntactic properties of these in my LFG97 paper
(see links on my web site), and show that *except for violating
lexical integrity* a head-movement analysis provides the best overall
explanation of the general properties of these structures.  I sketch
how LFG can capture the generalizations by means of "head-sharing" in
f-structure, preserving lexical integrity.  My work here owes much to
John Mugane's 1996 University of Arizona Ph.D. thesis _Bantu
Nominalization Structures_.  Others are working actively in this area;
see my paper for some references and leads (incomplete still).  Bob
Borsley has a forthcoming book of articles on this topic (which will
include a revision of my paper), again from various theoretical
perspectives: Robert Borsley (ed). (forthcoming) Syntactic Categories,
Academic Press.