Interim summary 2 for asymmetries

bingfu bingfu at SCF-FS.USC.EDU
Tue Mar 10 05:50:07 UTC 1998

Dear netters,
	So far, there are following 12 netters who have contacted me on
the issue. I am still in the procedure of corss-checking, organizing and
analyzing the data they provided.  Some correspondents and me have had
emails forth and back several rounds.  Thanks all of them!!

Jon Aske,
Matthew S Dryer,
Gisbert Fanselow,
Sungshim Hong,
Richard Kayne,
Natalia Kondrashova,
Elena Maslova,
Víctor Vázquez Martínez,
Ma. Teresa Macias Rabago,
Petr Roesel  roesel,
Takaku Tsunoda  tsunoda,
Theo Vennemann,

Here is another short interim summary and further specification.

Concerning my previous three questions.

Is there any recent discussion on the topic in the literature.
I already know Tomlin's book. "Basic Word Order", I need to know
something newer.  Do you know any other left-right asymmetry of
word order variation?

The only relevant reply is the one provided by Tsunoda.

Are there any languages that behave like English, which
rigorously permits only V DO Ad but not V Ad DO, unless
clearly motivated by processing ease (in heavy NP shift)?

No clear reply.  But several netters pointed that [V DO Ad] is less marked
than  [V Adv DO] in their native languages, including Russian.

In your native languages, what is the basic order
for V, DO, Ad?

An interesting reply is that Maslova points out in Russian, the basic
order is [Ad V DO]. In fact, this is exactly true of Chinese.

The data is quite complicated, but some tendencies seem to exist.  For the
"He walked slowly with a stick in the garden for three hours yesterday",
it seems that the order
[--three times ---yesterday----walked----]
has never shown up yet?   If any netter find this order, particularly in
basic order, please let know.

	Jon Aske raises the isse of explanation.  I do have some rough
tentative explanation in mind.  Simply speaking, these left-right
asymmetries are determined mainly by the interplay between the SEMANTIC
RELEVANCE (more semantically relevant and inhernt dependents tend to close
to the head) and REFERENTIAL PRECEDENCE (more identifiable and referential
units tend to precede less identifiable and referential units.)
	Specifically, if the two tendencies are harmonious to each other,
then the order is expected to be rather consistent and stable, or to be
tested frequently, probably becoming the basic order.  On the other hand,
if the two tendencies are contradictory, the order will inconsistent and
unstable, and hard to predict, unlikely to be grammaticalized
into a basic order. The underlying reason can
be analogized to the famous staring sentence in Anna Karenina by Leo
Tolstoy,  "all happy families are like one another; each unhappy family is
unhappy in its own way."  Most of the left-right asymmetries I mentioned
seem to be accountable for, at least partly, by the interplay of the two
	For example, give S is normally more referential than O, and O is
more relevant to the subcategorization of verb, OSV is not likely to exist
because it is motivated neither by seantic relevance nor referential
precedence. On the other hand, the frequent [S[OV]] and [S[VO]] are
motivated by both tendencies.
	In the orders of V, O and Adv, semantic tendency requires for O to
be adjacent to V, i.e. to fllow Adv, but referential tendency requires for
Adv to follow O, given nouns are normally more referential than adverbs.
Thus, in [V O Adv], both two requirements are met, thus, the order is
dominant.  In contrast, when both O and Adv precede V, the two
cannot be met simultaneously, thus, indetermination occurs.  When O and
Adv are on the two side of V, the two orders don't differ with respect to
semantic relevance.  It remains a mysery why [Adv V O] seems more common
than [O V Adv].  There must some be other factors that interfere with the
referential ordering tendency.  Is there is a possible branching tendency
such that if a head has two dependents on each side, the preferred
branching is [X [H Y]] (H is a head)??  Or a tendency such that postverbal
position is more appropriate than postverbal position for focusing??

Concerning methodological discussion: How much volume of data is required
for substantial generalizaion?
	My answer is medial, neither too much nor too less.  When the
great Russian chemist Mendeleev discover the periodic system of elements,
there were only about 50 known elements, which are, however, enough to
display the significant pattern rather clearly, for a ready observer.
Chemists don't need to wait until all elements discovered.  In fact,  the
great value of Mendeleev's periodic table lies in its 'vacant' cells,
which guided chemists to discover new elements.
	 Of course, too few data cannot show real underlying patterns.
But it does not mean that the more the data are, the better for
discovering the pattern.  Human's span of atention and ability to process
information are very limited.  Too much volume of complicated data would
obscure rather than clarify the pattern sometimes.
	For discovering language universals, Generalists prefer to study
few lnguages but in depth.  Topologists emphasize on wide range of
languages. The right answer may be in the middle ground.  Some tense
between the two methods is healthy.  Like to dig holes, if the opening is
too wide, we cannot effectively dig into depth.  On the other hand, if the
opening is too narrow, we cannot effectively dig deep either.  For manual
digging, perhaps about 45 degrees slope is appropriate.  Of course,
the help of machines, the situation is different.  Likewise, with the help
of computers and mathematics method, we now can handle much larger volume
of data.

						Bingfu Lu

More information about the Lingtyp mailing list