The Big Mess Construction: interactions between lexicon and
Kim Jong-Bok and Peter Sells
Kyung Hee University, South Korea; University of London, UK
email@example.com or firstname.lastname@example.org; email@example.com
The so-called Big Mess Construction (BMC) exemplified in corpus examples like It was [so]
prominent a punctuation (*so a prominent punctuation) in the landscape is peculiar in that it
has a predeterminer adjective followed by an NP with the indefinite article a/an. The BMC
construction can be introduced only by a limited set of degree words like so, as, too, how,
this, that, and so forth, as seen from a somewhat underdeveloped country/*somewhat
underdeveloped a country or a very big house/*very big a house. Van Eynde (2007) and Kay
and Sag (2008) attribute such idiosyncratic properties of the BMC to the existence of a special
construction specified with constraints like the following:
SELECT NP[MARKING a]
This indicates that English employs the ‘adpd-cx’ (‘adjectival pre-determiner construction’)
that forms the combination of a degree expression with an AP like so prominent, which in turn
‘selects’ an NP marked with an indefinite article like a. This then basically allows so
prominent to be combined with a punctuation. Such a construction-based analysis basically
claims that the grammatical properties of the BMC depend on constructional constraints like
(1). However, considering the behavior of similar degree expressions like (2) make us wonder
if the combinatorial properties of the BMC really hinge only on the constructional constraints:
(2) a. Why have I got [such] a terrible collection of letters? vs “[What] a fussy little man he
is”, said Mary.
b. That closeness makes a [more] interesting choreography. vs That closeness makes
[more] interesting an choreography.
Unlike so, expressions like such or what disallow a predeterminer adjective but combine with
an indefinite NP directly. Meanwhile, more or less allows both ordering. In addition, as in I
have [so] much to thank you for or Tell me [what] options are available, the degree
expressions so or what can combine with a bare NP. To capture such different behaviors of the
degree expressions such or more, a purely construction-based approach are forced to
introduce other constructions different from the one in (1). We can observe what matters in
such discontinuous constructions is the lexical properties of each degree expression, rather
than constructional constraints. That is, so, such, and more are different in what they can
combine with (eg, so big (of) a mess vs such (*of) a big mess): it is hard to assume their
ordering restriction hinges on constructional constraints.
Our approach follows Van Eynde (2007) and Kay & Sag (2008) in that degree words are
functors selecting their head, but differs from them in that degree expressions can select more
than one dependent: a modifier phrase, a nominal, and even an extraposed CP clause as
represented in the following:
(3) FORM <so>
SYN SELECT <AP, (NP[MARKING a])>
SYN SELECT <(NP[MARKING a])>
For example, the lexical information for so specifies that it can select a modifier AP and an
optional indefinite NP marked with a/an, in addition to the extraposed CP. The present
analysis thus attributes the properties of the discontinuous modifier to the interaction of head-
functor phrase and lexical properties of the degree expressions. This way of lexical treatment
will bring us several welcoming results. For example, the EXTRA(POSITION) value
lexically selected by the degree adverb can easily explain why some degree words can
introduce another dependent as in It was [so] prominent a punctuation in the landscape [that
one was positively drawn towards it]. This dependency is also triggered by the degree word so
not from a construction. The analysis, without introducing constructions similar to (1) for so
or more, can also predict variations in the uses of so as in The room was [so] dark or in No
life is [so] hard [that you can’t make it easier]. We can easily extend the present head-functor
and lexicalist-based analysis to explain other discontinuous examples like [too] slowly [to get
there by noon] or a [faster] response to the issue [than was expected]: too and the -er
comparative select not only one but two elements: the immediately following head slowly or
response and the dependent VP[inf] and the than phrase.
It is true that every language employs a certain number of constructions that can be
hardly predicted from general grammar rules. However, we need to be careful for the
proliferation of ‘constructions’ since there exist many constructions that come from the lexical
properties and further general constructions, rather than from specific ones.
Kay P. & I. Sag (2008). Not as hard a problem to solve as you might have thought. Paper
given at Construction Grammar Conference. University of Texas, Austin.
Van Eynde, F. (2007). The big mess construction. In Proceedings of the HPSG07 Conference.
CSLI On-line Publications.