attributive and predicate adjectives (summary and proposal)

Fri Aug 21 20:09:50 UTC 2009

Dear Colleagues,

Thanks to everyone who replied to my query a while back about relating
attributive adjectives (the red house) to predicate adjectives (The
house is red).  The reference list below will serve as a summary of
the citations I received; if I left anything out it was an accident,
so please let me know.  What appears below is a brief overview and a
sketch of a proposed analysis.  Comments welcome.

Steve
--------------

Adjectives as Predicates and Attributive Modifiers

Stephen Wechsler
August 21, 2009

Most proposals regarding attributive and predicative adjectives
involve a lexical alternation between an adjective form selecting a
SUBJ and one that does not.  For languages in which they differ
morphologically this is plausible (see e.g. Broadwell 2007 on
Zapotec).  For languages where the forms are identical, a lexical rule
or other lexical mechanism for capturing this putative alternation
might be defensible if it can be shown to feed some other lexical
process such as de-adjectival derivations.  I am not aware of any
evidence of this kind.

In the absence of positive evidence for such a zero-derivation
Adjective-to-Adjective rule, I would propose a lexically unified
‘constructional’ account, in which the same word ‘red’ appears in both
‘the red barn’ and ‘the barn is red’ and the differences fall out from
the differences in the syntactic environments.  Unified accounts fall
into two varieties: (i) the adjective selects a SUBJ in neither
construction; (ii) the adjective select a SUBJ in both constructions.

(i) SUBJ in neither.  Attia (2008) suggests that in e.g. ‘The barn is
red’, [red] has the closed function PREDLINK rather than the open
function XCOMP.  Subject agreement (in languages that have it) is done
constructionally, with equations on the AP node.  But anaphoric
binding patterns suggest that predicate adjectives have subjects:

0. a. Mary wondered how angry at himself/*herself John would be.
  (therefore 'John' is the SUBJ of 'angry')
    b. Mary wondered which pictures of himself/herself John would like.

(ii) SUBJ in both.  Zweigenbaum (1988) analyzes French postnominal AP
modifiers as SUBJ-selecting.  His evidence includes infinitival
complements of the adjective.  Note the M.SG agreement on ‘gentil’,
which follows straightforwardly if ‘capable’ is a subject control
adjective.

1. [un garçon capable d’être   gentil ]NP
     a   boy      capable of’being nice.M.SG

What about prenominal adjectives in a language like English?  Can
these be analyzed as SUBJ-selecting?  Alsina (p.c.) suggested this in
reply to my query; I believe this option may have been discussed in
ParGramm once upon a time.  I pursue this idea here, and invite your
comments.

Coppock (2009, ch. 5) argues in detail that those prenominal
adjectives that can alternatively be used predicatively are
*semantically* predicative-- even when appearing prenominally.  For
example, ‘red’ can be predicative, ‘mere’ cannot:

2. a. a red barn   ~  the barn is red
   b. a mere barn ~  *the barn is mere

Coppock notes that ‘red’ (but not ‘mere’) is a semantic predicate in
the sense of Montague 1974: it is type <e,t>.  Interestingly, there is
evidence for this semantic type even when it is prenominal (Coppock
2009, 168, citing Beesley 1988):  “Adjectives describing semantic
predicates would be predicted to be incompatible with sentence-level
adverbs, because sentence-level adverbs describe the status of
propositions, not predicates. As Beesley shows, this prediction is
borne out ...”

3. The obviously red barn collapsed.
4. *The obviously mere barn collapsed.

This suggest a SUBJ for ‘red’ but not ‘mere’, which would also account
for the contrast in 2 above, without need for a separate predicativity
feature. In LFG this comes out as:

red, Adj
(^PRED) = 'red<(^SUBJ)>'

mere, Adj
(^PRED) = 'mere'

        N’
     /            \
   AP              N’
! e (^ XADJ)       |
(!SUBJ) = ^        barn
    |
   red

        N’
   /          \
AP              N’
! e (^ADJ)      |
  |             barn
  mere

(Zweigenbaum 1988 notes that this ((!SUBJ) = ^) leads to cyclic
f-structures, and this this can be avoided by doing it HPSG-style,
with a HEAD feature.)

Coppock also notes that only the predicative type licenses
one-anaphora (citing Beesley 1988):

5. a. That’s a red box, and that’s a blue one.
   b. *That’s an utter fool, and that’s a fat one.

This semantic observation may be converted to a syntactic argument.
It seems to be the case that English nominal ellipsis is possible only
with (some) semantically predicative Adjectives (the few, the proud,
the uninsured, the good, the bad, the elderly, the rich, etc.):

6. a. Medicaid is for [the (very) poor]NP.
   b. [The poorest]NP live near the tracks.

7. a. *Some fools are welcome at my party, but not the utter.
   b. *The gods make trees, but the mere write poems.
       (“the mere people”)

('very' and -est in 6 shows they are adjectives, not nouns.)  If the
adjective selects a SUBJ, then it can have an optional equation:

poor, Adj
(^PRED) = 'poor<(^SUBJ)>'
{ (^SUBJ PRED) = 'people' }

‘utter’, ‘mere’ lacks a SUBJ so such an equation is not possible there
(it would require an inside-out function application).

A last argument for attributives taking SUBJ is that in languages
where the adjective shows gender/number agreement, that agreement
feature would then uniformly have SUBJ as its agreement trigger.

REFERENCES

Andrews, Avery D. 2009. ‘Grammatical’ vs. ‘Lexical’ Meaning
Constructors for Glue Semantics. Unpublished draft. School of Language
Studies, The Australian National University.

Arnold, Doug, and Louisa Sadler. 1992. Noun-modifying adjectives in
HPSG. Department of Language and Linguistics, University of Essex.

Attia, Mohammed. 2008. A Unified Analysis Of Copula Constructions In
LFG. Ed. M. Butt and T. H. King. Proceedings of LFG08: 89.

Avgustinova, Tania. 2006. A Functional Typology of Copular “Be”:
Towards an HPSG Formalisation. In Proceedings of the HPSG06
Conference, ed. Stefan Müller. CSLI On-line Publications.
http://cslipublications.stanford.edu/HPSG/7/avgustinova.pdf.

Barrier, Nicolas. 2002. Une MétaGrammaire pour les adjectifs du
français. TALN 2002, Nancy.

Beesley, Kenneth R. 1982. Evaluative Adjectives As One-Place
Predicates Dm Montague Grammar. Journal of Semantics 1, no. 3-4:
195-249.

Broadwell, George Aaron. 2007. Lexical sharing and non-projecting
words: The case of Zapotec adjectives. Proceedings of LFG07: 87-106.

Butt, M., T. King, M. E. Niño, and F. Segond. 1999. A Grammar Writer's
Cookbook. Number 95 in CSLI Lecture Notes. Stanford, CA: CSLI
Publications.

Coppock, Elizabeth. 2009. The Logical And Empirical Foundations Of
Baker’s Paradox. PhD Dissertation, Stanford University.

Dalrymple, Mary, Helge Dyvik, and Tracy H. King. 2004. Copular
Complements: Closed or Open? In The LFG 04 Conference, ed. Miriam Butt
and Tracy H. King, 188-198.

Müller, Stefan. 2002. Complex predicates: Verbal complexes,
resultative constructions, and particle verbs in German. CSLI
Publications.

Van Eynde, Frank. 2006. NP-internal agreement and the structure of the
noun phrase. Journal of Linguistics 42, no. 01: 139-186.

Whitman, Neal. 2005. Category neutrality: a type-logical
investigation. Routledge.

Zweigenbaum, Pierre. 1988. Attributive adjectives, adjuncts, and
cyclic f-structures in Lexical-Functional Grammar. Paris: Département
Intelligence Artificielle et Médecine, Paris 6.