What’s the difference between a nuclear physicist and a heavy drinker?
Resolving the bracketing paradox
Zoë Belk
Linguistics (Supervisor: Ad Neeleman)
Division of Psychology and Language Sciences, UCL
[email protected]
Introduction: What is a bracketing paradox?
A new variety of bracketing paradox
Bracketing paradoxes occur when the meaning of a word or phrase
doesn’t correspond to how it’s pronounced. Well-known examples include
nuclear physicist and transformational grammarian.
The meaning of the first phrase is “someone who does nuclear
physics”, not “a physicist who is nuclear” – nuclear and physic(s) form a
semantic unit, represented as [[nuclear physic] ist].4
However, for phonological reasons, the suffix –ist must attach only
to physic(s) and not to nuclear physics. The appropriate phonological
structure is [nuclear [physic –ist]].4
è There is evidence for both structures, so how can we resolve
this paradox?
Examples like hard worker and heavy drinker, which are derived
from verbs, have been argued to demonstrate a special relationship
between adjective and noun.1,2
However, I argue that these examples are bracketing paradoxes,
although different from traditional examples.
•  One phonological form, but two meanings à at least one mismatch
•  Both meanings are semantically compositional, pace Larson, Cinque
•  There’s evidence for both bracketings
How different is a nuclear physicist from a heavy drinker? Can we use
the same solution for both types of paradox?
A non-precedence approach to LF
rebracketing
A non-precedence approach to LF
rebracketing
Analysing traditional
bracketing paradoxes
N
AfN
N
•  Mismatch between syntax and PF
A
•  Mapping Principle to associate the different structures3
nuclear
N
N
Syntax & LF
N
A
N
nuclear
physic
AfN
A
ist
nuclear
ist
(14) Syntax
a.PF
(14)&
A non-precedence
rebracke
physic
PF
N
physic
N
A
AfN
•  No! The affix must attach to the head of the word/phrase in the
syntax
à The only possible syntactic structure is below left
à The mismatch is between syntax and LF
•  I propose that the adjective can undergo very local movement (that
doesn’t leave a trace), as long as the head doesn’t change and
c-command relations between non-heads remain intact
Zoë Belk
UCL
PhD
Day
2014
6
Zoë
Belk
UCL
PhD
Day
2014
•  Adhering to these conditions, which I call Information 6
Preservation, ensures that only the change below can occur:
ist
N
N
Analysing verbal bracketing paradoxes
N
hard
N
AfN
hard
V
work
In Dutch, prenominal modifiers appear with a declensional schwa roughly
physic
ist
in the following
circumstances:
•  Prenominal adjectives must be conjugated with a declensional
schwa when they are part of a definite DP. In the absence of a
determiner, they must not be conjugated
b.
N
A N
A
A primer on the Dutch declensional schwa
nuclear
a.N
N
VAf
work
N
er
N
b.
V
AfN
er
A
hard
N
N
V
AfN AfN
V er erAfN
AN V
hard
LF
work
work
ist
A
N while LF is a
In this case, the PF and syntactic structures
are isomorphic,
In this case,
the PF and syntactic structures are isomorphic, while LF is a
mismatch.
mismatch.
nuclear
physic
analysis requires
that adjectives
and adverbs
are underlyingly the
TheThis
structure
of Dutch
bracketing
paradoxes
This analysis
requires
adjectives
and adverbs
areverbal
underlyingly
same category,
but that
explains
several characteristics
bracketingthe
paraN of
same category,
doxes. but explains several characteristics of verbal bracketing paraVerbal
BPs: Given a string like (15-a),
Traditional
BPs: between noun and AP
doxes.
Adjacency
is explained.
Schwa:
No schwa:
N(16-a)
the only
possible
rebracketing
be A
(15-b).
Similarly
(16-b).
Adjacency
between
noun
and AP would
is explained.
Given
a string
likeand
(15-a),
S y n t a Similarly
c t i c a l l y, (16-a)
t h e and
m o d(16-b).
ifier
Syntactically,
the modifierwould
does be (15-b).
the only
possible rebracketing
de beroemd*(-e) gitarist
Hij speelt klassiek(*-e) gitaar.
(15)
a. schwa
[bald because
[heavy [drinker]]]
=)
nuclear a schwa,
requires
as it is in the
not
require
a
no
the famous(DECL) guitarist
he plays classical(DECL) guitarZoë Belk
N
Af
UCL
PhD
Daydrink]er]
2014
8N
b.
[bald
[[heavy
(15) determiner
a. [bald
[heavy [drinker]]]
=) same configuration as normal N
is present
in the N+A
(definite)
(indefinite)
constituents:
a. [[heavy
[heavydrink]er]
[bald [drinker]]]+A
=)
constituent:
b.(16)[bald
physic
ist
1
b. [heavy [bald drink] er]
But bracketing paradoxes show unexpected (and contrasting) behaviour:
(16)
a. [heavy [bald [drinker]]] =)
(22)
Traditional BPs:
Verbal BPs:
de klassiek(*-e) gitarist
the classical(DECL) guitarist
(definite)
de hard*(-e) werker
the hard(DECL) worker
(definite)
1
a.
N
b.
NN
b.This[heavy
[bald
er]
explains
whydrink]
the paradoxical
reading is only available with (15): in
Syntax
Syntax so the
(16), bald could be rebracketed, but to drink baldly is meaningless,
AAonly available
N N with (15): in
N
Af
This explains
why
the
paradoxical
reading
is
N
rebracketed reading is unavailable.
(16), bald could be rebracketed, but to drink baldly is meaningless, so the
harde V N AfNAf
ist
beroemde
rebracketed
reading
is
unavailable.
N
A
N
4
Some predictions of LF rebracketing
werkgitar er ist
Verbal bracketing paradoxes look just like other N+A combinations,
klassiek
gitar
4.1 predictions
Dutch data
4 Some
of LF rebracketing
but traditional bracketing paradoxes disallow a schwa where we
Verbal bracketing
paradoxes
requiremodifiers
the schwa:
would otherwise expect it.
In
Dutch,
prenominal
appear
with a declensional
schwa in certain
The differences in the behaviour of the
declensional
schwa in 1traditional
4.1 Dutch
data
contexts.
A partial
approximation
of thiswhat
rule we
is aswould
follows:expect from
and verbal bracketing
paradoxes
is exactly
(23)
Conclusions: So what?
a.
de elegant*(-e)
danser
the two
analyses
sketched
above,
and
cannot
be explained
by treating
In Dutch,
prenominal
modifiers
appear
with
a
declensional
schwa
in certain
(17)
Prenominal
adjectives
must
be
conjugated
with
a
declensional
schwa
the elegant(-decl) dancer
1
the two
of
paradox
uniformly.
contexts.
A types
partial
approximation
of
this
rule
is
as
follows:
when
they
are
part
of
a
definite
DP.
In
the
absence
of
a
determiner,
b. de hard*(-e)
werker
they must not be conjugated.
worker must be conjugated with a declensional schwa
(17)the hard(-decl)
Prenominal adjectives
The behaviour of the Dutch declensional schwa highlights the
Prenominal
modifiers
must
appear
a declensional
in a definite
when
they are part
of aReferences
definite
DP.with
In the
absence of schwa
a determiner,
different syntactic structures of traditional and verbal bracketing Again, this is predicted
rebracketing analysis. Here, the affix
DP: must by
they
notthe
be LF
conjugated.
paradoxes. The analyses presented here of the two types of paradox
is attached to
the verb in the syntax, rather than the phrase as in traditional
1Cinque, G. (2010). The Syntax of Adjectives: A comparative study. Cambridge, MA: MIT
predict exactly this difference.
1
Prenominal
modifiers
must
appear
with
declensional
schwa
in
a
definite
This
is athe
verymodifier
much
simplified
version
ofa the
rule,
whichprenominal
may
additionally
be underbracketing paradoxes,
and
behaves
like
any
other
Press.
1
Bracketing paradoxes are at the intersection of sound, meaning
going
language
change.
This
version
is
detailed
enough
for
our
purposes.
2Larson, R. (1995). Olga is a beautiful dancer. M.s.
DP:
modifier in a3 definite DP. The appearance of the schwa shows that the approand syntactic structure. They provide insight into how these different
Sproat, R. (1988). Bracketing paradoxes, cliticization and other topics: The mapping between
priate PF bracketing
is
the
same
as
any
other
prenominal
adjective
+
noun
aspects of language interact. By looking at these interactions, we can
syntactic
and
phonological
structure.
In
M.
Everaert,
M.
Trommelen,
and
R.
Huybregt
(Eds.),
1
This
is
a
very
much
simplified
version
of
the
rule,
which
may
additionally
be
underMorphology
and Modularity,
pp. 339–360.
Berlin:
De Gruyter
Mouton.
The
schwa
is
in
fact
predicted
under
the
analysis
presented
begin to understand the principles of language acquisition, as well combination.
as
going
language
change.
This
version‘lexically
is detailed
enough
for our
4Williams,
E. (1981).
“On
the notions
related’
and ‘head
of apurposes.
word’.” LI. 12(2): 245–274
what happens when language breaks down.
here. This behaviour is the opposite of traditional bracketing paradoxes,
showing that the two cannot be of the same kind.
(24)
N