WHAT IS A SYNTACTICON?
Joseph
E. Emonds, October 2002
Kobe-Shoin
Women’s University
I.
INTRODUCTION: HUNDREDS OF PARAMETERS?
This precis presents
some of the ideas developed and defended mainly in Chapters 1, 3 and 4 of my
book Lexicon and Grammar: the English
Syntacticon. Chapter and section numbers standing alone below refer to that
work.
The
idea of a Syntacticon naturally extends a line of research initiated by Borer
(1984) and Manzini and Wexler (1987) and fruitfully pursued in Ouhalla (1991).
...Borer's approach to parametric variation...associates parameters with individual
lexical items, as part of the information included in their lexical entries, rather
than with the principles of UG [Universal Grammar]. ...the nature of the
lexical information which determines parametric variation [is] nothing other
than the usual type of information
relating to selection and grammatical features,... it is not information
which is available over and above the familiar type of lexical properties;
rather, these properties themselves determine parametric variation.
One can hypothesise...that possibly functional
categories and substantives belong to two separate modules of the mind/ brain.
...there should in principle be a distinction between two notions of the
lexicon, a grammatical lexicon which
contains functional categories and which belongs to the domain of UG, in
the sense that its categories are determined by UG, and a mental lexicon which contains substantives and which exists
independently of UG, that is an autonomous module of the mind/brain (the
conceptual system). (Ouhalla, 1991, 7-10)
(4.3) Language-particular syntax resides entirely
in lexical specifications, namely inherent and contextual feature combinations
associated with closed class items.[1]
Observation. In thirty years of syntactic research
on English and French, certainly close syntactic relatives (head-initial, no
pro-drop, subjects not freely post-verbal, similar WH-movement, overt articles,
subject agreement, etc.), I find no grammatical morphemes which are exact
translations, i.e. English and French differ by perhaps 300 independent
parameters.
We can thus answer Lasnik's (1991)
“troubling new question,” “why do children acquire languages so slowly?”
Because, like linguists who model language formally, children must explicitly learn formal specifications for hundreds of grammatical items for each
language.
Children learn their grammatical lexicon
in say 3 years. But generative grammar has not developed anywhere near
satisfying analyses of more than a few grammatical items even for much studied
English, nor a contentful restricted theory of a formal lexicon. So we might
turn the question around, “why do generative grammarians analyze lexical items
so rarely?”
This pervasive linguistic fact of
children learning hundreds of
language-particular closed class items seems not to have attracted the
interest—hardly even the notice—of generativists. This study tries to specify
exactly what children learn when they acquire language-particular syntax.
II. PURELY SEMANTIC VS. COGNITIVE
SYNTACTIC FEATURES
(1) Purely semantic features. Let L = the
four lexical categories N, V, A and P. Only
these categories contain items differentiated by purely semantic features f, defined
as those features with no role in derivational computation.
There are many of these f. Hence
only these grammatical categories are called open classes.
I call this store of (only) four classes
of items the Dictionary. This is
Ouhalla's mental lexicon. This study does not address issues of formalizing and
acquiring the Dictionary.
(1.5) a. Canonical
position of syntactic features. UG associates a very few cognitive
syntactic features F with each syntactic category B. These B are the canonical positions of the F.
b.
Canonical realization of features.
Features and categories contribute to Logical Form (LF) only in canonical
positions.
Thus for LF interpretation, (i) F must be
on some matching B, and (ii) any head B0 must be the head of a BP.
Examples
of probable UG matches of cognitive syntactic features with categories: Syntactic
features F: Matching categories B:
Tense and modal features I
STATIVE vs. ACTIVITY V
PERFECTIVE (aspect) V
Quantifier features; DEF D
or NUM
LOCATION and PATH features P
ANIMATE, COUNT, ABSTRACT N
+INHERENT
(“individual vs. stage level”) A
Since
the features F are relatively few, they can distinguish relatively few distinct
morphemes.
I
call this smaller store of items the Syntacticon.
This is Ouhalla's grammatical lexicon.
(1.7) Closed classes. A closed grammatical
class X (including L = N, V, A, P) is one whose members have no purely semantic
features f, but only cognitive
syntactic features F. That is, the
Syntacticon is composed of all and only
the closed classes of items.
Corollary:
Unique syntactic behavior.
Since each Syntacticon item is a unique combination of syntactic features in
its language, each item typically has unique syntactic behavior.[2]
(1.8)
Categorial
Uniformity. The bar notation categories Xj do not vary across
languages, but the subcategories (features) Fi in each language's Syntacticon can.
Canonical
matching of categories with features is often but not always unique:
(2) Certain features may cross-classify the
major syntactic categories. Presumably, a [+N] nominal feature subsumes N, D
and A, while [-N] (-nominal) subsumes V, I and P.
(3) The feature +WH occurs on high
functional categories in projections of the [+N] categories D and A: which book, what day, how often. Probably
WH occurs on N and A as well: You bought
a big what? He seems very what to you?
(4)
The deictic feature +PROXIMATE
is similar to +WH: { this / that }
+ { bread / tall }.
(5) COMPARATIVE features occur in modifiers
of several categories: more interesting,
more into Zen, more of a man (Bresnan, 1973).
Although
many features F have unique canonical
positions, (3)-(5) show that Universal Grammar provides some F with more than
one possible host. Thus, the English Syntacticon matches a feature POTENTIAL
with I and A (can, able), but the
French one with V (pouv-).
The possibility of semantic features f in addition to syntactic F makes
lexical categories “open.” The question arises as to whether every member of N,
V, A and P must have f distinct from F. In fact, nothing
prevents leaving (2) as a one-way implication: N, V, A and P need not have f. Indeed, certain subclasses of N, V, A
and P behave empirically like non-lexical classes, exhibiting post s-structure
insertion contexts and unique syntactic behavior (cf. Emonds, 1985, Ch. 4, and
1987). The lexical categories are thus
like other categories: each has a subset of say up to twenty or so elements
fully characterized by cognitive syntactic features F and lacking purely
semantic features f.
Cognitive syntactic features and purely
semantic ones share some properties, namely (a)-(c) in the table below. But
once we find a defining difference
(4.75d) for the two types, many other clear-cut differences emerge. For
justifications of (f) through (k), see Ch. 3.
What follows here sketches the
syntactically important differences (l) and (m). Cf. also Chs. 4-7.
(4.75) DICTIONARY SYNTACTICON
a.
Items with both cognitive and purely syntactic features F:
yes yes
b.
Cognitive features F canonically realized on UG hosts: yes yes
c.
Insertion possible at the beginning of a derivation: yes yes
d. Items with purely
semantic features ƒ:
YES NO
e. Grammatical categories
in the inventory: N, V, A, P ALL
f.
Open classes; coining and neologisms for adult speakers: YES NO
g.
Bound morphemes have inherent stress and head compounds: YES NO
h.
Interface with non-linguistic memory and culture: YES
NO
i.
Full suppletion inside paradigms (Emonds, 1985, Ch. 4): NO YES
j.
Certain phonetically zero morphemes (4.11): NO YES
k.
Items conform phonologically to core vocabulary (Hannahs, 1995):NO YES
l. Items with alternatively
realized features: NO YES
m. Insertion also possible
during syntax and at PF: NO YES
My
reading on aphasia has been limited, but it seems to me to point to a very
restrictive claim:
(6) Broca’s
area relates to processing only indirectly; it mainly houses the Syntacticon.
III.
SYNTACTICON MEMBERS OF CLOSED CLASSES OF N, V, A AND P (4.75e)
The
conception of the Syntacticon takes issue with the standard a priori division
of “functional” vs. “lexical” categories, perfunctorily inherited from
structuralist and distributionalist grammars.
Put
bluntly, the most central Syntacticon or “functional” categories are in fact N,
V, A and P. Members of the closed subsets of open categories L0 can
be called "grammatical" L0.
These
grammatical L0 differ from each other by cognitive syntactic
features and syntactic subcategorization frames (Emonds, 2000, Ch. 2), but they
have no purely semantic features f.
English
grammatical verbs include be, have, do, get, go, come, let, make, say,
and probably further elements like put,
bring, take and want.
Its
grammatical nouns include one, self, thing, stuff, people, other(s),
place, time, way, reason.
The
vague distinction between grammatical and lexical or “contentful” prepositions
now becomes a predicted sub-case of a more general lexical category theory.
While grammatical P have no f, some lexical P (curiously akin to
compounds) are specified with purely semantic f: downstairs, nearby, overhead, outside, aboard, ashore, aside,
askance, sideways, etc.[3]
Though
I haven’t investigated this area, grammatical
A doubtless include many, few, good,
bad, well, very, such, so (he seems/
remains/ sounds [A so ]) and low numerals, which are declined as
As in Czech and Russian (Veselovská, 2001).
All
(and only) grammatical verbs, nouns, adjectives and prepositions have unique
syntactic behavior—none acts syntactically like any other. This follows from
being in the Syntacticon.[4]
Examples
of grammatical English L0 with syntactic features F but no purely
semantic features f:
(7) (i) the
activity verb [V do ]
(ii) the
stative verb [V be ]
(iii) the basic PATH preposition [P to ]
(iv) the
basic non-locational preposition [P of ]
(v) a
derivational suffix [N -ing ]
in derived nominals formed from V stems (cf. Ch. 4)
(vi)
a derivational suffix [A -ed ] in passive adjectives formed from
V stems (cf. Ch. 5)
IV.
INSERTING A SYNTACTICON ITEM AT 3 LEVELS (4.75m)
(2.18) Extended
Classical Subcategorization. A
lexical frame @, X , +___Y is
satisfied if and only if Y is a cognitive syntactic feature of a lexical head
of a complement in XP.[5]
(4.15) Syntacticon
entry for do/ do so: do, V, { +__ [DP N, -ƒ ]/ +___ so } ( +__ [PP to^ANIM])
(8) Mary
did { something mean, nothing else, this, more than expected, all that was
needed, some more stuff, the thing you criticized her for } (to Bill).
Now clean this apartment, and Bill
will do so again next month.
This
productive use of do is unavailable
if the head of do’s object is an open
class noun:
(9) *Mary did { a mean insult, inattention,
unexpected behavior, her great affection, some cleaning,
a bad arrangement, another home video } (to Bill).
*do more courage, *do a high pulse, *do a gallon of
turpentine, *do the heat, *do a look
Exactly
when in a derivation is a typical Syntacticon item like (4.15) inserted? This
is not fully worked out in Lexicon and
Grammar. But some analyses since that time (Emonds, 2001; Veselovská 2001; Whong-Barr, 2002) seem to converge
on the following specification:
(10) Syntactic
Lexicalization. LF-interpreted lexical items of category B (lacking
semantic f) are inserted at the end
of processing the smallest phase (or
cyclic domain) β that contains B.
Even
so, (10) precedes selection of
β--or extraction from β--in the next highest phase (domain).
In
addition to (10), the open class Dictionary co-opts Syntacticon items for
idioms and other meanings using purely semantic f: get going, let (something)
be, have at (it), have (one’s) way, get it (= understand), come (= have an orgasm), a must do, the have nots, a would be
star, etc.
Not
surprisingly then, the verb do appears
in several combinations which include purely semantic ƒ. These are listed in
the open class Dictionary, akin to idioms. A diagnostic for a Dictionary use of
do is that the optional to-phrase in the Syntacticon entry
(4.15) is excluded:
(11) (i) do
= clean:
Now
please do this apartment thoroughly (*to Mary).
(ii) do
= visit as a tourist:
We
haven't done Nara yet. But we will do it (*to them) next week.
(iii) do
= cover as a media story:
That
TV channel did the earthquake again (*to the local residents).
(iv) do in (someone)
The head teacher may do
that student in (*to his parents).
Open
class items are inserted at the outset of classically conceived derivations. In
fact, only such deep insertion allows transformational grammar to capture
relations such as active-passive.
(4.7) Deep
Lexicalization. Items which have non-syntactic, purely semantic features f must satisfy lexical insertion
conditions (just) before syntactic processing of the smallest cyclic domains
(phases) containing them. Recall, such f occur
only on N, V, A and P.
A
third use of English do is as an “auxiliary.”
As analyzed in Chomsky (1957), do
also appears outside of VP (under today’s category I) to “support” the
inherently suffixal English Tense morphemes. Thus, the Syntacticon entry for do (4.15) must be supplemented:
(12) Do-support: do, V, + ___ [ I, -MODAL ][6]
What
makes do-support work empirically in Syntactic Structures is that it follows other transformations (Chomsky,
1995, Ch. 4, note 22). What forces this delay of do-support?
For
the answer, we need to ask a related question: what leads to the ACTIVITY
(non-stative) sense of do in its
canonical use (4.15), and why is this lacking with do-support? Since most verbs have this sense, ACTIVITY is
presumably just the unmarked LF interpretation of the category V.
(13) Canonical
interpretation in LF. Each value of X0 has a canonical interpretation in LF, which
for V is ACTIVITY. For P it is LOCATION in space or time, etc.
Now
the "meaningless" auxiliary do
must not be subject to (13), given well-known facts that it imposes no
+ACTIVITY sense on a main V (resemble
remains stative) or ellipted VP:
(14) Mary does not resemble her mother.
John didn't know the answer, but Mary did
[VP Ø ].
Sam does resemble his father, doesn’t he?
In
contrast, the VP do so, in which do is under VP, is always +ACTIVITY:
(15) Mary does not resemble her mother, but she
will (*do so soon).
John didn't know the answer, but Mary
{did/ *did so}.
Sam resembles his father, and Jim
{does / *does so} too.
The
semantic vacuity of do-support
correctly follows from Canonical Realization (1.5b): the head category V can
contribute to LF only in its canonical position as head of VP.[7]
(4.21) Phonological
Lexicalization. Items specified in terms of uninterpreted (contextual and
alternatively realized) features enter a derivation subsequent to operations contributing to LF.
We
turn in the next section to alternatively realized features. Since auxiliary do has only an uninterpreted V category
and a contextual frame, it is inserted in PF (“late”) by virtue of (4.21).
Do-support makes us ask: how far away from
its canonical position can a category be realized?
V.
ALTERNATIVE REALIZATION BY CLOSED CLASS ITEMS (4.75l)
Many
paradigms empirically justify head
movements such as French finite V to I, I to C inversion, Dutch V movement
in sequences of final Vs, and Semitic N to D movement.
However,
head movement as widely used is a non-constrained device. It should apply only
when general, and be excluded as an option for only “some” members of a
category. In particular, English has no
V to I movement for do, be, or have (Emonds, 1994).
More
generally, syntactic features F (not semantic f) may be realized essentially “one lexical head away” (higher or
lower) than their canonical position, and no further.
We
first define lexical head. By virtue of Syntactic and Phonological
Lexicalization, the highest structural heads and the highest lexical heads
don’t necessarily coincide early in a derivation:
(4.27) Lexical
Head/Projection. Let Y0 be the highest lexically filled head in
Zj. Then Y0 is the lexical
head of Zj, and Zj is a lexical projection of Y0.
Heads
are also required in English to be rightmost in Z0 and leftmost in
ZP.
(4.20)
Alternative Realization (AR). A
syntactic feature F canonically matched in UG with category B can be
alternatively realized in a closed class grammatical morpheme under X0, provided X0
and B are both lexical heads of sister constituents. (Slightly generalized from Ch. 4)
F
can in fact be V itself. Auxiliary do
alternatively realizes V (=B) under an I0 (= X0) that is
a sister of VP (=Bj); I take X0 to be the head of itself.
The nodes in the AR relation are in bold.
_files/image001.gif)
_files/image002.gif)
(16) IP
_files/image003.gif){kind=small}
_files/image004.gif){kind=small}
DP I'
VP
_files/image005.gif)
_files/image006.gif){kind=small}
_files/image007.gif){kind=small}
John [I, -PAST] = X0
_files/image008.gif){kind=small}
V’
_files/image009.gif){kind=small}
_files/image010.gif){kind=small}
V
[I, -PAST] NEG
V = F = B DP
do es not know Mary
AR
can purge syntax of all “lowering” rules and hidden vestiges of these reworked
as untestable and invisible “LF raising” rules. It moreover eliminates
non-explanatory proposals that fully inflected stems are somehow generated by a
separate mechanism prior to a derivation.
To
see this, consider finite verbal inflections such as English –ed. As is typical of inflections, it
appears in syntactic positions associated with its host V. The best way to
express this is to say that a verbal inflection is a V. No special principle of “inflectional morphology” is
needed.
The
canonical position of Past is on I, and its position on V thus another example
of Alternative Realization, this time "lower" (AR nodes are again in
bold). F = PAST and B = I.
(4.25) Past
Tense: ed, V, +PAST, +V___
_files/image011.gif)
_files/image012.gif){kind=small}
(4.26) IP
_files/image013.gif){kind=small}
DP I'
_files/image014.gif){kind=small}
_files/image015.gif){kind=small}
Ann [I, PAST]
= [B, F] VP = XP
_files/image016.gif){kind=small}
_files/image017.gif){kind=small}
_files/image018.gif){kind=small}
Ø V
= X0 DP
_files/image019.gif){kind=small}
V [V, PAST] papers
burn ed
We
note the features of –ed are either
(i) alternatively realized, i.e. PAST, or (ii) contextual (+V___). It follows
from Phonological Lexicalization (4.21) that –ed is inserted in PF.[8]
Morphology
specialists usually object to extending Lieber’s (1980) right hand head rule to
inflection, claiming that the head of a VP as in (4.26) must be the V burn. That is, burn rather than –ed
selects complements, assigns accusative case, is the semantic head of VP at LF,
etc.
But
according to (4.27), burn is the head
of VP in syntax and LF in (4.26), since –ed
is absent.
The
following principles account for why I is empty if its features are undergo AR:
(4.35) Invisible
Category Principle (ICP). If all
marked canonical features F on B are alternatively realized by AR (4.20),
except perhaps B itself, then B can be empty.
The
ICP doesn't itself require that B must be empty. The obligatory zeroing of e.g.
I in (4.26) is performed by a separate principle that has effects outside AR:
(4.36) Economy
of Derivation. Two deep structures that differ only by empty categories not
interpretable at LF count as equivalent. Of equivalent deep structures, prefer
the derivation with the fewest insertions of free morphemes.
AR
defines the distance at which (only) closed class items can realize
non-canonical features.
VI.
ABSENCE OF CONTENT FEATURES
It
is natural to include apparent English “be-raising”
in a discussion of do-support and
affix movement. Emonds (1994) gives several arguments that English has no V to
I raising for be or any other V. More
generally, head movement applies to entire classes of X0 or not at
all.
I
claimed in (13) that ACTIVITY is simply the name given to the canonical LF
interpretation of the category V. This means that only its opposite STATIVE
must be marked on individual V. Emonds (2002) proposes a lexical notation for
this: the null symbol Ø among a lexical category’s syntactic features means,
not null in syntax, but rather absence
of content at LF.
(17) Absence
of Content. The syntactic feature Ø is a canonically matched with L = N, A,
V, P and means that L0 does not receive a canonical interpretation
at LF (V = ACTIVITY, etc.).
This
notation “L0, Ø” gets rid of superfluous names such as ACTIVITY,
LOCATION, etc.
Thus,
a P canonically indicates spatial or temporal LOCATION in LF. But the English P
of does not. So while the lexical
category for at is just P, that of
the marked item of is P, SOURCE, Ø.
Consequently,
the Syntacticon entry for be must be
(16):
(18) Verbal
copula: be, V, Ø, contextual frame(s)[9]
Now
let’s consider Ø to be a syntactic feature F like any other. Hence it can be
alternatively realized on I; (17) provides one of five English examples of this
(am, are, is, was, were).
(19) Inflected
copula: were, I, PAST, -MODAL, PLURAL, Ø
The
above principles (4.35)-(4.36) now explain why (19) allows the highest V in VP
to be empty rather than spelled out as be
when its only marked canonical feature is Ø, i.e. STATIVE in traditional terms.
This empty category V with inflected copulas gives the effect of “be-raising.”
For
why a not between I and V doesn’t
“block” the appearance of inflected copulas, see note 8. PLURAL in (19) is of
course an alternatively realized feature of the subject DP.
The
present system treats do as an unmarked V and be/ have as the least marked
V.
There
is a question of why inflected copulas
don’t occur with other stative verbs, wrongly “doubling” them (They { did/ *were } not seem nice). Ch. 4 allows unmarked AR only to permit a “saving” in
Economy or rescuing a derivation, but treats this issue of doubling too
briefly.
I
would say the AR system needs investigation and clarification in this important
area.
VII.
REVIEW OF TRI-LEVEL INSERTION (4.75m)
I
have distinguished three main types of
features that interact differently with respect to lexical insertion,
syntactic derivation and LF. The English grammatical verb do undergoes all three types. This section gives the English P to as another item with multi-level
insertion.
(i) Items
with purely semantic features f,
present only on the categories L = N,
V, A, P. They are used in LF but not in syntax and are not present on the closed classes of grammatical L.
(20) Examples:
do = clean, do = visit, do = make a media
event, do in = finish someone off
These transitive do are used in Dictionary entries, and all reject indirect objects;
cf. (11).
(21)
Deep
to with idioms: to, P, PATH, +___ DP; used in idioms with semantic ƒ
(22) Mary couldn't hold a candle to Sue.
Sam put the question to the jury.
Mary made passionate love to Bill.
Bill came to slowly.
Consonant
with (21), dative movement depends on P being empty in the syntax (Emonds,
1993):
(23)
*Mary couldn't hold Sue a candle.
*Sam put the jury the question.
*Mary made Bill passionate love.
(ii) Items
with no f but with cognitive
syntactic features F in canonical positions, which can occur with all syntactic categories. They play a
central role in both syntax and at LF.
(24) Examples: do so as a pro-VP for an ACTIVITY VP; do whose object N lacks semantic ƒ, possible with a to-phrase (do something unexpected to Bill). See (4.15) just before (8) above.
(25)
Syntactic
to interpreted as PATH: to, P, PATH, +___ DP
(26) They should move more city offices { to our
neighborhood/ next to the park }.
John pulled the toy { to his mother/ up
the stairs }.
Mary turned the papers over to her
lawyer.
Dative
movement is also excluded when P indicates a continuous PATH rather than simply
introducing (and case-marking) a Goal DP (Larson, 1988, section 5). This
interpretation of P is expressed by to
being inserted in the PP “phase” prior to dative movement in the VP phase.
(27) *They should move our neighborhood more
city offices.
*John pulled his mother the toy.
*Mary turned her lawyer over the papers.
(iii) Items with only purely syntactic features F', also possible with
all syntactic categories. Such F’ can be contextual features, features
alternatively realized in non-canonical positions, and an “absence of content”
combination [X0, Ø]. None play any role in LF.
(28)
Examples: do under I with TENSE as generated by do-support (12). Purely syntactic features also fully characterize
the items be in (18) and past tense –ed in (4.25).
(29)
To inserted as PF case-marker: to, P, PATH, Ø, +__ DP
Verbs
with frames ___DP^DP are forced to have PP complements so P can provide the
second DP with case. Heads of these Ps need to assign oblique case but should
be invisible in LF, which is what (29) achieves. Precisely these PPs give rise
to English dative movement (Emonds, 1993).
(30) a. They
should send more city officials to our neighborhood.
John
handed the toy to his mother.
Mary sent the papers
over to her lawyer.
b. They
should send our neighborhood more city officials.
John
handed his mother the toy.
Mary sent her lawyer
over the papers.
(31) Unified
entry for to: to, P, PATH, ( Ø ), +__ DP
As
discussed more in Emonds (2002), the feature Ø on some L exempts that L from
being selected by some higher frame ___L. For example, English of is [P, Ø] and V that select PPs (put, dash, glance, head) don’t select of-phrases; verbs such as try, persuade,
etc. select infinitival VPs but not stative verbs; expletives are [D, Ø] and
don’t satisfy +___DP.
The
Vs allowing dative movement as in (31) then select their indirect objects as
DPs not PPs.
(4.8) Tri-Level
Lexical Insertion. Lexical items from the Syntacticon, in accord with their
feature content, can be inserted at different stages of a derivation of a
sentence Si, via the Dictionary (“deep structure”), during the
syntactic derivation, and during phonology.[10]
_files/image020.gif)
_files/image021.gif)
Dictionary of entries with features f; Syntacticon of closed class categories
_files/image022.gif){kind=small}
interface
with non-linguistic conceptual including N, V, A, P;
elements limited structures; only N, V, A, P; coining and to cognitive syntactic features F
compounding
to build vocabulary
_files/image023.gif){kind=small}
_files/image024.gif){kind=small}
_files/image025.gif){kind=small}
syntactic phonological
derivation derivation
_files/image026.gif){kind=small}
Lexical
choices (Si) è è è è è è “Spell Out” è è è è è è
Phonological
from
open classes Form, PF(Si)
binding and co-indexing conditions
Logical
Form, LF(Si)
The accessibility of
the Syntacticon at different levels of the derivation agrees with a major
characteristic of Borer’s (1991) “Parallel Morphology.” Indeed our ideas have
developed along many of the same lines for over a decade using similar
argumentation. Our differences are to a great extent (by no means entirely)
terminological. I do not see why the word “parallel” should be used in either
of our models, if no second derivation besides the syntactic one is implied.
And as indicated here throughout, the Syntacticon is not limited to bound
morphemes or “morphology.”
(4.12) Types of insertion from the Syntacticon:
INSERTION
LEVEL FREE MORPHEMES BOUND MORPHEMES
Prior
to syntactic closed
class X with non-productive derivational
computation
on a specialized
meanings, morphology with specialized domain
(“phase”) and parts
of idioms lexical
meanings
During
syntactic closed
class grammatical productive derivational
computation,
prior words with LF
syntactic morphology; cf. sections
to
Spell Out features;
cf. Ch. 6 4.6,
4.7.2, 4.7.3 and Ch. 5
During
PF closed
class grammatical inflectional morphology;
computation,
after words required by
the EPP, sections
4.4, 4.7.1 and
Spell
Out case-marking,
etc. cf. Ch. 7 Ch. 5
REFERENCES
Borer, H. (1984) Parametric Syntax, Foris, Dordrecht.
---- (1991) “The
Causative-Inchoative Alternation: a Case Study In Parallel Morphology,” Linguistic Review 8, 119-158.
Bresnan, J. (1973) “The
Syntax of the Comparative Clause in Englis,.” Linguistic Inquiry 28, 119-164.
Chomsky, N. (1957) Syntactic Structures, Mouton, The Hague.
---- (1965) Aspects of the Theory of Syntax. MIT
Press, Cambridge.
---- (1995) The Minimalist Program. MIT Press,
Cambridge.
Corver, N. and H. van
Riemsdijk (2001) Semi-lexical Categories:
the Function of Content Words and the Content of Function Words, Mouton de Gruyter, Berlin.
Emonds, J. (1985) A Unified Theory of Syntactic Categories,
Foris, Dordrecht.
---- (1987) “The
Invisible Category Principle,” Linguistic
Inquiry 18, 613-631.
---- (1993) “Projecting
Indirect Objects,” Linguistic Review
10, 211-263.
---- (1994) “Two
Conditions of Economy,” in G. Cinque et al., eds., Paths toward Universal Grammar: Papers in Honor of Richard Kayne,
Georgetown University Press, Washington, D.C.
---- (2000) Lexicon and Grammar: the English Syntacticon,
Mouton de Gruyter, Berlin.
---- (2001) “The Flat
Structure Economy of Semi-Lexical Heads,” in Corver and van Riemsdijk (2001).
---- (2002) “Formatting
Lexical Entries: Interface Optionality and Zero.” Theoretical and Applied Linguistics at Kobe Shoin 5, 1-22.
Grimshaw, J. (1979)
“Complement Selection and the Lexicon,” Linguistic
Inquiry 10, 279-326.
Hannahs, S. J. (1995) Prosodic Structure and French
Morphophonology: Linguistische Arbeiten 337, Niemeyer Verlag, Tuebingen.
Larson,
R. (1988) “On the double object
construction,” Linguistic Inquiry 19,
335-391.
Lieber, R. (1980) The Organization of the Lexicon, MIT doctoral
dissertation.
Lasnik, H. (1991)
“Language Acquisition and Two Types of Constraints,” Behavioral and Brain Sciences 14.
Manzini, R. and K.
Wexler (1987) “Parameters, Binding Theory and Learnability,” Linguistic Inquiry 18, 413-444.
Ouhalla,
J. (1991) Functional Categories and
Parametric Variation, Routledge, London.
Veselovská,
L. (2001) "Agreement Patterns of Czech Group Nouns and Quantifiers,"
in Corver and van Riemsdijk (2001).
Whong-Barr,
M. (2002) “Small Clauses are Bigger,” in Proceedings
of the Linguistics Society of Korea