Winter School in Language and Speech

Winter School in Language and Speech
Technologies, 2012
Rovira i Virgili University, Tarragona
Geoffrey K. Pullum
University of Edinburgh
January 2012
Tarragona, Spain
Geoffrey K. Pullum (Edinburgh) WSLST 2012 Jan 2012 1 / 163

The English language: preliminaries
What we mean by English: grammar, style, dialects, etc.

What English is like: brief typology
Structure of The Cambridge Grammar of the English Language
Categories and functions
Lexical categorization: V, N, D, Adj, Adv, P, Sbr, Cdr, Intj
Syntactic features: [auxiliary], [definite], [number], [pronoun], ...
Projection to phrases: heads and dependents.
Complements and adjuncts
Subjects and objects
Canonical clauses

The Cambridge Grammar
The Cambridge Grammar of the English Language

by Rodney Huddleston and Geoffrey K. Pullum
is abbreviated here as CGEL.

The Cambridge Grammar of the English Language

by Rodney Huddleston and Geoffrey K. Pullum
is abbreviated here as CGEL.
Useful elementary textbook based on CGEL:

A Students Introduction to English
Grammar a guide to CGEL in 300 pages.

Structure of CGEL:
Chapter 1: introduction
Chapter 2: overview of the content of the entire book
Chapters 37: lexical categorization (N, V, Adj, Adv, Prep)
Chapter 8: adjuncts
Chapters 916: non-canonical clauses (negation, subordinate
clauses, non-declarative clauses, etc.)
Chapter 17: anaphora and ellipsis
Chapter 18: inflectional morphology
Chapter 19: derivational morphology (word formation)
Chapter 20: punctuation

First topic: verbs and auxiliaries
Verbs and the AUXILIARY/LEXICAL distinction
The verbal inflectional system
The Catenative-Auxiliary analysis

Lexical and auxiliary verbs
Two syntactic classes of verbs (they are not semantic!):
A special subsets, the auxiliary verbs
The rest, called the lexical verbs
Warning: The traditional definition of auxiliary verbs

as helping verbs is toxic! Stay away from it!
(In CGEL terms: it overlooks the non-core uses of auxiliaries.)

English verb inflection
The unique paradigm of be:
PRIMARY FORMS
NEUTRAL NEGATIVE
1st sg 3rd sg other 1st sg 3rd sg other
present am is are arent isnt arent
preterite was were wasnt werent
irrealis were werent
SECONDARY FORMS
plain form past participle gerund-participle
be been being

The wrong wrong wrong way to describe regular verbs:
PRIMARY FORMS
NEUTRAL NEGATIVE
1st sg 3rd sg other 1st sg 3rd sg other
present fill fills fill
preterite filled filled filled
irrealis filled filled filled
SECONDARY FORMS
plain form past participle gerund-participle
fill filled filling

The disastrous threesome lie/lied/lied (tell untruths), lie/lay /lain (be

recumbent), and lay /laid/laid (deposit):
plain present lie lie lay

PRIMARY 3rd sg present lies lies lays
preterite lied lay laid
plain form lie lie lay
SECONDARY gerund-participle lying lying laying
past participle lied lain laid

English verb inflection: LEXICAL VERBS
The six paradigmatic forms that have to be recognized

for a regular lexical verb like fill:
filled preterite
PRIMARY fills 3rd singular present
fill plain present
fill plain form
SECONDARY filling gerund-participle
filled past participle

The distinction between finite and non-finite clauses ALMOST BUT

NOT QUITE aligns with the distinction between primary and secondary
verb inflectional forms:
VERB FORM CONSTRUCTION CLAUSE

PRIMARY I am kind.
IMPERATIVE Be kind. FINITE
PLAIN FORM SUBJUNCTIVE that I be kind
INFINITIVAL for me to be kind
GERUND - PPL . being kind NON - FINITE
PAST- PPL . been kind

The distinction between finite and non-finite clauses ALMOST BUT

NOT QUITE aligns with the distinction between primary and secondary
verb inflectional forms:
VERB FORM CONSTRUCTION CLAUSE

PRIMARY I am kind.
IMPERATIVE Be kind. FINITE
PLAIN FORM SUBJUNCTIVE that I be kind
INFINITIVAL for me to be kind
GERUND - PPL . being kind NON - FINITE
PAST- PPL . been kind

But not all forms have different shapes:
fill ed preterite
PRIMARY fill s 3rd singular present
fill plain present
fill plain form
SECONDARY fill ing gerund-participle
fill ed past participle

There are actually no verbs with exactly 6 shapes.
Some irregular verbs (with past participle identical to plain form)

have just 3 shapes for the 6 paradigm forms:
3rd sing present gerund-participle
hits hitting
plain present plain form
hit
preterite past participle

Regular verbs, like bake, have 4 shapes

for the 6 paradigm forms:
bakes baking
bake
baked

Several dozen verbs with irregular past participle suffixes,

like take, have 5 shapes for 6 paradigm forms:
takes taking
take
took taken

English verb inflection: AUXILIARY VERBS
And now for the auxiliary verbs. As we saw, be

overcrowds the chart with extra forms:
is(nt) being
am / are(nt) be
was(nt) / were(nt) been

The negative forms really need a separate column.

The negative forms really are words, not word sequences
(Zwicky & Pullum, Cliticization versus inflection: English nt,

Language 59 [1983], 502-513)
Key syntactic evidence: single word inversion in interrogatives
We should go. Should we go?

We shouldnt go. Shouldnt we go?
We shouldve gone. *Shouldve we gone?
We oughta go. *Oughta we go?

For the non-modal auxiliary verb have (the one that expresses perfect
tense), the paradigm is the same as its transitive lexical verb
counterpart, except for the additional negative primary forms, and the
fact that it has no past participle:
PRIMARY SECONDARY
3rd sg pres neutral 3rd sg pres negative gerund-participle
have havent having

plain present neutral plain present negative plain form
have havent have

preterite neutral preterite negative past participle
had hadnt *

Auxiliary do is similar, except there are no grounds for saying it has a

gerund-participle or a plain form, so it has no secondary forms:
PRIMARY SECONDARY
3rd sg pres neutral 3rd sg pres negative gerund-participle
does doesnt *
plain present neutral plain present negative plain form
do dont *
preterite neutral preterite negative past participle
did didnt *

The modal auxiliary verb will has one less distinction than auxiliary do
there is no distinct 3rd singular present; moreover, all secondary
forms are missing:
PRIMARY SECONDARY
present neutral present negative gerund-participle
*
will wont
plain form
preterite neutral preterite negative *

past participle
would wouldnt
*

The modal may is defective in one more form (present negative):
PRIMARY SECONDARY
*
may *
plain form

past participle
might mightnt
*

Must is even more defective (no preterite), but still has an inflectional
paradigm:
PRIMARY SECONDARY
*
must mustnt
plain form

past participle
* *
*

Such defectiveness seems to be genuinely morphological.

Consider the data for the modal dare:
No one dare tell him. [present neutral]

We darent tell him. [present negative]
No one dared tell him before. [preterite neutral]
We darednt tell him before. [no preterite negative]
Not daring tell him, we left. [no gerund-participle]
I thought no one would dare tell him. [plain form]
No one has dared tell him yet. [past participle]

So the paradigm seems to be this:
PRIMARY SECONDARY
*
dare darent
plain form
preterite neutral preterite negative dare

past participle
dared *
dared

The modal need is different again. The data seem to be:
No one need tell him. [present neutral]

We neednt tell him. [present negative]
No one needed tell him before. [no preterite negative]
We needednt tell him before. [no preterite negative]
Not needing tell him, we left. [no gerund-participle]
I thought no one would need tell him. [plain form]
No one has needed tell him yet. [no past participle]

So the paradigm seems to be this:
PRIMARY SECONDARY
*
need neednt
plain form
preterite neutral preterite negative need

past participle
* *
*

Certain special properties, the NICE properties, distinguish lexical

from auxiliary verbs:
N Special Negation syntax, preceding clause-negating not

(He is not in), and special negation morphology with the nt suffix
(He isnt in).
I Special Initial (Inverted) position in independent polar

interrogatives (etc.): Do you love me?
C Special Code interpretation: Complement Omission Denotes

Ellipsis (Yes, you are [ ])
E Special Emphasis phonology heavy stress signals polarity

emphasis (But you cn help!)

The NICE properties are found with certain non-core uses of

auxiliaries. Illustrating just with inversion:
Is he [PP in? ]
Have you [NP any idea? ]
Would you rather [Clause I didnt come? ]
The underlined verb is not helping any main-clause lexical verb.
These items invert, but take PP or NP or full tensed Clause

complements (bracketed).
If the underlined words are in Aux, what is head of VP?

What is a catenative complement?
VP-internal
subjectless
non-finite
clausal complement
not an object
not predicative
not ascriptive
not specificational
allows arbitrary chaining

VP
PPP
P
V VP
PPP
P
might V VP
PPP
P
have V VP
PP

P
P
helped V VP
HH
H
avoid V AdjP
ZZ
seeming foolish

VP VP

V VP V VP

ought to be qualified thought to be qualified

As with lexical verbs, some auxiliary verbs take a bare infinitival

complement and some take a to-infinitival:
taking bare infinitival taking to-infinitival

lexical have; hear ; help; let; dare; hope; like; need;
make; see . . . seem; try ; want . . .
auxiliary can; dare; do; may ; is (modal use); ought
must; need; shall; will
Note also these idiomatic combinations:
be + going and BrE have + got: to-infinitival

had + better and would + rather / sooner / as soon: bare infinitival

The catenative-auxiliary analysis claims the auxiliaries are all verbs

and there is one clause per verb. Each clause can be separately
negated:
NEGATED :
I have always taken bribes from lobbyists. Neither.
I have not always taken bribes from lobbyists. Have.
I have always not taken bribes from lobbyists. Take.
I have not always not taken bribes from lobbyists. Both.

The old dependent-auxiliary analysis, where auxiliaries are little

non-verbal markers prefixed to the verb in a clause, has no explanation
for negation facts such as those just illustrated.
Nor can it handle various other facts to do with temporal specification

and constituent structure.
On this topic, read CGEL Ch 3, pp. 12141220.

Nouns, Determinatives, and NPs
Plan for this part of the lecture:
Nouns and their definition
The category of Determinatives
The Determiner function
Prenominal attributive modifiers
Pronouns, and why they are really nouns

Defining nouns
The thing-word caricature
Lowth, A Short Introduction to English Grammar (1762): A noun is

the name of any thing conceived to subsist, or of which we have
any notion
Murray, English Grammar (1795): A noun is the name of any

thing that exists, or of which we have any notion
Curme, English Grammar (1925): A noun . . . is the name of a

living being or lifeless thing
Garner, in The Chicago Manual of Style, 15th edition (2003): A

noun is a word that names something
250 years of saying the same old same old!

Defining nouns
What the general public thinks today
I got through 12 years of state funded schooling with the sum total of
my grammatical knowledge being Nouns are thing words, verbs are
doing words, and adjectives are describing words. I suspect we never
covered adverbs.
John Wilkins, on his blog
Evolving Thoughts (14 June 2008)

Defining nouns: history without evolution
We declared war on terror. We declared war on terror its not even a

noun, so, good luck. After we defeat it, Im sure well take on that
bastard ennui.
Jon Stewart, in a graduation address
at the College of William and Mary
I think uh the con th... the phrase the war on terror uh as...
as if there could be a war on an adjective, I mean its its just or an
adverb it doesnt really work.
Stefan Halper, Senior Fellow at the Carnegie Council on Ethics and
International Affairs in New York, interviewed on radio WHYY

Defining nouns
They dont all denote things
fire: process of combustion (rapid oxidation)

absence: not being there
emptiness: nothing being there
failure: not managing to
lack : failing to be provided when needed
method: how (by what means) something is done
assistance: being there with supportive actions

Defining nouns
They dont all denote things
fire: process of combustion (rapid oxidation)

absence: not being there
emptiness: nothing being there
failure: not managing to
lack : failing to be provided when needed
method: how (by what means) something is done
assistance: being there with supportive actions
We dont get our definition of Noun from metaphysics;

we get our metaphysics from the use of our nouns.

Nouns
What they do have in common
Four inflectional forms: singular and plural plain case singular

and plural genitive case
Serve as heads of NPs with functions like Subject, Object,

Predicative Complement, Complement of Preposition
Preceded by determinatives like the or a(n)
Modified by (preceding) adjectives like big
Modified by (following) relative clause modifiers like

who I didnt recognize

Nouns
An irregular lexeme with full paradigm
Inflectional forms of child

PLAIN GENITIVE
SINGULAR child childs
PLURAL children childrens

Nouns
An irregular lexeme with full paradigm
Inflectional forms of child

PLAIN GENITIVE
SINGULAR child childs
PLURAL children childrens
Phonologically, regular nouns are only half as complex:
Inflectional forms of dog
PLAIN GENITIVE
SINGULAR /dag/
PLURAL /dagz/

Determinatives as a distinct category
The qualifies a noun definition of Adjective leads all dictionaries to

include among the adjectives:
(i) both the articles (the, a(n)
(ii) both demonstratives (this, that)
(iii) all quantifiers (all, some, most, several, . . . )
(iv) all numerals (one, two, . . . , 243, . . . )
(v) the dependent genitive forms of pronouns (my , your , . . . )
Treating these as Adjectives makes a complete hash of the syntactic

properties of that category.

The proposal defended here (and in CGEL):
Determinative is a category of words (like Noun or Preposition),

distinct (but not disjoint) from Adjective, containing about 35 basic
items plus all the numerals.
Determiner is a function (like Subject or Head).
Determinatives often (but not always) serve as Determiner.
The Determiner function is often (but not always) filled by a

Determinative.
Genitive NPs can also serve in Determiner function.

A Grammar of Spoken English (Harold Palmer and F. G. Blandford,

1939) cites some criteria for distinguishing adjectives from a class of
determinatives (Fr. adjectifs dterminatifs) from the class of
adjectives:
(a) Determinatives cannot be used predicatively.
Alert residents objected. The residents were alert.
All residents objected. 6 *The residents were all.

(b) Determinatives only rarely express grade (comparison).
Those who succeeded were more motivated than the others.
*Those who succeeded were more most than the others.

(b) Determinatives only rarely express grade (comparison).
Those who succeeded were more motivated than the others.
*Those who succeeded were more most than the others.
(c) Determinatives only rarely take modifiers, and in particular do not

take intensifying modifiers.
[Very alert] residents objected.
*[Very all] residents objected.

Four more reliable criteria also distinguish Adjectives from

Determinatives:
(d) Determinatives do not stack, or even (for the most part) co-occur
in NP structure.
Responsible, alert, intelligent residents helped out.
*These, the, all residents helped out.

(e) Adding a Determinative can make a singular count noun into a

grammatical NP but adding an adjective cannot:
I heard [NP this guy] was asking about me.
*I heard [NP thin guy] was asking about me.

(f) Most Determinatives occur in the fused determiner-head

construction, where a Determinative serves as Determiner and
Head simultaneously and thus constitutes a whole NP. No
adjective does this.
Id like to hear more about [NP this].
*Id like to hear more about [NP thin].

One further minor point is that there is a phonological condition

sufficient to guarantee non-adjectivehood:
(h) Adjectives can never begin with phonological //, but

determinatives can, and some do (that, the, this).
As a result of a historical accident, initial // is never found in Nouns,

Verbs, or Adjectives.

There are about 35 basic Determinatives:
a(n) a few a little all + another

any both certain+ each either
enough every few + little + many
much neither no one + said +
several some such+ sufficient + that
the this various+ we + whatever +
whatsoever + what + whichever which+ you +
Words with superscript + belong to other categories as well.

Words in boldface italics have varying inflectional forms.

The Determiner of an NP is intuitively an initial subconstituent that

fixes certain properties like definiteness and quantification and make it
fit for use semantically as an argument.
In semantic terms, a Determiner combines with the property

denotation of a nominal expression to form a full NP meaning (a
generalized quantifier, under many accounts).
The Determiner function in English is filled by either a Determinative

(this house) or an NP in the genitive case (the presidents house).

The Determiner of an NP is intuitively an initial subconstituent that

fixes certain properties like definiteness and quantification and make it
fit for use semantically as an argument.
In semantic terms, a Determiner combines with the property

denotation of a nominal expression to form a full NP meaning (a
generalized quantifier, under many accounts).
The Determiner function in English is filled by either a Determinative

(this house) or an NP in the genitive case (the presidents house).
What does Specifier mean in post-1970 transformational grammar?

Not clear. But probably Determiner i.e., its regarded as a function,
not a category, and certainly not a lexical category.

The DP Hypothesis
The DP Hypothesis is not merely that some phrases have D as Head.
CGEL agrees that there are phrases with D as Head such as hardly
any , just about all, almost every :
NP

Det: Head:
DP Nom

Mod: Head: Head:
AdvP D N
Head: any onions

Adv
hardly

The DP Hypothesis
The DP Hypothesis is not merely that some phrases have D as Head.
CGEL agrees that there are phrases with D as Head such as hardly
any , just about all, almost every :
NP

Det: Head:
DP Nom

Mod: Head: Mod: Head:
AdvP D AdjP Nom

Mod: Head: all Head: Head:
AdvP Adv Adj N
Head: about civilized societies

Adv
just
The DP Hypothesis
But the DP hypothesis is not that some phrases have D as Head; it is

that phrases like the archbishop and this bicycle have the D as
Head.
In a phrase like the king of France, it is claimed that the is the Head.
The rest of the phrase, king of France, is a dependent, specifically a
Complement.
Once informally proposed by John Lyons, the DP Hypothesis was

revived in 1987 in the MIT doctoral dissertation of Steven Abney.

Attributive Modifiers
English allows pre-head modification of nouns by adjuncts of various

categorial types:
Adjective Phrase (AdjP):

those nice new green sandals you bought
Nominal (Nom, i.e. N): a labor union organizer
Gerund-participial Verb Phrase (VPGPL ):

several quietly sleeping children
Past-participial Verb Phrase (VPPPL ):

the universitys hastily developed pay reduction plan

With the ordering of attributive modifiers we find a phenomenon that

appears to be syntactic at first but is, I think, truly semantic through
and through or at least, sensitive to both semantic and syntactic
properties.
For example, in nice new green sandals it seems hard to reorder any
of the adjectives:
nice new green sandals

?? nice green new sandals
?? new nice green sandals
?? new green nice sandals
?? green nice new sandals
?? green new nice sandals

One suggested operational semantics for the ?? prefix would be

profoundly unlikely to get more than a tiny handful of non-spurious
Google hits.
For example: 100 instances of friendly little white + Noun can be found
using Google; but for ?? little white friendly , only 10 hits, most spurious.

The conditions on sequencing of attributive modifiers are not strict like

the condition that says the Determiner precedes the Head. They are
labile constraints: violating them creates special effects, or lowers
acceptability, but does not clearly mark the result as not being English.
Early modifiers (determinatives such as numerals; superlative

adjectives; ordinal adjectives; primacy adjectives) have a strong
tendency to precede residual modifiers (all others):
the two vital facts [D + residual]

the largest known meteorite [sup + residual]
the second abortive attempt [ord + residual]
a key new proposal [prim + residual]

And within the residual modifiers, there is a weaker tendency to adhere

to an order like this:
evaluative modifiers before
general modifiers before
age modifiers before
color modifiers before
provenance modifiers before
manufacture modifiers before
type modifiers
CGEL cites this NP:
an attractive tight-fitting new pink Italian lycra womens swimsuit

Shifting any modifier leftward in its sequence tends to suggest it is

being used contrastively with the following constituent presupposed:
pink Italian swimsuit if Italian swimsuits are under discussion;
Italian pink swimsuit if pink swimsuits are under discussion.

Pronouns and determinatives
The inflectional paradigms of the personal pronoun lexemes
NOM ACC DEP GEN IND GEN REFL

I me my mine myself
you your yours yourself
he him his himself
she her hers herself
it its itself
we us our ours ourselves
you your yours ourselves
they them their theirs themselves

Pronouns and determinatives
A small overlap: there are two personal determinatives.
PRONOUN USES DETERMINATIVE USES

I do all right. *I man do all right.
You do all right. *You man do all right.
He does all right. *He man does all right.
She does all right. *She woman does all right.
It does all right. *It puppy does all right.
We do all right. We men do all right. both OK
You do all right. You men do all right. both OK
They do all right. *They men do all right.
*The do all right. The women do all right.
*A does all right. A linguist does all right.
*Every do all right. Every puppy does all right.

Adjectives, Adverbs, their phrases, and their functions
Plan for this part of the lecture
The disastrous function / category confusion
The functions served by adjective phrases
Criteria for identifying adjectives
Criteria for identifying adverbs
A note on zero-derived adverbs in non-standard dialects
The Categorial Exclusivity assumption
The Adjective-Adverb Identity thesis

The function / category confusion
For two hundred years or more the following two notions have been
persistently confused in discussions of English grammar:
What kind of constituent is this? category
What role does this constituent play? function

For two hundred years or more the following two notions have been
persistently confused in discussions of English grammar:
What kind of constituent is this? category
What role does this constituent play? function
Category can be listed (for words) in a dictionary: it is independent of

the syntax of any particular sentence.
Function is not, and cannot be, a lexical property. It is entirely relative

to the syntax of some particular expression.

Just to make sure the disastrous confusion continues for another

hundred years or so, the Merriam-Webster dictionaries actually use
the term function for (lexical) category!
Main Entry: 1 pig
Pronunciation: pig
Function: noun
Inflected Form: -s
Usage: often attributive
Etymology: Middle English pigge
1 : a young swine of either sex that has not
reached sexual maturity; broadly : a wild or
domestic swine see HOG 1a . . .

Just to make sure the disastrous confusion continues for another

hundred years or so, the Merriam-Webster dictionaries actually use
the term function for (lexical) category!
Main Entry: 1 pig
Pronunciation: pig
Function:Category!! noun
Inflected Form: -s
Usage: often attributive
Etymology: Middle English pigge
1 : a young swine of either sex that has not
reached sexual maturity; broadly : a wild or
domestic swine see HOG 1a . . .

The functions of adjectives and AdjPs
Adj serves, of course, as Head of AdjP. (When nothing hangs on it we

often say adjective when strictly we mean AdjP.)
The two most important functions in which AdjPs serve are:
1. Attributive Modifier in the structure of NP (n.b.: stackable):
[NP those [Nom totally stupid [Nom red [Nom pants ] ] ] ]
2. Predicative Complement in the structure of VP:
Those pants [VP look really stupid ]

The functions of adjective phrases and AdjPs
There are a few less significant functions of AdjP:
Postpositive Modifier in the structure of NP:
Id never seen [NP anything so stupid ]
External Modifier in the structure of indefinite NP:
Id never seen [NP so stupid a [Nom pair of pants ]]
(Marginally, perhaps also Subject function: Totally stupid was my main

impression.)

The functions of adjective phrases
Traditional grammar crucially fails to recognize that
occurrence as Attributive 6= membership in Adj
In the same spirit as the thing word definition of Noun, traditional

grammarians define adjective as a word that modifies or qualifies or
adds to the meaning of a noun.
This is vague, semantically-tinged function talk. As a definition, it is
hopeless.
Consider: The good die young.
Two adjectives, no nouns. Modifying a noun is not necessary.
But as we shall see, it is also not sufficient.

Dictionaries often list various nouns with Adj as a second category.
Merriam-Webster is typical:
Main Entry: 2 head
Pronunciation: hed
Function:Category!! adjective
Etymology: Middle English heved, hed, from
heved, hed, n.
1 : of, relating to, or for a head or the head
2 : PRINCIPAL, CHIEF, LEADING, FIRST hhead choristeri
hhead cooki
3 : situated at the head hhead wall* hhead sailsi
4 : coming from in front : meeting the head as it is
moved forward hhead seai hhead tidei

Consider the modifiers in NPs like Alaska residents, Beatles

recordings, California girls, Dell laptops, Edinburgh weather . . .
If serving in Attributive Modifier function is sufficient to determine

adjectivehood, then the Adjective category will have to include
all place names in the world (Sheffield steel)
every company name (IBM stock , Toyota truck )
every number name (their prestigious 10025 zipcode)
every name of a chemical element or other substance (aluminium
foil, gold ring, oil painting, plutonium bomb)
every plant name (mahogany table, grass verge) . . .

Under the traditional view the list of adjectives will never end.
Adjective will be an open category, even larger than Noun.
And there will be no distinction in grammatical properties between the

two.

M-W is cautious in defining adjective, claiming only typicality, not
criteriality:
. . . typically used as a modifier of a noun to denote a

quality of the thing named . . .

M-W is cautious in defining adjective, claiming only typicality, not
criteriality:
. . . typically used as a modifier of a noun to denote a

quality of the thing named . . .
But amusingly, M-W also has an entry for adjective as an adjective!

The primary sense given is:
being an adjective han adjective wordi : functioning as an ad-

jective han adjective clausei : fitting or suitable to an adjective
hadjective uses of nounsi hadjective inflectionsi
A classic confusion over what adjective means, in a dictionary entry

where adjective is wrongly analysed as an adjective!

Criteria for adjectivehood
How do we define adjective, if not by reference to the function

Attributive Modifier?
In any language, the adjectives (if there are any) are a grammatically
distinct class of words including the simplest and most direct ways of
denoting one-dimensional and stative properties such as being
good, bad, large, small, new, old, black, white, etc.
But that is applicable only in universal grammar.
In English we can, and must, be more specific.

Adjectives usually denote static and gradable properties, and

many basic ones inflect for grade (old, older , oldest).
Uninflectable adjectives, if gradable, express grade with pre-Head

more and most.
No adjectives show agreement for person or number.
Adverbs function as pre-head Modifiers of Adjectives (unusually

intelligent, insanely great).
Typical AdjPs can serve as both Attributive Modifier (big boy ) and
Predicative Complement (looks big).
The Complements that Adjectives select are typically PP or

Clause almost never NP.

AdjP can serve as Adjuncts (i.e., Modifiers or Supplements).
When such an Adjunct is fronted, it requires a target of predication.
If no such target is available, we get the dangling modifier effect

useful for diagnosing adjectives:
Away is a Preposition but afraid is an Adjective:
Away from home, John behaved properly. [no target needed]

Afraid of us, John behaved properly. [target is John]
Away from home, there was just work. [target unneeded]
Afraid of us, there was just work. [BAD NO TARGET]

The predicativity test is also useful in distinguishing adjectives from

adverbs. The general principle is: Prepositions and Adverbs always
have some uses as non-predicative modifiers; adjectives never
do.
Again is an Adverb. Awake is an Adjective.
Again, we scarcely knew what to do. [no target needed]

Awake, we scarcely knew what to do. [target is we]
Again, it snowed heavily. [target unneeded]
Awake, it snowed heavily. [BAD NO TARGET]

The strange adjective worth
Does not inflect for grade (despite being short).
Can never be used attributively.
Selects an NP complement, not PP.
The complement is obligatory.

The strange adjective worth
Compare with the typical adjective worthy :
The lexeme worth The lexeme worthy
worth, *worther , *worthest worthy , worthier , worthiest
*Its a worth project. Its a worthy project.

It was worth my time. *It was worthy my time.
*It was worth of my time. It was worthy of my time.
*It is certainly worth. It is certainly worthy.

The words due, like, near , opposite, and unlike are also puzzling, with
a complex mix of Adjective and Preposition properties. But arguments
can be given that
due and opposite were adjectives but have evolved into

prepositions;
like and unlike are sometimes prepositions and sometimes

adjectives with NP or Clause complements;
near may be dually categorized, but is probably prepositional

(though it inflects for grade!).

And now...
Adverbs

Criteria for adverbhood
VP modification may be expressed by NP, PP, or AdvP:
Modifier type NP PP AdvP

L OCATIVE next door in the USA locally
T EMPORAL (next) Monday on Monday soon
D URATIVE a long time for ages lengthily
M ANNER the same way in haste hastily
Modifying a verb is therefore virtually useless as a criterion for

membership in the Adverb category. Yet it is the sole criterion
traditional grammars generally give.

Adverb is a heterogeneous category in the languages that have it. The

adverbs include expressions for
additive and restrictive focusing (also, even, only ),
degree (very , really , nearly ),
aspectuality (still, already , yet),
seriality (again),
connection (however , therefore, thus),
frequency (always, never , often),
modality (perhaps, probably , certainly ),
time (later , soon, recently ),
manner (quickly , easily , better ),
and other semantic categories.

English has a modest number of basic, underived adverbs:
again, almost, already , also, always, anyway , as, even, ever ,

how, however , indeed, just, long, maybe, never , often, only ,
otherwise, perhaps, please, quite, rather , sometimes, soon, still,
therefore, though, thus, too, very , well, yet
A few (early , hard, ill, late, right, well, . . . ) are formed from adjectives
by zero derivation (sometimes with a meaning change). But the vast
majority are derived by suffixation of ly .

But being derived by the ly suffix is not a criterion for adverbhood.

Many adjectives are derived from nouns by means of the same suffix:
beastly , brother ly , cowardly , curmudgeonly , deathly ,

father ly , friendly , kingly , mother ly , musicianly , painter ly ,
princely , queenly , sister ly , sprightly
And there are other ly adjectives like kindly and lively .

Jackendoff, in X Syntax (1977), suggested a key distinction between

adjectives and adverbs:
Jackendoffs complementability criterion

Adjectives take complements; adverbs do not.
Jackendoffs claim is not true.
Jackendoff may have been misled by the small size of the classes of
complement-taking adverbs.
There are several of such classes. They take the same complements
as the related adjectives.

A few examples of complement-taking adverbs (from the WSJ corpus)

operates [AdvP quite [Adv separately ] [PP from the rest of the
company ] ]
the lower portion could move [AdvP [Adv independently ]

[PP of the upper part ] ]
[AdvP [Adv luckily ] [PP for them, ] ] Mr. Keswick decided not to call
their bluff
it will be treated [AdvP no [Adv differently ] [PP from more permanent

trade developments ] ]
an unregulated concern that operated [AdvP [Adv similarly to a bank

trust department ] ]

A note on non-standard zero-derived adverbs
It is well known that many non-standard dialects use words with

adjectival form in adverb roles. Some song lines:
! Love me tender [Ken Darby; attributed to Presley & Matson]
! Treat me nice [Leiber & Stoller]
! Hurts so bad [Randazzo, Wilding & Hart]
! my love does it good [McCartney]
It is commonly thought that this indicates confusion of adjectives and

adverbs. Not true, I believe.

A note on non-standard zero-derived adverbs
What is involved seems to be merely a slight broadening of the class of

zero-derived adverbs in English mentioned above: early , hard, right,
etc.
And (I believe) even in the dialects that have a broader class of such
adverbs, the zero-derived ones are not used as pre-head modifiers:
! Drive real carefully now! ! Drive real careful now!

! She carefully drug it outside. *! She careful drug it outside.
(Sociolinguists: Is this claim correct? Can it be verified from actual

records of non-standard English conversation?)

The categorial exclusivity claim
There is a long history of traditional grammarians and modern linguists

stating this generalization for Standard English:

Adjectives modify nouns; adverbs modify non-nouns.
Almost any reference work at random any English grammar,

dictionary, or terminology guide will repeat this statement.

On this view, adjectives and adverbs (in Standard English) do not

overlap in function:
adjective modifier adverb modifier

modifying a noun a nice man *a nicely man
modifying a verb *she sings nice she sings nicely

The AdjectiveAdverb identity thesis
Given the exclusivity claim, an obvious hypothesis arises.
John Lyons proposed it in 1966: that adverbs are merely

positional variants of the corresponding adjectives.
A decade later Joseph Emonds (1976, 1213) suggested the same

thing: that the adverbs formed with ly are simply adjectives in a
verb-modifying rather than a noun-modifying function.

Andrew Radford (1988) suggests the term advective for the A

category.
And Ingo Plag (2003, p. 196) explicitly asserts that ly is an inflectional

suffix occurring on just those occurrences of adjectives that do not
modify nouns.
Mark Baker (2003, 230257) makes the radical proposal that there are
only three universal categories: N, V, and A. N must refer; V must
assign thematic roles to N; and A must occur only where neither N or V
can.

But it is not in fact true that an adverb can act as Modifier of a category
if and only if an adjective cannot.
We begin by setting aside two distractors.
First, ignore the sky above, the weather outside, the room
downstairs.
Traditionally these would be adverbs modifying nouns. (Henry Sweet

noted this.)
But in fact overwhelming evidence shows that the underlined words

are prepositions (claim to be justified later).

Second, ignore hardly anyone, precisely nothing, almost everybody .
These would appear to have adverbs modifying nouns if

(i) pronouns are a special kind of noun, and
(ii) anyone, nothing, and everybody are pronouns.
But CGEL rejects (ii).
CGEL holds that anyone, nothing, everybody , etc., are compound

determinatives.
And it is normal for determinatives to take adverbs as modifiers.

Harder to set aside are cases that appear to have adjectives

modifying other adjectives, where modification by the related
adverbs either expresses a different sense or is unacceptable.
ADJECTIVE MODIFIER SEMANTICALLY DISTINCT ADVERB

cold sober 6= coldly sober
plain stupid 6 = plainly stupid
bloody good 6= bloodily good
silky smooth 6= silkily smooth

And here the adverb versions seem outright unacceptable:
ADJECTIVE MODIFIER UNACCEPTABLE ADVERB

blind drunk 6= ?blindly drunk
filthy rich 6 = ?filthily rich
pretty cruel 6 = ?prettily cruel
black British 6 = ?blackly British

Another context in which both adjectives and adverbs occur is in the

complement of the copula:
ASCRIPTIVE COMPLEMENT :
The way she dressed was elegant.
( She elegantly performed the act of dressing herself.)
SPECIFYING COMPLEMENT
The way she dressed was elegantly.
( Her dress was elegant in style.)

ASCRIPTIVE COMPLEMENT
It was rude that she spoke to me.
( The fact of her speaking to me constituted rudeness.)
It was rudely that she spoke to me.
( She spoke to me in a rude manner.)

It was rude that she spoke to me.
( The fact of her speaking to me constituted rudeness.)
It was rudely that she spoke to me.
( She spoke to me in a rude manner.)
It was clever that they used flashbacks.
( Using flashbacks was a clever idea.)
It was cleverly that they used flashbacks.
( They used flashbacks in a clever way.)

But the really convincing evidence comes from a construction

mentioned briefly in Huddleston & Pullum, A Students Introduction to
English Grammar (2005, 123n not covered in CGEL).
Adverbs (or AdvPs) can postmodify nouns in NPs:
I express my profound disappointment at [NP the governments

refusal yet again to take the high road and bring forth a motion to
allow parliament to sit in committee of the whole.]
Yet again isnt modifying take the high road; its modifying the noun
refusal: the government has yet again refused. This is an AdvP
modifying a noun.

Other examples:
[NP The NHS and other health organisations internationally] clearly

need methodologies to support benefit analysis of merging healthcare
organisations.
[NP The unique role globally of the Australian Health Promoting Schools
Association, as a non-government organization specifically established
to promote the concept of the health promoting school,] is described.

Other attested examples:
the winner recently of both a Gramophone award and the Royal

Philharmonic Society Award for Best Chamber Ensemble
the people locally
the weather recently
the centerpiece visually of the film

The bottom line:
AdjectiveAdverb Identity is false

It is impossible to sustain the categorial identity claim for the
Adjective and Adverb categories.
Categorial Exclusivity is false

Not even the claim that the sets of modified constituents are
mutually exclusive for Adjective and Adverb can be sustained.
Whatever the facts may be for some other languages, Adj and Adv are
two quite distinct categories in Standard English.

Prepositions, Particles, and Subordinators
Prepositions, Adverbs, and Particles:

The Jespersen/Klima/Fraser/Emonds arguments
Subordinating conjunctions as prepositions:

The Hunter/Jespersen/Geis thesis
The CGEL category of Subordinators

Prepositions, Adverbs, and Particles
What is the lexical category of words such as in, up, and down ?
Every dictionary of English gives the wrong answer.
All say Preposition (prep) and Adverb (adv), and give them separate
entries despite copious semantic overlap.

The relevant dictionary entries for down (ignoring the noun meaning
soft feathers and the verb meaning defeat, drink, etc.) tend to say
this sort of thing:
down adv 1 a from a higher point to or toward the ground or base; b

from a higher to a lower position; 2 in a direction conventionally the
opposite of up . . .
down prep 1 a in a descending direction along; b from a higher to a

lower position upon or within; 2 in a direction conventionally the
opposite of up . . .

CGEL claims that in, on, off , up, down, round, through, etc., are never
adverbs.
Consider how implausible the traditional claim actually is.
It says that these words are adverbs, despite the fact that
(i) they are not derived from adjective stems like most adverbs;

adverbs.
(ii) they are all homophonous with specific prepositions; and

adverbs.
(ii) they are all homophonous with specific prepositions; and
(iii) they are synonymous with those homophonous prepositions.

A syntactic argument against the traditonal view
What is more, these words do not perform the usual modifying

functions of adverbs:
He was bleeding internally.
*He was bleeding in.
She was upwardly mobile.

*She was up mobile.
Mike fell immediately.

Mike fell down.
Mike immediately fell.
*Mike down fell.
These are adverbs without the properties of adverbs!

The Klima/Fraser/Emonds arguments
Several relevant syntactic arguments were given in a classic article of

the classical transformational era (citing earlier observations by
Edward Klima and Bruce Fraser):
Joseph E. Emonds (1972), Evidence that indirect object

movement is a structure-preserving rule.
Foundations of Language 8: 546561.
Reprinted 1973 in Gross, Halle & Schtzenberger (eds.),
The Formal Analysis of Natural Languages (Mouton).

1. Subcategorization of intransitives
Some intransitive verbs like glance and dart syntactically select a
directional phrase as an obligatory complement:
The lizard darted into that hole.

The lizard darted.
These so-called adverbs can fill that complement role:
The lizard darted in.
But in general adverbs cannot:

The lizard darted immediately.

2. Subcategorization of transitives
Certain transitive verbs like put and sneak select a directional phrase
as obligatory second complement:
Well have to sneak some beer into the dorm.

Well have to sneak some beer.
These so-called adverbs can fill in that complement role:
Well have to sneak some beer in.
But in general adverbs cannot:
Well have to sneak the beer quietly.

Interim conclusion
Conclusion from arguments 1 and 2:
in and similar words are not adverbs.

3. Into the bin with it etc. These constructions involve a directional

PP (underlined) followed by a with-headed PP:
Into the bin with it!

Up the stairs with you!
Some so-called adverbs can also occur as the directional, but

adverbs in general cannot:
Down with capitalist greed!

Out with the old, in with the new.
*Fiercely with capitalist greed!
*Slowly with the old, immediately with the new.

4. Fronted directional PPs with verbs of motion
Directional PPs switch with subjects of verbs of motion:
The terrified townspeople ran into the church.
Into the church ran the terrified townspeople.
The same alternation is found with the so-called adverbs, but not with
adverbs in general:
The terrified townspeople ran in.
In ran the terrified townspeople.
The terrified townspeople ran immediately.

Immediately ran the terrified townspeople.

5. Modification with right
Certain items such as right, straight, and bang occur in modern
Standard English as pre-head modifiers only with prepositions:
There is a McDonalds right dominating the square.
There is a McDonalds right the same place.
There is a McDonalds right local.
There is a McDonalds right locally.
There is a McDonalds right on our doorstep.
Yet sentences like these are fine:
She leaned over and fell right in.

The price went right down the next day.
Does this road go right through?

Interim conclusion
Conclusion from arguments 3, 4, and 5:
in and similar words head preposition phrases (PPs).

An additional argument
6. In and out
The traditional analysis has to put in and out, counterintuitively, in
entirely different categories:
It looks like were in luck. [in + NP]

It looks like were out luck. [*out + NP]
It looks like were out of luck. [out + PP]
I keep my diamonds in the safe. [in + NP]

I left my diamonds out the safe. [*out + NP]
I left my diamonds out of the safe. [out + PP]
Well be in trouble if he leaves. [in + NP]

Well be out trouble if he leaves. [*out + NP]
Well be out of trouble if he leaves. [out + PP]

Prepositions that never have NP complements
We can now see that many items are best categorized as prepositions
even though they are strictly intransitive, i.e., they never take NP
complements:
The lizard darted [PP away. ]
Well have to sneak some beer [PP back. ]
Lets hold the meeting [PP right here. ]
Make sure you come [PP straight home. ]
It went [PP right outside. ]
Dear Lord, grant me your precious gift of patience preferably

[PP right now. ]

Prepositions that never have NP complements
Words like ahead and abroad are prepositions that never take NP
complements, whereas again is an adverb. And it is easy to
demonstrate this:
In 2008 he got arrested again. [Adv at end of VP]

In 2008 he again got arrested. [Adv OK before V]
In 2008 he got arrested abroad. [PP at end of VP]

*In 2008 he abroad got arrested. [PP bad before V]
Conclusion: abroad is not an Adverb.

Particles
And what of particles, in the sense of the verb-particle construction?
They are light intransitive PPs with special linear positioning

privileges.
PARTICLES NOT PARTICLES

away at (strictly transitive)
back between
down downstream
in indoors
out outside
up upstairs

Particles
Plausible structures for a simple verb-particle construction:
VP VP

V PP NP V NP PP

take P+ the trash take the trash P+
out out

The traditional category subordinating conjunction
Traditional grammar recognizes a category of conjunctions divided

into subordinating conjunctions and coordinating conjunctions.
Cogent arguments were presented against the composition of the

former category at least 225 years ago.
John Hunter argued before the Royal Society of Edinburgh in 1784 that
it was not sensible for after to be placed in different categories in these
two examples:
I came after he departed.

I came after his departure.

The idea Hunter advocated was argued again by Jespersen 140 years
later (The Philosophy of Grammar, 1924) and again half a century
after that by Michael Geis (MIT Ph.D. dissertation, 1970):
Different verbs take different kinds of complement

(NP complement, Clause complement, no complement, etc.), yet,
Jespersen observes, no one thinks of assigning them to different
parts of speech.
Prepositions should be treated analogously, as subcategorized for

complements of different types.
Dual categorization should be resorted to only when the

grammatical facts require it.

Jespersen even finds a sentence in Thackeray where after has a

complement consisting of an NP coordinated with a content clause:
After [ the Baden business and he had dragged off his wife to
Champagne ], the Duke became greatly broken.
CGEL (1327, [15iiib]) cites a better-sounding and more recent attested

case, where the clause comes first:
After [their rubber plantation failed, and
her husbands death on the Upper Rewa in 1885 ], she maintained
her three young children with a tiny store.
Given such examples, dividing after into preposition and conjunction

instances becomes actually impossible.

Genuine multiple categorization
A word like cosy really does have three lexical entries:
a tea cosy (N) SINGULAR PLURAL

PLAIN cosy cosies
GENITIVE cosys cosies
keep us cosy (Adj) PLAIN COMPARATIVE SUPERLATIVE

cosy cosier cosiest
cosy up to her (V) PLAIN FORM cosy

3 RD SING . PRESENT cosies
PRETERITE , PAST PARTICIPLE cosied
GERUND - PARTICIPLE cosying

Spurious multiple categorization
But it is syntactic madness to claim that since belongs to three different

categories, which is what the traditional account does.
Ive loved her ever since [adv ]
Ive loved her ever since our first meeting [prep]
Ive loved her ever since we first met [conj]

Spurious multiple categorization
Typical dictionary information for since:
since adv
1 : continuously from a time in the past until the present . . .
2 : before the present time
3 : after a time in the past
since prep
1 : in the period after a specified time in the past
2 : continuously from a specified time in the past
since conj
1 : after the time in the past when
2 : up to the present time from the time in the past when
3 : for the reason that; because of the fact that special

A better account of since
Rough draft of a more sensible dictionary entry:
since prep
1 continuously from a certain time in the past until the present, the time
being specified by NP complement (empty since March) or Clause
complement (empty since they left) or the context (empty ever since
empty ever since that time).
In older usage, may modify a verb (a wall, since removed), or (espe-
cially if modified by long) may mean simply earlier than the present
(abandoned long since).
2 (with Clause complement only) for the reason that, or because of
the fact that (We must assume it, since there is no other explanation).

The category of Subordinators
Not all traditional subordinating conjunctions are prepositions.

CGEL posits a small class of subordinators.
a small class (half a dozen items)
markers of syntactic subordination
no independent meaning
often optional or substitutable
never function as Head
(Complementizer is a bad name for these. They dont always form

Complements. Modifiers such as relative clauses are also introduced
by Subordinators.)

Recapitulation
Notice that in these lectures I am rejecting both

the DP hypothesis (that the determinative rather than the nominal
body is Head in the king of France) and
the CP hypothesis (that the subordinator rather than the clausal
body in that nobody cares).
Nouns, Verbs, Adjectives, Adverbs, Prepositions, and Determinatives

head NPs, VPs, AdjPs, AdvPs, PPs, and DPs, respectively (though
the king of France is not a DP).
Subordinators and Coordinators never head phrases at all.

Declarative clause subordinator: that
default marker of a declarative content clause
has no meaning at all
normally unaccented
usually optional when marking an internal complement, but not

when marking a Subject clause
obligatory in that-relative before Subject gap
never functions as Head (lexical items select tensed or plain form

verbs, but not simply that as subordinator)

Interrogative clause subordinator: whether
marks a clause as an interrogative content clause
no clear meaning independently of clause interrogativity (?)
can be replaced by if when marking an internal complement, but

not when marking a Subject clause or in the whether or not
exhaustive conditional construction
never functions as Head

Other subordinators:
for when initial in infinitival clauses that have subjects for you to do
a thing like that (the word is also a preposition, of course, in other
contexts)
if in extraposed irrealis declarative content clauses, e.g. It would

be great if he were able to join us (compare *If he were able to join
us would be great) but note that conditional if is a preposition
possibly infinitival to also fits in this category (but would be an

anomalous member)

Nothing is gained by forcing all Subordinators into the P category (as

Emonds 1985 proposes).
But the prepositions traditionally called subordinating conjunctions

(although, because, conditional if , lest, since, though, unless, etc.) are
different from Subordinators:
independent lexical meaning
cannot be omitted or replaced without altering meaning
select complements the way Heads of phrases do

These prepositions all differ in complement selection:
OBJECT of - BARE that- PRED NO

NP PP CLAUSE CLAUSE COMP COMP
although + +
at +
because + +
down + +
given + +
lest +
out (+) + +
provided + +
since + + +

The difference with Subordinators, if CGEL has it right, is that they

dont select anything. They have no subcategorizational properties,
and take no complements.
They are meaningless markers of subordinate clause status, often

optional, quite idiosyncratic in syntactic behavior.
In X-bar theory terms, they are exceptions to the general principle for
categories in the lexicon: they do not found bar-level projections, they
are never heads.
In The staff were sure that the students were happy, the head of the
underlined complement clause is not that; it is the students were
happy .

Relative Constructions and Unbounded Dependencies
The typology of subordinate clauses
Types of relative clause
The diversity of wh-phrase uses
Unbounded dependencies
The implausibility of movement

The traditional view (shown here in green to remind you that it is

completely wrong) is that subordinate clauses are of three types:
Nominal or noun clauses: act like nouns can be subjects
and objects
Adjectival or adjective clauses: act like adjectives they
modify nouns
Adverbial or adverb clauses: act like adverbs they modify
verbs
This whole classification is a mistake.

First, this is another classification based on function.
As such, it does not separate clauses appropriately.
A clause like that he was guilty can belong either to the nominal
clauses or the adverbial ones:
That he was guilty seemed obvious. [nominal]
I was disappointed that he was guilty. [adverbial]
Why posit two kinds of clause, here rather than two functions
(external complement [= Subject] and internal Complement)?

Notice also:
Caroline was late, as she so often is. [adverbial]
Caroline was late, as her car wouldnt start. [adverbial]
Here there are two kinds of clause:

she so often is (structurally incomplete)
her car wouldnt start (structurally complete)
They are traditionally classified the same as adverbial because the

intuitively conceived function is modifying the adjective late.

Nominal or noun clauses arent like nouns.
They can be subjects, though there are limits:

*Is that he doesnt like you a problem?
They are not (normally) direct objects:

*He denied categorically responsibility.
He denied categorically that he was responsible.

Nominal or noun clauses arent like nouns.
They are never objects of prepositions:

*I dont approve of that you smoke.
Some verbs take nominal clauses but not nouns:

I used to marvel that he could remember all our names.
*I used to marvel his memory for names.
I wonder whether anyone cares.
*I wonder peoples apathy.

Adjectival clauses arent like adjectives:
They are never allowed prenominally (*a that I like car ).
They dont take adjective modifiers (*very that I like).
and so on.

Adjectival clauses arent like adjectives:
They are never allowed prenominally (*a that I like car ).
They dont take adjective modifiers (*very that I like).
and so on.
And adverbial clauses arent clauses at all, but PPs.
Nothing about the traditional classification really works.

The right classification of English subordinate clauses:
FINITE SUBORDINATE CLAUSES

content clauses
relative clauses
comparative clauses
NON - FINITE SUBORDINATE CLAUSES

infinitival clauses
participial clauses


content clauses
declarative interrogative exclamative
relative clauses
wh-relatives th-relatives bare relatives
comparative clauses
scalar inequality scalar equality
non-scalar inequality non-scalar equality


content clauses
declarative interrogative exclamative
relative clauses
wh-relatives th-relatives bare relatives
comparative clauses
scalar inequality scalar equality
non-scalar inequality non-scalar equality

Content clauses illustrated:
declarative content clause

(that) no one had ever told me
interrogative content clause

closed interrogative:
whether anyone could ever survive a trip to Mars
open interrogative:
what surviving a trip to Mars would call for
exclamative content clause

what a great accomplishment it has been

The classification of relative clauses
Formal types of relative clauses illustrated:
wh-relative clause:
which they had been invited to [NP ]
to which they had been invited [PP ]
th-relative clause:
that they had been invited to [NP ]
bare relative clause:

they had been invited to [NP ]

Relational types of relative clauses
integrated relative clause:

a party which we had been invited to [NP ]
a party to which we had been invited [PP ]
supplementary relative clause:

their party, which we had been invited to [NP ]
cleft relative clause:

It was the anthropology department
whose party we had been invited to [NP ]
fused relative construction:

I protested against what we had been invited to [NP ]

Which formal types can instantiate which relational types?
INTEGRATED SUPPLEMENTARY CLEFT FUSED
wh- yes yes yes yes
th- yes (very rare) yes no
bare yes no yes no

OPEN INTEGR . SUPPL . FUSED

INTERROG . RELATIVE RELATIVE RELATIVE
who X X X ?*
whom X X X
[+hum] whose X X X
[hum] whose X ?
what X X
which X X X
where X X X X
when X X X ?X
how X ?X
why X X ?
while X

Whose with non-human reference:

open interrogative: I asked whose hinges squeaked.
integrated relative: Oil any door whose hinges squeak.
supplementary relative: ? that door, whose hinges squeaked
fused relative: Oil whose hinges squeak.
What:
open interrogative: I didnt know what he was doing.
integrated relative: the things what he was doing
supplementary relative: his idea, what he had not tested
fused relative: I was suspicious about what he was doing.

Why :
open interrogative: I didnt know why he was yelling.
integrated relative: the reason why he was yelling
supplementary relative: my painful foot, why I was yelling
fused relative: I was suspicious about why he did it.
How:
open interrogative: I didnt know how he would respond.
integrated relative: the way how he would respond
supplementary relative: the usual way, how I do it
fused relative: I was suspicious about how he did it.

Unbounded dependencies: syntactic structure
The structure of We are deeply indebted to him:

Clause

Subj: Head:
NP VP

we Head: Comp:
V AdjP

are Adjunct: Head:
Adv AdjP

deeply Head: Comp:
Adj PP

indebted Head: Comp:
P NP
to him

Representing grammatical structure
The structure of Him, we are deeply indebted to :
Clause

Prenucleus: Nucleus:
NP Clause

him Subj: Head:
NP VP

we Head: Comp:
V AdjP

are Adjunct: Head:
Adv AdjP

deeply Head: Comp:
Adj PP

indebted Head: Comp:
P NP
to

A hypothesis about plain vanilla unbounded dependencies:
When the path is all Head and Comp links, unbounded dependencies
are always entirely grammatical.
When Adjunct interrupts the path, the result is always somewhat

ungrammatical.

How many paths are there to describe?

How far from the fronted complement (e.g., what)
can the stranded preposition be, in grammatical sentences?
I dont know what the FBI looked at .


I dont know what you imagine the FBI looked at .


I dont know what you imagine he thinks the FBI looked at .


I dont know what you imagine he thinks shell say the FBI
looked at .


looked at .
I dont know what you imagine he thinks shell say we told them
the FBI looked at .


looked at .
I dont know what you imagine he thinks shell say we told them
the FBI looked at .
Clearly there is no grammatically fixed limit.

Model-theoretic syntax
We can use a description language based on symbolic logic to express

statements about the structure of trees directly.


(x, y )[Clause(x) VP(y )]

There is a point labeled Clause and a point labeled VP in the tree.



(x)[Clause(x) (y )[(y < x)]]

There is a point labeled Clause right at the top of the tree.
(Note: y < x means that y is immediately above x and linked to it.)



(x)[Clause(x) (y )[(y < x)]]

There is a point labeled Clause right at the top of the tree.
(Note: y < x means that y is immediately above x and linked to it.)
(x, y , z)[((z < x) (z < y ) (x 6= y ) (NP(x)) (VP(y ))) (x y )]]
An NP node (x) precedes a VP node (y ) if they are both immediately

under some third node (z).

In a more expressive type of language known as monadic

second-order logic (MSO) we can also say things like this:


there is a set of points Z that are all on one branch


there is a set of points Z that are all on one branch

(Z )(x, y )[(Z (x) Z (y )) ((x < y ) (y < x))]

Z is a continuous chain of points


(x, y , z)[((Z (x)) (Z (z)) (x < y ) (y < z)) Z (y )]


(x, y , z)[((Z (x)) (Z (z)) (x < y ) (y < z)) Z (y )]
Z is a chain of Comp and Head links


(x, y , z)[((Z (x)) (Z (z)) (x < y ) (y < z)) Z (y )]
Z is a chain of Comp and Head links

(x, y )[(Z (x) Z (y ) x < y ) (Head(x, y ) Comp(x, y ))]

So in MSO we can give a very precise definition of the following notion

of a properly located vacancy:
A vacancy is properly located in a clause if and only if there is a
continuous chain of Comp and Head links leading down to it from the
top of the clause.


top of the clause.
For non-standard dialects (which allow Wheres it at?):
Prenucleus where can accompany a clause with a properly
located noun phrase vacancy that immediately follows at.


top of the clause.
For non-standard dialects (which allow Wheres it at?):
Prenucleus where can accompany a clause with a properly
located noun phrase vacancy that immediately follows at.
For all dialects:
a Prenucleus noun phrase can accompany a clause with any
properly located noun phrase vacancy.

Three pieces of computational good news about MSO:


1 D ECIDABLE SATISFIABILITY
For any description in an MSO language, a computer program can
figure out whether it is consistent whether there could be a tree
that conforms to that description.


2 T REE RECOGNIZABILITY
A property of trees is describable in MSO if and only if a
tree-crawling bug with finite memory could recognize that property.


2 T REE RECOGNIZABILITY
A property of trees is describable in MSO if and only if a
tree-crawling bug with finite memory could recognize that property.
3 S TRING RECOGNIZABILITY
If a grammar describing some set of sentences can be given as
an MSO description of a set of trees, then for any string of words it
can be determined by a computer program (quite rapidly) whether
it is grammatical or not.

Winter School in Language and Speech

Hochgeladen von

Dokumentinformationen

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Winter School in Language and Speech

Hochgeladen von

Copyright:

Verfügbare Formate

Winter School in Language and Speech

Geoffrey K. Pullum (Edinburgh) WSLST 2012 Jan 2012 1 / 163

What we mean by English: grammar, style, dialects, etc.

Geoffrey K. Pullum (Edinburgh) WSLST 2012 Jan 2012 2 / 163

The Cambridge Grammar of the English Language

Geoffrey K. Pullum (Edinburgh) WSLST 2012 Jan 2012 3 / 163

The Cambridge Grammar of the English Language

Useful elementary textbook based on CGEL:

Geoffrey K. Pullum (Edinburgh) WSLST 2012 Jan 2012 3 / 163

Geoffrey K. Pullum (Edinburgh) WSLST 2012 Jan 2012 4 / 163

Verbs and the AUXILIARY/LEXICAL distinction

The verbal inflectional system

The Catenative-Auxiliary analysis

Geoffrey K. Pullum (Edinburgh) WSLST 2012 Jan 2012 5 / 163

Two syntactic classes of verbs (they are not semantic!):

A special subsets, the auxiliary verbs

The rest, called the lexical verbs

Warning: The traditional definition of auxiliary verbs

Geoffrey K. Pullum (Edinburgh) WSLST 2012 Jan 2012 6 / 163

The unique paradigm of be:

Geoffrey K. Pullum (Edinburgh) WSLST 2012 Jan 2012 7 / 163

The wrong wrong wrong way to describe regular verbs:

Geoffrey K. Pullum (Edinburgh) WSLST 2012 Jan 2012 8 / 163

The disastrous threesome lie/lied/lied (tell untruths), lie/lay /lain (be

plain present lie lie lay

Geoffrey K. Pullum (Edinburgh) WSLST 2012 Jan 2012 9 / 163

The six paradigmatic forms that have to be recognized

Geoffrey K. Pullum (Edinburgh) WSLST 2012 Jan 2012 10 / 163

The distinction between finite and non-finite clauses ALMOST BUT

VERB FORM CONSTRUCTION CLAUSE

Geoffrey K. Pullum (Edinburgh) WSLST 2012 Jan 2012 11 / 163

The distinction between finite and non-finite clauses ALMOST BUT

VERB FORM CONSTRUCTION CLAUSE

Geoffrey K. Pullum (Edinburgh) WSLST 2012 Jan 2012 12 / 163

But not all forms have different shapes:

Geoffrey K. Pullum (Edinburgh) WSLST 2012 Jan 2012 13 / 163

There are actually no verbs with exactly 6 shapes.

Some irregular verbs (with past participle identical to plain form)

3rd sing present gerund-participle

Geoffrey K. Pullum (Edinburgh) WSLST 2012 Jan 2012 14 / 163

Regular verbs, like bake, have 4 shapes

3rd sing present gerund-participle

Geoffrey K. Pullum (Edinburgh) WSLST 2012 Jan 2012 15 / 163

Several dozen verbs with irregular past participle suffixes,

3rd sing present gerund-participle

Geoffrey K. Pullum (Edinburgh) WSLST 2012 Jan 2012 16 / 163

And now for the auxiliary verbs. As we saw, be

3rd sing present gerund-participle

was(nt) / were(nt) been

The negative forms really need a separate column.

Geoffrey K. Pullum (Edinburgh) WSLST 2012 Jan 2012 17 / 163

The negative forms really are words, not word sequences

(Zwicky & Pullum, Cliticization versus inflection: English nt,

Key syntactic evidence: single word inversion in interrogatives

We should go. Should we go?

Geoffrey K. Pullum (Edinburgh) WSLST 2012 Jan 2012 18 / 163

have havent having

have havent have

Geoffrey K. Pullum (Edinburgh) WSLST 2012 Jan 2012 19 / 163

Auxiliary do is similar, except there are no grounds for saying it has a

Geoffrey K. Pullum (Edinburgh) WSLST 2012 Jan 2012 20 / 163

preterite neutral preterite negative *

Geoffrey K. Pullum (Edinburgh) WSLST 2012 Jan 2012 21 / 163

The modal may is defective in one more form (present negative):