                       A GUIDE TO LINK TYPES
                          Davy Temperley

INTRODUCTION. This file contains an alphabetical list of all link
types used in the dictionary "2.0.dict". All subscript types are
explained also; however, only the actual link names are listed
alphabetically. (For example, the subscript type "Ds" will be
explained somewhere in the entry for "D".) All post-processing
features are also explained, under the entries for the links that are
involved.

Since the original release of the parser, a number of the link-names
have been changed. Our aim has been to make the link-names and
subscript names as mnemonically significant as possible, while still
keeping them concise. In many cases the mnemonic significance is
obvious: "A" stands for adjective, "D" stands for determiner, and so
on. In a few cases we have adopted a character to arbitrarily stand
for something in a number of different link-names. "G" generally
stands for proper nouns; "B" stands for fronted objects; "J" stands
for prepositional objects; "E" stands for adverbs; and "X" stands for
punctuation.

The first section below contains brief (1-2 sentence) descriptions
of each link type. The second section contains much more thorough
explanations.


                      I. LINK TYPES AT A GLANCE


A connects pre-noun ("attributive") adjectives to following nouns:
	"The BIG DOG chased me", "The BIG BLACK UGLY DOG chased me".

AA is used in the construction "How [adj] a [noun] was it?". It
	connects the adjective to the following "a".

AF connectives adjectives to verbs in cases where the adjective
	is fronted, such as questions and indirect questions: "How
	BIG IS it?"

AL connects a few determiners like "all" or "both" to following
	determiners: "ALL THE people are here".

AN connects noun-modifiers to following nouns: "The TAX PROPOSAL
	was rejected".

AZ connects the word "as" back to certain verbs that can take 
	"[obj] as [adj]" as a complement: "He VIEWED him AS stupid".

B serves various functions involving relative clauses and questions.
	It connects transitive verbs back to their objects in cases
	like relative clauses and questions ("WHO did you HIT?"); it
	also connects the main noun to the finite verb in subject-type
	relative clauses ("The DOG who CHASED me was black").

BI connects form of the verb "be" to certain idiomatic expressions:
	for example, cases like "He IS PRESIDENT of the company".

BT is used with time expressions acting as fronted objects: "How many
	YEARS did it LAST?".

BW connects "what" to various verbs like "think", which are not really
	transitive but can connect back to "what" in questions: "WHAT
	do you THINK?"

C links conjunctions to subjects of subordinate clauses ("He left 
	WHEN HE saw me"). it also links certain verbs to subjects
	of embedded clauses ("He SAID HE was sorry").

CC connects clauses to following coordinating conjunctions ("SHE left
	BUT we stayed").

CO connects "openers" to subjects of clauses: "APPARENTLY / ON Tuesday ,
	THEY went to a movie".

CQ connects to auxiliaries in comparative constructions involving
	s-v inversion: "SHE has more money THAN DOES Joe".

CX is used in comparative constructions where the right half of the
	comparative contains only an auxiliary: "She has more money
	THAN he DOES".

D connects determiners to nouns: "THE DOG chased A CAT and SOME BIRDS".

DD connects definite determiners ("the", "his") to number expressions
	certain things like number expressions and adjectives acting
	as nouns: "THE POOR", "THE TWO he mentioned".

DG connects the word "The" with proper nouns: "the Riviera", "the
	Mississippi".

DP connects possessive determiners to gerunds: "YOUR TELLING John to
	leave was stupid".

DT connects determiners to nouns in idiomatic time expressions: "NEXT
	WEEK", "NEXT THURSDAY".

E is used for verb-modifying adverbs which precede the verb: "He 
	APPARENTLY not COMING".

EA connects adverbs to adjectives: "She is a VERY GOOD player".

EB connects adverbs to forms of "be" before an object or prepositional
	phrase: "He IS APPARENTLY a good programmer".

EC connects adverbs to comparative adjectives: "It is MUCH BIGGER"

EE connects adverbs to other adverbs: "He ran VERY QUICKLY".

EF connects the word "enough" to preceding adjectives and adverbs: "He
	didn't run QUICKLY ENOUGH".

EI connects a few adverbs to "after" and "before": "I left SOON AFTER 
	I saw you".

EN connects certain adverbs to expressions of quantity: "The class has
	NEARLY FIFTY students".

ER is used the expression "The x-er..., the y-er...". it connects the
	two halfs of the expression together, via the comparative words
	(e.g. "The FASTER it is, the MORE they will like it").

FM connects the preposition "from" to various other prepositions: "We
	heard a scream FROM INSIDE the house".

G connects proper noun words together in series: "GEORGE HERBERT WALKER
	BUSH is here."

GN (stage 2 only) connects a proper noun to a preceding common noun 
	which introduces it: "The ACTOR Eddie MURPHY attended the event".

H connects "how" to "much" or "many": "HOW MUCH money do you have".

I connects certain words with infinitive verb forms, such as modal 
	verbs and "to": "You MUST DO it", "I want TO DO it".

IN connects the preposition "in" to certain time expressions: "We did it
	IN DECEMBER".

J connects prepositions to their objects: "The man WITH the HAT is here".

JG connects certain prepositions to proper-noun objects: "The Emir
	OF KUWAIT is here".

JQ connects prepositions to question-word determiners in "prepositional 
	questions": "IN WHICH room were you sleeping?"

JT connects certain conjunctions to time-expressions like "last week":
	"UNTIL last WEEK, I thought she liked me".

K connects certain verbs with particles like "in", "out", "up" and the
	like: "He STOOD UP and WALKED OUT".

L connects certain determiners to superlative adjectives: "He has THE
	BIGGEST room".

LE is used in comparative constructions to connect an adjective to the
	second half of the comparative expression beyond a complement
	phrase: "It is more LIKELY that Joe will go THAN that Fred will 
	go".

M connects nouns to various kinds of post-noun modifiers: prepositional
	phrases ("The MAN WITH the hat"), participle modifiers ("The
	WOMAN CARRYING the box"), prepositional relatives ("The MAN
	TO whom I was speaking"), and other kinds.

MG allows certain prepositions to modify proper nouns: "The EMIR OF
	Kuwait is here".

MV connects verbs and adjectives to modifying phrases that follow, like
	adverbs ("The dog RAN QUICKLY"), prepositional phrases ("The dog
	RAN IN the yard"), subordinating conjunctions ("He LEFT WHEN he
	saw me"), comparatives, participle phrases with commas, and 
	other things.

MX connects modifying phrases with commas to preceding nouns: "The DOG,
	a POODLE, was black". "JOHN, IN a black suit, looked great".
	
N connects the word "not" to preceding auxiliaries: "He DID NOT go".

ND connects numbers with expressions that require numerical determiners:
	"I saw him THREE WEEKS ago".

NF is used with NJ in idiomatic number expressions involving "of":
	"He lives two THIRDS OF a mile from here".

NI is used in a few special idiomatic number phrases: "I have BETWEEN
	5 AND 20 dogs".

NN connects number words together in series: "FOUR HUNDRED THOUSAND people
	live here".

NR connects fraction words with superlatives: "It is the THIRD BIGGEST
	city in China".

NS connects singular numbers (one, 1, a) to idiomatic expressions requiring
	number determiners: "I saw him ONE WEEK ago".

NW is used in idiomatic fraction expressions: "TWO THIRDS of the students
	were women".

O connects transitive verbs to their objects, direct or indirect: "She
	SAW ME", "I GAVE HIM the BOOK".

OD is used for verbs like "rise" and "fall" which can take expressions
	of distance as complements: "It FELL five FEET".

OF connects certain verbs and adjectives to the word "of": "She ACCUSED
	him OF the crime", "I'm PROUD OF you".

OT is used for verbs like "last" which can take time expressions as
	objects: "It LASTED five HOURS".

P connects forms of the verb "be" to various words that can be its
	complements: prepositions, adjectives, and passive and
	progressive participles: "He WAS [ ANGRY / IN the yard / CHOSEN /
	RUNNING ]". 

PF is used in certain questions with "be", when the complement need of
	"be" is satisfied by a preceding question word: "WHERE ARE you?",
	"WHEN will it BE?"

PP connects forms of "have" with past participles: "He HAS GONE".

Q is used in questions. It connects the wall to the auxiliary in simple
	yes-no questions ("///// DID you go?"); it connects the question
	word to the auxiliary in where-when-how questions ("WHERE DID
	you go").

QI connects certain verbs and adjectives to question-words, forming 
	indirect questions: "He WONDERED WHAT she would say".

R connects nouns to relative clauses. In subject-type relatives, it
	connects to the relative pronoun ("The DOG WHO chased me was
	black"); in object-type relatives, it connects either to the
	relative pronoun or to the subject of the relative clause 
	("The DOG THAT we chased was black", "The DOG WE chased was
	black").

RS is used in subject-type relative clauses to connect the relative
	pronoun to the verb: "The dog WHO CHASED me was black".

RW connects the right-wall to the left-wall in cases where the right-wall
	is not needed for punctuation purposes.

S connects subject nouns to finite verbs: "The DOG CHASED the cat":
	"The DOG [ IS chasing / HAS chased / WILL chase ] the cat".

SF is a special connector used to connect "filler" subjects like "it"
	and "there" to finite verbs: "THERE IS a problem", "IT IS
	likely that he will go".

SFI connects "filler" subjects like "it" and "there" to verbs in
	cases with subject-verb inversion: "IS THERE a problem?",
	"IS IT likely that he will go?"

SI connects subject nouns to finite verbs in cases of subject-verb
	inversion: "IS JOHN coming?", "Who DID HE see?"

TA is used to connect adjectives like "late" to month names: "We did
	it in LATE DECEMBER".

TD connects day-of-the-week words to time expressions like "morning":
	"We'll do it MONDAY MORNING".

TH connects words that take "that [clause]" complements with the word
	"that". These include verbs ("She TOLD him THAT..."), nouns
	("The IDEA THAT..."), and adjectives ("We are CERTAIN THAT").

TI is used for titles like "president", which can be used in certain
	cirumstances without a determiner: "AS PRESIDENT of the company,
	it is my decision".

TM is used to connect month names to day numbers: "It happened on
	JANUARY 21".

TO connects verbs and adjectives which take infinitival complements to
	the word "to": "We TRIED TO start the car", "We are EAGER TO
	do it".

TQ is the determiner connector for time expressions acting as fronted
	objects: "How MANY YEARS did it last".

TS connects certain verbs that can take subjunctive clauses as
	complements - "suggest", "require" - to the word that: "We
	SUGGESTED THAT he go".

TY is used for certain idiomatic usages of year numbers: "I saw him on
	January 21 , 1990 ". (In this case it connects the day number to
	the year number.)

U is a special connector on nouns, which is disjoined with both the
	determiner and subject-object connectors. It is used in idiomatic
	expressions like "What KIND_OF DOG did you buy?"

UN connects the words "until" and "since" to certain time phrases like
	"after [clause]": "You should wait UNTIL AFTER you talk to me".

V connects various verbs to idiomatic expressions that may be non-adjacent:
	"We TOOK him FOR_GRANTED", "We HELD her RESPONSIBLE".

W connects the subjects of main clauses to the wall, in ordinary 
	declaratives, imperatives, and most questions (except yes-no 
	questions). It also connects coordinating conjunctions to
	following clauses: "We left BUT SHE stayed".

WN connects the word "when" to time nouns like "year": "The YEAR WHEN
	we lived in England was wonderful".

WR connects the word "where" to a few verbs like "put" in questions like
	"WHERE did you PUT it?".

X is used with punctuation, to connect punctuation symbols either to
	words or to each other. For example, in this case, POODLE
	connects to commas on either side: "The dog , a POODLE , was 
	black."

Y is used in certain idiomatic time and place expressions, to connect
	quantity expressions to the head word of the expression: "He 
	left three HOURS AGO", "She lives three MILES FROM the station".

YP connects plural noun forms ending in s to "'" in possessive
	constructions: "The STUDENTS ' rooms are large".

YS connects nouns to the possessive suffix "'s": "JOHN 'S dog is black".

Z connects the preposition "as" to certain verbs: "AS we EXPECTED, he was
	late".


		II. A DETAILED EXPLANATION OF LINK TYPES


A       connects pre-noun ("attributive") adjectives to nouns. Any
        number of adjectives can be used; all connect to the noun.

                     +----A----+
                     |    +-A--+
        	     |    |    |
        	The big black dog ran 
        
        Nouns thus have optional "@A-" connectors, conjoined with "D-"
        connectors and with their main "S/O/J" complex.

        Many adjectives take complements such as clauses or
        infinitival phrases; but such complements may not be used
        when the adjective is being used prenominally ("The man was
        eager to go", "*The eager to go man is here"). Thus the
        A- connectors on adjectives must be disjoined with complement
        connectors like TO+, TH+, and QI+.

        Some adjectives, such as superlatives and number adjectives
        ("biggest", "first"), must be used with a definite determiner
        such as "The" or "His". See "L".

        All hyphenated expressions are treated as adjectives, and
        may be used either attributively or predicatively; they
        therefore carry both A+ and Pa- (e.g., "bone-headed").

        A few adjectives can act only as predicative adjectives, not
        prenominal ones ("asleep", "alone"). These have no A+
        connector. 

        Many participles can also act as prenominal adjectives; these,
        also, have A+ connectors. The situation here is complicated.
        Present participles of intransitive verbs have A+ connectors;
        those of transitive verbs do not ("The sleeping child", "*The
        hitting child"). Many passive participles have A+ connectors;
        this applies only to transitive verbs, since only transitive
        have passive forms ("The destroyed building"). (A few
        intransitive past participles also take A+: "The fallen
        horse".) Past participles of complex verbs which require
        phrasal complements do not carry A+ (*"The hoped agreement");
        those of complex verbs which can take direct objects carry A+
        in some cases, not in others ("The reported incident", "*The
        seen man"). (Note: Do not confuse participle-adjectives with
        full-fledged adjectives which happen to be participles, like
        "amused" and "annoying": "It is very annoying", "*It is very
        destroyed". Full-fledged adjectives can take modifiers such as
        "very"; see "EA".)

        A+ connectors on adjectives (but not participle-adjectives)
        are conjoined with an optional Xc+; this allows commas to be
        inserted after any prenominal adjective in a list ("The big,
        bad, ugly bear"). (This also allows doubtful usages such as
        "The big, bear", "The big black ugly, bear".) See "Xc".

        Attributive adjectives are also allowed on proper nouns with 
        cost 2. See "D: Determiners and Adjectives on Proper Nouns".

AA      is used in the construction "How big a dog was it?"

                 +---HA---+    
                 +EAh+-AA-+-Ds-+
                 |   |    |    |
                How big   a   dog was it        

        The article "a" thus carries "Ds+ & {AA- & HA-}".  Adjectives
        carry "AA+", disjoined with their A+ (used in prenominal
        adjectives) and "Pa-" (used in predicative adjectives).

        This construction uses the "EA-" on adjectives, which is
        also used for modifying adverbs like "very" and for
        questions like "How big is it". Similarly, the "EAh+" on "how"
        is also used for ordinary adjectival questions like "How big
        is it". "How" therefore carries "EAh+ & {HA+}". Since the HA+
        is optional, there is a danger of the unwanted construction
        "*How big dogs were they". This is prohibited in
        post-processing: see "EAh".

AF 	connects adjectives to verbs in cases where the adjective
        is "fronted", such as questions and indirect questions.

                     +-AF-+                    +------AF---------+
        	     |    |                    |                 |
 		How big  is it   I wonder how big he wants it to be

        Verbs that can take adjectival complements, like "be", "seem",
        and "make [obj]", have AF- disjoined with their Pa+ connectors
        (and any other complement connectors).  Adjectives have AF+
        connectors disjoined with their A+ and Pa-. The constructions
        above can only be used with the question-word "how" modifying
        the adjective: "*Very big is it". This is enforced by
        connector logic: the sentence must connect to the wall
        somehow. "How" has a W- connector; adjectives and adjectival
        adverbs like "very" do not. "How" also has a QI- connector,
        for use in indirect questions like ex. 2 above. Connector
        logic ensures that the only way the AF connector can be used
        is if it connects through "how" to something on the left.

        Moreover, as shown by the examples above, if the AF occurs
        in the main clause, s-v inversion must occur; if not (i.e.,
        if it is an indirect question), s-v inversion may not occur.
        This is enforced in post-processing; see "SI".

        If the adjective has complements, these must occur after the
        subject of the sentence: "HOW certain ARE you that he is
        coming?" This is enforced by the ordering of the elements on
        adjective expressions: "AF-" precedes "(TH+ or TO+...)". 
        ("*How certain that he is coming are you" seems questionable; 
        we reject it.) 

AL	connects a few determiners like "all" and "both" to 
        following determiners.

                 +--------Sp-----+
                 +-----Jp--+     |
                 +-AL-+-D--+     |
                 |    |    |     |
        	All  the people are here

        Words like "all" are unusual. They act like determiners in a
        way; they must agree with the noun in number ("All" may take
        mass or plural but not singular nouns: "*All the dog died").
        But they may precede another determiner like "the" or "his".
        Thus we allow them to a make a J+ connection to the noun,
        treating it like a prepositional object. We also subscript J-
        connectors on nouns, giving singular nouns "Js-" and
        mass/plural nouns "Jp-"; "all" is then given "Jp+". This
        prevents "*All the dog died".

        A further problem here is that not just any determiner may be
        used with "all": "*All some/many people are here". Thus we
        require "all" to a make connection to the determiner as well;
        determiners that can be used with it carry "{AL-} & D+".

        Note that "all" can also be used with no additional determiner:
        "All people are good". For this we give "all" an ordinary D+
        connector.
        
AN	connects noun-modifiers to nouns.
     
                 +-----------D-----+
                 |    +-----AN-----+
        	 |    |	    +--AN--+--S--+
                 |    |     |      |     |
        	The income tax proposal was rejected

        Any singular noun may be used as a noun-modifier. All the
        noun-modifiers of a phrase are connected to the main noun of
        the phrase. Any number of noun modifiers may be connected.
        Thus nouns have a "@AN-" connector, conjoined with the rest of
        their expressions. Noun modifiers therefore always connect in
        parallel, rather than serially (sometimes this is
        counterintuitive, as in "income tax proposal").

        Noun-modifiers must normally occur after ordinary adjectival
        modifiers: "The stupid tax proposal was rejected", "*The tax
        stupid proposal was rejected". However, there are exceptions
        to this: "city clerical worker", "New York municipal
        bonds". For this reason, we require in stage 1 that any AN
        connections be made closer to the noun than any A connections;
        at stage 2, however, we allow AN connections further away than
        A connections. Nouns therefore have the following:

        dog:  ({@AN-} & {@A- & {[[@AN-]]}} & ...
        
        In general, noun-modifiers may not be plural: "*The taxes
        proposal was rejected", "*I made an eggs sandwich". However,
        one does sometimes see plural noun modifiers: "arms control",
        "sales division", "weapons violations charges". Here again, we
        use stage 2, giving plural nouns "AN+" as a stage 2 connector:

        dog: ({@A-} & D- or (S+....)....) or AN+
        dogs: ({@A-} & D- or (S+....)....) or [[AN+]]

        Noun modifiers may also not take determiners or post-nominal
        modifiers "*The tax on liquor proposal was rejected" (unless
        hyphenated: see "Pa"). Thus the AN+ on nouns is disjoined 
        from the rest of the expression.

        Proper nouns also have an AN connector, allowing them to act
        as modifiers to nouns: "The Smith tax proposal was
        rejected". (Proper-noun modifiers are allowed to follow
        ordinary noun modifiers, which is incorrect: "*The tax Smith
        proposal was rejected" is accepted.)

AZ	connects the word "as" back to certain verbs that can take 
        "object-as-adjective" as a complement:

                      +---AZ---+
                      +-O--+   +--Pa+
                      |    |   |    |
        	He viewed him as stupid

        "As" thus has AZ- conjoined with Pa+ (used in ordinary predicate
        uses of adjectives). Verbs like "view" and "characterize" have
        "O+ & {AZ+}", disjoined with other complement connectors.

B       is used in a number of situations, involving relative clauses
        and questions. It is most often used with transitive verbs.
        Transitive verbs have an O+ connector, which can be satisfied
        by a noun to the right. However, there are also various ways
        in which a word to the left may satisfy this need, such as
        relative clauses and questions; thus transitive verbs have a
        "B-" disjoined with their "O+". B connectors are also used
        to link the main noun of a subject-type relative clause to the
        verb. And they are used for so-called prepositional
        prepositioning, in which the object-need of a preposition 
        is satisfied by a preceding word: "The MAN we talked TO is here".

        B connectors interact very heavily with post-processing. It
        will be noted that there are many kinds of subscripted B+
        connectors; many of these subscript distinctions are used only
        for post-processing, not for controlling actual linkages.

        _B- on Verbs_
        Transitive verbs have B- connectors disjoined with their
        O+ (and any other complement connectors they may have such
        as TO+ or TH+). Every finite verb also has a B- disjoined with
        its S+, for use in subject-type relative clauses. This B- is
        conjoined with RS+. Ordinary transitive verbs thus have the 
        following:

        	destroyed: (S- or (RS- & B-)) & (O+ or B-);

        _Relative Clauses_
        B is used in restrictive relative clauses (i.e. those without
        commas), to connect the main noun to the verb of the relative
        clause (whether the relative clause is subject-type or 
        object-type):

                     +--B------+
                     +-R-+-S---+
                     |   |     |
        	The dog  I chased was black

        	     +----B----+
        	     +-R-+-RS--+
                     |   |     |
        	The dog who chased me was black

        Bs and Bp are used to enforce noun-verb agreement in subject-type
        relative clauses; these are exactly analagous to "Ss" and "Sp".
        So, "The dog who chases me is black", "The dogs who chase me 
        are black" are accepted; "*The dog who chase me is black" is 
        rejected.

        See "R" for a fuller explanation of relative clauses.

        _Questions: B#w, B#m_
        Bsw and Bpw connectors are used for object-type questions, in
        which the object is a simple question word like "which", "what",
        "who", or "whom". Such words therefore have "B*w+" connectors.

        	 +-----Bsw---+
                 |   +---I---+
                 |   +SI-+   |
                 |   |   |   |
        	Who did you see yesterday

        Bsm and Bpm connectors are used for object-type questions, in
        which the object-phrase contains a noun:

        	        +----Bsm---+
                        |  +---I---+
                  +D**w-+  +-SI+   |
        	  |     |  |   |   |
        	Which dog did you buy
        
        This construction uses the same B- connectors on verbs that
        object-type relative clauses use. However, the B#m+ connector
        on nouns used is _not_ the one used in relative clauses. Note
        that in the above construction, the Bsm satisfies the main
        requirement of "dog"; "dog" need not (in fact may not) also
        serve as a subject or object of a clause (*Which dog ran in
        the park did you buy?). Thus, whereas the "R+ & B#+" complex
        on nouns is optionally conjoined with the main "S+ or O-..."
        complex, the B#m+ is disjoined with the main complex.

        	dog: (R+ & Bs+) ...& ((S+ & {Wd-...}) or O- or Bsm+...)
        
        B#m connectors may only be used with question-word determiners
        ("what", "which", "whose", "how[many/much]"): "*The dog did you
        buy", "*The dog you bought". This is enforced by the fact that
        the wall has to connect to the sentence somehow. Normally,
        this is done through the subject of the sentence; nouns have a
        "Wd-" conjoined with their "S+". Notice, however, that the
        "Bsm+" on nouns is disjoined from the Wd-. When the B#x is
        being used, the sentence cannot connect to the wall unless it 
        does so through the determiner - and the only determiners that
        have W- connectors are question-word determiners. (Question-word 
        determiners also have QI- connectors, used in indirect
        questions: see below.) (Object-type questions also require 
        subject-verb inversion: see "SI".)

        "What" and "which" can act either as question-word determiners
        or as complete noun-phrases; and as complete noun-phrases they
        may occur in either subject- or object-type questions. Thus
        they carry (B*w+ or S**w+ or D**w+) & (QI- or W-). (See also 
        "S**w".)

        _Indirect questions_
        The above discussion of B connectors in questions applies
        to indirect questions as well.

                          +---B(s)--+
                +-S--+-QI-+    +S(s)+
                |    |    |    |    |
        	I wonder who Dave  hit

        Noun-phrase question-words have "(W- or QI-) & B*w+": the W- 
        is used in questions, the QI- in indirect questions. See "R".
        
        _B links involving dependent clauses_
        Suppose the verb making the B connection is in a dependent
        clause:

                 +------------------B--------+
                 |   +---I---+               |     
                 |   +SI+    +-C--+-S--+--I--+
                 |   |  |    |    |    |     |
                Who do you think Bill will bring

        This construction is handled perfectly well under the current
        arrangement. Similarly with relative clauses and indirect
        questions: "I wonder who you think Bill will bring", "The
        man who I think you met yesterday is here".

        There is a problem, however: there are constraints on the
        way that a B link can be made out of a dependent clause.
        Specifically, a B link cannot be made to a word that is
        within a subordinate clause (1), an indirect question (2),
        or a relative clause (3).

                 +----------------B(s)------------+
                 |                   +--Cs-+--S(s)+
                 |                   |     |      |
            1.* Who did you leave because Bill mentioned?

                 +-----------B(s)------------+
                 |                 +--Cs+S(s)+
                 |                 |    |    |
            2.* Who do you wonder why Bill  hit?


                 +--------------B(r)-----------+
                 |                 +-----B(r)--+
                 |                 +--R--+RS(r)+
                 |                 |     |     |
            3.* Who do you know someone who likes?

        This hold true whether the outer construction is a direct
        question, an indirect question (ex. 4-6 below) or a relative
        clause (ex. 7-9).

        	4.*I wonder who you left because Bill mentioned
        	5.*I wonder who you wonder why Bill hit
        	6.*I wonder who you know someone who likes
        
        	7.*The man who you left because Bill mentioned is here
        	8.*The man who you wonder why Bill hit is here
        	9.*The man who you know someone who likes is here

        Linkages are found for all of these sentences, under the
        current arrangement. They are weeded out in post-processing.
        In each case, a domain is started at the beginning of the
        dependent clause. Notice that the domain started will then
        spread back through the B connector. However, the subordinate
        domains started in these cases are of different kinds. Verbs
        that take clausal complements (like "said") have Ce
        connectors, which start 'e' domains; question words and
        conjunctions, however, have Cs connectors, which start 's'
        domains, and relative pronouns have "R" connectors, starting
        'r' domains. We then dictate that 's' and 'r' domains simply
        are not allowed to stretch back before the root word (the left
        end of the starting link). (We call such domains "bounded
        domains".) Thus the incorrect sentences above are prohibited.
        
        As described above, finite verbs also have "B-" connectors
        for use in subject-type relative clauses. These cannot be
        used, however, unless the finite verb can also make a RS
        connector to the left. The only way this can happen in
        questions is if the question contains a subordinate clause:
        "Who do you think will come?" (See "RS".) 

        _More about Dependent Clauses in Indirect Questions_
        Dependent clauses within indirect questions are, again,
        handled quite naturally:

                          +------Bsw(s(e))----------+
                +-S--+-QI-+    +S(s)+-Ce(s)-+S(s(e))+
                |    |    |    |    |       |       |
        	I wonder who Dave  thinks Bill     hit

        Note that in this case, it is very important that the
        B#w link be a "restricted link": the 'e' domain begun
        by the Ce must not be traced through it, otherwise it
        would spread to the rest of the sentence.

        Another false positive arises here at the linkage level:
                     
                    +-----------O(s)----------+
                    |            +-QI-+D**w(s)+
                    |            |    |       |
        	I read I don't know which   book

        Again, this is handled by post-processing. Since 's' domains
        are "bounded", they may not reach back before the root word. 
        Thus the above construction is prohibited in post-processing.
        
        _Transitive adjectives_
         Certain adjectives, when used predicatively, can take
        transitive infinitives as complements: "It is easy to use".
        Such adjectives take B+ conjoined with TOt+; see "TOt". 

        _"Whatever", "whoever"_
        Bsd is used for words like "whatever" and "whoever". These
        words may take object-type relative clauses: however, they
        simultaneously serve as the subject or object of a main clause.
        Therefore, they take "(Bsd+ or Ss*d+) & (Os- or Ss+)".

                   +-----------Ss-------+
        	   +------Bsd--------+  |
        	   |		     |  |
        	Whatever you want to do is fine

        Since the "whatever" phrase seems to constitute its own clause 
        here, a domain must be started for such constructions. Therefore 
        we make "Bsd" domain starting. A domain must also be started for 
        the corresponding subject construction:

                   +-----------Ss-------+
        	   +------Ss*d-------+  |
        	   |		     |  |
        	Whatever pleases you is fine
        
        Therefore we make "Ss*d" domain-starting as well.
        	
        _Comparatives_        
        In comparatives, the second half of the comparative can
        take a transitive verb with no object: "I have more books
        than she has", "She has as many friends as I have". A B+ is
        needed here to satisfy the "B- or O+" requirement on
        transitive verbs. Thus "than" and "as" have a "Bc" connector.
        Post-processing ensures that this is only used in certain
        kinds of comparatives: "*I am smarter than she has". See 
        "MV: Comparatives".

        _Prepositional Prepositioning_
        99% of the uses of B connectors involve hooking to a verb on
        the right: either satisfying the object-need of a transitive
        verb (most often) or the subject-need in "RS" constructions 
        (see "RS"). B- connectors appear analogously on prepositions,
        however, when the object-need of a preposition can be
        satisfied by a word on the right. For this construction to be 
        valid, the preposition involved must be modifying a verb,
        not a noun (i.e., it must be using an MVp- connector, not an
        Mp-):
        	
        	1.Who did you talk to
        	2.Which room did you sleep in
        	3.*What country did you meet a man from
        	4.*How many legs did you see an insect with
        	5.Who did you take a picture of (!)

        Sentences 3 and 4 are perhaps grammatical, but not if we take
        the preposition to modify the noun. For this reason, we do not
        directly disjoin B- with J+, as we disjoin B- and O+; rather, 
        prepositions carry

        	(J+ & (Mp- or MVp- or ...)) or (MVp- & B-)

        (The exception is "of", as shown by ex. 5; here, B- must be
        conjoined with Mp-.) Beyond this, B- on prepositions can be
        used in all the ways it is used on verbs: in relative clauses
        ("The man who I talked to is here"), with transitive
        adjectives ("He is easy to talk to"), with "whatever"
        ("Whatever you like to work with is fine"), and so on.

        _Prepositional-Object Relative Clauses_
        A final use of B occurs in relative clauses where the focus of
        the clause is the object of a preposition modifying the subject.

                              +----Bpj-----+
                       +--MX--+-M--+-Jr+RS-+
                       |      |    |   |   |
        	The doctors, many of whom are surgeons, were angry


                                 +--------Bpj-----+
                      +-----MX---+-M-+-Jr-+Cr-+-S-+
                      |          |   |    |   |   |
        	The book, the author of which I know personally, is good

        We already have a mechanism for allowing noun phrases surrounded
        by commas to modify other noun phrases: this is the MX connector
        (e.g., "The doctor, a good friend of mine, is here"). We use that
        same mechanism here. In this case, however, the modifying noun
        acts like the antecedent of a relative clause. It may either act
        as a subject, in which case it must be followed by an ordinary
        finite verb phrase; or it may act as an object, in which case it
        must be followed by a phrase containing a transitive verb but no
        object. The "B/RS" system is already in place to handle this. In
        relative clauses, the relative pronoun provides the needed "RS+"
        to allow for the following finite verb; in object clauses, it
        makes a "Cr" link to the subject of the relative clause. In this
        case, "which" and "whom" can serve the same function. The only
        difference between this usage and an ordinary relative pronoun
        is that here the relative pronoun is also acting as a
        prepositional object. Thus we give it "Jr".

        	which: (R- or Jr-) & (RS+ or Cr+);

        We must then allow the main noun of a comma-modifier phrase to
        act as a relative antecedent. Thus we give nouns an optional 
        "B#j+", conjoined with their MX-:

        	dog: ...(S+ or O- or J- or ({Bsj+} & Xc+ & Xd- & MXs-))

        There are several false positives to be avoided here. First, we
        must prevent the Bsj being used without the appropriate relative
        pronoun. Subject constructions ("The doctors, many of them are
        surgeons"), will be prevented anyway, because there is no "RS"
        to connect to the verb. But object constructions must be prevented:

        	      +---MX----+---Bsj-+
                      |         |       |         
        	The book, the author I know personally, is excellent

        (Actually this construction will be accepted anyway; the comma phrase
        will be treated as an ordinary noun-phrase, a noun followed by
        a relative clause, as indeed it is.) Secondly, we must prevent
        the "which" from being used in the wrong place:

                             +-Mp-+-Jr+----Cr--+
                             |    |   |        |
        	I saw the author of which the book was excellent

        (This parse too will be accepted anyway, using an Mj construction,
        but in any case the Jr parse is redundant and should be avoided.)
        We solve these problems in post-processing: we require that the
        Jr must occur in the same group as B*j. (This goes both ways:
        a Jr requires a B*j, and a B*j requires a Jr.)

BI	is used to connect forms of the verb "be" to some idiomatic
        expressions. Forms of "be" therefore have BI+, directly
        disjoined with their other connectors:

        	are: S- & (Pg+ or Pv+ or Pp+ or .... or BI+)

        Some title words like "chairman", "president", "head", etc., can
        serve as complements of "be" without taking an article. For such
        words, special dictionary entries have been created:

        	president.i chairman.i: {@AN-} & BI-;

        Indirect questions can act as complements of "be" as well, but
        only when the subject of "be" is a word like "question":

        	       +-Ss*q+BIq+
                       |     |   |    		       
        	The question is who killed Nicole
        	?The murderer is who killed Nicole

        The S connectors on words like "question" are thus given a
        special "s*q" subscript. Question words are given a BIq-
        connector. P.P. then dictates that these BIq connectors cannot
        be used unless a Ss*q is present in the group. Question words
        have BIq- directly disjoined with QI- and W- connectors, used 
        in direct and indirect questions.

        The phrase "because (clause)" can be a complement with "be" as
        well, but only if the subject is "this" or "that". "This" and 
        "that" are thus given Ss*b+; "because" is given BIh-; and
        P.P. dictates that a BIh requires a Ss*b. (These words
        are also among the subjects that can take "(be) (indirect
        question)" as a predicate; Ss*b is thus added to Ss*q 
        in post-processing as a connector that permits BIq.)

BT	is used with time expressions acting as fronted objects, as
        in "how many years did it last". See "OT".
        
BW	connects "what" (as a direct or indirect question word) to 
        various verbs like "think", "decide", which are not really
        transitive verbs but which can connect to "what" in object-
        type questions.


        	  +----BW-----+
                  |  +----I---+
                  |  +SI-+    |
        	  |  |   |    |
        	What do you think?

        	*I thought a good idea today.

        BW is also used with verbs like "tell" and "ask", which can
        connect to "what" as if it were an indirect object: "what did
        you tell him"?

        "Whatever" can be used in this way as well: "Whatever you think
        is fine".

C	connects conjunctions and certain verbs with subjects of clauses.
        It is therefore used only in embedded and subordinate
        clauses, not main clauses.

                    +---C--+                          +-C-+
                    |      |                          |   |
        	I told him I was angry      Call me when you are ready

        Every noun, nominative pronoun, and every other potential
        subject has a "C-" conjoined with its "S+" connector
        (but not its O-, J-, etc.). The C- is directly disjoined with
        a Wd-, which is used in main clauses:

        	dog: ({C- or Wd-} & S+) or O- or J-...

        When a dependent clause is begun, the subject usually
        makes a C connection to the left. There are two exceptions.
        When an object-type relative clause occurs with an omitted
        relative pronoun ("The man you met is here"), the subject of
        the relative clause makes an R- connection, not a C- connection.
        The reason for this concerns the use of "CO": see "CO".
        Secondly, in indirect object-type questions, the subject of the
        indirect-question clause makes no left connection at all. The
        "C-" connection on nouns must therefore be optional. 
        
        _Different kinds of C+_
        Ce is used for verbs that take clausal complements, also
        known as "embedded clauses": "tell", "assume", "think",
        etc.. Such verbs therefore have "Ce+" disjoined with their
        other complement connectors (TH+, TO+, O+, etc.). All verbs
        that can take "Ce+" can also take "TH+": "I assumed we would 
        go", "I assumed that we would go". The reverse is not true,
        however: "*I asserted/whispered/retorted we should go".

        Cs is used in several kinds of subordinate clauses. It is used
        with certain conjunctions, like "when" and "after": "The man I
        saw after I left your party is here." (Some other conjunctions
        do not take Cs; see "Wd".) Usually conjunctions that take Cs+
        can act either precede or follow the clause they modify ("When
        I saw you, I left"; "I left after I saw you". They thus take
        "Cs+ & (MVs- or CO+)".

        In many cases, such conjunctions may also take noun-phrases or
        participles as objects: "The man I saw ( after lunch / after
        running ) is here"; thus they have "Cs+ or J+ or Mv+ or Mg+",
        as appropriate. Notice that conjunctions that take nouns as
        objects can in general modify nouns also. Some, like "after",
        can modify nouns, as well as taking nouns as objects ("The
        party after the lecture was good"); in this sense, they are
        essentially acting as prepositions, and take Mp-. This raises
        the question of whether they can take Mp- and Cs+ in
        conjunction: "?The party after Fred graduated was excellent". 
        We allow this, but the expressions could easily be rewritten 
        to prevent it.

        Cs is also used for certain nouns that take clausal complements, 
        like "way" and "time": "I remember the time I went to
        London". Such nouns therefore have "Cs+ or @M+....", conjoined
        with their main "S+ or O+..." complex. Cs is also used in 
        where/when/how indirect questions: "I wonder where they will live".
        Such question words therefore have "QI- & Cs+". (In direct questions
        of this kind, s-v inversion must take place; therefore no C
        connection is made. See "Wq".)

        _Reasons for the C+ distinctions_
        The reason for making the distinction between Ce and Cs relates to
        "bounded domains". In relative clauses and questions (direct and
        indirect), a transitive verb can make a B connection to a preceding
        noun-phrase; however, there are constraints on how this may be
        done.  A B link may be made to a word within an embedded
        clause ("I wonder who Joe thinks Bill hit"), but not to a word
        within a subordinate clause ("*I wonder who Joe cried when
        Bill hit."), nor to a word within a relative clause or indirect
        question. To enforce this, Ce connectors (found in embedded clauses)
        start 'e' domains, Cs connectors (found in conjunction-linked
        subordinate clauses, and some indirect questions) start 's'
        domains; we then dictate that B links can extend out of 'e'
        domains, but not 's' domains. See "B: B links involving
        dependent clauses" for further explanation.

        A "B" link may not be made to a word within a subordinate
        clause, from outside that clause. However, it is perfectly fine to
        have a conjunction-connected subordinate clause within a relative
        clause, as long as the B link is not inside it:

        	*The man I cried when John hit is here
        	The man I hit when John cried is here

        _Ca_ is used in indirect adverbial questions:
                           
                     +--R-+--Eeh-+--Ca-+
                     |    |      |     |
        	I wonder how quickly John ran

        Adverbs that can be used in this way have "EEh- & (Ca+ or
        Qe+ or MVa-...)". (Qe is used in direct adverbial questions.)
        Ca can only be used in indirect questions. This is enforced
        because if the sentence must connect to the wall, and can only
        do so through "how". ("How" could make a direct "W" connection	
        to the wall, but this would trigger post-processing constraints
        which would prevent "Ca" from being used.) Like Cs connectors 
        (used in other indirect questions), Ca starts an 's' domain,
        thus putting the indirect question clause in its own group.

        _Cr_ is used only in the obscure "Noun-modifying prepositional-
        object relative clause" construction: see B*j.

CC	is used to connect clauses to coordinating conjunctions. 

                  +--------------CC-----------+
                  +--S--+--MV--+-C-+-S-+      +-W---+--S--+
                  |     |      |   |   |      |     |     |
             3.	John screamed when I arrived but   Sue   left 		

        CC is used with coordinating conjunctions only, and it links to
        the subject of the previous main clause. Subordinating conjunctions,
        by contrast (like "when" and "after"), link to the main verb of the
        previous clause, main or dependent. See "W" for more explanation.
        
        CCq is used with verbs like "say", which can be used in a quotation
        or paraphrase:
                        +-------------CC----------------+
                        |             +-------Xd--------+
                        |             |         +---S---+-Xc-+
                        |             |         |       |    |
        	The President is busy , the spokesman said \\\\\	

        Verbs that can take quotational complements of this kind have
        "CC+", directly disjoined with their other complement connectors.
        The verb must connect to commas on either side (or to the right-
        hand wall); see "Xc".

        say.v: (S- or I-...) & (O+ or Ce+ or TH+ or (Xd- & Xc+ & CCq-);

        Verbs that can be used this way can usually also be placed 
        elsewhere in the sentence - either after an opener (ex. 1), or
        after the subject phrase (ex. 2). In the first case, they use
        COq; in the second case, they use Eq. 

                        +-------------S---------------+
                        |                       +-Eq--+
                        |                       |     |
        1.	The President , the spokesman said , is busy

                +---------------------CO--------------------+
                |                              +-----COq----+
                |                              |            |
        2.	At the moment , the spokesman said , the President is busy

        Verbs such as "say" therefore have the following:

        	(S- or I-) & (O+ ... or (Xd- & Xc+ & (CCq- or Eq+ or COq+)));

        The category of verbs that can take quotations like this cuts 
        across other verb categories. For that reason, it seemed
        simpler to designate new dictionary entries for
        quotation-taking verbs; these entries have only the complement
        connectors used in quotation (CCq, Eq, and COq). Such entries 
        are subscripted with ".q" ("say.q", etc.).

        Quotation complement connectors - Eq, CCq, and COq - are all 
        domain-starting. The structures created in such sentences are
        rather unusual, however. With domains, the usual principle is
        to have embedded clauses nested inside the domains of main
        clauses. In cases of quotation, it would seem (semantically
        anyway) that the quoting verb (e.g. "say") is the main clause,
        and the quotation itself is dependent on it. Therefore, where
        possible, we try to make the quoting expression start a domain
        which includes the quoted expression. This is easy in the case
        of Eq and COq; both are simply made domain-starting links.
        It is more difficult in the case of CCq, given the link structure; 
        there, the quoting expression is to the right, and cannot
        easily start a domain which will spread back to the quoted
        expression. So, in the case of CCq, the CCq again starts a
        domain, but this causes the quoting expression to be nested in
        the quoted one rather than vice versa.

                        +-------------CC------------------+
                        |               +-------Xd(e)-----+
                 +---D--+--S---+-Pa--+  |         +---S(e)+Xc(e)+
                 |      |      |     |  |         |       |     |
        	The President is   busy , the spokesman said  \\\\\	

CO	is used to connect "openers" to subjects of clauses:

                     +----------CO------------+
        	     |                        |
        1.	Apparently,                  they went to a movie
        2.     	    On Tuesday,	             they went to a movie	
        3.	Although they were tired,    they went to a movie	
        4.	  Leaving the kids at home,  they went to a movie	
        5.	Abandoned by their parents,  they went to a movie
        6.   Still upset about Joe,          they went to a movie

        Various kinds of words have CO+ connectors: adverbs (ex.1),
        prepositions (ex.2), conjunctions (ex.3), participle
        phrases (ex. 4-5), and adjective phrases (ex.6).

        Openers may take commas; almost all words with CO+ therefore have 
        an optional "{{Xd-} & Xc+}". See "Xc". 

        Nouns have optional @CO- connectors, conjoined with their 
        C- and W- connectors, conjoined in turn with S+:

        	dog: ...({({@CO-} & (W- or C-)) or C-} & S+) or O- or J-...)

        Thus a CO link is always made to the subject of a clause.

        There are constraints on the way clauses can take openers. Main
        clauses can always take openers, as can subordinate and
        embedded clauses (ex. 1-3 below); relative clauses can only 
        do so if they involve a relative pronoun (ex. 4 and 5); 
        indirect questions cannot (ex. 6):

        	1. Screaming furiously, Fred left the room	
        	2. I was about to leave when, screaming furiously,
        		Fred attacked me
        	3. I told her that, after the party, I would meet her
        	4. This is the man who, in some ways, I would like to hire
        	5. *This is a man, in some ways, I would like to hire
        	6. *I wonder who, on Tuesday, John had lunch with
        	
        Recall that the clauses at issue here take different connectors.
        Main clauses make a W connection; embedded and subordinate
        clauses make a C. Indirect questions subjects take no left-branching
        connector at all. Relative clause subjects make C connections
        when a relative pronoun is present; otherwise they make R
        connections to the previous noun. Note that in the above 
        expression for "dog", C- and W- are conjoined with CO+; C- is not.
        Thus, when a noun subject is making a C- connection backwards,
        or when it is making no connection at all, it may not take an
        opener. Therefore ex. 5 and 6 are rejected.

        A further distinction is necessary here. Participle phrases can
        modify main clauses; they are very rarely found, however, on 
        dependent clauses (either embedded or subordinate).

        	Shouting loudly, Fred ran out of the room
        	*John told us how shouting loudly, Fred ran out of the room
        	*John had just entered when shouting loudly, Fred ran 
                        out of the room	

        For this reason we give participle phrases COp+; we then create
        two CO- connectors on nouns, one conjoined with W-, the other 
        with C-. The one with C- is subscripted COd-:

        	{({@CO-} & W-) or ({@COd-} & C-) or Cc-} & S+
        	
        _Participles as Openers_ A further point is needed about
        participle openers. They appear to take complements in the
        same manner as ordinary participles. There is a problem here,
        however. While participles always make a connection to the
        left to an auxiliary, they sometimes make a further-left
        connection to a fronted object, for example:

        	      +-------B------+
                      |    +---Pg----+    
        	      |    |         |
        	What book are you reading

        For this reason, the Pg or Pv on participles must be to the left 
        of the complement expression on participles. However, openers may
        not be used in this way. Moreover, on openers, the CO
        connection (directly disjoined with the Pg or Pv) must be
        made further to the right than any complement connections:

        	   +----------CO---------+
                   +---+                 |
                   |   |                 |
        	Saying he was innocent, John left the room
                Angered by what he saw, John left the room

        	   +------------------------+
        	   +---CO-+                 |
        WRONG:     |      |                 |
        	Saying,  John left the room he was innocent
        	Angered, John left the room by what he saw

        Here, then, the CO+ must be to the right of the complement
        connectors on the expression. For this reason, we must fully
        disjoin the CO and Pg/Pv connectors on passive and progressive
        participles.  This is necessary in any case for progressive
        participles in the case of gerunds; for gerunds, also, the S+
        must be to the right of the complement connectors on the
        expression. (See "Ss*g".) We also directly disjoin COp+ here
        with MVx- (used for comma participle phrases modifying verbs),
        and MX*p- (used for comma participle phrases modifying nouns).

        hitting: (Pg- & (O+...))   or    (O+...) & (Ss*g+ or COp+ or
                                            MVx- or MX*p-);

        angered: (Pg- & {@MV+})    or    ({@MV+} & (COp+ or MVx- or MX*p-);

CQ	is used to connect to auxiliaries in cases where the
        subject and auxiliary are inverted, and where the complement
        need of the auxiliary is satisfied by a word behind it. Such
        situations include comparatives, with "more...than" or
        "as...as", as well as other situations involving "as".

                                   +-CQ-+
        		           |    |
        	He has more money than does Joe

        Normally, subject-verb inversion is tightly constrained; it may
        only occur in certain types of questions. To allow it in situations
        such as these, we must add "CQ" to a special list of connectors 
        in post-processing which permit (and in fact require) s-v inversion.
        See "SI". See also "MV: Comparatives: Constrained Uses with
        Separate Clauses". 

CX      is used in comparative constructions like the following:

                                   +--CX--+
        		           |      |
        	He has more money than I do

        Normally an auxiliary like "do" or "have" must be followed by
        a participle. In cases like the above, however, this need can
        be satisfied by a preceding comparative word like "as" or
        "than". Auxiliaries thus carry

                do: S- & (I+ or CX-...) 

        See "MV: Comparatives: Constrained Uses with Separate Clauses".

D 	connects determiners to nouns.

                 +-D-+
        	 |   |
        	The dog died

        The first two subscript places on D links relate to number agreement.
        consider the following simplified entries.

        	the: D+;
        	a: Ds+;
                some: Dm+;
        	many: Dmc+;
        	much: Dmu+;

        	dog: Ds- & ...;
        	dogs: {Dmc-} & ...;
        	water: {Dmu-} & ...;
        	war: {Dm-} & ...;
        	
        Essentially there are three categories of noun and article:
        singular, mass, and plural.  The first subscript place
        distinguishes between singular ("s") and everything else
        ("m"); the second place distinguishes between plural ("c") and
        mass ("u") (for "countable" and "uncountable"). Nouns and
        articles which are singular-only have Ds; those which
        are plural-only have Dmc; those which are mass-only have Dmu;
        nouns which may be singular or mass have D*u+; determiners which
        may be plural or mass have Dm+; and determiners which may be
        mass, plural or singular have D+. (A few nouns, such as "fish",
        may be plural or singular; for these we create multiple 
        dictionary entries.)

        The third subscript place on D connectors relates only to
        post-processing. D##w connectors are used for
        question-determiners: "which", "what" and "whose". The 
        w triggers post-processing constraints relating to 
        question-inversion; see "SI".

        D##w connectors may only be used in questions: "*I bought
        which eggs today". This is enforced because the D##w
        on question-words is conjoined with (W- or QI-); it must
        make a link back to either the wall or an indirect-question
        verb like "wonder".

        _Determiners and Adjectives on Proper Nouns_
        In general, proper nouns may not take determiners. However, there
        are a number of exceptions. A number of proper nouns may take
        the determiner "the": "The Emir of Kuwait died", "The Supreme 
        Soviet met today". For these we use the DG connector; see
        "DG". Beyond this, however, one quite often sees proper names
        taking determiners, for example with brand-names or with people.

        	?The new David Letterman is a happy, secure David Letterman.
        	?I bought a Toyota to carry my Macintosh and several IBMs

        Thus we give proper nouns D- at stage 2. Note that in the first
        case the proper noun carries an adjective as well; this is also
        not uncommon in more colloquial writing. Thus proper nouns carry

        	[[{@A-} & {D-}]] & (S+ or O- or J-...)

DD	is used to connect definite determiners ("the", "his", "John's")
        to number expressions and adjectives acting as nouns.
        
        DD connects determiners to number expressions:

        	 +-DD-+--D--+
        	 |    |     |
        	My three sisters are coming next week
 
                      +---------S-------+
                 +-DD-+-M-+             |
                 |    |   |             |
        	The  two in the window are very attractive

        In the first case above, the number expression is really
        acting as the determiner (there must be number agreement with
        the subject: "*My three sister is coming"); the other
        determiner is superfluous. In the second case, the number 
        acts as a noun-phrase; note that it may take prepositional 
        phrases, relative clauses, etc., just like an ordinary noun. 
        In either case, the number does not require a DD connection; 
        it is optional. Numbers therefore have the following:

        	three: {DD-} & (Dmc+ or 
        	({@M+ or ...} & (S+ or O-...)));

        Only certain determiners may be used here: "*A three sisters
        are coming","*Many three students are coming".

        DD is also used to connect determiners to adjectives when
        they are being used as self-contained noun-phrases. Here again,
        the definite determiner is most often used; possessive
        determiners can perhaps be used (we allow them); but
        singular, plural-mass determiners are incorrect, as is use
        with no determiner.
                                                 
                                 +----O-----+       +---O----+
                                 |     +-DD-+       |    +-DD+
                                 |     |    |       |    |   |
        	This law will benefit the rich and hurt the poor

        	?This law will benefit our rich and hurt our poor
        	*This law will benefit a rich and hurt some poor
        	*This law will benefit rich and hurt poor

        Adjectives may be used in this way as subjects ("The poor will
        suffer"), objects ("It will affect the poor"), or prepositional
        objects ("It applies to the poor"). When used as subjects,
        adjectives act as plural forms ("The poor are/*is going
        to suffer"). This construction is most often seen with a few 
        adjectives such as "rich", "poor", "powerful", "meek", and 
        "famous". We allow it with all adjectives; however, since it 
        is quite rare, we make it a stage 2 construction. Adjectives 
        therefore have:

        	poor: A+ or (Pa- & <complement>) or 
        		[[DD- & (Sp+ or SIp- or O- or J-)]];

DG	connects the word "the" with proper nouns. In general, proper nouns 
        may not take determiners; but we do allow proper nouns to take
        the determiner "the". This allows things like "The Emir of Kuwait 
        died", "The Supreme Soviet met today", but not "*I have known many 
        Kuwaits", "*We need a Soviet". Other uses of determiners with 
        proper nouns are allowed at stage 2; see "D: Determiners with Proper
        Nouns".

DP	is used to connect possessive determiners to gerunds in cases
        where the gerund is taking its normal complement: 

                         +----------Ss*g----+
                         +----TOo--+        |
                 +---DP--+-O---+   |        |
                 |       |     |   |        |
        	Your telling John to leave was a mistake

        See "S**g: Gerunds: The Complement / No Determiner Case."

DT	connects determiners with nouns in certain idiomatic time
        expressions like "next week" and "last Tuesday".

        	      +-------MVp---------+
        	      |               +DTi+
        	      |               |   |
        	I'm going to London next week		

        This might be a good place for a general discussion of time
        expressions.

        English contains many kinds of time expressions. The most common
        uses of these expressions are analogous to prepositional phrases
        (indeed, many time expressions are prepositional phrases). Time
        expressions therefore often connect to the rest of the sentence 
        with the same connectors as prepositionsl phrases: MVp+ (when
        following a verb), Mp+ (when following a noun phrase), and CO+ (when
        preceding a clause).

        Many time expressions are single words or simple prepositional
        phrases ("yesterday", "on Tuesday"), but many others are
        idiomatic in construction, and use special-purpose connectors 
        like JT, DT, and Yt: "It happened three weeks ago / last 
        Tuesday morning / two days after the party". (Certain place
        expressions like "five miles away", which use Yd, are similar
        in this regard). In such cases, decisions must be made, sometimes
        rather arbitrarily, about which word in the time expressions
        will serve as the head, i.e., will make the MV connection to
        the rest of the sentence. 

        When time expressions contain a preposition or conjunction, that
        word is almost always the head of the expression. Often the
        head word is modified by a quantity expression, using a Yt
        connector (see "Y"). There are many time expressions, however,
        which do not use prepositions or conjunctions, such as this one:
        	      	       
        	I'm going to London next week

        With a phrase like "next week", we make "week" the head of the
        expression (we create a special "week.i" dictionary entry for
        such nouns); we use DTi to attach a word like "next" (or "last"). We
        then give "week.i" the usual prepositional phrase expressions: MVp-,
        Mp-, CO+.

        	week.i: DTi- & (MVp- or Mp- or CO+);
        	next: DTi+;

        	      +-------MVp---------+
        	      |               +DTi+
        	      |               |   |
        	I'm going to London next week		

        A phrase like "next week" can also be used a noun phrase. 

        	1.I'll be free after next week
        	2.Next week would be good
        	3.Are you ready for next week

        With certain prepositions, time expressions are very common as
        objects: "after", "by", "until" (ex. 1). For these we use JT 
        (see "JT").

        Other noun-phrase uses of time expressions are less common,
        and rather colloquial (exx. 2-3). Therefore we give "week.i" the
        usual "S+ or O- or J-" complex for nouns, but we make it stage 2.

        week.i DTi- & (MVp- or Mp- or CO+ or JT- or [[S+ or O- or J-]]);

        With phrases like "this week" and "every week", we must do
        something similar. However, there is an important difference
        between "this week" and "next week". "This week" is a
        well-formed noun phrase on its own, while "next week" is not. 
        Therefore the noun-phrase uses of "this week" (like exx. 1-3
        above) will take care of themselves, simply using the ordinary
        noun entry "week.n". 
 	
        	  +-D-+--S-+
                  |   |    |
        	This week would be good

        We need only worry about the prepositional phrase usages: "I
        can see you this week". Therefore, for determiners like "this"
        aand "every", as well as phrases like "the_next" and
        "the_previous", we create a DTn connector, conjoined with the
        prepositional-phrase connectors of "week.i" but not the
        noun-phrase connectors:

        	week.i: ((DTn- or DTi-) & (MVp- or Mp...)) or 
        		(DTi- & (JT- or [[S+ or O-...]])

        	this: DTn+;
        	next: DTi+;

        (There are a few other determiners like "one" and "all" which
        can be be combined with some time-units and not others: "one
        day"/"*one week", "all day/*all month". For these we create
        one-word idiom entries. The same distinction must be drawn
        here, however, between phrases that are well-formed noun
        phrases - "one day" - and those that are not - "all day".)

        Phrases can also be formed with all these determiners (both
        the DTn+ ones and the DTi+ ones) with month and
        day-of-the-week names. Again, such expressions occur most
        commonly as prepositional phrases or JT objects (exx. 4-6),
        but can also occur as noun phrases (exx. 7-8).  Again, we make
        the noun-phrase usages stage 2.

        	4. We should meet this Tuesday
        	5. You should do it before next January
        	6. Last Monday, I saw Fred
        	7. Next Tuesday would be good
        	8. Are you ready for next Monday

        Unlike words like "week", day and month names can be used as noun
        phrases with no determiner at all (exx. 9-10). Day names are often
        objects of "on", and month names are often objects of "in";
        these are so common that it seems sensible to make them stage
        1 usages; therefore we create special "ON" and "IN" connectors
        for them (ex. 11).

        	9. Monday would be good
        	10. Are you ready for April

        	       +---MV---+-ON-+
        	       |        |    |
        	11. I saw Fred on Monday		

        Finally, month names can be used with dates and years. For this
        we use TD and TY. For this purpose, special categories have been
        created for numbers which are common as years and days-of-the-month.

        			      +--TY-+
                           +--ON-+-TD-+ +-Xd+-Xc+
                           |     |    | |   |   |
        	I saw him on January 21 , 1990 \\\\\

        TD is used to connect day-of-the-week names to word like "evening":
        "I saw him Tuesday evening". TA is used to connect adjectives like
        "late" to month names: "We did it in late December". There are some
        other subtle distinctions regarding time expressions which we will
        not go into here; the following sentences illustrate some of them.

        	I saw him Monday
        	*I saw him January
        	I saw him January 21
        	*January 21, I saw him
        	We did it on Jan 21
        	*We did it in Jan
        	*I saw him on early January 21
        	Monday's concert should be good
        	This Monday's concert should be good
        	This week's concert should be good
        	The Monday concert should be excellent
        	*The this week concert should be good
        	I saw him in January 1990
        
E	is used for verb-modifying adverbs which precede the verb:

        	       +---E---+
        	       |       |
        	he apparently is not coming

        This is perhaps a good place for a general discussion of adverbs.

        _Types of adverb_
        There are a number of types of adverb. Some kinds of adverb
        are quite specific in the way they may be used: adjective-
        modifiers ("EA"), adverb-modifiers ("EE"), comparative-
        modifiers ("EC"), number-modifiers ("EN"). These kinds of
        adverbs are highly constrained in their use; thus we
        assign a specific connector type to each. (See specific
        connector entries for discussion.)

        Other kinds are less constrained in the way they can be used:
        adverbs of manner, clausal adverbs, and time adverbs.  Adverbs
        of manner refer specifically to the manner in which an action
        is done: "He laughed loudly", "She ran quickly". Clausal
        adverbs refer in some way to the clause as a whole: "John is
        apparently coming", "John is also coming", "John is actually
        coming". Time adverbs give information about the time of the
        action: "John is soon coming".  Each of these types of adverbs
        has a variety of usages, some of which overlap; therefore they
        have some connectors in common. However, the usage of each
        category is slightly different; there is also some variance
        within the categories.

        Several kinds of connectors are used with manner/clausal/time
        adverbs. E connects adverbs to following verbs. Thus every
        verb has an optional "@E-", conjoined with the rest of its
        expression. MVa connects adverbs to preceding verbs or
        adjectives (see MVa). (In a series of verbs, "He will want to
        have tried to do it", only the last one will have an MV+
        available for use.) EB connects adverbs to forms of "be", when
        the "be" word is connecting to an object or prepositional
        phrase. And CO connects adverbs to a following
        subject-noun-phrase.

        (A somewhat fuzzy distinction is assumed here between adverbs
        and prepositions. Generally prepositions take objects, can
        modify nouns, and can be complements of "be", whereas adverbs
        differ in all three respects. However, there is some gray area
        between the two. See "MVp" for more discussion.)

        _Manner adverbs_ 
        Manner adverbs may occur after the verb: "He ran quickly, she
        laughed loudly". They may also occur at the beginning of the
        sentence: "Quickly he ran", "Loudly, she laughed". Or, they
        may occur before the verb (or before any of the verbs, if
        there are several): "She had quickly opened the door"; "She
        quickly had opened the door". They may not usually occur after
        forms of "be": "*She was loudly in the kitchen". Therefore,
        most manner adverbs carry "MVa- or E+ or CO+". However, a few
        manner adverbs do not take E+ or CO+, like "properly" and
        "outright": "*You should properly do it", "*Properly you
        should do it".

        _Clausal adverbs_
        Clausal adverbs may almost always occur before the verb:
        "He almost/probably/fortunately closed the door". They can
        almost never be used following the main verb: "*He closed the
        door almost/probably/fortunately". They can be used before the
        subject in some cases, not in others: "Probably/fortunately/
        apparently he closed the door", "*Barely/simply/ever/almost
        he closed the door". A few can only occur before the subject:
        "Maybe he is coming", "*He is maybe coming", "*He is coming maybe".
        Generally they can occur after forms of "be": "They are
        apparently/probably/fortunately good programmers."
        In short, the norm for clausal adverbs is "E+ or CO+ or
        EB+", but there are many exceptions. Note that there are also a
        few adverbs which can be used either as clausal or manner
        adverbs, like "clearly" and "sadly". These adverbs thus take
        "MVa- or E+ or CO+ or EB+", and when used with E+ or CO+, they
        are ambiguous: "Clearly, he read the speech".

        (Many clausal adverbs _can_ follow the verb with commas: thus
        they take "E+ or CO+ or EB+ or (Xd- & Xc+ & MVa-)". See "MVa".)

        A special category of clausal adverbs is words like "chemically",
        and "financially". These would appear to be clausal adverbs in
        that they modify the entire clause rather than the verb, and
        like clausal adverbs, they can occur before the clause before
        the verb, or after forms of "be" (exs. 1, 2 and 3 below).
        However, unlike most clausal adverbs, they may follow
        the main verb (ex. 4), and they may also modify adjectives 
        (ex. 5). Thus such adverbs take "MVa- or E+ or CO+ or EA+
        or EB+".

        	1. Biochemically, the experiment was well-designed
        	2. We biochemically altered the materials
        	3. It was biochemically a good experiment
        	4. The experiment was well-designed biochemically
        	5. We need to get some biochemically valid results

        _Time adverbs_
        Many time adverbs (sometimes, often, recently, soon) can be
        used either after the verb, before the verb, after "be", or
        before the subject: "Sometimes, we have chicken", "We
        sometimes have chicken", "We have chicken sometimes", "We are
        sometimes in the garden". Such adverbs thus take "MVa- or E+
        or CO+ or EB+".  However, a few ("always", "never", "rarely")
        can only take E+ or EB- ("*Always he is late", "*He is late
        always").
        
        _Other kinds of adverbs_
        Other more specialized uses of adverbs are explained
        elsewhere: adverbs modifying adjectives and other adverbs (see
        "EA", "EE"), those modifying number expressions ("EN"), those
        modifying comparatives ("EC", and those modifying
        prepositional phrases (see "MVl"). The fact that a number of
        words belong to different combinations of these categories
        explains the large number of adverb categories in the
        dictionary.  A further complication is that some adverbs can
        be modified by adverbial adverbs like "very"; others
        cannot. Those that can have "{EE-}" conjoined with their other
        connectors; such adverbs can also be part of adverbial
        questions ("How quickly did you run"), and therefore must also
        take Ca+ and Qe+.  See "EE", "Ca", "Qe".

        _High-cost Uses of Adverbs_
        Some uses of adverbs are more rare, and are therefore given a
        cost of 2, making them stage 2 usages. It was mentioned that
        certain adjectives, like "very", can modify adjectives: "He
        is very skillful". However, one also sometimes sees ordinary
        manner adverbs modifying adjectives:

        	The cellist's delicately melodic style contrasted
        	with the fiercely abrasive tone of the violin and
        	the pianist's violently percussive chords.

        To allow this, we give manner adverbs high-cost "EA+"
        connectors. Note that adjectives only have one (non-multiple)
        "EA-" connector; therefore, if an adjective takes an adverb in
        this way, it cannot also take an ordinary "intensifying"
        adverb like "very".  This seems correct: "*The delicately very
        melodic tone of the cello was beautiful."

        Adjectives may also sometimes be modified by clausal adverbs
        and time adverbs.

        	1. He was often friendly
        	2. The often underpaid administrators used to resent
        	the invariably rude students and the understandably
        	impatient professors.

        Recall that clausal adverbs take "EB+" connectors, and
        forms of "be" have "EB+" connectors; thus in a sentence like
        1 above, "EB" is used (see "EB"). In ex. 2, however, no EB+ connector
        is available. Thus we must allow the adverb to connect
        directly to the adjective. We therefore give adjectives 
        "@E-" connectors, conjoined _only_ with their A+ connectors,
        not with their Pa-. Since this usage is fairly rare, we again
        make it a stage 2 usage. This yields the following:

        rude: {EA-} & (({[[@Ec-]]} & A+) or (Pa- & <complement>);

        This requires several further comments. First, why is it
        that clausal/time adverbs attach to adjectives with "E+",
        and manner adverbs attach with "EA"? For one thing, as
        mentioned above, manner adverbs seem to take the place of
        ordinary adjectival adverbs like "very" and "quite", whereas
        clausal and time adverbs do not:

        	The often very underpaid administrators used to resent
        	the invariably quite rude students and the understandably
        	rather impatient professors.

        Furthermore, when time/clausal adverbs are used as well as
        an adjectival adverb (or a manner adverb), the time/clausal
        adverb must come first:

        	The cellist's sometimes very melodic tone contrasted with
        	the largely rather percussive chords of the piano

        	*The cellist's very sometimes melodic tone contrasted with
        	the rather largely percussive chords of the piano

        Both of these distinctions are enforced by the solution 
        described above. Finally, a problem arises here: manner
        adverbs also take E+ connectors, so given the above expression
        for adjectives, a sentence like "The fiercely percussive
        piano chords" will receive two parses. To prevent this, we
        give manner adverbs "Em+", and we give adjectives "Ec-",
        as shown.

EA	connects adverbs to adjectives. Certain adverbs can modify
        adjectives ("very", "quite", "relatively"); these have EA+
        connectors. Adjectives have optional "EA-" connectors.
     
                    +----Pa-----+
        	    |   +--EA---+
        	    |	|       |
        	He is pretty stupid

        EAh connects to adjectives in the same manner as "EA". EAh+
        occurs only on the word "how", and is used in adjectival
        questions:

            +-Wq-+EAh-+-----AF-----+
            |    |    |            |
           ||	How stupid can you be

        (Also indirect questions: "I wonder how stupid he can be.""
        	
        On adjectives, "EA-" is optionally conjoined with "Pa+ or A+
        or AF-".  It would seem, then, that there is nothing to
        prevent EAh being used with Pa- (exs. 1 and 2 below) or A+
        (ex. 3):

                        +---Pa--+
                    +-S-+  +-EAh+  <-- ?
                    |   |  |    |
        	1. *He is how stupid.

        	2. *He is I wonder how stupid.
        	3. *The how stupid man is here.

        This problem is analagous to that of question-word determiners
        ("*He bought I wonder which book"), and is prevented in
        exactly the same way. Like "D**w+" on "which", EAh+ on "how"
        is conjoined with "R- or Wq- or Ws-", and therefore must make
        a link back either to an verb taking indirect questions
        ("wonder", "know") or to the wall. Therefore, in ex. 1 and 3,
        no linkage is found. QI- also starts an 's' domain, which is
        bounded: i.e., it is prohibited from spreading back to the
        left of its root word. Thus ex. 2 is prohibited in post-
        processing. In practice, then, EAh only occurs with AF, never
        with A+ or Pa+.

        There is one problem, however. Usually question-word
        determiners can be used in noun-focused (subject- or
        object-type) questions: "Which dog did you chase", "Which dog
        chased you". Similar linkages might be imagined, involving EAh
        links:

                    +-EAh+-A--+
                    |    |    |
        	1. *How big dogs did you chase
        	2. *How big dogs chased you.

        However, such constructions are clearly incorrect.  Thus these
        must be prevented somehow. Another point: Notice that the noun
        here is plural. With singular nouns, a determiner is required,
        so no such linkage can be formed (*"How a big dog did you
        chase" would involve crossing links).  However, there is a way
        of expressing this thought: ex.3 below.  With subject-type
        questions, though, even this construction seems wrong (ex.4):

        	3. How big a dog did you chase?
        	4. *How big a dog chased you?

        Other questions with "how" seem fine: subject- or object-type
        question like 5 and 6, involving "how many" or "how much",
        or adjectival questions like 7.

        	5. How many dogs did you chase?
        	6. How many dogs chased you?
        	7. How big was it?

        The rule seems to be that if a question is "adjective-focused"
        (that is, if the degree of the adjective is what is being
        asked about), it must be a "be" type question (using AF),
        rather than a subject-type (S) or object-type (B) question,
        unless it is an object-type question using "how (adj) a
        (noun)" (as in ex. 3). (We must also of course enforce s-v
        inversion in questions, and prevent it in indirect questions;
        this involves other connectors. See "SI: Questions requiring
        s-v inversion".)

        How do we enforce these constraints? First we must allow the
        rather odd construction in ex. 3. To do this we use the HA and
        AA connectors (see "AA" for a full explanation):

        	how: (EAh+ & {HA+}) 
        	a: EAh- & AA+;

        This yields the following:

                 +---HA---+    
                 +EAh+-AA-+-Ds-+
                 |   |    |    |
                How big   a   dog was it        

        But this allows both 3 and 4; 4 must be prevented. Moreover,
        plural constructions, both subject and object- type (ex. 1 and
        2), must be prevented too. (We cannot make the HA+ obligatory
        on "how"; EAh is often used on its own, as in "How big is
        it".) We prevent these constructions in post-processing. We
        simply state that EAh connectors can only be used when one of
        a list of connectors is present in the group: either AF (as in
        ordinary adjectival questions) or Bsm (as in ex. 3).  Thus we
        weed out all constructions where the noun of the "how" phrase
        is either a subject or where it is a plural object.  This
        applies in the same way to indirect questions.  (Note that
        with indirect questions analogous to ex. 2 above - "I wonder
        how big dogs chase you" - another interpretation is possible
        in which "how" is analagous to "when"; this is correct and is
        accepted.)

EB	connects adverbs to forms of "be" before an object
        or prepositional phrase: 

                    +------------O-----------+
                +-S-+--EB--+                 |
        	 |  |      |                 |
        	He is apparently a good programmer

        Forms of "be" therefore have an optional "EB+". Note that
        "EB+" is conjoined with MVp+ or O+, but not with Pg+ and Pv+;
        this is because present and passive participles have @E-
        connectors, so they can connect to adverbs using these.

        Certain adverbs can also be used in comma modifiers, following
        the first comma:

                   +------MX-------+
        	   | +----Xc-------+                  
                   | +--EB--+      +-----Xd----+
                   | |      |      |           |
        	A man, apparently in a bad mood, was there

        	A man in a bad mood was there	
        	*A man apparently in a bad mood was there

        As the third example shows, if an adverb is to be used in such
        situations, the modifying phrase must be surrounded by commas.
        For this reason, it is simplest to make the adverb attach directly
        to the preceding comma. (This is the only case where words attach
        to each other _via_ a comma.) We then give commas "{@EB+} & Xc+".
        (See "X" for more on commas.) 

        Now the only problem is that while many adverbs that can follow "be"
        can also follow commas, some cannot: "He is really a good player",
        "*John, really a good player, beat everyone". Thus we give such
        adjectives "EBm-", and we give commas "EBc+".

        Regarding the kinds of adverbs that take EB+, see "E: Types of
        Adverb".

EC	connects adverbs and comparative adjectives: 

                    +---Pa---+
                    |   +-EC-+
                    |   |    |
        	It is much bigger

        Comparative adjectives have optional EC- connectors conjoined
        with their "Pa- or A+".  The same class of adverbs that modify
        comparatives - "much", "somewhat", "a little" - can also
        modify comparatives like "more", "less", and "fewer": "I like
        him much more now", "We have much fewer students now", "I earn
        much less now").  Notice that comparatives can be so modified
        whether they are acting as adverbs, noun phrases, or
        determiners; thus they have EC- optionally conjoined with the
        rest of their expression.

        ECn is used in noun-focused comparative questions, like
        "How much more money do you have". See "EEh".

        ECa is used in adjective-focused comparative questions like
        "How much more efficient are they". See "EEh".

        EEx links adverbs to the word "much". Many adverb-modifying
        adverbs, like "very", can also modify the word "much". "How"
        also has an EEh connector; but this connector has special	
        significance for post-processing (see "EEh"). Therefore we
        prevent the EEh from linking to "much"; instead, we create
        a special "H" connector for this purpose.

EE	connects adverbs to other adverbs. Some adverbs can
        modify other adverbs ("very", "quite"); these carry EE+
        connectors. Adverbs which can be modified in this way (some
        can not) take EE-.

                   +---MVa---+
        	   |    +-EE-+
        	   |    |    |
        	I ran very quickly	

        EE can also be used with E ("He very quickly left"), CO
        ("Very quickly, he left") and EB ("He is very clearly a
        good programmer"). 

        _EEh: Adverbial questions_
        EEh is used to connect "how" to adverbs, in adverbial
        questions (direct or indirect): "How quickly did you run",
        "I wonder how quickly you ran". Adverbs that can be used
        in questions like this thus take "{EE-} & (Qe+ or Ca+ or MVa-...)".
        "Qe" is used in direct questions; "Ca" is used in indirect
        questions.

                             +---I---+
                 +Eeh-+--Qe--+-SI+   |
                 |    |      |   |   |
        	How quickly did you run

        Eeh is therefore analagous to EAh (see "EAh"). The use of EEh
        is constrained by post- processing in a very similar way. The
        "EE-" on adverbs is conjoined with MVa-; thus an adverb might,
        in principle, make both an EEh and an MVa connection: "*I ran
        how quickly". In practice, however, EEh is only usable with
        Qe+ (used in direct questions) and Ca+ (in indirect
        questions). EEh+ on "how" is conjoined with "R- or Wq- or
        Ws-".  When EEh is used with MVa, either "how" cannot make the
        connection it needs to the left, or else the bounded-domain
        constraint is violated. See "EAh".

        _How Much, How Much More_
        "Much" can also be used with "how" as a determiner or
        noun-phrase (see "H").  It can also make an EC+ connection to
        "more"; "more" may then act as a noun-phrase (ex.1) or as a
        determiner. Alternatively, "more" can act as a comparative
        adverb modifying a sentence or a following adjective.  In this
        case, however, a different linkage is used (ex.2 & 3):

                 +--H-+ECn+
                 |    |   |
             1.	How much more did you earn

                 +EEh-+ECa+
                 |    |   |
             2.	How much more can you run
             3.	How much more efficient is your program		

        In either case, "how" makes a W connection to the wall (in
        direct questions) or an QI connection in indirect question. In
        that case, why do we distinguish between "H" and "EEh" at all?
        The reason is that there are complex constraints on the way
        this phrase is used. Consider the following:

        	1. How much more money do they have
        	2. How much more money will be coming in
        	3. How much more efficient is their program
        	4. *How much more efficient programmers do they have
        	5. How much more efficient a programmer do they have
                6. *How much more efficient programmers work for them
        	7. *How much more efficient a programmer works for them

        A pattern emerges here, similar to the pattern that emerged with
        "EAh":

        	8. How many dogs did you chase
        	9. How many dogs chased you
        	10. How big is it
        	11. *How big dogs did you chase
        	12. How big a dog did you chase
        	13. *How big dogs chase you
        	14. *How big a dog chased you

        When the focus of the question is an adjective, either
        directly or indirectly ("How efficient", "How much more
        efficient"), then it must be a "be"-type ("AF") question; it
        may not be a a subject- ("S") or object-type ("B") question;
        the exception is object-type singular questions like ex. 12
        ("How big a dog"). Exs. 8 and 9 are not adjective-focused;
        ex. 10 contains an "AF" link; thus these are fine.  We already
        have a mechanism for ensuring that in adjective-focused
        questions like 10-14, an AF link must occur, except in cases
        like ex. 12. (see "EAh".)  Thus the constraints are set up
        just the way we want: we just have to find a way of triggering
        them in the right cases, i.e., in "adjective-focused
        questions"; ex. 3-7 above.  The problem is that in cases 8-14,
        adjective-focused questions are easily characterized; they all
        contain an "EAh".  With exs. 1-7, they are not so easily
        characterized.  We need to make sure that the
        adjective-focused questions, and these only, all contain a
        certain connector. Then we can have that connector trigger the
        same the same constraints that are triggered by EAh.  This we
        do through connector logic; the connector we use is
        EEh. (Irrelevant connectors and subscripts are omitted here.)

                how: (QI- or W-) & (H+ or EEh+);
        	much: (H- & (ECn+ & Dmu+)) or (Ee- & (ECa+))
        	more: (ECn- & (Dmu+ or S+ or B+)) or (ECa- & EA+)

        If "more" forms an EA+ link with an adjective - implying an
        adjective-focused question - an ECa link with "much" is
        required; an EEh link with "how" is then formed (if any link
        with "how" is formed); and the "adjective-focused question"
        constraints are applied. Thus 4, 6, and 7 above are
        prohibited; 3 and 5 are allowed.  However, if "more" is acting
        as a determiner or noun-phrase (as in 1 and 2 above), its ECn+
        connector is used; the ECn- on "much" is therefore used, a H
        link with "how" is formed, and the "adjective-focused
        question" constraints are not enforced.

        (Recall that EEh, as well as being used in adjective-focused
        comparative questions, is also used in adverbial questions:
        "How quickly did he do it?" Thus Ca and Qe - which occur in
        adverbial questions - are added to the list of "contains_one"
        connectors that EEh requires in its group.  Notice also that
        "How much more" may be used in yet another way as an adverbial
        phrase: "How much more did he do it?" Thus "more", like other
        adverbs, must carry "Qe+ and Ca+" conjoined with its ECa-.
        Further, "more" may itself modify an adverb: "How much more
        quickly did he do it?" Thus "more" must also carry EE+, like
        other adverb-modifying adverbs.)

        EAx connects certain adverbs to the word "many", when it
        is _not_ being used in questions. The adverbs that can
        modify adjectives ("very", "relatively") can also
        modify "many". EA is also used by "how" to connect to
        adjectives; "how" has "EAh+" for this purpose. However, 
        "EAh" has a special use in post-processing: it signifies
        that an adjective-focused question is present (see EAh).
        To prevent such connections from forming in the phrase
        "how many", we give "many" an EEx- connector and create
        a H connector for linking "how" and "many". 

EF	is used to connect the word "enough" to adjectives and adverbs.
        
                    +-Pa+
                    |   +-EF-+
                    |   |    |
        	He is good enough
                    
                        +-----A------+
                        +--EF-+      |
                        |     |      |
              He is a good enough player	

        In this case, the word "enough" expresses the degree of the
        adjective or adverb, similar to preceding adjectival adverbs
        like "very", which use EA+.  Moreover, "enough" cannot be
        combined with such a preceding adverb: "He is very good
        enough". Therefore EF+ on adjectives and adverbss must be 
        disjoined with EA-:

        	good: {EA- or EF+} & (A+ or Pa- or ...)
                quickly: {EE- or EF+} & (CO+ or MVa- or ...)

EI	connects a few adverbs to "after" and "before", such as "soon",
        "immediately", and "shortly".

                         +-EI-+
                         |    |
        	I left soon after I saw you
        
EN      connects certain adverbs to expressions of quantity: "We have
        nearly 100 students", "I have about 50 dollars". Number words
        therefore carry an optional EN-.

                                +----EN----+
                                |          |
                It will cost almost 400 million dollars
                They died    almost 400 million years ago

        Note that the EN always connects to the head word of the
        number expression: i.e., the word that connects it to the
        rest of the sentence.

ER	is used in the construction "the X-er..., the Y-er...":
            
              +------ER----------+
         +-DG-+                  +-DG-+
         |    |   	         |    |
        The better it is,       the more people will use it
        The more people use it, the better it is

        Such constructions always use comparative adjectives or
        adverbs. They consist essentially of two similar phrases,
        attached together by an ER connector. (Any phrase that can
        occur in the first half can occur in the second, as the above
        examples suggest.) The comparative adjective is always the
        "head" of each phrase. Each comparative adjective must have
        the capacity to serve as either the first or second half of
        the expression. The complement connectors (i.e., those that
        connect within the phrase) are the same, whether the adjective
        is serving as a "first-half" or a "second-half".  Such
        adjectives therefore have:

        bigger: ( complement ) & DG- & ((Wd- & Xc+ & ER+) or ER-)
        		
        			         ----------------      ^ 
        		                  for first-half  for second-half
        			            phrase	     phrase

        The DG- connects to the definite article "the"; the Xc+
        connects to a comma; the Wd- to the wall; and the ER
        from one phrase to the other.

        The connectors used within the phrase are mostly those
        discussed elsewhere:
        
        	The bigger it is	AF+
        	The more you run	Cs+
        	The more you earn	B+
        	The more money you earn Dm*w+
        						
        An exception is the following construction:

                      +-TR--+--U---+
        	      |     |      |
        	The better the computer , the faster the program

        For such constructions, the word "the" has "TR- & U+".
        Nouns have U-; this special connector is disjoined with
        all other optional and mandatory connectors on nouns
        (except AN- and A-); see "U". 

FM	connects the prepositon "from" to various other prepositions.

        	        +--MVp-+--FM--+----J----+
        		|      |      |         |	
        	John screamed from inside the house
        			   under the bed
        			   behind the car
        			
        	*John screamed from with the dog

        "From" is unusual in that it can take many prepositional
        phrases as objects (rather than noun phrases). "From"
        therefore has "(FM+ or J+) & (MVp- or Mp- or CO+...)";
        it can serve this function whether it is acting as a
        verb modifier, noun modifier, opener, etc.. Prepositions
        that can serve as objects of "from" have "J+ & (MVp- or Mp-
        or CO+ ... or FM-)".

G	connects proper nouns together in series.

                   +--G---+---G--+--G--+
                   |      |      |     |
                George Herbert Walker Bush is here

        Any number of proper nouns may thus be strung together 
        to make a proper noun phrase. The last noun in the sequence 
        then serves as the head of the phrase.

        G is only used for the linking of proper nouns. When proper
        nouns (or common nouns) modify common nouns - "The Dole proposal"
        - AN is used.
        
        Any word, when capitalized, may be used as a proper noun: "I
        saw The today". "I had lunch with And Of From Smith." The
        exception is at the beginning of sentences. A word which is
        listed in the dictionary in an uncapitalized form, and which
        is used at the beginning of a sentence, will be treated only
        as an uncapitalized form (although it might in theory be
        intended as a proper noun). We had to do this, otherwise EVERY
        word beginning a sentence would be a potential proper noun -
        which would be ridiculous. ("*The died" would be accepted.)

        However: certain words are common uncapitalized words, but are
        also used as names, and thus should be recognized as names at
        the beginning of sentences: "Bill", "Pat", "Sue", etc.. A
        special category was created for these.

        Numbers are rarely used as proper nouns phrases, although they
        occasionally are ("301 is a great class", "I live in 509").
        We do not allow this, since it would create a huge number of
        false positives. However, numbers are sometimes used as part
        of a multi-word proper noun expressions: "Fahrenheit 451",
        "Die Hard 3", etc.. Therefore we give numbers "G- & (G+ or S+
        or O-...)" as a stage 2 usage.

GN	is a stage 2 connector used in expressions where proper nouns
        are introduced by a common noun, with or without a determiner:

           +----D-----+----GN-----+
           |     +-A--+     +--G--+----S----+
           |     |    |     |     |         |
          The famous actor Eddie Murphy attended the event
         	     Actor Eddie Murphy attended the event

        The proper noun (or the last in a series of proper noun words
        - see G) is the head of the expression, making an S, O, or J
        connection to the rest of the sentence. On common nouns, GN+
        is fully disjoined with the main S/O/J complex and all the
        connectors for post-noun modifiers; but it is conjoined with
        the @A- connector (for adjectives), as shown by the first
        example above, as well as the @AN-, for pre-noun noun
        modifiers ("The adventure movie actor Eddie Murphy").  As for
        determiners, nouns in this situation can take a definite or
        possessive determiner ("My friend John") but not an indefinite
        one ("*An actor Joe Smith"); therefore we use the DD
        connector. This yields:

        dog: {@AN-} & {@A-} & ((D- & {@M+}... & (S+ or O- or J-...)) 
        or (DD- & GN+))

H       connects "how" to "much" or "many". "Much" and "many" can serve 
        as determiners ("How many books do you have"); they can also
        serve as independent noun-phrases ("How many do you have").
        Unlike most determiners, they can also be preceded by the word
        "how" in either case, creating a question or indirect
        question:

                 +--H-+-Dmc-+
                 |    |     |
        	How many books do you have

        Therefore they have H- connectors, optionally conjoined with
        their D+ connectors and also with their main noun
        complexes. However: when the H- is being used on "many" and
        "much", a question situation is created (direct or
        indirect). This introduces constraints on word-order, mainly
        relating to subject-verb inversion. For this reason, H+ on
        "how" is conjoined with QI- (used in indirect questions) or
        Wq- or Ws- (used in direct questions).  These connectors
        trigger post-processing rules; these are explained in
        "SI". "H" is also in a post-processing category along with
        "D##w", relating also to s-v inversion; see "SI: Questions
        without s-v inversion".
        
I	connects certain verbs with infinitives.

                                                 +---I----+
                +-S-+--I-+                    +S-+-O--+   |
                |   |    |                    |  |    |   |
        	I must  go to the store       I made him  go

        Modals have I+ connectors. Certain other verbs also have I+
        connectors, sometimes conjoined with O+ connectors, like
        "make" and "see".

        The word "to" also has an "I+", conjoined with "TO-", 
        used in infinitives. 

        Infinitive verb forms have "I-", conjoined with their
        complement connectors. In every case except "be", the
        infinitive form is the same as the plural form; therefore the
        same expression can be used.

        "Ii" connectors are used by pp to enforce the correct use of
        "filler-it" and "there". See "SF: Filler-it". 

        "Ia" is used in object-type infinitival indirect
        questions. Here, a "to"+infinitive construction occurs, but in
        this case - unlike other "to"+infinitive constructions - the
        "to" is unable to connect back to another word.

                           +--B---+
                           |   +Ia+
                           |   |  |
        	I wonder what to buy
                              
        In infinitival questions with "where/when/how" - "I wonder
        where to go" - the question word instead makes a TOn connection
        with "to". See "TOn".

IN	is used to connect the preposition "in" to certain idiomatic
        time expressions.
 	
                          +----IN----+
                          |   +--TA--+
                          |   |      |
        	We did it in early December

        See "DT" for more discussion of time expressions.        

J 	connects prepositions to their objects.

        	     +-Mp+----J---+    +-----MVp----+-J-+
                     |   |        |    |            |   |
        	The man with the hat chased the dog on Tuesday

        Proper and common nouns, accusative pronouns, and determiners
        that can act as noun- phrases have "J-" disjoined with their
        S+ and O- connectors.

        Prepositions have "J+ & (Mp- or MVp-)". "Mp" is used for
        prepositions modifying nouns; "MVp" is used for prepositions
        modifying verbs and adjectives.  Prepositions may also have
        other connectors, disjoined with J+, such as Mg+, Mv+, and
        QI+; see "MV: Other Uses of MVp and MVs".
        
        Jw is used to connect prepositions to noun-phrase
        question-words in the construction "To whom were you
        speaking?"  The construction formed here is very similar to
        that formed for determiner-question-words, as in "To which
        person were you speaking?" See "JQ" for an explanation. Jw is
        also used in prepositional relative clauses, like "The room IN
        WHICH I was working was cold". See "Mj".

        Jr is used only with "noun-modifying prepositional-object
        relative clauses". See "B*j".
        
JG	connects certain prepositions ("of" and "for") to proper-noun
        objects.              
        		       +-MG--+--JG-+
                               |     |     |
        	The National Academy of Science is meeting

        Proper nouns have an optional "MG+"; certain prepositions like
        "of" and "for" have "MG- & JG+". Since MG attaches a modifying
        preposition to a noun, is analagous to M; and JG, which
        attaches a preposition to its object, is analagous to J. But
        note that MG+ on prepositions is conjoined only with JG+, not
        with J+. Thus proper nouns can only be modified with
        prepositional phrases using other proper nouns: "The National
        Academy of Science is meeting", "*The National Academy of
        science is meeting".  (Similarly, proper nouns taking JG can
        only modify other proper nouns; but since they also have J-
        connectors, they can use these to modify ordinary nouns: "My
        book on Germany is excellent".)

JQ	is used to connect prepositions to question-words in constructions
        like the following:

                 +-------Qd----+
                 +-----J---+   |
            +-Wj-+-JQ+--D--+   +-SI-+
            |    |   |     |   |    |
           ||	In which room were you working

        This requires some explanation. Consider the following
        simplified expressions:

 		in: {JQ+} & J+ & (MVp- or Mp- or (Qd+ & Wj-)...)
        	which: (R- & (RS+ or C+)) or 
        		((QI- or W-) & (B+ or S+)) or (JQ- & D+);

        Notice, first, that the preposition uses the same J+ to
        connect to "room" that it uses in an ordinary prepositional
        phrase (for example, with a MVp- or Mp-). In ordinary
        prepositional phrases, however, no JQ link is made; thus the
        JQ+ on "in" must be optional. Notice also that the complex
        used on "which" is disjoined from the connectors it uses in
        other kinds of questions (direct and indirect) and relative
        clauses.

        But given this expression for "in", what prevents use of JQ in
        ordinary prepositional phrases: "I sat in which room"? This is
        prevented by post-processing. The Wj linking the preposition
        to the wall starts a domain, in which it is included; we then
        require that JQ connectors must have a Wj connector in the
        same group. We also have to ensure that in constructions like
        the above, a question-determiner _is_ used: "*In that room
        were you working". To do this, we also insist in
        post-processing that a group with a Wj must contain a
        JQ. Notice, also, that the preposition here makes a Qd
        connection to the auxiliary. This connector, which starts a
        domain, is one of the ones that permits (and in fact requires)
        subject-verb inversion in post-processing (see SI). But this
        serves another function too; it isolates the JQ and J
        connectors in the group. This ensures that the post-processing
        rules relating to Wj and JQ will not be satisfied by a JQ
        occurring later in the sentence ("*In the room were you
        working on which desk?" (We should note that groups started by
        Wj, unlike others, normally do not contain whole clauses.)
        Thus the following domain structure is generated:

                  +---Q(j(m))----+
                  +-----J(j)-+   |
            +Wj(j)+JQ(j)+D(j)+   +SI(j(m))+
            |     |     |    |   |        |
           ||	 In  which room were     you working
        	
        Such a construction might be also be formed in indirect
        questions: "I wonder in which room he was sleeping". This
        usage would require extra paraphernalia, however, and seems
        almost non-existent, so we exclude it.

        A similar linkage is formed in questions where the question-word
        acts not as a determiner but as a noun-phrase:

                 +---Qd----+
            +-Wj-+-Jw+     +-SI-+
            |    |   |     |    |
           ||	 To whom  were you speaking

        Question-words that can be used in this way - "which" and
        "whom" - have Jw- connectors disjoined with everything else.
        In this case, then, the usual "J+" connector on the
        preposition is used, but it is given a 'w' subscript; no "JQ"
        connection is made. "Jw" is then added to "JQ" in the list of
        connectors which may satisfy the demand of "Wj"; and, in turn,
        "Jw" may only occur if a "Wj" is present.

JT      connects certain conjunctions to time-expressions that are not
        well-formed noun phrases, like "last week". and "this week".

                  +-----CO-------+
                  +----JT----+   |
                  |    +--DT-+   |
                  |    |     |   |
        	Until last week, I thought she liked me

        With time expressions that are well-formed noun phrases like
        "this week", no JT is necessary; the phrase can serve as a
        prepositional object using its ordinary J connector.  See
        "MVp: Time Expressions".

K	connects certain verbs with particles like "in", "out", "up",
        and the like.
                          +-K-+
                          |   |
        	The man came up	

        	*The man arrived up

        Particles that can be used in this way have "K-" disjoined
        with everything else. (Most are also prepositions; a few, like
        "away", are not.)

        We distinguish between verbs that can take particles and
        those that do not, but among particle-taking verbs, we do
        not distinguish between specific verb-particle pairs: we
        allow "We sorted them out/*up", "We put them in/*over".

        Verbs that take particles may be transitive ("pick"),
        intransitive ("come"), or trans/intrans ("move"). With
        transitive verbs, the particle may either precede or
        follow the direct object: "We picked the dog up"; "We
        picked up the dog". This yields the following expression.

        	pick: (S- or ....) & ((K+ & O+) or ((O+ or B-) & {K+}));
        				 
        				^ particle        ^ particle
        			      precedes object     follows object

        Note that the particle is always optional. However: if
        we made the particle optional in both the "pre-object"
        case and the "post-object case", then "We picked the dog"
        would receive two parses. So it must be obligatory in one
        case, optional in the other. 

        A further complication: with transitive verbs, the particle
        may always follow the object, but it may precede it only if
        the object is a full noun-phrase, not a pronoun: "We picked it
        up," "*We picked up it". We enforce this using
        post-processing.  In the expression for transitive verbs, the
        O+ in the pre-object case is subscripted "O*n+". Pronouns are
        then subscripted "Ox-". "Oxn" connectors are then prohibited
        in post-processing.

L	connects certain determiners to superlative adjectives. 
 
        	       +------D----+
        	       +--L--+     |
                       |     |     |
               He has the biggest room
        
        In most cases, when a determiner-adjective-noun phrase occurs,
        both the determiner and the adjective attach to the noun.
        Superlative adjectives are different, however.  Superlatives
        be used with determiners, and only certain determiners:

        	This is their biggest room
        	*This is a biggest room
        	*They have biggest rooms
        	
        To enforce this, it seems easiest to simply make the
        superlative connect to the determiner. Determiners that can
        perform this function carry "{L+} & D+". Superlative
        adjectives carry "L-".

        As well as superlatives, other adjectives are in this category
        such as "own", "next", and "same".  Number words like "third"
        and "fifteenth" are also included: "This is the fifteenth book
        I've read", "*This is a fifteenth book I've read", "*This is
        fifteenth book I've read".

        Numbers can also be used here, on the left end of an L link:
        "The five biggest cities are in China". As in other cases with
        determiners followed by numbers ("The five cities we saw
        were..."), a DD connector is used: numbers thus carry "{{L+} &
        DD-} & (Dmc+ or S+...)".

                      +------Dmc----+
                 +-DD-+--L--+       +-S--+
        	 |    |     |       |    |
        	The five biggest cities are

        Note that the L+ on numbers is optionally conjoined with the
        DD-; it may not be used unless the DD- is used. This prevents
        "*Five biggest cities are in China".

        On determiners that take L+, it is optionally conjoined not only
        with D+ but with DD+. This allows "The biggest five cities..."

LE      is a special connector used in comparative constructions to
        connect an adjective to the second half of the comparative
        expression beyond a complement phrase:

        	             +------------LE---------+
        	             |                       |     
        	It is more likely that Joe will go than that Fred will go

        See "MV: Comparatives".

M	connects nouns to various kinds of post-nominal modifiers without
        commas, such as prepositional phrases, participle modifiers,
        prepositional relatives, and possessive relatives. (Phrases of
        these kinds with commas use MX; see "MX".)

        Ordinary relative clauses do not use M+, but rather "R+ & B+"
        (see "B: relative clauses").  Nouns therefore have optional
        "@M+" connectors, conjoined with their main "S+ or O+ or
        J+..." expression and with their "{@A-} & D-". Certain
        determiners that can act as complete noun-phrases also carry
        @M+ ("everyone", "many"), as do numbers. Pronouns do not. This
        yields

        	The man from London is here
        	Some of the programmers are very good
        	Five of the programmers are very good
        	*He of the programmers is very good

        _"M" modifiers used with relative clauses_

        Note that words that take @M+ connectors have the
        following expression:

                ...{@M+} & {R+ & B+ & {[@M+]}}...

        Nouns frequently take a relative clause, or one or more
        prepositional phrases; they rarely take both.  Occasionally
        they do, however: "The picture that I showed you of John",
        "The picture of John that I showed you". Therefore we have to
        allow "@M+" and "R+ & B+" both conjoined and disjoined, in
        either order; this is what the above expression
        provides. There is still something rather arbitrary about
        this, however. We allow any number of M connections
        (prepositional phrases, participle modifiers, etc.) to be
        made, or one relative clause followed or preceded by any
        number of M connections, but not multiple relative
        clauses. This was necessitated by the fact that relative
        clauses require two conjoined connectors, and there is no way
        of allowing indefinitely many "R+ & B+" connections to a
        word. In practice, however, multiple relative clauses are
        extremely rare.

        _Mp: Prepositional phrases modifying nouns_

        "Mp" is used for prepositional phrases modifying nouns.
        Prepositions therefore have Mp- connectors directly disjoined
        with MVp- (used for prepositional phrases modifying verbs) and
        conjoined with J+ (used for prepositional objects). (Some
        prepositions can take other kinds of objects besides
        noun-phrases, and thus have other connectors conjoined with
        "MVp- or Mp-": see "MVp".)  Almost all words that have Mp-
        also have MVp-; one exception is "of", which can modify nouns
        but not verbs ("*The dog ran of the yard").  (A few verbs can
        be modified by "of": see "OF".)

        It was mentioned that many determiners and numbers also
        have M+ connectors, as well as nouns. Thus the sentences
        below are all linked in essentially the same way:

                            +-Mp+---J------+
                            |   |          |
        	1. The salaries of the programmers are excellent
        	2.         Some of the programmers are excellent
        	3.         Five of the programmers are excellent

        This may seem counterintuitive. In ex. 1, the subject
        of "are" is "salaries"; in ex. 2 and ex. 3, it is really
        "programmers". We see no need to make this distinction,
        however. As well as the special use of "some" and "five"
        shown here, these words can act as independent noun-phrases
        taking no modifying phrase or quite different modifying
        phrases (ex. 4 and 5); and even constructions like ex.
        2 and 3 above sometimes arise where it is quite clear that
        "some", not the prepositional object, is the real subject
        (e.g., ex. 7).

        	4. Some are excellent
        	5. Some with doctorates are excellent
        	6. Some holding doctorates are excellent
        	7. (Most of the pictures of the executives are 
        		terrible, but) 
        		some of the programmers are excellent

        _Ma: adjectival modifiers_
        Ma connects nouns with post-nominal adjectival modifiers. 

                                     +-Ma--+
                                     |     |      
        	These are people unhappy about the economy
        	This is a trial certain    to attract attention

        This is only correct when the adjective has some kind of
        complement or modifier attached to it: "*These are people
        unhappy", "*This is a trial certain". We enforce this in
        post-processing. Ma connectors start a domain; post-processing
        then requires that a group containing an Ma contain one of a
        list of connector-types, either complement connectors like
        "TO" or "TH", or the prepositional phrase connector MVp.

        _Mv and Mg: Participle Modifiers_
        Mv connects nouns with passive participles:

                     +--Mv-+
        	     |     |       
        	The dog chased by the man died

        Mg connects nouns with present participles:

                     +--Mg-+
                     |     |
        	The dog chasing the man died

        These are sometimes known as "participle modifiers". Every
        passive verb form has a Mv-, directly disjoined with its
        V- (used in normal passive constructions). Every present
        participle has an Mg-, directly disjoined with Pg- (used
        in normal present participle constructions). Thus in
        complex verbs, whatever complement is normally required
        or allowed by the participle (O+, TH+, TO+, etc.) will be
        required or allowed here. (Recall that passive verb-forms
        have different complement expressions from active forms.)

        	The dog chased by the man was black
        	*The dog chased the man was black
        	*The dog chasing was black
        	The dog chasing the man was black
        	People expecting to see the President will be disappointed
        	*People expecting will be disappointed

        If Mv- is directly disjoined with Pv- in every case, and Mg-
        with Pg-, why have two different connector types?  The reason
        relates to post-processing. Post-processing divides a sentence
        into groups of links, in which each group corresponds,
        roughly, to a clause: the set of links involved in a
        subject-verb expression (as well as any direct object,
        indirect object, or prepositional phrases). If the clause
        contains a embedded clause, this forms its own group. In some
        cases that participles are used, they belong to the same
        subject-verb expression as the previous links (ex. 1 below).
        In cases of participle modifiers, however, the participle
        indicates the beginning of a new subject-verb expression
        (ex. 2). 

                                        +----J----+
                 +-D---+-S-+--Pv-+--MV--+   +--D--+
                 |     |   |     |      |   |     |
        	The study was mentioned in the journal

                                                +--J(e)--+
                                +-Mv(e)-+-MVp(e)+   +D(e)+
                                |       |       |   |    |
        	I've read the study mentioned   in the journal

        Thus Mg and Mv connectors must start new domains;
        Pv and Pg connectors must not.

        This distinction is particularly important in cases where we
        use post-processing to enforce subject-verb constraints, such
        as "filler" uses of "it" and "there". Post-processing must
        know that in cases of participle modifiers, the verb in the
        participle modifiers does not apply to the subject of the main
        clause; thus it knows that "There seems to have been a study
        mentioned in the journal" is correct, while "There seems to
        have mentioned a study in the journal" is incorrect. See "SF:
        Filler-it".

        _Mv+, Mg+ used on conjunctions_
        Some conjunctions can also take present and past participles
        instead of noun-phrases and clauses: "We saw a movie ABOUT
        CATCHING dogs", "We entered BY CLIMBING in the window",
        "He will respond WHEN QUESTIONED". Mv and Mg are used
        here to connect the conjunction to the participle. 
        
        Again, the question arises here: why use Mg+ and Mv+ rather
        than Pv+ and Pg+? This is a difficult question, and again,
        relates only to post-processing.  Recall that Mg and Mv begin
        new domains; they imply that a new subject is in force. In
        many cases this seems to hold true with participles modifying
        prepositions and conjunctions.  This is particularly true when
        the preposition modifies a noun (ex. 1), but also sometimes
        when it modifies a verb (ex. 2).

                                 +---Mp-+---Mg-+
        	                 |      |      |
             1.	We saw a great movie about catching dogs
        	           
                       +-------MVp---+-Mg---+
                       |             |      |
             2. She criticized us   for   chasing the dog
             3.?There was a problem while catching the dog

        In other cases, however, it seems that the previous subject is
        still implied:

             4. We talked about catching dogs
             5. We angered her by chasing the dog
             6. The man died while chasing the dog

        At the moment, this has little effect on well-formedness.
        The only cases it matters are where post-processing checks are
        used to control the uses of "there". If we used Pg connectors
        rather than Mg, then sentences like ex. 3 above would be 
        prohibited; perhaps they should be. In any case, for now,
        Mg and Mv are used in all such cases. 

        _Mr: Possessive Relatives_
        Mr is used for relative clauses involving "whose":

        	     +-Mr-+-Ds*w+-S--+
        	     |    |     |    |
        	the dog whose owner died was black

                                +---Bsm--+
                     +-Mr-+-Ds*w+     +S-+
                     |    |     |     |  |
        	the dog whose owner John hit was black

        The two constructions shown here are exactly analagous
        to the uses of "whose" (and other question-determiners
        like "which") in indirect questions. A combination of 
        link logic and post-processing rules enforces very
        tight constraints on the use of Ds#w and Bsm (and hence
        of Mr as well); see "B#n, B#m".

        Like Mg and Mv connectors and relative clauses (R+ & B+),
        Mr connectors start a new domain. See Mg and Mv for an
        explanation.

        _Mj: Prepositional-object relative clauses_
        Mj is used for relative clauses in which the main
        noun is the object of a preposition: 

                          +---Cs--+
                     +-Mj-+-Jw-+  |
                     |    |    |  |
             1.	The man   to whom I was speaking was tall

                          +------Cs-----+
                          +----J----+   |
                     +-Mj-+--JQ-+-D-+   |
                     |    |     |   |   |
             2.	The man   to whose wife I was speaking was tall

        The "JQ" and "Jw" connectors used here are also used in
        prepositional questions ("To whom were you speaking"). See
        "JQ" and "Jw".

        Prepositions have "Mj- & Cs+" conjoined with J+ (which forms
        the "Jw" link with "whom"), and disjoined with "MVp- or Mp- or
        (Wj- & Qd+)..." (used in prepositional questions).  This
        raises possible problems. In relative constructions like the
        above, the prepositional object used must be a relative
        pronoun like "whom" (*The man to the man I was speaking was
        tall", "*The man to John's wife I was speaking was tall".
        Furthermore, we must prevent "whom" and "whose" being used in
        any old way: "*I was speaking to whom", "*I was speaking to
        whose wife".  These problems have been dealt with already in
        connection with prepositional questions: e.g., "To whom were
        you speaking?".  There, we have Wj - the connector connection
        the preposition to the wall - start a domain whose group
        contains only the initial prepositional phrase of the
        sentence; we then require that every group containing a Wj
        contain a Jw or a JQ, and that Jw and JQ may occur only in
        groups containing a Wj. Here, then, all we have to do is add
        "Mj" to the list of connectors that require JQ or Jw and
        satisfy the requirement of JQ and Jw.  The same constraints
        that apply in prepositional questions then automatically apply
        here.

        Note that this system very naturally permits more nested
        constructions like "The dog to the wife of whose owner you
        were speaking is here".

MG	allows certain prepositions to modify proper nouns. We do
        not allow most prepositions to modify proper nouns (*"John
        with the big nose is here") but we do allow it with "of" and
        "for": "The Society for Creative Anachronisms is meeting",
        "The Ongle of Bongle Dongle died today"). MG must be used with
        JG, forcing the object of the preposition also to be
        capitalized ("The Society for Creative Anachronisms is
        meeting", "*The Society for creative anachronisms is
        meeting").

MV      connects verbs (and adjectives) to modifying phrases like adverbs,
        prepositional phrases, time expressions, certain conjunctions,
        "than"-phrases, and other things.

        		 +----------------MV-----------+
        		 +------MV-----+               |
        		 +-MV--+       |               |
        		 |     |       |               |
        	The dog ran quickly through the park with a bone

        Any number of prepositional phrases or adverbs may attach to a
        verb. 

        _@MV+ on Verbs_
        The way @MV+ connectors are combined with complement connectors
        on verbs is quite complex. Simple intransitive, transitive and
        optionally-transitive verbs simply have an optional "@MV+";
        with transitive and optionally-transitive verbs, this is
        conjoined with the O+ connector:

        	destroy: (Sp- or I-) & (O+ or B-) & {@MV+};

        (Normally, prepositional phrases and the like follow the
        object in this situation, but sometimes a prepositional phrase
        is inserted before the object: "We destroyed the garage with axes
        on Tuesday", "?We destroyed with axes on Tuesday the garage".
        See below for explanation.)
        
        Some verbs connect to other verbs - modals (which take
        I+), auxiliaries (which take I+, PP+, or Pg+) or verbs that
        take infinitival complements (TO+). Such verbs may NEVER
        make an @MV+ connection beyond the verb they attach to.
        Thus in the sentence below, "in London" can connect only to
        "run", not to "would", "prefer", "be" or "appointed".
                     
                                                    +---------MV---+
                    +-PP--+-TO--+-I+--Pv--+-TO--+-I-+              |
                    |     |     |  |      |     |   |              |
             1. He had expected to be appointed to run the project in London

        Consider the following expressions:

        	expect: ({@MV+} & TO+) or ((O+ or B-) & {@MV+});
        	appointed: ...{@MV+} & {TO+};

        "Expect" has an @MV+ conjoined with TO+, but it must connect
        _before_ the complement; this is explained below.  Note that,
        since "expect" is also an ordinary transitive verb ("I
        expected the appointment"), it has an @MV+ conjoined with its
        O+ connector (which, again, connects _beyond_ the direct
        object), but this is disjoined with the TO+. With "appointed",
        the @MV+ conjoined with the TO+ is, again, before the TO+.
        Notice that the TO+ is optional here: thus "appointed" may be
        used with no complement at all, in which case it may take an
        MV connection ("He was appointed in London").

        The same applies to complement connectors like TH, TS, and
        C (attaching to clauses, regular or subjunctive) and
        QI (attaching to indirect questions). Again, such verbs
        cannot make MV connections beyond their complements. Thus
        the sentence below receives only the parse shown.

                                                      +---TOo-+
            +-S-+-TH-+-C-+-S---+--QI--+---C-+-S--+-I--+-O-+   +-I+MV+
            |   |    |   |     |      |     |    |    |   |   |  |  | 
        2. He said that he wondered where John would ask him to go on Monday

        Notice that phrasal complements of this kind, since they
        necessarily contain verbs, will themselves have available @MV+
        connectors at the end.

 	The point is that on any single verb expression, or in any
        clause involving verb-attached embedded clauses (not
        subordinate clauses and relative clauses), adverbial and
        prepositional modifiers placed at the end can only attach to
        the final verb in the expression. However, linkage logic
        insures that every verb expression or subordinate clause will
        eventually end with a verb that does _not_ connect to a
        further verb or clause, and thus will have an available
        "@MV+".  By doing it this way, we greatly reduce the number of
        linkages that would be found on long sentences; and in most
        cases the modifier applies to the final verb anyway. (Although
        not always. In ex. 2 above, for example, "on Monday" seems to
        apply to "go", but might also apply to "ask" or "wonder".)

        _Other @MV+ connectors on Verbs_
        In most cases, with complex verb expressions or verb-linked
        subordinate-clause expressions, prepositional phrases and the
        like occur at the end.  However, they may also in some cases
        occur in the middle. Verbs that take complements such as TO+,
        Pg+ or I+ occasionally take prepositional phrases _before_ the
        complement. And verbs that take clausal complements (TH+, C+
        or QI+) sometimes take prepositional phrases before their
        complements also.  Some of these uses are rather questionable
        (TH+ and TO+ are best; Pg+, I+, QI+, and C+ are more
        doubtful), but for now we allow them all.

                      +---------------------+
                      +--MV---+             |
                      |       |             |
        	He attempted for many years to be a concert pianist
        	?We discussed at that time hiring a new secretary
        	He announced  on Monday    that he was hiring Smith
        	I   asked him on Tuesday    who had been hired

        As mentioned, transitive verbs have "@MV+" connectors
        following their "O+" connectors. However, one occasionally
        sees a prepositional phrase inserted before the object:

        	We destroyed with axes the garage that our grandfather 
        		had built

        We therefore give transitive verbs an optional "@MV+" before
        the "O+", in addition to the one following. Since this
        construction is rare, we give it a cost of 2, making it a
        Stage 2 construction.  Moreover: in such cases, the object
        phrase is never a pronoun; it is always a full noun
        phrase. ("*We destroyed with axes it"). We already have a
        special O+ subscript, O*n+, that can be used only with noun
        phrases, not with pronouns (see "K" for an explanation).  Thus
        we use that here as well. This yields the following:

        	destroy: ([[@MV+ & O*n+]] or O- or B+) & {@MV+};

        A final case to be considered is verbs that take an object
        plus some other complement. Some verbs take "O+ & TOo+" (for
        object+infinitive constructions), "O+ & TH+" (for
        object+clause constructions), and the like. In such cases, one
        _never_ sees a prepositional phrase inserted before the
        object. One occasionally sees it inserted after the object and
        before the TH+ or TOo+.

        	*I told on Tuesday John to do it
        	I told John on Tuesday to do it

        Such verbs therefore have

        	(O+ or B-) & {@MV+} & {TOo+ or TH+...};

        The same applies to verbs that take two objects; a prepositional
        phrase may be inserted between the two objects, but never before
        the first object:

        	I gave him on Tuesday an expensive present
        	*I gave on Tuesday him an expensive present

        In cases where the verb may take two objects, we must also
        prevent the second object from being a pronoun (see "O"); 
        and in any case where the preposition precedes an object, 
        we must prevent that object from being a pronoun (see above). 
        In such	cases we use O*n+, which may not connect with pronouns.
        This yields:

        	gave: ... & ((O+ & O*n+) or (O+) or (O+ & [[@MV+]] & O*n+) or
        	([[@MV+]] & O*n+))

        This can be somewhat condensed into the following:

        	gave: ... & ((O+ & {{[[@MV+]]} & O*n+} or [[@MV+ & O*n+]]);

        This then yields the following judgments:

        	I gave my brother an expensive present
        	I gave him an expensive present
        	I gave an expensive present
        	I gave it
        	*I gave my brother it
        	I gave him for his birthday an expensive present 
        		(stage 2)
        	I gave my brother for his birthday an expensive present
        		(stage 2)
        	*I gave him for his birthday it
        	I gave for his birthday an expensive present (stage 2)
        	*I gave for his birthday it
        	*I gave for his birthday him an expensive present
        	*I gave for his birthday my brother an expensive present
        	*I gave for his birthday my brother it		
        
        _@MV+ on adjectives_
        MV is also used to attach modifying phrases to adjectives.
        This is only allowed when the adjective is being used
        predicatively: "He is HAPPY ABOUT his job", "*He is a happy
        about his job man". As with verbs, "@MV+" is optionally
        conjoined with complement connectors like "TO+" and "TH+" on
        adjectives, and is to the left of the complement connectors.
        Thus when adjectives participate in long phrases like "He
        wants to be certain that the John is eager to go", the "@MV+"
        on "certain" and "eager" cannot be used.

        _Words taking MV-: adverbs and prepositions_
        If a word or expression has an "MV-", that means it can modify
        a verb that precedes it. A number of different subscripts are
        used to identify different kinds of modifiers.  For the most
        part, these subscripts distinctions are just provided for the
        sake of clarity. They are rarely used to constrain linkages,
        although a few subscripts have special properties either in
        post-processing or at the linkage level.

        MVa connects adverbs and to verbs; adverbs thus have "MVa-"
        connectors. Many adverbs can take optional commas; they thus
        carry "{Xc+ & Xd-} & MVa-" (see "Xc"). Some "clausal" adverbs
        can _only_ be used in this position with commas:

        	*He is angry apparently.
        	He is angry, apparently.

        These adverbs thus carry "Xc+ & Xd- & MVa-". 

        MVp connects prepositions to verbs; prepositions have "(MVp-
        or Mp- or Pp-) & J+" (The J+ connects to the prepositional
        object; the Mp- is used when the preposition modifies a noun;
        the Pp- is used when the prepositional phrase is a complement
        of "be").

        For the most part, prepositions and adverbs form clearly
        distinct natural kinds.  Prepositions take an object, can
        modify nouns, and can be complements of "be"; adverbs differ
        in all these respects. But there are many words and phrases
        which are like adverbs in some ways (e.g., they take no
        object), but like prepositions in others. "Alone",
        "everywhere" and "downtown" can be complements of "be" and can
        modify nouns; "backwards" and "somewhere" can be
        "be"-complements but cannot modify nouns. In any case, whether
        something is labeled MVp or MVa has few practical
        consequences.

        It will be noted that we do not distinguish between the many
        kinds of prepositional phrases: time expressions ("on Monday"),
        place expressions ("in the plane"), manner expressions ("with
        skill and discretion"), special verb- or adjective- 
        complement expressions ("She prepared for the meeting", "She is
        angry about the decision"). There seems to be no reason to make
        these distinctions for our purposes. One problem is that there
        are certain verbs which seem to require a certain prepositional
        complement: "He hoped for an agreement", "*He hoped on Tuesday",
        "*He hoped". We could solve this problem by creating special
        connectors for such prepositions; at present, we simply give
        such verbs optional "@MV+" connectors, thus allowing 
        questionable sentences like the ones above.
        	
        MVs- is used for certain conjunctions (i.e. words that can
        take clauses), such as "while", "because", and "now_that". Note
        that some conjunctions, such as "after", can take either objects or
        clauses as complements; we usually treat these as prepositions,
        giving them "MVp". Note also that coordinating conjunctions like
        "although" or "but", which seem to connect two equal phrases,
        take "CC-", not "MVs-"; thus they connect to the subject of the
        preceding clause, not the main verb. This is because such 
        conjunctions may not be used in relative clauses (unlike other
        conjunctions and prepositions, which may be so used):

        	The woman you said you liked on Tuesday is here
        	The woman we saw after we left the party is here
        	*The woman you said you liked but she was too intelligent is
        		here
        
        See "W" for further explanation.	

        _Other uses of MVp and MVs_ As well as taking noun-phrases
        (using J+) and clauses (using C+), some prepositions and
        conjunctions may take present participles, passive participles
        or indirect questions:

                     +----MV-----+-Mg-+
                     |           |    |
        	I yelled at her for going to the party

                      +--MV-+-Mv-+
                      |     |    |
        	She cried when asked about it

                     +-MV--+-QI-+
                     |     |    |
        	I talked about how to use the program		

        Such words have "(J+ or Mg+ or Mv+ [as appropriate]) & MVp-".
        Some such uses are extremely unconstrained: almost any sentence
        can take a phrase like "by _ing" or "after _ing". Others are
        much more constrained, like "for _ing" (*"I went to the store
        for getting some milk"), and most indirect question uses ("*I went
        to the store about how to use the program"). At present we have
        no way of controlling this. 

        Note that when participles are used with conjunctions, we use
        Mv and Mg rather than Pv and Pg. See "M: Mv and Mg used with	
        conjunctions".

        _MVi_
        MVi is used to connect infinitival phrases to verbs and
        adjectives when they mean "in order to": "He went to the store
        to get some bread". Thus "to" carries "(MVi- or TO-...) &
        I+". (TO is used when infinitival phrases act as verb or
        adjective complements: "He wanted/expected/was eager to go.")
        Any verb or adjective may take such a phrase; it has a cost of
        1, however. 

        _MVl_
        A few adverbs can modify prepositional phrases, conjunctions
        and adverbs: "partly", "even", "largely". Notice that these
        adverbs do not simply modify the previous clause: they
        require a following phrase.

                    +---MVl--+-MVp-+
                    |        |     |
        	He did it largely in his spare time
        	He did it largely voluntarily
        	He did it largely because he wanted to do it
        	*He did it largely

        Such adverbs take "MVl- & (MVp+ or MVa+ or MVs+)". Somewhat
        counterintuitively, then, the prepositional phrase connects
        to the rest of the sentence _through_ the adverb; it no
        longer makes a direct connection. (Notice the subscripts here
        prevent such adverbs from linking in sequence: "*He did it
        largely partly in his spare time".)

        _MVx_
        MVx is used to connect verbs to certain modifying phrases
        surrounded by commas.  These include some kinds of phrases,
        namely passive and progressive participle phrases, that _must_
        be surrounded by commas when they modify verbs (exx. 1-4
        below).

                      +---MVx--+
                      |   +-Xc-+-------Xd------+
                      |   |    |               |               
        1.	John left , carrying a dog   \\\\\
 	2.	John left , followed by Bill \\\\\		
        3.	*John left carrying the dog
        4.	*John left followed by Bill

        5.	John left ,   with the dog   \\\\\
        	
                      +---MVp---+
                      |         |
        6.	John left     with the dog

        MVx is also used with prepositional phrases, which need not be
        surrounded by commas (see exx. 5-6 above).  When they are not,
        however, they use MVp to connect to the verb, not MVx.
        Prepositions therefore have

        with:	J+ & (MVp- or Mp- or Pp- or (Xc+ & Xd- & (MVx- or MX-));

        To connect to the commas, Xc and Xd are used; see "Xc".

        The MVx- on passive and progressive participles is disjoined
        with with their normal left-pointing connectors (Pg- & Mg- for
        progressives, Pv- & Mv- for passives). However, the MVx- on
        participles must be to the right of the complement connectors
        on the expression; the Pg/V/M connectors must be to the
        left. Therefore the MVx connectors on participles must be
        fully disjoined with the Pg/V/M connectors.  In this respect
        MVx- is much like MX*p- and COp+; see "CO: Participles as
        Openers" for more explanation.

        _Special Purpose MV- connectors_
        A number of MV connector types are used to trigger post-processing
        constraints. 

        MVt, MVm, MVy, and MVz are used in comparatives. The subscripts
        are used by post-processing to enforce a variety of constraints
        on the way comparatives are used. See "MV: Comparatives" below.

        MVh is used to attach the word "that" to verbs, in "so...that"
        "such...that" constructions: "He was so angry that he left".
        "She was such a good programmer that they had to keep her."
        These uses of "so" and "such" are specially subscripted with
        "k" (EAxk, EExk, Dm#k, Ds#k).  Post-processing ensures that
        MVh is only used when a special "so" or "such" connector is
        present.  In addition, MVh is a stage 2 connector; it is only
        considered in stage 2, thus preventing a huge number of
        spurious parses.

        The only use of subscripts to actually constrain linkages
        (with the exception of the post-processing features described
        above) is as follows.  There are certain uses where only a
        prepositional phrase may be used. Certain prepositions seems
        to take other prepositions as objects. They cannot take any
        kind of adverb as objects, though.

                     +-MVp-+MVp+MVp+
        	     |     |   |   |
        We were drinking  over out by the lake
        *We were drinking over
        *We were drinking over happily

        Such prepositions therefore carry MVp+ as a possible right-branching
        connector along with J+, Mg+, etc..

        _Comparatives_

        This is perhaps a good place to describe the system's handling
        of comparatives. Comparatives are constructions involving
        "more...than", or "as...as". Each comparative construction has
        two sides: a first half (involving "more" or "as") and a second half
        (involving "than" or "as"). (The two different "as"'s are given
        different entries in the dictionary, labeled "as.y" and "as.z",
        and we shall refer to them that way here.) We will begin by
        describing the "more...than" case; the "as...as" case is quite
        similar, though somewhat simpler.

        A wide variety of constructions are possible on both halves of
        a comparative expression.
 
        1.He is more intelligent  THAN	John		all
        2.He is bigger			John is   	1,2
        3.He runs more quickly		John does       all except 1,2,11
        4.I have more CDs		is John         1,2
        5.He earns more money		does Fred       all except 1,2,11
        6.He plays with more skill	for money	all
        7.More people attended the partyhe was last year1,2
        8.He earns more			last year       all
        9.He did it for more		attractive	1
        10.He plays more for pleasure	Fred earns      4,5,6,7,8,10
        11.He is more a scholar		came to the concert 4,5,6,7,8,10
 	12.He plays football more 	I work          3,12
        				you have tapes  3,4,12
        				elegantly       3,9
        				he said he earns 4,5,6,7,8,10
        				he said he was  1,2
        				I had expected  all
        				had been expected all
        				expected        all

        It seems that there is tremendous freedom in the way "more" 
        phrases and "than" phrases can be formed. However, there are
        strong constraints on the way they may be combined. By each
        "than" phrase above are listed the possible "more" phrases that they
        may be combined with. These constraints are not easy to enforce
        for our system through ordinary link logic. Nor are they easy
        to enforce with simple post-processing logic; very often the
        "than" phrase is in an embedded clause (as in several of
        the examples above), and therefore not in the same group as the
        "more" phrase. Therefore, our enforcement of these rules relies
        on a rather elaborate combination of link logic and post-processing.
        First we will describe the construction of "more" phrases; 
        then we will turn to the more complex construction of "than" 
        phrases.

        _The Construction of "More" Phrases_
        In the "more" phrases above, the word "more" is usually
        serving a pretty clear function. In the first case above, "He
        is more intelligent...", for example, it is serving as an
        adjectival adverb. This is clear because it may not be
        combined with another such adverb: *"He is more very
        intelligent". Thus we give it specially subscripted "EA"
        connector:

                    +---Pa------+
                 +S-+   +---EAm-+
                 |  |   |       |
        	He is more intelligent than...

        The other uses of "more" are similar; in each case, it is simply
        acting like a determiner, noun phrase (object or prepositional
        object), or adverb, and can simply link to the rest of the 
        sentence as if it were a normal member of these categories.
        In each case, however, the connector is specially subscripted
        (solely for post-processing purposes; this will be explained
        below).

        He is more intelligent		EAm+
        He runs more quickly		EEm+
        I have more CDs			Dm*m+			    
        He earns more money		Dm*m+			    
        He plays with more skill	Dm*m+			    
        More people attended the party	Dm*m+
        He earns more			Om-			    
        He did it for more		Jm-
        He does it more for pleasure	MVm-			    
        He is more a scholar		EB*m-

        A few other words can also perform the function of "more",
        marking a phrase as the left half of a comparative. These
        include adjectives such as "bigger" and adverbs such as
        "better" and "further". Again, in other respects these
        words are just like others in the same category. Therefore
        we give them ordinary adjective or adverb expressions but
        with special subscripts.

        He is taller		Pam
        He is a taller man	Am
        He plays better		MVm	

        Finally, the words "less" and "fewer" are very similar in
        function to "more", and have the same connectors. "Less"
        is identical to "more" in functioning as an adverb, mass
        noun phrase, and mass determiner; "fewer" duplicates the
        functions of "more" as a plural noun phrase and plural
        determiner.

        _The Construction of Than Phrases_
        The connection of "than" to the rest of the sentence is
        simple; it has an MVt- connector, and thus connects to the
        main verb of the sentence like a prepositional phrase. The MVt
        on "than" is restricted by post-processing; it cannot occur
        unless one of the "more" connectors listed above appears.
        This prevents "than" phrases from occurring in non-comparative
        contexts (*He is intelligent than me), and also in "as"
        contexts (*He is as intelligent than me).

        Now, how do we construct "than" phrases, and how do we enforce
        the constraints between "than" phrases and "more" phrases?  It
        is useful here to distinguish between two types of "than"
        phrase.  Some types are "unconstrained"; others are
        "constrained". We will discuss the unconstrained ones first.

        With some "than" phrases, there seems to be almost no limits on
        the kind of "more" phrases they can be combined with. One such
        type is noun phrases. 

        	He (is more intelligent) (runs more quickly) (earns more)
        		than John

        It seems here that such a complement can be used with any
        kind of "more" phrase. Therefore we give "than" an O*c+
        connector. No post-processing is needed here; we have already
        required that _some_ kind of "more" phrase must be used
        every time "MVt" is used. So far, then we have

        	than: MVt- & (O*c+...)
        			
        Another kind of "unconstrained" use of "than" is with a
        prepositional phrase.  Again, it seems that any use of "more"
        is possible:

        He (is more intelligent) (runs more quickly) (earns more)
        	than in the past

        Thus we give "than" an Mpc+, allowing it to connect to any
        prepositional phrase. Finally, "than" may take a phrase
        like "was expected", or "I had expected", combined with
        any use of "more":

        He (is more intelligent) (runs more quickly) (earns more)
        	than (I had expected) (was expected)

        For this, "than" is given an Zc connector. Verbs like
        "expect" have Z-, disjoined with their other complement
        connectors. (Z is also used in subordinate "as"-phrases
        like "He left, as I expected". See "Z".)

        	expect: (S-...) & (TH+ or TO+ or Z-)		
        							    
        This yields
        		       +----Z----+
        		       |   +--S--+
        	               |   |     |                		
        	He earns more than I expected			
    
        The second case, "He earns more than was expected", is more
        problematic, and requires a further SFsic connector on
        "than". See "Z" for more explanation.

        _Constrained "Than" Phrases_
        Other "than" phrases are more constrained. These divide into	
        two groups: those where the "than" phrase contains a clause,
        and those where it does not. Regarding the latter case, 
        a "than" phrase may contain an adjective, but only if the
        "more" phrase contains a predicative adjective as well as
        the adverb "more". "Than" is therefore given Pafc+.

                   +-----Pa-----+
        	   +EB*m+       |        +---Pafc-+
        	   |    |       |        |        |
        	He is more intelligent than attractive
        	*He has more money than attractive
        	*He is smarter than attractive
        	*More intelligent people came than attractive

        The "than" phrase may contain an adverb or prepositional
        phrase, but only if the "more" phrase also does, with "more"
        acting as an adverb.  Thus we give "than" MV*c+.

        	     +--MVp/a--+
                     +-MVm+    |         +-MVp/ac+
        	     |    |    |         |       |
        	He plays more for money than    for pleasure
        	He plays more quickly   than    elegantly

        	*He has more CDs than elegantly
        	?He has more CDs than for pleasure
        
        (The last sentence is perhaps valid if "for pleasure" is being
        construed in an unconstrained sense: "He has more CDs than he
        does for pleasure". And we allow this.)  Finally, the "than"
        phrase can be a mass or plural noun phrase, but only when
        "more" is acting as a determiner. In this case, the "than"
        seems to fulfill both the determiner and the main demand of
        the noun; therefore we use the "U-" connector on nouns, which
        overrides both these demands.

                        +-Dmum+     +U*c+
                        |     |     |   |
        	He has more money than time
        	?He has more money than a hobby

        (Here again, an unconstrained interpretation is possible--and
        is allowed--with both sentences.) How do we prevent these
        constrained "than" uses from being mixed and matched: *"He
        plays more quickly than intelligent", *"He is more intelligent
        than quickly"?  In each case, the particular words that are
        needed in the "more" phrase, and the particular use of "more",
        can be identified by their link-names.  Therefore, we can
        simply insist in post-processing that each "than" connector
        may only be used with the appropriate "more" connectors.
        Thus, "Pafc" demands "Pa" as well as "EB*m"; "MVpc" and "MVac"
        demand either "MVp" or "MVa"; and"U*c" demands "Dm*m".

        A "than" phrase can also contain a subjectless clause:

        	1.More people attended the party than came to the concert
        	2.More money was committed than was available
        	3.*More money was committed than were available

        To handle this, we give "than" an S**c connector. No new
        domain is started here. For this usage, "more" must either be
        serving as a noun-phrase (as in 1 above), or as determiner of
        a noun (as in 2); and in the latter case, the verb of the
        "than" clause must agree in number with this noun (see 3
        above). The number of the noun in the "more" phrase is
        indicated by the D connector; if it is a "Dmum", the noun is
        singular, whereas if it is "Dmcm", the noun is plural.  If the
        "than" connects to a singular verb (as in 1), a Ss*c will be
        formed. In post-processing, Ss*c demands an Om, Jm, or Dmum;
        an Sp*c demands an Om, Jm, or Dmcm. In this way we enforce
        number agreement.
        
        So far, then, we have:

        than: MVt- & (O*c+ or Mpc+ or ({SFsic+} & Zc+) or S**c+ or
        U*c+ or Pafc+)

        _Constrained Uses with Separate Clauses_

        In all the constrained cases discussed so far, the "more"
        phrase and the "than" phrase are in the same post-processing
        group. This seems logical, since groups usually correspond to
        clauses, and in these cases both phrases seem to be in the
        same clause. Indeed, it is necessary for both phrases to be in
        the same group for the post-processing rules to work. In other
        cases, however, the "than" phrase seems to contain its own
        clause.

        1.	He is more intelligent than I am
        2.	He earns more than I earn
        3.	I run more quickly than he does
        4.	*He is more intelligent than I do
        5.	*He earns more than I am
        6.	*He runs more quickly than I earn

        In the first case, we can use the AF- connector on verbs like
        "be" and "seem" which take adjective complements (used also in
        adjectival questions: "how intelligent is he?"). In the second
        case, we can use the B- connector on verbs, used with fronted
        objects in questions and relatives ("What do you earn?").  In
        the third case, we need to create a special connector.
        Normally, auxiliary verbs and modals like "do" and "can"
        require a main verb: "*I do", "*I can". In this case, however,
        the comparative seems to satisfy this need. Therefore, we give
        auxiliaries "CX-" disjoined with their main verb (I+ or PP+)
        connectors.  This yields the following linkages:

  					 +--AF-+
        		                 |     |
        	He is more intelligent than  I am

  					 +--B--+
        		                 |     |
        	He earns more money    than I earn

  					 +--CX-+
        		                 |     |
        	He runs more quickly   than  I do

        So far, then, "than" has the following:

        than: MVt- & (O*c+ or Mpc+ or ({SFsic+} & Zc+) or S**c+ or
        U*c+ or Pafc+ or AFc- or Bc- or CX-)

        These "than" phrases are clearly constrained, as shown by
        sentences 3-6 above. But in order to enforce these constraints
        as we do in the earlier cases, the "than" phrase and the
        "more" phrase have to be in the same group. Since normally
        groups correspond to clauses, it is not at all clear that the
        two phrases would be in the same group in these cases.  In
        simple cases such as the ones above we could perhaps prevent a
        new domain from being started at the "than" phrase.  However,
        in other cases, the "than" phrase may include an embedded
        clause: "He is smarter than I think he is".  Normally, the C+
        on "think" would start a new domain in this case.  But if that
        happens, how do we enforce post-processing constraints between
        the "more" phrase and the "than" phrase?

        Consider the following structure:

                     +-------MVto----+
                     +------O---+    +--Bc(e)-+
                     |    +Dmum-+    |   +S(e)+
                     |    |     |    |   |    |
        	He earns more money than I  earn

        Notice first of all that the "than" is making an "MVto"
        connector to the left, rather than the "MVt" described
        earlier. This connector starts an 's' domain, which contains
        the whole "than" phrase. (The "Bc" is a restricted link, as
        are the AFd and CX links described above; this prevents the
        subordinate group from spreading back to the rest of the
        sentence.) What we want to do is enforce a constraint on the
        "B" connector. The "Bc" connector may only be used if a "Dm#m"
        is present in the "more" phrase. The "Bc" connector and the
        "Dm#m" are not in the same group, so this cannot be enforced
        directly. Note, however, that the "Bc" connects to "than",
        which is making the "MVto" connector, and this MVto _is_ in
        the same group as the "Dm*m". Therefore, we can use ordinary
        link logic to ensure that the Bc only occurs with MVto:

        than: (MVt- & (O*c+ or Mpc+ or...)) or (MVto- & Bc+);
        
        and we can then use post-processing to ensure that MVto only
        occurs when a Dm*m is present in the same group.  In effect,
        then, we use link logic to communicate a post-processing
        constraint from one group to another.

        A similar process is used with the CX and AF cases. CX
        requires a determiner, noun-phrase, or adverb use of "more";
        AF requires a adjectival-adverb use (or a comparative
        adjective like "bigger" or "better"). CX+ and AF+ on "than"
        are each conjoined with a special form of "MVt" (MVtp and
        MVta, respectively). In both cases a new group is started for
        the "than" phrase; but in both cases, the "MVt*" is in the
        outer group, thus its use (and indirectly the use of the CX
        and AF) can be constrained depending on the links present in
        the "more" phrase. By doing this, we can handle not only
        simple cases like those above, but also more complex cases
        where the "than" phrase is in an embedded clause, like "He
        earns more money than I thought he did".
        
        In cases where the than-phrase contains "be" or an auxiliary,
        subject-verb inversion may occur:

        	He is more intelligent than am I
        	He runs more quickly than do I

        To allow the first of these, we give "than" a "PF" connector.
        Forms of "be" have PF-; this is used in certain cases where
        the complement demand of "be" is fulfilled by something
        preceding.  To handle the second, we give auxiliaries "CQ-"
        connectors, conjoined with their SI connectors (unlike CX-,
        which is conjoined with their S connectors). Auxiliaries thus
        have:

        	do: (S- & (I+ or CX-...)) or (SI- & (I+ & CQ-))

        (Since both CQ and PF occur with s/v-inversion, they
        must be added to the "compatible with inversion" and
        "requires inversion" lists in post-processing.)
        
        In terms of the constraints on their use, the PF case is
        identical to the AF case described above; the CQ case is
        identical to the CX case.  We already have a system in place
        for enforcing the right constraints for AFd and CX; thus PFc
        and CQ can simply be disjoined with them. This yields:

        than: ... or (MVta- & (AFd+ or QIc+)) or (MVtp- & (CX+ or CQ+))

        One more case of than-phrases in a separate clause is easy to
        deal with. A "than" phrase may contain a complete clause:

                        +--EEc-+     +-Cc-+
        		|      |     |    |
        	I ran more quickly than  he painted his house
        	
        This usage is constrained: the "more" phrase must must be
        acting either as an adverbial-adverb or a verb adverb ("I ran
        more").  In such a case, it is most logical to give "than" a
        C+ connector. C connectors are not (normally) in the domains
        they start; thus the Cc here is in the same group as the
        "more" phrase, and the constraints can be easily enforced.

        _Comparatives Involving Adjectival Clausal Complements_
        A special kind of comparative involves a "more" phrase with
        an adjective plus its complement; the "than" phrase then
        includes a similar complement, a similar main clause, or both:

 	I am more confident that Joe will come than I am that Fred will go
                           			    that Fred will go
        					    I was
 	It is more likely that Joe will come than it is that Fred will go
        					  that Fred will go
        				          it was
        *					  Fred is that Joe will go

        So far, we have made no provision for "than" phrases to take
        such complements at all. We have dealt with the case of
        more-phrases containing adjectives: "He is more intelligent
        than I am". As described above, an AFd+ connector on "than" is
        used for attaching to the verb of the "than" phrase; this is
        conjoined with a MVta-. We now add complement connectors like
        those on adjectives: THc+, TOic+, and TOc+. Either the "be"
        phrase may occur, or the complement phrase, or both. This then
        yields:

        than: ... or (MVta- & (AFd- or ({AFd-} & (THc+ or TOic+ or TOc+))));

        We have already enforced that the MVta will only occur when
        the "more" phrase contains an adjective modified with "more".
        Various other constraints must be enforced here. If the
        subject of the first phrase is filler "it" (that is, if there
        is a adjective-complement phrase present that can only be used
        with "it"), then the subject of the second phrase must be as
        well.  Therefore, we add the complex MVti- & AFdi+. This works
        the same way as the MVtx complex, except that MVti is only
        allowed with filler-it, while the MVta is only allowed with
        ordinary subjects. Thus we prevent the final incorrect
        sentence above. This yields:

        than: ... or (MVta- & AFd+) or (MVti- & AFdi+) or 
        ((MVta- & {AFd+}) or (MVti- & {AFdi+}) & (THc+ or TOic+ or 
        	TOc+ or (TOt+ & B+)))

        However, there is a problem here. Normally, in order to
        enforce these constraints, the MVta must be in the same
        group as the "more" phrase. But when an adjective takes
        a complement such as TOi or TH, the complement usually
        starts a new group:
 
            +----Pa---+                ?--MV--+
         +SF+   +--EAc+-TH-+--C-+S(e)+        | 
         |  |   |     |    |    |    |        |
        It is more likely that John left    than it is that...

        and there is no way for anything to the right to make an
        MV connection back to the outer group. Thus, in the sentence
        above, there is no way for the MVta to be in the same group
        as the outer group of the "more" phrase. 

        To fix this, we give adjectives "LE" connectors, disjoined
        with their complement connectors (if any). "Than" can then
        either make an LE link to the adjective of the "more" phrase
        (if it is taking a complement) or an MVta (if it is not).
        The sentence above therefore gets the following linkages:

            +----Pa---+----------LE-----------+--AF--+
         +SF+   +--EAc+-TH-+--C-+S(e)+        |      |
         |  |   |     |    |    |    |        |      |
        It is more likely that John left    than it is that...
        
        Now the LE connector _is_ in the same group as the "more"
        phrase. So it takes the place of the MVta, enforcing 
        the same constraints: the "more" phrase must contain an EAc,
        and if the "more" phrase contains a filler-it subject, the
        "than" phrase must also.

        A further constraint is that when the "than" phrase contains
        a complement, it must be of the same kind as that in the
        "more" phrase: 

        *It is more important to go than it is that Fred goes. (TOi/TH)
        *It is more important that Fred goes than to stay. (TH/TO)
        *It is more pleasant to use than it is to go. (TOt/TOi)

        To enforce this, The complement connectors on "than" are given
        special subscripts; p.p. insists that these connectors are
        only used when the corresponding connector occurs in the same
        group. (For this to work, the complement connector of the
        "than" phrase must be in the same group as that of the "more"
        phrase. However, we want the "AFd" to be in a different
        group. Thus, rather than having the MVta or LE connectors
        start the group as we do with MVto and MVtp, we have the AFd
        connector here start the group.) This yields the following
        structure:

                                          +----THc----+  
            +----Pa---+---------LE--------+-AF(e)+    |
         +SF+   +--EAc+-TH-+--C-+S(e)+    |   +SF+    |
         |  |   |     |    |    |    |    |   |  |    |
        It is more likely that John left than it is that...

        _Comparatives with "As"_ 
        Comparatives with "as" are quite similar to those with
        "more...  than", only somewhat simpler. The word "as"
        corresponds to "more"; a different word "as", as it were,
        corresponds to the word "than". These two "as"'s are called
        "as.y" and "as.z", respectively.

        It was mentioned that "than" connects to the rest of the
        sentence using a subscripted MV connector (or rather one of
        several: MVt, MVto, MVta, MVtp, MVti).  Post-processing
        carries a list of all the subscript types that may occur with
        "more": EAm, EEm, Dm*m, etc.. If none of these connectors are
        present, than no "more" phrase has occurred, so no "than"
        phrase may occur either: the MVt connectors may not be
        used. Note that a "than" phrase can not be used with an "as"
        phrase, nor vice versa:

        	*He earn as much money than she earns
        	*He earns more money as she earns

        Thus "as.y" connectors must be subscripted differently from
        "more" connectors. Once this is done, we can make a separate
        list of "as.y" connectors; we then give "as.z" "MVz*"
        subscripts. We can then ensure that these subscripts are used
        only with the "as.y" connectors, and the "than" subscripts are
        used only with "more" connectors.

        Many of the same connectors occur with "as.y" as occur with
        "more". With adjectives and adverbs, "as.y" serves as an
        adverb, exactly as "more" does:

        He is as smart		EAy
        He runs as quickly	EEy

        In other cases, "as.y" must connect with "much" or "many" to
        serve the same function as "more". When acting as a verb
        adverb, a noun phrase, or a mass determiner, it must attach
        to "much". When acting as plural determiner, it must attach
        to "many". This uses the "AM" connector. "Much" and "many"
        have "AM-":

        much: AM- & (Dmuy+ or Oy+ or Jy+ or MVy-);
        many: (AM- & Dmcy+);

            +-----O------+             +---Oy---+          +---MVy-+
            |  +-AM-+Dmuy+             |    +-AM+          |   +AM-+
            |  |    |    |             |    |   |          |   |   |
        I have as much money	    I earn as much	I walk as much

        Notice that it is then "much" (or "many") that makes the
        specially subscripted O, J, MV and D connections that will be
        used in post-processing.

        The expression for "as.z" is very similar to that for "than".
        It was mentioned that different subscripts are used for the
        MV- connecting "as" to the left. Beyond this, however, most of
        the connectors for "as.z" phrases are exactly the same as
        those for "than", and are used in the same way. For example,
        "as.z" can form unconstrained phrases using "O*c+" and "Mpc",
        just like "than". It can also make constrained,
        domain-starting phrases using "MVzo- & Bc+" (this is the same
        as the complex on "than", except that a "z" subscript is used
        instead of a "t"). With "than", p.p. ensured that MVto was
        used only when certain connectors were present in the "more"
        phrase: D**m, Om, Jm.  For "as.z", we simply add the
        corresponding "as.y" connectors to this same list (Dmuy, Dmcy,
        Oy, Jy), and we likewise make MVzo require one of those
        connectors on the list. (There is no danger that an "as"
        connector will be used with a "more-than" connector; we have
        already stipulated that an "than" connector requires a "more"
        connector rather than an "as.y" one, and similarly with
        "as.z".)

        The main difference with "as...as" is that its uses are
        somewhat more limited than "more...than":

        	He is more intelligent than attractive
        	*He is as intelligent as attractive
        	He did it more quickly than carefully
        	?He did it as quickly as carefully

        "As" is also a preposition and conjunction, and thus can be
        used in this capacity in many of the kinds of sitations where
        it is used as a comparative. (For this purpose it has yet a
        third dictionary entry, "as.p".) 

        	Joe is coming, as (I have to leave / I expected / a surprise)

        For this reason, many comparative sentences with "as" get
        multiple parses, and many strange-sounding comparative
        constructions will be accepted with "as" as a preposition.

MX	connects nouns to post-nominal noun modifiers surrounded by commas.

                     +----MX--+
                     |  +-Xd--+-----Xc------+
                     |  |     |             | 
        	The dog , a poodle          ,     
        	The dog ,    who was black  , barked loudly
        	The dog ,   with a big nose , barked loudly
        	The dog ,   chasing the cat , barked loudly
        	The dog ,   annoyed by Fred , barked loudly
        	The dog ,   angry at Fred   , barked loudly	

        The MX connectors used here have different subscripts
        depending on the kind of modifying phrase. Most kinds of comma
        modifier phrases correspond to kinds of no-comma modifier
        phrases, which use M rather than MX. However, every word that
        takes MX- requires a comma on either side of its phrase; such
        words therefore take (Xd- & Xc+). See "Xc".

        Nouns may take any number of comma modifiers:

             +--------------MX------------------+
             +--------MX-------+                |
             +----MX---+       |                |
             |  +--Xd--+-Xc+-Xd+-----Xc------+Xd+-------------+
             |  |      |   |   |             |  |             | 
        The dog , a poodle , with black hair , who was pretty , ...

        Nouns can take without-comma modifiers as well as with-comma
        modifiers. However, any comma modifiers must follow any
        no-comma modifiers:

        	The dog with black hair, a poodle, was pretty
        	*The dog, a poodle, with black hair was pretty
        	The dog that I bought, a poodle, with black hair, was pretty
        	*The dog, a poodle, with black hair, that I bought was pretty 

        Nouns therefore have

        	dog: {@M+} & {R+ & Bs+ & {[[@M+]]}} & {@MX+}...

        _Noun phrases as modifiers_
        Nouns can take noun phrases as modifiers with commas. (In
        this case the commas are obligatory; there is no corresponding
        without-comma phrase.) Nouns therefore have MX-, disjoined with
        their main S/O/J complex:

        	dog: Ss- or SIs+ or Os- or Js- or (Xc+ & Xd- & MX-);
        
        Proper nouns can act as noun modifiers too ("My professor,
        Ms. Smith, is very good"); thus they also have
        MX-. Stand-alone determiners are given MX- too, although this
        is rarely used (?"John and Bill, some of my friends, are
        here").

        Prepositions and progressive and passive participles also
        carry MX-. These usages correspond exactly to Mp, Mv, and Mg,
        used for prepositions and participle phrases without
        commas. As with Mv and Mg, the MX- on participles must start a
        domain; thus it is subscripted MX*p-, and this connector is
        domain-starting.

        Relative clauses may also act as comma-modifiers. These differ
        slightly from ordinary no-comma relative clauses, however.
        First of all, the relative pronoun is obligatory, unlike in
        no-comma phrases.  Secondly, the relative pronoun must be
        "who" or "which"; "that" is not allowed.

        	My friend, who you met yesterday, is here
        	*My friend, you met yesterday, is here
        	*My friend, that you met yesterday, is here

        For this reason, we treat comma relative clauses quite
        differently from no-comma ones.  The relative pronoun acts as
        the head of the phrase. In subject relatives, it makes an
        ordinary subject (S) connection to the verb (unlike in
        no-comma relatives, where it makes a RS connection). In object
        relatives, it makes a B connection to the verb (whereas in
        no-comma relatives it makes a C connection to the relative
        subject). Therefore it behaves just like a question
        phrase. Indeed, the MX- on "who" is directly disjoined with
        the W- and QI- connectors used in indirect questions.

        who: (B*w+ or S**w+) & (Ws- or Wq- or QI- or (Xd- & Xc+ & MX*r-));

        	     +--MX--+-S-+
                     |      |   |
        	My friend, who was drunk, left the party

        	     +--MX--+----B----+
        	     |      |         |
        	My friend, who John hated, left the party

        (The subscripts on B+ and S+ are irrelevant to MX-.) One
        complication here is that in subject relative clauses, the
        verb of the relative clause must agree with the main noun:
        "*My friend, who were drunk, left the party."  To handle this,
        we use post-processing. MX+ connectors on nouns are
        subscripted MXs+ or MXp+ for plural and singular.  This
        creates a MXsr or MXpr link with the relative pronoun; these
        links are domain-starting.  The verb that the relative pronoun
        connects to will then form an "Ss*w" or "Sp*w" link. We then
        dictate that a group containing a MXpr must contain a Sp*w;
        one containing a MXsr must contain a Ss*w.

        Prepositional relatives can also act as comma modifiers.  For
        this, prepositions have MX*j-. These connectors act exactly
        like Mj-; see "Mj". Finally, adjective phrases can serve as
        comma modifiers; adjectives therefore carry MX*a-, which is
        similar in function to Ma-.

N	connects the word "not" to preceding auxiliaries.

                    +------+
                    +-N-+  |
                    |   |  |
        	He had not gone
        	He was not speaking to Fred

        "Not" is similar to an ordinary adverb, in that it can be
        inserted between an auxiliary and a participle ("He had
        quickly gone", "He was quietly speaking to Fred"). With
        adverbs, we use E in this situations to connect the adverb to
        the following participle.  However, adverbs may modify such
        participles in any of their uses; for example, in uses of
        progressive participles as openers or noun modifiers ("A man
        quietly speaking to Fred looked up"); "not" may not be used in
        this way ("*The man not speaking to Fred looked up"). Therefore
        we have "not" connect not to the following participle, but to
        the previous auxiliary, using N. Auxiliaries therefore have
        N+ conjoined with their participle connectors (I, Pp, Pg,
        Pv).
        
        The word "not" may also follow the verb "be" with adjectives and
        prepositional phrases; for this we can use EB. See "EB".

ND 	connects numbers with certain expressions which require numerical
        determiners:

                            +-ND--+-Yt-+
        	            |     |    |
        	I saw him three weeks ago

        Also "The store is FIVE MILES away", "FIFTY PERCENT of them
        were women", "THREE OTHER people are coming", "These are the
        TWO BIGGEST buildings in the city". In each of these cases,
        the word on the right of the link - not the number - is used
        to link the word to the rest of the sentence.

        Numbers can also be used to modify the word "more", when it is
        acting as a plural determiner or noun: "Three more are
        coming", "Three more people are coming", "*Three more money is
        needed". Thus "more" has ND- optionally conjoined with its
        plural-determiner and plural-noun connectors.

        Notice that ND is not used for ordinary noun-phrase number
        expressions: "Fifty people came to the party", "Fifty came to
        the party". For these purposes, numbers have ordinary 
        plural determiner (Dmc) and noun (S/O/J) connectors. ND is
        only used when a numerical determiner is required:

        	I saw him three weeks ago
        	*I saw him the weeks ago
        	*I saw him John's weeks ago
        	These are the two biggest buildings
        	*These are the some biggest buildings

        ND on numbers is directly disjoined with the plural determiner
        and noun-phrase connectors, and conjoined with other connectors
        like NN+ and EN-, used in building larger number expressions:

                            +-----EN-----+
                            |     +--NN--+--ND--+
        	            |     |      |      |
        	I saw him about fifty million years ago

NF	is used with NJ in idiomatic number expressions involving "of":

                                +-NJ-+
                  +-NW-+---NF---+ +NS+
                  |    |        | |  |
        He lives two thirds    of a mile from here
        He lives one third     of a mile from here (uses NS instead of NW)
        He died three quarters of a century ago

        He lives      3/4      of a mile from here

        Fractional words like "thirds" and "third" have NF+, as do 
        fractional numbers like "3/4". "Of" has the complex "NF- & NJ+"
        disjoined with everything else. Singular words like "mile" (not
        plural forms: "*He lives two thirds of 5 miles from here") 
        have an optional NJ-. 

        NF and NJ are similar to ND in that they are only used in
        idiomatic expressions such as time as place expressions.
        Fraction words like "third" can also act as noun phrases (with
        the right determiners), and thus they carry "S/O/J"; they can
        also take modifying phrases such as "of" phrases using
        "M-". For this, no special apparatus is needed:

                    +-------S--------------+
                +NS-+-M--+----J---+        |
                |   |    |        |        |
        	A third  of the population is illiterate
              Two thirds of the population is illiterate (NW instead of NS)
        
NI	is used in a few special idiomatic number phrases:

                          
                          +----NIc----+
                          +--NId--+   |
                          +-NIa+  |   +--+
        	          |    |  |   |  |
        	I have between 5 and 20 dogs
        	*I have between dogs and cats

        The first word of the expression basically connects all the
        words together; but the last word of the expression (a number)
        connects the phrase to the outside world. NIa is also used in
        a few idiomatic number expressions: "He is aged 70", "He is on
        flight 714".

NN	connects number words together in series. The last word in the
        series then connects to the rest of the sentence - either
        making a D connection to a noun, an S/O/J connection (in which
        case the number expression acts as a noun phrase), an ND
        connection, etc..

        	  +--NN-+---NN---+--NN---+----D--+
        	  |     |        |       |       |
        	Four hundred thousand million people live here

        Numerals and single-word numbers under 100 ("3", "1000",
        "three", "eighty") have NN+, disjoined with "D+ or ND+ or
        S+...". Thus they must either connect forward to another
        number word, or serve as the head of the number expression;
        they cannot connect backward to another number word. The words
        "hundred" and "thousand" have "(NN- & NN+) or ({NN-} & (ND+ or
        Dmc+ or S+...)".  They must connect backwards to another
        number; they may either connect forwards to another number or
        serve as the head of the expression. This yields

        	Four people live here
        	*Four three six people live here
        	*Million four people live here
        	*Million four hundred people live here
        	*Million people live here
        	Four hundred million people live here

        Whatever numbers is serving as the head of the expression may
        also make a DD connection back to a definite or possessive
        determiner: "THE four hundred MILLION people living in New York 
        will suffer." See "DD". 

NR	connects fraction words with superlatives:

                       +------F----+
                       |    +--NR--+
        	       |    |      |
        	It is the third biggest city in China

        Superlative adjectives connect back to the previous determiner,
        rather than connecting forward to the noun like most adjectives.
        See "L".

NS 	serves a similar function to ND, only for singular expressions.
        (See "ND".) The words "one", "1", and "a (an)" therefore carry "NS+";
        singular words like "week", used in idiomatic number expressions,
        carry NS-. 

        		  +-NS+
        	          |   |   
        	I saw him a week ago

NW	is used in idiomatic number expressions. It is exactly like ND, except
        that whereas both word-numbers ("five") and numerical numbers ("5")
        have ND+, only word numbers have NW+. The only use of it right now
        is in building fraction expressions:

                 +-NW-+
        	 |    |
        	Two thirds of the students are women

O	connects transitive verbs to direct or indirect objects:

                          +----O----+
                          |         |
        	The dog chased the cat

        Some verbs have optional O connectors; they may or may not be
        transitive ("We moved"; "We moved it"). Some verbs can take
        an object followed by some other complement; such verbs have
        other connectors like Pa, TOo, I, Pg, or TH+ optionally conjoined
        with O+: 

                     +---?---+
                     +-O-+   |
                     |   |   |
                We made him do it   (I)
                We saw  him running   (Pg)
                We find him stupid   (Pa)
                We told him to leave   (TOo)
                We told him he was in trouble   (TH)

        Other verbs have two O+ connectors, one or both of which may
        be optional ("I gave him five dollars", "I gave five
        dollars"). In this case, the first object may either be a
        pronoun or a noun; however, if it is a noun, the second may
        not be a pronoun: "I gave him the money", "I gave John the
        money", "*I gave John it", "*I gave him it". This is parallel
        to the case of particles; in transitive verbs which take
        particles like "up" or "out", the particle may not precede a
        pronoun ("We sorted out them").  The O*n+/Ox- subscripts,
        developed for that purpose, are used here as well. The second
        O connector on two-object verbs has O*n+; pronouns have Ox-;
        "Oxn" is prohibited in post-processing. (See "K".)

        Os and Op connectors mark nouns as being singular or plural. The
        main reason for this is to enforce the correct use of "there is"/
        "there are". See "SF: 'There' as subject".

        Osi and Opi are used in the construction:

                      +------Bs----+
                      +----R---+   |
                   +SF+-Osi+   +-RS+
        	   |  |    |   |   |
        	1.It  is John who wants to do it

        Forms of "be" are thus able to take a noun (proper or common)
        plus a relative clause. Forms of "be" thus carry

        is: (Ss-) & (Pg+ or Pv+ or O*t+ ... or (Osi+ & R+ & Bs+) or
        	(Opi+ & R+ & Bp+)

        Note that the O connectors used in this expression are different
        from the ordinary O connector used in "They are professors".
        Note also that there must be number agreement between the
        noun and the verb of the relative: "*It was my father who wanted
        to do it", "*It was my parents who wants to do it". This is
        enforced by the linkage expressions above. There are other
        constraints on this construction as well:

        	2.*The man was John who did it
        	3.*I saw John who did it
        	4.*It wanted to be John who did it

        In this construction, the subject must be "it" (see ex.2); the
        verb must be "be" (ex.3); and there are constraints on the
        other verbs that may be used in the expression (ex.4). These
        are the same constraints that are already in place for the use
        of "filler-it". Thus we must simply add "Osi" and "Opi" to the
        list of connectors that require "filler-it" as subject, and
        everything else follows automatically. (See SF.)

OD	is used for a few verbs like "rise" and "fall" which can take
        expressions of distance as complements. 

        	    +----OD----+
                    |     +-ND-+
        	    |     |    |
        	It fell five feet

        Words like feet therefore have "OD-". (The special ".i" entry
        for distance nouns, used in other distance expressions, is
        used here.)  In financial writing, one often sees such verbs
        used with "points", or just ordinary numbers; therefore we
        allow this also:
        
        	GM stock fell five points
        	GM stock fell 2 1/2

OF	connects certain verbs and adjectives to the word "of". In
        most cases "of" can not modify verbs: "*I ran of the
        house". With a few verbs it can however: "I accused him of the
        crime", "I just thought of something". It can also modify
        certain adjectives: "I'm proud of you". Such verbs or
        adjectives have OF+ disjoined with their other complement
        connectors. "Of" has "J+ & (OF- or Mp-)"; to connect to its
        object, the usual "J+" is used.

ON	is used to connect the preposition "on" to certain time expressions.

        			      +--TY-+
                           +--ON-+-TD-+ +-Xd+-Xc+
                           |     |    | |   |   |
        	I saw him on January 21 , 1990 \\\\\

        See "DT" for more discussion of time expressions.

OT	is used for a few verbs like "last" which can take time
        expressions as objects:

        	     +-----OT----+
          	     |     +--ND-+
        	     |     |     |
        	It lasted five years
        	?I've been working on this five years
        	*It lasted five books
        
        (Constructions like the second above are sometimes seen; we
        disallow them.)  In questions, the object of such verbs may be
        fronted; in such cases, BT is used. In such cases, the phrase
        "how many" must precede the noun. This is analagous to an
        ordinary "how many" object-type question, like "how many dogs
        did you chase". BT is analogous to the usual B; TQ is
        analogous to Dmc.

                           +-----BT----+
                 +-H--+-TQ-+           |
                 |    |    |           |
        	How many years did it last

        Thus we give "many" "H- & TQ+"; we give "years.i" "TQ- & BT".

P	is used to link forms of the verb "be" to various words that
        can be its complements: prepositions, adjectives, and passive
        and progressive participles.

        	+S-+Pp-+
                |  |   |
        	He is in the yard (Pp)
        	He is running (Pg)
        	He was chosen (Pv)
        	He was angry (Pa)

        Some of these connectors, particularly Pg and Pa, are used
        also with other verbs that take complements of these kinds.

        _Pp_
        Pp is used to attach forms of "be" to prepositions.
        Prepositions thus have "Pp-" directly disjoined with other
        connectors used for attaching prepositional phrases to things
        Mp- (used for phrases modifying nouns), MVp- (used for
        phrases modifying verbs), CO+ (used for openers), and so on.

        _Pg_
        Pg connects verbs that take present participles with present 
        participles.

                    +--Pg--+
        	    |      |
        	I enjoy running

        A number of verbs - "be", "enjoy", "like", "hate", "remember" -
        take present participles as possible complements; such verbs
        have "Pg+" disjoined with other complement connectors (like
        O+, TO+, etc.). A few words take both objects and present
        participles: "I saw him leaving". Such verbs take "O+ & {Pg+}".

        Present participles can also be used with no preceding verb in
        so-called participle modifiers: "The dog chasing John was
        black". Mg is used here, not Pg; this distinction relates to
        post-processing. Present participles can also be used as
        subjects ("Playing the piano is fun"); such "gerund" usages
        use "Ss*g" connectors. See "Ss*g".

        Pgf is used by P.P. to control the use of "it" and "there". See SF.

        _Pv_
        Pv is used to connect forms of "be" to passive participles:

        	      +-Pv+
                      |   |
        	John was hit

        Form of "be" have "Pv+" disjoined with their other complement
        connectors (O+, Pg+, etc.).

        Since the passive form of a verb is always the same as the
        past participle form, the same expression can be used for
        both: the "Pv-" connector is thus disjoined with the "PP-".
        However, the connectors conjoined with Pv- are quite different
        from those conjoined with PP-. First of all, only transitive
        verbs have Pv connectors (*"He was arrived").  Moreover, the
        Pv connector must be disjoined with the O connector on such
        verbs, to prevent "*He was hit the dog".

        When verbs take complement connectors such as "TH+", "TO+",
        and "QI+", the Pv- must usually be disjoined:

                   +--Pv-+
                   |     |
        	I had known of the problem
        	I had known that it was a problem
        	I had known what was happening

              *	John was known of the problem
              *	John was known that it was a problem
              *	John was known what was happening

        The complication here is that, frequently, such constructions
        are permissible when the subject is "it". 

        	It was known that it was a problem
        	It was known what was happening

        We already have a mechanism in post-processing for ensuring
        that certain complement connectors ("THi", "QIi") are only
        used with "it" as the subject (see "SF"); so these can
        be used here. This produces:

        	known: (T- & (O+ or QI+ or TH+ or C+ ....)) or
        	       (Pv- & (QIi+ or THi+...));

        A further complication is that sometimes certain complements
        are permitted _only_ with the passive, for example: "He was known
        to be clever": "*I knew him to be clever". This yields:

        	known: (T- & (O+ or QI+ or TH+ or C+ ....)) or
        	       (Pv- & (QIi+ or THi+ or TO+...));

        If a verb can take an object plus another complement, such as
        an infinitive (O+ & TOo+) or clause (O+ & TH+), the Pv- must
        be disjoined with the O+, conjoined with the other complement
        connector:

                    +--TOo-+
                    +-O-+  |
                    |   |  |
        	I told him to go
                    
                    +-Pv+-TO-+
        	    |   |    |
        	He was told  to go

        	*He was told him to go

        this yields

        	told: (T- & ((O+ or B-) & {TH+ or QI+ or TOo+...})) or
        	      (Pv- & {TH+ or QI+ or TO+});

        (Note that for the passive, "TO" is used rather than "TOo". The
        function of "TOo" is to indicate to post-processing that a new
        subject is in force, by starting a new domain; but with the
        passive form, a new subject is _not_ in force. In "He was told
        to go", "he" is the implied subject of "go".)

        Sometimes one encounters what might be called a "prepositional
        passive". In most cases, a passive cannot be constructed out
        of a verb+preposition phrase: "I went to the house", "*The
        house was gone to"; "I threw a stick at the dog", "*The dog
        was thrown a stick at"; "We ate in the park", "*the park was
        eaten in". There are a few cases of common verb+preposition
        expressions, however, where such passives can be constructed:
        "I've been yelled at, gossiped about, lied to, and trifled
        with". We simply treat these as idiomatic, non-separable
        expressions, similar to passive forms of transitive verbs:

        	yelled_at lied_to: Pv- & {@MV+};

        _Participle Modifiers_
        In participle modifiers - that is to say, in cases where
        the passive participle modifies a noun directly, like
        "The dog chased by Fred was black" - Mv is used, not Pv.
        See "Mv".

        _Pa_
        Pa connects certain verbs to predicative adjectives:

        	     +-S--+-Pa-+
        	     |    |    |
        	The dog was black

        Only certain verbs carry Pa+ connectors ("be", "seem", "look",
        "taste"). A few carry Pa+ conjoined with O+, such as "make"
        and "keep": 

                    +---Pa---+
                +-S-+-O-+    |
                |   |   |    |
        	I made him happy
        
        A few adjectives can act only as prenominals, not predicatives
        ("former", "other"); these have only A+ connectors, no Pa-.

        Many adjectives can take phrasal complements when used in
        predicative position: "She is eager to go", "It is not clear
        who will be hired", "I am certain Joe did it", "He is fond of
        cookies". On such adjectives, Pa+ is conjoined with TO+, TOi+,
        TH+, Ce+, QI+, or OF+ connectors, as appropriate. Pa+ is also
        conjoined with @MV+, allowing prepositional or adverbial
        modifiers ("She is happy with her job"). In all these cases,
        the modifying phrase is optional ("fond" is an exception: "*He
        is fond").
                
        Paf connectors are used for post-processing, to control the
        use of "filler" subjects like "it" and "there". See "SF:
        Constraints on "Filler"-Only Phrases."

        Pa*j is used for verbs like "make" (mentioned above), which
        take object+adjective ("I made him happy"). In such cases, the
        adjective applies to the direct object, not the previous
        subject; thus a new domain must be started which includes the
        O and Pa links but not the S. Pa*j links therefore start "urfl
        domains". Pa*j is exactly analagous to TOo and I*j; see "TOo".

PF	is used in certain direct and indirect questions with "be". 
        Normally forms of "be" require a complement to the right; but
        they can also be satisfied by a question-word to the left like
        "where", "when" or "how".

              +-Wq+-PF-+-SI+
              |   |    |   |
             ||	Where are you

                        +--PF----+
                  +--QI-+    +-S-+
                  |     |    |   |
             I wonder where you are

        Forms of "be" thus have PF- directly disjoined with "O+ or Pg+
        or Pv+...".

        The question-words "where" and "when" have PF- conjoined with
        Wq- or QI-. In questions, then, a Wq- connection to the wall
        is made, enforcing s-v inversion (see "SI"); in indirect questions,
        an QI connection is made, preventing s-v inversion.

        Note that other verbs cannot be used in this way: "*Where ran
        you", "*What think you", "*What have you?" The first two cases
        are prevented because ordinary verbs don't have SI+
        connectors; the third is prevented because, although "have"
        has an SI+, it is conjoined only with PP+ (used in
        past-participles) and not with O+ and B- (used with direct
        objects). Only on forms of "be" is the SI+ conjoined with the
        entire complement expression.

        PF is also used in cases of fronted prepositional and participle
        phrases. 
         
        	  +---------PF--------+-SI-+
        	  |                   |    |
        	Among the candidates was John Smith, a professor
        	Carrying the box     was a small child
        	?Described by John   was a new program

        Such phrases vary widely in their acceptability, depending on
        the participle or preposition; we allow all participles and
        most prepositions to be used this way, but only at stage 2.
        (These phrases are not to be confused with openers. Openers
        attach to the subject of the main clause, which does not have
        s-v inversion and can have any verb: "Carrying the box, John
        left". Fronted prepositional or participle constructions
        involve s-v inversion; the phrase must attach to a form of
        "be".) In such cases, the fronted phrase must attach to the
        wall. S-v inversion is necessary here ("*Among the candidates
        John Smith was"); thus the participle makes a Wq connection to
        the wall, enforcing s-v inversion in post-processing.

        	among:    J+ & (Mp- or MVp- or ... [[Wq- & PF+]])
        	carrying: (Pg- & (O+...))   or    
        	    ((O+...) & (Ss*g+ or COp+ or [[Wq- & PF+]]));

        For participles, note that the Wq- & PF+ complex is directly
        disjoined with the COp+ and Ss*g+, and is to the right of the
        complement expression; it is fully disjoined with the Pg,
        which is to the left of its complement expression.  See "CO:
        Participles as Openers" for an explanation.

PP	connects forms of "have" with past participles:

                    +-PP-+
        	    |    |
        	He has gone

        Forms of "have" have "PP+" connectors; past participles have
        "PP-" connectors, conjoined with their complement connectors
        (O, TH, TO, etc.). Since the past participle form of the verb
        is usually the same as the simple past form (which uses an "S"
        connector), we can usually use the same expression for both.

        	died arrived moved purchased: (S- or PP-) & [complement];

        In cases where the simple past and the past participle are
        distinct, however, we must use separate expressions.

        	began went forsook: S- & [complement];
        	begun gone forsaken: PP- & [complement];

        The past participle is, in every case, the same as the passive
        form (where there is one); but the passive complement
        expression is usually quite different from the past-participle
        one, so the past-participle connector PP and the passive
        connector Pv may not be directly disjoined. (See "Pv".)

        PPi is used by post-processing to control the use of "it" and
        "there"; see "SF: Filler-it".
        	
Q	is used in questions, in several different ways. It is
        used to connect to auxiliaries in simple s-v inversion
        questions, when there is no preceding object. In some cases
        the auxiliary connects to the wall (ex. 1 below).  In other
        cases it connects to a question-word like "when", "where", or
        "why" (ex. 2).

                 +-Qd+
        	 |   |            
        	||  Are you going to to the movie

                 +-W+-Q-+
                 |  |   |
        	|| Why did you go

        In noun-focused questions, either object-type or subject-type,
        the question-word connects to the verb ("What did you buy",
        "Who bought that"). In the former case, there is s-v
        inversion, but no "Q" connection is made. Thus the "Q-" is
        optional on auxiliaries.

        Post-processing ensures that when a question-word is used in
        object-type or when-where-why questions (not subject-type
        questions), the outer group of the sentence contains some kind
        of SI; this is because in all such cases, a Wq is used to
        connect to the wall, and this requires an SI in the same
        group. P.P. further ensures that SI connectors can only be
        used when a Wq is present (see "SI"). However, in simple
        yes-no questions, question inversion must occur, yet no
        question word is present. Thus the Q+ on the wall is
        subscripted "Qd", and this is added to the list of connectors
        that permit s-v inversion.

        (Note that the Q- on auxiliaries is conjoined with SI+, but
        disjoined with S-; by link logic it can only be used with
        SI+. Post-processing tightly constrains the use of SI, preventing
        it from occurring in indirect questions, relative clauses, etc.;
        the use of Q is thus automatically constrained as well.)

        Q is also used for questions with prepositions: "IN which room
        WERE you working", "TO whom WERE you speaking". See "JQ",
        "Jw".

        _Adverbial questions: Qe_
        Qe is used to connect adverbs to following auxiliaries in
        adverbial questions:

                             +---I---+
                 +Eeh-+--Qe--+-SI+   |
                 |    |      |   |   |
        	How quickly did you run

        Qe can only be used in questions ("*Very quickly did you
        run").  This is enforced by the fact that the sentence must
        connect to the wall, and can only do so through "how". Again,
        the use of Qe in indirect questions is prevented because Q- on
        auxiliaries is conjoined with SI, whose use is constrained by
        post-processing.

        One false positive must be addressed here, however:

                 +------------B-------+
                 |           +----I---+
                 |    +--Qe--+-SI-+   |
                 |    |      |    |   |
              *	Who quickly did John hit

        We prohibit this in post-processing by requiring that a
        group containing a Qe must contain an EEh.

QI	connects certain verbs and adjectives to question-words, forming
        indirect questions:
        	
        	       +--QI--+
        	       |      |
        	1. I wonder what book he will read    	
        	2. I wonder what will happen
        	3. I wonder what he will do
        	4. I wonder when he will come
        	5. I wonder how big it is

                            +--QI--+
                            |      |
        	I am not certain where he is

        Question-words thus have "(QI- or W-)". (The W- is used in
        direct questions, to attach to the wall). This is conjoined
        with various connectors.  Determiner question-words, like 1
        above, use D**w+; subject-type questions (ex.2) use Ss*w+;
        object-type questions (ex.3) use B*w+; where/when/how
        questions (ex.4) use Cs+, PF+ or TOn+; adjective questions
        (ex.5) use EAh+. See entries for these individual
        connector-types.

        Verbs ("wonder", "know", "ask") and adjectives ("certain",
        "clear") that take indirect questions have QI+, disjoined with
        other complement connectors.

        QIi is used by P.P. to enforce the correct use of "filler"
        "it". Certain phrases can only be used with "it" as the	subject
        ("It is not clear what will happen"; *"I am not clear what will
        happen"); this is enforced in post-processing. "QIi" is therefore
        directly analagous to "THi" (see "THi").

        QI#d is, similarly, used by post-processing. The issue here is
        whether or not the QI connector should begin a domain. In all
        indirect questions, it is clear that the indirect question
        should be in a different group from what precedes.  With
        "where/when/how" constructions ("I wonder when he will come");
        the question-word makes a Cs connection to the new subject,
        beginning an 's' domain; thus the QI connector need not start
        a domain itself. In all other cases, however, the links
        extending to the right of the question-word (B, S, D, etc.)
        are not normally domain-starting. In these cases, then, the QI
        link has to start the domain. Thus we make "QI#d+" a
        domain-starting link, and assign it to question-words like
        "what" and "who"; "where", "when" and "how" have the
        non-domain-starting "QI+".

R	connects nouns to relative clauses. In subject-type
        relatives, it connects to the relative pronoun. In object-
        type relatives, it connects either to the relative pronoun
        or the subject of the relative clause if the relative pronoun is
        omitted.

                     +-----B------+
                     +-R--+C-+-S--+    
                     |    |  |    |                       
             1. The dog that I chased was black

        	     +----B----+
        	     +-R-+RS---+
                     |   |     |
             2. The dog who chased me was black

                     +--B------+
                     +-R-+-S---+
                     |   |     |
             3. The dog  I  chased was black

        Consider the following simplified expressions:

        	dog: ...(R+ & B+)... & (({C- or R-} & S+) or O+ or...);
        	who: R- & (C+ or RS+);

        When a relative pronoun makes a R connection back to a main
        noun, it must make either a C connection to the subject of a
        new clause (in an object-type relative - ex.1 above), or a RS
        connection to a finite verb (in a subject-type relative -
        ex. 2 above). Note that, conjoined with their S+ connectors,
        nouns have not only C- but R-.  Thus, in object-type
        relatives, the main noun can connect directly to the subject
        of the relative clause; no relative pronoun is necessary
        (ex. 3 above).

        The "R+ & B+" complex on nouns is disjoined with the "@M+".
        Nouns frequently take a relative clause, or one or more
        prepositional phrases; they rarely take both. (See "M".)  We
        also do not allow multiple relative clauses. These sound all
        right in some cases ("The movie I saw that I told you about"),
        but in practice seems to be extremely rare.

        _Other Words with "(R+ & B+)"_ 
        Some pronouns and determiners which act as noun phrases have a
        "R+ & B+" complex also: "ALL who apply for the job will be
        considered", "We need SOMEONE who can program well".

        Relative clauses may also occur within commas ("Dave, who you
        met yesterday, is here". In this case, however, a very
        different structure is formed. See "MX*r".

        _Relative Clauses and Post-Processing_
        When a relative clause is created, an 'r' domain is begun, by
        the R connector on the main noun. This 'r' domain spreads
        through the relative clause, and then back through the B
        connector hooking back to the main noun.

                     +-----------S-------+
        	     +-----B(r)-+        |
        	     +R(r)+RS(r)+-O(r)+  |
                     |    |     |     |  |
        	The dog  who chased  me was black        
        
        The point of the 'r' domain is to include all the connectors
        involved in the relative clause.  By ordinary domain logic,
        however, after spreading back through the B connector, the
        domain would then continue to spread to the S connector in the
        main clause and everything else to the right. This is clearly
        undesireable. We therefore create a special list of
        "restricted links" in post-processing. These are links through
        which domains are traced no further, when they extend back to
        the left of the root word (the word on the left end of the
        domain-starting link: the main noun, in this case).

        _Cycles in Relative Clauses_
        It will be noted that the linkages of relative clauses involve
        "cycles"; more links are present than are necessary than to
        simply connect all the words.  Why is the cycle needed? In
        object-type relatives, it seems natural for the main noun to
        connect to the verb of which it is the implied object. But by
        requiring a connection to the subject of the relative clause
        as well, we can prevent question inversion at the linkage
        stage ("*The dog did John chase was black"); hence the C+ on
        relative pronouns.  Once we require the relative pronoun to
        connect to the right in object-type relatives, we must let it
        connect to the right in subject-type relatives as well; hence
        the "RS" connector. (There is another motivation for the RS
        connector, involving embedded clauses within relatives: see
        "RS".)

        As always with cycles, however, there is a danger here of
        strange ill-formed sentences arising, such as "The dog I died
        John chased was black". To prevent this, we include Bp and Bs
        in a list of "must_be_connected_without" connectors.
        Post-processing then insists that the sentence must be fully
        connected even when these connectors are removed: in other
        words, it insists that they must be part of a cycle.

        It was mentioned that R is used when the noun connects to the
        subject of the relative clause rather than to the relative
        pronoun.  R is also used for infinitival phrases following
        nouns: "The TEAM TO beat is Miami". This construction is very
        rare and is given a cost of 2.

RS       is used in subject-type relative clauses. 

        	     +----B----+
        	     +-R-+RS---+
                     |   |     |
        	The dog who chased me was black
        
        Finite verbs have a B- connector disjoined with their S-
        connectors which connects to the main noun of a relative
        clause.  This B- is conjoined with a RS-; when verbs make a B
        connection back to a noun, they must form a RS connection as
        well. This is supplied by the relative pronoun.

        	died arrived: (S- or (RS- & B-));
        	who: (R- & (RS+ or C+);

        See "R" for a full discussion of relative clauses.
        
        A complicated situation arises when the verb of a subject-type
        relative is itself contained in an embedded clause.

                     +------B------------+
                     +-R-+--C-+-S--+-RS--+
                     |   |    |    |     |   
                The dog who John said chased me was black

        The "who" makes a C connection to the outer subject of the
        relative constituent, just as if it was an object-type
        relative.  But in this case, the "B" on the main noun is
        connecting to the relative verb as a subject, not an
        object. The verb therefore requires a RS connection.
        Moreover, the outer verb of the relative constituent, "said",
        requires a complement (normally it makes a "TH" or "C"
        connection to the subject of a embedded clause: "John SAID
        HE was coming"; "*John said"). Thus we give such verbs a "RSb"
        connector, conjoined with their other complement connectors.
        (In this situation, the RS connector must start a domain; this
        is the reason for the "b" subscript.)

        Notice that in the above case - unlike other subject-type
        relatives - the relative pronoun may be omitted: "The dog
        John said chased me was black". This follows naturally from
        the current system.

        _Other "RS" Constructions_
        Because the relative pronoun may be omitted here, and because
        (either via the relative pronoun or directly) the main noun is
        making a clause connection to the subject of the relative
        clause, this situation has a lot in common with object-type
        relative clauses; yet the main noun is the implied subject of
        the embedded clause, not the object. In this way, other
        "B+" connectors can be used that are normally used in
        object-type constructions, for example, those on
        question-words, "whatever"-type words, and "transitive"
        adjectives (the latter case is a little weird, but we allow
        it):

        	Who do you think hit John
        	Whatever you think will work is fine
        	John is easy to think hit Joe
        	
RW	connects the right-hand wall to the left-hand wall in cases
        where the right-hand wall is not needed for punctuation
        purposes. See "X: Comma Phrases at the Beginning and End
        of Sentences".

S	connects subject-nouns to finite verbs:

                     +-S---+
        	     |     |
        	The dog chased the cat

        Ss connects singular nouns to singular verb forms ("The dog chases 
        the cat"); Sp connects plural nouns to plural verb forms ("The dogs
        chase the cat"). Simple-past forms do not distinguish between
        singular and plural ("The dog chased the cat", "The dogs chased
        the cat"); thus an unsubscripted S- is used.

        S##t is used by PP to control sentences like "The problem is
        that John is coming"/"*The dog is that John is coming". See "THb".

        S#i, S#x, and S##i are used to control the use of the pronoun
        "I". "I" normally acts like a plural noun ("I run" / "*I
        runs"), with two exceptions: "I was" / "*I were", and "I am" /
        "*I are".  To control this, we simply give "I" an Sp*i+
        connector; we give "are" and "were" an Spx- connector; and, in
        post-processing, we outlaw Spxi connectors. We also give "am"
        an Spi- connector, and outlaw Spi connectors; "We am" is thus
        prevented, but "I am" forms an Spii link and is thus allowed.

        Ss#w is used for question-words like "who" that can act as
        noun-phrases in subject-type questions: "Who is coming?"
        This subscript serves a rather arcane purpose in post-processing.
        See D##w.

        _Ss*g: Gerunds_
        Ss*g is used for gerunds: "-ing" forms of verbs that act like nouns 
        or heads of noun phrases.

                   +------Ss*g-----+            
                   |               |            
        	Playing the piano is fun	

        This is perhaps a good place for a general discussion of gerunds.

        The use of gerunds is extremely problematic. It involves a huge
        twilight zone of strings which vary rather gradually in their
        grammaticality. There are two basic questions to be answered.
        1. What are the rules governing the construction of the
        actual gerund phrase? 2. What are the rules governing its
        larger context?

        _How can gerund phrases be constructed?_ 
        Gerund phrases can be constructed in two basic ways: A) with
        the normal complement of the verb, and no determiner; or B)
        with a determiner (and other common modifiers often used with
        nouns), but without the normal complement.  Let us call these
        the c/nd case and the nc/d case, respectively.

        _The Complement / No Determiner Case_
        A. Gerunds can be used with their normal complement and no 
        determiner:

        	Sleeping is fun
                Chasing dogs is fun
        	Telling John to leave won't help the situation
        	Telling John that you hate him will destroy your relationship
        	Graduating from college first will make you more marketable

        Whatever requirements apply to the complement as the verb is
        normally used apply to the gerund also (but only in the c/nd 
        case: the nc/d case is quite different, as described
        below):

        	*I talked John
        	*Talking John is fun
        	*I told to do it
        	*Telling to do it won't help the situation
        	*I like to chase
        	?Chasing is fun

        (A possible exception is cases like the last one, where the
        complement is simply omitted from a verb with otherwise
        requires one.  This case will be discussed later.)

        (With strain, gerunds may also be used in this way with a
        possessive determiner: "Your telling John to leave was a
        mistake". For this we use DP; see "DP".)

        To handle cases such as these, we could directly disjoin Ss*g+
        with Pg- on "-ing" forms of verbs (the Pg- is used in present
        participles). There is a problem here, however. On present
        participles, the complement expression must occur to the right
        of the main "Pg- or Mg-" expression. This is because when "B"
        connectors are used, they may need to hook back beyond the Pg
        link:

                 +------B------+
                 |    +---Pg---+
                 |    |        |
        	What are you doing

        With gerunds, however, the reverse is true: the Ss*g link must link
        to the right beyond whatever complement links are made (O, TH,
        etc.):
                   +----Ss*g----------+
                   +-TO-+             |
                   |    |             |
        	 Trying to kiss Susan was stupid

        therefore the entire "Pg" expression must be disjoined with the
        entire "Ss*g+" expression. Because of this, it seems clearer simply
        to treat the present participle and the gerund as two different
        words: the present participles are listed under "trying.v",
        "chasing.v"; the gerunds are listed under "trying.g", "chasing.g".
        The Ss*g+ expression contains the entire complement of the verb.

        So far, then, the usual expression for the ".g" entry for
        gerunds is simply:

        	[normal complement] & Ss*g+;

        _The Determiner/No Complement Usage_
        Gerunds may also be used with a determiner.  In this case,
        they may not take their normal complements.  The situation
        here is complex.

        Take a simple transitive verb like "hit". When a determiner is
        used, taking a simple direct object is clearly wrong (ex. 1).
        Taking no complement sounds questionable (ex. 2). The usual
        thing is to take an "of" phrase (ex. 3).

        	1.*The hitting dogs is fun
        	2.?The hitting is fun
        	3.The hitting of dogs is fun 

        With intransitive verbs, having no complement sounds funny
        also; here again, it's normal for an "of" phrase to be
        used. (Notice that with intransitive gerunds, "of X" names the
        subject of the verb; with transitive gerunds, "of X" names the
        object.)

        	4.?The graduating changes the situation	
        	5.?The sleeping can ruin a lecture
        	6.The graduating of Fred changes the situation	
        	7.The sleeping of students can ruin a lecture

        How about complex verbs? Here again, straightforward use 
        of the complement is definitely wrong (ex. 8-11). Use with
        no complement is iffy (ex. 12-13), as is use with no
        complement (ex. 14-18).

        	8.*The telling John to leave was stupid
        	9.*The showing how to use the program seemed 
        		to interest people
        	10.*The attempting to go to the party really angered Joe
        	11.*The demonstrating that our program 
        		could handle complex sentences impressed people

        	12.?The telling was unfortunate
        	13.?The demonstrating seemed to impress people

        	14.?The telling of John was stupid
        	15.?The showing of the program seemed to impress people
        	16.*The attempting of John to go angered John
        	17.*The telling of John to go was stupid
        	18.?The demonstrating of our program's abilities 
        		impressed people

        Sentences like the last five above are perhaps not so much
        incorrect as unnecessary; in most cases, we have nouns 
        such as "demonstration" and "attempt" which we use instead.

        So, when a determiner is present, using the normal complement
        is wrong with transitives and complex verbs. Having no
        complement sounds doubtful with transitives and complex verbs,
        and also with intransitives (where it corresponds to the
        normal use of the verb).  Using "of" phrases is fine with
        transitives, okay with intransitives, doubtful with complex
        verbs. (This means that with some complex verbs that cannot be
        transitive, like "wish" and "hope", there is no really good
        way of using the gerund with a determiner - which I think is
        true.)  For the moment, we ignore most of these subtle
        distinctions. We allow any gerund to be used with a
        determiner; we allow use with "of" at stage 1 (using OF+); we
        allow use without "of" at stage 2; and we disallow any use of
        the normal complement.  This yields
        	
        	(D- & (OF+ or [[()]])) & Ss*g+;

        When gerunds are modified by determiners, they may also be
        modified by adjectives, relative clauses, and participle
        modifiers. Other determiners may be used, such as
        possessives. Singular-only determiners like "a" cannot be
        used; gerunds seem to act like mass nouns in this sense.  Even
        some mass-noun determiners like "some" and "most" sound
        doubtful, but for now we allow them.

        	The sleeping of students described by Fred is a big problem
        	The sleeping of students I told you about is a big problem
        	The insensitive/frequent/habitual sleeping of students 
        		is a big problem
        	His hitting of the dog didn't help matters
        	?Some hitting of dogs will solve the problem
        	?Most hitting of dogs is unecessary

        This yields:

        {@A-} & Dmu- & (OF+ or [[()]]) & {R+ & Bs+...})) & Ss*g+;

        _Gerund Phrases with Neither Determiner nor Complement_
        What about gerund phrases with neither a determiner nor a complement?
        
        	Running is fun
        	?Chasing should be disallowed
        	?Telling was a bad idea

        With intransitive verbs like "run", such a usage will be
        allowed anyway; an "empty" complement is one possible use of
        the verb. Even with verbs like "chasing" and "telling",
        however, one occasionally sees use of the verb with neither a
        determiner or a complement. We therefore allow this at stage
        2; we incorporate it into the nc/d expression. The expression
        then becomes

        {@AN-} & {@A-} & (Dmu- or [[()]]) 
        & (OF+ or [[()]]) & {@M+} & {R+ & Bs+...} & Ss*g+;

        _Noun-modifiers on Gerunds_
        Adjectival-noun modifiers of gerunds pose a problem.  One
        sometimes sees these used either with a determiner or without.

        	Drug running has fueled the economy here for many years
        	The drug running here has fueled the economy 
        		here for many years
        	?Dog hitting is a big problem
        	?The dog hitting is a big problem

        However, one never sees both a noun-modifier on a 
        gerund and a complement:

        *Dog chasing cats is a big problem
        *Student complaining that the rules are unfair is very common
        
        Therefore, we treat noun-modifiers as part of the nc/d usage
        (which, as described above, does allow stage 2 usage without a
        determiner).

        {@AN-} & {@A-} & (Dmu- or [[()]])
        & (OF+ or [[()]]) & {@M+} & {R+ & Bs+...} & Ss*g+;

        _The Use of Gerunds_
        So much for the way gerund phrases are constructed. Once a valid
        gerund phrase has been constructed, how may it be used?

        It may be noted that, in the above sentences, I have used gerund
        phrases as subjects connecting to a wide variety of verbs. There are
        certainly a large number of verbs that gerunds can be subjects for,
        ranging from absolutely clear-cut cases, to somewhat metaphorical
        cases, to highly metaphorical cases. Many verbs which normally refer
        to actions, which one would think could not be performed by other
        actions, are often used with gerunds. The only kind that are never
        used are those that imply physical actions which are never used
        figuratively, and those that imply a state of mind: e.g. a
        propositional attitude, emotion, etc..

        	Inviting John will cause problems 
        	Inviting John may well destroy the party
        	Inviting John says to Cathy that you don't like her
        	*Inviting John kicked Fred in the pants
        	*Inviting John knows that Fred won't come
        	*Inviting John hopes that he won't come

        There seems to be no good way of limiting these uses,
        especially as we do not do so for ordinary nouns (i.e., we
        allow "The invitation of John knows/hopes that Fred won't
        come").  So, we allow gerunds - both in the nc/d case and the
        c/nd case - to make S links to any verb. (Clearly, both types
        act as singular nouns, not plurals: "*Inviting John cause
        problems.")

        As prepositional objects, gerunds can be used quite freely
        also. Some prepositions take gerunds extremely freely and
        commonly. The nc/d usage sounds acceptable with almost any
        preposition; the c/nd usage is extremely common with some
        prepositions like "by" and "about", less common with others.

        	I caused a problem by inviting John
        	?I caused a problem by the inviting of John
        	I should have talked to you before inviting John
        	?I should have talked to you before the inviting of John
        	We had a discussion about inviting John
        	We had a discussion about the inviting of John
                ?This led to inviting John
                This led to the inviting of John

        However, the c/nd usage has already been handled here in
        another way, using Mg (see "M: Mg and Mv used with
        Conjunctions").  We need only address the nc/d case
        here. (This is another reason for disjoining the nc/d case and
        the c/nd case.) Thus we add J- to the nc/d expression:

        (([normal complement] or @AN-) & Ss*g+) or 
        ({@AN-} & {A-} & (Dmu- or [[()]])... & (Ss+ or J-));	

        How about ordinary verb objects? Here, the c/nd case usually
        sounds quite wrong. There are a few exceptional complex verbs
        where it occurs commonly - more with the c/nd usage than the
        nc/d usage:

        	I hate/enjoy/recommend inviting John
        	?I hate/enjoy/recommend the inviting of John to parties

        For c/nd cases of this type we use Pg. With most other verbs,
        gerunds are used very rarely as objects. It seems that gerunds
        are used figuratively less often as objects than they are as
        subjects. When they are used, it is more often with the nc/d
        usage.

        	We completed/defended/fought the inviting of John
        	*We completed/defended/fought inviting John	
        	We made the inviting of John impossible
        	We expected the inviting of John to cause problems
        	?We expected inviting John to cause problems
        	We gave the inviting of John a thorough discussion
        	*We gave inviting John a thorough discussion
        	*We kicked/believed/shouted/promised/sold the inviting 
        		of John
        	*We kicked/believed/shouted/promised/sold inviting John

        For the moment, we simply allow all nc/d usages as objects,
        and forbid all c/nd usages (except for a few verbs that use
        Pg). This yields the following:

        (([normal complement] or @AN-) & Ss*g+) or 
        ({@AN-} & {A-} & (Dmu- or [[()]])... & (Ss+ or J- or O-));		

        There are a few gerunds that are used as objects with a
        variety of different verbs. Many of these are sports, bodily
        functions discussed in medical terms, and activities which are
        proscribed or illegal. 

        	That portion of the brain controls speaking, 
        		reading, and writing
        	*The law controls selling cigarettes 
        	The disease can cause swelling and itching
        	*The recession caused losing jobs 
        	The law would prohibit/allow/restrict/affect 
        		smoking/fishing in public places
        	*The law would prohibit/allow/restrict/affect 
        		selling cigarettes / eating fish

        The usual usage here is without either a complement or a
        determiner. Recall that this usage is covered under the c/nd
        usage. Therefore, since we allow gerunds to be used as ordinary
        objects under the c/nd usage, these cases will be covered as
        well.

        _"Urfl-only" domains_
        One final point about gerunds. In the c/nd usage of gerunds,
        the gerund phrase appears to constitute a verb expression
        which implies a different subject from the rest of the
        sentence. By the logic of post-processing, then, it seems
        sensible to include the gerund phrase in its own domain. When
        the gerund is acting as a subject, the gerund phrase includes
        everything starting from the gerund word and tracing to the
        right, _underneath_ (and not tracing through) the Ss*g
        link. This domain structure does not correspond to any
        existing domain structure - either normal domains or "urfl"
        domains (see "TOo"). Thus we create a special domain structure
        especially for this purpose: "urfl-only".  Ss*g links then
        start 'd' domains, which are "urfl-only".

                   +--------------Ss*g--------+-----O---+
                   +---MVp(d)--+---J(d)--+    |  +---D--+
                   +-O(d)-+    |    +D(d)+    |  |   +-A+
                   |      |    |    |    |    |  |   |  |
        	Telling John about the party was a bad idea

        ("Urfl" domains include everything "under the root link from
        the left", as well as including everything traced from the
        word on the right end of the root link, like ordinary
        domains. "Urfl-only" domains include _only_ the links that are
        under the root link and reachable from the left.)

        In the nc/d usage, gerunds seem to act much more like simple
        nouns. Therefore there is no need for them to create new
        domains. Thus ordinary S connectors are used here.

        _Other Special S Connectors_
        Ss*t is used for a few nouns that can take "be+that" as
        predicates:

        	The idea was that we would go to London
        	*The vacation was that we could go to London

        We enforce this distinction in post-processing. See "THb".

        Ss*b is used for a few subjects - "it", "this", and "that" -
        that can take the predicate "(be) because (clause)": "This is
        because he is stupid". See "BI". Ss*q is used for subjects
        like "question" that can take the predicate "(be) (indirect
        question)": "The question is why he did it". See "BI".

SF	is a special connector used for certain "filler" subjects
        like "it" and "there". It interacts heavily with
        post-processing.  Post-processing is used both to enforce that
        certain predicates may not be used with "filler" subjects, and
        also that certain predicates may only be used with such
        subjects. SF is also used with a few special phrases like "to"
        and "that" phrases when used as subjects.

        _"Filler-it"_
        Many verbs and adjectives take complements like "to+infinitive"
        or "that+clause":  (see "TO" and "TH")

        	3. I expect that he will go
        	4. I am glad that he is going
        	1. He wants to go
        	2. He is eager to go

        However, there are certain adjective-complement and
        verb-complement phrases that may only be used with the subject
        "it".

        	5. It is likely that John will go
        	6. *John is likely that John will go
        	7. It seems that John should go
                8. *John seems that John should go
        	9. It was suggested that John would go
                10. *John was suggested that John would go
        	11. It is important to go
                12. *John is important to go

        It can be seen in the dictionary that the verbs and adjectives
        on the left of "that" in exs. 5-10 - "seems", "likely" - have
        specially subscripted THi+ connectors, unlike "expect" and
        "glad" which have unsubscripted TH+ connectors. Moreover,
        "wants" and "eager" have TO+ connectors; "important" has,
        instead, a "TOi+" connector.  "It" also has a special "SFsi"
        connector; verbs which "It" may connect to directly are given
        "SFs-" connectors, disjoined with their "S-". ("It" also has
        an ordinary "Ss" connector; see below). Thus the connectors
        that may only be used with "it", as well as "it" itself, are
        marked with special connectors. We use this information in
        post-processing to make the distinctions noted above.

        Recall that post-processing divides the links of a sentence
        into groups, corresponding roughly to subject-verb expressions.
        In post-processing, we insist that any group which contains a
        THi connector or a TOi connector must contain an SFsi
        connector. In this way, we are able to reject ex. 6 above
        while accepting ex. 5.
        
        	+SFsi+--Paf-+-THi+--Ce+-S(e)-+I(e)+
        	|    |      |    |    |      |    |
        	It  is   likely that John  will  go

        	+-Ss-+--Paf-+-THi+--Ce+-S(e)-+I(e)+
        	|    |      |    |    |      |    |
              *John  is   likely that John  will  go

        Of course, "it" may also be used referentially, as an ordinary
        pronoun: "It is black with a long tail". Thus "it" has the
        following:

        	it: SFsi+ or Ss+ or O- or J-;

        A sentence like ex. 5 will thus get two parses, one with 
        SFsi and one with Ss; the one with Ss will be rejected.	This
        is not merely vanity, but has an important function, as
        described below.

        _Constraints on "Filler"-Only Phrases_
        Predicate phrases that may only be used with "it" as the
        subject, such as "likely that" and "important to", might be
        described as "filler-only".  There are other constraints on
        the way such phrases are used, beside the fact that "it" must
        be the subject. Only certain verbs and adjectives may be
        combined with them:

        	*It seems to be likely that John will go
        	*It is glad to be likely that John will go

        Thus, as well as enforcing that "it" is the subject of such
        phrases, we must ensure that they are not used in combination
        with predicates like "glad" and "wants to". Another way of
        thinking about this is that there is a "filler-it" which may
        not be used with "referential-only" predicates and a
        referential-"it" which may not be used with "filler-only"
        predicates. This is in fact the approach we use. The "S"
        connector on "it" corresponds to referential uses, the "SF" to
        "filler" uses.  We have already insisted that ""filler" only"
        predicates (like "likely that") are not used with "S"; we now
        must require that "referential-only" phrases are not used with
        "SF".  This is again enforced with p.p. rules. The verb and
        adjective connectors (I, T, Pg, Pa, Pv, TO) on the verbs and
        adjectives that may be used with "filler-it" ("be", "seem")
        are subscripted with "i" (yielding Ii, PPi, Pgi, Pai, Pvi,
        TOi); all others, like those on "want" and "glad", are left
        unsubscripted. (Note that "it" can only make a direct SF
        connection to certain verbs anyway; thus the use of non-
        referential "it" is partially controlled at the linkage
        level.)  PP rules then dictate that when an SFsi is present in
        a group, the only forms of verb connectors (I, T, Pg, Pa, and
        TO) which are allowed are those subscripted with an "i". The
        following linkage is thus rejected, because the group
        containing an SFsi also contains unsubscripted "Pa" and "TO"
        links.

                +SFsi+-Pa-+-TO+Ii+-Paf-+-THi-+--Ce+S(e)+I(e)+
                |    |    |   |  |     |     |    |    |    |
               *It  is  glad to  be likely that  Joe  will go

        Many sentences with "it" receive two parses, one with "S" and
        one with "SF". The above sentence, for example, also receives
        this linkage:

                +-Ss+-Pa-+-TO+Ii+-Paf-+-THi-+--Ce+S(e)+I(e)+
                |   |    |   |  |     |     |    |    |    |
               *It is  glad to  be likely that  Joe  will go

        This linkages is also rejected, however, because the "THi"
        requires an "SFsi". If a sentence contains "filler"-only links
        (in a group with "it"), the "S" parse will be invalid; if it
        contains referential-only links, the "SF" parse will be
        invalid. In this case, then, one linkage fails to meet the
        "SF" requirements, the other fails to meet the "S"
        requirements, and the sentence is rejected.

        Note that the division of the sentence into domains is
        essential here. Consider the sentence 

                +SFsi+-Paf+THi-+--Ce-+S(e)+Ii(e)+Pa(e)+TO(e)+I(e)+
                |    |    |    |     |    |     |     |     |    |
        	It  is likely that John will    be   glad   to   go

        If the sentence were not divided into domains, the
        post-processing rules would see that there are unsubscripted
        "Pa", "TO" and "I" connectors, which are incompatible with
        "SFsi" (as well as a THi connector, incompatible with the "S"
        usage of "it"); thus the sentence would be rejected. With
        domains, however, the parser knows that "THi" goes with "It"
        and "glad to go" goes with "John", and everything is okay. The
        same thing applies also, of course, with relative clauses,
        subordinate clauses, and the like: "the weather was terrible,
        but we thought it was likely that John, who was an excellent
        sailor, would get the boat safely back to the shore". Such
        constructions would obviously wreak havoc with post-processing
        rules unless the clauses were clearly demarcated.

        The same applies with sentences with "to":

                +SFsi+--Paf--+-TOi-+I(e)-+
                |    |       |     |     |
        	it   is important to    go

        In this case, the "TOi" on "important" must begin a new
        domain, so that the post-processor knows that "go" does not
        relate to "filler-it" (with which it is incompatible). The
        tricky thing here is that "adjective+to" constructions do not
        always start new domains: in sentences like "He is ready to
        go", the infinitive clearly relates to the subject preceding
        the adjective. For such adjectives, we use unsubscripted
        TO. (Note that there are a few adjective-infinitive
        constructions, such as "certain" and "likely", that do not
        start new domains with "to", but which are compatible with
        "filler-it": "It is certain to be important to go". Such
        adjectives take "TOf".)

        _"There" as a subject: SFst and SFp_	
        SFst and SFp are used when "there" is acting as a subject:

                  +-SFst-+
        	  |      |
        	There seems to be a problem.

                  +-SFp-+
                  |     |
        	There seem to be problems.

        "There" therefore has "SFst+ or SFp+"; verbs which can connect
        directly to "there" have "SFst-" or "SFp-", as appropriate.
        It may seem odd to distinguish between singular and plural
        "there", since the forms are the same; this relates to
        post-processing.

        There are constraints on the way "there" can be used as a
        subject, which are enforced mainly through post-processing.
        This is similar to the case of "filler-it", only simpler. Like
        "filler-it", "there" is compatible only with certain verbs and
        adjectives:

        	There seems to have been a problem
        	There might appear to be a problem
        	*There is eager to be a dog
        	*There wants to try to be a dog

        These are precisely the verbs and adjectives that are
        compatible with "filler-it".  These verbs have already been
        specially subscripted; post-processing rules enforce that
        groups which contain unsubscripted connectors of these types
        may not also contain a "SFsi" (used with "filler-it").  Thus
        we simply make such connectors incompatible with "SFst" and
        "SFp" as well. In this way, we prohibit incorrect sentences
        like those above. See "filler-it".  Note, however, that,
        unlike "it", "there" is always a "filler" subject; it
        therefore has no ordinary "S" connector.

        There are other constraints on "there". The "filler-only"
        phrases used with "it" (likely that, important to) may not be
        used with "There": "*There is likely that John is
        coming". Moreover, when "there" is acting as a subject, there
        must be an object in the clause ("*There seems to be likely"),
        and there must be number agreement between the object and the
        verb ("*There seem to be a dog here"). This is again enforced
        through post-processing. Forms of the verb "be" have O*t+
        connectors. "O-" connectors on singular nouns (used in
        ordinary direct-object links: see "O") are given "s"
        subscripts; those on plural nouns have "p" subscripts.  When a
        form of "be" connects to an Os- or Op- connector on a noun, an
        Opt or Ost link is thus created:

                  +-SFp-+            +-Ost-+
                  |     |	     |     |
              *	There seem to have been a dog here

                  +-SFst+             +-Ost-+
                  |     |	      |     |
        	There seems to have been a dog here

        PP rules then dictate that every SFst must have an Ost in its
        group and every SFp must have an Osp in its group. In the
        process, we also ensure that phrases used with "filler" it,
        like "likely that", do not occur with "there"; it turns out
        that there is no way for a group to contain a direct object
        and an predicative adjectival phrase like "likely that" at the
        same time.

        As discussed with "it", the use of post-processing domains is
        important here. Some nouns take "to" phrases as complements:
        "There was an effort to revive the bill".  A verb like
        "revive" in this case is fine, because it does not relate to
        the subject "there" ("*There seems to have revived the
        bill"). But in order for the post-processing rules to know
        this, the TOn link from the noun ("an EFFORT TO revive") must
        begin a new group, letting the parser know that a new subject
        is in force.

        _"Special Subjects": SFsx_
        SFsx is used for a few phrases that can act as subjects under rather
        constrained circumstances: "that"-clause phrases, "to"-infinitive
        phrases, and "where-when-why" phrases.
        
                  +---------SFsx---+
                  +Ce-+-S--+       |
        	  |   |    |       |
        	That Joe  is angry is not surprising

        	 +--------SFsx----+
                 +-I--+           |
        	 |    |           |
        	To invite Bill would be a mistake		
        
        	  +------SFsd---+
        	  +--Cs-+       |
                  |     |       |
        	Where they went is a mystery

        In each case, the subject phrase is a kind of phrase that
        occurs frequently as a verb complement. Thus we simply
        directly disjoin and conjoin "SFsx" on the head word with
        whatever connectors are used for verb complements. With
        "that", for example:

        	that: Ce+ & (TH- or SFsx+)

        As with gerunds, constructions of this kind seem to vary quite
        gradually in their grammaticality (See "Ss*g").  However, they
        are much rarer than gerunds, and seem much more constrained.

        	Finishing college would make you more marketable
        	?To finish college would make you more marketable
        	The graduating of Fred changes the situation
        	?That Fred graduated changes the situation
        	
        Most often, such special subject phrases are used with the
        verb "be" and a few other verbs and adjectives ("seem",
        "appear", "likely"). these are the same predicates that may be
        taken by "filler-it" and "there". We already have a system in
        place in post-processing for restricting the predicates used
        with "filler-it" and "there" (which use "SFsi" and "SFst",
        respectively). (See "SF: Filler-it".)  For now, then, we
        simply apply those same constraints to "SFsx". (One difference
        is that we allow forms of "be" to take direct objects under
        these circumstances: "That John graduated is a problem".)
        This solution could probably be improved, however.

        There are further constraints on the use of "special subjects". They
        may not occur in relative clauses (ex. 1), and they may not invert
        with their auxiliary (exx. 2-3).

        	The problem that John graduated is is very large
        	Is that John graduated a problem?
        	Is to graduate a good idea?		

        To solve the first problem, we give special subjects a restricted
        "clause" expression conjoined with their "SF+" connectors, omitting
        the normal "C-" connector. To solve the second problem, we simply 
        do not give special subjects "SI-" connectors.

SFI	connects "filler" subjects "it" and "there" to invertible
        verbs in questions with s-v inversion:

        	+SFI+
                |   |
        	Is it likely that John will go
        	Is there going to be a problem

        Both the use of "it" and "there" and the use of subject-verb
        inversion are highly constrained by post-processing; see "SF"
        and "SI". However, there is nothing problematic about SFI; SFI
        is to SF precisely as SI is to S.

SI	is used in subject-verb inversion:
          
                 +----Pg---+          +-----I------+
                 +-SI-+    |          +-SI-+       |
                 |    |    |          |    |       |
        	 Is John coming      Who did John see

        Only verbs which may be inverted (i.e. modals and auxiliaries)
        have SI+ connectors. On such verbs, the SI is disjoined with
        the S-, and is conjoined with whatever complement connectors the
        verb may be used with when it is inverted. For example, forms of
        "have" may be inverted when they are taking a past participle, but
        not when they are taking a direct object. Thus they have

        have: (SI+ & PP-) or (S- & (O+ or TO+ or PP+));

        This yields:

        	They have finished it
        	Have they finished it
        	They have dogs
        	*Have they dogs
        	They have to go
        	*Have they to go
        	
        The use of SI and S connectors is highly constrained. In many
        situations, subject-verb inversion may not occur; in some
        situations, it must occur. In many cases, the enforcement of
        this involves post-processing.

        _Questions requiring subject-verb inversion_ In some cases,
        s-v inversion _must_ occur: in object-type questions ("*Which
        dog you hit", "Which dog did you hit"), or with question words
        like "where" ("*Where you will go", "Where will you go"). This
        is enforced by post-processing.  When a question word begins a
        sentence, it must make a Wq connection to the wall (or a Ws
        connection: see below). (There is no other way for the wall to
        connect to the sentence.)  The Wq connection starts a 'm'
        domain, and is included in the domain.  We then require that a
        group with a "Wq" contain contain some kind of SI connector.
        This prevents "*Which dog I hit".

        This solution works equally well when the question contains an
        embedded clause: "What do you think he did?", "*What you
        think he did?" In both cases, "What" begins a group;
        in both cases, the link between "you" and its finite verb is
        in the outer group of this domain. Thus this group is required
        to contain an "SI" connector of some kind.

                  +----------Bsw(m(e))--------------+
                  |    +----I(m)--+                 |
           +-Wq(m)+    +SI(m)+    +Ce(m(e))+S(m(e))-+
           |      |    |     |    |        |        |
           ||   What  do    you think      he      did
        
        A similar situation arises with adjectival questions:

                     +-----AF----+
                     |   +---I---+
             +-Wq+EAh+   +SI-+   |
             |   |   |   |   |   |            
            || 	How big will it be

        Here again, s-v inversion must be enforced ("*How big it
        is"). We therefore give "how" a Wq- connector; this is then in
        the same group as whatever S or SI connector is in the outer
        group, and p.p. insists that it must be an SI.

        S-V inversion must also be enforced with the question words
        "where" and "when" and "how" (when used in this way):
        "Where(/when/how) will you go", "*Where you will go".  This is
        done with simple connector logic. Unlike with object-type
        constructions, the question word here is not making a B
        connector to the rest of the sentence; it must find some other
        way to connect. For this purpose, the question words
        "where", "when", "why" and "who" have Q+ conjoined with Wq-;
        and Q- on verbs is conjoined with SI+, disjoined with S-.

        	where: (Wq- & Q+)...
        	have: ({Q-} & SI+ & ...) or ({C-} & S+ & ...)

        Thus we allow "Where have you gone"; we prevent "*Where you
        have gone".

        _Cases where s-v inversion may not occur_
        In many cases, s-v inversion is prohibited: in relative
        clauses (*The dog who did you buy was black) and subordinate
        clauses ("I left the party after did you see Fred"). No
        linkages are found for these; again, there is no way for the
        illegally-inverted segment to connect to the rest of the
        sentence. (The one case where s-v inversion may occur unwanted
        is in indirect questions; this problem is discussed below.)

        _Simple s-v inversion ("yes-no") questions_        
        Questions may also be formed by simply inverting the verb and
        the auxiliary ("Are you coming", "Did you go"). In this case
        the question must make some kind of connection to the wall; we
        use the Q connector for this, giving the wall Q+.  The problem
        here is that no question word is used; thus there is no Wq
        present, indicating that s-v inversion may occur.  Thus the Q+
        on the wall is subscripted "Qd", and this is added to the list
        of connectors that permit s-v inversion. (See "Q".)
        
        _Questions without s-v inversion_
        Post-processing cannot insist on subject-verb inversion in all
        question-word questions. Subject-type questions do not contain
        s-v inversion:
                 
                 +-S--+
                 |    |
        	Who  hit John
        
        To allow this, question words like "who" and "which" may also
        make a Ws connection to the wall. Ws is exempted from the
        p.p. constraints applied to Wq; a group containing a Ws need
        not contain an SI connector (indeed, they may not; see below).

        	who: (S+ or B+) or (Ws- or Wq-);
        	which: (D**w+ or S+ or B+) or (Ws- or Wq-);

        (The special subscript on the D+ of "which" will be explained
        below.)  But in that case what will prevent illegal sentences
        being formed using the Ws: "*Who John hit", and "*Where John
        goes"? As described above, the second case is prevented by
        link logic; the sentence cannot form. For the first case, we
        require p.p.: we dictate that a group with a Ws may not
        contain an B#m. (We must also prevent the redundant parses
        resulting from Ws being used in "where/when/why" questions
        with s-v inversion. We do this by simply not giving these
        words Ws.)  In short: if there is a question-word present in
        the main clause, there must either be a Wq, in which case s-v
        inversion is enforced, or there must be a Ws, in which case
        B#m connections are prohibited in p.p. and "where-when-why"
        questions do not form; the result is that with Ws all s-v
        inversion questions are prevented either at the linkage stage
        or in post-processing; Ws is therefore used with all and only
        non-s-v questions, Wq with all and only s-v questions.

        There is one final problem: enforcing s-v inversion in
        object-type questions with embedded clauses.

                             +----------Bsm(m(e))----+
           +-Ws(m)+D**w(m(e))+   +S(m)+Ce(m)+S(m(e))+
           |      |          |   |    |      |       |
         * ||   Which      dog  you think   you     hit

        Here, "which" can use its Ws connection.  The parser then
        "thinks" it is a subject-type question; it finds no s-v
        inversion, and no B connector in the outer group, therefore it
        accepts the sentence. To prevent this, we assign question-word
        determiners ("which" and "what" a D**w+ subscript; we then
        stipulate that a group with a Ws must contain a "D**w"
        connector. We further make the B#m link _not_ a restricted
        link; the D**w is restricted, however. (See "B: Questions: B#w,
        B#m".)  In this case, then, the 'e' group spreads back to
        include the "D**w" connector; the group containing the Ws
        therefore no longer contains a D**w; and the sentence is
        rejected. (In effect, we want post-processing to "know" when
        something is a well-formed subject-type question. It only
        knows this is the case if a) there is an S in the outer group
        of the sentence, and b) the D connector of that subject is
        also in that group.)

        (One more little annoyance: We've said that groups with Ws
        correspond to subject-type questions. We've insisted that they
        may not contain SI links; and, to avoid the false positive
        just discussed, we've insisted that they must contain a
        D**w. Now the only problem is, they sometimes don't contain a
        D**w; sometimes the question word is itself the subject: "Who
        is coming?".  So we give question-words specially subscripted
        "S**w" connectors; and then we say, a group with a Ws must
        contain _either_ a D**w or an S**w.)

        _Indirect Questions_
        It was mentioned that in most cases where s-v inversion is
        prohibited (relative clauses and the like), it is prevented at
        the linkage level. There is one case where it is not, however,
        namely indirect questions.

                        +----B(s)-----+
                        |   +----I(s)-+
                    +-R-+   +SI(s)+   |
                    |   |   |     |   |
              * I know who did   you hit

        We prevent this by saying that an SI connector may only occur
        when Wq or Qd is present.  Wq connections can only form
        between a question-word and the wall; Qd can only form between
        an auxiliary verb and the wall.  Thus the above construction
        will be rejected.  (With "where/when/how" questions, this is
        accomplished at the linkage level. QI- on these words is
        disjoined with Q+; thus there is no way for the question word
        to connect to a s-v-inverted clause.)  The same applies to
        "how [adjective]" questions: "*I wonder how big is it" is
        therefore rejected.

        _Other Uses of SI Besides Questions: SI*j_
        There are a few other situations besides questions where SI is
        used.  In subjunctive clauses, a nominative noun phrase is
        used (e.g., "he" rather than "him"). In such cases, however,
        is most convenient to connect it to the preceding "that",
        rather than the following verb. Therefore we use SI here.
        
                             +----I--+
                      +--TS--+-SI-+  |
                      |      |    |  |
        	I suggested that  he go

        We also use SI in quotation constructions with certain verbs like
        "say":

        		+------CCq------+----SI----+
        	        |               |          |
        	The President is busy, said the spokesman

        As described above, the use of SI is usually tightly
        constrained by post-processing. Rather than try to adjust the
        post-processing constraints to handle these situations, we
        avoid the problem. We simply give the SI+ a "*j" subscript
        here; thus the post-processing constraints do not apply, and
        SI*j can be used anywhere that a linkage can be found. Verbs
        (and the word "that") that take subjunctive clauses thus carry
        "SI*j+".

TA 	is used to connect adjectives like "late" to month names: 
        
                          +----IN----+
                          |   +--TA--+
                          |   |      |
        	We did it in early December

        See "DT" for more discussion of time expressions.

TD	connects day-of-the-week words to time expression like "morning",
        "afternoon", and "evening". The day-of-the-week word serves as
        the head of the expression, and makes an MV, CO or other
        connection to the rest of the sentence. See "DT" for more
        discussion of time expressions.

TH	connects words that take "that [clause]" complements with the
        word "that". These include verbs ("I assured him that"),
        nouns ("The idea that...") and adjectives ("We are certain
        that...").

                     +---TH----+
                     |         |
        	I assured him that I would finish the project
        	The idea      that we would hire John is preposterous
        	I'm certain   that he could do a good job

        The word "that" therefore has "TH- & Ce+". The Ce+ connects to
        the subject of a following clause (see "C").

        Verbs that can take "that"+clause have TH+ disjoined with
        their other complement connectors (O+, TO+, etc.). With many
        such verbs, the "that" is optional; they thus carry "TH+ or
        Ce+", allowing a direct connection to the subject of the
        subordinate clause.  A few verbs can take object+"that" ("I
        told him that I was angry"); such verbs have "O+ & {TH+}."
        Adjectives like "certain" have "TH+" disjoined with other
        complement connectors. Nouns like "idea", "opinion", and
        "argument" have TH+ conjoined with {@M+} (used in
        prepositional phrases) but disjoined with {C+ & B+} (used in
        relative clauses); this is somewhat arbitrary.

        THb connectors are used to connect forms of "be" to "that".
        The reason for this is that only certain nouns can serve as
        the subject in such cases: "The problem is that John is
        coming", "*The dog is that John is coming". To enforce this,
        we give such nouns "Ss*t+" connectors Instead of the usual
        "Ss+".  We then insist in post-processing that a group
        containing a THb must contain an Ss*t.

        THi connectors are used with certain adjectives ("important")
        which may only take "that+clause" when "filler-it" is the
        subject.  This distinction involves post-processing.  See
        "SF: Filler-it" for further explanation.

TI	is used for titles like "president" and "chairman", which
        can be used in certain circumstances without a determiner:
        after the preposition "as" or after verbs liked "name" or "elect".


                +--------------CO-------------+ 
                +--TI--+-Mp---+               |
                |      |      |               |
        	As president of the company, it is my decision

                         +----TI---+ 
        		 |         |             
        	He was named president of the company

        There are special dictionary entries for the no-determiner
        usage of these nouns:

        president.i chairman.i: {@M+} & (BI- or (Xd- & Xc+ & MX-) or TI-);

        As this expression shows, such nouns can also be complements
        of "be" ("He is president of the company"), using BI-; and
        they can also form comma expressions modifying a noun ("John
        Smith, president of the company, ..."), using MX-.

TM	is used to connect month names to day numbers:

                             +-ON-+-TM-+
                             |    |    |
        	It happened on January 21

        See "DT" for more discussion of time expressions.
        
TO	connects verbs and adjectives which take infinitival complements
        to the word "to".
 
                    +---TO--+
        	    |       |
        	I tried     to start the car
        	We intend   to be firm
            We are eager    to do it

        The word "to" then makes an I link to an infinitive verb
        form. "To" therefore carries "TO- & I+".

        Verbs which take "to"+infinitive have "TO+" connectors.  Many
        such verbs can also take other kinds of complements; simple
        objects (O+), clauses with or without "that" (TH+/ Ce+),
        indirect questions (QI+), and so on. Such connectors are
        disjoined with "TO+". In some cases, the verb may take no
        complement at all "We hesitated"; in others, some kind of
        complement is obligatory (*"We intend"). Some verbs take a
        direct object plus an infinitive ("We told him to go"); these
        verbs do _not_ use TO+, but rather TOo+. See "TOo".

        Some adjectives also take "TO+". This is only the case 
        for usages where the same subject is implied before and
        after the adjective: "We were ready to go", but not
        "It is important to go". See "TOi".

        There are other situations involving "to"+infinitive where
        specially subscripted TO connectors are used: transitive
        adjectives ("He is easy to hit"; see "TOt"), indirect
        questions ("I wonder where to go"; see "TOn") and nouns that
        take to+infinitive ("We made an effort to go"; see "TOn").
        The word "to" (and only this word) carries unsubscripted TO-,
        it is directly conjoined with I+. The reason for the
        distinctions between TO, TOt, TOn TOi, and TOo relates to
        post-processing.  Recall that post-processing divides the
        links of a sentence into groups, corresponding roughly to
        subject-verb expressions. In some uses of infinitives -
        specifically, those described above which use unsubscripted TO
        - the infinitive simply continues the subject-verb expression
        that precedes it. Indeed, one can use a number of infinitives,
        all relating to the same subject: "I hope to be ready to
        consent to try to do it." In other cases (those specially
        subscripted TO connectors), the infinitive implies a different
        subject from what precedes ("I told him to go", "He is easy to
        hit"). In these cases the link connecting "to" with the
        infinitive must start a new domain. This is particularly
        important in sentences like the following, where
        post-processing is used to enforce the use of certain verbs.

        	There is certain to be a problem
        	*There is eager to be a problem
        	There might be an opportunity to refuse to do it
        	*There might refuse to be an opportunity to do it

        In addition, certain uses of infinitives are only permitted
        with "filler-it" as the subject: "*John is important to
        go". See "SF" for explanation of all this.

        It is important to note that we allow any kind of
        sentence to take an infinitival phrase, meaning "in order
        to": "We bought some eggs to make cookies". See "MVi".
        We give this connector a cost of 1. Thus many sentences
        which involve incorrect uses of "filler"-only phrases
        ("John is important to do it") will receive valid linkages
        using "MVi". 

        _Infinitival complements of transitive verbs: TOo_
        Some verbs can take an infinitival complement as well as a
        direct object. In such cases, TOo is used.

                     +---TOo--+
                     +--O--+  |
                     |     |  |
        	I advised him to go

        It will be seen that every "TOo" on verbs is conjoined with a
        preceding "O+". (Usually the TOo+ is optional: "I advised
        him".)  Note that in such situations, the infinitival verb
        relates not to the main subject ("I" in this case), but to the
        direct object of the verb ("him"). This is unlike other
        infinitival complements of verbs ("I hesitated/wanted/tried to
        go") where the preceding subject remains in force; for these
        situations, we use "TO".

        Like TO, TOo connects a verb to the word "to'. The reason for
        distinguishing between them relates entirely to
        post-processing.  TOo begins a new domain, thus telling
        post-processing that a the infinitive verb relates not to the
        preceding subject but to a new subject. In addition, TOo links
        start a special kind of domain, "urfl domains". Ordinary
        domains contain everything that can be reached from the right
        end of the domain-starting link. "Urfl domains", however, also
        include whatever can be reached from the left end of the
        domain-starting link and b) is underneath that link.  (Hence
        the name: "urfl" stands for "under root from left".) In the
        case, below, then, an ordinary domain started by the "C" link
        would contain only the "D" link. An "urfl" domain started by
        the C link would contain the C, the D, and the B.

                    +---C---+
                +-A-+-B-+   +-D-+
                |   |   |   |   |
        	bla bla bla bla bla		

        This is useful in the case of verbs which take objects and
        infinitives. Recall that the logic of domains is that links
        relating to a single subject-verb expression should be
        contained in a group. In this case, the infinitive relates to
        the object of the preceding verb; thus we want the O link to
        be in the same group as the infinitive. TOo starts "x"
        domains, which are "urfl":
        
                     +--TOo(x)+
                +-S--+-O(x)+  +I(x)+
                |    |     |  |    |
        	I   told  him to  go

        We use this to control uses of "filler-it" and "there".  "It"
        and "there" have OX connectors, which are used only in cases
        like this. The use of verbs and adjectives with "filler-it"
        and "there" is highly constrained; we have a complex apparatus
        for enforcing these rules in post-processing (see "SF"). The
        same rules apply when "it" and "there" are used with
        Obj+infinitive constructions.

                      +-TOo(x)-+
                      +OX(x)+  +Ii(x)+Paf(x)+
                      |     |  |     |      |
        	I expected it  to   be    easy to  use the program
              *I expected John to   be    easy to use the program

        Once we have the adjective/verb connectors in the same group
        as their implied subject, as in this case, we can simply apply
        the same constraints to "OXt" (for "there") that we apply for
        "SFst", and the same constraints to "OXi" that we apply to
        "SFsi", and everything else follows naturally.

        _Other Kinds of TO Connectors: TOn, TOi, TOt_
        TOn is used with nouns that take infinitival complements: "The
        EFFORT TO finish the program was successful".  (Only certain
        nouns can take such complements: "*The computer to finish the
        program was fast".) With such nouns, the TOn is conjoined with
        the @M+ (used in prepositional phrases and some other kinds of
        modifiers), disjoined with the "(C+ & B+)"; this is perhaps
        rather arbitrary.

        TOn is also used in indirect questions, to connect question
        words to "to": "I wonder WHERE TO go". (In object-type
        indirect questions, like "I wonder what to do next", no TOn
        connection is made; see "Ia".) Question words such as "where"
        therefore have "R- & (TOn+ or Cs+...)".

        TOi is used with adjectives that take infinitival complements
        but which take "filler-it" as the main subject.

                       +-TOi-+
                       |     |
        	It is fun   to try to beat the program

        In such cases, there are constraints on the verbs that may be
        taken by "it" ("*It tries to be fun to beat the program"), but
        not on the verbs in the infinitival phrase.  This is enforced
        with post-processing; see "SF: Filler-it".

        TOn and TOi both start new domains. They are therefore like
        TOo, but unlike TO, which is used with adjectives and verbs
        that take infinitive complements but which does not start a
        new domain. The reason is that the infinitive and what follows
        usually imply a new subject, rather than relating to the
        subject that precedes. This is important, for uses of
        "filler-it" and "there".  See "SF".

        TOt is used for certain adjectives which take transitive
        infinitival constructions.

                         +----B-----+
                         +-TOt-+-I--+
                         |     |    |
        	John is easy   to  hit

        Such adjectives take "TOt+ & B+", disjoined with their other
        complement connectors (TO+, THi+, etc.). 

        TOf connectors are used to enforce the correct use of non-
        referential "it". "TOf" connectors are used with adjectives
        which do not start new domains, but which may also be used
        with "filler-it" and "there".  See "SF: Filler-it".

TQ	is used as the determiner connector for time expressions
        acting as fronted objects, as in "how many years did it
        last". See "OT".

TS	is used in subjunctive constructions. It connects certain
        verbs that can take subjunctive clauses as complements -
        "suggest", "require" - to the word "that".  Such verbs have
        TS+, disjoined with other complement connectors (TH+, TO+,
        etc.). The word "that" has "TS- & SI*j+ & I*j+". Thus "that"
        connects to a subject (all nouns and nominative pronouns have
        SI- connectors) and to an infinitive verb; the subject and
        verb do not connect to each other.
        
                             +----I--+
                      +--TS--+-SI-+  |
                      |      |    |  |
        	I suggested that  he go

        SI connectors are mainly used in questions where there is
        question inversion ("Did he go", "Where did he go", "Who
        did he see"). The use of SI is highly constrained in
        post-processing, and is usually only permitted in questions;
        to avoid this, we give the SI connectors here a special
        subscript. See "SI*j".

        TS is also used for certain adjectives that take subjunctive:
        "It is IMPORTANT THAT he go". TSi is used here, since only
        the the "filler-it" may be used as a subject. 

        Notice that the "I+" connector on "that" is also subscripted:
        "I*j+". This connector starts an "urfl domain" (see "TOo".)
        The domain structure below is thus formed:

                             +-I*j(x)----+
                      +--TS--+SIs*j(x)+  |
                      |      |        |  |
        	I suggested that     he go

        We thereby capture the intuition that the SI link and the
        I link in a subjunctive clause form part of a single
        subject-verb expression, distinct from the previous subject-
        verb expression.

TY	is used for certain idiomatic usages of year numbers.

        			      +--TY-+
                           +--ON-+-TD-+ +-Xd+-Xc+
                           |     |    | |   |   |
        	I saw him on January 21 , 1990 \\\\\

        See "DT" for more discussion of time expressions.

U	is a special connector used with nouns. In most cases,
        nouns have both a determiner requirement ("D-") and a main
        subject-object requirement ("S+ or O- or J-").  In a few
        cases, however, a single word appears to satisfy both
        requirements:

                       +--U---+
                       |      |
        	What kind_of dog did you buy	

        "Dog" would seem to be acting like the prepositional object of "of"
        here; but, unusually, it requires no determiner. Therefore
        we create a "U-" connector, disjoined with both its "D-"
        and its "S+ or O- or J-..." complex.

        Other uses include the following:

        		               +--U--+
                                       |     |
        	We spend four dollars per student
        	We spend four dollars  a  student

        For the latter case, we give the word "a" the expression
        "[[Mp- or Us+]]". Since this usage is rare compared to
        the other uses of "a", we make it stage 2.

        U*t is used in comparatives; see "MV: Comparatives".

UN	connects the words "until" and "since" to certain time phrases
        like "after (clause)" and "before (clause)". "Until" and
        "since" are the only words that can take such phrases as
        complements ("*We waited before/by/in after the movie"); hence
        this special connector.

V	is used for attaching various verbs to idiomatic expressions
        that may be non-adjacent. Each verb has its own subscript.

                   +---V----+
                   |        |
        	I took him for granted  (Vt)
        	I held him responsible  (Vh)
        	He did     nothing but complain  (Vd)

W	is used to attach main clauses to the wall. Almost all kinds
        of main clauses - declaratives, most questions (object-type,
        subject-type, where/when/why, and prepositional), and
        imperatives - use a W of some kind to attach to the wall. The
        only exception is "yes-no" questions, which attach to the wall
        with Q. See "Q".

          +---W------+
          |          |
        /////	The dog ran (Wd)
        /////	    Who did you hit (Wq)
        /////	    Who is coming (Ws)
        /////	     To whom did you speak (Wj)
        /////	     Go away (Wi)

        Note that the wall is automatically inserted at the beginning
        of every sentence, and is then treated like a normal word; by
        the connectivity rule, therefore, it _must_ make some kind of
        connection to the sentence. The wall thus has "W- or Q-". 

        W is also used to attach clauses back to coordinating
        conjunctions in declarative sentences; coordinating
        conjunctions thus have "CC- & (Wd+ or Wq+ or Ws+)". CC is used
        to make a link back to the subject of the previous main
        clause.

        _Declarative Sentences: Wd_
        Wd is used in ordinary declarative sentences, to connect the
        main clause back to the wall (or to a previous coordinating
        conjunction). Nouns carry Wd-, optionally conjoined with their
        S+ connectors. Wd- on nouns is directly disjoined with C-
        (used in dependent clauses) and R- (used in some relative
        clauses); see "C-".

        	dog: (({@CO-} & Wd-) or ({@CO-} & C-) or R+) & S+

        _Questions: Wq, Ws, Wj_ 
        Wq, Ws and Wj are used to connect many types of questions to
        the wall: subject questions (Ws), object questions (Wq),
        where/when/why questions (Wq), adjectival questions (Wq), and
        prepositional questions (Wj).  Each of these connector types
        interacts heavily with post-processing.  See "SI" for an
        explanation of Wq and Ws; see "JQ" for an explanation of Wj.

        _Imperatives: Wi_
        Wi is used to connect imperatives to the wall.
 
             +-W-+
             |   |
          /////  Go away

        Imperative verb forms have "Wi-", conjoined with their
        complement connectors. Since the imperative verb form is
        always the same as the infinitive form (and the plural,
        in every case except "be"), the same expression can be
        used.

        _Coordinating Conjunctions_
        There are a number of words that serve to link clauses together:
        coordinating conjunctions like "and" and "but", and subordinating
        conjunctions like "after" and "because".

                  +--CC----+-Wd+
                  |        |   |
        	John left but he returned later

                       +-MVs+-Cs+
                       |    |   |
        	John left after I saw you

        Note that subordinating and coordinating conjunctions use very
        different linking structures. First of all, both the left-pointing
        and right-pointing connectors on the conjunctions are different;
        "but" has "CC- & Wd+", "after" has "MVs- & Cs+". Secondly, 
        coordinating conjunctions connect back to the subject of the
        previous clause, subordinating conjunctions to the verb.
        There are several reasons for making these distinctions.
        First of all, coordinating conjunctions may not be used in
        relative clauses:

        	*The man I tried to hit but John stopped me is here
        	*The man I tried to stop John but he hit is here
        	*The man I hit but John comforted is here

        (There are other constraints on relative clauses: the main
        noun of a relative clause may not link to something inside an
        embedded clause.  We handle this using Ce and Cs; see "Ce".)
        So, we need to prevent these constructions. Coordinating
        conjunctions have another related property. They may be used
        to connect clauses in sequence, like subordinating
        conjunctions. But whereas subordinating conjunctions seem to
        link in a nested way, with each modifying the last,
        coordinating conjunctions seem to "leap" over any preceding
        subordinating conjunctions:

                  +------------+-C-+-S-+------+-C---+--S--+
                  |            |   |   |      |     |     |
             1.	John screamed when I arrived after Sue   left (seems right)

                                   +---- ? ---+
                  +------------+-C-+-S-+      +-W---+--S--+
                  |            |   |   |      |     |     |
             2.	John screamed when I arrived but   Sue   left (seems wrong)

                  +-------------CC------------+
                  +------------+-C-+-S-+      +-C---+--S--+
                  |            |   |   |      |     |     |
             3.	John screamed when I arrived but   Sue   left (seems right)

        We handle this in the following way. In the first place,
        coordinating conjunctions link to the left not with MVs-, like
        other conjunctions, but with CC-.

        	and but: CC- & W+;
        	dog: {R- or C- or (W- & {CC+})} & S+...;

        Note that subject nouns may make a CC connection to the right,
        but only if a W is being made to the left (i.e., if the noun
        in a subject of a main clause), not if a C is being made. In
        other words, while subordinating conjunctions connect to the
        main verb of the nearest clause to the left, coordinating
        conjunctions connect to the _subject_ of the nearest _main_
        clause to the left.  Thus ex. 3 above is allowed, but ex. 2 is
        prevented. The problem with relative clauses is solved also.
        In relative clauses, the main subject of the relative always
        makes either a C- or an R- to the left, and neither one is
        conjoined with CC+; so no coordinating conjunctions can
        appear.

        Note that the above expressions also allow coordinating
        conjunctions to link clauses in sequence:

                  +-------CC---+--W-+---CC--+-W-+
                  |            |    |       |   |
        	John screamed and Fred ran but Dave cried

        Coordinating conjunctions may also connect directly to the
        wall: "And John screamed". Thus they carry a "Wc"
        connector. Furthermore, a coordinating conjunction may link to
        a following question, rather than to a declarative
        clause. They may not, however, link from a question to a
        declarative clause:

       	I know you don't like Joe, but why did you send him that nasty note
       	*Why did you send Joe that nasty note, but I know you don't like him

        Thus we give such conjunctions the following:

        	(CC- or Wc-) & (Wd+ or Wq+ or Ws+ or Qd+);

        Another reason for distinguishing between W and C is that
        certain openers like participle openers may be used in main
        clauses but not dependent ones; see "COp".

WN	is used to connect "when" phrases to time-nouns like "year"
        and "day". In such cases, WN is acting like a noun-modifier;
        such modifiers would normally use M-. However, "when" cannot
        normally be used in this way, as the second sentence shows;
        only with time nouns. Thus time-nouns have an option WN+
        (along with an optional @M+); "when" has WN- conjoined with
        Cs+, connecting to a subordinate clause.

                     +--WN-+
                     |     |
        	The year when I lived in England was wonderful

              *The school when I lived in England was wonderful

WR	is used to connect the word "where" to a few verbs like "put".
        "Put" usually requires an object plus a prepositional phrase
        as a complement; but this prepositional phrase requirement can
        be satisfied by the preceding word "where" in questions.

        	*I put it

                  +----WR------+-O-+
        	  |            |   |
        	Where did you put it

X	is used to connect punctuation symbols to words. Xc connects
        a word to a comma to the right; Xd connects a word to a comma
        to a left. Xx is used with colons and semi-colons.

        _Comma Phrases_
        A wide variety of words and phrases can be used with commas on
        either side.  Some are noun-modifiers, using MX to connect to
        a previous noun. Some are verb modifiers, using MV to connect
        to a previous verb or E to connect to a following one. Some
        are "openers", using CO to connect to a following clause
        subject. And some are quoting expressions ("he said"), which
        can be inserted at various places in the sentence. In each of
        these cases, the head word of the expression must link to
        commas on either side.

                  +-MX---+   
                  |  +Xd-+-----Xc-------+
                  |  |   |              |
        1.	John , a doctor		, is here	 
        2.	John , with his mother  , is here
        3.	John , carrying the dog , left the room
        4.	John , who you know     , is here
        5.	John ,  to whom I spoke , is here

                  +--MV--+
                  |  +Xd-+-----Xc-------+
                  |  |   |              |		
        6.   We left , carrying the dog , and Fred followed
        7.   We left , quietly	        , and Fred followed     
        8.   We left , with John	, and Fred followed
        9.   We left , when we saw him  , and Fred followed

    		             +-------CO-------+
                         +Xd-+-----Xc-------+ |
                         |   |              | | 	     
        10. He said that , after the party  , he had gone home
        11. He said that , eventually	    , he had gone home
        12. He said that , after he saw us  , he had gone home
        
                            +---Xd-+--Xc-------+
                            |      |           |
            	    He left , he said	       , and Joe followed   (CCq-)
        	       John , he said	       , left the party     (Eq+)
        14. After the party , he said	       , he left	     (COq+)

        In each case, the head word must connect most closely to
        the words within the comma expression; then to the commas
        themselves; then to any words outside. So, for example, in
        ex. 2 above, prepositions have

        with:	J+ & (.... or (Xc+ & Xd- & MX-))

        	^  	       ^      ^     ^
         link to prep.    links to commas  link to previous 
          object			   noun

        In some cases, the commas are obligatory (such as noun
        modifiers of nouns, like ex. 1: "*John a doctor is here"). In
        other cases the commas are essentially optional, but for
        various reasons, we decided to use a different connectors for
        with-comma and without-comma expressions (for example,
        prepositional phrases modifying nouns take M without commas,
        MX with commas). In other cases, the commas are optional and
        the same connector is used with or without; this is the case,
        for example, with manner adverbs modifying verbs. Such adverbs
        therefore take "{Xc+ & Xd-} & MVa-".

        One special case must be discussed. On opener phrases, the
        following comma is optional; the preceding comma is also
        optional, and is only permitted when there is a following
        comma. 

        	He said that after the party he went home
        	He said that after the party, he went home
        	He said that, after the party, he went home
        	*He said that, after the party he went home

        Openers therefore take "{{Xd-} & Xc+} & CO+".

        _Comma Phrases at the Beginning and End of Sentences_
        It was said that, with the exception of openers, expressions
        generally require either two commas or none; it is not
        permitted to have only a comma at the beginning or only one at
        the end.  One very important exception to all this must be
        mentioned, namely, comma expressions at the beginnings and
        endings of sentences. The case of phrases at the beginning has
        been dealt with, since these are always openers; there the
        beginning comma may be omitted.  With phrases at the end,
        however, the linkage expressions proposed above will require a
        comma at the end of the sentence, and this is clearly wrong. A
        comma is not only not required; it is incorrect. (The same
        applies to openers; the expressions provided will allow a
        comma at the very beginning of the sentence.) As described
        above, then, the system will accept the incorrect sentences 1
        and 2 below, and reject the correct sentence 3.

        		 +-Xd+-----Xc-------+
        		 |   |              |
        1.	*He left , carrying the dog ,
        2.	*        , After the party  , he left
        3.	He left  , carrying the dog

        Regarding the false positives 1 and 2, we simply allow these.
        Rules of the form "word X is not permitted at the beginning
        or end of a sentence" do not seem to arise in syntax generally,
        only in punctuation; therefore we have not considered it
        important to accommodate them. (Clearly, a system for weeding
        out sentences beginning or ending with commas could easily be
        devised.) What is important is to allow sentence 3 above. This
        we do by installing a "right-hand wall". The right-hand wall has
        an "Xc-" connector; this may satisfy the demand of a comma-phrase
        head-word for a following comma, when the phrase occurs at the
        end of the sentence. Ex. 3 above will therefore be accepted:

        		 +-Xd+-----Xc-------+
        		 |   |              |
        	 He left , carrying the dog \\\\\	

        (This punctuation usage is the only function of the right-hand
        wall. In cases where the punctuation function is not needed,
        we give the right-hand wall an "RW-", and the left-hand wall
        an "RW+"; thus the two walls connect to each other, preserving
        connectivity.)

        In many cases, comma phrases can be used in sequence. These
        can either be cases of one comma phrase nested inside another
        (ex. 1 below), or one comma phrase following another, both
        modifying the same previous word (ex.2) or the same
        following word (ex.3); or one comma phrase
        modifying a previous word, and another modifying a following
        word (ex.4).

        	        +--------------------Xc----+
       	      +------MV-+         +---MX-----+     |
       	      |   +-Xd--+         |  +---Xd--+-Xc+ |
              |   |     |         |  |       |   | |
       	1.He left , carrying the dog ,  a poodle , , and Fred followed

        	  +---------MX--------+
        	  +---MX---+	      |
                  |  +-Xd--+-Xc-+  +Xd+---Xc-----+
                  |  |     |    |  |  |          |
        2.	John , a doctor ,  , who you met , is here

        3.	After he saw John, Fred said, he left the party
        4. 	Although I liked my doctor, Mr. Smith, later, 
                        I chose a different one

        The structures shown for exx. 1 and 2 are those that would be
        required given the expressions above. However, this is
        obviously wrong.  Each comma requires a phrase before and
        afterwards; but multiple commas in a row are not required,
        indeed (again) they are not permitted. What seems to happen is
        that a single comma fulfills the demands of more than one
        phrase. In ex. 1, the final comma fulfills the Xc demand of
        both phrases; in ex. 2, the comma at the end fulfills the Xc
        demand of the previous phrase and the Xd demand of the
        following one. We handle this by giving the comma the
        following expression:

        	COMMA: {@Xc-} & (Xc- or Xd+)

        Now, as well as making an obligatory Xd or Xd connection, it
        can make any number of additional Xc connections as well.
        For exs. 1 and 2 above, then, this yields:
        
        		      +----------------Xc-------+
        	    +------MV-+         +---MX---+      |
        	    |   +-Xd--+         |  +-Xd--+--Xc--+
                    |   |     |         |  |     |      |
        	He left , carrying the dog , a poodle   , and Fred followed

        	  +---------MX-----+
        	  +---MX---+	   |
                  |  +-Xd--+-Xc-+Xd+---Xc-----+
                  |  |     |    |  |          | 
        	John , a doctor , who you met , is here

        There are three problems here. First, when a comma is serving
        the demands for more than one phrase, it is creating a cycle;
        it is linking two phrases together that are indirectly linked
        in some other way. Cycles are dangerous; they may allow
        non-cyclic expressions to form where they are not wanted
        (for example, "John, a doctor I saw Fred, a doctor, is here").
        To prevent these, we modify the comma expression slightly:

        	COMMA: ({@Xca-} & (Xd+ or Xc-))

        In post-processing, we then add "Xca" to the "must be connected
        without" list. This means, in effect, that any time an Xca
        connection is made (i.e., any time a single comma is being	
        used in more than one phrase), the sentence must be connected
        even without that Xca; this is another way of saying that Xca
        may only be used in cycles.

        It is clear that the function of fulfilling several Xc+
        demands may also be performed by the right-hand wall: "We
        left, carrying the dog, a poodle". Thus the right-hand wall is
        given not just Xc-, but "{@Xca-} & Xc-".

        A second problem involves post-processing. The groups of
        links formed in post-processing are supposed to correspond to
        clauses. In some cases it seems appropriate for comma
        phrases to form groups; for example, in relative clauses
        ("John , who you met , is here"). Again, however, a problem
        arises with multiple comma phrases. Consider the following
        sentence. Here, if left unchecked, the domain started by the
        first MX connector will spread via the second comma to the
        second comma phrase ("a doctor"), which is clearly not part
        of the relative clause, but rather a direct modifier of the
        main noun. What's worse, it will then spread back through
        the second MX connector and then through the rest of the
        sentence.

        	  +-----------------S(r)------------+
        	  +-----------MX(r)--------+	    |
        	  |     +---Xca(r)--+      |	    |
        	  +MX(r)+--B(r)--+  |      |        |
        	  |  +Xd+  +S(r)-+  +Xd(r)-+Xc(r)+  |
        	  |  |  |  |     |  |      |     |  |
        	John , who you  met , a doctor   , is here

        We solve this problem by creating a "ignore_these_links" list
        in post-processing. Links in this list are simply ignore by
        post-processing; that is, no domains are traced through
        them. We add "Xca" to this list. Then, in the above sentence,
        the "r" domain started by the first MX will not spread through
        the Xca to the second comma phrase; it will be confined to the
        relative clause.

        A third problem involves false positives. Return to the cases
        discussed above, involving consecutive commas: "John , who you
        met , , a doctor , is here". We have allowed for the correct
        version of this sentence, in which the functions of the
        consecutive commas are all served by a single comma. But we
        have not prohibited the consecutive-comma sentence, which is
        of course incorrect. Here again, we have not worried much
        about this false positive, since the kind of rule involved
        ("Do not allow more than one X in a row") seems to arise only
        in punctuation; presumably it could easily be dealt with if it
        were important to do so.

        _Xx: Colons and Semi-Colons_
        Xx is used with colons and semi-colons. Unlike comma phrases -
        where a word is usually the head of the phrase, with the comma(s)
        acting as appendanges - with colons and semi-colons, the two
        phrases are attached together _through_ the punctuation symbol.
        More specifically, the "Xx" attaches the colon or semi-colon back 
        to the wall.

          +-----Xx-----+        
          +-Wd--+      +Wd+
          |     |      |  |
        /////	I left ; Joe stayed

          +----------Xx--------+        
          +-Wd--+              +Wd+
          |     |              |  |              
        /////	I have an idea : we should go
        		         that John should go
        	                 plastics

        With semi-colons, the following phrase must be a clause;
        semi-colons thus carry "Xx- & W+". With colons, the following
        phrase can be a clause, "to-" or "that-" phrase, or noun
        phrase; thus colons have "Xx- & (W+ or J+ or TH+ or TO+)".

Y	is used in certain idiomatic time and place expressions. It
        connects quantity expressions to the head word of the expression
        (i.e., the word that connects to the rest of the sentence):
                     
                     +------MVp-------+
                     |          +--Y--+
                     |          |     |
                We swam three miles away
        	We swam three miles from the shore
        	We swam three weeks ago
        	We swam three hours after we saw you

        Such expressions can usually be used as openers, verb-modifying
        prepositions, or noun-modifying prepositions "We swam three
        weeks ago", "Three weeks ago we swam", "The party three weeks
        ago was great". The word carrying Y- is thus generally conjoined
        with (MVp- or CO+ or Mp-).

        Words that carry Y+ usually also carry {ND-}; this connects
        back to a number expression. In some cases the ND- is optional
        ("We swam after we saw you"); in other cases it is obligatory
        (*"We swam ago"). In some cases, the word carrying Y+ must
        take a prepositional object or clause as well: "*We swam three
        hours after"; such words therefore have "(J+ or Cs+) & Y+".
        In other cases no connection may be made: "*We swam three
        miles away the shore."
        
        Yt is used in time expressions. Words like "ago" and "after"
        have "Yt-"; "days", "weeks", etc, have Yt+, as well as
        idiomatic expressions like "a_long_time". Yd is used in
        distance expressions: "miles" and "feet" carry Yd+,
        prepositions like "from" and "behind" carry Yd-.

        Ya is used for a few adjectives that can take spatial
        expressions as modifiers ("he is three feet tall"). Ye is used
        in the expression "We swim every three weeks". Here "every" is
        treated as the head; but in this case the number expression
        follows the head instead of preceding it. Yx is used in
        expressions like "I did it three times a week". Here the "a
        week" expression is treated as the head.

YP	is used in possessive constructions to connect plural noun 
        forms ending in s to "'". See also YS.

                        +-YP-+--D-+
        		|    |    |
        	The students ' wishes have been neglected.

YS      connects nouns to the possessive suffix "'s". "'s" then acts 
        as a determiner, making a D connection to a noun.

                  +-YS-+-D-+
        	  |    |   |
        	John  's  dog is black

        Plural nouns take only an "'"; "The students' dogs are black".
        This uses a YP connector. The dictionary thus includes the
        following:

        	's: YS- & D+;
        	': YP- & D+;

        A possessive pronoun ("John's") can also act as a complete
        noun phrase: "John's is black". Thus the expressions "'s" and
        "'" also carry the main expression carried by nouns "(S+ or O-
        etc.)". This usage is rare, however; we therefore make it a
        stage 2 usage.

        All proper and common singular nouns have YS+; all plural
        nouns have YP+. (YS is also used for plural forms that don't
        end in s: "Men's legs are longer than women's".)  Nouns which
        form possessive determiners in this way - whether singular or
        plural - take their own determiners and adjectives just like
        ordinary nouns (ex. 1-6 below). However, they may not take
        post-modifiers like relative clauses (ex. 7-8); nor, of
        course, may they act as subjects or objects. Thus YS+ and YP+
        on nouns must be conjoined with "@A- & D+", disjoined with
        everything else.
        
        	1.The rich student was tall
        	2.The rich student's car was black
        	3.*Student was tall
        	4.*Student's car was black
        	5.Students are tall
        	6.Students' cars are black
        	7.The man I met was tall
        	8.*The man I met's car was black

        In a few weird cases, a noun can take both a possessive
        determiner and another determiner; these cannot be handled ("A
        girls' school" is rejected).

Z	connects the preposition "as" to certain verbs.

                 +-----CO--+
                 +--Z--+   |
                 |     |   |
        	As I said, I like broccoli

        Some verbs which take clausal complements - "threaten", "say",
        "mention" - can also be used in constructions like the above;
        in such cases, the complement requirement of the verb - which
        may be mandatory ("*I said") - is satisfied.  Such verbs
        therefore have "Z-" disjoined with their other complement
        connectors (TH+, Ce+, etc.). The word "as" has "Z+" disjoined
        with "Cs+" (used in conjunction phrases) and J+ (used for
        prepositional objects). Z+ is conjoined with "CO+ or MVs-";
        this allows such phrases as closers as well: "I like broccoli,
        as I said".

        as.p: (Z+ or J+ or Cs+) & (CO+ or MVs-);

        The Z connector can also be used in comparative expressions:

        	He was not as late as I expected
        	?He was not as late as I said

        "As.z" (the second "as" in a comparative expression) therefore
        has Z+ as well. However, some verbs can be used in this
        construction only as openers and closers, not as comparatives
        (such as "said" above).  Therefore we subscript the connector
        on "as.z" Zc+; verbs with can not take Z in comparatives have
        Zx-.

        "As" phrases of this kind can also be used without a subject:

        	He likes broccoli, as was expected
        	He earns as much as was expected

         Here, the phrase "was expected" has two demands:
        "is" demands a subject, and "expected" demands a complement.
        "As" must fulfill both of these demands. So "as" has
        a subject connector, optionally conjoined with the Z+. 
        							
        	as.z: ({SFsic+} & Zc+

        			   +----Z-----+
 			           +SF-+--Pv--+
  			           |   |      |
        	he earns as much  as was expected

        "Than" may also be used in this way. The use of comparatives
        is highly constrained by post-processing: see "MVx".

        The "Zc" link here starts a domain, which includes only the
        "than" or "as" phrase.  Note also that the subject connector
        her is an "SF", not a "S", and that it is subscripted.  In
        post-processing, we then enforce the same restrictions on the
        verbs that may occur with "SFsic" that we enforce with the
        filler subjects "it" and "there". Thus we allow "...than
        seemed to be expected", but prohibit "...than wanted to be
        expected", etc.. See "SF: Filler-it" for more explanation.


