Empty categories

As much as possible, empty categories are annotated the same way as overt ones. In particular, all empty categories have phrasal labels, and empty categories other than traces also have POS tags.

The following types of empty categories are distinguished:

Trace of movement

Unlike other empty categories, traces do not have POS tags.

Traces are co-indexed with their antecedents. In cases where an antecedent is not associated with a trace in the syntax (as is possible in connection with ellipsis, notably sluicing), the absence of an ordinary index is highlighted by giving the antecedent a dummy index "x". See Sluicing for examples.

Dash tags with information about grammatical functions and similar information are indicated on the trace, not on the antecedent.

Not all instances of movement are explicitly indicated. In particular, raising is not annotated differently from subject control in the general case. See Trace of A-movement for exceptions.

The following types of traces are represented:

Trace of wh- movement

Wh- traces are represented as *T*. 
( (CP-QUE-MAT (WNP-1 (WPRO What))
              (IP-SUB (BEP are)
	              (NP-SBJ (PRO you))
		      (VP (DAG doing)
		      	  (NP-OB1 *T*-1)))
	      (PUNC ?)))

In instances of pied-piping, a phrasal wh- constituent recursively dominates another phrasal wh- constituent. 

( (CP-QUE-MAT (WPP-1 (P In)
                     (WNP (WD which) (N city)))
              (IP-SUB (MD would)
	              (NP-SBJ (PRO they))
		      (VP (HV have)
		          (VP (PP *T*-1)
			      (VBN lived))))
	      (PUNC ?)))

The following clause types exhibit wh- movement. See the documentation on the individual clause types for examples.

always CP-CMP (comparative)
CP-EOP (empty operator construction, infinitival relative)
CP-FRL (free relative)
CP-REL (ordinary relative)
CP-TMC (tough movement construction)
sometimes CP-ADT (adjunct)
CP-QUE-MAT (direct question)
CP-QUE-SUB (indirect question)

Trace of other A'-movement

Traces of A'-movement other than wh- movement are represented by *ICH* (mnemonic for "interpret constituent here"). These include topicalization (leftward movement of phrases to a pre-subject position) and extraposition (rightward movement). For the moment, ICH traces are implemented systematically only for some cases of topicalization:

Trace of A-movement

Traces of local A-movement (passive or subject-to-subject raising within an IP) are not indicated. But traces of long-distance A-movement (raising out of an IP) are indicated by *. 
( (IP-MAT (NP-SBJ-1 (PRO It@))
          (VP (BEP @'s)
	      (VP (VAN called)
	          (IP-ECM (NP-SBJ *-1)
		          (NP-PRD (N-COMP (NPR Jezebel) (N sauce))))))
	  (PUNC .)))

( (IP-MAT (NP-SBJ-1 (PRO They))
          (VP (BEP are)
	      (VP (VAN believed)
	          (IP-ECM (NP-SBJ *-1)
			  (TO to)
		          (VP (BE be)
			      (ADJP-PRD (ADJ guilty))))))
	  (PUNC .)))

Trace or no trace?

Traces of movement are only assumed in connection with positions available in the overt syntax. For instance, in degree and comparative constructions, the degree complement is more closely associated with the degree head than with the dimension adjective. However, the degree complement is generally ungrammatical in the position next to the degree head. 
  too good to be true
* too to be true good

  more expensive than it is now
* more than it is now expensive

In such a case, our annotation does not include an extraposition trace. 

( (ADJP (ADVP (ADVR too)			← not like this
	      (IP-INF *ICH*-1))
	(ADJ good)
	(IP-INF-1 (TO to)
		  (VP (BE be)
		      (ADJP-PRD true)))))

( (ADJP (ADVP (ADVR too))			← like this
        (ADJ good)
	(IP-INF (TO to)
		(VP (BE be)
		    (ADJP-PRD true)))))

( (ADJP (QP (QR more)				← not like this
	    (CP-CMP *ICH*-1))
	(ADJ expensive)
	(CP-CMP-1 (WADJP-2 (WPRO 0))
		  (C than)
		  (IP-SUB (NP-SBJ (PRO it))
			  (VP (BEP is)
			      (ADJP-PRD *T*-2)
			      (ADVP-TMP (ADV now)))))))

( (ADJP (QP (QR more))				← like this
	(ADJ expensive)
	(CP-CMP (WADJP-1 (WPRO 0))
		(C than)
		(IP-SUB (NP-SBJ (PRO it))
			(VP (BEP is)
			    (ADJP-PRD *T*-1)
			    (ADVP-TMP (ADV now)))))))))

A trace is included only if the phrase moves from the lowest position available in the syntax to some higher position. 

( (IP-MAT (NP-SBJ (PRO It))
	  (VP (BED was)
	      (ADJP-PRD (ADJR better)
			(CP-CMP (WADJP-1 (WPRO 0))
				(C than)
				(IP-SUB (NP-SBJ (PRO it))
					(VP (BEP is)
					    (ADJP-PRD *T*-1)
					    (ADVP-TMP (ADV now))))))
	      (ADVP-TMP (ADV then)))
	  (PUNC .)))

( (IP-MAT (NP-SBJ (PRO It))
	  (VP (BED was)
	      (ADJP-PRD (ADJR better)
			(CP-CMP *ICH*-1))
	      (ADVP-TMP (ADV then))
	      (CP-CMP-1 (WADJP-2 (WPRO 0))
			(C than)
			(IP-SUB (NP-SBJ (PRO it))
				(VP (BEP is)
				    (ADJP-PRD *T*-2)
				    (ADVP-TMP (ADV now))))))
	  (PUNC .)))

Empty subject

The following types of empty subjects are distinguished:

Empty expletive subject

Empty expletive subjects are the silent counterparts of existential (EX) or impersonal pronoun (PRO) subjects with NP or clausal associates. 
( (IP-MAT (NP-SBJ (EX *exp*))
	  (VP (BEP Is@)
	      (NEG @n't)
	      (VP (NP-LGS (D a)
			  (ADJP (ADJ whole))
			  (N lot))
		  (VAN left))))
  (PUNC .)))

( (IP-MAT (NP-SBJ-1 (PRO *exp*))
	  (VP (BEP Was)
	      (ADJP-PRD (ADJ difficult))
	      (IP-INF-1 (TO to)
			(VP (VB deal)
			    (PP (P with)
				(NP (D the)
				    (N aftermath))))))
	  (PUNC .)))

( (IP-MAT (NP-SBJ (NS People))
	  (VP (BEP are)
	      (ADJP-PRD (ADJ happy)
			(PP (P with)
			    (NP (PRO it)))))
	  (PUNC ,)
	  (PAREN (IP-MAT (NP-SBJ (PRO *exp*))
			 (VP (VBP seems)
			     (CP-ADV (C-ADV like)))))
	  (PUNC .)))

Empty thematic subject

For the moment, empty thematic subjects are represented by *.

In future releases, we would like to distinguish, if it is possible to do so systematically, between:

Examples of each type are given below. 



( (IP-MAT (CONJ And)
	  (NP-SBJ (PRO he@))
	  (VP (HVP @'s)
	      (VP (BEN been)
		  (PP (P to)
		      (NP (NP-POS (PRO$ my))
			  (N place)))
		  (NP-TMP (QP (Q several))
			  (NS time))))
	  (PUNC .)))

( (IP-MAT (CP-ADV (C-ADV When)
		  (IP-SUB (NP-SBJ (PRO I))
			  (VP (BED was)
			      (NP-PRD (D a) (N child)))))
	  (PUNC ,)
	  (NP-SBJ (PRO *con*))					← *con*
	  (VP (VBD come)
	      (PP (P to)
		  (NP (NP-POS (NP-POS (PRO$ my))
			      (N$ daddy's)))))
	  (PUNC .)))




( (IP-MAT (NP-SBJ (D That@))
	  (VP (BEP @'s)
	      (NP-PRD (D the)
		      (ADJP (ADJ only))
		      (N place)
		      (CP-REL (WADVP-1 (WADV 0))
			      (C-REL 0)
			      (IP-SUB (NP-SBJ (PRO I))
				      (VP (ADVP-TMP (ADV ever))
					  (DOD did)
					  (VP (VB go)
					      (ADVP-DIR *T*-1)))))))
	  (PUNC ,)))

( (IP-MAT (NP-SBJ (PRO *pro*))
	  (VP (ADVP-TMP (ADV Never))
	      (DOD did)
	      (VP (VB go)
		  (PP (P to)
		      (NP (QP (Q no))
			  (ADJP (ADJ other))
			  (N school)
			  (PP (P but)
			      (ADVP (ADVP (ADV right))
				    (RP up)
				    (ADV there)))))))
	  (PUNC .)))

Arbitrary subject in ECM infinitive

The arbitrary subject of an infinitive governed by an ECM verb is labelled *arb*. Though arbitrary in reference, this subject is not PRO (= the subject of control complements), but rather corresponds to an overt oblique argument. 
( (IP-MAT (CONJ but)
	  (NP-SBJ (PRO you))
	  (VP (ADVP-TMP (ADV never))
	      (VBD heared)
	      (IP-ECM (NP-SBJ (PRO *arb*))
		      (VP (VB tell)
			  (PP (P of)
			      (NP (D a)
				  (N-COMP (N man) (N doctor)))))))
	  (PUNC .)))

Other empty category

See also Empty categories that are neither traces nor empty subjects are represented by 0. By default, empty categories are avoided if possible (see HAD for a common case). 
( (IP-MAT (CONJ And)
	  (NP-SBJ (PRO we@))
	  (VP (MD @'d)
	      (VP (VP (VB gather)
		      (NP-OB1 (D them)
			      (ADJP (ADJ little))
			      (NS chickens)))
		  (PUNC ,)
		  (CONJP (CONJ and)
			 (VP (VB put)
			     (NP-OB1 (PRO 0))				← empty object pronoun
			     (PP (P in)
				 (NP (D that=there) (N kettle)))))))
	  (PUNC .)))

( (CP-QUE-MAT (INTJ Well)
	      (PUNC ,)
	      (ADVP (ADV now))
	      (PUNC ,)
	      (IP-SUB (DOD did)
		      (NP-SBJ (PRO you))
		      (VP (VB help)
			  (NP-OB1 (NP-POS (PRO$ your))
				  (N daddy))
			  (IP-INF (TO 0)				← empty infinitival marker
				  (VP (VB rob)
				      (NP-OB1 (D the) (NS bees))))
			  (ADVP-TMP (ADV ever))))
	      (PUNC ?)))


( (IP-MAT (NP-SBJ (PRO I))
	  (VP (BED =uz)
	      (VP (GTG getting)
		  (NP-OB1 (NP (N milk))
			  (CONJP (CONJ and)
				 (NP (N butter))))
		  (PP (P from)
		      (NP (D a)
			  (N lady)
			  (CP-REL (WNP-1 (WPRO 0))			← empty relative pronoun
				  (C-REL that)
				  (IP-SUB (NP-SBJ *T*-1)
					  (VP (VBD lived)
					      (ADVP-LOC (ADVP (ADV right))
							(ADV there)))))))))
	  (PUNC ,)))

( (IP-MAT (CONJ and)
	  (NP-SBJ (PRO you))
	  (VP (MD would)
	      (VP (GT get)
		  (NP-OB1 (D the)
			  (NS clothes)
			  (CP-REL (WNP-1 (WPRO 0))			← empty relative pronoun
				  (C-REL 0)    				← empty complementizer
				  (IP-SUB (NP-SBJ (PRO you))
					  (VP (VBD needed)
					      (NP-OB1 *T*-1))))
			  (PP (P for)
			      (NP (D the) (NS children))))))
	  (PUNC .)))

The position of empty categories

Given that they are silent, the linear position of empty categories in the parse cannot always be uniquely determined. The following rules apply.