Course:LING300/Constituency

Constituents and why they matter

Constituents are the bread and butter of syntax. The basic premise of syntactic analysis is that every sentence of every human language can be broken down into units called "constituents". And conversely, every sentence of every human language is built by combining constituents with each other. Although most linguists now take for granted the existence of constituents, it is in fact a relatively new idea, and was introduced into linguistic analysis about 80 years ago in the 1940s by Rulon S. Wells in a paper entitled Immediate Constituent Analysis. (Although 80 years might not seem so recent, it's recent compared to other ideas about human language, such as the claim that there are categories; the category idea has been around for a couple thousand of years, and in the Western tradition of scholarship, dates back to Aristotle.)

Constituency tests as a form of experimentation

Once the existence of constituents as the organizing units of sentences was recognized, attention turned to developing constituency tests for individual languages. These tests are sometimes called diagnostics. Just as a medical test (also called a diagnostic test) is used to detect a condition, a constituency test or diagnostic is used to determine how a sentence is structured.

Constituency tests are mini-experiments

A constituency test is a form of experimentation, and is designed to deliberately perturb sentence structure by manipulating one variable in the sentence, and holding everything else in the sentence constant. According to whether the result of this experimental perturbation is "positive" or "negative", then the syntactician can draw conclusions about the structure of the sentence. This is just like a medical test, which can yield either a positive of negative result. For language, a constituency test is positive if the result of applying it yields a well-formed ("grammatical") sentence. Conversely, a constituency test is negative if the result of applying it yields an ill-formed ("ungrammatical") sentence.

Interpreting the results of a constituency test

While a positive result confirms the existence of a constituent, a negative result is more difficult to interpret. In other words, just because a particular constituency test fails to apply, this does not necessarily indicate that the target string is not a constituent. That's why it's necessary to apply several constituency tests.

Constituency tests range over syntactic minimal pairs

Applying a syntactic test, and this includes constituency tests, always always a yields syntactic minimal pair. In other words, two examples are compared:

a. the base string;

b. the permuted string that arises from applying the test.

Some examples are given below of syntactic minimal pairs that arise from applying constituency tests.

(1) substitution test: pro-form substitution
    a. [That girl in the blue dress] put a picture on your desk.
    b. [She] put a picture on your desk.

(2) movement test: clefting
    a. [That girl in the blue dress] put a picture on your desk.
    b. [That girl in the blue dress] is the one who put a picture on your desk.

(3) coordination test: conjunction
    a. [That girl in the blue dress] put a picture on your desk.
    b. [[Maya] and [that girl in the blue dress]] put a picture on your desk.

Constituent structure as a form of knowledge

Constituent structure is more than a useful fiction for syntacticians. There is experimental evidence that speakers have tacit knowledge of constituent structure. This knowledge is detectable in tasks that involve speech perception and lexical recall, and is also discernible in neural activity.

Constituent structure causes "auditory illusions"

Studies by Ladefoged & Broadbent (1960), and Fodor & Bever (1965) show that if speakers are asked to listen to a randomly placed click while they are listening to a sentence, they perceive the click as being close to a major constituent boundary, even when it isn't. This "auditory illusion" persists even if speakers know where the click actually is (because they've placed it there themselves); they still perceive it as being near a constituent boundary. (See Sportiche, Koopman & Stabler §3.10.1 for details.)

Constituent structure affects word recall

Studies by Johnson (1965) show that speakers prone to make more errors int task that require them to recall a specific word, if that word occurs at the beginning of a major constituent. (See Sportiche, Koopman & Stabler §3.10.2 for details.)

Constituent structure is reflected in neural activity

Syntactic processing is associated with specify types particular type of neural activity. (See Sportiche, Koopman & Stabler §3.10.3 for details.)

Production of syntactic errors: localized to the prefrontal area of the brain (Indefrey et al. 2001)
Detection of syntactic anomalies: associated with distinctive neural activity (Kang et al. 1999; Featherson et al. 2000; Embick et al. 2001)
Size of syntactic constituents: associated with an increase in neural activation (Pallier, Devauchelle, and Dehaene 2011)

Some common constituency tests

Constituency tests apply to all languages. Some tests are completely general, in that they can be used in any languages, but other tests are language-specific and only apply to some languages. (Again this is similar to medical tests, some of which can be used for all humans, while others are used to test specific conditions in some humans.) In trying to detect a constituent, it is useful to try to use several constituency tests. The table below summarizes which constituency tests can be used to test for what.

Type of test	Test	Constituent
Substitution	substitution with a pro-form	NP, DP, PP, VP, CP
	substitution with a null string	DP, VP
Movement	Topicalization	DP, PP, VP, CP
	Clefting	DP, PP
	Pseudo-clefting	DP, PP, AP, VP, CP
	Wh-movement	DP, PP, AP
	Heavy constituent shift	DP, PP
Coordination	and, or, but not	X(P)

Following is an overview of the more widely used constituency tests, with examples of how they apply to English.

Substitution tests

Substitution tests are of two types: (i) those that involve replace a string with a pro-form; (ii) those that involve eliding or deleting a string (i.e. replacing a string with a null string).

Substitution with a pro-form

Pro-forms fall into difference classes according to the constituent that they substitute for. All languages have pro-forms, but one must determine for a given language, what the inventory of pro-forms is in that language. Substitution with a pro-form can be used to identify NP, DP, PP, VP, and CP.

Examples of pro-forms in English

(1) pro-NP: one substitutes for NP
    a.  Maya bought that [_NP big red shirt ]
    b.  Maya bought that [_NP one ]

(2) pro-DP: personal pronouns and reflexives substitute for DP
    a. Maya bought [_DP that big red shirt ]
    b. Maya bought [_DP it ]

(3) pro-PP: there substitutes for PP (Caution, there also functions as a pro-DP)
     a. Maya put the book [_PP on the shelf ]
     b. Maya put the book [_PP there ]

(4) pro-VP: do so substitutes for VP
    a. Maya will edit the Wikipedia article hastily, but Emily will [_VP edit the Wikipedia article ] carefully.
    b. Maya edited the Wikipedia article hastily, but Emily [_VP do so ] carefully.

(5) pro-CP: so, this, and that substitute for CP
    a. Maya said that she would leave early, and Emily said [_VP that she would leave early too ]
    b. Maya said that she would leave early, and Emily said [_VP so ] too.

Substitution with a null string

Technically, ellipsis or deletion is equivalent to substituting a substring with a null string (i.e., a string that contains nothing). A null string is symbolized as [ e ], where e stands for "empty category". In English, substitution with a null string is only found with VP. In other languages it also occurs with DP, such empty DPs are labelled pro (called "small pro or null pronoun)), and they are analyzed as phonologically null pro-DPs.

Examples of VP-elipsis in English

(1) VP-ellipsis with statement/response sequence
    a. Maya really does [_VP like paintings by Picasso] !.
    b. She does not [_VP e] !

(2) VP-ellipsis with coordination
    a. Maya will [_VP go to the movies] and Francis will [_VP go to the movies ] too.
    b. Maya will [_VP go to the movies] and Francis will [_VP e ] too

(3) VP ellipsis with tag question
    a. Maya will not eat gluten, will she [_VP not eat gluten]?
    b. Maya will not eat gluten, will she [_VP e]?

Movement tests

Topicalization

Topicalization places a constituent at the beginning of the sentence. In many languages, including English, topicalization is associated with special prosody and a pause (represented in the orthography as a comma ","). A sentence with topicalization always has a counterpart sentence where topicalization has failed to apply. The topicalization test is used to identify DP, PP, VP, and CP.

Examples of topicalization in English

(1) DP topicalization
    a. That girl in the blue dress will put [_DP the picture of the Queen ] on your desk before tomorrow.
    b. [_DP The picture of the Queen ], that girl in the blue dress will put [_DP e ] on your desk before tomorrow.

(2) PP topicalization
    a. That girl in the blue dress will put a picture of the Queen [_PP on your desk ] before tomorrow.
    b. [_PP On your desk ], that girl in the blue dress will put a picture of the Queen [_PP e ] before tomorrow.

(3) VP topicalization
    a. That girl in the blue dress will [_VP put a picture of the Queen on your desk ] before tomorrow.
    b. [_VP Put a picture of the Queen on your desk ], that girl in the blue dress will [_VP e  ] before tomorrow.

(4) CP topicalization
    a. Surely Maya must already know [_CP that you leave tomorrow ].
    b. [_CP That you leave tomorrow ], surely Maya must already know [_VC e ].

Clefting

Clefting removes a constituent from its source sentence and places it after an it BE sequence. The general formula for a cleft sentence is:

[_CP … X …] → it BE X [_CP that … e … ].

In a cleft sentence, X corresponds to the focused constituent. Clefting is used to identify DP and PP In all languages. In some languages (but not English), it can also be used to identify VP. (VP clefting is sometimes called predicate clefting.)

Examples of clefting in English

(1) DP clefting
   a. Maya want to look at [_DP your notes ] after class.
   b. It's [_DP your notes' ] that Maya want to look at [_DP e ] after class.

(2) PP clefting
   a. Maya want to look at your notes [_DP after class ].
   b. It's [_PP after class ] that Maya want to look at your notes [_DP e ].

Pseudo-clefting

Pseudo-clefting takes a constituent XP from its source sentence and places it at the end of its source sentence after BE or DO, and then combines a wh-expression together with the sentence remnant. The general formula for a pseudo-cleft sentence is:

[_CP … XP …] → [WH-XP [_CP that … e … ]] BE/DO XP

In a cleft sentence, XP corresponds to the focused constituent. Pseudo-clefting is used to identify (a focused) DP, PP, AP, VP, and CP. (See SKS §3.7.3 for discussion of speaker variation regarding pseudo-clefting.)

Examples of pseudo-clefting in English

(1) DP pseudo-clefting
    a. Maya wantd to look at [_DP your notes ] after class.
    b. [What Maya wants to look at [_DP e ] after class] is [_DP your notes ].

(2) PP pseudo-clefting
    a. Maya wants to go [_PP to Greece ] for her holidays.
    b. [Where Maya wants to go [_PP e ] for her holidays] is [_PP to Greece ].

(3) AP pseudo-clefting
    a. Maya has always been [_AP very kind to animals ].
    b. [What Maya has always been [_AP e ]] is [_AP very kind to animals ].

(4) VP pseudo-clefting
    a. Maya will [_VP put a picture of the Queen on your desk ].
    b. [What Maya will do [_VP e ]] is [_VP put a picture of the Queen on your desk ].

(5) CP pseudo-clefting (tensed CP)
    a. Maya promised [_CP that she would be gentle ].
    b. [What Maya promised [_CP e ]] is [_CP that she would be gentle ].

(6) CP pseudo-clefting (infinitival CP)
    a. Maya wants [_CP (for Emily) to leave ].
    b. [What Maya wants [_CP e ] is] [_CP (for Emily) to leave ].

Wh-movement

Wh-movement is a term used in generative grammar to describe content questions, which in English involved the use of words that typically begging with wh, e.g. who, what, where, when, why. Forming a wh-question actually folds together two constituency tests. First, the Wh-word is a pro-form that substitutes for a constituent. Second, in many languages, including English, wh-expresssions can be moved or displaced to the beginning of the sentence; this is called "wh-movement. Wh-movement can be used to identify DP, PP, and DegP.
In addition to being a useful test for constituency, wh-movement plays a pivotal role in the analysis of the property "movement" (also called the "displacement property"". See Unit 3 for discussion, as well as Chapter 10 of Sportiche, Koopman and Stabler.

Examples of wh-movement in English

(1) Wh-movement of DP
    a. Henry bought [_DP these poetry books ] last week.
    b. Henry bought [_DP which books ] last week?
    c. [_DP Which books ] did Henry buy [_DP e ]  last week?

(2) Wh-movement of PP
     a. Henry put those poetry books [_PP beside the table] yesterday.
     b. Henry put those poetry books [_PP where] yesterday?
     c. [_PP Where ] did Henry put those poetry books [_PP e ] yesterday?

(3) Wh-movement of DegP
    a. Maya is [_DegP very proud of her accomplishments ].
    b. [_DegP How proud of her accomplishments ] is Maya [_DegP e ]?

Heavy constituent shift

Heavy constituent shift applies to constituents that are "heavy", where heaviness is determined by the relative size of the phrase. Phrases that contain just one word, for example pronouns like it and there are not heavy, and so are not subject to heavy constituent shift. But phrases that contain several words are heavy, and so can undergo heavy constituent shift. Although not all constituents undergo heavy constituent shift, only strings that are constituents do. With heavy constituent shift, a constituent is shifted from its normal position to the end of the sentence; this test is used to identify DP and PP constituents.

Terminological note: In the early literature of generative grammar, "heavy constituent shift" is called "heavy NP shift". More recent analyses analyze NP as DP (see section on X-bar Theory). In addition, not only do nominal phrases (DPs) undergo heavy constituent shift, but so too do PPs. For this reason, the more general term — heavy constituent shift — is now used.

Examples of heavy constituent shift in English with DP

(1) pro-DPs do not undergo heavy constituent shift
    a.  I sent [_DP it ] to you.
    b. *I sent to you [_DP it ].

(2) relatively complex DPs can, but need not, undergo heavy constituent shift
    a. I sent [_DP the recipes from the paper ] to you.
    b. I sent to you [_DP the recipes from the paper ].

(3) very complex DPs must undergo heavy constituent shift
    a.  I sent [_DP the recipes from the paper that I told you about yesterday ] to you.
    b.  I sent to you [_DP the recipes from the paper that I told you about yesterday ].

Examples of heavy constituent shift in English with PP

(1) pro-PPs do not undergo heavy constituent shift
    a.  I sent Maya [_DP there ] this morning.
    b. *I sent Maya this morning [_DP there ].

(2) relatively complex PPs can, but need not, undergo heavy constituent shift
    a. I sent Maya [_DP to the store across the street ] this morning.
    b. I sent Maya this morning [_DP to the store across the street ].

(3) very complex DPs must undergo heavy constituent shift
    a. *I sent Maya [_DP to the store across the street and down the hill ] this morning.
    b.  I sent Maya this morning [_DP to the store across the street and down the hill ].

Coordination tests

The coordination of two constituents of type X yields a larger constituent of the same type. For example, the conjunction marker and combines elements of the same to with each other:

[_X(P) [ X(P) and X(P) ] ]

Other markers of coordination include or and but not. Like and, they have the property of combining two constituents of type X to yield a larger constituent of the same type. This means that it is possible to have a general formula for coordination:

[_X(P) [ X(P) COORD X(P) ] ]

Because coordination requires that the elements it combines be of the same type, it can be used as a test for constituent structure.

All languages have some way of marking coordination, but languages differ in the logical properties of these coordinating particle. For example, English coordinating particles are are completely general in that they use the same coordinating particle for all categories. In other languages, coordinating particles may be restricted to nominal and clausal constituents, i.e., to DP and CP, and VP-level coordination is marked in a different way, for example by a serial verb constructions

Examples of how the coordination test applies to English

(1) Examples of XP coordination
    a. The [_NP [_NP big dog ] and [_NP small cat ]] chased each other.
    b. [_DP [_DP Me ] and [_DP Maya ]] went hiking together last week.
    c. I've been [_PP [_PP to Teslin ] and [_PP to Skagway ]].
    d. Maya will [_VP [_VP play the violin ] and [_VP dance a jig ]].
    e.  I know that [_TP [_TP Maya will play the fiddle ] and [_TP Emily will dance a jig ]].
    f. I know [_CP [_TP that Maya will play the fiddle ] and [_TP that Emily will dance a jig ]].

(2) Examples of X⁰ coordination
    a. The big [_N [_N cats ] and [_N dogs ]] chased each other.
    b. [_D [_D These ] and [_D those ]] cars are energy-efficient.
    c. I've been [_P [_P up ] and [_P down ]] the stairs.
    d. Maya will [_V [_V walk ] and [_V run ]] to school.
    e. Maya [_T [_T can ] and [_T will ]] play the fiddle.
    f. I don't know [_C [_C that ] or [_C if ]] Maya will play the fiddle.

Ellipsis

Using constituency tests to detect the complement/adjunct distinction

Constituency tests can be used to detect the difference between a complement and an adjunct.

Using word order to detect complement and adjuncts

While complements have a fixed position, adjuncts do not.

Complements have a fixed position

English is a head-initial language; this means that the head of a phrase precedes its complement. This predicts that the complement appears in a fixed position in English. In particular, the [Head-Complement] word order is the only well-formed word order in English. This is confirmed by the following data;

(1) V_HEAD DP_COMPL
    a.  [_V.HEAD see] [_DP.COMPL the groom]
    b. *[_DP.COMPL the groom] [_V.HEAD see]

(2) P_HEAD DP_COMPL
    a.  [_P.HEAD with] [_DP.COMPL the groom]
    b. *[_DP.COMPL the groom] [_P.HEAD with]

(3) N_HEAD PP_COMPL
    a.  [_N.HEAD sister] [_PP.COMPL of the groom]
    b. *[_PP.COMPL of the groom] _N.HEAD sister]

(4) A_HEAD PP_COMPL
    a.  [_A.HEAD proud] [_PP.COMPL of the groom]
    b. *[_PP.COMPL of the groom] [_A.HEAD proud]

Adjuncts do not have a fixed position

As modifiers, adjuncts can precede or follow the phrase that they modify. In other words, adjuncts do not have a fixed position.

Predicate modifiers

(1) a. VP_PRED Adv
       Lucy [_VP.PRED opened the door] [_AdvP carefully]
    b. Adv VP
       Lucy [_AdvP carefully] [_VP.PRED opened the door]

(2) a. VP_PRED PP_MOD
       Lucy [_VP.PRED opened the door] [_PP.MOD without hesitation]
    b. PP_MOD VP_PRED
       Lucy, [_PP.MOD without hesitation], [_VP.PRED opened the door]

(3) a. PP_PRED Adv
       Lucy is [_PP.PRED on the road] [_AdvP reluctantly]
    b. Adv PP_PRED
       Lucy is, [_AdvP reluctantly], [_PP on the road]

(4) a. PP_PRED PP_MOD
       Lucy is [_PP.PRED on the road] [_PP.MOD with reluctance]
    b. PP_MOD PP_PRED
       Lucy is, [_PP.MOD with reluctance], [_PP.PRED on the road]

(5) a. NP_PRED Adv
       Lucy is [_NP.PRED chair of the meeting] [_AdvP occasionally]
    b. Adv NP_PRED
       Lucy is, [_AdvP occasionally], [_NP.PRED chair of the meeting]

(6) a. NP_PRED PP_MOD
       Lucy is [_NP.PRED chair of the meeting] [_PP on occasion]
    b. PP_MOD PP_PRED
       Lucy is, [_PP on occasion], [_NP chair of the meeting]

(7) a. AP_PRED Adv
       Lucy is [_AP.PRED proud of her brother] [_AdvP frequently]
    b. Adv AP_PRED
       Lucy is [_AdvP frequently] [_AP.PRED proud of her brother]

(8) a. AP_PRED PP_MOD
       Lucy is [_AP.PRED proud of her brother] [_PP on occasion]
    b. PP_MOD AP_PRED
       Lucy is, [_PP on occasion], [_AP proud of her brother]

Using ellipsis to detect complements and adjuncts

Practice exercises for constituency tests

Before completing these practice exercises, review Chapter 3 of Sportiche, Koopman & Stabler.

Recognizing Constituency Tests

Consider the following minimal pairs of sentences. For each pair indicate which constituency test has been applied and which constituent it identifies. The first example is done for you by way of illustration.

Replacement tests: These tests include replacement via ellipsis ((also called "deletion") or replacement via a pro-form. Replacement tests are also called substation tests.

    (1) a. The new students will study Russian.
        b. The new students will too.

Sample answer: (1a) is an example of replacement via ellipsis. The constituent that is identified is VP, specifically study Russian. Because the VP constituent is targeted, this test is called "VP-ellipsis" or "VP-deletion".

    (2) a. This client might buy that red sports-car
        b. This client might buy that one.

    (3) a. Lucy’s older brother will raise cattle; Maya’s older brother will raise sheep.
        b. Lucy’s will raise cattle; Maya’s will raise sheep.

    (4) a. The girl with the red hair should put the heavy book on the table.
        b. The girl with the red hair should put the heavy book there.

    (5) a. The girl with the red hair should put the heavy book on the table.
        b. The girl with the red hair should put it on the table.

    (6) a. The girl with the red hair should put the heavy book on the table.
        b. She should put the heavy book on the table.

    (7) a. Lucy will write her memo quickly.
        b. Lucy will do so quickly.

    (8) a. Lucy will write her memo quickly.
        b. Lucy will write hers quickly.

Movement tests: Movement tests (also called displacement tests) include topicalization, clefting, pseudoclefting, and wh-movement.

    (9) a. Lucy will write her memo quickly.
        b. Write the memo quickly, Lucy will.

    (10) a. Lucy will write to her cousins in Italy.
         b. To her cousins in Italy, Lucy will write.

    (11) a. Lucy will visit her cousins in Italy.
         b. Her cousins, Lucy will visit in Italy.

    (12) a. Lucy must think that Maya will visit soon.
         b. That Maya will visit soon, Lucy must think.

    (13) a. Lucy must think that Maya will visit soon.
         b. That Maya will visit soon, Lucy must think.

    (14) a. Lucy bought a red sportscar.
         b. It’s Lucy who bought a red sportscar.

    (15) a. Lucy bought a red sportscar.
         b. A red sportscar is what Lucy bought.

    (16) a. Lucy bought a red sportscar.
         b. Buy a red sportscar is what Lucy did.

    (17) a. Lucy bought a used sportscar from the dealership.
         b. It’s from the dealership that Lucy Lucy bought a used sportscar

    (18) a. The girl with the red hair put the heavy book on the table.
         b. What did the girl with the red hair put on the table?

    (19) a. The girl with the red hair put the heavy book on the table.
         b. Where did the girl with the red hair put the heavy book?

    (20) a. The girl with the red hair put the heavy book on the table.
         b. Who put the heavy book on the table?

    (21) a. Lucy will be amazingly tall.
         b. How tall will Lucy be?

    (22) a. Lucy wrote her memo very quickly.
         b. How quickly did Lucy write the memo?

Coordination tests: Coordination test include conjunction with and and disjunction with or.

    (23) a. Lucy might read these books during the holidays.
         b. Lucy might read these books and those magazines during the holidays.

    (24) a. Lucy might read these books during the holidays.
         b. Lucy might read these books and magazines during the holidays.

    (25) a. Lucy will read the book.
         b. Lucy will read the book or see the movie.

    (26) a. Lucy might read the book.
         b. Lucy might read the book but will certainly see the movie.

    (27) a. Sarah must think that Lucy will leave.
         b. Sarah must think that Lucy will leave or that Maya will stay.

    (28) a. Sarah must think that Lucy will leave.
         b. Sarah must think that Lucy will leave and Maya will stay.

    (29) a. Lucy saw the incredibly tall woman.
         b. Lucy saw the incredibly tall woman and short man.

    (30) a. Lucy saw the incredibly tall woman.
         b. Lucy saw the incredibly tall and amazingly strong woman.

    (31) a. Lucy has lived in New York.
         b. Lucy has lived in New York and in Los Angeles.

Applying Constituency Tests

Show how the constituency tests apply to the following example sentences. The first example is done for you by way of illustration.

 (1) Lucy lived in New York.
     a. test: pro-form replacement, subject DP
     b. test: pro-form replacement, PP
     c. test: movement via clefting, PP

Sample answer for (1)

The result of applying the pro-form replacement to the subject DP is:

(1) a. She lived in New York.

The result of applying the pro-form replacement test to PP is:

(1) b. Lucy lived there.

The result of applying the cleft movement test to PP is:

(1) c. In New York is where Lucy lived.

(2) Lucy will read the contract very carefully.
    a. test: pro-form replacement, VP
    b. test: wh-movement, AdvP
    c. test: VP topicalization

(3) Lucy will be proud of her daughter.
    a. test: pseudocleft, AP
    b. test: topicalization, AP
    c. test: DP-ellipsis

(4) Lucy will be proud of her daughter.
    a. test: pseudocleft, AP
    b. test: topicalization, AP
    c. test: DP-ellipsis