PRO (linguistics)

PRO (pronounced big pro, to distinguish it from pro, pronounced little/small pro) is an empty category whose existence is first postulated in classical Government and Binding Theory.

In recent literature the existence of PRO has been argued for without the help of the PRO Theorem and government theory. It is posited that PRO is in complementary distribution with overt subjects because it is the only item that is able to check "null case" from a non-finite T, namely the 'to' in control infinitives. By positing this, the problems arising from the PRO Theorem are avoided. The problem was that PRO serves as the subject of the control clause and must, as such, be case-marked. However, following the traditional Government and Binding framework, case-marking only applies under government. Thus, PRO could not be case-marked since it must be ungoverned. If PRO receives null-case, it is licensed to appear in subject position.

Motivation for PRO

There are several independent pieces of linguistic theory which motivate the existence of PRO. These motivating theories are outlined below.

The Extended Projection Principle

Due to the Extended Projection Principle (EPP) requiring that all clauses must have a subject there is strong arguments that the idea of a null pronominal or PRO is apparent. If EPP did not agree with the idea of PRO then it would be difficult to explain a grammatical sentences such as:

(1) John promised Bill to control the situation.
    Johnj promised Bill [CP[TP PROj to control the situation]]
(2) John convinced Bill to sleep.
    John convinced Billk [CP[PROk to sleep]]
    (adapted from: Koopman, Sportiche and Stabler 2014: 247 (29), 251 (47b))

In the examples above, the top sentences in (1) and (2) show the surface sentences, whereas the bottom expansions show that PRO serves as the subject of the non-finite clauses, [_CP to control the situation] and [_CP to sleep], thereby satisfying the EPP-feature of T (occupied by the infinitival 'to' in the examples). Since (1) is a subject control sentence, PRO is co-indexed with its antecedent 'John', and since (2) is an object control sentence, PRO is co-indexed with its antecedent 'Bill'.

Theta Criterion

It is understood that every verb has theta roles and that under the Theta Criterion every theta role must be present in the structure of the sentence. These theta roles must consist of a syntactic category even when there is no overt subject; therefore, PRO would help to satisfy the Theta Criterion. For example:

(1) a. John promised Mary to examine the patient.
    b. John promised Mary [PRO to examine the patient].

In this example sentence the verb "examine" has the lexical entry, examine: V <DP_agent DP_theme>; therefore, "examine" must have a DP (determiner phrase) as an agent and a DP as a theme; however, in sentence (1)a it is seen that there is not a DP as an agent, which does not satisfy the Theta Criterion. In this case, PRO is realized as the appearance of a null subject, as shown in sentence (1)b. Sentence (1)b satisfies the Theta Criterion by having PRO as the DP_agent in the sentence and [the patient] as the DP_theme.

Binding Theory

The conclusion that non-finite clauses must have a phonetically null subject is also motivated to satisfy basic binding theory. This is because non-finite clauses may contain anaphors which would lack a local antecedent if the null subject PRO was not present.

(16)(a) It's important [PROi to prepare myselfi properly for the exam]
    (b) It's important [PROj to prepare oneselfj properly for the exam]
(Radford 2004: 111)

The reflexive pronouns myself and oneself need a local antecedent for the sentences to be grammatical. PRO can fill this antecedent role, as shown by the co-indexations; however, PRO itself has no local antecedent in these examples, illustrating that PRO can share reference with an external referent as in example 16 a, or have an arbitrary reading as in example 16 b.

Nominal Agreement

Further evidence that non-finite clauses do have a phonetically null subject is given by the fact that predicate nominals must agree with the subject of the clause in a copular clause.

(13)(a) They want [their son to become a millionaire/*millionaires]
    (b) He wants [his sons to become millionaires/*a millionaire]

(14)(a) They want [PRO to become millionaires/*a millionaire]
    (b) He wants [PRO to become a millionaire/*millionaires]
(Radford 2004: 110)

In example 13 the number of the predicate nominal, a millionaire/millionaires, must agree with that of the copular subject, their son/his sons, for the phrase to be grammatical. In (14) the predicate nominal appears to be in agreement with the subject of the verb want. This can be explained simply if PRO is co-referenced with the top level subject, then the predicate nominal is simply in agreement with PRO (the subject of the copular clause), just as it would be if an overt subject had been introduced.

Theoretical Characterization of PRO

The PRO Theorem

The interpretation of PRO may be either dependent on another noun phrase, like anaphors, or arbitrary, like pronominals. That is why in terms of features PRO may be described as [+anaphor, +pronominal]; however, this set of features would pose a problem for Binding Theory: an anaphor must be bound in its governing category, whereas a pronominal must be free in its governing category. The solution that Chomsky proposed was the so-called PRO Theorem: PRO must be ungoverned. In other words, PRO cannot be governed and that is why it cannot have a governing category. Under this consideration, the features of PRO no longer conflict with the principles of Binding Theory, though developments in Binding Theory since 1981 have presented significant challenges to with Chomsky's postulation. For example, if PRO is ungoverned, then it must not be case-marked; however, it has been observed that in Icelandic, PRO appears to be case-marked, and is thus governed. Some more recent texts characterize PRO without any references to the PRO theorem.

Null Case of PRO

It has been argued that PRO does in fact have case, which is checked by non-finite, T. This is illustrated by ungrammatical sentences such as (1), (2) and (3) below, where PRO may be used in the specifier position of the TP (Tense Phrase) in which other DPs, such as the proper nouns in the examples, cannot be used.

(1) *Kerry attempted [Bill [T to] study physics].
(2) *Kerry persuaded Sarah [Bill [T to] study physics].
(3) *It is not easy [Bill [T to] study physics].
(adapted from Martin 2001: 144 (13))

Due to the fact that the subject of the non-finite T must satisfy the case checked by T, and the observation that this case cannot be satisfied by a pronounced DP, it is argued that these non-finite T's (and -ing clausal gerunds), check for Null Case, and that the only DP reconcilable with such a case is PRO. This can be seen by replacing the DP, Bill, with PRO in any of the ungrammatical sentences (1), (2) and (3) above, and finding that these sentences are now considered grammatical.

(4) Kerry attempted PRO to study physics.
(5) Kerry persuaded Sarah PRO to study physics.
(6) It is not easy PRO to study physics.
(adapted from Martin 2001: 144 (13))

It is furthermore argued that Null Case is the only case assignable to PRO, and that PRO is the only DP to which Null Case may be assigned. These assertions have since been challenged by certain data which appear to demonstrate that PRO may carry case other than Null Case.

Distribution of PRO

The distribution of PRO in infinitival clauses can be demonstrated by the following data. PRO has different distributions depending on whether it appears in an obligatory control or non-obligatory control context.

Distribution Under Obligatory Control

(1) a. *It was expected PRO to shave himself.
(1) b. *John thinks that it was expected PRO to shave himself.
(1) c. *John’s campaign expects PRO to shave himself.
(Hornstein 1999: 73)

Examples 1 a, b, and c demonstrate that PRO's antecedent must be present, local and must c-command PRO.

(1) d. John expects PRO to win and Bill does too. (= Bill win)
(Hornstein 1999: 73)

Example 1 d demonstrates that only the sloppy reading is allowed under VP ellipsis.

(1) e. *Johni told Maryj PROi+j to wash themselves/each other.
(Hornstein 1999: 73)

And example 1 e demonstrates that PRO may not have split antecedents.

Distribution Under Non-Obligatory Control

(2) a. It was believed that PRO shaving was important.
(2) b. Johni thinks that it is believed that PROi shaving himself is important.
(2) c. Clinton’si campaign believes that PROi keeping his sex life under control is necessary for electoral success.
(Hornstein 1999: 73)

In contrast to PRO under obligatory control, examples 2 a, b, c illustrate that PRO under non-obligatory control may have an antecedent which is not local, does not c-command PRO, or which is not present at all.

(2) d. John thinks that PRO getting his resume in order is crucial and Bill does too.
(Hornstein 1999: 73)

Example 2 d demonstrates that both the sloppy and strict readings are permitted. That is Bill may think that John having his resume in order is crucial, or that Bill thinks that having his own resume in order is crucial.

(2) e. Johni told Maryj [that [[PROi+j washing themselves/each other] would be fun]].
(Hornstein 1999: 73)

And example 2 e demonstrates that PRO allows split antecedents under non-obligatory control.

Movement of PRO

There is discussion of whether PRO moves to Spec-TP position in non-finite clauses. This movement would exist to satisfy the EPP properties of the tense marker to, however it is argued by Baltin that PRO does not move and that the tense marker to does not have an EPP feature. This assertion is argued against by Radford. The following data from Radford illustrates the point of contention.

(61)(a) They don't want [to see you]
(61)(b) They don't want [CP [C ∅] [TP [T to] [VP PRO [V see] you]]]
(61)(c) They don't want [CP [C ∅] [TP PRO [T to] [VP PRO [V see] you]]]
(Radford 2004: 309)

If to has an EPP feature then the theory would predict that PRO would move to the Spec-TP position as in example 61 c. Baltin argues that this is not the case as placing PRO between want and to would block the contraction of want and to into wanna (They don't wanna see you). Radford argues that this contraction can be achieved while still allowing to to have an EPP feature by first having to cliticise onto the null complementizer ∅ then having this compound cliticise onto want. Radford justifies this by arguing that PRO must appear in Spec-TP position to satisfy the binding properties of certain sentences.

(66)(a)[CP [C ∅][TP [T to][AUXP themselves [AUX be] [VP [V indicted] PRO]]]]
(66)(b)[CP [C ∅][TP PRO [T to][AUXP themselves [AUX be] [VP [V indicted] PRO]]]]
(Radford 2004: 311)

In the above data moving PRO to Spec-TP position is necessary for it to c-command themselves, which in turn is necessary to satisfy the binding principles and have PRO be coreferenced with themselves.

Cross-linguistic differences in PRO

Occurrences of PRO have been discussed and documented with regards to many languages. There are many languages which provide evidence of PRO consistent with the current theory, but the attempt to describe the behavior of PRO in some languages has given rise to significant issues with PRO as a result of conflicts with evolution in Government and Binding Theory.

English

(1) Johnj promised Mary [PROj to control himself]
(2) John convinced Billk [PROk to sleep]
    (Koopman, Sportiche and Stabler, 2014: 247 (30), 251 (47)) 

Icelandic

(1) Ólaf langar ekki til [að PRO vera ríkur].6 
    Olaf.A.M.SG longs not for to N be rich.N.M.SG 
    'Olaf doesn't want to be rich.' 
(2) Hana langar ekki til [að PRO vera kalt]. 
    her.A longs not for to D be cold.DFT 
    'She doesn't want to be (feeling) cold.' 
(3) þessi saga var skrifuð til [að PRO vera lesin/*lesið].
    this story.N.F.SG was written for to N be read.N.f.sg/*dft 
    'This story was written to be read.' 
    (Sigurðsson and Sigursson, 2008: 407 (4, 7))

An important fact of PRO in Icelandic, in contrast to English, is that it appears to be case-marked.

Romance languages

Spanish

(1) [Sin PROi saber por qué] A Juani le gusta María.
    Without to know why to Juan-dat 3s-dat likes María-nom
    'Without knowing why, Juan likes María.'
    (Montrul 1998: 32 (12))

PRO in adjunct clauses in Spanish are controlled by dative experiencers (which in the deep structure are the highest theta-role) because they fill the subject position. Dative experiencers (see Theta role) were also very common in Old English.

French

(2) Le parachutisme effraye Pierrei [avant même de PROi y avoir été initié].
    'Skydiving scares Peteri [even before PROi being initiated to it].'
    (Montrul 1998: 33 (13))

In French, as in Spanish, PRO can be controlled by dative experiencers in object position in an adjunct clause. The structure in sentences like that in (2) above can lead to an ambiguous interpretation; however, the adjunct clause in this sentence can never be controlled by the sentence subject, le parachutisme ('skydiving'), because it is inanimate. Due to pragmatics, this would be impossible.

Alternative Theories

A Movement Theory of Control

Norbert Hornstein has proposed that control verbs can be explained without resorting to PRO, and as such that PRO can be done away with entirely. This theory explains obligatory control with movement, and non-obligatory control with pro (little pro). This alternative theory of control was in part motivated by adherence to the minimalist program.

Working Assumptions

The Movement Theory of Control is predicated on the following principles.

(18) a. θ-roles are features on verbs.
     b. Greed is Englightened Self-Interest.
     c. A D/NP "receives" a θ-role by checking a θ-feature of a verbal/predicative phrase that it merges with.
     d. There is no upper bound on the number of θ-roles a chain can have.
     e. Sideward movement is permitted.
(Hornstein 1999: 78)

Principle d is of particular importance as it allows the traces of a single D/NP to acquire more than one θ-role allowing it to simultaneously satisfy the θ-criterion for multiple positions, eg. the subject of the non-finite embedded clause and the subject of the matrix verb. Hornstein argues that there is insufficient empirical evidence that a chain must be restricted to a single θ-role and that allowing multiple θ-roles per chain is the null hypothesis.

Obligatory Control As Movement

These principles allow control verbs to be explained by movement and what had previously been interpreted as PRO can instead be interpreted as a residue equivalent to an NP-trace.

(19) a. John hopes to leave.
     b. [IP John [VP John [hopes [IP John to [VP John leave]]]]]
(Hornstein 1999: 79)

This way the chain of Johns occupies the θ-roles of who is hoping and who is leaving, allowing the interpretation of obligatory control as movement and eliminating the need for PRO as a subject.

Non-Obligatory Control as pro

With the need of PRO eliminated under obligatory control, Hornstein argues that it follows naturally to eliminate PRO altogether from the theory as it behaves equivalently to little pro under non-obligatory control. In non-obligatory control an overt embedded subject may be introduced or omitted, and omitting the embedded subject may result in an arbitrary reading. Additionally, the overt subject may not be moved out of the embedded clause.

(43) a. It is believed that Bill’s/pro shaving is important.
     b. *Bill’s is believed that shaving is important.
(44) a. It is impossible for Bill/pro to win at roulette.
     b. *Bill is impossible to win at roulette.
(Hornstein 1999: 92)

Little pro may behave as a definite pronoun or indefinite (similar to the English one), which allows it to be distributed in the same way non-obligatory control PRO is observed to be distributed. This along with the observation that non-obligatory control arises when movement is not permitted places it in complementary distribution with obligatory control, as it is explained as movement. Since non-obligatory control occurs when movement is not permitted, it may be treated as an 'elsewhere' case with little pro being inserted as a last resort measure to rescue the derivation if an overt subject is missing.

Criticism

Since the publication of this movement theory of control some data has been discussed which it does not explain, challenging the completeness of the movement theory of control.

Imoaka argues that scrambling out of a split control clause is incompatible with the movement theory of control as constructed in Japanese by Takano and Fujii. Imoaka argues for a theory of Equi-NP Deletion to explain control and claims that such a theory is empirically superior as it successfully explains the problematic data as well as the data previously explained by the movement theory of control.

See also

• Control
• Pronoun
• Binding Theory
• Empty category