Syntactic Control Research Paper

These conventions will be followed throughout the article.

Mary convinced Billy to do her homework.                                                                                                  (1)

Mary_i convinced Billy_j [Ø_j to do her_i homework].                                                                                             (1a)

In (1) the noun phrase Billy controls the reference of the empty subject of the infinitival complement clause. Every native speaker of English will understand (1) as meaning that Billy—and not Mary herself—will do her homework. The challenging linguistic problem is to find a principled account for why in (1) the main clause object is the controller, why the main clause subject cannot have this function, and why the empty subject cannot just refer to a third person not mentioned in the sentence at all.

Sentences such as (1) exhibit the notion of obligatory control, i.e., a case in which the empty subject must be coreferential with a main clause constituent. Obligatory control contrasts with free (also called arbitrary or nonobligatory) control as in:

[Ø to wait in the crowded airport] would be a mistake.                                                                                     (2)

In the subject clause in (2), the reference of the understood subject is not determined by any other noun phrase in the sentence but depends on the discourse context.

1. Some Control Phenomena

In this research paper, the term ‘control’ will be understood as meaning ‘obligatory control.’ The following nonexhaustive list gives some sentences that exemplify obligatory control.

(a) Infinitival Purpose Clauses

Johni bought a book_j [Ø_i to amuse himself with]. (adapted from Nishigauchi 1984, p. 229)                                (3)

Mary_i bought the little boy_j a toy [Ø_j to amuse himself with]. (adapted from Nishigauchi 1984,

1. 229) (4)

(b) Infinitival Complement Clauses

John_i begged Mary_j [Ø_j to leave].                                                                                                                    (5)

Mary_i begged [Ø_i to leave].                                                                                         (6)

(c) Gerundial Complement Clauses

Paul_i thanked Mary_j for [Ø_j writing the report].                                                                                      (7)

Paul_i thanked Mary_j for [Ø_i having been admitted to the tennis club].      (8)

Most work on control has been done on nonfinite complement clauses, i.e., on clauses exemplified by sentences (5) through (8). The remainder of the article will therefore be concerned with control in nonfinite complement clauses.

2. A Short History Of Control

In the early days of generative grammar in the 1960s and 1970s, it was generally believed that control is a syntactic phenomenon, i.e., that the controller of the empty subject in the subordinate clause can be determined on a purely structural basis. More recently, it has become increasingly evident that semantic and pragmatic factors play an equally, if not more, important role in control.

2.1 Syntactic Approaches

The syntactic approach to control assumes that control relations can be determined on a purely structural basis. The most influential proposal along these lines was made by Peter Rosenbaum (1967, 1970). He postulated a Minimal Distance Principle (MDP), which predicts that the reference of the (empty) subject of a nonfinite complement clause depends on the closest available controller in the main clause. In sentence (1) above, the main clause noun phrase closest to the empty subject Ø is Billy. Indeed, this noun phrase controls the reference of the empty subject, and the MDP thus makes the right prediction.

There are, however, many sentences for which the MDP is falsified. Difficulties emerge with constructions that display more than one potential controller in the main clause, as in (9) and (10):

Mary_i persuaded Bill_j [Ø_j to clear her_i desk].                                                                                        (9)

Mary_i promised Bill_j [Ø_i to clear her_i desk].                                                                                          (10)

For sentence (9) the MDP correctly selects the closest noun phrase in the main clause, Bill, as controller. Persuade belongs to the class of control verbs that normally select the main clause object as the controller. However, for sentence (10) the MDP wrongly selects Bill as controller; in fact, the main clause subject Mary is the controller. Promise belongs to the class of control verbs that usually select their subject as controller. Subject control verbs of the promise class constitute a problem for the MDP and have to be marked as exceptional in the lexicon.

Another problem that the MDP faces derives from the fact that one and the same verb may have either subject or object control:

Bill_i thanked the coach_j for [Ø_j selecting him_i].                                                                                      (11)

Bill_i thanked the coach_j for [Ø_i having been admitted to the tennis club].            (12)

In (11), which can be regarded as the default case, the controller is the coach (object control), whereas in (12) a control switch or shift takes place: the most likely referent of the implicit subject Ø is Bill (subject control). The MDP wrongly predicts that in both sentences the coach should be the controller.

Other syntactic control theories have been proposed by Joan Bresnan (1982) and Emmon Bach (1979) in a logical framework. What these theories have in common is that they regard syntactic structure as the decisive factor for determining the control relationship. However, there is a substantial body of evidence that suggests strongly the importance of semantic and pragmatic factors in the determination of control relations.

2.2 Semantic Approaches

A comparison of sentences (9) and (10), which look alike structurally, invites the conclusion that their differing control properties are caused by the meanings of the verbs persuade and promise, respectively. To persuade somebody to do something means that the addressee of an act of persuasion is led to believe that he or she should do something. This prospective action is expressed in the infinitival complement clause. Since in an active sentence such as (9) the addressee is the grammatical object of the main clause, it is not surprising that this constituent functions as controller of the empty subject of the subordinate clause. In contrast, a promise involves a commitment of the speaker to a future course of action. In an active sentence such as (10), the ‘promiser’ is the grammatical subject and one would therefore expect the subject to be controller of the understood subject of the complement clause.

The most complete account that acknowledges the importance of the meaning of control verbs is probably Sag and Pollard’s (1991) control theory. These authors use semantic criteria to distinguish three classes of control verbs: (a) the order / permit class, which typically induces object control; (b) the promise class, which typically induces subject control; and (c) the want / expect class, which again induces subject control. The order/ permit and the promise classes conceptually imply actions, whereas verbs of the want expect class express what Sag and Pollard call an ‘orientation’ in their complement clause.

2.3 Pragmatic Approaches

Since the mid-1980s, the insight has grown that, in addition to the meaning of control verbs, the context in which a sentence is uttered and world knowledge about the participants’ status play a significant role in assigning the coreferential links between controller and controllee (e.g., Comrie 1984, 1985, Ruzicka 1999; see Pragmatics: Linguistic).

These additional pragmatic factors are illustrated by the following German sentences (adapted from Wegener 1989):

Der Anwalt_i versprach dem Haftling_j [Ø_j bald aus dem Gefangnis entlassen zu werden].             (13)

‘The lawyer promised the prisoner that he ( = the prisoner) would be released from jail soon.’

Der Haftling_i versprach seiner Frau_j [Ø_i bald aus dem Gefangnis entlassen zu werden].               (14)

‘The prisoner promised his wife that he ( = the prisoner) would be released from jail soon.’

The most plausible interpretation for both (13) and (14) is that the prisoner will be released from jail. These interpretations follow from the speech act verb versprechen, ‘promise,’ the status of the referents of the noun phrases der Anwalt, dem Haftling, and seiner Frau, and the meaning of the passive complement clause. A promise implies a promiser who is committed to a future action, and a promisee who benefits from the action of the promiser. This semantic–pragmatic scenario can be applied to the interpretation of sentences (13) and (14) and the most plausible controller can be inferred (cf. Panther and Kopcke 1993, Panther 1994).

In (13) the lawyer is the promiser; hence the lawyer is committed to some future action. However, the complement clause does not express an action explicitly. A step toward a reasonable interpretation is to assume that the complement clause expresses the result of an action performed by the promiser. The action itself has to be inferred. Furthermore, the result of the action is beneficial to the prisoner. Therefore, the most plausible controller is the noun phrase dem Haftling. On this account, the assignment of the controller is not a matter of finding the noun phrase that is syntactically closest to Ø in the matrix clause, but a matter of common-sense knowledge about the roles of lawyers and prisoners.

For (14), a coherent interpretation results if it is assumed that the prisoner can behave in such a way that he causes his release from jail. Again, the complement clause is interpreted as the result of an action—an action carried out by the prisoner to the effect that he will be released from prison. His wife can be regarded as the beneficiary of this action.

The interpretations sketched above are not the only ones possible. On the assumption that the lawyer is in jail, (13) could be interpreted as conveying that the lawyer will be released from jail; and (14) could have the interpretation, ‘The prisoner promised his wife that she would be released from jail soon’ given the background knowledge that the prisoner’s wife is in prison. It is in the nature of the pragmatic approach to control that different background assumptions have an impact on what is the most plausible controller of Ø.

Similar observations have been made for Chinese by Huang (1994, p. 153), who contrasts the following sentences (referential subscripts and bracketing have been added):

Xuesheng_i daying laoshi_j [Ø_i mingtian jiao zuowen].                  (15)

‘The pupil promises the teacher that he ( = the pupil) will hand in the essay tomorrow’

Laoshi_i daying xuesheng_j [Ø_j mingtian fang yi tian jia].              (16)

The teacher promises the pupil that he ( = the pupil) will have a day off tomorrow’

The control verb daying, ‘promise,’ can be used either in the ‘unmarked,’ i.e., normal, sense ‘commitment-to-action, as in (15), or in the ‘marked’ sense ‘commitment-to-permission,’ as in (16). These senses are derived pragmatically; they constitute ‘preferred’ interpretations that can be inferred on the basis of the meaning of the speech act verb daying and world knowledge about teacher–pupil interactions in the institutional framework of a school.

To summarize, the interpretation of control sentences relies not only on syntactic and semantic information but also, to a substantial degree, on world knowledge and contextual clues about the utterance situation.

3. Language-Specific Coding Differences Of Control Relations

One might surmise that the mechanisms of control are very similar across languages, but on closer inspection it turns out that there are language-specific differences in the way that control structures are coded. A selective list of these coding differences is given below (for a detailed account, see Ruzicka 1999).

3.1 Modal vs. Non-Modal Coding

For many speakers of English the following sentence is ambiguous:

Johnny_i asked his mother_j [Ø_i/j to go to the movies].                                                                                          (17)

Sentence (17) can mean either that Johnny wants his mother to go to the movies or that Johnny asks for permission to go to the movies himself. This latter interpretation seems to be the most likely one, assuming that Johnny’s mother has authority over Johnny (see Sect. 2.3).

As Ruzicka (1999, p. 45) points out, the ‘ask-for-permission’ (subject control) reading is also available alongside the ‘ask-for-action’ interpretation (object control) in Spanish, although the concept of permission is not coded explicitly.

Marıa_i le_j pidio a Juan_j [Ø_i hablar con los muchachos].                           (18)

‘Marıa asked Juan that she be allowed to talk to the boys’

The ‘ask-for-permission’ interpretation is not possible, however, if the same sentence is translated literally into German:

Hanschen_i bat seine Mutter_j [Ø_j ins Kino zu gehen].                                                                                          (19)

‘Johnny asked his mother to go to the movies’

Sentence (19) can only mean that Johnny’s mother is supposed to go to the movies. In order to express the ‘ask-for-permission’ sense, with Johnny as controller, German speakers have to resort to an utterance that codes permission by means of the modal auxiliary durfen:

Hanschen_i bat seine Mutter_j [Ø_i ins Kino gehen zu durfen].                    (20)

‘Johnny asked his mother to be allowed to go to the movies’

3.2 Causative vs. Non-Causative Coding

Another important coding difference between English and languages like German, Russian, and Czech is exemplified by

Peter_i persuaded his father_j [Ø_j to be examined by a doctor].                   (21)

Sentence (21) means that Peter’s father should undergo a medical examination. This sense is, how- ever, not explicitly coded in sentence (21), but is nevertheless conventionally conveyed by it. What is coded in the subordinate clause is the result of the action, not the action itself. If sentence (21) is translated literally into German, for many speakers ungrammaticality results and, for those who accept the sentence, an interpretation with Peter as controller is more plausible:

?Peter_i uberredete seinen Vater_j [Ø_i von einem Arzt untersucht zu werden].                    (22)

‘Peter persuaded his father that he ( = Peter) should be examined by a doctor’

The only way of rendering the intended message of (21) in German or Russian is to code his father as an active participant who engages in an action which results in a medical examination by a doctor:

Peter_i uberredete seinen Vater_j [Ø_j sich von einem Arzt untersuchen zu lassen].              (23)

‘Peter persuaded his father to have himself ( = his father) examined by a doctor’

Petr_i ugovoril otca_j [Ø_j podvergnut’ sebja osmotru vrac]. (Ruzicka 1999, p. 58)               (24)

‘Peter persuaded father that he ( = father) should undergo a medical examination’

Both German and Russian have to code the agent role of father explicitly. This is done by the causative verb lassen in German and the action verb podvergnut’ sebja ‘submit oneself ’ in Russian.

3.3 Implicit Control vs. Explicit Control

Another important difference between English and languages such as German and Russian concerns the ability for a conceptually present controller to remain unexpressed. In German it is possible to say:

Der Professor_i bat Ø_j [Ø_j den Zeitschriftenartikel zu kopieren].              (25)

‘The professor requested (of some person) that this person photocopy the journal article’

In (25) the controller of the understood subject remains unexpressed. The person or group of persons supposed to comply with the request is indeterminate. This phenomenon may be called implicit control (cf. Panther 1997).

However, a word-for-word translation of (25) would not have the same effect in English:

The professor_i asked Ø_j [Ø_i to photocopy the journal article]       (26)

As the referential subscripts indicate, the only possible interpretation of (26) is that the professor himself wants to photocopy the journal article and asks for permission to do so. In German, however, the translation of (26) requires modal coding with durfen ‘be allowed to’ (see Sect. 3.1).

Another example of implicit control is (27):

Der Arzt_i riet Ø_j [Ø_j den Fettanteil in meinem Essen zu verringern].                   (27)

‘The doctor advised (someone) that this person should reduce the amount of fat in my diet’

The literal translation of (27) in English leads to ungrammaticality (marked by an asterisk):

*The doctor_i advised Ø_j [Ø_j to reduce the amount of fat in my diet].                    (28)

These examples illustrate a general tendency in English: in obligatory infinitival control constructions the controller has to be made explicit, whereas in languages such as German and Russian it can remain unexpressed. However, sometimes there is a way to circumvent the principle in English by resorting to a gerund construction. A possible translation of (27) is:

The doctor_i advised Ø_j [Ø_j reducing the amount of fat in my diet].                      (29)

4. Conclusion

Human languages employ various means to keep track of referents on the sentence and the discourse level. Control relations can be seen in this larger framework of coreference phenomena in general. As with other coreference relations, it turns out that the interpretation of control structures involves a complex interplay of syntactic, semantic, and pragmatic parameters.

Bibliography:

1. Bach E 1979 Control in Montague Grammar. Linguistic Inquiry 10: 515–31
2. Bresnan J 1982 Control and complementation. Linguistic Inquiry 13: 343–434
3. Comrie B 1984 Subject and object control: Syntax, semantics, pragmatics. Berkeley Linguistics Society: Proceedings of the Tenth Annual Meeting. Berkeley Linguistics Society, Berkeley, CA
4. Comrie B 1985 Reflections on subject and object control. Journal of Semantics 4: 47–65
5. Huang Y 1994 The Syntax and Pragmatics of Anaphora: A Study with Special Reference to Chinese. Cambridge University Press, Cambridge, UK
6. Nishigauchi T 1984 Control and the thematic domain. Language 60: 215–50
7. Panther K-U 1994 Kontrollphanomene im Englischen und Deutschen aus semantisch-pragmatischer Perspektive. Narr, Tubingen
8. Panther K-U 1997 An account of implicit complement control in English and German. In: Verspoor M, Lee K D, Sweetser E (eds.) Lexical and Syntactical Constructions and the Construction of Meaning. Benjamins, Amsterdam
9. Panther K-U, Kopcke K-M 1993 A cognitive approach to obligatory control phenomena in English and German. Folia Linguistica 27: 57–105
10. Postal P M 1970 On coreferential complement subject deletion. Linguistic Inquiry 1: 439–500
11. Rosenbaum P S 1967 The Grammar of English Predicate Complement Constructions. MIT Press, Cambridge, MA
12. Rosenbaum P S 1970 A principle governing deletion in English sentential complementation. In: Jacobs R, Rosenbaum P S (eds.) Readings in English Transformational Grammar. Ginn, Waltham, MA
13. Ruzicka R 1999 Control in Grammar and Pragmatics. Benjamins, Amsterdam
14. Sag I A, Pollard C 1991 An integrated theory of complement control. Language 67: 63–113
15. Wegener H 1989 ‘Kontrolle’—semantisch gesehen: Zur Interpretation von Infinitivkomplementen im Deutschen. Deutsche Sprache 17: 206–28