Voice, impersonal construction, and zero lexeme: formalization of crucial notions

3.1 Voice

Definitions 1 and 2, as well as the calculus of voices2, are based on Mel’čuk (2006: 181–262); additional explanations and data concerning voices can be found there. Note, however, that the present exposition proposes refined formulations.

The notion of voice is underlied by the notion of diathesis.

The other names used in the literature for diathesis are alignment and linking.

Generally, L can have several diatheses, associated each with some of L’s different inflectional wordforms. One of these diatheses characterizes L’s lexicographic form: it is L’s basic diathesis, while all the others are non-basic.

Let this be emphasized: L’s non-basic diatheses retain the same SemAs and the same relationship between them as those present in the basic diathesis. In other words, all L’s diatheses have the same semantic valency. Increasing or decreasing the semantic valency of L – that is, adding SemAs (causatives and applicatives) or removing SemAs (decausatives) – results in producing a new, i.e. derived, lexeme Lʹ, different from L; these lexical-syntactic phenomena are excluded from consideration here. The same is true of changing the relationship between SemAs (which happens, for instance, with the reciprocal).

According to the above characterization of diathesis, a voice2 does not essentially affect L’s propositional meaning,^[2] that is, it does not modify its semantic valency. At the same time, it can change L’s meaning in a non-essential way – that is, by adding to it particular semantic components, such as, for instance, ‘negative impact on the Sem-actant Y’ (the adversative passive in a series of East-Asia languages).

Take a semantically bi-actantial transitive verb, e.g., [to] SHAVE, with a basic diathesis as follows:^[3]

John ‘X’, I, Subj is shaving Alan ‘Y’, II, DirО.

A lexical unit L with this basic diathesis can have exactly twelve different diatheses: the basic diathesis plus eleven non-basic ones. The latter are obtainable by the following three operations applied to the basic diathesis: permutation of DSyntAs (with respect to the SemAs), suppression of DSyntAs, and referential identification of SemAs.

1. Permutations (including zero permutation) produce four diatheses:

   

2. Suppressions add three diatheses:

   

3. Permutations together with suppressions create another two diatheses:

   

4. And, finally, referential identification of SemAs produces three diatheses more:

   

As a result, we have twelve logically possible voices2, that is, voice1 grammemes. They are illustrated below: with English expressions built on the sample sentence John is shaving Alan and its voice modifications, as well as with actual examples from languages that have the voice2 in question.

The names of the voices2 are constructed as follows:

1. Absence of manipulation (of the basic diathesis)  : active

2. Permutation of DSyntAs involving DSyntA I    : passive

   • involving only DSyntA I             : –”–, partial

   • involving both DSyntA I and II         : –”–, full

     • heightening the rank of DSyntA II         : –”–, promotional

     • lowering the rank of DSyntA I, but

       • without promoting DSyntA II          : –”–, demotional

3. Suppression of DSyntAs            : suppressive

   • suppressing both DSyntAs I and II        : –”–, absolute

4. Identification of SemAs             : reflexive

   • suppressing both DSyntAs I and II         : –”–, absolute

1. ‘active’ [act]:

zero modification of the basic diathesis (‘John is shaving Alan’)

2. ‘full promotional passive’: [F_PROM_PASS]

bilateral permutation of DSyntAs I and II (‘Alan is being shaved by John’)

3. ‘partial demotional passive’: [P_DEM_PASS]	demotion of DSyntA I to III, with DSyntA II retained in place (‘It is shaving Alan by John’)

A verb in the partial (= unilateral) demotional passive does not have a DSyntA I at all. In a language that requires the obligatory agreement of a Main Verb with the SSynt-subject, this verb form automatically receives – at the SSynt-level – a dummy subject, that is, a semantically empty pronoun, which can be overt, as Ger. ES in (7), or zero, as Ukr. Ø ( 3 , neu ) SG EMPTY in (8):

As will be shown, the verb form in the Lingala sentence (2e) is also the partial demotional passive.

An interesting case of the partial demotional passive is found in the Nilotic language Maasai (Mel’čuk 2006: 276–283; the acute symbol ′ stands for high tone):

The semantically empty (= dummy) zero SSynt-subject in (9) is a plural empty pronoun, while other similar dummies shown in this subsection are in the singular. The wordform yíóók ‘we’ – the DirO – is in the nominative case, since the Main Verb in Maasai features an ergative construction: with the SSynt-subject in the oblique case and the DirO, in the nominative. The wordform ɪltʊŋaná ‘people’ is in the oblique case, which is expected from an Agentive Complement.

Since the partial demotional passive verb form lacks DSyntA I, in a language where the SSynt-subject is necessary (because of the agreement of the Main Verb), this voice2 implies the impersonal construction, that is, the use of a dummy pronoun. Although formally correct, the name impersonal can be misleading, since semantically this construction is 100 % personal – in the sense that it is possible only with a human semantic Agent; see additional remarks in the next item.

4. ‘full demotional passive’: [F_DEM_PASS]

demotion of both DSyntAs I and II (‘Is shaving by John at Alan’)

For more details on this form, see Appendix, Comment [1], p. 19.

A verb in the full demotional passive also lacks the DSyntA I, like a verb in the partial demotional passive. In a language where a SSynt-subject is obligatory, a F_DEM_PASS form also receives a dummy subject at the SSynt-level – if, of course, it is not in an invariable form, which does not agree with anything, as is the case in Lithuanian, in examples (10) and (11).

The full demotional passive is more current with intransitive verbs. Thus, Lithuanian allows it for any intransitive verb whose SemA 1 = ‘X’ denotes a person:

The full demotional passive of intransitive verbs is also quite usual in Latin:

Full demotional passive is known as the impersonal passive, because this form cannot have a semantically full SSynt-subject: either it has an empty (= dummy) subject, or no subject at all.

In this connection, let it be emphasized that the term impersonal is not quite convenient: as shown by Frajzyngier (1982), in many languages – Indo-European, Turkish, Arabic – a full demotional passive form necessarily has a human (i.e., personal) Agent: indefinite people. However, for lack of a better term and in conformity with the longstanding tradition, the adjective impersonal can be used to characterize this passive, hoping that the present warning is sufficient. For more on the term impersonal, see Subsection 3.2.

The fact that the full demotional passive necessarily has a human Agent is covered by the semantic rule that supplies the corresponding grammeme to the verb in the DSynt–structure:

‘some people←1–P–2→Y’ ⇔ L(‘P’)_{(V)F_DEM_PASS}–II→L(‘Y’).

In other words, a verb can take the full demotional passive grammeme only if its Sem-actant 1 = ‘X’ is ‘people’.

5. ‘subjectless suppressive’: [SBJL_SUPPR]

suppression of DSyntA I (‘Be shaving Alan/him’)

In Polish the SBJL_SUPPR verb form has no other tense or mood – it is invariable.

In both (13) and (14), no SSynt-Subject, even a dummy one, is possible: the verb form does not show any number-personal agreement; neither is an AgCo expressible.

In Spanish, the situation is slightly different:

A Spanish SBJL_SUPPR form has a zero dummy SSynt-subject Ø ( 3 ) SG EMPTY , with which the Main Verb agrees.

The subjectless suppressive, just as the full demotional passive, implies an impersonal clause – by definition, since this verb form cannot have a semantically full SSynt-subject, expressing SemA 1 = ‘X’. However, all the cases of subjectless suppressives known to date presuppose a human Agent: thus, the Polish sentence Zruinowano most lit. ‘«They» destroyed the.bridge’ means that the bridge was destroyed by people, not by a flood or an avalanche (or else by animals). The case is the same as is with the full demotional passive; a similar semantic rule must be foreseen: ‘some people←1–P–2→Y’ ⇔ L(‘P’)_{(V)SUBJL_SUPPR}–II→L(‘Y’).

6. ‘objectless suppressive’: [OBJL_SUPPR]

suppression of DSyntA II (‘John is shaving [someone/people]’)

In the first sentence of this pair, the DirO is obligatory: if it is not overtly expressed, the verbal form means ‘deceives him/her/them’, where the elided DirO is clear from the context. In the second sentence, no DirO (= the expression of the Patient) is possible; the verb refers to a usual activity rather than a specific action.

7. ‘absolute suppressive’: [ABS_SUPPR]

suppression of both DSyntAs I and II (‘There is shaving’)

The suffix -kan expresses the subjectless suppressive (with the indefinite personal actor, as in Estonian and Polish): for instance, Kin+tuks+kán lit. ‘Me-hit-«they»’. = ‘«They» hit me’. Here the marker of the objectless suppressive -nin and that of the subjectless suppressive -kan combine to produce the absolute suppressive marker -ninkan. (The DirO of the verb mātaxī ‘charge’ in its basic diathesis designates the people who are charged the price, and the designation of the price is an oblique object.)

8. ‘agentless promotional passive’: [AGTL_PROM_PASS]	promotion of DSyntA II, with suppression of the former DSyntA I (‘Alan is being shaved’)

The expression of the Agent is impossible in traditional style.

9. ‘patientless demotional passive’: [PATL_DEM_PASS]	demotion of DSyntA I, with suppression of the former DSyntA II (‘There is shaving by John’)

No known example.

10. ‘objectless reflexive’: [OBJL_REFL]	referential identification of SemAs, with suppression of DSyntA II
	(‘John is shaving himself’)

11. ‘subjectless reflexive’: [SUBJL_REFL]	referential identification of SemAs, with suppression of DSyntA I (‘There is shaving himself by John’)

The reflexive form nusiskusta is invariable; it does not agree with anything, so there is no SSynt-Subject, not even a dummy. The marker of the subjectless reflexive is a combination of three affixes: the reflexive prefix si-, the past passive participle suffix -t and the suffix of invariable form (“neuter gender”) -a.

12. ‘absolute reflexive’: [ABS_REFL]	referential identification of the SemAs, with suppression of both the DSyntA I and II (‘Be.shaving.oneself’)

The suffix of the subjectless suppressive -no and the clitic of the objectless reflexive się are combined into the absolute reflexive marker.

The use of passives and suppressives is controlled by the communicative structure of the sentence: Thematization, Focalization, Backgrounding; but the reflexives are imposed strictly by semantics: by the identification of the Agent and the Patient (the Agent acts upon himself). Nevertheless, in innumerable cases, natural languages confound passives, suppressives and reflexives, using for them the same expressive means; suffice it to mention the notorious Indo-European middle voice. This fact largely justifies the inclusion of reflexives into voices2.

3.2 Impersonal construction

Two terms are current in linguistic publications: impersonal construction and impersonal verb;^[5] both need a definition.

Three families of impersonal constructions are logically possible:

1. The Main Verb in the given form does not allow an SSynt-subject at all; for instance, Estonian and Polish subjectless suppressives in (13) and (14).

2. The Main Verb in the clause under consideration has as its SSynt-subject a semantically empty lexeme (= a dummy); for instance:

   1. English meteorological verbs (such as [to] snow and [to] dawn);

   2. Spanish subjectless suppressives in (15);

   3. English sentences with a dummy it and a quasi-subjectival dependent, which is an infinitive phrase (It is necessary to act immediately ), or a subordinate clause (It pleases me that the members of the House support this important bill.).

3. The Main Verb in the clause under consideration has as its SSynt-subject a lexeme with the signified of the above special type: for instance, Rus. Ø ( 3 , neu ) SG ʻ elements ′ in (22b).

For instance, here are two impersonal verbs: [to] rain (It rained hard.) and Rus. tošnitʹ ‘nauseate’ (Eë tošnit lit. ‘It her nauseates’. = ‘She is nauseated’.).

The adjective impersonal should be used in any other combination with caution. Thus, the term *impersonal pronoun is infelicitous: as an example of a typical “impersonal pronoun” we often find Eng. one (as in One never knows .), Fr. on and Ger. man – while these pronouns denote only persons! The only pronoun that could be legitimately called impersonal is the empty it and its equivalents in other languages; however, it is more reasonable to call it what it is: an empty pronoun. On the other hand, the term impersonal passive can be accepted: it denotes a verb form that controls an impersonal construction. (But mind the warning concerning the term impersonal in Subsection 3.1, Item 3, p. 9.)

3.3 Zero lexemes

A zero lexeme, as any overt lexeme, can be semantically full or semantically empty.

All zero lexemes I know of are nominal pronouns that can be SSynt-subjects only; all of them impose agreement on the Main Verb. As a stock example, consider two zero lexemes in Russian (Mel’čuk 1974, 1988: 303–337, 2006: 469–516):

1. Ø ( 3 ) PL ʻ people ′ ‘indefinite people’, glossed «they»;

2. Ø ( 3 , neu ) SG ʻ elements ′ ‘natural forces ≈ elements’, glossed «elements» («stixii», in Russian).

In case the fence was toppled by something that the Speaker chooses not to indicate, he can say Zabor byl povalen ‘The fence was toppled’ [in any possible way], using the passive without the Agent Complement (agentless promotional passive, No. 8 in Subsection 3.1, p. 13); if the fence fell on its own, the sentence Zabor povalilsja lit. ‘The fence toppled itself’ is used – with a different verb povalitʹsja ‘fall on its own’.

For more on the lexeme Ø ( 3 , neu ) SG ʻ elements ′ , see Appendix, Comment 2.

In addition, Russian also has a semantically empty zero lexeme, quite similar in its use to the impersonal Eng. it: Ø ( 3 , neu ) SG EMPTY   .

Russian is, of course, not the only language to feature zero lexemes in the role of SSynt-subjects. Thus, one finds zero lexemes in Bantu languages – for instance, in Kinyarwanda (Kimenyi 1980: 184–190; c = /č/, j = /ž/; the acute symbol ′ stands for the high tone).

Kinyarwanda has three semantically full zero pronouns: Ø ( II ) ʻ people ′ ‘indefinite people’, Ø ( XIV ) ʻ time ′ ‘time of the day’ and Ø ( XVI ) ʻ weather ′ ‘weather’. All these lexemes have their semantic sources in the starting semantic structure.

1. The indefinite-personal pronoun Ø ( II ) ʻ people ′ , of noun class II (ba-): human plurality; semantically, it is more or less equivalent to Fr. on and Ger. man, as well as to the zero pronoun Ø ( 3 ) PL ʻ people ′ in Russian (and other Slavic languages); it is glossed as «they».

1. The pronoun Ø ( XIV ) ʻ time ′ , of noun class XIV (bu-):

1. The pronoun Ø ( XVI ) ʻ weather ′ , of noun class XVI (ha-):

Beside these semantically full zero lexemes, Kinyarwanda has two semantically empty zero lexemes: Ø ( VIII ) EMPTY   and Ø ( XVI ) EMPTY   . Note a contrast with dummy subjects in English, French, German or Russian: a dummy pronoun Eng. it, Fr. il, Ger. es or Russian Ø ( 3 , neu ) SG EMPTY   controls one and the same agreeing form of the Main Verb, so that each of these languages has only one dummy pronoun. Unlike these languages, Kinyarwanda uses two different zero dummy pronouns, which belong to different noun classes and impose different agreements on the Main Verb.

This characterization of zero lexemes in Kinyarwanda is intended to buttress the proposal for a zero lexeme in Lingala, another Bantu language.