Passive and Antipassive in a Functional Description of French Reflexive Verbs 0 . Functional Principles

0. Functional Principles Writing a grammar, or just doing grammatical description, presupposes some kind of guiding principle. The guiding principle of most traditional reference grammars of the standard European languages has been the classification offered by the traditional parts of speech or word classes: nouns, pronouns, adjectives, verbs, prepositions ... The architecture of the work seems almost to follow automatically from such a list. One guiding principle among several of our work on French which emerges as more and more forceful is the idea of cross-linguistic functional categories, i.e. the common functional content of structurally often quite diverse constructions of different languages. Now, for one thing, French does not have a passive in the same sense as Latin does, i.e. a morphologically anchored grammatical process. Yet, the resemblances between Latin and French passives are simply too persuasive to be ignored. And by looking for common functional content, in spite of structural differences resemblances are not at all rare either! one is often lead to classify the material in ways different from the tradition and to identify patterns not previously recognised as grammatical patterns, and sometimes even to solve traditional puzzles and paradoxes. This kind of “squinting grammar” to use Jespersen’s term can be quite fruitful and lead to genuine insights, hopefully not only in the grammarian’s mind. This I hope to demonstrate by the example of reflexive constructions in French, which group neatly as passives and antipassives. An analysis which would never have been possible without a good deal of squinting: what grammarian working only on the major European languages has ever heard of the antipassive? 75

The danger of the approach is of course that the straight-jacket of Latin grammar is replaced by an unprincipled promiscuity of all kinds of categories from the grammars of exotic languages.But this danger should be constrained by two principles.First, by the requirement of the existence of a functional category, i.e. the already mentioned common functional content of different grammatical structures.Secondly, by the awareness that the grammaticalisation of structural patterns occurs in different degrees in different languages: what is e.g. a genuine grammatical process (i.e. an obligatory choice) in one language, may only be an option found with a handful of lexemes in the next, even an option which exhibits only scant superficial internal unity.
Anyway, I think there is a good chance of making sense out of the idea of cross-linguistic functional categories.Such categories could for instance be the ways in which different languages proceed to get rid of Agent-and Patient-arguments under certain pragmatic or textual circumstances.These procedures are, however great the structural differences from language to language, good candidates for the status of cross-linguistic functional categories.And they are well-known in most languages, the second however mostly in exotic languages: they are called respectively PASSIVE and ANTIPASSIVE.

Passive
The category of voice is constituted by a systematic variation in the pairing of semantic rôles (SR) with grammatical relations (GR).The standard transitive construction looks like this: (1) SR: In such a two participant clause, one participant can be highlighted by marginalising or removing the other.By the marginalisation of a participant is meant the process by which it is so to speak pushed into the peripheral part of the clause by not being linked to -or by losing -one of the central GRs, subject or object.It is instead, if expressed at all, expressed as an oblique or a circumstantial phrase.
This is what typically happens in the passive: the Agent is demoted, i.e. removed or marginalised and realised as a circumstantial.Since the clause needs a subject, the Patient is consequently promoted and takes over the subject relation: (2) Passive: By the Patient's usurpation of the subject relation, there is no object relation left in the clause and the resulting structure is an intransitive construction.The passive is, as is well known, realised as a predicative construction in French (cf.Sørensen 1986), which has been fully grammaticalised in the sense that it occurs with almost all transitive verbs1 : (3) Julie a ouvert la fenêtre.'Julie opened the window' La fenêtre a été ouverte (par Julie).
'The window was opened (by Julie)' But the same, or at least a very similar, content can be expressed by a reflexive construction.One important difference is however, that the Agent is obligatorily suppressed in that case: (4) La fenêtre s'est ouverte Ø. 'The window (was) opened' This construction is semi-grammaticalised: many, but not all, transitive verbs allow it.

Antipassive
The less well-known antipassive voice -which until recently has only been recognised in ergative languages -is the mirror image of the passive, because what it does is to remove or marginalise the Patient, thereby concentrating the content of the clause on the Agent: (5) Antipassive: Most of the general content of the antipassive follows from this concentration upon the Agent by the demotion of the Patient.Instead of stating that the Agent performs an action directed towards a Patient, it is stated that the Agent is involved in an activity which is or isn't relative to a Patient, but this Patient is not integrated into the verbal situation in any central rôle: it has no central GR.It is the Agent and its willful and intentional activity, sometimes in vain, which is highlighted.
The different more specific shades of meaning often associated with the antipassive thus follow from the demotion of the Patient (cf.especially Bittner 1987, Cooreman 1994): the Patient is unimportant because it is non-specific or unidentifiable, or it is obvious, therefore often not mentioned at all in the antipassive; if expressed, it is seen as only partially affected by the Agent's activity, or not affected at all, the Agent's activity being an attempted or fruitless, even imaginary, activity; the action intended by the Agent is represented as not carried through to its conclusion, and so on.
The antipassive can be illustrated by the following example from West Greenlandic (Bittner 1987:195) The crucial differences between the transitive and the antipassive versions are the following: in a., the Agent is ergatively marked, in b. it is in the absolutive; in a. the Patient is in the absolutive, in b. in the instrumental.And most importantly, in a., the verb is inflected transitively and agrees with both participants, whereas in b. it is inflected intransitively (with a Ø antipassive affix, according to Bittner's analysis) and agrees only with one participant, the Agent in the absolutive.
One can make a case for the view that the alternation between transitive and prepositional constructions of verbs in Danish is a case of antipassive too, the more so since certain verbs in the prepositional (= antipassive) construction are inflected intransitively, i.e. with auxiliary vaere 'be' instead of the transitive have 'have ', cf. Durst-Andersen & Herslund (1996).And the Greenlandic examples translate neatly into Danish transitive and prepositional respectively: (6') a.
The difference between these two is clearly in accordance with the notional characterisation of the antipassive above.The alternation is not however, as is the case in West Greenlandic, fully grammaticalised, although widespread and apparently gaining ground.But this equation with the antipassive has given many important clues to the interpretation of the prepositional construction of Danish transitive verbs.Such alternations are lexically determined and not systematic grammatical choices, but they also exhibit the typical antipassive features: the marginalisation of the Patient and the loss of the object relation whereby the clause concentrates on the Agent.As mentioned above, the antipassive clause states that the Agent is involved in an activity which may be relative to a Patient, but this Patient is not integrated into the verbal situation with a central GR.It is the Agent's willful and intentional activity which is highlighted.This is also what traditional grammars say about pairs such as (7), to the extent that they say anything about them at all (but see Gougenheim 1970).

Antipassive and Reflexive
The crucial feature of the antipassive is thus the demotion of the Patient and hence the suppression of the object relation, the consequence of which is the creation of an intransitive clause.Another obvious way of getting rid of a Patient is to identify it with the Agent.Certain languages use accordingly their antipassive to express reflexivity (cf.Foley & Van Valin 1985:339 f., Cooreman 1994:83, Palmer 1994:185, Lazard 1994:239).And similar facts are found in Danish where the aforementioned prepositional (antipassive) construction is used when the verbs are used reflexively: ( And here we approach something which looks like a grammaticalised antipassive in French.

Passive reflexive:
(14) Les cuisses de grenouilles se mangent avec les doigts.'Frogs' legs are eaten with the fingers' Apart from the fact that different descriptions have partly diverging classifications and labels for the different types 2 , the classification con-ceals two major problems.The first problem is that such a classification tends to destroy the unity of reflexivity and to make believe that reflexive constructions can be made to mean almost anything; the very unity of the se-constructions vanishes, which is sought remedied by treating the entire field of se-constructions as an exponent of the middle voice (Stéfanini 1971) So the reflexive c. cannot simply be derived from the transitive d., which is however the fundamental tenure of the current analysis of ordinary reflexives, i.e. they are variants of transitive clauses where subject and object happen to be identical.One way of solving such problems is to start all over and look for functional categories which might yield a better understanding of the reflexive field.One thing that all reflexive clauses have in common is that they are intransitive, and as intransitive you can possibly get in French: they all inflect with the auxiliary être.Now, we saw above that "agent fantôme"' by Boons et al. (1976).Lamiroy (1993) calls 4. 'moyen', which seems to correspond more or less to the tradition, and 5. 'passif'.Melis (1990) calls 4. 'non agentif' and 5. 'médio-passif'.Desclés et al. (1986) call 4. 'medio-passive' and 5. 'passive'.4. is called 'ergatif' by Zribi-Hertz (1987), cf.Lagaé (1990), and Grimshaw (1982) calls 4. 'inchoative' and 5. 'middle'.Add to this that not all writers classify the same examples in the same categories.
the common denominator of passive and antipassive was their detransitivising effect.If we interpret the reflexive constructions as the meeting ground of passive and antipassive, we get another and, I submit, better and more coherent picture of this field.

Antipassive Reflexives
This is not the time nor the place to discuss all reflexive verbs and constructions, so I have to limit the discussion to some rather clear types which, like the example in (15) above, are problematic or intractable in the five-way classificatory system.The following cases seem to adapt rather well to an antipassive interpretation.

Ordinary reflexives
The ordinary reflexives which are always seen as unproblematic are in fact, as soon as you move away from the standard text-book examples of se laver 'wash' and se raser 'shave', quite varied and often quite difficult to reconcile with a standard reflexive meaning.They are for one thing often highly metonymic.And the standard paraphrases seldom or never work, cf. ( 16 It is obvious that standard paraphrases like ( 16) would be quite nonsensical in cases like these.The shades of meaning associated with socalled "ordinary" reflexives seem always to be derivable from the core meaning of the antipassive.Functionally, such reflexive clauses are antipassive and whatever metonymic meaning or shades of meaning are involved fall out from the basic meaning of the antipassive and the content of the involved lexemes.They are in fact instances of what Cooreman (1994:52) identifies as the use of the antipassive where the Patient is entirely predictable or obvious.

Psychological verbs
The description of the antipassive in 3. above revealed, but did not discuss, a distinction within antipassives which is crucial for the understanding of certain psychological or emotional verbs.By identifying the Patient with the Agent -which was seen as one way of getting rid of the Patient in the antipassive -what happens is really a kind of promotion of the Patient: the demotion of the Patient-rôle is performed by the promotion of the Patient-phrase to the Agent-rôle and thereby to the subject relation.Foley & Van Valin (1985) accordingly distinguish backgrounding and foregrounding antipassives.The case of psychological reflexive verbs in French is a case of foregrounding antipassive.Compare the following: (


The foregrounding antipassive (3.), shares some important properties with the passive and constitutes the transition between antipassive and passive reflexives.

Passive Reflexives
As is well known, French has a regular passive construction with the past participle of transitive verbs and the verb être, cf.(3) above.Beside this construction, there are two different reflexive constructions which are also passive-like -albeit variously labelled by different grammarians, cf.note 2 -in the sense that they conform to the over-all passive schema of (2): demotion of the Agent and the consequent promotion of the Patient to subject.With one qualification, however: not only is the Agent demoted, it is so thorougly demoted that it cannot be expressed in either of them.The promotion of the Patient is what is common to the passive and the foregrounding antipassive.

Passive
This construction can be described by the following schema: (26) Reflexive Passive: The Agent cannot be expressed, but its presence in the argument structure is attested by the possible presence of adverbs which can only be construed as qualifying an Agent (cf. the "agent fantôme" of Boons et al. 1976 The fundamental characteristic of this construction is its general or even generic, maxim-like, meaning.This meaning is well in accordance with the fact that the understood, but unexpressed and unexpressable Agent is maximally general.Sentences like these can in fact often be paraphrased by transitive sentences with the general human subject pronoun, on (cf.Lamiroy 1993:66).The generic aspect is explicitly brought out by a dislocation with cela: (28) Les vitres, ça se brise avec enthousiasme.'Windows that Refl break with enthusiasm' Les erreurs, ça se paie.'Errors that Refl pay' and by the fact that it mainly occurs in the present and the imperfect tenses.

Medio-passive
This type of reflexive construction is characterised by the schema (29), where a change takes place in the argument structure itself, where the Agent is totally removed: the verbal action is presented as happening by itself and not even adverbs which could be seen as qualifying an absent Agent are permitted.To the active transitive schema corresponds an intransitive schema with no Agent.To the causative meaning of the active corresponds an uncausative meaning of the reflexive.So rather than medio-passive, I shall call this type uncausative: The construction is lexically severely constrained in the sense that not all verbs allow it, only those whose combination with a proper Patient permit a reading like the one mentioned above, i.e. of an action or process which, so to speak, happens by itself:

Summary and Conclusion
The The French reflexive constructions of the above picture, where the different types of antipassive and passive reflexives are spread out, depict a scale of agentivity from the most Agent-promoting antipassive con-struction (i.e. the backgrounding with suppression of the Patient) to the uncausative where there is no trace left of the Agent.So they conform to the following over-all scale: The two major types of reflexives, antipassive and passive, meet in the cases where the object of the transitive construction is the subject of the reflexive.What the two patterns have in common besides this, is, functionally speaking, the fact that both highlight one of the two participants of the transitive construction, the Agent in the case of the antipassive, the Patient in the case of the passive.Both types create intransitive structures and this is signalled in the only way in which it is possible in French, namely by the intransitive inflexion of the verb by way of the auxiliary être.
The different reflexive constructions are however, as is well known, lexically constrained.But as shown by Kazenin (1994), there is often a clear connection between the semantic content of a verb and the kind of alternations it allows.So the more information on the Agent a verb conveys, the greater its tendency to allow or prefer Agent-preserving operations such as the antipassive.Conversely, the more specific information it conveys on the Patient, the more Patient-preserving operations it will permit, such as the passive or the uncausative.Within the French reflexives, the Agent-oriented verbs, i.e. verbs describing the Agent's manipulation of something (cf.Kazenin 1994:149), have the backgrounding antipassive constructions, whereas the verbs allowing the foregrounding antipassive, the passive or the uncausative typically are verbs describing changes of state of the Patient without conveying much information on a possible Agent's bringing about of such changes.
Certain communication verbs have reflexive uses which are not easily derived from their transitive use without setting up yet another type, viz. a "metonymic" reflexive: