"impersonal" passives with agents ("demi-digest")

Björn Wiemer Bjoern.Wiemer at UNI-KONSTANZ.DE
Thu Aug 30 15:56:29 UTC 2001

Dear members of the list,
about three weeks ago I sent a letter requesting for information on the 
evolution of passive morphology and passive constructions. I was astonished 
to get so many detailed replies. The reason why I am sending this digest of 
what I have gotten is, inter alia, that I had to digest the replies myself. 
Before presenting a summary I want to thank those who have supplied me with 
information on Celtic, Old Indic, Indo-Aryan, Fennic, Germanic and some 
other languages. Not being a specialist in ancient Ide. languages, I must 
immediately comment that the replies concerning especially them are 
bewildering insofar as the data meant to sustain the chronologic primacy of 
"impersonal" vs. "personal" passives seem to me to a certain extent 
contradictory (see below). This is but a tiny consolation for someone like 
me who has been dealing with Slavic and Baltic, where documented language 
history is not "deep" enough to give empirical evidence for answers to the 
questions I sent 3 weeks ago.
         I will repeat my letter from August, 9th, and give summaries in 

>Date: Thu, 09 Aug 2001 13:38:52 +0200
>From: Björn Wiemer <pop05540 at popserver.uni-konstanz.de>
>Subject: "impersonal" passives with agents
>Dear ALT'ers,
>I would be interested to know if anybody of you knows anything more 
>concrete on the following specific topic: by an "impersonal" passive we 
>traditionally understand a construction in which the verb (or some form 
>derived from it, e.g. a participle) bears morphology associated to the 
>passive, but doesn't have an NP which could be characterized as the 
>subject of the respective clause. More specifically, there is no NP in the 
>nominative case (for NOM-ACC languages) which would trigger agreement in 
>the finite verb or the verbal noun (e.g. gender, number for the participle).
         A terminological remark sent to me by Matt Shibatani: "impersonal 
passive" should be defined as a verbal construction with passive morphology 
that does not have a REFERENTIAL subject. In his opinion, dummy subjects 
(as in Germanic languages and French) should be considered as grammatical 
subjects insofar as they agree in number (and gender, if there is any) with 
the predicate (finite verb). "Impersonal passive" comprises thus (a) 
passives with dummy subjects, (b) passives without any syntactically overt 
subject-NP (or pronoun).

>  As far as I know, it has remained hitherto unclear whether so-called 
> "impersonal" passives arise earlier than "personal" passives (i.e. 
> passives with an agreeing NP). Although the diachrony of passives, for 
> instance, in Germanic languages seems to indicate that passives arose 
> first from transitive verbs (or, more generally, from verbs with more 
> than one argument, with the highest-ranking argument being demoted), 
> before passive morphology started being applied to intransitive verbs 
> (more precisely: verbs with only one argument). For this reason, we might 
> assume that "impersonal" passives -- the natural consequence of the 
> demotion of the single argument of an intransitive verb -- are 
> diachronically secondary.
>         There remain, however, three principal questions not "covered" by 
> this kind of data:
>1. In order for a lower-ranking argument ("patient" or the like) to become 
>promoted to a syntactic subject (pivot, controller etc.) there first must 
>occur demotion of the highest-ranking argument ("agent" or the like). 
>Thus, logically promotion presupposes demotion, and there are plenties of 
>examples in different languages (known at least since Comrie 1977 on 
>"spontaneous demotion") in which only demotion of the highest-ranking 
>argument occurs, without promotion of a lower-ranking one. This looks like 
>a "passive half-way". The point is, however, how these purely demotional 
>constructions appeared, i.e. in which diachronic relationship do they 
>stand in comparison to "promotional passives" (the "normal" ones)? Can 
>anybody give me more concrete and reliable information on this matter?
         According to Boris Zakharyin, Sanskrit conforms to the picture of 
Germanic (see above). He writes: "In Early Sanskrit passive formations were 
very much limited (statistically and grammatically also - as they could be 
formed on the basis of the so called 'present stems' - that is, they were 
used - though rarely - in Presence and Imperfektum only), later their use 
has become widened and simultaneously passives from intransitives (mostly 
from change of position or state verbs) started being used."
         Matt Shibatani suggested that Celtic and Germanic would be primary 
sources to establish the chronology of impersonal and personal passives. 
Germanic data seem to confirm rather the order personal > impersonal (see 
above). As for Celtic, I must admit that I have been hitherto unable to put 
together all the details which are discussed by specialists in relevant 
publications. For the time being I will just give a list of what colleagues 
have hinted at:
1. a downloadable paper by Michael Noonan: 'Subjectless clauses in 
Irish'  (http://www.uwm.edu/~noonan).
2. Stenson, N. 1981: Studies in Irish syntax. Tuebingen (only synchronic)
3. Thurneyssen, R. 1980: A Grammar of Old Irish. Dublin. (a kind of 
standard book, but not with the relevant information I have been searching for)

I also give a part of a letter from Elisa Roma: "Old Irish had an 
impersonal passive which agreed with the object (not
morphologically promoted) only in number, but not in person. 1st and 2nd 
persons were expressed through infixed pronouns, i.e. accusative pronouns. 
Agents could (but need not) be expressed through prepositional phrases, 
and  the impersonal was normally used with intransitives. In Modern Irish 
the situation is similar, but slightly different: the impersonal 
(historically older) passive doesn't show any number agreement with the 
object, which is still not morphologically promoted (though this can be 
seen only with (some) pronouns, since the nominative-accusative case 
distinction is lost for nouns). But, more interestingly, agents cannot be
expressed with this old impersonal passive, and a new periphrastic passive 
has been coined (so-called "substantive" verb + past participle), with 
which an agent can be overtly expressed through a prepositional phrase."
         Nolan Brian <Brian.Nolan at itb.ie>, who does not believe that 
impersonal passives are derived from personal ones, ramifies his view by 
Irish data. He has sent me an interesting, yet unpublished paper. Maybe, he 
will be willing to send it to other interested persons, if you write to him 

>2. Passive morphology applied to transitive verbs -- or, more precisely, 
>to verbs with an ACC-object in the active -- may also render constructions 
>without a nominatival subject, i.e. "purely demotional passives", with the 
>object keeping its ACC marking. (This, of course, holds only for languages 
>which have rich enough case morphology.) Such a kind of "demi-passive" (it 
>might be argued that they are active, since the ACC-object is retained) is 
>considered to be a rather rare phenomenon world-wide (cf., for instance, 
>Shibatani 1998). And one might argue that it is a late development in the 
>evolution of passives.
>         For instance, the Polish construction with the petrified 
> -no/to-participle (former neuter sg. of the nominal declension) and a 
> possible ACC-object (e.g., Dano mu.DAT.SG.M ksiazke.ACC.SG.F lit. '(It) 
> was given him a book') is a rather late development, not known to Common 
> Slavic. The same holds for Polish "reflexive impersonals" with an 
> ACC-object (Gosciom.DAT.PL.M pokazuje.3.SG.PRS sie.RM wille.ACC.SG.F lit. 
> '(It) is shown to the guests the cottage' = The cottage is shown to the 
> guests), which is an even later development (most probably by analogy to 
> -no/to).
>         Does anyone know of investigations which can show how such 
> "passives" with retained (or regained ?) active government have arisen 
> and in which diachronic relation do they stand to "real" passives, both 
> from transitive and intransitive verbs?
         'Demi-passives' appeared much later than both personal and 
impersonal passives: in Indic they are attested only from the 13-14th c. 
(B. Zakharyin), when they "undoubtedly copied Middle or even New Indo-Aryan 
constructions" (L. Kulikov). Language contact seems to have triggered at 
lot here.

>3. I have recently come across an author who gives some Sanskrit examples 
>of impersonal passives (from one-place verbs) with an agent phrase in the 
>instrumental; cf.
>(1)     gamyate maya    lit. '(it) is going by me' / '(there) is (some) 
>going by me'
>(2)     supyate tvaya   lit. '(it) has been slept by you'
>(3)     gatam anena     lit. '(it) has been gone by him'.
>Unfortunately, there are no references (neither to sources, nor to 
>linguists), so that I am not sure how such examples should really be 
>qualified. Maybe, someone of you is able to give me some information on 
>such kind of impersonal passive with agent phrases? And, again, do we know 
>anything about their diachronic relationship with respect to (a) agreeing 
>(promotional) passives and (b) impersonal passives with an ACC-object?
         Agents with passives of intransitive verbs (i.e. pure demotional 
passives) seem to be widespread. They are attested in Latin (acriter 
pugnatum est ab utrisque "it was fought by both parties"), in Sanskrit 
"this construction became increasingly popular; in particular for motion 
verbs in the past one almost only finds impersonal forms with instrumental 
agents" (S. Luraghi). In Indic it gave rise to an ergative pattern (this 
pattern might, however, have been influenced by Indo-Aryan; cf. L. 
Kulikov's information on the source of Indic/Hindi demi-passives). 
Interestingly enough, however, the Polish construction cited above has not 
developped into an ergative pattern, nor has Lithuanian, where we have 
impersonal passives from virtually any verb, for instance:
(4)     Musu            cia     seniai          gyven-t-a
         we.GEN.PL       here    long.ago        live.PPP.NEUT
lit.    ,by us it was lived here a long time ago'
Shibatani (1998) rightly noticed that this "passive" in fact has begun to 
serve another function, namely: inferential evidentiality. Thus, It would 
be interesting to know more about passives becoming evidentials (of some 
sort). Does anybody know more on this topic?
         Some people (e.g., Schmalstieg) have tried to argue that such a 
pattern as in (4) testifies an "ergative past" of (East) Baltic, but the 
evidence is very tiny, if not speculative at all. One could also object 
that the marking pattern in (4) does not fit to an ergative one, because 
the single argument MUSU is in an oblique case and, thus, behaves 
differently than normal "subjects" of intransitive verbs, which are in the 
nominative (and trigger agreement with the finite verb and participles). To 
the contrary, it behaves like the Agent of a transitive verb; cf.:
(5)     Musu    buvo            atliktas 
sunkus                  darbas
                 COP.PRT perform.PPP.M.SG.NOM    hard.M.SG.NOM   work.M.SG.NOM
lit.    ,By us (some) hard work was done'.
--> the patient (= lower-ranking argument) agrees with the predicate.
After all, a purported development from an ergative system to the passive 
would run counter to an implicative hierarchy according to which passive 
may develop into ergatives, but no cases of an evolution in the opposite 
direction have been attested (Haspelmath 1990 and #1142 in the Universals 
Archive at Konstanz University).
         As for modern Romance, constructions with the reflexive marker like
(6)     Qui si mangia bene      lit. 'Here (it) is eaten well'
are not at all genetically related to Latin syntax. Kemmer (1993:178f.) 
writes that they appeared in the 13-16 c. and that only after them personal 
passives appeared. Furthermore, the impersonal construction with ACC object 
(= 'demi-passive')
(7a)    Qui e’ si.rm legge.sg troppi libri.pl
lit.    ,Here (it) is RM read.SG many books.PL'
was diachronically anterior to the personal
(7b)    Qui si.rm leggono.pl troppi libri.pl
lit.    ,Here RM read.PL many books.PL' = ,Many books are read here'.
If this is correct, the direction of development was exactly opposed to the 
"reflexive" passive in Polish (and other Slavic languages) in comparison to 
the 'demi-passive' marked by the RM. The same holds for the analytic 
passive compared to the -no/to-construction.
         Unfortunately, I have not found anywhere information about the 
agent: could it be added in a construction like (7a)? As for modern 
Italian, Silvia Luraghi explained to me that an agent phrase can be added 
(e.g., Si combatte da entrambe le parti aspramente ;It was fought 
vigorously by both sides', compare the Latin ex. above). This agent phrase, 
however, does usually not coincide with the normal "da", but with a PP ("da 
parte di") which seems still to be on its way into the system of secondary 
         I couldn't figure out whether this is (or has ever been) possible 
also with transitive verbs leaving the object where it would be in the 
active. In Polish this is now impossible, but in Ukrainian (which calqued 
it from Polish and shows a more archaic stage) one can (occasionally) add 
an agent in the instrumental; e.g.,
(8)     Grjadku         cybuli          bulo            zryto 
         (garden)bed.F.ACC.SG    onion.F.GEN.SG COP.PRT  dig_up.PPP.NEUT 
lit.    ,(it) was digged_up the onion bed by a mole'
i.e.    ,The onion bed was digged up by a mole'.

This is all I can present at the moment in a "half-digest" (or 
"demi-digested") way. If anybody wants some more references or details on 
Slavic and Baltic, I am ready to answer as far as I will be able. With best 
wishes and regards,
Bjoern Wiemer.

Dr. Bjoern Wiemer
Universitaet Konstanz
FB Sprachwissenschaft / Slavistik
Postfach 55 60, D 179
D- 78457 Konstanz

tel.: ++49 / 7531 / 88 -2582
fax: ++49 / 7531 / 88 -4007
e-mail: Bjoern.Wiemer at uni-konstanz.de
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.linguistlist.org/pipermail/lingtyp/attachments/20010830/3e7ee43d/attachment.htm>

More information about the Lingtyp mailing list