Christo Moskovsky & Alan Libert 101

Journal of Universal Language 5

September 2004, 101-149

Possibilities for Passives in Natural and

Artificial Languages
Christo Moskovsky & Alan Libert
University of Newcastle

Abstract
The Passive Voice is a category which we find in the large majority
of natural languages, and also in most artificial languages. The first
major part of this paper offers a survey of passive constructions in a
number of typologically distinct natural languages, with the basic
aim of presenting the (prospective) artificial language constructor
with the range of functional and formal properties of the Passive
Voice which occur in natural languages. This survey shows that
regardless of the fair amount of variation in the morphosyntactic
form of passives that we find in different natural languages,
crosslinguistically passives are remarkably uniform in inevitably
occurring as a grammatical category marked (synthetically or
analytically) on the verb; they are also remarkably uniform in relation
to the basic function they perform: passivization inevitably involves
demotion of a primary clausal term (the Subject) and in most of the
cases also promotion of a non-primary term. The next part of the
paper offers an overview of Passive Voice formation and function in
artificial languages, which will provide the language constructor with
a good idea of some of the ‘design’ decisions taken with regard to this
grammatical category. Finally, the paper briefly discusses various
design issues in relation to economy, explicitness/ambiguity,
102 Possibilities for Passives in Natural and Artificial Languages

functionality, and learnability and presents some specific

recommendations with regard to the possible design of passives in an
artificial language.

Keywords: passive, voice, verb, morphology, syntax, artificial

language

1. Introduction

Constructors of artificial languages (ALs) face a number of

interesting and potentially challenging design problems. 1 However,
before specific problems (relating, e.g., to structural, functional,
meaningful, etc. aspects of the language under construction) are dealt
with, the AL constructor must address one very fundamental
question: should the language he is involved in designing be
modelled on natural language? We believe that there are at least two
valid reasons why it should: one is conceptual-philosophical, the
second is cognitive. In relation to the former, natural language is,
universally, the only comprehensive system for communication

1
We use the following abbreviations in glosses of examples:
ABL – ablative DO - direct object PART – participle
ABS – absolutive ERG – ergative PASS – passive
ACC – accusative GEN – genitive PL - plural
ACT – actor IMPERS - impersonal PREP - preposition
AL – artificial language INF – infinitive PRES - present
ANTIPASS – antipassive INSTR – instrumental PST - past
AOR – aorist IO - indirect object PSTPART – past participle
ART – article IMPERF-imperfect PV – passive voice
ASP – aspect NOM – nominative REFL – reflexive pronoun or clitic
AUX – auxiliary verb OPrep - object of preposition SG - singular
CIRC – circumstantial PART - participle SU – subject
CL – clitic PERF – perfect TOP – topic
DAT - dative
Also, “OT” indicates that a translation is by us, rather than being made by the
source. We have sometimes not indicated the internal structure of words where this
was not relevant. Sometimes we have left out a translation when the word for word
gloss was sufficient for understanding the meaning of the example.
 Christo Moskovsky & Alan Libert 103

among people: wherever there are people, they use natural language
as their principal (and, most of the time, only) means to communicate
with each other, and while sometimes there may be, in addition to
language, some other non-linguistic symbolic communication system,
its communicative potential and the range of conceptual domains in
which it is used would be extremely limited. Given the universality of
natural language, one almost inevitable conclusion would be that it
must be well suited to perform the function of human communication,
alongside a range of other important functions,2 and that some of its
basic properties are likely to be essential to those functions. The issue
of learnability should also be of prime concern to the AL constructor:
in view of the fact that in most normal circumstances ALs would
acquired as a second language, then learners could be expected to
cope in a more efficient manner with the learning of a new
communication system which bears at least some of the basic
properties of a communication system they are already familiar with.3
In a series of papers (see Libert & Moskovsky 2002, 2003), we are
considering some of the most common grammatical categories of
natural language and how they are represented in a range of natural
and artificial languages. 4 One such grammatical category is the
Passive Voice (PV): authors involved in typological studies of natural
language (e.g., Keenan 1985, among many others) acknowledge that
a distinct (morphological) form associated with (but, often, not only)
2
There is a growing recognition of the key role that language plays in relation to
thinking, especially in those (usually conscious) thinking processes which
Vygotsky has described as “higher mental functions” (see, e.g., Williams 1989).
Some (e.g., Gleitman & Liberman 1995) have also emphasized the importance of
language in relation to the attainment and storage of knowledge (e.g., as a medium
of representation for knowledge in the human mind).
3
It has now been established that (adult) second language learners find it much
harder to become proficient in second languages which are typologically distant
from their native language (Ringbom 1987). This would suggest that it would be
even harder to attain competence in a communication system which is radically
different from natural language.
4
We exclude from the set of artificial languages fictional languages, such as Klingon
104 Possibilities for Passives in Natural and Artificial Languages

a passive meaning can be found in the vast majority of human

languages. Such facts strongly suggest that the PV is an indispensable
property of human language.

2. Natural Languages

In the most typical case, sentences in PV are analysed as derived

from corresponding sentences in the Active Voice, and this derivation
minimally involves applying some specialised verbal morphology
(often unique to PV) and, frequently, also change in the order of
sentence constituents:

(1) { NP1 – V – NP2 }ACTIVE → { NP2 – V-PASS – (NP1)}PASSIVE

A suitable example from English to illustrate this is provided in (2):

(2) { [NP1 The Council] – will consider – [NP2 your proposal] }ACTIVE
→ { [NP2 Your proposal] – will BE considerED – (by [NP1 the
Council])}PASSIVE

PV formation can involve the use of an auxiliary verb (AUX)

together with a specific form of the main verb (most commonly, a
participle), as in the example above, or only a specific (passive)
verbal morpheme, as in the Kinyarwanda example below:5

(3) a. Abagóre ba-a-boon-ye mweébwe.

Women they-PST-see-ASP you-PL
‘The women saw you.’

5
In Keenan’s (1985) terms, “periphrastic” and “strict morphological” passives,
respectively.
 Christo Moskovsky & Alan Libert 105

b. Mweébwe mw-aa-boon-y-w-e n’âbagóre.

you you-PST-see-ASP-PASS-ASP by-women
‘You were seen by the women.’ (Kimenyi 1988:363)

In other words, passivization involves derivation of a

(morphologically and/or syntactically) marked construction (M) from
a (morphologically and/or syntactically) unmarked or neutral
construction (Ø) (Comrie 1988):6

(4) ØACTIVE → MPASSIVE

Semantically, the basic difference between the active and the

passive member of the pairs of sentences above is that the semantic
roles associated with the arguments of the verb are differentially
aligned with syntactic functions (Huddleston & Pullum 2002:1427):
in the passive sentences, the subjects (SUs) are associated with the
semantic role of the direct objects (DOs) of the corresponding active
sentence, while SUs of the active sentences take up an oblique
syntactic function in the passive equivalent (or are eliminated
altogether).
It is widely agreed that one of the most typical discourse functions
of passivization is change of the thematic focus in the sentence: from
the entity denoted by SU NP in the active sentence to the entity
denoted by Object NP (and, perhaps, in part to the action denoted by
the verb). This change of thematic focus is commonly accomplished
by way of promoting one of the arguments of the verb and demoting,
or even completely eliminating, another argument (Keenan 1985,
Palmer 1994).
Change of thematic focus and/or promotion of an argument is by
far not specific to passive constructions (as in (5b)): we also find it in

6
But note our example (14) from Tagalog which probably represents a type of
language in which Active and Passive Voice are equally marked.
106 Possibilities for Passives in Natural and Artificial Languages

other derived constructions such as (5c) and (5d) which literature

commonly defines as cases of “topicalization” and “left-dislocation”,
respectively:

(5) a. The Council will consider your proposal next week.

(active)
b. Your proposal will be considered by the Council next week.
(passive)
c. Your proposal the Council will consider ___ next week.
(topicalization)
d. As for your proposal, the Council will consider it next week.
(left dislocation)

The most obvious formal difference between (5b) and (5c,d) is

that (5b) involves a morphological marking on the verb which is
specific to PV, while sentences such as (5c,d) don’t; in other words,
the verb in constructions like (5c,d) remains unchanged, i.e. in its
basic unmarked active form. Another significant difference can be
observed in relation to the grammatical function of the promoted
constituent: passivization involves a change of grammatical function,
while topicalization does not. There are a range of other important
differences. As Keenan (1985:244, among others) has noted, in cases
of topicalization the promoted constituent occupies a sentence
position which is “unusual” and is specifically created (e.g., by way
of adjunction) for that purpose, while in passivization the promoted
constituent takes up one of the basic sentence positions, that of the
Subject. Significantly, the promoted constituent then assumes all the
properties (e.g., morphological, syntactic, discourse, etc.) typically
associated with subjects.7 One of the consequences of this is that a

7
E.g., the ability to operate as a syntactic and/or pragmatic pivot, to operate as an
antecedent in an anaphoric relationship, etc. The reader should bear in mind that the
SU NP is, syntactically, in many ways unique: it is the only structural node in the
 Christo Moskovsky & Alan Libert 107

non-SU constituent in the active sentence becomes a SU in the

passive equivalent, and it acquires the higher thematic status that
sentence subjects have by virtue of their position. However, this
thematic status is in no way exceptional and is definitely less
prominent than the thematic status of a topicalized constituent: in this
sense, thematically the NP your proposal is much more of a topic in
topicalized constructions like (5c) than in PV constructions like (5b).
Keenan also suggests that PV must be a much more fundamental
grammatical category than topicalization or left-dislocation, because
while it is possible to conduct all major syntactic operations, such as
nominalization, question formation, relative clause formation, etc.,
on PV constructions, there are severe restrictions on the types of
syntactic operations that can be conducted on topicalized or
left-dislocated constructions. Furthermore, as Palmer (1995) has
pointed out, in passivization the promoted constituent is most
commonly an argument NP 8 , while topicalization can involve
practically any syntactic constituent.
Another issue that needs to be addressed with regard to the scope
of the grammatical category of PV is the distinction between “true”
PV and “stative” PV9. This distinction is most obvious in languages
which have two different morphological forms for the two types of
PV, as is the case in German:

(6) a. Der Tisch wird gedeckt. (‘true passive’10)

the table becomes covered
‘The table is being laid.’

clause which c-commands all other nodes in the same clause, but is not
c-commanded by another node.
8
This could be argued to hold even for sentences such as (i), in which the preposition
in is reanalyzed with the verb.:
(i) This bed has not been slept in.
9
The terms verbal and adjectival passive are also commonly used in literature.
10
The respective German terms are Vorgangspassive and Zustandspassive.
108 Possibilities for Passives in Natural and Artificial Languages

b. Der Tisch ist gedeckt. (‘stative passive’10)

the table is covered
‘The table is laid.’

Nedjalkov & Jaxontov (1988:47) discuss another interesting

example (originally in Jespersen 1924:274) which is particularly
illustrative of the difference between ‘stative’ and ‘true’ PV:

(7) Als ich um fünf kam, war die Tür geschlossen,

when I at five came was the door locked
aber ich weiß nicht, wann sie geschlossen wurde.
but I know not when it locked became
‘When I came (home) at five, the door was locked but I don’t
know when it was locked.’

As these examples show, in German different auxiliary verbs are

used with the two types of PV: werden ‘become’ and sein ‘be’.
German reference grammars describe the grammatical construction
in (6a) as a passive voice expressing an action, and the grammatical
construction in (6b) as a passive voice expressing a state. In
languages like English which generally use one AUX for the
formation of PV (and which seems to be the more common case
crosslinguistically), certain sentences can potentially yield both
‘stative’ and ‘true’ PV interpretation:

(8) The house was deserted.

This sentence can be interpreted as a statement about a specific

property (“deserted”) that a house has, which would be the ‘stative’
interpretation; alternatively, it can be interpreted as a statement about
the occurrence of an event involving the house (i.e., that its occupants
left it), which would be the ‘true passive’ interpretation. It seems that
the two different interpretations are derived from the ambiguous
 Christo Moskovsky & Alan Libert 109

categorial status of past participles, which can be used as (main) verbs

and also as adjectives: when the participle is used as a form of the verb,
then the ‘true passive’ interpretation is derived; when the participle is
used as an adjective, we get the ‘stative’ interpretation. With regard to
the latter, it is unclear in what ways sentences such as (8) differ from
sentences with a copular verb and a predicative adjective:

(9) The house was big.

In both cases they perform the grammatical function of

Subject-complement, and in both cases we can use the predicative
adjective attributively:

(10) a. the big house

b. the deserted house

In the light of this, it does not really seem justified to treat ‘stative’
passives as a genuine case of PV, and therefore, they will not be
considered in this paper. Another construction (see (11)), which is
sometimes described as “Middle Voice”, will also be ignored here:

(11) The door opened.

While semantically there may be good reasons to describe (11) in

different terms from a sentence like My aunt arrived, in the strictly
formal approach adopted here both can only be analysed as
intransitive sentences in the active voice.
Before we complete these brief introductory notes, a few words
about the notion of ‘antipassive’ are in order. Constructions referred
to with the term ‘antipassive’ are normally found in Ergative-
Absolutive (ERG-ABS) languages; in a number of aspects, such
constructions can be treated as a mirror image of PV constructions in
Nominative-Accusative (NOM-ACC) languages. Within the latter, in
110 Possibilities for Passives in Natural and Artificial Languages

the ‘core’ unmarked (i.e., active) clause, the primary term is normally
marked for the NOM and can loosely be described as ACTOR (or the
entity involved in enacting the verbal action); the term loosely
denoting the PATIENT (or the entity being acted on) would usually
(but not always) take ACC marking; in clauses involving a single
term (i.e., intransitive clauses), the primary term is the NOM SU. In
these languages, passivization involves promoting a non-primary
term to SU position and demoting a primary term to a position with a
peripheral, e.g., oblique, status (see (2-3)). In contrast, in the typical
‘core’ unmarked clause in an ERG-ABS language the primary term is
normally marked for ABS and can loosely be described as denoting
the PATIENT (or the entity being acted on), while the ACTOR would
normally be marked for ERG (see (12a) below); in clauses involving
a single term (i.e., intransitive clauses), the primary term is the ABS
SU. In the corresponding ‘antipassive’ construction, the non-primary
term (the ACTOR) is promoted to SU position and is accordingly
marked for ABS, while the primary term of the active sentence (the
PATIENT) is demoted to a peripheral (oblique) position and is
marked for a different morphological case: Dative (DAT) in this case:

(12) Yidi (Australia)

a. Wagudya-ngu bunya-ø giba:l.
man-ERG woman-ABS scratch (PAST)
‘The man scratched the woman.’
b. Wagu:dya-ø giba-:dyi-nyu bunya:-nda.
man-ABS scratch-ANTIPASS-PST woman-DAT
‘The man scratched the woman.’
(Dixon 1977, cited in Foley & van Valin 1985:338)

Even though the promotion/demotion of terms taking place in

‘antipassive’ constructions like (12b) above goes, as far as the
ACTOR-PATIENT dichotomy is concerned, in the opposite direction
than the promotion/demotion of terms in typically PV constructions,
 Christo Moskovsky & Alan Libert 111

the phenomenon is essentially the same: a primary term is demoted

and a non-primary term is promoted.11 In light of this, there does not
seem to be a valid reason to treat ‘antipassive’ constructions
differently from standard PV. Within such an approach, one relevant
question is whether we should go as far as calling them “passive” (as
suggested by Palmer 1994:158). The view taken here is that the term
‘antipassive’ should be preserved, because it appears to capture the
essence of these constructions: ‘antipassivization’ in ERG-ABS
languages seems to involve exactly what occurs when a NOM-ACC
PV construction is converted into a corresponding active construction,
i.e. an ACTOR/AGENT is promoted and a PATIENT is demoted. In
this regard, the term ‘antipassive’ appears to be very appropriate.
Another solid reason why the term ‘antipassive’ should be preserved
is to distinguish constructions like (12b) above from what seem to be
typically PV constructions like (13b), which we also find in
ERG-ABS languages (sometimes in the same language alongside
‘antipassive’ constructions):

(13) Tzotzil (Mayan, Mexico)

a. S-mil-ox-ø Xan li Petal e.
3SG(ERG)-kill-PST-3SG(ABS) John DET Peter
‘Peter killed John.’
b. Mil-bil-ø yuun Petal li Xan e.
kill-PASS-3SG(ABS) by Peter DET John
‘John was killed by Peter.’ (Foley & van Valin 1985:312)

In the following two sections, we shall consider formal and

functional aspects of the typology of PV constructions in natural
language, and then in the second part of this paper we shall present an
overview of PV constructions in a range of artificial languages (ALs).

11
Other terms frequently used in the same sense are ‘foregrounding’ and
‘backgrounding’, respectively.
112 Possibilities for Passives in Natural and Artificial Languages

2.1. Typology of Passive Voice Formation in Natural Languages

As can be expected, there is a substantial amount of variety in

relation to how different natural languages form their PV
constructions, which sentence positions (e.g., DO, IO, OPrep,
Adverbial, etc.) can undergo passivization, and the range of functions
that a PV construction can perform. Before these issues are briefly
discussed below, it is important to note that there are at least two
aspects in which PV constructions are, crosslinguistically,
remarkably uniform: PV is invariably a formal feature of the verb
phrase, and it invariably involves demotion of a ‘core’ term (or
argument): most commonly the clausal subject. Also, in all (but one)
of the natural languages that we have reviewed for the purposes of
this paper, we find the characteristic dichotomy between an
unmarked Active Voice and a marked Passive Voice construction
(see (4) above); the only exception we are aware of is Tagalog, in
which Active and Passive voice seem to be equally marked:

(14) Tagalog
a. S[um]ampal ng lalake ang babae.
[ACTIVE]-slap DO man TOP woman
‘The woman slapped the man.’
b. S[in]ampal ng babae ang lalake.
[PASSIVE]-slap ACT woman TOP man
‘The man was slapped by the woman.’(Keenan 1985:252)

One typologically important distinction which we already briefly

touched upon is between synthetic and analytical PV (‘strict
morphological’ and ‘periphrastic’ in Keenan’s (1985) terms): in the
former, PV formation involves a passive morphological marker on
the verb as, e.g., in (3), repeated here:
 Christo Moskovsky & Alan Libert 113

(3) b. Mweébwe mw-aa-boon-y-w-e n’âbagóre.

You you-PST-see-ASP-PASS-ASP by-women
‘You were seen by the women.’

Another language which uses synthetic PV is Turkish:

(15) Cevap yaz-mak iin kâğt-la kalem kullan-l-r.

answer write-INF for paper-and pen use-PASS-PRES
‘Paper and pencil may be used to write the answer.’
(Shibatani 1985:828, corrected)

It seems that in the vast majority of languages using synthetic PV

there is a single passive morphological marker which is used to mark
PV, regardless of which clausal term is involved in the passivization
(e.g., DO, IO, or something else). We are aware, however, of at least
one language which uses different morphological markers for PV
constructions involving different clausal terms:

(16) Malagasy
a. Nividy ny vary ho anny ankizy ny vehivavy.
bought-ACT the rice for the children the woman
‘The woman bought rice for the childen.’
b. Nividin ny vehivavy ho anny ankizy
bought-PASS the woman for the children
ny vary.
the rice
‘The rice was bought by the woman for the children.’
c. Nividianan ny vehivavy ny vary ny ankizy.
bought-CIRC the woman the rice the children
‘The children were bought rice by the woman.’
d. Nividianan ny vehivavy ny vary ny vola.
bought-CIRC the woman the rice the money
‘The money was used to buy rice by the woman.’
114 Possibilities for Passives in Natural and Artificial Languages

(Keenan 1972:172-3, cited in Palmer 1994:17)

(16a) is the unmarked active clause from which various PV

constructions are derived. (16b) involves the most basic type of PV,
in which the DO is promoted to SU position. (16c) and (16d) involve
promotion of IO and INSTR, respectively.12 The reader will note that
the morphological marker used for the passivization of the DO is
different from the morphological marker used for the passivization of
the IO and the INSTR.
Before we move on to consider the analytical (or ‘periphrastic’)
type of PV construction, we should note that there are languages in
which the passive morpheme is also used to express reflexive and
middle meaning:

(17) Western Armenian

a. Namag- k r-v-ets-av Mari-e-n.
letter-the write-v-AOR-3SG Mari-ABL-the
‘The letter was written by Mari.’
b. Vartan- hak-v-ets-av.
Vartan-the wear-v-AOR-3SG
‘Vartan dressed (himself).’
c. Tur- kots-v-ets-av.
door-the open-v-AOR-3SG
‘The door opened.’ (Haig 1982:162-5)

Such data should not be regarded as very unusual or surprising, given the
substantial overlap which we find crosslinguistically between passive,
reflexive, impersonal, and middle constructions (briefly discussed below).
As Keenan has noted, while most natural languages would make
use of either one or the other, occasionally we find languages using
both. Whether one is used or the other can be determined by aspect, as

12
Keenan has described the constructions in (16c, d) as ‘circumstantial’ voice.
 Christo Moskovsky & Alan Libert 115

in the following:

(18) Latin
a. Pugnatur uno tempore
fight-3SG-PRES-PASS one-SG-AB time-SG-ABL
omnibus locis.
all-PL-ABL place-PL-ABL
‘There is fighting at one time in all places.’
(Caes. B.G. 7, 84, in Palmer 1994:127)
b. Dareus (ab Alexandro) victus est.
Darius (by Alexander) conquered is
‘Darius was conquered (by Alexander).’ (Keenan 1985:251)

Most of the European languages are like English with regard to

PV formation: it involves a combination of two grammatical
elements: an AUX (usually be) and a (participial) form of the main
verb. More often than not, these grammatical elements are not
specific to PV: for instance, in English, the AUX be is also used to
mark Progressive aspect, while the Past Participle is used in the
Perfect Aspect. There are languages (e.g., Bulgarian), in which we
find two distinct forms of the past participle: one for use in active
constructions (e.g., with the perfect aspect) and another strictly for
PV constructions:

(19) a. Bashta mu e postroil nov dom.

father his-CL is build-PASTPART(ACT) new home
‘His father has built a new home.’
b. Nov dom beshe postroen
new home was build-PASTPART(PASS)
(ot bashta mu).
from father his-CL
‘A new home was built (by his father).’
116 Possibilities for Passives in Natural and Artificial Languages

Again according to Keenan (1985), PV constructions of the

analytical type can be further subdivided into four categories,
depending on the type of AUX involved: 1) be/become (see (20a)
below, as well as (2) and (6) above); 2) verbs of reception (see (20b)
below); 3) verbs of motion (see (20c) below); 4) verbs of
experiencing (see (20d) below).

(20) a. Persian
Ahmed košté šod.
Ahmed killed become
‘Ahmed was killed.’ (Keenan 1985:257-60)
b. Tzeltal
La y-ich utel (yuun s-tat) te
PST he-receive bawling out because his-father
Ziak-e.
ART Ziak-ART
‘Ziak got a bawling out (from his father).’ (ibid.)
c. Hindi
Murgi mari gayee.
chicken killed went
‘The chicken was killed.’ (ibid.)
d. Vietnamese
Quang bi (Bao) ghet.
Quang suffer (Bao) detest
‘Quang is detested (by Bao).’ (ibid.)

It is not uncommon for a language to make use of more than one

of these auxiliary verbs. Indeed, this is the case in Persian, where we
find both ‘become’ (as in (20a) above) and ‘go’ used as a PV
auxiliary:

(21) a. Ali loget-ra be kar bord.

Ali word-DO to work take
 Christo Moskovsky & Alan Libert 117

‘Ali used the word.’

b. Loget be kar reft.
word to work went
‘The word was used.’ (ibid.)

In English itself, we find the verb ‘get’ sometimes used to express

passive meaning, as in the following example:

(22) Both doctors got reprimanded by the hospital board.

(Huddleston & Pullum 2002:1443)

It appears, however, that this use of ‘get’ is highly restricted in

that there are a range of semantic and pragmatic limitations on it13.
Also, there is not complete agreement in relation to the actual status
of ‘get’ in these constructions: is it an auxiliary verb or not?
Huddleston & Pullum (ibid.) have argued that constructions like (22)
are actually complex sentences consisting of a main clause [Both
doctors got X], which is in the Active voice, and a subordinate
complement clause [X PRO reprimanded by the hospital board],
which is in the PV.14
Our overview will remain incomplete if we fail to discuss the use
of reflexive forms (e.g., reflexive clitics) to express passive meaning.
In the natural languages we have considered in this paper, marking
13
Huddleston & Pullum (2002: 1442) note, for instance, that constructions like (22)
are only found with what they call ‘dynamic verbs’.
14
Huddleston & Pullum (2002: 1430) call these ‘bare passives’. Here is another
example they discuss:

The guy [mauled by our neighbour’s dog] is in intensive care.

While superficially the bracketed relative clause is incomplete in that it has no

overt subject and no AUX, they are recoverable and therefore it is not completely
clear whether treating them in terms of ‘bare passives” is the most appropriate
analysis for them. If this analysis is, nevertheless, adopted, then these ‘bare’
passives would constitute a case of synthetic PV.
118 Possibilities for Passives in Natural and Artificial Languages

PV with a reflexive form is a very common device, alongside the

“true” PV marker (be that ‘periphrastic’ or ‘strict morphological’).
The following example from Russian illustrates the two types:

(23) a. Kalitka byla otkryta

gate-NOM was open-PERFECTIVE-PST PART
Olegom.
Oleg-INSTR
‘The gate was opened by Oleg.’
b. Kalitka otkrylas’ Olegom
gate-NOM open-IMPERFECTIVE-REFL Oleg-INSTR
‘The gate was opened by Oleg.’ (Siewerska 1985:247)

Passive-reflexive forms are very common in the Slavic languages:

(24) Czech
Zivne latky se pobleuji filtrem.
Nutritive substances REFL absorb-3PL filter-INSTR
‘Nutritive substances are absorbed by filter.’ (ibid.:246)

(25) Polish
Owe przesądy dzisiaj inaczej się
These prejudices-ACC today differently SELF
Interpretuje.
interpret
‘These prejudices are interpreted differently today.’
(ibid.:262)

(26) Bulgarian
Vasheto predlozhenie shte se razgleda ot
your-the proposal will self-CL consider from
saveta.
council-the
 Christo Moskovsky & Alan Libert 119

‘Your proposal will be considered by the Council.’

We also find passive-reflexive forms in the Romance languages:

(27) Italian
a. Si comprano due penne.
REFL buy-3PL two pens
‘Two pens are bought.’
b. Questo giornale si legge ogni
This newspaper REFL read-3SG-PRES each
mattina da morning by.
moltissima gente very many people
‘This newspaper is read every morning by lots of people.’
(Palmer 1994:143)

(28) Spanish
Se curó a los brujos.
REFL cured-3SING to the-PL sorcerers-PL
‘The sorcerers were cured.’ (ibid.:144)

(29) French
La porte s'est ouverte.
the door REFL-is opened
‘The door opened.’ (ibid.:143)

It should be noted that in the Romance language these forms are

more commonly used to express impersonal (e.g., (27a) and (28))
and/or “middle” (e.g., (29)), rather than “true” passive, meaning. In
fact, in contrast to Slavic, these languages do not easily permit
expression of the agent in passive-reflexive constructions of the type
exemplified above, and sentences like the one in (27b) seem to be quite
rare. In relation to this, it should also be noted that crosslinguistically
there seems to be a substantial overlap among constructions expressing
120 Possibilities for Passives in Natural and Artificial Languages

passive, reflexive, middle and impersonal meaning (for an extensive

discussion of this issue, see, e.g., Shibatani 1985). Interestingly, with
all of them the reflexive marker (a pronoun, a clitic or a bound
morpheme) performs the same function: it changes the
subcategorization frame of the verb by reducing the number of its
arguments; as we shall see shortly, this is also one of the most
important functions of passivization (according to some, the principal
one). 15
The wide-spread occurrence of passive-reflexive forms in natural
languages should not necessarily be seen as very unusual or
surprising, given that reflexivity inherently involves a “passive”
component: in the most typical case of ‘true’ reflexivity, the entity
denoted by the SU NP initiates and carries out an action which is
directed back onto the same entity, making the SU not only an Actor
(e.g., Agent or Experiencer), but also a Patient:

(30) John described himself as an honest man.

(Libert & Moskovsky 2002:105)

The action denoted by the verb in a reflexive construction can be

analysed as consisting of two distinct components: an active and a
passive one. Curiously, the device reflexive constructions in some
languages employ to convey a strictly passive meaning is to somehow
block the active component of the action. For example, in Bulgarian
this is accomplished by restricting the semantic category of SUs of
passive-reflexive constructions to only inanimate NPs (Maslov
1956): the referent of an inanimate NP SU cannot normally initiate
and/or perform an action, and this way only the second (passive)

15
Most commonly, this involves detransitavization of a transitive verb, but can also
occur with intransitive verbs, which eliminates the SU Actor, effectively making
the sentence impersonal (see below).
 Christo Moskovsky & Alan Libert 121

component of the reflexive action remains in force16:

(31) Sluchaj-jat se opisa v presa-ta.

case-the self-CL described in press-the

‘The case was described in the Press.’

Finally, a brief note about the distribution of passive-reflexive and

true passive constructions within the same language. It seems that a
semi-complementary relationship holds between the two types of
passive: the passive-reflexive type seems to be used almost
exclusively with imperfective verbs (or verbs in the imperfective
aspect), while true PV can be used with both perfective and
imperfective verbs (Nedjalkov & Jaxontov 1988). Another
significant difference between the two is manifested in the fact that
non-reflexive PV constructions can have a ‘stative’ interpretation
alongside a ‘true passive’ interpretation, while passive-reflexive
constructions can only have a ‘true passive’ interpretation.

2.2. Typology of Passivizable Positions

In the most common case, the DO NP of the active sentence

becomes the SU of the PV sentence while the SU of the active
sentence is assigned an oblique (e.g., prepositional) status or is
removed altogether, as in (2), repeated here for convenience:

16
There are rare cases in which a reflexive construction with an animate SU can have
a strictly passive meaning, but then there are other factors which preclude active
participation of the SU in the verbal action, e.g., the semantics of the verb:

(i) Toj se rodi v bogato semejstvo.

He self-CL bore in rich family
‘He was born in a rich family.’
122 Possibilities for Passives in Natural and Artificial Languages

(2) The Council will consider [your proposal]DO →

[Your proposal]SU will be considered (by the Council).

In English PV constructions the provision of the Actor/Agent (the

SU NP of the active sentence) is optional, however there are many
natural languages which do not allow it. The following example is
from Latvian, but Keenan (1985:249) lists a number of other
languages with the same situation.

(32) Es tieku macits (*no mates)

I am taught (by mother) (ibid.)

Keenan also stresses that even in languages which in principle

allow the expression of the Actor/Agent in PV constructions,
agentless passives are quite often preferred. In view of the fact that
one of the main functions of passivization is actually the demotion of
the term expressing the Actor/Agent (in NOM-ACC languages), this
preference for agentless passives is hardly surprising.
Passivization of positions different from DO (e.g., IO, OPrep,
Adverbial, etc.) is possible, but is relatively rare. Among the
European languages at least, English seems to be rather exceptional
in allowing types of PV constructions like the following:

(33) Bill will be offered an executive position in the company. [IO]

(34) The problem was dealt with very efficiently. [OPrep]

Not surprisingly, such constructions have been subject to

extensive research and different analyses have been proposed for
their treatment. In relation to sentences like (34), e.g., it has been
proposed that the preposition is re-analysed together with the verb,
forming a verbal complex [V + P], effectively turning the NP into a
DO. It is not among the goals of our paper to consider the merit of
 Christo Moskovsky & Alan Libert 123

such proposals. What is important for us is that, regardless of the

linguistic analysis we adopt, PV constructions of the types
exemplified in (33-4) do not exist in most European languages, and
are generally rare. This is not to say that languages not allowing PV
constructions like (33-4) do not have the potential to promote a
non-DO NP: they usually do, but would employ a different
mechanism. One way languages can do that is by using an impersonal
PV construction:

(35) German
a. Es wurde ihm geholfen.
It-IMPERS became him-DAT helped
‘He was helped.’
b. Ihm wurde geholfen.
Him-DAT became helped
‘He was helped.’
c. Gestern wurde ihm geholfen.
Yesterday became him-DAT helped
‘He was helped yesterday.’

Crosslinguistic evidence indicates that practically any sentence

constituent (including temporal and locative nominals) can be
promoted (see, e.g., the Malagasy examples in (16) above). Consider
the following examples from Kinyarwanda (Kimenyi 1988:361-2):

(36) a. Umugóre y-á-báa-ye perezida.

woman she-PST-be-ASP president
‘The woman became a president.’
b. Perezida y-á-báa-w-e n’ûmugóre.
President he-PST-be-PASS-ASP by woman
‘It is the woman that became a president.’
(literally: ‘A president was become by the woman.’)
124 Possibilities for Passives in Natural and Artificial Languages

(37) a. Umagabo a-ra-geend-a ijoro.

Man he-PRES-travel-ASP night
‘The man is travelling in the night.’
b. Ijoro ri-ra-geend-w-a n’ûmugabo.
night it-PRES-travel-PASS-ASP by man
‘It is in the night that the man is travelling.’
(literally: ‘The night is being travelled in by the man.’)

In the first of the examples above, it is the complement NP of a

copular verb that undergoes passivization, while in the second
example we see a temporal nominal promoted as the SU of the
corresponding PV construction. We find another interesting example
in Palauan (W. Austronesian) (Foley and van Valin 1985:316-17):

(38) a. a ngelek-ek a smeer er a tereter.

DET child-my sick-INSRT with DET cold
‘My child is sick with a cold.’
b. a tereter a l-seer er ngiy a
DET cold PASS+3SG-sick with 3SG DET
ngelek-ek.
child-my
‘With a cold is being sick by my child.’

Even from the more ‘permissive’ perspective of the English

language, these are very unusual examples.
Some natural languages allow for passivization of intransitive
active sentences, which does not involve promotion of a non-primary
term, but only demotion of the primary term, effectively making the
sentence impersonal (also see the Latin example in (18a):

(39) Bulgarian
a. Njakoj e minaval po
someone is pass-PSTPART(ACTIVE) along
 Christo Moskovsky & Alan Libert 125

tazi pateka skoro.

this trail soon
‘Someone has passed long this trail recently.’
b. Minavano e po tazi pateka skoro.
pass-PSTPART(PASS) is along this trail soon
‘Someone has passed long this trail recently.’
(literally: ‘(It) has been passed along this trail recently.’)

Finally, there are languages like Japanese in which passivization

sometimes involves not the promotion of an existing non-primary
term, but the creation of a new term; this is the case with the so-called
‘adversity passives’ (Kuno 1973:23-4, in Palmer 1994:130):

(40) a. Tuma ga sin-da.

wife NOM die-PST
‘The wife died.’
b. John ga tuma ni sin-are-ta.
John NOM wife by die-PASS-PST
‘John was adversely affected by his wife’s death.’
(literally: ‘John was died by his wife.’)

2.3. Typology of Passive Voice Functions

There is general agreement that syntactically passivization

involves promotion of a non-primary term and demotion of a primary
term. In fact, some (e.g., Shibatani 1985) have argued that the latter is
more fundamental than the former, because some PV constructions
may not involve promotion of a non-primary term (as is the case with
PV constructions derived from intransitive active sentences (see (39)
above)), but all PV constructions involve demotion of a primary term.
As we noted above, passivization typically involves a change in
the subcategorization frame of the verb by reducing the number of its
arguments. Quite commonly, this is an operation of detransitivizing
126 Possibilities for Passives in Natural and Artificial Languages

the transitive verb in the active sentence. This is accomplished either

by applying passive morphology or by using a reflexive marker
(pronoun, clitic, or a morpheme). In fact, there are languages in which
a common detransitivizing morpheme is also used as the formative
marker for PV:

(41) Ainu (Japan)

a. Kamui umma raike.
Bear horse kill
‘A bear killed a horse.’
b. Umma kamui orowa a-raike.
horse bear from PASS-kill
‘The horse was killed by a bear.’ (Shibatani 1985: 824)

As also noted above (see (40) above), in some rare cases

passivization involves the creation of a completely new argument, i.e.
one which did not exist in the active version.
From a discourse perspective, the promotion of non-primary
terms leads to changes in the discourse structure of the sentence: the
promoted term acquires a greater thematic focus. Likewise, the
demotion of the primary term takes away (some of) its thematic value.
As we indicated earlier, in a large number of languages the
ACTOR/AGENT associated with the primary term of the active
sentence is rarely expressed, or even cannot be expressed at all, in the
corresponding PV construction. As many authors have noted (e.g.,
Jespersen 1924), this makes PV constructions particularly suitable for
discourse contexts in which the ACTOR/AGENT is unknown or is
unimportant and/or the speaker wishes to focus on the PATIENT (and,
perhaps, the verbal action) rather than the ACTOR/AGENT. This is
one of the reasons why in many languages PV constructions can yield
an impersonal interpretation.
This completes our necessarily brief discussion of PV in natural
languages. Next we shall how this grammatical category is
 Christo Moskovsky & Alan Libert 127

represented in a range of artificial languages.

3. Artificial Languages

We shall discuss only a fraction of the more than a thousand

artificial languages which have been created, but this will, we hope,
give a general idea of the choices AL designers have made with
respect to the passive. A small number of ALs simply do not have a
passive.17 For example, Wald (1909) says of pan-kel: “There is no
passive”; his example of how it is to be replaced is that I was loved
would be expressed by the pan-kel equivalent of One loves me (ibid.).
Similarly, concerning frater2 Bartlett (1998) states,

There is no passive construction as such. One may use an

impersonal construction with ANTROP as subject, or, if even that
will not serve, one may use the null pronoun KO to indicate that
the object receives the action of the verb without specifying any
subject or agent. Alternatively, in many cases one may use the
reflexive pronoun AFTO to convert an actual or implied passive
form: “The book is (found) on the table” becomes BIBLO DONE
UREKE AFTO EPI MENSA (literally, “the book confers
finding itself on the table”).

One might also note that while Unitario does have a passive, the creator
of the language states that it “should be avoided” (Pleyer 1990:105).
17
This situation is to be distinguished from that when a full grammar of an AL has
not (apparently) been presented and passives are not discussed and do not occur in
texts in the language, e.g., Fitusa as described by Rosenblum (1935). For such ALs
we cannot know whether there would have been a passive if all the details of the
language had been presented. This is again different from the situation when the
grammatical description of the AL does not mention passives, but where they
occur in texts, as happens in Jones (1972) about Eurolengo. Here we are concerned
with ALs for which it is explicitly stated that they do not have a passive.
128 Possibilities for Passives in Natural and Artificial Languages

3.1. Synthetic and Analytic Passives

The majority of ALs do have a passive, and we can ask about the
general form of it: is it (always) synthetic, (always) analytic, or
sometimes synthetic and sometimes analytic, depending on the tense,
mood, etc.? A fourth possibility is for an AL to have both synthetic
and analytic forms, in free variation.
A fair number of ALs only mark passive forms synthetically in all
tenses/aspects and moods (including participles and infinitives), or at
least in all tenses/aspects/moods which are synthetic in the active
voice. 18 These appear to include the Blue Language, Bopal,
Communicationssprache, Dil, Gilo, Kosmos,19 Langue international
neo-latin, Lingua, Menet’s lang. universelle, Mondlingvo, Mundolinco,
Myrana, Olingo (hyphens), Oz, Pantos-Dimou-Glossa, Parla, Perio,
Qosmiani, Unish, Veltparl, Volapük, Völkerverkehrsprache,
Weltsprache (Eichhorn), and Weltsprache (Volk and Fuchs). One
must be careful when including a language in this list, since there are
some languages, both natural and artificial, which generally have
synthetic passives, but which have a few forms which are analytic, i.e.
one must look at the entire verbal paradigm. Some ALs have a
passive marker which apparently is connected with a hyphen, e.g.,
Balta’s oj- as in al oj-alog ‘I am seen’ (Couturat and Leau
1903/1979:19). It precedes the tense marker, e.g., a- for the present
tense, which is clearly a prefix, i.e., is not connected with a hyphen. In
such cases one might wonder why only one of these types of markers
is hyphenated.
18
Of course one would only expect passives in all tenses/aspects and moods to be
synthetic if all active forms were analytic. The significant question really is
whether there are any analytic passive forms when the corresponding active forms
are sythethic.
19
It may be that Kosmos has some analytic passive forms as well (and analytic active
forms), given the following remark by Couturat & Leau (1903/1979:379): “One
can use the verb esomi ‘to be’ with the participles of other verbs to render diverse
nuances of the latter.”
 Christo Moskovsky & Alan Libert 129

There are many ALs, including the most widely used ones, which
mainly have analytic passives, but again one must take care when
classifying, since many ALs generally have analytic passives but
have a few synthetic forms, usually or always participles and/or
infinitives. There appear to be very few, if any, ALs whose passive
forms are all analytic. One might think that all the passive forms of
Interlingua (IALA) are analytic, given the following remark by Gode
and Blair (1951:38):

The passive infinitives and participles are similarly composed of

the infinitive and participles of the auxiliary with the past
participle of the main verb following.

esser survelitate ‘to be watched’

essente surveliate ‘being watched’
essite surveliate ‘been watched’

However, the past participle can occur alone as an adjective with a

passive meaning, e.g., Le pluvia impellite per le vento entra trans le
fenestra rumpite ‘The rain driven by the wind comes in through the
broken window’ (ibid.:34), and so one could argue that not all the
passive forms of the language are analytic. The same situation might
hold in many or most other ALs which appear to have an entirely
analytic passive system. One the other hand, it could be said that the
words in question are not forms of the verb, but adjectives, and thus
that all verb forms are analytic. Most of the passive forms of
Zahlensprache are analytic, but the (present) infinitive and participles
are synthetic; however, again one might claim that the participles are
not verbal forms: Couturat and Leau label the passive participle forms
which they give as nouns, adjective, and adverb.
Esperanto is a language which is mainly analytic in passive
formation, but with possible exceptions: most passive forms consist
of a form of esti ‘to be’ and a passive participle, e.g., Ili estis amataj
130 Possibilities for Passives in Natural and Artificial Languages

de ĉiuj ‘The were loved by all’ (Wells 1969:10), but the passive
participles themselves are synthetic, e.g., the present passive
participle amata ‘(being) loved’ (c.f., the present active participle
amanta ‘loving’).
Some ALs have (mainly) analytic verbal systems, not only in the
passive but also in the active. In at least some languages of this sort
the passive could not be said to be more marked than the active or
derived from it. As we shall see, the same is true for some languages
with (mainly) synthetic passives. There are thus ALs in which both
voices are (largely/mainly) synthetic, those where the active is
(mainly) synthetic and the passive is (mainly) analytic, and those in
which both voices are (mainly) analytic; however, we know of no
ALs which have (mainly) synthetic passives and (mainly) analytic
actives. In Hély’s language all forms of all verbs, with the exception
of ser and der, meaning ‘to be’ and ‘to do’ respectively, seem to be
analytic, having as their first element the latter of these verbs in the
active voice, and the former in the passive and neutral.20
Spokil has three ways of constructing passive forms, one synthetic
and two analytic, a fact which becomes less surprising when we learn
that there are two methods of forming active verbs, one synthetic and
one analytic. Active tense/aspect/mood forms can involve affixes,
e.g., arbai and arbei are present and past indicative forms of arbe ‘to
work’, or they can consist of an infinitive followed by a word
containing l- and one or two tense/aspect/mood suffixes, e.g., arbe
lai is another present indicative form of arbe. 21 One can build a
synthetic passive from a synthetic active form by means of -en-,
which follows the root and precedes the tense/aspect/mood suffixes,
e.g., movenai is a present passive indicative form of move ‘love’. One
kind of analytic passive is parallel in structure to the analytic active,
20
This is one of a small number of ALs which have another possibility besides active
and passive. Couturat and Leau (1907/1979:17) say of the neutral in Hély’s
language, “statifs, d’état” (“stative, of state”).
21
In fact there is a third way of constructing active forms in the secondary tenses.
 Christo Moskovsky & Alan Libert 131

except that instead of the infinitive of the main verb there is the
passive participle, e.g., moveno lai. The other analytic passive
consists of a form of ve ‘to be’ (which will bear the tense/aspect/mood
marking) and then the passive participle, e.g., vai moveno.
Some ALs have both synthetic and analytic passives in at least
some tenses, moods, etc. This is true of Europal, although one may
get the impression from Weisbart’s (1972:8) description that the
synthetic forms have some sort of priority or are preferable: “The
passive forms are formed by attaching the indicative endings to the
present passive participle: i salutadan ‘I am greeted’. The forms iu
san salutad are also permissible.” (San is the present indicative of ‘to
be’, i is a shortened form of iu ‘I’.) Passive forms in Orba can be
constructed with a form of the auxiliary ese ‘to be’ and the past
participle. Orba also has a synthetic passive (at least of some
tenses/aspects), formed by attaching tense/aspect/mood and
person/number agreement markers as suffixes to the root. For
example, the 1st person present passive indicative of lemse ‘to love’
is ebo lemso or lemsobo.
There are both synthetic and analytic passives in Communia.
Synthetic finite forms are marked by the suffix -r, placed at the end of
active forms, e.g., the present and future active indicative forms of
laudare ‘to praise’ are lauda and laudaro respectively, while their
passive equivalents are laudar and laudaror. Passive infinitives are
formed in a different manner: they end with -i instead of the -e ending of
the active infinitives, e.g., laudari. Passive participles are also formed in
a different way: compare the active and passive present participles
laudant and laudandi, and the future active and passive participles
laudaront and laudarondi. The analytic passive involves a form of esere
‘to be’ and a participle, probably the “absolute participle”.22
The order of the parts of analytic passives is generally for the

22
Couturat & Leau (1903/1979:411) say, “The author allows in addition an absolute
participle … which can be used in the active as in the passive …”
132 Possibilities for Passives in Natural and Artificial Languages

auxiliary verb to precede main verb form, as in most western

European languages. However, in Unitario passives the main verb
precedes the auxiliary, e.g., ego laudante estas, ‘I am praised’ (Pleyer
1990:105), “to set the passive apart from the other compound tenses”
(ibid.); compare ego habas laudante ‘I have praised’ and ego sum
studiendo ‘I am studying’ (ibid., OT).
In many natural languages with analytic passives, the main verb
form does not need to be immediately after the auxiliary verb. The
same possibility is explicitly permitted in American: “The passive
voice is formed by using the verb, to be, as an auxiliary verb, with and
before (though not necessarily the next word before) the perfect
participle” (O’Connor 1917:16). Below is an example of a passive
construction in American with the two parts separated.

(42) … Francıán ůx ı muć dıcedo

… French also is extensively spoken (ibid.:37)

On the other hand this may not be allowed in Zahlensprache, if the

following statement is accurate: “The passive is formed with the
auxiliary verb sumum (to be), followed immediately by the infinitive
from which one has cut off the final -m” (Couturat and Leau
1903/1979:101).

3.2. Uniform and Non-uniform Passive Formation

We have seen that some ALs are not uniform in their passive
formation. Even among ALs that are entirely synthetic or analytic in
this respect we find some which vary in how they make passive forms.
We have already seen that in Communia synthetic passive infinitives
and participles are marked differently than the other synthetic passive
forms. Perio forms passives in three ways, depending on the
tense/aspect/mood: in most cases -t is added at or near the end of the
active form; since the forms in question already have a t in this
 Christo Moskovsky & Alan Libert 133

position there will be a double or long tt, e.g., mi vidit ‘I see’, mi viditt
‘I am seen’. The passive of the present infinitive, present participle,
and past participle are built by attaching -h to the end of the active
form, e.g., vidi ‘to see’, vidih ‘to be seen’. Unless there is an error in
the able of verb forms in Couturat and Leau (1907/1979:6), the past
passive infinitive is formed in yet another way, by the deletion of t
from the active equivalent: vidati is ‘to have seen’ while vidai is ‘to
have been seen’
Dil seems to be an AL which is entirely uniform in forming
synthetic passives; the passive marker is always the suffix -i-, placed
between the root and any other suffixes, e.g., loben ‘to praise’, lobien
‘to be praised’; lobem ‘praises (3SG INDIC)’, lobiem ‘is praised’.
There are also ALs which lack passives in some moods, etc. As far
as we can determine, all passive forms in Hom-idyoma are part of the
“[c]ompound conjugation with the passive participle in ‘ata’”
(Cárdenas 1923:102) (with the exception of this participle itself), and
Cárdenas (ibid.:104) states that this conjugation “has no imperative”.
Perhaps this should not be a surprise, since one cannot control what is
done to one; it should be more surprising that so many languages,
both natural and artificial, have passive imperatives. However, one
might argue that passive imperatives are not to be interpreted
literally: when one says, “be seen”, what one means is ‘create a state
of affairs such that you are seen’ or something of the sort.
It is also theoretically possible for a language to have passive
forms with no corresponding active forms. This seems to be the case
for Idiom Neutral, which has a passive gerundive, e.g., amand
‘who/which must be loved’, but no active gerundive. (This is the
same situation as in Latin.)
Another kind of non-uniformity involves verbal irregularity or
different conjugations, i.e., where it is not different tenses, moods, etc.
which show differences in passive formation, but where the choice of
passive marking depends on the verb involved. However, such
differences generally do not involve only the passive, i.e., if there are
134 Possibilities for Passives in Natural and Artificial Languages

irregularities or more than one conjugation of verbs, they will show

differences in tense, mood formation as well as passive formation.
Therefore we shall not discuss this kind of variation here, with one
type of exception, involving, one might say, the most radical
difference possible: in some ALs, as in many (or most) natural
languages, verbs can differ in whether they have passive forms at all.
We believe that this would be common among ALs, but rarely is it
explicitly stated that there are no passive forms of certain verbs –
perhaps it did not occur to some designers to state this e.g., for ‘to be’
even though they would not allow passive forms of it, or they did not
think it necessary to state such a prohibition. Foulk (1937:37) is
explicit about this in Americai Speak, saying, “Only verbs that take
an object in the active voice have a passive voice” (our translation
from Americai Speak). Concerning Esperanto Wennergren (2003:
section “Pasivigeblaj frazoj”) states “Only objective [i.e., transitive]
verbs can be passivized”.23
On the other hand, it would be striking if an AL has passive forms
of e.g., ‘to be’. According to Couturat & Leau (1907/1979:7), the
passive of sasi ‘to be’ in Perio is sasih ‘to become’. However, one
might question whether this is really a passive, or simply a form with
passive morphology which does not have a passive meaning, unlike,
e.g., (38b), which could be argued to be involve a passivized copular
verb.24 In Weltsprache (Volk and Fuchs) the verb son ‘to become’

23
This is a prescriptive statement. For the vast majority of ALs there is not much
need to distinguish between prescriptive and descriptive statements, since they
saw so little use. Esperanto, as the most widely used AL is different in this regard.
Consider what Wennergren (2003:ibid) later says: “Sometimes because of the
influence of a national language the passivization of a completely intransitive verb
can appear. Such sentences are to be avoided, because they go against the most
basic principle of the passive”. He then quotes a passage from a translation by
Zamenhof from German into Esperanto which contains a passive of an
instransitive verb and says “The sentence is bad, although by Zamenhof”.
24
In their chapter on Ulla Couturat & Leau (1907/1979:92) give a form for the
passive of este ‘to be’, este esten, but they do not comment on it and the second
 Christo Moskovsky & Alan Libert 135

looks like the verb ‘to be’, san, with passive marking (i.e., the vowel
-o- instead of -a-), but it is not described by Couturat and Leau
(1903/1979:266-7) as the passive form of the latter.

3.3. Deponent Verbs

There are a fair number of ALs which are based partly or (almost)
entirely on Latin. Given that such languages are generally attempts to
simplify Latin, one might wonder whether deponent verbs, i.e. verbs
with passive form and active meaning, have been eliminated in them.
This is the case in Hély’s language: Couturat & Leau (1907/1979:18)
state, “The deponents would be restored to the active form (confiter,
or better confesser), as in the Romance languages”. This also seems
to be true of Kosmos, since Couturat and Leau (1903/1979:376) say,
The deponent verbs of Latin are treated as if they had the active form
(in -o). E.g.: imitomi, sequomi.” Likewise in Linguum Islianum,
“the deponent verbs are restored to the active form: imitare, mirare,
…” (ibid.:543). One the other hand, SPL has deponent verbs,
although Dominicus (1982:87) says, “in SPL, the deponent verbs can
be used either in their original passive form or in a simplified active
form.”

3.4. Passivizable Positions in ALs

One may wonder to what extent ALs allow promotion of indirect

and prepositional objects; we have seen in section 2.2 that some
natural languages allow this or an even wider range of clausal terms
to be promoted. Foulk (1937:180) states that “A verb in the passive
voice never has an object” (our translation from Americai Speak),
which seems to rule out the passivization with a ditransitive verb. We

verb does not have the form one would expect, given that passive participles end in
-ed, e.g., ested. We do not know whether this is an error.
136 Possibilities for Passives in Natural and Artificial Languages

know of no clear examples of the promotion of an indirect object (or

of any other constituent than a direct object) in any AL, and we know
of no explicit statements that it is permitted, but this may be partly
due to the fact that AL designers often do not discuss the syntax of
their languages in great detail. Lojban seems to have (something like)
the promotion of a wide range of positions, referred to as
“conversion”. However, this is said not to be passivization: “The
effect is similar to what in English is called ‘passive voice’. In Lojban,
however, a conversion is not ‘passive’” (Nicholas & Cowan 2003).

3.5. Prefixes, Suffixes, and the Location of the Affix

The location of the passive marking affixes relative to the root and
to other affixes varies. A prefix is used for this purpose in Bopal, Gilo,
Letellier’s language, Lingua, Mondlingvo, Myrana (which also has
an active prefix), Unish, and Volapük, while the passive of many
languages involves a suffix, e.g., Dil, Menet’s langue universelle,
Mundolinco, Pantos-Dimou-Glossa, Parla, Uropa, and Veltparl. In
Oz the passive marker, -l-, appears to be an infix: according to Elam
(1932:9) it is placed “after the vowel of the stem as ep itQks yaf I see
them; yaf itQlks ep they are seen by me.” The letter <Q> is this
language stands for a vowel, and we take the root meaning ‘see’ to be
Qks.
The order is passive-tense/aspect-root in Bopal (with the mood
and person/number agreement markers being suffixes)(recall that this
is also the order in Balta with its hyphenated passive marker). Myrana
has the order person/number-voice-tense/aspect-root (with the mood
markers again being suffixes). Unlike these languages, in Parla the
passive marker, -et-, is adjacent to the root, e.g., labeti ‘to be worked’,
labetai ‘to have been worked’ (-i and -a- being the infinitive and past
markers respectively). As we have mentioned, in Spokil synthetic
passives also involve a suffix which is immediately after the root, and
which precedes tense/aspect/mood markers. The Veltparl passive
 Christo Moskovsky & Alan Libert 137

suffix goes immediately after the root, and thus is before the mood
suffix, while the tense markers are prefixes. In Weltsprache
(Eichhorn) the passive suffix is between the tense/aspect suffix and
before the number agreement suffix (this language having number
agreement but not person agreement). We thus see that AL designers
have chosen a variety of different positions relative to other
morphemes in the verb for their passive/voice affixes, sometimes
before the root and sometimes after, sometimes closer to the root than
other verbal affixes and sometimes further away than them.

3.6. Choice of Auxiliary Verb

Among languages which have analytic passives, there are

differences in which auxiliary verb is used. We can draw a basic
distinction between those languages which have an auxiliary whose
only function is the formation of passives and those in which the
auxiliary verb involved in passive formation has other functions. The
former group is smaller. Arulo has the auxiliary verb verdar whose
only function apparently is passive formation; it is followed by the
present passive participle, which ends in -ata, e.g., el verdas vexata
‘she is (being) teased’ (Talmey 1925:12).25 The reason for this was
apparently to prevent ambiguity. However, the verb ‘to be’ can be
used for the same purpose: Talmey (ibid.:13) states, “Also esar may
be used as auxiliary to form the passive, but verdar is preferable; …
li esos (verdos) defetata, they will be defeated.”26
According to Couturat & Leau (1903/1979:207), in Dilpok “Le

25
Arulo has a synthetic version of this, with verd- as a suffix, followed by a verbal
ending, e.g., li punisverdis ‘they were punished’ (ibid.:13).
26
Again there is a synthetic version, with es- suffixed to a participle from which the
adjectival ending -a has been removed, and followed by a verbal ending, e.g., Deo
amatesas ‘God is (being) loved’ (ibid.:13). However, Talmey (ibid.:13) says that
“This passive is not recommendable; it produces forms that are too long;
reprezentatesabas, has been represented”.
138 Possibilities for Passives in Natural and Artificial Languages

passif se forme avec l’auxiliaire eri ‘etre’ et le participe passif: mi ere

loved = je suis aimé”. (The passive is formed with the auxiliary eri
‘to be’ and the passive participle: mi ere loved = I am loved”.)
However, there is another verb ‘to be’ in the language, esi, which is
used as a copula, e.g., mi ese glad ‘I am glad’ (ibid.). This may
indicate that ere is reserved for passive constructions.
INTAL has an auxiliary verb fi which apparently only is used in
passive forms, or more specifically in “the passive of the action”
(Weferling 1974:13),27 which we would take to mean the true passive,
while “the passive of the state” (ibid.) (i.e., we believe, the stative
passive) involve es ‘to be’. Weferling (ibid.) asserts that “It is
absolutely necessary to distinguish [between the two types of
passive] … in the international language”.28 The examples he gives
(ibid.) are Me fi konvikat ‘I become convinced’, Le porte fi klozat cak
vespre ‘The door is closed every evening’ (OT), and Le porte es
klozat dum le tot nokte ‘The door is closed through the whole night’
(OT).
Among languages of the second type, i.e., in which there is no
dedicated passive auxiliary, there are differences in the other
functions of the passive auxiliary verb (c.f., English vs. German).
Most languages of this type use ‘to be’ in passive forms (and
presumably it is also used as a copula). Such languages include
Americai Speak, Anglo-Franca, Dutton Speedwords, Internasionalo,
Latino sine Flexione, Latinulus, Lingua Internacional, The Master
Language, Nov Latin, Noviilatiin, Solresol, Spelin, Suma, and
Zahlensprache. (Note that this fact is not surprising if an AL is
(closely) based on English and/or a Romance language.) The passive
forms of Universal-Latein involve the auxiliary fieri ‘to become’, e.g.,

27
We have translated all quotations from this source from Weferling’s original INTAL.
28
Compare the remark in Gode and Blair (1951:38) concerning Interlingua (IALA):
“Note that the passive as such cannot be distinguished from a form of esser ‘to be’
with a participial adjective. Io es maritate may mean ‘I am being married’ and ‘I
am being married.’”
 Christo Moskovsky & Alan Libert 139

fieram laudarate ‘I am praised’. This is probably due to the influence

of the designer’s presumed native language, German.

3.7. Choice of Main Verb Form

ALs with analytic passives differ in which main verb form is used
in passives. Many ALs, like many natural languages, use a perfect or
past (active) participle as the form of the main verb, or to put it
another way, the perfect/past active participle and the passive
participle have the same form. These include American, Latino sine
Flexione, Latinulus, Mundolingue, Nov Latin, and Novilatiin. Lingua
Komun has an active and a passive participle. Couturat & Leau
(1903/1979:482), our source for this language, do not specify the
tense/aspect of these, but they say, “the (anterior) composed tenses
are formed by means of the auxiliary haver and the passive
participle; the tenses of the passive, by means of the auxiliary eser
and the same participle.” Thus Lingua Komun represents the same
situation, although described in a different manner in our source.
Apparently in contrast with these languages Antivolapük
indicative (and some other?) passive forms involve an infinitive, e.g.,
io es aimer ‘I am loved’ (Couturat & Leau 1903/1979:445). However,
the infinitive is not a typical (Indo-European) infinitive in terms of its
use: it also occurs in some active forms, e.g., io aver donner ‘I have
given’ (ibid.), and for that matter, one could say, in active present
forms, since e.g., io donner means ‘I give’. Further, our source for
Antivolapük, Couturat & Leau (1903/1979), says nothing about
participles in the language, so one could say that forms such as aimer
function as both infinitives and participles and therefore that its
passive is not really different from those of the languages just
mentioned. The same situation seems to hold of Suma: there is only
one synthetic form for verbs, which functions as an infinitive,
indicative, main verb form in passive constructions, and main verb in
past forms, e.g., ma oki mu ‘I see him’ (Russell 1966:2), ma te oki ‘I
140 Possibilities for Passives in Natural and Artificial Languages

am seen’.
Langue Catholique, on the other hand, can be more plausibly
argued to have passives which use infinitive forms of the main verb
(and this is how it is described by Couturat & Leau
(1903/1979:439-440)), since it has participles which are distinct from
infinitives, as the former end in -ante and -ente. 29 However, the
infinitives are the same in form as finite verb forms, i.e. indicative,
imperative, and subjunctive forms, e.g., ama is the imperfect active
infinitive/indicative/subjunctive/imperative of ‘to love’. The
imperfect passive of this verb is è ama ‘was/were loved’, è being the
present tense of ‘to be’.30 In any case, the form involved is not a
participle, so Langue Catholique does represent a different way of
forming passives than many ALs (and natural languages) with
analytic passive forms.
Faiguet de Villeneuve’s Langue Nouvelle represents another
possibility for constructing analytic passives: (a form identical to) the
present active indicative is the form employed (after a form of sas ‘to
be’), e.g., the present passive indicative is sa dona ‘am/are/is given’,
and in this language the present active indicative is different in form
from the present infinitive: the former ends in -a, e.g., dona, the latter
in -as, e.g., donas.31
It is only a minority of ALs with analytical passives which have a
special form of the main verb which is used only as a passive
participle (and not as a perfect participle and/or as some other form),
just as it is a small minority of such languages which have a dedicated

29
Alberto Liptay, the designer of Langue Catholique, “foresees two systems of
conjugations” (Couturat & Leau (ibid.:439); the description here may apply to
only one of them.
30
The form which Couturat & Leau (ibid.:400) give for the present passive of this
verb indicates that this form may be created in a different way: amè. (They say that
this is “for ame-è”).
31
The 2nd person singular/plural imperative also has the same form as the present
indicative (while the 1st and 3rd person imperatives have the same form as the
present subjunctive).
 Christo Moskovsky & Alan Libert 141

passive auxiliary. Dutton Speedwords is such a language: Dutton

(1951:64-5) says:

It is a weakness of English construction―the same principle is

also followed in certain other languages―that the endings ‘-d’
and ‘-ed’ are alike used (a) to express the past tense of what are
called weak verbs, and (b) the passive voice … In Speedwords a
definite distinction is made, e.g., J yga l bu (past tense) [‘I finished
the book’], but L bu r e gad by Peter (passive voice, future tense)
[‘ The book will be finished by Peter’], the passive voice indicator
in Speedwords being the single-letter suffix -d. […] The
distinction is made because Speedwords are intended for the use
of other nationals as well as the English-speaking peoples. Once
this is achieved the present barriers to free international
correspondence will be removed for all time.

As far as we can determine, in Spelin the form of the main verb

used in passives, which ends in -ed, e.g., i bi miled ‘I am loved’ , is
only used in the passive construction, even though Couturat & Leau
(1903/1979:474) called it “the past (passive participle)”, and even
though it apparently was derived from the English participle in -ed,
which is the past participle and so has another function. We say this
because the past and pluperfect forms in this language do not involve
a form in -ed; in fact they are synthetic, e.g., mile ‘has/have loved’,
mila ‘had loved’. The main verb form used in passives is
Zahlensprache apparently is distinct from all other verbal forms,
since it is the infinitive (which also functions as the present
indicative) without the latter’s last segment: e.g., lamorem ‘to love’,
sumum lamore ‘to be loved’.

3.8. Multiple Uses of Affixes

Some ALs with synthetic passives are similar to some of the ALs
142 Possibilities for Passives in Natural and Artificial Languages

discussed in sections 3.6 and 3.7 in that an element involved in

passive formation also has another function, i.e., the passive affix can
mark other categories.
In Gilo the prefix ad- marks passive verb forms:

(43) a. Tom angez Peter. ‘Tom angers Peter.’

b. Peter adagnez Tom. ‘Peter is angered by Tom.’
(Giles 2000: sec. 7.1.1)

The same element is also a preposition, and marks agents,

although it “is not required if the person/thing doing the action of the
verb immediate[ly] follows the verb” (ibid.). This explains why it
does not precede Tom in (43b) above; compare the sentences below:

(44) a. an piskis fata ad-og-oz u tri

one fish fat PASS-give-PST to three
filis blaka ad du peil ajoma.
cats white by two men old
‘One fat fish was given to three black cats by two old
men.’
b. an piskis fata u tri filis blaka adogoz du peil ajoma.
‘One fat fish to three black cats was given by two old
men.’ (ibid.: Appendix A)

In (44a) we have both the prefix ad- and the preposition ad, but
the latter does not need to occur in (44b), because again the agent
phrase is adjacent to the verb. Apparently the prefix is not required in
all passive sentences, given the following sentence:

(45) u il og-oz bøl ad An.

to him give-PST ball by Ann
 Christo Moskovsky & Alan Libert 143

‘To him was given the ball by Ann.’ (ibid.: sec. 3.5.3)32

It is not clear to us why the prefix has been/can be left out, and
perhaps this is a mistake on Giles’ part; he says that “This form is
‘grammatically correct’ but its use is not recommended” (ibid.),
although the absence of ad- may not be the reason for this. Ad can
also have an agentive type meaning for a phrase modifying a noun
e.g., buk ad Dikenz ‘a book by Dickens’ (ibid.: sect. 7.1.1) (c.f., buk
adritoz Dikenz ‘a book written by Dickens’ (ibid.)).33
The same general situation holds in Uhish: the passive marker is
be-, which as an independent word marks the agent of a passive, e.g.,
Buk beskribed be me ‘The book was written by me’ (Lee 2002:63).
The Uropa affix -ym- marks passive verb forms and objects.
Donisthorpe (1913:21) gives the following explanation for this
sharing of function:

It was explained in the Second Lesson that “ym” = “in relation

to.” Oma-vid e ym arba = the seeing-man is related to the tree; or,
as we say, The man sees the tree: Oma vide arbam; (“arbam” =
arba-ym). But if the seeing man is related to the tree, the tree must
be related to the seeing-man. If “oma vide arba-ym,” then “arba
yme oma-vide,” which may be written “arba vidyme oma,” or as
we say in English, “the tree is seen-by the man.” It appears, then,
that the root-idea in the “passive verb” is the root idea in the
“objective noun.” And, if the idea is the same, the same word
should be used to express it. That word is “ym.” “Ym” makes
nouns objective (or “accusative”) and it makes verbs objective (or

32
Note that this sentence is a possible example of a passivized indirect object,
although it is not entirely clear.
33
Ad is used to translate ‘for’ in the translation of “If as one people, speaking the
same language, they have begun to do this, then nothing that they wish will be
impossible for them”: If, biz pei aga, piez po ama, uli ezgidioz ce, øka zafi fi uli
wixuz buz oppos ad uli. (Giles 2000: Appendix C).
144 Possibilities for Passives in Natural and Artificial Languages

“passive”).

3.9. The Markedness/Derivedness of the Passive

As we have seen, in most natural languages, passives are more

marked than actives. There are two ways in which this is manifested:
in some languages a higher proportion of passive forms than of active
forms are analytic and in some languages passive forms are derived
from active forms by the addition of an affix. In ALs as well passives
generally seem more marked than actives, and in the same two ways.
However, there are some ALs in which one could not say that the
passive is more marked than the active. We already mentioned the
fact that in Hély’s language both active and passive forms are analytic.
In Mundolinco, as far as we can determine, all active and passive
forms are synthetic, and passive forms contain the suffix -r. However,
most active forms also have a suffix, -s, which occurs in the same
position as -r would, and so the passive apparently is not derived from
the active, and so is not more marked than it.34 In Myrana, in most
tenses/aspects both actives and passives (at least of most verbs) bear
voice markers (t- for the former and sh- for the latter35), e.g., the
imperfect active indicative of punir ‘to punish’ is mitapun, while the
corresponding passive is mishapun, so again the passive might not
appear more marked that the active. However, in the present tense the
active forms does not contain t- (or the present tense marker -e-),

34
The imperative, at least the form given in Couturat & Leau (1907/1979:46),
parlende, does not have the suffix -s. We do not know whether there are
corresponding passive forms of it. One might speculate that it does not bear the
active suffix since, as discussed above, strictly speaking a passive imperative is
semantically odd, and perhaps lacking in this language, and so the designer of the
language did not feel the need to give the active imperative forms the usual active
marker, since there were no corresponding passive forms to be kept distinct from
them.
35
These voice markers are derived from the verbs tedje ‘to do/make’ and sheje ‘to
become’.
 Christo Moskovsky & Alan Libert 145

while the passive forms still have sh-, e.g., the present active
indicative of punir is mipun, and its passive counterpart is mishepun.
Therefore one might claim that underlyingly in this language the
active is more basic than the passive.
In Weltsprache (Volk & Fuchs) the difference between active and
passive forms is marked by the vowel of the final syllable: the former
have -a- or -ä-, the latter -o- or -ö-. One could argue that in this
language the passive is not more marked that the active.

4. Recommendations

First of all, we recommend that an AL have a passive. Although

one might argue that an AL without one will be simpler, its existence
gives a language more flexibility and more possibilities to express the
subtle differences of meaning connected with thematic focus. As in
other areas, the language designer may be caught between
incompatible demands or desires, e.g., neutrality or internationality
vs. recognizability. In terms of pure ease of learning, some of the
choices we have discussed may not show a clear preference. Consider,
for example the choice of synthetic vs. analytic passives. The major
western European languages all have analytic passives and even
Latin had some analytic passive forms. Does this mean that ALs, at
least those whose potential speaker base would consist largely of
those who have such a language as a mother tongue, should avoid
synthetic passives? The answer is not clear to us, but we do not
believe that synthetic passives would be unduly difficult to learn
compared to an analytic passive, which may even be more difficult,
given that one has to learn which two forms, e.g., which participle, is
involved in their formation.
We would however recommend uniformity in passive formation:
the passive forms of the AL should be either all synthetic or all
analytic and we do not see any compelling reason to allow both
146 Possibilities for Passives in Natural and Artificial Languages

synthetic and analytic passives, and in fact we wonder why some

designers have done this; perhaps the reason is to permit stylistic
variation. We think that there should be only one set of forms of the
passive, all constructed in the same way (meaning with the same affix,
if a synthetic passive is chosen), and that these forms should have
precisely one function, that of marking the passive (e.g., they should
not also mark middle or impersonal constructions).
We have seen that natural languages differ considerably in which
positions they allow to be passivized. Here we would take a middle
course, neither restricting passivization to direct objects nor allowing
almost every NP in a sentence to become the subject of a passive.
Although the latter choice would yield a wide variety of possibilities
for thematic focus and stylistic variation, it could also produce many
sentences which would be difficult for the majority of speakers to
parse, given that most natural languages are not so free in this regard.
We have found it difficult to make a principled decision about which
positions should be passivizable: at least direct and indirect objects
should be able to be promoted, as well as objects of prepositions if the
preposition has undergone reanalysis to be joined to the verb and the
objects of verb-particle constructions, e.g., John looked up the word
–> The word was looked up by John. Possibly all arguments should be
passizivable, and the distinction between passivizable and
non-passivizable positions should be along the lines of the argument
vs. adjunct distinction. This option would lead to some sentences
which were strange to some speakers, but the most striking and
unusual instances of passivization from natural language would be
ruled out.
Finally, we believe that the passive should be marked in relation
to the active, i.e., if the language in question has a synthetic passive,
that this passive should be constructed by adding an affix to the
equivalent active form (rather than having affixes for both the active
and the passive). The vast majority of natural languages have passive
forms which can be said to be derived from active forms, and this may
 Christo Moskovsky & Alan Libert 147

reflect some underlying notion that actives are more basic than
passives, i.e., one might say, that the subject position (in
nominative-accusative languages) is generally associated with agents
and vice-versa. If this is the case, perhaps the marking of voices
should reflect which one is felt to be basic or unmarked and which
one is felt to be derived or marked.

References
Bartlett, P. 1998. Frater2. Available at URL <http://www.smart.net/~
bartlett/ frater2.html>.
Cárdenas, C. 1923. Hom-Idyomo (2nd edition). Leipzig: Fischer & Wittig.
Comrie, B. 1988. Passive and Voice. In Shibatani (ed.), 1-23.
Couturat, L. & L. Leau. 1903, 1907/1979. Histoire de la Language
Universelle and Les Nouvelles Langues Internationales. Hildesheim:
Georg Olms Verlag.
Dixon, R. 1977. A Grammar of Yidin. Cambridge: Cambridge University
Press.
Dominicus, R. 1982. SPL. Wisconsin: Dominicus Publishing House.
Donisthorpe, W. 1913. Uropa. Guildford: W. Stent & Sons.
Durie, M. 1988. The So-called Passive in Acehnese. Language 64, 104-13.
Dutton, R. 1951. Teach Yourself Dutton Speedwords. London: The English
Universities Press.
Elam, C. 1932. The Case for an A Priori Language. Cincinnati, OH: The
Open Sesame Press.
Foley, W. & R. van Valin. 1984. Functional Syntax and Universal Grammar.
Cambridge: Cambridge University Press.
Foulk, R. 1937. Americai Spek. New York: J. J. Little & Ives.
Giles, A. 2000. Gilo Grammar. Available at URL <http://users.classicfm.
net/alangiles/Grammar.htm>
Givón, T. 1979. On Understanding Grammar. New York: Academic Press
(cited in Palmer 1994).
Gleitman, L. & M. Liberman. 1995. The Cognitive Sciences of Language:
Introduction. Language: An Invitation to Cognitive Sciences 1.
Cambridge, MA: MIT Press.
Gode, A. & H. Blair. 1951. Interlingua: A Grammar of the International
148 Possibilities for Passives in Natural and Artificial Languages

Language. New York: Storm Publishers.

Haig, H. 1982. Passivization in Modern Western Armenian. In Hopper &
Thompson (eds.), 161-176 (cited in Palmer 1994).
Hashimoto, M. 1988. The Structure and Typology of the Chinese Passive
Construction. In Shibatani (ed.), 329-354.
Haspelmath, M. 1990. The Grammaticalization of Passive Morphology.
Studies in Language 14, 25-72.
Hopper, P. & S. Thompson. (eds.). 1982. Studies in Transitivity. Syntax and
Semantics 15. New York: Academic Press (cited in Palmer 1994).
Huddleston, R. & G. Pullum. 2002. The Cambridge Grammar of the English
Language. Cambridge: Cambridge University Press.
Jacobs, R. 1976. A Passive Continuum in Austronesian. Papers from the
Parasession on Diachronic Syntax 118-25. Chicago, IL: Chicago
Linguistics Society (cited in Shibatani 1985).
Jespersen, O. 1924. The Philosophy of Grammar. London & New York:
Allen & Unwin.
Jones, L. 1972. Eurolengo. Newcastle upon Tyne: Oriel Press.
Keenan, E. 1972. Relative Clause Formation in Malagasy and Some Related
and Some not so Related Languages. In Peranteau et al. (eds.), 169-189.
Keenan, E. 1985. Passive in the World’s Languages. In Shopen (ed.),
243-81.
Kimenyi, A. 1988. Passiveness in Kinyarwanda. In Shibatani (ed.), 355-86.
Kozinsky, I., V. Nedjalkov, & M. Polinskaya. 1988. Antipassive in
Chuckchee: Oblique Object, Object Incorporation, Zero Object. In
Shibatani (ed.), 651-706.
Kuno, S. 1973. The Structure of the Japanese Language. Cambridge, MA:
MIT Press.
Lee, D.-Y. 2002. A Comparison of Unish Grammar with Esperanto. Journal
of Universal Language 3.2, 57-74.
Libert, A. & C. Moskovsky. 2002. On the Form and Function of Reflexives
in Artificial Languages. Journal of Universal Language 3, 97-132.
______________________. 2003. A Survey of Relative Pronouns and their
Uses in Natural and Artificial Languages. Journal of Universal
Language 4, 61-116.
Maslov, J. 1956. Očerk Bolgarskoj Grammatiki [Sketch of Bulgarian
Grammar]. Moscow: Izdatel’stvo Literatury na Inostrannyx Jazykax.
Nedjalkov, V. & S. Jaxontov. 1988. The Typology of Resultative
Constructions. In Nedjalkov (ed.), 3-62.
Nedjalkov, V. (ed.). 1988. Typology of Resultative Constructions.
 Christo Moskovsky & Alan Libert 149

Amsterdam: Benjamins.
Nicholas, N. & J. Cowan. 2003. What is Lojban? The Logical Language
Group. Available at URL <http://www.lojban.org/publications/level0/
brochure/book1.html>.
O’Connor, C. 1917. American: The New Pan-American Language. Buffalo,
New York: Hansauer-Jones.
Palmer, F. 1994. Grammatical Roles and Relations. Cambridge: Cambridge
University Press.
Peranteau, P., J. Levi, & G. Phares. (eds.). 1972. The Chicago Which Hunt.
Chicago, IL: Chicago Linguistics Society.
Pleyer, M. 1990. Unitario per Medio de Pictoroyn. Bensheim: Unitario
Press.
Ringbom, H. 1987. The Role of the First Language in Foreign Language
Learning. Clevedon: Multilingual Matters.
Rosenblum, B. 1935. Fitusa: Die Geburt einer Sprache. Self-pulbihsed,
Basel.
Shibatani, P. (ed.). 1988. Passive and Voice. Amsterdam: Benjamins.
1985. Passives and Related Constructions. Language 61, 821-48.
Shopen, T. (ed.). 1985. Language Typology and Language Description.
Cambridge: Cambridge University Press.
Siewerska, A. 1988. The Passive in Slavic. In Shibatani (ed.), 243-89.
Song, N. 1987. Empathy-based Affectedness and Passivization.
Transactions of the Philological Society 1987, 74-89 (cited in Palmer
1994).
Talmey, M. 1925. Arulo: Text Book of the Universal Language. New York:
Ilo Press.
Wald, M. 1909. Pan-kel (4th edition). Self-published, Gross-Beeren.
Weferling, E. 1974. Standard-Gramatika del International Auksiliari
Lingue (6th edition). Self-published, Braunschweig.
Weisbart, J. 1972. Europal (Europäisch): Kurze Grammatik und
Begrundung. Berlin: Michaelis & Neumann.
Wennergren, B. 2003. PMEG: Plena Manlibro de Esperanta Gramatiko
(version 12.2). Available at URL < http://bertilow.com/pmeg/>.
Williams, M. 1989. Vygotsky’s Social Theory of Mind. Harvard
Educational Review 591, 108-126.