http://seelrc.

org/glossos/
The Slavic and East European Language Resource Center [email protected]

Greville G. Corbett
University of Surrey

Agreement in Slavic*
1. Introduction

Agreement in Slavic has attracted and challenged researchers for many years. Besides numerous theses
and articles in journals and collections on the topic, there are also several monographs, usually devoted
to a single language, sometimes broader in scope.1 One aim is to give a synthesis of this research,
demonstrating both the complexity of the topic and the interest of some of the results (section 2). Such
a synthesis is complicated by the liveliness of current work, which is both deepening our
understanding of the scale of the problems and trying to bring formal models closer to being able to
give adequate accounts of well-established phenomena. A further aim, then, is to outline this current
work (section 3). Finally the paper suggests a prospective of promising and challenging directions for
future research, some which arise naturally from the directions of earlier and current work, some
which are less obvious, depending on cross-disciplinary collaboration (section 4). As preparation for
the main sections, we first consider the terms we require and the advantages which the Slavic family
provides for research on agreement.

*
The support of the ESRC (grants R000236063 and R000222419) and of the ERC (grant ERC-2008-AdG-230268
MORPHOLOGY) is gratefully acknowledged. I also wish to thank Dunstan Brown, Iván Igartua and the participants at the
workshop “Comparative Slavic Morphosyntax”, especially Wayles Browne, for comments on an earlier version. This
overview was prepared for publication after the Workshop and has been updated since; I thank Claire Turner for help in the
preparation of the revised version. Newer work is referenced in the final paragraph; I apologize for any inadvertent
omissions. Glossing is used to clarify the point at issue (it is not full glossing), following the conventions of the Leipzig
Glossing Rules (http://www.eva.mpg.de/lingua/resources/glossing-rules.php) with the following abbreviations:
ACC accusative N neuter
AUX auxiliary NEG negative
DAT dative NOM nominative
DEF definite article NM.PERS non-masculine personal
DU dual PL plural
EMPH emphatic particle PST past
F feminine POSS possessive adjective suffix
FUT future Q question particle
GEN genitive REFL reflexive
INS instrumental SG singular
M masculine 1 first person
M.PERS masculine personal 2 second person
n total number of examples 3 third person
1
These are Popov (1964), Vanek (1970), Skoblikova (1971), Crockett (1976), Corbett (1979a, 1983a), Iomdin (1990) and
Schmidt and Lehfeldt (1995). I have made available a fuller bibliography on the topic, at:
http://www.surrey.ac.uk/LIS/SMG/projects/agreement/agreement_bib_unicode.htm#Slavonic.

© 1999-2010 The Slavic and East European Language Resource Center

Glossos is the registered trademark of Duke University. All rights reserved.
 GREVILLE G. CORBETT 2

1.1. Basic definitions

Defining agreement is difficult (but see recent work noted in section 3.1 below). As Anderson (1992:
103) remarks: ‘this is a quite intuitive notion which is nonetheless surprisingly difficult to delimit with
precision’. Steele points to the ‘systematic covariance between a semantic or formal property of one
element and a formal property of another’ (1978: 610). The essential notion is the covariance or
matching of feature specifications between two separate elements, such as subject noun phrase and
verb. We should then clarify whether the determination of the form of anaphoric pronouns falls within
agreement. Barlow (1988/1992: 134-52; 1991) reviews the literature and concludes that there are no
good grounds for distinguishing between agreement and antecedent-anaphora relations. Indeed, most
mainstream work on agreement uses the term in this wider sense, to include pronouns, and we shall do
so too. We shall call the element which determines the agreement (say the subject noun phrase) the
controller. The element whose form is determined by agreement is the target. The syntactic
environment in which agreement occurs is the domain of agreement. And when we indicate in what
respect there is agreement, we are referring to agreement features (or categories). Thus number is an
agreement feature, it has the values singular, dual, plural and so on. As these terms suggest, there is a
clear intuition that agreement is asymmetric (a point taken up in section 2.4.1).

1.2. Variation across the Slavic family

We should consider briefly how Slavic agreement appears from a typological vantage, and then
consider the range of possibilities within the family.

1.2.1. Agreement in Slavic from a typological viewpoint

From a broad typological viewpoint, the Slavic languages are very similar, showing the characteristics
one would expect from conservative Indo-European languages. This is particularly true in respect of
agreement, where they show just the traits one might consider typical from traditional works on
agreement: obligatory agreement with the subject, agreement in gender and number within the noun
phrase, and so on (a summary is given at the beginning of section 2). There is an extensive literature
on agreement in Slavic, including several monographs (see the bibliography mentioned in footnote 1),
and yet there is a good deal still to be done; suggestions for future directions are given in section 4. In
fact, since the basics are well established, and moderately well understood, this makes the Slavic
languages a good basis for taking our understanding of agreement further. Particularly since they
provide the linguist with almost ‘laboratory conditions’.

1.2.2. ‘Laboratory conditions’

 AGREEMENT IN SLAVONIC 3

The challenge of humanities research is analysis in the face of an overwhelming number of variables.
We shall see that in agreement, there are indeed substantial numbers of factors which can vary
independently. Typically scientists then can take two approaches. One is idealization — eliminating
some of the variables: we may deal with one language only, with a particular structure, with a
particular genre. The second approach is to control for particular variables. The chemist can quite
directly manipulate the variables in experiments. For linguists this is possible to some extent (in the
way we choose the speakers we investigate, for example). An alternative approach here is to look for
situations in which experimental conditions are provided for us (Pateman 1987: 258); for instance,
sometimes a set of related dialects will provide variation in a particular factor of interest, so that we
can see the results of manipulating a particular variable ‘on the ground’.
The Slavic languages provide a remarkable laboratory, in two ways. First in terms of their
structures. The categories we are discussing here show slight variation from language to language: if
we find that a particular factor or variable is particularly important, then it may well be possible to find
a Slavic language which has this factor and another language, for comparison, which does not. The
second way in which the Slavic languages provide near laboratory conditions is in terms of their
status. We have those with millions of monolingual speakers at one end of the scale, and at the other
end we have Sorbian, all of whose adult speakers are bilingual (the great majority are speakers of
Upper Sorbian, while the position of Lower Sorbian is grave).

2. The ‘State of the Art’

We shall review the established generalizations which hold across the Slavic family. Typically we find
agreement within the noun phrase in number and gender. Finite verbs generally agree with their
subject in person and number. Past tenses are frequently formed with the so-called l-participle, which
creates a more interesting situation: here the auxiliary verb shows agreement in person and number,
while the participle shows agreement in number and gender. Some Slavic languages, such as Russian,
use a null form for the verb ‘be’ in its present tense, so that the former participle is the sole form in the
past tense; it may be said therefore that Russian verbs agree in person and number in the present, but in
number and gender in the past. Various types of pronoun, including the relative pronoun, also show
agreement with their antecedents, in number and gender.
The description given so far covers a large proportion of the instances of agreement in Slavic —
the more straightforward cases. However, there are several instances in which more than one
agreement form may be found. The different constructions have in common a choice between
agreement determined by the form, syntactic agreement, or by the meaning, semantic agreement. As
we shall see, the choice may be influenced by the agreement controller (section 2.1) and by the target
(section 2.2). We show that the various influences on the choice operate independently of each other
 GREVILLE G. CORBETT 4

(section 2.3), before going on to review what is established about agreement features (section 2.4).
And then, in section 2.5, we consider constructions which do not fall readily within the description
given above, and which are surprising typologically.

2.1. The effect of controllers

When there is a choice of agreement, this is normally made possible by the controller. There are
certain controller types (within Slavic and beyond) which regularly permit agreement choices, and we
discuss these in section 2.1.1. And then there are certain factors which range over different controller
types, and which favour one or other agreement choice; we treat these in section 2.1.2. (‘Favouring’ a
particular choice may include favouring it to such a degree that the alternative is excluded in some
circumstances.)

2.1.1. Controller types

Agreement rules are frequently formulated as though a controller’s features were constant, that is, that
all agreements will be identical. In fact, we regularly find agreement choices: a given controller allows
two (occasionally three)2 agreement possibilities. It is important to be clear about the possible
meanings of choice here. The Russian noun para ‘couple’, as we shall see in section 2.2.1.1, takes
feminine singular and plural agreements. In one sense, then, it allows a choice of agreement, in a way
that a noun like kniga ‘book’ does not. However, for any given target the form is determined; the
speaker does not have a choice. Other controllers, like Russian vrač ‘doctor’ when denoting a woman,
allow two possibilities for the same target type, thus the predicate may be masculine or feminine. For
both controllers simple-minded agreement rules would be inadequate.
The choices arise from a mismatch of semantic and formal properties of the controller. The
controller may have the semantics expected of a particular number or gender but a form which is
normally associated with a different specification. Controllers which allow agreement choices may be
classified as in Table 1.

Table 1. Types of controllers which induce agreement mismatches

controller type example

lexical item(s) Serbo-Croat deca ‘children’
lexically restricted construction masculine nouns quantified by
numeral ‘two’, ‘three’ or ‘four’ in
Serbo-Croat
construction conjoined noun phrases

2
Instances where there are three possibilities are discussed briefly in section 4.1.4.
 AGREEMENT IN SLAVONIC 5

This classification turns out to be too crude, as we shall see in the following sections. There are indeed
individual lexical items which induce agreement choices, like Russian para ‘couple’, which denotes a
plurality but has the form of a singular noun. There are also large sets of such items, for instance
Russian nouns like Russian vrač ‘(female) doctor’, that is nouns for professions, denoting women but
having the morphology of masculine nouns. Then there are constructions which are lexically
restricted; this is clear for the Serbo-Croat example given in the table. There are also more open-ended
quantified expressions. Next there are constructions like the associative where the head noun may be
drawn from a large subset of those denoting humans, hence it is lexically restricted but the restrictions
are quite generous. Finally there are construction types whose structure invokes agreement options, but
which appear not to be lexically restricted, such as conjoined noun phrases. (Even here, however, we
find that the noun phrases which are conjoined tend to be headed by noun phrases of the same type, all
animate or all inanimate.) Thus the types given in Table 1 represent pointers for the following sections,
but the main conclusion is that controllers which allow agreement choices range all the way from
unique lexical items to open-ended constructions.

2.1.1.1. Lexical items

We might expect that the features associated with a given lexical item in a given use could be stated
just once, and that the problem would then be to formulate appropriate agreement rules. However,
there are several lexical items for which this is not the case, but which allow agreement choices; such
items are known as ‘hybrids’. These may relate to number, or gender, or both, and they arise from a
mismatch between the meaning of the noun and its morphological form.3 An example of a number
mismatch has been mentioned already, namely Russian para ‘couple’, which has the morphology of a
singular, but denotes more than one. As we shall see (section 2.2.1.1), it takes singular agreements,
except of the personal pronoun. Another example of a number mismatch is provided by Old Church
Slavic družina ‘company’ and similar nouns, which most often take singular attributive modifiers and
plurals in other positions (see Huntley 1989: 24-25 for details). For gender mismatches we may take
the Czech děvče ‘girl (colloquial), which takes neuter agreements, except for the personal pronoun,
which may be neuter or feminine (Vanek 1970: 87-88). There are also various honorific titles, which
take feminine and masculine agreements in Polish, neuter and feminine in Russian. A considerably
researched type of controller is Russian nouns like vrač ‘doctor’, when denoting a female. Since such
nouns have the morphology typically associated with masculines, but denote females, a complex
pattern of masculine and feminine agreements occurs (Corbett 1991: 183-184, 231–232 and sources

3
For additional data and examples of items discussed in this section see Corbett 1983a; for titles pp. 23-24, for Serbo-Croat
gazda ‘landlord, master, boss’ pp. 14-17 and references there, for Serbo-Croat deca ‘children’ pp. 76-88, for Russian
značitel´noe lico ‘important person’ pp. 25-26, and for vrač ‘(female) doctor’ pp. 30–39.
 GREVILLE G. CORBETT 6

there). And then there are nouns which show gender mismatches in the plural: Serbo-Croat gazda
‘landlord, master, boss’ and similar nouns, which are now established as masculine in the singular, but
which allow masculine and feminine agreements in the plural; and Polish nouns like łajdak ‘wretch’,
which take a combination of non-masculine personal and masculine personal agreements (Corbett
1991: 233-236 and references there). A truly remarkable instance is Serbo-Croat deca ‘children’ which
takes feminine singular, neuter plural and masculine plural agreements. All these items show patterns
of agreement which are in accord with the Agreement Hierarchy (section 2.2.1.1).
Two general points are worth noting. First, these examples may comprise individual lexical items
(even single items in the use of a particular individual, as in the case of the special agreements found
with značitel´noe lico ‘important person’ by Gogol´), or relatively large numbers of nouns, as in the
case of nouns like Russian vrač ‘(female) doctor’. In the latter situation, though the system of
agreements may be the same, we must not assume that the actual frequency of the different options
will be the same from item to item. Quite the opposite: there is evidence that vrač ‘(female) doctor’
and buxgalter ‘(female) accountant’ behave rather differently. And second, while the reason for these
agreement choices is to be found in a mismatch between semantics and morphology, such a mismatch
is not a sufficient condition for an agreement choice. Thus Russian djadja ‘uncle’ (like similar nouns)
denotes a male but belongs to the morphological class whose members are usually feminine. The
semantics overrides the morphology, such that the noun is straightforwardly masculine; for agreement
purposes it behaves just like otec ‘father’.

2.1.1.2. Honorifics

The use of address forms is well known, and there is an extensive literature. When we have a pronoun
like Russian vy ‘you’ used in this way, there are interesting agreement effects. Since the pronoun is
plural, it takes some plural agreements; as shown by the verb in this Russian example:

(1) vy xot-ite ...

2PL want-2PL
‘You (polite) want ...’

However, the pronoun is being used to address a single individual, and some singular agreements are
found (as usually in the Russian long form adjective):

(2) vy molčaliv-aja
2PL silent-SG.F
‘You (polite) are silent’

A comparison of the agreement patterns in the different Slavic languages will be given in section
2.2.1.2. In a sense this is still an example of an agreement choice being lexically determined, but it is a
 AGREEMENT IN SLAVONIC 7

special use of the lexical item. We can extend the claim about honorifics, in that not only pronouns are
affected. This latter usage is syntactically rather different and will be considered in section 2.1.1.5.

2.1.1.3. Quantifier phrases, especially numeral phrases

In this section we examine a type of phrase where the choice of predicate agreement in Slavic shows
considerable variation. We shall concentrate here on numeral phrases; data on other quantifiers in
Russian will be given in Table 14 below. Often two forms are possible. This is more like a
construction than the instances considered in sections 2.1.1.1 and 2.1.1.2, but the actual quantifier
involved has a dramatic effect on the agreement found. At this point we shall continue discussing
controller types. (The choice is also affected by controller factors: animacy of the subject, and its
position relative to the predicate; for now we abstract away from those and concentrate on the
substantial influence of the quantifier.) Table 2 gives data on predicate agreement with noun phrases
headed by various numerals in the different Slavic languages.

Table 2. Predicate agreement with numeral phrases in Slavic

‘2’ ‘3’ ‘4’ ‘5’-‘10’ ‘100’

West Slavic:
Czech PL PL PL sg sg
Slovak PL PL PL PL/sg sg
Sorbian DUAL PL PL PL/sg sg
Polish 99% PL 91% PL ||100% PL 7% PL
(n=123)4 (n=43) (n=15) (n=68)
South Slavic:
Old Church Slavic DUAL PL PL (PL)/sg
Bulgarian PL PL PL PL PL
Macedonian PL PL PL PL PL
Serbo-Croat 97% PL 89% PL 83% PL 7% PL
(n=735) (n=249) (n=133) (n=1161)
Slovene DUAL PL PL sg sg
East Slavic:
Ukrainian 83% PL 79% PL 74% PL 38% PL 21% PL
(n=208) (n=150) (n=34) (n=45) (n=14)
Belarusian 92% PL 78% PL 63% PL 39% PL || 50% PL
(n=219) (n=67) (n=16) (n=49) (n=2)
Russian 86% PL 77% PL 76% PL 50% PL
(n=541) (n=247) (n=68) (n=220)

4
Thus in the texts scanned there were 123 examples of phrases with the numeral ‘two’ controlling predicate agreement, of
which 99% (rounded to the nearest whole number) showed plural agreement.
 GREVILLE G. CORBETT 8

In Table 2, DUAL (where available) and PL(URAL) represent semantic agreement. When a cell has a
single entry (e.g. ‘PL’), this indicates that the form is used in the majority of instances, though not
necessarily all. Thus in Slovene, the plural is normal with ‘three’ and ‘four’, but the singular may be
used in expressions of time. Where we do not have more precise data, these few exceptions are
ignored (time expressions also account for some of the singular forms with ‘two’, ‘three’ and ‘four’ in
other languages). A gap indicates a lack of data. The judgements and statistics in the table come from
Suprun (1969: 175–187) unless otherwise stated, and were discussed in Corbett (1983a: 220–224).
In Slovak, with the numerals ‘five’-‘ten’ the plural is used with masculine personal forms and
otherwise the singular; according to Suprun, exceptions account for less than one per cent of the
examples. Sto ‘100’ takes the singular (Ján Bosák and Ľubomir Ďurovič personal communications).
Sorbian preserves the dual; otherwise agreements are broadly similar to those of Slovak (Suprun
1963a). The Polish figures are calculated from the data given in Suprun (1963b); instances where the
numeral itself is in the genitive are excluded. The final Polish entry is for numerals of all types from
‘five’ up to ‘999’. Suprun also gives seven examples of agreement with phrases with tysiąc ‘1,000’, all
singular. Suprun (1961: 81-86) provides data on Old Church Slavic: for ‘five–ten’, he has ten
examples of singular predicates, six of plural predicates and two where one source has singular
agreement and another has plural. According to Večerka (1960: 197), however, in the Gospels the
singular is used in the overwhelming majority of instances, hence the plural entry is bracketed. Suprun
also gives an example with sъto ‘100’, one with tysęšta ‘1,000’ and one with tьma ‘10,000’; all three
have singular agreement. The Serbo-Croat statistics are taken from Sand (1971: 51-52, 73); the figure
for dva ‘two’ includes examples with oba ‘both’; ‘two’, ‘three’ and ‘four’ include compound numerals
ending in ‘two–four’ and the remaining figure is for all other numerals above ‘four’. Slovene
judgements are from Vincenot (1975: 196) and Suprun (1969, 176). The final entry for Ukrainian
includes examples with sorok ‘40’ as well as sto ‘100’.
Let us turn to the pattern revealed by Table 2. The South Slavic languages Bulgarian and
Macedonian differ from the others in using the plural with all numerals above ‘one’ in almost all
instances. Other Slavic languages use both singular and plural. The remaining South Slavic languages
(Old Church Slavic, Serbo-Croat and Slovene) use the dual (when available) with phrases with the
numeral ‘two’; otherwise they show a strong preference for plural agreement for quantified phrases
with the numerals ‘two’-‘four’, and for singular agreement (though with varying degrees of tolerance
towards the plural) with numerals from ‘five’ and above. In several languages the distinction between
‘two’-‘four’ on the one hand, and ‘five’ upwards on the other, is fairly sharp. However, the statistics
for Serbo-Croat and Polish show that here the division is not absolute. The overall picture is clear: the
higher the numeral the more likely is singular agreement. The form which is semantically justified
becomes more likely the lower the numeral. This is clearly true in the straightforward cases like
 AGREEMENT IN SLAVONIC 9

Slovak. The statistical data too support this claim, apart from two minor inconsistencies (indicated in
the table with ||). These two cases need not concern us as the sample size for both numerals is small.
Even apart from these, it is not the case that there is a statistically significant difference between every
pair of successive numerals in each language. However, statistical advice is that the pattern is so
overwhelming that statistical tests of significance are superfluous. Apart from the two exceptions
mentioned, the rank order of the numerals according to the frequency with which they take plural
agreement is the same in the different languages and this order is inversely related to numerical value.
Given that the lower the numeral is, the more likely it is to take semantically justified agreement,
why should this be? The groups which we quantify with larger numbers are the groups which are less
individuated and conversely are more likely to be viewed as a unit. For this reason they are more likely
to be encoded grammatically as a noun. And as a result, when there is a grammatical choice, the higher
are more likely to be treated somewhat more like nouns. Russian četyre knigi ‘four books’ is ‘more
plural’ than pjat´ knig ‘five books’, and in a sense tri knigi ‘three books’ is ‘more plural’ than četyre
knigi ‘four books’; we are better able to individuate three items than four (Corbett 2000: chapter 6). An
extended discussion of possible structures giving rise to these agreement options with some numerals
in different Slavic languages can be found in Franks (1995: 93-219); he adopts a GB approach,
drawing on Pesetsky (1982) and Babby (1987).

2.1.1.4. Associatives

An interesting construction variously called the ‘group plural’, ‘representative plural’ and the
‘associative’ (the term we shall used) has been highlighted by Moravcsik (1994). Typically we find
forms consisting of a nominal plus a marker, which denote a set comprised of the referent of the
nominal (the main member) plus one or more associated members: Hungarian János-ék (János-
ASSOCIATIVE) ‘John and his family/friends’. Slavic has no special marker of this type but some
languages allow the use of the ordinary plural morphology with this same effect. A particularly
interesting case for us is found in the Talitsk dialect of Russian (Bogdanov 1968). In this dialect, a
plural verb can be used with a singular noun phrase, to indicate reference to a person or persons
besides the one indicated directly. That is to say, the associative is marked not by a marker on the
nominal, but by plural agreement:

(3) Góša pr´ijéxal´i !

Gosha arrived.PST.PL5
‘Gosha and his family have arrived’

5
Bogdanov’s transcription has been transliterated in these examples.
 GREVILLE G. CORBETT 10

This was used when the named person arrived with his wife and children; the fact that more than just
one person is involved is shown in this dialect exclusively by the agreement. (Bogdanov makes clear
that these examples are not examples of honorific usage, and that there is a singular/plural opposition
in examples like those given, the singular having the straightforward non-associative meaning.) This
plural agreement does not extend into the noun phrase, and so conflicting agreements can be found in
the same sentence:

(4) moj brat tam tóža žýl´i

my.SG brother[SG] there also lived.PL
‘my brother and his family also lived there’

The question of different possible agreements with the same controller is one we return to in section
2.2.1.1.
Bogdanov (1968: 69-70) points out the possibility of this construction with a noun phrase headed
by a third person pronoun:6

(5) on tútyka fp´ir´ót nas pr´ijéxal´i

he here before us arrived.PL
‘he and his family came here before we did’

Thus again the range of possible controllers is quite large.

2.1.1.5. Respected nouns

As noted above, plural agreement may be the only indicator of honorific usage, as in this example of a
maid talking in turn of her master and mistress:

(6) <<Mamen´ka plačut, — šepnula ona vsled uxodivšej Elene, a

mother cry.PL whispered she after leaving Elena and

papen´ka gnevajutsja ... >>

father be.angry.PL
“Your mother is crying”, she whispered after Elena, who was leaving, “and your father is angry
...” (Turgenev Nakanune, 1860)

Here the plural verbs with singular subjects indicate that the speaker is showing respect for the people
referred to. This demonstrates that in cases like this the controllers cannot be restricted to particular
lexical items, but that a range of noun phrases may be involved. (For evidence that this construction

6
Here we are interested in the pronoun as the controller of the agreement. Bogdanov does not specifically state that this
type of construction is not found with other pronouns, but the implication (and expectation) would be that such forms
would be impossible. If so, at first sight, this would be a problem for the attempt to integrate constraints on possible
numbers with the Animacy Hierarchy. The problem that associatives appear to pose for a typology of number systems is
dealt with in Corbett and Mithun (1996).
 AGREEMENT IN SLAVONIC 11

follows the constraints of the Agreement Hierarchy, see Corbett 1983a: 24-25, and for sources on the
construction in Belarusian, Czech, Polish, Slovak, Slovene and Ukrainian see 1983a: 41n8.)7

2.1.1.6. Conjoined noun phrases

An agreement controller consisting of conjoined noun phrases may well give rise to an agreement
option. It may allow agreement with both or all the conjuncts, and it may allow agreement with just
one conjunct. The latter type is frequent both in texts and in naturally occurring discourse. When
agreement is with one conjunct it is almost always with the nearest. (As we shall see in section 2.5.3,
some languages, exceptionally, allow agreement with the first conjunct when it is not the nearest.)
Here is a typical example of agreement with the nearest conjunct from Russian:

(7) Teper´ na nej by-l sinij kostjum i novaja

now on her be-PST[SG.M] (dark) blue dress(M)[SG] and new

belaja bluzk-a ...

white blouse(F)-SG.
‘She was now wearing a blue dress and a new white blouse ...’ (Vojnovič)

In this example the verb agrees just with the conjunct sinij kostjum ‘blue dress’, and the nearest
conjunct is also the first. The more significant examples are those where the nearest are first are
distinct. Here is a clear example from Cassubian (Stone 1993: 784):

(8) Odraz-a i strach czierowô-ł jego postępkama

revulsion.(F)-SG and fear(M)[SG] direct-PST[SG.M] his actions
‘Revulsion and fear directed his actions.’

Here again the genders of the nouns make it clear that agreement is with the nearest. The alternative is
for agreement to be with all the conjuncts, as in this Slovene example (Lenček 1972):

(9) Tonček in Marina sta prizadevn-a

Tonček.M and Marina.F be.DU assiduous-DU.M
‘Tonček and Marina are assiduous’

Agreement is with both conjuncts, and the gender and number resolution rules specify the form of the
target as dual and, where appropriate, masculine. We discuss resolution rules in section 2.4.3 below.
For many of the Slavic languages the number resolution rule simply specifies plural, and in some there
7
Wayles Browne points out that the construction is found in Kajkavian Serbo-Croat too (personal communication), as in
the song:
(i) Mamica su štrukl-e pek-l-i
Mummy AUX.PL dumpling-PL bake-PST-PL.M
‘Mummy was baking dumplings.’

Note the masculine plural is used.

 GREVILLE G. CORBETT 12

is no place for gender resolution since gender is not distinguished in the plural. More examples of
agreement with the nearest conjunct and with all conjuncts can be found in section 2.2.1.1.

2.1.1.7. Comitative constructions

This construction together with alternative agreement possibilities is found in some but not all the
Slavic languages. We may illustrate it from Belarusian (Bukatevič et al. 1958: 292):

(10) Dzed z unukam laviŭ rybu

grandfather with grandson catch.PST.SG.M fish
‘Grandfather and grandson were fishing’

(11) Brat z sjastroju pajš-l-i ŭ tèatr

brother with sister go-PST-PL to theatre
‘Brother and sister went to the theatre’

The head noun in the nominative case may control the agreement ((10) — syntactic agreement) or
there may be agreement with the expression as a whole ((11) — semantic agreement, as determined by
resolution rules, discussed in section 2.4.3.4). As we might expect, semantic agreement is less likely
with comitative expressions than with conjoined noun phrases, as shown by Russian data.

Table 3. Comparison of agreement with comitative phrases and

conjoined noun phrases in contemporary Russian8

comitative phrases conjoined noun phrases

SG PL %PL SG PL %PL
literature 5 4 44 28 57 67
press 15 15 50 30 723 96

2.1.2. Controller factors

These are factors relating to controllers but which range over different controller types. Two are well
established, and their interaction is also moderately well researched.

2.1.2.1. Animacy

There is a massive amount of evidence, primarily from text counts but also from work with
consultants, that controllers referring to animates are more likely to take semantically justified

8
From Corbett (1983a: 154-155). The literary corpus consists of Panova’s Sputniki (1946) and Nekrasov’s Kira
Georgievna (1961); the figures for the language of the press come from Graudina, Ickovič and Katlinskaja (1976: 31, 346).
There is an unfortunate transposition of headings in my original table: the data are correct above. The difference between
the two constructions is more marked than these figures suggest, since comitatives are almost entirely limited to animate
noun phrases, and animacy, as we shall see, is a factor favouring semantic agreement. If the examples of conjoined noun
phrases were restricted to animates, the percentage plural agreement would be higher.
 AGREEMENT IN SLAVONIC 13

agreement than are those referring to inanimates. The evidence comes from different Slavic languages,
and involves various quantified expressions and conjoined noun phrases (for a survey and sources see
Corbett 1983a: 110-132, 139, 143-146). To give one example, Patton examined a large corpus of 19th
and 20th century Russian literary texts, and a sample from Pravda, for examples of predicate
agreement with quantified subjects. From her data (1969: 35, 63, 148, 160) the following may be
calculated:9

Table 4. The effect of animacy on agreement with quantified expressions in Russian

SINGULAR PLURAL PERCENT PLURAL

ANIMATE 790 1293 62
INANIMATE 1047 740 41

We shall see further evidence for the effect of animacy in section 2.1.2.3. While animacy is
firmly established as a controller factor, there is less work on how it might be subdivided. For instance,
abstracts are less likely to control semantic agreement than are concrete noun phrases, as can be seen
in agreement with conjoined noun phrases (Timberlake 1993: 865).

2.1.2.2. Precedence

There is also strong evidence that controllers which precede their targets are more likely to take
semantically justified agreement than are those which follow (for the argument for treating this as a
controller factor see Nichols, Rappaport and Timberlake 1980, and commentary in Corbett 1983a: 137,
154, 175). Again there is evidence from different Slavic languages, and it involves quantified
expressions, conjoined noun phrases and comitative phrases (Corbett 1983a: 107-150 passim). To give
just one part of the evidence: Sand examined a large corpus of Serbo-Croat texts (literature of the
1960s, non-fiction 1951-1968 and the newspaper Politika 1969-1970). The largest controller type
investigated was the numerals from pet ‘five’ upwards. Table 5 has been drawn up from her data
(1971: 73-75):

Table 5. The effect of precedence on agreement with quantified expressions

(involving ‘5’ and above) in Serbo-Croat

SINGULAR PLURAL PERCENT PLURAL

subject-predicate 249 61 20
predicate-subject 830 21 2

9
The effect of the calculation is to combine her main and secondary corpus and to eliminate examples with plural
determiners (such as èti ‘these’, which make a plural predicate obligatory).
 GREVILLE G. CORBETT 14

There will be further evidence for the effect of precedence in the next section 2.1.2.3. While it is
easiest to show the effect of precedence in subject-predicate domains, with different types of controller
subjects, it is also relevant to controllers of attributive modifiers.

2.1.2.3. The interaction of animacy and precedence

We have seen that controllers which refer to animates are more likely to take agreement forms with a
greater degree of semantic justification than are those referring to inanimates. Similarly, controllers
which precede their targets are more likely to take agreement forms with a greater degree of semantic
justification than are those which follow. Since these two controller factors are independent, we can
cross-classify for them. Table 6 records agreement with a set of quantifiers in a selection of Russian
literary texts of the last two centuries (details in Corbett 1983a: 150-153.)

Table 6. Predicate agreement with quantified expressions in Russian

ANIMATE INANIMATE
SG PL %PL SG PL %PL
subject-predicate 11 48 81 21 20 49
predicate-subject 24 23 49 70 18 20

Both animacy and precedence exert a major influence on the agreement form selected. The plural, the
form with greater semantic justification, is more likely if the subject is animate and if it precedes the
predicate. With both factors exerting an influence, the likelihood of semantic agreement is greatest,
with neither factor it is lowest, and with just one it falls in the middle.
As similar pattern, though with different percentages, can be found with conjoined noun phrases.
In table 7, the data are taken from modern literary texts, from Russian (1930-1979) and from Serbo-
Croat (a corpus of short works by Ivo Andrić).10

Table 7. Predicate agreement with conjoined noun phrases

subject type ANIMATE INANIMATE

word order SG PL % PL SG PL %PL
subject-predicate Russian 0 115 100 10 57 85
Serbo-Croat 0 21 100 3 32 91
predicate-subject Russian 14 75 84 82 32 28
Serbo-Croat 7 16 70 46 16 26

It is evident that both factors favour resolution. When both are present, Russian and Serbo-Croat (in
the samples here) require the plural form (which is the resolved form, the semantically justified form).

10
Details in Corbett (1983a, pp. 105-35 especially p. 130 on Russian, and pp. 139-140 and p. 101 on Serbo-Croat).
 AGREEMENT IN SLAVONIC 15

When either one of the factors is present, the plural form is found in a significantly higher proportion
of the cases than when neither is present. In Russian the two factors are of about equal weight, and in
Serbo-Croat precedence appears to be the more important factor. Thus both animacy and precedence
have a substantial effect on agreement choices, ranging over different controller types.

2.2. The effect of targets

As we have seen, particular controllers make agreement choices possible, with varying degrees of
openness to the choice. And controller factors, which apply to different types of controller, exert a
clear influence on the choice. We now investigate the considerable impact which targets also have on
that choice. First we consider the different target types we can distinguish (section 2.2.1) and then we
look at target factors which range over the different types (section 2.2.2).

2.2.1. Target types

We shall consider target types in terms of two hierarchies, starting from the higher level and moving to
the more detailed.

2.2.1.1. The Agreement Hierarchy

We begin with the largest syntactic domains, and here agreement options are constrained by the
Agreement Hierarchy (Corbett 1979b; 1983a: 8-41, 81–86; 1991: 225-241). This hierarchy was
proposed on the basis of data from a range of languages, but the strongest evidence comes from
detailed analysis of Slavic. Four types of agreement targets can be distinguished:

attributive < predicate < relative pronoun < personal pronoun

Figure 1. The Agreement Hierarchy

Possible agreement patterns are constrained as follows:

For any controller that permits alternative agreement forms, as we move rightwards along
the Agreement Hierarchy, the likelihood of agreement forms with greater semantic
justification will increase monotonically (that is, with no intervening decrease).

As an illustration of the type of data covered by the Agreement Hierarchy, consider agreement with
numeral phrases in Serbo-Croat involving the numerals ‘two’, ‘three’ and ‘four’. These require a
special form of masculine nouns, a survival of the dual number which is synchronically equivalent to
the genitive singular. Attributive modifiers to such nouns must take the ending -a; it has been argued
 GREVILLE G. CORBETT 16

that it should be analysed synchronically as a neuter plural.11 However it is analysed, this -a form
represents syntactic agreement.

(12) dva dobr-a čovek-a

two good-PL.N men-SG.GEN
‘two good men’

In the predicate the neuter plural form (syntactic agreement) and the masculine plural form (semantic
agreement) are both possible:

(13) ov-a dva čovek-a su dobr-a/dobr-i

this-PL.N two men-SG.GEN be.PL good-PL.N/good-PL.M
‘these two men are good’

The relative pronoun is also found in both forms:

(14) dva čovek-a koj-a/koj-i ...

two men-SG.GEN who-PL.N/who-PL.M ...
‘two men who ...’

The personal pronoun must stand in the masculine plural form oni (*ona is unacceptable). We
therefore find syntactic agreement in attributive position, both types of agreement of the predicate and
relative pronoun, and only semantic agreement of the personal pronoun. We can go further, in that
there are figures for the relative frequency of the two forms in the positions where there is an option.
These are derived from Sand (1971: 55-56, 63) and presented in Table 8:

Table 8. Percentage distribution of masculine plural (versus

neuter plural) forms in Serbo-Croat

attributive predicate relative personal

pronoun pronoun
percentage showing plural 0 18 62 100
(semantic) agreement (n = 376) (n = 32)

Table 8 shows a monotonic increase in the likelihood of agreement forms with greater semantic
justification.
We will analyse three further sets of data (from Corbett 1983a: 12-25). First we consider
agreement with Russian para ‘couple, man and woman’. In the following example we find singular
agreement of the attributive adjective, predicate and relative pronoun:

11
Corbett (1983a: 13-14, 89-92). Browne (1993: 373) labels the form on nouns and on agreement targets the ‘234’ form.
Sometimes it is labelled the ‘paucal’; this latter usage it better avoided, since the paucal is a number form parallel to
singular, plural and dual, and is used for a small number. The Serbo-Croat form is not like this; it is not possible to use
čovek-a ‘man-SG.GEN/234’ independently to refer to a small number of men.’
 AGREEMENT IN SLAVONIC 17

(15) ...by-l-a izjaščn-aja vljublenn-aja para, za kotor-oj vse s ljubopytstvom

be- PST-SG.F elegant-SG.F loving-SG.F couple after which-SG.F all with curiosity

sledili i kotor-aja ne skryvala svoego sčast´ja: on tanceval tol´ko s nej,

followed and which-SG.F NEG hid own happiness he danced only with her

i vse vyxodilo u nix tak tonko, očarovatel´no, čto tol´ko odin

and everything turned.out at them so delicately charmingly that only alone

komandir znal, čto èt-a para nanjat-a Lloyd-om igrat´ v ljubov´ za

captain knew that this-SG.F couple employed-SG.F Lloyds-INS to.play at love for

xorošie den´gi i uže davno plavaet to na odnom, to na drugom korable.

good money and already long.time sails now on one now on other ship.

‘there was an elegant loving couple, who everyone watched with curiosity and who did not hide
their happiness: he danced only with her and with them everything turned out so delicately and
charmingly that only the captain knew that this couple was employed by Lloyds for good money
to play at being in love and had already being sailing for some time on different ships.’ (Bunin,
Gospodin iz San-Francisko)

What of the personal pronoun? Here the form is the plural oni, as in (15) but also more clearly in this
example (with the derived form paročka ):

(16) Krome Mariny, avtobusa dožida-l-a-s´ kakaja-to paročka.

besides Marina for bus wait-PST-SG.F-REFL some.SG.F couple

Im, kak vidno, bylo vse ravno — pridet avtobus ili ne pridet.
3PL.DAT as evident was all equal will come bus or NEG come.FUT.3SG

‘Besides Marina, there was a couple waiting for the bus. It was evidently all the same to them
whether the bus came or not.’ (Laskin, Kak togda)

Thus we find syntactic agreement in all positions except the one on the extreme right of the
hierarchy.12
Consider now conjoined noun phrases. In attributive position in Russian we normally find
singular agreement:

(17) Èt-a vzyskatel´nost´, samokritičnost´ tože raspolaga-l-i k nemu

this-SG.F exactingness(F) self-criticalness(F) also dispose-PST-PL to him
‘This exactingness and self-criticalness also disposed me favourably towards him.’
(Černov, Introduction to Smol´janinov, Sredi morennyx xolmov)

12
Examples are scarce, however. The Uppsala corpus has just two relevant examples; the agreements are as described
above.
 GREVILLE G. CORBETT 18

However, the plural is also possible:

(18) Marija zadumalas´ ob ostavlenn-yx muže i dočeri: kak oni

Maria thought about left.behind-PL husband and daughter how 3PL.NOM

tam, čto s nimi?

there what with 3PL.INS?
‘Maria thought about the husband and daughter she had left behind, and wondered how they
were and what was happening to them.’ (Maksimov, Karantin)

There is an interaction of controller factors here; as we saw in section 2.1.2.1, the fact that the
controller is animate is a factor in favour of semantic agreement. Overall (taking animate and
inanimate together) in attributive position the singular is more likely; in a sample of literary prose
(n=44) 14% of the examples had plural agreement.
In the predicate, again both forms are found. Example (19) shows singular agreement, and plural
agreement is illustrated in (17) above, and in (20):

(19) By-l-a u nego ešče gitar-a i samoučitel´ k nej

be-PST-SG.F at him also guitar(F)-SG and manual(M)[SG] for it
‘He also had a guitar and a manual for it.’ (Vojnovič, Putem vzaimnoj perepiski)

(20) Pečk-u mne počini-l-i brigadir i muž

stove-ACC 1SG.DAT mend-PST-PL brigade.leader and husband

kladovščic-y, kotor-ym ja postavi-l za èto vypit´.

storekeeper-GEN who-PL.DAT I provide-PST[SG.M] for this to.drink
‘The stove was mended by the brigade leader and the husband of the storekeeper, and in return I
provided them with drinks’ (Amal´rik, Neželannoe putešestvie v Sibir´)

Here the plural is more likely; again using a sample of literary prose (n=290) 71% of the
examples showed plural agreement (as in (17) and in (20)). Relative pronouns are almost always plural
in this construction, as in (20). However, there are rare examples of singulars:

(21) Kazalos´, on skol´zit po žizni s toj že stremitel´nost´ju i

seemed he glides through life with that very swiftness(F) and

neprinuždennost´ju, s kak-oju pero ego skol´zit po bumage.

ease(F) with which-SG.F.INS pen his glides over paper
‘It seemed he glided through life with the same swiftness and ease with which his pen glides
over paper.’ (Adamovič, Introduction to Nabokov, Zaščita Lužina)

The personal pronoun (as in (18) above) is normally in the plural.

There is more to be said about conjoined noun phrases (section 2.4.3). At this point the important
thing is that conjoined noun phrases in Russian allow agreement options but in accord with the
Agreement Hierarchy.
 AGREEMENT IN SLAVONIC 19

The last topic is one we have touched on already, namely associative plurals. We discussed
example (4) from the Talitsk dialect of Russian, repeated for convenience as (22):

(22) moj brat tam tóža žý-l´-i

my[SG] brother[SG] there also live-PST-PL
‘my brother and his family also lived there’

Here we find singular (syntactic) agreement of the attributive modifier and plural agreement in the
predicate. No examples of relative pronouns are given by Bogdanov, however, the personal pronoun in
this construction is plural (Bogdanov 1968: 71):

(23) Pra Kuz´mú my šypka ab´is´n´ít´ tóža n´e móžym, paš´imú on´í
about Kuz´ma 1PL.NOM much explain also NEG can because 3PL.NOM

n´e p´íšut vam

NEG write 2PL.DAT
‘We can’t tell you much about Kuz´ma either, because they don’t write to you.’

Again agreements in this construction adhere to the hierarchy. The data from these different
constructions are summarized in Table 9. This constitutes only a small part of the supporting evidence
(see Corbett 1979b, 1983a: 8-41, 1991: 225-260; Leko 1986: 203-212, Huntley 1989 for further data).

Table 9. The Agreement Hierarchy: evidence from four controllers and controller types

attributive predicate relative pronoun personal

pronoun
‘two’-‘four’ plus neut pl neut pl/(MASC PL) (neut pl)/MASC PL MASC PL
noun in Serbo-Croat [MASC PL 18%] [MASC PL 62%]
Russian para sg sg sg PL
Russian conjoined sg/(PL) (sg)/PL [PL 71%] (sg)/PL PL
NPs [PL 14%]
Associative
construction sg PL no data PL
(Russian dialect)

Note: Lower case indicates syntactic agreement, and upper case SEMANTIC AGREEMENT;
parentheses indicate a less frequent variant.
It can be seen that this wide variety of agreement options is indeed constrained by the Agreement
Hierarchy.

2.2.1.2. The Predicate Hierarchy

We now focus on the predicate, which was one position on the Agreement Hierarchy. In a paper
drawing largely but not exclusively on Slavic data, Comrie (1975) showed how honorific plural
 GREVILLE G. CORBETT 20

pronouns may take singular or plural agreement, but that this variation is constrained by what I shall
call the ‘Predicate Hierarchy’:

verb < participle < adjective < noun

Figure 2. The Predicate Hierarchy

Reformulating Comrie’s proposal we may claim that:

For any controller that permits alternative agreement forms, as we move rightwards along
the Predicate Hierarchy, the likelihood of agreement forms with greater semantic
justification will increase monotonically (that is, with no intervening decrease).

In subsequent research I investigated evidence for all the Slavic languages, for agreement with
honorific pronouns, and the results are given in summary form in Table 10.

Table 10. Agreement with honorific vy in the Slavic Languages

finite verb participle adjective noun

West Slavic:
Czech pl (pl)/SG (pl)/SG SG
Slovak pl pl/(SG) SG SG
Lower Sorbian pl pl pl/SG SG
Upper Sorbian pl (pl)/SG (pl)/SG SG
Polish dialects pl pl/SG pl/SG SG
South Slavic:
Bulgarian pl pl [96%] SG [97%] SG
n=167 n=163
Macedonian pl pl (pl)/SG SG
Serbo-Croat pl pl pl/(SG) SG
Slovene pl pl/(SG) pl(SG) SG
East Slavic:
Ukrainian pl pl/(SG) (pl)/SG SG
Belarusian pl pl SG SG
Russian pl pl short form long form SG13
pl [97%] SG [89%]
n=145 n=37

Note: lower case indicates syntactic agreement, and upper case SEMANTIC AGREEMENT.
Parentheses indicate less frequent or less preferred variants. The sources of the percentage
figures are given in Corbett (1983a: 42-59).

13
Though this is rare, even predicate nominals may be involved. This appears to have been found in nineteenth century
Russian, in the speech of the less educated:
(i) Izmennik-i vy, čto li?
traitor-PL 2PL.NOM that Q
‘Are you a traitor then?’
(Čexov, Xolodnaja krov´ 1887; quoted in Vinogradov and Istrina 1954: 520)
 AGREEMENT IN SLAVONIC 21

Again there is great variation, but the overall pattern is very clear and it is fully in accord with the
Predicate Hierarchy.14

2.2.2. Target factors

Target factors are those which range over different target types. There are four cases to discuss.

2.2.2.1. The Predicate Hierarchy of Individuation

It has been known for some time that predicate type has a role in influencing agreement choices.
However, it is not clear how this fits with the other factors we have considered. The clearest data come
from Robblee (1993b), so we will look at her results first and then return to the question of how they
relate to the Predicate Hierarchy just discussed. Robblee reports that predicates form a hierarchy of
individuation, which she motivates from other phenomena as well as agreement, including the genitive
case marking of subjects in negated sentences (Robblee 1993a). There are three main classes, each
split into two; the reader is referred to Robblee’s work for justification of these, but the examples in
Table 11 give an indication of membership. The six subtypes represent increasing degrees of inherent
individuation of the predicate.15 “A predicate of low inherent individuation may be attributed to and
thus occurs with many more kinds of arguments than a predicate of high individuation. For instance,
the predicate byt´ ‘be’ regularly occurs with subject noun phrases that are abstract, and also with those
that are concrete. In contrast, only noun phrases denoting concrete objects normally occur as the
subject of the stative predicate krasnet´ ‘redden [intrans.]’.” (Robblee 1993b: 425).
The question is whether this hierarchy of predicates is relevant to agreement. Robblee took a
corpus of eight works of Russian prose published from 1976 to 1988. She extracted instances of
predicate agreement with quantified noun phrases including either a numeral or one of neskol´ko
‘several’, malo ‘few’ or nemalo ‘several, more than a few’ (for details see Robblee 1993b), giving 373
relevant examples. The results are given in Table 11.

14
Naturally we should consider how the two hierarchies combine. This is not straightforward: the Predicate Hierarchy
forms a sub-hierarchy within the Agreement Hierarchy. For data and discussion see Corbett (1983a: 76-93).
15
Individuation as a controller factor came up in section 2.1.1.3.
 GREVILLE G. CORBETT 22

Table 11. Agreement with quantified noun phrases according to predicate type
(Robblee 1993b: 428; her percentage figures rounded to whole numbers)

Examples Total Percent

with singular examples singular
Subtype Ia byt´ ‘be’ 71 76 93
Subtype Ib, e.g. proizojti ‘occur’ 42 47 89
CLASS I (“Inversion”) subtotal 113 123 92
Subtype IIa, e.g. stojat´ ‘stand’ 66 122 54
Subtype IIb, e.g. krasnet´ ‘redden’ 16 38 42
CLASS II (“Intransitive”) subtotal 82 160 51
Subtype IIIa, e.g. rabotat´ ‘work’ 4 13 31
Subtype IIIb, e.g. udarit´ ‘hit’ 8 77 10
CLASS III (“Agentive”) subtotal 12 90 13
OVERALL TOTAL 207 373 56

The results are clear; syntactic (singular) agreement is most common with byt´ and successively less
common with more individuated predicates.
The effect of the predicate is substantiated convincingly. However, we need to disentangle the
different factors at work. Thus Robblee’s first class comprises ‘inversion predicates’ (a Relational
Grammar term). Among other properties, these predicates are more likely to appear in predicate-
subject structures than are other predicates; and predicate-subject word-order disfavours semantic
agreement (section 2.1.2.2); it would be helpful, therefore, to have a count in which the factor of word-
order is held constant, in order to isolate the effect of the predicate type. Robblee provides this in a
later paper; she takes the same 373 examples as in Table 11 and cross-classifies her three main
predicate types with word-order.

Table 12. Singular agreement with quantified noun phrases

according to word order and predicate type
(Robblee 1997: 235; her percentage figures rounded to whole numbers)

SV word order VS word order TOTAL VS/SV

CLASS I 11/13 85% 102/110 93% 113/123 92% 1.1
(“Inversion”)
CLASS II 16/43 37% 66/117 56% 82/160 51% 1.5
(“Intransitive”)
CLASS III 2/55 4% 10/35 29% 12/90 13% 7.9
(“Agentive”)

Thus of the class I (inversion) predicates, of the 13 found with subject-verb word-order, 11 (i.e. 85%)
had singular agreement. As we already knew, singular agreement is more like with verb-subject order
than with subject-verb order (and we find this with each class of predicate). But equally, if we keep the
word-order constant and consider the class of predicate, then we see that singular is most likely with
 AGREEMENT IN SLAVONIC 23

inversion predicates, less so with intransitives and least likely with agentives. The effect of word-order
increases as we move down the Predicate Hierarchy of Individuation, as we see n the last column,
derived by dividing the percentage singular agreement with verb-subject sentences by the percentage
with subject-verb sentences.16 Here then we have clear evidence that this hierarchy has an effect
independent of word-order.
The question which remains is how this hierarchy relates to Comrie’s Predicate Hierarchy.
Robblee’s Predicate Hierarchy of Individuation provides a cross-cutting classification, as becomes
clear when we consider non-verbal predicates. A few of these, such as vidno ‘visible’ are in class Ib
(Robblee 1993a: 216), while the majority are lower on the hierarchy (1993a: 230).17 In Comrie’s
Predicate Hierarchy, which has a syntactic and morphological basis, verbs and non-verbs are fully
separated. Thus Robblee’s hierarchy can be seen as a target factor, ranging over the predicate types
defined in Comrie’s hierarchy. It would be of great interest to know more about the interactions
between the two, in particular to know more about how adjectives behave in structures which allow
agreement choices.18

2.2.2.2. Stacking

The remaining target factors are syntactic in nature. Like the last factor discussed, they range over
different target types, but not necessarily all target types. The clear cases of stacking involve
attributive modifiers. As an instance, Serbo-Croat nouns like gazda ‘‘master’, when in the plural,
permit both masculine and feminine modifiers. If we find stacked modifiers, usually both take the
same form. This is not always so, as in this example (Marković 1954: 95):

(24) ov-i privatn-e zanatlije

this-PL.M private-PL.F artisans
‘these private artisans’

Both agreement possibilities are found together. According to Leko (1986: 216) many speakers would
not accept this, preferring ove ‘this-PL.F’. What matters here is that those who do accept different
forms in stacked modifiers have them as in (24), with the form with greater semantic justification, the
masculine, further from the controller (we do not find the reverse: *ove privatni zanatlije). The
constraint is as follows:

16
Robblee’s figures in this column are correct: my rounding of percentages means that checking that column requires
recalculation of the percentages themselves.
17
Overall, predicative nouns and adjectives would be more individuating than transitive verbs, so the most individuating in
the Predicate Hierarchy of Individuation (Karen Robblee, personal communication).
18
It is known that, when other factors are held constant, adjectives favour semantic agreement by comparison with verbs
(Corbett 1983a: 163–170).
 GREVILLE G. CORBETT 24

If stacked targets show different agreement forms, the target further from the controller will
show the form with greater semantic justification

To incorporate a constraint of this type means allowing the agreement shown by one target to be
constrained in part by that of another target. It is claimed that it will apply equally to any stacked
targets, and so would apply to stacked relative clauses. These are rare in Slavic and as yet no relevant
examples have been found.

2.2.2.3. Parallelism

This constraint clearly applies to different target types. Two targets are said to be parallel when they
fill the same syntactic slot in relation to the same controller. Normally we find the same agreement
form for both, but this is not always the case, as in this example from Serbo-Croat:

(25) Sarajlije su igral-e bolje i gotovo potpuno dominira-l-i terenom.

Sarajevans AUX.3PL played-PL.F better and almost completely dominate-PST-PL.M field
‘The Sarajevans played better and dominated the field almost completely.’
(Oslobođenje 27.2.1953, quoted by Marković 1954:96)

The two verbal predicates are parallel. The controller of both is Sarajlije ‘Sarajevans’, a noun like
gazde ‘landlords’. The nearer target shows syntactic agreement while the further shows semantic
agreement. The following constraint applies:

If parallel targets show different agreement forms, then the further target will show
semantic agreement.

This constraint ranges over different target types; for further examples of predicates and an example of
relative pronouns see Corbett (1983a: 71-74). An alternative way of looking at parallel targets is to say
that they are minimally stacked, and so this constraint is a sub-case of that in the previous section.
These latter two constraints operate at sentence level; they refer to the simultaneous presence of
two targets. They are linked to a more general corpus level regularity which is the effect of ‘real’
distance.

2.2.2.4. ‘Real distance’

It has been claimed, more from data from outside Slavic than from within, that for any particular target
type, the further it is removed from its controller, the greater the likelihood of semantic agreement.
The examples almost always concern targets following the controller. Thus a target moved increasing
far after its controller becomes increasingly more likely to take semantic agreement. In the other
direction the position is not clear (data are harder to find) but the reverse effect is likely. It appears
then that greater ‘real distance’ accentuates the effect of the word-order factor.
 AGREEMENT IN SLAVONIC 25

2.3. Independence of influences

We have seen that the factors that range over controller types are independent of those types; they can
in principle affect each of them.19 Even within controller types, variation is independent of the
controller factors: for instance, the difference in the likelihood of semantic agreement with different
quantifiers is independent of the factors of animacy and word-order (some evidence is reported in
Corbett 1986: 1016 and in Robblee 1997: 237). Equally target types and factors are independent of
each other. Finally, influences on controllers are in principle independent of those arising from targets.
However, the independence of all these influences has not been demonstrated in detail. It does suggest,
however, that agreement is much more complex than is generally recognized.

2.4. Features

In this section we first consider which are the agreement features (section 2.4.1), then we look at
regularities in the way they are expressed (section 2.4.2), before examining the complex issue of
feature resolution (section 2.4.3) and the question of default values (2.4.4).

2.4.1. Person, number and gender

The three features which are indisputably agreement features are somewhat different in nature. Gender
is an inherent feature of the noun. It is found on the target, say the adjective, as a consequence of its
presence in the noun (overt or covert). Thus an ending marking gender on an adjective has nothing to
do with the lexical meaning of the adjective. A somewhat similar situation obtains for person; person
is an inherent feature of the pronoun, but not of the verb. Number is more difficult. It is an inherent
feature of some nouns: those which are only singular (like Serbo-Croat hrabrost ‘courage’) or only
plural (like Russian šči ‘cabbage soup’) impose this feature value on their modifiers. However, a
considerable proportion of the nouns in Slavic languages can be associated with both (or all) numbers.
In straightforward examples involving such nouns, the number feature appears to relate primarily to
the noun; the property denoted by the adjective is not affected by the change in number. The three
agreement features are all nominal; they are what Zwicky (1992: 378) calls the ‘direct features’ of
nouns and noun phrases. As Nichols shows, they have an interesting hierarchical relationship: gender
is the one which is most prone to be marked only by agreement; number is quite likely to be marked
only in this way, but this never occurs with person (1992: 160-162). Further discussion of the relations
between the three features can be found in Bybee (1985: 22–24, 28-33) and Wunderlich (1993).

19
‘In principle’ because in some instances there is no opportunity for such influence; for instance, the factor of animacy
cannot affect individual lexical items which always denote animates or always inanimates.
 GREVILLE G. CORBETT 26

Traditional accounts of Slavic languages also include agreement in case. In a phrase like
Russian: v novom avtomobile ‘in a new car’, the adjective and noun stands in the same case, but this
covariance differs from that found with gender, number or person. Case is not a feature of the noun: it
is imposed on the noun phrase by government by some other syntactic element. Thus the noun and
adjective are in the same case because it is imposed equally on both. This is not agreement, if we take
seriously the question of asymmetry. Following that view, we would not recognize case as an
agreement feature, though we should recognize that it interacts strongly with agreement features.
There is, however, a problem with this view. It concerns Polish expressions like the following
(Dziwirek 1990: 147):

(26) Sześć kobiet by-ł-o smutn-ych

six woman[PL.GEN] be-PST-SG.N sad-PL.GEN
‘Six women were sad.’

The verb is third singular neuter ‘by default’ (see section 2.4.4); the adjective appears to agree in
number and case with the quantified noun within the subject noun phrase (this is not the only
possibility; according to Dziwirek (1990: 158n16) the neuter singular is found in ‘informal spoken
Polish’). This construction is in any case difficult to analyse, but it suggests that we may have to allow
for agreement in case.20

2.4.2. Exponents of features

Since the major interest of Slavic agreement is in the syntax, we have concentrated on this. But the
morphology too deserves some attention (for a general account of the morphology of agreement,
showing the wider typological frame, see Corbett 1998). We might have imagined that agreement
features could be stated just at the level of the language. But, of course, we cannot simply say that a
particular language has gender agreement. There is likely to be variation among the elements
identified as agreement targets, as in this example from Upper Sorbian:

(27) wón je pisa-ł

he AUX.3SG write-PST.SG.M
‘he wrote’

Here the finite verb agrees in number and person, while the participle agrees in number and gender.
While observing differences between word classes in respect of the agreements they may show, we
might expect to treat word classes as internally uniform in respect of their agreement potential. This
too is an oversimplification since there are instances of systematic differences within word classes.
Thus Russian verbs agree with their subject in person and number, except in the past tense, which
20
Another possible instance is the secondary predicates of Old Church Slavic (Wayles Browne, personal communication).
 AGREEMENT IN SLAVONIC 27

agrees in gender and number. This generalization holds for all verbs. But there are also instances of
idiosyncratically different agreement possibilities within word classes. Macedonian adjectives show
this clearly:

Table 13. Agreement of Macedonian adjectives (Friedman 1993: 266-267)

MASCULINE FEMININE NEUTER PLURAL gloss

nov nova novo novi new
kasmetlija kasmetlija kasmetlija kasmetlii lucky
taze taze taze taze fresh

Nov ‘new’ and similar adjectives distinguish three genders and two numbers; those like kasmetlija
‘lucky’ agree in number but not gender, while taze ‘fresh’ and adjectives like it are indeclinable.

2.4.3. Resolution rules

A problem where Slavic data have been important is that of ‘resolution rules’, the rules which
determine the number, gender and person of the target for controllers consisting of conjoined noun
phrases. I have dealt with these previously at some length and the interested reader may consult that
work in Corbett (1983a: 177-214; 1983b; 1983c; 1991: 261-306); see also Wechsler and Zlatić (2003:
171-195) for an important contribution, and Corbett (2006: 238-263) for a new overview. The account
here will cover just the main points. Let us first take Slovene for illustration. If a masculine singular
and a feminine singular are conjoined, it is the gender and number resolution rules which specify the
form of the target, say the predicative adjective, as masculine dual (data from Lenček 1972):

(28) Tonček in Marina sta prizadevn-a

Tonček.M and Marina.F be.DU assiduous-DU.M
‘Tonček and Marina are assiduous.’

This example illustrates the point that resolution rules do not operate only to resolve feature clashes
but can also operate when conjuncts share features (singular in this example). The next shows the
same thing with relation to gender:

(29) to drevo in gnezdo na njem mi bosta ostal-a v spominu

that tree.N and nest.N on it 1SG.DAT AUX.FUT.3DU remain-DU.M in memory
‘That tree and the nest on it will remain in my memory.’

Though both conjuncts are neuter, the gender resolution rules specify masculine agreement (as we
shall see later). It is important to recall that resolution is generally not obligatory; instead agreement is
often with one conjunct only, and so resolution is not involved (section 2.1.1.6).
 GREVILLE G. CORBETT 28

Number and person resolution are the easier ones; in Slavic and elsewhere, if resolution occurs,
then two singulars (in Old Church Slavic, Slovene and Sorbian) require a dual and in all other
instances a plural is used (see Corbett 1983b: 177-178, 189-190 for a complication here). And for
person, the presence of a first person determines first person agreement, failing that a second person
will give second person agreement (and failing both of these conditions the third person arises by a
general default).21
Let us return to the most interesting type, namely gender resolution. This may follow two distinct
principles: the syntactic principle or the semantic principle. Gender resolution by the syntactic
principle operates according to the syntactic gender of the conjoined items, irrespective of their
meaning. Gender resolution by the semantic principle involves reference to the meaning of the
conjoined elements, even if this implies disregard for their syntactic gender. This gives us two types;
there is a third (mixed) type, since some languages use interesting combinations of the two principles.
We shall consider these three types in turn (sections 2.4.3.1-2.4.3.3). Since this overview was first
presented there has been interesting progress on gender resolution. The data are presented here
according to the old typology and the pointers to new ideas are given in section 2.4.3.3. Then, having
seen various instance of gender resolution, we shall consider further the typology of resolution systems
(section 2.4.3.4).

2.4.3.1. Syntactic gender resolution

Syntactic gender resolution is found in Slovene. In Slovene, a masculine conjoined with a feminine, as
we saw in (28) above, or with a neuter, as in (30) takes a masculine predicate:

(30) Tonček in to dekletce sta prizadevn-a

Tonček(M) and that little.girl(N) be.DU assiduous-DU.M
‘Tonček and that little girl are assiduous.’

When a feminine and a neuter are conjoined, the masculine is still found (examples (31), (32) and (33)
are from Priestly (1993: 433), the remainder are from Lenček 1972):

(31) Milka in njeno tele sta bi-l-a zunaj

Milka(F) and her calf(N) AUX.3DU be-PST-DU.M outside
‘Milka and her calf were outside.’

21
The problems of incorporating resolution rules into linguistic theory have been partly addressed, particularly within the
framework of Generalized Phrase Structure Grammar (Farkas and Ojeda 1983; Morgan 1984; Sag, Gazdar, Wasow and
Weisler 1985: 152-155).
 AGREEMENT IN SLAVONIC 29

Similarly, two conjoined neuter singulars take a masculine dual, as we saw in (29).22 The way in which
the feminine/neuter dual form can result from the resolution rules is if two feminines are conjoined:

(32) Milka in njena mačka sta bi-l-i zunaj

Milka(F) and her cat(F) AUX.3DU be-PST-DU.F outside
‘Milka and her cat were outside.’

The gender resolution rules are as follows:

1. if all conjuncts are feminine, then the feminine form is used;

2. otherwise the masculine is used.

The number resolution rules determine when the dual and when the plural form are to be used. As this
is so, the rules just given will also account for gender resolution when the plural results. Thus in (33),
all the conjuncts are neuter, but the masculine plural form is required:

(33) Dve telet-i in eno žrebe so bi-l-i zunaj

two foal(N)-DU and one calf(N) AUX.3PL be-PST-PL.M outside
‘Two foals and a calf were outside.’

Again, the feminine is possible only if all the conjuncts are feminine:

(34) Marina, Marta in Marjanca so prizadevn-e

Marina(F) Marta(F) and Marjanca(F) be.3PL assiduous-PL.F
‘Marina, Marta and Marjanca are assiduous.’

The important point for our typology is that in the rules given there is no recourse to semantic
factors; the syntactic gender is the sufficient determining factor.

2.4.3.2. Mixed gender resolution

Polish has three forms for gender agreement in the singular; in the plural there is a division into
masculine personal (abbreviated ‘M.PERS’) and the remainder (non-masculine personal, ‘NM.PERS’).
When in conjoined structures none of the conjuncts is headed by a masculine personal noun, then the
non-masculine personal is used (Rothstein 1993: 732–733):

(35) Basia i Marysia przynioś-ł-y sałatę

Basia(F) and Marysia(F) brought-PST-PL.NM.PERS salad
‘Basia and Marysia brought salad.’

If a masculine personal noun heads one of the conjuncts then the masculine personal form is used:

22
We follow Lenček here, who gives the fullest and clearest account. For a possible complication with examples like (29)
see Corbett (1983a: 212n6).
 GREVILLE G. CORBETT 30

(36) Janek i Marysia przynieś-l-i ciastk-a

Janek(M) and Marysia(F) bring-PST-M.PERS.PL pastry-PL.ACC
‘Janek and Marysia brought pastries.’

The resolution rules required appear to be as follows:

1. if at least one conjunct is masculine personal, then the masculine personal form is used;
2. otherwise the non-masculine personal form is used.

Rules like these can be found in numerous descriptions, and they operate in other West Slavic
languages. However, Polish is actually more interesting. Consider this example from Doroszewski
(1962: 237):

(37) Hania i Reks bawi-l-i się piłk-ą

Hania(F) and Rex(M) play-PST-PL.M.PERS REFL ball-INS
‘Hania and Rex were playing with a ball.’

There is no masculine personal conjunct in (37), since Reks, a dog, is masculine but not personal, and
yet the predicate is masculine personal. Sentences like (37) have been discussed at length. The best
data are provided by Zieniukowa (1979), who gives responses to a questionnaire by 31 young people
in their upper teens.23 For a sentence comparable to (37), only two speakers used the non-masculine
personal form (and one used a different construction). The masculine personal form, as in (37), is the
preferred form. It cannot result simply from the presence of the noun Hania denoting a person, since in
(35) both conjuncts denoted humans but a non-masculine personal form was used. It is worth checking
whether having a masculine animate conjunct is sufficient: in the following example both conjuncts
are masculine animate:

(38) pies i kot jedl-i na podwórzu

dog(M) and cat(M) eat-PST-PL.M.PERS on yard
‘The dog and the cat were eating in the yard.’

Zieniukowa found that the masculine personal form (as in the example) was the majority choice;
however, seven speakers chose the non-masculine personal form. Thus masculine animates are less
likely to produce a masculine personal form than masculine animate plus a feminine denoting a
human. Even the combination of feminine denoting a human conjoined with masculine inanimate can
result in a masculine personal form:

23
Compare Weiss (1985: 354-355).
 AGREEMENT IN SLAVONIC 31

(39) Mama, córeczka i wózek ukaza-l-i się nagle

Mother(F) daughter(F) and pram(M) show-PST-PL.M.PERS REFL suddenly
‘The mother, daughter and pram appeared suddenly.’

In this example speakers were equally divided between the masculine personal and the non-
masculine personal ukazały (one speaker chose neither).
The rules required to cover these examples (and other types described in Corbett 1983a: 197–
200) are as follows:

1. if the subject includes a masculine personal conjunct, the predicate will be in the masculine
personal form;
2. (optional) if the subject includes the features masculine and personal, whether these are
syntactic or semantic, the predicate may be in the masculine personal form;
3. (optional) if the subject includes a masculine animate conjunct, the predicate may be in the
masculine personal form;
4. otherwise the predicate will be in the non-masculine personal form.

The first rule, which accounts for the form in (36), is straightforward. The optional Rules 2 and 3 both
represent relaxations of Rule 1: in Rule 2 the conditions apply to the subject as a whole rather than to a
single conjunct and, more surprisingly, they allow semantic or syntactic features or a combination of
these. Rule 3, on the other hand, retains the restriction to a single conjunct but reduces the requirement
from personal to animate. Rule 2 accounts for the form in sentence (39) while Rule 3 permits (38). It is
significant that when both Rule 2 and Rule 3 can apply, as in (37) then for those Zieniukowa consulted
the masculine personal form is almost obligatory. When none of these rules apply, the non-masculine
personal form is assigned by Rule 4, as in sentence (35). The rules refer both to syntactic gender and to
semantic criteria. Thus Polish stands between the clearly semantic gender resolution and the syntactic
type.

2.4.3.3. Semantic gender resolution

Slavic does not have a language with semantic gender resolution, but Serbo-Croat gives fascinating
signs of a system that may be moving that way. The resolution rules seem to be as in Slovene, that is,
if all the conjuncts are feminine then the feminine form is used and otherwise the masculine form is
used. Given at least one non-feminine conjunct there is no problem — the masculine form is the
resolved form. Here feminine and neuter are conjoined:

(40) Znanje i intuicija su kod njega

knowledge(N) and intuition(F) AUX.3PL with him

sarađiva-l-i i dopunjava-l-i se...

work.together-PST-PL.M and supplement-PST-PL.M REFL
‘Knowledge and intuition worked together in him and supplemented each other...’
 GREVILLE G. CORBETT 32

(Andrić, Travnička Hronika)

Similarly, when neuters are conjoined we find a masculine plural predicate:

(41) Njegovo mesto u razvitku kasabe i njegovo

his place(N) in development town.GEN and his

značenje u životu kasabalij-a bi-l-i su

importance(N) in life inhabitants-PL.GEN be-PST-PL.M AUX.3PL

onakv-i kako smo ih napred ukratko opisali.

such-PL.M as AUX.1PL them before briefly described
‘His place in the development of the town and his importance in the life of the inhabitants were
such as we described them briefly before.’ (Andrić, Travnička Hronika)

When all the conjuncts are feminine, then we would expect feminine agreements, as we find in the
following example:

(42) Opreznost, suptilnost i pedanterija tih bezbrojnih

discretion(F), subtlety(F) and pedantry(F) these innumerable

poruka zbunjiva-l-e su mladića ...

assignments.GEN perplex-PST-PL.F AUX.3PL young.man
‘The discretion, subtlety and pedantry of these innumerable assignments perplexed the young
man ... ’ (Andrić, Travnička Hronika)

So far the examples can be handled by the simple rules, found in Slovene (if all conjuncts are
feminine, agreements will be feminine, otherwise masculine). And this must have been the earlier
situation in Serbo-Croat. More recently in Serbo-Croat we find the use of masculine agreements in
instances not sanctioned by the old rules. Gudkov (1965) claimed that a masculine predicate is
possible, even though all the conjuncts are feminine, provided that at least one of them is headed by a
noun of the i-stem type (with no ending in the nominative singular):

(43) Vređa-l-i su ga nebriga i lakomislenost Tahir-beg-ov-a.

offend –PST-PL.M AUX.3PL him carelessness(F) and capriciousness(F) Tahir-beg-POSS-F
‘Tahir-beg’s carelessness and capriciousness offended him.’ (Andrić, Travnička Hronika)

Masculine agreement here is not obligatory, as example (42) shows: both masculine and feminine
agreements are both found. The gender resolution rules are similar to those required for Slovene, but
we must allow for the first rule to be optional. The interesting point is the nature of the condition on
this optionality. Gudkov suggested a morphological condition, so our rules would be along these lines:
 AGREEMENT IN SLAVONIC 33

1. if all conjuncts are feminine, then the feminine form will be used; (if at least one of the
conjuncts is a noun of the i-stem declension, then this rule is optional);
2. otherwise the masculine will be used.

Such a rule would be quite remarkable, because agreement rules normally refer to syntactic or
semantic categories. The condition referring to a noun of a particular declensional type would be a
considerable weakening of the theory of agreement.24 It turns out that the morphological condition
does not cover all the relevant instances. Gudkov himself subsequently pointed out occasional
examples in which subjects headed exclusively by feminine nouns in -a take masculine agreements
(1974: 61):

(44) štula i štaka bi-l-i su sve što

wooden.leg(F) and crutch(F) be-PST-PL.M AUX.3PL all that

je tadašnja medicina mogla da mu pruži.

AUX.3SG of that time medicine could that him.DAT offer
‘A wooden leg and a crutch were all that medical science of that time could offer him.’
(Popović)

Gudkov was quite right; I can add this confirming example:

(45) Žustrina sa kojom je pisao i lakoća sa kojom je

speed(F) with which AUX.3SG wrote and ease(F) with which AUX.3SG

nalazio reči i poređenja zagreja-l-i su ga, i

found words and comparisons warm-PST-PL.M AUX.3PL him and

konzul oseti nešto kao olakšanje.

consul felt something like relief
‘The speed with which he wrote and the ease with which he found words and comparisons
encouraged him, and the consul felt a kind of relief.’
(Andrić, Travnička Hronika)

Even if we accepted the weakening implied by the morphological condition, we would still be
unable to account for some of the examples involved, including (44) and (45). The significant point is
that in all the examples with feminine conjuncts but masculine agreement, the noun phrases denote

24
A more serious weakening would be to suggest that the rule should be stated in phonological terms (which would then
infringe the principle of phonology-free syntax (Pullum and Zwicky 1988: 278). The rule would then refer to the presence
of a noun ending in a consonant (since the nouns of the i-stem declension typically end in a consonant, in the nominative
singular like the typical masculines). However, there are also nouns like misao ‘thought’ which belong to the i-stem
declension yet end in a vowel. This final vowel alternates with a consonant, but in a not fully predictable way (see Corbett
1983a: 190 for details). Hence the phonological condition is inadequate.
 GREVILLE G. CORBETT 34

inanimates. I have found no examples of masculine agreement with feminine nouns denoting
persons.25 Therefore the condition can and should be stated as a semantic one:

1. if all conjuncts are feminine, then the feminine form will be used; (if the conjuncts denote
inanimates, then this rule is optional);
2. otherwise the masculine will be used.

We still need to explain why the majority of examples with feminine conjuncts and masculine
agreements involve a noun of the i-stem declension. This declension includes a large proportion of
abstract nouns, and practically no animates. When one collects genuine examples of conjoined noun
phrases it is striking that the overwhelming majority have conjuncts of the same semantic type (all
animate or all inanimate). Thus when a feminine noun of the i-stem declension is one of the conjuncts
then there will normally be no animates in the subject. This means, in turn, that the use of the feminine
agreement form will have no semantic justification (unlike its use with animate conjuncts which
denote females). What seems to be happening is that Serbo-Croat is is moving from gender resolution
rules operating on syntactic conditions towards a semantic system. We may represent the development
as follows:

Stage 1 (earlier Serbo-Croat, as still in Slovene)

1. all conjuncts feminine → feminine

2. otherwise → masculine

This is a simple syntactic system.

Stage 2 (present Serbo-Croat)

1. all conjuncts feminine → feminine (optional for inanimates)

2. otherwise → masculine

It is not clear whether the option should relate to inanimates or non-humans. In either case the crucial
point is that the rules include a reference to semantics, and so we have moved from a syntactic to a
mixed system. These rules can be formulated differently:

Stage 2 (alternative formulation)

1. all conjuncts female → feminine

2. all conjuncts feminine → feminine (optional)
3. otherwise → masculine

25
Ljubomir Popović informs me (personal communication) that he has found rare examples even of this type in written
text.
 AGREEMENT IN SLAVONIC 35

A possible development is for the optional rule to be dropped.

Stage 3 (hypothetical)

1. all conjuncts female → feminine

2. otherwise → masculine

If this development occurs, then we shall have had the development from a syntactic system, through a
mixed system, to a semantic system. (For further discussion of resolution in Serbo-Croat see Leko
(1986: 220-243), and for the earlier period of Slavic see Šul´ga (1997).)
The importance of semantic considerations should give pause for thought here. Perhaps too the
rather special rules for coordinate noun phrases, which are a phenomenon somewhat peripheral to the
syntactic system should concern us. Wechsler and Zlatić (2003: 171-95) made the interesting
suggestion that across languages animate noun phrases are subject to semantic resolution, while
inanimates are subject to syntactic resolution. In all languages, even those like Slovene and Serbo-
Croat, gender resolution is semantically driven. A key piece of evidence concerns the behaviour of
hybrid nouns (§2.1.1.1) in coordinate structures. They cite Farkaş and Zec (1995) for this; in fact the
observation had been made earlier by Megaard (1976: 95), though its significance was not recognized
then. We may rethink their suggestion as follows: resolution must be based in part on semantic criteria;
it may additionally be based on syntactic criteria. This then gives us two pleasing effects. First, gender
resolution is more like person and number resolution. Second, gender resolution directly reflects
gender assignment, since gender assignment is always based on semantic criteria, which may or may
not be supplemented by formal criteria; for details of this new typology see Corbett (2006: 258-263).

2.4.3.4. More on the typology of resolution systems

There are further types of feature which might have been expected to have a role, and their exclusion
from gender resolution (and usually from resolution in general) allows us to constrain further the
possible types of resolution system.
First there are no rules of the type: ‘if there is a first person feminine conjunct then ...’ or ‘if there
is a neuter dual conjunct then ...’. Gender resolution needs to refer only to gender (and equally person
resolution refers only to person, and number resolution refers only to number).26 It might appear that

26
Smith, Tsimpli and Ouhalla (1993: 316-317) report on an attempt to teach (among other things) impossible resolution
rules to Christopher, a polyglot savant, and to a control group. Their invented language Epun has these (impossible) rules:
1st singular plus third singular feminine is resolved as third plural feminine;
1st singular plus third singular masculine is resolved as second singular.
(Gender is distinguished only in the third person.)
In translating from English to Epun, Christopher followed the first rule, but generalized it to the second case and used the
third plural masculine there. Of the four subjects used as controls (all first year undergraduate students of linguistics, one
managed the impossible agreements, the other three used the forms expected in genuine natural languages (the first person
plural) in both the situations above.
 GREVILLE G. CORBETT 36

the resolution rules are completely independent of each other. However, while they are independent in
their formulation, they are not independent in their operation; in other words, they operate as a set or
not at all. Agreement may be with one conjunct or with all conjuncts; if the latter, that is if resolution
operates, then all applicable resolution rules must operate. There cannot, say, be resolution in gender
but not in person. For more on the interdependence of resolution rules see Corbett (1983b: 182-183,
2003).
Resolution is also independent of the agreement target. That is to say, there are no resolution
rules of the type ‘if all conjuncts are neuter, then adjectival targets take masculine agreement while
verbal targets take neuter agreement.’ Information about the target cannot be part of a resolution rule.
What does differ is the likelihood of resolution as compared with agreement with the nearest conjunct.
Resolution is a particular case of semantic agreement. The distribution of resolution (semantic
agreement) versus agreement with the nearest conjunct (syntactic agreement) is therefore constrained
by the Agreement Hierarchy (see section 2.2.1.1).
A third type of conceivable conditioning factor which is never employed is that of the
construction: resolution is ‘construction independent’. As we saw in section 2.1.1.7, some Slavic
languages have a comitative construction where agreement is possible with the governed noun phrase
as well as with the head (for this reason they are sometimes called quasi-comitatives). When this is the
case, resolution in comitative constructions will always be as in conjoined constructions. Thus Polish
has a comitative construction which allows agreement with both noun phrases (resolution), (Dyła
1988: 386):27

(46) Jurek z Jankiem posz-l-i na spacer

Jurek(M) with Janek(M) go-PST-PL.M.PERS for walk
‘Jurek and Janek went for a walk.’

(47) Ewa z Jankiem posz-l-i na spacer

Ewa(F) with Janek(M) go-PST-PL.M.PERS for walk
‘Ewa and Janek went for a walk.’

(48) Ewa z Marią posz-ł-y na spacer

Ewa(F) with Maria(F) go-PST-PL.NM.PERS for walk
‘Ewa and Maria went for a walk.’

As Dyła points out, resolution rules operate to give the same results as with conjoined
expressions: where we have a masculine personal conjunct then masculine personal agreement results

27
See also Szupryczyńska (1991). For a small amount of Slovene data see Lenček (1972: 61-62), and for a discussion of
the semantics of these constructions in Russian and Polish see McNally (1993).
 AGREEMENT IN SLAVONIC 37

(as in (46) and (47)).28 It should be noted, however, that once again the likelihood of resolution as
opposed to agreement with just one element (the head noun phrase in this instance) does depend on the
construction.

2.4.4. Default values

Agreement targets which are morphologically able to agree typically must agree in Slavic, even when
their controller lacks the appropriate features (as when it is a complete clause or an infinitive
construction)29 or is totally lacking (as in impersonal constructions). In such circumstances default
values appear, normally third person, singular and neuter.
These defaults may differ from the normal occurrences of third person neuter singulars in two
ways. First, they may in some circumstances be morphologically distinct. Thus Ukrainian predicative
adjectives have -o for ‘neutral agreement’, the failure to agree, as opposed to -e for the neuter singular.
And Russian uses the pronoun èto ‘this’ rather than ono ‘it’ for antecedents that are not genuinely
neuter singulars. And second, they differ syntactically, in that conjoining controllers which take
default agreements simply leads to the same default agreement (not neuter plural).
This line of research was continued by Dziwirek in a Relational Grammar account (1990); she
shows that numeral phrases in Polish can fill the subject position (that is, she argues against the
impersonal sentence analysis) but that the predicate agreement (neuter singular) represents default
values. Her analysis is discussed in Stroińska (1992). Defaults also figure prominently in accounts of
Serbo-Croat numeral phrases: the GB account of Franks (1995: 114-115) and the HPSG account of
Wechsler and Zlatić (1997).

28
A further construction type is what Schwartz (1988) calls verb-coded coordinations: she illustrates the construction from
various languages, including Polish (1988: 54; two typographical errors have been corrected):
(i) posz-l-i-śmy z matką do kina
go-PST-PL.M.PERS-AUX.1PL with mother to cinema
‘mother and I went to the cinema’
In this construction, one of the conjuncts is omitted. However, the verb form (masculine personal) shows the form expected
from gender resolution, provided the first person (speaker) is male. If the speaker is female, then the form would be:
(ii) posz-ł-y-śmy z matką do kina
go-PST-NM.PERS.PL-AUX.1PL with mother to cinema
‘mother and I went to the cinema’
Compare with the following, also with a female speaker:
(iii) posz-l-i-śmy z ojcem do kina
go-PST-PL.M.PERS-AUX.1PL with father to cinema
‘father and I went to the cinema’
We again find the forms predicted by the normal resolution rules. (I am grateful to Katarzyna Jaszolt and Roland Sussex
for help with the data.)
29
See Browne (1990) on Serbo-Croat and (1998) on South Slavic more generally.
 GREVILLE G. CORBETT 38

2.5. Typologically unusual domains

The agreement domains typical for Slavic, discussed briefly at the beginning of section 2, are not
remarkable typologically. There are, however, some agreement constructions which are more unusual,
and we consider them in turn here.

2.5.1. Possessives in Upper Sorbian and Slovak

Perhaps the most remarkable instance of agreement in Slavic is found in Upper Sorbian, where the
possessive adjective can control an attributive modifier, as in this example (from Fasske 1981:
382-383):

(49) moj-eho muž-ow-a sotr-a

my-SG.M.GEN husband-POSS-SG.F.NOM sister-SG.F.NOM
‘my husband’s sister’

In (49), the possessive suffix -ow- may be thought of as marking the phrase mój muž ‘my husband’. To
it is added the inflectional marker for nominative singular feminine, showing agreement with the head
noun sotra ‘sister’. The particularly interesting form is mojeho; this is masculine since muž ‘husband’,
which is the source of mužowa, is masculine. It is singular for the same reason (the formation of the
possessive adjective requires a singular referent). Thus we have the possessive adjective as a controller
of agreement, taking another attributive modifier as its target, which is a totally unexpected agreement
domain. The construction has been discussed in detail in Corbett (1987, 1995) and so it will be just
noted here; see those sources for references and for the distribution of the construction in the Sorbian
dialects see Fasske (1996: 66-73). This Upper Sorbian construction is indeed remarkable; the only
other modern Slavic language which has it, and to a more limited extent, is Slovak. Control of the
relative pronoun by the possessive adjective is much more common, while control of the anaphoric
pronoun is general in Slavic (except for Polish, where it is limited).

2.5.2. ‘Collaborative’ agreement

The loss of the dual number has led to lower numerals being involved in complex constructions in
various Slavic languages. Here we consider such numerals in Russian. The numerals dva ‘two’, tri
‘three’ and četyre ‘four’, when themselves in the nominative, take a noun in the genitive singular. Dva
‘two’ has the feminine form dve:

(50) dv-e sosn-ý

two-F pine-SG.GEN
‘two pines’
 AGREEMENT IN SLAVONIC 39

The special interest here is that the numeral governs the form of the noun, requiring it to be genitive
and singular (the stress in this example makes the noun unambiguously genitive singular). It is only
because the noun is singular that there can be agreement in gender, since gender is not distinguished in
the plural in Russian. So the genitive singular noun, required to be in that form by the numeral, in turn
acts as the controller for gender agreement of the numeral, hence the term ‘collaborative agreement’.
For further discussion of the headedness relations here see Corbett (1993).30 The problems discussed in
sections 2.5.1 and 2.5.2 were treated together as ‘reluctant controllers’ in Corbett (1986).

2.5.3. ‘Distant’ agreement

In agreement with conjoined noun phrases, as we saw in section 2.1.1.6, there are typically two
possibilities (within Slavic and more widely): agreement with all conjuncts (in which case the
operation of resolution rules is required) or agreement with just the nearest conjunct. There is,
however, a further rare possibility. It is distant agreement, that is to say agreement with the first
conjunct, which, with subject-verb word order, is not the nearest. Examples occur in Slovene:

(51) knjig-e in peres-a so se podraži-l-e

book(F)-PL and pen(N)-PL AUX.3PL REFL get.dear-PST-PL.F
‘Books and pens have become more expensive.’ (Lenček 1972: 59)

This must be an instance of agreement with the first conjunct; agreement with all would require the
masculine plural (a result of the resolution rules, section 2.4.3). Similar examples occur in Serbo-
Croat; the most extensive source is Megaard (1976), see also Leko (1986: 230). (It has also been
claimed that in Čakavian dialects of the 16th-17th centuries agreement could be with the most
important conjunct, even if this was not the nearest or the first, Glavan 1927-28: 143–145; the
evidence is very limited.) Since these examples have been misunderstood outside Slavic circles it is
worth stressing that distant agreement is rare, and that agreement with the nearest noun phrase is much
more common.

2.5.4. ‘Back’ agreement

Some traditional grammars state that, in sentences consisting of subject noun phrase, copula and
nominal predicate, besides the expected agreement with the subject, agreement of the copula with the
noun phrase in the predicate may also be found. This phenomenon is called attraction, or back or
backward agreement. Given that predicate-subject order is common in Slavic, it is of course not
sufficient merely to find examples of agreement with the postverbal noun phrase. The evidence is

30
And for the special situation in Serbo-Croat see Corbett (2009).
 GREVILLE G. CORBETT 40

assessed in Corbett (1986; see 1019-1020 for discussion of Russian data; see also Crockett 1976: 406-
407). The following Czech sentences appear to be good prima facie evidence for back agreement (the
construction is noted in Vanek (1970: 53):31

(52) jedna a dvě jsou tři

one and two be.PL three
‘one and two are (make) three’

(53) jedna a tři jsou čtyři

one and three be.PL four
‘one and three are four’

(54) dvě a tři je pět

two and three be.SG five
‘two and three are five’

(55) tři a tři je šest

three and three be.SG six
‘three and three are six’

The regularity is that if the numeral to the right (in our examples) of the copula is ‘two’, ‘three’ or
‘four’, then the copula takes plural agreement, while if it is ‘five’ or above, then the singular is found.
Moreover, there are syntactic tests to show that the numeral in question is part of the predicate
(Corbett 1986: 1002-1003). More needs to be done to specify the conditions under which this type of
agreement can occur in different Slavic languages. This is made more difficult by the fact that its
‘habitat’ is being eroded by the rise of the instrumental predicate. However, the mere existence of back
agreement would be problematical for some theoretical frameworks.

3. Innovations and ongoing research

Here we consider three broad areas: questions of definition (section 3.1), formal models (section 3.2),
and the question of features (section 3.3).

3.1. Definitions

Two strands of work deserve attention here: the basic problem of defining agreement, and the interest
of the phenomena which make that definition difficult.

3.1.1. Attempts to define agreement

As suggested in the quote from Anderson, it is a genuinely difficult task to define agreement in a clear
and consistent way, separating it from allied but different phenemena. Work which should be noted
31
The acceptability of the sentences given according to Vanek’s account was confirmed by three consultants: Robert
Slonek, Magda Newman and Otto Pick.
 AGREEMENT IN SLAVONIC 41

here, as being particularly relevant to Slavic problems, is that of Kibrik (1977), Lehfeldt (1980) and
Mel´čuk (1993). We shall do no more than note the problem and the fact of ongoing work here, since
there is a book-length survey of research into the question by Schmidt and Lehfeldt (1995), and a
recent ‘canonical’ approach to the issue in Corbett (2006).

3.1.2. ‘Border phenomena’

Work on definitions highlights constructions which are of particular interest. Thus all the instances
discussed in section 2.1 above are likely to be problematic for at least some definitions of agreement.
Such ‘border phenomena’ deserve special attention. One which is of special interest is the clitic-
doubling found in Bulgarian (Scatton 1993: 228-229, 234-235, and Rudin 1986), and in Macedonian.
Consider this example from Macedonian (Friedman 1993: 285, see also 291):

(56) Kuče-to ja kaca mačka-ta

dog-DEF.SG.N 3SG.F.ACC bites cat.DEF.SG.F
‘the dog bites the cat’

In Macedonian, if the object is definite then an object clitic pronoun is required. Depending on one’s
definition of agreement this is either agreement conditioned by definiteness, or it is not yet a case of
agreement but is an interesting instance of a source of future agreement. Either way, the development
of this construction, and the differences between Bulgarian and Macedonian in this area, are of special
interest (see Friedman 1994 and references there).

3.2. Formal models

Naturally we wish to describe the agreement patterns of Slavic with an appropriate formal model. Let
us consider its desired properties. There is an intuition that agreement is asymmetric. Some accounts
of agreement capture this intuition directly by copying feature specifications from the controller to the
target. These feature-copying approaches face several problems: the controller may be absent (as in
pro-drop languages), or it may be present but be underspecified, something which occurs frequently
with pronouns (Barlow 1988/1992: 30–43; his arguments are developed in Pollard and Sag 1994: 62-
67); or the feature specifications on the controller and the target may simply not match. Unification-
based accounts, where agreement can be seen as a matter of cumulating partial information from the
controller and the target, have much better prospects (Shieber 1986: 21-22, Barlow 1988/1992: 22-45,
but see Bayer and Johnson 1995 for problems). Consider just one of the arguments, starting from the
following example:
 GREVILLE G. CORBETT 42

(57) Ja bud-u dovolen/dovol´n-a

I be.FUT-1SG satisfied[SG.M]/satisfied-SG.F
‘I will be satisfied’ (man/woman talking).

In accounts based on a rule of feature-copying, we need to say that Russian and other Slavic languages
have two pronouns ja, one masculine and one feminine, which happen to be phonologically identical.
In a unification-based approach, we could have the following feature structures, for the female
instance (the first for the pronoun and the verb, and the second for the predicative adjective):

⎡number: singular ⎤ ⎡number: singular ⎤

(58) ⎢⎣ person: 1st ⎥⎦ ⎢⎣gender: feminine ⎥⎦

These feature structures can be unified, since they are compatible, to give the following structure:

⎡number: singular ⎤
(59) ⎢ person: 1st ⎥
⎢⎣gender: feminine ⎥⎦

Thus the information is cumulated from different parts of the structure. These approaches to language
may be called ‘constraint-based approaches’ (Shieber 1992: 1); they specify, as constraints, that
particular feature structures must unify.
However, specifically for representing agreement, this leaves the question of asymmetry. In
Generalized Phrase Structure Grammar this notion is reintroduced by the Control Agreement Principle
(based on Keenan 1974), which specifies possible controllers and targets, and gives them different
statuses (see Gazdar, Klein, Pullum and Sag 1985). In Head-Driven Phrase Structure Grammar the
asymmetry is captured through ‘anchoring’; gender, number and person features are anchored to real
world entities through noun phrase indices, even though they may be expressed morphologically other
than on the noun phrase (see Pollard and Sag 1994: 60-99, and compare Kathol 1999).
A recurrent problem is that we find mismatches — instances where controller and target may
realize feature values which do not unify neatly (we saw numerous examples in sections 2.2.1.1 and
2.4.3). We have seen areas where the pattern is clear, the data constrained by the Agreement Hierarchy
(section 2.2.1.1), the data constrained by the Predicate Hierarchy (section 2.2.1.2) and predicate
agreement with numeral phrases (Table 2 in section 2.1.1.3). But these constraints, which have
substantial cross-linguistic support, do not fit readily into current accounts of agreement (Barlow 1991,
Pollard and Sag 1994: 58, Kathol 1999).32 Let us concentrate on just one, the Agreement Hierarchy.

32
See also Kirby (1999: 92-96) for recent discussion of the Agreement Hierarchy in terms of the emergence of language
universals.
 AGREEMENT IN SLAVONIC 43

When developing a more refined account of agreement, which might address the kind of data
covered by the Agreement Hierarchy, a common first move, as Barlow (1991 points out) is to split
agreement into two different phenomena: for instance, in LFG it is grammatical vs anaphoric
agreement (Bresnan and Mchombo 1987). Whatever the merits of splitting agreement into two
phenomena, this does not solve our problem. We cannot simply say that, for instance, where there is a
choice of agreement options, semantically justified agreement will be found within the structures
which LFG calls anaphoric agreement. First, because the divide between syntactic and semantic
agreement is not necessarily clear-cut. We saw this with agreement in Serbo-Croat (Table 9), there is a
choice in predicate position and for the relative pronoun. Second, at the extremes, there can still be a
choice: the noun phrase must surely come under ‘grammatical agreement’ if agreement is split, and yet
we can find semantic agreement here (example (18)); conversely the personal pronoun would be
expected to fall under anaphoric agreement, and yet syntactic agreement can be found here (for
instance, neuter agreement with Czech děvče ‘girl (colloquial), Vanek 1970: 87-88). Moreover, it is
not at all clear that the problem is being tackled in terms of the right component. The constraints we
have been discussing, in particular the Agreement Hierarchy, are violable at the ‘sentence level’. The
Agreement Hierarchy does not necessarily rule out specific sentences, as this Serbo-Croat example
shows.

(60) Dva tim-a, koj-a se nalaze u donjem delu tabel-e,

two team-SG.GEN, which-PL.N REFL find in lower part table-GEN,

Radnički i Olimpija, u Kragujevcu na teškom terenu igra-l-i

Radnički and Olimpija, in Kragujevac on difficult pitch play-PST-PL.M

su prljavo i nesportski.
AUX.3PL dirtily and unsportingly
‘Two teams, which find themselves in the lower part of the (league) table, Radnički and
Olimpija, on a difficult pitch in Kragujevac played dirtily and unsportingly.’
(Politika 9.XII.1969, from Sand 1971: 63-65)

Here we have semantic agreement of the predicate, but syntactic agreement of the relative pronoun.
We might have expected that it would be impossible to find semantic agreement at a point to the left
on the Agreement Hierarchy occurring together with syntactic agreement at a point to the right. This
example shows that that expectation is false. Such examples do occur, if relatively rarely. The
important claim is that at the level of the corpus the constraints of the Agreement Hierarchy will hold.
That is to say, overall the likelihood of semantic agreement will be greater with targets to the right on
the hierarchy (which is indeed the case with this type of controller in Serbo-Croat, as the data in Table
9 demonstrate) but this constraint need not apply at the level of the individual sentence, and so
 GREVILLE G. CORBETT 44

instances like (60) are not ruled out. (See Johnson and Postal 1980: 677-687 for a discussion of
sentence-level versus corpus-level constraints.)
In HP too, agreement is split, into index agreement, syntactic agreement and pragmatic
agreement (Pollard and Sag 1994: 60-62). Working within this framework, and attempting to take
account of the many instances of agreement choices, Wechsler and Zlatić (1998, 2001, 2003, partly
prefigured in Zlatić and Wechsler 1997) consider the types of information in lexical entries and
possible mismatches between them. They suggest that lexical entries include four types of information:

declension ⇔ concord ⇔ index ⇔ semantics

In a completely regular noun all the types of information correlate. But others may break the chain at
any of the three points given. Wechsler and Zlatić account for some of the interesting problems of
Serbo-Croat and they formalize their account, providing a fragment in HPSG. There are still few
analyses which address part of the complexity of the data and given a genuinely formal account.

3.3. Understanding agreement features

Many are interested in agreement as a syntactic problem. However, there is a tradition within Slavic
studies of interest in the agreement features themselves, particularly in gender, but also in number.
There is important work on establishing the features involved, by Zaliznjak (1964, 197333) and by
Gladkij (1969, 1973a, 1973b). And there are survey articles on the relevant features, like that of
Naylor (1978).

3.3.1. Genders and subgenders

Gender is a fascinating category, which I have reviewed both specifically in relation to Slavic (Corbett
1988) and more widely (Corbett 1991), and will not consider at length here. Since the Slavic systems
of gender assignment are of interest typologically (gender depends in part on inflectional class) a
formal account of the Russian system is given by Fraser and Corbett (1995) in the Network
Morphology framework. The rise of animacy is considered in section 4.3.2 below. A formal approach
to the question of subgenders, concentrating on Polish also within the Network Morphology
framework, is provided by Brown (1998).

33
While primarily on case, Zaliznjak (1973) is a major paper on the general issue of establishing the features and values
required for analysing a given language.
 AGREEMENT IN SLAVONIC 45

3.3.2. Number

Here too an overview of the Slavic systems is available, with extensive references (Corbett 1994) and
a general typology can be found in Corbett (2000). A bibliography of research on number in Slavic is
available.34

4. Topics for future investigation

There are several areas which clearly merit research. The fact that so much has already been done
makes it more likely that work in Slavic can lead to wider theoretical advances.

4.1. The typology of agreement systems

In some respects Slavic agreement systems are unremarkable typologically; after all, many languages
have subject-predicate agreement, agreement in gender and number within the noun phrase, and so on.
And yet there are points where Slavic is unusual, and these clearly deserve special attention.

4.1.1. Possessives in Upper Sorbian and Slovak

As noted in section 2.5.1, agreement of attributive modifiers with possessive adjectives is quite
remarkable. It deserves further work, particularly given the uncertain situation of Upper Sorbian.

4.1.2. Distant agreement

This was discussed in section 2.5.3; it is unusual in Slavic and unusual cross-linguistically; little more
is known. Megaard (1976: 80-81) suggests that distant agreement is possible in Serbo-Croat only if the
conjuncts are of the same number — all singular or all plural; see also Leko 1986: 230 for discussion
of the phenomenon. For this topic even the basic facts have not yet been established.

4.1.3. ‘Back’ agreement

With back agreement too (section 2.5.4) we have not got the essential data: which languages allow
back agreement, under what circumstances, optionally or obligatorily.

4.1.4. Triple agreement possibilities

When discussing alternative agreement possibilities, there is a tendency to assume there must be just
two options. This is normally the case, but there are also situations in which there are in principle at
least three possible agreements. Phrases headed by the Russian numeral tysjača ‘thousand’ can,

34
At http://www.surrey.ac.uk/LIS/SMG/number.html
 GREVILLE G. CORBETT 46

according to different authors, take feminine singular, neuter singular and plural agreement of the
predicate. But whether any individual speaker has three possibilities is not clear. Or there are
conjoined noun phrases in Serbo-Croat where the head nouns are feminines denoting abstracts: in
principle we might expect it to be possible to have agreement with just the nearest (feminine singular),
normal resolution (feminine plural), or the new type of resolution (masculine plural).35
Such three-way possibilities are relatively rare, and for this reason deserver a closer look. Are we
dealing merely with two binary choices which happen to converge, or does the third option introduce
anything new, in terms of variability, speaker uncertainty or even avoidance of certain constructions?

4.2. Modelling of variation

Slavic languages make it evident that agreement options are not temporary instabilities but are
endemic in agreement systems. The interactions of several competing factors lead to complex patterns
of variation. We should look to accounts which model the variation we observe. We might expect
statistical approaches to have a place. We should also expect models to be testable. This is happening
in morphology (thus Network Morphology accounts of Slavic data, like that of Corbett and Fraser
1993 have an implemented fragment, written in the lexical knowledge represent language DATR,
which can be used to check computationally that the theory does indeed give the output claimed). We
should expect that formal syntactic models too should more often be supported by implemented
versions rather than by hand-waving.

4.3. Diachronic perspectives

It is natural to ask how agreement systems change over time. The Slavic family provides interesting
data on a least four aspects of the question: the directionality of change, the rise of new feature values,
the rise of object agreement and the development of resolution rules.36

4.3.1. ‘There and back’ changes

There are plenty of examples in the literature of change, over varying periods of time, from an initial
state to a final state. We may come to expect that that is how change must be. But Slavic provides
evidence for change proceeding in one direction and then turning back. Consider these data on
predicate agreement in Russian.

35
Serbo-Croat deca ‘children’ takes feminine singular, neuter plural and masculine plural agreements; however, according
to the analysis in Corbett (1983a: 76-85), for any given target in a specified agreement domain there are never more than
two options with this controller.
36
For change in the conditions on agreement in Russian over the last two centuries, on-line searches can be run on the
Short-term Morphosyntactic Change Database (Krasovitsky, Brown, Corbett, Baerman, Long and Quilliam 2009).
 AGREEMENT IN SLAVONIC 47

Table 14. Predicate agreement with quantified subjects in Russian: 18–20th centuries

18th century 19th century 20th century

quantifier SG PL %PL SG PL %PL SG PL %PL
‘two’-‘four’ 55 357 87 54 357 87 146 710 83
‘five’-‘ten’ 19 34 64 74 82 53 110 110 50
collectives (e.g. dvoe ‘two’) 25 41 62 23 82 78 35 255 88
complex numbers (e.g. 48 26 35 71 62 47 103 65 39
dvadcat´ ‘20’)
compound numbers (e.g. sorok 45 24 35 72 71 50 57 40 41
pjat´ ‘45’)
neskol´ko ‘a few’ 105 28 21 151 139 48 137 78 36
mnogo ‘many’, skol´ko ‘how 42 0 0 not given 282 9 3
many’, stol´ko ‘so many’

These data (derived from Suprun 1969: 185, 188) show the particular quantifier involved has a major
influence on the form of predicate agreement, as discussed in section 2.1.1.3. What is of special
interest here is the change in the agreement options over time. The table shows that if we consider
complex numerals, compound numerals, or the quantifier neskol´ko ‘a few’, we find a rise in plural
agreement from the eighteenth to the nineteenth century, and a drop again from the nineteenth to the
twentieth century. Thus the change for these quantifiers moves in favour of semantic agreement and
then back in favour of syntactic agreement.

4.3.2. The rise of animacy

Animacy is an agreement category, reflected in agreement of the adjective and relative pronoun, and
also of the predicate verb in some languages. It is a sub-gender (Corbett 1991: 165-167), narrower in
range than the three main genders. It is a more recent addition to the system than the main genders,
and assignment to the animate sub-gender is (still) largely based on semantics. There are languages
like Russian where, with very few exceptions, only nouns which are semantically animate can be
grammatically animate. Then there are others in which animacy is becoming less clearly semantic;
thus for some speakers of Polish, in some syntactic environments, a whole range of nouns denoting
inanimates can be treated as animate, for example banan ‘banana’ (Wertz 1977: 57-59). If we look at
the agreement targets affected by animacy, then in the south-west (Serbo-Croat) we see that only the
singular of masculines is affected. In the north-east (Russian), all the plurals are affected in addition
(Huntley 1980).
There are further developments in some West Slavic languages, most notably in Polish, where
there is a distinction in the plural between nouns denoting male humans and all others. Assignment is
fairly strictly semantic, and the expression involves the nouns and the agreeing forms (within the noun
phrase and in the predicate). Here the masculine personal has unique forms of its own — they are not
 GREVILLE G. CORBETT 48

syncretic with others. Let us look at Upper Sorbian, which has the complexity of Polish together with
the dual number. It is described by Ermakova (1976) and Fasske (1981: 399–413); the relevant data
are given in Table 15.

Table 15. Genders and subgenders in Upper Sorbian

I II III IV V
NOM SG dobry susod dobry kóń dobry štom dobra žona dobre słowo
ACC SG dobreho susoda dobreho konja dobry štom dobru žonu dobre słowo
ACC DU dobreju susodow dobrej konjej dobrej štomaj dobrej žonje dobrej słowje
NOM PL dobri susodźi dobre konje dobre štomy dobre žony dobre słowa
‘good neighbour’ ‘good horse’ ‘good tree’ ‘good woman’ ‘good word’
MASCULINE MASCULINE MASCULINE FEMININE NEUTER
PERSONAL ANIMATE INANIMATE

The target gender forms illustrated in Table 15 (selected from Upper Sorbian’s three numbers and
seven cases) allow us to establish five agreement classes (I–V). The nominative and accusative
singular agreement forms provide the evidence to separate out agreement class IV (which is the
feminine gender) and agreement class V (the neuter). Of the remaining agreement classes, II differs
from III only in the accusative singular. Both have dependent (only syncretic) target forms, equivalent
to the genitive and nominative respectively. They are thus subgenders, masculine animate and
inanimate respectively. Agreement class I differs from them both in the accusative dual, a dependent
target gender form, and in the nominative plural. (It differs in other forms not included in Table 15.)
This agreement class (I), the masculine personal, can be treated as a separate gender since it varies in
four forms out of 21, (seven cases multiplied by three numbers, though there is considerable
syncretism) and it has an independent target gender form, the nominative plural (also arguably the
nominative dual). This system is presented in Figure 3.

singular dual plural

eho eju i
y

y ej e

Figure 3. The gender system of Upper Sorbian

 AGREEMENT IN SLAVONIC 49

The variety of animate and personal forms in Slavic, many of which are of relatively recent
origin, present a wonderful opportunity to see the rise of a new gender; it is easier to see how genders
are born by looking at the recent history of Slavic than by speculating on the earlier period of Indo-
European. For progress in this area specifically with regard to Russian see Krys´ko (1994) and
Timberlake (1997).

4.3.3. Clitic doubling

Clitic doubling was discussed briefly in section 3.1.2. It appears to be the beginning of object
agreement in Slavic. As such it deserves careful research. The differences already apparent in
Bulgarian and Macedonian, and the comparison with similar languages with clitics but no clitic
doubling, make it a potentially fruitful area.

4.3.4. The development of resolution rules in Serbo-Croatian

As we saw in section 2.4.3.3, the resolution rules of Serbo-Croatian are at a pivotal point where the
system type appears to be changing to what would be a new type for Slavic. This change would be
worth documenting.

4.4. Psycholinguistic work

Agreement has of late been proving of interest to psycholinguists; see, for example, Bock and Miller
(1991), Vigliocco, Butterworth, Semenza (1995) and Clahsen and Hansen (1997). Often they work on
systems which are rather less complex than those of Slavic. However, Kehayia, Jarema, and Kądziela
(1990) include Polish as one of the languages in a cross-linguistic comparison. Given the existence of
agreement systems in Slavic languages which vary in respect of quite small features, Slavic might well
prove fertile ground. We should hope to see increasing collaboration with psycholinguists.

4.5. The acquisition of agreement and agreement categories

Given the complexity of the factors involved in agreement and the interest of the agreement categories
for cognitive development, it would be natural to investigate their acquisition. There is fascinating
information on the acquisition of Polish in Smoczyńska (1985: 629-630, 637-639, 641-642, 644-646).
There is much more work that could be done in terms of acquisition, particularly with comparisons
across the Slavic languages. It is demanding work, requiring careful separation of the categories being
investigated and ideally a large corpus of acquisition data.
 GREVILLE G. CORBETT 50

4.6. Conversational data

Work on agreement in Slavic has in the main been firmly based on empirical data, as in the work of
Patton (1969), Suprun (1969), Sand (1971) and Robblee (1993b) for example. However, the data have
most often been from written language. With the increasing availability of large corpora, we can look
to further worthwhile work of this type. But detailed work on conversational data would be of
particular value. Initial research shows that agreement choices, which are of considerable theoretical
interest, occur relatively commonly in spoken language. A corpus of about 49,000 words of spoken
Russian (Zemskaja and Kapanadze 1978) was scanned for examples of three agreement choices: those
involving conjoined noun phrases, quantified expressions and relative kto ‘who’ (with a plural
antecedent). There were 22 examples, in other words examples occurred more frequently than once in
2,500 words. In addition, there were instances of other agreement choices which were not included in
the count. More generally, however, spoken data are where we must look for an understanding of the
function of agreement.

4.7. The function of agreement

Given its pervasiveness and complexity, it is reasonable to ask what agreement does. This question is
to be understood as relating to the agreement system as a whole. The old answer is that it introduces
redundancy, so that if part of the message is lost, owing to noise in the communication channel, there
is a greater chance that the original message can be reconstituted. This may well be part of the answer.
Researchers have also pointed to a more specific function of agreement, namely its role in allowing the
speaker to keep track of referents in a discourse, by means of the agreement categories (see Lehmann
1982: 233, 1988; Foley and Van Valin 1984: 327). This view is consistent with that of Barlow
(1988/1992: 3, 7), according to whom the controller and target ‘instigate discourse referents’. Pollard
and Sag (1988: 242) also see the role of agreement as being to keep track of referents in discourse. The
old and new suggestions as to the role of agreement may be seen as complementary. It may well be
that its functions and importance may vary considerably from language to language.
There are also more specific questions, relating to particular parts of the agreement system.
Concerning the function of agreement choices, Robblee (1993b: 433–437) considers singular and
plural agreement with quantified noun phrases and argues that agreement expresses the speaker’s view
of the event in terms of individuation, with plural agreement having an individuating function and
singular agreement being used for deindividuation. And then there is the relation between agreement
morphology and the possibility of dropping subject pronouns. Franks (1995: 297-298), in a discussion
of null subject phenomena, points out that in those Slavic languages where the copula and the past
tense auxiliary have ‘full-fledged’ agreement systems (including person agreement), that is in West
 AGREEMENT IN SLAVONIC 51

and South Slavic, subject pronouns are standardly omitted. And in East Slavic, where there is no
copula and past tense auxiliary showing person agreement, subject pronouns are not standardly
dropped. However, a study based on texts with translations in different Slavic languages shows the
picture to be more complex and gradient (Seo 2001).

5. Conclusion

Agreement in Slavic is an area that has already been relatively well researched. We have reasonable
accounts of the different agreement systems, and ample evidence of the pervasive nature of options in
agreement systems. The nature of the work that has been done makes Slavic agreement a potentially
excellent area for new research, of various kinds. For instance, we know a good deal about individual
factors which affect agreement choices, but much less on how they interact. We know something
about the adult systems, but rather little about how they are acquired and what their function is. And
the Slavic languages are sufficiently similar and sufficiently different to provide an attractive research
laboratory.
Research on agreement continues steadily. In particular, the conditions on different agreement
options in Slavic prove as fascinating as ever. They are tackled in an HPSG framework in a series of
papers, concentrating on Serbian/Croatian, by Stephen Wechsler and Larisa Zlatić, notably in their
2001 paper. The various papers are drawn together in Wechsler and Zlatić (2003). Various interesting
options in South Slavic are also documented by Mladenova (2001). Her data are directly relevant to
the Agreement Hierarchy, which is specifically tested in Leko (2000), using Bosnian data and in
Igartua (2004), looking at Old Russian. The relevant evidence on the Agreement Hierarchy, from
Slavic and beyond, is summarized in Corbett (2006). Agreement in Czech is discussed both by
Panevová and Petkevič (1997) and by Veselovská (2001). The diachrony of agreement is the subject of
Igartua (2000). A particularly welcome departure is the use of psycholinguistic techniques to elucidate
agreement, as in Nicol and Wilson (2000) and Rusakova (2001).

References

Anderson, Stephen R. 1992. A-Morphous Morphology (Cambridge Studies in Linguistics 62).

Cambridge: Cambridge University Press.
Babby, Leonard H. 1987. Case, prequantifiers, and discontinuous agreement in Russian. Natural
Language and Linguistic Theory 5. 91-138.
Barlow, Michael. 1988. A situated theory of agreement. Doctoral dissertation, Stanford University,
Stanford. Published 1992, New York: Garland.
 GREVILLE G. CORBETT 52

Barlow, Michael. 1991. The Agreement Hierarchy and grammatical theory. In: Laurel A. Sutton,
Christopher Johnson and Ruth Shields (eds) Proceedings of the Seventeenth Annual Meeting of
the Berkeley Linguistics Society, February 15-18, 1991: General Session and Parasession on the
Grammar of Event Structure, 30–40. Berkeley: Berkeley Linguistics Society, University of
California.
Bayer, Sam and Johnson, Mark. 1995. Features and Agreement. In: Proceedings of the 33rd Annual
Meeting of the Association for Computational Linguistics. Morristown, NJ.
Bock, Kathryn and Miller, Carol A. 1991. Broken agreement. Cognitive Psychology 23.45-93.
Bogdanov, V. N. 1968. Osobyj slučaj dialektnogo soglasovanija skazuemogo s podležaščim po smyslu
i kategorija predstavitel´nosti. Naučnye doklady vysšej školy: filologičeskie nauki no. 4. 68-75.
Bresnan, Joan and Mchombo, Sam A. 1987. Topic, pronoun, and agreement in Chichewfla. Language
63.741-82.
Brown, Dunstan. 1998. Defining ‘subgender’: virile and devirilised nouns in Polish. Lingua 104.187-
233.
Browne, Wayles. 1990. Kako kongruirati sa infinitivom? Sistem pojmova za opis slaganja u
srpskohrvatskom jeziku. Naučni Sastanak Slavista u Vukove Dane: Referati i saopštenja 20,
2.157-163.
Browne, Wayles. 1993. Serbo-Croat. In: Bernard Comrie and Greville G. Corbett. (eds) The Slavonic
Languages, 306-387. London: Routledge.
Browne, Wayles. 1998. Vsjako načalo e trudno: Traps in lesson I in the South Slavic languages. In:
Mila Dimitrova-Vulchanova, Lars Hellan, Ivan Kasabov and Ilyana Krapova (eds) Papers from
the Second Conference on Formal Approaches to South Slavic Languages, Sofia 26-29
September 1997 (=Trondheim Working Papers in Linguistics 31).
Bukatevič, N. I., Gricjutenko, I. E., Miževskaja, G. M., Pavljuk, N. V., Savickaja, S. A. and
Smaglenko, F. P. 1958. Očerki po sravnitel´noj grammatike vostočnoslavjanskix jazykov.
Odessa: Odesskij gosudarstvennyj universitet im. I. I. Mečnikova. [Reprinted as Slavistic
Printings and Reprintings no. 139, 1969, The Hague, Mouton.]
Bybee, Joan. 1985. Morphology: a study of the relation between meaning and form (Typological
Studies in Language 9). Amsterdam: Benjamins.
Clahsen, Harald and Hansen, Detlef. 1997. The grammatical agreement deficit in specific language
impairment: Evidence from therapy experiments. In: Myrna Gopnik (ed.) The Inheritance and
Innateness of Grammars, 141-60. New York: Oxford University Press.
Comrie, Bernard. 1975. Polite plurals and predicate agreement. Language 51.406-18.
Corbett, Greville G. 1979a. Predicate Agreement in Russian (Birmingham Slavonic Monographs, 7).
University of Birmingham, Department of Russian Language and Literature.
 AGREEMENT IN SLAVONIC 53

Corbett, Greville G. 1979b. The agreement hierarchy. Journal of Linguistics 15.203–24.

Corbett, Greville G. 1983a. Hierarchies, Targets And Controllers: Agreement Patterns in Slavic.
London: Croom Helm.
Corbett, Greville G. 1983b. Resolution rules: agreement in person, number and gender. In: Gerald
Gazdar, Ewan Klein and Geoffrey Pullum (eds) Order, Concord and Constituency (Linguistic
Models, 4), 175–206. Dordrecht: Foris.
Corbett, Greville G. 1983c. Slaganje predikata sa više subjekata u srpskohrvatskom jeziku. Naučni
Sastanak Slavista u Vukove Dane: Referati i saopštenja 12, 1.93–102.
Corbett, Greville G. 1986. Agreement: a partial specification, based on Slavonic data. Linguistics 24,
no. 6. 995-1023.
Corbett, Greville G. 1987. The morphology/syntax interface: evidence from possessive adjectives in
Slavonic. Language 63.299-345.
Corbett, Greville G. 1988. Gender in Slavonic from the standpoint of a general typology of gender
systems. Slavonic and East European Review 66.1–20.
Corbett, Greville G. 1991. Gender. Cambridge: Cambridge University Press.
Corbett, Greville G. 1993. The head of Russian numeral expressions. In: Greville G. Corbett, Norman
M. Fraser and Scott McGlashan (eds) Heads in Grammatical Theory, 11-35. Cambridge:
Cambridge University Press.
Corbett, Greville G. 1994. Systems of grammatical number in Slavonic. Slavonic and East European
Review 72.201-217.
Corbett, Greville G. 1995. Slavonic’s closest approach to Suffix Copying: the possessive adjective. In:
Frans Plank (ed.) Double Case: Agreement by Suffixaufnahme, 265-282. New York: Oxford
University Press.
Corbett, Greville G. 1998. Morphology and agreement. In: Andrew Spencer and Arnold Zwicky (eds)
The Handbook of Morphology, 191-205. Oxford: Blackwells.
Corbett, Greville G. 2000. Number. Cambridge: Cambridge University Press.
Corbett, Greville G. 2003. Types of typology, illustrated from gender systems. In: Frans Plank (ed.)
Noun Phrase Structure in the Languages of Europe (Empirical Approaches to Language
Typology EUROTYP 20-7), 289-334. Berlin: Mouton de Gruyter.
Corbett, Greville G. 2006. Agreement. Cambridge: Cambridge University Press.
Corbett, Greville G. 2009. Morphology-free syntax: two potential counter-examples from Serbo-Croat.
In: Steven Franks, Vrinda Chidambaram and Brian Joseph (eds) A Linguist's Linguist: Studies in
South Slavic Linguistics in Honor of E. Wayles Browne. Bloomington, Indiana: Slavica.
Corbett, Greville G. and Fraser, Norman M. 1993. Network Morphology: A DATR account of Russian
inflectional morphology. Journal of Linguistics 29.113-142.
 GREVILLE G. CORBETT 54

Corbett, Greville G. and Mithun, Marianne. 1996. Associative forms in a typology of number systems:
evidence from Yup’ik. Journal of Linguistics 32.1-17.
Crockett, Dina B. 1976. Agreement in Contemporary Standard Russian. Cambridge, Mass.: Slavica.
Doroszewski, W. 1962. O kulturę słowa: poradnik językowy. Warsaw: PIW.
Dyła, Stefan 1988. Quasi-comitative coordination in Polish. Linguistics 26.383-414.
Dziwirek, Katarzyna. 1990. Default agreement in Polish. In: Katarzyna Dziwirek, Patrick Farrell and
Errapel Majías-Bikandi (eds) Grammatical Relations: A Cross-theoretical Perspective, 147-61.
Stanford: Stanford Linguistics Association/CSLI.
Ermakova, M. I. 1976. Sopostavitel´noe opisanie sistem soglasovatel´nyx klassov i morfologičeskaja
klassifikacija suščestvitel´nyx v sovremennom serbolužickom jazyke. In: Slavjanskoe i
balkanskoe jazykoznanie: Problemy morfologii sovremennyx slavjanskix jazykov, 40-63.
Moscow: Nauka.
Farkas, Donka and Ojeda, Almerindo. 1983. Agreement and coordinate NPs. Linguistics 21.659–673.
Farkaş, Donka F. and Zec, Draga 1995. Agreement and pronominal reference. In: Guglielmo Cinque
and Giuliana Giusti (eds) Advances in Roumanian Linguistics (Linguistik Aktuell 10), 83-101.
Amsterdam: John Benjamins.
Fasske, Helmut. 1981. Grammatik der obersorbischen Schriftsprache der Gegenwart: Morphologie.
Bautzen: VEB Domowina Verlag.
Fasske, Helmut. 1996. Sorbischer Sprachatlas: 15: Syntax. Bautzen: Domowina Verlag.
Foley, William A. and Van Valin, Robert D. 1984. Functional Syntax and Universal Grammar.
Cambridge: Cambridge University Press.
Franks, Steven. 1995. Parameters of Slavic Morphosyntax. New York: Oxford University Press.
Fraser, Norman M. and Corbett, Greville G. 1995. Gender, animacy and declensional class
assignment: a unified account for Russian. In: Geert Booij and Jaap van Marle (eds) Yearbook of
Morphology 1994, 123-50. Dordrecht: Kluwer.
Friedman, Victor A. 1993. Macedonian. In: Bernard Comrie and Greville G. Corbett (eds) The
Slavonic Languages, 249-305. London: Routledge.
Friedman, Victor A. 1994. Variation and grammaticalization in the development of Balkanisms. In:
Katharine Beals, Jeannette Denton, Robert Knippen, Lynette Melnar, Hisami Suzuki and Erica
Zeinfeld (eds) CLS 30: Papers from the 30th Regional Meeting of the Chicago Linguistic
Society: 1994: 2: The Parasession of Variation in Linguistic Theory, 101-115. Chicago:
Chicago Linguistic Society,.
Gazdar, Gerald; Klein, Ewan; Pullum Geoffrey K., Sag, Ivan A. 1985. Generalized Phrase Structure
Grammar. Oxford: Blackwell.
 AGREEMENT IN SLAVONIC 55

Gladkij, A. 1969. K opredeleniju ponjatij padeža i roda suščestvitel´nogo. Voprosy jazykoznanija no.
2.110-23.
Gladkij, A. 1973a. Popytka formal´nogo opredeleniju ponjatij padeža i roda suščestvitel´nogo. In: A.
A. Zaliznjak (ed.) Problemy grammatičeskogo modelirovanija, 24-53. Moscow: Nauka.
Gladkij, A. 1973b. An attempt at the formal definition of case and gender of the noun. In: F. Kiefer
(ed.) Mathematical Models of Language, 159-204. Stockholm: Skriptor.
Glavan, V. 1927-1928. Kongruencija u jeziku starih čakavskih pisaca. Južnoslovenski filolog 7.111-
159.
Graudina, L. K., Ickovič, V. A. and Katlinskaja, L. P. 1976. Grammatičeskaja pravil´nost´ russkoj
reči: opyt častotno-stilističeskogo slovarja variantov. Moscow: Nauka.
Gudkov, V. [P.] 1965. Dodatak pravilima slaganja predikata sa više subjekata. Književnost i jezik 12.
60-61.
Gudkov, V. P. 1974. Prilog o pravilima kongruencije. Književnost i jezik 21.58–61.
Huntley, David. 1980. The evolution of genitive-accusative animate and personal nouns in Slavic
dialects. In: J. Fisiak (ed.) Historical Morphology (Trends in Linguistics: Studies and
Monographs 17), 189-212. The Hague: Mouton.
Huntley, David. 1989. Grammatical and lexical features in number and gender agreement in Old
Bulgarian. Paleobulgarica 13. 21–32.
Igartua, Iván. 2000. Osobennosti razvitija soglasovanija v istorii russkogo jazyka. Russian Linguistics
24.287-320.
Igartua, Iván. 2004. Sobiratel´nye suščestvitel´nye i ierarxija soglasovanija v drevnerusskom jazyke.
Die Welt der Slaven 49.229-46.
Iomdin, Leonid L. 1990. Avtomatičeskaja obrabotka teksta na estestvennom jazyke: model´
soglasovanija. Moscow: Nauka.
Johnson, David E. and Postal, Paul M. 1980. Arc Pair Grammar. Princeton: Princeton University
Press.
Kathol, Andreas. 1999. Agreement and the syntax-morphology interface in HPSG. In: Robert Levine
and Georgia Green (eds) Studies in Contemporary Phrase Structure Grammar, 209-260.
Cambridge: Cambridge University Press.
Keenan, Edward. 1974. The functional principle: generalizing the notion ‘subject of’. In: Michael W.
La Galy, Robert A. Fox and Anthony Bruck (eds) Papers from the Tenth Regional Meeting,
Chicago Linguistic Society, April 19-21, 1974, 298–309. Chicago: Chicago Linguistic Society.
Kehayia, Eva; Jarema, Gonia and Kądziela, Danuta. 1990. Cross-linguistic study of morphological
errors in aphasia: evidence form English, Greek, and Polish. In: Jean-Luc Nespoulous and Pierre
Villiard (eds) Morphology, Phonology and Aphasia, 140-55. New York: Springer.
 GREVILLE G. CORBETT 56

Kibrik, A. E. 1977. O sootnošenii ponjatija sintaksičeskogo podčinenija s ponjatijami soglasovanija,

upravlenija i primykanija. In: Problemy teoretičeskoj i èksperimental´noj lingvistiki: Sbornik
statej (Publikacii otdelenija strukturnoj i prikladnoj lingvistiki, under the general editorship of V.
A. Zvegincev, vyp. 8), pp. 161-179. Slightly amended version in A. E. Kibrik 1992. Očerki po
obščim i prikladnym voprosam jazykoznanija (universal´noe, tipovoe i specifičnoe v jazyke), 102-
123. Moscow: Izdatel´stvo Moskovskogo universiteta.
Kirby, Simon. 1999. Function, Selection and Innateness: the Emergence of Language Universals.
Oxford: Oxford University Press.
2009
Krasovitsky, Alexander; Brown, Dunstan; Corbett, Greville G.; Baerman, Matthew; Long, Alison; and
Quilliam, Harley. Database of Short Term Morphosyntactic Change. [available at
<http://www.surrey.ac.uk/LIS/SMG/STMC/> ]
Krys´ko, V. B. 1994. Razvitie kategorii oduševlennosti v istorii russkogo jazyka. Moscow: Lyceum.
Lehfeldt, Werner. 1980. Upravlenie, soglasovanie, primykanie v russkom jazyke: popytka
operacional´nogo opredelenija. Russian Linguistics 4. 249-267.
Lehmann, Christian. 1982. Universal and typological aspects of agreement. In: Hansjakob Seiler and
Franz J. Stachowiak (eds) Apprehension: Das sprachliche Erfassen von Gegenständen: II: Die
Techniken und ihre zusammenhang in Einzelsprachen, 201-67. Tübingen: Narr.
Lehmann, Christian. 1988. On the function of agreement. In: Michael Barlow and Charles A. Ferguson
(eds) Agreement in Natural Language: Approaches, Theories, Descriptions, 55-65. Stanford:
Center for the Study of Language and Information, Stanford University.
Leko, Nedžad. 1986. Syntax of noun headed structures in Serbo-Croatian and corresponding phrasal
structures in English. Ph.D. dissertation, Indiana University. Distributed by University
Microfilms, Ann Arbor, reference 86–28003.
Leko, Nedžad. 2000. Syntactic vs. semantic agreement in the Oslo corpus. In: Tracy Holloway King
and Irina A. Sekerina (eds) Annual workshop on Formal Approaches to Slavic Linguistics 8: The
Philadelphia Meeting 1999 (Michigan Slavic Materials 45), 259-78. Ann Arbor, MI.: Michigan
Slavic Publications.
Lenček, Rado. 1972. O zaznamovanosti in nevtralizaciji slovnične kategorije spola v slovenskem
knjižnem jeziku. Slavistična revija 20.55-63.
McNally, Louise. 1993. Comitative coordination: a case study in group formation. Natural Language
and Linguistic Theory 11.347-379.
Marković, Svetozar V. 1954. O kolebljivosti slaganja u rodu kod imenica čiji se prirodni i gramatički
rod ne slažu (i o rodu ovih imenica). Pitanja književnosti i jezika (Sarajevo) 1. 87–110.
Megaard, J. 1976. Predikatets kongruens i serbokroatisk i setninger med koordinerte
subjektsnominalfraser. Unpublished dissertation, University of Oslo.
 AGREEMENT IN SLAVONIC 57

Mel´čuk, Igor´ A. 1993. Soglasovanie, upravlenie, kongruèntnost´. Voprosy jazykoznanija no. 5.16-58.
Mladenova, Olga M. 2001. Neuter designations of humans and norms of social interactions in the
Balkans. Anthropological Linguistics 43.18-53.
Moravcsik, Edith. 1994. Group plural — associative plural or cohort plural. Email document,
LINGUIST List: Vol-5-681. 11 June 1994. ISSN: 1068-4875.
Morgan, Jerry L. 1984. Some problems of agreement in English and Albanian. Proceedings of the
Tenth Annual Meeting of the Berkeley Linguistic Society, 233–247. Berkeley: University of
California Press.
Naylor, Kenneth E. 1978. Some observations on the grammatical categories of the Slavic declensions.
In: Henrik Birnbaum (ed.) American Contributions to the 8th International Congress of Slavists,
I: Linguistics and Poetics, 532-552. Columbus, Ohio: Slavica.
Nichols, Johanna. 1992. Linguistic diversity in space and time. Chicago: University of Chicago Press.
Nichols, Johanna; Rappaport, Gilbert and Timberlake, Alan. 1980. Subject, topic and control in
Russian. Proceedings of the Sixth Annual Meeting of the Berkeley Linguistics Society, 372-386.
Berkeley: Berkeley Linguistics Society, University of California.
Nicol, Janet and Wilson, Rachel. 2000. Agreement and case-marking in Russian: A psycholinguistic
investigation of agreement errors in production. In: Tracy Holloway King and Irina A. Sekerina
(eds) Annual workshop on Formal Approaches to Slavic Linguistics 8: The Philadelphia Meeting
1999 (Michigan Slavic Materials 45), 314-27. Ann Arbor: Michigan Slavic Publications.
Panevová, Jarmila and Petkevič, Vladimír. 1997. Agreement in Czech and its formal account. In: Uwe
Junghanns and Gerhild Zybatow (eds) Formale Slavistik, 321-33. Frankfurt: Vervuert Verlag.
Pateman, Trevor. 1987. Philosophy of Linguistics. In: John Lyons, Richard Coates, Margaret Deuchar
and Gerald Gazdar (eds) New Horizons in Linguistics 2, 249-267. London: Penguin.
Patton, Helen B. 1969. A Study of the Agreement of the Predicate with a Quantitative Subject in
Contemporary Russian. Ph.D. dissertation, University of Pennsylvania. Distributed by University
Microfilms, Ann Arbor, 7O-7839.
Pesetsky, David. 1982. Paths and categories. Unpublished PhD dissertation, MIT.
Pollard, Carl and Sag, Ivan A. 1994. Head-Driven Phrase Structure Grammar. Chicago: University of
Chicago Press.
Popov, K. 1964. Sintaktičnoto săglasuvane v băl garski ezik. Sofia: Nauka i izkustvo.
Priestly, T. M. S. 1993. Slovene. In: Bernard Comrie and Greville G. Corbett (eds) The Slavonic
Languages, 388- 51. London: Routledge.
Pullum, Geoffrey K. and Zwicky, Arnold. 1988. The syntax-phonology interface. In: Frederick J.
Newmeyer (ed.) Linguistics: The Cambridge Survey: I: Linguistic Theory: Foundations,
255-280. Cambridge: Cambridge University Press.
 GREVILLE G. CORBETT 58

Robblee, Karen E. 1993a. Predicate lexicosemantics and case marking under negation in Russian.
Russian Linguistics 17.209-236.
Robblee, Karen E. 1993b. Individuation and Russian agreement. Slavic and East European Journal
37.423-441.
Robblee, Karen E. 1997. The interaction of word order, agreement and case marking. In: Anne-Marie
Simon-Vandenbergen, Kristin Davidse and Dirk Noël (eds) Reconnecting Language:
Morphology and Syntax in Functional Perspectives, 227-248. Amsterdam: John Benjamins.
Rothstein, Robert A. 1993. Polish. In: Bernard Comrie and Greville G. Corbett (eds) The Slavonic
Languages, 686-758. London: Routledge.
Rudin, Catherine. 1986. Aspects of Bulgarian Syntax: Complementizers and WH Constructions.
Columbus, Ohio: Slavica.
Rusakova, Marina V. 2001. Imennaja slovoforma flektivnogo jazyka (soglasovanie v russkom
atributivnom slovosočetanii). Unpublished PhD dissertation, St Petersburg: Gosudarstvennyj
pedagogičeskij universitet im. A. I. Gercena.
Sag, Ivan A.; Gazdar, Gerald; Wasow, Thomas and Weisler, Steven. 1985. Coordination and how to
distinguish categories. Natural Language and Linguistic Theory 3.117–171.
Sand, Diane E. Z. 1971. Agreement of the Predicate with Quantitative Subjects in Serbo-Croatian,
Ph.D. dissertation, University of Pennsylvania. Distributed by University Microfilms, Ann
Arbor, 72–17,420.
Scatton, Ernest A. 1993. Bulgarian. In: Bernard Comrie and Greville G. Corbett (eds) The Slavonic
Languages, 188-248. London: Routledge.
Schmidt, Peter and Lehfeldt, Werner. 1995. Kongruenz, Rektion, Adjunktion: Systematische und
historische Untersuchungen zur allgemeinen Morphosyntax und zu den Wortfügung
(slovosočetanija) im Russischen (Specimina Philologiae Slavicae, Supplementband 37).
München: Otto Sagner.
Schwartz, Linda. 1988. Conditions for verb-coded coordinations. In: Michael Hammond, Edith
Moravcsik and Jessica Worth (eds) Studies in Syntactic Typology (Typological Studies in
Language 17), 53-73. Amsterdam: Benjamins.
Seo, Seunghyun. 2001. The frequency of null subject in Russian, Polish, Czech, Bulgarian and Serbo-
Croatian: an analysis according to morphosyntactic environments. PhD dissertation, Indiana
University. Distributed by UMI, Ann Arbor, reference 3038515.
Shieber, Stuart M. 1986. An Introduction to Unification-Based Approaches to Grammar (CSLI lecture
notes 4). CSLI, Stanford University, Stanford.
Shieber, Stuart M. 1992. Constraint-Based Grammar Formalisms: Parsing and Type Inference for
Natural and Computer Languages. Cambridge, MA: MIT Press.
 AGREEMENT IN SLAVONIC 59

Skoblikova, E. S. 1971. Soglasovanie i upravlenie v russkom jazyke. Moscow: Prosveščenie.

Smith, Neil; Tsimpli, Ianthi-Maria and Ouhalla, Jamal. 1993. Learning the impossible: the acquisition
of possible and impossible languages by a polyglot savant. Lingua 91.279-347.
Smoczyńska, Magdalena. 1985. The acquisition of Polish. In: Dan I. Slobin, (ed.) The Crosslinguistic
Study of Language Acquisition: I: The Data, 595-686. Hillsdale, N.J.: Lawrence Erlbaum.
Steele, Susan. 1978. Word order variation: a typological study In: Joseph H. Greenberg, Charles A.
Ferguson and Edith A. Moravcsik (eds) Universals of Human Language: IV: Syntax, 585-623.
Stanford: Stanford University Press.
Stone, Gerald. 1993. Cassubian. In: Bernard Comrie and Greville G. Corbett. (eds) The Slavonic
Languages, 75 -794. London: Routledge.
Stroińska, Magda. 1992. Numerals and agreement in Polish. Canadian Slavonic Papers 34.429-444.
Šul´ga, M. V. 1997. Slavjanskij grammatičeskij rod: privativnaja oppozicija. Voprosy jazykoznanija
no. 3. 26-39.
Suprun, A. E. 1961. Staroslavjanskie čislitel´nye. Frunze: Kirghiz State University.
Suprun, A. E. 1963a. O soglasovanii skazuemogo s podležaščim, vključajuščim količestvennye
čislitel´nye v serbo-lužickix jazykax. Serbo-lužickij lingvističeskij sbornik, 138-53. Moscow.
Suprun, A. E. 1963b. Zametki po sintaksisu pol´skix čislitel´nyx. Pytannja slovjans´koho
movoznavstva (L´vov) 7-8.135-45.
Suprun, A. E. 1969. Slavjanskie čislitel´nye (stanovlenie čislitel´nyx kak osoboj časti reči). Minsk:
Belorussian State University.
Szupryczyńska, Maria. 1991. Związki składniowe w konstrukcjach z tzw. „podmiotem
towarzyszącym”. In: Maciej Grochowski and Daniel Weiss (eds) "Words are physicians for an
ailing mind" (Sagners Slavistische Sammlung 17), 415-420. Munich: Otto Sagner.
Timberlake, Alan. 1993. Russian. In: Bernard Comrie and Greville G. Corbett (eds) The Slavonic
Languages, 827-886. London: Routledge.
Timberlake, Alan. 1997. Čemu jesi slěpilъ brat svoi: templates and the development of animacy.
Russian Linguistics 21.49-62.
Vanek, Anthony L. 1970. Aspects of subject-verb agreement. Edmonton: Department of Slavic
Languages, University of Alberta. [Republished 1977 in the series Current inquiry into language
and linguistics, 23. Edmonton: Linguistic Research.]
Večerka, R. 1960. K sintaksisu imen čislitel´nyx v staroslavjanskom jazyke. In: V. Georgiev (ed.)
Ezikovedsko-etnografski izsledvanija v pamet na akademik Stojan Romanski , 195-208. Sofia:
Bulgarian Academy of Sciences.
 GREVILLE G. CORBETT 60

Veselovská, Ludmila. 2001. Agreement patterns of Czech group nouns and quantifiers. In: Norbert
Corver and Henk van Riemsdijk (eds) Semi-lexical Categories: The Function of Content Words
and the Content of Function Words (Studies in Generative Grammar 59), 274-320. Berlin: Mouton
de Gruyter.
Vigliocco, Gabriella; Butterworth, Brian and Semenza, Carlo. 1995. Constructing subject-verb
agreement in speech: the role of semantic and morphological factors. Journal of Memory and
Language 34.186--215.
Vincenot, C. 1975. Essai de grammaire slovène. Ljubljana: Mladinska knjiga.
Vinogradov, V. V. and Istrina, E. S. (eds) 1954. Grammatika russkogo jazyka: II: Sintaksis, čast´
pervaja. Moscow: AN SSSR.
Wechsler, Stephen and Zlatić, Larisa. 1997. Case and agreement with Serbian quantified NP’s. Paper
read at the 16th International Congress of Linguistics, Paris, July 1997.
Wechsler, Stephen and Zlatić, Larisa. 1998. A theory of agreement and disagreement. In: B. K.
Bergen, M. C. Plauche and A. C. Bailey (eds) Proceedings of the Berkeley Linguistics Society
1998, 280-291. Berkeley: Berkeley Linguistics Society, University of California.
Wechsler, Stephen and Zlatić, Larisa. 2000. A theory of agreement and its application to Serbo-
Croatian. Language 76.799-832.
Wechsler, Stephen and Zlatić, Larisa. 2003. The Many Faces of Agreement. Stanford: CSLI.
Weiss, Daniel. 1985. Frau und Tier in der sprachlichen Grauzone: diskriminierende Strukturen
slavischer Sprachen. In: Werner Lehfeldt (ed.) Slavistische Linguistik 1984. Referate des X.
Konstanzer Slavistischen Arbeitstreffens, Konstanz, 11. mit 14. 9. 1984. 317-358. Munich: Otto
Sagner.
Wertz, C. A. 1977. The number of genders in Polish. Canadian Slavonic Papers 19.50–63.
Wunderlich, Dieter. 1993. Funktionale Kategorien im Lexicon. In: Frank Beckmann and Gerhard
Heyer (eds) Theorie und Praxis des Lexikons, 54-73. Berlin: de Gruyter.
Zaliznjak, A. A. 1964. K voprosu o grammatičeskix kategorijax roda i oduševlennosti v sovremennom
russkom jazyke.Voprosy jazykoznanija no. 4.25-40.
Zaliznjak, Andrej A. 1973. O ponimanii termina ‘padež’ v lingvističeskix opisanijax. In: Andrej A.
Zaliznjak (ed.) Problemy grammatičeskogo modelirovanija, 53-87. Moscow: Nauka, [Reprinted
in: Andrej A. Zaliznjak. 2002. Russkoe imennoe slovoizmenenie: s priloženiem izbrannyx rabot
po sovremennomu russkomu jazyku i obščemu jazykoznaniju, 613-47. Moscow: Jazyki
slavjanskoj kul´tury.]
Zemskaja, E. A. and Kapanadze, L. A. (eds) 1978. Russkaja razgovornaja reč´: teksty. Moscow:
Nauka.
 AGREEMENT IN SLAVONIC 61

Zieniukowa, J. 1979. Składnia zgody w zdaniach z podmiotem szeregowym we współczesnej

polszczyźnie. Slavia Occidentalis 36.117-129.
Zlatić, Larisa and Wechsler, Stephen. 1997. Mixed agreement in Serbian: a constraint based approach.
In: Brian Agbayani and Sze-Wing Tang (eds) The Proceedings of the Fifteenth West Coast
Conference on Formal Linguistics, 521–535. Stanford: CSLI.
Zwicky, Arnold M. 1992. Jottings on adpositions, case inflections, government, and agreement. In:
Diane Brentari, Gary N. Larson and Lynn A. MacLeod (eds) The Joy of Grammar: A Festschrift
in honor of James D. McCawley, 369-83. Amsterdam: Benjamins.