Andrej Malchukov & Andrew Spencer (eds.

)
The Handbook of Case

Spatial Cases
Denis Creissels
Université Lumière (Lyon 2)
[email protected]

(second version, submitted 23/10/06)

Acknowledgements

I thank the following colleagues for their comments and suggestions: Gilles Authier, Misha Daniel, Bernard Fradin,
Sylvain Patri, and an anonymous reviewer.

1. Introductory remarks

1.1.Definitions

In this chapter, the term case is taken in its traditional meaning of inflectional category-system
(and the individual categories or values of that system) expressing dependency relations involving
NPs.1
A spatial relation involves two percepts, a Figure (or Theme, or Trajector) and a Ground (or
Location, or Landmark), the Figure being perceived as located or in motion relative to the
Ground.2
A spatial case is an inflected form of nouns or NPs distinct from the absolute form available for
the extra-syntactic function of pure designation, and apt to fulfill one of the following functions
without the addition of an adposition:
– non-verbal predicate, or predicative complement of a copula, specifying the location of an
entity,
– verb satellite specifying the location of an event,
– argument of motion verbs specifying the source, path, or destination of the movement.
It may happen that the same case form has some uses corresponding to this definition along with
non-spatial uses (see sections 4 and 5).
Some authors have advocated that all cases can be viewed as locative expressions (see in
particular Hjelmslev 1935, Hjelmslev 1937, Anderson 1971, Cook 1989). Since we are concerned
with cases that have concrete spatial meanings (and not with cases having only non-spatial uses
derivable from abstract locative meanings) this question will not be further developped here.

1 For a discussion of the various extensions of the term case encountered in the literature, see Haspelmath (this
volume).
2 For a general approach to the study of the linguistic expression of spatial relations, see e.g. Jackendoff 1983,
Langacker 1987, Jackendoff & Landau 1992, Svorou 1994, Pederson 1995, Pederson & al. 1998, Talmy 2000. Shay &
Seibert 2003 provides a collection of papers exploring the variety of the linguistic means of expressing spatial relations
in typologically diverse languages.

–1–
 1.2. Simple and complex spatial cases

Morphologically complex spatial cases may include a formative having an independent existence
as a spatial case marker, combined with an extension, i.e., a formative that has no independent
existence (for example, Basque directional -ra-ntz ‘towards’ and terminative -ra-ino ‘up to’ combine
allative -ra with two extensions that exist only in combination with the allative marker).
It may also happen that none of the formatives involved in the formation of complex spatial case
marker has an independent existence. In Akhvakh, all spatial cases arise from a combination of
markers for relative orientation with markers for destination, location or source, but none of the
distinctions underlying this system involves zero marking.
Spatial cases arising from a combination of 3 formatives are attested. In some Daghestanian
languages, cases markers encoding path consist of a first formative for relative orientation, a second
formative that by itself carries an ablative meaning, and an third formative, responsible for the
meaning of path, that exists only as an extension of the ablative marker. In Tsez, the possible
addition of a distal marker is responsible for the existence of spatial cases consisting of 4 formatives
(Comrie & al. 1999).

1.3. Problems in separating spatial cases from spatial adpositions

It is not possible to discuss here the criteria according to which descriptive linguists decide to
treat simple markers encoding spatial meanings and located at the edge of noun phrases as spatial
cases or spatial adpositions. In the case of complex markers, three options are available
(combination of two affixes, of an affix and an adposition, or of two adpositions), and the criteria
that may lead to choose between these three options cannot be discussed here either:

‘Although one can easily separate different layers of case marking in a particular language, as in Hindi for instance,
it can be difficult to determine whether a single layer of case marking in a particular language is affixal or
adpositional.’
(Blake 1994:11)

Consequently, in the search for typological generalizations concerning spatial cases in the
narrow sense of this term, one must always keep in mind that there is some degree of arbitrariness
in the distinction between cases affixes and adpositions as it is recognized in the descriptions of
individual languages. It is nevertheless interesting to take this distinction into account, because
spatial relation markers that are clearly affixal lend themselves to some generalizations that do not
apply to those that are clearly adpositional and vice versa.

1.4. Spatial cases in languages devoid of case contrast between core syntactic terms

Most of the time, noun forms identifiable as spatial cases according to the definition put forward
in the introduction are observed in languages in which morphological variations of nouns
contribute to mark the contrast between syntactic core termes (nominative vs. accusative, or
absolutive vs. ergative), but there are exceptions. Tswana has no case contrast between syntactic
core terms, but Tswana nouns have a locative form. Nahuatl is another case in point.

1.5. Case languages devoid of spatial cases stricto sensu

A language may have case inflection without having spatial cases. For example, none of the three
cases of Classical Arabic (nominative, accusative, and genitive) can be used by itself in spatial
functions.
In case languages devoid of spatial cases proper, the functions typical of spatial cases are
expressed by means of adpositional phrases. This is a common situation among the modern Indo-

–2–
European languages that have not entirely eliminated the Indo-European case inflection. For
example, in Russian and other Slavonic languages:
– the ancient locative case has become a prepositional case, that can only be used in combination
with some prepositions,
– the case resulting from the fusion of I.E. genitive and ablative still assumes the genitive
function by itself, but assumes the ablative function in combination with some prepositions only,
– similarly, the accusative case, which originally assumed the allative function by itself, occurs in
allative function in combination with prepositions only.
Case forms used in such conditions may still contribute to the expression of spatial meanings (in
Russian, both za + accusative and za + instrumental encode the relative orientation BEHIND, but
the choice between accusative and instrumental case encodes the distinction between destination
and location).3 It may however also happen that the choice of a particular case is simply required by
the adposition (in Russian, ot ‘from’ is necessarily followed by a noun phrase in the genitive case).
In the old I.E. system, most clearly attested in Sanscrit, three case forms of nouns qualified as
spatial cases, at least in some of their uses. Latin illustrates the transition towards another type, in
which the same case forms can contribute to the expression of spatial meanings within the frame of
prepositional phrases only: in Latin, with some nouns, the accusative case assumed an allative
function, and the ablative case assumed locative and ablative functions, without the help of any
preposition,4 but for most nouns, the expression of spatial meanings necessitated the presence of a
preposition.
In languages that undergo such an evolution, it is common that isolated spatial case forms of
some nouns, when they are not accompanied by noun dependents, continue to be used by
themselves with spatial meanings, becoming thus spatial adverbs.

1.6. Spatial cases and semantic classes of nouns

Two semantic classes of nouns frequently have particularities in relation with spatial cases:
Geographical names,5 and nouns referring to humans.
Geographical names often have a ‘lighter’ spatial marking than most other nouns and tend to be
more conservative in evolutions affecting the expression of spatial relations. This is quite obviously
the consequence of their predisposition to represent the reference point in a spatial relation, and of
the frequency of their use as spatial complements or adjuncts. In Latin, the nouns that maintained
spatial uses of prepositionless ablative and prepositionless accusative were mainly town names. In
Tswana, names of towns or countries have no locative form, and occur in the absolute form in
contexts in which, with very few exceptions, other nouns must take the locative form. In Hungarian,
some town names maintain an ancient locative ending -ett/ött/ott that has been eliminated from
regular noun inflection.6
Nouns referring to humans, or more generally to animate beings, may have special forms for
spatial cases. In some languages, such nouns are incompatible with spatial cases (for example, in
Armenian, they do not have locative case). This follows from a general tendency to express spatial
relations with human beings as the reference point in an indirect way, through a genitival
construction (‘at N’s place’). The grammaticalization of such constructions may lead to the creation
of spatial adpositions specifically used with human nouns (like French chez < Latin casa), but in
some variants of this scenario, the result may be the emergence of spatial cases specifically used
with human nouns; one may for example imagine the grammaticalization of a construction in which

3 Abraham 2003 discusses case assignment with respect to the doubly-governing prepositions in German, showing the
weaknesses of the traditional presentation of this question.
4 Some of these nouns (but not all) maintained also a locative form distinct from the ablative.
5 Common nouns characterizable as ‘natural locations’ (such as house, or village) often show the same tendencies as
geographical names with respect to the expression of spatial relations.
6 This ancient locative suffix subsists also in the inflection of spatial postpositions – see section 3.

–3–
a spatial case affix is attached to a genitive interpreted as an elliptical variant of the construction
N’s (place).
Basque cumulates both particularities: contrary to the other cases of Basque, which have the
same form for all nouns, spatial cases have special forms (a) for animate nouns, and (b) for
geographical names. Interestingly, the spatial case endings of Basque are shorter for geographical
names than for ordinary nouns, and longer for animate nouns.
‘Endingless locatives’ (which in at least some cases can be analyzed as bare noun stems in
locative function) are well-attested in some ancient I.E. languages, but the nouns that have this
particularity do not seem to constitute a natural semantic class (for Hittite, see Neu 1980), and the
very notion of ‘endingless locatives’ in I.E. is not entirely clear.
An interesting case of grammaticalization of the relation between semantic classes of nouns and
the encoding of spatial meanings is attested by Central Bantu languages, in which affixes that are
basically noun class prefixes (and as such, primarily attach to subsets of nouns that have in common
an inherent locative meaning) are also used with nouns belonging to other classes with a function
similar to that of spatial case markers (Grégoire 1977).

1.7. Adverbs and adpositions inflected for spatial cases

In languages in which nouns have a case inflection including a subsystem of spatial cases, spatial
adverbs and adpositions commonly have possibilities of variations similar to noun case inflection,
but limited to spatial cases. This can be viewed as an indication that, diachronically, such adverbs or
postpositions originate from nominal forms inflected for case –see section 3.

2. Semantic distinctions expressed by spatial cases

2.1. Preliminary remark

When analyzing the semantic distinctions expressed by spatial cases, cases that assume spatial
functions in a productive way along with uses typical of some non-spatial case must be considered
on a par with specialized spatial cases, and in this respect, current terminology may be misleading.
For example, Turkish provides a typical illustration of the tripartite distinction locative / allative /
ablative, but the case assuming the allative function is also used as a dative marker, and is
designated simply as ‘dative’ in many Turkish grammars, whereas in other languages which have
basically the same situation, a case very similar to Turkish ‘dative’ is described as an ‘allative’ or
‘directive’ case having also dative functions.

2.2. Case systems with a unique spatial case

Some languages have a unique spatial form of nouns used equally, without the help of any
adposition, for verb dependents expressing location, source of movement, or destination of
movement (Tswana, Nahuatl). Such forms are commonly designated as locative.7
But there are also languages with a unique spatial case, designated as locative, expressing
location only. For example, Baltic languages (Latvian, Lithuanian) encode location by means of
prepositionless locative, but use prepositions governing non-spatial cases for destination or source.

7 Creissels 2006 examines the strategies used to encode the distinction between location, destination, and source, in
languages that do not encode this distinction by means of adpositions or case affixes. See also Wälchli & Zuñiga 2006.

–4–
 2.3. Types of distinctions structuring the systems of spatial cases

Spatial case systems tend to be structured along two dimensions: the distinction between
location, destination, source, and path on the one hand, and reference to particular types of spatial
configurations (or relative orientations) on the other hand. The terms configuration and relative
orientation, taken here as synonymous, refer to distinctions of the type expressed by the choice
between in, on, at, behind, under, etc. in English.
For a formal analysis of the semantics of locative expressions, see Kracht 2002.

2.4. Unidimensional spatial case systems

Spatial case systems encoding distinctions between types of configurations only, without any
distinction between location, destination, source, and path, do not seem to be attested.
Unidimensional spatial case systems tend to be organized according to a tripartite distinction
between location, destination of movement, and source of movement. Locative / allative / ablative
are the terms most commonly used to describe such systems, at least when the case forms in
question are used predominantly in spatial functions.8 Another possible terminology, mostly used
however in descriptions of bidimensional spatial case systems, is essive / lative / elative.9
In systems of spatial adpositions, the conflation of location with destination is quite common.
Among spatial case systems, this conflation is found for example in Kryz, but this is not a common
situation, which is somewhat unexpected, given the diachronic relation between case affixes and
adpositions.
A few languages have a locative-ablative vs. allative distinction: Kanuri, Dinka, Old Georgian; in
Latin, only part of the nouns that maintained spatial uses of the ablative case had a distinct locative
form. Contrary to what we observed in relation with the location-destination conflation, this second
type of conflation is equally rare in case systems and in adposition systems.

2.5. Bidimensional spatial case systems

Bidimensional spatial case systems combine an indication on relative orientation with a

distinction of the type locative / allative / ablative (or essive / lative / elative).10 The indication on
relative orientation carried by spatial cases is most of the time limited to three basic configurations
that can be symbolized as IN, ON, and AT.11
The following chart gives the terms used to label the 9 spatial cases of Hungarian analyzable as
combining the two dimensions location / destination / source and IN / ON / AT.12

IN ON AT
location -ban/ben (inessive) -(o/e/ö)n (superessive) -nál/nél (adessive)
destination -ba/be (illative) -ra/re (sublative) -hoz/hez/höz (allative)
source -ból/ből (elative) -ról/ről (delative) -tól/től (ablative)
8 Allative and ablative are used with a more restricted meaning in the description of some bidimensional systems of
spatial cases –see section 2.5.
9 Note that essive is also used, for example in Hungarian grammars, as a label for non-spatial cases with the meaning
‘as’, ‘in capacity of’.
10 Recent works on Daghestanian languages, which have particularly elaborated systems of spatial cases, use the terms
localization for what is called here configuration or relative orientation, and direction for the choice between location,
destination, source, and path.
11 This is in accordance with a general principle according to which, ‘Where inflectional case and adpositions co-occur
in a language, the adpositional system normally exhibits finer distinctions than the inflectional system’ (Blake 1994:11).
12 In a consistent use of Latin roots, sublative should be replaced by superlative, since sub- refers to UNDER-
configuration. The choice of this term by Hungarian grammarians was probably motivated by the desire to avoid
homonymy with superlative as an inflected form of adjectives.

–5–
 More elaborated bidimensional systems of spatial cases are found in Daghestanian languages13
and in Burushaski. In such systems, the second dimension may include a 4th term specifically used
to encode path (translative, or perlative), and the first dimension may include additional
distinctions, for example :
– series encoding configurations more commonly encoded by means of adpositions or locational
nouns: UNDER, BEHIND, or IN FRONT OF,
– distinct series for two varieties of IN-configuration, ‘in filled, dense space’ (The stone is in the
ground) vs. ‘within an empty or closed space’ (The water is in the glass),
– distinct series for two varieties of ON-configuration, according to distinctions such as vertical
vs. horizontal contact, or contact due to gravity vs. adherence (see Ganenkov 2005 on Daghestanian
languages, Tiffou 1999:190 on Burushaski).
The following chart gives possible labels for the spatial cases of Avar as they are currently
presented in descriptions of Avar, with 5 series of 4 cases each:14

ON AT IN1 UNDER IN2

location -da -q -ɬ -L’ -<b>
(superessive) (apudessive) (intereressive) (subessive) (inessive)
destination -de -qe -ɬe -L’e -<b>-e
(superlative) (apudlative) (interlative) (sublative) (illative)
source -dasa -qa -ɬa -L’a -sa
(superelative) (apudelative) (interelative) (subelative) (inelative)
path -dasan -qan -ɬan -L’an -san
(supertranslative) (apudtranslative) (intertranslative) (subtranslative) (intranslative)

The semantic complexity of such spatial case markers is more or less reflected in their form.
In Hungarian it is not possible, in a strictly synchronic analysis, to isolate in each spatial case
marker a formative for the distinction location / destination / source and another for the distinction
IN / ON / AT, but the superessive marker is the only one constituting the direct reflex of an ancient
case marker, and most of the others (in particular, those encoding IN-configuration) still have
forms suggesting that they originally combined two distinct formatives.

13 On Daghestanian case systems in general, see Daniel & Ganenkov, this volume. On Daghestanian spatial case
systems, see Comrie 1999, and for a diachronic approach Alekseev 2003. Detailed presentations of Daghestanian
systems can be found in Authier to appear (Kryz), Charachidzé 1981 (Avar), Haspelmath 1993 (Lezgian), Kibrik (ed.)
1996 (Godoberi), Kibrik & al. (eds.) 2001 (Bagvalal), Kibrik & Testelec (eds.) 1999 (Tsakhur), Sumbatova & Mutalov
2003 (Dargwa), van den Berg 1995 (Hunzib).
14 This inventory calls for the following two remarks, which unfortunately cannot be sufficiently developped here :
(a) The 1st series, traditionally described as encoding ON-configuration, tends to take the status of series unmarked for
configuration, the postposition t’ad being used to encode specifically ON-configuration.
(b) The basic meaning of the two series labelled here IN1 and IN2 is currently described as involving the distinction in
dense space vs. in empty space. <b> in the first two forms of the 5th series is a gender-number marker agreeing with
the absolutive term of the construction in which these forms fulfill the function of locative adjuncts or allative
complements, conventionally given in the form it takes for ‘singular, non-human’. This last series has no formative of its
own, but its formation involves a special stem, whereas the suffixes of the other series attach to the same ‘oblique stem’
as non-spatial case markers (for example, the oblique stem of ruq’ ‘house’ is ruq’aɬ-, but the stem of this word for the
suffixes of the 5th series of spatial cases is roq’o-). In addition to that, this series exists only with a limited set of nouns,
which casts some doubts on the decision to analyse these forms as inflected forms of nouns rather than locative adverbs
derived from nouns. Diachronically, it seems reasonable to assume that this 5th series represents a layer of spatial case
marking anterior to the emergence of the bidimensional system of spatial cases. Additional evidence supporting this
hypothesis is the similarity between this series and the case inflection of adverbs/postpositions such as hani<b> ‘here’,
ki<b> ‘where ?’, horL’o<b> ‘in the middle (of)’, χadu<b> ‘behind’, žani<b> ‘inside’ (among which some at least
may be ancient locational nouns subsisting only in spatial case forms).

–6–
 In Avar, 3 series of spatial cases quite transparently result from the concatenation of a
configuration marker and a marker encoding the location / destination / source / path distinction,
but the segmentation of the remaining two series is more problematic. In such cases, it can be
assumed that evolutions have blurred the original composition of complex markers, or that the
system conflates distinct diachronic layers of case marking (see in particular note 12 on the 5th
series of spatial cases of Avar).
A plausible source of such systems is the morphologization of constructions involving a case-
marked locational noun or adposition expressing configuration –see section 3.
In bidimensional systems in which spatial case markers can be more or less straightforwardly
segmented into two formatives, it may happen that:
– either the absence of the second formative encodes location, contrasting with non-void
formatives encoding source and destination (Avar, Burushaski),
– or the configuration marker is optional, and its absence indicates that the relative orientation is
not specified (such a system is reconstructed for Proto-Finnish – see section 6).
A frequent problem in describing bidimensional spatial case systems is that one of the series may
tend to lose its specific meaning, and to function as the default choice when the specification of a
particular configuration is not important from a communicative point of view. For example,
Akhvakh has 5 series of spatial cases; three of them quite obviously encode specific configurations
(IN, AT and UNDER), but the other two have uses that lead to the conclusion that one of them
(characterized by a formative -q-) has lexicalized (in the sense that its productivity is very limited,
and the nouns that have this series of spatial cases must be listed in the lexicon), whereas the other
(characterized by a formative -g-) is the default choice to encode spatial relations without referring
to a particular configuration. A similar problem has been mentioned for Avar in note 11.
In the absence of a series used by default when no particular configuration is intended, the
choice of a particular series may involve lexical properties of nouns. Kracht 2002:6 rightly observes
that Hungarian has a mezőn ‘in the meadow’ (superessive) but a kertben ‘in the garden (inessive),
and that a hajón ‘on the ship’ (superessive) ‘is the neutral way of saying that you are on the ship,
regardless of whether you are in a cabin or on deck’.

2.6. Tridimensional spatial case systems

Tridimensional spatial case systems are exceptional, but Tsez attests the possibility to combine
the two dimensions commonly found in Daghestanian systems with a binary distinction ±distal.

2.7. The expression of path

In addition to the common tripartite distinction location / destination / source, a few languages
(e.g., Yup’ik) have a fourth unanalyzable marker specifically used to encode path (translative, or
perlative), but the following situations are much more common:
– ablative markers are often used to encode, not only source, but also path (Basque, Turkish,
Armenian, Lezgian);
– path may also be encoded by a complex case marker consisting of an ablative marker plus an
extension in the sense of section 1.2 (Avar and other Daghestanian languages);
– path may be encoded by means of the combination of a spatial case and an adposition
(Hungarian);
– instrumental case markers or adpositions often have the expression of path as one of their
possible uses, even in languages that have a distinction between an instrumental case/adposition
and an ablative case (in Azerbaidjani, at least with some nouns, the ablative suffix -dAn and the
comitative-instrumental postposition ile can equally encode the role of path).

–7–
 2.8. Other uncommon types of spatial cases

Páez has a unique allative case -na and a unique ablative case -hu, but four locative cases
expressing an obligatory distinction between four types of posture: standing (-te), lying (-ka),
hanging (-khe), and leaning (-su).
In languages that have spatial cases, terminative (‘up to’) is commonly encoded by means of an
adposition governing the allative case, or a complex case marker resulting from the addition of an
extension to an allative marker (Basque -ra-ino, Azerbaijani -(y)A-cAn), but Hungarian has an
unanalyzable terminative marker (-ig). A terminative case is also mentioned in Estonian grammars,
but it is a complex marker based on the genitive case.
Directional (‘towards’) is encoded in Basque by means of an extension of the allative marker.

2.8. Common types of spatial cases in uncommon types of organization

From a strictly synchronic point of view, Georgian can be analyzed as having an inventory of
seven spatial cases.15 All of them are complex markers (‘cas secondaires’ in Vogt’s terminology),
but their meaning is entirely determined by the second formative (which is historically a cliticized
postposition), and their first formative is a non-spatial case initially governed by the postposition
from which the second formative originates:
– superessive-superlative: dative + -ze
– inessive-illative: dative + -ši
– adessive-allative: dative + -tan
– ablative: genitive + -gan
– elative: instrumental + -dan
– directional: genitive + -k’en
– terminative: adverbial + -mde

3. Spatial case forms of locational nouns, spatial adpositions,

and spatial adverbs
In languages that have unidimensional systems of spatial cases, relative orientation is often
transparently expressed by means of locational nouns inflected for spatial cases and taking as their
complement a noun phrase (often in the genitive case), as in Basque:

gain ‘top’ azpi ‘bottom’ ondo ‘side’ etc.

location gain-ean azpi-an ondo-an
destination gain-era azpi-ra ondo-ra
source gain-etik azpi-tik ondo-tik

The grammaticalization of such situations may lead:

– in a first stage, to systems of spatial postpositions inflected for spatial cases;
– in a second stage, to bidimensional systems of spatial cases.
This is precisely what occurred in Hungarian. This language has a large inventory of
postpositions with a case inflection consisting of three forms encoding the distinction location /
destination / source:

15 see Vogt 1971; most Georgian grammars do not consider these complex markers as cases, but are not explicit on the
nature of their second formative.

–8–
 BESIDE FRONT BACK etc.
location mell-ett el-ött mög-ött
destination mell-é el-é mög-é
source mell-ől el-ől mög-ül

These postpositions are quite obviously, from a diachronic point of view, spatial case forms of
locational nouns. But synchronically, due to relatively recent changes in the case system of
Hungarian, they cannot be recognized as such, even when the noun they originate from still exists
(for example, mell-ett/-é/-ől is clearly cognate with mell ‘chest’, but none of the three forms of this
postposition coincides with any of the case forms of the noun mell).
The cliticization of some of these postpositions resulted in the two-dimensional spatial case
system of Hungarian presented in section 2.5.16
In a variant of this scenario, nominal forms or phrases including spatial case markers
grammaticalize in contexts in which they are not combined with a complement NP, becoming thus
spatial adverbs, and the adverbs created in this way may maintain a case inflection limited to spatial
cases, and often irregular from a synchronic point of view, due to the fact that, once they have
frozen as adverbs, these forms are not affected by subsequent changes affecting regular noun
inflection.
For example, Basque has three deictic place adverbs, hemen/hon- ‘here’, hor ‘there (near
hearer)’, and han ‘over there’, with the following possibilities of case inflection:

locative hemen hor han

allative hona horra hara
ablative hemendik hortik handik

The etymological connection with the three demonstratives (hau(r)/hon-, hori, and hura/ha-) is
clear, but in the present state of the language, these forms cannot be recognized as case forms of
demonstratives, or of phrases including demonstratives.
Adverbes inflected for spatial cases, often cognate with inflected postpositions, are very common
in Hungarian. For example, the postposition el-ött / el-é / el-ől expressing the relative orientation
FRONT is cognate with the adverb elöl / elő-re / elöl-ről expressing exactly the same meaning.
Akhvakh is another typical example of a language with a large inventory of such adverbs.

4. Syncretisms between core syntactic cases and spatial cases

4.1. Allative-dative syncretism

Not surprisngly, given the prototypical meanings currently recognized to cases, the allative-
dative syncretism is by far the most common syncretism involving a core syntactic case and a spatial
case. Even in languages that have a specialized dative marker, it is not rare that allative can replace
dative in its most typical function (e.g., Akhvakh), and the shift allative → dative is admittedly a
common source of dative markers.

16 Note that the Hungarian postpositions whose cliticization gave rise to the case system of present-day Hungarian still
exist as words in complementary distribution with case suffixes: in Hungarian, with the only exception of the accusative,
case suffixes cannot attach to personal pronouns, and this impossibility is compensated by the use of postpositions
inflected for person. For example, ‘by me’ cannot be expressed as *én-nél (én ‘I’, -nél ‘adessive’), but only as nálam, 1st
person of a postposition cognate with the case suffix -nál/nél; this potposition differs from ordinary postpositions in that
it exists only in forms having a personal ending – for a detailed analysis, see Creissels 2006b.

–9–
 4.2. Allative-accusative syncretism, and unmarked allative

An allative-accusative syncretism in languages that have a specialized dative case is attested in

ancient Indo-european languages (Sanscrit, Latin).
A possible explanation is a first change by which an allative marker extends its use to dative and
subsequently accusative functions (as attested in the history of the Romance preposition a < Latin
ad in Spanish, or in Sardinian), the chain allative-dative-accusative being subsequently broken by
the creation of a new dative marker from a different source. However, this hypothesis is entirely
speculative, and the following observations on unmarked allative suggest rather a direct functional
explanation of such situations.
Unmarked allative (i.e., the use of the absolute form of nouns in allative function without the
help of any adposition) is observed in several languages that have also direct objects devoid of any
mark of their function (for example, in Maale, an Omotic language in which a marked case form of
nouns is used for subjects, and the absolute form is used for objects). The situation of such
languages is therefore functionally similar to that of languages using the same case-marked form of
nouns for direct objects and complements of verbs of movement expressing destination. In some
cases (for example, in Armenian), the unmarked allative can be explained by the loss of the
nominative vs. accusative distinction in a system originally characterized by an allative-accusative
syncretism. But unmarked allative is also found in languages in which there is no evidence that an
accusative case ever existed, and other explanations must be considered.
For example, Tswana has a locative form of nouns indistinctly used for nouns in the roles of
location, source, and destination, but the verbs of movement assigning to their complement the role
of destination can also optionally take complements encoding destination that bear no mark of
their role, and are treated in fact as direct objects (in particular, they can pronominalize in the same
way as direct objects).
The tendency to treat nouns encoding destination or source of movement as the second core
term of transitive constructions is sporadically attested in many languages (as in English: He
entered / leaved the room), and seems particularly strong for nouns in allative function, which may
explain the situation of languages in which unmarked allative is the rule.17

4.3. Locative-dative syncretism

Mongolian has a locative-dative syncretism, but uses a different marker (traditionally analyzed
as a postposition governing the absolute form of nouns) for allative. A similar situation existed in
Old Hittite.

4.4. Spatial cases and genitive

The ablative-genitive syncretism is common in adposition systems (French de), and therefore
would be expected to be found in case systems too, but examples are not easy to find. Such a
syncretism is postulated in the history of Greek, but in Classical Greek, the ablative-genitive case
was no more a spatial case, since it assumed spatial functions productively in combination with
prepositions only.
The same can be said of the locative-genitive, allative-genitive, and locative-allative-genitive
syncretisms: such syncretisms are relatively common in adposition systems (for example, Bulgarian
na illustrates the locative-allative-genitive syncretism), but seem to be rare in case systems.

17 More generally, on the possibility to treat NPs encoding spatial relations as core syntactic terms, see Dimmendaal
2003.

– 10 –
 4.5. Spatial cases and ergative

Dixon 1994 signals the existence of a locative-ergative syncretism in some languages, but does
not mention the possibility of an ablative-ergative syncretism. This is somewhat unexpected, since
ablative-instrumental and instrumental-ergative are among the most common case syncretisms, and
in many languages, obliques representing agents (in particular, in passive constructions) occur in
the ablative case. An illustration of the ablative-ergative syncretism can however be found in Kryz,
where the ergative suffix coincides with the second formative of spatial cases encoding source.

5. Other uses of spatial cases

5.1. Oblique arguments

Arguments of individual verbs may occur in case forms that have otherwise the status of spatial
cases. Such case assignments can often be viewed as resulting from a metaphorical extension of the
spatial meaning of the case in question. For example, verbs with the meaning ‘be afraid (of)’
commonly assign ablative case to their complement.

5.2. Spatial cases in constructions involving valency changes

Ablative is used in some languages for NPs representing the demoted subject in passive
constructions (Greenlandic). This use is clearly related to the ablative of cause – see 5.3.3.
Yup’ik uses the ablative case for demoted objects in antipassive and applicative constructions,
and the allative case for demoted transitive subjects in constructions involving the introduction of
an additional participant in subject role.
Spatial cases may be involved in non-volitional agent constructions. For example, Lezgian marks
the non-volitional agent with the adelative case.

5.3. Spatial cases used to mark verb satellites with various circumstantial meanings

Cross-linguistically, the ablative-instrumental syncretism is particularly common.

Case markers specialized in the expression of temporal meaning (such as Hungarian -kor) are
not common. Spatial cases are widely used to encode temporal relations, which is commonly viewed
as a metaphorical extension of their basic meaning.
Locative of state, allative of transformation, allative of purpose, and ablative of cause, are also
relatively common metaphorical extensions of the use of spatial cases.
Basque has a secondary case marker based on allative ( -ra-ko) for NPs used as verb dependents
with a meaning of purpose.

5.4. Partitive

Hungarian illustrates the use of ablative to express partitive.

Basque has a partitive marker -(r)ik resulting from the specialization of an allomorph of the
ablative marker -tik.

5.5. Comparative

The use of ablative to mark adjective complements expressing comparison is relatively common.

– 11 –
 5.6. Spatial cases governed by adpositions

It may happen that case markers apt to encode spatial relations without the help of any
adposition have also uses in which they are governed by an adposition (Basque zehar ‘throughout’
combines with NPs in the locative case, Turkish kadar ‘until’ combines with NPs in the allative case,
sonra ‘after’ combines with NPs in the ablative case, etc.)

5.7. Spatial cases and the syntactic role of noun dependent

Spatial cases are rarely used to mark NPs in the role of noun dependents.
Some languages have ‘adjectivizers’ systematically used to convert words or phrases typically
used as verb satellites (including NPs marked for spatial cases) into noun dependents (Basque
-ko,18 Turkish -ki, Akhvakh -se).

6. Non-spatial cases derived from spatial cases

The generalization of the use of adpositions with originally spatial cases is not the only way by
which spatial cases can undergo a change in their status. Another possible scenario is that, after
developping non-spatial uses, spatial cases tend to be used predominantly in their new function,
and to be replaced by other spatial cases (or by adpositions) for the expression of spatial meanings.
This is quite obviously a very common type of change in the evolution of case systems, and,
diachronically, syncretisms such as those described in sections 4 & 5 often constitute intermediary
stages in evolutions converting spatial cases into markers of core syntactic relations, or into cases
expressing other types of circumstantial meanings.
For example, the system of spatial cases of Finnish originally consisted of nine cases organized in
three series with the meanings ‘unmarked for relative orentation’, ‘IN’ (originally characterized by a
formative -s-), and ‘AT’ (originally characterized by a formative -l-), but the first series has
developped non-spatial uses (modal < locative, partitive < ablative, and transformative < allative),
and is now used only residually in spatial functions.19

References

Abraham, W. 2003. The myth of doubly governing prepositions in German. In Shay, E. & U.
Seibert (eds.), Motion, direction and location in languages. 19-38. Amsterdam / Philadelphia:
John Benjamins.
Alekseev, M.E. 2003. Sravnitel’no-istoričeskaja morfologija naxsko-dagestanskix jazykov. Kategorii
imeni. Moskva: Academia.
Anderson, J.M. 1971. The grammar of case: towards a localistic theory. Cambridge: Cambridge
University Press.
Authier, G. To appear. Eléments de la langue kryz, dialecte d’Alik. Paris: Société de Linguistique
de Paris.
Blake, B. 1994. Case. Cambridge: Cambridge University Press.

18 Traditional Basque grammars wrongly identify -ko as a ‘local genitive’ case. In fact, -ko attaches to NPs already
inflected for case, and has a purely syntactic function of converting verb satellites into noun dependents. The source of
the error is that, due to morphophonological rules, when -ko attaches to NPs in the locative case, the locative suffix may
not be apparent.
19 The three cases of this series are designated in Finnish grammars as essive, partitive, and translative.

– 12 –
Charachidzé, G. 1981. Grammaire de la langue avar (langue du Caucase Nord-Est). Paris: Jean
Favard.
Comrie, B. 1999. Spatial cases in Daghestanian languages. Sprachtypologie und
Universalienforschung 52-2. 108-117.
Comrie, B. & M. Polinsky. 1998. The great Daghestanian case hoax. In Siewierska, A. & J. J. Song
(eds.), Case, typology and grammar. 95-114. Amsterdam / Philadelphia: John Benjamins.
Comrie, B., M. Polinsky, & R. Rajabov. 1998. Tsezian languages. In Harris, A. & R. Smets (eds.),
The languages of the Caucasus. London: Curzon.
Cook, W.A. 1989. Case grammar theory. Washington: Georgetown University Press.
Creissels, D. 2006a. Encoding the distinction between location, source, and destination: a
typological study. In Hickmann, M. & S. Robert (eds.), Space in Languages: linguistic systems
and cognitive categories. 19-28. Amsterdam / Philadelphia: John Benjamins.
Creissels, D. 2006b. Cas et postpositions en hongrois. Bulletin de la société de linguistique de Paris
101. 225-272.
Daniel & Ganenkov (this volume) …
Dimmendaal, G.J. 2003. Locatives as core constituents. In Shay, E. & U. Seibert (eds.), Motion,
direction and location in languages. 91-109. Amsterdam / Philadelphia: John Benjamins.
Dixon, R.M.W. 1994. Ergativity. Cambridge: Cambridge University Press.
Ganenkov, D.S. 2005. Kontaktnye lokalizacii v naxsko-dagestanskix jazykax i ix tipologičeskie
paralleli. PhD thesis. State University of Moscow.
Grégoire, C. 1975. Les locatifs en Bantou. Tervuren: Musée Royal de l’Afrique Centrale.
Haspelmath, M. 1993. A grammar of Lezgian. Berlin / New York: Mouton de Gruyter.
Haspelmath (this volume) …
Hjelmslev, L. 1935. La catégorie des cas: Etude de grammaire générale I . Kopenhagen:
Munksgaard.
Hjelmslev, L. 1937. La catégorie des cas: Etude de grammaire générale II. Kopenhagen:
Munksgaard.
Jackendoff, R. 1983. Semantics and cognition. Cambridge : MIT Press.
Jackendoff, R. & B. Landau. 1992. Spatial language and spatial cognition. In Languages of the
mind: Essays on mental representation. 99-124.
Kibrik, A.E. (ed.). 1996. Godoberi. München / Newcastle: Lincom Europa.
Kibrik, A.E., K.I. Kazenin, E.A. Ljutikova & S.G. Tatevosov (eds.). 2001. Bagvalinskij jazyk:
gramatika, teksty, slovari. Moskva: Nasledie.
Kibrik, A.E. & Ja.G. Testelec (eds.). 1999. Èlementy caxurskogo jazyka v tipologičeskom
osveščenii. Moskva: Nasledie.
Kracht, M. 2002. On the semantics of locatives. Linguistics and Philosophy 25. 157-232.
Langacker, R. 1987. Foundations of cognitive grammar, Volume 1: Theoretical prerequisites.
Stanford: Stanford University Press.
Neu, E. 1980. Studien zum endungslosen Lokativ des Hethitischen. Innsbruck: Institut für
Sprachwissenschaft der Universität Innsbruck.
Pederson, E. 1995. Language as context, language as means: spatial cognition and habitual
language use. Cognitive linguistics 6 (1). 33-62.
Pederson, E., E. Danziger, D. Wilkins, S. Levinson, S. Kita & G. Senft. 1998. Semantic typology and
spatial conceptualization. Language 74. 557-89.
Shay, E. & U. Seibert (eds.). 2003. Motion, direction and location in languages. Amsterdam /
Philadelphia: John Benjamins.
Sumbatova, N.R. & R.O. Mutalov. 2003. A grammar of Icari Dargwa. München / Newcastle:
Lincom Europa.
Svorou, S. 1994. The grammar of space. Amsterdam / Philadelphia: John Benjamins.
Talmy, L. 2000. Towards a cognitive semantics (2 vol.). Cambridge: MIT Press.
Tiffou, E. 1999. Parlons Bourouchaski. Paris: L’Harmattan.

– 13 –
van den Berg, H.E. 1995. A grammar of Hunzib (with texts and lexicon). München / Newcastle:
Lincom Europa.
Wälchli, B. & F. Zúñiga. 2006. Source-goal (in)difference and the typology of motion events in the
clause. Sprachtypologie und Universalienforschung 59-3. 284-303.

– 14 –