Location and Locality
Henk van Riemsdijk & Riny Huybregts
Tilburg University
abstract
An account of spatial semantics cannot fail to encompass three basic notions: location,
change of location, and shape. While shapes can be considered to be properties of
objects with a spatial dimension, locations and paths are relations between such objects.
Most of the semantic and morphosyntactic literature, therefore concentrates on locations
and paths. Without pursuing the intricacies of the semantics and logic of these notions,
we take them to be basic and address the question of how these notions are reflected in
syntax and morphology. There are indeed languages in which there is a direct
grammatical correlate of the notions location and path. Accordingly, and following
Jackendoff (1983) and Koopman (1993), we take the abstract structure of a spatial
phrase in the verbal domain to be [V' Vo
[PP DIRo
[P' LOCo
[N' No
] ] ] ]. Our purpose here
is to present new evidence for such a structure based on locality considerations. A
robust notion of locality (heads involved in a syntactic relation R must be hierarchically
adjacent) predicts †R(V,PDIR
), †R(PDIR
,PLOC
), †R(PLOC
,N), *R(V,PLOC
), *R(PDIR
),
*R(V,N).
© M.A.C. Huijbregts & H.C. van Riemsdijk
Tilburg University
Humanities GM / CLS
P.O. Box 90153
NL - 5000 LE TILBURG
phone: +31 13 4662773 fax: +31 13 4663110
e-mail: riemsdyk@kub.nl riny.huijbregts@kub.nl
www: http://cwis.kub.nl/~fdl/general/people/riemsdh/
Van Riemsdijk & Huijbregts Location and Locality
Page 2 of 23
1. LOCATIONS AND PATHS
1
1.1. Introductory Remarks
An account of spatial semantics cannot fail to encompass three basic notions:
location, change of location (or path, transition, direction), and shapes. While shapes
can be considered to be properties of objects with a spatial dimension, locations and
paths are relations between such objects. See Levelt (1996) for enlightening
discussion. But most of the literature, both semantic (cf. Asher & Sablayrolles (1994),
Gruber (1965, 1976), Jackendoff (1983, 1990), Talmy (1983)) and morphosyntactic
(cf. Bierwisch (1988), Hjelmslev (1935-37), Koopman (1993)), therefore concentrates
on locations and paths. For our present purposes, without pursuing the intricacies of
the semantics and logic of these notions, we will take them to be basic and turn to the
question of how they interface with grammar. In other words, we will discuss a
number of aspects of the way in which these notions are reflected in syntax and
morphology.
From the perspective of morphology and syntax, one reason why space may not have
been among the most prominent research topics is that in many of the more familiar
languages the notions having to do with locations and paths are expressed by means
of prepositions (or postpositions). These are closed class items and languages tend to
have relatively few of them, often augmented by a few derived expressions, a type of
complex prepositions of the type in front of. Adpositions show little or no overt
systematicity in their shape or in their behavior. There is no property that above and
below share to the exclusion of other prepositions, for example. And it is hard to see
why Dutch distinguishes achter and na ('behind' and 'after'), but not 'in front of' and
'before' (both expressed by voor). In addition, there is an abundance of quite
mysterious collocational restrictions: why are pictures 'on the wall' in English but 'at
the wall' in Dutch?
Fortunately, there are a few languages that are more revealing in this regard. In
particular, there are languages that yield evidence that there is a grammatical correlate
of the notions location and path. More specifically, we will argue that there are
syntactically and morphologically specifiable positions in grammatical structure
which can be identified as location (LOC) and path (PATH or DIR). In the next two
subsections we will discuss two pure cases of this, the 'extreme' morphological
solution as found in Lezgian (section 1.2.) and the 'extreme' syntactic solution as
found in German (section 1.3.). Section 1.4. translates these findings into an
architecture for the morpho-syntactic representation of spatial expressions, essentially
as follows: [V' Vo
[PP DIRo
[P' LOCo
[N' No
] ] ] ].
1
The present article has grown out of research carried out in the framework of a discussion
group in which Norbert Corver, Reinhard Muskens, Craig Thiersch and Elias Thijsse also participated.
Early versions were presented by Van Riemsdijk to audiences at the University of Trondheim, Keio
University, Tokyo, Tsukuba University and Hokkaido University, Sapporo. A version closer to the
present one was presented by both authors at the Conference on Logic, Linguistics and Computation in
Batumi, Georgia in September 1999 and at the 16th
Annual Meeting of the Israeli Association for
Theoretical Linguistics at Tel Aviv in June 2000. Our thanks to all these audiences for stimulating
discussions.
Van Riemsdijk & Huijbregts Location and Locality
Page 3 of 23
In section 2. we turn to various types of evidence in favor of such structures.
Specifically, we will address the question whether the relations between the four
heads in question are constrained by locality. For example, could some rule or
principle relate DIR and N if an intervening LOC is present? In this section, we
present evidence that locality is obeyed at all levels. More specifically, we will
examine the predictions that ensue from, on the one hand, a hierarchically ordered
sequence of four heads (V - DIR - LOC - N), and, on the other, a robust notion of
locality to the effect that only adjacent heads in such a hierarchically ordered
sequence can enter into morphological or syntactic relationships, non-adjacent ones
being blocked. In section 2.1. we discuss the positive prediction, that is, the prediction
that the relationships N-LOC, LOC-DIR and DIR-V are permitted and are found. In
section 2.2. we show that the relationships predicted to be impossible (N-DIR, LOCV
and N-V) are indeed not found.
1.2. The Morphological Extreme: Lezgian
It has been known at least since Hjelmslev (1935/37) that the most extended spatial
case systems are found among the Daghestanian languages in the Caucasus.
Hjelmslev's own work was on Tabassaran, renowned among connoisseurs for being
the richest among them, though some of the others (including Lak, Archi, Tsakhur
and Udi, cf. Schulze (1983, 1985)) are not far behind.2
For our purposes, however, it
will suffice to look at the comparatively limited system of Lezgian, not in the least
because for this language there is an extensive modern grammar (Haspelmath (1993)).
A spatial expression in Lezgian is built up according to the following scheme.
(1) stem - stem augmentative suffix - locative morpheme - path morpheme
An example of such a spatial expression is shown in (2).
(2) sew - re - qh
- aj (postelative case)
bear - augm. - behind - from
'from behind the bear'
This part of Lezgian morphology, then, is entirely and transparently agglutinative,
even to the extent that phonological interaction between the different morpheme slots
is minimal. We need not be concerned with the stem-augmentative suffix here, but it
emerges from an example like (2) that there is a strict separation between locative
suffixes and suffixes expressing motion, direction, orientation, paths. As a matter of
fact, the language has five suffixes to express different locations and three suffixes to
2
Tabassaran is reported to have up to 48 locative cases. But, as will be seen directly when we
examine Lezgian, the figure is misleading because it results from the combinatorial
possibilities of far fewer suffixes. Where the 15 cases of Lezgian result from the combination
of one morpheme slot with a three-way choice and another slot with a five-way choice,
Tabassaran has 7x3x2 = 42 or 8x3x2 = 48 possible combinations (depending on whether one
looks at the Southern or at the Northern dialect). See Comrie & Polinsky (1998) for details.
These authors also report that Tabassaran is amply superseded in richness by Tsez. They
analyze Tsez as having a 7x2x4 = 56 case system; furthermore, each of these combinations
may or may not have the equative-1 case suffix added to it, bringing the total up to 112
distinct local case expressions.
Van Riemsdijk & Huijbregts Location and Locality
Page 4 of 23
express motion (including the Ø-morpheme to indicate absence of motion). This
yields a 3x5 matrix and 15 spatial cases. Of these, for reasons unknown, one, the
indirective ('into') is missing from the language. The overall system, then, can be
represented as in (3).3
The terminology for the case names is Haspelmath's.
(3)
1st morph.: AT BEHIND UNDER ON IN
2nd morph.: -w -qh
-k -l -Ø
AT -Ø Adessive Postessive Subessive Superessive Inessive
FROM -aj Adelative Postelative Subelative Superelative Inelative
TO -di Addirective Postdirective Subdirective Superdirective (Indir.)
By way of illustration, here is a full paradigm of a noun with all the possible case
forms:
(4) Absolutive: sew the bear
Ergative: sew-re the bear
Genitive: sew-re-n of the bear
Dative: sew-re-z to the bear
Adessive: sew-re-w at the bear
Adelative: sew-re-w-aj from the bear
Addirective: sew-re-w-di toward the bear
Postessive: sew-re-qh behind the bear
Postelative: sew-re-qh-aj from behind the bear
Postdirective: sew-re-qh-di to behind the bear
Subessive: sew-re-k under the bear
Subelative: sew-re-k-aj from under the bear
Subdirective: sew-re-k-di to under the bear
Superessive: sew-re-l on the bear
Superelative: sew-re-l-aj off the bear
Superdirective: sew-re-l-di onto the bear
Inessive: sew-re in the bear
Inelative: sew-re-aj ( sewräj) out of the bear
Indirective: -----------
3
It would be a mistake to conclude from such a table that the semantics and use of these cases
is as regular as their morphology. On the contrary. First of all, the language has a large number of
postpositions as well, and many of these are primarily spatial in nature and are frequently used, perhaps
more frequently than the spatial cases, to express purely spatial relations. Second, the spatial cases,
when used in a purely spatial sense, often have a meaning that does not correspond directly to their
place in the system. To cite one example (Haspelmath (1993:98)), the superessive expresses not only
the location 'on' but also the direction 'onto'. Third, the spatial cases have many non-spatial uses. These
may be temporal, of course, but also, for example, instrumental uses. The latter are primarily expressed
by the addirective case, but also by the superdirective. Lastly, there are multitudes of instances where
these cases are used in the manner of prepositional objects in apparently quite unpredictable ways. See
also Comrie & Polinsky (1998) for a discussion of morphological vs. semantic transparency.
Van Riemsdijk & Huijbregts Location and Locality
Page 5 of 23
We may conclude that the morphological system of Lezgian has two neatly separated
morpheme slots, one for Location and one for Direction.4
1.3. The Syntactic Extreme: German
Evidence that separate positions for Location and Direction must also be posited in
syntax comes from complex adpositional phrases in German. An argument to this
effect was originally presented in Van Riemsdijk (1990). In German, alongside
'normal' prepositional PPs such as (5)
(5) a. auf demDAT Tisch (onLOC the table)
b. auf denACC Tisch (ontoDIR the table)
we also find PPs with two adpositional elements, usually one initial and one final,
whence the name 'circumpositions.'
(6) a. auf das Dach hinauf
b. auf das Dach hinunter
c. auf das dach hinüber
The postpositional element in these PPs consists of a locational adposition (auf = 'on',
unter = 'under', über = 'over') with a deictic morpheme hin- or her- attached to it. Hinmeans
away from, and her- means towards, the speaker or the focal reference point in
the conversation. The preposition indicates the location where the movement
described ends, viz. on the roof, while the postpositional element describes the
orientation of the path: hinauf for upward motion, hinunter for downward motion, and
hinüber for a motion across from somewhere. Accordingly the meanings of the three
PPs in (6) can be depicted as in (7).
4
The three-way system for DIR in Lezgian is mirrored in languages like Finnish and
Hungarian. But four-way systems are found as well. In such four-way systems, the fourth type of path
is 'via', 'past'. Languages with such a system include Inuit (cf. Bok-Bennema (1991) and references
cited there, in particular Bourquin (1891) and Warlpiri (cf. Hale (1986)). The terminology varies. One
variant is: locative - allative - ablative - vialis; another: locative - terminalis - distantialis - prosecutive;
and Hale uses locative - allative - elative - perlative. In view of the close conceptual similarity between
space and time and the way these notions are expressed in language, it might be tempting to describe
the four-way system in terms of two binary aspectual features that are independently motivated, for
example:
+ TELIC - TELIC
+ INCHOATIVE ‘via’ ‘from’
- INCHOATIVE ‘to’ ‘at’
The idea is that the reference point is an end point ([+telic]) in the to case, a starting point
([+inchoative]) in the from case, and when being passed as in the via case it is both. But the suggested
link to aspectual notions remains tenuous at best.
Van Riemsdijk & Huijbregts Location and Locality
Page 6 of 23
(7)
This seems to indicate quite directly that the preposition corresponds to LOC and the
postposition to DIR. This is confirmed by examples in which the prepositional
element is missing. The prediction would be that such examples describe the
orientation of the path only, while being neutral with respect to the location. Indeed,
consider (8).
(8) Die Schnecke kroch das Dach hinauf
the snail crept the roof up
This example means that the snail is engaged in an upward motion along the roof, but
the location with respect to the roof is unspecified. First of all, it is not clear that the
motion will ever end at some particular point (though, pragmatically, roofs tend to be
finite in length/height). And second, the snail, due to its inherent properties, might be
engaged in this upward motion on top of the roof or underneath it (cf. (9)). The fact
that the snail must be in direct contact with the roof we again take to be determined by
our knowledge of the physical world.
(9)
So far so good. These examples appear to present straightforward evidence in favor of
two separate positions inside complex spatial expressions, one to express Location
and one to express Direction or Path. We believe this to be the correct conclusion,
even though in many cases the neat separation between LOC and DIR is blurred or
simply absent. Take a simple case like (10).
(10) Auf das Dach mit der Antenne! ('Onto the roof with the antenna!')
6a
6b
6c
Van Riemsdijk & Huijbregts Location and Locality
Page 7 of 23
Here, quite clearly, the top of the roof is the endpoint of the motion, and hence a path,
though unspecified as to its orientation, is implied. Nevertheless, all of this seems to
be expressed by the prepositional element alone, though we might, of course, assume
the presence of a null-postpositional element.5
1.4. The Functional Architecture of Extended Nominal Projections
On the basis of the above observations, we may conclude that spatial phrases have, in
essence, a structure as given in (11), where the order of the functional heads may vary
from language to language, and perhaps even within one language.6
In this we follow,
in essence, Jackendoff (1983), Koopman (1993) and Zwarts (1995).The overall
approach takes as a point of departure the notion of extended projection, as developed
in Grimshaw (1991) and Van Riemsdijk (1990), but the specific implementation we
choose is the one proposed in Van Riemsdijk (1990, 1998b), in which intermediate
phrases do not have the status of maximal projections. This is why there is no NP in
(11). For details and arguments, the reader is referred to the references given.
(11)
We take this type of structure to underlie both the complex morphological spatial
expressions of a language like Lezgian and the analytical, circumpositional, spatial
phrases of a language like German. By way of illustration, we will assume that a
circumpositional phrase like (12) is, essentially derived as in (13).7, 8
5
Depending on the specific choice of adposition, other options may be simply unavailable. The
generic directional preposition nach ('to'), for example, can only be used as a pure preposition and
never has any kind of postpositional element associated with it. Others, like zu ('to') may take a
postpositional element, as in zu mir herauf ('up and toward me'), but the deictic element can sometimes
stand alone, as in zu ihm hin ('to him and away from the speaker'). It is not surprising to find such
specific conditions associated with closed class items, and they do not affect the overall conclusions
regarding the architecture of spatial phrases.
6
We leave entirely open the question as to whether a universal word order should be assumed,
with all variation being derived by movement processes, or whether language specific ordering (a
directionality parameter of some kind) may be assumed.
7
Again, we remain agnostic about the underlying order of the functional heads (head-initial vs.
head-final) and consequently on whether any movement processes are involved in deriving (13).
Similarly, nothing hinges on the specific implementation of case assignment/checking that is
incorporated in (13).
8
We ignore the question of where the deictic morpheme comes from and assume it to be part of
the element filling the DIR-slot. It is tempting to speculate, however, that there might be an additional
position involved here. Recall (footnote 2) that Tabassaran and Tsez are thought to have three distinct
slots for the morphemes making up complex local case expressions, cf. Comrie & Polinsky (1998). In
addition to LOC and DIR, Tabassaran has a slot for a morpheme which expresses the distinction
Pmax
PDIR
o
P’
PLOC
oN’
No
Van Riemsdijk & Huijbregts Location and Locality
Page 8 of 23
(12) auf das Dach hinunter
(13)
Similarly, we assume that a Lezgian spatial expression like (14) is derived in the
manner shown in (15).9
(14) sew - re - qh
- aj
bear - augm. - behind - from
between 'precise location' and 'general location'. This is used to distinguish, for example, 'to' and
'towards', or 'from a precise point' vs. 'from the general area of.' We might call such a morpheme slot
AREAL. In Tsez, this AREAL feature appears to be incorporated into the DIR series. But Tsez, in turn,
has a third slot to express 'distal' (as opposed to neutral or close). This category of distality can be taken
to be conceptually quite close to the hin-/her-distinction in German. We might, therefore, assume that
there is a third functional position, call it DIST or DEIX, to accommodate such cases. Interestingly, the
DIST position in Tsez is between LOC and DIR, which would be compatible with the situation in
German.
9
The remarks made in the footnote 7. apply here as well, though with one proviso. It has been
argued in Van Riemsdijk (1998a) that adjunction of terminal elements to form agglutinative structures
operates under strict linear adjacency. If that is correct, then a head-final structure such as the one
depicted in (15) must be assumed to be underlying.
Pmax
PDIR
o
P’
PLOC
o N’
CASE N’
[+acc]
DET
No
hinunter
auf
das
Dach
Van Riemsdijk & Huijbregts Location and Locality
Page 9 of 23
(15)
2. LOCALITY
We now turn to the relations between the heads posited in such an architecture. A
simple notion of locality predicts that some of these relations can exist and that others
are blocked. This can be schematized as follows:
(16)
N LOC DIR V
2.1.1.
2.1.2.
2.1.3.
2.2.1.
2.2.2.
2.2.3.
In other words, we predict that the three adjacent links between these four heads are
permitted, while the three non-adjacent relations are blocked by locality. In the
remainder of this section, we will address the evidence that bears on these predictions.
The relevant section headings are shown in (16).
Pmax
PDIR
o
P’
N’
No AUGM
PLOC
o
sew-
-re
-qh
-aj
Van Riemsdijk & Huijbregts Location and Locality
Page 10 of 23
2.1. What is allowed
2.1.1. N - LOC
It is a fact about English that pictures are ON the wall, not, for example, AT the wall
or IN the wall. In Dutch, on the other hand, pictures are AT the wall (aan de muur).
There may be deep reasons for such facts and such differences,10
but for the time
being we might as well say that there is a specific, and at least partly unpredictable,
dependency between nouns such as wall and the preposition one has to use when
talking about some object that stands in a particular spatial relation to it. With wall it
is on, with sentence it is in (as in there is an eror in this sentence), etc. The important
fact, for our purposes here, is that once the choice of a locational preposition is fixed,
the rest follows. In other words, in a directional context we say put a picture on the
wall and insert an eror into this sentence, not *put a picture at the wall or *insert an
eror onto the sentence (cf. also the discussion in section 2.1.3.). Thus, once the type
of location is correctly selected, the choice of directional extensions is free (though it
may depend on the verb, of course).
This type of dependency is particularly clear with geographical proper names.
Holland is a flat country. And to many people, its minimal elevations are hardly
noticeable. Nevertheless distinctions are made in the language along such dimensions.
One example is the contrast between the Betuwe and the Veluwe. The Betuwe is an
area between the rivers Rijn and Waal, two major arms of the Rhine delta. And the
Veluwe is a sandy, heathy nature preserve north-east of the Betuwe. With the Betuwe,
Dutch uses in ('in'), whereas with the Veluwe Dutch uses op ('on'). To those familiar
with these areas or with some schooling in Dutch geography, this is not necessarily a
surprise. The Betuwe, is entirely flat and it is bordered, both in the north and in the
south, by massive dikes against flooding by the two rivers. From the perspective of
those dikes, then, and even from the perspective of the rivers themselves, the Betuwe
is a geographical depression, almost a valley. The Veluwe, on the other hand, could
almost be considered mountainous for Dutch standards in that it features elevations of
up to a few dozen meters. This being said, we must nevertheless note that people use
the corresponding prepositions correctly even if they have never been there, never had
any Dutch geogrphy lessons at school, and only know that these are stretches of land
where people can live, walk, ride, etc. So it is a piece of lexical knowledge, at least
partly idiosyncratic, that we have the following contrast:
(17) Jan woont in de Betuwe vs. Jan woont op de Veluwe
(“John lives in de Betuwe”) (“John lives on the Veluwe”)
But once this contrast is lexically fixed, the choice of adpositions for the various paths
follows automatically:
10
In this case the crucial notion, presumably, is the fact that a picture on a wall is in contact
with the wall. So, on is used even for ceilings, for example there is a fresco on the ceiling, even though,
in a sense, the fresco is underneath the ceiling and certainly not on top of it. Similarly, the Dutch
preposition aan, used for walls, can be used for ceilings as well. But the relevant type of contact is hard
to define in view of the fact that a stain, physically almost impossible to distinguish from a fresco, is
'on the ceiling' (op het plafond) rather than 'at the ceiling' (aan het plafond). It remains to be seen how
much of this is predictable from regular conceptualizations and how much is simply lexically encoded.
Van Riemsdijk & Huijbregts Location and Locality
Page 11 of 23
(18) AT: in de Betuwe (in) vs. op de Veluwe (on)
TO: de Betuwe in (into) vs. de Veluwe op (onto)
FROM: uit de Betuwe (out of) vs. van de Veluwe (from)
VIA: door de Betuwe (through) vs. over de Veluwe (over, across)
These are by no means isolated cases. In Finnish, we are told, cities differ as to
whether internal or external cases are used.11
Helsinki is referred to by means of
internal cases, while Turku is used with external cases. The paradigm is as follows:
(19) a. Helsingi-ssä in Helsinki inessive
b. Helsingi-stä out of Helsinki elative
c. Helsinki-in into Helsinki illative
(20) a. Tampere-e-lla at Tampere adessive
b. Tampere-e-lta from Tampere ablative
c. Tampere-e-lle to Tampere allative
Similarly, it appears that the way in which the location Ukraine is referred to has been
changed since the old regime was toppled and the Ukraine became independent.12
The
'old' way was to say 'on the Ukraine' and 'onto the Ukraine' while now they say 'in the
Ukraine' and 'into the Ukraine':
(21) a. >" I8D"4>, vs. >" I8D"4>J
na ukraine na ukrainu
on Ukraineprep on Ukraineacc
'on Ukraine' 'onto Ukraine'
b. & I8D"4>, vs. & I8D"4>J
v ukraine v ukrainu
in Ukraineprep into Ukraineacc
'in Ukraine' 'into Ukraine'
Finally, in Lezgian, Haspelmath (1993) reports that the location of certain Lezgian
villages is rendered by means of the superessive, while the other villages as well as all
non-Lezgian villages are referred to by means of the inessive. And again, the
directional cases have to be chosen accordingly. One of the 'special' villages is
Kasumkent. So, given that 'in Kasumkent' is expressed by means of the superessive,
the notion 'from Kasumkent' is correspondingly expressed by means of the
superelative: 13
(22) a. Q'asumxür.e-l b. Q'asumxür.e-l-aj
Kasumkent.AUGM-SUPER(-ESS) Kasumkent.AUGM-SUPER-EL
("in Kasumkent") ("from Kasumkent")
11
Again, some of this may have a conceptual origin: a larger city may be more easily
conceptualized as having an interior than a small town, but once grammaticalized, if a small town
grows into a big city, the grammar will not necessarily follow suit. In addition, there are morphological
constraints. For example, place names ending in -la take the internal cases rather than the external
ones. Thus, Mikkola takes the inessive Mikkolassa rather than the adessive Mikkolalla, probably as a
result of some kind of morpho-phonological dissimilation.
12
Thanks to Masha Yelenevskaya and Ben Hermans for pointing this out to us.
13
Recall that the indirective constitutes an accidental gap. More generally, the directive cases
are mostly used in temporal meanings in the present-day language. Other case forms tend to take over
in such cases. Thanks to Martin Haspelmath (p.c.) for elucidating this point. This is why the directives
have been omitted from the text paradigm.
Van Riemsdijk & Huijbregts Location and Locality
Page 12 of 23
Before we leave this topic, it should be pointed out that there are cases in which the
precise nature of the path appears to be determined by the noun. As an example,
consider the expression 'out of the blue' in the meaning of 'appearing quite suddenly
with no apparent origin.' You might say that the elative path is certainly determined
by the choice of the deadjectival noun blue. Indeed, there is no corresponding
expression 'into the blue' nor 'in the blue'. But this does not count as a real
counterexample since the location is lexically determined (that is, not free) as well.
After all you cannot say 'from the blue' either in anything near the intended meaning.
Hence, we have a tight lexical dependency ranging from N via LOC to DIR.14
We conclude that the lexical dependencies between specific choices of nouns and the
spatial adpositions or cases that go with them confirms the proposed architecture for
spatial phrases under a simple and robust notion of locality.
2.1.2. LOC - DIR
Turning now to the relation between LOC and DIR, consider the issue of mixed
systems, that is systems in which free and bound morphemes interact, unlike in the
'pure' systems of Lezgian and German described in section 1. Given the fact that we
have a binary choice both for LOC and for DIR, there is a total of four possibilities, as
shown in (23).
(23)
N LOC DIR
1. stem bound bound Lezgian, cf. section 1.2.
2. stem bound free Navajo, cf. section 2.2.1.
3. stem free bound Turkish, cf. sections 2.1.2. and 2.2.1.
4. stem free free German, cf. section 1.3.
Cases 1 and 4 were discussed above. Case 2 we turn to in section 2.2.1. below. Here
we focus on case 3. The most straightforward position to take on case 3 would be to
say that it cannot exist. Indeed, if free morphemes that represent the category LOC are
typically adpositions, and if adpositions are typically uninflected, then we predict that
such structures do not occur. Let us see why this is so. The LOC-element, being free,
will not attach to the noun. But what about the DIR-element? It must find a host, since
it is a bound morpheme. But where can it find a host. By hypothesis, the LOCelement
is an uninflected adposition, so it cannot serve as a host. And the noun is
unavailable because of locality considerations: LOC intervenes between N and DIR.
Given the assumptions, we believe this is the correct prediction.
Suppose, however, that the LOC element is inflectable. Then a totally different
situation arises. For now, the bound DIR-element can attach to the free LOCmorpheme
to be expressed. Typically, in such a situation the LOC-element is not an
adposition but a nominal element.15
This is indeed the situation that we find in
Turkish and many other languages. Consider the following examples.
14
We believe that this way of looking at things comes close to the notion of grammatical 'span'
as developed in Williams (in prep.)
15
This is not to deny the fact that such LOC-nouns are often degenerate nouns that may well
have the status of adpositions in many relevant respects.
Van Riemsdijk & Huijbregts Location and Locality
Page 13 of 23
(24) a. a—aç-tan Turkish
tree-ABL ('from the tree')
b. a—aç-ta
tree-LOC ('in the tree')
c. *a—aç-ta-tan
What these examples show is that the locative and the ablative case morphemes are in
complementary distribution. Both cannot be adjoined to a single noun stem. This
means that a DIR-morpheme, when used without a LOC-morpheme, can be a bound
morpheme. The point, however, is that when both are present the LOC-morpheme
must be a free morpheme, in fact a nominal element that requires the genitive case on
the noun it modifies.
(25) a. a—aç-2n arka-s2-ndan b. a—aç-2n arka-s2-nda
tree-GEN back-3sg-ABL tree-GEN back-3sg-LOC
('from behind the tree') ('behind the tree')
What these examples show is that bound morphemes in an architecture like the one
defended here must attach to an adjacent host. If no appropriate host is available in the
adjacent position, either another element capable of hosting the affix must substitute
for it, or there simply is no converging form available.
2.1.3. DIR - V
In assessing whether a syntactic relationship between DIR and V can exist, there are,
as before, two main aspects to take into account: selection and movement. On the
former, we can be brief. It is quite clear that many verbs select a directional
complement. We have to be quite careful, however. Take a verb like put in examples
like:
(26) a. put the plate on the table
b. put the problem behind you
c. put the bird in the cage
The problem is that English PPs of the type used here are ambiguous between a
purely locative meaning and a directional one. That the PP can be directional all by
itself is shown by the PP-with-NP construction:
(27) On the table with those plates!
That is, the PP does not need a verb to acquire the directional meaning. PPs can be
disambiguated, of course, by means of the addition of to, but then we see a partial
trade-off: when the governing verb (or the constructional context) explicitly imposes a
directional meaning, the addition of to may feel somewhat redundant or contrastive in
some cases:
(28) a. ?put the plate onto the table
b. ?put the bird into the cage
c. Into the cage with that bird!
Things are somewhat clearer in a language like German, in which the two senses are
differentiated by the choice of case: dative for locative expressions and accusative for
Van Riemsdijk & Huijbregts Location and Locality
Page 14 of 23
directional ones. And here we see that a verb like stellen (partly equivalent to put)
takes directional PPs exclusively.
(29) a. Er stellt den Teller auf denacc (*demdat) Tisch
he puts the plate on the table
b. Sie stellt den Stuhl unter denacc (*demdat) Tisch
she puts the chair under the table
Note, however, that the mild tradeoff that we observe here can be much more radical.
In such radical cases, DIR is typically expressed as part of the verb, while LOC is
expressed on the PP. Consider first an example from Classical Greek:
(30) a. kata-tithénai epì chthonós (6"J"-J42X<"4 ,BÂ P2@<`H)
down-put on floor ('put down on the floor')
b. aph-ístamai (apò) basiléoos (•N-\FJ":"4 ("BÎ) $"F48XjH)
off - position from king ('distance oneself from the king')
In (30a) the element specifying the direction (kata) is attached to the verb as a prefix,
while the location is expressed by means of the preposition epi, which is part of the
nominal projection. In (30b) we have a case where LOC and DIR are conflated into a
single preposition. Here we have the choice of just having a verbal prefix or having
the prefix as well as a kind of copy of the same element in the nominal projection,
thereby (artificially, as it were) separating out LOC and DIR.
Essentially the same situation is found in Latin, as shown in the following examples.
(31) a. extra fines e-gredi
beyond the limits out-step
b. (trans) flumen trans-nare
across river across-swim
(32) a. Caesar milites castris in/eduxit (Lehmann, 1992)
('Caesar led his soldiers into/out of their camp')
b. Caesar milites flumen traduxit
('Caesar led his soldiers across the river')
We leave open the question of whether this tradeoff is the result of head movement or
of some lexical form of reduction of redundancy.
Turning now to Dutch and German, we observe that the DIR-elements that were
discussed in section 1.3. can be attached to the verb. The following examples from
German and Dutch show this.
(33) a. …weil sie das Klavier auf den dritten Stock hinauf hätten tragen sollen
because they the piano on the third floor upward had carry should
'because they should have carried the piano up to the third floor'
b. …weil sie das Klavier auf den dritten Stock hätten hinauf-tragen sollen
(34) a. …omdat hij zijn fiets de vrachtwagen in wilde duwen
because he his bicycle the truck into wanted push
'because he wanted to push his bicycle into the truck'
b. …omdat hij zijn fiets de vrachtwagen wilde in duwen
Van Riemsdijk & Huijbregts Location and Locality
Page 15 of 23
On most accounts,16
the postpositional element originates as part of the directional PP,
is separated from that PP and adjoins to the governing verb. Depending on the
specific properties of the verb cluster and the language in question, the DIR-element
may end up non-adjacent to its remnant PP somewhere inside the verbal cluster, as
shown in the b-sentences.
Let us turn now to an even more radical case. Yucatec Maya is reported to lack the
means of expressing direction as part of the nominal projection entirely (cf.
Bohnemeyer (1998) and Goldap (1992)). That is, there is a neat separation: DIR
attaches to (or is attracted by) V, while LOC attaches to (attracts) N. Consider first the
pure locatives in (35).
(35) a. ti, le pìinoh-o, b. ti, u pàach le pìinoh-o,
Loc Def tree-D2 Loc Poss.3sg Posterior Def tree-D2
'at the tree' 'behind the tree' (literally: 'at back of the tree')
According to Bohnemeyer (1998), "inactive motion" verbs lexicalize punctual
change-of-location of a figure (theme) with respect to a ground object but no
continuous locomotion along a trajectory is defined with respect to the ground object.
Put more simply, location is expressed as part of the nominal projection, but the path
is expressed as part of the verb, though not in the form of a separate morpheme:
(36) a. le kàaro-o' h òok ti' le kàaha-o'
Def cart-D2 Prv enter(B.3sg) in/Loc Def box-D2 enter = go.to
'the cart entered [in] the box'
b. le kàaro-o' h hóok' ti' le kàaha-o'
Def cart-D2 Prv exit(B.3sg) in/Loc Def box-D2 exit = go.from
'the cart exited [in] the box'
c. le kàaro-o' ti' yàan ti' le kàaha-o'
Def cart-D2 Prv exist(B.3sg) in/Loc Def box-D2 exist = be.at
'the cart is in the box'
This separation, this division of labor between the verb and the nominal projection is
so strong that it persists in the Spanish spoken by the Yucatec Maya, as pointed out in
Lehmann (1992).
16
It has occasionally been argued that the postpositional element is a verbal particle from the
start and hence never a part of the PP. But if that were true, the NP would presumably have to be the
direct object. In an example like (31), however, there already is a direct object, so we would have a
kind of double object construction with the direct object preceding the second object. This seems
highly implausible. For further discussion, see Van Riemsdijk (1978), Broekhuis (forthcoming).
Among the relevant facts, note further that the directional object in an example like (34) cannot be
passivized, even when there is no other object around:
(i) *De vrachtwagen werd in gereden
the truck was into driven
'the truck was being driven into'
This example is instructive in yet another way. For inrijden does exist as a particle verb and as such has
the meaning of 'breaking in' as said of a (new) car. And indeed on the meaning 'the truck was broken
in' the example (i) is perfectly grammatical, cf. Van Riemsdijk (1978: 92).
Van Riemsdijk & Huijbregts Location and Locality
Page 16 of 23
(37) a. el ratón salió en su agujero YM-Spanish
the mouse left in its hole
b. el ratón salió de su agujero Standard Spanish
the mouse left from its hole
Consider now the question of how Yucatec Maya would express a trajectory with
both an explicit starting point and an explicit goal such as (38).
(38) the mouse went from the hole into the box
The answer is that it is impossible to express this without having two full-fledged
verbs. That is, a paraphrase of the type 'Juan left LOC-A; Juan arrived LOC-B' must
be used:
(39) a. le ch'o'-e' h hóok' ti' le áaktun-e';
Def mouse-D3 Prv exit(B.3sg) Loc Def hole-D3
b. (le ch'o'-e') h òok ti' le kàaha-o'.
Def mouse-D3 Prv enter(B.3sg) Loc Def box-D2
If we interpret the unique presence of DIR on (or in) the verb as the result of a
movement operation, that is, if we assume that DIR is incorporated into the verb, then
this surprising fact follows directly from across-the-board (ATB) considerations.
Consider, in fact, the following putative underlying structure:
(40)
GO FROM IN HOLE TO IN BOX
V DIR LOC N DIR LOC N
For both DIR-elements to incorporate into a single verb, they would have to do so in
an ATB-fashion (cf. Ross (1986), Williams (1978)). But in order for that to be
possible, they would have to be identical, which they are not. Hence, this derivation is
not available.17
17
Alternatively, one might assume that FROM and TO are incorporated in successive steps. But
first of all, it is not clear how the two directional phrases could be interpreted in any way other than as
Van Riemsdijk & Huijbregts Location and Locality
Page 17 of 23
2.2. What is not allowed
As a cautionary remark, we would like to point out that claims about negative
evidence must be taken with considerable grains of salt. We do not pretend to have
studied large numbers of languages in search for counterevidence. Our main purpose
here has been to suggest a new strategy of enquiry, one which we hope will help
guide future typological research. We therefore believe the evidence presented below
to be suggestive but not necessarily conclusive.
2.2.1. N - DIR
By hypothesis, a configuration in which DIR is attached (or at least adjacent) to N in
the presence of LOC could only arise if DIR attracts N or if DIR is incorporated into
N. Both processes are predicted to be blocked by locality. In order to see this,
consider again the Turkish examples presented in section 2.1.2. above, repeated here
as (41/42).
(41) a. a—aç-tan Turkish
tree-ABL ('from the tree')
b. a—aç-ta
tree-LOC ('in the tree')
c. *a—aç-ta-tan
tree-LOC-ABL ('out of the tree')
(42) a. a—aç-2n arka-s2-ndan b. a—aç-2n arka-s2-nda
tree-GEN back-3sg-ABL tree-GEN back-3sg-LOC
('from behind the tree') ('behind the tree')
What would constitute counterevidence to our claim? Since the structure of Turkish
extended nominal projectsions is agglutinative, we do not need to consider
incorporation. Instead, we focus on attraction. Suppose, then, that the noun is attracted
directly by DIR. We already know that the LOC and DIR morphemes are in
complementary distribution. Indeed, (43) is as ungrammatical (or perhaps even worse)
than (41c).
(43) *a—aç-tan-ta
tree-ABL-LOC
But if we choose the analytical way of expressing the location, DIR can still not be
attached to the head noun, regardless of whether we use a genitive or not:
(44) a. *a—aç-2n-dan arka-s2 b. *a—aç-tan arka-s2
tree-GEN-ABL back-3sg tree-ABL back-3sg
In both cases, by hypothesis, LOC would have had to be crossed over in order for DIR
to be attached diredctly to N. This, then, confirms our prediction.
(asyndetically) coordinated. That means that successive instances of head movement, applied to each
DIR-head separately, would violate the Coordinate Structure Constraint. This is, in fact, the essence of
ATB-movement. In addition, it has been proposed that multiple incorporation is blocked as a matter of
principle anyway (cf. Kayne (1994: §§ 3.3. & 3.4.)).
Van Riemsdijk & Huijbregts Location and Locality
Page 18 of 23
A somewhat more complex argument along similar lines comes from Navajo (cf.
Kaufman (1992)). In Navajo, subjects agree with the predicate in person and number,
as shown in (45).
(45) a*áán tl*éé* Navajo
hole 3sgS.be.dark
'the hole is dark'
However, when the subject is a locational noun, it is also possible to use a special
kind of locative agreement. If this mode of expression is chosen, LOC is either
marked both on the noun and on the predicate, or on the predicate alone. Having just a
locative 'subject' without overt locative agreement marking on the predicate is
ungrammatical, as shown in (46).
(46) a. a*áán-góne** ha-tl*éé*
hole-in Loc-be.dark
'it is dark in the hole'
b. a*áán ha-tl*éé*
c. *a*áán-góne** tl*éé*
For objects, an identical situation is found. In other words, alongside 'regular'
person/number agreement with the object, as in (47a) it is also possible to have
locative agreement with objects as in (47b-d). And again, agreement is obligatory,
though the copy of LOC on the noun is optional.
(47) a. ashkii bikooh yi-yii-»ts
boy canyon 3sgO-3sgS.Perf-see
'the boy saw the canyon'
b. ashkii bikooh-góyaa hoo-»ts
boy canyon-down.in Loc-3sgS.Perf-see
'the boy saw the canyon'
c. ashkii bikooh hoo-»ts
d. *ashkii bikooh-góyaa yi-yii-»ts
Let us now turn to directional expressions. The DIR-marker yah is a free morpheme.
And by hypothesis it intervenes between the locative marked noun and the predicate.
Hence, we predict that in the presence of the DIR-marker locative agreement is
blocked. As (48) shows, this is exactly right. The locative marker must appear on the
noun, while agreement on the predicate leads to ungrammaticality.
(48) a. ashkii hooghan-góne** yah ííyá
boy house-in into 3sgS.Perf.go
'the boy went into the house'
b. *ashkii hooghan-góne** yah ha-ííyá
2.2.2. LOC - V
We start examining this case by looking at an example that involves what might be
called semantic incorporation. The verb stay strongly implies the presence of a
location in which the staying takes place but that location need not be expressed
Van Riemsdijk & Huijbregts Location and Locality
Page 19 of 23
overtly. Suppose we assume, then, that LOC is semantically incorporated into such a
verb, regardless of whether the location is also expressed overtly. Note now that with
such a verb the origin explicating the situation before the staying starts cannot be
expressed.
(49) a. John stayed (here)
b. *John stayed (here) from Rotterdam
c. John came from Rotterdam and stayed (here)
Such evidence is quite shaky, however, not in the least because it is very unclear
whether any syntactic operation is involved in this type of semantic incorporation.
What would be a more syntactic piece of evidence? A very straightforward line of
reasoning can be derived from the behavior of directional phrases in Dutch and
German. For ease of reference (cf. section 1.3. and section 2.1.3.) we will concentrate
on German here. German, like Dutch, has what is generally referred to as particle
incorporation. Particle incorporation can apply both to locative and to directional
particles. Consider (50/51). The b-sentence is the non-incorporated version, while the
c-sentences show that the particle is attached to 'its' verb and has inverted into the
verb cluster.
(50) a. Du bleibst oben
you stay up(stairs)
b. …weil du oben hättest bleiben sollen
because you upstairs had stay should
'because you should have stayed upstairs'
c. …weil du hättest oben bleiben sollen
(51) a. Du kommst herauf
you come upstairs
b. …weil du herauf hättest kommen sollen
because you upstairs had come should
'because you should have come upstairs'
c. …weil du hättest herauf-kommen sollen
With this in mind, there is an easy way to test our prediction. As was shown in section
1.3., German has a whole range of circumpositional phrases in which both a location
and a direction are expressed. In view of the above examples (50/51) we might expect
both the LOC and the DIR element in such circumpositional phrases to be available
for incorporation into the verb. But in actual fact only DIR can incorporate, again in
full obeyance of locality.
(52) a. Er springt [auf das Dach hinüber]
he jumps on the roof across
'he jumps across onto the roof'
b. …weil er [auf das Dach hinüber] hätte springen können
because he on the roof across had jump could
'because he could have jumped across onto the roof'
c. …weil er [auf das Dach ___ ] hätte hinüber-springen können
d. *…weil er [ ___ das Dach hinüber] hätte auf-springen können
Van Riemsdijk & Huijbregts Location and Locality
Page 20 of 23
(53) a. Sie fährt [unter der Brücke durch]
she drives under the bridge through
'she crosses underneath the bridge (driving)'
b. …weil sie [unter der Brücke durch] wird fahren müssen
because she under the bridge through will drive must
'because she will have to drive through/underneath the bridge'
c. …weil sie [unter der Brücke ___ ] wird durch-fahren müssen
d. *…weil sie [ ___ der Brücke durch] wird unter-fahren müssen
We conclude that the pattern of incorporation possibilities encountered here is in full
accordance with locality as applied to the architecture of locational phrases we are
defending here.
2.2.3. N - V
This prediction, the impossibility of a direct relation between the noun in a locational
expression and the verb, may well appear strange upon first consideration. After all,
semantic selection does seem to apply between a verb and a noun even when the noun
is in a locational expression:
(54) a. dive into the pool vs. #dive into the surface
b. climb on(to) the top vs. #climb on(to) the horizon
The verb dive apparently selects locational objects that have an interior, and the verb
climb selects locational objects which require vertical locomotion in order to reach
them. But two remarks are in order here before we conclude that such examples
constitute counterevidence to our claim. First, notice that both verbs actually
subcategorize for an (optional) directional goal-phrase. In other words, the lexical
specification of these verbs must also include the presence of DIR and LOC, hence, in
that sense no locality violation is involved. Second, notice that what we are talking
about here is semantic selection. Now, the real semantic content of a locational phrase
is concentrated in the lexical head. The LOC and DIR elements, while by no means
semantically empty, are quite bleached and unspecific in their meaning. It would not
be unreasonable to assume that semantic selection ignores functional heads. In fact,
this is what undoubtedly has to be assumed anyway. After all determiners do not
block the semantic selection of the noun in a direct object construction either.18
Another way of approaching this issue is to examine noun incorporation in relation to
nouns in locational expressions. The prediction, again, is quite clear: any such
incorporation process should be blocked. We do in fact believe this to be the correct
prediction, though we have not studied noun incorporation in any great detail. Baker
(1988:81ff) does mention the fact that locatives sometimes appear to be incorporated.
Baker cites the following Mohawk examples from Hewitt (1903):
(55) H~kare' n•ñ' ia' -e- 'heñt - ~ra'ne' ka-'heñt-owan“' Mohawk
after now TL-3F-field-reached PRE-field-large
'Then, after a while, she reached a grassy clearing that was large'
(56) O' na'karoñtot“' nene' karoñto' ne dji teieia-'hia-tha' ?
what PARTITIVE-PRE-tree-SUF PRE-tree-SUF where IMP-stream-cross-INSTR
'What kind of tree is used to cross the stream there?'
18
For more extensive discussion, see for example Corver & Van Riemsdijk (to appear).
Van Riemsdijk & Huijbregts Location and Locality
Page 21 of 23
While some kind of semantic locatives do seem to be incorporated here, they are, in a
sense, like direct objects too, as Baker remarks. Indeed, in English one would say
reach a field or cross a stream. More significantly, there is no sign of any locative
expression containing any visible trace of either DIR or LOC outside the complex
verb. We may thus conclude, with Baker, that true locative noun incorporation does
not exist, in conformity with the point we are making here.
2.3. Conclusions
This concludes our survey of what elements can and cannot be directly related by
syntactic processes in complex locational phrases. We have argued that an an
architecture of locational phrases in which the nominal projection is extended by a
locative and a directional functional head, in that order, is well motivated, not only by
the existence of morphological and syntactic patterns in which such a structure is
quite apparent, but also from a variety of locality considerations.
Van Riemsdijk & Huijbregts Location and Locality
Page 22 of 23
3. REFERENCES
Asher, N. & P. Sablayrolles (1994) >A Compositional Spatio-Temporal Semantics of
French Motion Verbs and Spatial PPs= in M. Harvey & L. Santelmann (eds.)
SALT IV, pp 1-15, Ithaca, N.Y.: Cornell University.
Baker, M.C. (1988) Incorporation. Chicago: University of Chicago Press.
Bierwisch, M. (1988) 'On the Grammar of Local Prepositions' in M. Bierwisch, W.
Motsch & I. Zimmermann (eds.) Syntax, Semantik, und Lexikon: Rudolf
Rçñi…ka zum 65. Geburtstag, pp 1-65. Berlin: Akademie Verlag.
Bloom, P., M.A. Peterson, L. Nadel & M.F. Garrett eds. (1996) Language and Space.
Cambridge, MA: MIT Press / Bradford.
Bohnemeyer, J. (1998) Time Relations in Discourse: Evidence from a Comparative
Approach to Yucatec Maya. Tilburg University Ph.D. thesis.
Bok-Bennema, R. (1991) Case and Agreement in Inuit, Dordrecht: Foris.
Bourquin, T. (1891) Grammatik der Eskimo-Sprache wie sie im Bereich der
Missionsniederlassungen der Brüdergemeinde and der Labradorküste
gesprochen wird. London.
Broekhuis, H. (forthcoming) Modern Grammar of Dutch: The PP, MGD Working
Papers, Tilburg University.
Comrie, B. & M. Polinsky (1998) 'The Great Daghestan Case Coax' in: A. Siewierska
& J.J. Song (eds.) Case, Typology, and Grammar, pp 95-114, Amsterdam:
John Benjamins.
Corver, N. & H.C. van Riemsdijk eds. (to appear) The Content of Function Words and
the Function of Content Words. Berlin: Mouton de Gruyter.
Goldap, C. (1992), 'Morphology and Semantics of Yucatec Space Relators,'
Zeitschrift fuer Phonetik, Sprachwissenschaft und Kommunikationsforschung
45.6, 612-625.
Grimshaw, J. (1991) >Extended Projection,= unpublished manuscript, Brandeis
University.
Gruber, J.S. (1965) Studies in Lexical Relations. Ph.D. Dissertation, MIT. Reprinted
in Gruber (1976).
Gruber, J.S. (1976) Lexical Structures in Syntax and Semantics. Amsterdam: North
Holland.
Hale, K. (1986) >Notes on World View and Semantic Categories: some Warlpiri
Examples= in: P.C. Muysken & H.C. van Riemsdijk (eds.) Features and
Projections. Dordrecht: Foris.
Haspelmath, M. (1993) A Grammar of Lezgian. Berlin: Mouton de Gruyter.
Hewitt, J. (1903) Iroquoian Cosmology. 21st
Annual Report of the Bureau of
American Ethnology. Washington, D.C.: Smithsonian Institution.
Hjelmslev, L. (1935/37) La catégorie des cas. Facs. edition 1972, München:
Wilhem Fink Verlag.
Jackendoff, R (1983). Semantics and Cognition. Cambridge: The MIT Press.
Jackendoff, R. (1990) Semantic Structures. Cambridge, MA: MIT Press.
Kaufman, E. (1974) 'Navajo Spatial Enclitics: A Case for Unbounded Rightward
Movement' Linguistic Inquiry 5: 507-533.
Kayne, R. (1994) The Antisymmetry of Syntax. Cambridge, Massachusetts: MIT Press.
Koopman, H. (1993) >The Structure of Dutch PPs= unpublished manuscript, UCLA.
Van Riemsdijk & Huijbregts Location and Locality
Page 23 of 23
Lehmann, C. (1992) "Yukatekische lokale Relatoren in typolgischer Perspektive,"
Zeitschrift fuer Phonetik, Sprachwissenschaft und Kommunikationsforschung
45.6, 626-641.
Levelt, W.J.M. (1996) 'Perspective Taking and Ellipsis in Spatial Descriptions' in P.
Bloom et al. (eds.) pp 77-107.
Riemsdijk, H.C. van (1978) A Case Study in Syntactic Markedness. Dordrecht: Foris.
Riemsdijk, H.C. van (1990) 'Functional Prepositions' in H. Pinkster & I. Genée (eds.)
Unity in Diversity: Papers presented to Simon C. Dik on his 50th birthday.
Dordrecht: Foris.
Riemsdijk, H.C. van (1998a) 'Head Movement and Adjacency' Natural Language and
Linguistic Theory 16, 633-678.
Riemsdijk, H.C. van (1998b) 'Categorial Feature magnetism: The Endocentricity and
Distribution of Projections' The Journal of Comparative Germanic Linguistics
2: 1-48.
Ross, J.R. (1986) Infinite Syntax. Norwood, New Jersey: Ablex Publ. Coorporation.
Schulze, W. (1983) Die Sprache der Uden in Nord-Azerbajdñan. Wiesbaden: Otto
Harassowitz.
Schulze, W. (1985) ‘Remarks on the Case System of the Lezgian Languages’ Folia
Slavica 7, 303-322.
Talmy, L. (1983) 'How Language Structures Space' in H. Pick & L. Acredolo (eds.)
Spacial Orientation: Theory, Research and Application. New York: Plenum.
Williams, E.S. (1978) "Across-the-Board Rule Application," Linguistic Inquiry 9.1,
31-43.
Williams, E.S. (in preparation) Representation Theory, book manuscript.
Zwarts, J. (1995) >Lexical and Functional Direction in Dutch= in: M. den Dikken & K.
Hengeveld (eds.) Linguistics in the Netherlands 1995. Amsterdam: John
Benjamins.