prepositions in discourse

42
is is a contribution from Functions of Languages 15:2 © 2008. John Benjamins Publishing Company is electronic file may not be altered in any way. e author(s) of this article is/are permitted to use this PDF file to generate printed copies to be used by way of offprints, for their personal use only. Permission is granted by the publishers to post this file on a closed server which is accessible to members (students and staff) only of the author’s/s’ institute. For any other use of this material prior written permission should be obtained from the publishers or through the Copyright Clearance Center (for USA: www.copyright.com). Please contact [email protected] or consult our website: www.benjamins.com Tables of Contents, abstracts and guidelines are available at www.benjamins.com John Benjamins Publishing Company

Upload: clarissa-ayres

Post on 08-Apr-2016

14 views

Category:

Documents


1 download

DESCRIPTION

prepositions

TRANSCRIPT

Page 1: Prepositions in Discourse

This is a contribution from Functions of Languages 15:2© 2008. John Benjamins Publishing Company

This electronic file may not be altered in any way.The author(s) of this article is/are permitted to use this PDF file to generate printed copies to be used by way of offprints, for their personal use only.Permission is granted by the publishers to post this file on a closed server which is accessible to members (students and staff) only of the author’s/s’ institute.For any other use of this material prior written permission should be obtained from the publishers or through the Copyright Clearance Center (for USA: www.copyright.com). Please contact [email protected] or consult our website: www.benjamins.com

Tables of Contents, abstracts and guidelines are available at www.benjamins.com

John Benjamins Publishing Company

Page 2: Prepositions in Discourse

English prepositions in Functional Discourse Grammar

Evelien KeizerUniversity of Amsterdam

Adpositions have always been problematic in terms of analysis and representa-tion: should they be regarded as lexical elements, with an argument structure, or as semantically empty grammatical elements, i.e. as operators or functions? Or could it be that some adpositions are lexical and others grammatical, or even that one and the same adposition can be either, dependent on its use in a par-ticular context? In Functional Grammar (Dik 1997a,b) adpositions are analysed as grammatical elements, represented as functions expressing relations between terms (referring expressions). Various alternative treatments have been proposed within FG, all of which, however, fail to solve all the problems, or address all the relevant questions involved. This article offers an analysis of English prepositions within the model of Functional Discourse Grammar (Hengeveld and Mackenzie 2006, 2008), based on the semantic, syntactic and morphological evidence avail-able and fully exploiting the novel features of this model.

1. Introduction1

Most linguistic frameworks (formal or functional, formalistic or non-formalistic) acknowledge the importance of differentiating between lexical (or content) ele-ments and grammatical (or form) elements. Functional Grammar (Dik 1997a,b) forms no exception, and distinguishes between the two classes of elements in un-derlying representation: whereas the former are analysed as predicates, the latter are analysed as operators or functions (e.g. Dik 1997: 159). Such an approach is un-problematic for those elements that are clearly lexical or grammatical. Thus, what-ever their background, linguists seem to agree that full verbs, nouns and adjectives are prototypical lexical elements, whereas affixes, auxiliaries and particles are gen-erally regarded as grammatical elements. There are, however also elements — in some cases, classes of elements — whose status as lexical or grammatical is less straightforward. One of the most problematic classes of elements in this respect

Functions of Language 15:2 (2008), 216–256. doi 10.1075/fol.15.2.03keiissn 0929–998X / e-issn 1569–9765 © John Benjamins Publishing Company

Page 3: Prepositions in Discourse

© 2008. John Benjamins Publishing CompanyAll rights reserved

English prepositions in Functional Discourse Grammar 217

is that of adpositions, which have been classified as either lexical or grammatical (depending on the linguistic framework in question), and sometimes as both (with some adpositions being classified as lexical and others as grammatical).

In all standard works on the theory on Functional Grammar (Dik 1978, 1989, 1997a,b) adpositional phrases are analysed as terms with a semantic function. This corresponds to the view taken in most grammaticalisation studies, where ad-positions are typically regarded as part of the grammatical inventory (e.g. Bybee 2003: 145; Heine and Kuteva 2002a: 3–4; Hopper and Traugott 1993: 4). Such an approach deviates from the position taken in both formal and cognitive semantics, as well as in many more syntactically oriented frameworks, where adpositions are often assumed to form a lexical category (alongside V, N and A). Unfortunately, the classification of adpositions as either lexical or grammatical is rarely justified by linguistic evidence (in the form of tests or a list of criteria) (see Keizer, 2007).

The controversial nature of adpositions is also evident from the fact that, in the course of time, various alternative treatments of adpositions have been pro-posed within Functional Grammar. As will be demonstrated, however, these ear-lier proposals fail to solve, or often even address, the basic problems involved. The aim of the present article is to propose an analysis and representation of English prepositions and prepositional constructions within the newly developed theory of Functional Discourse Grammar (Hengeveld and Mackenzie 2006, 2008) which provides answers to the following questions:2

1. Is it plausible to assume that all prepositions in English are grammatical ele-ments and that the wide range of meanings they express are represented in underlying structure by a very limited number of highly abstract semantic functions?

2. If, on the other hand, we choose to analyse prepositions as predicates, should this be true of all prepositions, or is there evidence to suggest some preposi-tions are better regarded as more lexical and others as more grammatical (e.g. Lehmann 2002: 1,8)?

3. Is it justified and/or feasible to distinguish between lexical and grammatical uses of one and the same preposition (e.g. on in the book on the table vs. his dependence on his wife)?

4. What kind of evidence can be used to determine whether prepositions are lexical or grammatical?

The article is organised as follows. Section 2 presents an overview of the various treatments of adpositions proposed in Functional Grammar so far. In Section 3 the arguments and evidence provided in these earlier proposals are evaluated and further evidence is brought in from semantics and cognition, morphology and syntax. Together, this information forms the basis for a new proposal for the

Page 4: Prepositions in Discourse

© 2008. John Benjamins Publishing CompanyAll rights reserved

218 Evelien Keizer

treatment of English prepositions and prepositional constructions within Func-tional Discourse Grammar, made in Section 4. Section 5 offers some conclusions. Finally, as it will be assumed that the answers to the questions addressed may differ from language to language, this article will be concerned with English data only.

2. Adpositions in FG: An overview

In standard Functional Grammar (henceforth FG), adpositions, as grammatical elements, are represented as semantic functions assigned to terms (i.e. referring expressions); the expression the mouse under the table in (1a), for instance, is given the representation in (1a′), (cf. Dik 1997a: 206–207):

(1) a. the mouse under the table a′. (d1x1: mouseN: {(d1x2: tableN)Loc})

In (1a′), under the table is represented as a term with the semantic function Loc(ation); within the expression the mouse under the table, this locative term is used in an attributive function, as the second restrictor on the term variable x1 (representing the referent of the entire term).3

This approach, however, turned out to be problematic in a number of respects. The first problem concerns the idea that adpositional phrases are represented as terms: whereas terms are defined as referring expressions, adpositional phrases typically fulfil a non-referring (modifying or predicative) function (as is clear from the representation in (1a′), where the curly brackets indicate a predicative func-tion). To account for the predicative function of terms with such semantic func-tions as location, direction, path and source, Dik (e.g. 1997a: 206–207) introduces a term-predicate formation rule, converting such terms into one-place predicates.4 This does not, however, solve the problem: the input is still a term referring to an entity. Take an example like the mouse under the table. The modifying phrase under the table is represented as a term, the table, with the semantic function Loca-tion. However, the table referred to is not itself the location of the mouse; instead the term the table refers to an entity in relation to which the location of the mouse is indicated. Nor would such a view be logically acceptable: in a sentence like The mouse under the table looked frightened, reference is made to a mouse and a table; the place, or location, on the other hand, is not referred to.5 Instead the property of being under the table is predicated of the mouse, a point which was convincingly made by Jackendoff (1983: 161–163), and which, within FG, has resulted in the introduction of the p-variable (Mackenzie 1992a; now the l-variable, Hengeveld and Mackenzie 2006, 2008).

Page 5: Prepositions in Discourse

© 2008. John Benjamins Publishing CompanyAll rights reserved

English prepositions in Functional Discourse Grammar 219

Secondly, if adpositions are to be regarded as grammatical elements, this means that they cannot have a meaning of their own: inserted by means of an ex-pression rule, they can only derive their meaning from the semantic function as-signed to the term as a whole. The number of semantic functions available is, how-ever, limited. As a result, such an approach cannot do justice to (a) the differences in meaning between different adpositions expressing the same semantic function (e.g. at, in, by, above, below, beside, opposite, across, between, before, behind, beyond and around, all of which express Location); (b) the large range of senses of one single adposition (e.g. the many senses of over described in various studies from cognitive grammar, such as Brugman 1988; Brugman and Lakoff 1988, Dewell 1994; Kreitzer 1997; Tyler and Evans 2001; see also evidence from psycholinguistic research on the functional elements determining the use of spatial prepositions, e.g. Coventry et al. 1994, Garrod et al. 1999, Coventry 2003).

Various attempts have been made to come up with a more satisfactory treat-ment of adpositions and adpositional phrases in FG. Thus it has been suggested that, to tackle the problem of underspecification, more semantic functions need to be distinguished to match the large number of adpositions. De Groot (1989: 13), for instance, introduces the semantic function Subessive to trigger the preposition under. As pointed out by Mackenzie (1992b:4), however, this would lead to a pro-liferation of semantic functions, without, however adding to the explanatory value of the theory (see also Samuelsdorff 1998: 273).

For most authors challenging the standard FG view of adpositions, however, the crux of the matter lies in the assumption that all adpositions are grammati-cal elements. Weigand (1990), for instance, proposes to represent prepositional phrases through a combination of semantic functions (e.g. Loc, Path) and a lim-ited number of prepositional predicates (e.g. on and under). A simplified example is given in (2):

(2) a. The mouse ran from under the table. a′. e run (ag m mouse) ([path, from] p under (y table)))

where the preposition under is analysed as a one-place predicate heading a locative expression ‘p’,6 which in turn is used as the argument with the semantic function Path (expressed as from). The advantage of this approach is that we need only a small set of semantic functions, as it is not necessary to have a separate semantic function for each locative or directional preposition.

Taking the idea of lexical prepositions one step further, Mackenzie (1992b, 2001) proposes to represent most prepositions as predicates. In addition, Mack-enzie distinguishes a small set of grammatical (spatial and temporal) preposi-tions, which he regards as direct realisations of the semantic functions Loc(ation), So(urce), Path, All(ative) and App(roach):7

Page 6: Prepositions in Discourse

© 2008. John Benjamins Publishing CompanyAll rights reserved

220 Evelien Keizer

(3) Loc: at (spatial and temporal) So: from (spatial and temporal) Path: via (spatial), for (temporal) All: to (spatial and temporal), until/till (temporal) App: towards (spatial)

All other prepositions are assumed to be stored in the lexicon as one-place predi-cates. These lexical prepositions can then combine with one of the five semantic functions to form a complex preposition (e.g. in + All = into). In addition, Mack-enzie’s proposal can accommodate such embedded prepositional constructions as in (4a) by analysing from as the direct realisation of the semantic function Source, which is assigned to a place-denoting term headed by the prepositional predicate under:

(4) a. from under the table a′. (d1pi: underP (d1xi: fj: tableN)Ref)So

Moreover, Mackenzie argues, such a system can account for the use of locative ad-verbs (as in I met him outside), as well as for the fact that some, but not all preposi-tions (i.e. only the lexical ones) allow adverbial modification (e.g. right behind the door).8

Another interesting contribution to the debate is that by Bakker and Siewier-ska (2002), who advocate the view that, since the distinction between lexical and grammatical elements (of any kind) is graded rather than strict, both ought to be included in the lexicon. This means that all adpositions, whether predicates or grammatical elements, will be assumed to have an entry in the lexicon, which will include either a meaning definition (in the case of lexical adpositions) or an ab-stract predicate (in the case of a grammatical adposition). Thus, following Mack-enzie (1992b), they consider above to be a lexical preposition, with “a complex set of semantic features, procuring it a full blown entry in the lexicon” (Bakker and Siewierska 2002: 160). The meaning postulate for above will therefore consist of a number of abstract locational predicates, as represented in (6a). The lexical entry for the grammatical preposition at, on the other hand, will contain only the ab-stract predicate LOC, corresponding to the semantic function LOC. Whether or not an adposition will appear as a predicate in the underlying representation of an expression will therefore depend on the complexity of its lexical entry.

(6) a. LEX.ENTRY= ‘above’, [P], ((xi)ZERO (xj)REF, …, [SUPERIOR (xi)ZERO (xj)GOAL: (VERTICAL)DIR & NOT (CONTACT (xi)ZERO (xj)GOAL)]

b. LEX.ENTRY= ‘at’, [P], ((xi)ZERO (xj)REF, …, [LOC (xi)ZERO (xj)REF]

Pérez Quintero (2004) differs from all these proposals in that she regards all adpo-sitions as lexical items, to be represented as one-place predicates. Her reasons for

Page 7: Prepositions in Discourse

© 2008. John Benjamins Publishing CompanyAll rights reserved

English prepositions in Functional Discourse Grammar 221

rejecting the standard approach are the same as those given by Mackenzie (1992b, 2001, 2002): firstly, even if it were possible in every utterance to trigger the correct adposition by extending the set of semantic function, such an approach would have very little explanatory value; secondly, since adpositions can be said to have ascriptive value in the sense that they designate a relation between two entities, and since, according to Mackenzie (2002: 3), availability for the communicative subact of ascription is the defining property of a predicate, adpositions qualify for predicate status (Pérez Quintero 2004: 158). Pérez Quintero sees no reason, however, to distinguish a set of grammatical adpositions: the fact that some adpo-sitions have a primitive meaning, she argues, does not necessarily mean that they cannot function as predicates (Pérez Quintero 2004: 159). Pérez Quintero does, however, allow for a grammatical use of certain adpositions. In that case, the adpo-sition performs a purely syntactic function — that of indicating case — and has no independent meaning. In English only three prepositions can be used this way: to, when used to express the semantic function of Recipient; by when used to express the semantic function of Agent; and of in such constructions as a man of honour (Pérez Quintero 2004: 163–164).

Most of the arguments put forward in these proposals are semantic in nature (‘primitive meaning’, ‘independent meaning’). Sometimes syntactic criteria have been used (combinatory properties, predicative use, modifiability), while some ar-guments are of a more general theoretical nature (lack of explanatory value). With the notable exception of Mackenzie (1992b), however, no attempt has been made to identify a set of criteria for systematically testing the lexical-grammatical status of adpositions. The next section will be a first step in that direction.

3. The evidence

3.1 Grammatical vs. lexical adpositions

3.1.1 Semantics and cognitionOne of the central questions in determining the lexical-grammatical status of ad-positions is that of whether or not adpositions — some or all — can be said to have lexical content. In an approach in which all adpositions are treated as gram-matical items, the answer is clearly ‘no’. All of the alternative proposals mentioned, however, argue in favour of a system in which there are grammatical as well as lexical adpositions.9 Such a view does, of course, beg the question of which ad-positions have meaning and which do not. Unfortunately, the various proposals, even within FG, offer different classifications. As we have seen, Weigand (1990) recognises only a limited number of lexical predicates (including on and under),

Page 8: Prepositions in Discourse

© 2008. John Benjamins Publishing CompanyAll rights reserved

222 Evelien Keizer

whereas Mackenzie classifies almost all English prepositions as lexical, reserving grammatical status for a small set only (at, from, to, towards, via, for and until/till). In other frameworks, we find other classifications. In Role and Reference Gram-mar (RRG; Van Valin and LaPolla 1997, based on Jolly 1993), be-at and be-via are regarded as primitives; all other prepositions are predicative and are analysed in terms of these primitives (e.g. from = BECOME NOT be-at). Dowty (1979), on the other hand, qualifies be-at, be-in and be-on as primitives. Clearly, there is no con-sensus on which (kinds of) criteria are to be used to distinguish the two classes.

Nor is it just a problem of which adpositions belong to which class; matters are further complicated by the fact that it is difficult, if at all possible, to determine where one category ends and the other begins. As is well-known from grammati-calisation studies, grammatical elements typically develop out of lexical elements, whereby the exact point of transition from one category to the other is difficult, perhaps even impossible, to establish.10 This means that even within a particular category of words, such as adpositions, some elements will be more lexical than others, and that, even if we are able to provide a satisfactory classification of the clearest cases, there may always be cases that defy straightforward categorisation.

Let us, however, for the purposes of this article, concentrate on the seman-tic evidence available for considering adpositions (some or all) as either gram-matical or lexical. As pointed out before, the standard approach of analysing all adpositions as semantically empty elements, expressing a limited number of se-mantic functions, is untenable because it leads to gross underspecification. This underspecification is due to the fact that the different prepositions have different meanings: in English, for instance, on, under, near and in do not merely indicate location; they each have their own specific meaning, which determines their ap-propriateness in a particular context. Now it may be that, generally speaking, these meanings are more abstract and schematic than those of most nouns or verbs; this does not, however, mean that they lack semantic content. It might, for instance, be argued that the primary meaning of a preposition like in (say, containment), is, in fact, much more specific than the meaning of a noun like thing or a verb like do; similarly, the meaning of under, even in its metaphorical use, seems to be more tangible than that of a highly abstract noun like sort. In both cases, it can be argued that the preposition makes a greater semantic contribution to the construction they appear in than the particular noun or verb (cf. Langacker 1987: 18–19).11

If there is one preposition that is generally claimed to have little semantic con-tent, it is the preposition of (e.g. Huddleston and Pullum 2002: 658; Lindstromberg 1998: 195; see also Pérez Quintero 2004: 164), partly because, unlike the other prep-ositions, it no longer has any concrete, literal sense, and partly because, possibly as a result of this, it can be used to indicate a wide range of sometimes very dis-parate meanings (including possession, relative-of, part-of, property-of, type-of,

Page 9: Prepositions in Discourse

© 2008. John Benjamins Publishing CompanyAll rights reserved

English prepositions in Functional Discourse Grammar 223

quantity-of, as well as age, size, form, source, content, depiction; e.g. Huddleston and Pullum 2002: 442,477). Clearly, it will be difficult to find a definition which would capture all of these meanings. The result would obviously be a very abstract meaning definition, like the one offered by Langacker (1992b:487; cf. 1992a: 296), according to whom “[i]t seems quite accurate to describe [all these relationships] as designating some kind of intrinsic relationship between the two participants”.

Interestingly enough, however, of is typically not included in the class of gram-matical prepositions. So which semantic criteria are used to decide on the lexical or grammatical status of a preposition? Mackenzie (1992b:10) justifies the recogni-tion of a small set of grammatical prepositions on the basis that they are ‘primitive’ elements, which defy further etymological or syntactic analysis (with reference to Kahr 1975: 43) (for syntactic evidence, see Section 3.1.2). But is this really true of each of the prepositions classified as grammatical by Mackenzie? According to Mackenzie (1992b:16), what the five spatial prepositions at, from, to, towards and via have in common is “the property that the Ref-argument of these prepositions is always understood as a (zero-dimensional) point in space”. Now, this has in-deed been acknowledged to hold for at; Herskovits (1986: 128–140), for instance, describes the functions of at as locating two entities at precisely the same point in space and construing them as geometric points (cf. Lee 2001: 23). In the case of from, via, to and towards, however, some further meaning element is clearly pres-ent, which would seem to disqualify them as primitives. To confuse matters even more, in RRG from and to are indeed analysed as decomposable lexical elements, whereas via is treated as a primitive (like at).

One might, however, wonder whether qualifying the referent of via and from as ‘zero-dimensional points in space’ is, in fact, justified. Although both preposi-tions can certainly be used in combination with non-dimensional entities, this is by no means a requirement. Lindstromberg (1998: 39–40) shows that the referent of spatial from, indicating the starting point of some path, may be precisely named (the non-dimensional argument in (7a)), or more or less vaguely located (by means of the one-, two- or three-dimensional arguments in (7b–d), respectively).

(7) a. Draw a line from point A to point B. b. Draw a line from somewhere on line A to line B.12

c. The rocker slowly rose from the surface of the moon. d. The noise came from the cave. (Lindstromberg 1998: 40)

Note also that neither the Oxford English Dictionary (OED) nor the Merriam-Web-ster Online Dictionary (Webster) restricts the meaning of from to non-dimensional arguments only. OED describes the relevant meaning of from simply as “indicating the place, quarter, etc. whence something comes or is brought or fetched”, while

Page 10: Prepositions in Discourse

© 2008. John Benjamins Publishing CompanyAll rights reserved

224 Evelien Keizer

Webster describes from as “indicating the starting point of a physical movement”. The same can be said of the preposition via. Both dictionaries describe its mean-ing as ‘by way of ’; OED adds to this “by the route which passes through or over (a specified place)”. Once again this explains why, like from, via allows for a three-dimensional argument. Lindstromberg (1998: 128–129) even regards this as the central meaning of via:

via is vague about the exact relationship of the path in relation to the Landmark [here: argument; MEK]. Thus, via can be used when the path goes through the Landmark (which is probably the central meaning), or when the path only grazes the Landmark, or even when the path goes near but does not touch the Land-mark.

Of the five spatial prepositions, this leaves us with at as the last surviving serious candidate for primitive (and as such grammatical) status. Unlike the other four prepositions, the primary meaning of at does seem to be location at some non-dimensional point in space. By not indicating the exact relationship between the two entities, at is indeed less specific than, for instance, in (which indicates that the argument is a container) or on (which indicates contact with a surface). Thus, to say of a person that he or she is in the church is more specific than to say he or she is at the church, which may also mean in the churchyard. The question is, how-ever, whether this makes at less meaningful, less of a content word, than in or on. Note also that it is not merely that at is more neutral or less specific in one or more respects than other locational prepositions. In addition, the requirement that the argument be conceptualised as a point in space places certain restrictions on its use, making it more specific in this respect. Consider the following examples:

(8) a. John is at the supermarket. b. John is in the supermarket.

As pointed out before, example (8a) does not actually require John to be inside the building, thus making its use less specific than (8b). At the same time, however, the requirement that the argument be conceived of as a point in space causes (8a) to be appropriate only when the speaker is at some distance from the supermarket — which, as a result, loses its dimensions — whereas (8b) is more likely to be uttered when the speaker is standing just outside (or even inside) the supermarket (Lee 2001: 23). The difference between the two prepositions is therefore not simply that the one is more abstract than the other, but that they have different meanings.13 As such, it is doubtful whether it is justified to analyse one of them as a lexical and the other as a grammatical element.

Finally, note that an analysis of at as the direct expression of the semantic function Loc suggests that at functions as a superordinate to all other locational

Page 11: Prepositions in Discourse

© 2008. John Benjamins Publishing CompanyAll rights reserved

English prepositions in Functional Discourse Grammar 225

prepositions, in the sense that it expresses an abstract feature shared by all other, more specific, locational prepositions. In that case, it should, of course, be possible to replace any of these more specific locational prepositions by at, thereby simply making the expression less informative. This prediction is, of course, not correct, precisely because at has a specific meaning of its own, at least one crucial feature of which (geometric point in space) is not shared by other locational prepositions.

In sum, it looks as though there is no semantic justification for classifying the five spatial prepositions at, from, via, to and towards as grammatical elements: semantically they have little in common (apart from being locational, they each have their own specific meaning); as such, there is nothing to set them apart, as a group, from other prepositions.

3.1.2 SyntaxOne of the most widely used arguments for regarding adpositions as a grammatical category is that they form a closed class (e.g. Samuelsdorff 1998). In many cases, however, this statement is weakened to allow for occasional additions. Thus Hud-dleston and Pullum (2002: 603) describe prepositions as a “relatively closed” class, while in Yule (1996: 76), too, we read that prepositions form a closed class because we “almost never” add new prepositions to the language. According to Langacker (1987: 19), on the other hand, such additions are in fact far from exceptional, and the class of adpositions is, in fact, “essentially open-ended”.14

More convincing evidence for classifying adpositions, or at least some adposi-tions, as grammatical elements may be provided by their behaviour with regard to modification. The underlying assumption is, of course, that grammatical elements, lacking lexical content, cannot be modified (e.g. Bybee et al. 1994: 7). However, at least some adpositions, so it seems, do allow for (adverbial) modification. Accord-ing to Halliday (1985: 188–189), for instance, some English prepositions can form groups by modification, as in right behind (the door), not without (some misgivings) and all along (the beach). Mackenzie (1992b) uses similar examples to support his proposal for distinguishing two classes of prepositions: a small set of grammati-cal prepositions, which do not allow for adverbial modification, and a larger class of lexical prepositions, which do. Thus, an expression like right behind the door would be analysed as in (9), where the adverb right functions as a restrictor on the prepositional predicate (and its argument), indicating “that the spatial relation holds with more than normal geometrical precision” (Mackenzie 1992b: 12):

(9) a. right behind the door a′. (f1: behindP (d1x1: f2: door)Ref : f3: rightAdv)

Mackenzie subsequently argues that his proposal correctly leads one to expect that right cannot be combined with any of the five grammatical prepositions. Admittedly,

Page 12: Prepositions in Discourse

© 2008. John Benjamins Publishing CompanyAll rights reserved

226 Evelien Keizer

it is possible to say He came right from London, but in that case, the adverb right should not, according to Mackenzie, be analysed as modifying the from-relation, but rather as a modifier indicating manner (modifying not the preposition, but the verbal predicate). In other words, in example (10a) right is analysed as a restrictor on behind me, while in (10b) it modifies the verbal predicate come:

(10) a. They came from right behind me. b. They came right from behind me.

Note, however, that in (10b) the adverb right is analysed as a restrictor not just on the preposition, but on the preposition together with its argument, an interpreta-tion which I think is correct (in most cases at least). This means that the restric-tor has, in fact, an entire PP in its scope, and that the expression right behind the door had perhaps better be represented as in (11) (notice the extra pair of brackets around behind the door):15

(11) ((f1: behindP (d1x1: f2: door)Ref): (f3: rightAdv))

If this is, indeed, the correct analysis of constructions of this kind, the presence of an adverb tells us nothing about the status of the preposition; even if the preposi-tion is believed to have a grammatical function, the PP is a lexical expression and can, as such, be modified. This may then also account for the fact that right (or similar adverbs) can, in fact, combine with such prepositions as at, towards and via as well. Consider, for instance, the following examples, where just and utterly occur in non-verbal predications headed by at:

(12) a. If you go up that road it’s just at the bottom of that <ICE-GB:S1A-071 #53:1:D>16

b. I’m not utterly at the bottom of the road <ICE-GB:S1A-020 #281:1:B>

Now, it might be argued that the adverbs in these examples function as modifiers of the entire predication rather than the PP; although the difference in interpreta-tion may be negligible, what would be modified would be the assignment of the property rather than the property itself. In other cases, however, such an explana-tion is less plausible. Consider in this respect the sentences in (13), where by far the most likely interpretation is that in which the adverbs right and only function to specify more precisely the locations at the back and at Tokyo University.

(13) a. You’ll find Miss Jardine right at the back there <ICE-GB:W2F-011 #11> b. Thank you for your letter proposing a branch of Dillons dealing with

customers’ orders only at or for Tokyo University. <ICE-GB:W1B-019 #15:2>

Similar examples can be found for the preposition toward, as shown in (14).

Page 13: Prepositions in Discourse

© 2008. John Benjamins Publishing CompanyAll rights reserved

English prepositions in Functional Discourse Grammar 227

(14) Their planned escape was cut short when they discovered that the only way back was the roadway the vehicle had come from — straight toward the buildings and the mysterious alien ship, still performing its excavation.

Another type of syntactic evidence that can be used to distinguish between lexi-cal and grammatical prepositions is that of mutual exclusivity: as pointed out by Bybee et al. (1994: 7), for instance, grammatical elements cannot co-occur with members of the same class. This is indeed one of the criteria used by Mackenzie to justify the distinction of a separate class of grammatical preposition. Thus, se-quences of prepositions like from under or to behind are analysed as combinations of a lexical preposition (under/behind) and a grammatical preposition (from/to) (Mackenzie 1992b: 11). For the sequence from under the table to behind the door Mackenzie therefore proposes the following underlying representation:

(15) a. from under the table to behind the door a′. (d1p1: f1: underP (d1x1: f2: tableN)Ref)So (d1p2: f3: behindP (d1x2: f4:

doorN)Ref)All

The analysis in (15b) leads us to expect that from can combine freely with all lexi-cal prepositions, but not with any of the other grammatical prepositions. This pre-diction does indeed seem to be borne out, since the sequences from at, from via, from to and from towards seem to be impossible.

Nevertheless, there are weaknesses in this line of argumentation. First of all, the analysis proposed would lead us to expect all grammatical prepositions to occur in similar combinations; as Mackenzie himself admits, however, such se-quences as in (15a) are basically limited to from (and occasionally to) + preposi-tion. Similarly, not all lexical prepositions can co-occur with from either. Taken together, these facts may lead one to suspect that there may be other reasons why only certain combinations are allowed — semantic reasons, for instance. One of these reasons may very well be that (spatial) from requires a locative argument (denoting typically, though not necessarily, a zero-dimensional point in space). This would explain the absence of the sequences from via, from towards and from to (where the second prepositions are analysed as grammatical elements indicat-ing Path, Allative and Approach), as well as such sequences as *?from up, *?from down, *?from across (where the second prepositions would be lexical prepositions denoting Path).17 Given the kind of relations designated by the two prepositions, it is difficult to find a plausible interpretation for these sequences.

This leaves us with the combination from at. The first question to be answered is whether this combination is indeed entirely unacceptable. A quick Google search shows that this seems not to be the case (although judgements may differ); some of the attested examples are given in (16).18

Page 14: Prepositions in Discourse

© 2008. John Benjamins Publishing CompanyAll rights reserved

228 Evelien Keizer

(16) a. Removing page headers from at the bottom and top of the page while printing the page.

(www.mail-archive.com/[email protected]/msg04647.html) b. Make a right turn from at the United Way sign into the parking lot.

(www.californiateenhealth.org/events_060104.asp) c. The Boat Crew Seamanship Manual can be downloaded from at the

Office of Auxiliary website (click on “Publications”). (www.cgaux.org/cgauxweb/public/tbaskaux.shtml)

Nevertheless, it is undeniably true that the combination from at is relatively rare, and that in many cases the presence of at may even seem superfluous. The same seems to be true for certain other combinations of from with a locative preposi-tion (from in, from on). So why is it that some combinations of prepositions are frequently used and perfectly acceptable, while others are rarely used and at best questionable? A number of factors may play a role here.19 One of these concerns the availability of a default interpretation in the given context of the PP intro-duced by the second preposition. Consider the contrast between ??from on the table, which is questionable (or at least exceptional), and from under the table, which is perfectly ok. One way to account for this difference is to assume that with an argument like table the on-relation is the default interpretation (the surface of the table being its most prominent feature). In other words, the expression from the table will be interpreted as meaning from on the table; as such, use of on would be redundant. The same would apply to the combination from in: from the build-ing will be interpreted as ‘from in(side) the building’; from his chair as ‘from in his chair’. Any other relation, however, must be coded explicitly (e.g. from behind, under, nearby the building/chair). The same explanation can be given for the rare occurrence of the combination from at: a PP like from the station, for instance, will always be interpreted as either ‘from at the station’ or ‘from in(side) the station’.

A second, related, factor that seems to play a role here is that of locational op-position: the possibility of expressing a second preposition following from seems to be determined by the availability of (inherent) contrast with some other loca-tional preposition (or, perhaps more generally, some other locational concept). Thus, the acceptability of from under the table can be accounted for by the presence of the opposition ‘on-under’; the same can be said for such combinations of from below, from behind, from inside etc. Note that this factor interacts with the avail-ability of a default reading: where, given the nature of the object(s) referred to, one of the two prepositions denotes the default relation, it is the non-default preposi-tion that is more likely to be expressed (under rather than on, behind rather than in front of etc.). Naturally, these default prepositions can also be explicitly expressed, in which case — not surprisingly — a fuller, more emphatic, form of the preposi-

Page 15: Prepositions in Discourse

© 2008. John Benjamins Publishing CompanyAll rights reserved

English prepositions in Functional Discourse Grammar 229

tion tends to be preferred (hence from inside the building, rather than from in the building; from on top of the table rather than from on the table).20

This second factor, too, helps to explain why the combination from at is rarely used, as at does not have a locational counterpart: the opposite of at seems to be ‘not-at’. Finally, such an account explains why from can quite easily be followed by at in those cases where at (or the PP introduced by at) is modified, as the presence of a modifier adds some the inherent opposition at lacks (hence from straight at home, from almost at the middle).

So let us continue by considering such expressions combining the presence of a modifier with that of a second preposition. By analysing the modifiers as having scope over the PPs as a whole, and/or by analysing all prepositions as predicates, we can also account for the following examples, where the grammatical spatial prepositions distinguished by Mackenzie co-occur with both a modifier and an-other preposition:

(17) a. Subject: HOW TO MAKE MONEY FROM RIGHT AT HOME (THOUSANDS) (http://vishnu.mth.uct.ac.za/omei/gr/BBS/Template/messages/msgs31703.html)

b. Work a long-bladed knife into the sidewall, then saw a zig-zag, wavy, or more detailed pattern, cut from almost at the center hole to right where the tire begins to curve towards the sidewall. (www.suite101.com/print_article.cfm/75/78343)

(18) a. Mostly, it’s light exiting the crown at oblique angles, as when the stone is tipped appreciably away from straight toward the viewer. (www.ganoksin.com/orchid/archive/200206/msg00067.htm)

b. But I hadn’t prepared myself for an x-ray picture, showing heavy calcium deposit from almost to the wrist to near my elbow. (www.chesapeakestyle.com/celebrate/mar04.html)

In (17) the preposition at occurs in combination with the preposition from. In be-tween the two prepositions we find an adverb, which can only be taken to modify either the following PP or, if we assume at, from, via, to and toward(s) to be lexical elements, just the second preposition. The examples in (18) show that this is also is true for the prepositions to and towards.

A last source of syntactic evidence to be considered here is the predicative use of locative adverbs, some examples of which are given in (19).

(19) a. John is in/out. b. All the shades were down. c. The kitchen is below.

Page 16: Prepositions in Discourse

© 2008. John Benjamins Publishing CompanyAll rights reserved

230 Evelien Keizer

Now clearly, there is a relation between these locative adverbs and prepositions, as in many cases we are dealing with the same lexical element. To account for this, Mackenzie (1992b: 12) introduces the following predicate formation rule, which changes the category of the input predicate from preposition to adverb, while re-ducing the number of arguments from one to zero:

(20) Locative Adverb Creation Input: ΦP (xi)Ref (e.g. below the deck) Output: ΦAdv (e.g. below)

What is interesting is that not all prepositions can serve as input to this rule; more specifically, that the grammatical prepositions distinguished by Mackenzie do not have an adverbial use (see (21a)). If we assume that locative adverbs are indeed created through a predicate formation rule, this can easily be accounted for: since these prepositions are not lexical elements, they cannot serve as input to the predi-cate formation rule. This does not explain, however, why the rule does not apply to a large majority of lexical prepositions either (on the relevant — spatial — reading; example (21b)):

(21) a. * John is at/from/via/to/towards. b. * John is under/over/above/beneath/between/against/through/beside/

among/amid/beyond.

In Section 4.2, I will come back to the use of Locative Adverbs and their relation to prepositions. I will argue that there is no need for the Locative Adverb Creation rule, for the simple reason that we are, in fact, dealing with one and the same lexi-cal element on both the prepositional and the adverbial use.

3.1.3 MorphologyIn English, prepositions do not easily combine with other elements to form complex elements; nor do they often form the basis of derivation or conversion processes. Occasionally, however, they do, in which case their morphological behaviour can be taken as an indication of their status as lexical elements.21 In this section I will first briefly consider compounds consisting of two prepositions, as well as combi-nations of prepositions with elements belonging to some other lexical category; subsequently I will consider a few cases of derivation and conversion.

Unfortunately, morphological evidence is scarce. This is partly due to the fact that the class of prepositions is relatively small since the number of spatial and temporal relations that require linguistic coding tends to be limited. As far as derivation is concerned, matters are further complicated by the fact that it is not always clear which category an element belongs to: is the over- in overact or overhang an adverb or a preposition? And what about in and out in the expression

Page 17: Prepositions in Discourse

© 2008. John Benjamins Publishing CompanyAll rights reserved

English prepositions in Functional Discourse Grammar 231

the ins and outs? More importantly, do we really have to assume different entries for such forms as over, in and out in the lexicon, one for the preposition and one for the adverb?

Nevertheless, it may be useful to look at the little evidence we have. English has a number of preposition compounds, some spelt as one word (into, onto, upon), some as two but clearly acting as one unit (out of, up to, off to). Mackenzie (1992b) offers a very elegant analysis of the obvious analogy in the compounds into and onto, which is argued to result from “an interaction between semantic function (Locative vs Allative) and lexical choice (in vs on)” (Mackenzie 1992b: 10). The analogy can, however, equally insightfully be accounted for if to, like in and on, is regarded as a predicate, in which case the complex prepositions into and onto simply result from predicate formation. This, I feel, may even be a more attractive analysis. After all, if into is derived from a combination of a lexical locative prepo-sition in and the semantic function All, this suggests that an extra step is required every time before the predicate into can be retrieved from the lexicon. This would mean that the processing of into and onto is more complex than that of compound (lexical) prepositions (e.g. upon, nearby). A more likely explanation may be to view into and onto as compounds of two lexical elements, stored in and retrievable from the lexicon. The same would hold for such fixed combinations as up to, down to, off to, on to, in from, out of, as well as for any other prepositional compounds.

Prepositions can, however, also combine with other kinds of elements. A fairly productive combination is (or was) that of the adverbs here, there and where and a preposition. Thus we have, for instance, thereby, therefore, thereafter, therein; here-by, hereinafter; but we also have thereto (formal, legal); hereto (formal, legal, dia-lectal) and whereto, as well as hitherto, and even hereat, thereat, whereat, herefrom, therefrom and wherefrom (formal and archaic/rare). Once again there is no reason to assume that the prepositional elements by, in and after in these compounds dif-fer in status from at and from: the most straightforward way to deal with these cases is to regard them all as resulting from predicate formation. The same would hold for the combinations toward(s), inward(s), outward(s), upward(s), downward(s), forward(s) and backward(s).

Prepositions also play a modest role in derivational processes. The most obvi-ous examples are those where the preposition functions as a prefix. Some of these are, in fact, quite productive, such as, for instance, over, which, can be used in combination with nouns (overcoat, overlord), verbs (overgrow, overlook),22 adjec-tives (overripe, overzealous) and adverbs (overmuch). The counterpart of over, un-der, is similarly productive (underdog, underestimate, underhand, underwater).23

Now obviously, not all prepositions can be used as prefixes. In addition to over and under, prepositions like out (quite productive with verbs), up (upgrade, update), on (on-base, on-beat), off (off-beat, off-hand) and by (with nouns, e.g. by-

Page 18: Prepositions in Discourse

© 2008. John Benjamins Publishing CompanyAll rights reserved

232 Evelien Keizer

line, bypass) can be found in this function. In Present Day English we do not, however, find at, from, via and towards as prefixes, nor for that matter most other prepositions (such as across, between, below, beneath, against, around or among). In OE and ME, on the other hand, some of these prepositions do occur as prefixes. The preposition at, for instance, was a frequent prefix in OE (meaning ‘at, close to, to’); it can also be found in some words in ME, such as at-stand(en) ‘to stand close to’, at-rech(en) ‘to reach to, get at’, at-fore(n) ‘before’ and at-hind(en) ‘behind’. Similarly, in ME, to used to occur as a prefix in combination with verbs, nouns, adjectives and adverbs (resulting in such forms as to-come ‘to happen/to arrive/to come to’, to-draught, ‘following, retinue/resort’, together and to-when ‘until what time/how long’). All these facts seem to suggest that these prepositions must be (or must have been) lexical elements which could function as the input of some predicate formation rule.

The same conclusion can be drawn on the basis of the few cases of conversion involving prepositions. Thus we have the nouns up(s) and down(s) and ins and outs, as well as the nominal forms to(s) and fro(s). It is true that in the latter cases we are dealing with fixed expressions (Mackenzie 1992b: 12), but at some point they must have been formed on the basis of the lexical elements in and out, and to and from.

Finally, it is possible, though rare, for a preposition to function as the root of a derivation. Examples are the adverb inly (‘inwardly, internally/intimately, closely, fully’) as well as the fairly recent (specialised) derived noun aboutness.

3.2 Grammatical vs. lexical use of adpositions

So far, we have been looking at arguments and evidence presented by linguists to show that all or some adpositions should be seen as grammatical or lexical. There is, however, another way of looking at the lexical/grammatical distinction, where-by this distinction applies to the use of adpositions rather than the adpositions themselves. Thus in various theoretical frameworks it is assumed that modifying PPs are introduced by lexical adpositions, whereas complements (of verbs, nouns or adjectives) are introduced by grammatical adpositions (in this context also re-ferred to as functional prepositions). Consider the examples in (22).

(22) a. the book on the table b. John’s reliance on his own ingenuity

In (22a) the PP on the table functions as a modifier of the non-relational nominal head table; in this construction the preposition on, indicating a spatial relation, is often assumed to be a lexical preposition with its own argument structure. In (22b), on the other hand, the PP on his own ingenuity is generally taken to be

Page 19: Prepositions in Discourse

© 2008. John Benjamins Publishing CompanyAll rights reserved

English prepositions in Functional Discourse Grammar 233

the complement of the relational (deverbal) head reliance; in this construction the preposition on does not indicate location and is typically analysed as a purely grammatical element, without argument structure. This is the position taken in recent work within generative grammar (see Corver and Van Riemsdijk 2001). A similar approach can be found in RRG, where a distinction is made between predicative and non-predicative uses of adpositions:

Adpositions in the periphery of the clause are always predicative, while non-pred-icative adpositions normally mark oblique core arguments. A given adposition may function either predicatively or non-predicatively, depending upon which verb it appears with; for example, from is non-predicative when it occurs with a verb like take, which licences a source argument, as in Sally took the book from the boy, whereas it is predicative with a verb like die, as in She died from malaria. (Van Valin and LaPolla 1997: 52)

One important implication of such an approach is that not only does the lexical/grammatical distinction not necessarily apply to an entire syntactic category (say adpositions), neither does it apply to specific members of a category (say, the prep-osition on) in all their occurrences; instead, the status of the adposition depends on the function of the PP it introduces (modifier or complement).

The distinction between modifiers and complements is, however, notoriously difficult to make. This is particularly true for modifier and complement PPs within the noun phrase, which appear in the same form and position. Various attempts have been made to draw up lists of defining features, semantic and syntactic, to distinguish between the two types of PP. The syntactic criteria suggested include restrictions on the number of complements (but not modifiers), coordination (possible between PPs with the same function, but not between complements and modifiers), use of the pronoun one (combinable with modifiers but not with com-plements), and differences with regard to mobility within the clause (see e.g. Hud-dleston 1984; Radford 1988; Huddleston and Pullum 2002). As shown in Keizer (2004, 2007), however, what are often given as rules are at best tendencies, and as such make very unreliable tests.

Nor is it altogether clear to what extent the prepositions introducing comple-ments and modifiers differ semantically. One of the semantic tests suggested to dis-tinguish between the two types of preposition concerns their selection. According to Huddleston and Pullum (2002: 440), for instance, complements must be li-censed by the head noun; i.e. in the case of a complement PP, the head noun de-termines the choice of preposition (or the limited range of permitted choices). The default preposition is of, and this is often the only possibility, as in the King of France. In some cases, particularly with deverbal nouns, some other preposition is selected (as in reliance on, collaboration with).

Page 20: Prepositions in Discourse

© 2008. John Benjamins Publishing CompanyAll rights reserved

234 Evelien Keizer

Although it is certainly true that in the case of complements the choice of preposition is generally speaking more restricted, this choice can hardly be used as a criterion for complement status of the PP. For one thing, we have seen that one and the same preposition can be used to introduce both complements and modifi-ers. We can have the king of France, where of is claimed to introduce a complement, as well as a king of wood or a king of considerable intelligence, where it introduces PP-modifiers; we can have the book on physics as well as the book on the table. Ac-cording to Huddleston (1984: 264), however, we can still tell which prepositions are grammatical and which are lexical, since lexical prepositions have their own (typically locative) meaning, whereas grammatical prepositions, lacking seman-tic content, are determined by the head noun (see also Huddleston and Pullum, 2002: 653ff). Now, this might be argued to apply to a preposition like on in the book on physics and the book on the table: clearly the meaning of on is locative in the latter construction, but not in the former. In many cases, however, the distinction between lexical and grammatical prepositions is not so straightforward. This is first of all due to the fact that it is not always possible to establish the lexical mean-ing of a preposition. The best example is, presumably, the preposition of, which at an early stage lost its original locative meaning (close to away/from, indicating source) and has since developed a host of meanings, including geographical ori-gin, possession, part-of, attribute-of, type-of, quantity-of and depiction-of (Hud-dleston and Pullum 2002: 658; see also 477; Lindstromberg 1998: 195). Thus, in its present-day use, of does not seem to have a literal (spatial) meaning; nor is it at all clear which of its uses is to be regarded as primary. Nevertheless, it is generally acknowledged that of can introduce both complements and modifiers. In the ex-amples in (23), for instance, the of-PPs are usually regarded as complements; those in (24), on the other hand, are generally taken to function as modifiers.

(23) a. the sister of Mary b. the house of her former husband c. the spire of the cathedral d. a glass of water e. the death of the emperor f. the conquest of Persia

(24) a. the wines of France b. a man of honour c. a girl of a sunny disposition d a boy of sixteen e. a frame of steel f. a matter of no importance

Page 21: Prepositions in Discourse

© 2008. John Benjamins Publishing CompanyAll rights reserved

English prepositions in Functional Discourse Grammar 235

In (24), Huddleston and Pullum (2002: 659) argue, the preposition of is not select-ed by the head and thus makes an independent contribution to the meaning. This claim may be justified for example (24a), where of, indicating geographical origin, comes closest to having its original locative meaning. It may also be justifiable in (24b), where, as pointed out by Huddleston (1984: 264), of can be paraphrased by means of the (lexical) preposition with and alternates with without (a man of/with-out honour). In examples (24c–f), however, it is much more difficult to maintain that of contributes independently to the meaning of the construction. In all these cases the lexical meaning of of seems to be much more elusive than, for instance, its meaning in examples (23b&c), where the semantic contribution of the preposi-tion (possession, part-of) is much more tangible.

In other cases, the claim that the prepositions used to introduce complements lack semantic content is even more doubtful. Huddleston and Pullum (2002: 661), for instance, acknowledge that on the basis of this criterion it is not always easy to draw the distinction between with-complements and with-modifiers. The primary meanings of with are generally assumed to be non-specific proximity (Lindstrom-berg 1998: 208) and accompaniment (Huddleston and Pullum 2002: 661). When used to express this meaning, with is usually taken to introduce PP-modifiers (a house with a garden, a mother with her child). Other closely related and very com-mon uses of with involve cooperation (my work with John; collaboration with other countries), while, as pointed out by Lindstromberg (1998: 212), with has a similar meaning when it accompanies communication and participation words, such as communication, discussion, consultation, conspiracy, talk, negotiation and partici-pation. In all these cases with clearly has a meaning related to the primary sense. As such, there seems to be no justification for classifying these uses of with as functional. Nevertheless, the PPs they introduce are typically analysed as comple-ments.

The opposite is also true: modifiers are not always introduced by lexical prepo-sitions. This, we have seen, is true of the preposition of, the meaning of which proved to be difficult to establish in the first place. Another case in point is the preposition by. Its primary sense is a locative one, indicating proximity in the hori-zontal plane (Lindstromberg 1998: 141; Huddleston and Pullum 2002: 655), which we do, indeed, find in PP-modifiers (example (25a)), though most uses of by when used to introduce PP-modifiers are metaphorical (e.g. example (25b)). When used to mark the Agent (example (25c)), by is generally regarded as functional, despite the fact that here, too, the PP it introduces is a modifier. (25) a. The belt smashed into the table by his side and his whole body flinched.

<ICE-GB:W2F-001 #135:1>

Page 22: Prepositions in Discourse

© 2008. John Benjamins Publishing CompanyAll rights reserved

236 Evelien Keizer

b. Douglas Hurd a quiet man by nature a diplomat by training couldn’t resist pointing out on Guy Fawkes Day that he’d had enough of the explosions of the past few days <ICE-GB:S2B-007 #16:1:C>

c. We had a lecture by that guy Rene Weis over there <ICE-GB:S1A-006 #20:1:B>

In sum, distinguishing between lexical and functional prepositions or uses of prepositions on the basis of their (lack of) semantic content is problematic be-cause the notion of semantic content is obviously graded. Neither will it do simply to say that complements are introduced by functional prepositions and modifiers by lexical prepositions, first of all because the distinction between complements and modifiers is similarly difficult to make (none of the tests suggested in the lit-erature turns out to be reliable; see e.g. Keizer 2004, 2007), and secondly because, as we have seen, there is no reason to assume a one-to-one relationship between complement/modifier status and the distinction between functional and lexical (uses of) prepositions.

4. English prepositions in Functional Discourse Grammar

4.1 Relevant features of Functional Discourse Grammar

4.1.1 A general outline of the modelNow that we have discussed several current approaches and reviewed the evidence, it is time to consider the possibilities offered by Functional Discourse Grammar (henceforth FDG) for a more unified and consistent treatment of adpositions and adpositional phrases. Before we can do so, however, a brief summary of the rel-evant features of the model may be required.

FDG (Hengeveld 2004, 2005; Hengeveld and Mackenzie 2006, 2008) is a func-tional-typological grammar model that is intended as the successor to FG (Dik 1997a,b). Although it shares many of the basic assumptions and general features of FG, it also deviates from FG in a number of important respects. One crucial difference concerns the fact that, unlike FG, FDG has a top-down organisation: the starting point of analysis is the Speaker’s intention, which, via the ordered operations of (pragmatic and semantic) formulation and (morphosyntactic and phonological) encoding, is gradually prepared for articulation. The operations of formulation and encoding are FDG’s main concern; they are part of the cen-tral component of the model, the grammatical component. This grammatical component does not, however, operate in isolation, but interacts with three fur-ther components: a conceptual component, containing information concerning the intention of the Speaker; a contextual component, containing linguistic and

Page 23: Prepositions in Discourse

© 2008. John Benjamins Publishing CompanyAll rights reserved

English prepositions in Functional Discourse Grammar 237

non-linguistic contextual information (including shared speaker-hearer long-term knowledge); and an output component, where phonetic, graphological, or signed articulation takes place (see Fig. 1).

Another feature that distinguishes FDG from FG is the internal organisation of the grammatical component, where four modules interact to produce the ap-propriate linguistic forms. Each module deals with a different level of linguistic or-ganisation. At the top of the model we find the pragmatic and semantic modules, rendering the Interpersonal and Representational levels of analysis, respectively. Both these levels are the output of the process of formulation. Information from these levels then enters the phase of encoding: first that of morphosyntactic en-coding, yielding a Morphosyntactic level, and finally that of phonological encod-ing, yielding a representation at the Phonological level.

Finally, these four levels not only receive their input from the other compo-nents and levels, but also from further information contained within the gram-matical component; i.e. from the language user’s knowledge of his/her language. This information, which takes the form of primitives, comes in three types. Firstly, there is information about the patterns available in the language; at the Interper-sonal and Representational levels, this information takes the form of frames (in-dicating acceptable valency structures), at the Morphosyntactic level of templates (yielding the possible word order patterns), and at the phonological level of pro-sodic patterns. Secondly, there is information about the items available: lexemes for formulation, auxiliaries for morphosyntactic encoding and (bound) mor-phemes for phonological encoding. Thirdly, each level has its own set of operators and functions, i.e. abstract triggers of grammatical processes.

For the purposes of this article, we only need to consider two levels of analysis: the Interpersonal level and, in particular, the Representational level. The Interper-sonal level deals with all the formal aspects of a linguistic unit that reflect its role in the interaction between Speaker and Addressee. In keeping with the overall architecture of FDG, the units of discourse relevant at this level are hierarchically organised. A simplified representation is given in (26):

(26) (MI: [(AI: [(FI: ILL)Φ (PI)Φ (PJ)Φ (CI: [(TI)Φ (RI)Φ … ])])])

At the highest level in this hierarchy we find the Move, which describes the en-tire segment of discourse relevant at this level. The Move consists of one or more (temporally ordered) Acts (AI, AJ), which together form its (complex) Head. Each Act in turn consists of an Illocution (FI), the Speech Participants (PI and PJ) and a Communicated Content (CI). Finally, within the Communicated Content, one or more Subacts of Reference (RI) and Ascription (TI) are ‘evoked’ by the Speaker.

Page 24: Prepositions in Discourse

© 2008. John Benjamins Publishing CompanyAll rights reserved

238 Evelien Keizer

The Representational level deals with the semantic aspects of a linguistic unit; it is the level of ‘denotation’, where the appropriate lexemes are chosen to describe entities in some (real or imagined) non-linguistic world. Since these entities are of different orders, the linguistic units at this level differ with respect to the ontologi-cal category they denote. The hierarchical structure of the Representational level is as follows:

(27) (pi: [(epi: [ (ei: [(fi)) (xi)ϕ (li)ϕ (ti)ϕ …])])])

Conceptual component

FormulationFramesLexemesInterpersonal andrepresentationaloperators

TemplatesGrammaticalmorphemesMorphosyntacticoperators

TemplatesSuppletive formsPhonologicaloperators

contextual

component

Interpersonal Level

Representational Level

Morphosyntactic Encoding

Morphosyntactic level

Phonological Encoding

Phonological level

Output component

Figure 1. FDG: general layout

Page 25: Prepositions in Discourse

© 2008. John Benjamins Publishing CompanyAll rights reserved

English prepositions in Functional Discourse Grammar 239

The most straightforward type of entity is the individual, a first-order entity (xi). Entities of this type are concrete, can be seen and touched, can be located in time and space, and can be evaluated in terms of their existence. State-of-affairs (SoAs) are second-order entities (ei); they can be located in space and time, and evalu-ated in terms of their reality. Third-order entities, or propositions, are mental con-structs (pi). They cannot be located in time or space; they can be evaluated in terms of their truth. In addition to these three orders of entities (already distinguished by Lyons 1977: 442–447), several more semantic categories can be distinguished. Thus, there are separate units denoting location (li) and time (ti); since the con-cepts of time and space cannot be reduced to any of the other entity types, but in-stead rather specify dimensions of those entity types, they are seen as constituting independent semantic categories (cf. Mackenzie 1992a for location and Olbertz 1998 for time). At the lowest level of the hierarchy, the entity denoted is a property or relation (fi). Properties and relations do not have independent existence but can only be evaluated in terms of their applicability. Finally, in between the proposi-tion and the SoA we find another special semantic category, the episode (epi). Episodes represent a combination of units of another semantic category; they may be defined as a semantically coherent set of SoAs.

The primitives selected at the Representational level include the lexemes (pred-icates) used to describe the entity denoted and the appropriate frames, specifying the number of arguments and their roles in the SoA. These roles are represented by means of a restricted number of semantic functions (indicated by φ in (27), e.g., Actor, Undergoer, Location).

4.1.2 The representation of adpositionsAt the Interpersonal level, the adpositional construction as a whole is analysed either as a subact of reference (as in I put the book on the table, where on the table is interpreted as ‘referring to’ a place) or as a subact of ascription (when used predicatively, as in The book is on the table). Note that by separating the referential and ascriptive functions of adpositional phrases from their denotation, and by analysing adpositional phrases as denoting places and times rather than objects, one of the problems of the standard FG account has been solved: since adposi-tional phrases are no longer analysed as terms referring to first order entities, they can be used either in a referential or in an ascriptive function — as such the rule of term-predicate formation is no longer required.

The crucial level of analysis for adpositions and adpositional phrases is, how-ever, the Representational level: first, because it is at this level that a distinction is made between different types of entity denoted (e.g., events, propositions, ob-jects or locations); secondly, and even more importantly, because it is here that the distinction between predicates and semantic functions is relevant. What we

Page 26: Prepositions in Discourse

© 2008. John Benjamins Publishing CompanyAll rights reserved

240 Evelien Keizer

know is that adpositional constructions are analysed as units denoting location or time, symbolised by the variables ‘l’ and ‘t’, respectively (Hengeveld and Mackenzie 2006). What still needs to be established, however, is whether to analyse adposi-tions (all or some) as predicates restricting these variables, as in (28a), or as seman-tic functions assigned to a term headed by a nominal predicate, as in (28a′):24

(28) a. (1l1: atP (1l2: (stationN))Ref) a′. (1l1: stationN)Loc

In addition, we need to consider the question of which frames are needed for the coding of adpositional constructions at the Representational level.

4.2 The proposal

4.2.1 Lexical prepositions and grammatical useThe proposal for the treatment of adpositions and adpositional phrases to follow will start from the assumption that all English prepositions are lexical elements, to be represented in the lexicon with a meaning definition. Such a treatment does justice to the fact that all adpositions, in most of their uses, have some measure of semantic content (Section 3.1.1), as well as to the fact that adpositions share a large number of syntactic and morphological properties with nouns, verbs and adjectives (Sections 3.1.2 and 3.1.3). In addition, it will be argued that a very lim-ited number of adpositions can, in some of their uses, more plausibly be regarded as grammatical elements, marking rather than describing the relation between a head and some argument or modifier. In English the only prepositions to be ana-lysed this way are of and by, and only in those cases where they are used to indicate the semantic functions of Agent (or Positioner, Force, Processed or Ø) or Patient; i.e. the semantic functions typically assigned to the first or second argument of a verbal predicate.

This means, first of all, that adpositional phrases are analysed as expressions with an adpositional first restrictor. Their prototypical denotation is a place, but PPs may also denote a time or some other abstract entity. They typically function as modifiers within a clause (e.g. I bought the book in Paris) or NP (e.g. the book on the table). If, however, their presence is required to complete the meaning of a verbal, nominal or adpositional first restrictor (head), they are to be analysed as arguments (e.g. of verbal predicate like put, a nominal predicate like father, or some other preposition, typically from).

However, before we turn to the representation of lexical adpositions, let us first explore in some more detail the idea that some adpositions can be used gram-matically, and that in English this use is limited to the prepositions of and by when used to introduce terms with the semantic function of Agent, Positioner, Force,

Page 27: Prepositions in Discourse

© 2008. John Benjamins Publishing CompanyAll rights reserved

English prepositions in Functional Discourse Grammar 241

Processed, Ø or Patient (i.e. the semantic functions that can be assigned to the first and second arguments of verbal and adjectival predicates; see Dik 1997a: 118). Consider, for instance, the use of the preposition of in NPs with a deverbal nomi-nal predicate. Some simple examples are given in (29):

(29) a. the fall of the city b. the destruction of the city

Following Dik (1989, 1997a; see also Dik 1985a, 1985b), I will regard the nominal predicates fall and destruction in these examples as derived from the correspond-ing verbal predicates to fall and to destroy, respectively, through a rule of predi-cate formation affecting the syntactic category and form of the derived predicate. According to this rule, both the valency of the input predicate and the semantic functions of the arguments are preserved in the process. The predicates fall and destruction in (29) are formed as follows:

(30) Verbal Noun Formation: a. Input: fallV (xi)Proc Output: fallN (xi)Proc b. Input: destroyV (xi)Ag (xj)Go Output: destructionN (xi)Ag (xj)Go

However, the rule of verbal noun formation does not only affect the form of the in-put predicate, but also that of the arguments. This is accounted for by the Principle of Formal Adjustment (Dik 1997b: 157–158):

(31) Principle of Formal Adjustment (PFA):25

Derived secondary constructions of type X are under pressure to adjust their formal expression to the prototypical expression model of non-derived, primary constructions of type X.

Modifiers within the term typically appear in the form of adjectives, genitives, adpositions and relative clauses. The PFA accounts for the fact that the Agent and Patient terms of a transitive input predicate (e.g. example (32a)) can appear as of- or by-PPs, or as genitival modifiers in the term (e.g. examples (32a′-a′′′)):

(32) a. Caesar destroyed the city. a′. the destruction of the city by Caesar a′′. Caesar’s destruction of the city a′′′. the city’s destruction by Caesar

In other words, the grammatical use of adpositions is restricted to those cases where the adposition expresses a semantic function which in the corresponding verbal construction is not expressed by means of an adposition but by some (other) gram-

Page 28: Prepositions in Discourse

© 2008. John Benjamins Publishing CompanyAll rights reserved

242 Evelien Keizer

matical means, such as word order, agreement or case. It seems therefore reasonable to assume that when they appear in the derived construction the prepositions of and by do not have semantic content, but are indeed simply expressions of the semantic relations in question. This approach is further corroborated by the fact that it is only these semantic functions (i.e. Agent, Positioner, Force, Processed, Ø or Patient in FG; now Actor and Undergoer) which in the nominal domain can be expressed by the likewise grammatical means of a genitive (e.g. examples (32a′′& a′′′)).

Nor does the fact that in these cases the PPs are represented as terms form a problem. Unlike in the case of locative PPs, the semantic functions of Agent, Pa-tient etc. do apply to the term itself. Thus, whereas in a construction like the mouse under the table, the term the table does not itself refer to the location, but rather as the object in relation to which the location of the mouse is defined, in an expression like the destruction of the city by Caesar, the terms the city and Caesar do refer to the Patient (Undergoer) and Agent (Actor), respectively, of the predication denoted.

Observe furthermore that the current proposal is in line with the observa-tions made earlier about the lack of a one-to-one relationship between the distinc-tion between complements and modifiers and the distinction between lexical and grammatical adpositions. Thus, as pointed out before, of does not always have a grammatical function: when used in combination with a non-derived head (as in the house of my parents, a man of honour, the spire of the cathedral, a girl of eighteen, etc.), the preposition of will be analysed as a lexical element, indicating an “intrin-sic relation” between the head and the argument of the preposition. Similarly, by can be used both grammatically and lexically; in both cases, however, it typically introduces modifiers.

Finally, we need to account for the fact that in most cases prepositions require an NP-complement (we cannot say *I saw the book on), but that many of them also can stand by themselves (e.g. John is in). With nominal heads, the opposite is true: while in most cases omission of the PP is, syntactically and semantically at least, unproblematic (we can refer to the book as well as to the book on the table), in other cases the PP is more or less required to complete the meaning of the head (we do not, as a rule, refer to a destruction). Note, once more, that this is not directly relat-ed to the lexical or grammatical status of the preposition (Sections 3.1.2 and 3.2).

Clearly what we need is a flexible system, one which allows us to use one and the same (adpositional or nominal) predicate with and without a complement. FDG seems to allow for the required degree of flexibility: unlike in FG, where predicates were stored in the lexicon along with a predicate frame, in the new model predicates and frames are stored separately. By allowing predicates to com-bine with more than one frame, the system becomes both more flexible and more efficient. What follows is a more detailed account of the some of the frames needed and the ways in which they can be combined.

Page 29: Prepositions in Discourse

© 2008. John Benjamins Publishing CompanyAll rights reserved

English prepositions in Functional Discourse Grammar 243

4.2.2 Frames in FDG4.2.2.1 Preposition frames. For the purposes of the present discussion, it will be argued that two prepositional frames need to be distinguished:

a. Type 1: a frame for PPs headed by a transitive preposition (i.e. a one-place prepositional predicate);

b. Type 2: a frame for PPs headed by an intransitive preposition (i.e. an avalent prepositional predicate).

In combination with the various frames for nominal (and verbal) constructions (some of which will be discussed below), these frames can account for the different constructions and combinations prepositions and prepositional phrases appear in.

Type 1: (33) PredP (αi)Ref

26

In this frame the preposition has its prototypical (relational) function: it indicates a relation between some entity or SoA and the argument of the preposition. Every preposition can be inserted in this frame. The argument is always assigned the se-mantic function Ref (as the entity with reference to which the relation indicated by the preposition is asserted to hold) and may be of any entity type (i.e. the preposi-tion phrase as a whole will denote a place, but the argument may denote any type of entity, e.g. an event (during the party) or a proposition (in this proposal)). This is the frame which is selected for PPs functioning as PP arguments or modifiers within the term. Some examples are given in (34):

(34) a. the book on the table a′. (l1: onP (1x1: tableN)Ref) b. the noise from under the table b′. (l1: fromP (l2: underP (1x1: tableN)Ref)Ref)

Type 2: (35) PredP

Here the preposition is used without an argument; this is the frame which is select-ed in the case of so-called prepositional adverbs. Consider the examples in (36):

(36) a. John is in. b. All the shades are down. c. The kitchen is below.

These constructions will be analysed as non-verbal predications the non-verbal predicates of which consist of an intransitive preposition only (cf. Lee 1998). In each case the argument is, to some extent, inferrable: in (36a) John is in the same

Page 30: Prepositions in Discourse

© 2008. John Benjamins Publishing CompanyAll rights reserved

244 Evelien Keizer

building as the speaker; in (36b) the shades are down in relation the vertical dimen-sion of the window/room; and in (36c) the kitchen is below either in relation to the position of the speaker, or in relation to the vertical dimension of the building.

The proposal has a number of advantages. First, it treats prepositions in the same way as any other lexical category in allowing for a transitive (relational) and an intransitive (non-relation) use. Secondly, by regarding the elements in, on and below here as prepositions rather than adverbs, it accounts for the fact that in many cases one and the same element is used in all cases. Recall that in Mackenzie’s (1992b: 12) proposal this parallelism was accounted for by the introduction of a predicate formation rule converting prepositions into adverbs. In the present pro-posal there is no need for such a rule: rather than assuming that in constructions such as (36) valency reduction has taken place, it will now be assumed that these constructions result from a combination of the prepositional element in question and the proper (intransitive) frame. Since prepositions are regarded as lexical el-ements, their functioning as non-verbal predicates poses no problem.27 Nor, as pointed out before, do we need a separate rule of Locative Adverb Creation.

Finally, as mentioned before, not all prepositions allow the use of this frame (e.g. *John is at/from, *The journey is via, *The mouse is under); and even if they do, not all of them can occur in all syntactic environments. Thus, unlike PPs of Type 1, the intransitive frame is hardly ever selected for modification or complementa-tion within the term (i.e. we do not have *the book on, *the mouse in or *the shades down; note, however, that the kitchen below and (on a temporal use) the day after are fine). As to why it is that not all prepositions can be used in this way, one can only speculate; as pointed out before, there is little reason to assume that this has anything to do with the status of the element as grammatical or lexical.

4.2.2.2 Noun frames. Let us finally consider how the approach advocated would work in the case of N + PP constructions. As it will not be possible to discuss all possible combinations of nouns and prepositional phrases, discussion will be con-fined to three common configurations, whose treatment may serve as a basis for other, more complex, combinations.

(i) Type 1: a frame for NPs headed by non-relational nominal predicates and modified by a prepositional phrase;

(ii) Type 2: a frame for NPs headed by relational (typically deverbal) nouns with one (inherited) PP argument; the preposition is selected by the head (inher-ited from the verb along with the argument);

(iii) Type 3: a frame for NPs headed by relational (typically deverbal) nouns fol-lowed by arguments introduced by of or by; these prepositions (in this use at

Page 31: Prepositions in Discourse

© 2008. John Benjamins Publishing CompanyAll rights reserved

English prepositions in Functional Discourse Grammar 245

least) are grammaticalised and as such expressions of semantic functions (Ac-tor / Undergoer).

Type 1: (37) PredN: (li: PredP (αi)Ref)

In (37) a non-relational nominal head is modified by a PP functioning as a (non-first) restrictor. As illustrated in (38a′), this PP-modifier is headed by a preposi-tional predicate which itself takes an argument (typically a term, but it may also be a PP or adverb; see e.g. Lee 1998: 140); this argument is assigned the semantic function Ref. The form of the prepositional predicate depends on the relation (loc-ative, temporal, or otherwise) between the head and the argument of the preposi-tion. Although the head may, in some cases, place (semantic) restrictions on the choice of preposition, the preposition is not selected by the head.

(38) a. the book on the table a′. (1x1: [bookN: (l1: onP (1x2: tableN)Ref)])

In constructions such as (38a), the preposition is represented as a lexical element. As such, it is assumed to have semantic content: it functions to indicate a relation (locative, temporal, or otherwise) between two entities. Note that the preposition is not necessarily used in its primary (literal) sense. Which interpretation is given to the preposition depends on the nature of the two entities; thus in a construction like the class on Tuesday the preposition will be given a temporal interpretation on account of the predicational head class (an event) and the temporal argument Tuesday.

Since the prepositions in these constructions are analysed as predicates, we can easily account for the fact that they can be modified, irrespective of whether the adverbs in question are analysed as modifying just the prepositional predicate or the PP as a whole. Opting for the latter solution (see discussion of example (10) above), we can (roughly) represent the NP in (39a) as in (39a′):

(39) a. the girl right at the back (cf. (13a)) a′. (1x1: [girlN: [(l1: atP (1x2: back)Ref): (rightAdv)]])

Similarly, co-occurrence with other prepositions, whatever the combination, is un-problematic: the first preposition simply takes the following PP as its argument. As can be seen in (40), this means that such constructions contain two Ref arguments: one a regular term functioning as the argument of the second preposition, the other a prepositional phrase functioning as the argument of the first preposition:

(40) a. the noise from behind the door a′. (1e1: [noiseN: [l1: fromP (l2: behindP (1x1: doorN)Ref)Ref]])

Page 32: Prepositions in Discourse

© 2008. John Benjamins Publishing CompanyAll rights reserved

246 Evelien Keizer

Moreover, such an approach can deal effortlessly with such examples as (41), which involve both modification and co-occurrence of prepositions:

(41) a. a close-up view from right at the edge (www.philaprintshop.com/niagara.html) a′. (1e1: [viewN: close-upA: [(l1: fromP ((l2: atP (1x1: edgeN)Ref): (rightAdv))

Ref)]]) b. heavy calcium deposit from almost to the wrist to near my elbow. (cf.

(18b)) b′. (1x1: [calcium depositN: heavyA: [(l1: fromP ((l2: toP (1x2: wrist)Ref):

(almostAdv))Ref) (l1: toP (l2: nearP (1x2: elbow)Ref)Ref)]])

Finally, the fact that prepositions can combine with other prepositions, as well as with other types of predicate, to form derived predicates (compounds as well as derivations) can also easily be accounted for: since all prepositions are predicates stored in the lexicon, they are available as input for predicate formation rules.

Type 2: (42) PredN (fi: PredP (αj)Ref)Ref

In Type-2 constructions, the PP functions not as a restrictor, but as a preposi-tional argument within the term. A possible instantiation of this frame would be the term in (43a), which would be given the underlying structure given in (43a′). Here the nominal (deverbal) first restrictor reliance is analysed as taking an PP argument, which it will be assumed to have inherited from the verbal predicate rely. This PP argument (on his ingenuity) will be assigned the semantic function Ref. It is headed by the prepositional predicate on, which, in turn, takes an argu-ment, here in the form of a term (his ingenuity); this argument, too, is assigned the semantic function Ref.

(43) a. his reliance on his ingenuity a′. (1e1 : [(relianceN (l1: onP (f1: [ingenuity: his])Ref)Ref): his])

A final difference between Type-1 and Type-2 constructions is that in the latter the nominal predicate selects (or has a strong preference for) a particular preposition as the head of the prepositional argument (e.g. reliance on, appeal for, reference to, interest in, conversation with, protest against). Although in these constructions the preposition can still have its primary, spatial meaning (e.g. their emergence from the room), in most cases the preposition is not used in its primary sense. Never-theless, it is still regarded as a lexical element (i.e. as a metaphorical extension, but still involving a spatial aspect). As in the case of non-prototypical uses of preposi-tions introducing modifiers (e.g. the class on Tuesday), the exact interpretation is triggered by the context: in an example like (43a) it is the co-occurrence of the

Page 33: Prepositions in Discourse

© 2008. John Benjamins Publishing CompanyAll rights reserved

English prepositions in Functional Discourse Grammar 247

preposition on with the abstract noun ingenuity which leads to the correct — non-primary — interpretation of on.

Syntactic evidence for analysing the prepositions in these constructions as predicates is scarce. Semantically, however, an analysis of these prepositions as predicates is far more plausible. First of all, we have seen that they are not meaning-less elements; quite often the relation denoted is direct linked to the primary sense of the preposition. At the same time it will be clear that where the prepositions express more abstract relations, it will be difficult (and cumbersome) to have all these relations represented by different semantic functions.

Observe finally that from a purely formal point of view, Type-1 and Type-2 frames yield identical constructions: in both cases the construction takes the form of a nominal head and one or more PPs. Nevertheless, different analyses are called for on account of the different status of the PPs. Although it is true that the criteria for distinguishing between argument (complements) and modifiers are far from clear-cut, the distinction is nevertheless a valid one, since the syntactic behaviour of (typical) modifiers differs significantly from that of (typical) arguments. Despite the many in-between cases, this difference therefore ought to be reflected in un-derlying representation.

Type 3: (44) a. PredN (αi)A/U intransitive input verb b. PredN (αi)U transitive input verb

In constructions of this type we are once again dealing with derived nominal pred-icates which function as one-place predicates. Where the input verb is an intran-sitive verb, the only argument of the verb (Actor, Undergoer) is inherited by the derived nominal predicate. The frame for such a construction is given in (44a); an example can be found in (45a).

(45) a. the arrival of the train a′. (1e1 : [arrivalN (1x1: train)U])

In the case of a transitive input predicate, it will be assumed that only the second argument of the verb (the Undergoer) is inherited; any other arguments of the verb (e.g. Actor, Recipient) will be regarded as (implied) modifiers in the derived nomi-nal structure (for a discussion, see Mackenzie 1996). The basic frames for con-structions of this kind is given in (44a); an instantiation of this frame can be found in (46a). It is, of course, possible, to express both the Undergoer and the Actor. In such constructions, the (optional) Actor PP will be given the status of a restrictor; its semantic function — inherited from the frame of the input verbal predicate — triggers the use of the grammatical element by. An example is given in (46b).

Page 34: Prepositions in Discourse

© 2008. John Benjamins Publishing CompanyAll rights reserved

248 Evelien Keizer

(46) a. the treatment of the patient a′. (1e1: [treatmentN (1x1: patient)U]) b. the treatment of the patient by the doctor b′. (1e1: [treatmentN (1x1: patient)U]: (x2: doctor)A)

The reasons for assigning different underlying representations to these PPs will by now be clear. As pointed out before, the argument-modifier distinction is based on differences in syntactic behaviour (concerning its position in and outside of the term, coordination, anaphora etc.) of the PPs in question. Only the first argument of nominal predicates derived from one-place verbal predicates, and the second argument of nominal predicates derived from two-place verbal predicates behave consistently as arguments of the derived nominal head. The status of the prepo-sition (lexical or grammatical) is not directly related to the argument-modifier distinction. Instead, grammatical status is confined to those expressions which appear as terms in the corresponding verbal constructions, i.e. the first and second arguments. Within the nominal domain these arguments take the form of PPs. However, since they fulfil the same function with regard to the head, it will be as-sumed that the prepositions in question (by and of) do not denote this function; instead they will be taken to express the (inherited) semantic functions Actor and Undergoer, their presence the result of the PFA.28 Analysing these expressions as terms is further justified by the fact that, unlike in the case of their locative coun-terparts, it is the entities referred to by these terms that function as the Actor, Undergoer etc.

5 Conclusion

In this article I have tried to show the need for a treatment of English prepositions and prepositional phrases that differs radically from the way these are treated in the standard FG approach, and which is also crucially different from any of the alternatives offered since then. I started by identifying the various problems and reviewing all the evidence and arguments used in previous proposals as well as ad-ditional evidence from semantics, syntax and morphology. Making use of the new possibilities offered by the model of FDG, I have subsequently proposed an analy-sis which, as I hope to have demonstrated, can remedy most of the shortcomings of previous attempts, while offering a unified and consistent treatment of preposi-tions and prepositional phrases. The most important features of this proposal are:

– Prepositions are lexical elements; they are predicates stored in the lexicon, provided with a meaning definition.

Page 35: Prepositions in Discourse

© 2008. John Benjamins Publishing CompanyAll rights reserved

English prepositions in Functional Discourse Grammar 249

– Prepositions can be transitive (prototypical use) or intransitive. This means we have two frames for prepositions, one in which the preposition functions as a one-place predicate and one in which it is avalent. In this way, prepositions are no different from the other syntactic categories (N, V and A).

– The arguments of prepositions are typically terms; they may, however, also take the form of prepositions or prepositional phrases, adverbs or clauses.

– Some prepositions (English by and of) allow for a grammatical use, their presence being a result of PFA. This use is restricted to those cases where the preposition heads a PP which functions as the inherited argument of a dever-bal/deadjectival nominal head with the semantic functions Actor or Undergo-eror. In those cases, the PPs are analysed as terms with semantic expressions, and the prepositions are grammatical elements expressing these semantic functions.

– Elements traditionally categorised as adverbs (in constructions like John is in) are simply to be regarded as instances of intransitive prepositions.

The current proposal can account for many of the specific properties of preposi-tions and prepositional phrases. The fact, for instance, that prepositional phrases can be used both referentially (as subjects or complements) and ascriptively (as modifiers or non-verbal predicates) can easily be accounted for. Similarly, the fact that prepositions and prepositional phrases can be modified is no longer a prob-lem; the same is true for the occurrence of sequences of two or more prepositions. In addition, the proposal can deal with those instances where the NP-arguments of prepositions are not expressed; the same is true for unexpressed PP-arguments of relational nominal predicates. Moreover, the proposal offers an explanation for the fact that many prepositions are identical in form to what were traditionally regarded as adverbs.

Finally, it has been shown that the analysis proposed provides a unified treat-ment for a number of constructions containing prepositions and prepositional phrases, including constructions with ‘locative adverbs’ (by distinguishing a frame for the intransitive use of prepositions) and nominal constructions with PP argu-ments and modifiers (by means of different frames for NPs headed by relational and non-relational nominal predicates).

Received 22 August 2006. Revised version 21 April 2007.Notes

1. I would like to thank Lachlan Mackenzie, as well as the two anonymous reviewers, for in-sightful comments on earlier versions of this article. All remaining errors and inconsistencies are, of course, entirely my own.

Page 36: Prepositions in Discourse

© 2008. John Benjamins Publishing CompanyAll rights reserved

250 Evelien Keizer

2. It will be clear that for any discussion of the lexical/grammatical status of an element, one needs to know what exactly is meant by these terms, and on the basis of which criteria the dis-tinction can be made. In this article, a number of semantic, syntactic and morphological criteria will be applied to prepositions to justify their qualification as grammatical or lexical. For a more general discussion of the lexical-grammatical dichotomy in FDG, see Keizer (2007).

3. For an illustration of how such an approach might apply to English temporal modifiers (in-cluding prepositional phrases), see Connolly (1994, 1995).

4. For discussion of this rule, see also Mackenzie and Hannay (1982: 47) and Mackenzie (1983).

5. It is, of course, possible to refer to places, as in Under the table is a good place for a mouse to hide. In that case, however, the PP functions not as a modifier, but as a participant in the SoA (with the semantic function 0).

6. Note that in other approaches (e.g. Zeller 2001; also the treatment of predicative/adjunct prepositions in RRG) prepositions are often regarded as two-place predicates, indicating a rela-tion between two arguments. Like Weigand, I prefer to see them as one-place predicates, re-quiring only one complement. When this position is filled, the PP as a whole can function as a modifier of some other predicate (the book on the table), as a predicate (the book is on the table) or as an argument (I put the book on the table).

7. For a similar treatment of spatial and temporal prepositional phrases in French, see François (1996).

8. Samuelsdorff (1998) also assumes there to be a small set of semantic functions that are direct-ly realised by a preposition. He rejects the idea of representing prepositions as one-place predi-cates because predicates, he argues, express a property of their argument, which prepositions do not: in the phrase under the table, under does not express a property of the table. I do not believe, however, that this is a valid argument. Relational predicates do not express a property of their argument; instead predicate and argument together express a property of some other entity (the entity they restrict or modify).

9. A similar view can be found in the analyses of prepositional phrases developed in other theoretical frameworks, e.g. Lexical Functional Grammar (Bresnan 1982, 2001), Generative Grammar (Van Riemsdijk and Williams 1986; Radford 1988; Corver and Van Riemsdijk 2001), Generalised Phrase Structure Grammar (Gazdar et al. 1985), Head-Drive Phrase Structure Grammar (Pollard and Sag 1984), as well as in traditional, descriptive approaches (e.g. Hud-dleston and Pullum 2002).

10. Note that although grammaticalisation studies emphasise the fact that the change from lexi-cal to grammatical is, indeed, a gradual one and that we are dealing with a continuum rather than with two clearly distinguished sets (e.g. Bybee 2003; Bybee et al. 1994; Heine and Kuteva 2002a, 2002b; Hopper 1991; Hopper and Traugott 1993; Lehmann 1989, 1995), they do, at the same time, continually classify elements as either lexical or grammatical, without, however, giv-ing any indication of where and how to draw the boundary. For a more principled approach, see Keizer (2007).

Page 37: Prepositions in Discourse

© 2008. John Benjamins Publishing CompanyAll rights reserved

English prepositions in Functional Discourse Grammar 251

11. Compare also Halliday and Hasan (1976: 274), who describe ‘general nouns’, such as thing, people, place or matter, as “borderline case[s] between a lexical item (member of an open set) and a grammatical item (member of a closed set)”.

12. Observe that such an analysis is corroborated by the fact that the second PP can be replaced by the pronominal forms here and there.

13. Lindstromberg’s choice of example here is unfortunate as it does not really prove his point; after all, the head of argument of the preposition, somewhere, still designates a specific point along the line. A better example would be He lives three miles from the border, where border does not designate a point but a line, even though the hearer may choose to interpret the construction as meaning ‘He lives three miles from a particular point along the border’. In that case, however, the aspect of non-dimensionality is not part of the meaning of from, but inferred from a combi-nation of lexical, contextual and general knowledge.

14. One of the reviewers offered an alternative explanation for the difference in meaning be-tween (8a) and (8b). According to him/her, at does not indicate an exact geometric relation-ship between the two entities; instead its meaning is functional rather than locational. Thus a sentence like John is at the piano is only appropriate if John is ‘positioned canonically to use the piano for its intended purpose’. Similarly, the reviewer would prefer to explain the difference between (8a) an (8b) as indicating a difference in the interaction between the entities involved: whereas at seems to indicate that John is engaged in the usual activities associated with super-markets (i.e. he’s shopping), in appears neutral with respect to his activities (i.e. he might have ducked in to get out of the rain). I quite agree with the reviewer’s interpretation of the sentence John is at the piano. However, this particular functional element of at seems to be present only in a restricted number of in-stances: e.g. only with [+human] subjects (note, for instance, its absence in such expression as the house at the corner or the light at the end of the tunnel) and only in certain combinations and contexts. I am not really convinced that it is this functional element that accounts for the differ-ence between (8a) and (9b); to me the use of at still seems to be primarily spatial here. Note in his respect also the following example (from Google): I was about three years old, and standing at the piano, which my sister was playing, and my eyes were at the keyboard level). Nevertheless, it will certainly be worthwhile to see how the various uses of at (its different functional elements) relate to the restrictions on its distribution. Unfortunately, I will not be able to do so in the pres-ent article.

15. An example of a recently added member of the class of prepositions may be contra, which in most dictionaries is not mentioned as a preposition (the OED gives a few examples, but the prepositional status of contra in these examples is disputable), but which is nowadays used as a transitive preposition (as in the unjust relationship of the US contra the world (Google) or Again, contra Lawrence, it is obvious that for Eliot the idea that modern western society should adopt sav-age customs is seen as ludicrous and reprehensible (BNC)). Other examples of prepositions that at some stage have entered the English lexicon are on top of and in front of (the latter becoming a complex preposition, losing its definite article, in the 19th century). Bakker and Siewierska (2002: 151ff.) describe the birth of a new Dutch preposition (richting ‘towards’).

16. The examples followed by a code beginning with ICE-GB are from the ICE-GB Corpus, a one-million-word corpus, tagged and parsed, of spoken and written English composed at the Survey of English Usage, University College London.

Page 38: Prepositions in Discourse

© 2008. John Benjamins Publishing CompanyAll rights reserved

252 Evelien Keizer

17. As pointed out by one of the reviewers, it is possible to use such combinations in expressions like He came from up North and The cat came from across the street. Note, however, that in these cases the PPs up North and across the street do not describe a path (as in He walked up the road or The cat ran across the street), but a location (the endpoint of a conceptualised path between the position of the speaker and that of some other entity).

18. Even more surprising (or questionable) may be the following examples, both of which are from the official websites of universities and organisations:

(ia) After completing his service in the Marine Corps, Brigham pursued graduate studies in chemical engineering from at the University of Oklahoma from 1956 to 1962. (www.stanford.edu/dept/news/pr/2004/brighamobit218.html)

(ib) Because of her expert knowledge, Dr Bennett is working close together with Radiologists and Radiographers from at the Leicester Royal Infirmary to further develop scanning meth-ods for NHS patients. (http://www.le.ac.uk/press/press/eyespecialists.html)

What seems to be happening in these examples, as well as in the examples in (16), is that the sec-ond preposition (at) is used as a place marker; in other words, it is used to indicate that the argu-ment of the first preposition (from) is not an object (the default interpretation), but a location.

19. I am grateful to one of the reviewers for his/her comments on this part of the article, which have forced me to reconsider some of the arguments used.

20. Note that on a truly contrastive reading, every place-denoting preposition can follow from, e.g. I wasn’t near the car park but at the car park.

21. In FG it has always been assumed that only lexical elements can be input to predicate for-mation rules. As one of the reviewers correctly points out, however, it is possible to for certain derivational affixes to appear in conversion processes (e.g. isms and ologies), while a great many of such affixes are semantically transparent (e.g. such neoclassical affixes as arch-, mono-, bio-, etc., but also such more mundane affixes as -ful and -less, un- and -able). This seems to clash with the FG treatment of such affixes as grammatical elements, triggered by the expression rules (e.g. e.g. Dik 1997a: 349; Dik 1997b: 168). The exact status of derivational elements is, however, far from unequivocal. In grammatical-isation studies, for instance, it is generally assumed that they have more lexical content than in-flectional affixes (e.g. Halliday 1961: 249; Bybee 1985; Hammond and Noonan 1988; Lehmann 1989; Hopper and Traugott 1993; Bybee, Perkins and Pagliuca 1994: 22). As such they may be seen as, at least, partially lexical; i.e. they can ‘be considered to serve a role between contentive and grammatical forms’ (Hopper and Traugott 1993: 4). This then may account for the fact that occasionally they may be used independently, and may even undergo conversion processes. Al-ternatively, it may be argued that sometimes a grammatical (or highly grammaticalised) element can occasionally be “upgraded” (degrammaticalised) for the occasion.

22. As pointed out by one of the reviewers, adding a prepositional prefix may increase the va-lency of a verbal predicate, as in look/overlook, live/outlive, go/undergo. This may be seen as sup-porting an analysis of prepositions as lexical elements.

23. It might be claimed that in some of these cases the prefix has adjectival or adverbial force. Where the preposition is used in its primary, spatial sense, however, there is good reason to as-

Page 39: Prepositions in Discourse

© 2008. John Benjamins Publishing CompanyAll rights reserved

English prepositions in Functional Discourse Grammar 253

sume that we are dealing with the basic, prepositional form (overcoat, overtower). With regard to under, OED specifically states that the prefix is always a preposition.

24. For practical reasons, the discussion to follow will concentrate on spatial prepositions, for which representation by means of the l-variable is unproblematic. In those cases where the preposition does not have a locational meaning, representation by means of an f-variable will be more appropriate (indicating that the preposition or PP ascribes some property f (e.g. the property for Mary in a present for Mary or with a handle in a cup with a handle) to an entity (the present/cup)).

25. PFA, of course, also applies to deadjectival nouns like height. This means that also in a con-struction like the height of the building, the preposition is regarded as having a grammatical use.

26. The variable α is used to denote any type of entity (generalising over the various variables at the Representational level).

27. Note that in an expression like John is in, the intransitive preposition in must undergo non-verbal predicate formation, to supply it with an argument position. This is necessary to account for the fact that the subject (John) is not the argument of the preposition (which is intransitive, with an implied argument like the building/the office/the house), but of the prepositional phrase as a whole (i.e. the intransitive preposition together with its implied argument).

28. Note that for the very same reason the agent PPs in passive constructions are best analysed as modifiers introduced by the grammatical element by.

References

Bakker, D. and A. Siewierska (2002) Adpositions, the lexicon and expression rules. In R. Mairal Usón and M.J. Pérez Quintero (eds.) New perspectives on argument structure in Functional Grammar. Berlin: Mouton de Gruyter. 125–177.

Bresnan, J. (ed.) (1982) The mental representation of grammatical relations. Cambridge, MA: MIT Press.

Bresnan, J. (2001) Lexical-functional syntax. Oxford: Blackwell.Brugman, C. (1983) The story of over. Bloomington, Indiana: Indiana University Linguistics Club.Brugman, C. and G. Lakoff (1988) Cognitive typology and lexical networks. In S. Small, G. Cot-

trell, and M. Tanenhaus (eds.) Lexical ambiguity resolution. Palo Alto, CA: Morgan Kauf-man. 477–507.

Bybee, J.L. (2003) Cognitive processes in grammaticalization. In M. Tomasello (ed.) The new psychology of language. Cognitive and functional approaches to language structure: Vol. 2. Mahwah, NJ: Erlbaum. 145–167.

Bybee, J.L., R. Perkins and W. Pagliuca (1994) The evolution of grammar: tense, aspect, and mo-dality in the languages of the world. Chicago: The University of Chicago Press.

Connolly, J.H. (199) On the generation of English temporal satellite terms. In E. Engberg-Ped-sen, L. Falster Jakobsen and L. Schack Rasmussen (eds.) Function and expression in Func-tional Grammar. Berlin: Mouton de Gruyter. 303–315.

Connolly, J.H (1995) On the generation of internally complex English temporal satellite terms. Working Papers in Functional Grammar 57.

Page 40: Prepositions in Discourse

© 2008. John Benjamins Publishing CompanyAll rights reserved

254 Evelien Keizer

Corver, N. and H. van Riemsdijk (2001) Semi-lexical categories. In N. Corver and H. van Ri-emsdijk (eds.) Semi-lexical categories. Berlin: Mouton de Gruyter. 1–19.

Coventry, K.R. (2003) Spatial prepositions, spatial templates, and ‘semantic’ versus ‘pragmatic’ visual representations. In E. van der Zee and J. Slack (eds.) Representing direction in lan-guage and space. Oxford: Oxford University Press. 255–267.

Coventry, K.R., R. Carmichael and S.C. Garrod (1994) Spatial prepositions, object-specific func-tion, and task requirements. Journal of Semantics 11: 289–309.

Dewell, T. (1994) Over again: Image-schema, transformations in semantic analyses. Cognitive Linguistics 5.4: 351–380.

Dik, S.C. (1978) Functional Grammar. Dordrecht: Foris.Dik, S.C. (1985a) Valentie en valentie operaties in Funktionele Grammatika. Tijdschrift voor

Taal- en Tekstwetenschap 5: 95–114.Dik, S.C. (1985b) Formal and semantic adjustment of derived constructions. In A.M. Bolkestein,

C. De Groot and J.L. Mackenzie (eds.) Predicates and terms in Functional Grammar. Dor-drecht: Foris. 1–28.

Dik, S.C. (1989) The theory of Functional Grammar. Dordrecht: Foris.Dik, S.C. (1997a) The theory of Functional Grammar: Vol. 1: The structure of the clause. Berlin:

Mouton de Gruyter.Dik, S.C. (1997b) The theory of Functional Grammar: Vol. 2: Complex and derived constructions.

Berlin: Mouton de Gruyter.Dowty, D.R. (1979) Word meaning and Montague grammar. Dordrecht: Reidel.François, J. (1996) Reference to space and time through local adpositions and appositional cases:

beyond the proposals of C. de Groot, J.H. Connolly and J.L. Mackenzie. Working Papers in Functional Grammar 62.

Garrod, S.C, G. Ferrier and S. Campbell (1999) In and on: investigating the functional geometry of spatial prepositions. Cognition 72: 167–189.

Gazdar, G., E. Klein, G. Pullum and I. Sag (1985) Generalised Phrase Structure Grammar. Ox-ford: Blackwell.

Groot, C. de (1989) Predicate structure in a Functional Grammar of Hungarian. Dordrecht: Foris.Halliday, M.A.K. (1985) An introduction to functional grammar. London: Arnold.Halliday, M.A.K. and R. Hasan (1976) Cohesion in English. London: Longman.Heine, B. and T. Kuteva (2002a) The world lexicon of grammaticalization. Cambridge: Cam-

bridge University Press.Heine, B. and T. Kuteva (2002b) On the evolution of grammatical forms. In A. Wray (ed.) The

transition to language, Oxford: Oxford University Press. 376–397.Hengeveld, K. (2004) The architecture of a Functional Discourse Grammar. In J.L. Macken-

zie and M.Á. Gómez González (eds.), A New Architecture for Functional Grammar. Berlin: Mouton de Gruyter. 1–21.

Hengeveld, K. (2005) Dynamic expression in Functional Discourse Grammar. In C. de Groot and K. Hengeveld (eds.) Morphosyntactic expression in Functional Grammar. Berlin: Mou-ton de Gruyter. 53–86.

Hengeveld, K. and J.L. Mackenzie (2006) Functional Discourse Grammar. In K. Brown (ed.). Encyclopedia of language and linguistics. Vol. 4. Oxford: Elsevier. 668–676.

Hengeveld, K. and J.L. Mackenzie (2008) Functional Discourse Grammar. Oxford: Oxford Uni-versity Press.

Herskovits, A. (1986) Language and spatial cognition: An interdisciplinary study of the preposi-tions in English. Cambridge: Cambridge University Press.

Page 41: Prepositions in Discourse

© 2008. John Benjamins Publishing CompanyAll rights reserved

English prepositions in Functional Discourse Grammar 255

Hopper, P.J. (1991) On some principles of grammaticization. In E.C. Traugott and B. Heine (eds.) Approaches to grammaticalization. Vols. 1 and 2. Amsterdam: Benjamins. 17–35.

Hopper, P.J. and E.C. Traugott (1993) Grammaticalization. Cambridge: Cambridge University Press.

Huddleston, R. (1984) Introduction to the grammar of English. Cambridge: Cambridge Univer-sity Press.

Huddleston, R. and G.K. Pullum (2002) The Cambridge grammar of the English language. Cam-bridge: Cambridge University Press.

Jackendoff, R. (1983) Semantics and cognition. Cambridge, MA: MIT Press.Jolly, J.A. (1993) Preposition assignment in English. In R.D. Van Valin (ed.) Advances in Role and

Reference Grammar. Amsterdam: Benjamins 275–310.Kahr, J.C. (1975) Adpositions and locationals: typology and diachronic development. Working

Papers on Language Universals 19: 21–54.Keizer, M.E. (2004) Postnominal PP complements and modifiers: a cognitive distinction. Eng-

lish Language and Linguistics 8/2: 323–350.Keizer, M.E. (2007) The English noun phrase: The nature of linguistic categorization. Cambridge:

Cambridge University Press.Keizer, M.E. (2007) The lexical-grammatical dichotomy in Functional Discourse Grammar.

Alfa-Revista Lingüística 51/2 (Advances in Functional Discourse Grammar): 35–56.Kreitzer, A. (1997) Multiple levels of schematization: a study in the conceptualization of space.

Cognitive Linguistics 8: 291–325.Lakoff, G. (1987) Women, fire and dangerous things: What categories reveal about the mind. Chi-

cago: University of Chicago Press.Langacker, R.W. (1987) Foundations of cognitive grammar: Vol. 1: Theoretical prerequisites. Stan-

ford: Stanford University Press.Langacker, R.W. (1992a) Prepositions as grammatical(izing) elements. Leuvense Bijdragen 81:

287–309.Langacker, R.W. (1992b) The symbolic nature of Cognitive Grammar: the meaning of of and

of-periphrasis. In M. Pütz (ed.) Thirty years of linguistic evolution: Studies in honour of René Dirven on the occasion of his sixtieth birthday. Amsterdam: Benjamins. 483–502.

Lee, D. (1998) Intransitive prepositions: are they viable? In P. Collins and D. Lee (eds.), The clause in English. Amsterdam: Benjamins, 133–147.

Lee, D. (2001) Cognitive linguistics: an introduction. Oxford: Oxford University Press.Lehmann, Chr. (1989) Grammatikalisierung und Lexikalisierung. Zeitschrift für Phonetik,

Sprachwissenschaft und Kommunikationsforschung 42: 11–19.Lehmann, Chr. (1995) Thoughts on grammaticalization. München: Lincom Europa.Lehmann, Chr. (2002) New reflections on grammaticalization and lexicalization. In I. Wischer and

G. Diewald (eds.) New reflections on grammaticalization. Amsterdam: Benjamins. 1–18.Lindstromberg, S. (1998) English prepositions explained. Amsterdam: Benjamins.Lyons, J. 1977. Semantics. 2 vols. Cambridge: Cambridge University Press.Mackenzie, J.L. (1983) Nominal predicates in a Functional Grammar. In S.C. Dik (ed.) Advances

in Functional Grammar. Dordrecht: Foris. 31–51.Mackenzie, J.L. (1992a) Places and things. In M. Fortescue, P. Harder and L. Kristoffersen (eds.)

Layered structure and reference in a functional perspective. Amsterdam: Benjamins. 253–276.Mackenzie, J.L. (1992b) English spatial prepositions in Functional Grammar. Working Papers in

Functional Grammar 46.

Page 42: Prepositions in Discourse

© 2008. John Benjamins Publishing CompanyAll rights reserved

256 Evelien Keizer

Mackenzie, J.L. (1996) English nominalizations in the layered model of the sentence. In B. Devriendt, L. Goossens and J. Van der Auwera (eds.) Complex structures: A functionalist perspective. Berlin: Mouton de Gruyter. 325–353.

Mackenzie, J.L. (2001) Adverbs and adpositions: the Cinderella categories of Functional Gram-mar. In M.J. Pérez Quintero (ed.) Challenges and developments in Functional Grammar. (Monographic section of Revista Canaria de Estudios Ingleses 42.) 119–135.

Mackenzie, J.L. (2002) The predicate in Function Grammar. In R. Mairal Usón and M.J. Pérez Quintero (eds.) New perspectives on argument structure in Functional Grammar. Berlin: Mouton de Gruyter. 1–38.

Mackenzie, J.L. and M. Hannay (1982) Predicational predicates and focus constructions in a Functional Grammar of English. Lingua 56: 43–57.

Olbertz, H. (1998) Verbal periphrases in a Functional Grammar of Spanish. Berlin: Mouton de Gruyter.

Pérez Quintero, M.J. (2004) Adpositions in FG: has Cinderella been invited to the ball? In H. Aertsen, M. Hannay and R. Lyall (eds.) Words in their places: Festschrift for J. Lachlan Mack-enzie. Amsterdam: Faculty of Arts, Vrije Universiteit Amsterdam.

Pollard, C. and I. Sag (1984) Head-driven Phrase Structure Grammar. Chicago: Chicago Uni-versity Press.

Radford, A. (1988) Transformational Grammar: A first course. New York: Cambridge University Press.

Riemsdijk, H. van and E. Williams (1986) Introduction to the theory of grammar. Cambridge, MA: MIT Press.

Samuelsdorff, P.O. (1998) Pronouns, adpositions, “adverbs” and the lexicon. In H. Olbertz, K. Hengeveld and J. Sánchez García (eds.) The structure of the lexicon in Functional Grammar. Amsterdam: Benjamins. 267–278.

Tyler, A. and V. Evans (2001) Reconsidering prepositional polysemy networks: the case of over. Language 77(4): 724–765.

Van Valin, R.D. and R.J. La Polla (1997) Syntax: structure, meaning and function. Cambridge: Cambridge University Press.

Weigand, H. (1990) Linguistically motivated principles of knowledge base systems. Dordrecht: Foris.

Yule, G. (1996) The study of language. Cambridge: Cambridge University Press.Zeller, J. (2001) Lexical particles, semi-lexical postpositions. In N. Corver and H. van Riemsdijk

(eds.) Semi-lexical categories. Berlin: Mouton de Gruyter. 505–549.

Author’s address

Evelien KeizerSpuistraat 2101012 VT AmsterdamThe Netherlands

[email protected]