Multi-preposition constructions in English

Evelien Keizer, Universiteit van Amsterdam, The Netherlands

Abstract

This paper examines the semantic and syntactic properties of English constructions

containing a verb followed by two (or three) prepositions (including expressions like to

walk out on, to talk someone out of, to go up to). First it is argued that, just like simple

verb-preposition constructions (e.g. to come across, to switch off or to refer to), multi-

preposition construction (MP-construction) come in various types. By applying a large

number of semantic and syntactic criteria to authentic examples, it is shown that a

distinction needs to be made between two major construction types – the composite-

predicate construction and the appositional construction – whereby the latter

construction type can be further divided into a resultative construction and a Verb + PP-

construction. Subsequently, FDG representations are offered for each of these

construction types, reflecting the differences between these types at the

Representational Level. Finally, some important implications for the theory of FDG are

considered concerning the categorization of lexical elements, in particular the viability

of (a) the distinction between particles, locative adverbs and prepositions (where they

all take the same form), (b) the distinction between grammatical and lexical

prepositions, and (c) the distinction between prepositions and conjunctions (where they

take the same form).

1. Introduction

Although a great deal of attention has been paid, in various theoretical frameworks, to

verb-preposition constructions,1 very little research has been done on constructions with

multiple prepositions.2 This paper intends to look at these constructions in some detail.

Like simple verb-preposition constructions, multi-preposition constructions (henceforth

MP-constructions) come in various kinds, illustrated in (1)-(4):

Construction I (cf. Collins Cobuild English Grammar, 1990:169-170):

Intransitive: (1) a. His girlfriend walked out on him.

b. I couldn’t put up with his paranoia.

Transitive: (2) a. We tried to talk her out of it.

b. I’m sure they put him up to it.

Construction II:

Intransitive: (3) a. John went up to the roof.

b. He walked out into the garden.

Transitive: (4) a. We put the junk down on the floor.

b. Sue left the chairs out in the garden.

In examples (1) and (2), referred to for the time being as Construction I, we find what

have traditionally been called phrasal verb constructions: fixed combinations of a verb

and two prepositional elements (often called particles or adverbs). The verbal complex

may take its own direct object (examples (2a&b)), but need not do so (examples

43

(1a&b)). In examples (3) and (4) (Construction II), the verb and the two prepositional

elements are not generally regarded as forming one semantic or syntactic unit. The exact

relation between the verb and the two prepositions, and the internal structure of the

prepositional unit is, however, far from clear.

Part of this paper will be based on some recent work on prepositions and verb-

preposition constructions in FGD (Keizer 2008, in press). It will be demonstrated that

although there are important parallels between simple verb-preposition constructions

and MP-constructions, these parallels are not perfect, and that MP-constructions

definitely merit their own treatment and analysis. In addition, I will discuss some

broader implications for the theory of FDG.

Before starting the discussion, I would like to comment briefly on the data and

method used. Throughout the paper, use will be made of authentic data from a variety of

sources: the ICE-GB-Corpus, the British National Corpus (BNC), the Corpus of

Contemporary American English (COCA) and the Internet (Google search). These data

will be used, first, to establish the semantic properties of the various MP-constructions

(see A. below). Secondly, the data are used to find out to what extent any semantic

differences between these constructions are reflected in their syntactic behaviour (see B.

below), in accordance with the basic principle of FDG that only those semantic

distinctions that are formally expressed in a language are relevant for the grammar of a

language:

[FDG] is form-oriented in providing, for each of the languages analysed, an account of only

those interpersonal and representational phenomena which are reflected in morphosyntactic or

phonological form. (Hengeveld and Mackenzie 2008:39; see also Hengeveld and Mackenzie

2008:15).

Previous research into MP-constructions, supplemented by an analysis of the

data, suggests that at least the following semantic and syntactic criteria are relevant to

an understanding of the internal structure of MP-constructions. In what follows these

criteria will therefore be applied to the constructions in (1)-(4):

A. Semantic criteria:

(i) To what extent does the verb retain its original meaning?

(ii) Does the verb have its usual valency?

(iii) Is there a resultative relation between the prepositional unit and any of the

arguments?

(iv) Does the verb ‘select’ a particular preposition?

B. Syntactic criteria:

(i) Does the construction allow ‘extraction’ (clefting, questioning and fronting) of

the prepositional unit?

(ii) Does the construction allow for coordination of the prepositional unit?

(iii) Does the prepositional unit occur independently in other syntactic

environments?

(iv) Can either of the prepositional elements be omitted (without affecting the

internal structure of the construction)?

(v) Does the construction allow sequences of more than two prepositions?

(vi) Does the construction allow for alternative word orders (reversibility)?

44

2. Earlier treatments

As mentioned before, MP-constructions have not been studied in much detail.

Hengeveld and Mackenzie (2008:232-233) analyse sentences like (5a) as involving a

directional argument (to the station) and a locative expression (down) modifying the

verb; the string went down is, in other words, regarded as forming a semantic unit:

(5) a. He went down to the station.

b. (fi: goV (fi): [(li: (fj: down (fj)) (li)) (fi)Ф])

Hengeveld and Mackenzie (2008:232)

Hengeveld and Mackenzie give the following reasons for adopting this analysis:

The close relationship of these modifiers with the verbs they modify shows up not only in the

fact that they are restricted in use to movement verbs, but also in that many languages encode

directional meanings lexically. (Hengeveld and Mackenzie 2008:232)

To illustrate the second point, Hengeveld and Mackenzie mention Spanish baj-ó ‘he

went down’ and German her-kommen ‘to come here’. One may doubt, however, the

validity of this argument: the fact that a particular language uses a single lexeme to

denote a certain concept does not mean that the combination of words used to denote

that same concept in another language must also form a semantic unit.3 The first reason

given for analysing went down as one unit is also problematic: as shown by the

examples in (6), modifiers like down, up and away are not restricted in use to movement

verbs:

(6) a. The next time I went down, I stayed down (BNC)

b. Safety experts recommend that children never sled head-first or while lying down on

their stomachs. (COCA)

c. we didn’t used to see much of it when we lived up in Manchester (BNC)

d. “I told you I’d be back, but perhaps I stayed away too long…” (BNC)

Moreover, as will be demonstrated in detail later on, there is compelling formal

evidence that in (5a) it is down to the station which forms one unit, rather than the

combination went down.

Within the generative framework, Den Dikken (1995:144) proposes an analysis

of sentences like (7a) as involving two Small Clauses. In SC2, the PP on the shelf is

predicative, θ-marking a book. In SC1, down is the SC-predicate, predicating over SC2.

This higher SC is the complement of the verb put:

(7) a. John put (down) the book (down) on the shelf

a’. [VP put [SC1 θ’ [XP down [SC2 the book [PP on the shelf]]]]]4

I agree with Den Dikken that the string the book down on the shelf forms one

constituent. I do not, however, agree with the internal structure Den Dikken assigns to

this phrase, as I find it semantically implausible (it would mean that ‘the book being on

the shelf’ is ‘down’). In what follows I will defend the view that on the shelf functions

as a further specification (appositional modification) of down.5

45

3. Keizer (in press)

In the treatment of verb-preposition constructions presented in Keizer (in press), three

different analyses are proposed to account for the distinctive semantic and syntactic

properties of each group. Since the treatment of MP-constructions will be based on

these analyses, I will start by giving a short overview of the data and proposed

representations. The following three types of construction were distinguished:

 The composite predicate construction: [V + P]

 The resultative construction: V + [P + NP]

 The V + PP-construction: V + PP

Examples (8) to (10) give some of the relevant syntactic features of the constructions

involved. Even a quick glance shows that, despite the superficial similarities between

the three groups (illustrated in examples (8a), (9a) and (10a)), they differ considerably

in syntactic behaviour:

(8) a. Pete came across the letters.

b. *Pete came the letters across.

c. *The letters are across.

d. Pete came across them.

e. *Pete came them across

f. *It was across the letters that Pete came.

g. *Pete came across the letters and across some old photographs.

(9) a. Pete switched off the lights.

b. Pete switched the lights off.

c. The lights are off.

d. *Pete switched off them.

e. Pete switched them off.

f. *It was off the lights that Pete switched.

g. *Pete switched off the lights and off the heating.6

(10) a. Pete depends on his parents.

b. *Pete depends his parents on.

c. *His parents are on.

d. Pete depends on them.

e. *Pete depends them on

f. It is on his parents that Peter depends.

g. Pete depends on his parents and on his friends.

Thus, reversing the order between preposition and NP (b-sentences) and reformulating the result

of the action in the form of a copular sentence (c-sentences) is possible in (9), but not in (8) and

(10). The NP can be replaced by a proform when it follows the preposition (d-sentences) in (8)

and (10), but not in (9); when the preposition follows the NP (e-sentences), using a proform is

possible in (9), but not in (8) and (10). Finally, clefting (f-sentences) and coordination (g-

sentences) of the ‘P + NP’-string yields a grammatical construction in (10) but not in (8) and

(9).

Analysis I: Composite predicate construction: [V + P]

Keizer (in press) analyses the expressions in (8) as composite predicate constructions,

i.e. as constructions with a composite predicate consisting of a verb and a preposition.

These constructions can be either intransitive or transitive. Other examples of such

46

combinations are the two-place composite predicates stumble across, take after, stand

by, make for, grow on, come by (for the intransitive construction in (11)) and the three-

place predicates draw into, hold against, let into (a secret), talk into, get through (the

test), keep off (the premises) (for the transitive construction in (12)).

Intransitive:

(11) a. Pete came across some letters.

b. (Past ei: (fi: [(fj: [V P] (fj)) (xi)A (xj)Ref] (fi)) (ei))

where V = ‘come’

P = ‘across’

fj = ‘come across’ (composite predicate)

xi = ‘Pete’ (first argument of ‘come across’ (Actor))

xj = ‘some letters’ (second argument of ‘come across’ (Reference)7)

Transitive:

(12) a. Sam showed us around the theatre.

b. (Past ei: (fi: [(fj: [V P] (fj)) (xi)A (xj)U

(xk)Ref] (fi)) (ei))

where V = ‘show’

P = ‘around’

fj = ‘show around’ (composite predicate)

xi = ‘Sam’ (first argument of ‘show around’ (Actor))

xj = ‘us’ (second argument of ‘show around’ (Undergoer))

xk = ‘the theatre’ (third argument of ‘show around’ (Reference))

In these constructions, the verb and the preposition behave semantically as one unit:

together they denote the relation that is asserted to hold between the arguments.

Syntactically, the tests of clefting and coordination (examples (8f&g)) clearly indicate

that the preposition and the noun phrase do not act as one syntactic unit. The fact that

adverbs can occasionally appear between the composite parts (and in the transitive

construction also the direct object NP) suggests that the verb and the preposition do not

form a single lexeme.

Analysis II: The resultative construction: V + [P + NP]

The examples in (9) are given a different analysis, in which the preposition functions as

a non-verbal predicate in a resultative construction. These constructions, too, have an

intransitive and a transitive form. Other prototypical examples of resultative verb-

preposition constructions include back down/off, barge in, boil over, break out, catch

on, die out, doze off, go back, melt away, lie down, stop by, stick around (for the

intransitive construction in (13)) and bring down, call off, drag out, give back, have on,

let in, leave behind, shut out (for the transitive construction in (14)).

Intransitive:

(13) a. Sue came in. (result: ‘Sue is in’)

b. (ei: (fi: [(fj) (xi)A (fk: [(li) (xi)] (fk))Res] (fi)) (ei))

where fj = ‘come’

li = ‘in’

fk = ‘Sue (being) in’ (second argument of ‘come’ (Result))

xi = ‘Sue’ (first argument of ‘come’ (Actor))

47

Transitive:

(14) a. Pete switched off the lights. (result: ‘the lights are off’)

b. (ei: (fi: [(fj) (xi)A (xj)U (fk: [(fl) (xj)] (fk))Res] (fi)) (ei))

where fj = ‘switch’

fl = ‘off’

fk = ‘the lights (being) off’ (third argument of ‘switch’ (Result))

xi = ‘Pete’ (first argument of ‘switch’ (Actor))

xj = ‘the lights’ (second argument of ‘switch’ (Undergoer))

Semantically, these constructions differ considerably from composite predicate

constructions. Firstly, the sentences in (13a) and (14a) inevitably result in a state (‘Sue

is in’ in (13); ‘The lights are off’ in (14)). Another semantic difference concerns the fact

that the actions denoted by the verb can quite easily be conceptualized independently

from the meaning of the preposition; verb and preposition do not form one meaning

unit.

Neither is it plausible to regard the preposition and the direct object NP as

forming one prepositional phrase; witness example (9b), where the preposition follows

the NP, and (9f&g), which show that the combination of P and NP cannot be clefted and

coordinated. Instead, the P and the NP are analysed as a non-verbal Configurational

Property (Sue in the garden, the lights off),8 expressing the result of the action denoted

(i.e. ‘switching’).

Analysis III: The V + PP-construction

The sentences in (10), finally, are analysed as straightforward V + PP-constructions.

Once again, a distinction can be made between intransitive and transitive constructions.

In addition, some verbs select a specific preposition (depend on), whereas other verbs

combine with a range of different prepositions (lead (in)to/from/up/off/across/over) etc.

Examples of the former can be found in (15) and (16):

Intransitive:

(15) a. He converted to Catholicism.

b. (Past ei: (fi: [(fj) (xi)A (fk)] (fi)) (ei))

where fj = ‘convert’

x

i = ‘he’

f

k = ‘to Catholicism’

Transitive:

(16) a. They converted him to Catholicism.

b. (Past ei: (fi: [(fj) (xi)A (xj)U (fk)] (fi)) (ei))

where fj = ‘convert’

x

i = ‘they’

x

j = ‘him’

f

k = ‘to Catholicism’

In all these construction the preposition and the following NP behave syntactically as

one prepositional phrase. Semantically, too, they function as independent units,

complementing – as a whole – the meaning of the verbal lexeme.

48

4 Multi-preposition constructions in FDG

Turning to MP-constructions, it will be argued that these, too, can be divided into

different groups. Here two major construction types can be distinguished; the second

type can be further divided into two subtypes:

 Construction I: The composite predicate construction (Section 4.1)

 Construction II: The appositional construction (Section 4.2)

o The resultative construction

o The V + PP-construction

4.1 Construction I: Composite predicate constructions

In this section it will be argued that constructions like those given in example (17) can

justifiably be regarded as composite predicate constructions. Section 4.1.1 provide the

semantic and some of the syntactic reasons for doing so (more details will be provided

in Section 4.2.1). Section 4.1.2 will suggest an underlying representation for these

constructions.

4.1.1 Characterization of the composite predicate construction

The first major construction type is exemplified by the sentences in (17) and (18):

(17) Intransitive:

a. I couldn’t put up with his paranoia. (= example (1b))

b. His girlfriend walked out on him.

(18) Transitive: