Particles as grammaticalized complex predicates

25





For a discussion of the relationship between grammaticalization and lexicalization, see e.g. Brinton (2002).
See also Asbury (2005) for a discussion of Hungarian, where the often-assumed dichotomy of adpositions-are-free and case-suffixes-are-bound is not as clearcut as in many other languages.
P(articiples) are a special case of A(djectives).
See also Lipka's semantic types of phrasal verbs where the CAUSE types are usually matched by a BECOME type. To take an example, the BECOME counterpart of (11) consists of verbs like (sun/news, truth/daughter in photograph) come out, (news) filter out, (anger) flame out, (news) leak out, (moon/ancient belief) peep out (Lipka 1972: 197-198).
Conversions as in (iv) and (v) are a much-noted phenomenon with particle-verbs; see e.g. Lipka (1972: 98-114).
Eg. in examples like Hi eodon up to þære dune (Num. 14:40), He eode ut on ðæt land (Gen. 24, 63). Ut and up in such phrases are either adverbs in the spec of a PP or heads themselves postmodified or complemented by a PP. See Elenbaas (2007) for a discussion.
The causative variant would be I ran my car over a bottle.
Non-spatial uses are found with eg. forth, but here the particle is an event-modifier, and does not conform to either the LCS in (17) or those in (16):

(i) Peter cnucode forð oð þæt hi hine inn leton (Hml. Th. i. 396, 34; Wlfst. 222 33)
Peter knocked forth until they him in let
'Peter kept on knocking until they let him in'

For such event-modifiers, see McIntyre (2001) and Los (2004).
Non-light verbs are rarer, but they do exist; eg. (i), with formian 'scour':

(i) formige man þone pytt clæne
scour one the well cleane
'scour the well clean, scour out the well'

Verbs like sing and work have of course not been dealt with yet in the MED.
Example (32) is probably the same construction as the OE example of (i):

(i) þa englas cwædon him to (Gen 19.17)
the angels spoke him to
'The angels spoke to him'

Structures like (32) and (i) are very reminiscent of complex verbs in Modern Dutch and German that appear to be postpositions rather than the result of grammaticalized complex predicates. German examples that ultimately derive from the same pattern are what Blom (2005) has termed "postpositional particle verbs" like anstarren 'stare at', and zulachen 'smile at'. As German has preserved its case endings, we can tell from the dative case of the objects of these verbs that they are the complement of the postposition rather than the object of the particle-verb combination.

The particle-verbs shut up, carry out, pick up and point out are the only Category 2 verb combinations in their list of the phrasal verbs with the highest frequency (Biber et al. 1999: 410).
Example (i), for instance, is an instance of the way-construction (Goldberg 1995):
(i) REGAN
Go thrust him out at gates, and let him smell his way to Dover. (Shakespeare, King Lear, Act III, Sc. vii)
Bettelou Los
Radboud Universiteit Nijmegen
Particles as grammaticalized complex predicates


Abstract

This paper argues that English phrasal verbs represent a grammaticalization, from Phrase to Head, of a complex predicate construction. Predicates and the particles of phrasal verbs share a number of striking quirks: syntactically, both may appear with "unselected objects" and, semantically, both may form idioms of which the meaning cannot be predicted from its separate parts. Particles cannot be analyzed as predicates synchronically, however, because they allow two word orders: V – NP – particle and V – particle – NP, whereas predicates only allow the first of these, and not the second; furthermore, the particle appears to "bleach" much more easily than predicates, probably because the prototypical predicates, adjectives, express properties, whereas particles (prepositions) express paths. EModE marks a significant point in the development of the particle verb system in that the verbs participating in the combination are no longer restricted to 'light' verbs but include deadjectival and denominal verbs, unergatives, and 'manner-of-motion' verbs.


Introduction

This chapter investigates the origins of English phrasal verbs. These verbs consist of a verb and an adverbial or prepositional element traditionally referred to as a particle. The observation that particles share many characteristics with complex predicates has a long history in the literature from at least the early fifties onwards (Anthony 1953: 86). It has been observed, for instance, that the two orders famously exhibited by phrasal verbs, V – NP – Particle (as in (1a)), and V – Particle – NP (as in (1b)), are also found with complex predicates (cf. (1c) with (1a), and (1d) with (1b)).

(1) a. He threw the remains of his dinner away.
b. He threw away the remains of his dinner.
c. He threw the documents in the dustbin.
d. *He threw in the dustbin the documents.
e. He threw all the documents containing incriminating evidence in the dustbin.
f. He threw in the dustbin all the documents containing incriminating evidence.

The fact that the particle away derives from the Prepostitional Phrase on weg and was therefore of the same category as the predicate in the dustbin further supports a diachronic link, as we will discuss in the next section. The main difference between the particle and predicate constructions in (1) is the fact that the second order, V – Predicate – NP, as in (1d), is a marked one, and most probably the result of extraposition of the NP. Biber et al. (1999: 930) note that (1c) is the regular order ("by far the most common option"), and we will refer to this order as the 'predicate order'. Biber et al. conclude that the other order, as in (1d and 1f), is triggered by considerations of end-weight; note that (1f), with its long NP, is acceptable, whereas (1d) is not. In the case of particles, however, it is the V – Particle – NP order that is the most frequent one. We will refer to this order as the 'particle order', and alternations such as (1a)-(1b), where end-weight is not a trigger, as 'particle syntax'. Biber et al. note that there is no single factor that governs the selection of one particular order over the other: end-weight is one, but there are others. They note that the 'particle order' is linked with a high degree of idiomaticity of the combination, as in (2a), whereas the 'predicate order' tends to occur primarily with particles with literal, spatial meanings, as in (2b) (both examples from Biber et al. 1999: 933):

(2) a. Now carry out the instructions. (Fiction)
b. The Germans carried the corpse out. (Fiction)

They note that in (2b) "the result of the action is that 'the corpse is out', while it certainly is not true that 'the instructions are out' as a result of the action in idiomatic [(1a)]" (Biber et al. 1999: 933). Such resultative meanings are typical of predicates, as we will see below, which explains the tendency for such spatial particles to have 'predicate' orders. I will argue in this chapter that these, and other similarities between particles and predicates point to a diachronic relationship. The 'particle order' of (1b) is a morphosyntactic sign that the predicate has grammaticalized: from a phrase (XP), it has become a head (X0), allowing incorporation into the verb so that verb and particle express a single verbal action and function as a single lexeme. 'Particle syntax' is not restricted to particle verb combinations but may include V+AP and V+PP combinations as illustrated in (3) (see also Fraser 1965: 82ff, Bolinger 1971: 37ff, Quirk et al. 1985: 734, 1167; Claridge 2000: 66-70, 153; Denison 1981: 36-37). The items in (4), once PPs, now appear to be adverbs or adjectives. Such a lexicalisation from phrase to head of the predicate shows that the complex predicate construction is grammaticalizing.

(3) a. break/blow/blast/cut/fling/push/rake/whisk open, cut/stop short, bleach white, blow/keep/make/sift clear, put straight, let/set free, think fit, cast/let/pry/shake/wrestle loose, strip naked etc.
b. bring to light, put in execution, take in hand, call to mind, call in question, take into consideration etc.

(4) carry aloft (< on loft), set alight (< on light), take apart (< on part), put awry (< on wry), carry around (< on round), keep asunder (< on sunder), set afoot (< on foot), etc.

The arguments advanced against identifying particles as grammaticalized predicates focus on the failure of many particles (most notoriously up) to function as independent predicates (He phoned me up versus *I am up), the lack of telicity in many cases (though telicity would be expected if particles were grammaticalized predicates, which are as a rule resultative), and the failure of constituency tests and topicalization. I will argue in this chapter that these same quirks are exhibited by predicates. Resultative complex predicates show various degrees of productivity, transparency and idiosyncrasy which mirror those of particle verbs, and easily acquire idiomatic meanings that only work in combination with specific verbs: cf. drink NP under the table versus *he is under the table (Goldberg and Jackendoff 2004).
If the similarity in behaviour of predicates and particles can be taken as a pointer to a shared diachronic origin, it is all the more interesting that clear signs of this affinity between the two do not emerge until EModE. The affinity can be traced into OE, but then only with prefixes, which were still productive in that stage of the language, and not with particles.


2. Particles and predicates

Origin of particles

Bolinger (1971:18) provides this list of particles:

(5) aback, about, above, across, after, again, aground, ahead, along, alongside, aloud, apart, around, aside, askew, astray, astride, atop, asunder, athwart, away, back, before, behind, below, between, by, down, forth, forward, home, in, off, on, out, over, past, round, through, to, together, under, underground, underneath, up.

The great majority represent grammaticalized (or better, lexicalized) prepositional phrases. Underground is a clear example, but there are many others: away from onweg, down from adown > ofdune. Others are homophonous with PE prepositions (by, in, off, on, over, through, to, under, up) so they are probably prepositional in origin, but have lost their NP complement. A popular synchronic view is to label particles 'intransitive prepositions' (Emonds 1976) or to say that their NP-complement has become 'defocused' and remains implicit for that reason (McIntyre 2004), or to talk about 'reduced prepositional phrases' (Lipka 1972: 17); examples are (6a-d), from Lipka (1972: 17):

(6) a. He put the kettle on [the fire]
b. He took the ring off [his finger]
c. He ran up [the stairs]
d. She took the book out [of the pocket]

Some of these items are possibly more adverbial than prepositional (especially out), and this has been remarked on many times in the literature (eg. Sroka 1972 who distinguishes 'adverbs', 'prepositions' and 'adverb-prepositions words'). Fraser (1965) and Fairclough (1965) use 'particle' precisely because it is difficult to draw a clear line (see also Lipka 1972: 19). The problem of classification in a way reflects the status of the preposition, or broader, the adposition, as a syntactic category. Is it a lexical or functional category, is it analogous with V? V and P can both assign case, whereas N and A cannot; and classifications like intransitive/transitive can be argued to apply not only to V but also to P. Prepositions are said to express case realisations (Emonds 1985) and as such to appear in the extended projection of N rather than in a projection of their own. Adpositions and morphological case would then be expressions of the same functional category.
A third group seems definitely adverbial in origin, but even here we find a trace of a preposition: forth, forward, out; possibly also home (=homeward). Finally, past and round are in origin a past participle and an adjective, respectively. If particles originate in a syntactic construction as complex predicates, this explains why they are mainly of the categories P and A: they are in origin predicates (PP and AP are typically predicate categories) but have grammaticalized (after having lexicalized, in the case of some PPs) into particles and prefixes.

2.2 Can particles be analysed as predicates?

The examples in (7) list some PDE instances of complex predicates collected by Rappaport Hovav & Levin (2001).

(7) a. Last night, the dog poked me [predawake] every hour to go outside (The Toronto Sun, 27 Nov. 1994, p. 6)
b. Sudse cooked them all [pred into a premature death] with her wild food. (P. Chute. 1987. Castine. Garden City, NY: Doubleday, p. 78)
c. She might employ it [her body] as a weapon – fall forward and flatten me [pred wafer-thin]. (Delia Ephron. 2000. Big City Eyes, New York: Bantam, p. 92)

At an abstract level, there is a subject-predicate relationship between the accusative object and the complex predicate or object complement, i.e. between me and awake, them and a premature death, and me and wafer-thin. A simplified representation is the structure in (8), with an Agreement Phrase, with the empty head Agr mediating between the object as the subject of the Agreement Phrase and its adjectival predicate:

(8) VP

V AgrP

NP Agr'
me
Agr AP
awake

Note that the NP me (the object) receives its thematic role from the predicate in this representation, and has accusative case not because that case is assigned by the verb but because the NP is the subject of the Agreement Phrase and accusative is the default case for subjects in verbless or non-finite constructions. This means that it is the predicate that licenses the object and not the verb, and this accounts for the phenomenon of 'unselected' objects, i.e. cases where the verb on its own, without the predicate, could not appear with the object (more about this phenomenon in the next section).
The failure of most particles to function as a predicate in a copular construction is often noted in the literature, eg. Zeller (1999): *he is up/it is up (cf. phone John up/eat up all the food etc.), but many predicates similarly fail to appear straightforwardly in a copular construction; cf. the variability of (9a-f), rewriting the predicates of (7) and (13) below as copular constructions:

(9) a. I am awake.
b. *They were into a premature death.
c. I am wafer-thin.
d. *The pub is dry.
e. *The daylights were out of the campers.
f. The dust is out of the sofa.

The non-acceptability of these predicates in copular constructions does not in itself, however, constitute a counterargument to an analysis along the lines of (8) for predicates, in which the predicate selects the object. McIntyre (2001) provides a number of other meanings that are "idiosyncratically restricted to a particular structural environment": the malefactive use of on in my cat died on me, my car broke down on me is not possible as a postnominal modifier (*an accident on me). Goldberg and Jackendoff (2004: 560-562, and references cited there)) provide many more examples (eg. under the table in drink NP under the table), and set out in detail the idiosyncrasy of other predicates, particularly the choice of PP or AP: stab/bat/put/batter/frighten/crush/scare/burn NP to death versus *dead, but he sang himself hoarse versus *to hoarseness or he ate himself sick versus *to sickness; and he sang himself to exhaustion versus *exhausted. Note that even semantically transparent, relatively non-idiomaticized predicates like into shape and to death cannot function as independent predicates: *he is into shape, he is to death. McIntyre (2004: 546) points out that directional PPs and continuous state-of-change comparatives (i.e. inherently eventive PPs/APs) are incompatible with copulas although they uncontroversially predicate over NPs: *I am to the station/colder and colder versus I walked to the station/I got colder and colder. He concludes that the copula is untrustworthy as a test for complex predicates (McIntyre 2004: 547), and we conclude that it does not constitute counterevidence to our claim that there is a diachronic relationship between predicates and particles.
The reason that (7) is less likely to be the correct analysis for verb-particle constructions is the fact that particles have undergone grammaticalization and have started to form a single unit with the verb. The order in (1b), the 'particle order', appears to require a morphological analysis in which particle and verb form a unit, in its most extreme form along the lines of (10):


(10) VP

Vmax NP

Vmin particle

An analysis as in (10) can account for formations like get-at-able and knocker-upper, but is, in this most extreme form, ultimately untenable without special stipulations because verbal inflectional endings still attach to Vmin and not to Vmax (=the V+particle compound). Blom (2005: 104) notes that the combination of the properties compositionality, conventionality, and productivity, all strikingly present in particle verbs, is in fact very reminiscent of word formation, especially in derivation. Particles are much like derivational morphemes in that it is possible to see patterns, but these patterns or rules do not apply with strict regularity but show idiosyncracies of various kinds (see Riehemann 1998 for German bar-derivation). Particles could be described as free derivational morphemes.
Particles, then, show evidence in Modern English both of being phrasal (as in (9)) and of being heads (as in (10)). They are fascinating as a field of study precisely because they seem to straddle the no-man's land between syntax and morphology: separable, but in combination input to word formation processes. A unitary analysis that generates both orders (1a) and (1b) and still makes intuitive sense is almost impossible (see Elenbaas 2007 for a proposal, and for discussion).
The 'particle order' as in (1b) precludes a predicate analysis along the lines of (8) as the only option for particle-verbs, but does not preclude a complex predicate origin of the construction, with the 'particle syntax' the result of grammaticalization. We will discuss the parallels between particles and predicates in the next sections.

2.3 Predicate quirks

2.3.1 Unselected objects
Lipka (1972: 197-212) lists the various meanings of phrasal verbs with out and up, of which (11) and (12) represent a selection, and notes that "[s]ince the particle is omitted for reasons of brevity, it is readily apparent that many collocations which are possible with the V[erb]P[article]C[ombination] are excluded for the simplex verb" (1972: 215): the object cannot in many cases be selected by the verb on its own.

(11) with out: CAUSE + BE + /+ APPARENT/
blurt (secret), bring (meaning of a passage/ young lady, book), dig (book), dope (specifications), drag (reason), draw (scarf), ferret (secret), figure (problem), find (sb/sth – Deleted), fish (coin), flush (dollars/tax evaders), hunt (old diary, hat), haul (old essay), jerk (fish, pistol), lay (cold meal, evening clothes), nose (rat, trail/scandal, evidence), point (pictures, the man/mistake/that...), puzzle (sth), rake (scandal), reckon (how much we will need), root (truffles/possessions), rout (bottle), scare (partridge), search (friend/insincerity), seek (sb, place, book, keymen, enemy bombers), smell (sb, witch/secret, opposition), smoke (intentions), sound (sb), spell (views), spy (secrets/land), track (development), trot (horse/knowledge, excuse), whip (knife, wallet), worm (secret). (Lipka 1972: 200)

(12) with up: CAUSE + BE + /+ APPARENT/
Call (scenes from childhood), conjure (spirits, visions of the past), cough (sth), dig (statue), fetch (anecdotes), hunt (old records, references, quotations), look(fast train), plow (arrowheads/secrets), raise (prophet), rake (diary/scandal, old quarrels, past), reckon (bill), root (sb), scare (game), scout (clients), (dog) scratch (bone), show (fraud, ignorance/rogue, impostor), turn (facts in an encyclopedia). (Lipka 1972: 206-207)

The same phenomenon of the unselected object is seen in complex predication: cases in which V + predicate select a different set of objects than V would do on its own. An example is (7b) above: them does not have the same thematic relationship to cook as the objects have that cook selects outside of a resultative construction: a meal etc. Other examples are presented in (13):

(13) a. They drank the pub dry (Spencer & Zaretskaya 1998).
b. The bears frightened the daylights out of the campers (McIntyre 2001: 144)
c. I beat the dust out of the sofa (McIntyre 2001: 144)
d. He worked his fingers to the bone.

2.3.2 Idiomaticity
The idiomaticity of phrasal verbs, and their very variable degrees of transparency and productivity, is often noted in the literature (eg. Lüdeling 2001; see also the findings in Biber et al. 1999: 412-413). What is less well known is that they share these features with complex predicates: pry and come will only combine with complex predicates that mean something like 'apart,' drive will only combine with complex predicates denoting "negative and extreme mental states" (Goldberg and Jackendoff 2004: 559):

(14) a. He pried it apart/open/loose/free/*flat/*straight
b. It came apart/open/loose/free/*flat/*straight (Goldberg and Jackendoff 2004: 559)

(15) a. He drove her crazy/nuts/bananas/to desperation/to drink/up the wall/ meshuga/ frantic
b. *He drove her happy/sick/silly/clean/calm/thin/sober (Goldberg and Jackendoff 2004: 559)

2.3.3 Telicity
Another insight from Goldberg and Jackendoff's (2004) article is that resultative complex predicates need not be telic. They distinguish four types of complex predicate (538, 540):

(16) a. Noncausative property resultative (eg. The pond froze solid)
Syntax: NP1 V AP/PP2
Semantics: X1 BECOME Y2
MEANS: [verbal subevent]
b. Causative property resultative (eg. Willy watered the plants flat)
Syntax: NP1 V NP2 AP3
Semantics: X1 CAUSE [Y2 BECOME Z3]
MEANS: [verbal subevent, here: Willy watered the plants]
c. Noncausative path resultative (eg. The ball rolled down the hill)
Syntax: NP1 GO Path2
Semantics: X1 GO Path2
MEANS: [verbal subevent]
d. Causative path resultative (eg. Bill rolled the ball down the hill)
Syntax: NP1 V NP2 PP3
Semantics: X1 CAUSE [Y2 GO Path3]
MEANS: [verbal subevent, here: Bill rolled the ball]

Although only the two causative types of (16b) and (16d) would traditionally be termed complex predicates, they are clearly related to their noncausative counterparts (16a) and (16c). The types that grammaticalize, however, are the paths (c-d), expressed by prepositions, rather than the properties (solid, flat) expressed by adjectives.
Goldberg and Jackendoff (2004) demonstrate that the aspect and/or aktionsart of the complex predicate is determined by that of the causative subevent, which in turn hinges on whether the predicate sets up an endpoint to the event or not. As there are predicates that do not set up an endpoint, resultatives are not necessarily always telic; cf. John went along the river, where along the river expresses a path without a specific endpoint (2004: 543). This means that there are also stative and atelic resultatives, and, after grammaticalization, stative and atelic particle-verb combinations (eg. look on, float by etc.). The fact that there are particles that are not resultative does not argue against a predicate origin: complex predicates do not need to be resultative either. Prototypical predicates and particles, however, are resultative.
Farrell (2005: 118) notes that "[t]he resultative V-DP-Adj construction appears to have the same basic structure as the V-DP-P construction. The key difference is that only the latter typically has a compound-verb paraphrase (i.e. turn on the lights as well as turn the lights on)", – i.e. the difference between the 'particle order'as in (1a) above, and the predicate order as in (1b). That difference is, however, crucial. If some aspects of particle verbs can be said to have been inherited from their predicate origins, their grammaticalization has led to them being different in other respects, most importantly in the 'particle order', bleached meanings and the fact that they make reference to a very abstract 'Path', the precise semantics of which are filled in pragmatically. We will discuss this in the next section.


3. Grammaticalization

3.1 Path predicates grammaticalize

One aspect in which particles do differ from predicates, and which must be of primary importance in charting the grammaticalization process they have undergone, is the fact that particles show extensive semantic bleaching compared to predicates. Consider the following Lexical Conceptual Structure as typical of (resultative) predicates (from Spencer & Zaretskaya 1998, in turn based on Jackendoff 1990), with (17) illustrating a predicate construction as in (7a):

(17) [CAUSE[ACT (x)], BECOME [W(y)]], by[V(x)]]

(18) [CAUSE[ACT (the dog)], BECOME [awake(me)]], by[poking(the dog)]]

The by-phrase in this notation equals the verbal subevent in Goldberg and Jackendoff's (2004) notation above (in (15)), and particle verbs fit into this means or manner phrase with varying degrees of acceptability, in a large part depending on the verb (for non-causative verbs, eg. unaccusatives, the LCS needs to pruned somewhat (cf. (16a)-(16c) above)). If we classify the verbs that occur in particle-verb combinations in terms of whether they make the best fit in an LCS like (16) or (17), we find that verbs expressing manner make the best fit:

(19) (i) transitives
(ii) unergatives: chop, knock, laugh, sing, work
(iii) 'Manner of motion' unaccusatives: run, lope, sprint, dash, rush, hurry, scurry, scramble (Slobin 2005: 316)
(iv) denominal verbs, derived from the instrument used in causing the object y to reach the state W: boot out, bowl over, branch out, brick up, buckle up, elbow out, fork out, hand over, pan out, patch up
(v) deadjectival/ denominal verbs constituting a conversion of the state W itself: back off/away, brazen out, cheer up, clear up/out/off/away, crack up, free up, gloss over, open up/out, parcel out, pretty up, round up/off
(vi) 'Light' verbs:
a. transitive: get, keep, let, make, place, put, set
b. unaccusative: come, go

Many transitives that express some activity like cooking and poking in (7) fit nonproblematically in the manner slot, and the same goes for the intransitive unergatives, in (ii).

(20) a. He chopped the tree down
b. [CAUSE[ACT (he)], BECOME [down (tree)]], by[chopping(he)]]

Manner-of-motion verbs in (iii) make good fits, too (in the non-causative pruned version of the LCS). The denominal verbs in (iv) specify manner too, but do not have an independent existence outside the particle verb combination, and they fit the LCS less well for that reason (eg. (21)):

(21) a. They elbowed me out
b. [CAUSE[ACT (They)], BECOME [out (me)]], by[elbowing(They)]]

When the verb does not specify manner, but is a conversion of the state W itself (the verbs in (v)) or a 'light' verb, as in (vi), they do not fit the LCS. The fact that specifying manner is apparently important for a good fit is interesting, as the manner-of-motion verbs in (iii) and the denominal verbs in (iv) are only robustly attested from EModE onwards, as we will see in section 4.3).
The point to note is that the less than perfect fit of some of these particle verbs also comes to the fore when these same verbs appear with genuine syntactic predicates, and is due to the contribution of the verb rather than to that of particle or predicate. What is different between particles and predicates is the degree of explicitness of W. The particle appears to be bleached to a degree that does not seem possible with genuine predicates. Adjectives, expressing properties, are less likely to bleach; but paths, with or without endpoints, may do so very easily. There is a clear link here with productivity: spatial resultatives appear to be totally productive in that any spatial PP that can be construed as a path can be used as a complex predicate (Goldberg and Jackendoff 2004: 558); APs such as free (of NP), clear (of NP), apart, open and shut, i.e. exactly the set of APs that may grammaticalize (see (3a)), are also fairly freely productive and are argued to be interpreted as "spatial being open configurations with some force-dynamic overtones" (Goldberg and Jackendoff 2004: 558). They do not only represent a property but a spatial configuration "affording free passage between the interior and exterior of the object" (Goldberg and Jackendoff 2004: 559). This insight, then, allows us to postulate the generalization that only path-predicates will grammaticalize.

3.2 The defocused complement of Prt

Paths leave much more room for pragmatic interpretation than properties, and part of the grammaticalization process is the fact that the complement of the preposition becomes, in McIntyre's (2004) words, 'defocused': in he took the ring off, the object can be pragmatically reconstructed with ease given our knowledge of rings and fingers; in other cases, the exact identification or reconstruction of the ground is either unimportant or infelicitous (cf. also the examples in (6) above):

(22) a. She took a newspaper in (=into the house).
b. She threw the remains of the dinner out (=out of the house).
c. Bill pushed Harry along (=along the trail) (Goldberg and Jackendoff 2004: 543)
d. Wait Long by the River and the Bodies of Your Enemies Will Float By (=by one's position; title of CD album by The Drones)

This phenomenon could be interpreted as loss of argument structure, on a par with the loss of argument structure we observe in the grammaticalization of verbs into auxiliaries. There is, however, a caveat here in that prepositions may not have a syntactic argument at an earlier stage but were located in the specifier of an NP in the local cases (instrumental, ablative, locative). They developed into prepositions only later (Vincent 1999):

(23) KP (24) PP

Spec K' Spec P'

K NP P NP
preposition Spec N' Spec N'
adposition
N N

Particles may well have split off from prepositions when the latter were still in the 'adposition' stage, and may never have had proper syntactic arguments – they were associated with a particular NP by virtue of occupying its specifier position. Particles like down, out, off and up may have acquired their prepositional use only recently: down is originally a PP that grammaticalized to a head, while OE ut and up do not show clear prepositional uses but are usually followed by prepositional phrases. What is clear is that the defocusing of the object allows the particle to be analysed as a Head rather than a Phrase, a typical grammaticalization effect.
Because the syntactic status of some of these NP complements of path-Ps is unclear, I will refer to this NP as the ground, a semantic rather than syntactic term, and to the object NP of the particle verb combination as the figure, following Svenonius (2003), after Talmy (1978). In a sentence like he took the ring off his finger, then, the ring is the figure and his finger is the ground of the particle off.

(25) he took the ring off his finger
figure prt ground

In particle verbs that result from a grammaticalization process of complex predicates, it is the figure that becomes the fully affected object of the particle verb combination; that figure traverses the path expressed by the particle.

3.3. Grounds as fully affected objects

In addition to the figure of the path becoming the object, there is another pattern in which it is the ground that surfaces as the object of the particle verb combination. An example is (26), which was discussed by Denison (2004) as an example of reanalysis: the preposition increasingly attaches itself to the verb, and its former complement (the ground) becomes the object of the particle verb combination.

(26) My car ran over a bottle (lying in the road)
[VP ranintr [PP over [NP a bottle]]]
[VP rantrans [part over] [NP a bottle]] (Denison 2004: 18)

It is another resultative predicate pattern, with its unaccusative verb conforming to the LCS in (16c), Goldberg and Jackendoff's noncausative path resultative. My car is here the figure, and starts out initially as the object of the verb (because the verb is an unaccusative). The earliest literature on particle verbs mention the phenomenon of object transfer ('Subjektvertauschung', 'Objektvertauschung': Hundsnurscher 1968: 124ff quoted in Lipka 1972: 94). Compare the object of the first and second of the following pairs, of which the first one is the figure of the particle, but the second one is the ground: water runs or pours out of the bucket, we brush the lint off a coat, rinse the dirt off the plates.

(27) a. das Wasser läuft aus/der Eimer läuft aus (Lipka 1972: 94)
the water runs out/the bucket runs out
b. John poured out the water/John poured out the bucket (McIntyre 2001)
c. Clear out mud (from a river)/clear out a river (by removing mud) (Lipka 1972: 94)
d. Brush the lint off/brush the coat off (Farrell 2004: 110)
e. Het vuil afspoelen/ de borden afspoelen (Blom 2005: 190)
the dirt off-rinse the plates off-rinse
'rinse off the dirt' 'rinse off the plates'

Synchronic similarities point to an affinity between particles and predicates. Could they point to a diachronic relationship? We will now look at the situation in earlier English.

4. Earlier English

4.1 Evidence for grammaticalization

OE particle verbs generally fit the resultative semantics of predicates, even more so than their PE counterparts because they are almost invariably transparent combinations of (transitive and unaccusative) verbs and particles, as in the following examples (both from Elenbaas 2007):

(28) þæt hi hine ut sceoldon wurpan (coeust, LS 8 (Eust) 168.173)
that they him out should throw
'that they should throw him out'

(29) & ærn swa feor up swa næfre ær ne dyde (ChronE (Plummer) 1014.28.1906)
and ran as far up as never before not did
'and ran up as far as he never did before'

Of all the verb groups listed in (19) only those in (vi) 'light' verbs, both transitive and unaccusative,occur robustly with particles in OE. Notably absent are the denominal or deadjectival conversions, and the finer-grained 'manner-of-motion' like run, jog, lope, sprint, dash, rush, hurry, scurry, scramble (Slobin 2005: 316). We will see that they only start to appear in EModE (Section 4.3).
Saying anything definite about the grammaticalization process is not easy. If we interpret particles that are heads rather than phrases as the result of grammaticalization, we could try to find evidence that they are exclusively phrases in OE and become heads only later on. Example (29) with up being premodified by swa feor 'so far' shows that OE particles certainly allow a phrasal analysis, but it cannot be established on the available data that they are exclusively phrasal. Even for PDE, most analyses acknowledge that particles can be heads and phrases (as evidenced by the fact that they often allow premodification by elements like just and right in the 'predicate order' but not in the 'particle order'). Evidence of verb raising clusters, where we might adduce head status if the particle is found adjacent to its verb (i.e. orders like þæt hi hine sceoldon ut wurpan, cf. (28) above), cannot be conclusive because of the possibility of Verb Projection Raising in OE. The only firm evidence for grammaticalization of predicate phrases into heads comes from the EModE examples of PPs like those in (3b) grammaticalizing. Claridge (2000: 138-140; 158) has some examples in her EModE corpus of the NP in these grammaticalizing PPs still occurring with some degree of premodification: take NP into [your most grave and wise] consideration.
If the order of (1b) with the verb and the particle adjacent can be analysed as a grammaticalized order in PE (with the particle as a head), can we perhaps deduce anything from the occurrence of this order? The grammaticalized 'particle order' is already the most frequent order in the first subperiod of the Middle English part of the Helsinki Corpus, and becomes even more frequent in the subsequent ones (Elenbaas 2007: 260). Biber et al. (1999: 932) report that with full noun phrases, the 'particle order' in written registers occurs in over 90% of all cases; in conversation, the rate is much lower (about 60%), although it is still the most frequent order. There are many other factors governing the selection of the 'particle' or 'predicate order', including focus (see also Dehé 2002), end-weight (extraposition of heavy objects) and idiomaticity (as was discussed in Section 1 above), whereas the extraposition of genuine syntactic predicates (as in (1d)) appears to be governed by only one factor, namely end-weight (Biber et al. 1999: 931). In all, the word order findings appear to tally with the hypothesis that the (1b) order is the result of grammaticalization. The fact that particle verb combinations with eg. up and ut are invariably transparent and spatial in OE (eg. Denison 1985, Elenbaas 2007), then, probably means that they are phrases rather than heads.

No predicate quirks in OE and ME?

A typical 'quirk' of particles and predicates that we discussed in section 2.3.1 was that of the 'unselected object', a very creative use of the resultative construction that we saw in (13a-c) where the particle-verb or predicate-verb combination occurs with an object that could not occur with the verb on its own. No cases of unselected objects in OE were found, but this is not surprising in view of the fact that for languages without native speakers we cannot rely on our intuitions of which objects verbs typically take and which objects are unselected, as we did in the case of (13a-c) where we know that the objects are unselected because you can't drink pubs, frighten daylights, or beat dust. To identify unequivocal unselected objects in a dead language like OE we have to rely on unergative (i.e. truly intransitive) verbs that cannot take any object on their own (as PDE work in (13d), which, though unergative, occurs with an object when there is a particle or predicate present: he worked out a solution, he worked his fingers to the bone). Unfortunately, OE particles seem to occur exclusively with unaccusative and transitive verbs, as we saw in the previous section, and I was not able to identify any cases of particles and objects occurring with unergative verbs. Unselected objects, then, are not a feature of OE particle verbs, but, interestingly, they are not a feature of genuine syntactic predicates in OE either – these, too, occur, as far as I have been able to establish, mainly with 'light' verbs (the OE counterparts of PDE come, get, go, keep, let, make, place, put, set – see Section 3.1 above). In (30) we have the light verb gedon 'do':

(30) þu ne miht ænne locc gedon hwitne oððe blacne
Lat. non potes unum capillum album facere aut nigrum
'you cannot turn one hair white or black'

A search of typical unergatives (the OE counterparts of verbs like dream, laugh, sing and work) did not bring up any combinations with predicates. Neither predicates nor particles, then, are used as creatively as they are in PDE.
Nor does ME yield much evidence of unergative verbs with either a complex predicate or a particle; neither the Middle English part of the Helsinki Corpus, the Middle English Dictionary or the OED offer any examples under unergatives like dream, laugh, sing and work, apart from variations on the idiom laugh NP to scorn (to hokere, to bismare, til/at/into hething), and the instance in (31) from the OED and the MED:

(31) þey haue an herbe ... þat makeþ men laughe hem selue to deþ (Trevisa Higden (Rolls) I.305)

We do get instances like (323), but they are probably postpositional rather than true particle-verbs:

(32) ofte he hire loh to (a1225 (?a1200) Lay. Brut 18542)
often he her laughed to
'he often laughed at her'

The phenomenon of unselected objects can be found, but in the prefixed verb in OE, not in the particle verb. Prefixes represent an older layer of grammaticalization, very much akin to the later particle system (see also Claridge 2000: 87), in which the grammaticalized element has become a bound morpheme, inseparable from the verb. It has 'frozen' in preverbal rather than postverbal position because early Germanic was an OV-language which at that stage probably did not have the verb-second operation that separate particle and verb in OE, Modern Dutch and Modern German. This allowed the grammaticalization process to proceed to its logical conclusion: the predicate and verb became a single lexical item (van Kemenade and Los 2003). These verbal prefixes have long been recognized as 'transitivizing' and even changing the case of the object. Compare for instance hliehhan 'laugh' which is an unergative verb but may occur with the object that is laughed at in the genitive, as in (33), with behliehhan 'deride' (lit. 'be-laugh') which is a fully transitive verb with its object in the accusative (as in (34)).

(33) ðonne we hliehað gligmonna unnyttes cræftes. (CP 34.231.4)
when we laugh jesters-gen useless tricks
'when we laugh at the useless tricks of jesters'

(34) Huru, ic swiðe ne þearf hinsiþ behlehhan (Guthlac 87: 1356-1357)
Indeed, I much not want departure-acc deride
'Indeed, I do not want to laugh at his death'

If we assume that the prefix be- was once the predicate of an Agreement Phrase in a structure like (8), with its subject (which later became the object of the prefixed verb) in the specifier, this subject would have had accusative case (the default case of subjects of verbless or non-finite constructions). This case is not mediated by the verb but by the predicate. The transitivizing effect of these prefixes, then, is an inheritance from their predicate origins. Deadjectival and denominal verbs, as in (21), which often do not have a simplex, are also a feature of prefixed verbs (van Kemenade and Los 2003), though not of OE or ME particle verbs.
We must conclude that predicate and particle quirks are not much in evidence in OE and ME. The first blossomings of both appear to date from the EModE period.

4.3 Particle verbs in EModE

Historical studies on particle-verb combinations in English seem to agree that the PDE situation with respect to particle verb combinations is reached in the EModE period, with the 16th and 17th centuries representing a high point in their development (eg. Brinton 1988; for phrasal verbs in Shakespeare, see Castillo 1994; see also Claridge 2000: 96-98 and the studies cited there). A marked difference with OE and ME is the appearance of other groups of verbs than transitives or 'light' verbs (eg. groups (i) and (vi) listed in (19). In (35) and (36) we have EModE examples of denominal verbs and in (37) an example of a deadjectival verb:

(35) Goe, sayes hee; trusse up your trinkets and be gone. The cooke, seeing no remedy, departed. (Helsinki Corpus: Robert Armin, A nest of Ninnies, p. 14, 1608)

(36) There was in the time of Will Sommers another artificiall foole, or jester, in the court, whose subtiltie heapt up wealth by gifts giuen him, for which Will Sommers could neuer abide him (Helsinki Corpus: Richard Madox's diary, p. 47, 1582)

(37) At 12 the east wynd began to fresh up which caused us to way upon the eb, but before we wer passed a lege yt faynted and we wer fayn to cast Anchor. (Helsinki Corpus: Richard Madox's diary, p. 135, 1582)

Other examples of such verbs appearing in combination with the particle up in the EModE part of the Helsinki Corpus were:

(38) clitch up, clap up, sprout up, reckon up, block up, treasure up, burble up, nuzzle up ('indulged'), prune up, knit up ('joined together in friendship'), mould up, dry up, rip up ('?think up'), dress up, work up, seal up, pluck up, clamber up.

Chronologically, then, there is a difference between the appearance of sets (vi) of the list in (19) ('light' verbs like the transitives set, turn, make, let etc.) and unaccusatives that can be described as 'core' motion verbs like rise, come, go etc. and the other sets. Set (vi) has appeared in particle-verb combinations since OE. The other sets, (i)-(v) – most notably manner of motion verbs like clamber, trip, wander, unergatives like work, sing, laugh, play, and denominal/deadjectival verbs – are very rarely found in OE in a particle verb combination, if at all, and they are the ones that exhibit the predicate quirks to the greatest advantage: unselected objects, verbs that are only attested in the particle-verb combination but have no independent existence, denominal and deadjectival verbs, idiomatic combinations. I have split these verbs up into 2 groups: Category 1 comprises set (vi), the set that has been attested with particles from the earliest period, whereas Category 2 comprises sets (i-v) that only start to appear robustly with particles in EModE. If we look at occurrences with the particle up in the EModE part of the Helsinki Corpus, we find both categories well represented in all three periods, although the share of Category 2 verbs goes up slightly (29.6 % of the total in E1, 30% in E2 and 32.3 % in E3). There is an increase in the overall numbers, too (Figure 1).

@@Insert figure 1 here

The denominal and deadjectival verbs of Category 2 were found in travel writings, diaries, plays, autobiographies and in one text on education, which seems to indicate that they are more prominent in informal genres; the size of the corpora are too small, however, to say anything definite. Biber et al. (1999: 407-413) focus primarily on the verbs and particles that have the highest frequencies, and are therefore almost exclusively Category 1 verbs, so that these tentative register findings for EModE cannot be compared to the PDE situation.
With respect to predicates the findings were consistent with the earlier situation in OE and ME: there were very few genuine syntactic predicates with verbs other than light verbs (turn, make etc.); I was unable to find any examples of the more adventurous complex predicates as discussed in Goldberg and Jackendoff (2004), or Rappaport Hovav & Levin (2001) (see (7) above). Whether it is the size of the corpus (with over 900,000 words about twice the size of the EModE part of the Helsinki Corpus), or the authorship (a creative writer taking a syntactic construction to extremes) , the situation is very different in Shakespeare's plays. Imaginative examples of various constructions abound. A search for the usual suspects (the unergative intransitive verbs dream, laugh, sing, work) immediately turns up genuine syntactic predicates with unergatives like the ones in (39):

(39) CLEOPATRA·
That time, – O times! –
I laugh'd him out of patience; and that night
I laugh'd him into patience; and next morn,
Ere the ninth hour, I drunk him to his bed;
Then put my tires and mantles on him, whilst
I wore his sword Philippan. (Shakespeare, Anthony and Cleopatra, Act II, Sc. v).

It seems, then, that the more striking parallels between predicates and particles that can be gleaned from Goldberg and Jackendoff (2004), particularly the phenomenon of the unselected object, are only in evidence from EModE onwards. Lipka (1972) states that not all languages that have particle verb combinations have deadjectival and denominal verbs; some, like Japanese, only have deverbal verbs, and OE seems similar in having only verbs that are attested as verbs independently, i.e. also as a 'simplex'; it is one of the peculiarities of PE (and the modern West-Germanic languages in general) that many particle verbs have no simplex (eg. peter out, brazen out, pretty up and many others). It looks as if English was more like Japanese in this respect in earlier times, as it is not until EModE that we start to find unergatives, manner of motion verbs, denominal and deadjectival verbs in any large numbers.
Talmy (1985) has proposed a binary typology along the dimension path-expressions. There are, he claims, verb-framed languages and satellite-framed languages; the former describes paths by a 'path' verb like exit while the latter describes paths by an element associated with the verb, like a particle or prefix; this explains why some languages have particle or prefix verbs and others do not. Slobin (2005, 2006) argues that satellite-framed languages typically have many more types of manner verbs than verb-framed languages, and that there may well be a diachronic dimension: because 'manner' is so easily encoded in these languages, there is "over time – a predisposition to attend to this domain" (Slobin 2005: 316). The more fine-grained the distinctions become, the more learners are geared to making these fine distinctions, which ultimately leads to impressive lists for, say, "types of rapid bipedal motion" like run, jog, lope, sprint, dash, rush, hurry, scurry, scramble etc. (Slobin 2005: 316). A count of innovative manner of motion verbs per century, based on the OED, seems to show an increase from 1500 onwards, but such results are difficult to interpret correctly because English was so intensely relexified (see Slobin 2006: 72). Manner of motion verbs are a good fit in an LCS like (17), and lead to expressive particle-verbs; they may account to some degree for the marked difference between the ME and the EModE periods. Although the origin of the rising popularity of the other Category 2 verbs seems less clear, many of them also encode 'manner', as we discussed in Section 3.1, and are consequently good fits in the LCS. It is this that may hold the key to their appearance in EModE.


5. Conclusion

The semantic and syntactic similarities of complex predicates and particles point to a common origin, with particles having grammaticalized from phrase to head. The order V – prt – NP shows this grammaticalization most clearly. The most striking quirks of the complex predicate construction, as observed by Goldberg and Jackendoff, i.e. unselected objects and idiomaticity (fixed combinations of verb and predicate), are also shared by particle verbs. The conclusion that these shared quirks are similarly the result of shared diachronic origins is problematic because they appear late, in EModE, both in particle verbs as in complex predicates. EModE seems to mark a significant point in the development of the particle verb system in that the verbs participating in the combination are no longer restricted to 'light' verbs but include deadjectival and denominal verbs, unergatives, and 'manner-of-motion' verbs.


References
Anthony, E.M. Jr. (1953). Test frames for structures with UP in modern American English. Ann Arbor: University Microfilms.
Asbury, Anna. (2005). "Adpositions as case-realisations." Leiden Papers in Linguistics 2.3, edited by Martin Salzmann and Luis Vicente, 69-92.
Biber, D., S. Johannsson, G. Leech, S. Conrad and E. Finnegan (1999). The Longman Grammar of Spoken and Written English. Harlow: Longman.
Blom, C. (2005). Complex Predicates in Dutch: Synchrony and Diachrony. LOT: Utrecht.
Bolinger, D. (1971). The Phrasal Verb in English. Cambridge, Mass.: Harvard University Press.
Brinton, L.J. (1988). The Development of the English Aspectual system: Aspectualizers and Postverbal Particles. Cambridge: Cambridge University Press.
Brinton, L.J. (2002). "Grammaticalization versus lexicalization reconsidered." In: English Historical Syntax and Morphology (Current Issues in Linguistic Theory Series 223), edited by Teresa Fanego, María José López-Couso and Javier Pérez-Guerra, 67-97. Amsterdam/ Philadelphia: Benjamins.
Castillo, C. (1994). "Verb-particle combinations in Shakespearean English: A syntactic study." Neuphilologische Mitteilungen 95: 439-51.
Claridge, C. (2000). Multi-Word Verbs in Early Modern English: A Corpus-Based Study. Amsterdam: Rodopi.
Dehé, N. (2002). Particle Verbs in English: Syntax, Information Structure, and Intonation. Amsterdam: John Benjamins.
Denison, D. (1981). Aspects of the History of English Group-verbs, with particular Attention to the Syntax of the Ormulum, diss. Oxford University.
Denison, D. (1985). "The origins of completive up in English." Neuphilologische Mitteilungen 86, 37-61.
Denison, D. (2004). "Do grammars change when they leak?" In: New Perspectives on English Historical Linguistics, Vol. I: Syntax and Morphology, edited by Christian Kay, Simon Horobin and Jeremy Smith, 15-29. Amsterdam: John Benjamins.
Elenbaas, M. (2007). The Synchronic and Diachronic Syntax of the English Verb-Particle Combination. Utrecht: LOT.
Emonds, J. (1976). A Transformational Approach to English Syntax. New York: Academic Press.
Emonds, J. (1985). A Unified Theory of Syntactic Categories. Dordrecht: Foris.
Fairclough, N.L. (1965) Studies in the Collocation of lexical Items with Prepositions and Adverbs in a Corpus of spoken and written Present-day English. MA thesis, University College London.
Farrell, D. (2005). "English verb-preposition constructions: constituency and order." Language 80: 96-137.
Fraser, B. (1965) An Examination of the Verb-particle Construction in English. Dissertation MIT.
Goldberg, Adele E. (1995). Constructions. Chicago: University of Chicago Press.
Goldberg, Adele E., and Jackendoff, R (2004). "The English resultative as a family of constructions". Language 80: 532-568.
Hundschnurscher, F. (1968) Das System der Partikelverben mit AUS in der Gegenwartsprache. Dissertation Universität Tübingen.
Jackendoff, R. (1990) Semantic Structures. Cambridge, Mass.: MIT Press.
Kemenade, A. van and B. Los (2003). "Particles and prefixes in Dutch and English." In: Yearbook of Morphology 2003, edited by G.E. Booij and J. van Marle, 79-117. Dordrecht: Kluwer.
Lipka, L. (1972) Semantic Structure and Word Formation: Verb-particle Constructions in Contemporary English. München: Wilhelm Fink.
Los, B. (2004). "From resultative predicate to event-modifier: The case of forth and on". In: Proceedings of the 12th International Conference on English Historical Linguistics, edited by Christian Kay, Simon Horobin & Jeremy Smith, 83-102. Amsterdam: John Benjamins.
Lüdeling, A. (2001). On Particle Verbs and Similar Constructions in German. University of Chicago Press: Chicago.
McIntyre, Andrew. (2001). "Argument blockages induced by verb particles in English and German: Event modification and secondary predication". In: Structural Aspects of Semantically Complex Verbs, edited by N. Dehé & A. Wanner, 131-164. Berlin: Peter Lang.
McIntyre, Andrew. (2004). "Event paths, conflation, argument structure, and VP shells". Linguistics 42, 523-571.
The Middle English Dictionary (1952-). Edited by Hans Kurath. Ann Arbor: University of Michigan Press.
The Oxford English Dictionary on CD-ROM (1989) 2nd edition. Ed. by J. A. Simpson & E.S.C. Weiner. Oxford: Oxford University Press.
Quirk, R., S. Greenbaum, G. Leech & J. Svartvik (1985) A Comprehensive Grammar of the English Language. London: Longman.
Rappaport Hovav, M., and B. Levin (2001). "An event structure account of English resultatives". Language 77.766-797.
Riehemann, S, Z. (1998). "Type-based derivational morphology."Journal of Comparative Germanic Linguistics 2, 49-77.
Slobin, D. I. (2005). "Linguistic representations of motion events: What is signifier and what is signified?" In: Outside-In – Inside-Out, edited by Costantino Maeder, Olga Fischer & William J. Herlofsky, 307-322. Amsterdam: John Benjamins.
Slobin, D.I. (2006). "What makes manner of motion salient? Explorations in linguistic typology, discourse and cognition." In: Space in Languages: Linguistic Systems and Cognitive Categories, edited by Maya Hickmann & Stéphane Robert, 59-81. Amsterdam: John Benjamins.
Spencer, A. & Zaretskaya, M. (1998) Verb prefixation in Russian as lexical subordination. Linguistics 36, 1-39.
Sroka, K. A. (1972). The Syntax of English Phrasal Verbs. (Janua Linguarum 129).The Hague : Mouton.
Svenonius, P. (2003). "Limits on P: Filling in holes vs. falling in holes". In Proceedings of the 19th Scandinavian Conference on Linguistics (Nordlyd 32), edited by Anne Dahl, Kristine Bentzen and Peter Svenonius, 431-445.
Talmy, L. (1978). "Figure and Ground in complex sentences". In: Universals of Human Language, edited by Joseph Greenberg, 625-649.Stanford, CA: Stanford University Press.
Talmy, L. (1985). "Lexicalization patterns: Semantic structure in lexical forms." In: Language Typology and Lexical Description, Vol. 3, Grammatical Categories and the Lexicon, edited by T. Shopen, 36-149. Cambridge: Cambridge University Press.
Vincent, N. (1999). "The evolution of c-structure: Prepositions and PPs from Indo-European to Romance." Linguistics 37: 1111-1153.
Zeller, J. (1999) Particle verbs, Local Domains, and a Theory of Lexical Licensing. Diss. Johann-Wolfgang-Goethe-Universität, Frankfurt am Main.