Against overt particle incorporation Jochen Zeller (Johann Wofgang Goethe-Universitaet Frankfurt) Verb-particle constructions in Dutch and German are an interesting borderli= ne case between morphology and syntax. On the one hand, they allow further=20 adjectival and nominal derivation: (1)=09German:=09Einarbeitung (in-work-ing; "training") =09=09unannehmbar (un-at-take-able; "inacceptable") (2)=09Dutch:=09uitsterving (out-die-ing; "extinction") =09=09opmerkzaam (up-notice-able; "attentive") On the other hand, particle verbs are syntactically transparent. For exampl= e, they can be separated from the verb in verb second (3) and verb raising=20 (VR)(4) contexts: (3)=09=09German: Peter laedt das Heu ab =09=09=09Peter loads the hay off (4)=09=09Dutch: =09dat Jan Maria op ti heeft probeeren te belleni =09=09=09that Jan Mary up has tried to call There is an ongoing debate in the literature whether particle verbs=20 should be analysed syntactically or morphologically. On the syntactic=20 side, proponents of the SC- approach (Kayne 1985, Hoekstra 1988, den=20 Dikken 1995, among many others) argue that particles are generated in a=20 syntactic head position and (optionally) combine with the verb via head=20 movement. Keyser/Roeper (1992) postulate an abstract clitic position=20 below V=F8 that can be occupied by a particle. On the morphological/lexical= =20 side, it has been argued that particle verbs are generated in the lexicon= =20 and inserted as complex V=F8s (Koster 1975; Bierwisch 1987; Johnson 1991;= =20 Neeleman/Weerman 1993; Wunderlich/Stiebels 1994).=20 =09In this paper I will argue that 1.) particle verbs are represented=20 syntactically, but that their underlying structure can be the input for=20 morphological word formation rules and that 2.) the particle remains=20 inside the (SC-)complement of the verb in overt syntax (provided that=20 Dutch and German are SOV) and does not incorporate before LF. I call the=20 latter proposal the Covert Incorporation Approach (CIA) to particles and advance the following arguments against the common view that particle and= =20 verb form a complex verbal head at some stage in overt syntax: (i) If a particle verb is analyzed as a complex V=F8, additional stipulations are required to explain why the particle cannot move with=20 the verb to C=F8 in verb second. Furthermore, such an analysis entails that= =20 a segment can be moved out of the particle verb, leaving a trace inside=20 the verbal head. However, it has been argued that excorporation is not a=20 legitimate instance of move alpha (Baker 1988; Kayne 1991). If one assumes= =20 alternatively that particle incorporation is optional (van Riemsdijk=20 1978) and does not apply in the case of verb second, it remains unclear=20 why overt particle incorporation is necessary at all. Particle stranding=20 in verb second is a natural consequence of the CIA, however, which does=20 not require excorporation, since the particle does not incorporate in=20 overt syntax. (ii) The particle always precedes the infinitival markers zu/te in German= =20 and Dutch:=09 (5)=09(a)=09German:=09abzuladen ("to load down") =09(b)=09Dutch:=09optebellen ("to call up") Since the infinitival marker in German and Dutch is generated in Infl=20 (Beukema/den Dikken 1989; Sabel 1996), the derivation of infinitival form= =20 of particle verbs either violates the Head Movement Constraint (the=20 particle skips the trace of the verb and left-adjoins to Infl) or=20 requires excorporation of a V-segment into Infl, followed by=20 incorporation of the remnant verbal head including the particle. In=20 contrast, the CIA predicts that the particle precedes the infinitival=20 marker, since it remains inside the SC when the verb moves to Infl. (iii) The intonation pattern of particle verbs differs radically from=20 that of other complex verbal heads. In complex verbs derived by=20 preposition incorporation, the stress is always on the verb. In=20 verb-particle constructions, however, the particle always bears stress=20 and acts like the head of other phrasal complements in this respect: (6)(a)ueber'fliegen (fly over)=09(b)'abfliegen (leave flying) (c)nach Hause= =20 =09=09=09=09=09=09=09=09fliegen =09over(Prep)-fly=09=09=09off (Prt)-fly=09 "fly home" (German) (6) is difficult to explain under the assumtion that both particle verbs=20 and prefixed verbs are complex V=F8s. However, since the particle is the=20 head of a SC-complement of the verb even in overt syntax, the=20 similiarities between (6)(b) and (c) follow naturally from the CIA. So far, the CIA explains the properties of particles in a much more=20 elegant way than an overt incorporation approach. Notice, however, that=20 no lexical material can intervene between the verb and the particle if=20 the verb has not moved further than Infl: (7)German: (a) Peter will ein Mauskript ti abschicken [von Chomsky]i=20 =09 Peter wants a manuscript off-send by Chomsky =09 (b)=09*Peter will ein Manuskript ti ab [von Chomsky]i schicken If extraposition involves right-adjunction to VP, the CIA does not=20 explain how (7)(b) is ruled out. However, in Truckenbrodt (1994) it is=20 argued that extraposition is subject to phonological restrictions. A=20 prosodic category can never be split by an extraposed phrase. Since=20 particle and verb, although syntactically separated, always form a=20 prosodic category, cases like (7)(b) are excluded on phonological=20 grounds. It is not necessary to assume that particle and verb form a=20 complex head. The fact that adjectives and nouns can be derived from particle verbs is= =20 compatible with the CIA. I assume with Lieber (1992) that the word=20 formation component has access to phrases as well as to X=F8-categories.=20 The nouns and adjectives in (7) are derived N=F8 and A=F8 heads which inclu= de=20 VPs. This view also allows it to associate the complex structure of a=20 particle verb with a single idiosyncratic meaning. The lexical entry=20 specifies a non-compositional meaning of the whole VP dominating verb=20 and particle. In this respect, particle verbs are similiar to phrasal=20 idioms and should be analyzed along the same lines (as suggested by=20 Jackendoff 1975, 1994, forthcoming). The strongest challenge to my proposal is VR in Dutch. If an embedded=20 infinitival form of a particle verb attaches to the matrix verb, the=20 particle can also move with the verb: (8) Dutch: =09dat Jan Maria ti heeft probeeren op te belleni =09=09that Jan Mary has tried-IPP up to call The past participle in (8) is replaced by the infinitival form of the=20 verb (the infinitivus pro participio-effect; IPP). This is taken as an=20 indication that matrix verb and embedded infinitive indeed form a complex= =20 head (den Besten/Rutten 1989). If this verb cluster is really derived by=20 verb raising, it contradicts my claim that the particle does not=20 incorporate overtly into the verb. However, note that so-called verb=20 projection raising (VPR) in Flemish (den Besten/Edmondson 1983;=20 Haegeman/Riemsdijk 1986; Wyngaerd 1989) has always posed a problem for a=20 verb raising analysis. In Flemish, non-verbal material can also appear in= =20 the verb cluster: (9)=09Ik ti heb willen [dat huis kopen]i =09I have wanted-IPP that house buy =09 Furthermore, Hoeksema (1991) has argued that VPR is still part of the=20 basic grammar of Standard Dutch (it still shows up in spoken Dutch). It=20 seems that the problem raised by (8) is not particular to my analysis.=20 Tentatively, I suggest the following solution: the verbal clusters in (8)= =20 and (9) are morphologically derived V=F8s; the matrix verb functions as a= =20 verbal affix similiar to the adjectival and nominal affixes that derive=20 nouns and adjectives from particle verbs in (1) and (2) (Bierwisch 1990).= =20 This view is supported by a yet unexplained observation: A particle=20 prevents a verb with an infinitival complement from triggering VR of the=20 embedded verb (van Riemsdijk 1978): (10) Dutch: (a)=09 omdat Jan de auto ti begon te wasseni =09 (b)=09 *omdat Jan de auto ti aanfing te wasseni =20 =09=09 that Jan the car began to wash=20 Since particle verbs are VPs, according to the CIA, they cannot function=20 as verbal affixes, and (10)(b) is ruled out.