- 1 Many thanks are due to my colleague Katarina Bartkova for her judgements on Hungarian, and the two (...)
1The study aims at examining the relationship between nominalization and aspect in Hungarian. In this language, aspect in the verbal domain is mostly expressed by the means of preverbs, a class of preverbal detachable particles. As we will see, preverbs are maintained in eventive derived nominals. Consequently, we examine and compare their syntactic and semantic behavior in verbal and nominal contexts. We show, after Knittel & Forintos-Kosten (2002), that preverbs are markers of boundedness, both for inner/lexical and outer/grammatical aspect. At the level of lexical aspect, they indicate a bound in the process, and appear as telicity markers; when modifying grammatical aspect, they mark perfectivity. We show that these values are maintained in case of nominalization. Consequently, derived event nominals in Hungarian keep the aspectual properties of the corresponding verbs, for inner as well as outer aspect, thus corroborating the idea, put forward by several authors (Engelhardt 2000; Alexiadou, Iordachioaia & Soare 2010; Iordachioaia & Soare 2008, 2009; Knittel 2011) that outer aspect is a valid category for event nominals.
2This article is organized as follows. In section 2, we examine preverbs in the verbal domain. We first discuss their aspectual values; then we describe their syntactic behavior and we provide a syntactic analysis accounting for it. Section 3 is dedicated to nominalizations. We present first the syntactic peculiarities of event nominals in Hungarian, and, turning to the study of preverbs in deverbal nominals, we show that their aspectual properties parallel those observed in verbal environments. Consequently, we suggest syntactic analyses accounting for the various cases observed.
3Preverbs are a class of preverbal elements that differ from prefixes by their detachability. Examples of sentences containing preverbs are given in (1-4):
4As shown by their translations in the above examples, the initial value of the preverbs át, el and be (vs meg (3)) is directional or locative. However, they interfere in the sentence in various ways. In (1), the preverb is responsible for a change in the argument structure, since it forces the presence of the locative complement a hídon ‘the bridge’; the same holds for examples (5), where be enables the presence of an accusative object (Nyéki, 1988: 144). In (2), the preverb el changes the verb meaning: whereas alszik means ‘to sleep’, elalszik is ‘to fall asleep’. Another case of meaning change is illustrated in (6).
5Contrary to the other examples, the preverbs in (3-4) have only an aspectual value, and are respectively a telicity (3) and a perfectivity (4) marker.
6In what follows, we will focus on the aspectual role of preverbs. A more complete description can be found in Kiefer (1994).
7According to Knittel & Forintos-Kosten (2002), preverbs can influence both inner aspect (i.e. lexical aspect or Aktionsart) and outer aspect, that is, grammatical aspect, or viewpoint in Smith’s (1991) terms. When modifying inner aspect, preverbs act as boundedness markers. They influence the verb meaning in introducing a bound in its aspectual structure. This is the case, for example, for the pair alszik ‘to sleep’/elalszik ‘to fall asleep’ (2). Whereas the simple form (that is, without preverb) is an activity (in Vendler’s 1967 sense), the preverbed form (2) is inchoative: it refers to the achievement constituting the initial bound of the activity. In the following examples, from Nyéki (1988: 139-140), the simple forms describe states; when preverbs are added, the sentences become inceptive; that is, they describe the telic events at the origin of the stative situations. Consequently, dynamic adverbs such as egyszer csak ‘suddenly’ (7) and lassan ‘slowly’ (8) become available only in the presence of a preverb.
8In the above examples, the preverbs indicate the presence of an initial bound. In other cases, they also enable reference to events with a final bound. This is shown in (9-12): the acceptability of át ‘for’ in the absence of a preverb and that of belül ‘in’ with a preverbed form indicates that the preverb acts as a telicity marker here.
9The examples in (9-10) differ from those in (11-12) in that they have agentive subjects. Conversely, the verbs érik (11) and fogy (12) are strictly anticausative. They cannot be construed with an ‘active’ subject, i.e. a cause or an agent. If a cause is to be introduced with these verbs, it is only as an ablative K(ase)P. Only their causative counterparts (meg)érel ‘to make ripe’ and (el)fogyaszt ‘to consume’ can support a cause subject. This is exemplified for megérik / megérel in (13):
10As we will see in 2.3. érik, as well as áll ‘to be at stop; to stand’ display anticausative and unaccusative properties respectively.
11Knittel & Forintos-Kosten (2002) analyze the forms without preverbs in (11-12) as gradual completion verbs (Bertinetto & Squartini 1995); as shown by the (b) examples, the preverbs convey telicity, and introduce the idea that the events display a final endpoint, whereas no such endpoint is implied in (a). This contrast accounts for the distribution of át ‘for’ (a) and belül ‘in’ (b). Now, an alternative description can be put forward, taking into account Hay, Kenny & Levin’s (1999) analysis. According to these authors, degree achievements can be telic as well as atelic depending on the presence of a bound in the change of state of their affected argument. In other words, the change in the state of the argument of érik ‘to ripen’ and of fogy ‘to run out’, when simple, is unbounded, as shown by the following examples:
- 2 The perfectivity of the complex forms is due to the past tense; in the present, the final bound is (...)
12The examples in (14) show that no final bound is implied when the preverb is absent2.
13As mentioned above, preverbs can also modify outer aspect, and turn imperfective sentences into perfective. Whereas telicity markers reflect the presence of a natural final endpoint to the process, perfectivity presents the situation as a whole, into which both the initial and final endpoints are included. The difference between telicity and perfectivity, then, lies in the fact that telicity implies a final bound (cf. (b) in (7-12)), whereas perfectivity conveys the idea that this final bound has been reached. This is the case in the accomplishment sentences in (15):
14Since it contains a quantized object, eszik az almát ‘eat the apple’ is an accomplishment. However, the expected postposition belül ‘in’ is impossible here since it would imply that the event reached its end (15a), an interpretation prevented by the imperfective. For belül to be possible, a preverb indicating perfectivity has to be used in the sentence, as in (15b). The perfective value of the preverb is confirmed in (15c). The preverbed form is indeed incompatible with a description of the event as still going on, showing that it is necessarily presented as a whole, including its final endpoint, which corresponds to the perfective viewpoint (Smith 1991). In the absence of the preverb, however, the continuation of the event is not excluded, due to the imperfective viewpoint.
15In the following examples, it is the postposition át ‘across’ (cf. (16a)) that is used as a preverb with the verb megy ‘to go, to walk’. It thus changes both the argument structure (from intransitive to transitive with a PP complement) and the aspect (from atelic to telic) of the initial verb, as shown by the contrast between (16b) and (16c). As in (15c), the preverbal position of the preverb expresses the perfective viewpoint. This is shown in (16d), where the second event (amikor delet harangoztak ‘when noon rang’) can only be understood as following the first (see also (20)), and not as occurring during it, which would be the case with the imperfective.
16We will come back to these data in the following sections.
17As the above examples show, preverbs normally appear as verbal prefixes; yet, contrary to prefixes, they can be detached from the verb, in which case they appear immediately after it:
18The postverbal position of preverbs relies on various factors. It can be due to the interrogative (17) and negative (18) structure of the sentence. In (19) it is forced by the focalization of a constituent, subject in (a), object in (b). Finally, in (20), in combination with a temporal subordinate clause, the postverbal placement of the preverb expresses the progressive aspect; the event described in the subordinate clause is understood as occurring during the process described in the main clause (see (16d) for a comparison).
19In the following section, we present a syntactic analysis of preverbs in verbal contexts based on the observations put forward by Knittel and Forintos-Kosten (2002).
20In their analysis of preverbs, Knittel and Forintos-Kosten (2002) suggest that the various influences that these elements can have on aspect and on argument structure is determined by the positions where they merge.
21According to these authors, preverbs have a bounding effect on verbal predicates (see section 2.1.). Yet, as we saw above, depending on the original (a)telicity of the verb, this bounding effect can affect either inner (i.e. lexical) or outer (i.e. grammatical) aspect.
22Recall first from examples (7-8) and (2) that preverbs can occur with stative verbs and activities, both event types being atelic (Vendler 1967, Smith 1991), that is, unbounded with respect to inner aspect. In such cases, the preverbs introduce inner boundedness, since they transform activities and statives into achievements with an inceptive or inchoative value. In these cases, the preverbs induce the presence of an initial bound. In other cases, including agentive intransitive and gradual predicates, it is the final bound which is introduced by the preverb, cf. (9-12).
23To account for this use, Knittel and Forintos-Kosten (2002) suggest, following a proposal by Egerland (1998), after Borer (1993), that the preverbs merge as heads of an aspectual projection between VP and vP, whose head encodes a [+b(ounded)] feature (see also Borer 2005, Travis 2009 for similar analyses). In more recent approaches, such as those developed by Harley (2009), Alexiadou & Anagnostopoulou (2009), Alexiadou, Iordachioaia & Soare (2010), where the verbal head is analyzed as an unspecified root √ associated with an eventive component vP, the representation in (21) can be suggested. The root incorporates v and the preverb in the course of its movement towards the higher inflectional projections.
24The structure [v [root Theme]]vP has been proposed by Landau (2010) for unaccusative verbs. In fact, verbs such as érik/megérik ‘to ripen’ or áll/megáll ‘to stand; to be at stop’/‘to stop’ react positively to the unaccusativity tests proposed for Hungarian in Halm (2012), namely the presence of incorporated subjects (Marácz 1989, Kiss 2002) (22) and the impossibility of inserting egyet ‘once’ (Kiefer 1992, Pinon 2001) in the sentences (23):
25When there is no preverb, the verb appears unbounded. Consequently, AspP is not projected. By contrast with (21), transitive and unergative verbs, that display agentive subjects, are characterized by the presence of a voiceP component, introducing the Agent argument (Kratzer 1996).
26Consequently, a verb such as megdolgozik ‘to work[+Telic]’ (10) will receive the analysis in (24):
27When preverbs act as perfectivity markers, as in (15), they also encode boundedness. However, in this case, boundedness applies to outer aspect, internal boundedness (i.e. telicity) being due here to the presence of a quantized direct object. Knittel and Forintos-Kosten (2002) propose to represent preverbs marking perfectivity as heads of a second aspectual projection, inflectional in nature, and dominating the agentive voiceP, as in (25) (= (15)):
28Since telicity is due to the object, we consider, with Knittel & Forintos-Kosten (2002), that Asp[+B]P is projected, and hosts az almát ‘the apple’ in its specifier.
29Finally, when the preverb is originally a postposition, but acts also as a telicity and a perfectivity marker, as in (16), the following representation can be suggested, after Knittel & Forintos-Kosten (2002):
30Here, the postposition incorporates into √, then moves with it to the upper verbal and aspectual heads.
31To sum up, Hungarian preverbs, depending on their merging position, can perform two different roles with regards to boundedness. When merging (or moving if they are originally postpositions) above vP, they act as telicity (i.e. inner boundedness) markers, whereas they are perfectivity (i.e. outer boundedness) markers when merging or moving above voiceP. Consequently, according to the present analysis, Hungarian does only display one class of preverbs, and not two, as postulated by Svenonius (2004) for Slavic or Germanic.
32After the examination of preverbs at the verbal level, we now turn towards their study as parts of nominalized verbs.
33In this section, we present first the general properties of nominalization in Hungarian. Then we turn towards the examination of the syntactic and semantic behavior of preverbs in nominalizations.
- 3 The difference between ás and és is due to vowel harmony.
34According to Laczkó (1985) and Szabolcsi (1994), the only productive, and most common manner of deriving event nominals from verbs in Hungarian is by the means of the suffix ás/és3, as shown in (27):
35Yet, as is often the case across languages, -ás/-és nominals can refer either to event nominals or to result nominals. Now, the distinction between events and results is crucial since, as shown by Grimshaw (1990), only event nominals can keep the argument structure of the verb they are built on. According to Szabolcsi, however, the dependencies of nominalized verbs provide a clue as to the eventive vs resultative nature of the nominalization. This is illustrated in the following examples, from Szabolci’s (1994) examples (14a-b):
- 4 The form -i is considered a suffix by Szabolcsi. However, since it adjectivizes whole PPs and occur (...)
- 5 Való is the only possible adjectivizing form for case-marked nouns (*Mari-hoz-i/Mari-hoz való: towa (...)
36In Hungarian, noun dependencies must be adjectivized. In other words, no PP or KP can appear in a structure whose head is a noun, independently of its nature (i.e. simple or derived). If it should be the case, the PP has to be adjectivized. This can be done in two ways. First, the adjectival suffix -i can be added as an affix on the PP4, as in (28a) ebéd után-i ‘after lunch-Adj’; second, the PP can be adjectivized by való (lit. ‘being’), a former present participle of ‘to be’.5 As noted by Szabolcsi (1990, 1994), the use of való is possible only if the deverbal nominal has an event reading, as shown in (28b), where ‘speaking’ is the only possible translation of felszólalás. If the suffix -i is used, the ambiguity between the eventive and the resultative readings remains (28a). Consequently, if the adjectivization of the dependencies is made by the means of való, the eventive reading is ensured.
37The examples in (28) also show that nominalizations syntactically behave like possessive DPs, and display the structure in (29):
38The Possessee is characterized by a Possessive marker (Knittel 1998), followed by an agreement marker whose person and number features reflect those of the Possessor, which appears at its left as a nominative DP (see Szabolcsi 1983, 1994).
39If the nominalized verb is transitive, the nominative case is generally assigned to the object, and the subject, if present, is realized by the means of a (adjectivized) PP with által ‘by’:
40As noted by Szabolcsi, Hungarian nominalizations qualify as ‘true nouns’, not as (verbal) gerunds. If they were verbal gerunds, they would keep the verbal property of assigning accusative to their direct object, and, since they would be modified by adverbs instead of adjectives, they would not display the adjectivization processes just described.
41In this section, we examine the behavior of preverbs in deverbal nominals, beginning with those that mark inner boundedness.
42As discussed in section 2.1, preverbs can indicate that an event is bounded, by designating either its initial or final point. In such cases, nominalization by the means of ás/és is generally possible for both the simple and the preverbed form. The nominals describe the same events as the verbal forms.
43Consider first the examples in (31), where the simple form is a state, the preverbed form being the achievement leading to it:
44In these examples, only the preverbed forms, whether verbal or nominal, are dynamic. This distinction is reflected by the fact that durative PPs have to be introduced by belül ‘in’ instead of át ‘for’ in the presence of the preverb (32). In the same manner, dynamic adverbs and adjectives are possible only in the presence of the preverb (33), see also (8):
45Yet, as noted by Maeinborn (2005, 2007), after Kratzer (1996), statives do not constitute a uniform class. Some of them display a Davidsonian argument, and, as such, qualify as events. If this is the case, the event component enables modifiers such as ein bisschen (German) ‘a little bit’, to get a temporal reading instead of a simple degree reading. The following example show that egy kis ‘a little’ (literally ‘one little’) has indeed a temporal reading when modifying allás:
46Consequently, we are dealing here with a stative eventive nominal. Yet, eventivity is not a sufficient condition for the presence of argument structure. In other words, the complex vs simple nature of the event nominal remains to be determined. If we consider, with Grimshaw (1990), that the arguments are obligatory with complex events, the grammaticality of the following example, where no argument is realized, shows that allás is a simple event:
47Note that the potential argument of állás is realized as a first person singular agreement marker on the adjective, showing that it is here an argument of the adjective, not of the noun, which would be the case if the suffix appeared on it, as in állás-amPoss.1Sg, lit. ‘my standing’.
48The nominalization of the simple verb being a simple event, it is expected that it is also the case for that of the complex verb, since the preverb has no influence on the argument structure here. Consequently, felállás ‘the rising’ should be able to occur in the same context as állás in (35), which is indeed the case:
49Parallel to áll(ás)/feláll(ás) are the cases of gugol(ás) ‘to be crouched’, ‘crouching position’/le-gugol(ás) ‘to crouch’/‘the crouching’ and ül(és) ‘to be seated’, ‘sitting position’/fel-ül(és) ‘to sit’, ‘the sitting’:
50Let us now turn towards telicity markers, that is, preverbs that indicate the presence of a final bound to the event.
51First, agentive intransitive verbs, such as (meg)dolgozik (10) or (meg)főz (9) do not produce eventive nominalizations, as shown by their impossibility to combine with át or belül adjectivized PPs (see also Szabolcsi 1994):
52We saw in (16) the opposition between the atelicity of megy (a hídon) ‘to go (on the bridge)’, and the telicity of átmegy a hídon ‘to cross the bridge’. The examples in (40) show that this opposition is maintained in the corresponding nominalizations:
53The possible presence of the agent in a PP headed by által shows that both menés and átmenés keep the agentive argument structure of the verb, and are thus complex events.
54Consider finally the case of degree predicates, whose telicity depends on the preverb, cf. érik/megérik ‘to ripen’ (11); fogy/elfogy ‘to run out’ (12). Both the simple and the complex verbs can be nominalized by the means of -ás/-és:
55As expected from the above examples, the initial (a)telicity value is maintained in the nominalization.
56Since these nominals are events, as shown by the fact that they display aspectual distinction, the question of their argument structure has to be raised. The examples in (45-46) show that, contrary to állás/felállás, fogyás/elfogyás and érés/megérés cannot give rise to well-formed sentences if deprived of their argument.
57Consequently, they can be analyzed as exhibiting an argument structure.
58Let us now summarize our observations on preverbs influencing lexical aspect. Three cases have been examined successively:
(i) simple form: stative
preverb form: inchoative achievement;
(ii) simple form: dynamic atelic (activity)
preverbed form: dynamic telic (accomplishment);
(iii) simple form: gradual atelic (activity) preverbed form: gradual telic (accomplishment).
59With the exception of unergative verbs, that do not give eventive nominalizations (39), the meaning and the lexical aspect of both simple and preverbed verbs are fully inherited by the derived nominals. Since the nominalization process does not modify inner aspect, it appears as a simple category change. As for the argument structure, two cases can be distinguished.
60For transitive structures of type (ii), the arguments are maintained in the nominalization process. This is not the case for (i): although eventive, the nominalizations are simple events. Conversely, the verbs in (iii) maintain their unique argument when nominalized. Recall from section 2.3. that verbs of both (i) and (iii) can be characterized as unaccusative. However, as was shown in (13a), only the verbs of class (iii) admit the insertion of ablative KP expressing the cause. This consequently leads to distinguish between anticausatives (iii), which admit cause KPs and keep their argument when nominalized, and “plain” unaccusatives, which do not.
61Recall from examples in (15) that preverbs act as perfectivity markers when combined with telic VPs. In this section, we examine their behavior in nominalizations.
62Consider the examples in (47-48):
63The examples in (a-b) show simple (a) and preverbed (b) verbs. As described in 2.1. above, these sentences differ in their outer aspect, with the preverb conveying perfectivity. As can be seen from examples (c-d), the corresponding nominalizations can also occur with (d) or without (c) preverbs.
64Since preverbs keep the same aspectual value with verbs and nouns when they modify inner aspect (cf. 3.2.1.), the question that arises now is whether the same phenomenon is observed for perfective preverbs. In other words, do perfective sentences produce perfective nominalizations in Hungarian? This question in turn leads to that of the presence of outer aspect in DPs whose heads result from a nominalization process, a hypothesis suggested by various authors for other languages (Engelhardt 2000; Alexiadou, Iordachioaia & Soare 2010; Iordachioaia & Soare 2008, 2009; Knittel 2011).
65On this issue, various observations have been put forward that evidence the perfective/imperfective distinction in nominalizations, and can prove useful to answer the above question.
66A first series of observations is due to Iordachioaia & Soare (2008, 2009). These authors observe that Romanian argument-taking nominalization can be built up by the means of two suffixes, namely the supine, which conveys imperfectivity, and the infinitive, which brings about perfective nominals. Supine nominals are characterized by a pluractional value (Lasersohn 1995, Van Geenhoven 2004, Laca 2006a,b); they can refer to a plurality of events without exhibiting plural marking.
67The following examples show that in Hungarian, the pluractional reading is available only for nominalizations of imperfective sentences, and cannot appear if a preverb is present:
68Example (49a) describes a situation where Peter usually kills ducks, or is in the habit of doing so. Observe here that no frequency adverb is needed to refer to a plurality of events, thus revealing that pluractionality is readily available without any overt marking. Conversely, only one killing event is referred to when the preverb is used (49b). The same contrast holds for (50): in the (a) example, the situation described is one where, for example, children at school regularly read in a book for one hour each time. In (b), the adjectivized PP egy éven át való ‘for one hour’ is unavailable, and has to be replaced by egy éven belül való ‘in one hour’ for the sentence to be well-formed. In this case, the adjectivized PP corresponds to the time needed to read the whole book. As in (49b), the sentence describes a single event.
69Consequently, one can conclude that the pluractionality effects displayed by simple nominalizations is a reflect of their imperfectivity, as is the case in Romanian.
70Now, it has been observed by Smith (1991) that achievement sentences do not admit the imperfective viewpoint. This incompatibility is due to the fact that the imperfective focuses on the internal stages of processes. Since achievements are punctual, and are deprived of duration, they do not have internal stages for the imperfective to focus on. If nominalizations without preverbs are indeed imperfective, they should be ungrammatical with achievements too. The unacceptability of (51a) shows that this is indeed the case:
71The killing of a duck is an achievement situation. As expected, its punctual character makes it unavailable for the imperfective. Note that, in such a case, no iterative reading can appear since killing a duck is a once-only event. In fact, the only possible reading of this example is one where the killing has taken some time; in other words, for the imperfective viewpoint to be possible, the event has to be construed as an accomplishment, instead of an achievement. No such derived reading is necessary if a preverb is present, as in (b), due to the fact that the perfective viewpoint is unproblematic with achievements.
72Consider finally the examples in (52) and (53). The questions in (52a) and (53a) contain nominalization of complex verbs. Consequently, if the perfective viewpoint is indeed conveyed here, a positive answer should express the fact that the event has been fully completed; conversely, the answer should be negative if the event has not reached its natural endpoint. The examples in (b) and (c) show that this prediction is indeed borne out:
73Since the questions contain perfective nominalizations, the answer has to convey the idea that the event has reached its final point to be positive.
74We can then conclude that when a preverb indicates perfectivity at the sentence level, it keeps the same value in case of nominalization. More generally, the examples provided in sections 3.2.1. and 3.2.2. show that the influence of the preverbs when combined with verbs does not vary when the verbs are nominalized. As we saw in 3.2.1. preverbs modify inner aspect in the same manner for verbs and the corresponding nominalizations, when they are possible. This observation is also true for outer aspect: when a preverb is a perfective marker in a verbal environment, it acts as such in the corresponding DP. We conclude consequently that (im)perfectivity is a feature appropriate for the description of Hungarian event nominals.
75In the following section, we put forward a syntactic analysis of preverbs in nominalizations.
76As we have just seen, the use of preverbs and the operation of nominalization do not influence each other. In other words, except for unergatives, the nominalization process does not modify the aspectual properties of the original verb, be it simple or complex. To account for this fact, a possible hypothesis is that the nominalization process takes place after the aspectual component has merged.
77As discussed above, preverbs encode both inner and outer aspect. In section 2.3, we suggested, after Forintos-Kosten & Knittel (2002), that preverbs are heads of Aspectual projections. The projection of inner aspect dominates the eventive projection vP (cf. (21)), that of outer aspect, (Asp2), stands above VoiceP (25, 26).
78Consider first nominalizations of the structures where the preverb modifies inner aspect. In section 3.2.1, we distinguished different cases, recalled in Table (54):
79(54)
Simple form
|
Preverbed form
|
Value of the preverb
|
Pv+verb type
|
Event
structure
|
Examples
|
state
|
achievement
(inchoative)
|
initial bound
|
unaccusative
|
yes
|
állás/felállás (standing/rising)
gugolás/legugolás
(being crouched/crouching)
ülés/felülés
(being seated/sitting)
|
activity
|
accomplishment
|
final bound
|
transitive
|
yes
|
menés/átmenés (going/crossing)
|
activity
|
accomplishment
|
final bound
|
unergative
|
no
|
főzés/megfőzés (cooking)
|
scalar activity
|
scalar accomplishment
|
final bound
|
anticausative
|
yes
|
fogyás/elfogyás (lowering/running out)
érés/megérés
(ripening/becoming ripe)
|
80For the first case, it can be suggested that the nominal layer nP, whose head hosts the suffix ás/és, dominates the event projection vP (Alexiadou 2009):
81Since these structures are nominalizations of unaccusatives (2.3.), voiceP is never projected.
82As we saw in 2.1. and 3.2.1. there is another verb class that displays the same structure as unaccusatives, namely that of anticausatives (e.g. érés/mégérés ‘ripening’). Consequently, the same representation can be suggested. However, anticausatives, contrary to unaccusatives, admit ablative KPs expressing the cause. To account for this difference, we admit with Alexiadou & Anagnostopoulou (2009), that the vP of anticausatives contain a Causative feature, able to legitimate such a KP.
83As is the case for verbs, AspP is not projected in unbounded forms such as allás.
84Let us now consider the case of átmenés ‘crossing’. Three further points have to be taken into consideration. First, the preverb is originally a postposition (16); second, the nominal retains the agent of the verb (41). Third, the preverb acts also as a perfectivity marker (26). Consequently, both VoiceP (Alexiadou 2009) and Asp2P have to be projected, as is the case for the corresponding verbal structure, the nominalization layer standing above it.
85Consequently, the deverbal nominal has the same argument structure as the base verb.
86The same projections can be postulated for preverbs modifying outer aspect of transitive verbs (becsomagol (47) ‘to pack’, elolvás ‘to read’ (48d)). In (57), as in (56), nP, the higher layer of the structure, is placed above Asp2P. Since it is the object that is responsible for the telicity of elolvasás, it moves to Spec,Asp[+B]P (cf. (25)).
87Despite all the parallelisms observed above, there is a major difference between the syntactic behavior of the preverbs in sentences and in nominalizations, which is predicted by our analysis. Recall indeed from examples (17-20) that preverbs must appear in postverbal position under certain conditions (focalization, negation, interrogation). However, such a phenomenon is never observed for preverbs in nominalizations. In other words, whereas a sequence such as (58a) is possible, (58b) is not. This is illustrated in (59).
88In the presence of negation, the preverb must appear after the verb ((59a), see also (18)). If the negation scopes on a DP containing a preverbed deverbal nominal as its head, as in (59b), the preverb must remain in its usual prefixal position. As shown by (c), it cannot appear after the derived noun.
89This variation between the verb and the noun is expected if we consider that the nominalization suffix creates a new lexical item, whose category is n instead of v/Voice. As a consequence, preverb extraction, which is a syntactic process, is blocked from under nP level.
90The impossibility of detaching the preverb from its host in case of nominalization also signals that TenseP is absent from nominal structures in Hungarian, contrary to what is suggested by Lecarme (2012) for various languages. As shown by the following examples, the detachment of preverbs such as át in átmegy ‘to cross’ indicates outer aspect variation.
91Knittel & Forintos-Kosten (2002) consider that, in Hungarian, Tense is imperfective; perfectivity is conveyed only in the presence of Asp2. The pv-V order results from a movement of the complex verb to Tense from Asp2 position. When the preverb appears at the right of the verb (60a), they consider that the verb alone has moved to Tense, leaving the excorporated preverb behind. This analysis also accounts for the perfectivity ambiguity observed in (61):
92Contrary to what happens for dolgozott in (61), nominalizations are never ambiguous for outer aspect; it is the presence of the preverb alone that determines perfectivity. Moreover, as we saw in (58-59), a nominal such as átmennés cannot display a detached preverb. Since no interaction between Tense and perfectivity can be evidenced in the case of nominalizations, there is no argument in favor of a nominal Tense in Hungarian.
93In this paper, we discussed and compared preverbs in the verbal domain and in deverbal eventive nominals. We showed, following Knittel & Forintos-Kosten (2002), that preverbs are markers of telicity, and also of perfectivity when they occur in telic environments. They can thus be described as boundedness markers. As we have shown, these observations also appropriately describe the role of preverbs in nominalizations. Consequently, it can be concluded that preverbs do not exhibit nor entail any variation when combined with verbs or deverbal nominals. To account for this behavior, we have suggested that the syntactic projection responsible for the nominalization process stands above those encoding inner and outer aspect. This relative positioning explains why preverbs cannot be separated from nominals, even though they admit postverbal placement in sentences.
94An important observation that has been made here is that Hungarian deverbal nominals can display perfective as well as imperfective aspect, further supporting the hypothesis that outer aspect can be present in derived event nominals. If our analysis is correct, Hungarian expresses perfectivity in the same manner for verbs and deverbal nominals, i.e. by the means of preverbs. This language thus differs from Russian, where verbal prefixes, although present in the nominal structure, are deprived of their perfective value (Tatevosov 2013), but also from French (Knittel 2011) and Romanian (Iordachioaia & Soare 2008, 2009), languages where the expression of outer aspect is conveyed by the means of the mass/count distinction. Thus, another indirect consequence of our analysis is that it explains why Hungarian deverbal -ás/-és nominals never pluralize (Szabolcsi 1994; Kenesei 2005); since the availability of number inflection in event nominals conveys perfectivity in a number of languages, it is expected not to be observed if this value is expressed by other means, such as preverbs.