- * I would like to thank Artemis Alexiadou, Gianina Iordǎchioaia, Fabienne Martin, Antje Rossdeutsche (...)
- 1 I leave aside the difference between [+eventive] and [-eventive] er-nominals and its relation to th (...)
1The literature on er-nominalizations has established the so-called ‘external argument generalization’: Thesenominals typically denote the external argument of the underlying verb irrespectively of the specific θ-role which this argument has (Rappaport Hovav & Levin 1992; Fabb 1984; Keyser & Roeper 1984; van Hout & Roeper 1998 among others). That is, we find agent and instrument er-nominalizations but also er-nominalizations denoting other types of external arguments such as causer, holder or experiencer as illustrated in (1) for English. 1
(1)
|
a.
|
He is a teacher.
|
(agent)
|
|
b.
|
This is a can-opener.
|
(instrument)
|
|
c.
|
Anger is a great diffuser of pent-up emotions.
|
(causer)
|
|
d.
|
He is a holder of a Visa or Master card
|
(holder)
|
|
e.
|
He is a lover of French cuisine.
|
(experiencer)
|
Across Germanic and Romance languages, the formation of external argument-denoting er-nominalizations is a fully productive derivational process. It was observed, however, that not all er-nominalizations obey the external argument generalizations. The examples in (2) seem to denote the theme, i.e. the internal argument of the underlying predicate.
(2)
|
a.
|
baker
|
(a baked potato)
|
|
b.
|
broiler
|
(a broiled chicken)
|
|
c.
|
scratcher
|
(a lottery ticket that is scratched)
|
|
d.
|
bestseller
|
(something that sells well)
|
Nominals such as in (2) have an interpretation that is close to the interpretation that the base verb receives in the middle construction. Thus, it was proposed that these nominals are in fact derived from the middle versions of the underlying verbs where the theme (the argument denoted by the er-nominals in (2)) is the (either base generated or derived) subject of the verb (Booij 1986; Rappaport Hovav & Levin 1992; Heyvaert 1998; 2003).
2Besides object-denoting er-nominals, we also find er-nominals denoting the complement of a preposition modifying the verb (where the preposition is often locational). For these types of er-nominals, it was also proposed that they can be subsumed under a middle-kind of analysis (at least in Dutch, Heyvaert 1998; 2003).
(3)
|
a.
|
diner
|
(a place to dine in)
|
|
b.
|
sleeper
|
(a train where one can sleep in)
|
|
c.
|
toploader
|
(a washing machine)
|
- 2 Some examples exist such as the German counterpart of (3c) and loanwords like ‘bestseller’.
- 3 But see Cabredo Hofherr (2000) who argues that the theme in German middles is an external argument.
- 4 Many of the object-denoting er-nominals in English are from specific verbal subclasses (cooking ver (...)
While examples as in (2) and (3) can be found in English and Dutch, they seem to be quite rare in German. 2 A reason for this difference could be that English and Dutch form morphologically unmarked middles while German marks its middles with the reflexive pronoun ‘sich’ (cf. Schäfer 2006; 2008 for a proposal which correlates this difference in morphological marking with a difference concerning the syntactic position of the theme in middles; in Dutch and English middles, the theme is a (derived) external argument in Spec,vP, while in German middles, it remains in its VP-internal base position). 3 It should, however, be noted that even in languages that allow the kind of -er nominalizations in (2) and (3), their formation is certainly not fully productive but such a nominal has to be accepted in the language community in order to be understood in the right way. 4 A speaker cannot arbitrarily form an er-nominal with the intention that this nominal denotes the object of the underlying verb (or the object of a verbal preposition) while this is always possible if the er-nominal is ought to denote the external argument of a verb. That is, while virtually every verb projecting an external argument allows an er-nominal denoting this external argument, only a small subset of verbs allows er-nominals to denote the internal argument. This suggests that object-denoting er-nominals are (in fact need to be) lexicalized.
- 5 The literature sometimes provides examples of er-nominals derived from alleged unaccusative verbs. (...)
- 6 Again, languages differ in productivity. English allows it only with nouns denoting cities or villa (...)
3Finally, we can also find er-nominals with adjectival stems (foreigner, loner), prepositional stems (upper, downer, insider), denominal stems (porker, Londoner, villager) or derived from measure words (fiver) (see Ryder (1999) for a collection of such examples). 5 Once again, it should be noted that such derivations are not fully productive in that we cannot use any adjective, preposition or noun to form a corresponding er-nominal. This does not mean that there are no interesting generalizations to be made about what kind of non-verb derived er-nominals are possible or not. As an example, noun-derived er-nominals are clearly restricted by the semantics of the noun; while some noun classes do not allow er-formation at all (e.g. animals: *doger, *cater, *birder), other noun classes are persistently compatible with er-formation and then, the reading these nouns receive is clearly determined by a stereotypical pattern. For example, er-nominals derived from nouns denoting civilizing places (cities, villages, countries …) denote people who live at this place (but not people who just work there, or have any other relation to the place). 6 Similarly, German er-nominals derived from company names denote the employees but not the people who, for example, buy the products of this company (e.g.Bank-er, BMW-ler, IBM-er, …).
- 7 Therefore, Alexiadou & Schäfer (2010) propose to relate the difference between external argument-de (...)
4To conclude, while the class of er-nominalizations which do not denote the external argument of a verb is certainly interesting and amenable to specific generalizations, it seems fair to say that only the formation of external argument-denoting er-nominalizations is really a productive derivational process within and across languages. 7
5In the next section, I turn to a further type of verb-derived er-nominals in German. While this type is restricted in productivity in that it is possible only with verbs from a very specific class, it turns out that, within this class, its formation is highly productive.
6In this section, I discuss a further type of er-nominalizations which I call ‘event-denoting er-nominalizations’. Event-denoting er-nominalizations are, as far as I know, restricted to German. While the existence of this type of nominalizations has been acknowledged sometimes in the literature in passing (e.g. Fanselow 1985), it has (once again, to my knowledge) never been discussed in detail. Especially, the restrictions on the formation of event-denoting er-nominalizations have not been discussed.
7As an illustration, look at the two examples in (4) and (5). These examples are ambiguous between a reading where the nominal denotes the external argument of the underlying verb (a) and a reading where the nominal denotes the event of the underlying verb (b). Importantly, the event reading expresses something like a ‘minimal event’: (4b) describes one single jumping cycle which starts when a person’s feet leave the ground and stops as soon as the feet touch the ground again. Similarly, (5b) expresses one short beeping sound. Note that English er-nominalizations only have the external argument-denoting reading while the event-denoting reading surfaces with zero-morphology.
(4)
|
|
ein Hüpfer
|
|
a.
|
a jumper (a person who jumps)
|
|
b.
|
a jump (a/one jumping event)
|
(5)
|
|
ein Piepser
|
|
a.
|
a beeper (an agent who beeps)
|
|
b.
|
a beep (a/one beeping event)
|
- 8 I identified more than 100 verbs that form event-denoting er-nominals.
The formation of event-denoting er-nominals is not an idiosyncratic phenomenon restricted to a small number of verbs. 8 Instead, it turns out that it is highly productive within a specific, well-defined class of verbs. As a first approximation, we find them within the following semantic verb classes (using the terminology of Levin 1993).
(6)
|
a.
|
Verbs of contact by impact
|
|
b.
|
Verbs of (light/sound/substance) emission
|
|
c.
|
Verbs of manner of motion and body internal motion
|
However, being a member of these classes is not sufficient. A closer inspection of the verbs within these three classes reveals that a verb must have a ‘semelfactive’ use in order to be able to form an event-denoting er-nominal. Before I show this, I shortly introduce one proposal in the literature to characterize semelfactives.
8According to Rothstein (2007; 2008), semelfactives are verbs denoting ‘single occurrence’ events. In addition, these verbs are homonymous with activity verbs denoting iterations of the single event. For example, the verb ‘knock (on the door)’ can either have a semelfactive reading where an object is brought in contact sharply with a door once, or it can have an activity reading which expresses an iteration of the single event, i.e. an object is brought in contact sharply with a door a number of times. More specifically, Rothstein proposes that activities are derived from semelfactives by the operation of s(ingular)-summing below.
(7)
|
S-summing (Rothstein 2007): (singular summing) sums activity events with no temporal gap between them and forms a new singular event out of this sum.
|
S-summing is the operation forming activities. All semelfactive predicates have in addition an activity reading but not all activity predicates have a semelfactive reading. Rothstein (2008: 46) explains the differences and similarities between semelfactives and activities on the basis of a comparison of the two predicates skip and walk (the highlighting is mine):
Events in the denotation of the activity predicates skip and walk are formed by S-summing from minimal events of skipping and walking. These predicates denote, respectively, the set of skipping and walking events closed under S-summing. The difference between them is that minimal events of skipping are naturally individuable or naturally atomic, while minimal events of walking are not. […] When the minimal events in the denotation of an activity predicate P are naturally atomic, or naturally individuable, then they are lexically accessible. […] A predicate P is naturally atomic if what counts as one instance of P is given as part of the meaning of P and is thus not context dependent. […] A naturally atomic entity is one whose unit structure is perceptually salient and given by the world.
As mentioned, all semelfactive predicates have, in addition, an activity use but not all activities also have a semelfactive use. The property of semelfactives to be naturally atomic allows us to identify systematic differences between the two types of predicates (cf. Rothstein 2007).
9Semelfactives (8a, b) can be counted in two ways: counting adverbials can count either the single event (the semelfactive version) or the s-summed iterations of the predicate (the activity version). With pure activities (8c) only extended events can be counted because the single event is not naturally atomic, i.e. it is not lexically accessible.
(8)
|
a.
|
John knocked twice.
|
(ambiguous)
|
|
b.
|
John jumped three times.
|
(ambiguous)
|
|
c.
|
She walked three times.
|
(not ambiguous)
|
Semelfactives can be iterated in two ways: Again and again can modify either the single event or the activity predicate. In the case of activities, only the extended event can be iterated. This leads to different implications about the time course of the iterated events. Naturally atomic events can be iterated without a break between the individual events. (9a) can, therefore, be understood as process which is ongoing for some time. With activities, which do not involve naturally atomic events, the iteration implies that there must be a temporal gap between the individual activity phases. (9b), therefore, cannot be understood as a process ongoing for some time.
(9)
|
a
|
She jumped again and again.
|
→
|
She jumped for several minutes.
|
|
b.
|
He ran again and again.
|
↛
|
He ran for several minutes.
|
In the next section, I apply such tests to the verbal classes identified in (6). As it turns out, only semelfactive verbs within these classes allow the formation of event-denoting -er nominalizations.
Table I lists a number of er-nominalizations derived from ‘verbs of contact by impact’. All the examples in the left column have two interpretations; they either denote the external argument of the underlying verb or the (minimal) event expressed by the verb (only the latter reading is indicated in the table). The examples in the right column, on the other hand, are not ambiguous. They only allow for the external argument-denoting interpretation but lack the event-denoting reading (indicated by the asterisks in front of the examples in the table).
Table I: Verbs of contact by impact
Semelfactives
|
Activities
|
Klopfer
Aufpraller
Piekser
Schubser
Stupser
Rempler Anrempler
|
(a knock)
(a bounce)
(a prick)
(a jostle)
(a nudge)
(a jostle) (a jostle)
|
*Hämmerer
*Prügler
??Stampfer
*Drücker
??Beisser
*Schieber
*Quetscher
|
(hammering-event)
(a birching event)
(stamping event)
(pressing-event)
(a biting event)
(pushing-event)
(a squeezing event)
|
A closer inspection oftableI reveals that the verbs underlying the nominals in the left column are semelfactives while the verbs underlying the nominals in the right column are activities. This is illustrated with two verbs, ‘klopfen’ (to knock) and ‘hämmern’ (to hammer) which clearly differ with respect to the tests introduced above.
10If we count the event as in (10), we get an ambiguous result with ‘klopfen’ (either an atomic event or an extended event is counted) but not with ‘hämmern’ (only an extended event can be counted).
- 9 The verb ‘anklopfen’ (at-knock), in contrast, is an activity and, in turn, the er-nominalization do (...)
(10)
|
a.
|
Er
|
klopfte
|
dreimal.
|
(ambiguous)9
|
|
|
He
|
knocked
|
three-times
|
|
|
b.
|
Er
|
hammertoe
|
dreimal.
|
(not ambiguous)
|
|
|
He
|
hammered
|
three-times
|
|
If we add iterative adjuncts like ‘immer wieder’ (lit.: always again) or ‘wieder und wieder’ (again and again), ‘klopfen’ is again ambiguous (11a); either the atomic event or the extended activity is iterated. The verb ‘hämmern’ in (12a) does not show this ambiguity; only the extended event can be iterated. This difference between ‘klopfen’ and ‘hämmern’ is stressed by the fact that only the iterated semelfactive event in (11a) is logically compatible with (11b) which involves an atelic temporal modifier. The iterated activity event in (12a) is logically not compatible with (12b) which again involves an atelic temporal modifier. The reason is that the atelic modifier suggests that the agent acts without a break but only a naturally atomic event can be iterated without an interruption. If we want to iterate an extended event, we have to assume that there is a break between the individual extended events, as otherwise we could not identify the beginning or the end of the individual extended events; but this interrupted scenario cannot be described with a ‘for some time’-adverbial. The c-examples show the same (in)compatibility between iterated events and modifiers which suggest that the agent acted without a break; again, the semelfactive verb in (11c) gives much better results than the pure activity verb in (12c).
(11)
|
a.
|
Er
|
klopfte
|
wieder und wieder.
|
(ambiguous)
|
|
|
He
|
knocked
|
again and again
|
(→)
|
|
b.
|
Er
|
klopfte
|
eine Zeit lang
|
(am Stück/ohne Pause).
|
|
|
He
|
knocked
|
some time long
|
(at a stretch/without interruption)
|
|
c.
|
Er
|
klopfte
|
wieder und wieder
|
ohne Unterbrechung.
|
|
|
He
|
knocked
|
again and again
|
without interruption
|
(12)
|
a.
|
Er
|
hämmerte
|
wieder und wieder.
|
(not ambiguous)
|
|
|
He
|
hammered
|
again and again
|
(↛)
|
|
b.
|
He
|
hämmerte
|
eine Zeit lang
|
(am Stück/ohne Pause).
|
|
|
He
|
hammered
|
some time long
|
(at a stretch/without interruption)
|
|
c.
|
#Er
|
hämmerte
|
wieder und wieder
|
ohne Unterbrechung.
|
|
|
He
|
hammered
|
again and again
|
without interruption
|
Table II lists a number of er-nominalizations derived from (different types of) ‘verbs of emission’. Again, the examples in the left column are ambiguous, denoting either the external argument of the underlying verb or the event expressed by the underlying verb, while the examples in the right column only allow for the external argument-denoting reading but lack the event-denoting reading.
Table II: Verbs of emission
Semelfactives
|
|
Activities
|
|
?Aufblitzer
Piepser
Klopfer
Rülpser
Quietscher
Krächzer
Juchzer
Träufler/Tropfer
Spritzer
|
(flashing-event)
(a beep)
(a knock)
(a belch)
(a jar)
(a caw)
(a crow)
(a drop)
(a splash)
|
*Leuchter
*Blinker
*Funkeler
*Pieper
*Schreier
*Rauscher
*Summer
*Rassler
*Bluter
|
(a shining event)
(a flashing event)
(a sparkling event)
(a puling event)
(a shouting event)
(a swoosh)
(a buzzing)
(a rattling)
(a bleeding event)
|
Once again, the verbs underlying the nominalizations in the left column but not the verbs underlying the nominalizations in the right column are semelfactives. This is illustrated exemplarily below. The examples in (13)-(15) show that the sound-emission verb ‘piepsen’ (to beep) is a semelfactive verb while ‘summen’ (to buzz) is an activity verb.
(13)
|
a.
|
Er
|
piepste
|
dreimal.
|
(ambiguous)
|
|
|
He
|
beeped
|
three-times
|
|
|
b.
|
Er
|
summte
|
dreimal.
|
(not ambiguous)
|
|
|
He
|
buzzed
|
three-times
|
|
(14)
|
a.
|
Er
|
piepste
|
wieder und wieder.
|
(ambiguous)
|
|
|
|
He
|
beeped
|
again and again
|
(→)
|
|
|
b.
|
Er
|
piepste
|
eine Zeit lang
|
(am Stück/ohne Pause).
|
|
|
|
He
|
beeped
|
some time long
|
(at a stretch/without interruption)
|
|
|
c.
|
Er
|
piepste
|
ohne
|
Unterbrechung
|
wieder und wieder.
|
|
|
He
|
beeped
|
without
|
interruption
|
again and again
|
(15)
|
a.
|
Er
|
summte
|
wieder und wieder.
|
(not ambiguous)
|
|
|
|
He
|
buzzed
|
again and again
|
(↛)
|
|
|
b.
|
Er
|
summte
|
eine Zeit
|
lang
|
(am Stück/ohne Pause).
|
|
|
He
|
buzzed
|
some time
|
long
|
(at a stretch/without interruption)
|
|
c.
|
#Er
|
summte
|
ohne Unterbrechung
|
wieder und wieder.
|
|
|
|
He
|
buzzed
|
without interruption
|
again and again
|
|
- 10 A reviewer has problems with the categorization of the light-emission verb ‘blinken’ (to blink/flas (...)
The same contrast can be found with the light-emission verbs in (16)-(18).10 ‘(Auf-) blitzen’ (to flash) which allows for the formation of an event-denoting er-nominalization is a semelfactive while ‘leuchten’ (to shine) which does not allow for an event-denoting er-nominalization is an activity.
(16)
|
a.
|
weil
|
die Lampe
|
dreimal
|
(auf-)blitzte.
|
(ambiguous)
|
|
|
as
|
the lamp
|
three times
|
flashed
|
|
|
b.
|
weil
|
die Lampe
|
dreimal
|
leuchtete.
|
(not ambiguous)
|
|
|
as
|
the lamp
|
three times
|
shined
|
|
(17)
|
a.
|
weil
|
die Lampe
|
wieder und wieder
|
(auf-)blitzte.
|
(ambiguous)
|
|
|
as
|
the lamp
|
again and again
|
flashed
|
(→)
|
|
b.
|
weil
|
die Lampe
|
eine Zeit lang
|
(ohne Pause)
|
(auf-)blitzte.
|
|
|
as
|
the lamp
|
some time long
|
(without stop)
|
flashed
|
|
c.
|
weil
|
die Lampe
|
ohne Pause
|
wieder und wieder
|
(auf-)blitzte.
|
|
|
as
|
the lamp
|
without stop
|
again and again
|
flashed
|
(18)
|
a.
|
weil
|
die Lampe
|
wieder und wieder
|
leuchtete.
|
(not ambiguous)
|
|
|
as
|
the lamp
|
again and again
|
shined
|
(↛)
|
|
b.
|
weil
|
die Lampe
|
eine Zeit lang
|
(am Stück/ohne Pause)
|
leuchtete.
|
|
|
as
|
the lamp
|
some time long
|
(at a stretch/without stop)
|
shined
|
|
c.
|
#weil
|
die Lampe
|
ohne Pause
|
wieder und wieder
|
leuchtete.
|
|
|
as
|
the lamp
|
without stop
|
again and again
|
shined
|
Finally, table III lists er-nominalizations from ‘verbs of manner of motion’ and ‘verbs of body internal motion’. Again, the examples in the left column are ambiguous, denoting either the external argument of the underlying verb or the event expressed by the underlying verb, while the examples in the right column only allow for the external argument-denoting reading but lack the event-denoting reading.
Table III: Verbs of manner of motion and body internal motion
Semelfactives
|
Activities
|
|
Wackeler
?Stakser
Hüpfer
Hopser
?Stolperer
?Schlenkerer
?Schwenker
Dreher
|
(a wiggling event)
(a stalker)
(a hopper)
(a hopper)
(a stumble)
(a swing)
(a swing)
(a turn)
|
*Schütteler
*Torkler
*Rutscher
*Schlitterer
*Gleiter
*Roller
??Wirbler
??Schaukeler
|
(shaking event)
(a tottering event)
(a slip)
(a sliding event)
(a sliding event)
(a rolling event)
(a spinning event)
(a swinging event)
|
Again, what is relevant for the event-denoting reading is the semelfactive nature of the underlying verb. ‘Hüpfen’ (to jump) occurs in the left column and shows semelfactive behaviour while ‘rollen’ (to roll) occurs in the right column and shows activity behaviour.
(19)
|
a.
|
Er
|
hüpfte
|
dreimal.
|
(ambiguous)
|
|
|
He
|
hopped
|
three-times
|
|
|
b.
|
Er
|
rollte
|
dreimal
|
(not ambiguous)
|
|
|
He
|
rolled
|
three-times
|
|
(20)
|
a.
|
Er
|
hüpfte
|
wieder und wieder.
|
(ambiguous)
|
|
|
|
He
|
hopped
|
again and again
|
(→)
|
|
|
b.
|
Er
|
hüpfte
|
eine Zeit lang
|
(am Stück/ohne Pause).
|
|
|
|
He
|
hopped
|
some time long
|
(at a stretch/without interruption)
|
|
|
c.
|
Er
|
hüpfte
|
ohne
|
Unterbrechung
|
wieder und wieder.
|
|
|
He
|
hopped
|
without
|
interruption
|
again and again
|
(21)
|
a.
|
Er
|
rollte
|
wieder und wieder.
|
(not ambiguous)
|
|
|
|
He
|
rolled
|
again and again
|
(↛)
|
|
|
b.
|
Er
|
rollte
|
eine Zeit lang
|
(am Stück/ohne Pause).
|
|
|
|
He
|
rolled
|
some time long
|
(at a stretch/without interruption)
|
|
|
c.
|
Er
|
rollte
|
ohne
|
Unterbrechung
|
wieder und wieder.
|
|
|
He
|
rolled
|
without
|
interruption
|
again and again
|
- 11 Antje Rossdeutscher (p.c.) suggests that besides being semelfactive, the events in event-denoting e (...)
- 12 Sometimes the eventive er-nominal is doubled or blocked by a morphologically zero-derived event nom (...)
To conclude, er-nominalizations in German can denote events if their source is a semelfactive predicate, i.e. if its event is atomic/individuable. 11/12
11Above, I identified two productive types of er-nominalizations in German. A number of theoretical questions follow from this result:
Q1
|
Are external argument-denoting -ers and event-denoting -ers formed by one and the same (derivational) morpheme?
|
- 13 A reviewer asks why this is the null hypothesis since we find -er also in other uses, for example a (...)
This should be the null hypothesis as long as we do not have clear (empirical or theoretical) arguments against it. 13 The goal of the second part of this paper is to investigate the consequences of (as well as some problems for) exactly this hypothesis. If we assume, therefore, that the answer to the above question is yes, then the following questions arise:
Q2
|
Why are event-denoting er-nominals restricted to German?
|
Q3
|
Why can’t we form event-denoting er-nominals from all verb classes but only from semelfactives?
|
Q4
|
Where do the different interpretations (external argument vs. event) come from?
|
Q5
|
Does -er have any meaning?
|
Q6
|
Is -er a nominalizer or an argument?
|
- 14 I find this proposal suspicious because it forces us to duplicate all the different Voice heads tha (...)
Starting with the last question, if external argument-denoting er-nominals and event-denoting er-nominals involve the same er-morpheme, we cannot follow the recent proposal in Baker & Vinokurova (2009) and assume that -er is the nominal version of Kratzer’s Voice head with the semantics in (22a) and the syntax in (22b). 14 This is so because Voice heads introduce (θ-roles for) external arguments but they do not introduce events (the latter is introduced by v or V below Voice).
- 15 This proposal violates the head-movement constraint.
- 16 The alternative that -er is a complementizer or a linker in a relative clause structure and that a (...)
Similarly, we cannot follow van Hout & Roeper (1996) who propose a structure as in (23) where « -er in nominalizations is base-generated in the Specifier of Voice and moves up with the verb to an empty N head, rather than being base-generated as the nominal head ». 15/16
For an alternative analysis, I follow Alexiadou & Schäfer (2010) and assume that -er is a nominalizer. I illustrate this proposal first for external argument-denoting er-nominalizationsand turn, afterwards, to event-denoting er-nominalizations. Alexiadou & Schäfer (2010) propose the structure in (24) for external argument-denoting er-nominalizations. The verbal event <e> is introduced by the v-head (see Alexiadou & Schäfer (2010) for arguments that an event is present in all types of external argument-denoting er-nominalizations). Voice introduces the external argument of the verbal event (Kratzer 1996) which I assume to be realized by a variable <x> in the present structure. (I leave it open here whether this variable is a PRO or some other element (cf. Roy & Soare to appear)). I also assume that an aspect head on top of VoiceP is present in these er-nominalizations though this is not relevant for the present discussion (see Alexiadou & Schäfer (2010) for motivation; see Ntelitheos (2006) for a slightly different perspective). Finally, a nominalizing n-head realized by -er takes the verbal structure as its complement. This n-head introduces a referential operator <R> which binds the closest variable (in an argument position; but see below). This is the external argument variable <x> which was introduced in Spec,VoiceP Therefore, the er-nominalization denotes the external argument of the verbal event. I assume that v, Voice and N combine morphologically by some process (head- or XP-movement) that I do not further investigate here.
- 17 Some semelfactives are transitive. The corresponding event nominals do not license complements. Not (...)
- 18 The licensing of by-phrases or possessors which are both o.k. with event-denoting er-nominals is no (...)
Turning to event-denoting er-nominalizations,I propose the structure in (25). Voice is missing in this structure (and probably aspect, too).17 The absence ofVoice is corroborated by the lack of control into purpose clauses illustrated with the examples in (25a-d) involving the semelfactive verb ‘piepsen’ (to beep). Active semelfactives (25a), their impersonal passive counterparts (25b) but also nominalized infinitives denoting the verbal event (25c) allow control into purpose clauses (see Alexiadou et al. (2011) for further arguments that German nominalized infinitives involve Voice). The event-denoting er-nominal in (25d), on the other hand, does not license control suggesting that it lacks Voice.18
(25)
|
a.
|
Ich
|
hörte
|
[wie eri piepste]
|
[um
|
PROi die Oma
|
zu schockieren].
|
|
|
|
I
|
heard
|
how he beeped
|
in-order
|
the grandma
|
to shock
|
|
|
|
‘I heard how he beeped in order to shock the grandma.’
|
|
|
b.
|
Ich
|
hörte
|
[wie (von ihmi)
|
gepiepst
|
wurde]
|
|
|
|
|
I
|
heard
|
how (by him)
|
beeped
|
was
|
|
|
|
|
[um
|
PROi
|
die Oma
|
zu
|
schockieren].
|
|
|
|
|
in-order
|
|
the grandma
|
to
|
shock
|
|
|
|
|
‘I heard how a beep was emitted (by him) in order to shock the grandma.’
|
|
|
c.
|
?Ich hörte
|
[seini Piepsen] [um
|
PROidie Oma
|
zu schockieren].
|
|
|
|
|
|
I heard
|
his beeping
|
in order
|
the grandma
|
to shock
|
|
|
|
|
‘I heard his beeping in order to shock the grandma.’
|
|
|
d.
|
*Ich
|
hörte
|
[seineni Piepser]
|
[um PROi
|
die Oma
|
zu
|
schockieren].
|
|
|
I
|
heard
|
his beep
|
in-order
|
the grandma
|
to
|
shock
|
|
|
‘I heard his beep in order to shock the grandma.’
|
|
Once again, the <R> operator introduced by -er needs to bind a variable. Crucially, in the absence of Voice, the verbal event variable <e> introduced in v is the closest variable which <R> can bind (see also fn. 21). Therefore, the nominal denotes the verbal event.
The structures in (24) and (26) suggest that there exists only one -er affix which is present in all er-nominalizations. -er is the realization of an n-head which introduces a referential argument slot <R> for the nominal it produces. This <R> is an operator which needs to bind a variable under minimality (closest c-command). The different meanings result from different binding relations which, in turn, result from different structures (basically the presence vs. absence of Voice). This provides an answer to Q4 above.
- 19 The existence of a derivational morpheme such as -er under the above characterization is not expect (...)
12Let’s turn to Q5. Obviously, <R>ER cannot be sensitive to the difference between entities and events. If <R>ER has any meaning, this meaning must be applicable in both, the domain of events and the domain of entities. But as we saw above not every type of event can be bound by <R>ER. The event must be semelfactive, i.e. naturally atomic.It seems reasonable to take this as a selectional restriction of <R>ER.19 As a slight alternative, I assume that <R>ER carries this semantic property and transfers it to the bound variable; a variable bound by <R>ER is forced to be interpreted as naturally atomic. Crucially, natural atomicity is a property that cuts across the nominal domain (the domain of entities) and the verbal domain (the domain of events). Natural atomic events are semelfactives (as well as change-of-state events (see below)). Natural atomic entities are count nouns (but see Rothstein (2010) and references there for refinements). As an answer to Q3 above I propose, therefore, that the meaning of <R>ER enters the derivation as follows:
-
(i) <R>ER binds the closest variable.
-
(ii) <R>ER has as its meaning the property to be inherently ‘naturally atomic’.
-
(iii) This denotation is assigned to/transferred to the bound variable. At LF, the ‘lexical-conceptual restriction’ of the variable bound by -er must be semantically/conceptually compatible with this property.
- 20 This could be implemented if activities and semelfactives involve different types of little v. A re (...)
As a result, the er-nominal denotes either a count noun or a semelfactive event. If the event expressed by a verb cannot be conceptualized as semelfactive (e.g. with activities like ‘to run’), a semantic mismatch occurs. More specifically, assume that such events can be bound in the syntax by <R>ER. Therefore, the <R>ER operator assigns to <e> the property to be naturally atomic. But since the lexical core of the verb ‘run’ cannot be conceptualized as naturally atomic, such a nominalization is semantically deviant: <R>ER wants to bind an atomic event but the event associated with verbs such as ‘run’ cannot be conceptualized as atomic and, therefore, the two do not fit in their interpretations. The resulting structure, while syntactically well-formed, is filtered out as incomprehensible due to a semantic mismatch at the conceptual level.20
13What about er-nominals derived from anticausatives expressing a change of state? Why don’t the examples in (27) denote the change-of-state events expressed by the underlying verbs? This is especially striking as change-of-state events are typically assumed to be atomic, i.e., similarly to an atomic or minimal jumping event we know quite well what an atomic or minimal change-of-state event would be like.
(27)
|
a.
|
*Brecher (break-er)
|
|
|
b.
|
*Schmelzer (melt-er)
|
(under the intended event-reading)
|
However, change-of-state verbs are bi-eventive, involving a process-part followed by a resultant state (cf. (28) using the terminology of Ramchand (2008)). Crucially, only the combination of these two sub-events makes change-of-state events atomic. If we add <R>ER on top of the structure in (28) it will bind the closest <e> in the process v-head; however, this event on its own is not naturally atomic and the structure is, therefore, filtered out as incomprehensible similar to cases involving mono-eventive activities such as ‘to run’ discussed above.
(28)
|
[ProcessP <e> [Result P <s>]]
|
(e.g. Alexiadou et al. 2006, Ramchand 2008)
|
Next, I try to account for Q2 above, i.e. for the question why languages other than German lack event-denoting er-nominals? I will concentrate my discussion on English. As mentioned, while English does not have event-denoting er-nominals, it nevertheless has event-denoting nouns that correspond to the semelfactive er-nominals in German. These English nouns are zero-derived.
(29)
|
a bounce, a knock, a beep, a jump, …
|
I propose that the nouns in (29) have exactly the same syntactic structure as the corresponding event-denoting e9r-nominals in German, i.e. the structure in (26) above. However, in English the n-head is spelt out in a different way in such a constellation. The framework of Distributed Morphology allows us to formulate that the Spell-Out of the n-head forming naturally atomic nouns can be realized differently depending on the syntactic context. Following Embick (2003), insertion of Vocabulary items is sensitive to locality. That is, the Spell-Out rules for n can make reference to its c-command domain as suggested by the two rules in (30). That is while German realizes the naturally atomic nominalizer as -er in any context, English realizes it as -er if it is located on top of Voice as in (24) but it realizes it as zero morpheme if it is located on top of an eventive v-head as in (26).
(30)
|
a.
|
Spell-out of n: Voice-Cycle
|
|
|
|
n
|
↔
|
-er/
|
{√BEEP, √JUMP, ...}
|
|
b.
|
Spell-out of n: v-Cycle
|
|
|
|
n
|
↔
|
-Ø /
|
{√BEEP, √JUMP, ...}
|
- 21 As mentioned, object-denoting er-nominalizations are not (productively) available across languages. (...)
14In this section, I want to discuss a potential challenge for the central hypothesis of this article that external argument-denoting er-nominalizations and event-denoting er-nominalizations in German are derived by one and the same nominalizer. A clear prediction of this hypothesis as it was developed in the previous section is that both types of er-nominalizations should always be naturally atomic. I argued that this is the case for all event-denoting er-nominalization; they only denote semelfactive events. But we also need to investigate whether the same holds for external argument-denoting er-nominalization. Is it correct that external argument-denoting er-nominals are always naturally atomic, i.e. count? The answer seems to be no. There exists a set of er-nominals which denote what seem to be mass expressions,typically artificial liquids. Following Sleeman & Verheugd (2004), I call these er-nominals ‘products’. The German examples of this type of er-nominals in (31a-c) are doubled by their English counterparts in (32a-c): 21
(31)
|
a.
|
(Nagellack-)entfern-er
|
(32)
|
a.
|
(nail polish) remov-er
|
|
b.
|
(Essig-/Teppich-)reinig-er
|
|
b.
|
(vinegar/carpet) clean-er
|
|
c.
|
(Foto-)entwickl-er/fixier-er
|
|
c.
|
(photo) develop-er/fix-er
|
Is this then an argument that there must be two er-nominalizers in German, one denoting external arguments which can be mass or count (this er-nominalizers would also exist in other Germanic and Romance languages) and one denoting naturally atomic events, i.e. events that are count? This would mean that German has two different nominalizers which are, by accident, both realized as –er (see also fn. 13). However, some observations argue in favour of the idea developed above that German has only one er-nominalizer.These observations give us a hint about how external argument-denoting -ers can sometimes circumvent the semantic restriction to be naturally atomic.
15While products have unequivocal mass-readings as shown in (33), it should be noted that the very same nominals can easily be used as count nouns as the use of the determiners ein (a), jeder (every), zwei (two) or diese (these) in (34a-c) indicates.
(33)
|
Ich
|
habe Wasser/Bier/Essigreiniger/Nagellackentferner
|
verschüttet
|
|
I
|
have water/beer/vinegar cleaner/nail polish remover
|
poured
|
(34)
|
a.
|
Ein/Jeder
|
Teppichreiniger
|
kann
|
verwendet
|
werden.
|
|
|
A/Every
|
carpet-cleaner
|
can
|
used
|
be
|
|
b.
|
Zwei (verschiedene)
|
Essigreiniger
|
|
|
|
|
|
Two different
|
vinegar-cleaners
|
|
|
|
|
c.
|
Diese Teppichreiniger
|
ähneln sich
|
in ihrer Wirkungsweise.
|
|
|
|
|
These carpet cleaners
|
are-similar
|
in their mode-of-action
|
|
|
It is well known that the mass-count distinction is not entirely clear-cut. Nouns which are basically count can be shifted to a mass-reading and nouns which are basically mass can be shifted to a count-reading (e.g. Chierchia 2010). The former shift, illustrated in (35), can be brought about by a process called ‘universal grinder’; the latter shift, illustrated in (36a, b) can be brought about by a process called ‘universal packager’.
(35)
|
a.
|
There is chicken in the soup
|
|
b.
|
There is apple on the floor
|
(36)
|
a.
|
I drank three beers.
|
|
b.
|
I like only three wines: Chardonnay, Pinot and Chianti.
|
The question is which process is responsible for the mass and count uses of products seen in (33) and (34). Are their count-uses derived from a basic mass interpretation via universal packager, or are their mass-uses derived from a basic count interpretation via universal grinder? Under the analysis developed in the previous section we would predict the latter direction to be correct. This view would be confirmed if products would not show all properties of clear instances of mass nouns. This seems to me to be the case, although judgements are very subtle: while unequivocal mass-nouns can be modified with quantifiers like all and most (37a, b), products slightly resist such modification (38a, b); thereby, they are similar to coerced count nouns (39a, b).
(37)
|
a.
|
Ich
|
habe alles
|
Wasser/alle Zahnpasta/
|
|
allen
|
Sand
|
weggewischt.
|
|
|
I
|
have all
|
water/
|
all toothpaste/
|
all
|
sand
|
wiped-away
|
|
b.
|
Das meiste
|
Wasser
|
/Die meiste
|
Zahnpasta/Der meiste
|
Sand wurde
|
weggewischt
|
|
|
|
The most
|
water/
|
the most
|
toothpaste/the most
|
sand was
|
wiped-away
|
|
(38)
|
a.
|
?(?)Ich
|
habe allen
|
Teppichreiniger/allen
|
Kleber
|
weggewischt.
|
|
|
|
I
|
have all
|
carpet-cleaner/ all
|
glue
|
wiped-away
|
|
|
b.
|
?(?)Der meiste
|
Teppichreiniger
|
/Der meiste
|
Kleber
|
wurde
|
Weggewischt.
|
|
|
The most
|
carpet-cleaner/
|
the most
|
glue
|
was
|
wiped-off
|
(39)
|
a.
|
??Der meiste Apfel/Das meiste
|
Hühnchen
|
wurde
|
weggewischt.
|
|
|
The most apple/The most
|
chicken
|
was
|
wiped-away
|
|
b.
|
??Ich habe
|
allen Apfel/alles Hühnchen
|
weggewischt.
|
|
|
|
I have
|
all apple/ all chicken
|
wiped-away
|
|
These data tentatively suggest that products are external argument-denoting (instrumental) er-nominalizations that are basically naturally atomic. More specifically, I propose that they start with a basic interpretation as a sort or a kind. Furthermore, as all count nouns, products can be coerced into mass. The special thing about products seems to be that, due to the fact that products are fluids, this shift from count to mass happens with less effort than with other count nouns.
16In this article, I discussed German event-denoting er-nominalizations and showed that these can be formed only from semelfactive verbs, i.e. they denote only semelfactive events. Similar to external argument-denoting er-nominals, the formation of event-denoting er-nominals is fully productive in that it is possible with every semelfactive verb in German.
17The existence of two productive types of er-nominalizations (external argument-denoting and event-denoting) poses a number of theoretical questions which I discussed in the second part of the article. Specifically, I investigated the hypothesis that there is only one er-nominalizer which introduces an <R>-operator that binds the closest variable. The meaning of this operator cuts across the verbal and the nominal domain: <R>ER has as its meaning the property to be inherently naturally atomic. This denotation is transferred to the bound variable and, at LF, the variable must be conceptually compatible with this property. The two readings (external argument-denoting and event-denoting) are the result of a structural difference; if Voice is present, then the <R> operator located in the n-head binds the external argument in Spec,Voice, if Voice is absent, <R> binds the event in v.
18To answer the question why only German but not English or Dutch have event-denoting er-nominalizations, I proposed that different Spell Out rules are at work in these languages. These rules are sensitive to the syntactic context in which a head occurs and, in the case of event-denoting nominals, the Spell Out rule of English (and Dutch) chooses a zero exponent for the n-head that is spelt out as -er in the context of Voice. German, on the other hand, spells out the relevant n-head as -er in both contexts.
19Finally, I discussed potential counterexamples to the claim that all er-nominalizations are atomic. I argued that er-nominals denoting products are basically external argument (instrument) denoting count nominals which can, however, be easily coerced to nominals with a mass-interpretation.