CC BY
23
М.Ю. Князев
Санкт-Петербургский государственный университет, 199034, Россия, г. Санкт-Петербург, Российская Федерация;
Высшая школа экономики в Санкт-Петербурге,
190121, Россия, г. Санкт-Петербург, Российская Федерация;
Московский государственный педагогический университет, 119991, г. Москва, Российская Федерация
Ограничение на агентивность как аргумент в пользу именной функциональной оболочки для сентенциальных актантов: экспериментальное исследование1
В статье предлагается единый для всех сентенциальных актантов в русском языке анализ с «DP-оболочкой». Выдвигается гипотеза о том, что в косвенных позициях сентенциальные актанты лицензируются нулевым предлогом. Приводится аргумент, основанный на ограничении на агентивность при придаточных с союзами что и чтобы. Приводятся экспериментальные свидетельства, опирающиеся на факторное определение ограничения на агентивность. На основе экспериментальных данных демонстрируется премущество анализа с нулевым предлогом по отношению к двум другим подходам.
Ключевые слова: сентенциальные актанты, ограничения на одушевленность, нулевой предлог, падеж, экспериментальные методы в лингвистике.
1 Работа над статьей велась в рамках проекта РНФ 16-18-02003 «Структура значения и ее отображение в системе лексических и функциональных категорий русского языка», реализуемого в МПГУ.
M. Knyazev
Saint Petersburg State University, Saint Petersburg, 199034, Russia;
Higher School of Economics - Saint Petersburg, Saint Petersburg, 190121, Russia;
Moscow State University of Education, Moscow, 119991, Russia
A general DP-shell analysis of clausal complements: Experimental evidence from the agentivity restriction on Russian cto - and ctoby-clauses2
This paper argues for a general DP-shell analysis of clausal complements in Russian. It is proposed that clausal complements are licensed by a null P in Caseless positions. The argument is based on an agentivity restriction on cto- and ctoby-clauses. Experimental evidence is presented that makes use of the factorial definition of the agentivity restriction. Two alternative accounts - in terms of a partial DP-shell and semantic coercion - are discussed. It is shown that the experimental results favor the null P account over the alternatives.
Key words: clausal complements, animacy restriction, null preposition, Case, experimental methods.
1. Introduction
The commonly-held view that clausal arguments, as opposed to DP arguments, do not need (structural) Case [Pesetsky, 1982; Pesetsky, Torrego, 2011] has been challenged by a number of proposals analyzing clausal arguments as optionally embedded in a (possibly null) DP-shell and thus needing Case at least sometimes [Davies, Dubinsky, 2009; Hartman, 2013; Kastner, 2015]. The evidence for this analysis comes from the fact that in a number of languages, including Russian, clausal arguments, e.g. cto-clauses, can be embedded in an overt DP-shell giving rise to the to, cto-clause construction [Khomitsevich, 2007; Hartman, 2013].
2 The study is supported by the RSF (Russian Science Foundation) project 16-18-02003 "Structure of meaning and its mapping into lexical and functional categories of Russian" at MSPU (Moscow State University of Education).
Building on these proposals, Knyazev (2016) advances a more radical claim arguing that (bare) cto-clause complements in Russian are always embedded in a null DP-shell, which is licensed by the "Last Resort" insertion of a null P in oblique/PP positions. The argument is based on the so-called agentivity restriction, namely the fact that cto-clause complements of speech act verbs occupying oblique positions force the agentive reading of the verb even when the verb allows the non-agentive reading with a to, cto-clause.3 The agentivity restriction follows from the interpretation of the null P as the relation hold of holding propositional content, which requires a sentient argument.
This paper considers two alternative explanations of the agentivity restriction that do not utilize the null P. According to the first one, the hold relation arises not by virtue of inserting a null P but as the result of the semantic coercion of the verb that happens whenever a PP/oblique-selecting verb combines with a cto-clause. According to the second one, cto-clauses project a DP-shell only with the non-agentive variants of the verbs while remaining CPs with the agentive variant. Thus the agentivity restriction follows from the violation of the Case requirement with the non-agentive variant without invoking P-insertion.
The aim of this paper is twofold. First, it provides experimental evidence for the agentivity restriction, reporting the results of a formal acceptability judgment study using a 2 x 2 factorial design (see, a.o., [Sprouse et al., 2016]). Second, it discusses the results of an experimental study of a similar agentivity restriction observed with ctoby-clauses [Pekelis, 2014], which provides indirect evidence for the null P account favoring it over the two alternatives discussed above.
The structure of this paper is as follows. In section 2 I introduce the agentivity restriction on cto-clauses and present experimental evidence for it. In section 3 I lay out the null P account and discuss two alternative accounts of the agentivity restriction. In section 4 I turn to the agentivity restriction on ctoby-clauses. I report the results of the experimental study of this restriction and show how these results provide evidence for the null P account. Section 5 concludes.
2. The agentivity restriction
2.1. Introducing the restriction
The argument for the null P account is based on the agentivity restriction, which concerns non-illocutionary uses of speech act verbs, illustrated in (1a)-(1b). Such uses express the natural-meaning relation, where the subject
3 The other argument comes from the so-called collocational restriction on c7o-clause complements of nouns. See [Knyazev, 2016, 2017] for details.
is a natural sign of the state of affairs expressed by the complement [Kissine, 2010]. Interestingly, in such uses cto-clauses are significantly degraded unless they are embedded in an overt DP-shell (realized as the demonstrative to) introduced by the selected P/oblique case; cf. the agentive use in (2).
(1) a. Это намекает *(на то), что бар для туристов.
Eto namekaet *(na to), Cto bar dlja turistov. this hints on it.Acc that bar for tourists
'This suggests that the bar is for tourists.'
b. Многое говорит *(o том), что здесь были поселения. Mnogoe govorit *(o tom), cto zdes' byli poselenija. a lot says about it.Loc that here were settlements 'Many things indicate that there were settlements here.'
(2) Ученые намекают / говорят, что здесь были Ucenye namekajut / govorjat, cto zdes' byli scientists say that here were поселения.
poselenija. settlements
'Scientists say/hint that there were settlements here.'
Crucially, all verbs that show the agentivity restriction disallow the realization of their clausal argument as an accusative DP, as shown in (3), and require an oblique/PP instead, cf. (1).4 By contrast, verbs that do realize their clausal argument as an accusative DP such as dokazyvat' 'prove' and podtverzdat' 'confirm' do not show the agentivity restriction, as seen in (4).
??говорит?
??govorit?
says
(3) Что это *намекает Cto eto *namekaet what.Acc this hints 'What does this suggest/indicate?'
(4) Это доказывает / подтверждает { нашу гипотезу / Eto dokazyvaet / podtverzdaet { nasu gipotezu / this proves confirms our hypothesis.Acc
4 Henceforth I will not distinguish between oblique and PP positions assuming, following [Pesetsky, 2013], that positions where oblique-case-marked DPs occur are PP positions by virtue of the fact that oblique cases are uniformly assigned by (possibly null instances of) the category P.
, что бар для туристов }. , cto bar dlja turistov }. that bar for tourists
'This proves/confirms our hypothesis/ that the bar is for tourists.'
The only exception to this pattern are verbs like ubezdat' 'convince' which do not show the agentivity restriction despite realizing their clausal argument as a PP, as shown in (5). Such verbs are special in that they have a causative component to their meaning and an experiencer argument (cf. 'convince' ~ 'cause to believe').5
(5) Это убедило их в том, что бар для туристов.
Eto ubedilo ix (v tom), cto bar dlja turistov.
this convinced them in it.Loc that bar for tourists 'This convinced them that the bar is for tourists.'
Thus the agentivity restriction is observed only with cto-clause arguments in oblique/PP positions with verbs that have no agent or experiencer argument.
2.2. Experiment 1
One may note that violations of the agentivity restriction do not yield absolute unacceptability. This leads to the possibility that the proposed restriction is not a real grammatical restriction but rather an epiphenomenon reflecting the perceived 'colloquial' flavor of cto-clauses in PP positions, coupled with the overall lower 'naturalness' of non-illocutionary uses of speech act verbs. To rule out this possibility, the agentivity restriction was experimentally investigated using a 2 x 2 factorial design with factors clause type and agentivity, as shown in (6) (the design followed studies of island effects such as [Sprouse et al., 2016]). The effect of the agentivity restriction on acceptability was measured by first calculating the sum of the effects of clause type [6a-6b] and agentivity [6a-6c] and then subtracting it from the total effect of the purported violation in (6d) [6a-6d] (obtaining the differences-in-differences, or DD-score). A grammatical effect was operationalized as a positive DD-score reflecting a statistically significant interaction between the two factors. This would be visually represented as an interaction plot with two non-parallel lines with a characteristic 'superadditive pattern' (i.e. the joint effect exceeds the sum of the two independent effects).
5 Another example is napominat' 'remind', which allows both natural sign and causative reading in its non-agentive uses. Crucially, it shows the agentivity restriction only in its natural sign reading. See [Knyazev, 2016; 2018] for details.
(6) a. agentive I to, cto
Иван намекает на то, nmo.
Ivan namekaet na to, cto.
Ivan hints on it.ACC that
b. agentive | cto
Иван намекает, что...
Ivan namekaet, cto...
Ivan hints that
c. non-agentive I to, cto
3mo намекает на то, umo.
Èto namekaet na to, cto.
this hints on it.ACC that
d. non-agentive | cto
Это намекает, Èto namekaet, this hints
cto.. that
In the experiment (n = 282) six verbs were tested comprised of five verbs showing the agentivity restriction (govorit' 'say', namekat' 'hint', napominat' 'remind', ukazyvat' 'indicate' and svidetel'stvovat' 'testify') and the verb ubezdat' 'convince' as the control case.6 Each verb was tested with two types of non-agentive subjects, i.e. full DP and pronoun, in four conditions as in (6) yielding 6 x 2 x 4 = 48 experimental sentences in sum.7 The participants were asked to give acceptability judgments on a 1-5 scale. For each of the 6 x 2 = 12 sets a two-way ANOVA (on z-scores) was performed and the DD-scores were obtained. The results are summarized in Table 1 and Figure 1.
As we see in Table 1, the first five verbs showed robust interaction and a DD-score at least as strong as 0.5 (the only exception is namekat' 'hint' with a full DP subject, where the effect is weak). Crucially, ubezdat' 'convince', which was not predicted to show the agentivity restriction, yielded a different pattern with only a weak effect with both types of subjects.
6 The verb grozit' 'threaten', which is also reported to show the agentivity restriction in [Knyazev, 2016], was not included in the experimental sentences because of the clear acceptability contrast.
7 See [Knyazev, 2018] for details and materials.
83
Языкознание
ТаЫе 1
ТИс ББ-зсогез ап(1Ню р-уа!ие« 1"ог Ше Шегасйоп оГ Ню Гайог« сьаше туре ап(1 л<;к.\п\п у
говорить govorit' 'вау' намекать патека!' 'ЫпГ напоминать тротта? 'геттсГ указывать икагууа!' 'тсНса1е' свидетельствовать 8у1с1е1еГ81уоуа1' ЧевШу' убеждать иЬегсЫ' 'сопутсе'
1и11 № ББ 0.84 0.34 0.54 0.70 0.99 0.32
Р <0.001 0.02 <0.001 <0.001 <0.001 0.03
рг0110ш1 ББ 1.36 0.54 0.74 1.14 1.29 0.29
Р <0.001 <0.001 <0.001 <0.001 <0.001 0.06
1.0 0.5 0.0 -0.5
n
I -1.0 I 1.0
I
0.5 0.0 -0.5 -1.0
govorit' namekat' napominat' ukazyvat' svidetel'stvovat' ubezdat'
i i k L Full NP
L V \\ A V
\ \ \s L \ \S L \
V \ \ \ \ N
\ > 1
3
—1 Agentivity. A A Agent -o NonAgent 1—
Pronoun
-- \ K i
\ V N \ \N L Vs \
\ V \ > > \ \ \ \ \S
\ V \ \ \ \ \ \
\ \ ) \ \ \ .............. }
) \ \
-t-
To,chto Chto To, chto Chto To,chto Chto To, chto Chto To, chto Chto To, chto Chto
Clause type
Fig. 1. The mean (normalized) ratings for cto- and to, cfo-clauses
To conclude, the results of the experiment largely confirm the existence of the agentivity restriction as an independent grammatical constraint.
3. The account and some alternatives
3.1. The null P account
In order to account for the agentivity restriction, Knyazev (2016) argues that cfo-clauses are embedded in a null DP-shell, which is licensed in PP/oblique positions by the 'Last Resort' insertion of a null P. The crucial assumption is that the null P is interpreted as the generalized relation hold obtaining between an attitude holder (realized as, whereas agent/experiencer argument of the higher verb), and the proposition (realized as the complement of the null P, henceforth Phold).8 It is further assumed that Phold incorporates into V with the semantic effect of predicate conjunction, as in (7).
(7) namekat' 'hint' + PHOLD:
Xp<s ^.Xx.Xe. hint (e) & Cause/Agent (e,x) & Theme (e,p) & hold (x,p)
8 Krapova (2010) argues for the insertion of null P in clausal complements of factive PP verbs in Bulgarian. The difference between Krapova's null P and Phold (see Below) is that the former is restricted to verbs that take preposition za 'for' and complements headed by the complementizer deto found in relative clauses, whereas Phold can substitute for any preposition. Note incidentally that Russian cto in c7o-clauses can also function as a relative complementizer.
The agentivity restriction follows from the licensing condition on PHOLD, namely that it requires an attitude holder argument, which is present in the agentive uses and absent in the non-agentive uses, as shown in (8).
(S) Иван / *это намекает, nmo он устал
Ivan I *èto namekaet, [pp Phold [dp ôto on ustal]]
Ivan this hints that he tired
-X-
-X-
-HOLD■
HOLD
The fact that verbs like dokazyvat' 'prove' and podtverzdat' 'confirm' do not show the agentivity restriction is expected since these verbs can license accusative Case and thus do not need a null P to license a cto-clause. The lack of the agentivity restriction with verbs like ubezdat' 'convince' is also expected as these verbs have an experiencer argument interpreted as an attitude holder.
Despite the empirical adequacy of the null P account of the agentivity restriction, it may raise conceptual objections as it postulates a null element. Thus it would be desirable to explore alternative explanations of the restriction that do not utilize null P.
S6
3.2. A semantic coercion account
The idea that the agentivity restriction stems from the sentience restriction imposed by the relation hold can also be implemented by the mechanism of coercion of the verb meaning without postulating a null P.
Suppose that verbs that take clausal arguments cannot by themselves take propositional (<s,t>) arguments and instead uniformly take individual (<e>) arguments. Whenever a verb takes a clausal argument, the argument is nominalized, its type shifted to <e> (see [Potts, 2002]). As a consequence, only verbs that take <e>-type/DP arguments, e.g. dokazyvat' 'prove', will be able to take clausal arguments 'natively'. By contrast, verbs like namekat' 'hint' that do not take <e>-type/DP arguments, will be unable to take a clausal argument (unless the relation is mediated by overt P, leaving the exact details of the composition aside), as shown in (9a), and will need some special mechanism for this. It is natural to assume that this mechanism is coercion, which augments the meaning of the verb by a relation R introducing a clausal argument, as in (9b).9 A good candidate for this argument-introducer is the same hold relation that we saw above.
(9) a. namekat' 'hint':
Xx.Xe. hint (e) & Cause/Agent (e,x) No argument slot!!
' The idea was suggested to me by Ora Matushansky (p.c.).
(9) b. NAMEKAT' 'hint': coercion
Xp<s ^.Xx.Xe. hint (e) & Cause/Agent (e,x) & Theme (e,y) & hold (x,p)
Whenever verbs like namekat' 'hint' take cto-clauses, this leads to coercion and thereby to the agentivity restriction. The proposed account yields the same results as the null P account without postulating any unpronounced structure.
3.3. A partial DP-shell account
The other alternative follows the insight of the null P account that cto-clause complements of the non-agentive variants of speech act verbs are DPs and that the agentivity restriction ultimately stems from a Case requirement violation. However, instead of invoking P-insertion to account for the availability of cto-clauses with the agentive variants, it assumes that clausal arguments in the agentive case are simply CPs and thus do not need Case.
This alternative follows [Kastner, 2015] in assuming that clausal arguments (that-clauses) come in two types depending on whether the verb is presuppositional or non-presuppositional. Presuppositional verbs presuppose the truth or existence of their propositional argument. These are factive verbs like regret, know, remember and the so-called response stance verbs like deny, accept, agree, which express speech acts undertaken with regards to a certain already-existing proposition. Non-presuppositional verbs express the bringing about of the existence of a proposition and include basic propositional attitude verbs like say, think, claim, suspect, etc. Kastner's main claim is that clausal complements of presuppositional verbs are embedded in a (possibly null) DP-shell, whereas complements of non-presuppositional verbs are CPs.
The agentivity restriction directly follows under Kastner's proposal. Given that speech act verbs in their non-agentive uses express the natural-sign relation between two facts [Kissine, 2010], their complements will be factive and thus DPs. Hence they will violate the Case requirement.10 As for the agentive variants, they express volunteered-stance speech acts and thus can be analyzed as CPs.
In order to decide between these three potential accounts of the agentivity restriction, we need to look at a similar restriction observed with ctoby-clauses.
10 One problem with this account is that the agentivity restriction is not as strong as is expected from a Case requirement violation. Note that the null P account does not suffer from this problem as the restriction stems from the violation of the licensing condition on the null P, which is semantic in nature.
4. Another agentivity restriction
4.1. Pekelis's observation
Pekelis (2014) observes that manipulative verbs in Russian express their complements as either infinitival or subjunctive (ctoby)-clauses, as shown in (10a)-(10b). However, when these verbs are used with non-agentive subjects, ctoby-clauses are degraded, as shown in (10b), cf. (10a). Pekelis lists four verbs pozvojat' 'allow', vynuzdat' 'force', zastavljat' 'force' and trebovat' 'require' as showing this restriction to varying degrees.
(10) a. Руководство / положение / это позволило нам
Rukovodstvo / polozenie / eto pozvolilo nam management situation this allowed us
увеличить расходы. uvelicit' rasxody.
increase.INF expenses
'The management/this allowed us to increase the expenses.'
b. Руководство / ??положение / *это позволило, чтобы мы Rukovodstvo / ??polozenie / *eto pozvolilo, ctoby my management situation this allowed that.SUBJ we
увеличили расходы. uvelicili expenses. increased rasxody
'The management/situation/this allowed us to increase the expenses.'
The restriction in (10b) shares non-trivial properties with the agentivity restriction on cto-clauses discussed in section 2. First, all four verbs listed by Pekelis fail to take accusative DPs in their non-agentive variant, as shown in (11). In contrast, verbs that do take accusative DPs such as predpolagat' 'presuppose' allow ctoby-clauses, as shown in (12).11 Second, Pekelis found that the restriction is weaker with non-pronominal subjects, as can be seen in (10b). This parallels the findings reported for cto-clauses in section 2.2.
(11) Что это Cto eto
what.ACC this.NOM
^позволяет / "?"?вынуждает / *заставляет / ??требует?
??pozvoljaet / ??vynuzdaet / *zastavljaet / ??trebuet?
allow / force / make / require
Intended: 'What does this allow/require?'
11 Other verbs in this class are dopuskat' 'permit' andpredusmatrivat' 'envisage'.
88
(12) Это Èto this
, чтобы , ctoby
предполагает predpolagaet presuposes они сразу oni srazu
{ мгновенное { mgnovennoe immediate
приняли prinjali
that.suBj they at once took 'This presupposes an immediate decision / that they take the decision immediately.'
решение / resenie / decision.Acc решение}. resenie}. decision
Given these similarities between the two agentivity restrictions, it is desirable to have a unified account of them.12 But before exploring such an account, I will discuss the results of an experimental study of this restriction.
4.2. Experiment 2
The agentivity restriction on ctoby-clauses was tested using a similar factorial design as in Experiment 1, as shown in (13).13
(13) a. AGENTIVE | INFIN
Иван позволил нам увеличить расходы.
Ivan pozvolil nam uvelicit' rasxody.
Ivan allowed us increase.INF expenses
b. agentive | ctoby
Иван позволил нам, чтобы...
Ivan pozvolil nam, ctoby...
Ivan allowed us that.suBj
c. NON-AGENTIVE | INFIN
Это позволило нам увеличить расходы.
Èto pozvolilo nam uvelicit' rasxody.
this allowed us increase.INF expenses
d. non-agentive | ctoby
Это позволило нам, чтобы...
Èto pozvolilo nam, ctoby.
this allowed us that.suBj
12 I leave the discussion of Pekelis's account of these facts for another occasion.
13 Note that both non-agentive variants were tested with the realized dative arguments for reasons of uniformity. In Pekelis's (2014) ungrammatical examples (see (10b)) the dative arguments were missing.
In the experiment (n = 70) the four verbs reported to show the restriction were tested with a pronominal and a non-pronominal (agentive) subject.14 The results of the experiment are summarized in Table 2 and Figure 2.
Operationalizing the grammatical effect as before, the experiment revealed the agentivity restriction only for pozvoljat' 'allow' and vynujzdat' 'force' (with the pronominal subject) shaded in Table 2, while no effect was found for zastavljat' 'make'. Trebovat' 'require' showed an altogether different pattern, presumably due to an independent confound and was removed from the analysis.15
14 Thus 4 verbs x 2 types of subject x 4 conditions = 32 experimental sentences, distributed across 4 lists (supplemented bay 6 fillers in each list) were tested. The experimental sentences and fillers (in Russian) with the raw judgments are listed in (i) and (ii).
(i) Тренер позволил ему выйти на поле в стартовом составе (4.38); Тренер позволил ему, чтобы он вышел на поле в стартовом составе (2.63); Состояние футболиста позволило ему выйти на поле в стартовом составе (4.61); Состояние футболиста позволило ему, чтобы он вышел на поле в стартовом составе (1.60); Родители позволили ему принять католическое крещение (4.35); Родители позволили ему, чтобы он принял католическое крещение (2.00); Это позволило ему принять католическое крещение (4.84); (iid) Это позволило ему, чтобы он принял католическое крещение (1.33); Иван заставил их купить билеты на другой рейс (4.89); Иван заставил их, чтобы они купили новые билеты на другой рейс (1.65); Ситуация заставила их купить билеты на другой рейс (4.54); Ситуация заставила их, чтобы они купили билеты на другой рейс (1.53); Муж заставил ее провести там несколько дней. (5.00); Муж заставил ее, чтобы она провела там несколько дней (1.82); Что-то заставило ее провести там несколько дней. (4.80); Что-то заставило ее, чтобы она провела там несколько дней (1.38); Жена вынуждает меня подать документы на развод (4.92); Жена вынуждает меня, чтобы я подал документы на развод (3.05); Все это вынуждает меня подать документы на развод (4.89); Все это вынуждает меня, чтобы я подал документы на развод (1.8); Начальник вынуждает меня подписать заявление об увольнении (4.90); Начальник вынуждает меня, чтобы я подписала заявление об увольнении (2.46); Обстоятельства вынуждают меня подписать заявление об увольнении (5.00); Обстоятельства вынуждают меня, чтобы я подписала заявление об увольнении (2.67); Президент требует от главы региона принять немедленные меры (3.94); Президент требует от главы региона, чтобы тот принял немедленные меры (4.45); Обстановка требует от главы региона принять немедленные меры (4.08); Обстановка требует от главы региона, чтобы тот принял немедленные меры (4.26); Лукашенко требует от чиновников вести расчеты с Россией в валюте. (4.68); Лукашенко требует от чиновников, чтобы оно вело расчеты с Россией в валюте (1.78); Это требует от белорусских чиновников вести расчеты с Россией в валюте (4.15); Это требует от белорусских чиновников, чтобы они вели расчеты с Россией в валюте (2.77).
(ii) Это разрешает пользователю менять пароли самостоятельно (4.11); Закон допускает работать не полный рабочий день (3.16); Строение тела предполагает, чтобы человек использовал подушку во время сна (2.33); Программа предусматривает вкладывать деньги в модернизацию музея (2.23); Устройство реактора предусматривает, чтобы авария не могла иметь разрушительных последствий (1.97); Это запрещает собственнику, чтобы квартира выставлялась на продажу (1.49).
15 As can be seen in (i), trebovat' 'require' is degraded with infinitival complements in both ™ its agentive and non-agentive uses, as opposed to the other there verbs tested.
O (i) { 1Ты требуешь / ''''Это требует } от меня ехать.
_о
т CJZ
{ ?Ty trebues' / "Eto trebuet } ot menja uexat'.
you require this requires from me leave. inf
Intended: 'You require / this requires me to leave.'
Table 2
The DD-scores and the p-values for the interaction of the factors clause type and agentivity
позволять pozvoljat' 'allow' вынуждать vynuzdat' 'force' заставлять zastavljat' 'make' требовать trebovat' 'force'
full NP DD 0.61 -0.15 0.03 0.21
p 0.02 not sign. not sign. not sign.
pronoun DD 0.70 0.60 0.20 -0.80
p 0.01 0.01 not sign. 0.01
Infin Chtoby Infin Chtoby Infin Chtoby Infin Chtoby Clause type
Fig. 2. The mean (normalized) ratings for infinitival and cfoby-clauses
Now let's see whether each of the three accounts discussed in section 3 is consistent with these results.
4.3. Discussion
Let's start form the partial DP-shell account, which derives the agentivity restriction from a Case requirement violation stemming from the DP-shell analysis of presuppositional complements. Given the semantics
of ctoby-clauses in examples like (10a)-(10b), it is unlikely that they are presupposed. Hence this account incorrectly predicts no agentivity restriction for ctoby-clauses.
As for the null P account and the semantic coercion account, both accounts correctly predict the agentivity restriction for verbs pozvol'at' 'allow' and vynuzdat' 'force'. In the agentive variant the ctoby-clause will be licensed either directly by the verb in the case of pozvol'at' 'allow', which licenses the accusative DPs (see (14a)), or via the relation hold' (a minimally different variant of hold for subjunctive complements) in the case of vynuzdat' 'force' (see (14b)). In the non-agentive variant, however, the ctoby-clause will lead to a clash between the non-agentive subject and hold' in the case of vynuzdat' 'force' (as shown in (14b) for the null P account; the semantic coercion account works similarly). The same is true for the non-agentive pozvoljat' 'allow' except that under the null P account PHoLD will be disallowed altogether as the verb does not take PP complements (see also the discussion below).
(14) a. Иван Ivan Ivan
позволяет pozvoljaet allows
мне mne me
это eto
this.ACC
b. Иван / *Это вынуждает меня, Ivan / *Eto vynuzdaet menja, Ivan this forces me
[pp PHOLD
чтобы...} Ctoby...
that.SUBJ
чтобы... ctoby...]
that.SUBJ
-X-
HOLD' -
-x-
HOLD'-
When it comes to zastavljat' 'make', which disallows both DP and PP complements, as shown in (15a), the predictions of the two accounts differ. The null P account correctly predicts no contrast between the agentive and the non-agentive variant since both disallow P-insertion for reasons of c-selection, as shown in (15b). Coercion, however, will fail to be blocked and hence the coercion account will incorrectly predict the agentivity restriction, as shown in (15c).
(15) a. *Иван заставляет меня { это / на это...}
*Ivan zastavljaet menja { eto
Ivan makes me.ACC this.ACC
b. С-selection
*Иван / *Это заставляет меня,
*Ivan / *Eto zastavljaet menja,
Ivan this makes me.ACC
/ na eto...} on this
чтобы...
[PP Phold' Ctoby...]
that.SUBJ
c. Иван / *Это ЗАСТАВЛЯЕТ меня, Ivan / *Eto ZASTAVLJAET menja, Ivan this makes me
i x hold' - hold'-
чтобы... ctoby... that.suBj
To conclude, the acceptability judgment pattern of zastavljat' 'make' favors the null P account over the semantic coercion account. The correlation between a purely syntactic property (the lack of selection for PP) and the lack of the agentivity restriction as defined by the factorial design is expected under the syntactic (null P) account but not under the semantic coercion account.
5. Conclusion
In this paper I argued - against the standard view - that clausal complements in Russian need Case by virtue of being embedded in a null DP-shell. I also argued that in 'Caseless' positions complement clauses are licensed by the "Last Resort" insertion of a null P. The evidence for this view comes from the agentivity restriction on both cto- and ctoby-clauses, which follows from the licensing requirement of the null P. I provided experimental evidence for the agentivity restriction by utilizing a 2 x 2 factorial definition of the agentivity restriction. I discussed two alternative accounts of the restriction - the partial DP-shell account and the semantic coercion account - and showed that only the null P account is fully consistent with the results of the experimentals.
References
Davies, Dubinsky, 2009 - Davies W., Dubinsky S. On the existence (and distribution) of sentential subjects. Hypothesis A / hypothesis B: Linguistic explorations in honor of DavidM. Perlmutter. D.B. Gerdts, J.C. Moore, M. Polinsky (eds.). Cambridge, MA, 2009. Pp. 111-128.
Hartman, 2013 - Hartman J. Varieties of clausal complementation. Ph.D. Diss. Massachusetts Institute of Technology, 2013.
Kastner, 2015 - Kastner I. Factivity mirrors interpretation: The selectional requirements of presuppositional verbs. Lingua. 2015. Vol. 164. Pp. 156-188.
Khomitsevich, 2008 - Khomitsevich O. Dependencies across Phases: From sequence of tenseto restrictions on movement. Ph.D. Diss., Utrecht University, 2008.
Kissine, 2010 - Kissine M. Metaphorical projection, subjectification and English speech act verbs. Folia Linguistica. 2010. Vol. 44. Pp. 339-370.
Knyazev, 2016 - Knyazev M. Licensing clausal complements: the case of cto-clauses. Doct. diss. Utrecht University, 2016.
Knyazev, 2017 — Исследование «слабого» грамматического ограничения методами экспериментального синтаксиса: пример придаточных с союзом что в функции сентенциального актанта существительного // Рема. 2017. № 1. С. 22-40. [Knyazev M. Studying a weak grammatical violation with experimental syntax methods: the case of sentential complements of nouns with the complementizer cto. Rhema. 2017. No 1. Pp. 22-40.]
Knyazev, 2018 - Князев М. Это говорит, что: Ограничение на сентенциальный актант с союзом что при деагентивных употреблениях глаголов речи // Вопросы языкознания. 2018. В печати. [Knyazev M. A restriction on dausal complements with the complementizer cto with non-agentive uses of speech verbs in Russian. Voprosy Jazykoznanija. 2018. To appear.]
Krapova, 2010 - Krapova I. Bulgarian relative and factive clauses with the invariant complementizer deto 'that'. Lingua. 2010. Vol. 120. Pp. 1240-1272.
Pekelis, 2014 - Пекелис О.Е. Инфинитив vs. придаточное с союзом чтобы: к вопросу о выборе способа оформления сентенциального актанта в русском языке // Вопросы языкознания. 2014. № 4. С. 13-45. [Pekelis О.Е. Infinitiv vs. pridatocnoe s sojuzom ctoby: k voprosu o vybore sentencial'nogo aktanta v russkom jazyke [Infinitive vs. ctoby-clause: Choosing the strategy of sentential argument marking in Russian]. Voprosy jazykoznanija. 2014. No. 4. Pp. 13-45.]
Pesetsky, 1982 - Pesetsky D. Paths and categories. Doct. Diss. MIT, Cambridge, MA, 1982.
Pesetsky, Torrego, 2011 - Pesetsky D., Torrego E. Case. Oxford Handbook of Linguistic Minimalism. C. Boeckx (ed.). Oxford, 2011.
Potts, 2002 - Potts C. The Syntax and Semantics of as-parentheticals. Natural Language and Linguistic Theory. Vol. 20. Pp. 623-689.
Sprouse et al., 2016 - Experimental syntax and the cross-linguistic variation of island effects in English and Italian. Natural Language and Linguistic Theory. 2016. Vol. 34. Pp. 37-44.
Статья поступила в редакцию 15.09.2017 The article was received on 15.09.2017
Князев Михаил Юрьевич - Ph.D. (лингвистика); старший преподаватель кафедры сравнительного литературоведения и лингвистики, Национальный исследовательский университет «Высшая школа экономики» в Санкт-Петербурге; ассистент кафедры общего языкознания, Санкт-Петербургский государственный университет; старший научный сотрудник Института современных лингвистических исследований, Московский государственный педагогический университет
Knyazev Mikhail Yu. - Ph.D. in Linguistics; Senior Lecturer of Chair of Comparative Literary Studies and Linguistics, Higher School of Economics - Saint Petersburg; Assistant of the Department of General Linguistics, St. Petersburg State University; Senior Researcher of Institute of Modern Linguistic Studies, Moscow State University of Education
E-mail: misha.knjazev@gmail.com
94
