The influence of grammatical features on linearization: Evidence from Slovenian Petra Mišmaš University of Nova Gorica
Introduction This paper explores the influence of grammatical features on linearization. The goal of the paper is to determine to which extent linearization is influenced by grammatical features. The Principle of Distinctness will be assumed, which claims two elements of the same type cannot appear inside the same phase of the derivation (Richards 2010). In order to linearize the structure within a phase, all elements must be distinct, which means two DPs must differ in at least one feature. I want to explore which features have the power to make two DPs distinct and what is the relation between the features – is there a hierarchy between the features, which are important at linearization. In order to investigate the effects of features, three different phenomena in Slovenian are explored: multiple Wh-fronting, Multiple Sluicing and double accusative constructions. Section 1 offers an overview of the Principle of Distinctness as proposed by Richards (2010) and the influence of features in the assumed theory. Section 2 gives examples from languages that are influenced by the Principle of Distinctness and based on the data assess the influence of features on linearization in these languages. This is achieved by looking at phenomena in which more than one DP is located in a single phase of the derivation. Section 2.1 is about multiple Wh-fronting in Serbian and Croatian, section 2.2 deals with Slovenian multiple Wh-fronting, section 2.3 with multiple Sluicing and section 2.4 with double accusatives in Slovenian. Section 3 discusses the results and outlines the proposed hierarchy of features and section 4 is the conclusion.
1 On Distinctness
Languages display different morphological dissimilation effects which they use to avoid accidental repetition of elements, because of the general prohibition for elements of the same type to appear to close together. These constraints are such that they “operate on both form 1
and content of morphemes, banning adjacent identity within a circumscribed domain” (Nevins 2010: 1). These constraints are known as repetition avoidance, morphological dissimilation, haplology, anti-homophony or the morphological OCP (Nevins 2010). One mechanism of repetition avoidance, which is in the center of our investigation, can be observed during the mapping from syntax to phonology. More precisely, at the point of SpellOut when syntactic structure has to be linearized. Other mechanisms, according to Nevins (2010), come into play after linearization and are divided into prosodification of linearized elements within the Spell-Out domain, M-Word formation and Allomorph selection. This paper deals with the first stage of mapping form syntax to phonology – the linearization of syntactic terminals and the influence grammatical features have at this point of the derivation. In order to assess the influence of features at this point and whether there is a hierarchy of the features at the point of linearization, we will first be looking at how linearization proceeds, which according to Richards (2010) is regulated by the Principle of Distinctness.
1.1 The Principle of Distinctness (Richards 2010) Richards (2010) bases his theory on the idea that some properties of syntax follow from the interaction between syntax and phonology, specifically on the ban which prohibits elements of the same type to be close together. In order to explain this ban, Richards assumes (following Chomsky 2000, 2001) that Spell-Out occurs cyclically during the derivation after the completion of strong phases, where strong phases include vP, CP, PP, and KP1 (i.e. kase phrase, the highest projection of the noun phrase). After each strong phase the material is sent to PF (Spell-Out domain) through Spell-Out. Richards (2010) also assumes that the head of a phase and its specifier (i.e. edge) are a part of the following phase for the purposes of the calculation of distinctness since they are linearized with the material of the higher phase (Chomsky 2000). Following Chomsky (2000), he also assumes that trees generated by syntax do not contain information about linear order and that linearization of nodes is performed by the operation of Spell-Out. It is also assumed that linearization is performed after Spell-Out by a version of Linear Correspondence Axiom (LCA) as proposed by Kayne (1994), by which a total order of terminals within a Spell-Out domain (phase) is established. Based on these assumption Richards (2010) formulates the principle of Distinctness: 1
When talking about English Richards (2010) focuses on DPs. In this paper we will also be looking at DPs (and not KPs).
2
(1)
Distinctness: If a linearization statement <α,α> is generated the derivation crashes. (Richards 2010: 5)
According to (1), a linearization statement is only interpretable when the nodes in the linearization statement are distinct from each other. Two nodes of the same type (i.e. <α,α>) in an asymmetric c-command relation cannot be linearized in the same Spell-Out domain. For example: the linearization statement , which comes from the linearization of the structure in (2) in which XP does not stand for any of the strong phases (i.e. vP, CP, PP, or KP)2, cannot be linearized and causes a crash:
(2)
XP 3 DP X’ John 3 X DP Mary Richards (2010) also states that Distinctness only affects functional heads which enter the
derivation as feature bundles and that Distinctness effects arise before vocabulary insertion. Functional heads are therefore represented only as feature bundles and, according to Richards (2010), languages are different with respect to the richness of feature bundles. The question then is, whether only categorical features are relevant to calculate distinctness or are there any other features which are responsible for making DPs distinct?
1.2 Two types of languages Richards (2010) divides languages into two groups according to which nodes count as being ‘of the same type’, that is, which nodes cannot be linearized in a particular language when they occur in one phase. For languages such as English, being of the same type means having nodes with the same label (e.g. ). Other languages are, according to Richards (2010), sensitive to the value of features like case and gender nodes like DP have. Richards (2010) focuses mainly on the English type, while this paper looks into languages that are
2
When X stands for one of the strong phases the linearization is not problematic: Since the head of a phase and its specifier are linearized with the material of the higher phase the two DPs are not be in the same phase and the derivation of (2) would be successful.
3
assumed to be of the second type, specifically into the importance of features in these languages. An example of a violation of the Principle of Distinctness in English is shown in (3), which shows sentences with exceptives:
(3) a. Every man danced with every woman, except John with Mary.
(Richards 2010: (6a))
b. *Every man danced with every woman, except John Mary.
(Richards 2010: (7a))
In (3b) both remnants are DPs and both are in the same Spell-Out domain forming a linearization statement . The ungrammaticality of (3b) follows from the Principle of Distinctness – two nodes in one Spell-Out domain are of the same type, which means that they cannot be distinguished from each other. They thus cannot be linearized and the linearization statement causes a crash. Example (3a) however is acceptable – the Principle of Distinctness is not violated since the two remnants are not of the same type – since PP is also a phase, linearization only sees a DP and a PP, which gives the linearization statement . Richards (2010) notes that there are languages that allow multiple DP remnants in multiple sluicing; an example of such a language is Japanese (other such languages include German, Dutch and Greek…). Richards (2010) claims that a language such as Japanese is sensitive to features other than labels (case, gender, animacy). This is shown below on Japanese multiple sluicing:
(4)
Watashi-wa dono otokonoko-ni-mo hoshigatteita subete-no hon-o every boy.DAT
I.TOP dare-ni
nani-o
wanted
every
ageta ga
book.ACC gave but
ka wasureta
who.DAT what.ACC Q forgot ‘I gave every boy all the books he wanted, but I’ve forgotten who what.’ (Richards 2010: (88a))
According to Richards (2010) in (4) the two DP remnants of multiple sluicing carry different case features and even if both Wh-phrases are in one Spell-Out domain, the linearization statement <[DP, DAT], [DP, ACC]> is distinct enough that linearization can proceed. (5) shows that when the case feature of the two Wh-words is the same, sluicing becomes impossible. As Richards claims this is because the two Wh-words are not distinct. 4
(5) a. [Sensei-o
hihansita] gakusei-ga koko-ni oozei iru kedo, dare-ga dare-o ka
Teacher.ACC criticized student.NOM here.DAT many be but
who.NOM who.ACC Q
oboeteinai remember.NEG ‘There are lots of students here who criticized teachers, but I don’t remember who who.’ (Richards 2010: (89a)) b. *[Sensei-ga
suki na] gakusei-ga koko-ni
teacher.NOM like
oozei iru kedo, dare-ga
student.NOM here.DAT many be but
dare-ga ka
who.NOM who.NOM Q
oboeteinai remember.NEG ‘There are lots of students here who like teachers, but I don’t remember who who.’ (Richards 2010: (89b))
In (5a) the DP remnants of multiple sluicing have different case marking. Richards (2010) claims that since the sentence is acceptable we can conclude different case features are enough for the two DPs to be distinct from each other, making it possible for linearization to proceed, suggesting the linearization statement is <[DP, NOM] [DP, ACC]>. In (5b) there are again two DP remnants of multiple sluicing and the sentence is unacceptable. Following Richards’ Principle of Distinctness, the conclusion is that the two DPs cannot be distinguished from each other, since they both have the same nominative case feature, forming a linearization statement <[DP, NOM] [DP, NOM]>. Case however is not the only feature that can make Japanese DPs distinct – one other feature that has this ability is animacy, the effects of which can be observed in (6):
(6)
[Doobutsu-ga suki na] hito-ga Animal.NOM like
koko-ni oozei iru kedo, dare-ga nani-ga
person.NOM here.DAT many be but
ka
who.NOM what.NOM Q
oboeteinai remember.NEG ‘There are lots of people here who like animals, but I don’t remember who what.’ (Richards 2010: (90))
(6) and (5b) differ in animacy features. In (5b) both DP remnants have the same animacy feature ([+animate]), while in (6) one DP has a [+animate] ant the other [-animate] feature. 5
Comparing the linearization statements, Richards (2010) concludes that (5b) has the linearization statement <[DP, NOM, +animate] [DP, NOM, +animate]>, while (6) has <[DP, NOM, +animate] [DP, NOM, -animate]>. Following the Principle of Distinctness, according to Richards (2010), the grammaticality of (6) can be explained by assuming that one different feature is enough for the two DPs to be distinguished from each other, making linearization possible, while in (5b) the DPs are indistinguishable from each other. Japanese examples raise questions about the importance of features in the Principle of Distinctness. In languages which are sensitive to features is one feature enough to make two phrases distinct? Are all features equally important when it comes to distinctness? Is there a ranking among various features? In order to answer these questions I give some more examples from languages that are sensitive to feature specifications of the linearized phrases in the next section.
2 The influence of features
In this section I discuss several examples in Slovenian and other languages in which multiple DPs3 occur in the same phase of the derivation. These examples will explore the influence of features in the Principle of Distinctness, specifically: What counts as distinct when it comes to features; is it enough if two DPs only differ in one feature or is it the case that more features have to differ and if one features is enough, is it important which features are different. The three different phenomena discussed in this section are multiple Wh-fronting, multiple sluicing and double object construction.
2.1 Multiple Wh-questions Richards (2010) discusses the influence of Distinctness on multiple Wh-fronting in languages such as Serbian and Croatian, which are said to display differences between each other with respect to which features in the two languages make fronted DPs distinct. In Serbian, case does make DPs distinct, but gender does not:
3
Following Progovac (1998) and Pereltsvaig (2007) it will be assumed that Slovenian has DPs.
6
(7) a. *Kojem
je čovjeku kojem
dječaku mrsko pomogati?
Serbian
which.DAT AUX man.DAT which.DAT boy.DAT boring help.INF “Which man doesn’t feel like helping which boy?” b. *Kojem
je čovjeku
kojoj
ženi
(Richards 2010: (106b)) mrsko pomogati?
Serbian
which.DAT AUX man.DAT which.DAT woman.DAT boring help.INF “Which man doesn’t feel like helping which woman?” c. Kojem
je čovjeku kojeg
(Richards 2010: (106c))
dječaka mrsko pozdraviti?
Serbian
which.DA AUX man.DAT which.GEN boy.GEN boring greet.INF “Which man doesn’t feel like greeting which boy?”
(Richards 2010: (106a))
In (7a) both fronted DPs are dative and both have the same gender feature (masculine). The sentence is, according to Richards (2010), ungrammatical and it seems that this is not because both gender features are the same, but rather because of the same case feature – we can conclude this by comparing (7a) to (7b) in which the DPs have the same case feature (dative), but different gender feature (masculine and feminine), which has no effect on the grammaticality. In (7c), on the other hand, the two DPs are both masculine, but have different case feature (dative and genitive) – the sentence is, according to Richards (2010), completely acceptable, which suggests that being distinguishable (or distinct) by case, but not gender, is more important in Serbian. Croatian acts slightly different from Serbian and the judgments for examples which are comparable to the ones in Serbian are different:
(8) a. ??Kojem
je čovjeku kojem
dječaku pomoči?
Croatian
which.DAT AUX man.DAT which.DAT boy.DAT help.INF “Which man is to help which boy?” b. ?Kojem
je
čovjeku kojoj
(Richards 2010: (107a)) ženi
pomoči
Croatian
which.DAT AUX man.DAT which.DAT woman.DAT help.INF “Which man is to help which woman?” c.
Kojem
je čovjeku
kojeg
dječaka pozdraviti?
(Richards 2010: (109)) Croatian
which.DAT AUX man.DAT which.GEN boy.GEN greet.INF “Which man is to greet which boy?’”
(Richards 2010: (108))
In (8a) the two DPs have the same case and gender features (dative, masculine) and the sentence is almost unacceptable (the differences in the acceptability are very subtle). But as 7
shown in (8b) the sentence improves if the two DPs differ at least in gender. Case seems to play a crucial role here too, since (8c) where the two DPs have different case but the same gender is completely acceptable. Based on the examples in (8) we can conclude that even in Croatian it is more important to have different case than to only differ in gender (since (8b) is less acceptable than (8c)). If we compare the data from Serbian and Croatian presented in Richards (2010) we can first conclude, that the two languages fall into the same group of languages as Japanese since the fronted Wh-phrases are all DPs and examples like (7c) and (8c) are still completely acceptable. In both languages case feature have an influence on the acceptability of examples: When case features on DPs are different the sentences are acceptable (Serbian (7a) and (7b) compared to (7c) and Croatian (8a) compared to (8c)). But the languages are different with respect to gender features: In Serbian gender features have no effect on the acceptability (comparing (7a) and (7b)) while in Croatian gender has an influence (comparing (8a) to (8b)) but this influence seems to be less strong than the influence of case, since (8c) is completely acceptable and (8b) is judged as less acceptable. This suggests a ranking of features that are important for distinguishing between two DPs and also that the influence of lower ranked features on acceptability is language specific (if gender is ranked lower than case but only has an effect in Croatian). Making a conclusion based solely on the examples above, we could assume, that case is ranked higher than gender. In order to determine that, more data is needed. The Slovenian data that will help us answer this question are presented in section 2. 2. Before we turn to Slovenian, I will report on a small survey. Since the data presented from Serbian in Richards (2010) is based on judgments of a few speakers, I have rechecked the data from Serbian presented in Richards (2010) with nine Serbian speakers. The judgments by the nine speakers are drastically different to the ones presented in Richards (2010):
(9) a. ?Kojem
je čovjeku kojem
dječaku mrsko pomagati?4
Serbian
which.DAT AUX man.DAT which.DAT boy.DAT boring help.INF “Which man is to help which boy?” b.
?Kojem
je čovjeku kojoj
ženi
mrsko pomagati?
Serbian
which. DAT AUX man. DAT which. DAT woman. DAT boring help.INF “Which man is to help which woman?”
4
In the retested examples the verb 'pomogati' was changed into 'pomagati', which is the common form of the verb, as suggested by one of the speakers.
8
c.
??Kojem
je čovjeku kojeg
dječaka mrsko pozdraviti?
Serbian
which. DAT AUX man. DAT which.GEN boy.GEN boring greet.ING “Which man is to greet which boy?”
Comparing these grammaticality judgments to the ones in Richards (2010) we can see that the judgments collected from the nine speakers show a mirror image to the data from Richards (2010). Speakers find example (9a) in which the two DPs have the same case and gender features and example (9b) in which the two DPs have the same case but different gender feature more acceptable than example (9c) in which the two DPs have different case and gender feature. This contradicts the claim, that DPs with different features are more acceptable than the ones with the same features. Also, the data from nine speakers speak against the higher ranking of case with respect to gender features. However the data in (9) is still problematic since Serbian speakers were not tested in a controlled environment, which means we cannot rely on accuracy of the data. In order to get accurate data I have tested speakers from Slovenian, which as Serbian and Croatian also has multiple Wh-fronting. Data from the experiment is shown in the next section.
2.2 Slovenian multiple Wh-fronting Data from Richards (2010) offers an insight into the influence of features, but in order to determine whether there is a ranking of features, we need to get more data, which I will get from Slovenian. We will first look at some general properties of Slovenian multiple Whfronting and then look at the results of an experiment testing grammaticality judgments.
2.2.1 Multiple Wh-fronting Golden (1997) in her work on Slovenian multiple Wh-fronting observed that Slovenian has the properties of both groups of Slavic languages with multiple Wh-fronting (i.e. languages with Multiply Filled Specifier, [+MFS], such as Bulgarian, and languages with no Multiply Filled Specifiers, such as Serbo-Croatian in which not all Wh-phrases are in SpecCP), which were determined by Rudin (1988). Golden (1997) proposes that Slovenian requires a [-MFS] CP structure, which was proposed for languages such as Serbo-Croatian, Czech and Polish by Rudin (1988):
9
(10) a.
CP 3 SpecCP (wh-DP) C’ 3 C … IP 3 wh-DP I’ 6
(Golden 1997)
Following Stepanov (1998), Bošković (2002) and Stjepanović (2003) another structure will be assumed. Bošković (2002) and Stjepanović (2003) claim that Serbian and Croatian have ‘real’ Wh-fronting (movement motivated by checking a [+wh] feature) only in some contexts and only in this contexts the highest Wh-phrase moves to SpecCP position. In other contexts there is no ‘real’ Wh-fronting but rather focus movement to a position immediately below CP (the AgrP position) (Stepanov 1998, following Bošković). Since only ‘real’ Wh-fronting displays Superiority effects and Slovenian does not display any Superiority effects, we can conclude, that all wh-DP are located in one strong phase (i.e. below CP) (following Stepanov 1998): (10) b.
CP 3 AgrP 3 wh-DP,wh-DP… IP 6
If we would assume (10a) we would not expect any Distinctness effects since the two fronted Wh-DPs would not be linearized in the same phase – the higher Wh-DP is located in the Specifier of CP which is a strong phase and Specifiers of strong phases are linearized with the material of the following phase. By assuming (10b) we can predict some Distinctness effects since both wh-DPs are located in a single phase (below CP) and therefore form a linearization statement. Slovenian shows that this prediction is borne out, which will be shown below.
2.2.2 Multiple Wh-fronting in Slovenian Slovenian acceptability judgments were gathered from fifteen speakers in a small experiment. The list of examples consisted of multiple Wh-fronting examples that had different case, gender and animacy combination, multiple Wh-sluicing examples and fillers.
10
The experiment included 36 examples of multiple Wh-fronting, 9 examples of multiple sluicing and 34 grammatical and ungrammatical fillers. The fifteen speakers were asked to judge the sentences using a 1 to 5 scale, 5 being completely acceptable and 1 being unacceptable. Examples were recorded and the speakers assessed the sentences after hearing them. They also had the sentences written down in the sheet on which they wrote down their judgments. As was reported for Croatian examples in Richards (2010), the differences in the acceptability are subtle but noticeable. In order to determine if the adjacency in linearization (being in the same phase) is one of the crucial factors for acceptability of Slovenian multiple Wh-questions, the tested exampled had all Wh-phrases fronted or just one of the Wh-DPs fronted, which is a valid option in Slovenian, because multiple fronting is optional in Slovenian (Golden 1997). Given that case appears to be the most important feature in Serbian and Croatian (according to Richards 2010), we can first compare examples with same or different case features on the fronted wh-words. Looking at the results, the first observation is that sentences with two wh-DPs which have the same case feature are in general less acceptable than sentences with DPs which have different case features. For example: double datives got average grades from 2.6 up to 3.3 while sentences with one dative DP and one genitive DP got responses from 3.2 up to 3.7.5
(11)
??Kateri punci
kateri tetki
ni
zoprno pomagati?
[Which girl].DAT.F [which aunt].DAT.F not-be annoying help.INF “Which girl is not annoyed to help which aunt?” (12)
?Kateri punčki
katere starke
ni
zoprno narisati?
[which girl].DAT.F [which old-lady].GEN.F not-be annoying draw.INF “Which girl is not annoyed to draw which old-lady?”
Example (12) got an average grade of 3.5 and is thus more acceptable than (11), which got an average grade of 3. From the point of view of the Principle of Distinctness this is not surprising assuming that case is relevant in distinguishing DPs in Slovenian. The linearization statement of (11) is <[DP, DAT, F] [DP, DAT, F]>, while for (12) it is <[DP, DAT, F] [DP,
5
We are interested in the contrast in the acceptability of examples with different values of features, so the marks (i.e. whether an example is graded by a star or a question mark) are assigned relative to each other (and not based on a certain scale).
11
GEN, F]>. In (12) the gender feature is apparently enough to make the two DPs distinguishable from each other. After observing the general difference between examples with DPs with different case features and DPs which have the same case feature, we can turn attention to examples with the same case features. Double accusatives in which both Wh-DPs have an inanimate grammatical feature (regardless of whether the two DPs were close or in different phases) were judged the worst with an average of 1.8:
(13)
*Kateri računalnik
uči
programer
kateri teorem?
[which computer].ACC.M.-AN teach programmer.NOM.M [which theorem].ACC.M.-AN “Which computer is being taught which theorem by the programmer?” (14)
*Kateri računalnik
kateri teorem
uči
programer?
[which computer].ACC.M.-AN [which theorem].ACC.M.-AN teaches programmer.NOM.M “Which computer is being taught which theorem by the programmer?” As both (14) and (13)6 were given comparably low grades (example (14) got on average 1.8) (14) seems to be unacceptable for reasons outside of the Principle of Distinctness: Double accusatives in Slovenian occur with the verb ‘učiti’ to teach and it seems that sentences where both the ‘what is being taught’ and ‘who is being taught’ are inanimate objects are lexically not preferred. That the reason for unacceptability is outside of the Principle of Distinctness can also be concluded based on example (13) in which the two DPs are not in one phase of the derivation and the sentence is still unacceptable. An argument that sentences such as (13) and (14) are ungrammatical for reasons outside of the Principle of Distinctness (assuming the Principle holds) is also that changing the number feature has no effect on the acceptability of these sentences (grammatical number was not tested in the experiment):
(15)
*Kateri računalnik
uči
programer
[which computer].ACC.M.-AN.SG teach programmer.NOM.M
6
Example (13) was not tested with speakers, but a comparable example was. Based on the tested example, which
got an average judgement of 1.9, a mark was assigned to (13).
12
katere teoreme? [which theorem].ACC.M.-AN.PL “Which computer is being taught which theorem by the programmer?” (16)
*Kateri računalnik
katere teoreme
uči
[which computer].ACC.M.-AN.SG [which theorem].ACC.M.-AN.PL teaches programer? programmer.NOM.M “Which computer is being taught which theorem by the programmer?”
Sentences with DPs with different number features are still unacceptable and if we assume that having at least one different feature on DPs is enough for a sentence to be acceptable this would mean that example (16) with two inanimate accusative DPs is unacceptable for reasons not related to the Principle of Distinctness (in (16) the linearization statement is <[DP, ACC, AN, SG] [DP, ACC, -AN, PL]>). Grammatical animacy feature has a bigger influence and the sentence in (16) is more acceptable than examples (13) and (14), with the average acceptability of 3.3:
(16)
? Katerega učenca
kateri teorem
uči
profesor?
[which student]. ACC.M.+AN [which theorem].ACC.M.-AN teaches professor “Which girl is Božo teaching which subject?”
This sentence is more acceptable, but it is not clear whether this is because of the lexical reasons related to the verb ‘učiti’ to teach (one DP is naming the person who is being taught, the other the subject which is taught) or is it because of the Distinctness – since the two DPs have different linearization statements: <[DP, ACC, M, +AN] [DP, ACC, M, -AN]>. With the examples like (13), (14) and (16) I was only testing for the effect of grammatical animacy, which can only be observed in Slovenian on masculine DPs in accusative singular (Toporišič 2004), but as we saw we cannot make any conclusions about the influence of grammatical animacy. In the experiment was not testing for the influence of semantic animacy.7 7
Semantic animacy could be tested on examples like (i) and (ii). Since Principle of Distinctness applies to functional heads which are also determined by semantic features, semantic features could also influence the grammaticality. Assuming the Principle of Distinctness (Richards 2010) we could predict that (i) is less acceptable than (ii), since both DPs in (i) have the same values for all the features and (ii) has different values for semantic animacy.
13
Looking at gender feature, we can observe that examples with the same case and different gender features seem to be slightly more acceptable than the ones with the same case and same gender DPs:
(17)
??Kateremu fantu
je
kateremu dedku
zoprno pomagati?
[which boy].DAT.M AUX [which grandfather].DAT.M annoying help.INF “Which boy is annoyed by helping which grandfather?” (18)
?Kateremu starčku
je
kateri teti
zoprno
pomagati?
[which old-man].DAT.M AUX [which aunt].DAT.F annoying help.INF “Which boy is annoyed by helping which aunt?”
A sentence comparable to (18) got an average response of 3, while (17) got an average response of 2.6. If we compare the linearization statements of the two examples, we can see that in (17) the linearization statement is <[DP, DAT, M] [DP, DAT, M]>, while in (17) it is <[DP, DAT, M] [DP, DAT, F]>. This means that in (18) the gender feature is enough to make the two DPs distinguishable from each other.8 Number also has an influence on the acceptability of examples in which DPs have the same case and gender features – such as (17). The influence of number was not tested in the experiment, but the intuition of a few speakers is that sentences (19) and (20) are more acceptable than (17):
(19)
?Kateremu fantu
je katerim dedkom
zoprno pomagati?
[which boy].DAT.M.SG AUX [which grandfather].DAT.M.PL annoying help.INF “Which boy is annoyed by helping which grandfathers?”
(i)
(ii)
Kateri punci kateri tetki ni zoprno pomagati? [Which girl].DAT.F [which aunt].DAT.F not-be annoying help.INF “Which girl is not annoyed to help which aunt?” Kateri punci kateri sobi ni zoprno pobarvati sten? [Which girl].DAT.F.+AN [which room].DAT.F.-AN not-be annoying paint.INF walls “Which girl is not annoyed to paint the walls of which room?”
I find example (ii) more acceptable but more speakers need to be tested to see if the prediction is borne out. Our data from the acceptability judgment test is also an argument against the structure of multiple Whquestions which was proposed by Rudin (1988) for Serbo-Croatian and Czech and Golden (1997) for Slovenian, which was shown in (10a). Since we have assumed Richards (2010) Principle of Distinctness, the graded acceptability of Slovenian multiple wh-questions can only be accounted for if both Wh-phrases are in a single phase, which is not the case for the structure in (10a). The structure in (10b), however, can account for the data. Assuming the Principle of Distinctness the data from Slovenian multiple Wh-fronting is an argument for structure (10b). 8
14
(20)
?Katerim fantom
je kateremu dedku
zoprno
pomagati?
[which boy].DAT.M.PL AUX [which grandfather].DAT.M.SG annoying help.INF “Which boys is annoyed by helping which grandfather?”
In (19) the linearization statement is <[DP, DAT, M, SG] [DP, DAT, M, PL]> and in (20) <[DP, DAT, M, PL] [DP, DAT, M, SG]>. These examples are more acceptable than (17) which suggest that number is a feature that can make DPs distinct, but since the sentences are less acceptable than examples with different case feature this suggests that number has a lesser influence. This difference in the acceptability of (17) and (18) suggests that Slovenian is the same type of a language as Croatian, shown in section 2.1, and what makes DPs the most distinct from each other seems to be different case features of DPs, however if two DPs have the same case feature and different gender feature the sentence is less, but still acceptable. From this we could conclude that there is a ranking of features in languages such as Slovenian and Croatian. The data on Serbian, section 2.1, is not conclusive since the data in Richards (2010) is drastically different from the judgments of speakers that were consulted for this paper. If the judgments of the nine Serbian speakers are such that they represent the intuition of majority of Serbian speakers, this would suggest that case is not ranked as high as it is in languages such as Slovenian or Croatian, but that it might be more important to have different gender feature. Unfortunately we don’t have enough data for such a conclusion, more work has to be done.
2. 2. 2 A note on the proximity of the two DPs One parameter that was tested with Slovenian speakers was if sentences with two Wh-DPs that have indistinguishable features are more acceptable if not all Wh-phrases are fronted. Sentences in which not all Wh-phrases are fronted are completely acceptable in Slovenian:
(21)
Kdo poje kaj? Who sings what Who is singing what?
15
According to Richards (2010) suppressing movement is one of the mechanisms used to avoid Distinctness violations (others include adding and deleting structure and movement), as shown for Croatian:
(22)
??Kojem
je čovjeku kojem
dječaku pomoči?
which.DAT AUX man.DAT which.DAT boy.DAT help.INF “Which man is to help which boy?” (23)
Kojem
je čovjeku pomoči kojem
dječaku?
which. DAT AUX man.DAT help. INF which.DAT boy.DAT ‘Which man is to help which boy?’
Avoiding movement of the lower Wh-DP divides the two DPs in different phases of the derivation, which means they do not form a linearization statement and Distinctness is not violated (Richards 2010). This mechanism was re-tested on Slovenian in which not all Wh-phrases have to move and which acts like Croatian with respect to the influence of features. Again, Slovenian behaves similarly to Croatian, since examples with non-distinguishable DPs close together are judged slightly worse than examples in which only one DP is fronted:
(24)
??Katero punco
katero snov
uči
Božo?
[which girl].ACC.F [which subject].ACC.F teaches Božo (25)
?Katero punco
uči
Božo katero snov?
[which girl].ACC.F teaches Božo [which subject].ACC.F “Which girl is Božo teaching which subject?”
This suggests that avoiding movement can improve the grammaticality of Slovenian multiple Wh-questions with Wh-phrase that have indistinguishable features.
2.3 Multiple sluicing An environment similar to the one in multiple Wh-fronting examples can be found in multiple sluicing. Following the standard understanding of sluicing, I assume that the DP remnants of the sluiced clause are located in one phase of the derivation, just above IP, and provide an environment in which we can check the influence of features. 16
Sluicing is an instance of clausal ellipsis – a sluice consists of a CP in which the sentential part (the IP) is elided (Merchant 2001 etc.):
(26)
John bought something. I wonder [CP what [ IP John bought t]]
(27)
I wonder [CP what [ IPJohn bought t]]
(Grebenyova 2007: (3))
Assuming that Slovenian Wh-movement is an instance of focus movement (as shown in section 2.2.1), the Wh-phrases are moved to a position below CP – in the Specifier of AgrP. The IP is then elided. The proposed structure is in (28):
(28)
CP 3 AgrP 3 IP wh-DP,wh-DP… 6
Given that Wh-phrases were assumed to be located in a single phase also in Wh-fronting, as discussed above, we can easily conclude that sluicing also involves Wh-phrases in the same phase. Richards' (2010) division of languages also holds for Sluicing. In English two phrases can appear together in sluicing as long as they have different labels. The ungrammaticality of (29b) can be explained using the Principle of Distinctness – there are two DPs in one phase of the derivation9, while in (29a) the two sluicing remnants have different labels (DP, PP) and are therefore distinguishable from each other.
(29) a. Everyone was dancing with somebody, but I don’t know [DP who] [PP with whom] b. *Everyone insulted somebody, but I don’t know [DP who] [DP whom] (Richards 2010: (87))
9
Merchant (2001) on the other hand has an example in which there are two DP remnants of sluicing, which implies English can allow multiple DPs in a single phase: (i) Everybody said he'd bring something different to the potluck, But I can't remember who what. (Merchant 2001: 113, fn. 4, ex. (ii))
17
Richards (2010) also discusses languages which allow multiple sluicing with two DPs, such as German, Japanese (shown in section 1.1), Dutch and Greek. In these languages, multiple sluicing is allowed as long the two DPs are distinct from each other, but the languages differ as to which features are relevant for the purpose of linearization. Greek, for example, distinguishes DPs on the basis of case and animacy (Richards 2010):
(30)
Kapjos
idhe kapjon,
alla dhen ksero
someone.NOM saw someone.ACC but not
pjos
pjon.
Greek
I.know who.NOM who.ACC
‘Someone saw someone, but I don’t know who whom.’
(Richards 2010: (96))
The grammaticality of (30) is explained assuming the Principle of Distinctness as we can say that the two DPs in (30) are distinct enough to be linearized in one Spell-Out domain, because they bare different case features. Examples of Slovenian multiple sluicing will be presented below.
2.3.1 Slovenian multiple sluicing Assuming the structure in (28) to be the structure of Slovenian multiple Wh-examples, we can check whether it is influenced by the Principle of Distinctness and if there is a pattern in the grammatical judgements, which would help us determine the influence of features. Again, fifteen speakers were tested on examples with different combinations for grammatical features for case, gender and animacy, but also examples in which the two remnants have different labels. Sentences in which the two sluices have different labels are unproblematic since they got an average grade of 4.7:
(31)
Neko punčko smo zamenjali
[z
drugo punčko], ampak ne vemo katero
[DP Some doll] AUX switched [PP with another doll] s
but
not know [DP which]
katero
[PP with which] “We switched a doll with some other doll but we do not know which doll with which doll.”
18
Examples such as (31) show that just like English Slovenian can easily linearize a DP and a PP even if the PP contains a DP that is (phonologically) identical to the DP with which the PP is linearized. Given that PP is a phase, linearization only looks at the outermost label. However, the story is more complicated with multiple DP remnants. First, we can observe case sensitivity, and at the same time, no sensitivity to phonological form (i.e. we are checking syncretic forms), which is shown in example (32).10 Sensitivity to case is also shown in (33), where the two Wh-words are also distinct morphologically:
(32)
Nekaj
pogojuje nekaj,
pa
Something.NOM conditions something.ACC, but
ne vem kaj
kaj.
not know what.NOM what.ACC
“Something is conditioning something, but I don’t know what conditions what.” (33)
Nečemu
smo nekaj
dodali, ampak ne spomnim se,
Something.DAT AUX something.ACC added, but
čemu
not remember REFL who.DAT
kaj what.ACC “We added something to something, but I don’t remember what to what.”
Sentences like (33) with different case features on the two DPs are as completely acceptable. This means, that assuming the Principle of Distinctness, the two DPs even though they share the same phase of the derivation, are distinct because they have different case features (the unproblematic linearization statement for (33) is <[DP, DAT] [DP, ACC]>. Judgments of examples in which the two DPs have the same case feature are presented below and in general these examples are less acceptable. Example (34) got an average response of 1.9; example comparable to (35) got an average of 2.2; and example (36) got a response of 2.7:
(34)
*Nekemu moškemu smo pomagali podariti darilo nekemu fantu [some
man].DAT
AUX
helped
give
gift
ampak ne vemo
[some boy].DAT but
not know
10
There are languages, such as Greek, which are sensitive to syncretism, however, this, according to Richards (2010) is not because of the phonological form, but because “syncretism must involve manipulation of the syntactic representation prior to lexical insertion” (Richards 2010: 48): (i)
*Ksero oti kathe agori espase ena parathiro, alla dhen ksero pio pio (Greek) I-know that every boy.NEUT broke one window.NEUT but not I-know which.NEUT which.NEUT ‘I know that every boy broke one window, but I don’t know which which’ (Richards 2010: (97a))
19
kateremu kateremu. which. DAT which. DAT (35)
*/?Neki punčki
smo pomagali podariti darilo neki starki
[Some girl].DAT AUX helped kateri
give
gift
ampak ne vemo
[some old-lady].DAT but
not know
kateri.
which.DAT which. DAT (36)
?Nekemu stricu
se ne ljubi
pisati
neki punci,
ampak ne vemo
[Some uncle]. DAT.M REF not feel-like write.INF [some girl]. DAT.F but not know kateremu
kateri.
which.DAT.M which.DAT.F “Some uncle does not feel like writing to some girl, but we don’t know which uncle to which girl.”
Examples with two DPs with the same case feature, (34)-(36), are generally judged worse than examples with different case features, (32), (33). Examples that have the same case feature but different gender feature, (36), is judged better than the ones that have the same case and gender feature, (33) and (34). Assuming the Principle of Distinctness, we can say that in the examples with same case and gender features the two DPs are indistinguishable from each other, for example the linearization statement for (35) is <[DP, DAT, F] [DP, DAT, F]>. In example (36), on the other hand, the two DPs are distinguishable by their gender feature, the linearization statement being <[DP, DAT, M] [DP, DAT, F]>. If this explanation is on the right track, than gender features effect the linearization and consequently the acceptability of multiple DPs in a single phase. If we compare the acceptability of examples (32) and (33) to examples (34) to (35) the results suggest, that case is still ranked higher than gender since examples with different case are more acceptable than examples with the same case but different gender (examples (32) and (33), which have different case, are more acceptable than (36), which has same case different gender). The experiment also checked the influence of grammatical animacy features in Slovenian multiple sluicing.11 Examples with the same case, gender and animacy features were graded as completely unacceptable (average response 1,7):
11
Semantic animacy was not tested. Additional tests should be conducted in order to investigate the judgments of sentences such as (i) and (ii), however, at first glance they seem to be equally unacceptable:
20
(37)
*Tone uči
nek računalnik
nek program
ampak ne
Tone teaches [some computer].ACC.M.-AN [some program].ACC.M.-AN but vem
not
katerega katerega.
know which.ACC which.ACC
The unacceptability of (37) could be explained also with the Principle of Distinctness. The critical linearization statement for (37) is <[DP, ACC, M, -AN] [DP, ACC, M, -AN]>, the two DPs are indistinguishable, they are both DPs, they have the same case feature, same gender and same animacy feature, but given that for comparable multiple Wh-fronting examples we said their ungrammaticality cannot be attributed to the Principle of Distinctness, but rather to some lexical properties of the verb učiti 'to teach', we cannot conclude the Principle of Distinctness was at play in (37). Changing the animacy of one of the two DPs, makes the sentences acceptable. In terms of the Principle of Distinctness this would mean that the example (38) produces a linearization statement where the two DPs can be distinguished: <[DP, ACC, M, +AN] [DP, ACC, M, -AN]>.
(38)
Tone uči
nekega fanta
neki program
ampak ne vem
Tone teaches [some boy].ACC.AN [some program].ACC.-AN but katerega
not know
katerega.
which.ACC.+AN which.ACC.-AN “Tone is teaching some boy some program, but I don’t know which program to which boy.”
(i)??
(ii)??
Neki punčki smo pomagali pomagati neki starki, ampak ne vem, [some girl].DAT.F.+AN AUX helped help.INF [some old-lady] DAT.F.+AN but not know kateri kateri. which.DAT.F.+AN which. DAT.F.+AN “We have helped some girl with helping some old lady, but we don’t know which girl and which oldlady.” Neki punčki smo pomali prebeliti stene eni sobi, ampak ne vem, [some girl].DAT.F.+AN AUX helped paint.INF walls [some room] DAT.F.-AN but not know kateri kateri. which.DAT.F.+AN which. DAT.F.-AN “We helped some girl paint the walls of some room, but we do not know to which girl and which room.”
21
Grammatical number on the other hand does not influence the grammaticality of double accusative sluices to the same extent as animacy (these examples were not tested with speakers):
(39)
?Tone uči
nek računalnik
neke programe
ampak ne
Tone teaches [some computer].ACC.M.-AN.SG [some program].ACC.M.-AN.PL but vem
katerega
not
katerega.
know which.ACC.SG which.ACC.PL “Tone is teaching the computers some programs but I don’t know which programs to which computer.”
(39) is more acceptable than (37), but less acceptable than (38). This again suggest that (37) is ungrammatical for reasons outside of Distinctness. Since we saw that sentences with two DPs which both have accusative and inanimate grammatical features might be unacceptable for reasons outside of Distinctness, we can check for the influence of number on other type of sentences with two DPs with the same case and gender:
(40)
?Neki punčki
smo pomagali podariti darila nekim starkam
[Some girl].DAT.SG AUX helped vemo kateri
give
gifts
ampak ne
[some old-lady].DAT.PL but
not
katerim.
know which.DAT.SG which. DAT.PL “We helped some girl give gifts to some old ladies, but we do not know which girl to which old ladies.” (41)
?Nekim punčkam
smo pomagali podariti darila neki starki
[Some girl].DAT.PL AUX helped vemo kateri
give
ampak ne
gifts [some old-lady].DAT.SG but
not
kateri.
know which.DAT.PL which. DAT.SG “We helped some girls give gifts to some old lady, but we do not know which girls to which old lady.”
Sentences in (40) and (41) are more acceptable than (35), but still not completely acceptable. This shows that grammatical number has some influence on the acceptability of sentences with two DPs with the same case and gender features (in (40) the linearization statement is 22
<[DP, DAT, F, SG] [DP, DAT, F, PL]> and we can see that number is the only different feature in the feature bundle and it can only be feature that makes two DPs distinct). The Slovenian multiple sluicing examples show that features do play a role in the linearization (that they are visible at the point of linearization) and that there is some ranking of features – examples with DPs that are distinguishable by case are the most acceptable. In order to look into the status of features further, the next section deals with double object in Slovenian – with the emphases on double accusatives, which allow us to focus on features other than case.
2.4 Double objects Since vP is a phase, double objects with their two DPs inside a single VP should also be subject to the Principle of Distinctness. Just like multiple sluicing, double object constructions also involve two DPs in a single phase of the derivation. In case of the double object construction this phase is the vP (cf. Larson 1988, Chomsky 1995, 2000, Richards 2010 etc.). The nominative DP poses no problem to the Principle of Distinctness, since it is on the edge of the phase (Specifier of vP), the lower two DPs (objects) however are interesting from the point of the theory, since both DPs are in the same strong phase. Following Richards (2010) languages such as English allow double objects. He contrasts low applicative (42a) to ‘prepositional datives’, (42b):
(42)
a.
I gave John a book.
(Richards 2010: 188a)
b.
I gave a book to John.
(Richards 2010: 188b)
(42b) is not problematic from the point of view of the Principle of Distinctness – the indirect object is a prepositional phrase and Richards (2010) assumes PPs are strong phases, which means the two DPs are located in two different phases. (42a) on the other hand does involve two DPs in the same transitive vP phase at one point of the derivation. But following Richards (2010) when a VP contains two DPs, low applicatives attach to VP (Pylkkänen (2000)), one of them must move to a higher Specifier of a phase head. This means that DP John in (42a) moves from the AppP to the specifier of transitive vP, which is a strong phase – because the DP John is now in specifier vP it linearizes with the material of the following phase and Distinctness is not violated. Slovenian, as English, also allows double objects: 23
(43)
Maja
pošilja pisma
Janezu.
Maja.NOM sends letter.ACC Janez.DAT “Maja is sending letters to John.”
As seen in example (43) double objects where two DPs differ in case are completely acceptable, which is can be a consequence of different features on the two DPs but also one DP moving on the edge of transitive vP (for a discussion of low and high applicatives in Slovenian see (Marvin 2011)). However, Slovenian also has sentences with two DPs, which have the same case feature, i.e. double datives and double accusatives. These examples will help us understand what role in Distinctness is played by gender and animacy features. The main focus will be on double accusatives since more Slovenian speakers find them acceptable (double datives are only acceptable in some dialects12).
2.4.1 Double accusatives Not all languages that distinguish DPs based on features allow double accusatives. In Japanese, for which it was shown in 1.1 that it makes distinction based on features, there is a ban on two accusatives in the same domain, which can be accounted for using Richards (2010) Distinctness, as it was shown by Nevins (2010). Japanese allows dative, (45), or accusative, (44), objects next to causative verbs. There is a semantic difference – dative objects express semantics of non-coercive causation or permission:
(44)
Taroo-wa Hanako-o
ikaseta
Taro.TOP Hanako.ACC caused-to-go “Taro made Hanako go.” (45)
(Nevins 2010: (14))
Taroo-wa Hanako-ni ikaseta Taro.TOP Hanako.DAT caused-to-go
12
(i)
(ii)
Vrgla ti bom žogo Zoji. thrown cl.2.SG.DAT AUX.1.SG. ball.ACC Zoja.DAT “I’ll throw Zoja the ball for you” ??Binetu bom vrgla žogo Zoji BineDAT AUX thrown ball.ACC Zoja.DAT “I’ll throw Zoja the ball for Bine.”
(Marvin 2011: (15a))
(Marvin 2011: (16))
According to Marvin (2011) both dative DPs can be in the same phase, however, these constructions will not be a part of the discussion on double objects in Slovenian.
24
“Taro had/let Hanako go.”
(Nevins 2010: (15))
With embedded transitive verbs, the causative verb only allow the dative object, which looses it’s semantics of non-coercive causation or permission (Nevins 2010).13
(46)
Taroo-wa Hanako-ni / *Hanako-o
kusuri-o
nomaseta
Taro.TOP Hanako.DAT/ Hanako.ACC medicine.ACC caused-to-drink Taro had/made/let Hanako drink the medicine.
(Nevins 2010: (16))
This shows that not all languages which are not just sensitive to labels, but to grammatical features, allow constructions such as double accusatives. Other languages, such as Greek and Slovenian, which are sensitive to features, allow double accusatives:
(47)
Didaksa
[NP ta pedhia]
[NP tin grammatiki ton Arxeon]
Greek
Taught-1sg [the children.ACC] [ the grammar.ACC the ancient] “I taught the children the grammar of Ancient Greek.” (Anagnostopoulou 2000: (11)) (48)
Tone
uči
Majo
matematiko.
Tone.NOM teaches Maja.ACC mathematics.ACC “Tone teaches mathematics to Maja.”
Such constructions present a problem for the principle of Distinctness as stated in Richards (2010) since double accusatives also have no applicative head (Anagnostopoulou 2000) and since we can assume that both accusative DPs occur in the same strong vP phase (as shown in (49) (taken from Anagnostopoulou 2000), and both share the same case feature, which makes them non-distinct (in (48) both objects have the same gender and case features, but it might be semantic animacy14 that makes the two DPs distinct).
13
Inserting the adverb makes no difference (i) *Taroo-wa Hanako-o tikarazukude kusui-o nomaseta Taro.TOP Hanako.DAT forcibly medicine.ACC caused-to-drink 14 In the experiment I was not testing for semantic animacy, but examples such as (43) show, that semantic animacy does make DPs distinct. Despite this, we will only be focusing on grammatical features and their influence.
25
(49)
vP 3 Agent v’ 3 v VP 3 DP goal V’ 3 V DP theme
A closer look at Slovenian double accusatives shows, that there are restrictions on accusative DPs and it seems that accusative DPs must differ in at least one feature. Examples with two accusatives of the same gender are unacceptable:
(50)
*Učitelj uči
Majo
Zofko Kvedrovo.
Teacher teaches Maja.ACC.F [Zofka Kveder].ACC.F (51)
??Učitelj uči
Toneta
Prešerna.
teacher teaches Tone.ACC.M Prešeren.ACC.M “The teacher teaches Tone about Prešeren.” (Receives an unwanted reading: “The teacher teaches someone named Tone Prešeren.”) (52)
Učitelj
uči
Majo
Prešerna.
Teacher teaches Maja.ACC.F Prešeren.ACC.M “The teacher teaches Maja about Prešeren.” The unacceptability of (50) and (51) can be accounted for using the Principle of Distinctness. The linearization statements of the double accusatives are undistinguishable, in example (50) the statement is <[DP, ACC, F] [DP, ACC, F]>.15 In (52) the two DPs have different features and the statement <[DP, ACC, F] [DP, ACC, M]> can be linearized.
15
Sentence (46) is more acceptable if the two accusative DPs are in different phases. In (i) the accusative DP Prešerna is left in vP, while Toneta is moved outside the vP, above the clitic: (i)
Toneta smo učili Prešerna. Tone.ACC.M.+AN AUX teach Prešeren.ACC.M.+AN “We have taught Tone about Prešeren.”
26
Example (52) shows that different gender feature can make two DPs distinguishable and it seems the same can be achieved with different animacy features:
(53)
*Tone uči
računalnik
šah.
Tone teaches computer.ACC.M.-AN chess. ACC.M.-AN (54)
Tone uči
Janeza
šah.
Tone teaches Janez. ACC.M.MAN chess. ACC.M.-AN “Tone teaches chess to Janez.”
One way to account for (53) is to claim that the two DPs are indistinguishable since both DPs have the same values for the features in the feature bundle, while in (54) the DPs have distinct linearization statement <[DP, ACC, M, -AN] [DP, ACC, M, +AN]> and the sentence is acceptable. But there are also other problems with double inanimate accusatives. As we mentioned with multiple Wh-fronting of accusative DPs – double accusatives occur next to the verb ‘učiti’ to teach and sentence in (53) might be unacceptable because of the lexical inanimate recipient who is being taught a subject (which is usually inanimate). Double accusative object also display effects of number, since sentences like (55) are more acceptable than the sentence in (51):
(55)
Učitelj uči
učence
Prešerna.
Teacher teaches students.ACC.M.+AN.PL Prešeren.ACC.M.+AN.SG “The teacher is teaching the students about Prešeren.”
This suggests that DPs can be different in any feature for the sentence to be acceptable, since the example (55) is acceptable and the only difference between the two accusative DPs is in the number feature. This means that grammatical number is one of the features that are important at linearization. What double accusatives have shown is that when two DPs have the same value of the case feature, the sentences are only acceptable when the two DPs are distinguishable in at least one other feature, which for Slovenian is gender, number or animacy (with DPs which have masculine gender feature).
27
3 Are values of features always important?
One question that is mentioned, but not answered in Richards (2010), is why some languages are sensitive to labels (e.g. English) and others to values of their grammatical features (e.g. Slovenian, Japanese…). One possibility that Richards (2010) mentions is the richness of Case morphology in languages such as German, Greek, Japanese (and we can add Slovenian) compared to English. However Richards (2010) also notes that such accounts were not successful in the past. In this section we will try to account for the difference between the two groups of languages. As shown in Richards (2010) Distinctness only affects functional heads, which enter the derivation as feature bundles, and that Distinctness effects arise before Vocabulary Insertion. Functional heads are therefore represented only as feature bundles and, according to Richards (2010), languages are different with respect to the richness of feature bundles, which content is defined with syntactic and semantic16 features (Harley and Noyer 1999). This paper only focuses on the syntactic (grammatical) features and I will try to show that the difference between languages happen because of the organization of grammatical features (which is also seen in the morphology of a language such as Slovenian). With respect to which features languages are sensitive to we can assume, based on the data so far (and also based on the data presented in Richards 2010), that there is a hierarchy between the features that languages take into account at linearization. The suggested hierarchy takes into consideration that some languages are sensitive to the ‘category’ feature – if two phrases have the same category (for example D) and form a linearization statement <[DP], [DP]> the sentence is ungrammatical; other languages look to other features. Based on data from Slovenian we can determine that these features include case, gender, animacy and grammatical number. Slovenian data has also shown that there is a hierarchy between these features, which can be observed in the grammaticality judgments from the experiment. Sentences in which the two DPs have different case feature, such as (56) below, are the most acceptable (for example, in the experiment sentences with double datives never got an average grade higher than 3.3 (regardless of gender of DPs) while sentences with a nominative DP and an accusative DP got grades from 3.4 to 4.6), which suggests case is ranked high. Looking again at sentences with two DPs that have the same case feature, but 16
Semantic animacy was not tested with speakers so its effects are not clear. Grammatical animacy was the only interest, since it is morphologically visible for male gender singular, while semantic animacy is not reflected in the morphology. To understand whether semantic features also play a role at linearization, more tests need to be conducted.
28
different gender on DPs, as (57) below, we can see, that on average these type of sentences got a slightly higher grade (average above 3.1 up to 3.3), than sentences with two DPs with the same case and gender (from 2.6 up to 3), as shown in (58).17
(56)
Kateri punčki
katere starke
ni
zoprno narisati?
[which girl].DAT.F [which old-lady].GEN.F not-be annoying draw.INF “Which girl is not annoyed to draw which old-lady?” (57)
?Kateremu starčku
je
kateri teti
zoprno
pomagati?
[which old-man].DAT.M AUX [which aunt].DAT.F annoying help.INF “Which boy is annoyed by helping which aunt?” (58)
??Kateremu fantu
je
kateremu dedku
zoprno pomagati?
[which boy].DAT.M AUX [which grandfather].DAT.M annoying help.INF “Which boy is annoyed by helping which grandfather?”
Lower acceptability of sentences with same case different gender combination suggests that case is ranked higher than gender. As we saw in discussion about animacy, nothing conclusive can be decided based on Slovenian data, since ungrammaticality of double inanimate accusatives might follow from lexical properties of the verb. Grammatical number also has an influence on linearization of DPs, but since the influence of number was not tested in the experiment, it is hard to determine the position of number in the hierarchy. Based on the data from the experiment we can assume the following hierarchy: (category) case, gender (animacy, gender). This is also supported by Croatian data from Richards (2010), shown in section 2.1. The question is then, how these features are organized into a hierarchy. I propose the features are organized as shown in (59).
(59)
category case ൦ ൪ gender (animacy/number)
Following Embick and Noyer (2007) and the theory of Distributive morphology this is the organization of feature bundles of abstract morphemes prior to Vocabulary Insertion. A DP is then represented as shown in (60), where nP includes the root, which does not include any
17
Sentences are again marked with respect to each other.
29
grammatical information (it does involve phonological features which we will leave aside). Semantic information is then also a part of the features of the abstract root, which means that we can expect that semantic animacy will influence grammaticality of a sentence (Embick and Noyer 2007).
(60)
DP 3 ሾD, featuresሿ nP 5 What I propose is that the features above are universally in a fix order, which is shown in (59), and the Principle of Distinctness checks the features of DPs, which are in the same phase of the derivation, top to bottom. The linearization in Slovenian therefore proceeds in the following manner: In (56) the corresponding linearization statement is <[DP, DAT, F, +/-AN] [DP, GEN, F, +/-AN]>. The comparison of the statement first checks the highest feature of both DPs, category, and then proceeds to the next level, case – the values for case are different making the nodes distinct. Further information is not important. This is shown in (61):
(61)
XP 3 DP X’ 3 6
D ۲܂ۯ ൦ ൪ FEM +/−AN
NP
3 DP 6 3
D ۵۳ۼ ൦ ൪ FEM
+/−AN
NP
In (57) the linearization statement is <[DP, DAT, M, +AN] [DP, DAT, F, +/-AN]> and the nodes are distinct on a lower, gender, level and the linearization is successful regardless of the fact that the grammatical animacy feature on the female DP is unvalued. This is shown below:
30
(62)
XP 3 DP X’ 3 6
D DAT ൦ ൪ ۻ
+/−AN
3
NP
DP 6 3
D DAT ൦ ൪ ۴۳ۻ
+/−AN
NP
In (58) the comparison proceeds beyond case, but no valued features are different <[DP, DAT, M, +AN] [DP, DAT, M, +AN]>. The derivation crashes. I do not claim the graded grammatical judgments in (56)-(58) mean that it universally holds that the lower the distinct feature is, the more ungrammatical the sentence. Evidence against such a claim, come from Japanese grammatical judgments shown in section 1.2 and is repeated below:
(63)
*[Sensei-ga
suki na] gakusei-ga koko-ni
teacher.NOM like
oozei iru kedo, dare-ga
student.NOM here.DAT many be but
dare-ga ka
who.NOM who.NOM Q
oboeteinai remember.NEG ‘There are lots of students here who like teachers, but I don’t remember who who.’ (Richards 2010: (89b)) (64)
[Doobutsu-ga suki na] hito-ga Animal.NOM like
koko-ni oozei iru kedo, dare-ga nani-ga
ka
person.NOM here.DAT many be but who.NOM what.NOM Q
oboeteinai remember.NEG ‘There are lots of people here who like animals, but I don’t remember who what.’ (Richards 2010: (90)) (65)
Watashi-wa dono otokonoko-ni-mo hoshigatteita subete-no hon-o I.TOP
every boy.DAT
wanted
nani-o ka wasureta what.ACC Q forgot 31
ageta ga dare-ni
every book.ACC gave but who.DAT
‘I gave every boy all the books he wanted, but I’ve forgotten who what.’ (Richards 2010: (88a))
When two DPs have the same case and animacy feature in Japanese, the sentence is ungrammatical, as shown in (63). In Japanese being distinct in a lower ranked feature, such as animacy, (64), does not mean, that the sentence is less grammatical than a sentence in which a higher ranked feature on two DPs is different, (65). On the other hand we can make a prediction that in languages that behave as Slovenian and have graded acceptability of sentences. In these languages sentences that have two DPs in a single phase and are more acceptable will have DPs that are different in a higher ranking feature. Sentences that have two DPs in a single phase and are less acceptable will have DPs that are distinct in a lower ranked feature. The remaining question is why there are two groups of languages. I suggest that in English not all features are valued, but are still present in the feature bundle – in DPs the only feature which is always valued is category and therefore the Principle of Distinctness only checks for category. Still there is some evidence for the presence of the unvalued features, specifically case can be valued (who–whom). When a high ranked feature such as case is present, it is taken into consideration:
(66)
Everybody said he'd bring something different to the potluck. But I can't remember who what.
(Merchant 2001: 113, fn. 4, ex. (ii))
In this case the Principle of Distinctness checks the features top to bottom and because the case features are different, the derivation is successful. Still the question is why for example number is never taken into consideration at linearization18 in English. One possible answer is, that it is to low in the hierarchy and since there are unvalued features on top of it, the Principle of Distinctness stops checking the hierarchy before reaching number. Using a hierarchy of features, in which some features are valued and others are not, we avoid the division of languages into two groups and rather attribute all languages and the variation between them to a difference in the value of features in the feature bundle. The proposed hierarchy also explains graded acceptability in languages such as Slovenian and Croatian. 18
There is no evidence in Richards (2010) for the acceptability of sentences such as (i) below: (i) Everyone invited everyone, except John Marys.
32
5 Conclusion
The paper shows that Slovenian follows the principle of Distinctness (Richards 2010). This is demonstrated with the data acquired from the grammaticality judgments of Slovenian speakers for multiple Wh-fronting, multiple sluicing and also with the data from double accusatives. This gives an additional argument for the Principle of Distinctness especially in multiple Wh-fronting, since the argument in Richards (2010) was based only on judgments of a few speakers for Serbian and Croatian. Slovenian also confirms the Principle of Distinctness in general. It was shown that based on the data form Slovenian and other languages we can conclude that languages do not behave differently with respect to the Principle of Distinctness. We can assume a hierarchy of features that form a class of features responsible for linearization. At the top of the hierarchy is the category feature, followed by case, gender, animacy and number. Languages differ as to which features have are valued and are taken into consideration at the point of linearization. The Principle of Distinctness then compares (top to bottom) the feature bundles of DPs that are in the same phase of the derivation. The proposed hierarchy explains the graded acceptability of Slovenian and Croatian examples since it ranks features, which make sentences more acceptable when they are distinct. When the distinction is in the higher features (e.g. case) the sentence is more acceptable, than when lower feature is distinct. We can also predict that other languages with graded acceptability will follow the proposed ranking. This means that in such languages sentences whith multiple DPs in a single phase will be more acceptable when the DPs are distinct in a feature which is higher in the hierarchy. On the other hand there are no arguments that all languages will have graded acceptability (for example Japanese only has acceptable or unacceptable sentences regardless of the feature that is different).
33
Appendix: List of items used in the experiment The experiment included 36 examples of multiple wh-fronting, 9 examples of sluicing and 34 fillers of which 12 were ungrammatical, (68)-(79) and 22 grammatical, (46)-(67).
I. Multiple Wh-fronting examples (1)
Kateremu fantu
je
zoprno pomagati kateremu stricu?
[Which boy].DAT.M AUX annoying help.INF [which uncle].DAT.M (2)
Kateremu fantu
je
kateremu dedku
zoprno pomagati?
[Which boy].DAT.M AUX [which grandfather].DAT.M annoying help.INF (3)
Kateri sosedi
je
fino pomagati kateri deklici?
[Which neighbor].DAT.F AUX good help.INF [which girl].DAT.F (4)
Kateri punci
kateri tetki
ni
zoprno pomagati?
[Which girl].DAT.F [which aunt]. DAT.F not-be annoying help.INF (5)
Kateremu žrebetu
se ne ljubi
nagajati kateremu teletu?
[Which stallion].DAT.N REF not feel-like tease.INF [which calf].DAT.N (6)
Kateremu teletu
se
kateremu žrebetu
ne ljubi
dati
hrane?
[which calf].DAT.N REFL [which stallion].DAT.N not feel-like give.INF food (7)
Kateremu kmetu
se kateremu teletu
ne ljubi
pospraviti hleva?
[Which farmer].DAT.M REFL [which calf].DAT.N not feel-like clear.INF stable (8)
Kateremu starčku
se
kateri teti
ne ljubi
slediti?
[which old-man].DAT.M REFL [which aunt].DAT.F not feel-like follow.INF (9)
Kateri teti
se
ne ljubi
pospraviti hleva kateremu teletu?
[which aunt].DAT.F REFL not feel-like clear.INF stable [which calf].DAT.N (10)
Kateri teti
se
ne ljubi
pomagati kateremu stricu?
[which aunt].DAT.F REFL not feel-like help.INF [which uncle].DAT.M (11)
Kateremu stricu
ni
odveč
objeti
katerega nečaka?
[which uncle].DAT.M not-be too-much hug.INF [which nephew].GEN.M (12)
Kateremu prijatelju
katerega fanta
ni
odveč
objeti?
[which friend].DAT.M [which boy] GEN.M not-be too-much hug.INF (13)
Kateri punčki
katere starke
ni
zoprno narisati?
[Which girl].DAT.F [which old-lady].GEN.F not-be annoyed draw.INF (14)
Kateri trgovki
ni
zoprno prodajati katere čokolade?
[Which merchant].DAT.F not-be annoyed sell.INF [which chocolate].GEN.F
34
(15)
Kateremu fantu
katerega žrebeta
ni
zoprno jahati?
[Which boy].DAT.M [which stallion].GEN.N not-be annoyed riding (16)
Kateri računalnik
uči
programer
kateri program?
[which computer].ACC.M.-AN teach programmer.NOM.M [which program].ACC.M.-AN (17)
Katerega strica
uči
Mojca kateri teorem?
[Which uncle].ACC.M.+AN teaches Mojca [which theorem]. ACC.M.-AN (18)
Kateri računalnik
kateri teorem
uči
programer?
[which computer].ACC.M .-AN [which theorem].ACC.M.-AN teaches programmer.NOM.M (19)
Katerega učenca
katerega avtorja
uči
učitelj?
[Which student].ACC.M.+AN [which author].ACC.M.+AN teaches teacher (20)
Katerega učenca
kateri teorem
uči
profesor?
[Which student].ACC.M.+AN [which theorem].ACC.M.-AN teaches professor (21)
Katero punco
uči
Božo katero snov?
[which girl]. ACC.F teaches Božo [which subject] ACC.F (22)
Katerega učenca
katero snov
učimo?
[Which student].ACC.M.+AN [which subject].ACC.F teach (23)
Katero punco
katerega avtorja
uči Miha?
[which girl].ACC.F [which author].ACC.M.+AN teaches Miha (24)
Katero učenko
uči
Božo katerega avtorja?
[Which student].ACC.F teaches Božo [which author].ACC.M.+AN (25)
Kateri dedek
je mislil
kateri vojak,
da je umrl?
[Which grandfather].NOM.M. AUX thought [which solider]. NOM.M that AUX died (26)
Kateri fant
kateri poba
pravi, da je bedak?
[Which boy].NOM.M [which guy].NOM.M says, that is a jerk (27)
Katero pleme
kateri minister
upa,
da pride mimo?
[which tribe].NOM.M [which misiter].NOM.M hopes, that comes by (28)
Katerega fanta
kateri članek
ne zanima?
[Which boy].ACC.M [which article].GEN.M not interested (29)
Katere punce
katera revija
ne privlači?
[which girl].ACC.F [which magazine].GEN.F not attract (30)
Katerega profesorja
ne zanima
kateri članek?
[which professor].ACC.M not interested [which article].GEN.M (31)
Kateri tetki
je
všeč katera tortica?
[which aunt].DAT.F AUX like [which cake].NOM.F 35
(32)
Kateri zdravnici
katera ordinacija
ni
všeč?
[which doctor].DAT.F [which examination-room].NOM.F not-be like (33)
Kateremu profesorju
katera snov
ni
všeč?
[which professor]DAT.M [which subject].NOM.F not-be like (34)
Kateri fant
katerega voditelja
posluša?
[Which boy].NOM.M [which anchor]. ACC.M listens (35)
Kateri fant
kateri dnevnik
bere?
[Which boy].NOM.M.+AN [which journal].ACC.M.-AN reads (36)
Kateri tiskalnik
kateri roman
tiska?
[Which printer].NOM.M.-AN [which novel].ACC.M.-AN prints
II. Multiple Sluicing Examples (37)
Nekaj
pogojuje
nekaj,
pa ne vem, kaj
kaj.
Something.NOM conditions something.ACC, but not know what. NOM what.ACC (38)
Neki punčki
smo pomagali pomagati neki starki,
[Some girl].DAT.F AUX helped kateri
ampak ne vem,
help.INF [some old-lady].DAT.F but
not know
kateri
which.DAT which.DAT (39)
Neka teta
je
oropala eno pošto,
a ne vemo, katera
[Some aunt].NOM.F AUX robbed [some post].ACC.F but not know which.NOM.F katero. which.ACC.F (40)
En fant
ima rad nekoga,
[Some boy].NOM.M loves (41)
pa ne vemo, kdo
somebody.ACC.M but not know who.NOM who.ACC
Nekemu moškemu smo pomagali podariti darilo nekemu fantu [some
koga.
man].DAT AUX helped
give
gift
ampak ne vemo
[some boy].DAT but
not know
kateremu kateremu. which.DAT which.DAT (42)
Nekemu stricu
se
ne ljubi
pisati
neki punci,
ampak ne vemo
[Some uncle].DAT.M REFL not feel-like write.INF [some girl].DAT.FF but not know kateremu
kateri.
which.DAT.M which.DAT.F (43)
Tone uči
nek računalnik
nek program
ampak ne
Tone teaches [some computer].ACC.M.-AN [some program].ACC.M.-AN but 36
not
vem
katerega katerega.
know which.ACC which.ACC (44)
Neki punčki
smo pomagali zapeti pesem o
[Some doll].NOM.F AUX help
sing
nekem dedku,
ampak ne vem
song about some grandfather but not know
kateri o kateremu. which about which (45)
Neko punčko
smo zamenjali z
drugo punčko, ampak ne vemo katero
[Some doll].NOM.F AUX switched with another doll s
but
not know which.DAT
katero
with which
III. Fillers (46)
Kaj kaj napoveduje?
(47)
Kateremu očetu se zdi prav otrokom pomagati?
(48)
Najmanjšemu teletu se lušta kravi nagajati.
(49)
Čigavi sestri se ljubi vsak dan Tončki pomagati?
(50)
Tistemu velikemu človeku se zdi prav hiši zgraditi balkon.
(51)
Debelemu mesarju se ne zdi prav kateremu pujsu oprati parkljev?
(52)
Katero učenko je učil Tone matematike?
(53)
Kateri fant je katero punco udaril s čim po glavi?
(54)
Jaz se ljubi svojo obleko zlikati.
(55)
Ta rjavemu žrebetu se je travni bilko.
(56)
Moj računalnik se ne ljubi mojega pisma natisniti.
(57)
Katera soseda je ponoči kradla kje paradižnik?
(58)
Katerega zdravnika ni nihče napadel?
(59)
Katerega učenca je profesor učil osla?
(60)
Čigavo zdravnico je napadel pobesneli bolnik?
(61)
Tista punčka pravi, da ne ve nič o nekem avtu.
(62)
Neki stric je ukradel točno določeno uro včeraj popoldan.
(63)
Tine je včeraj učil katero učenko katero snov?
(64)
Kakšno je sladico spekla tista tetka?
(65)
Katero knjigo je kupil Tone čigavemu bratu?
(66)
Kateri ta prvi izvod knjige si kupil svojim mami?
(67)
Tone pravi katerega fanta, da ne vidi? 37
(68)
Profesor je učil učenca osla.
(69)
Čigava punca katerega je fanta gledala z velikimi očmi?
(70)
Čigava je kateri konček padel po stopnicah?
(71)
Katero voznico je policija ustavila vso pijano?
(72)
Jaz se ljubi svojo obleko zlikati.
(73)
Katera mama se očetovi sestri pomaga?
(74)
Katerega računalnika se ne ljubi natisniti pisma?
(75)
Kam je Tone rekel, kje so Mihata napadli?
(76)
Suhemu mesarju se pujs oprati parkljev da.
(77)
Peter je povedal, da so zaposlili nekoga iz ene od afriških držav, ampak ne vem katere.
(78)
Zadnjič sem poskusila en dober sladoled, ampak ne vem katerega.
(79)
Koga je Peter rekel, ko je Marija udarila?
38
References
Adger, David and Peter Svenonius. 2010. Features in minimalist syntax. In The Oxford Handbook of Linguistic Minimalism, ed. Cedric Boeckx, 27–51. Oxford, New York : Oxford University Press. Anagnostopoulou, Elena. 2000. Two classes of double object verbs: The role of zero morphology. Paper presented at the 23rd GLOW Workshop on Null/Overt Morphology, Bilbao, April 19. Bošković, Željko. 1999. On multiple feature-checking: Multiple wh-fronting and multiple head movement. In Working Minimalism, ed. Samuel Epstein and Norbert Hornstein, 159187. Cambridge, MA: MIT Press. Bošković, Željko. 2001. On the nature of the syntax-phonology interface. Clitization and related phenomena. Amsterdam: Elsevier. Bošković, Željko. 2002. On multiple wh-fronting. Linguistic Inquiry 33: 351-383. Chomsky, Noam. 1965. Aspects of the theory of syntax. Cambridge, MA.: MIT Press. Chomsky, Noam. 1995. The Minimalist Program. Cambridge, MA.: MIT Press. Chomsky, Noam. 2000. Minimalist inquiries: The framework. In Essays on minimalist syntax in honor of Howard Lasnik, ed. Roger Martin, David Michaels and Juan Uriagereka, 89– 155. Cambridge, MA: MIT Press. Embick, David and Rolf Noyer. 2006. Distributed Morphology and the Syntax/Morphology Interface. In The Oxford handbook of linguistic interfaces, ed. Gillian Ramchand and Charles Reiss, 289-324. Oxford; New York : Oxford University Press. Golden, Marija. 1997. Multiple wh-questions in Slovene. In: Formal Approaches to Slavic Linguistics: The Cornell Meeting, 1995, ed. Wayles Browne, Ewa Dornisch, Natasha Kondrashova, and Draga Zec, 240-266. MI: Michigan Slavic Publications, Ann Arbor. Grebenyova, Lydia. 2007. Sluicing in Slavic. Journal of the Slavic Linguistic Society 15: 4981. Halle, Morris and Alec Marantz. 1994. Some key features of Distributed Morphology. In Papers on phonology and morphology, ed. Andrew Carnie and Heidi Harley, 275-288. MIT Working Papers in Linguistics 21. Cambridge, MA: MIT, MIT Working Papers in Linguistics. Harley, Heidi and Rolf Noyer. 1999. State-of-the-Article: Distributed Morphology. GLOT 4.4. 3-9. 39
Kayne, Richard. 1994. The antisymmetry of syntax. Cambridge, MA: MIT Press. Larson, Richard K. 1988. On Double Object conctructions. Linguistic Inquiry 19: 335-391. Lasnik, Hovard. 2007. Multiple sluicing. Ms. LING 819. Marvin, Tatjana. 2011. High and Low Applicatives in Slovenian and Cross-linguistically. Slovensi jezik-Slovene linguistics studies 8: 95-113. Merchant, Jason. 2001. The Syntax of Silence. Sluicing, Islands, and the Theory of Elipsis. Oxford: Oxford University Press. Nevins, Andrew. 2010. Morphophonological Dissimilation, Morphosyntactic Dissimilation and the Architecture of Exponence. For The Morphology and Phonology of Exponence, ed. Jochen Trommer. Draft of December 24, 2010. Pereltsvaig, Asya. 2007. On the Universality of DP: A View from Russian. Studia Linguistica 61(1): 59-94. Progovac, Ljiljana. 1998. Determiner phrase in a language without determiners. Journal of Linguistics 34: 165–179. Pylkkänen, Liina. 2000. What applicative heads apply to. In Proceedings of the 24th Annual Penn Linguistics Colloquium, 1–13, ed. by Michelle Minnick, Alexander Williams and Elsi Kaiser. Richards, Norvin. 2009. Locality and Linearization. The Case of Kinande. Geko Kenkyu. 136: 75-92. Richards, Norvin. 2010. Uttering Trees. Cambridge, MA: MIT Press. Rudin, Catherine. 1988. On multiple questions and multiple fronting. Natural Language and Linguistic Theory 6:445-501. Stepanov, Arthur. 1998. On wh-fronting in Russian. In NELS 28, ed. by Pius N. Tamanji and Kiyomi Kusumoto, 453–467. Amherst: University of Massachusetts, GLSA. Stjepanović, Sandra. 2003. Multiple wh-fronting in Serbo-Croatian matrix questions and the matrix sluicing construction. In Multiple Wh-fronting, ed. Cedric Boeckx and Kleanthes K. Grohmann, 255-284. Amsterdam: John Benjamins. Toporišič, Jože. 2004. Slovenska slovnica. Maribor: Obzorja.
40
