Inverse attraction in Ingrian Finnish relative clauses.
Ingrian Finnish relative clauses allow a syntactic effect known as inverse attraction (or else attractio inversa, hereafter IA). (2) Under IA, the head of the relative clause acquires the syntactic marking which corresponds to the relativized position, i.e. it gets marked for case as if it belonged to the subordinate clause. For example, in (1), the word lammas 'sheep' is marked for genitive case, which is the case it would normally receive in the sentence 'I bought a sheep'. As the word 'sheep' occupies the subject position in the main clause, it would otherwise be expected to get nominative marking.
(1) NOM [right arrow] GEN (3)
lampa-n minka mia eilen ost-i-n loiko sheep-GEN what.GEN I.NOM yesterday buy-PST-1SG lie.PRS.3SG koi-n luon home-GEN near
'The sheep I bought yesterday is lying in front of the house'
In (1), the head noun and the relative pronoun are marked for the same case. It could therefore be supposed that the head noun agrees in case with the relative pronoun. This hypothesis is sometimes used as a part of the definition of IA. However, an alternative analysis is possible. The head and the relative pronoun could get their case independently from the same predicate. These two major analyses are considered in more detail bellow.
IA has been attested in a number of dead languages. The best known in this sense are Latin (example 2) (Touratier 1980 : 147-211) and Ancient Greek (Grimm 2005 : 78-92). IA is also reported in Hittite, Old Persian, Oscan and Umbrian (Hahn 1964), Vedic and Sanskrit (Gonda 1975 : 195), Middle High German (Pittner 1995) and Modern Church Slavonic ([TEXT NOT REPRODUCIBLE IN ASCII] 1619 : 238).
(2) Latin, NOM [right arrow] ACC
Urb-em qu-am statu-o vestr-a est city-ACC.SG which-ACC.F.SG found-PRS.ACT.1SG your-NOM.F.SG be.PRS.3SG
'The city which I found is yours' (Vergilius, Aeneis I 573)
To the best of my knowledge, almost nothing has been written on the "classical" variety of IA in living languages. (4) Apart from Ingrian Finnish, IA occurs in at least one more living language variety, namely, substandard colloquial Russian: (5)
(3) colloquial Russian, NOM [right arrow] DAT
TeM KOMy Ha[??]o y[??]e 3Ha[??]T those.DAT.PL who.DAT.SG is_necessary already know.PRS.3PL
'Those who need to know it already do' (http://www.sdelanounas.ru)
A less incontrovertible type of IA is attested in at least two living language varieties, Modern Persian (Lehmann 1984 : 185; Aghaei 2006 : 72-76, 90-95) (6) and East Franconian German (Fleischer 2006 : 229). In these varieties, the relative clause does not contain a relative pronoun. Therefore the "attracted" case on the head noun cannot be regarded as borrowed from the relative pronoun and can only be handled in terms of the relativized position. Cf. (4) for a similar construction in Old English (Harbert 1983):
(4) Old English, NOM [right arrow] ACC
pone stan pe [??]a wyrhtan awurpon pes is geworden the.ACC stone COMPL the workers rejected that.NOM is become on paere hyrnan heafod on the corner head
'The stone that the workers cast off, that has become the head of the corner' (Luke 20:17, as cited in Harbert 1983 : 552, the glosses are retained)
Examples like (4) could become and do become a ground to claim that IA does not normally involve any agreement or "assimilation" between the head and the relative pronoun (Lehmann 1984 : 185). Alternatively, such examples could be regarded as containing a different construction. See 6.1 for further discussion.
To sum up, IA has only been attested in Indo-European languages, predominantly if not exclusively dead ones. Ingrian Finnish is neither, which makes it ideal to check the existing hypotheses on the structure of relative clauses under IA.
The paper is structured as follows. Section 2 contains the basic data on relativization in Ingrian Finnish. These data make it possible to describe the properties of relative clauses subject to IA as opposed to other types of relative clauses: "regular" externally headed (section 3) and internally headed (section 4) relative clauses. Section 5 is a description of other properties of IA, primarily those concerning the syntactic position of the head noun and the relative pronoun. Section 6 includes a summary of the properties (6.1), an overview of possible interpretations (6.2) and some further issues (6.3).
2. Ingrian Finnish relative clauses: basic information
Ingrian Finnish relative clauses are predominantly finite. Participial relativization is possible but comparatively infrequent, and, as opposed to Standard Finnish, Ingrian Finnish ma-forms (the so-called agent participle) cannot be used adnominally. (7)
Finite relative clauses in Ingrian Finnish normally contain one of the two nominal relative pronouns, kuka who or mika 'what'. This is in line with the description of Southeastern Finnish dialects by Paakkonen (1988 : 29-30). The Standard Finnish relative pronoun joka 'which' is almost never used in Ingrian Finnish. The words kuka and mika are not restricted in their use to relative clauses and are also used as interrogative pronouns. In both cases, the pronouns can be used either independently (mika what, which) or attributively, as in mika pere what family, whichever family .
The basic distinction between kuka and mika is that between people and other referents, though kuka is occasionally used with reference to animals. This distinction holds only for the substantive use of the pronouns, while in attributive function mika what is used more widely. In relative clauses, as opposed to questions, the system is further complicated by a set of conditions under which mika can refer to people and (most rarely) kuka can refer to inanimates.
"Agreement mismatches" in number are also attested under certain conditions. Some restrictions on these "mismatches" are discussed below (4.2, 4.3, 4.5).
The main positional types of relative clauses in Ingrian Finnish are post-nominal clauses (immediately following their heads), extraposed clauses (following the main clause), correlatives (preceding the main clause) and free (i.e. headless) relatives.
In Mullonen 2004, which is a collection of Ingrian Finnish texts, the frequency distribution is as follows: (8)
In sections 3 and 4, IA-constructions are compared respectively with externally headed (postnominal and extraposed) and internally headed relatives (in Ingrian Finnish represented only by correlatives). (9)
3. IA vs. "regular" externally headed relative clauses
The present study did not reveal any contexts in which IA is obligatory. Thus, in all relevant examples the "attracted" case can be replaced with a more expected one, cf. (1) and (5).
(5) lammas minka mia eilen ost-i-n loiko koi-n luon sheep.NOM what.GEN I.NOM yesterday buy-PST-1SG lie.3SG home-GEN near
'The sheep I bought yesterday is lying in front of the house'
The converse does not hold. IA is ungrammatical in a wide range of contexts. Thus, all other factors held equal, out of 174 externally headed relative clauses in Mullonen 2004, only 7 could contain IA. In all other cases, either one of the conditions described below (3.1-3.3) is not met or the expected case marking coincides with that of the relativized position. In the latter case IA is impossible (or at least unobservable), because the replacing marking would otherwise coincide with the replaced marking.
Unfortunately, none of these 7 clauses actually contained IA in the examined publication.
3.1. The head of the IA-construction cannot occur postverbally:
(6) talo-n luon loiko lammas / *lampa-n minka mia house-GEN near lie.PRS.3SG sheep.NOM sheep-GEN what.GEN I.NOM eilen ost-i-n yesterday buy-PST-1SG
'In front of the house, there is a sheep I bought yesterday'
A similar restriction is observed in most languages with IA, see, among others, Hahn 1964 and Harbert 1983. (10)
3.2. The head of the IA-construction cannot be a complement of anything except the main predicate. The relevant structures include pre- and postpositional phrases (7), noun phrases (8) and infinitive phrases.
(7) ilman ihmis-ta / *ihminen mika man pois without man-PART man.NOM what.NOM go.PST.3SG away miu-l on ikava I-AD be.PRS.3SG boring.NOM
'I miss the man who left'
(8) OK ihmise-n / *ihminen nime-n mika minu-le soitt-i man-GEN man.NOM name-GEN what.NOM I-ALL call-PST.3SG mid unohut-i-n I.NOM forget-PST-1SG
'I forgot the name of the man who called me'
Cf. Aghaei 2006 : 95, where Modern Persian IA is shown to be impossible when the head is a complement of a preposition.
3.3. IA is not compatible with extra position of the relative clause, which is possible and quite frequent in other cases (see section 2). In (9), the relative clause is extraposed, therefore the head cannot undergo IA.
(9) OK lammas / *lamp a -n loiko koi-n luon sheep.NOM sheep-GEN lie.PRS.3SG home-GEN near minka mia eilen ost-i-n what.GEN I.NOM yesterday buy-PST-1SG
'The sheep I bought yesterday is lying in front of the house'
A similar restriction is noted in Aghaei 2006 : 76 for Modern Persian.
The three restrictions listed above could be summarized in terms of information structure. Thus, according to Harbert (1983), IA implies that the relative construction (i.e. both the head and the relative clause) should be in the topic position, where no case is assigned.
However, this hypothesis proves wrong, at least for the Ingrian Finnish data. It is neither necessary nor sufficient for the head to be in the topic rather than in the focus. When the postverbal head is topical, the IA is still impossible (10). On the other hand, the focalized preverbal head in (11) can receive the attracted case. (11)
(10) kons tullo se ihmine / *si-ta ihmis-ta when come.PRS.3SG that.NOM man.NOM that-PART man-PART ke-ta mia kutsu-i-n who-PART I.NOM call-PST-1SG
'When will the man I invited come?'
(11) NOM [right arrow] PART OK vast ta-ta ihmis-ta ke-ta kavva ei suote-ttu only this-PART man-PART who-PART long NEG feed-PC_IPS.PST voi-p suuva ta-ta suotava can-PRS.3SG eat.INF this-PART food.PART
'Only a man who hasn t been fed for a long time can eat this food'
Even more crucially, in Ingrian Finnish, the IA-construction can be coordinated with an NP. This NP is marked for the case required in the main clause. It can either precede (12) or follow (13) the relative construction. These data exclude not only the analysis based on the topic position, but any analysis according to which the IA-constructions are always left-dislocated.
(12) PART [right arrow] AD
OKmiu-n siso-ja-in ja naise-l ke-l mia ela-n I-GEN sister-PART-P1SG and woman-AD who-AD I live-1SG mid tutvust-i-n I.NOM introduce-PST-1SG
'I introduced my sister and the woman at whose house I m living to each other'
(13) NOM [right arrow] PART ihmis-ta ke-ta sia na-i-t ja man-PART who-PART you.NOM see-PST-2SG and miu-n veiko-in o-vat ustav-i-a
I-GEN brother.NOM-P1SG be-PRS.3PL friend-PL-PART
'The man you saw and my brother are friends'
Touratier observes the coordination of IA-constructions with NPs in Latin. He argues that such coordination could only be possible because it is of a special semantic kind. According to Touratier, the conjunction in these examples means as well as', and the following phrase represents a separate proposition with an elided predicate. It is thus not properly embedded in the main clause (Touratier 1980 : 182, 203). However, it could not be the case in (12) and (13) for two reasons. First, the verb in (13) is in the plural, while both conjuncts are singular. Therefore, elliptical analysis is at least problematic. Second, the main clause in (12) and (13) describes a symmetrical relation in which at least two participants are required. Therefore, the conjuncts necessarily belong to one proposition (cf. the English translation in (13) and its counterpart under elliptical analysis #The man you saw is a friend and my brother is a friend).
As follows from the above, the IA-constructions need not be in the topic position. Probably the more accurate way to describe the restrictions 3.1-3.3 is in terms of linear structure. Thus, the forbidden structures are those in which the relative clause is preceded by the element which assigns case to the head of the relative clause. (12)
4. IA vs. correlatives
Another structure which is close to IA-construction is a correlative clause with an internal head (14). As in the case of IA, the marking of the head and the relative pronoun correspond to the same syntactic position.
(14) OK minka lampa-n mia ost-i-n (se) loiko what.GEN sheep-GEN I.NOM buy-PST-1SG that.NOM lie.PRS.3SG talo-n luon house-GEN near
'The sheep I bought is lying in front of the house'
In most works on IA-constructions, the latter are more or less explicitly compared to correlatives, see, among others, Lancelot 1696 : 279-280 and Bhatt 2005b : 2. The two types of clauses are sometimes even claimed to be equivalent, apart from the word order, cf. for instance Hahn 1964. However, the difference between IA-constructions and correlatives in Ingrian Finnish goes far beyond the word order.
4.1. IA is possible in appositive relative clauses, i.e. the relative clauses which do not restrict the reference of the head noun (15a). Ingrian Finnish correlatives cannot be appositive (15b). (13)
(15a) NOM [right arrow] PART miu-n isa-ja-in ke-ta luo-ti loiko polnitsa-s I-GEN father-PART-P1SG who-PART beat-IPS.PST lie.PRS.3SG hospital-IN
'My father, who has been beaten, is in the hospital'
(15b) *ke-ta miu-n isa-ja-in luo-ti loiko polnitsa-s who-PART I-GEN father-PART-P1SG beat-IPS.PST lie.PRS.3SG hospital-IN
'My father, who has been beaten, is in the hospital'
4.2. In Ingrian Finnish relative clauses, "agreement mismatches" (14) in number are possible and widely attested. (15) In many cases, a head in the plural can be followed by a relative pronoun in the singular. This is true for regular postnominal clauses as well as for clauses with IA (16a), while in correlative clauses such mismatches are ruled out (16b).
(16a) ALL [right arrow] NOM OK oppila-t kuka muohdstu-i-O/muohdstu-i-t pupil-PL.NOM who.NOM be_late-PST-3SG be_late-PST-3PL anne-tti kaks give-IPS.PST two.NOM
'The pupils who were late were graded "poor"'
(16b) OK kutka / *kuka oppila-t muohdstu-i-t who.PL.NOM who.NOM pupil-PL.NOM be_late-PST-3PL anne-tti kaks give-IPS.PST two.NOM
'The pupils who were late were graded "poor"'
4.3. With some collective nouns, such as per(h)e 'family', kamanta 'team', ruhma 'group', "semantic agreement" in number and animacy is allowed. Thus, in regular postnominal constructions and in IA-constructions, the word pere family can be combined with the animate relative pronoun kuka who in the plural (17a).
(17a) PART/GEN [right arrow] nom OK pere kutka ella-t nuapuri-n talo-s family.NOM who.PL.NOM live.PRS-3PL neighbor-GEN house-IN mid kutsu-i-n louna-i-l' I.NOM call-PST-1SG dinner-PL-ALL
'I asked the family who live next door to dinner'
However, in the corresponding correlatives, only the inanimate relative pronoun mika 'what' can be used:
(17b) *kutka pere ella-t/ella who.PL.NOM family.NOM live.PRS-3PL live.PRS.3SG nuapuri-n talo-s mia kutsu-i-n louna-i-l' neighbour-GEN house-IN I.NOM call-PST-1SG dinner-PL-ALL
'I asked the family who live next door to dinner'
(17c) OK mika pere ella nuapuri-n talo-s what.NOM family.NOM live.PRS.3SG neighbour-GEN house-IN mid kutsu-i-n louna-i-l' I.NOM call-PST-1SG dinner-PL-ALL
'I asked the family who live next door to dinner'
4.4. IA is compatible with "agreement mismatches" in case, i.e. the head noun and the relative pronoun can be marked for different cases which correspond to the same syntactic position.
Thus, for instance, both partitive and genitive are used to mark the direct object, and examples like (18a) are acceptable for most speakers (see also (29) below).
(18a) NOM [right arrow] GEN
?ihmise-n ke-ta na-i-t eilen ella naapuri-s man-GEN who-PART see-PST-2SG yesterday live.PRS.3SG neighbor-IN
'The man you saw yesterday lives next door'
In correlative clauses such mismatches are absolutely unacceptable:
(18b) ke-ta OK ihmis-ta / *ihmise-n mia nai-n who-PART man-PART man-GEN I.NOM see-1SG miu-st nduttija I-EL please.PRS.3SG
'I like the man I see'
Such attraction is not restricted to case forms. Ingrian Finnish has some locative adverbs which can only be used to denote direction and others which denote location sensu sticto. Thus, as evidenced by (19)--(20), ulkon outdoors can only mark the position in space, while ulos is used to mark the direction of movement.
(19) mia ruao-i-n ulkon I.NOM work-PST-1SG outdoors[location]
'I worked outdoors'
(20) mia e-n lahe ulos / *ulkon I.NOM neg-1sg go outdoors[direction] outdoors[location]
'I am not going outdoors'
The expected head in (21) would be the adverb of direction ulos. However, it is replaced with the adverb of location ulkon. As in other examples of IA, the formal properties of the head are determined by the subordinate clause. However, no case marking is present, therefore the standard explanation that the head borrowed the case of the relative pronoun is inapplicable to (21).
(21) (adverb of direction) [right arrow] (adverb of location)
ulkon mi-s mia koko uamu-n ruao-i-n outdoors[location] what-IN I.NOM all morning-GEN work-PST-1SG mia enna e-n lahe I.NOM anymore NEG-1SG go
'I am not going outdoors anymore, where I have been working all morning'
4.5. When the head of a relative clause refers to the owner of a place and this place is the relativized participant, a mismatch in animacy and case can occur (22a). In correlatives, such mismatches are ungrammatical (22b).
(22a) AD [right arrow] ABL
si-l't ihmise-l't mi-st mia tul-i-n that-ABL man-ABL what-EL I.NOM come-PST-1SG on sorja talo be.PRS.3SG beautiful.NOM house.NOM
'The man I ve come from (lit. 'where I ve come from') has a beautiful house'
(22b) ke-l't / *mi-st ihmise-lt mia tul-i-n who-ABL what-EL man-ABL I.NOM come-PST-1SG on sorja talo be.PRS.3SG beautiful.NOM house.NOM
'The man I've come from has a beautiful house'
4.6. IA-constructions (23a), but not correlatives (23b) are compatible with quantifiers in the head. (16) The quantifier can be the head (24) or a part of it (23a).
(23a) nom < all OK kaik-i-l ihmis-i-l' ke-l' mia kiruta-n all-PL-all man-PL-ALL who-ALL I.NOM write.PRS-1SG kirjo-i-O ella-t Suome-s letter-PL-PART live.PRS-3PL Finland-IN
'All the people to whom I write letters live in Finland'
(23b) *ke-l' kaik-i-l' ihmis-i-l' mia kiruta-n who-ALL all-PL-ALL man-PL-ALL I.NOM write.PRS-1SG ella-t Suome-s live.PRS-3PL Finland-IN
'All the people to whom I write letters live in Finland'
(24) NOM [right arrow] EL OK kaiki-st mi-st muo luaje-ti ol mukava all-EL what-EL we.NOM talk-IPS.PST be.PST.3SG interesting.NOM
'Everything we discussed was interesting'
4.7. IA-constructions (25a), but not correlatives (25b), are compatible with demonstratives (se 'that', tama this') in the head. Cf. a similar observation for Latin in Touratier (1980 : 153). The demonstrative can be the head or a part of it.
(25a) NOM [right arrow] ALL se-l' ihmise-l' ke-l' mia kiruta-n that-all man-ALL who-ALL I.NOM write.PRS-1SG ella Suome-s live.PRS.3SG Finland-IN
'The man to whom I write lives in Finland'
(25b) *ke-l se-l' ihmise-l' mia kiruta-n who-ALL that-ALL man-ALL I.NOM write.PRS-1SG ella Suome-s live.PRS.3SG Finland-IN
'The man to whom I write lives in Finland'
5. Other properties of IA
This section lists the properties of IA-constructions that do not help to distinguish the latter from the other types of relative clauses, but could still be revealing in the discussion of this phenomenon.
The relative clause that provokes IA can be c o o r d i n a t e d with another relative clause. In (26), the case marking on the head is determined by the first relative clause and does not meet the case requirements of the main clause and the second relative clause. This structure suggests that IA can operate in non-constituents, because in most analyses, the head and the first relative clause do not form a constituent.
(26) NOM [right arrow] PART
si-ta ihmis-ta ke-ta mia na-i-n ja kene-n kans that-PART man-PART who-PART I.NOM see-PST-1SG and who-GEN with
mia lua-i-n tullo mei-l viera-i-hen I.NOM talk-PST-1SG come.3SG we-ALL guest-PL-ILL
'The man whom I saw and talked to is going to pay us a visit'
5.2. Positions within the main and dependent clauses
5.2.1. One of the most discussed questions in the studies on IA is the availability of IA-construction depending on the case requirements in the main and subordinate clauses. As noted in numerous studies, the statistical data on Latin and Ancient Greek seem to indicate that IA is impossible in some configurations. The relevant conditions are the case "before attraction" (the case required in the main clause) and the case "after attraction" (the case required in the subordinate clause).
The restrictions are often summarized by means of the Case-Markedness Hierarchy, which looks as follows:
(27) NOM < ACC < DAT < GEN ... (Bhatt 2005b : 2)
The hierarchy predicts that the more marked cases (those to the right) will not be replaced by the less marked cases (those to the left). Thus the case shifts in (28a) are expected to be grammatical, while the shifts in (28b) are expected to be ungrammatical.
(28a) OK NOM [right arrow] ACC, OK NOM [right arrow] DAT, OK ACC [right arrow] DAT ...
(28b) *ACC [right arrow] NOM, *DAT [right arrow] NOM, *DAT [right arrow] ACC.
Counterexamples have been noted in the literature, see e.g. Bhatt 2005b, where the hierarchy (27) is claimed to be inapplicable to IA. However, the literature on IA en masse ignores these examples or treats them as occasional, see, among others, McCreith 1988, Pittner 1995 and Grimm 2005. The statements on the possibility of different attractions are only based on text data and thus can contain gaps.
In Ingrian Finnish, the cases which correspond most closely to the accusative are 1) the genitive and the partitive (in the singular) or 2) the nominative and the partitive (in the plural) in the object position. The case functionally closest to the dative is the allative. The last position on the hierarchy (genitive-marked possessor) is discussed in 3.2 and 5.2.2.
In view of these adjustments, the hierarchy (27) is not respected in Ingrian Finnish. Most frequently (though not exclusively) it is violated when the subject position is relativized, as in (16a), (17a). The attraction of the kind "DAT [right arrow] ACC" is also possible. In (29), this shift is accompanied by a "case mismatch" (see 4.4). The head receives direct object (partitive) marking from the subordinate clause, however, its case does not correspond to the case of the relative pronoun.
(29) ALL [right arrow] PART
lampa-i-t a mitka mia ost-i-n mia anno-i-n sheep-PL-PART what.PL.NOM I.NOM buy-PST-1SG I.NOM give-PST-1SG
'I gave some hay to the sheep (pl) I bought'
A possible way to save the hierarchy in (27) would be to claim that the nominative case in examples (16a) and (17a) appears independently of IA. It could be analysed as the case of the topic, the so-called nominativus pendens. Under this analysis, example (29) would involve a shift from the nominative of the topic to the partitive.
In fact, some speakers do use structures where the nominative of the head is not determined by either the main clause or the subordinate clause (30). However, these structures are less acceptable in general and much less frequent (see Table 2). (17)
(30) ??lammas minka mia ost-i-n anno-i-n hein-i sheep.NOM what.GEN I.NOM buy-PST-1SG give-PST-1SG grass-PL.PART
'I gave hay to the sheep (sg) I bought'
Moreover, the nominative on the head of a relative construction can be used when this construction is the second conjunct in a coordinated structure (31). The resultant structure includes two topicalized conjuncts, and only the first one is marked for the case internal to the clause (adessive). Such structures with nominativus pendens are impossible elsewhere, while in sentences with IA-constructions a similar phenomenon is observed, see (10)-(11) above. Therefore, the case shift in (31) seems to exclude the explanation based on nominativus pendens.
(31) AD [right arrow] NOM
OK miu-n siso-lla-in ja se naine kuka ella I-GEN sister-AD-P1SG and that.NOM woman.NOM who.NOM live.PRS.3SG
nuapurtalo-s suntumapaiva on uhtenpain neighbour_house-IN birthday.NOM be.PRS.3SG on_the_same_day
'My sister and the woman who lives next door have birthday on the same day'
To sum up, the hierarchy (27) does not hold in Ingrian Finnish. Most of the exceptions could be explained by reference to nominativus pendens; however, given the statistical data, this analysis does not seem plausible.
5.2.2. Another parameter which proves important is the degree of embeddedness of the relative pronoun.
In Ingrian Finnish, as in Standard Finnish, the genitive phrase can be either more embedded (as a part of an NP or a PP) or less embedded (as a direct object). As examples (32)-(33) show, IA-constructions with the more embedded genitive phrase are evaluated as slightly less felicitous. Table 3 shows that such cases of IA are also rarer in translations from Russian ([chi square], P < 0.01).
(32) NOM [right arrow] GEN lapse-n minkan jat-i-n kotti ka-i-O itke-ma child-GEN what.GEN leave-PST-1SG home.ILL go-PST-3SG cry-SUP.ILL
'The child I left home started crying'
(33) NOM [right arrow] GEN
laps/?lapse-n kene-n kans mia mdn-i-n viera-i-hen child.NOM child-GEN who-GEN with I.NOM go-PST-1SG guest-PL-ILL
ol huvilla be.PST.3SG content.P3
'The child whom I took on a visit was happy'
6.1. The main properties of IA-constructions
The main properties of IA-constructions are summarized in Table 4. In the table, IA-constructions are contrasted with related structures with the same pronouns, kuka 'who' and mika 'what'.
The columns "correlatives" and "regular externally headed relative clauses" correspond to the two major relativization strategies in Ingrian Finnish. The column "relative clauses with IA" contains the data on the construction which is the topic of this paper. The relevant data are discussed in detail in sections 3-4.
One more column is added to include the data on questions in which mika what or kuka who are interrogative adjectives. Thus, for instance, the interrogative kutka (who.PL.NOM) cannot be used with a collective noun (34). These data were not discussed in the previous sections.
(34) OK mika / *kutka pere ella ta-s talo-s? what.NOM who.PL.NOM family.NOM live.PRS.3SG this-IN house-IN
'What family lives in this house?'
The table includes properties of two kinds. Properties 1 and 4 are claims that are either true (+) or false (-) for a certain construction. All the other properties correspond to structures which are possible in some constructions (OK) and impossible in others (*). The shaded cells in the table correspond to the properties which are inapplicable to a given group or impossible by the definition of the group.
For a discussion of properties 2-3, see section 3.
Properties 4-7 show that in all that concerns agreement correlatives pattern with questions, whereas IA-constructions pattern with regular externally headed relative clauses. The position of relative/interrogative pronouns in internally-headed correlatives and questions like (34) is normally analyzed as adnominal. For both IA-constructions and externally headed relative clauses, the analysis is not as obvious. Thus, in Bianchi 1999 and 2000, their derivation includes a stage where the pronoun is used adnominally. However, properties 4-7 show that in these cases, the agreement is much less strict. This difference in agreement properties (between correlatives and questions on the one hand and IA-constructions and externally headed relative constructions on the other hand) is easiest to describe as that between the adnominal and pronominal use of relative pronouns. In the first case, the relative pronoun forms a DP (determiner phrase) with a noun and is subject to agreement. In the second case, the relative pronoun has the distribution of a DP on its own and the relation with the noun is that of congruence. (18) Therefore, the relative pronouns in IA-constructions are used pronominally, not adnominally.
The data on "agreement mismatches", especially case mismatches (6-7), are important for identifying the limits of what is to be considered an IA-construction. These data suggest that the head noun normally acquires its "attracted" case not by agreement or assimilation with the relative pronoun, otherwise the two cases would coincide. The head gets its marking as if it were in the subordinate clause and occupied the relativized position. Therefore, the "classical" IA in languages with relative pronouns is probably fully identical with a similar effect in languages without relative pronouns, such as Modern Persian, East Franconian German and Old English (see section 1). In both cases, the head acquires its marking in the subordinate clause. The relative pronouns independently receive case marking in the same position if they are present.
Properties 8-10 are indicative of lesser embeddedness of the head noun into the relative clause in IA-constructions, as opposed to correlatives. Thus, quantifiers and demonstratives do not receive interpretation in the subordinate clause and semantically belong only to the main clause. For example, the relative construction Everything we discussed in (24) does not imply that we discussed everything. Appositive relative clauses are often claimed to be more or less syntactically independent of the head noun. They are sometimes analyzed as attached at a higher level of syntactic structure or coordinated with the head noun, see the overview and analysis in De Vries 2002 : 203-231. The data on agreement mismatches (4-7) seem to point in the same direction. The syntactic position of the head in correlatives is a normal position for a noun in the argument structure of a predicate. As proposed above, the relative pronoun in IA-constructions has the distribution of a DP. Therefore, it occupies the relativized position in the subordinate clause. The hypothetical position of the head in this clause cannot be licensed by the argument structure and is at least supplementary. To sum up, the head noun is more external to the subordinate clause in IA-constructions, than in correlatives.
The data discussed in 3-5 are sufficient to discuss a number of hypotheses on the why and how of IA which have been proposed in the literature so far. The hypotheses discussed below are partly compatible with each other, i.e. they do not always exclude each other.
Hypothesis 1: IA is a result of the fact the speaker has not finished planning by the time he starts speaking and does not know what case to use for the head noun in the initial position.
This hypothesis seems incompatible with the coordinated structures in which the "attracted" head is preceded by a noun marked for its position in the main clause (12), (31). The required case is used on the first conjunct and thus is already clear to the speaker by the time the head of the relative clause is pronounced.
Hypothesis 2: The head noun receives its attracted case because it is moved to a position where no case can be assigned and needs one for the derivation to converge (Harbert 1983).
This hypothesis does not seem to handle the facts of coordination (12), (13), (31). If case can be assigned to one of the conjuncts, it is not clear why this assignment could be blocked for the other conjunct. (19)
Hypothesis 3: IA-construction is always syntactically isolated (Touratier 1980 : 203, 210; Lehmann 1984 : 185).
This hypothesis is also incompatible with coordination (12), (13), (31). Example (11), where the head is in the focus, also seems problematic for this approach.
Hypothesis 4: IA-constructions are differently linearized correlatives (Hahn 1964; Bianchi 1999; Bhatt 2005 : 2).
See section 4 for a detailed discussion of the differences between IA-constructions and correlatives.
Hypothesis 4': In particular, Bianchi in 1999 : 92; 2000 : 66-67 proposes the following structures for different kinds of relative clauses:
(35a) Latin, correlative clause [CP [DP qui [[NP vir]].sub.i] [CP ... [t.sub.i]]] who man
(35b) Latin, IA-construction [CP [DP [NP VIR] [[DP qui [t.sub.NP]]].sub.i] [CP ... [t.sub.i]]]
(35c) Latin, externally-headed relative construction
[DP [D.sup.0] [CP [DP [NP vir] [[DP qui [t.sub.NP]]].sub.i] [CP ... [t.sub.i]]]]
As mentioned above, this analysis has problems with contrasting IAconstructions and correlatives. However, even if this issue is ignored, the properties discussed in section 4 are still problematic to match the structure in (35b).
First, as shown above (4.1), Ingrian Finnish IA-constructions are found with appositive relative clauses. According to Bianchi's analysis, the IP (inflectional phrase) of appositive relative clauses is moved to the specifier of the external determiner at Logical Form (LF). However, as can be seen in (35b), the IA-construction does not have an external determiner in this approach, cf. (35c).
Second, the heads with demonstratives (4.7) and quantifiers (4.6) are problematic. Bianchi suggests in (1999 : 93) that the demonstrative in such constructions originates inside the relative clause. However, in (2000 : 68) she argues against this suggestion. (20) This analysis seems even less plausible for the quantifiers, which are most obviously never interpreted within the relative clause.
The case mismatches (4.4-4.5) are also problematic for this approach. Hypothesis 5: IA-construction is a clause (Hahn 1964; Lehmann 1984 : 185) and/or is not a DP (Bianchi 1999 : 92).
The greatest problem with this approach (unresolved in the present paper) is that it implies that a syntactic position can be filled by two nominal phrases. As shown above, the head noun and the relative pronoun cannot be reduced in the analysis to a single noun phrase (or a single DP, to be more specific).
However, this approach seems to be the only one to cope with the existing case assignment rules. Therefore, the conclusion that the IA-construction is a clause is inevitable, though a slight amendment is suggested in the following section.
6.3. Related properties of Ingrian Finnish relative clauses
IA is likely a manifestation of low nominalization of relative clauses in terms of Lehmann (1984; 1986). The level of nominalization is the degree to which the clause behaves like a nominal rather than like a clause. Other manifestations of low nominalization in Ingrian Finnish include at least two more phenomena: "adverbial" relative constructions and coordination of the relative clause and the main clause.
An example of an "adverbial" relative construction is given in (36). In (36), the relative construction does not occupy any obvious nominal position in the main clause, as could be expected of a noun phrase. As reflected in the translation, it can, however, be interpreted as an adverbial modifier.
(36) lello / OK lello-n minkan isa ott-i-O toy.NOM toy-GEN what.GEN father.NOM take-PST-3SG
laps ka-i-O itke-ma child.NOM go-PST-3SG cry-SUP.ILL
'When the father took away the toy, the child started crying'
Thus, the relative clause operates not at the level of an NP, but at the clause level. According to Lehmann (1988 : 189-192), the level at which the clause is attached is one of the parameters of embeddedness. (21)
Lehmann observes a similar construction in Latin (1984 : 351; 1988 : 187), which is another language with IA. (22)
(37) Calcem partiario coquendam qui dant, ita lime.ACC.SG partly burn.GERV.ACC.SG who.NOM give.3PL so
'If lime is given for burning by shares, it is given thus' (Lehmann 1988 : 187)
In Ingrian Finnish, such clauses are relatively frequent. In Mullonen 2004, 51 clauses of this kind were attested, which is about 17 % of relative clauses in the publication. (23)
(38) Mi-s talo-s ol' pien talous ja what-IN house-IN be.PST.3SG little.NOM household.NOM and
paljo tytto-loi-O, ni mdnn-ii-t sit piija-ks many girl-PL-PART so go-PST-3PL then servant-TRL
If in a certain house there was a small household and many girls, then [they] got a job as servants (lit. In which house the household was little ..., then ...') (Mullonen 2004 : 68) (24)
Another manifestation of low nominalization is the ability of a relative construction to be coordinated with the main clause by means of an adversative conjunction: (25)
(39) oppila-t kutka kirjutt-i-t paremmast pupil-PL.NOM who.PL.NOM write-PST-PL well.COMP
a merki-t sa-i-vat heiko-mma-t but mark-PL.NOM get-PST-3PL weak-COMP-PL.NOM
'The pupils who wrote [the work] better got, however, worse marks'
Usually the level of nominalization is used to characterize a well-defined group of relative clauses, such as correlative clauses or relative clauses with certain relativization markers. However, the phenomena listed above do not seem absolutely independent, though they apply to structurally diverse relative constructions: a "regular" headed relative clause and an IA-construction in (36) and a correlative in (38). Probably, the level of nominalization could be also used to characterize a whole relativization system at a certain period of time. Thus, it could be claimed that Ingrian Finnish relativization system as a whole shows a tendency for low nominalization.
Further evidence for this claim includes quantitative data. In Ingrian Finnish, adjoined relative clauses (correlatives and extraposed clauses) are very frequent. As mentioned in section 2, these two groups constitute as much as 30 % of relative clauses in Mullonen 2004. According to Lehmann (1988 : 672), adjoined relative clauses are normally the least nominalized. The relative frequency of strategies cannot be attributed to any type of relative clauses, but seems an important characteristic of the relativization system.
Therefore, IA probably manifests low nominalization in Ingrian Finnish relativization system. This suggestion is closest to Hypothesis 5 above, according to which IA-construction is a clause. However, it differs from the latter in at least two crucial aspects.
1) The proposed approach presupposes some gradualness. The binary distinction between a clause vs. a non-clause does not allow to explain why IA-constructions could differ from correlatives.
2) This approach allows to reflect the connection between IA and the phenomena listed above in this section. These phenomena could be described as co-existing in the system of relativization at a certain stage and reflecting similar tendencies. However, an absolutely unified syntactic analysis for the corresponding structures is not necessary, because a language can include structures nominalized to varying degrees.
1/2/3--1st/2nd/3rd person; ABL--ablative; ACC--accusative; ACT--active; AD--adessive; ALL--allative; COMP--comparative; COMPL--complementizer; DAT--dative; EL--elative; F--feminine; GEN--genitive; GERV--gerundive; ILL--illative; IN--inessive; INF--infinitive; IPS--impersonal; NEG--negative; NOM--nominative; P--possessive; PART--partitive; PASS--passive; PC--participle; PL--plural; PRS--present; PST--past; SG--singular; SUP--supine; TRL--translative.
I am indebted to Natalia Kuznetsova and Mehmet Muslimov for their help at various stages of this study. I also wish to thank Oleg Belyaev and the anonymous reviewer for their thorough comments on this paper.
The study was supported by the Russian Foundation for Humanities (project 11-04-00172a).
Aghaei, B. 2006, The Syntax of Ke-Clause and Clausal Extraposition in Modern Persian, Austin (PhD Thesis. University of Texas at Austin).
Bader, M., Bayer, J. 2006, The Mental Representation of Case.--Case and Linking in Language Comprehension. Evidence from German, Dordrecht (Studies in Theoretical Psycholinguistics 34), 115-137.
Bhatt, R. 2005a, Correlative Clauses. Paper presented at LOT Summer School, Leiden.
--2005b, In search of Case Attraction. Paper presented at LOT Summer School, Leiden.
Bianchi, V. 1999, Consequences of Antisymmetry. Headed Relative Clauses, Berlin--New York.
--2000, Some Issues in the Syntax of Relative Determiners.--The Syntax of Relative Clauses, Philadelphia, 53-81.
Corbett, G. G. 2006, Agreement, Cambridge.
Fleischer, F. 2006, Dative and Indirect Object in German Dialects. Evidence from Relative Clauses.--Datives and Other Cases. Between Argument Structure and Event Structure, Amsterdam--Philadelphia, 213-238.
Fowler, H. W. 1994, A Dictionary of Modern English Usage, Hertfordshire.
Gonda, J. 1975, Selected Studies 1. Indo-European Linguistics, Leiden.
Grimm, S. M. 2005, Lattice of Case and Agentivity, Amsterdam (MSc Thesis, Universiteit van Amsterdam).
Hahn, E. A. 1964, Relative and Antecedent.--Transactions and Proceedings of the American Philological Association 95, 111-141.
Hakulinen, A., Vilkuna, M., Korhonen, R., Koivisto, V.,
Heinonen, T., Alho, I. 2004, Iso suomen kielioppi, Helsinki. (http://scripta.kotus.fi/visk/etusivu.php).
Harbert, W. 1983, A Note on Old English Free Relatives.--Linguistic Inquiry 14, 549-553.
Izvorski, R. 1997, Subject Free Relatives in Null-Subject Languages. Evidence from Slavic.--Formal Approaches to Slavic Linguistics. The Cornell Meeting, 1995, Ann Arbor, 267-288.
Lancelot, C. 1696, Nouvelle methode de messieur de Port Royal pour appren dre facilement la langue latine, Paris.
Lehmann, Ch. 1984, Der Relativsatz. Typologie seiner Strukturen. Theorie seiner Funktionen. Kompendium seiner Grammatik, Tubingen (Language Universals, Series 3).
--1986, On the Typology of Relative Clauses.--Linguistics 24, 663-680.
--1988, Towards a Typology of Clause Linkage.--Clause Combining in Grammar and Discourse, Amsterdam--Philadelphia, 181-226.
Meillet, A., Vendryes, J. 1948, Traite de grammaire comparee des langues classiques, Paris.
Mullonen, M. 2004, Elettiinpa ennen Inkeris. Naytteita inkerinsuomalaisista murteista, Petroskoi.
Parker, F., Riley, K., Meyer, Ch. 1988, Case Assignment and the Ordering of Constituents in Coordinate Constructions.--American Speech 63, 3.
Paakkonen, I. 1988, Relatiivisanan valinta, Helsinki.
Pittner, K. 1995, The Case of German Relatives.--Linguistic Review 12, 197-231.
Pogorelec, B. 1955/1956, Sintakticne napake v solskih nalogah.--Jezik in slovstvo 1, 202-210.
Touratier, Ch. 1980, La relative. Essai de theorie syntaxique (a partir de faits latins, francais, allemands, anglais, grecs, hebreux, etc.), Paris.
Vries, M. de 2002. The Syntax of Relativization, Utrecht (PhD Thesis).
--2003, [Rev.] V. Bianchi, Consequences of Antisymmetry Headed Relative Clauses.--Lingua 113, 173-181.
[TEXT NOT REPRODUCIBLE IN ASCII].
[TEXT NOT REPRODUCIBLE IN ASCII].
[TEXT NOT REPRODUCIBLE IN ASCII].
MARIA KHOLODILOVA (Saint-Petersburg)
Institute for Linguistic Studies, Russian Academy of Sciences
(1) The Ingrian Finnish data were collected in 2011-2012 in Central Ingria (Gatchina District). The data were obtained from 19 speakers of Ingrian Finnish.
All the examples (unless stated otherwise) were elicited as translations from Russian or offered to the speakers for a grammaticality judgment task. In the latter case, examples are prefixed by grammaticality marks (OK, ?, ?? or *), which represent the average acceptability rates (from fully acceptable to fully unacceptable).
(2) Inverse attraction owns its name to another effect, called relative attraction. Under relative attraction, the relative pronoun gets marked for the same case as the head. IA works the other way round, hence the word inverse.
(3) IA is well attested in my data. In translations from Russian, IA-constructions are used in 107 examples out of 290, where they could occur (see section 3 for the constraints on IA). The constructions were used in translations by 17 speakers out of 19. The two speakers who did not use the construction in the collected data rated it as acceptable.
(4) Cf.: "As far as I can see, this construction [attractio relativi.--M. Kh.] is only found [in] dead languages, as is the inverse phenomenon: attractio inversa" (De Vries 2003 : 178).
Some data are available on the errors which follow the lines of IA in German (Bader, Bayer 2006), English (Fowler 1994 : 68) and Slovene (Pogorelec 1955/1956 : 208). Probably, it means that this phenomenon is present in substandard varieties of these languages, though the data are not sufficient to exclude the "slip of the tongue" explanation.
(5) The Russian construction seems to be basically restricted to false free relatives, i.e. relatives without an N in the head.
(6) The Modern Persian IA is also different from the usual patterns in that the head in IA-constructions is always marked with the accusative, though the relativized position in the subordinate clauses may require prepositional marking (Aghaei 2006 : 74).
(7) A description of their adnominal use in Standard Finnish is provided by Hakulinen, Vilkuna, Korhonen, Koivisto, Heinonen, Alho (2004 : [section] 525).
(8) The estimate is based on the full sample of relative clauses in Mullonen 2004, which includes about 251 examples. The "adverbial" relative clauses (see 6.3) are left out of the calculations.
(9) Another tempting parallel, which is not considered bellow, is that between positions available for IA and non-matching in free relatives. For further information on matching see e.g. Izvorski 1997 and Vries 2002 : 227-231.
(10) This constraint does not hold in English error data (see footnote 4) (Fowler 1994 : 68). Nor does it seem unexceptional in colloquial Russian, though a strong tendency does exist.
(11) The idea to use these contexts (questions and sentences containing focus particles) to distinguish between focalized and postverbal positions, which coincide in the majority of cases, is borrowed from Izvorski 1997.
(12) This wording is closest to the one proposed in Meillet, Vendryes 1948 : 643.
(13) According to Bhatt (2005a : 10), this restriction on the semantics of correlatives is universal.
(14) The term is quoted because the relationship between the head and the relative pronoun is widely regarded as congruence. However, the mismatches are more often discussed in terms of agreement, as, for instance, in Corbett 2006.
(15) Within the system of Ingrian Finnish relative pronouns, special plural forms are only attested in the nominative, therefore in all oblique forms this kind of "mismatch" is obligatory. Such forms are used in all sorts of plural contexts, including questions and correlative constructions with the head noun in the plural:
(i) OK ke-ta toveritso-i-O muo kutsu-tti tullo-t iltasil who-PART friend-PL-PART we.NOM call-IPS.PST come.PRS-3PL in.the.evening
'The friends we invited will come in the evening'
(16) Floating quantifiers with the attracted case are, however, ungrammatical:
(i) NOM [right arrow] ALL
ihmisil ke-l mia kiruta-n ella-t OK kaik / *kaik-i-l man-PL-ALL who-ALL I.NOM write.PRS-1SG live.PRS-3PL all.NOM all-ALL Suome-s Finland-IN
'All the people to whom I write live in Finland'
(17) Some difference between the two groups is predictable under the analysis in point (if the nominative in both cases is treated as nominativus pendens). In the second group (27), the topicalization can be followed by case attraction. In the first group (16a), this is impossible because the case of the head and the relative pronoun coincide. However, the frequency of case attraction to cases other than nominative is about 60 % at most. Thus, the difference between the two groups is still statistically significant ([chi square], P < 0.01).
(18) Cf. also the Agreement Hierarchy in Corbett 2006 : 207, according to which relative pronouns are less subject to agreement mismatches than attributes.
(19) Actually, this argument is not as strong as it may appear, because a similar asymmetry is suggested by English coordinated structures, such as she and me. Cf. Parker, Riley, Meyer 1988, where such structures are used to claim that English coordination blocks case assignment. However, in Ingrian Finnish such asymmetry in the marking of conjuncts is never attested in structures other than IA-constructions. Therefore, if Harbert's (1983) analysis were adopted, it would mean that by the time IA operates the head noun has no case. It would mean, in turn, that the presence of a relative clause can remove the case of a conjunct for some reason independent from IA. This assumption seems highly improbable.
(20) In 2000 : 68, Bianchi observes that the "attracted" case of the demonstrative in most attested examples is the accusative. She suggests that it is an "external" case assigned to left-dislocated NPs, and thus, actually, no IA takes place. However, in Ingrian Finnish, left-dislocated NPs do not normally receive accusative case and the demonstrative does not restrict the number of cases available for IA, see, for instance, (20a) and (23a).
(21) Lehmann also suggests a connection between IA and the "adverbial relative clauses", though on somewhat different grounds (1984 : 351).
(22) At a certain stage, constructions like (36) were also possible in Russian, see [TEXT NOT REPRODUCIBLE IN ASCII] 2004 : 137. According to the data of the Russian National Corpus (ruscorpora.ru), such constructions with the pronoun KTO 'who' were relatively frequent in the first half of the 19th century and then gradually fell out of use.
(23) A relative clause was classified as "adverbial" in two cases: 1) if the main clause does not contain a position that could ni 'so' and/or sit 'then', which are normally used after adverbial clauses. Example (36) represents both cases.
(24) The original transcription is retained.
(25) A similar construction is possible in Votic, see Caoo 1963 : 268.
Table 1 Positional types of relative clauses in Ingrian Finnish postnominal clauses 54% extraposed clauses 15% correlatives 15% free relatives 16% Table 2 The frequency of nominative marking that does not correspond to the case frame of the main clause head head not ratio of in NOM in NOM heads in NOM relative pronoun in nom (16a) 13 7 0.65 relative pronoun not in nom (30) 6 33 0.15 Table 3 More vs. less embedded genitive phrases in relative constructions that occupy the subject position IA present IA absent ratio of IA less embedded 14 10 0.58 (direct object) more embedded (possessor; 2 21 0.09 postposition dependent) Table 4 Some properties of the clauses with pronouns kuka 'who' and mika 'what' questions correlatives 1. the case of the head noun + + and the relative pronoun always corresponds to the same position (1) 2. the relative construction follows the word that assigns case to the head in the main clause (3.1, 3.2) 3. extraposition (3.3) 4. the case of the head noun and the relative pronoun + + always coincides (4.4) 5. kuka (who.SG.NOM) with a noun in plural (4.2) * * 6. kutka (who.PL.NOM) with * * collective nouns (4.3) 7. mika 'what' in internal locative cases with a head * * referring to the owner of a place (4.5) 8. quantifier in * the head (4.6) 9. se 'that', tama 'this' * in the head (4.7) 10. appositive RC-s (4.1) * relative regular clauses externally with IA headed relative clauses 1. the case of the head noun + - and the relative pronoun always corresponds to the same position (1) 2. the relative construction * OK follows the word that assigns case to the head in the main clause (3.1, 3.2) 3. extraposition (3.3) * OK 4. the case of the head noun and the relative pronoun - - always coincides (4.4) 5. kuka (who.SG.NOM) with a noun in plural (4.2) OK OK 6. kutka (who.PL.NOM) with OK OK collective nouns (4.3) 7. mika 'what' in internal locative cases with a head OK OK referring to the owner of a place (4.5) 8. quantifier in OK OK the head (4.6) 9. se 'that', tama 'this' OK OK in the head (4.7) 10. appositive RC-s (4.1) OK OK
|Printer friendly Cite/link Email Feedback|
|Date:||Jun 1, 2013|
|Previous Article:||Finnicisms in the place-names of Ergeme Parish (Latvia).|
|Next Article:||Studies in Uralic etymology I: Saami etymologies.|