2.1 Noun phrases in Finnish

A sizeable chunk of the evidence discussed in this article is based on Finnish, a case marking language with fifteen morphological case suffixes. I will start by reviewing some elementary facts concerning the Finnish nominal system.

I assume that Finnish nominal phrases are composed of optional overt determiners or demonstrative pronouns (D), quantificational expressions (Q), numerals (Num), adjective phrases (APs), prenominal genitive modifiers (possessives, thematic arguments of the nominal predicate), a nominal head (N), and postnominal complements such as PPs or CPs, in that order:

1. (3) ne kaikki kolme tärkeää Peka-n oivallusta those.d all.q three.num important.ap Pekka-gen insights.n⁰ [_CP jotka auttoivat häntä] which helped him ‘all those three important insights of Pekka, which helped him’

I also assume that determiners/demonstratives, quantifiers and numerals are themselves heads, projecting determiner phrases, quantifier phrases and numeral phrases, respectively. Adjective phrases are inserted inside of the NP. Finnish is often said to lack the category of articles, but instead of an article the nominal projection may be headed by a demonstrative (glossed as D in this article). The underlying structure of Finnish DPs, assumed throughout this article, is thus as shown in (4).

1. (4)

   

One immediate concern here is that the syntax of the noun phrase in general is a controversial matter (Alexiadou, Haegeman & Stavrou Reference Alexiadou, Haegeman and Stavrou2007). What example (3) shows is that determiners/demonstratives, quantifiers, numerals, adjectives, nominal heads and nominal complements all fit together into the Finnish noun phrase, and the analysis in (4) reflects this. Apart from the noun head, all these elements are optional. Whether covert heads exist when the corresponding phonological exponents are missing, in particular whether the articleless noun phrases in Finnish are always DPs, is irrelevant to the ensuing discussion.

Another controvery concerns the question of whether the functional heads D, Q, and Num all c-command the nominal head as shown in (4), or whether they first project DP, QP and NumP and then these phrases c-command. The evidence presented in this article suggests that at least the numeral head c-commands the noun, while the question of whether D and Q also do so is much less important. I will assume that they c-command the noun head as well.

2.2 Case distribution in Finnish

In Finnish, case distribution affects demonstrative pronouns, quantifiers, numerals, all adjectives and the nominal head itself. Nominative, accusative, partitive, genitive and inherent/lexical cases undergo case distribution. Thus, whatever the explanation of case distribution may be, the nature of the case feature itself (i.e. structural vs. non-structural) is not relevant.Footnote ⁴

1. (5)

   1. (a) Minä katsoin si-tä yh-tä pien-tä punais-ta talo-a. I.nom watched that-par one-par small-par red-par house-par ‘I watched that one small red house.’

   2. (b) Minä näin se-n yhde-n piene-n punaise-n talo-n. I.nom saw that-acc one-acc small-acc red-acc house-acc‘ I saw that one small red house.’

   3. (c) Se yksi pieni punainen talo seisoi that.nom one.nom small.nom red.nom house.nom standed tie-n vieressä. road-gen beside.ine ‘That one small red house stood besides the road.’

   4. (d) se-n yhde-n piene-n punaise-n talo-n ove-n that-gen one-gen small-gen red-gen house-gen door-gen kahva. knob.nom ‘the knob of the door of that one small red house’

   5. (e) sii-nä yhde-ssä piene-ssä punaise-ssa talo-ssa. that-ine one-ine small-ine red-ine house-ine ‘in that one small red house’

Based on the data in (5), one could conjecture that case distribution inflects all elements within DPs and concerns all case features. This would be too simple. First, when the DP contains another phrase, such as an adjective phrase, not all elements of that phrase are assigned case by the external matrix element. For instance, the case of the complement of a participle phrase (AP) is not influenced by the case distribution inside of its DP host. It is the participle head, marked as VA in the example below, which controls the case of the complement of its own phrase.

1. (6) Pekka näki [_AP auto-a osta-neen] isä-nsä tulevan kotiin. Pekka saw car-par buy-va.past.acc father-acc.px/sg3 come home ‘Pekka saw his father, who had bought a car, come home.’

If the adjective head within a DP is modified by another adjective, that adjective takes genitive case; this occurs independent of the case of the DP as a whole or of the case of the adjective head.

1. (7) Pekka katsoi valtava-n iso-a laiva-a.Pekka watched huge-gen big-par boat-par‘Pekka watched a boat that was terribly big.’

Case distribution does not affect the complement of the noun head irrespective of the type of the complement. The data below shows this for an inherent case complement (8a), CP complement (8b) and clausal non-finite complement (8c).

1. (8)

   1. (a) Pekka katsoi [valtava-n iso-a laiva-a Helsingistä]_par.Pekka watched huge-gen big-par boat-par Helsinki.ela‘Pekka watched a boat from Helsinki that was terribly big.’

   2. (b) Pekka pohti [hypoteesi-ä jostaesitelmöitsijäpuhui]_par.Pekka thought hypothesis-par which presenter talked‘Pekka thought about the hypothesis that the presenter was talking about.’

   3. (c) Pekka pohti [tarve-tta-an kirjoittaakirja-n]_par.Pekka thought desire-par-px/sg3 write.A book-acc‘Pekka thought about his desire to write a book.’

Another way in which the simple generalization that case is distributed to all elements within DPs breaks down is when the DP contains an element that itself assigns a case. Demonstrative pronouns or adjectives never do this in Finnish, but numerals and other quantificational expressions representing quantities different from one do. These heads assign partitive case to their complements (Vainikka Reference Vainikka, Nelson and Manninen2003):

1. (9)

   1. (a) yksi minuuttione minute. nom ‘one minute’

   2. (b) nolla minuutti-a zero minute-par.sg ‘zero minutes’

   3. (c) puoli minuutti-a half minute-par.sg‘half a minute’

   4. (d) tuhat minuutti-a thousand minute-par.sg ‘thousand minutes’

Estonian and Inari Saami, historically related Finno-Ugric languages, show the same pattern:

1. (10)

   1. (a) Estonian (Rutkowski Reference Rutkowski, Holmer, Svantesson and Viberg2001: 7, ex. (10)) kaks sõdurittwo soldier.par ‘two soldiers’

   2. (b) Inari Saami (Nelson & Toivonen Reference Nelson, Toivonen, Nelson and Manninen2003: 326, ex. (7a)) čiččâm poccud láá tobbeen seven reindeer.par are there ‘there are seven reindeers.’

This phenomenon is well documented also in Slavonic languages such as Polish (Rutkowski Reference Rutkowski, Holmer, Svantesson and Viberg2001, Reference Rutkowski2002; Rappaport Reference Rappaport2002, Reference Rappaport, Bański and Przepiórkowski2003), Serbo-Croatian (Wechsler & Zlatić Reference Wechsler and Zlatić2000) and Russian (Babby Reference Babby1987). The interest here is in what happens when a DP assigned with a structural case other than partitive (say, accusative) contains a numeral that assigns partitive case to its own complement. Because of case distribution, this leads to a situation in which the complement of the numeral is seemingly assigned two structural cases: accusative by the matrix element and partitive by the numeral. In this situation the case assigned by the numeral to its complement overrides the matrix case assignment:

1. (11)

   1. (a) [Ne kaksi pien-tä auto-a]_NOM seisoivat tiellä. those.∅ two.∅ small-par car-par.sg standed road.ade ‘Those two small cars stood at the road.’

   2. (b) Minä löysin [ne kaksi pien-tä auto-a]_acc. I found those.∅ two.∅ small-par car-par.sg ‘I found those two small cars.’

The elements above the numeral in the DP structure (i.e. D, Q; see (4) above) are seemingly assigned no case at all, and hence they are glossed as ∅ in the above example. I will return to this phenomenon presently. It turns out that not all words in the DP are assigned the same case: if the DP contains another head which assigns its own case within the DP, this effect overrides a structural case assigned by the matrix element. As will be discussed later, the story is different if the DP is assigned an inherent or lexical case. Furthermore, I will later show that the above pattern is also sensitive to one of the ϕ-features of the nominal head, namely number.

Also, in Russian, Polish and Serbo-Croatian certain DP-internal numerals or quantificational expressions function as case assigners. The basic pattern of case distribution in these languages resembles that of Finnish: certain numerals or quantifers assign case to the elements they c-command. In the following examples, the numerals osiem ‘eight’ and piec/pjat' ‘five’ assigns genitive case, the so-called ‘genitive of quantification’, or gen(Q), to the elements they c-command. The demonstrative pronoun and the numeral appear in their bare form.

1. (12)

   1. (a) Polish (Rappaport Reference Rappaport, Bański and Przepiórkowski2003: 128, ex. (14)) Widzimy osiem dziewczynek. we.see eight.∅ girls.gen.pl‘We see eight girls.’

   2. (b) Polish (Franks Reference Franks2002: 151, ex. (16)) Te pięć kobiet poszło do domu. these.∅ five.∅ women.gen went.sg to home ‘These five women went home.’

   3. (c) Russian (Franks Reference Franks1995: 95, ex. (1b)) pro pjat' interesnyx knig about five interesting.gen.pl books.gen.pl

The Finnish numeral yksi ‘one’ behaves differently from other numerals in that it does not assign partitive case to its complement, while it itself exhibits case declension. Analogous to (11b), we thus have (13a) in which case is distributed to every nominal element; the bare numeral and partitive case would lead to ungrammaticality (13b).

1. (13)

   1. (a) Minä näin [se-n yhde-n piene-n auto-n]_acc. I saw that-acc one-acc small-acc car-acc ‘I saw that one small car.’

   2. (b) *Minä näin [se-n yksi pien-tä auto-a]_acc. I saw that-acc one.∅ small-par car-par

This behavior of yksi remains unexplained. One possibility is that it has started to behave like a determiner in Finnish, much like Italian uno ‘one’, and thus it belongs to a different grammatical category than the other numerals (see Laury Reference Laury1997). Rappaport (Reference Rappaport, Bański and Przepiórkowski2003) proposes that the numeral ‘one’ in Russian, which behaves similarly to Finnish and Serbo-Croatian with respect to case distribution, is an adjective. Following Rutkowski (Reference Rutkowski, Holmer, Svantesson and Viberg2001, Reference Rutkowski2002), who discusses similar data from Polish and Estonian, I will refer to yksi as an A-numeral (‘adjectival numeral’) and all the other cardinal numerals assigning partitive case as Q-numerals (‘quantificational numerals’).Footnote ⁵ Unless otherwise stated the analysis here concerns only Q-numerals.

One further detail concerning the Q-numerals in Finnish requires a comment. As shown in (9b–d) above, the complement of the numeral, i.e. the noun head, must occur in singular. The Q-numeral which values partitive case to this complement is also in singular. On the other hand, if the numeral designates a plurality, then the demonstrative must be in plural. Therefore, when we want to take the ϕ-features into account, a pair such as (11a–b) will be glossed as (14a–b).

1. (14)

   1. (a) [Ne kaksi pien-tä auto-a]_nom seiso-ivat tiellä. those.pl.∅ two.sg.∅ small-sg.par car-sg.par stand-past.3pl road.ade ‘Those two small cars stood at the road.’

   2. (b) Minä näin [ne kaksi pien-tä auto-a]_acc. I saw those.pl.∅ two.sg.∅ small-sg.par car-sg.par ‘I saw those two small cars.’

This effect is most clearly seen in a construction in which the nominal head is a plurale tantum noun that has only a plural form, e.g. sakset ‘scissors’. The construction is ungrammatical when such a nominal is c-commanded by a singular numeral, since the inherent feature [plural] of the nominal head disagrees with the feature [singular] of the numeral, see (15a–c). Thus, if the head noun is inherently plural, only the plural numeral can be combined with it, see (15d–e).

1. (15)

   1. (a) *kolme sakse-a three.∅ scissor-par.sg

   2. (b) *kolme saks-ia three.∅ scissor-par.pl

   3. (c) *kolme sakse-t three.∅ scissor-nom/acc.pl

   4. (d) kolme-t sakse-t three-nom/acc.pl scissor-nom/acc.pl

   5. (e) *?kolme-t saks-ia three-nom/acc.pl scissor-par.pl

Note, in (15e), that the plural numeral cannot function as a case assigner, hence it constitutes an A-numeral.

The singular Q-numeral agrees with its complement in ϕ-features, but it does not agree with the demonstrative. The agreement with the demonstrative is ‘semantic’, so that a Q-numeral not denoting a plurality, such as puoli ‘half’ (9c), will require a singular demonstrative. But this semantic dimension does not affect the ϕ-features of the Q-numeral itself, which is always singular.Footnote ⁶

One last demonstration that the Q-numeral is singular is that when the numeral phrase is the subject of a finite verb, the finite verb does not inflect in the plural, but in the singular instead (16).

1. (16) kolme auto-a aja-a tiellä three car-par.sg drive-sg road ‘three cars drive on the road’

2.3 Elements above the Q-numeral

In this section I will argue, following Brattico (Reference Brattico2008), that while the Q-numeral assigns partitive case to its complement, it is the matrix assigner which assigns case to the nominal elements above the Q-numeral in the DP structure.

In example (11), the determiner and the quantifier above the Q-numeral do not show overt case marking. However, consider the data in (17). The whole DP is headed by se ‘it’, the singular form of the demonstrative ne ‘those’. Examples (17a–b) show that the singular demonstrative does inflect overtly for matrix case, both accusative and nominative, while the Q-numeral still assigns partitive case to its complement lower in the structure. Example (17c) shows that when the DP occurs in the object position, the nominative form of se cannot be used.Footnote ⁷

1. (17)

   1. (a) Minä odotin [se-n puoli tunti-a]_acc. I waited that-acc half hour-par ‘I waited that half hour.’

   2. (b) [Se puoli tunti-a]_nom kului nopeasti. that.nom half hour-par lapsed fast ‘That half an hour lapsed fast.’

   3. (c) *Minä odotin [se puoli tunti-a]_acc. I waited that.nom half hour-par

One point about this data may require clarification. As rightly pointed out by an anonymous reviewer, it is possible that the demonstrative pronoun se-n ‘that’ is assigned genitive case rather than accusative case. This is because in the singular, the genitive and the accusative suffixes are homonymous (see footnote 4). If the demonstrative is in genitive, we cannot conclude that it has been assigned case by the matrix element.

This hypothesis is incorrect, however, as shown by the so-called negative test. According to this test, if the matrix clause in Finnish is negated, it forces partitive case on the object of the same clause or on the thematic subject of the embedded infinitive (MA-infinitive in the example below). To see this, compare (18a–b). As can be seen in (18c–e), genitive case does not undergo this alteration.

1. (18)

   1. (a) Pekka näki Merja-n/ hän-et lähte-mässä. Pekka saw Merja-acc him-acc leave-ma ‘Pekka saw Merja/him leaving.’

   2. (b) Pekka ei nähnyt Merja-a/ hän-tä lähte-mässä. Pekka not see Merja-par he-par leave-ma ‘Pekka did not saw Merja/him leaving.’

   3. (c) Pekka näki Merja-n/ hän-en lähte-vän. Pekka saw Merja-gen he-gen leave-va ‘Pekka saw Merja's/his leaving.’

   4. (d) Pekka ei nähnyt Merja-n/ hän-en lähte-vän. Pekka not saw Merja-gen he-gen leave-va ‘Pekka did not saw Merja's/his leaving.’

   5. (e) *Pekka ei nähnyt Merja-a/ hän-tä lähte-vän. Pekka not saw Merja-par he-par leave-va ‘Pekka did not saw Merja's/his leaving.’

Next we apply this test to the DPs headed by a demonstrative. In a construction like (17a), the demonstrative pronoun takes the partitive case when the sentence is negated. Thus, it must be in accusative rather than in genitive:

1. (19)

   1. (a) Minä odotin [se-n puoli tunti-a]_acc. I waited that-acc half hour-par ‘I waited that half hour.’

   2. (b) Minä en odottanut [si-tä puol-ta tunti-a]_par. I not wait that-par half-par hour-par

   3. (c) *Minä en odottanut [se-n puoli tunti-a]_par. I not wait that-acc/gen half hour-par

   4. (d) *Minä en odottanut [se-n puol-ta tunti-a]_par. I not wait that-acc/gen half-par hour-par

Let us return to case assignment above the Q-numeral. Another important piece of evidence comes from the behavior of Finnish adjectives. In Finnish, adjectives can occur either below or above the numeral within the DP. When the adjective is located above the numeral, it is assigned matrix case (20a, c). When it is located below the numeral, it is assigned partitive case (20b, d).

1. (20)

   1. (a) Minä odotin pitkästyttävä-t kolme minuutti-a. I waited boring-acc three.∅ minute-par ‘I waited the boring three minutes.’

   2. (b) Minä odotin kolme pitkästyttävä-ä minuutti-a. I waited three.∅ boring-par minute-par ‘I waited the boring three minutes.’

   3. (c) Minä söin pilaantune-en puoli leipä-ä.I ate rotten-acc half bread-par ‘I ate a rotten half of a bread.’

   4. (d) Minä söin puoli pilaantunut-ta leipä-ä.I ate half rotten-par bread-par ‘I ate half of a rotten bread.’

The following data shows that the distribution of the accusative and partitive case correlates with word order:

1. (21)

   1. (a) *Minä odotin kolme pitkästyttävä-t minuutti-a. I waited three.∅ boring-acc minute-par ‘I waited the boring three minutes.’

   2. (b) *Minä odotin pitkästyttävä-ä kolme minuutti-a. I waited boring-par three.∅ minute-par ‘I waited the boring three minutes.’

An additional reason to think that we need to distinguish the morphosyntactic environment above the numeral from the one below of the numeral concerns ϕ-features. Recall that the Q-numeral is itself in singular, triggering the singular form of the noun head in its complement. The demonstrative may be in plural, however. Since an adjective can occur both above and below the numeral, it is possible to examine whether its ϕ-features change as a function of its position. As shown in (22), they do.

1. (22)

   1. (a) ne kaksi pilaantunut-ta leipä-ä those.pl two.sg rotten-par.sg bread-sg

   2. (b) *ne kaksi pilaantun-eita leipä-ä those.pl two.sg rotten-par.pl bread-par.sg

   3. (c) ne pilaantune-et kaksi leipä-ä those.pl rotten-acc.pl two.sg bread-par.sg

   4. (d) *ne pilaantunu-t kaksi leipä-ä those.pl rotten-acc.sg two.sg bread-par.sg

An adjective below the Q-numeral must be in singular, while an adjective above the Q-numeral must agree with the demonstrative.

2.4 Locality and case assignment

Overall the generalization covering Finnish case distribution data appears to be that each of the nominal elements inside of the noun phrase is assigned its case features, individually, by the closest c-commanding assigner.

1. (23) Locality and case assignment

   All case assignees (e.g. D, Q, N, A) are assigned their case features by the closest c-commanding case assigner (e.g. T, v, Num, P).

The generalization is depicted in (24), where the matrix verb V assigns its case to the determiner/demonstrative pronoun, while the Q-numeral assigns partitive case to the adjective head and to the nominal head. In both cases, the assignment takes place under a local c-command relation.

1. (24)

   

Consider the data to see how (23) captures case distribution patterns in Finnish. Phrase-structure (24) itself shows the basic case distribution pattern with a Q-numeral, as illustrated in (17a–c), (20a–d) and (21a–c). In Finnish, the prenominal adjective phrase can contain a sequence of complements in addition to the adjective head. The matrix assigners or assigners in the projectional spine of the noun phrase (e.g. the Q-numeral) assigns case to the adjective head, but not to any of its complements inside of the AP (6). This follows from (23) since it is the adjective head, or some functional structure therein, that is the closest assigner for its own complement(s). Recall that in Finnish, many adjective heads assign object case (accusative or partitive). The adjective phrase thus constitutes a closed morphosyntactic environment within the DP, an objective case assignment sandwiched between nominative case distribution:

1. (25) se [auto-a osta-va] mies that.nom car-par buy-va.nom man.nom ‘that man, who was buying a car’

Case distribution does not extend to a complement of the noun head (8a–c), a phenomenon that follows from (23) if we assume that nominal heads, or some functional structure therein, are assigners. There is some evidence that the noun head assigns the genitive in Finnish (Vainikka Reference Vainikka1989, Reference Vainikka, Holmberg and Nikanne1993; Brattico & Leinonen Reference Brattico and Leinonen2009). If so, then the nominal head is the closest assigner for its own complement.

2.5 Case distribution in Slavonic

It is of some interest to examine whether a similar locality principle holds in other languages. Similar data emerges from a number of Slavonic languages, including Russian, Polish and Serbo-Croatian.Footnote ⁸ Russian resembles Finnish in that cardinal numerals greater or equal to five (pjat') assign case to the nominal element following them (Corbett Reference Corbett1978, Reference Corbett1983; Babby Reference Babby1987; Franks Reference Franks1994). In Russian, the Q-numeral assigns the genitive case, while in Finnish it assigns the partitive. On the other hand, the numeral odin ‘one’ declines like an adjective. These patterns can thus be compared to Finnish numerals. As in Finnish, the genitive assignment overrides matrix assignment and the case assignment appears to be based on locality in both languages (Babby Reference Babby1987, Franks Reference Franks1994).

Russian numerals five and above assign genitive plural, whereas the so-called paucal numerals oba ‘both’, dva ‘two’, tri ‘three’ and četyre ‘four’ assign the genitive singular. This latter pattern parallels the Finnish data, in which the singular partitive is likewise governed. In the Slavonic literature this exceptional number is sometimes called ‘paucal’ as it can be distinguished by various linguistic tests and by its historical development. These numerals also have additional properties which relate to ϕ-feature agreement in terms of animacy and gender (Corbett Reference Corbett1978, Reference Corbett1983). It thus seems that whereas Finnish numerals can be neatly partitioned into two classes, i.e. adjectival numerals and quantificational numerals, in Russian either there exist three distinct classes or there is a noun-adjective continuum (Corbett Reference Corbett1978, Reference Corbett1983).

In Finnish, the case-assigning functional heads do not themselves show overt case suffixes. This is also true of Russian. Thus, Franks says that for the genitive-assigning numerals ‘it is actually unclear whether these should be regarded as syncretic nominative/accusative forms or as caseless quantifiers’ (Franks Reference Franks1994: 601, note 4). Examples (26a–c) illustrate these properties in Russian, with Finnish equivalents in (27):

1. (26) Russian (Franks Reference Franks1994: 600, ex. (1))

   1. (a) Ivan kupil odnu mašinu. Ivan bought one.acc.sg car.acc.sg ‘Ivan bought one car.’

   2. (b) Ivan kupil tri mašiny. Ivan bought three.∅ cars.gen.sg ‘Ivan bought three cars.’

   3. (c) Ivan kupil pjat' mašin. Ivan bought five.∅ cars.gen.pl ‘Ivan bought five cars.’

1. (27) Finnish

   1. (a) Ivan osti yhde-n auto-n. Ivan bought one-acc car-acc

   2. (b) Ivan osti kolme auto-a. Ivan bought three.∅ car-par.sg

   3. (c) Ivan osti viisi auto-a. Ivan bought five.∅ car-par.sg

Other Slavonic languages show similar properties. In Polish, cardinal numerals higher than four assign genitive case to their complement in what appears to be nominative subject position (28a) and accusative object position (28b).Footnote ⁹

1. (28) Polish (Rutkowski 2001, ex. (7a–b))

   1. (a) [Osiem koni]_nom je. eight horses.gen eat ‘Eight horses eat.’

   2. (b) On kocha [osiem koni]_acc. he loves eight horses.gen ‘He loves eight horses.’

Similar phenomenon is attested in Czech and Serbo-Croatian:

1. (29) Czech (Franks Reference Franks1995: 136, ex. (17)) Pět pánu přišlo five men.gen.pl arrived.n.sg

1. (30) Serbo-Croatian (Franks Reference Franks1995: 97, ex. (6))

   1. (a) za osam dana in eight days.gen.pl

   2. (b) sa pet djevojaka with five girls.gen.pl

   3. (c) vlasnik pet malih kuća owner five small.gen houses.gen.pl ‘the owner of five small houses’

The data from Finnish demonstratives and adjectives (e.g. examples (17), (20) and (22)) supports the proposition that the elements above the Q-numeral are assigned by the matrix element, whereas elements below the Q-numeral are assigned by the Q-numeral itself. Matters become slightly more complex, however, when we look at the Slavonic languages. In Polish, a demonstrative that occurs above the Q-numeral is assigned accusative case, not the matrix case, in all direct case contexts:

1. (31) Polish (Rutkowski Reference Rutkowski2002: 7, ex. (6a–b))

   1. (a) *ci pięciu męczyzn these.nom five.∅ men.gen

   2. (b) tych pięciu męczyzn these.acc five.∅ men.gen ‘these five men’

In Serbo-Croatian and Russian, demonstratives and some other prenumeral elements are in some instances assigned genitive case and not the matrix case (Franks Reference Franks1995: 100–102). My impression of the Slavonic literature is that many authors try hard to maintain the idea that assignment is based on local government; the above deviant constructions could be explained away as involving for example different grammatical structure (Babby Reference Babby1987) or grammatical movement (Corbett Reference Corbett1979). It has also been argued that the DPs in (31a–b) occur only in positions where accusative case is assigned to the DP; hence (31b) does not violate (23) (Franks Reference Franks1995: 132–135, Rutkowski Reference Rutkowski2002).

2.6 Two-part models of Finnish case distribution

Let us turn to the two-part models of case. Hypothesis (23) suggests that case assignment is a process that occurs between words, not between words and phrases. Assuming the two-part model as a background, the simplest such hypothesis assumes that the case feature is first assigned to the nominal head N° on the basis of the syntactic position of the DP, and this feature is then shared by other elements within the determiner/nominal phrase.

The present data shows that the matrix element does not assign case features to the nominal head when the DP contains a numeral; rather, the overt case feature is distributed to the nominal elements above the numeral, while the numeral assigns partitive case to the nominal head. In other words, the presence of the numeral cuts the expected case-related connection between the matrix verb and the nominal head.

One way to explain the data in the framework of this hypothesis is to assume that the numeral functions as the nominal head of the noun phrase, with the partitive-marked constituent occurring as its complement. Several reasons speak against this hypothesis.

First, even if the numeral were the nominal head of the DP, it still does not show case agreement with the matrix element. Rather, a numeral which assigns partitive case always takes a ‘bare form’, that is, a form without any case suffixes. Yet this cannot be because the numeral could not inflect for case, since the numeral can decline like the nominal head. Thus compare (32a–b), where (32a) shows the Q-numeral assigning partitive case and (32b) shows the same numeral with a t-suffix, encoding plural nominative/accusative, which does not assign partitive to its complement.

1. (32)

   1. (a) ne kolme auto-a those three.sg car-sg.par ‘those three cars’

   2. (b) ne kolme-t auto-t those three-pl.nom/acc car-pl.nom/acc ‘those aggregrates of three cars’

In Finnish the numeral representing a cardinal quantity therefore has two forms, the Q-numeral and the A-numeral; the numeral yksi ‘one’ has only the A-numeral form. The correct generalization here appears to be that if the numeral inflects for matrix case, then it cannot assign partitive case, and vice versa, if the numeral assigns partitive case, it will appear in its bare form. But if the numeral declines, so does the nominal head. There is no evidence that the Q-numeral would receive its case feature from the external matrix case assigner.

Second, while the numeral is an optional constituent within the DP, it cannot appear alone in the DP, apart from elliptical constructions; rather, it always requires a complement noun. Third, the numeral cannot be complemented with a CP that can occur as a complement to nominal heads. This is shown in (33a–b).

1. (33)

   1. (a) mies jonka tapasin man who.acc.sg met ‘a/the man, who I met’

   2. (b) ??kolme jotka tapasin three who.pl met ‘the three who I met’

Adjectives which occur within the DP always modify the nominal head and never the numeral. In Finnish, the adjective parillinen is lexically ambiguous: it can mean either ‘even-numbered’ or ‘something which has a pair’. If this adjective actually modified an odd-numbered numeral, the expression below would be contradictory; this reading is nevertheless impossible.

1. (34) ne parilliset kolme sukkaa those even-numbered three socks ‘those three socks which have pairs’ ‘*those even-numbered three socks’

Normal adjectives do not seem to be capable of modifying the numeral, only the true nominal head.Footnote ¹⁰

One attractive feature of the hypothesis that case is assigned to the noun head is that it could unify case distribution with ϕ-feature distribution by taking the view that both features are distributed by the same grammatical mechanism: an upward percolation or, alternatively, feature checking or feature sharing. However, if we assume that the numeral functions as the nominal head of the NP, we lose this property, since as I pointed out earlier (see example (14) above) the numeral which assigns partitive case is in the singular, whereas the elements above the numeral can be both in the plural (35a) and in the singular (35b).

1. (35)

   1. (a) ne pitkästyttävä-t kolme minuutti-a those.pl boring-pl three.sg minutes-par.sg ‘those boring three minutes’

   2. (b) se pitkästyttävä puoli minuutti-a that.sg boring.sg half.sg minute-par.sg ‘that boring half a minute’

There is no grammatical number agreement between the Q-numeral and the demonstrative (35a); rather, there is singular agreement between the nominative head and the numeral.Footnote ¹¹

The last argument against assuming that the numeral constitutes the head of the DP is based on the fact that syntactic and semantic selectional requirements hold between the matrix verb and the noun, not between the matrix verb and the numeral. In examples (36–37), the numeral does not affect the semanticality of the clause, but the nature of the noun head does.

1. (36)

   1. (a) #Minä odotin kolme talo-a. I waited three house-par ‘I waited three houses.’

   2. (b) #Minä odotin talo-a. I waited house-par ‘I waited a house.’

   3. (c) Minä odotin kolme minuutti-a. I waited three minute-par ‘I waited three minutes.’

   4. (d) Minä odotin minuuti-n. I waited minute-acc ‘I waited a minute.’

1. (37)

   1. (a) Minä nukuin kolme minuutti-a. I slept three minute-par.sg ‘I slept three minutes.’

   2. (b) #Minä nukuin kolme talo-a. I slept three house-par.sg ‘I slept three houses.’

Let us turn now to the DP-hypothesis, according to which there are two case mechanisms: case assignment to DP and an additional case concord rule. One way to maintain the DP hypothesis in the face of the numeral data is to assume that the numeral selects another DP and then assigns partitive case to that DP. This is the model proposed by Vainikka (Reference Vainikka, Holmberg and Nikanne1993). The problem with this hypothesis is that it posits rules ‘DP→D+NumP’ and ‘NumP→Num+DP’ which predict structures of the form ‘D Num D Num D Num...N’. Such expressions are not attested. Examples (38a–c) provide several conceivable instances, all ungrammatical. I have found no evidence that without manipuluating the construction further one could insert quantifiers or determiners below the numeral.Footnote ¹²

1. (38)

   1. (a) *ne kaksi nii-tä kaike-t auto-ja those two these-par.pl all-acc.pl cars-par.pl

   2. (b) *ne kaksi si-tä kaikki-a auto-a those two these-par.sg all-par.sg cars-par.sg

   3. (c) *se puoli yh-tä pitkä-ä minuutti-a that half one-par.sg long-par.sg minute-par.sg

Another problem with this hypothesis is that it, too, requires the numeral to function as the head of the higher NP. This hypothesis was rejected above. These and several additional arguments against the DP-hypothesis are discussed in Brattico (Reference Brattico2008).

Given these difficulties in fitting the two-part model to the data, I turn next to Kayne's (Reference Kayne2005) one-part model. According to this model, neither DPs nor nominal heads play a special role in the syntactic assignment mechanism. Instead, each assigner is capable of encapsulating a morphosyntactic environment for whatever lies in its own complement. The data seen so far behaves exactly as expected under this hypothesis; indeed, the descriptive principle (23) constitutes one version of the one-part model.

There are several paths one might take after making this assumption explicit. In the rest of this article, I will examine several additional empirical facts and on this basis propose two generalizations which interact with the Kaynean case assignment mechanism to provide the best model for all the available data. Another plausible approach would be to develop Kayne's proposal into a fully formal model within the context of some more particular linguistic framework. I will not attempt this task in this article due to the additional data that would require discussion before any such model should be attempted. I will only make two brief comments concerning the relevance of the present data to the minimalist program (Chomsky Reference Chomsky1995).

The standard assumption within minimalism has been that case is implemented (i) by feature checking and (ii) by following the two-part model which assumes that case is first assigned to the noun head in syntax and distributed together with ϕ-concord (Bošković Reference Bošković2006). The present data is noncommital with respect to the assumption that case is implemented by means of case checking instead of case assignment/assignment. All the evidence we have seen so far is compatible with the assumption that the closest c-commanding assigner checks the case of the nominal elements it c-commands. This would represent a variation on the one-part hypothesis, not a departure from it.

The hypothesis (ii) stems from the lexicalist orientation of the MP. Since in many cases the nominal head bears the case feature checked by the matrix element, it has been a straightforward conclusion to assume that it has a distinguished role in case checking. Thus, when finite T agrees with the DP in case, there is a checking relation between T and the nominative case feature of the nominal head. This assumption must be abandoned: if there is any checking relation between T and the nominal elements within the DP containing the numeral, this relation must obtain between T and some element(s) above the numeral, since only those elements bear nominative case. But I think this problem can be solved if we assume that instead of such case checking, case is assigned by means of local valuation (Chomsky Reference Chomsky and Kenstowicz2001, Reference Chomsky, Freidin, Otero and Zubizarreta2008). We can assume that the case features of the nominals are assigned (valued) by the closest assigners, often functional heads, or simply as soon as a possible assigner is merged.Footnote ¹³

3.1 Case Hierarchy

The data we have seen so far suggests that each nominal element is assigned its case independently of the rest and that this mechanism obeys a locality effect in which the closest relevant element determines the case of another element. Thus, each phonological word within the nominal phrase is assigned case depending on the c-commanding (‘governing’) relations.

But this cannot be all. Consider the behavior of numerals in both Finnish and Russian. Here we cite the Finnish facts, which are identical to the Russian facts in the relevant aspects (see Babby Reference Babby1987 for the latter). As we have seen, the non-structural case suffix is distributed among the nominal elements (39).

1. (39) nii-ssä kolme-ssa piene-ssä talo-ssa those-ine three-ine small-ine house-ine ‘in those three small houses’

I assume here for the sake of simplicity that the inessive case is assigned either by the main verb, as specified in its subcategorization frame, or by a covert prepositional element which heads the DP (see Nikanne Reference Nikanne, Holmberg and Nikanne1993 for a version of the latter hypothesis). However, as example (39) demonstrates, the numeral ceases to assign partitive case to its complement when it is assigned inherent case by the preposition. Thus, examples (40a–c) are all ungrammatical.

1. (40)

   1. (a) *nii-ssä kolme-ssa pien-tä talo-a those-ine three-ine small-par house-par ‘in those three small houses’

   2. (b) *niissä kolme pien-tä talo-a those-ine three.∅ small-par house-par ‘in those three small houses’

   3. (c) *niissä kolme piene-ssä talo-ssa those-ine three.∅ small-ine house-ine ‘in those three small houses’

The same phenomenon is observed when the DP containing the numeral is assigned genitive (41a–c) or partitive case (41c–d):

1. (41)

   1. (a) [kolme-n talo-n]_gen ryhmä three-gen house-gen group ‘a group of three houses’

   2. (b) *[kolme-n talo-a]_gen ryhmä three-gen house-par group

   3. (c) *[kolme talo-a]_gen ryhmä three.∅ house-par group

   4. (d) [niitä kolme-a talo-a]_par those.par three-par house-par ‘those three houses’

   5. (e) *[niitä kolme talo-a]_par those.par three.∅ house-par

Only when the DP is assigned nominative or accusative can the numeral assign partitive case to its complement. Russian shows that the pattern is not restricted to Finnish. According to Babby (Reference Babby1987) and Franks (Reference Franks1994), Q-numerals within noun phrases in positions assigned inherent or lexical case do not assign genitive case to their complements. In such contexts all elements decline in the inherent case, in instrumental case in the example below:

1. (42) Russian (Franks Reference Franks1994: 601, ex. (3a–b))

   1. (a) Ivan vladeet odnoj fabrikoj. Ivan owns one.inst.sg factory.inst.sg ‘Ivan owns one factory.’

   2. (b) Ivan vladeet tremja fabrikami. Ivan owns three.inst.pl factories.inst.pl ‘Ivan owns three factories.’

In Polish direct case contexts, too, a numeral higher than four assigns genitive case to the adnominals within its own complement (43a–b), while this effect disappears in inherent or lexical case contexts (43c).

1. (43) Polish (Rappaport Reference Rappaport, Bański and Przepiórkowski2003: 124, ex. (3a–b))

   1. (a) Te miłe dziewczyny bawią się na podwórzu. these.nom.pl nice.nom.pl girls.nom.pl play.pl. refl on courtyard ‘These nice girls are playing in the courtyard.’

   2. (b) Sześć miłych dziewczynek bawi się na podwórzu. six nice.gen girls.gen play.pl. refl on courtyard ‘Six nice girls are playing in the courtyard.’

   3. (c) On ufa siedmiu koniom /*koni.(Rutkowski Reference Rutkowski, Holmer, Svantesson and Viberg2001, ex. (8)) he trusts seven.dat horses.dat horses.gen ‘He trusts seven horses.’

The pattern can also be found in Estonian (Rutkowski Reference Rutkowski, Holmer, Svantesson and Viberg2001) and in Inari Saami (Nelson & Toivonen Reference Nelson, Toivonen, Nelson and Manninen2003). In sum, in Finnish the closest assigner (numeral) does not assign its case to the element it c-commands if there is a structurally higher case assigner which assigns lexical, inherent, genitive or partitive case.

Babby (Reference Babby1987) explains the data by positing a Case Hierarchy, in which if an element E is c-commanded by a lexical case assigner, this effect overrides the effects of genitive and nominative/accusative assigners, while a genitive assigner overrides the effects of nominative/accusative assigners.

1. (44) A case-taker is assigned case by the c-commanding assigner. A case hierarchy (lexical case>genitive case>structural case) determines which of the competing cases is manifested overtly.

The general idea behind this principle is that structural case is assigned differently than are other case features. Babby, for instance, assumes that structural case is assigned at s-structure, inherent case features at d-structure; moreover, inherent case features cannot be erased during the derivation. Franks (Reference Franks1995) and Rutkowski (Reference Rutkowski, Holmer, Svantesson and Viberg2001) follow the same model in its essentials.

I would like to make few comments before accepting this hypothesis. First, note that although the Case Hierarchy can be argued to be necessary for the explanation of case distribution, it may not be sufficient. In many grammatical frameworks a given linguistic element can be c-commanded simultaneously by several assigners assigning structural case, for example, both v (or the transitive finite verb) and finite T (or C). The above Case Hierarchy does not resolve the case conflict between nominative and accusative case. The domain of case assignment must thus be restricted in some way in addition to being regulated by c-command relations. Babby (Reference Babby1987) assumes that these additional restrictions derive from locality, so that the more local structural case assigner wins the case competition. The present principle (23), or an equivalent system of locality, is therefore necessary.

Moreover, recall that in Finnish only the nominative or accusative singular Q-numeral assigns partitive case to its complement. All other numeral forms behave essentially like adjectives: they are case-takers and thus they decline like adjectives, and they do not assign case by themselves. Case assignment seems to be controlled both by the properties of the local c-commanding assigners and by the properties of the nominal head. Again, it has not escaped notice that something similar is found in the Slavonic languages, in which the ϕ-features of the noun head affect the morphosyntax of the numeral (Franks Reference Franks1995, Rappaport Reference Rappaport, Bański and Przepiórkowski2003). Whatever exactly is going on here, the Case Hierarchy by itself seems insufficient to explain the pattern.

To me the most interesting problem for the Case Hierarchy is the following. In the mainstream Slavonic tradition, the Case Hierarchy has been reduced in various ways to the difference between structural and inherent/lexical case features. This approach has to be revised in some form for Finnish, since the distinction between structural and inherent/lexical case does not align perfectly with the shift in the case concord pattern. Thus, while nominative and accusative contexts allow the partitive numeral, all other case contexts disallow it. The problem is that by all syntactic tests the genitive and partitive cases count as structural cases (Vainikka Reference Vainikka1989, Reference Vainikka, Holmberg and Nikanne1993, Reference Vainikka, Nelson and Manninen2003; Brattico & Leinonen Reference Brattico and Leinonen2009). Similar data emerges from Serbo-Croatian, to be discussed below.

I would like to draw three conclusions from this discussion. A case hierarchy of some sort looks impossible to avoid. Some case features are ‘weaker’ than others. The case hierarchy by itself does not explain all the data; a locality principle is required in addition. I will suggest that a certain kind of case hierarchy (section 3.2) together with the locality principle (23) provides the best description of the data. Finally, the idea that the case hierarchy is based on a distinction between structural and inherent case features is problematic because it does not apply to Finnish. In Finnish, the case hierarchy draws a line between nominative/accusative and the rest of the case features, while this is not the same distinction as the one between the structural and inherent case features. In the next section, I present an alternative formulation of the case hierarchy which tries to take this fact into account.

3.2 Rethinking the Case Hierarchy

In seeking for another solution, there is an independent generalization which seems to cover both situations, namely case assignment being overriden by inherent, genitive and partitive case assignment, and being overriden by the plural number feature of the nominal. The Q-numeral which assigns partitive case in Finnish is a marked head which has the following three special properties: (1) it assigns partitive case, (2) it is grammatically in the singular, and (3) it does not show overt case suffixes. Let us call this numeral the ‘bare numeral’. Its exceptional behavior follows from an independently motivated generalization proposed by Stowell (Reference Stowell1981). Many assigners such as complementizers, prepositions and other functional heads are not themselves assigned overt case, and thus they too appear to be ‘bare forms’. The opposite is also true to some extent for Finnish, since if some element is assigned overt case, it cannot by itself serve as a case assigner. This is true of Finnish demonstrative pronouns and adjectives. This explains the peculiar behavior of the Finnish numeral yksi ‘one’ which does not assign partitive case to its complement, contrary to other numerals, but is itself assigned overt matrix case. Finally, some grammatical elements, such as most adverbs and grammatical particles, seem at least prima facie to fall outside of the assigner–assignee system in Finnish. The following ‘Case-Resistance Principle’, discussed in Stowell (Reference Stowell1981) and Bošković (Reference Bošković2006: 526), thus brings us a long waytowards an explanation of the Finnish facts.

1. (45) Case-Resistance Principle

   No grammatical element can be simultaneously assigned overt case and function as a assigner.

Keeping this generalization in mind, let us turn to the problem with the plural marking. Recall that the Finnish numeral loses its ability to function as a assigner when it is marked for plural. This is shown in (46a–c).

1. (46)

   1. (a) kolme mies-tä three.∅ men-par.sg ‘three men’

   2. (b) kolme-t miehe-t three-acc/nom.pl men-acc/nom.pl ‘three men’

   3. (c) *kolme-t mies-tä three-acc/nom.pl men-par.sg

However, the Finnish t-form in (46b) also carries nominative and accusative case features, and the t-ending is thus not just a plural marker (Nelson Reference Nelson1998: 40–42). Indeed, with pronouns the t-form encodes accusative case unambiguously.

1. (47)

   1. (a) Hän näki minu-t. he.nom saw me-acc ‘He saw me.’

   2. (b) Minä näin häne-t. I.nom saw him-acc ‘I saw him.’

The t-forms listed in (46b–c) are likewise marked overtly for case. Further evidence for this assumption can be derived from the behavior of nominal ϕ-features. Consider the fact that when the DP as a whole is in the singular and is assigned either nominative or accusative, only the bare form of the numeral is possible:

1. (48)

   1. (a) Pekka odotti se-n puoli minuutti-a. Pekka waited that-acc.sg half.∅ minute-par.sg ‘Pekka waited that half a minute.’

   2. (b) *Pekka odotti se-n puole-n minuuti-n. Pekka waited that-acc.sg half-acc.sg minute-acc.sg ‘Pekka waited that half a minute.’

   3. (c) Se puoli minuutti-a kului nopeasti. that.nom.sg half.∅ minute-par.sg lapsed fast ‘That half a minute lapsed fast.’

   4. (d) *Se puoli minuutti kului nopeasti. that.nom.sg half.nom.sg minute.nom.sg lapsed fast ‘That half a minute lapsed fast.’

However, if the DP containing the numeral is in the plural form and is assigned nominative or accusative, either the bare form or the non-bare form can occur (49a–d):

1. (49)

   1. (a) Pekka osti ne kolme-t auto-t. Pekka bought those.acc.pl three-acc.pl cars-acc.pl ‘Pekka bought those three cars.’

   2. (b) Pekka osti ne kolme auto-a. Pekka bought those.acc.pl three.∅ cars-par.sg ‘Pekka bought those three cars.’

   3. (c) Ne kolme-t auto-t hajosivat. those.nom.pl three-nom.pl car-nom.pl broke.down ‘Those three cars broke down.’

   4. (d) Ne kolme auto-a hajosivat. those.nom.pl three.∅ cars-par.sg broke.down.3pl ‘Those three cars broke down.’

Thus, the reason why the t-marked plural numeral cannot function as a assigner in Finnish cannot be because the DP as a whole is in the plural. Rather, the reason is the Case-Resistance Principle: the t-marked form of the numeral, realizing also the case features, cannot function as a assigner. It is easy to see that this generalization also subsumes the case where a nominal element is assigned an inherent or lexical case: an element that is assigned overt inherent/lexical case cannot function as a assigner, thereby making the numeral unable to assign partitive case.

Before continuing, one point needs to be sorted out. As shown above, in plural nominative and accusative contexts there are two possible forms for the numeral, the form which carries overt case features and the bare form. The choice carries a semantic function. From the semantic point of view, a bare numeral phrase such as two socks refers to a plurality. This interpretation is triggered irrespective of the singular form of the plural/nominal head. When the nominal head is marked for plural as well, we get a layered double plurality: once from the numeral irrespective of its form, and once from the nominal head. In a DP containing both the numeral and the plural suffix, we get a reading where the DP refers to two pluralities:

1. (50)

   1. (a) Pekka osti kolme sukka-a. Pekka bought three.∅ socks-par.sg ‘Pekka bought three socks.’

   2. (b) Pekka osti kolme-t suka-t. Pekka bought three-acc.pl sock-acc.pl ‘Pekka bought three pairs of socks.’

In (50b), the DP encodes plurality twice: the numeral three plus the plural suffix. This results in the reading ‘three aggregrates (i.e. pairs) of socks’. In (50a), there is only one feature encoding plurality, resulting in the simpler reading ‘three socks’. The choice between the bare form and the inflected form within a plural DP thus follows from independent semantic considerations. Accordingly, we can say that the choice of the type of the numeral carries semantic function.

In sum, there is some evidence that assigners have a tendency to be realized as bare forms.Footnote ¹⁴ What thus divides the set of case features into a ‘hierarchy’ in Finnish is the fact that while the nominative and accusative cases have the option of remaining covert, all other case features must be expressed overtly. This explains why all case features except the nominative and accusative prevent the bare form from appearing and, therefore, also prevent the case assigner from occurring. Four additional pieces of evidence support this explanation.

First, in some Finnish grammatical constructions, a full DP, when inserted into a position assigned by accusative case if occupied by a pronoun, is expressed by a zero-derived bare form (sometimes called ‘zero-accusative’). This phenomenon is shown in (51).

1. (51)

   1. (a) Meidän täytyy löytää talo (/hän-et). We must find house.∅ him-acc ‘We must find a house/him.’

   2. (b) Talo (/hän-et) löydettiin. house.∅ him-acc found.pass ‘A house/he was found.’

   3. (c) Etsi talo (/hän-et)! find house.∅ him-acc ‘Find a house/him!’

This phenomenon never occurs with other object case features or inherent case contexts in Finnish. For example, when a finite sentence is negated in Finnish, the object appears in the partitive case. Partitive case is never neutralized in these contexts:

1. (52)

   1. (a) Meidän ei täydy löytää talo-a/ *talo. We not must find house-par house.∅ ‘We must not find a house.’

   2. (b) Talo-a/ *talo ei löydetty. house-par house.∅ not found.pass ‘A house was not found.’

   3. (c) Älä etsi talo-a/ *talo! don't find house-par house.∅ ‘Don't search for the house!’

Thus, the zero-derived bare forms can appear in a context that is otherwise assigned nominative or accusative case, but this option is not available for genitive, partitive or inherent case. There is thus an asymmetry between the nominative/accusative and other case features as regards the zero-derived forms. I propose that the bare Q-numeral is a similar, zero-derived from of the accusative (and nominative).

Second, some cross-linguistic evidence for this principle can be found from Polish numeral phrases. As shown in Przepiórkowski (Reference Przepiórkowski, Grover and Vallduví1996), in Polish the following generalization holds. When a DP containing a numeral such as five occurs in a context where the DP is assigned nominative or accusative case, then the demonstrative and the numeral take a bare form which (i) is ambiguous between the nominative and accusative and which (ii) assigns genitive case to the nominal head. When the DP is assigned any other case, then the demonstrative, the numeral and the noun head are assigned that case unambiguously, and the numeral does not assign any separate case to the nominal head. What is surprising about these facts is that they mirror the Finnish data almost completely even though Polish and Finnish are historically unrelated languages. They therefore provide a compelling evidence that the nominative and accusative have at least the special property that they may remain morphologically ambiguous or morphologically unrealized in some contexts.

Third, Wechsler & Zlatić (Reference Wechsler and Zlatić2001) provide similar data from Serbo-Croatian. In Serbo-Croatian, certain borrowed proper names as well as certain quantifiers do not decline and thus they behave like bare ‘frozen forms’. These proper names are not assigners either; rather, what they share with assigners such as quantifiers in this language is that they are bare forms. This allows us to study the effects of the Case Hierarchy independently of case assignment. The authors first show that, with the exception of dative and instrumental, these proper names may occur within NPs that are assigned case features, and that these case features are shared by other nominal elements such as demonstratives. However, the proper names remain caseless and are thus considered as lexically unspecified for case. In the case of dative or instrumental case features, however, a proper name cannot occur unless some other grammatical element, such as a preposition, carries that case feature. The authors thus propose the following Dative/Instrumental Case Realization Condition (page 547):

1. (53) If a verb or noun assigns dative or instrumental case to an NP, then that case must be morphologically realized by some element within the NP.

This is the same principle we have laid down in this section except that it singles out the dative and instrumental as cases that must be realized overtly, to the exclusion of others.

The fourth argument in favor of this version of the Case Hierarchy is based on the observation that the nominative stands apart as being a suffixless case. There is no need to present elaborate arguments to show that the nominative case does not need to be expressed overtly, as it corresponds to the zero morpheme.

Here it is also worth recalling once more the main motivation for rejecting an explanation in terms of lexical vs. structural case: In Finnish, partitive and genitive are structural cases by all relevant syntactic and semantic tests, yet they behave like inherent cases with respect to case distribution. On the other hand, these cases differ from the nominative and accusative in that, unlike the nominative and accusative, they cannot remain covert.

3.3 Summary

In sum, I suggest that the facts concerning case distribution in Finnish and in several Slavonic languages fall under the following three descriptive generalizations.

1. (54)

   1. (a) Morphosyntactic Asymmetry

      A case-assigner must be a caseless bare form in terms of its overt morphosyntax (adopted from Stowell Reference Stowell1981).

   2. (b) Case Hierarchy

      Case features are organized into a Case Hierarchy which requires that some case features be realized overtly (adopted from Babby Reference Babby1987; Franks Reference Franks1994, Reference Franks1995; Wechsler & Zlatić Reference Wechsler and Zlatić2001).

   3. (c) Case Valuation

      All lexical elements are assigned their overt morphological case by the local c-commanding assigner. A given element may be assigned several case features at once (Blake Reference Blake2001: 107–109, Sadler & Nordlinger Reference Sadler and Nordlinger2006) or just one (Kayne Reference Kayne2005, Brattico Reference Brattico2008).

The principle of Morphosyntactic Asymmetry captures the fact that assignment is asymmetric at least at the level of overt morphosyntax. While it is still unclear to me whether this principle as such should be assumed as part of the UG, it works well with a significant range of data, Finnish in particular.

The Case Hierarchy organizes case features into a hierarchy in terms of whether they must be realized overtly. Babby (Reference Babby1987) thinks that this distinction coincides with the distinction between structural vs. non-structural case; but this hypothesis is problematic for Finnish, in which strong arguments have been presented that both genitive and partitive case are structural (Vainikka Reference Vainikka1989, Reference Vainikka, Nelson and Manninen2003), and for Serbo-Croatian, where dative and instrumental case are distinguished from the rest.

Kiparsky (Reference Kiparsky and Peters1972, inter alia) notes that there is a tendency for the semantic relevance of the case features to correlate with the Case Hierarchy, so that the semantically more relevant forms tend to be expressed overtly. Wechsler & Zlatić (Reference Wechsler and Zlatić2001) rely on this generalization in their explanation of the Serbo-Croatian data, and Rutkowski (Reference Rutkowski2002) in his explanation of the Polish and Estonian data. Wechsler & Zlatić suggest that the dative and instrumental in Serbo-Croatian have more semantic content than the other case features, and must thus be retained at surface structure. It is not entirely wrong to say, conversely, that the nominative and accusative in Finnish contain ‘less semantic content’ than many other case features. On the other hand, sincepartitive and genitive cases seem to be structural cases as well (Vainikka Reference Vainikka1989, Brattico & Leinonen Reference Brattico and Leinonen2009), and the partitive in particular is devoid of any semantic content (Vainikka Reference Vainikka, Nelson and Manninen2003), this suggestion perhaps cannot be maintained in any very strong sense. Moreover, the relevant notion of the semantic content of a case feature, and the exact mechanisms of how this notion influences the grammatical and morphological realization of case features, need to be worked out. While Rutkowski thinks that it coincides with semantic markedness and ‘cognitive, perceptual complexity’, Lapointe (Reference Lapointe, Barlow and Ferguson1988) proposes a universal semantic ranking of case features. Meanwhile, since the distinction must be encoded in the syntax in any case, we could borrow the strong/weak distinction from a generative theory: a weak case feature is one which does not need to be expressed overtly, while a strong case feature is one that must be. Languages will then differ in terms of which case features are weak and which are strong.

The Case Hierararchy must leave room for a considerable amount of cross-linguistic variation. The principle itself requires case to be realized overtly, but it does not say how this realization takes place. Thus, in some languages the case feature can be realized on any nominal element as long as it is realized on one. The Australian language Gooniyandi is such a language (McGregor Reference McGregor1990). In other languages, one nominal element may realize several case features at once. See Blake (Reference Blake2001) for a summary of the cross-linguistic variation concerning case realization.

The hypothesis of one-part case assignment captures the idea that case distribution and case assignment are based on the same grammatical principle. At the heart of this principle lie two interwoven ideas, namely that (i) such assignment is based on locality, a property that is pervasive in natural language grammar, and (ii) assignment concerns individual words rather than phrases.

All three principles express a uniform theme: they suggest that case assignment is a ‘surface effect’. Thus, an element cannot assign anything if it shows overt assignment itself (54a); some case features must be realized overtly (54b); and finally, assignment establishes relations between functional elements and words, not between functional elements and syntactic phrases (54c). Moreover, if these principles regulate syntactic configurations in Finnish, we cannot escape the conclusion that there are constraints having to do with overt morphosyntax which filter out otherwise legitimate syntactic possibilities.