Two types of resumptive pronouns in polish relative clauses

This paper discusses two types of resumptive pronouns found in Polish relative clauses: (i) adjacent resumptives and (ii) embedded resumptives. It will be argued that adjacent resumptives are truncated forms of the relative operator, whereas embedded resumptives are ‘regular’ resumptive pronouns found in other languages like Hebrew and Russian. Support for this claim will come from analyzing the differences between adjacent and embedded resumptives, and analyzing the similarities between adjacent resumptives and relative operators. Cross-linguistic data involving the interaction of relative clause formation and resumption, as well as the interaction of cliticization and resumption will provide additional support for the above claim.


Introduction
In this paper I will discuss the properties of resumptive pronouns in Polish relative clauses. It will be argued that Polish has 'regular' embedded resumptive pronoun constructions, like those found in Hebrew, Russian, and English. However, it will be also shown that Polish has another type of resumptive pronoun, only present in one type of relative clause. It will be argued that this resumptive pronoun is in fact a truncated form of the relative operator. The paper will concentrate on constructions like the one below: (1) Marysia zna chłopców, co ich Ania lubi

Mary knows boys that them Anne likes 'Mary knows some boys that Ann likes'
What is interesting about (1) is that the resumptive pronoun is adjacent to the relative marker (Mykowiecka 2000, Fisiak 1978, Pesestky 1998). This configuration is only possible in relative clauses headed by a complementizer like relative marker: co, but not in relative clauses headed by an operator: który. 1 Consider the example below: *(2) Marysia zna chłopców, których ich Ania lubi Mary knows boys whom them Anne likes 'Mary knows some boys who Ann likes' 1 I will discuss briefly the differences between both types of relative clauses later in the paper.
For a full discussion and arguments for considering który to be an operator see Szczegielniak (2005). The operator który is inflected for number/gender/case/person and can be translated as 'which'. The marker co is not inflected and is a homonym of 'what' in Polish. However, resumptive pronouns are possible in który-relatives. They just have to be embedded. When embedded, resumptives are possible in both types of relative clauses: (3) a.

Ania lubi
Anne likes 'Mary knows some boys who I know Ann likes' b.
Marysia zna chłopców, co ja wiem że (ich) Ania Mary knows boys that I know that (them) Anne lubi likes 'Mary knows some boys that I know that Ann likes' I will call resumptive pronouns that can occur next to the relative marker pronouns adjacent resumptives as opposed to embedded resumptives, which require embedding in Polish and other languages (see Borer 1984, Erteschik-Shir 1992, Boeckx 2003).
It will be proposed that adjacent resumptive pronouns are actually truncated/cliticized forms of the operator który. Thus constructions like (4a) have the underlying form of (4b). 2 2 As pointed out by an anonymous reviewer, Pesetsky (1998) following Fisiak et.al. (1978) reports that resumptive pronouns in co-relatives are impossible in subject position, and when they carry accusative case. The former claim is correct, and I will provide an account why this is so in section 4. However, the claim that accusative resumptives in co-relative clauses are *(i) Ten samochód co Janek go widział wczoraj zniknął this car that Janek it saw yesterday disappeared 'This car that Janek saw yesterday disappeared' Adjacent resumptives, however, require adjacency to the relative marker co. Hence, (ii) is perfectly fine: Ten samochód co go Janek widział wczoraj zniknął this car that it Janek saw yesterday disappeared 'This car that Janek saw yesterday disappeared' The ungrammaticality of (i) stems from the Focus/Topic restrictions that allow the subject to be sandwiched between two relative markers. I will argue that (i) is ungrammatical since it is derived from (iii) which is also ungrammatical (see hypothesis 5 below).

*(iii)
Ten samochód co Janek którego widział wczoraj zniknął this car that Janek which saw yesterday disappeared 'This car that Janek saw yesterday disappeared' Two Types of Resumptive Pronouns in Polish Relative Clauses 2005), which is subsequently followed by truncation of the wh-part of the operator. 3 (5) Adjacent resumptives are truncated operators derived by the elimination of the wh-component.
In the section below, I will discuss resumption in Polish in more detail and show that embedded resumptives and adjacent resumptives have different properties. In section 3, I compare the properties of relative clauses with adjacent resumptives and relative clauses with both co and który markers in order to show that the former is derived from the latter (the hypothesis in 5). In section 4, I examine the morphological operation of relative pronoun truncation. In section 5, I provide an account of the difference between Russian and Polish as far as adjacent resumptives are concerned.

Resumption in Polish
In Polish, both co and który relatives allow embedded resumptive pronouns, both in subject and object relative clauses (the slash between co and który indicates 'either or'). The above discussion shows that embedded resumptives are not limited to relative clause constructions in Polish. However, resumption in these other types of A-bar movement can only be carried out via one type of resumption that involves embedded resumptives.
The above data strongly suggests that resumptive pronouns adjacent to the relative marker co are not the same pronouns that we find in embedded resumptive constructions. Following the hypothesis in (5), I propose that adjacent resumptives are clitic/truncated forms of the relative marker który and that adjacent pronoun constructions are derived from co plus który constructions (see 4a and 4b).
An If adjacent resumptives were to be derived from embedded resumptives, the lack of subject adjacent resumptives would be hard to account for. Note that subject pronouns in Polish can undergo climbing (footnote 4).
Thus the subject/object contrast cannot be captured by assuming that subject embedded pronoun cannot raise to a higher clause.
In this section I have shown that adjacent resumptives and embedded resumptives are two different kinds of resumptive pronouns. In the next section, I will show that relative clauses with adjacent resumptives behave like relative clauses that contain both co and który relative markers.
Two Types of Resumptive Pronouns in Polish Relative Clauses

Relative clauses with adjacent resumptives compared to co plus który relatives
Before I explore the similarities between co-relative clauses with adjacent resumptives and relative clauses containing the marker co and the operator który, let me examine the evidence that który is actually a relative pronoun and an operator. I will do this by examining the ability of head noun interpretation inside the relative clause in relative clauses that have który as a relative marker (with or without co) and relative clauses that do not have it.
Following Szczegielniak (2005), I assume that Polish has two types of relative The relative marker co is never inflected for anything in its role as a relative marker. It is also used exclusively in non-complement relative clauses. Complement relatives, as well as subordinate clauses utilize the complementizer że. In Polish, a comma before the relative marker does not indicate an appositive reading.
Following Szczegielniak (2005), I propose the following generalization how relative clauses in Polish are derived: According to Szczegielniak (2005), co-relatives do not allow an appositive reading, but do allow degree/amount readings (Carlson 1977), allow the breaking up idioms, induce Condition C-violations (the head noun reconstructs and induces a Condition C violation), allow the interpretation of a reflexive inside the relative clause, allow wide scope of the head noun, finally the relative does not behave as adjunct in co-relative clauses. In contrast, któryrelatives exhibit the opposite behavior that indicates that the head noun cannot be interpreted inside the relative clause. Take for example the breaking up of idioms. Idioms can be relativized in co-relatives, but not in który-relatives: (24) a. Słów co on nie rzucał na wiatr words that he not throw on wind 'Empty promises that he did not make' ??b. Słów których on nie rzucał na wiatr words which he not throw on wind 'Empty promises that he did not make' ??c. Słów co których on nie rzucał na wiatr words that which he not throw on wind 'Empty promises that he did not make' Following Szczegielniak (2005), I will assume that co plus który relatives behave like który-relatives. Both types of relative clauses behave in way that indicates that the head noun is not interpreted inside the relative clause (see 21). Hence, we observe: (i) the ability of head nouns to 'escape' Condition C effects, (ii) the ability of a given relative clause to license a restrictive meaning, and (iii) the inability to license a degree reading (for a full list see 23).
In the reminder of this section it will be shown that constructions with adjacent resumptives and co plus który relatives behave identically as far as the interpretation of the head noun is concerned. I will explore the ability to license appositive readings, the ability to escape Condition C effects, and the inability to license degree readings of co plus który and adjacent resumptive pronoun constructions. It will be shown that co plus który and adjacent resumptive pronoun constructions pattern together as far as head noun reconstruction in contrast to co-relative clauses that do not have an operator. Chierchia and McConnell-Ginet (1990) argue that appositive relative clauses are background assertions, and authors such as Emonds (1979), Sells (1985, Demirdache (1991), Del Gobbo (2003 have shown that appositive relative clauses are independent sentences. Co plus który constructions allow both a restrictive and appositive reading. The example below shows that this is also the case in adjacent pronoun constructions. A relative clause with co and The above examples support the claim that (25a) is actually derived from (25b) and that both constructions are derived via operator movement (see example 21), as opposed to (25c) which has to be derived via head noun raising (Szczegielniak 2005).

Constructions involving Condition C violations are another good
indicator of whether head noun reconstruction had taken place or not. The Rexpression 'John' can 'escape' a Condition C violation in co plus który constructions. This is also true for adjacent resumptive constructions, but not for bare co relatives (i.e. relatives without any kind of resumptive pronoun).
Consider the examples below: 9,10 (26) a. Condition C effects provide support to the claim that the head noun in co plus który and adjacent resumptive relatives can be interpreted outside the RC, whereas relative clauses containing just co seem to force head noun reconstruction. This in turn supports the claim that (26a) is derived from (26b) and that example (26c) is derived in a different fashion (head noun raising).
A third example where we see head noun reconstruction not taking place is in cases where a degree reading is not possible. Carlson (1977) was the first to observe that relatives can have degree/amount readings in addition to restrictive ones. Degree/amount relative clauses behave differently from regular restrictive relative clauses. Degree relatives indicate the degree of quantity, not identity of substance. For example in English, we have the following contrast (Grosu andLandman 1998, following Heim 1997): (27) a. It will take us the rest of our long lives to drink the champagne that/Ø they spilled that evening b. It will take us the rest of our long lives to drink the champagne which they spilled that evening Example (27a) can be a restrictive relative or a degree/amount relative.
In the latter case, we get identity of quantity and not of substance. Authors like Carlson (1977), Sauerland (1998), Heim (1997), Grosu and Landman (1998) have argued that in order to have a degree/amount reading the part of the DP 'champagne' that depicts the amount of champagne has to be in some way interpreted inside the RC. I will assume that in order to arrive at a degree reading, the head noun has to be interpreted inside the relative clause.
Polish relative clauses containing co plus który allow for an identity reading only (Szczegielniak 2005). This is also true for relative clauses containing adjacent resumptives, but not for relative clauses headed just by co: The degree reading data clearly indicates that head noun reconstruction is not only optional, but probably impossible in both co plus który and adjacent resumptive constructions. This again supports the claim in (5) that (28b) is derived from (28a) and that (28c) has a different derivation.
The above facts indicate that co plus który relative clauses and relative clauses with adjacent resumptives pattern in the same way as far as head noun reconstruction is concerned. However note that non-adjacent resumptive pronoun constructions behave identically to adjacent resumptive constructions and co plus który relatives in that they: (i) do not allow degree readings; (ii) permit the escape of Condition C effects; and (iii) license appositive meanings.

Relative pronoun truncation
In order to establish a more comprehensive picture of how adjacent resumptives are formed, I will explore the morpho-phonological relationship between the resumptive pronoun form and the który relative pronoun. As can 11 It is not a mystery why embedded resumptives would block head noun reconstruction in co relative clauses. As was shown earlier, embedded resumptives alleviate island effects, thus it is likely that reconstruction is blocked in constructions containing embedded resumptives. In fact, cliticization has to take place after który raising since it will be argued that the element któr-triggers obligatory raising of the relative pronoun in co plus który relatives, and optional raising in który relatives. This would account for the contrast between (33a) and (33b) woman(acc)that man(nom) recognized whose yesterday Janek zna od lat Janek knows for years 14 The (b) example improves if the material between co and który is focused. However, this would then imply a derivation where there is movement into the space between the two relative markers and not który remaining in situ. Obviously, this also leaves the question why the operator when not accompanied by co can remain in-situ as in (34).
'A woman who a man recognized yesterday Janek knows for years' I argue therefore that adjacent resumptives are clitic forms of the który relative pronoun in co plus który constructions, and cliticization takes place after the relative pronoun has raised out of its base position.

Cross-linguistic predictions -the case of Russian
There is a prediction that in languages where there are two ways of introducing a relative clause but no possibility of combining them, there should be no adjacent resumptive pronouns. This arguably could be the case in This correlates with the fact that in Russian there are no čto plus kotoryj relative clauses: (37) *a. Sobaka, čto kotoraja guliala vo dvore, byla golodnaja.
Dog that which walked in yard was hungry 'The dog that walked in the yard was hungry' *b. Sobaka, čto kotoruju my našli včera, byla golodnaja.
Dog that which we found yesterday was hungry 'The dog we found yesterday was hungry' Thus in Russian because there are no čto plus kotoryj relative clauses then there are no adjacent relative pronouns since there is nothing to cliticize next to čto. If adjacent resumptives were derived via movement of an embedded pronoun this difference between Polish and Russian would be a mystery. Especially if we adopt the proposal that the two types of Russian relatives are derived in the same way as Polish ones.

Conclusion
In this paper I have shown that adjacent resumptive pronoun constructions in Polish co-relative clauses are derived from relative clauses containing both the relative marker co and the relative operator który. Adjacent resumptives are a product of truncation of the wh-element of the operator. This explains the identical behavior of relative clauses with adjacent resumptives and co plus który relative clauses. It accounts for the fact that there are no object adjacent resumptive pronouns and for cases of multiple resumption.
Finally the proposal put forward here accounts for the lack of adjacent resumptives in Russian čto-relative clauses.