A comparison of the modal dać się structure with the dispositional middle in Polish ∗

The paper compares the modal dać się structure with the dispositional middle in Polish. It is argued that the two structures are similar as regards argument realization, i.e. in both constructions, the theme argument appears in the structural subject position. The two structures also have a dispositional meaning in common. However, they show a number of differences. They differ in the presence of a syntactically active agent, their aspectual properties, the availability of episodic interpretations, the obligatory presence of an adverbial modifier, and verb class restrictions. Although these differences seem to argue against a common syntactic derivation for the two structures analysed here, they do not preclude classifying the modal dać się structure as a subtype of the dispositional middle. If middles are seen as a notional category, understood as a special meaning that different grammatical structures can have, along the lines postulated by Condoravdi (1989), then the modal dać się structure can be subsumed under the label of middle. In fact, it is argued that the modal dać się structure represents Type II middles in Ackema and Schoorlemmer’s (2005) typology, and it shows properties typical of lassen-middles in German (Pitteroff 2014).


Introduction
The verb dać 'give' appears in a wide variety of structures in Polish, including ditransitive, causative, impersonal and modal ones.Out of these four configurations, the ditransitive dać 'give' has been most frequently analysed in the literature (cf., for instance, Topolińska 1993, andCitko 2011).The distribution and properties of the remaining three constructions listed above have recently been examined by von Waldenfels (2012Waldenfels ( , 2015) ) in a corpus study based on the data from Russian, Polish and Czech.
The aim of this paper is to take a closer look at the structure containing the verb dać 'give', followed by the pronoun się 'self' and the infinitival complement, the so-called modal dać się construction, and compare it with another type of structure, namely the dispositional middle in Polish.Since the modal dać się construction and the middle share a surface syntactic structure and a dispositional meaning, one might be inclined to classify them as belonging to the same category.The comparative study carried out in this paper demonstrates that in spite of a number of common traits, the two constructions under scrutiny exhibit some distinct syntactic properties that cast doubt on a uniform syntactic treatment of the two structures.However, on the grounds of a common dispositional meaning that these structures exhibit, it is possible to classify them as belonging to the middle category, understood in notional terms, along the lines of Condoravdi (1989).
Although the syntactic properties of the modal dać się structure and the dispositional middle are examined here in detail, we do not intend to analyse the derivation of either of them (for an analysis of the syntax of the modal dać się construction, cf.Bondaruk 2015).
The paper consists of five sections.In section 2, the two types of structure to be analysed in the paper are briefly described and characterised.Section 3 focuses on the middle construction from a cross-linguistic perspective in order to prepare the ground for a juxtaposition of the Polish middle and the modal dać się construction, undertaken in section 4. First, in section 4.1, the similarities between the two types of structure are pointed out, mostly relating to their surface syntactic structure, the implied agent, dispositional meaning, and generic interpretation.Subsequently, in section 4.2, the focus is on the differences between the two constructions concerning their aspectual properties, the availability of the agent, the optionality of the adverb, and verb class restriction.Section 5 provides the conclusions.

The data to be analysed
It has been noted above that the modal dać się structure comprises the verb dać 'give', followed by the pronoun się, homphonous with the reflexive się, and the infinitival complement, as demonstrated in (1) below: 1   (1) Te koszule dają się (łatwo) prać. 2 these shirts-nom.plgive-3pl się easily wash-inf 3 'These shirts can be washed easily.' 1 Von Waldenfels (2012: 153) calls a structure such as (1) 'a modal passive'.However, we prefer to call it simply 'modal', without referring to its alleged passive character.Adopting the term 'modal passive' would force us to explain in what sense these sentences are passive, since they do not exhibit any passive morphology. 2 The pronoun się is glossed throughout as 'się', instead of being treated as a reflexive pronoun, as it does not have a reflexive function in the structures analysed in the paper (we owe this remark to an anonymous reviewer). 3The following abbreviations have been used in the paper: acc -accusative, dat -dative, inf -infinitive, instinstrumental, nom -nominative, non-vir -non-virile, pl -plural, and sg -singular.
Sentence (1) exhibits an inanimate surface subject te koszule 'these shirts', which determines the φ-features of the verb as the 3 rd person, plural, and which corresponds to the theme argument of the verb in the infinitival complement, namely prać 'wash'. 4The pronoun się seems to be identical to the surface subject.
Sentence (1) has a modal meaning, as confirmed by its English translation which contains the modal verb can.The modality involved in (1) is called dispositional and will be elaborated on in Section 4.1.3below.The brackets around the adverb łatwo 'easily' in (1) imply its optionality.Although the English rendering of (1) makes use of the passive verb form, the verb in the Polish version of example (1) is active, and bears no trace of passive morphology (cf.footnote 1).
The modal structure in (1) seems to be similar to the causative permissive structure, as in (2) below: (2) Marek daje się (łatwo) zapraszać na przyjęcia.Mark-nom.sggive-3sg się easily invite-inf to parties 'Mark lets himself be invited to parties.' In (2) the verb dać 'give', co-occurring with an animate subject Marek 'Mark', has a permissive meaning equivalent to pozwalać 'let'.The difference in meaning notwithstanding, the modal and the causative structures in (1) and ( 2), respectively, seem to have an identical syntactic surface structure, since both have a verb agreeing in φ-features with the nominative case marked subject, and they show the pronoun się co-referential with the subject.The syntactic similarity between the modal and the causative permissive structure will be revisited in section 4.1.3.
Moreover, it is worth pointing out that besides the modal dać się structure as in (1) above, in which the verb agrees with the surface subject, there exists an impersonal modal construction, as in (3): (3) Tę koszulę daje się (łatwo) prać.this shirt-acc.sggive-3sg się easily wash-inf 'This shirt can be washed easily.' In (3) the verb appears in the default 3 rd singular form, and the DP that precedes it does not represent a surface subject, but rather a complement of the verb prać 'wash', as it occurs in the accusative, not the nominative case. 5However, from the point of view of semantics, (3) is identical with (1), as both of them have a modal meaning, equivalent to a modal verb such as można 'can'.
(i) Te koszule dawały się prać (łatwo).these shirts-nom.pl.non-vir gave-3pl.non-virsię wash-inf easily 'These shirts could be washed easily.' In (i) the verb agrees with the subject in the 3 rd person, plural number and non-virile gender. 5In example (i) below with the plural DP in front of the verb, we cannot easily determine the case form of the DP, as it is syncretic between the nominative and the accusative.No such syncretism appears in the singular, and for this reason the singular DP is used in (3) above: The main focus of the paper is on structures such as (1), and reference will be made to the other two constructions in (2) and (3) only when they are relevant to the discussion.
Furthermore, the modal dać się structure in Polish closely resembles an analogous structure found in German, the so-called lassen-middle (the term taken from Fagan 1992, Ackema and Schoorlemmer 2005, and Pitteroff 2014), illustrated in (4) below, taken from Pitteroff and Alexiadou (2012: 214): 6   (4) Das Auto lässt sich (angenehm) fahren.the car-nom let-3sg się comfortably drive-inf 'The car drives carefully.' The only difference between the Polish modal structure in (1) and the German lassenmiddle in (4) lies in the matrix verb itself, which in the former is dać 'give', while in the latter it corresponds to lassen 'let'.Otherwise the two structures are identical as regards the verbal agreement with the nominative case-marked surface subject, the presence of the pronoun się co-referential with the subject, the infinitival complement in the complement clause, and the theme role of the surface subject with respect to the verb in the infinitive.Actually the similarities between the two structures in Polish and German have made von Waldenfels (2012Waldenfels ( , 2015) ) conclude that due to the language contact between West Slavic languages and German, the verb lassen had a role to play in shaping the development of the Slavic grammaticalised give. 7What is more, the term lassen-middles used in relation to structures such as (4) in German implies their treatment as a subtype of the middle constructions proper.This, in turn, makes one wonder whether the Polish modal dać się structure can be regarded as an instance of the middle construction, and thus justifies a comparison between these two structures undertaken in this paper.
Middle constructions in Polish, like in other languages (cf.section 3 below), represent generic modal statements about the understood object (Lekakou 2005: 10, Ackema andSchoorlemmer 2005: 140), as exemplified in (5): (5) Te koszule piorą się łatwo.these shirts-nom.plwash-3pl się easily 'These shirts wash easily.' (i) Te koszule daje się (łatwo) prać.these shirts-nom/acc give-3sg się easily wash-inf 'These shirts can be washed easily.' In (5) the notional object of prać 'wash' corresponds to the surface subject koszule 'shirts', which is associated with a theme theta role.It is also the surface subject that determines the verbal agreement in (5).Just like the modal dać się structure, introduced above, the middle in (5) contains the pronoun się.The presence of an adverb łatwo 'easily' in (5) is obligatory, in contradistinction with (1), where the manner adverb is just optional.In the middles such as (5) the valency of the verb has been manipulated, i.e. the internal argument is found in the subject position and the external argument is only implied (Levin 1993: 25-26), but is not overtly manifested, which makes this structure similar to the passive.8However, the middle in Polish never shows any passive morphology.9Compare (5) with its passive equivalent in (6): (6) Te koszule zostały (łatwo) uprane.these shirts-nom.plwere easily washed 'These shirts were washed easily.' The passive sentence in (6) contains the auxiliary zostać 'get/become', followed by the passive participle of the verb prać 'wash'.The distinct morphology that the middle and the passive exhibit seems to argue against analysing them in the same way (but see Stroik 1992, 1999, and Hoekstra and Roberts 1993 for an A-movement analysis of both passives and middles).
Moreover, middle constructions such as (5) are sometimes referred to in the literature as dispositional middles (cf.Alexiadou and Doron 2011: 26), as the generalisations they express are true on account of the inherent properties of their subjects (cf.Lekakou 2005).In other words, in the dispositional middle, some property inherent in the subject facilitates the action expressed by the verb (Lekakou 2005), for instance in (5) above, these shirts have some inherent property that makes washing them easy.The term 'dispositional middles' will be used interchangeably with the label 'middles' throughout this paper.
Another structure similar to middles corresponds to anti-causatives (or inchoatives), which typically express a change of state, as in (7) below: (7) Te koszule uprały się (łatwo).these shirts-nom.plwashed się easily 'These shirts washed easily.'In (7), like in (5) above, the valency of the verb has been affected, i.e. the external argument of the verb prać 'wash' is not syntactically represented, and the internal argument surfaces in the superficial subject position.In contradistinction to passives and middles, anti-causatives are normally taken to lack the overt or implied presence of an agent (Levin 1993: 26, Levin and Rappaport Hovav 1995, Reinhart 1996, 2000, 2002. Ackema and Schoorlemmer 2005).However, this claim is contradicted by Polish data such as (8), in which the agent argument is overtly manifested in the form of a dative DP: (8) Te koszule uprały mi się (łatwo).these shirts-nom.plwashed me-dat się easily 'These shirts washed easily for me.' Kibort (2004: 183) argues that the dative in sentences such as (8) does indeed correspond to the agent or causer of the action expressed by the verb.The possibility of having an implied agent/causer in anti-causatives makes them similar to middles (as well as passives); the difference between them lies in the fact that middles obligatorily imply an agent, while anticausatives do not (cf.Kibort 2004: 203).Moreover, as has been stated above, middles have a generic meaning, whereas anti-causatives refer to a spontaneous occurrence or an unintentional/accidental/involuntary action (Kibort 2004: 217).Actually, Kibort (2004: 204) argues that middles have the same morpho-lexical structure as anti-causatives, and the two differ from each other only at the level of semantics, presumably in Lexical Conceptual Representations (for a similar view, cf.Ackema and Schoorlemmer 1994).In a similar vein, Schäfer (2008: 220) argues that middles of transitive verbs represent generic unaccusatives, since, as first observed by Hale and Keyser (1987), languages make use of identical morphological marking in the case of middles and unaccusatives.In Schäfer's (2008) account the middle formation is parasitic on unaccusatives, which makes it possible to derive middles in the syntax, without resorting to any lexical operations specific to middle formation (cf.Fagan 1992, Ackema and Schoorlemmer 1994, 2005, Marelj 2004).
Since our main goal in this paper is to compare the modal dać się structure with the dispositional middle in Polish, we remain agnostic as to whether the derivation of middles in Polish takes place in the lexicon or in the syntax proper.

Two types of middles from a cross-linguistic perspective
Since middles constitute a point of reference against which the properties of the modal dać się structure can be established here, and since the lassen-middle in German (cf.(4) above) bears a close resemblance to the Polish modal dać się structure, it seems worthwhile to briefly examine the properties of the middle construction itself.The cross-linguistic perspective adopted in this section is meant to bring forth the complexity surrounding this construction and the futility of any attempt at its uniform characterization.
In the literature, two types of middle have been distinguished (cf.Marelj 2004, Lekakou 2005, Ackema and Schoorlemmer 2005).Type I middles do not allow a syntactically active Agent, lack passive morphology, require adverbial modification, and show lexical restrictions on the verbs that can be found in the middle sentence.Middles of Type I are attested in English, Dutch and German.The properties of Type I middles are listed in (9) below, reproduced after Ackema and Schoorlemmer (2005: 133): a.The external argument of the non-middle counterpart of the middle verb cannot be expressed as a (9) regular DP-argument in the middle.
b.If the non-middle counterpart of the middle verb has a direct internal argument role, the subject of the middle sentence carries this role.
c.The middle verb is stative, non-episodic.The middle sentence is a generic statement.It expresses the fact that the argument mentioned in (b) has a particular individual-level property, or that events denoted by the verb or the verb-argument combination have a particular property in general.
The statements in (9a) and (9b) relate to the argument realization in the middle construction, while (9c) concerns the semantics of the middle.( 9a) and ( 9b) specify what has already been shown in section 2 in relation to Polish middles, namely the fact that the internal argument typically surfaces in the syntactic subject position in the middle construction, whereas the agent argument is not overtly present.Turning to (9c), in addition to the generic character of the middles, it also stresses their stative aspectual nature and the fact they ascribe a property to their subject.Both the syntactic and semantic properties listed in (9) for Type I middles will be tested against the modal dać się structure in section 4 below.
Type II middles, on the other hand, show the opposing characteristics; they do have a syntactically active implicit agent, the adverbial modification is not obligatory, in many languages they are morphologically identical with the passive (for instance, in Greek and Italian), they place less stringent restrictions on the verb, and can have an episodic meaning.Type II middles can be found in Greek, Italian, French, and Serbo-Croatian.Ackema and Schoorlemmer (2005: 154) observe that Type I middles are parasitic on simple active structures, whereas the occurrence of Type II middles depends on the availability of the reflexive-marked passive construction. 10Pitteroff (2014: 30) argues that the split between Type I and Type II middles, briefly characterised above from a cross-linguistic perspective, has its language internal manifestation in German.For him, German canonical middles represent Type I, whereas lassen-middles, as in (4), belong to Type II. 11In section 4.3, an attempt will be made to confront Pitteroff's (2014) claim with the Polish dispositional middle and the modal dać się structure.
The lack of uniform syntactic behaviour within the class of middles attested in various languages mentioned above seems to support the conclusion drawn by Lekakou (2005: 50) that "there is no cross-linguistically coherent syntactic sense of the 'middle'".This conclusion basically accords with Condoravdi's (1989) claim that the middle is a notional category which is associated with a particular interpretation, but not with any specific syntactic structure.

The modal dać się structure vs. the dispositional middle construction
The conclusion reached at the end of section 3 points towards the non-uniform nature of the class of dispositional middles, cross-linguistically. This, in turn, might indicate that in Polish the term 'middle' may likewise refer to a wider range of data than pointed out in section 2. The main objective of this section is to check whether the modal dać się construction can be regarded as a subtype of dispositional middles.First, in section 4.1, some similarities between the two structures under scrutiny will be mentioned, while in section 4.2 the differences between them will be analysed.Finally, in section 4.3 an attempt will be made to determine whether the differences that the two structures show exclude the possibility of subsuming the modal dać się structure under the label of dispositional middles in Polish.

Similarities between the modal dać się structure and the dispositional middle in Polish
The modal dać się structure and the Polish dispositional middle exhibit a number of similarities, which concern argument realization (cf.section 4.1.1),the implied agent (cf.section 4.1.2),dispositionality (cf.section 4.1.3),and generic interpretation (cf.section 4.1.4).

Argument realisation
The first similarity between the two structures under consideration, already apparent from the overview of the data in section 2, relates to argument realisation.As noted in section 3, middles cross-linguistically show the understood internal argument in the surface subject position.This argument realisation can be found in both Polish dispositional middles and the modal dać się structure.In both ( 10) and ( 11) below, which instantiate the dispositional middle and the middle dać się structure, respectively, the theme argument, i.e. te książki 'these books', surfaces as the subject: (10) Te książki czytają się łatwo.12these books-nom.plread-3pl się easily 'These books read easily.' (11) Te książki dają się (łatwo) czytać these books-nom.plgive-3pl się easily read-inf 'These books can be read easily.' In dispositional middles and the modal dać się structure, the understood object acts as the subject, as it bears the nominative case and triggers verbal agreement (cf.( 10) and ( 11)).

Implied agent
It has been noted in section 2 that dispositional middles have an implied agent.The question is whether the modal dać się structure is similar in this respect.In order to test this, let us make use of the test proposed by Fellbaum (1986) and used in relation to German by Pitteroff (2014: 33-4).The test involves the use of a modifier such as łatwo 'easily', whose occurrence is linked with the semantic presence or absence of an agent.In (10) and (11) above, the modifier łatwo 'easily' is found, and in both cases the meaning is the same, i.e. 'One does not have to put much effort into reading these books'. 13This, in turn, indicates that both dispositional middles, as in (10), and the modal dać się structure, as in (11), show the semantic presence of an agent.
Additionally, Marelj (2004), following Siloni (2003), argues that the instrument role is available only if the explicit or implicit subject is present.When applied to the dispositional middle and the modal dać się structure, the test, based on the instrument theta role, indicates the presence of an implied subject in both these structures, as confirmed by ( 12) and ( 13) below: (12) Ta koszula łatwo się prasuje żelazkiem na parę.this shirt-nom easily się irons iron-inst for steam 'This shirt irons easily with a steam iron.' (13) Ta koszula daje się (łatwo) prasować żelazkiem na parę.this shirt-nom gives się easily iron-inf iron-inst for steam 'This shirt can be ironed easily with a steam iron.'Both ( 12) and ( 13) are licit with the instrument phrase żelazkiem na parę 'with a steam iron', which serves as evidence for the presence of an implied subject in the two structures under scrutiny.
The question of whether the agent is syntactically present in both structures analysed here is more difficult to answer.Let us first note that both dispositional middles and the modal dać się structure can host a dative nominal, in a way similar to anti-causatives such as (8) above, as confirmed by ( 14) and (15) below: (14) Te samochody prowadzą się łatwo nawet niedoświadczonym kierowcom.these cars-nom.pldrive-3pl się easily even inexperienced drivers-dat 'These cars drive easily even for inexperienced drivers.' (15) Te samochody dają się (łatwo) prowadzić nawet niedoświadczonym kierowcom.these cars-nom.plgive-3pl się easily drive-inf even inexperienced drivers-dat 'These cars can be driven easily even by inexperienced drivers.' In both ( 14) and ( 15) the dative DP is not a beneficiary, but an agent.In the literature dative DPs like the one in ( 14) are taken to represent involuntary agents/experiencers which are The only interpretation available for (iii) is as follows: 'The door can be opened without any effort'.
(17) Jankowi te teksty łatwo się tłumaczą.(Cichosz 2014: 40) John-dat these texts-pl easily się translate-3pl 'To these texts translate easily.'Cichosz (2014) argues that the dative in ( 17) can be interpreted as an involuntary (out of control) agent of the action, as well as a beneficiary.She suggests that the dative DP is generated high, in contradistinction to beneficiary dative DPs.For her, high datives are adjuncts adjoined to TP.Her analysis predicts that high datives, being TP-adjuncts, should be possible either in the sentence initial or final position.This claim, however, is problematic in the light of the data such as (18) below, where the dative appears in the clause medial position: (18) Te samochody wszystkim prowadzą się łatwo.these cars-nom.pleveryone drive-3pl się easily 'These cars drive easily for everyone.' In (18) the dative DP wszystkim 'everyone' corresponds to the agent of driving, but it cannot be taken to represent a high dative, which makes Cichosz's (2014) account untenable.No problem of this kind arises in the high applicative approach, as a high applicative head can be merged within the VP, as originally proposed by Pylkkänen (2008).
As for the dative DP in (15), it does not seem to correspond to an involuntary agent/experiencer, but rather stands for a genuine agent which actively causes the event of driving.This is different from the corresponding middle in ( 14), and hence calls for a different syntactic treatment.The problem of the syntactic representation of an implicit agent in both of the structures under consideration is addressed in section 4.2.2.

Dispositional meaning
The dispositional meaning, which is typical of middles, has already been pointed out in section 2, and has been frequently taken to be a distinctive properly of this structure (Fagan 1992, Ackema andSchoorlemmer 2005, inter alia).To recall, the dispositional meaning involves the ascription of a property to a subject which holds by virtue of an inherent property that the subject has (Lekakou 2005, Pitteroff 2014: 42).According to Lekakou (2005) and Menéndez-Benito (2013), dispositional readings arise in the presence of a covert possibility modal that selects for a particular type of circumstantial modality (Kratzer 1991).The middle in ( 19) below has a dispositional meaning, which can be paraphrased as in ( 20):   Gehrmann (1983: 12), who notes that the dispositional meaning is enforced by the presence of an adverb, although this is just a tendency, not a regularity, as noted by von Waldenfels 2012:163).Nonetheless, (22) sounds totally degraded as a continuation of the middle structure in ( 19).This clearly shows that the middle can have just a dispositional meaning, while the modal dać się structure can convey a dispositional, as well as a permissive meaning.The ambiguous character of the modal dać sie structure has also been noted by von Waldenfels (2012: 162-164).

Generic interpretation
Middles are considered to be generic, and therefore they express regular occurrences, rather than specific events (Keyser and Roeper 1984, Condoravdi 1989, Marelj 2004, Lakakou 2005, among others).Genericity is also a property of Polish middles, as confirmed by the fact that they lack an actuality entailment, in contradistinction to episodic statements.The difference is apparent when we compare sentences ( 23) and ( 24) below (a similar test is used for lassenmiddles by Pitteroff 2014: 38): (23) Te podłogi czyszczą się łatwo, ale jeszcze nikt ich nie czyścił.these floors-nom.plclean-pl się easily but yet nobody them not cleaned 'These floors clean easily, but nobody has cleaned them yet.' (24) Ewa wyczyściła podłogę, #ale podłoga nie została wyczyszczona.
Eve cleaned floor but floor not was cleaned '#Eve cleaned the floor, but the floor hasn't been cleaned.' The continuation added after the comma in ( 24) results in the sentence being incongruous, whereas the middle construction in ( 23) can be so continued without giving rise to a contradiction.This indicates that the actuality entailment is absent in middles, in contradistinction to active sentences, which, in turn means that the former can be classed as generic statements.
The modal dać się structure patterns in the same way as middles as regards the actuality entailment, as demonstrated in ( 25): (25) Te podłogi dają się (łatwo) czyścić, ale jeszcze nikt ich nie czyścił.these floors-nom.plgive-pl się easily clean-inf but yet nobody them not cleaned 'These floors can be cleaned easily, but nobody has cleaned them yet.' The grammaticality of (25) indicates the lack of actuality entailment in the modal dać się structure, which proves its generic nature.
To sum up, the discussion in this section has shown that the middle construction is always generic.The modal dać się structure clearly contrasts with the middle in that it can also be associated with an episodic meaning.

Differences between the modal dać się structure and the dispositional middle in Polish
In this section the focus is on the most important differences between the two structures analysed here, which include aspect (section 4.2.1),availability of a syntactically represented agent (section 4.2.2),verb class restrictions (section 4.2.3), and the presence of an adverb (section 4.2.4).However, before analysing each of these differences in detail, let us note the most noticeable difference between the middle and the modal dać się structure.The difference relates to the number of verbs, i.e. the latter contains an additional predicate dać 'give', whereas the former exhibits just one predicate.The verb dać 'give', as has already been noted in section 2, takes an infinitival verb as its complement.

Aspect
In the literature, middles are regarded as stative (Keyser and Roeper 1984, Condoravdi 1989, Fagan 1992, Ackema and Schoorlemmer 2005, among others).Pitteroff (2014: 36) argues that German lassen-middles are also stative.Let us check whether Polish dispositional middles and modal dać się structures can also be treated as stative.Dispositional middles such as (10), repeated for convenience below as (29), cannot be used to answer the question in (28) below, which supports their stative nature: However, the aspectual properties of the modal dać się structure as in (11), repeated for convenience in (30) below, are different.When used in an appropriate context, such as (31) below, sentence (30) can be a possible answer to the question in (28).
(30) Te książki dają się (łatwo) czytać these books-nom.plgive-3pl się easily read-inf 'These books can be read easily.'Te książki były zaszyfrowane, ale właśnie złamaliśmy kod i teraz te książki dają się (?łatwo) czytać.( 31) 'These books were encrypted, but we have broken the code, and now the books can be read (easily).' This time, there is no difference in aspectual properties, depending on whether the adverb is present or not (although the version with the adverb is slightly degraded, as signaled by the question mark next to the adverb in (31)).Both versions in (30), with and without the adverb, can be used as an answer to the question in (28), which argues against the stative nature of the modal dać się structure in Polish.
Moreover, dispositional middles in Polish are incompatible with specific time adverbials, as can be seen in ( 32) below: 18 (32) #Te książki czytają się łatwo o czwartej rano.these books-nom.plread-3pl się easily at 4 a.m.'#These books read easily at 4 am.'However, this kind of modification is tolerated in the modal dać się structure, no matter whether the adverb is present or not.This is shown in (33) below: (33) Te książki dają się (?łatwo) czytać o czwartej rano.these books-nom.plgive-3pl się easily read-inf at 4 a.m.'These books can be read easily at 4 a.m.' In (33) the modification by the time adverbial is perfectly licit, and ( 33) is only slightly worse when the manner adverb is present.The grammaticality of (33) and the semantic illformedness of (32), once again, show that the aspectual properties of the modal dać się structure and the dispositional middle are different in Polish.Whereas the latter is always stative, the former does not have to be so.

Availability of the agent
It has been stated in section 3 that dispositional middles of Type 1, cross-linguistically, do not have any syntactically active agent.The evidence used to support this claim relies on the presence of the by-phrase, agent oriented adverbs, and control into infinitival adjunct clauses (cf.Marelj 2004, Ackema andSchoorlemmer 2005, among others).Since the test based on control does not seem to point towards the actual syntactic presence of the subject, as extensively argued by Marelj (2004: 121-124), but rather serves to prove that the implicit subject is semantically active, it is taken to be unreliable in establishing the syntactic presence of the subject.Consequently, only the first two tests mentioned above will be applied to the two Polish structures under consideration in order to look for the similarities and differences they might show as regards the syntactic presence of the agent argument. 18One of the reviewers argues that middles can sometimes co-occur with specific time adverbials, as in (i) below: (i) Te pączki sprzedają się świetnie rano, ale po południu w ogóle.these donuts-nom.plsell-pl się well morning but in afternoon at all 'These donuts sell well in the morning, but they do not sell at all in the afternoon.'On the basis of the grammaticality of (i), the reviewer argues that middles do not necessarily have a generic reading, but may be regarded as describing events.Although this might be true, we seem to entertain a different hypothesis.Following Krfika et al. (1995), we believe that generic sentences do not have to refer to timeless truths, but instead the time when a certain property holds can be restricted (cf.Greenberg (1998), who discusses generic statements temporally restricted).When viewed from this perspective, (i) is not to be treated as episodic, but is a case of temporally restricted genericity, i.e. it is a property of donuts that they sell well in the morning.
The possibility of having an agent-oriented adverb świadomie 'deliberately' in a modal structure such as (37), and its impossibility in a middle construction such as (36) indicate that an agent is syntactically present in the former, but is absent in the latter structure.
To recapitulate, the by-phrase test and the test based on agentive adverbs yield contrasting results for the dispositional middle and the modal dać się structure, which allows us to draw the conclusion that the former structure lacks a syntactically represented agent (but cf.Golendzinowska (2004: 110), who argues that Polish middles have a null agentive argument syntactically represented as pro), while the latter type of structure has an agent active in the syntax.

Verb class restrictions
In the literature, there is a lot of discussion as regards the verb classes which allow the middle formation (cf., for instance Ackema and Schoorlemmer 2005).Fagan (1992) argues that only accomplishments (e.g. read, sell, cross, break, etc.) and activities (e.g. drive, run, play, push, smoke, etc.) in the sense of Vendler (1967) are compatible with the dispositional middle, while achievements (e.g.recognize, realize, lose, find, etc.) and states (e.g.like, love, hate, possess, have, etc.) are not.
When tested against the Polish data, the claim that achievement verbs cannot participate in the middle formation turns out to be problematic, as can be seen in ( 38): (38) Te klucze gubią się łatwo.these keys-nom.pllose-pl się easily 'These keys get lost easily.'In (38) the verb gubić 'lose' appears, which belongs to the achievement class, but nonetheless it can be used in the middle sentence.However, other achievement verbs cannot be inserted in the middle structure, as can be seen in (39) below: (39) *Te błędy zauważają się łatwo.these mistakes-nom.plnotice-pl się easily 'These mistakes get noticed easily.'Moreover, stative verbs can never be found in dispositional middles in Polish, as exemplified in ( 40 In contradistinction to the Polish dispositional middle, the modal dać się structure does not seem to obey the lexical restrictions mentioned above.Both achievement and state verbs are possible in this type of structure, as can be seen in ( 41), and (42), respectively: (41) Te błędy dają się (łatwo) zauważyć.these mistakes-nom.plgive-pl się easily notice-inf 'These mistakes can be easily noticed.' (42) Warszawa da się (łatwo) lubić.Warsaw-nom.sg gives się easily like-inf 'Warsaw can be easily liked.'However, the modal dać się structure is not entirely unrestricted.Firstly, the structure under scrutiny is impossible with some stative verbs such as mieć 'have' or znać 'know', as shown in ( 43) and ( 44 Secondly, unaccusative verbs can never be found in the modal dać się structure, as confirmed by ( 45): (45) *Te kwiaty dają się (łatwo) zwiędnąć.these flowers-nom.plgive-pl się easily wither 'These flowers can wither easily.'Consequently, it seems that the modal dać się structure is subject to lexical restrictions that are different from those attested in the dispositional middles in Polish.

Adverbial modification
As has already been noted in section 2, dispositional middles in Polish and cross-linguistically require the obligatory modification by a manner adverb, such as łatwo 'easily', which specifies the way in which the eventuality expressed by the verb can be carried out. 21Other possible adverbs include trudno 'with difficulty', dobrze 'well', źle 'badly', etc.In the literature adverbs of this type are treated as those that select a covert experiencer (Roberts 1987, Lekakou 2005).The omission of manner adverbs in the middle construction results in the loss of the dispositional generic interpretation, as can be seen in ( 46) and (47) below: (46) Te koszule piorą się źle.these shirts-nom.plwash-pl się badly 'These shirts wash badly.' 21  Grimshaw and Vikner (1993) argue that the presence of obligatory adjuncts is linked with the event structure of the verb.They note that "the adverbial requirement for middles seems to be narrower than would be expected for obligatory adjuncts" (Grimshaw and Vikner 1993: 145), and they leave this problem as an open question.
(47) Te koszule piorą się.these shirts-nom.plwash-pl się 'These shirts are being washed.'Whereas (46) represents a middle sentence, whose meaning can be paraphrased as follows: by virtue of their inherent property these shirts wash badly, (47) refers only to the ongoing event, and means that these shirts are now being washed.
In the modal dać się structure the adverb is always optional, as can be seen in ( 48) below, and in all of the instances of this structure provided so far (cf.for instance (1), (11), and (15) above).

Is the modal dać się structure a subtype of dispositional middles in Polish?
Having presented the main similarities and differences that the modal dać się structure shows in comparison with the dispositional middle in Polish, we can now take stock and determine whether the former can be treated as a subtype of the latter.Out of the four points of convergence between the two structures listed in section 4.1, only the ones concerning argument realization, implied agent and dispositional meaning seem to be valid.The generic meaning, included among the similarities between the two structures, does not seem to actually represent one, as the modal dać się structure can give rise to a generic as well as an episodic meaning, in contradistinction to the middle structure which is always generic.The differences between the two structures relate to 1) aspect -stative in the case of middles, but eventive in the modal dać się structure, 2) the syntactic representation of the agent -present in the modal dać się structure, but absent in middles, 3) different verb class restrictions in the two structures analysed here, and 4) modification by the adverb -obligatory in the middle, but optional in the modal dać się structure.All the differences just listed allow us to draw the conclusion that the two structures differ considerably as regards their syntactic structure.However, as has been noted in section 3 above, following Condoravdi (1989), inter alia, middles tend to form a notional category, i.e. "a special interpretation certain syntactic configurations may give rise to, rather than a distinct grammatical construction" (Pitteroff 2014: 30).If viewed from this perspective, the modal dać się structure, which has a dispositional interpretation, can be treated as a subtype of dispositional middles.In fact, when confronted with the two types of middles, mentioned in section 3, the modal dać się structure seems to show the properties typical of Type II middles, on account of the syntactic presence of an implied agent, the optionality of the adverb, its episodic meaning, and the less stringent restrictions on the verb class than those found in Type I middles.In this respect, the modal dać się structure closely resembles lassen-middles in German, treated by Pitteroff (2014) as Type II middles.The difference that the modal dać się structure shows in comparison with Type II middles is the lack of passive morphology, which is also typical of German lassenmiddles.The modal dać się structure is also often ambiguous between the dispositional and permissive meaning, which might suggest that the two are syntactically similar (a point further developed in Bondaruk 2015)

Conclusions
The paper has aimed at presenting the syntactic and semantic properties of the modal dać się structure by comparing it with the dispositional middle in Polish.Since in the literature both dispositional middles and lassen-middles have been treated as members of the class of middles, it seems justified to examine the possibility of conflating these two types of structure in Polish under the same label .It has been argued that the modal dać się structure shows the same argument realization as the dispositional middle, i.e. the presence of the theme argument in the syntactic subject position.Moreover, both structures have the same dispositional meaning.However, it has been demonstrated that there are more differences between the two structures under scrutiny than actual points of convergence.First of all, although both can have an implied agent which is semantically active, only the modal dać się structure has an agent active in the syntax, as well.Secondly, the dispositional middle is stative, in contradistinction to the modal dać się structure, which can be eventive.Thirdly, although both structures can be generic, the modal dać się construction can also be episodic.Fourthly, a restricted set of verbs can appear in dispositional middles, while in the modal dać się structure a wider range of verbs can be found.Finally, the manner adverb modification is indispensable for the middle interpretation to be possible, while the adverb may be missing in the modal dać się construction without affecting its acceptability.Taking all these differences into account, it has been concluded that the two constructions have distinct syntactic structures.However, the dispositional meaning they share makes it possible for them to be subsumed under the label of the middle construction, understood as a notional category, along the lines first postulated by Condoravdi (1989).Within Ackema and Schoorlemmer's (2005) typology of middles, the dispositional middle in Polish belongs to Type I middles, while the modal dać się structure can be classed as Type II middles.
(i) Drzwi otwierają się łatwo.door-nom.plopen-3pl się easily 'The door opens easily.' (ii) Drzwi łatwo się otwierają.door-nom.pleasily się open-3pl 'The door opens easily.'Sentence(i) can have the following paraphrase: 'The door opens without any effort', while (ii) has a different meaning, namely: 'The door opens by itself'.However, the difference in the placement of the adverb in (10) and (11) above does not result in any difference in meaning.Cf. also (iii) below with (i) and (ii):(iii) Drzwi dają się łatwo otworzyć.door-pl give-3pl się easily open-inf 'The door can be opened easily.' nom.pl clean-pl się easily 'These floors clean easily.' by virtue of the material they are made of, the floors clean easily (20)Likewise, the corresponding modal dać się structure, provided in (21), gives rise to a dispositional reading depicted in (20) above.
) czyścić.these floors-nom.plgive-pl się easily clean-inf 'These floors can be cleaned easily.'However, in addition to the dispositional meaning, (21) can also have a permissive meaning, which surfaces in the continuation of (21), provided in (22) below: 17 (22) bo mamy dobre środki czyszczące.because we-have good agents cleaning 'because we have good detergents.' (22) sounds slightly better as a continuation of (21) if the manner adverb in (21) is missing (cf.
nom.pl read-3pl się easily 'These books read easily.' is ungrammatical, as the stative verb lubić 'like' cannot participate in the middle formation.