Agent Demotion in German and Polish

This paper provides a contrastive account of agent demotion in German and Polish. While agent demotion is a relatively broad term that is often used including mere backgrounding, the focus is on lexical and morphosyntactic means that allow for agents to be omitted entirely, such as different forms of the passive voice, reflexive constructions, generic pronouns or unaccusative verbs with impersonal subjects (among others). It is argued that passive constructions consist-ing of an auxiliary and a participle seem to exhibit much higher frequencies in German, while Polish commonly uses subjectless constructions such as agentless reflexives, certain modal elements or the -no/-to-construction. This confirms the more central position of German within the SAE Sprachbund and accounts for transfer phenomena that may arise when speakers of either language become learners of the other.


Introduction
Agent demotion has been met with considerable interest in language typology, sociolinguistics and discourse analysis. In particular, scholars working on languages for specific purposes (LSP) have studied agent demotion in great detail, as it is thought to be a typical characteristic of academic, technical or bureaucratic discourse. The option to background or omit information on a given topic is a crucial aspect of language in its socio-political context. Furthermore, typologists have shown considerable differences in how even closely related languages behave with respect to agent demotion. The following paper explores this topic by focusing on two neighbouring languages in the centre of Europe: German and Polish. Three research questions are adressed: What means do these languages employ to omit agents? In what respects are they similar, where and how do they differ? How can they be situated within the broader areal context? To answer these questions, the paper provides a thorough review of the existing literature on agent demotion in German and Polish and examines examples from written discourse. This contrastive description constitutes an important basis for a sequence of studies that will explore the acquisition of German as an academic language in the context of Polish universities.
In order to do so, it will first be necessary to outlay the theoretical groundwork. Section 2 therefore presents a definition and a short typological overview of agent demotion. Furthermore, based on a sociolinguistic and discourse-analytical approach, it is explained where and why agents are omitted in actual language use. Thereafter the respective constructions in German and Polish are described in detail, using examples from (mostly) academic writing and written media. It is important to note that the focus is descriptive rather than theoretical. It shall not be attempted to provide an in-depth analysis within the framework of a major grammatical theory. Finally the two languages are compared in order to answer the above-mentioned research questions. Conclusions shall be drawn on possible implications for learners of either language, especially with respect to an LSP context.

Agent demotion: Typological and sociolinguistic perspectives
Any theoretical discussion of the concept of agent demotion should begin by clarifying what the term agent actually means. In his paper Agency in Language, Duranti (2005: 453) provides a working definition: Agency is here understood as the property of those entities (i) that have some degree of control over their own behavior, (ii) whose actions in the world affect other entities' (and sometimes their own), and (iii) whose actions are the object of evaluation (e. g. in terms of their responsibility for a given outcome). (Duranti 2005: 453) Dowty (1991: 571f.), building on prototype theory and the concept of fuzzy categories (Rosch 1973 i. a.), draws attention to the fact that thematic roles such as agent and patient are not "discrete categories at all, but rather are cluster concepts" and gives the following list of contributing properties for the agent Proto-Role: Prototypical agents show all or at least a majority of these properties, while more peripheral ones might exhibit only one or two. The advantage of such a taxonomy is its ability to encompass a large number of situations along the continuum between prototypical agents and patients (van Valin 2005: 58) without a necessary case-by-case categorization. As illustrated by the above examples, definitions (or approximate definitions) of the term agent generally involve concepts like volitionality, control and effect on other entities. Thus, the subject the lady in (1a) and (1b) can be considered to be a prototypical agent, with (1c) being less prototypical and (1d) lacking agency entirely.
(1) a. The lady beats the driver. b. The lady is eating chicken soup. c. The lady is meditating. d. The lady is dying.
Having established what agents are, what does agent demotion mean? Definitions range from the de-emphasis of an agent by placing it in an oblique (or non-subject) syntactic position up to the option of omitting it entirely (cf. Solstad/Lyngfelt 2006: 8f. andSiewierska 2008: 121). The focus of this paper is on morphosyntactic structures and lexical items which do the latter.
There is a wide variety of constructions available to background or omit the agent -both crosslinguistically and within individual languages. One generally accepted approach or definition is, as has been hinted at, nonexistent (cf. Malchukov/Siewierska 2011). This is reflected by the variety of terminology in the literature: Scholars have used terms such as "impersonalization" (e. g. Malchukov/Siewierska 2011), "agent demotion" (e. g. , "subject demotion" (e. g. Blevins 2003), "agent defocusing" (e. g. Myhill 1997), "agent backgrounding" (e. g. Fried 2006), "deagentivization" (e. g. von Polenz 198,very common in German linguistics) or "mitigation of agency" (Duranti 2005), to name but a few. While there is some overlap, they stem from different traditions and are not synonymous. As Malchukov/ Siewierska (2011: 1) rightly point out, there have been "difficulties in identifying impersonal constructions on a cross-linguistic basis". Siewierska (2008: 116-125) provides a useful distinction between two main approaches to what she calls "impersonalization": a formal one and a functional one. The former focuses on the nature of the subject and refers to constructions with non-referential subjects or without any subject whatsoever, such as generic pronouns (2), generic reflexives (3)  The latter approach, on the other hand, is centred on the agent or instigator (which is Siewierska's preferred term) and additionally grasps constructions with a referential subject which is not identical with the instigator, such as regular passives (5), inchoative/unaccusative verbs (6) or verbal nouns (7), among others.
(5) All options were tried. For the purpose of this paper, agent demotion shall be defined quite losely, following the second -functional -approach to what Siewierska calls impersonalisation: the omission of referential agents from the surface of text and discourse leaving their identity entirely implicit. This implicitness of omitted participants can be explained using frame semantics, as introduced by Charles Fillmore (1976 i. a.). Lexical items, but also more complex syntactic entities activate so-called frames, in which speakers' world knowledge about the situation, its participants and their inter-relations is organised. For instance, the noun migration evokes a frame that includes not only the migrating individuals, but also countries of provenance as well as destinations. Thus, in a sentence like Migration is among the most significant issues of our times in which none of these elements is mentioned, their existence can be inferred. The same holds for all of the constructions mentioned above. For this reason, such terms as impersonal or impersonalisation, which are often used to describe agentless structures (even when they clearly imply agents that can only be human beings) shall be avoided in this paper. Instead, the term indeterminate-personal is used, since the agentless constructions that are discussed here are in fact very much personal, merely leaving the identity of the agent indeterminate (cf. Helbig/ Buscha 2004: 150 for similar thoughts on the German passive).
In discourse analysis, which often examines the representation of social actors in texts, it has long been established that agent demotion can be used for what may be called their "suppression" (cf. van Leeuwen 2008: 29-31) 2 . Leaving the identity of actors unmentioned or ISSN 1615-3014 95 implicit can be due to alleged irrelevance or lack of knowledge on the speaker's part (cf. Jespersen 1924: 167f.), but very often it is used to obscure (social) responsibilities (cf. van Dijk 2008: 166). In political discourse as well as in journalism, this option is a very powerful means to (mis)represent events in accordance with one's own ideology (cf. Fairclough 1992: 181 andPaltridge 2012: 29-33). Moreover, agentless constructions have often been associated with LSP and particularly with formal, written registers such as bureaucratic, legal or academic style because of a generally accepted demand for objectivity and anonymity (cf. Oksaar 1998: 397). It is thus unsurprising that passive constructions have been shown to be less common in child than in adult speech (cf. Jisa et al. 2002: 175).
The passive voice is arguably among the most famous and most commonly discussed forms of agent demotion -possibly due to its wide spread in large Western European languages. The typical SAE (Standard Average European) construction, which is most common in central Europe, consists of an auxiliary and a resultative participle 3 (cf. In many European languages -particularly at the fringe of the SAE Sprachbund -such as Spanish, Portuguese, Polish or Slovenian (cf. Rivero 2002;Jisa et al. 2002 andRivero/Sheppard 2003) the auxiliary-participle passive is rivaled by another form of agent demotion: indeterminate-personal reflexives, sometimes referred to as reflexive passives or passive reflexives (cf. Lekakou 2006: 172 andCennamo 2011: 182). In these constructions, verbs inflected for the third person singular are combined with reflexive morphosyntax. This can be seen in example (3) for Spanish and (11) for Slovenian: 3 The term resultative is chosen here in accordance with Nedjalkov/Jaxontov (1988: 6) who use it for "verb forms that express a state implying a previous event" and Haspelmath (1994: 159) who defines resultative participles as "expressing a state that results from a previous event". Traditional terms such as past participle or passive participle shall be avoided since these verb forms neither always express past tense nor passive voice in the respective languages. 4 For the sources of examples (9)-(69), see references.
ISSN 1615-3014 96 (11) Slovenian Verjame se, da je onesnaženost zraka kriva […] believe.3SG.PRES itself that it is air pollution guilty 'It is believed that air pollution is responsible.' Of course, these are only two frequent examples of constructions used to omit agents, alongside which each language exhibits a number of others. In the following it will be shown which lexical and morphosyntactic means German and Polish -two Indo-European languages in the centre of Europe -employ to leave agents indeterminate. 5

Agent demotion in German
This outline shall mostly concentrate on finite structures. As has been explained above, the broad definition of agent demotion that is used here also includes structures such as nominalisation. These more peripheral ways of demoting the agent shall only briefly be touched upon. At first, the focus shall be on the verb phrase and different mechanisms of voice and valency reduction. Subsequently, it will be shown which structures outside the verb phrase are relevant to agent demotion.

Passive constructions
German has a number of constructions that are commonly identified under the label Passiv, all of which follow the aforementioned SAE pattern auxiliary + resultative participle. The one that has traditionally received most attention is the so-called Vorgangspassiv 'procedural passive' formed with werden 'to become'. As its name suggests, it presents actions or processes as ongoing (cf. Helbig/Buscha 2004: 144-149), as shown by the following example: (12) German Seit 2015 wurde im Bezirk keine einzige Sozialwohnung gebaut since 2015 became in the district not a single social housing unit built 'Since 2015 not a single social housing unit has been built in the district.' Lasch (2016: 124-126) connects the procedural nature of this construction to the lexical meaning of werden, including it among what he calls constructions of commutation.
The statal passive (Zustandspassiv) is formed with the verb sein 'to be'. Unlike the procedural passive, it has a resultative meaning and refers to the state after the completion of a process (cf. Zifonun et al. 1997: 1810f. and Helbig/Buscha 2004): (13) German (Auer 1998: 183) Der Vorlesetext ist als Anhang beigefügt. the reading text is as an appendix attached 'The reading text is attached as an appendix.' Together with similar constructions using the verbs bleiben 'to remain' (14), scheinen 'to seem' (15) and aussehen 'to appear' (among others), Lasch (2016: 122-124) includes the statal sein passive in a group for which he uses the term ascription, again pointing to the fact that the ISSN 1615-3014 97 meaning of the respective passive constructions is derived from the lexical meaning of the verb as a copula with adjectives or nouns. (15) German Seine Karriere schien beendet. his career seemed finished 'His career seemed to be over.' These regular passive constructions can also refer to actions caused by non-human entities (for instance animals). However, this is not always the case. With verbs referring to volitional actions by human agents, both the procedural and the statal passive can be used without a subject, resulting in an indeterminate-personal construction (cf. Helbig/Buscha 2004: 150-152 andEngel et al. 1999: 653). It is worth noting that in this case the werden passive may even be formed from intransitive verbs (cf. Zifonun et al. 1997Zifonun et al. : 1792Zifonun et al. -1796. In many European languages such as English there is no indeterminate-personal passive, as can be seen in the translations below. Verbs that take an indirect object in the dative can also be passivised with the auxiliaries bekommen/kriegen 'to receive, to get' (cf. Helbig/Buscha 2004: 167f. andLasch 2016: 79-84 Similarly to the sein passive, the second sentence expresses an after-state, referring to the woman's hair after the dyeing process. The last construction with the resultative participle that deserves to be mentioned here is the one with gehören 'to belong'. It is semantically close to the werden passive due to its procedural nature, but combines it with deontic modality (cf. Lasch 2016: 84-87). This can be explained by its relation to its relation to the modal directional usage of gehören in such sentences as Das gehört in den Müll 'This belongs in the trash' (cf. ibd.: 434-454).

Modal indeterminate-personal constructions
German also possesses a number of indeterminate-personal constructions that -similar to the aforementioned gehören passive -have a modal meaning component. One of them is the typically Germanic zu ‚to' + infinitive (von Polenz 1985: 185 andAbraham 2006: 17-19 The next group of modal indeterminate-personal constructions that deserve mentioning are those involving the verb lassen 'to let', which, as Lasch (2016: 88) puts it, "drives tears into every grammarian's eyes" due to the remarkable dynamics it shows in language use. An indepth analysis shall therefore not be attempted here. Lassen can be used reflexively with personal subjects and an infinitive, expressing a deontic or circumstantial flavour (cf. Madbouly Selmy 1993: 109-112)  Reflexive lassen is also frequently construed without any subject whatsoever, or simply es 'it' as a dummy subject in a way that is not unlike subjectless passive constructions (cf. Zifonun et al. 1997Zifonun et al. : 1855Zifonun et al. -1857

Reflexive and unaccusative constructions
Modal lassen + infinitive is not the only instance in which reflexiveness plays a major role. Agentless reflexives -sometimes identified with the term Medium 'middle voice' (cf. Hentschel/Weydt 2013: 114) -are another relevant construction that needs to be mentioned here. Helbig/Buscha (2004: 164) note that they demote the agent even further than the passive voice, presenting situations as accidental and devoid of agency. It occurs both with impersonal subjects and dummy es (which appears as a clitic 's in gg): (33) German (Auer 1998: 191) Bei ihnen finden sich in der Spontansprache für /sp/ 93 Belege […] with these find REFL.3PS in spontaneous speech for /sp/ 93 instances […] 'Regarding these, there are 93 instances of /sp/ to be found in spontaneous speech.' (34) German Im Herzen Villingens wohnt sich's besonders teuer in the heart of Villingen lives REFL.3PS-it particularly expensive 'In the heart of Villingen it's particularly expensive to live.' It has been pointed out that these indeterminate-personal reflexives are not particularly frequent in German (cf. Hentschel/Weydt 2013: 123).
In addition to voice and valency operations, it is also possible to omit agents by using unaccusative verb phrases (cf. Madbouly Selmy 1993: 67f.) including phrasal verbs (cf. Helbig/ Buscha 2004: 163f.). As the following examples show, even if a human agent is not overtly mentioned, the verbs auffallen and in Vergessenheit geraten always evoke frames in which such participants are present, even if they cannot appear in the subject position. In this context, so-called labile verbs are also worth mentioning. These verbs, which are common in English and somewhat less so in German, can both be used unaccusatively or transi-tively with an agent subject (cf. Haspelmath 1993: 92). Verbs in this group include kochen 'to boil' (37), brechen 'to break' and schmelzen 'to melt', among others.

(37) German
Ich koche die Milch. Die Milch kocht. I boil the milk the milk boils 'I boil the milk. The milk boils.' Unaccusative constructions also include copula verbs such as sein, bleiben and werden. Combined with predicative adjectives that evoke a frame containing an agent -such as deverbal adjectives (cf. Helbig/Buscha 2004: 165f.) -they can be used to omit this agent. Like the aforementioned passives with participles, the construction grammarian Lasch (2016: 123) counts examples such as the following as constructions of ascription and commutation.

(38) German
Sein Tod ist bedauerlich, war aber absehbar. his death is regrettable was but foreseeable 'His death is unfortunate, but was to be expected.'

Non-prototypical subjects and other forms of agent demotion
The previously mentioned constructions involve verb phrases that can not have a prototypical agent subject. As has been indicated in the preceeding chapter, demotion is also possible with agentive non-reflexive verb phrases. One way of doing this is using non-referential or generic pronouns such as man 'one' (42) or jemand 'someone' (cf. Madbouly Selmy 1993: 42-44 andHelbig/Buscha 2004: 231-233). Es 'it' as a dummy subject is particularly common with meteorological or natural processes that are impersonal by definition, but can sometimes demote agents (cf. ibd.: 5). Moreover, it has also been noted that personal pronouns can be used generically (cf. Madbouly Selmy 1993: 51f.).

adds up one but all instances for the adult speakers […] 'If you add up all instances of the adult speakers…'
Apart from these indeterminate-personal pronouns, an agent subject position can also be filled with impersonal nouns (cf. Madbouly Selmy 1993: 72-75), in an often metaphorical or metonymical way, as the famous gun advocacy slogan in (43) shows. A particularly common type of this impersonalisation (the term impersonalisation truly fits here!) is what has been called Subjektschub 'subject shift' in German linguistics (von Polenz 1985: 187), referring to the use of nouns such as text, article, theory etc. instead of personal nouns like author. This is exemplified in (44). Moreover, nominalisations (cf. Bzdęga 1980: 39 andHelbig/Buscha 2004: 260), infinitives (cf. Madbouly Selmy 1993: 106-108) or subordinate clauses (cf. Helbig/ Buscha 2004: 590f.) are also sometimes used similarly. In both of the examples below, the subjects in bold print do not refer to the actual agent, but to an instrument that is used by them to accomplish the action. As has been briefly touched upon above, derivation of nouns (and similarly adjectives or participles) can also serve to omit agents. Present participles with zu (45), which are related to modal zu-infinitives are a particularly interesting case of this. They are sometimes called gerundives (Helbig/Buscha 2004: 165 andHentschel/Weydt 2013: 132)  As mentioned in section 2, agent demotion is generally associated with standard or formal language. In German, this has been often been argued to be true for passive constructions with werden (cf. Roelcke 2003: 61), which are thought to be particularly common in academic discourse or LSP (cf. Helbig/Buscha 1996: 147), while being avoided in so-called Leichte Sprache 'simple language', a variety developed for accessibility (cf. Baumert 2016: 193). It is important to note that the periphrastic werden passive with its opposition to the statal sein passive is not particularly old: Arguably under Latin influence, it only became fully grammaticalised in Modern High German and has been gradually increasing in frequency starting from the 16th century (cf. Eichinger 1987: 136;Roelcke 2003: 42 andRoelcke 2009: 41), possibly paralleling the evolution of the text genres it is commonly connected with. The surge of rationalism and objectivity in the enlightenment era may have contributed to this development (cf. Eßer 1997: 94). All in all it has become clear that the German language exhibits a considerable variety of tools for agent demotion. This variety is perhaps wider and more complex than many reference grammars suggest.

Agent demotion in Polish
Like in other Slavic languages, sentences that do not have a subject are relatively common in Polish (cf. Kibort 2011: 357 andPanzer 1999: 236-238). Guścin (1973: 48) defines subjectless sentences as follows: Można (…) dla języka polskiego zdefinować zdanie bezpodmiotowe jako takie, w którym nie występuje rzeczownik (lub inne części mowy w jego funkcji) w nominatiwie. 'For the Polish language subjectless sentences can be defined as ones in which no nominative noun (or another part of speech with such a function) is present.' (1973: 48) Historically, the concept of subjectless sentences has often been rejected in favour of analysing the respective structures as elliptical (cf. Galkina-Fedoruk 1958: 41-50 i. a.), possibly because it contradicts traditions in western logics (cf. Guścin 1973: 48). Avoiding an in-depth explanation of the topic due to lack of space, this paper shall adopt the notion of subjectlessness as it can elegantly account for a number of constructions in the Polish language which play an important role in agent demotion. Similarly to the previous chapter on German, the discussion shall start with non-agentive verb phrases and then move on to structures beyond the verb phrase level.

Passive and related constructions
The typical SAE passive construction with an auxiliary and a resultative participle also exists in Polish (cf. Bartnicka et al. 2004: 356-357). It is important to note that the Slavic aspectual opposition between perfective and imperfective verbs interacts with its formation. While perfective verbs and their participles refer to completed processes, imperfect ones and their participles present a process as ongoing (cf.

(47) Polish
Elektromobilność jest opłacalna tylko wtedy, kiedy jest dotowana electromobility is profitable only then when is subsidised 'Electromobility is only profitable when it is subsidised.' The formation of the passive voice is further complicated by the fact that zostać is itself a perfective verb and thus only has past and future forms. In the present tense, perfective verbs therefore need to be combined with its imperfective counterpart zostawać (cf. Engel et al. 1999: 649-650). Apart from this procedural passive, Polish also possesses a statal alternative that presents processes as already having been completed. It is formed with być and the resultative participle of perfective verbs: Wypadek pod Głogowem. Droga była zablokowana accident near Głogów road was blocked 'Accident near Głogów. Road was blocked.' This kind of passivisation is restricted to transitive verbs with an accusative complement as well verbs with a complement in the instrumental case. Reflexive verbs cannot be passivised (cf. Engel et al. 1999: 649-651 and 658f.). A subjectless passive with an indeterminate-personal reading can be formed with być and imperfective verbs that are unergative -particularly household activities that obviously evoke frames including a human agent (cf. Kibort 2011: 357-393) -such as in the following example: (49) Polish Wygląda na to, że dawno tam nie było sprzątane. looks at this that long time there not was cleaned 'It looks like no cleaning has happened there in a long time.' Apart from być and zostać and their imperfective forms, it has been pointed out that stać się (perf.)/stawać się (imperf.) 'to become' is a third copula verb that may occur with participles (cf. Szupryczyńska 1980: 35-38). However, this usage is hardly mentioned in present reference grammars and does not seem to play a major role in agent demotion. Moreover, mieć 'to have' is also sometimes brought up as a possible recipient passive auxiliary in sentences like the following (cf. Korytkowska 1993: 172): (50) Polish Trenerka miała zrobioną fryzurę na "mokrą włoszkę". coach had done haircut at "wet look" 'The coach had her hair done in the wet look.' Regardless of whether such a categorisation is reasonable or not, this construction definitely serves to omit agents.

Indeterminate-personal constructions
Apart from these typical SAE-style passives, Polish resultative participles appear in an additional construction of agent demotion: The famous -no/-to, which is formed by adding the adverbial suffix -o to the participle (cf. Doros 1964: 107-109 andSadowska 2012: 421, 427). Always referring to past actions, its meaning is indeterminate-personal and unlike the aforementioned passives it can be formed from unaccusative verbs as well (cf. Bzdęga 1980: 49 andKibort 2008: 263-265 While there is no overt subject in this construction, it actually shows some syntactic behaviour of sentences that do contain subjects (cf. Kibort 2001: 266).
Furthermore, there is a number of indeterminate-personal constructions that do not involve resultative participles, but verbal nouns, infinitives or derived adjectives. Among these are unaccusative phrasal verb constructions with verbal nouns such as znajdować zastosowanie 'to find use' (cf. Engel et al. 1999: 659): (53) Polish (Engel et al. 1999: 659) Nowe środki znajdują zastosowanie new means find use 'New means find use.' Verbal nouns can also be used with do 'to' to express circumstantial modality (cf. Engel et al. 1999: 661f. andBzdęga 1980: 51): Jest do kupienia Bugatti Veyron is to buying Bugatti Veyorn 'A Bugatti Veyron is for sale.' A similar circumstantial flavour is expressed by the modal existential construction, which Polish shares with many other Eastern and Southern European languages (cf. Šimík 2011: 32). In this construction, an existential predicate is combined with an interrogative element and infinitive: (55) Polish Nie było co zbierać. not was what to gather 'There was nothing to gather.' Polish also possesses deverbal adjectives that express circumstantial modality and can be used with the copula być in the following way (cf. Bzdęga 1980: 52): Ich cel jest zrozumiały […] their goal is understandable […] 'Their goal is understandable.' A further mechanism that plays a major role in agent demotion is reflexivisation. Polish often employs reflexive się to create the inchoative counterpart to a causative verb, presenting actions as haphazard and devoid of agentivity (cf. Mecner 1992: 30-34;Engel et al 1999: 662 andBerdychowska 2013: 71). For instance, in the following example the transitive verb zamykać 'to close' is reflexived to omit the identity of who closes the bank branches: (57) Polish Mamy coraz mniej bankomatów, zamykają się też oddziały. we have fewer and fewer ATMs close REFL also branches 'We have fewer and fewer ATMs and branches are also closing.' As suggested in the introduction, Polish is among the languages at the fringe of the SAE Sprachbund in which indeterminate-personal reflexives are very common. They are formed with się and a verb inflected for third person singular neuter (cf. Kibort 2008: 263 andSadowska 2012: 429) and are subjectless as patients remain in the accusative case (cf. Prenner 2019: 343). The following three examples show this. Interestingly, indeterminate-personal reflexives can even be formed from verbs that are reflexive themselves (cf. Patejuk/Przepiórkowski 2015: 274), as in (60). It is often pointed out that the meaning is equivalent to to that of the -no/-to construction (cf. Kibort 2008: 270-274  One specific instance of this construction that deserves mentioning here is the modal da się + infinitive 'it is possible to ' (cf. Bzdęga 1980: 51). The verb dać literally means 'to give', but functions like 'to le' when used with an infinitive. In a corpus study, Sansò (2006: 238-245) distinguishes three stituation types that are relevant for agent demotion: patient-oriented processes, bare happening and agentless generic events. Patient-oriented processes "represent […] a two-participant event from the point of view of the patient" (ibd.: 238), bare happening conceptualises the "event depicted by the verb as a naked fact, at the lowest level of elaboration" (ibid.: 240) without focusing either of its participants and in agentless generic events "an agent, usually human, is understood to exist, but is defocused because of its genericity". As he shows in his study, whereas the passive voice seems to be the default with patient-oriented processes, -no/-to coincides with bare happening and indeterminate-personal reflexive are the most common form with agentless generic events (ibd.: 255).
Not only does Polish have sentences without a subject: Like other Slavic languages, it can also form sentences without a finite verb. One example of this are so-called defective modals such as można 'it is possible' (62), warto 'it is worthwile', wolno 'it is allowed' or trzeba 'it is necessary ', among others (cf. Sadowska 2012: 445). They are constructed with infinitives and have indeterminate-personal semantics. A copula is needed in the past and future tenses, as shown by (63): (62) Polish (Nowak 2016 : 339) […] można wyróżnić trzy grupy argumentów […] it is possible to distinguish three groups of arguments 'It's possible to dstinguish three groups of arguments.' (63) Polish Nie wolno będzie przewozić innych osób […] not allowed will be to transport other people […] 'It will not be allowed to transport other people.' Predicative adjectives/adverbs or nouns are used with infinitives in a similar way (cf. Bzdęga 1980: 40f.). Examples are trudno 'difficult' (64) czas 'time' (65) and nie sposób 'no way' (58). The indeterminate-personal reflexive construction is uncommon with the sensory verbs widzieć 'to see', słyszeć 'to hear' and czuć 'to feel'. Instead, Polish uses the defective infinitives widać, słychać and czuć in sentences without a finite verb, expressing the notion that something is to be seen, heard or felt (cf. Bartnicka et al. 2004: 514-515 andSadowska 2012: 369). In a parallel to the indeterminate-personal adverbs and nouns mentioned above, the copula być needs to be added in the past and future tenses, as shown by second of the following two examples: Finally, before turning to forms of agent demotion outside the verb phrase, it is necessary to mention that personally inflected verbs can be used in an indeterminate-personal way (cf. Panzer 1999: 236-238). The third person plural and the second person singular are particularly common (cf. Berdychowska 2013: 91).

Non-prototypical subjects and other forms of agent demotion
There are a number of ways in which agents can be omitted by occupying the subject position in active sentences in non-prototypical ways. One of these options is the use of indefinite pronouns (cf. Sadowska 2012: 47) or simply the noun czlowiek 'human' (cf. Berdychowska 2013: 91). The metonymical use of impersonal nouns as agents is rarely mentioned in the literature, but seems to occur as well. This is shown by the following two examples in which an article stands for its author and cars for the people who drive them: Apart from all of these forms of agent demotion at sentence level, Polish can also demote agents through derived nouns and adjectives, including the resultative participles mentioned above (cf. Bzdęga 1980: 52 andSadowska 2012: 421).
Indeterminate-personal style and particularly the passive voice has generally been linked to literary or academic prose (cf. Gajda 2001: 189;Kita 2013: 185 andBądziński 2019: 389f.) while guidelines for accessibility-focussed plain language generally recommend to avoid it (cf. Piekot/Zarzeczny/Moroń 2019: 202). However, subjectless constructions other than the passive and -no/-to also seem to be frequent in colloquial varieties (cf. Konopka 2003: 681). Historically, these two structures are relatively young: The past indeterminate-personal -no/-to developed no earlier than in the 17th century (cf. Klemensiewicz 1976: 432). The passive voice is reported to be relatively rare in Middle Polish texts (until the outgoing 18th century), with third person plural and indeterminate-personal reflexives ocurring much more frequently (cf. ibd.: 422f.). In fact, the prevalence of such forms is still reported for contemporary Polish (cf. Bzdęga 1980: 49f.). All in all, it has become clear that just like German, Polish exhibits a great ISSN 1615-3014 109 variety of constructions for demoting the agent. The next chapter will point out common patterns and differences.

Comparison and conclusions
To some extent, German and Polish exhibit very similar constructions of agent demotion. This comes as no surprise since they are not only genetically related, but have been areally close for a very long time. Both languages have typically Indo-European resultative participles which are combined with auxiliaries to form passive constructions. In both languages unaccusativity and reflexiveness play some role, as do indeterminate-personal or impersonal NPs as agent subjects. However, Polish and German are seemingly different in the way these different constructions are distributed.
Whereas in German, the passive voice seems to be central to agent demotion, in Polish indeterminate-personal reflexives and other non-passive constructions are much more grammaticalised and seem to dominate not only in spoken language (cf. Bzdęga 1980: 48), but in fact even in the media and in scholarly discourse (cf. Koneczna 1958: 91 andSadowska 2012: 427). The lower prominence of the passive voice comes as no surprise since it grammaticalised much later than in German, likely boosted by increased cultural and political influence from German-speaking Prussia and Austria starting from the late 18th century (cf. Weiss 1982). Similarly, do 'to' + verbal noun has been pointed out to result from German influence (cf. Weiss 1987 and1988). It is also worthwile to note that German -like Dutchemploys the passive voice to encode both what Sansò (2006: 266) calls patient-oriented processes and bare happening, while in Polish the latter overwhelmingly correlates with a different construction, namely -no/-to. This underscores that the German passive might be further down the grammaticalisation path than the Polish one, being used in a larger range of contexts. Moreover, in German the options of forming sentences without a subject are heavily restricted, whereas Polish exhibits a considerable variety of subjectless constructions. Not being a pro-drop language, the former often relies on es 'it' as a dummy subject (for instance in reflexive lassen constructions), unlike the latter (cf. Bzdęga 1980: 40-43) (for instance in comparable da się + infinitive). However, Konopka (2003: 681) points out that there are in fact dialects that do use ono 'it' in a similar way. German also exhibits forms of agent demotion that are completely unparalleled in Polish, most notably zu + infinitive and labile verbs. indeterminate-personal reflexives (strongly grammaticalised)

Agent demotion below and beyond the verb phrase
indeterminate-personal and impersonal agent subjects derived (verbal) nouns and adjectives generic use of pronouns/ personal verb forms All in all, it can be concluded that while both languages have a participial passive, in German it has been argued to be more frequent and grammaticalised than in Polish, which relies more on other constructions. Thus, simple areal feature maps as in Haspelmath (2001Haspelmath ( : 1496, which divide languages in a binary manner and ignore differing levels of frequency and grammaticalisation sometimes fail to provide an accurate image of a Sprachbund. The comparative account given in this paper confirms the generally assumed core position of German within SAE, with Polish being located at the periphery (cf. Haspelmath 2001Haspelmath : 1502. For future research, it would be interesting to compare the distribution of the different constructions in different registers, using a quantitative approach. How does this complex picture of similar and dissimlar structures affect Polish learners of German? Bzdęga (1980: 53) concludes at the end of his paper that certain difficulties resulting from the apparent differences are to be expected. Indeed, diverging frequencies in the respective languages might result in Polish speakers overproducing reflexive constructions -or other formally or functionally similar constructions -in German. 7 Moreover, native Speakers of Polish might sometimes struggle to distinguish between the statal and procedural passives in German. More general issues that affect many of the German constructions might be the correct usage ISSN 1615-3014 111 of dummy es and the so-called Satzklammer 'sentence bracket', with finite verbs occupying the second position in a main clause and infinite components such as participles or infinitives coming last. In standard Polish there are no equivalents to these phenomena. Finally, those constructions that are unparalleled in Polish, such as zu + infinitive in German, might also be underproduced. A number of corpus studies shall explore academic texts in German language from both Poland and German-speaking countries to show how the vast number of constructions that have been described in this paper are used in actual L1 and L2 writing. It will then be possible to state whether the assumptions above prove to be true.