Agentivity in Human Impersonal Constructions in Polish and Russian

1.1 A (very) short introduction and information about the structure of this book

This book presents a synchronic study on the applicability and acceptability of four different impersonal constructions in Polish (-no/-to, reflexive, and 3pl construction) and Russian (only 3pl construction). The relevant constructions show different degrees of applicability with different verbs independent from the animacy of the demoted subject. The main idea of this book is that the grammaticality of these constructions depends on the degree of agentivity of the implicit subject entailed by the verb. For the description of the verb-semantics, we use Dowty’s proto-role model (1991), which lists five entailments for the proto-agent role: [volition], [sentience], [movement], [causation], and [independent existence]. For the other proto-role, the proto-patient, Dowty also lists five entailments. This model operates by summing up the above-mentioned entailments to identify how much ‘agent-like’ or ‘patient-like’ a semantic role is. Having equipped the reader with a literally more than rough guide, in what follows, I will shortly discuss the structure of this book:

In Chapter 2 I will give a detailed description of morphological, semantic, and stylistic differences of all four relevant constructions. Chapter 3 first gives an overview over thematic roles and thematic role approaches, and in particular focusses on two thematic role models: (i) Dowty’s model (1991) and (ii) Primus’s model (1999a). Chapter 4 offers a detailed description of Dowty’s framework of verbal entailments (in our terminology ‘agentivity features’), which will serve as working basis in the empirical part of this book. In Chapter 5 a corpus study for the Polish -no/-to and the reflexive construction will be presented with the aim to explore the behavior of agentivity features under the circumstances of the mentioned constructions and the distinction between transitive and intransitive verbs. Chapter 6 is devoted to a thorough exploration of the feature [sentience] by means of four acceptability judgment studies involving the personal passive, the -no/-to, the reflexive, and the 3pl construction in Polish. Moreover, it shall be answered whether the three constructions fulfill the same discourse functions by assuming that similar results in the acceptability judgment studies are evidence for it. By means of two acceptability judgment studies – one for the Polish ←15 | 16→reflexive construction and one for the Russian 3pl construction – Chapter 7 is concerned with the exploration of the feature [movement].

←16 | 17→As already commented, the common denominator of the four mentioned constructions is that they all demote a human subject. Human subjects rank high on the animacy scale, which is why the demotion of human subjects typically implies special morpho-syntactic devices, and the relevant constructions are such a device. Unlike Russian, and as illustrated in the examples from the NKJP, Polish seems to have three quasi-synonymous constructions (examples 1−3 above). The nature of these constructions has been the topic of several linguistic studies. In many of the studies about the relevant constructions in Polish, the authors have focused on a certain type of construction, for example, the reflexive construction (Krzek 2010; Meyer 2010), 3pl construction (Siewierska & Papastathi 2011), no/-to construction (Puzynina 1993), etc. This book addresses, inter alia, the question of how the relevant constructions differ morphologically and semantically, which is why it is the first empirically based study to include the Polish -no/-to, the reflexive, and the 3pl construction, and for Russian the 3pl construction.

The common term for the constructions under investigation in Slavic linguistics is indeterminate-personal constructions (from Russian неоределённо-личные предложения). This is not entirely accurate since these constructions are clearly impersonal in the sense of Siewierska (2008a: 116) adopted here: They all lack a canonical subject, that is, a subject in the nominative case. Along with this term, a variety of other terms have been proposed by linguists for this type of construction. One of them is the term HIPs (human impersonal pronouns) which was proposed by Gast and van der Auwera (2013). It includes all kinds of impersonalization strategies where a pronoun with human reference is involved. One strategy is the use of impersonal pronouns which are widespread in (some) Indo-European languages, for example, in German man or in French on. In comparison, Slavic languages use different strategies for this kind of expression, for example, čovek (Bulg.) or ljudi (Russ.), but they are less widely distributed (Gast & van der Auwera 2013: 2). One of the most remarkable features of HIPs is that they do not introduce discourse referents, which is regarded as a defining property. According to Gast and van der Auwera, HIPs introduce a “variable ranging over human entities without referring to any human referent in particular” (ibid: 5−6). However, as noted by Bunčić (2019), the category of HIPs is a rather broad category and includes such generic (or in terms of Cabredo Hofherr (2003) “universal”) readings like the 2sg in Russian:

Another term used in the literature is R-impersonals (e.g., Siewierska 2011; Malchukov & Ogawa 2011), which simply describes the fact that there is no referential subject to locate (‘R’ means ‘reduced in referentiality’). Siewierska defines R-impersonals as follows:

R-impersonals have the appearance of regular, personal constructions but feature a subject which is human and non-referential. The non-referential human subject may be expressed lexically, pronominally or by the whole construction. The subject of lexical R-impersonals is typically the word for ‘person’ or ‘people’ […] (Siewierska 2011: 57−58).

The fact that there is no referential subject is in sharp contrast to personal constructions, where grammatical subjects are normally referential, that is, they have absolute reference (Keenan 1987: 102). However, throughout this book, I will use the term arbs, which is short form for “constructions with arbitrary ←18 | 19→interpretations”, as proposed by Malamud (2013). I consider it the most suitable term for these constructions, as it includes all the above-mentioned constructions under investigation: the -no/-to construction, the reflexive impersonal construction, and the 3pl impersonal in both languages. While arb constructions can have generic readings, they must be able to provide one of the following four readings as summarized by Cabredo Hofherr (2003): (i): specific existential reading, (ii) vague existential reading, (iii) inferred existential reading, and (iv) corporate reading. In contrast to that, impersonal constructions like the 2sg construction cannot provide any of these readings, and for this reason they are not arbs.

For the title of this book, I have chosen the term ‘Human Impersonal Constructions’ for two reasons. First, it is closer to the more common term ‘impersonal constructions’, and the addition of ‘human’ clearly signals the involvement of such an underlying agent. Second, while the specified term ‘arbs’ is most appropriate, it is not commonly used. To provide interested readers with more general information, ‘Human Impersonal Constructions’ is used in the title for convenience.

1.4 Unavailability and reduced acceptability of arbs in connection with certain verbs

Besides relevant and obvious morphosyntactic differences, the three constructions under question in Polish are characterized by stylistic distinctions. While the 3pl construction in Russian covers all notions, the situation is different with regard to Polish. In Polish the 3pl construction is considered more colloquial than the other two options (the -no/-to construction or the reflexive construction). Regarding the restrictions of the arbs discussed in the literature, there is an agreement among linguists (e.g., Bogusławski 1984; Bunčić 2018: 103; Kibort 2008: 267; Laskowski 1984: 147; Mel’čuk 1974: 350; Padučeva 2012: 29, etc.) that these constructions can only be used with an implicit subject that is obligatorily [+​human]. To account for the semantic constraints of arb constructions, we refer to Silverstein’s animacy scale (1976: 122) according to which the following hierarchy must be applied for general nouns: human > animate > inanimate. In contrast to the -no/-to arb and the 3pl arb, a semantic property of the reflexive arb is that the feature [+​human] can be overridden, and a non-human animate subject can potentially be assumed. However, the construction is never used to refer to inanimates; the subject must be interpreted obligatorily as [+​conscious] (Krzek 2011: 71). Also, a human implicit subject is not always a sufficient prerequisite for arbs. Arbs were the subject of numerous discussions ←19 | 20→(e.g., Puzynina 1993; Wiese 1973) insofar as not all verbs can serve as a basis for them, and this is particularly true for the -no/-to arb. Linguists tried to investigate the semantic boundaries of the -no/-to arb by adopting morphological and semantic perspectives. For instance, Puzynina (1993) and Bogusławski (1984) highlight that preferably verbs denoting activities can serve as basis for the -no/-to forms. Kibort (2004: 277) observed that in predicates which have a stative and a non-stative reading like być ‘be’, the -no/-to arb seems to prefer a volitional participant. So far, no theory has offered a plausible explanation why some verbs are barred from this construction, even if the basic requirement of a human agent is fulfilled. Two studies provide initial evidence in this respect: a study with intransitive Polish verbs on the acceptability of the -no/-to arb (Bunčić 2019) and a study with intransitive German verbs on the acceptability of the man impersonal (Kretzschmar & Primus: in press). Both studies report that some verb groups yielded better ratings as opposed to other verb groups. This finding has given rise to the hypothesis that arb constructions depend on a prominence relation in the sense of Himmelmann and Primus (2015), in this case involving the agentivity of the verb (which will be the topic of Chapter 4). The main idea pursued in this book is that the availability and grammaticality of Polish and Russian arbs do not only depend on animacy but also on agentivity, which, as already mentioned in the beginning, is motivated by the verb’s semantics.

Since verb semantics is the starting point for the solution of this riddle, in Chapter 3 I will take a closer look at the basics of thematic role approaches operating with taxonomic lists of semantic properties. This chapter is devoted particularly to the notion of agentivity and the semantic role of the agent. Dowty (1991) operates in his framework with two maximally distinct poles: the proto-agent and the proto-patient. This model will form the central basis of how agentivity is understood and analyzed in this book. There is a “preliminary list” of five entailments, but for our purposes it is sufficient to use only the first four entailments, namely [volition], [sentience], [movement], and [causation], and refer to them as ‘agentivity features’. These features can occur not only in different combinations but also in isolation from each other. In addition to Dowty’s model, I will also focus on Primus’s model (1999a, 2006), which operates with three roles: the proto-agent, the proto-patient, and additionally the proto-recipient. Generally, Primus’s proto-properties show a significant overlap with Dowty’s lists of predicate entailments, but there are also some important innovations, which will be discussed. The goal of this chapter is to offer a discussion of both models and extract the most important aspects of the models for a thorough investigation of agentivity in arb constructions in Polish and Russian.

Chapter 4 gives an overview of the frameworks that will serve as the working basis in the three following empirical chapters (5, 6 and 7). The central ideas will be taken from Dowty’s (1991) prototype approach, and the notion of linguistic prominence will serve to refine Dowty’s approach. Dowty’s prototype model, which operates with feature aggregation, will be applied as empirical basis for the analysis of agentivity. By summing up agentivity features, the strength of agentivity can be identified, and within Dowty’s framework this is the appropriate means of ranking agentive roles along a prototypicality cline. For the identification of agentivity features, reliable linguistic tests are required, which will be presented.

Dowty’s approach is formulated construction-independent, that is, it should be true for a certain verb x in any linguistic construction. The above-mentioned acceptability test for the -no/-to arb in Polish by Bunčić (2019), however, shows that intransitive verbs with only one agentivity feature turned out to be better than verbs with two agentivity features. According to Dowty’s approach the verb that accumulates a higher number of features should have been better irrespective of the linguistic construction. What is even more surprising about Bunčić’s study is that the same verbs entailing three, two, and one agentivity feature(s), when embedded in an active construction, scored equally well, and that there are practically no differences in acceptability. The diverging results for the arb and the active construction disconfirm role prototypicality. Moreover, the fact that verbs with one agentivity feature turned out to be better than those with two features in the study by Bunčić (2019) raises the question of whether agentivity features have to be weighted or ranked. A related question is whether some of the features can be decomposed further. One possible solution to explain these phenomena is to look for an alternative to feature aggregation that still addresses the basic principles of Dowty’s and Primus’s frameworks but explains the diverging clines and the discrepancy in acceptability between verbs with different numbers of agentivity features. In other words, a more flexible approach is needed. A plausible explanation for diverging acceptability clines might be role prominence, as proposed by Kretzschmar et al. (2019). More precisely, they consider feature prioritization a possible method. In this approach some or one feature(s) have priority over others in constraints for certain grammatical constructions. We will follow this proposal and test it with further acceptability judgment studies with special focus on the agentivity features [sentience] and [movement] (Chapters 6 and 7).

Bunčić (2018) showed in a corpus study that there is a correlation between the number of agentivity features of verbs and the frequency of certain arb constructions using Polish and Serbo-Croatian corpora. In Chapter 5 I will use the same method as in the study by Bunčić and run a larger corpus study (using the Araneum Polonicum Maius) for Polish. Altogether, thirty-one intransitive and transitive verbs in Polish will be analyzed. Again, the starting point will be the simple feature-aggregational model in Dowty’s sense as described above, which serves to assign the selected verbs into different groups. The aim of this chapter is to explore the behavior of agentivity effects in the -no/-to arb and the reflexive arb. The verbs will be analyzed in both aspects: the present, future, and past tenses, and the conditional, respectively.

The following objectives are set:

• • The main aim is to find out which agentivity features are relevant for the frequency of the arb constructions.
• • A related question is whether it is solely the number of features that counts or some features have priority over other features and therefore might have to be weighted or ranked. Based on the collected data, I will try to define a means of measuring the weight of agentivity features.
• • I will try to answer these questions with regard to the respective constructions. This means that properties could show construction-specific behavior, which would result in different kinds of weightings and rankings of the respective features across constructions.

1.7 Exploration of the features [sentience] and [movement]

Chapter 6 will explore the notion of the agentivity feature [sentience], and Chapter 7 will concentrate on the feature [movement].

For the investigation of [sentience] four acceptability judgment tests with Polish native speakers were carried out. The test design of those studies is very similar to a recently conducted study by Kretzschmar et al. (2019). In this study they provided initial evidence that role-related acceptability clines for the same set of verb classes differ as a function of syntactic construction in German, thus disconfirming the prototype approach. Furthermore, it shall be investigated whether [sentience] should be decomposed into several features, for example, differentiating in perception, emotion, and cognition (as proposed e.g., by Van Valin 1999; Viberg 2001). By decomposing this feature, I want to verify whether ←22 | 23→verbs representing the mentioned sub-features might behave differently and yield different results in acceptability judgments.

Since the semantic status of [movement] is unclear, by means of two acceptability judgment tests (one for the Polish reflexive arb and one for the Russian 3pl arb¹), the semantic characteristics of the feature [movement] will be explored. The ambiguous status of [movement] is illustrated by the following, quite different, definitions: Dowty (1991: 552) and Primus (2012b: 25f.) attribute [movement] to any form of activity of the participant in question, also for subtle activities like in look at, sneeze, bleed, and vomit and not only to such activities that imply a change of position. In a different passage Dowty (1991: 602) assumes a “small local movement on the part of the agent participant” when the agent participant carries out dynamic actions, for example, throwing a ball. Obviously, according to Dowty (1991) and Primus (2012a, b), there are different manifestations of [movement]: (i) internal movement within the agent participant, (ii) a small local movement, and (iii) movement as the change of position. This raises the question of whether the binary concept of [movement], as proposed by Dowty, can be substantiated or whether it should rather be conceived of as a gradual ←23 | 24→category. The underlying hypothesis in the two studies assumes that [movement] as change of position is a gradual category: verbs like ‘run’ denote a faster locomotion than verbs like ‘walk’ or ‘pace’. In the studies motion verbs were classified according to the degree of velocity they represent. In sum, three degrees of speed are distinguished: fast motion [+​fast], neutral motion [±fast], and slow motion [−fast]. We expect verbs with the feature [+​fast] to be more agentive than verbs with [±fast] or [−fast]. Verbs marked as [+​fast] should therefore be more acceptable in arbs.

2.1 The problem with ‘impersonal’ constructions

The purpose of this chapter is to analyze impersonal constructions in Polish and Russian and in particular the function of arb constructions. The guiding questions in this first chapter are theory-driven and are as follows: Do impersonal constructions have to meet structural requirements and, if so, which ones? What is the essence of an ‘impersonal’ construction? Are there different types of impersonal constructions and thus different strategies of impersonalization?

A great deal has been written about impersonal constructions and/or impersonal verbs, and one goal of this chapter is to provide an overview of impersonalization strategies. The terms ‘impersonal construction’, ‘impersonal verb’, and ‘subjectless construction’ often have blurred boundaries and are used synonymously in linguistics for the description of various phenomena, including impersonal constructions in the strict sense, passives and impersonal verbs (the last of which I will term ‘true impersonals’).

The term ‘impersonal’ itself is only reluctantly used by many scholars, who cite terminological issues and argue the term to be misleading. Rather the term ←25 | 26→is used as a ‘label of convenience’, and it is also considered to be ‘a time-honored term’. It can be applied both to a group of verbs and expressions defined by their semantics as well as to a certain type of construction described from a syntactic perspective (Méndez-Naya & López-Couso 1997: 186). The first logical step building on the above-mentioned facts is to delimit impersonal constructions that are determined by the semantics of the verb as such from those impersonal constructions which put the acting subject in the background or completely demote it and thus are primarily impersonal from a syntactic viewpoint.

A typical definition of impersonal constructions, mostly based on the specific characteristics of ‘impersonal verbs’, can be found in Bußmann (2008: 765). According to this definition, the logical subject is not expressed by the grammatical subject. This also includes inherent impersonal verbs like ‘rain’ or ‘shine’:

Unpersönliche Konstruktionen. Syntaktische Konstruktionen, in denen das logische Subjekt nicht durch das grammatische Subjekt ausgedrückt wird, wie es vor allem der Fall ist bei Unpersönlichen Verben (Es mangelt ihm an Sensibilität), bestimmten Passivkonstruktionen (Es darf gelacht werden) […].

‘Impersonal constructions. Syntactic constructions in which the logical subject is not expressed by the grammatical subject, as in the case of impersonal verbs (He lacks sensibility), and certain passive constructions (Laughing is permitted) […].’

Another definition in which syntactic and discourse-pragmatic properties play an important role since the subject has no referential function is the following one:

Subjektbildung der 3. Person ohne referentielle Bedeutung, im Dt. durch das unpersönliche es ausgedrückt. (Werner 1988: 938)

‘Subject formation of the 3^rd person without referential meaning, expressed in German by the impersonal es.’

What is referred to as the “logical subject” is in many approaches “divided into the semantic-pragmatic roles of agent, experiencer, and beneficiary” and clearly lies outside the syntactic level (“Was dagegen gemeinhin als ‘logisches Subjekt’ bezeichnet wird, zergliedert sich […] in die semantisch-pragmatischen Rollen des Agens’, Experiencer und Benefizienten und gehört mithin einer ganz anderen Ebene an”, Wiemer 1995: 312). Many linguists “who define impersonal verbs from a semantic perspective”, as Méndez-Naya and López-Couso (1997: 186) summarize, “point to the existence of two clearly distinct classes”: (i) “verbs denoting natural phenomena, such as […] rain”, (ii) “verbs which refer to events […] outside the volitional control of an experiencer (see McCawley 1976, 194)”. The second group consists of verbs that usually denote “physical, mental or emotional experience” or “verbs denoting needs, obligations and happenstance” ←26 | 27→(Naya & López-Couso 1997: 186). “The most widespread opinion” among linguists “is that only weather verbs are genuine members of the” ‘impersonal’ group and therefore “are labelled really impersonal verbs (van der Graaf (1904, 1)) or ‘impersonals proper’ (Mustanoja (1960, 433))”, as Méndez-Naya and López-Couso (1997: 186) summarize.

The second group of impersonals, which are clearly distinct from those in the first group, includes constructions formed with a personal verb with reduced reference, denoting an action that is exclusively tied to human entities, as exemplified in the following sentences:

According to Siewierska (2008a: 115), the notion of impersonality has been well studied in the context of Indo-European languages and especially in Indo-European diachronic linguistics. To my knowledge, the first work in Slavic linguistics devoted to ‘impersonals’ is by Miklosich (1883)². This work is a comprehensive study of subjectless sentences in Slavic. In a nutshell, Miklosich (1883: 2) observed that subjectless sentences display the following properties:

[…] wird das Subject nicht nur nicht ausgedrückt, sondern nicht einmal gedacht. Nach einem solchen Subjecte kann nicht gefragt werden, und die Frage kann bei solchen Sätzen nur lauten: was geschieht? In allen solchen Sätzen wird ein Vorgang ausgedrückt, ohne dass das wirkende Subject genannt wird: das Verbum tritt völlig subjectlos auf.

[…] the subject is not only not expressed, but not even thought. Such a subject cannot be asked for, and the question in such sentences can only be: what happens? In all such sentences a process is expressed without the acting subject being named: the verb occurs completely subjectless. (Miklosich 1883: 2)

←27 | 28→In this exhaustive 1883 study, Miklosich describes a wide range of varieties of impersonal constructions, not only in Slavic languages but also in other Indo-European languages. Miklosich analyzes the impersonal constructions according to the use of the verb and its form, respectively. This results in a differentiation into four groups. Group 1: sentences containing an active verb;³ Group 2: sentences containing a reflexive verb, for example, in German sich kümmern ‘care’; Group 3: sentences containing a passive verb;⁴ and Group 4: sentences with a noun and the verb esse.

Group 1:

Group 2:

Group 3:

In the introduction to their extensive collection of papers, Malchukov and Siewerska (2011: 3) summarize that impersonal constructions have frequently been the subject of theoretical approaches or models of grammar and the resulting controversial issues: in Generative Grammar (GG), the subject-based notion of impersonality has been described by Chomsky (1981) as a phenomenon of a ‘syntactic zero’ (pro, PRO, NP trace; operator trace). Consequent expansions on this topic deal with expletive zero subjects, EPP violations (Extended Projection Principle), and non-canonical subjects, as reflected in the works of various linguists, for example, Svenonius (2002), Mendikoetxea (2008), or Biberauer et al. (2010). The same subject-based approach has been used in other frameworks as well, including LFG (Lexical Functional Grammar) and Kibort’s (2008) adaptation of it. Mel’čuk (1979) offered an unusual approach to impersonal constructions, postulating several zero elements (zero lexemes), including ∅^elements (for a 3sg neut. argument) and ∅^people (for 3pl argument)⁵. This approach is embedded in ←29 | 30→the wider context of Meaning-Text Theory (MTT), which was developed by Žolkovskij and Mel’čuk (1965), suggesting that language consists in a mapping from semantics (the content/meaning) to its form or text (phonetics). Mel’čuk’s notion of zero lexemes is an extension of the general notion of zero signs, and a zero lexeme contains only what Mel’čuk (1995: 178) calls a ‘zero lex’. Siewierska (2008a) also addresses the characteristics of impersonal constructions. What she observes is that the term ‘impersonal’ can be interpreted and analyzed in different ways, depending on the perspective and on which linguistic level is in focus. She summarizes that with a focus on structure, ‘impersonal’ can signify the absense of a canonical subject, whereas from a functional perspective ‘impersonal’ can mean that the agent is defocused (Siewierska 2008a: 116). In the agent-defocusing strategy the agent is understood, as Siewierska (2008a: 121) notes, in a broad sense. The agent is conceived of as the “causal participant of an event”, and the terms actor, instigator, and initiator are often used synonymously to render this function (Siewierska uses the term instigator). By defocusing the agent, its prominence is lowered in comparison with what it would have been in a normal, personal active construction. According to Siewierska (2008a: 121), “the defocusing may involve (a) the non-elaboration or under-elaboration of the instigator, (b) the demotion of the instigator from its typical subject or topic function or (c) both demotion and non-elaboration”. Constructions that fulfil the characteristics mentioned in (a) are constructions without a fully referential subject (ibid.). Moreover, from the agent-defocusing perspective, impersonality is understood as a decision made by the speaker in relation to the construction of an event in response to the effects of discourse (Malchukov & Siewierska 2011: 2).

From the subject-centered point of view, a subject is canonical if it is fully referential and displays the relevant morphosyntactic properties of a subject (ibid.). In this respect, Siewierska’s analysis of a ‘canonical subject’ is largely consistent with Keenan’s (1976) definition of a prototypical subject. According to Keenan (1976), a prototypical subject should be: (i) a referential argument, (ii) a definite NP (iii) topical, (iv) animate, and (v) agentive. This is clearly in opposition to the qualities that constitute a subject in impersonal constructions. From the subject-oriented point of view, constructions which have been identified as impersonal have a subject to which the following properties apply: (i) the subject is not fully referential, (ii) the subject “does not display canonical subject properties”, (iii) the subject “is not a verbal argument but merely a place filler manifesting no semantic or referential properties, i.e. an expletive subject”, (iv) the subject is covert (Siewierska 2008a: 116).

←30 | 31→Another rather discourse-based approach is offered by Gast and van der Auwera (2013: 136), who describe the process of impersonalization (the concept was borrowed from Blevins (2003) but adapted in a slightly different way) as follows:

Impersonalization is the process of filling an argument position of a predicate with a variable ranging over sets of human participants without establishing a referential link to any entity from the universe of discourse.

In the following sections of this chapter, I will take a closer look at different functional varieties of impersonal constructions, based on Siewierska’s (2008a, 2008b) functional typology and on Kibort’s (2008) LFG-based classification, since they provide comprehensive lists of impersonal constructions. The presented theories are not radically divergent, neither do they overlap entirely, but they offer different views and analyses on impersonality and represent a wide range of impersonal constructions. Subsequently, I will present a semantics-based approach to impersonal constructions and impersonal verbs (see Section 2.6.3).

2.3 Siewierska’s four types of impersonal constructions