Preface

[0.1] Errors and omissions are unavoidable in scientific writing. They are the writer’s equivalent of statistical Type I and Type II errors, respectively. And just as with statistical data, I have worked hard to keep all errors and omissions to a minimum in relation to the number of justified statements and judgements. Still, I need to start with a major disclaimer: this book is a work in progress. The current version undoubtedly contains numerous errors, omissions, inaccuracies and wrong generalisations. I can say this with certainty because I have been changing, adding and deleting details up to the last moment before publication of the current version of this book. And I do not expect it to stop here. Actually, the work-in-progress status is intended to be taken quite literally: I plan to update and revise this book regularly in the future. Any progress can be followed online at github.com/cysouw/diathesis. I welcome any suggestions for improvement, which can be submitted as an “issue” on that website, or, even better, as a “pull request” including proposed changes and corrections.

[0.2] This book is about German grammar, but the book is written in English. These two decisions have a purely personal background. First, the idea to write about German grammar arose in the context of me teaching German grammar at the Philipps-University in Marburg. Actually, the diversity of diathesis could, and should, be investigated in the same depth in other language besides German. Second, the book is written in English first and foremost because I personally feel more comfortable writing in English than in German. Also, I think that the current approach to diathesis is also of interest to readers that do not care too much about all minute details of German grammar. And for the readers that are interested in those details of the German language, I assume that they both have a working knowledge of English (so they can read this text) and of German (so they can understand the German examples). For that reason I decided to omit any interlinear glossing of the examples. Most examples are simple enough to be understood even with just an approximate understanding of German. Adding interlinear glossing throughout would be an enormous undertaking, which I think is not worthwhile given the intended readership.

[0.3] This book is written in Pandoc Markdown. Markdown (commonmark.org) allows for clean and readable raw text, while Pandoc (pandoc.org) provides easy transformations of the text into beautiful output, for example in formats like XeLaTex/PDF or HTML. I have used various extensions for Pandoc (“filters” in Pandoc-parlance), for example to format and number linguistic examples. More information on these filters can be found on the GitHub webpage mentioned above.

[0.4] The many lists, examples and subsections of this book make it a serialised database, and I have included many cross references to connect related parts throughout. To read the resulting hypertext I would urge the reader to try out an electronic version, either PDF or HTML. I personally have become really enamoured with the HTML version as it allows for easy searching and for quick forward-and-backward jumping through the text using our already internalised muscle-memory from web browsing. Also, the advances in CSS-styling have progressed to such an extend that the layout of HTML is almost approaching LaTeX sophistication, while adding a responsive/adaptive design (Marcotte 2010). The HTML version of this book is prepared as a single standalone file that can be used offline. This file will be opened in a web browser, even when the file is saved locally. It just uses the web browser as a text-rendering engine.

[0.5] Many thanks to Martin Haspelmath, Simon Kasper, and especially to Jens Fleischhauer for extremely helpful comments and detailed suggestions. The proofreaders at the Language Science Press did magnificent work: many thanks to Agnes Kim, Annemarie Verkerk, Elliot Pearl, Felix Kopecky, Jeroen van de Weijer, Katja Politt, Lea Schäfer, Lisa Schäfer, Patricia Cabredo Hofherr, Yvonne Treis and one additional anonymous proofreader. Of the many students that have participated in my lectures and seminars (and had to endure my work-in-progress) I would like to explicitly thank Annika Besser, Dennis Beitel, Franziska Beyer, Patricia Bier, Katja Daube, Milena Gropp, Eric Ilten, Jens Jakob, Christina Jann, Vanessa Lang, Katrin Leinweber, Rieke Hänche, Philip Ossowski, Nico Reinicke, Kristina Splanemann and Stella van den Berg. They wrote very useful term papers related to the topics discussed here and provided valuable suggestions, feedback and corrections. And finally I would like to thank all people at the Digitales Wörterbuch der deutschen Sprache (DWDS) as listed at dwds.de. Such easy-access resources are reinvigorating linguistic research enormously.

1 Setting the scene

1.1 The daunting diversity of diathesis

[1.1] The quintessential example of German diathesis, discussed in every grammatical description of this language, is the werden+Partizip passive construction (10.146 a). The crucial characteristic that makes this a diathesis is that the state-of-affairs as described by the passive is fundamentally the same as in the corresponding active (1.1 b). Yet, while the circumstances remain the same, the grammatical structure and the communicated perspective differ between the two expressions.

[1.2] This general approach to diathesis, viz. alternating sentence structures that express approximately the same state-of-affairs, is applicable to a large number of grammatical phenomena in German. All in all, almost 250 different German diatheses are described in this book, some highly productive, some only attested for a handful of verbs. The main goal of this book is to present this wealth of grammatical possibility in a unified manner, while at the same time attempting to classify and organise this diversity.¹ I will make no attempt to fit all the hundreds of constructional possibilities of the German language into any specific grammatical framework, although the collection of diatheses presented here might be taken as a modelling-challenge for your favourite grammatical theory.

[1.3] It might come as a surprise that there are so many different diatheses in German, but my impression is that in this respect German is no exception among the world’s languages. I expect that all languages have a similar abundance of different ways in which to construe a sentence around a lexical predicate. In a sense, a diathesis allows for the expression of a distinct perspective on the event described, something that is arguably a common desire of any language user.

[1.4] By way of an introduction, consider the following five illustrative examples of diathesis.² Many examples in this book contain masculine nouns, not because of laziness on my behalf, but because their definite articles overtly show the different German cases (der, des, dem, den). Notwithstanding this grammatical preference, I will try to use examples with as much diversity as possible throughout this book. Some verbs, like beginnen ‘to start’ (1.2 a), allow for a passive-like construction without any werden auxiliary, often called “anticausative”. Other verbs, like schießen ‘to shoot’ (1.2 b), allow for an alternation between an accusative and a prepositional phrase, often called “antipassive”. Further, there are many different kinds of diathesis marked by a reflexive pronoun, like the reflexive antipassive with beklagen ‘to lament’ (1.2 c). Diathesis is also frequently marked by a prefix, like the applicative between stammen aus and entstammen ‘to descend from’ (1.2 d). Lastly, many light-verb constructions show diathesis, for example the sein+zu‑In­fi­ni­tiv passive, as illustrated below with the main verb lösen ‘to solve’ (1.2 e).

1.2 Defining diathesis

[1.5] A diathesis is defined here as special kind of alternation between two different clause constructions. To properly define diathesis, I will first define “alternation” in general. The definition of “clause alternation” will then be established on that basis. Finally, a diathesis will be defined as a special kind of clause alternation.

[1.6] Hidden in these succinct definitions there are various grammatical concepts that will be expanded upon in the subsequent sections. The more general aspects of clause alternation are discussed in this chapter, directly below the following definitions, while the details of the analysis and the classification of diathesis are postponed to the next chapter.

Alternation

[1.7] An alternation (or simply “grammatical marking”) is defined as follows:

• An alternation consists of two different linguistic constructions, i.e. there are two alternants.
• Typically, the lexical root (or possibly a larger linguistic entity) does not change in the alternation, i.e. there is some fundamental linguistic material in both alternants that remains the same.
• Typically, the difference between these alternants is overtly indicated by the presence of some additional linguistic material.
• Typically, the overt marking introduces a direction into the alternation, i.e. one alternant consists of less/shorter/zero linguistic elements (“less mark­ed”) and one consists of more/longer/overt linguistic material (“more mark­ed”). In some situations the direction of the alternation remains undecidable, i.e. the alternation is equipollent.
• The collection of roots (or possibly larger linguistic entities) to which the alternation applies creates an empirical linguistic class. Such an empirical class often has a semantic/functional interpretation, but not necessarily so.

[1.8] Alternations include basic morphological oppositions, like singular vs. plural (1.3 a) and in general comprise any opposition of grammatical forms, like synthetic present vs. analytic perfect (1.3 b). Alternations also exist in syntax as oppositions between different sentence structures, like a plain transitive nominative+accusative structure alternating with an intransitive reflexive anticausative with obligatory adverbial (1.3 c).

[1.9] There is a thought-provoking and almost philosophical issue here, which I will not further explore, namely whether the basis of grammatical analysis are the constructions themselves or the alternations between constructions. The approach taken here is that the alternations are the more crucial entities. I consider alternations as the morphosyntactic equivalent to phonological minimal pairs. Alternations are also useful in the practice of grammatical description. The meaning/function of a construction by itself is often hard to describe in full, while the meaning/function of an alternation can simply be described by the difference in meaning between the alternants. Only the crucial facets that are added by the alternation need to be captured – a task that often is already difficult enough. Similar intuitions about the importance of alternations have led to the development of syntactic transformations (cf. Harris 1957).

Clause alternation

[1.10] Based on the above definition of an alternation, a clause alternation is defined as follows:

• In a clause alternation both alternants are monoclausal, i.e. both alternants contain a single main predicate. For German, the crucial definitional property of monoclausality will be a phenomenon called “coherence” (see Sec­tion 1.3.1).
• In both alternants, the main predicate consists of the same lexeme, i.e both contain the same verb and the meaning of the lexical verb does not (substantially) change between the alternants (see Sec­tion 1.3.2).
• The lexical roles of the verb do not change between the alternants. Lexical roles are an inherent part of the meaning of a verb and alternations do not change them. Note, however, that there might be new roles introduced or some roles might be left unexpressed (see Sec­tion 1.3.3).
• A specific clause alternation is only applicable to a restricted set of verbs. The list of applicable verbs constitutes the domain of application of the alternation. Any semantic/syntactic characterisation of these applicable verbs should be seen as a summary of the domain of application, not its definition (see Sec­tion 1.3.4).
• The meaning/function of the whole clause changes between the alternants. This functional difference is the whole point of having a clause alternation in the first place (see Sec­tion 1.3.5).

[1.11] Clause alternations are widespread when auxiliaries are introduced, like modal müssen ‘have to’ (1.4 a), see Sec­tion 11.4.7. However, clause alternations are attested with many more different kinds of marking, like the verb particle auf‑ marking completeness of the action (1.4 b), see Sec­tion 6.7.9, or the somewhat mysterious “free” reflexive sich with verbs like ansehen ‘look at’ (4.24 c), see Sec­tion 7.4.4. Arguably, the special word order in German subordinate clauses (viz. with the finite verb in clause-final position) can also be regarded as a clause alternation (1.4 d).

Diathesis

[1.12] Based on the notion of a clause alternation, a diathesis (sometimes also known as “valency alternation”) is defined as follows:

• A diathesis is a clause alternation in which at least one of the lexical roles has a different form between the clausal alternants, i.e. at least one of the roles is remapped to a different grammatical expression. In other words, a diathesis is a clause alternation with a co-occurring role-marking alternation (see Sec­tion 2.4). In opposition to diathesis, a clause alternation without any role remapping is called an epithesis (see Chapter 4).
• Remapped roles do not substantially change their meaning, i.e. “who does what to whom” does not change. However, the details of the relation between the main predicate and the remapped roles can (and normally will) change. For example, the role’s influence on the action, or its affectedness by the predicate, might change. At the extreme, the remapping of a role can consist in the addition of a completely new role or in the complete removal of an existing role (see Sec­tion 2.2).
• As is true for all alternations, a diathesis has a direction from a formally “less marked” to a “more marked” alternant. However, in some special instances a diathesis will have no overt marking other than the change in marking of the remapped roles itself. This will be called a covert (or equipollent, or bare) diathesis. The different ways of marking a diathesis will be the basic organisational structure of this book (see Sec­tion 2.3).
• As a practical restriction, I will only consider diatheses in which one alternant is a basic clause with a single finite verb. In principle, when there is a diathesis between a basic clause and structure A, and another diathesis between the same basic clause and structure B, then there is also a diathesis between structures A and B. However, such diatheses are not included here. It is a major goal of this book to present a precise description of all possible role remappings relative to the basic clause structure (see Sec­tion 2.7).

[1.13] The prototypical example of a diathesis is the werden+Partizip passive as in (1.5 a), see Sec­tion 10.5.15. Also widely acknowledged is the bekommen+Partizip dative passive as shown in (1.5 b), see Sec­tion 10.5.21. However, the diversity of diatheses in German goes well beyond such light-verb constructions. There is also, for example, a passive-like diathesis marked with a reflexive pronoun as shown in (1.5 c), see Sec­tion 7.5.7, or an applicative diathesis marked with a prefix be‑ as shown in (1.5 d), see Sec­tion 8.8.9.

1.3 Definitional details

1.3.1 Monoclausality and coherence

[1.14] Diatheses are defined here as alternations between clauses with the same main verb. However, a single clause in German can contain multiple verb forms, for example when auxiliaries are used. It is crucial to strictly distinguish between expressions in which a multi-verb construction is monoclausal and when it is not. Only monoclausal constructions will be considered in this book.

[1.15] To define monoclausality, I will use the fact that the finite verb in German is placed at the end of a subordinate clause. The dummy main sentence Es ist bekannt, dass ‘it is known that’ will be used to force such a subordinate construction. The position of the finite verb in the subordinate clause can then be used to identify the boundary of the clause. Concretely, everything that can occur before the finite verb still belongs within the clause. In contrast, everything that has to come after the finite verb belongs to a different clause. For example, the sentence in (1.6 a) will turn into (1.6 b) in a subordinate construction. The finite verbform gehe now occurs at the end of the sentence. In this example it is not possible for anything to follow after this finite verb, as shown by the ungrammaticality of (1.6 c). So, the original sentence in (1.6 a) is a single clause.

[1.16] Sentences with this characteristic will be called (syntactically) coherent, following Bech (1955; see also Kiss 1995 for an in-depth discussion).³ This usage of the term “coherence” is slightly confusing, because it is used here as a technical term from the syntactic literature, completely independent from the pragmatic usage of the term “coherence” for contextual interconnectedness. Coherent constructions are considered to be monoclausal. For example, coherence is attested in auxiliary constructions with participles (1.7) and infinitives (1.8). Such constructions are thus monoclausal.

[1.17] In contrast, constructions with zu and an infinitive are sometimes coherent, e.g. (1.9) with the finite verb geben ‘to give’, and sometimes non-coherent, e.g. (1.10) with the finite verb behaupten ‘to claim’. The coherent construction in (1.9 a) is thus monoclausal, while the non-coherent construction in (1.10 a) consists of two clauses.

[1.18] All diatheses in this book are monoclausal (by definition). Besides trying to list all German diatheses, I will also catalogue all monoclausal alternations without any change in role marking (i.e. without diathesis). Such alternations will be called epitheses. This book can thus also be read as a collection of all German monoclausal constructions, with or without role remapping. Somewhat unexpectedly, the number of alternations with role remapping (i.e. diathesis) is about twice as large as the number of monoclausal alternations without changes in role marking (i.e. epithesis).

[1.19] When the above definition of monoclausality is strictly followed, then there turn out to be dozens of verbs that can be used as the finite “auxiliary” in a coherent multi-verb clause. When used as finite auxiliaries, these verbs are grammaticalised, i.e. they shed much of their lexical meaning when used in a multi-verb construction. Such grammaticalised verbs are classified into different groups and referred to by many different names in the German grammatical literature, for examples Hilfsverb, Kopulaverb, Modalverb, Modalitätsverb, Halbmodalverb (Eisenberg 2006a), Nebenverb (Engel 1996: 406), Funktionsverb (Polenz 1963 cited in Kamber 2008: 34), Strukturverb (Weber 2005), or Stützverb (Seelbach 1991 cited in Kamber 2008: 34). I will not pursue the question here how to classify these verbs into different kinds. I will simply refer to the whole group of these auxiliary verbs as light verbs.⁴ The same English term “light verb” has recently also been used as a translation of the German term Funktionsverb (Fleischhauer & Gamerschlag 2019; Fleischhauer 2021; Fleischhauer & Hartmann 2021). Actually, both that use and my use of the term “light verb” are instances of a more general overarching concept. The Funktionsverb can be specified as a “light verb with a nominal main predicate”. In contrast, in this book the term “light verb” is used for a “light verb with a verbal main predicate” (cf. Sec­tion 13.2.4). All light verbs that will be discussed in this book are shown in alphabetical order in (1.11).

1.3.2 Grammaticalisation of lexical meaning

[1.20] It is not always immediately clear whether a verb remains the same verb in a clause alternation. For example, the verb trinken means ‘to drink’ when used as a transitive verb (1.12 a). In contrast, when used intransitively it contains a strong insinuation that the drinking includes too much alcohol, so it might better be translated as ‘to be an alcoholic’ (1.12 b). In effect, the transitive and the intransitive use of trinken have a different meaning. However, in this case, the special intransitive meaning is probably best analysed as a conversational implicature because the suggestion of alcoholism can be suppressed given the right context (1.12 c).

[1.21] In general, when the same lexical verb is used in different alternating constructions, then there is (of course) a difference in meaning between the two occurrences. However, ideally this difference is completely induced by the alternation and not by the lexical verb itself. Yet, it is extremely common for the combination of a lexical verb with the surrounding construction to grammaticalise into a new meaning. For example, the verb auftreten means something like ‘to act’ as an intransitive (1.13 a), but ‘to kick open’ as a transitive (1.13 b). Both meanings originate as metaphorical extensions from the meaning ‘to step on something (by foot)’.

[1.22] Likewise, historical processes can lead to current homophony of two different lexemes. For example, this appears to be the case with the verb abhauen, which has a transitive meaning ‘to cut off’ (1.14 a). However, it has attained another usage during the course of the 20th century as an intransitive meaning ‘to run away’ (1.14 b), probably based on a southern German dialectal meaning of hauen ‘to go, to walk’.⁵ Pfeiffer (1993), entry hauen at https://www.dwds.de/wb/etymwb/hauen, accessed 12 December 2018.

[1.23] In between those extremes (i.e. conversational implicature as with trinken and different lexicalisation as with abhauen) there are various intermediate stages of semantic separation. For example, the verb hängen ‘to hang’ can be used as a regular intransitive verb with a location (1.15 a). However, the specific combination with the preposition an can also have a special meaning ‘being emotionally attached to something’ (1.15 b). In this case it seems most appropriate to interpret the combination hängen an as a separate lexicalisation, although the path of the grammaticalisation to this new interpretation can still be intuitively grasped.

[1.24] As with all grammaticalisation, it is often difficult to decide where to draw the line on the continuum between implicature (trinken), metaphorical extension (auftreten), contextual lexicalisation (hängen an) or completely different lexicalisation (abhauen). I tend to be rather lenient in allowing slightly different meanings still to be counted as diatheses of the “same” verb. However, I will exclude clear examples of grammaticalised lexicalisation and treat those as separate verbs.

1.3.3 Lexeme-specific lexical roles

[1.25] A crucial aspect of diathesis is that the lexical roles of a specific verb do not change, only the grammatical encoding of the roles is modified. For example, the verb füllen ‘to fill’ occurs in various monoclausal constructions (1.16) but the roles of (i) filler, (ii) filled container and (iii) filling substance remain the same. The different grammatical forms that are used to express these roles are indexed with the corresponding subscripts in the examples below.

[1.26] Already from this example it becomes clear that it is often really difficult, if not impossible, to attach a specific valency to a verb. Given that most (and possibly all) German verbs show some kind of diathesis, I reject the traditional notion of a fixed valency belonging to a specific verb (see Sec­tion 2.2 for a more extensive discussion). Alternatively, I propose that it is possible to list all lexeme-specific roles (or equivalently lexical roles) as a fixed characteristic of each lexical verb. The following three criteria will be used to determine the lexical roles of a verb.

[1.27] First, each role that is case-marked in at least one clause alternant is a lexeme-specific role. In the example of füllen in (1.16), each of the three roles is marked as nominative, accusative or dative in at least one of the alternants, so all three roles are lexeme-specific. Various exceptions and stipulations to this criterion are discussed in Sec­tion 5.2.

[1.28] Second, all obligatory prepositional phrases are lexeme-specific roles. There are different kinds of such lexeme-specific prepositional phrases. First, some verbs obligatorily require such a location, like sich befinden (1.17), see Sec­tion 6.3.3. Second, some diatheses introduce an obligatory location, like the caused-motion diathesis with waschen (1.18), see Sec­tion 6.8.8. This diathesis turns an accusative object Hose ‘trousers’ (1.18 a) into an obligatory location (1.18 b,c).

[1.29] Third, there is a large class of lexeme-specific prepositional phrases that I will call governed prepositions, like with arbeiten an ‘to work on’ (1.19 a). These prepositional phrases are not obligatorily present (1.19 b). However, the intuition behind governed prepositions is that these prepositions are lexically determined by the verb and often strongly grammaticalised both semantically and structurally. For example, the preposition an with the verb arbeiten ‘to work’ (1.19 a) is semantically strongly bleached with none of its local meaning remaining. Syntactically, the prepositional phrase in arbeiten an can be paraphrased with a complement clause of the form daran, dass (1.19 c). The possibility of such a syntactic paraphrase will be used as the main characteristic to identify governed prepositions. Various exceptions and stipulations to this criterion are discussed in Sec­tion 6.2.

[1.30] Under this approach, there are a few lexical roles that appear to be very widespread, up to the point of seemingly being universally applicable to all verbs. If that would be the case, then it would defeat the idea of lexeme-specific roles. However, on closer inspection it appears that there are no roles that apply to all verbs. The closest contender is, arguably, the role of causer, which can be introduced to almost any German verb by using the lassen+In­fi­ni­tiv diathesis (1.20), see Sec­tion 11.6.2. However, there is a small group of verbs that do not allow for this diathesis, like gefallen ‘to like’ (1.21). This shows that even virtually universal roles are still lexically determined in German.

[1.31] Such lexeme-specific roles are called “verb-specific semantic roles” in Van Valin (2004) and “microroles” in Hartmann et al. (2014). The obvious next step (as discussed in both these papers) is to group such microroles into clusters of semantic/thematic meso­roles, i.e. widespread roles like agent, patient, experiencer, etc. Such semantic roles are used constantly in contemporary linguistics, but they are surprisingly ill-defined. For example, given a random German verb like füllen as exemplified at the start of this section, it is not clear at all what should be its semantic roles, and what criteria should be used to determine them. Hartmann et al. (2014) and Cysouw (2014) use cross-linguistic data to approach this problem. However, such an approach does not allow for language-specific definitions, which is the problem here.

[1.32] As a solution, I propose to use the applicability of a diathesis as a criterion for the language-specific determination of semantic mesoroles. For example, a “German patient” might be defined as the group of those lexical roles that are changed from accusative to nominative in the werden+Partizip passive diathesis. Note that this definition is not supposed to satisfy all intuitions that surround the notion “patient” in linguistics. Quite to the contrary, the proposal is to define a semantic role like “patient” on the basis of applicable diatheses and adapt any intuitions to that definition (or, if that feels too radical, simply use a different name for the newly defined semantic role, like Leidtragende). Even more general, a semantic role could also be defined by the cross section of multiple diatheses. The determination of suitable (combinations of) diatheses to define semantic roles for German will not be pursued here, but left for a follow-up investigation.

1.3.4 Domain of application and verb classes

[1.33] A widespread assumption in linguistic analysis is that most alternations (including diatheses) have a sensible domain of application. This is the idea that there is some rationale, often a kind of semantic characterisation, explaining which roots allow for a specific kind of linguistic marking (e.g. only transitive verbs allow for a werden passive).⁶ For example, compare the infamous sentence on the first page of Levin (1993: 1) “the behavior of a verb, particularly with respect to the expression and interpretation of its arguments, is to a large extent determined by its meaning”. In contrast to that claim, Levin in practice uses form to establish classes of verbs, not meaning. That practical approach is also taken here. Although I also believe that many alternations, including diatheses, mostly behave sensibly, I would advise not to expect too much (semantic) regularity to be hidden in grammatical structure. Many grammatical regularities have unexpected exceptions, be it because of haphazard diachronic change or through creative analogical extension. Even stronger, some grammatical marking appears to be almost completely without (semantic) rationale, like the assignment of plural allomorphy in German. If that can happen in morphology, it can also happen in syntax.

[1.34] As a practical approach to determine the domain of application (say, which verbs allow for the werden+Partizip passive) I propose to always first enumerate as many examples as possible. In other words, first empirically establish a verb class of those verbs that happen to be possible with an alternation. In this book, I will not be satisfied with four or five cases that suggest a neat semantic characterisation for a specific diathesis. In contrast, I will list as many as possible further examples, idiosyncratic as they may be. Only after such a formal listing of the domain of application (which ideally needs much more corpus research than I have been able to perform here) is it possible to investigate the presence of any (semantic) rationale. In many cases there might be a (partial) rationale for the attested group of verbs, but it is just as likely (and just as interesting) to have to conclude that there apparently is none.

[1.35] Taking this principle one step further, I propose to define the domain of application by the concrete listing of all examples. Any (semantic) characterisation is then always a post-hoc generalisation, not a definition. This radically lexicalist interpretation is the approach that I will follow in this book. For each diathesis I will list as many example verbs as possible that take part in the diathesis (for some I will reverse the approach and list verbs that are not possible). These lexical lists (i.e. the empirical verb classes) are the definition of the domain of application. For some diatheses I will speculate about semantic generalisations, but I consider these generalisations always to be secondary to the concrete listing of examples. My semantic generalisations are thus never a causal explanation.

[1.36] This principle of definition by listing even holds for questions of productivity. For example, when somebody would propose a nonce-verb like flurchten to be a new verb for the German language in an experimental setting, then its meaning is partially defined by stipulating what kind of diatheses it can take part in. For example, the following constructions might, or might not, be chosen in examples that contextualise the new verb. The choices made will strongly influence the interpretation of the new verb.

1.3.5 Functional analysis

[1.37] Intimately connected to the domain of application (i.e. which verbs allow for which diathesis) is the question as to the meaning/function of a specific diathesis. In essence, this question asks for a description of the difference in meaning between the two alternants of a diathesis. For example, what is the difference between the transitive schließen (1.23 a) and the corresponding reflexive anticausative sich schließen (1.23 b), cf. Sec­tion 7.5.2. Although the answer might seem obvious for some diatheses, it turns out to be extremely difficult to give a concise description of such differences for most diatheses, and I will regularly refrain from trying to provide such a descriptions. Each diathesis is actually its own research project, preferably investigated using predictive corpus analysis (cf. the large literature on the English dative alternation, or for German De Vaere, De Cuypere & Willems 2018).

[1.38] There are two empirical pieces of information that are crucial for such a functional analysis of a diathesis. The first key data point is the actual list of verbs that allow for a specific diathesis (i.e. the verb class as defined by a diathesis, see the previous Sec­tion 1.3.4). The second important consideration is any restriction on the kind of nominals that can be used to fill the roles of a diathesis (i.e. nominal classes as defined by a diathesis). Both these classes can be established empirically by collecting and analysing a corpus of examples of a specific diathesis.

[1.39] The problem of a functional description for a diathesis becomes even larger with the realisation that there are many dozens of diatheses, often highly similar to each other. For example, it is really difficult (cf. Schäfer 2007; Kurogo 2016) to characterise the difference between the reflexive anticausative, like with schließen above in (1.23), and the unmarked anticausative, like with kochen in (1.24), see Sec­tion 5.5.5.

[1.40] The problem of providing a concise functional description for a grammatical construction is not restricted to diatheses. Clause alternations that do not have any role-remapping are also in need of a functional description (see Chapter 4 for a survey). An example is the phenomenon of “free” reflexives, illustrated with ansehen in (1.25), see Sec­tion 7.4.4. Although there is no role-remapping in this alternation (and thus no diathesis), it turns out to be really difficult to describe the difference between (1.25 a) without reflexive and (1.25 b) with reflexive. All clause alternations, with or without role-remapping, are in need of a functional analysis, and most such analyses will need substantial further research.

1.4 Method

[1.41] Methodologically, I regard the approach in this book as an attempt to unify grammatical research with lexicographic research, two aspects of linguistics that are often considered to be separate inquiries. In contrast to such a separation, I would like to propose a view of linguistics that might be called grammar of the lexicon (cf. Levin 1993: 2–4, but one might just as well include all of construction grammar here). In this approach, each grammatical phenomenon should always be linked to individual occurrences, either types (lexicon) or tokens (corpus). It is my experience from compiling the current compendium of diatheses that identifying and characterising a specific grammatical structure is really just “step one” of grammatical research. Only by trying to find more examples, with different lexemes and in different contexts, it becomes clear how prominent and varying a grammatical structure really is.

[1.42] As as rule-of-thumb, I propose the 10-in-10-rule as “step two” of grammatical research. If you think that a particular construction is widespread, or maybe even typical for a specific class of words (e.g. typical for intransitives), then take 10 minutes to search for examples, either using your own intuitions or in one of the many online databases or corpora. Resources like the Digitales Wörterbuch der deutschen Sprache (DWDS), the various online offerings of the Institut für Deutsche Sprache (IDS), or just a Google “exact search” are literally just a click away and completely rejuvenate linguistic research. If you are not able to list 10 examples in 10 minutes, then go back to the drawing board and reconsider your intuitions.

[1.43] Such quick-and-dirty 10-minute surveys of course still only represent a preliminary step of grammatical analysis. It is this level of detail that I have aimed for in this book (although I have often spend way more than 10 minutes searching for examples of a specific diathesis). Each sub-subsection about an individual diathesis lists lexemes that can be used with this diathesis. It includes examples of the diathesis, either from my own intuitions or from any of the above mentioned databases. I plan to expand and fine-tune these lists in future revisions of this book.

[1.44] However, the real research is only yet to come. “Step three” would be the in-depth investigation of individual diatheses by sampling examples from corpora and formulating predictive parameters to explain their usage. For example, De Vaere et al. (2018) investigate the dative antipassive (see Sec­tion 6.7.12) for just the single verb geben ‘to give’. Now, there are hundreds of diatheses and hundreds of verbs listed in this book, so there are tens of thousands of similar research projects just waiting to be tackled.

1.5 Previous research

[1.45] The current attempt to present an all-encompassing survey of German diatheses builds on a rich scholarly tradition (with many scholarly precursors to be cited in appropriate places throughout this book). An early precursor is Hans Conon von der Gabelentz (1861), which includes an astonishlingly modern take on the typological diversity of passive-like constructions. A direct inspiration for the current work is the highly influential survey of diathesis for English by Levin (1993), followed by a similar attempt for German by Sauerland (1994). A recent cross-linguistic survey of valency and diathesis in this tradition is edited by Malchukov & Comrie (2015), which also includes data on German (Haspelmath & Baumann 2013). A recent pan-Germanic survey of different kinds of diathesis is available in Kasper (2023).

[1.46] Mostly independently from this comparative work, there is a long tradition in the German grammatical literature to investigate diathesis, e.g. as “Konversenverhältnis der Aktanten” (Eroms 1980: 24; cf. Heringer 1968). An early attempt at a survey of various diatheses is presented by Höhle (1978). Basic summaries of German diathesis in the context of valency can be found in Eroms (2000: Ch. 10) or Ágel (2000: Ch. 6). There also are a few monographs about specific German diatheses (e.g. Leirbukt 1997; Holl 2010; Jäger 2013) and recently some corpus studies into the effect of specific diatheses on individual verbs have appeared (De Vaere, De Cuypere & Willems 2018; Imo 2018; Dux 2020: Ch. 6).

[1.47] Diathesis is of course closely linked to the concept of valency, so the groundbreaking valency dictionary for German by Helbig and Schenkel (1969) deserves mentioning. They identify the problem that certain verbs can be used in different constructions, but diatheses are not investigated consistently in their dictionary. Another highly influential valency dictionary for German, edited by Schumacher (1986; also the precursor Engel & Schumacher 1978), discusses passive diatheses for all verbs that are included. I regard the current survey of German diatheses as a next step to extend such valency dictionaries into even more all-encompassing dictionaries that discuss all possible clause constructions for each verb.

1.6 Structure of this book

[1.48] The structure of this book is somewhat unusual for a scholarly monograph. It is not a narrative with a painless beginning, a sturdy middle and a satisfying conclusion. Rather, this book is open ended, and it does not have a gratifying closure at the end. That is by design. It is an encyclopaedia after all.

[1.49] The book consists of three different parts. First, Chapters 1 and 2 discuss the major theoretical considerations that guide the collection and organisation of the data. This is the part to read for insights into the nature of diathesis and sundry topics. Second, Chapters 3 and 4 are the closest to what this book has to offer in the way of a conclusion. Each of these chapters presents an astonishingly long list of grammatical constructions, summarising a selection of the major linguistic structures as identified in the remainder of this book. These summaries are intended to be a quick entry into the actual German language data, with ample cross-references to the following encyclopaedia. Finally, Chapters 5 to 13 make up the actual encyclopaedia, with separate sub-subsections about individual phenomena. Each of these sub-subsections is a fragment of a research project, sometimes mundane, but often full of surprising avenues for future research.

[1.50] Going through the chapters in more detail, this first chapter summarised the basic definitional properties of clauses and clause alternations. The next Chapter 2 describes in detail how diatheses are analysed and classified. Chapter 3 presents a summary of the about 80 major German diatheses and proposes German names for those structures. As a small extra, Chapter 4 summarises the about 40 major epithetical structures and provides an unconventional approach to the analysis of verbal categories of German.

[1.51] Chapters 5 to 13 are the core part of this book. These chapter are database-like texts, discussing each clause structure separately. All of these chapters have exactly the same internal structure. Sometimes certain sections are empty, because there is no grammatical structure in that category. However, the headers of those empty sections have been preserved for the sake of parallel numbering across chapters. The following structure is used in all data chapters:

• Section X.1 is a concise introduction into the kind of diatheses discussed.
• Section X.2 is a detailed discussion of the relevant grammatical definitions.
• Section X.3 lists deponent structures
  (i.e structures that look like diathesis, but there is no alternation).
• Section X.4 lists epitheses
  (i.e. alternations with the relevant structure, but there is no diathesis).
• Section X.5 lists diatheses with subject demotion.
• Section X.6 lists diatheses with promotion to subject.
• Section X.7 lists diatheses with object demotion.
• Section X.8 lists diatheses with promotion to object.
• Section X.9 lists symmetrical diatheses.

[1.52] In these data chapters, each diathesis has its own sub-subsection with a unique section header that can be used as a name to refer to the diathesis (e.g. Sec­tion 10.5.15 on the “werden+Partizip Passive”). Many of these unique names are quite boring, but hopefully descriptively useful for future reference. More imaginative German names are only added for the major diatheses. Individual verbs are listed with each diathesis to show the extent of its applicability. Often a section with further examples is added to illustrate the diathesis. Neither these lists of verbs nor the examples are intended to be exhaustive in any way. Sometimes an additional section with notes is provided to discuss idiosyncrasies of individual verbs. All this information should be read as a first step towards a more in-depth research into individual diatheses and into the constructional possibilities that are available to individual verbs.

2 The structure of a diathesis

2.1 Analytical dimensions

[2.1] The central concepts to analyse the structure of a diathesis are valency (Sec­tion 2.2), voice (Sec­tion 2.3) and diathetical operation (Sec­tion 2.4). These concepts will be discussed extensively in this chapter. Two new concepts are introduced as well, namely stacking (Sec­tion 2.5) and chaining (Sec­tion 2.6). Finally, I will present an extensive discussion about naming diatheses at the end of this chapter (Sec­tion 2.7).

[2.2] A diathesis (as defined in Sec­tion 1.2) is an alternation between two different clausal construction. Each of the alternants show a different mapping of (grammatical) expressions onto (semantic) roles. Such an alternation is what is called a “diathetical operation” in Zuñiga & Kittilä (2019: 4), in contrast to the term “diathesis” being used for each individual mapping between expressions and roles. However, because I will only consider diathetical operations between an unmarked basic clause and a marked alternant, I have decided to simplify the terminology in this book. Both the alternation itself (Zuñiga & Kittilä’s “diathetical operation”, e.g. “causativisation”) and the derived alternant (Zuñiga & Kittilä’s “diathesis”, e.g. “causative”) will be referred to here simply as a diathesis, from Greek διάθεσις ‘placement in order, (re)arrangement’.

[2.3] A side-effect of this approach is that “active” is not a diathesis, but simply the unmarked counterpart of a diathetical operation. Even stronger, I will refrain from using the term “active” because it immediately conjures up “passive” as its antithesis. This opposition is too much of an oversimplification as “passive” is just one of the hundreds of possible diatheses. Also, the “active” does not necessarily describe an action, so content-wise this term is also ill-fitting.

[2.4] As an alternative to “active”, I will use the term basic clause as the unmarked base of comparison for all clause alternations. A basic clause is a clause with a single finite verb from, either in the German Präsens or Präteritum tense.⁷ Strictly speaking, a basic clause can also be a clause with a single finite verb in the Konjunktiv I or Konjunktiv II. However, because these are rather rare nowadays I have hidden this possibility in this footnote. All other verb forms, including the Perfekt and the other traditional German tense-aspect distinctions, are all derived clauses, i.e. the result of some kind of clause alternations. A diathesis is a special instance of a derived clause that also exhibits a diathetical operation. In constrast to a diathesis, a derived clause without any diathetical operation will be called epithesis, from Greek ἐπίθεσις ‘placement upon, imposition’. Epithesis is grammatical marking “on top of” a basic clause.⁸ There appears to be a rarely used alternative meaning of the term “epithesis” in linguistics to indicate the addition of a sound to the end of the word, i.e. a special kind of epenthesis, see for example http://www.websters1913.com/words/Epithesis, accessed 23 December 2022. A summary of all major epithetical constructions will be presented in Chapter 4.

[2.5] The actual linguistic marking of a diathesis (for example using verbal morphology or auxiliaries) is called grammatical voice. The crux of this term is that a voice is the language-specific linguistic expression used to mark the diathesis. For example, a diathesis in German can be “voiced” by a reflexive pronoun (Chapter 7) or by a light verb with a zu‑In­fi­ni­tiv (Chapter 12), and so on. The different kinds of grammatical voice in German are used here as the central organisational characteristic for the separation of the data chapters. An overview of the different German voices is given in Sec­tion 2.3.

[2.6] A major objective of research into diathesis is to describe the connection between a diathetical operation and its semantic effects on the meaning of the clause.⁹ Somewhat confusingly, Zuñiga and Kittilä (2019: 3) use the term “transitivity” (following Hopper & Thompson 1980) for such semantic effects. However, the effects of diathetical operations seem to far surpass differences in transitivity. For example, when an accusative argument is remapped onto a prepositional phrase (“antipassive”) this often implies less involvement of the participant. The details of such semantic effects for each of the hundreds of diatheses in this book is an important and fascinating topic, for which I can currently offer only limited insights throughout. In practice, I assume that each diathesis as described here has some kind of semantic/pragmatic effect. However, the concise specification of these effects needs much more detailed research, which has to be provided by subsequent work. Wherever I can, I will observe tendencies and propose hypotheses for such future research.

[2.7] Lastly, I prefer to use the term “diathesis” over the frequently attested term “valency alternation”, although in practice both terms can mostly be used interchangeably. There is only a difference when using a highly simplistic interpretation of the term “valency”, namely as indicating the number of arguments of a verb. There are many diatheses in which the number of arguments does not change between the alternants. For example, in (2.1) both sentences have the same (surface) valency, having three arguments: a nominative case, an accusative case and a prepositional phrase. Yet, there clearly is a role-remapping between both sentences. Strictly speaking, “number-of-arguments” valency alternations are then just a subset of all possible diatheses.

2.2 Valency

2.2.1 Utterance valency and lexical roles

[2.8] Valency is traditionally interpreted as a fixed constructional characteristic of a lexical verb, e.g. the verb geben ‘to give’ is said to be ditransitive. A central thesis of this book is that this conception of valency is too limited. Individual verbs can (and normally will) be used in many different constructions with different valency (i.e. most verbs show some kind of diathesis). Consider for example the verb wehen ‘to blow’. Such weather verbs are often considered to have zero valency, which in German is characterised by an obligatory non-phoric es pronoun (2.2 a). However, the same verb can just as well be used as an intransitive (2.2 b,c), as a transitive with an accusative object (2.2 d), or even as a ditransitive with a dative and accusative object (2.2 e). Note that the prepositional phrases in (2.2 c,d,e) cannot be left out and their obligatory presence might be used to argue for argument-status of these prepositional phrases. The example in (2.2 e) then will be an example of the verb wehen with a valency of four.

[2.9] There is a recurrent tendency in the literature to try and reduce such variation to a single valency per verb (viz. its “real” or “underlying” valency), and various strategies are employed to arrive at such a prototypical valency (see e.g. Welke 2011: Ch. 9 for a survey). That will not be the approach taken here. Instead, valency is proposed to be a characteristic of a specific utterance, not of a specific verb. So, the examples in (2.2) can simply be assigned an utterance valency from zero (2.2 a) to four (2.2 e) even though they all use the same lexeme wehen as their main verb.

[2.10] As a replacement of the traditional lexeme-specific notion of valency (e.g. geben is ditransitive), I propose to use the notion of lexeme-specific lexical roles (e.g. geben has lexical roles “giver”, “givee”, “given object”, etc.). Lexical roles are participants that are treated as an argument of utterance valency in at least one of the possible diatheses of a verb (see the next section for the complete definition of such arguments). The existence of such lexical roles is solely determined by the verb and does not change with different sentence constructions around the verb. Looking back at the example in (2.2 e), the following lexical roles of the German verb wehen ‘to blow’ can be established:

1. blower: the blowing air, Fahrtwind ‘headwind’.
2. blowee: Object affected by the blowing air, Mütze ‘cap’.
3. blowing location: Location affected by the blowing air, Kopf ‘head’.
4. blowing location possessor: Possessor of the locational object, in (2.2 e) the dative mir ‘my’. This role is necessarily the possessor of the location Kopf ‘head’ affected by the blowing.

[2.11] Additionally, it is of course possible to define a notion of lexeme valency. A straigtforward approach would be to take the utterance valency of a basic clause (cf. paragraph 2.4) as the definition of lexeme valency. However, in general lexeme valency has to be a much more complex construct. For example, lexeme valency can be defined as the collection of all attested utterance valencies for a specific lexeme. To be precise, this lexeme valency also has to include an indexation of the lexical roles across all arguments. This addition is important to distinguish between, for example, the lexical valency of kochen ‘to cook’ (2.3) and essen ‘to eat’ (2.4). Both can occur with a transitive and intransitive utterance valency, but the lexical roles that can be used in the intransitive differ. Specifically, the patient-role “eaten object” of essen cannot be used as nominative subject in the intransitive (2.4 c), which is different from kochen (2.3 c). In effect, these verbs have a different lexeme valency.

2.2.2 Arguments of utterance valency

[2.12] The lexical roles of a specific verb are defined as those participants that are treated as argument in at least one of the possible utterance valencies of this verb. So, to identify lexical roles, a strict definition of the arguments of utterance valency is needed. The following four kinds of arguments can be identified. First, all case marked noun phrase constituents are arguments, with a few exceptions that will be discussed in Sec­tion 5.2. Basically, case-marked arguments (2.5 a) can be questioned by the question pronouns wer or was, including their case forms wem, wen and wessen (2.5 b,c). Further, case-marked arguments can be pronominalized by personal pronouns (2.5 d) or the indefinite pronouns (irgend)jemand or (irgend)etwas (2.5 e).

[2.13] Second, prepositional phrases (2.6 a) are arguments of an utterance when they can be paraphrased by a complement clause of the form da(r)+preposition, dass (2.6 b). All details of the difficult question when to treat prepositional phrases as arguments in German are discussed in Sec­tion 6.2.

[2.14] Third, there exist obligatory prepositional phrases, though they are not very widespread in German and mostly designate a location. Some verbs always expect such a location-role, like sich befinden ‘to be located’ (2.7 a,b), see Sec­tion 6.3.3 and 7.3.4.

[2.15] More widespread are diatheses that introduce an obligatory local role, like with brechen ‘to break’ (2.8), see Sec­tion 6.8.8. This example crucially shows that arguments should be determined as part of the clause structure, not the lexeme structure. It is perfectly possible to use the verb brechen without a prepositional phrase (2.8 c), but only with a different lexical role in the accusative, viz. Felsen takes the role “broken object”, while Loch has the role “location where the breaking took place”.

[2.16] Finally, all complement clauses are arguments (2.9 a,b). Complement clauses can be questioned by was (2.9 c) and pronominalised by a definite pronoun es (2.9 d) or an indefinite pronoun (irgend)etwas (2.9 e). Complement clauses are thus syntactically highly similar to case marked noun phrases. Caution should be taken when interpreting pronominalised examples like (2.9 d,e), because it is not immediately obvious whether the pronouns are replacing a case-marked noun phrase or a complement clause. For example, with the verb hoffen (2.9 e) it is not possible to replace the pronoun es with a noun phrase, though a governed prepositional phrase with auf is possible (2.9 f). The impact of such complement-clause arguments has not (yet) been thoroughly investigated in this book.

2.2.3 es Arguments

[2.17] A further kind of utterance-valency argument can be instantiated by es, the 3rd person nominative/accusative pronoun in the neuter gender. There are four uses of this pronoun that have to be distinguished, the last of which is particularly important for the analysis of diathesis. First, the most obvious use of the pronouns es is for anaphoric reference (phoric es). The next two uses do not have argument-status (viz. correlative and position-simulating es). Most interestingly in the current context, the fourth usage of es does not have anaphoric reference, but will still be counted as an argument (viz. valency-simulating es). I will illustrate these four options below.

[2.18] First, es can be used for anaphoric reference to neuter nouns, typically with gender agreement as shown in (2.10). There are many variants of such phoric usage, extensively discussed by Czicza (2014: Ch. 2).

[2.19] Second, another kind of referential es occurs with some non-finite complement clauses. By definition, a complement clause replaces an argument (2.11 a), but in some examples a pronoun es remains in place of the original argument, side by side with the complement clause (2.11 b). This is known as a correlative es (Czicza 2014: 79ff.).

[2.20] Third, the pronoun es is also used to fill the first sentence position in front of the finite verb (Vorfeld in the German grammatical terminology), because there is a strong regularity in German that this position cannot be left empty (except in imperatives and in yes/no questions). Word order is rather flexible in German, and it is often possible to have no lexical content in the Vorfeld. In such sentences, the pronoun es has to be used to fill the Vorfeld, as shown in (2.12 b). This is known as a position-simulating es (Czicza 2014: 115).

[2.21] Finally, there are also constructions that obligatorily include the pronoun es in the sentence as part of the valency of the utterance. The main reason for such a pronoun es is that there is a strong regularity in German that a nominative subject has to be present in each sentence (with very few exceptions, see below). Note that “subject” is defined here strictly for German as the nominative noun phrase that shows agreement with the finite verb. When there is no such subject available, then the pronoun es is used to fill this gap. This is known as a valency-simulating es (Czicza 2014: 115). In the analysis of diatheses in this book, such valency-simulating es is alway explicitly noted.

[2.22] In constructions with a valency-simulating es, as exemplified in (2.13 a,b), the pronoun es can occur in the Vorfeld (2.13 a), seemingly parallel to the position-simulating usage (2.12 b). However, when another constituent is placed in the Vorfeld, the pronoun es in (2.13 a) cannot be dropped, but has to occur elsewhere in the sentence, typically immediately after the finite verb (2.13 b). This post-verbal retention of the pronoun es is a typical sign for the valency-simulating use.

[2.23] In a very limited set of constructions the expected valency-simulating es is not present, resulting in sentences without any formal nominative subject. Some of these examples are historical idiosyncrasies (2.14), see Sections 6.5.3 and 5.9.3, respectively.

[2.24] However, there are a few impersonal diatheses that completely remove the subject but still have no valency-simulating es. Specifically, the following diatheses result in sentences that do not have any nominative subject.

• (2.15 a) impersonal werden+Partizip, see Sec­tion 10.5.1
• (2.15 b) impersonal lassen+Infinitiv+Reflexiv, see Sec­tion 11.5.1
• (2.15 c) impersonal sein+In­fi­ni­tiv, see Sec­tion 11.5.3
• (2.15 d) impersonal sein+zu‑In­fi­ni­tiv, see Sec­tion 12.5.1

[2.25] There are also a few rare cases in which there is a valency-simulating es in what appears to be an accusative case (2.16), see Sections 6.3.5 and 9.7.3, respectively. These constructions need a more in-depth investigation.

2.2.4 Adjuncts

[2.26] Adjuncts, the counterparts of arguments, are phrases that are not specifically induced by the main verb of a clause. Typically, such adjuncts are adverbial prepositional phrases, see Sec­tion 6.2.2. Just like with arguments, adjunct status should not be linked to a lexical verb itself, but to the clause construction in which it is used.

[2.27] For example, the verb tanzen ‘to dance’ is typically considered to be an intransitive verb with optional (adjunct) location prepositional phrases (2.17 a,b). However, there is a crucial difference between the two locations in these two examples, see Sec­tion 6.8.1. The static location im Saal ‘in the hall’ (2.17 a) remains optional in the perfect (with auxiliary haben), see (2.17 c,d), while the dynamic location in den Saal ‘into the hall’ becomes obligatory in the perfect (with auxiliary sein), see (10.83 e,f). So, the obligatory location in (2.17 e) is an utterance argument (and as a consequence, the role of “path” is a lexical role of such verbs of movement like tanzen).

[2.28] Adjuncts are, by definition, optionally present, so there is a natural connection to them being unexpressed. A central unanswered problem is whether there is a crucial distinction between constructions in which a participant is obligatorily absent (i.e. impossible to express) vs. optionally absent (i.e. possibly not expressed). In most diatheses that involve absence, the whole point is that there is an alternation between absence and presence of a lexical role (e.g. in all diatheses that involve a drop or addition). The problematic cases are differences like passive vs. anticausative, which by definition are distinguished by possibility vs. impossibility for the agent to be expressed. This difference is highly volatile, i.e. it often differs from lexeme to lexeme whether it is possible or just dispreferred for an agent to be expressed.

2.3 Voice

[2.29] The formal linguistic marking of a diathesis, for example by verbal morphology or auxiliaries, is called grammatical voice (following Zúñiga & Kittilä 2019: 4). The different kinds of grammatical voice in German establish the basic organisational framework of this book. Each of the data chapters discusses a specific kind of grammatical voice, listing all diatheses using that “voicing”. The nine data chapters can be grouped into four kinds of grammatical voices:

• zero-marked “covert” diatheses (Chapters 5-6)
• diatheses expressed by reflexive pronouns (Chapter 7)
• diatheses expressed by preverbs, preverbials or adverbials (Chapters 8-9)
• diatheses expressed by light-verb constructions (Chapters 10-13)

[2.30] The first two chapters deal with diatheses that are not overtly marked as such, i.e. they deal with covert diatheses. Because there is no marking on either of the two alternants, it is often difficult to discern a direction in such equipollent alternations. In Chapter 5 I will discuss diatheses that only differ in the marking of case, for example unmarked anticausatives like (2.18). Chapter 6 deals with unmarked diatheses in which at least on of the alternants is a prepositional phrase, for example unmarked antipassives like (2.19).

[2.31] The contribution of reflexive pronouns for the marking of diathesis is discussed in Chapter 7. A central claim in this chapter is that ‘self-inflicting’ reflexive reference (2.20) does not count as diathesis in German. In contrast, there are various other diathetical constructions in German that use reflexive pronouns without such self-inflicting reflexive reference, like the antipassive in (2.21). In such diatheses the presence of a reflexive pronoun is the actual marking of the diathesis, it is not signalling that subject and object are the same participant. An important generalisation about diatheses with reflexive pronouns is that they are always demotions.

[2.32] In Chapter 8 I will turn to preverbs, i.e. verbal prefixes that in German grammar are known as Verbpräfixe and Verbpartikeln. Syntactically, these are different kinds of elements, but from the perspective of diathesis they appear to function rather similar. The most widespread diathesis marked by such preverbs is an applicative, like with be‑ in (2.22). Because of the bound morphological structure, these diatheses show a strong tendency to grammaticalise into a large variety of different kinds of diathetical operations. A central generalisation of the diatheses discussed in this chapter is that the resulting sentence structures after a preverb diathesis is mostly transitive (especially nominative+accusative).

[2.33] Closely related to preverbs are resultative preverbials that induce diathesis, like the applicative with leer‑ in (2.23). There exist also diatheses induced by evaluative adverbials, like the reflexive anticausative with a manner specification gut in (2.24). Although these two kinds of elements, resultatives and evaluatives, occur in rather different kinds of diathesis, for convenience both phenomena are combined into a single chapter on adverbial-like diatheses in Chapter 9.

[2.34] A large number of diatheses use light verbs in combination with a non-finite form of the lexical verb. A somewhat surprising insight is that light-verb diatheses always involve a role-change of the nominative subject. I distinguish four different kinds of light verb constructions, to be discussed in four different chapters. Chapter 10 discusses light verb construction with participles, like the infamous werden+Partizip passive (2.25).

[2.35] The next three chapters describe different combinations of light verbs with lexical verbs in the infinitive. Chapter 11 discusses light verbs with straight infinitives, like the lassen+In­fi­ni­tiv causative (2.26).

[2.36] Chapter 12 investigates light verbs with zu plus an infinitive, like for example the sein+zu‑In­fi­ni­tiv passive (2.27). The combination of zu with an infinitive is arguably completely grammaticalised and is considered here to be yet another non-finite verb form of German, alongside Partizip and In­fi­ni­tiv. I propose to call it the zu-In­fi­ni­tiv.

[2.37] Finally, Chapter 13 looks at the combinations of prepositions, article and an infinitive, like the halten+am‑In­fi­ni­tiv causative (2.29). In such constructions, the preposition and article are obligatorily fused (an+dem>am) and this fused combination cannot be separated from the infinitive. Such completely grammaticalised constructions appear to be rather recent in German and are often considered substandard. Only a few of such combinations pass all the tests for complete grammaticalisation. When all tests apply, then I consider the combination of preposition+article+infinitive to be yet another non-finite verb form of German, alongside Partizip, In­fi­ni­tiv and zu-In­fi­ni­tiv. I propose to call it the Präpositions­in­fi­ni­tiv.

2.4 Diathetical operations

2.4.1 Abbreviations used

[2.38] A diathetical operation is a change that happens to the marking of the participants in a diathesis. One of the central definitional properties of a diathesis is that the coding of at least one of the participants has to change, for example a participant erstwhile coded with an accusative turns into a prepositional phrase. Because the role of the participant remains the same (again, by definition), such a change amounts to the mapping of a role onto a different grammatical form. I will call this process role-remapping.

[2.39] In the analysis of diathetical operations in this book I will use the following abbreviations, as summarised in Figure 2.1. First, grammatical expressions, i.e. actual grammatical forms as identified in traditional German grammar, are abbreviated with single letters, shown at the right side of the figure. For case-marked noun phrases I will use the easily recognisable capital letters nadg for nominative, accusative, dative and genitive respectively. As argued earlier (see Sec­tion 2.2.2) there are also prepositional phrases that express lexically determined roles. These will also be abbreviated with capital letters: l for obligatory locations and p for governed prepositional phrases.

[2.40] Lower-cased letters are used for non-argument participants in the clause: ‘p’ for non-governed prepositional phrases and ‘g’ for adnominal genitives. Adnominal genitives become relevant because in some diatheses a newly introduced participant is inherently the possessor of another participant (see paragraph 2.133). Absence of a specific role will be indicated by a dash. Lower-cased ‘a’ and ‘d’ are only used in Sec­tion 7.4 to indicate accusative and dative reflexive pronouns in referential usage. As described in much detail in that chapter, it is important to distinguish between reflexive pronouns in German that refer to a lexical role (i.e. “referential” or “real” reflexive constructions) and reflexive pronouns that mark a diathesis without referring to a separate role themselves. Only the former reflexive pronouns, those that are (doubly) marking a role, are abbreviated by lowercased ‘a’ or ‘d’.

[2.41] Besides single-letter abbreviations I will also use capitalised three-letter abbreviations for a more abstract level of analysis. As summarised at the left side of Figure 2.1, the grammatical expressions are grouped into sets of grammatical macroroles, mostly along familiar lines. However, it is crucial to realise that these macroroles are defined here as a superset of language-specific German grammatical expressions. There is no abstract metalinguistic (universal) definition assumed. The current grouping is not necessarily the best or most optimal grouping, but this grouping has emerged to be useful to organise the large diversity of diatheses in this book.

[2.42] The notion subject (sbj) is used for nominative phrases that show agreement with the finite verb.¹⁰ There are various other nominatively marked phrases in German grammar which are not included under this heading, e.g. the nominative in nominal or equational predication like der Täter in Er ist der Täter. The other case-marked governed phrases are combined as cased objects (obj). The cased objects together with the prepositional objects (pbj) form a superclass of objects. Non-governed phrases are separated in overtly expressed adjuncts (adj) and unexpressed roles omitted (ø). Although I will use this five-way distinction throughout this book (sbj, obj, pbj, adj, ø), the five steps are not equidistant. The macroroles obj and pbj are rather closely related, and likewise are adj and ø intimately linked. Collapsing these pairs results in the traditional subject-object-adjunct distinction.

[2.43] There are some indications that the group of cased objects (obj) might be fruitfully separated into core (accusative) and non-core (dative/genitive). This would simplify the analysis of case change in object chains (Sec­tion 2.7.5), the antipassive hierarchy (Sec­tion 6.7) and the case-marking of the reflexive pronoun (Sec­tion 7.3). However, in the majority of diatheses all three cases seem to behave as a uniform group, so I did not consistently pursue this separation.

[2.44] It is imperative to realise that the macroroles are defined in a language-specific way for German grammar as groupings of language-specific German expressions (e.g. adj is defined as being either a non-governed prepositional phrase or an adnominal genitive). The names that are used (e.g. “object” or “adjunct”) deliberately conjure up general cross-linguistic associations, but it remains to be seen whether similar definitions as used here are also useful for other languages. I will refrain from any cross-linguistic speculation in this context here.

2.4.2 Remapping of roles

[2.45] All diatheses in this book will be organised and categorised in various levels of abstraction using the abbreviations as summarised in Figure 2.1. The following levels of abstraction will be used to arrange the diatheses in each chapter.

[2.46] level 1: diathesis. On the lowest level, each diathesis is summarised in its own sub-subsection. The establishment of an individual diathesis is not always obvious, and each diathesis in this book is already a conscious categorisation (which could be wrong). It has actually been a voyage of discovery in the preparation of this book to decide when to consider a set of examples to be a single diathesis. Very often erstwhile single diatheses turned out to be better analysed by a separation into various different diatheses, and vice versa. Although I am rather confident in the quality of the current decisions, I expect that further refinements are necessary in the future.

[2.47] level 2: remapping pattern. The role-remapping of each diathesis is analysed using the single-letter abbreviations (nadgpl pgad‑) from Figure 2.1. A remapping is specified as an ordered listing of grammatical expressions for roles, both before and after the diathesis. For example, [na | ‑n] is a diathesis that involves two roles that are marked N(ominative) and A(ccusative) before the diathesis but unexpressed (“–”) and N(ominative) after the diathesis, respectively. Because there are many diatheses with this same pattern, this characterisation is already an (implicit) classification.

[2.48] level 3: local group. Groups of diatheses with similar remapping and similar semantics within each chapter can be grouped together as a local group. These groups are rather ad-hoc and mainly represent a useful summary to streamline the presentation. Local groups are indicated by similar names for the diatheses.

[2.49] level 4: macrorole pattern. The remapping of each local group is structurally analysed in terms of the three-letter macroroles (sbj, obj, pbj, adj, ø) from Figure 2.1. For example, the remapping from above [na | ‑n] includes both a change from n to being omitted (i.e. sbj › ø) and a change from a to n (i.e. obj › sbj). These two macrorole changes can be combined into a single macrorole pattern obj › sbj › ø.

[2.50] level 5: promotion/demotion. On the most abstract level, all diatheses are separated into chapter-subsections of either demotion or promotion (with only very few diatheses being symmetrical exchanges). Basically, each remapping is evaluated on the macrorole hierarchy (2.29) with role-remapping upwards being promotion and role-remapping downward being demotion. Note that there is a crucial additional criterion necessary, because the majority of diatheses consist of chains of two coinciding remappings (see Sec­tion 2.6 on the notion of “chains”). In such chained remappings, the largest jump on the macrorole hierarchy defines a diathesis as being demotion or promotion. When both jumps are equally large, then the diathesis is symmetric.

[2.51] For example, the diathesis in (2.30) is analysed as a remapping pattern [na | ‑n], see Sec­tion 7.5.2. This should be read as follows: there is an alternation between a clause with na arguments (nominative, accusative) and a clause with only n marking (nominative). The relative order of these letters is crucial, as the order of the roles remains fixed in this notation, e.g. the second letter on the left (a for accusative) corresponds to the second letter on the right (n for nominative). The dash on the right indicates that the corresponding n on the left is not expressed. Note that the actual linear arrangement of the letters is flexible, as long as both sides of the alternation remain in the same order, i.e. [an | n‑] would be the same remapping pattern as [na | ‑n]. The pattern [na | ‑n] is an implicit categorisation, because there are many other diatheses that have exactly the same pattern (see e.g. Sec­tions 5.5.5, 9.5.2, 10.5.12).

[2.52] Although there is a reflexive pronoun in (2.30 b), this pronoun is not included with a lower-cased ‘a’ in the remapping pattern [na | ‑n], because this reflexive pronoun does not refer to a separate role. The verb schließen ‘to close’ implies at least two different roles, the “closer” and the “closed object”, expressed as nominative and accusative in (2.30 a), respectively. In (2.30 b) only the role of “closed object” is expressed as nominative. The reflexive pronoun does not refer to any other role. I interpret the reflexive pronoun in (2.30) as a marker of the diathesis itself (see Chapter 7 for an extensive discussion), so there is actually an overt direction in the markedness from unmarked (2.30 a) to reflexive-marked (2.30 b). The vertical bar “ | ” in the middle of the remapping pattern [na | ‑n] implies this direction in markedness from left to right, i.e. left side describes the unmarked alternant and the right side the marked alternant. Reordering the remapping pattern around the vertical bar would result in a completely reversed diathesis [‑n | na].

[2.53] The diathesis in (2.30) is one of various examples of a local group that are all called “reflexive antipassive”. Other diatheses in this group include examples like (2.31) with an additional governed preposition, analysed here with the remapping pattern [nap | ‑np], see Sec­tion 7.5.6. All diatheses in this local group have the same macrorole pattern, namely obj › sbj › ø, i.e a cased object is turned into nominative subject, which is omitted (i.e. unexpressed).

[2.54] This diathesis is a combination of two different remappings obj › sbj and sbj › ø, with the first being a promotion on the macrorole hierarchy and the second a demotion. Crucially, because the demotion part (sbj › ø) is a larger jump on the hierarchy than the promotion part (obj › sbj), the complete combination is categorised as a demotion.

[2.55] So, in summary, the role-remapping of the diathesis (2.31) is categorised as summarised below. This information also informs the place in the book where this diathesis will be discussed: Reflexive voice is Chapter 7, demotion that includes the subject in the macrorole pattern is always Section 5 within each chapter, and consequently, this diathesis can be found with the heading obj › sbj › ø named “reflexive antipassive” in Sec­tion 7.5.6.

1. diathesis: reflexive antipassive+governed preposition
2. remapping pattern: [nap | ‑np]
3. local group: reflexive antipassive
4. macrorole pattern: obj › sbj › ø
5. promotion/demotion: demotion
6. voice: reflexive marking

2.5 Stacking

2.5.1 Combining diatheses

[2.56] Different clause alternations (both diatheses and epitheses) can be applied one after the other, forming stacks of diatheses and/or epitheses. The term “stacking” is introduced here explicitly in opposition to “subordinating”. Subordination leads to non-coherent multi-clause constructions, while stacks always remain coherent and thus monoclausal. My impression is that much of modern syntactic theory could be drastically simplified by strictly distinguishing between stacking and subordinating.

[2.57] Stacked diatheses can lead to convoluted role-remappings. A beautiful example of such stacking of diatheses is given by Dixon (2014: 252) for the Amazonian language Paumarí. Here, the root noki‑ ‘to see’ is transparently related to the meaning ‘to show’ through a series of derivational diatheses, viz. noki‑ ‘to see’, noki-a‑ ‘to be visible’, na-noki-a‑ ‘to become visible’, and finally na-noki-a-hi‑ ‘to make become visible’ i.e. ‘to show’.

[2.58] German does not have that many morphologically bound mechanisms for diathesis, though there are incidental examples that come close. For example, the verb liegen ‘to lie’ changes with ablaut to legen ‘to lay’ (see Sec­tion 5.6.3), which in turn can take a preverb to form be-legen ‘to cover’ (see Sec­tion 8.7.13). However, when the perspective is broadened beyond bound morphology and all different kinds of German diathesis are considered, then it turns out that stacking of diatheses is extremely widespread.

[2.59] In many cases, the different steps in a stack can be easily disentangled by carefully observing the formal marking of the diathesis. For example, the construction in (2.32 c) includes both a preverb be‑ and a reflexive pronoun sich and it turns out that these are applied in turn to make a stack of two diatheses. Starting with the verb antworten ‘to answer’ with a governed preposition auf (2.32 a), the applicative preverb be‑ changes the prepositional phrase to an accusative (2.32 b), see Sec­tion 8.8.9. Subsequently, the reflexive anticausative turns the accusative into a nominative and drops the nominative agent (2.32 c), see Sec­tion 7.5.2.

[2.60] Clause alternations are applied one after the other, i.e. the order of the alternations in a stack is of crucial importance in most examples (unordered stacks exist, but are unusual, see Sec­tion 2.6.4). Basically, a stack is just a list of clause alternations applied one after the other. Syntactically this is just a linear sequence of application, i.e. there is no branching possible with stacking. Semantically this means that each subsequent clause alternation has scope over the previous one.

[2.61] A stack can be written down using a symbol like +> to indicate the additive (+) and sequential (>) nature of the combination. The stack above (2.32) can then be analysed as: (2.32 a) +> be‑ applicative +> reflexive anticausative = (2.32 c). This notation leads to concise analyses, as shown for example in (2.33) for the difference between the sentences (2.33 a,b) and (2.33 c,d).

[2.62] With unmarked (“covert”) diatheses such stacks can sometimes be tricky to tease apart. As an example, consider the arguably somewhat artificially constructed example in (2.34) using the verb schneiden ‘to cut’. It starts off in (7.85 a) as a basic transitive construction with a nominative and accusative argument. Yet, after various twists and turns it ends up on (2.34 f) with a nominative, an accusative, a dative and an obligatory location prepositional phrase, while the original agent Arzt is not even expressed.

[2.63] Teasing this stack apart, there are five different diatheses, concurrently showing that the verb schneiden has at least five different lexeme-specific roles. As defined in Sec­tion 2.2.2, each role that appears as a case-marked constituent in at least one diathesis is a lexeme-specific role, and all of the following participants are case-marked in the stack of diatheses (2.34):

• the cutter Arzt ‘physician’
• the cut object Nagel ‘nail’
• the possessor of the cut object Patient ‘patient’
• the result of the cutting Schlitz ‘groove, slit’
• the instrument doing the cutting Fräser ‘milling cutter’

[2.64] The five diatheses (and the corresponding role-remappings) are the following:

• (2.34 b), in antipassive: changing the cut object Nagel from accusative to prepositional object, see Sec­tion 6.7.9.4.
• (2.34 c), possessor raising: changing the possessor of the cut object Patient from adnominal genitive into dative, see Sec­tion 6.8.12.
• (2.34 d), object exchange: adding a new accusative object Schlitz as the result of the cutting, see Sec­tion 6.8.8.
• (2.34 e), adjunct addition: adding a syntacticaly optional instrument Fräser, see Sec­tion 6.2.4.
• (2.34 f), instrument anticausative: turning the instrument Fräser from prepositional phrase to nominative, see Sec­tion 6.5.5.

2.5.2 Fixed stacks

[2.65] There are a few examples of diatheses that look like stacks of two diatheses, but on closer inspection it turns out that the intermediate construction does not exist. A few major examples of such fixed stacks are exemplified below.

[2.66] There is an infamous anticausative diathesis that needs a reflexive pronoun, which is attested for a large, but restricted group of verbs like schließen ‘to close’ (2.35 a,b), see Sec­tion 7.5.2. A completely different group of verbs also has an anticausative diathesis with a reflexive pronoun, but only with an additional evaluative adverbial. This is for example attested with verkaufen ‘to sell’ (2.35 c,d), see Sec­tion 9.5.2. In this case, the diathesis is marked by both the reflexive pronoun and the presence of an adverbial, and neither is possible without the other. Such a combination of two obligatorily co-occurring formal marking strategies is called a fixed stack.

[2.67] Various diatheses between a verb, like fassen ‘to grasp’ (2.36 a), and its pre­verb-alternant, like befassen ‘to be concerned with’ (2.36 b), additionally need a reflexive pronoun, see Sec­tion 8.7.4. So here we have a fixed stack of a reflexive pronoun and a preverb together that mark the diathesis.

[2.68] Also some light verb alternations show fixed stacks. For example, there is a very widespread causative diathesis using the light verb lassen with an infinitive (2.37 b), see Sec­tion 11.6.2. Additionally, the combination of lassen+In­fi­ni­tiv and a reflexive pronoun leads to a passive alternation (2.37 c), which does not make sense as being derived from the causative (2.37 b). It seems better to consider the combination of lassen+Infinitiv+Reflexiv as a fixed stack, see Sec­tion 11.5.5.

2.6 Chaining

2.6.1 Beyond solitary remapping

[2.69] Many diatheses just remap a single role. Such diatheses are called isolated diatheses here. However, there are also many diatheses in which more than one role is remapped. I distinguish the following kinds of combined role-remappings, of which only the first is frequently attested.

• chained diathesis (Sec­tion 2.6.2): Two roles change their formal marking, forming a chain in which one role changes its form from X to Y, while the other role changes its form from Y to Z. This results in a chain (x › y › z).
• multi-chained diathesis (Sec­tion 2.6.3): Three (or possibly even more) roles change their formal marking, forming a longer chain of connected changes.
• disjunct diathesis (Sec­tion 2.6.4): Two or more roles change their formal marking, with no overlap between marking.

[2.70] Chained diatheses are surprisingly frequent in German, and my impression is that this pervasiveness extends to many other languages beyond German. In a chained diathesis the result of the first remapping is the start of the second remapping. This can be conceptualised as a “push” or “pull” chain in which one remapping induces another. The prevalence of such chains is probably caused by two general tendencies of language structure, namely distinctness and default marking. These tendencies are formulated here as hypotheses for language structure in general, beyond the specifics of German.

[2.71] First, the tendency for distinctness causes language to disprefer multiple constituents with the same structure in a single clause. For example, the German languages tends to prevent two accusatives in the same clause. In effect, if a diathesis would gives rise to such a duplication, then the duplicated constituent is preferably “pushed” out to another kind of marking. Second, the principle of default marking induces languages to mark at least one of its constituents as the “default” in each clause. For example, in German the nominative subject has to be present in almost every clause. As a result, if a diathesis removes this preferred constituent, then another constituent is typically “pulled” into this kind of marking. It remains to be further investigated whether these two forces really exist, and whether the two tendencies can be teased apart.

2.6.2 Chained diatheses

[2.72] In German, chained diatheses typically occur when the nominative subject is involved in the diathesis. There can only be a single nominative subject in a German clause, and it is highly unusual to have a sentence without a nominative subject. This implies that any diathesis involving the nominative subject typically includes two remappings, namely one from something else to nominative and a second remapping of the erstwhile nominative to something else.

[2.73] A prototypical example of a chained diathesis involving the nominative subject is the werden passive (2.38). Here, the erstwhile accusative Kuchen ‘cake’ is turned into a nominative, while the erstwhile nominative Lehrling ‘apprentice’ is removed (or optionally retained as a von prepositional phrase). So, we have a chain consisting of the role-remappings obj › sbj and sbj › adj.

[2.74] Diatheses without involvement of the nominative subject are more flexible, in that both isolated and chained diatheses are common. A typical example of a chained diathesis is an object exchange induced by the preverb be‑ (2.39). In this example, a prepositional phrase für ihre Freundin ‘for her friend’ is remapped to an accusative (adj › obj) while the erstwhile accusative Essen ‘food’ is turned into a prepositional phrase (obj › adj).

[2.75] Among the chained diatheses there is a group of frequently recurring remapping patterns. Because of their frequency, it is highly useful to give them specific names. Such names are widespread in the literature, e.g. anticausative for obj › sbj › ø or passive for obj › sbj › adj. A survey of the various names used in this book will be pursued in Sec­tion 2.7.

2.6.3 Multi-chained diatheses

[2.76] multi-chained diatheses consist of combinations of more than two role-re­map­pings that occur in a sequence. This occurs frequently as the result of a stack of multiple diatheses, but only very rarely in a single diathesis. As an example arising from a stack of multiple diatheses consider taking a verb like lesen ‘to read’ (2.40 a) and applying a stack of two diatheses (2.40 b,c). This leads to a chain of three role-remappings. First, the preverb diathesis with vor‑ (2.40 b) leads to the addition of a dative argument dem Jungen, i.e. a role-remapping ø › obj, see Sec­tion 8.8.8. On top of that, the bekommen dative passive (2.40 c) promotes this dative to subject and removes the original subject, i.e. a role-remapping obj › sbj › ø, see Sec­tion 10.5.21. Combined, these two diatheses lead to a multi-chained role-re­map­ping ø › obj › sbj › ø.

[2.77] Such multi-chained diatheses that are the result of diathesis-stacking are widespread. However, I know of only two diatheses with a multi-chain that cannot be decomposed into a stack of separate diatheses. Both these “fixed” multi-chain diatheses appear to occur with just a few idiosyncratic verbs, so this phenomenon really seems to be dispreferred in German.

[2.78] First, the preverb diathesis from erben ‘to inherit’ to enterben ‘to disinherit’ (2.41), see Sec­tion 8.6.14, contains three linked role-remappings for (i) the originator of the inheritance Vater ‘father’ (adj › sbj), (ii) the receiver of the inheritance Junge ‘boy’ (sbj › obj) and (iii) the inheritance Schreibtisch ‘desk’ (obj › ø).

[2.79] Second, the verb schmecken ‘to taste’ (2.42), see Sec­tion 6.5.6, allows for two different constructions with three linked role-remappings for (i) the tasted substance Pfefferminze ‘peppermint’ (obj › adj), for (ii) the tasted dish Suppe ‘soup’ (adj › sbj) and for (iii) the taster Koch ‘cook’ (sbj › ø).

2.6.4 Disjunct diatheses

[2.80] disjunct diatheses consist of a combination of multiple role-remappings that are not linked to each other. Just as with the multi-chained diatheses from the previous section, disjunct diatheses regularly occur as the result of stacking of diatheses. In contrast, they are very rare in individual diatheses.

[2.81] When multiple diatheses are stacked, i.e. when they are applied sequentially on top of each other, they are sometimes structurally independent (and thus unordered). For example, the verb waschen ‘to wash’ (2.43 a) can be used in a object-exchange diathesis (2.43 b) in which the role of washee Hemd ‘shirt’ is turned from an accusative into a location (obj › pbj) and a new accusative object is introduced for the role of the object to be removed Fleck ‘stain’ (ø › obj), see Sec­tion 6.8.8. Independent of this chained diathesis, the possessor of the object Nachbar ‘neighbour’ can be raised to genitive (2.43 c), i.e. a possessor applicative (adj › obj), see Sec­tion 6.8.13.

[2.82] There are only a few incidental examples of such disjunct diatheses without stacking. The following four examples all only occur with a very limited number of verbs. First, the verb deuten can be used both to mean ‘interpret’ (2.44 a) and ‘forebode’ (2.44 b) with a rather transparent connection between the two. However, the role-remappings are quite complex, see Sec­tion 6.5.11.

[2.83] Second, some preverbs lead to disjunct diatheses, like with schweigen ‘to remain silent’ and verschweigen ‘to conceal’ (2.45), see Sec­tion 8.8.16.

[2.84] Further examples are a few verbs of naming like schimpfen ‘to scold’ (2.46), see Sec­tion 6.8.6. The disjunct diathesis in (2.47) is less clear, as it might be better analysed as a stack, see Sec­tion 6.5.3.

[2.85] The only more widespread disjunct diathesis is the caused-motion construction that can arise with some apparently intransitive verbs like schwitzen ‘to sweat’ (2.48). This diathesis introduces two roles at once: a result of the sweating Fleck ‘stain’ and an obligatory location of the result Hemd ‘shirt’, see Sec­tion 6.8.3.

2.7 Naming

2.7.1 Names for macrorole patterns

[2.86] Throughout the introductory chapters I have used various names for diatheses, like passive, antipassive, applicative or causative. These names have a long history in the typological grammatical literature (cf. Mel’čuk 1993; Wunderlich 1993; Wunderlich 2015; Dixon & Aikhenvald 2000; Dixon 2014; Haspelmath & Müller-Bardey 2004; Kulikov 2011; Malchukov 2015: 96ff.; Zúñiga & Kittilä 2019). Although I have been using these terms as if their meaning is clear, this is often far from the truth. Many different terms and definitions have been proposed in the literature, and different terms have at times been used for the same phenomena. For example, the original proposal for the term “antipassive” is already 50 years old (Silverstein 1972: 395), but the same phenomenon is also known as deaccusative (Geniušė 1987: 94) or antiapplicative (Haspelmath & Müller-Bardey 2004: 1132; Scheibl 2006: 371). Conversely, the term antipassive is also attested referring to a slightly different phenomenon of the drop of an object (Scheibl 2006: 372–373).

[2.87] In this section I will describe in more detail how these names are used and defined in the current book about German diatheses. The names for diatheses will here always refer to a macrorole pattern, i.e. to the highly abstract classification of a diathesis in terms of sbj, obj, etc. as defined in Sec­tion 2.4.2. For example, the term “anticausative” will be used as a name for the macrorole pattern obj › sbj › ø. Such macrorole patterns are strictly defined here in a language-specific way for German, so care should be taken when applying the same names to different languages.

[2.88] One widespread term that I will avoid is the term “middle” (and likewise the Latinate equivalent term “medium”). This term for a diathesis is already attested as μεσότης in the oldest known Greek grammatical text, the τέχνη γραμματική of Dionysius Thrax, and it has become a mainstay in the grammatical literature ever since.¹¹ Thrax writes: διαθέσεις εἰσὶ τρεῖς, ἐνέργεια, πάθος, μεσότης “there are three diatheses, active, passive and middle” (Uhlig 1883: 48). The phenomena that are called “middle” in the literature are highly variable, and there is no consensus about what kind of diathesis this term is supposed to designate, other than something that is neither active nor passive (see Zúñiga & Kittilä 2019: 168–177 for a thorough summary of the complex philological history of the term middle/medium). Such a broad and ill-defined term is not useful for a detailed analysis of the large variety of attested role-remappings in German.¹² It is of course possible to add yet another definition for the term “middle” to the wide variety of already existing uses (e.g. Inglese 2023), but to me this only increases the confusion.

[2.89] The discussion about the different names for macrorole patterns will be split into four parts. First, the next two sections will present names for diatheses involving the nominative subject. Subsequent sections will discuss diatheses not involving the subject. In both discussions, a central distinction will be made between isolated diatheses and chained diatheses (cf. Sec­tion 2.6).

2.7.2 Isolated subject diatheses

[2.90] Isolated diatheses that involve a nominative subject do not show much variation in German. The most widespread kind is the drop of the subject (sbj › ø), i.e. the complete removal of the role marked as nominative subject without any further accompanying role-remapping or reintroduction of a new subject. This is typically attested with intransitive verbs: after removing the single available role, there is no other role introduced to fill the structural subject position. Semantically, such diatheses put the focus on the activity as described by the verb itself, so I propose to call them insubjective diatheses. Note that there is a strong tendency for every German sentence to formally have a nominative subject with verb agreement. Consequently, such insubjective diatheses regularly (but not always) result in the presence of a valency-simulating nominative pronoun es (see Sec­tion 2.2.3).

[2.91] An insubjective diathesis is attested with verbs like stinken ‘to stink’ (2.49), see Sec­tion 5.5.1. In a sentence like es stinkt the pronoun es can of course simply be an anaphor, like in (2.49 b). In such a sentence, the role of “stinker” is still present and there is no diathesis at all. However, in other contexts (2.49 c) the verb stinken is used without implied subject. This is typically attested in contexts in which some odour is attested, but the originator is not known.

[2.92] Another example of a insubjective diathesis is illustrated with the verb leben ‘to live’ (2.50), see Sec­tion 9.5.1. Many such intransitive verbs can be used without a subject in a habitual sense, but this is only possible with an obligatory adverbial qualification like gut (2.50 b,c).

[2.93] Also the so-called impersonal passive consisting of werden+Partizip (2.51), see Sec­tion 10.5.1, is an example of a insubjective diathesis, in this case even without any valency-simulating es.

[2.94] A different kind of isolated subject diathesis is subject demotion of the nominative subject to a prepositional phrase. An example is the geben+zu‑In­fi­ni­tiv (2.52), see Sec­tion 12.5.4. In this diathesis, the subject is demoted to an optional non-governed prepositional phrase (sbj › adj). The demotion is the only role-remapping that is happening in this diathesis, so I propose to call such a diathesis a desubjective.

[2.95] The other isolated subject diatheses are only attested in incidental examples in German, like a subject demotion to a governed preposition (sbj › pbj) with fehlen shown in (2.53), see Sec­tion 6.5.2.

[2.96] Isolated subject addition (ø › sbj) is very rare in German, partly because it would need an unmarked construction without any subject to start off with. A possible example is the addition of a subject that seems possible with some weather verbs like donnern ‘to thunder’ (2.54), see Sec­tion 5.6.1.

2.7.3 Chained subject diatheses

[2.97] Chained diatheses that involve the nominative subject are widespread in German (in contrast to the infrequent occurrence of isolated diatheses as discussed previously). Figure 2.2 presents an overview of the different terms that I will use for these diatheses. The bold-faced terms are used for widely attested diatheses, while the other kinds of diatheses are only incidentally found. There is currently no evidence in German for the existence of the remappings that are left empty in the figure. There appears to be a preference for various kinds of demotion (i.e. the upper right corner of the figure), which fits nicely with the known typological preference of German for anticausative constructions (Haspelmath 1993: 101; Nichols, Peterson & Barnes 2004: 189).

[2.98] The upper right part of Figure 2.2 are demotions, the lower left part are promotions, and on the diagonal are examples of symmetrical diatheses. I will discuss all types in this order.

2.7.3.1 Demotions

[2.99] The most extreme kind of demotion is an anticausative (obj › sbj › ø). The typical characteristic of an anticausative is the complete removal of the nominative subject that is the causer of the action/state of the clause. Filling the syntactic gap, a case-marked argument (typically the accusative) is promoted to subject. This is a widespread kind of diathesis. An example is the reflexive anticausative with verb like schliessen ‘to close’ (2.55), see Sec­tion 7.5.2.

[2.100] Very similar to an anticausative is the passive (obj › sbj › adj). The main difference between the two (a distinction which is often difficult to delimit) is that for a passive the original subject is still implied and can optionally be overtly expressed (2.56). In contrast, for an anticausative the original subject is completely removed and a phrase like by itself can typically be added. As an example of a passive diathesis in (2.56) is the bekommen Rezipientenpassiv in which a dative is promoted to subject Sec­tion 10.5.21

[2.101] A conversive (obj › sbj › pbj) looks similar to a passive, except that the prepositional phrase is a lexically governed preposition, so it has a more object-like grammatical status. An example is the verb empören ‘to appall’ (2.57 a) with the reflexive diathesis sich empören über ‘to be outraged about’ (2.57 b,c), see Sec­tion 7.5.7. The term conversive is adapted from Kulikov (2011: 380).

[2.102] For the next diathesis, I propose the term fabricative (pbj › sbj › ø) based on Lat. fabrica ‘plan, trick, workmanship’. This term is used for a diathesis in German in which a fabricated product can be expressed either as a governed prepositional phrase or as a nominative subject. This diathesis occurs for example with various verbs of emotional interactions like überraschen ‘to surprise’ (2.58 a), see Sec­tion 6.5.7. To understand this diathesis, a distinction is needed between the role of the “fabricator” (here: Lehrer, ‘teacher’) and the role of the “fabricated product”, which induces the emotion (here: Aufgabe, ‘assignment’). The mit prepositional phrase that expresses the fabricated product in (2.58 a) is a governed preposition (2.58 c). The diathesis promotes this fabricated product to nominative subject and the fabricator is removed from the expression (2.58 b). The experiencer in the accusative mich remains unchanged.

[2.103] A similar kind of diathesis will be called conciliative (adj › sbj › ø) based on Lat. conciliator ‘intermediary, mediator’. In a conciliative an external object (typically an instrument) is promoted to subject (2.59), see Sec­tion 6.5.5. The conciliative and fabricative in German both regularly use a prepositional phrase with mit, but the grammatical status is clearly different. The mit phrase in a conciliative is an optional adjunct (2.59), while the mit phrase in a fabricative is a governed preposition (2.58). This grammatical difference is paralleled by a functional difference in the role that is promoted to subject: a conciliative concerns a (typically tangible) instrument that is used by an agent, while a fabricative promotes a (typically intangible) creation that is produced by the agent.

2.7.3.2 Promotions

[2.104] The most widespread promotion to subject attested in German is the diathesis with role-remapping ø › sbj › obj, called novative here (based on Lat. novare ‘renew, refresh’). This role-remapping is best known as “causative”, but this semantic characterisation does not hold for all examples of this diathesis. Various other novative diatheses exist in which the new nominative is not a causer but an experiencer, opinionator, permitter or assistant.

[2.105] Semantically, the most widespread kind of novative adds a new causer to the construction, like with the diatheses between brennen ‘to burn (intransitive)’ and verbrennen ‘to burn (transitive)’ (2.60), see Sec­tion 8.6.1. Such a diathesis is aptly called a causative.

[2.106] The sehen+In­fi­ni­tiv diathesis (2.61), see Sec­tion 11.6.6, adds a new nominative subject and the old nominative is turned into an accusative. This diathesis is thus structurally an example of a novative (ø › sbj › obj). However, the newly added nominative is not a causer. The new role is better described as an experiencer, so this diathesis can semantically be called an experientive. Similar constructions are also attested with light-verbs hören, fühlen, and spüren.

[2.107] The finden+Partizip diathesis (2.62), see Sec­tion 10.6.4 also adds a new nominative subject while the old nominative is turned into an accusative. The role of the new nominative is best characterised as somebody having an opinion, so this diathesis can semantically be called an opiniative. The main verb is typically a patientive intransitive predicate like scheitern, ‘to fail’, see Sec­tion 10.2.5. Similar constructions also exist with light verbs wissen, sehen and glauben.

[2.108] The lassen+In­fi­ni­tiv diathesis (2.63), see Sec­tion 11.6.2 is also structurally a novative (ø › sbj › obj). This diathesis has multiple possible interpretations, among them also a causative reading (2.63). However, in the example in (2.64) the newly added nominative is allowing the action to happen, not causing it, so this diathesis can semantically be called a permissive. This second interpretation typically happens with agentive intransitive predicates like schlafen ‘to sleep’, see Sec­tion 10.2.5. However, note that in both examples the other interpretation is also possible, albeit only is specially crafted contexts.

[2.109] Finally, the lehren/helfen+In­fi­ni­tiv diathesis (2.65), see Sec­tion 11.6.12, is a novative in which the role of the new subject is more of an assistant than a real causative. Therefor it is called here an assistive novative. Note that both lehren and helfen can also be used with zu‑In­fi­ni­tiv, but then the constructions are not coherent, so those constructions are not included among the diatheses.

[2.110] The novative with extended demotion (ø › sbj › pbj) is extremely rare in German. The name is adapted from Kulikov (2011: 388) to denote a diathesis in which the demotion accompanying the novative is not just sbj › obj but sbj › pbj. The diathesis between freuen ‘to be pleased’ and erfreuen ‘to please’ (2.66) might be an example because mit is a governed preposition (2.66 c), see Sec­tion 8.6.11.

[2.111] Slightly more widespread, a novative with extreme demotion (ø › sbj › adj) is a novative diathesis that almost completely removes the erstwhile subject. This is attested in an interesting group of constructions using light verbs like finden with a participle and a transitive main verb like aufheben ‘to preserve’ (2.67), see Sec­tion 10.6.8. With this diathesis, there is a new opinionator introduced, just like with the opiniative above (see paragraph 2.107). However, the erstwhile nominative subject is now demoted to an optional prepositional phrase.

[2.112] The remaining types of promotions are extremely rare. A reversed passive (adj › sbj › obj) demotes the subject to object and at the same time promotes a new subject from an erstwhile adjunct role. An example in German is the diathesis from erben ‘to inherit’ to enterben ‘to disinherit’ (2.68 a,b), see Sec­tion 8.6.14. This is semantically very close to a causative ø › sbj › obj in which the newly introduced causer can sometimes be expressed as an adjunct (2.68 c,d). This affinity between a reversed passive and a causative is reminiscent of the affinity between a passive and an anticausative. In both pairs, the difference amounts to a switch between the closely related macro-role of an optional adjunct (adj) and being completely unexpressed (ø).

[2.113] Finally, a reversed conversive (pbj › sbj › obj) differs from a reversed passive in that the prepositional phrase is a lexically governed preposition, as can be identified by a possible da(r)+Präposition, dass paraphrase. This is for example attested with the diatheses between staunen über ‘to marvel’ and erstaunen ‘to amaze’ (2.69), see Sec­tion 8.6.12.

2.7.3.3 Symmetrical subject diatheses

[2.114] Completely symmetrical diatheses involving the subject are rare in German. A perfectly symmetrical inversive (obj › sbj › obj) is a diathesis that switches subject and object. This term is proposed by Malchukov (2015: 99–100) inspired by the so-called “inverse” marking found in Algonquian languages. An inversive diathesis is designated as a “symmetric conversive” by Kulikov (2011: 380). An example of an inversive is the diathesis between wundern ‘to puzzle’ and bewundern ‘to admire’ (2.70), see Sec­tion 8.9.5.

[2.115] Much more widespread in German are diatheses in which a nominative/accusative construction is inverted into a dative/nominative construction. This is for example attested for the bleiben+zu‑In­fi­ni­tiv diathesis (2.71), see Sec­tion 12.9.1. Because dative and accusative are both classified here as obj, this counts as an inversive diathesis. However, when a separation between core case (accusative) and non-core case (dative/genitive) would be pursued (see paragraph 2.43), then this diathesis would be an example of demotion. There are two remappings, namely down from sbj to non-core-obj and up from core-obj to sbj. When non-core is taken as being lower on the macrorole hierarchy (2.29) then the biggest jump is the jump down, which is the definition of demotion (see Sec­tion 2.4.2). Instead of adding a completely new set of categories I propose to simply split inversive into two subtypes and call this phenomenon a demoted inversive.

[2.116] The opposite promoted inversive promotes a dative/genitive into a nominative subject, and demotes the erstwhile nominative to an accusative. This is illustrated with the haben+In­fi­ni­tiv diathesis in (2.72), see Sec­tion 11.9.2.

[2.117] At the other extreme, a commutative (ø › sbj › ø) completely removes the old subject and introduces a completely new role as subject. I propose this term on the basis of Lat. commutare ‘exchange, replace’. A German example of such a diathesis is the geben+Partizip construction (2.73), see Sec­tion 10.9.3. Note that the subjects in the two sentences do not have to be the same participant.

[2.118] The two other symmetrical diatheses in between the two extremes are even rarer. A preposition inversive (pbj › sbj › pbj) is similar to an inversive, but the exchange is with a governed preposition. This is arguably attested in the diathesis between strahlen ‘to shine’ and erstrahlen ‘to gleam’ (2.74), see Sec­tion 8.9.6.

[2.119] Finally, an example of an adjunct commutative (adj › sbj › adj) is possibly attested with the verb wimmeln ‘to swarm’ (2.75), see Sec­tion 6.9.1.

2.7.4 Isolated object diatheses

[2.120] The situation with object diatheses is reversed from the previously discussed subject diatheses. With object diatheses, isolated diatheses are much more widespread and they occur with a wide variety of role-remappings, see Figure 2.3. In contrast, chained object diatheses are less widespread and can mostly be analysed as a combination of multiple isolated diatheses.

[2.121] The top right diatheses in Figure 2.3 are demotions, while the bottom left ones are promotions. The bottom right of the figure is left completely empty because these remappings are not diatheses anymore, but simply optional marking. There is a strong tendency for object demotions in German to be either unmarked, or marked by reflexive pronouns, while the object promotions are typically marked by preverbs or resultative preverbials.

[2.122] The exception to this generalisation are the so-called locative and delocative diatheses. With those, promotions (locatives) are formally unmarked, while demotions (delocatives) are typically marked by preverbs or resultative preverbials. A possible explanation for this apparent markedness reversal is that the adding or removing of location phrases should not be seen as a change in valency (“diathetical operation”), but as the marking of the diathesis itself (“voice”). This would be a parallel to the addition/removal of directionals (see Sec­tion 9.2.5).

[2.123] I will discuss the different role-remappings from Figure 2.3 in four subsections. First, I will summarise the various kinds of applicatives and antipassives (mid left and mid top), then the objectives and deobjectives (top right and left bottom), followed by locative and delocative diatheses (mid bottom and mid right), and finally the symmetrical exchanges (on the diagonal).

2.7.4.1 Applicatives & antipassives

[2.124] Applicatives and antipassives are very similar, though reversed. applicatives (adj › obj) change a prepositional phrase into a case-marked phrase, while antipassives (obj › adj) convert a case-marked phrase into a prepositional phrase. Given this affinity, instead of antipassive it might be better to call such remappings “antiapplicative” (e.g. Haspelmath & Müller-Bardey 2004: 1132) or “deapplicative” (in line with the other names below).

[2.125] By removing or adding an object, applicatives and antipassives change the transitivity of the sentence. However, because case marking in German is nominative/accusative aligned, changes in transitivity are not reflected in the marking of the subject. This is crucially different from languages with ergatively aligned case marking, in which antipassives also include a change in the marking of the subject, namely from ergative to absolutive (and vice versa with applicatives). Terminologically, these two situations might be distinguished by using the term “deapplicative” for nominative/accusative languages and reserve “antipassive” for ergative/absolutive languages. I decided against that distinction and the term “antipassive” will be used throughout in this book with this explicit caveat.

[2.126] Applicatives occur frequently with the addition of a preverb, like in the alternation between steigen auf ‘to climb’ and besteigen ‘to mount’ (2.76), see Sec­tion 8.8.9.

[2.127] Antipassives in German are often unmarked (see further below), but an example of an antipassive with a clear markedness direction is the alternation between treffen ‘to meet’ and reflexive sich treffen mit ’to meet with (2.77), see Sec­tion 7.7.3.

[2.128] The object of applicatives and antipassives is typically an accusative, but datives can also be targeted. An example of a dative applicative is the alternation between stammen aus ‘to hail from’ and entstammen ‘to be descended from’ (2.78), see Sec­tion 8.8.14. An example of a dative antipassive is the covert alternation of berichten ‘to report’ (2.79), see Sec­tion 6.7.12.

[2.129] In the discussion of diatheses in this book I consistently distinguish governed applicatives (pbj › obj) and governed antipassives (obj › pbj) when the prepositional phrase is a governed preposition (see Sec­tion 6.2.1). An example of a governed applicative is the diathesis between arbeiten an ‘to work on’ (with a governed preposition an) and bearbeiten ‘to edit, adapt’ (2.80), see Sec­tion 8.8.10. An example of a governed antipassive is the diathesis between beklagen ‘to lament’ and sich beklagen ‘to complain’ (with a governed preposition über) (2.81), see Sec­tion 7.7.4. However, the differentiation between the governed and non-governed applicative/antipassive does not currently allow for any promising semantic or structural generalisations, so this differentiation might grammatically be unnecessary to explain German sentence structure.

[2.130] There are a many diatheses with a role-remapping between adjunct and object that do not have any overt indication of a direction. Without explicit marking it is difficult to decide whether such diatheses are cases of (applicative) promotion (adj › obj) or (antipassive) demotion (obj › adj). For the sake of organisation in this book I classify such covert alternations on the basis of (debatable) semantic intuitions and parallels to other overtly marked diatheses.

[2.131] Most covert diatheses with an alternation between prepositional phrases and case-marked arguments are classified here as antipassive, like in the alternation between schießen auf ‘to aim at’ and schießen ‘to shoot’ (2.82), see Sec­tion 6.7.9. This is also widespread with datives (2.83), see Sec­tion 6.7.12. In such examples, I judge the case-marking to be more basic than the prepositional phrase.

[2.132] In contrast, there is a widespread alternation between datives and beneficiary für prepositional phrases (2.84) that I classify as an applicative, see Sec­tion 6.8.10. In this example the beneficiary dative seems to be the derived construction.

[2.133] There is a further kind of covert diathesis with a dative object, conventionally called possessor raising. In such diatheses there is an alternation between a possessor (typically expressed as an adnominal genitive) and a dative (2.85). The dative can alternate with the possessor of a nominative subject (see Sec­tion 5.8.3), an accusative object (see Sec­tion 5.8.4) or an obligatory location (see Sec­tion 6.8.12). Following widespread convention, I classify these diatheses as promotion (adj › obj)

[2.134] These two covert kinds of dative applicative (viz. beneficiary and possessor applicative) are semantically and structurally clearly distinct. The datives that show a possessive alternation (2.85) are semantically experiencers. In contrast, datives that alternate with für prepositional phrases (2.84) are semantically beneficiaries. In especially crafted context it might be possible to evoke either reading for the same sentence (2.86).

2.7.4.2 Objectives & deobjectives

[2.135] A deobjective diathesis (obj › ø) is a diathesis that drops an object, i.e. a role cannot be expressed anymore (the term is taken from Haspelmath & Müller-Bardey 2004: 1131). A deobjective drop is illustrated in (2.87) with an alternation from kaufen ‘to buy’ to einkaufen ‘to shop’, see Sec­tion 8.7.2 for an extensive discussion.

[2.136] A special variant of a deobjective occurs with verbs that apply to the body, like verbrennen ‘to burn’ (2.88). In such constructions, a reflexive pronoun is necessary. This diathesis is called endoreflexive (Haspelmath 1987: 27–28), see Sec­tion 7.7.1 for an extensive discussion.

[2.137] An objective diathesis (ø › obj) is a diathesis that adds a new object, i.e. a completely new role is introduced in the form of an object. An example of an overtly marked object addition is the alternation from zaubern ‘to perform magic’ to verzaubern ‘to enchant’ (2.89). In this example the new object is simply an added patient to an erstwhile intransitive action. Such object additions are frequently attested with preverbs like ver-, see Sec­tion 8.8.1.

[2.138] A semantically special kind of diathesis introduces a new added result object. Such an objective diathesis adds an object that is the result of performing the activity described by the predicate. An overtly marked example is presented in (2.90) with the diathesis between arbeiten ‘to work’ and the inherent reflexive sich etwas erarbeiten ‘to acquire something through work’, see Sec­tion 8.8.6. The result of the work is added as an object in (2.90 b).¹³ Here I consciously avoid the term “resultative” for this phenomenon to avoid confusion. First, I already use the term “resultative” in this book for a special class of preverbial adjectives (see Sec­tion 9.2.6). Second, the term “resultative” is also frequently used in the literature for an aspectual concept, namely to indicate a special kind of state induced as the result of performing the predicate (e.g. Nedjalkov 1988).

[2.139] Objectives and deobjectives are frequently attested without any overt marking (cf. ambitransitive/labile verbs), and in such “covert” diatheses it is difficult to establish a direction. As already noted above, for the sake of organisation in this book I classify such covert alternations on the basis of (often debatable) semantic intuitions and parallels with other overtly marked diatheses. For example, the verb stören ‘to disturb’ (2.91) can be used both with and without an accusative object, see Sec­tion 5.7.1. This is classified here as a deobjective diathesis. Such unmarked object drops are also attested with datives, see Sec­tion 5.7.4, and with governed prepositions, see Sec­tion 6.7. The dropping of an object is also often used to put the focus on the action itself, but then it is typically attested with an adverbial, see Sec­tion 9.7.1 for an extensive discussion.

[2.140] In contrast, the verb stottern ‘to stutter’ is classified here as an example of a covert object addition (2.92), although there is no formal differentiation from the previous example of a covert object drop (2.91). The intuition is that stottern is basically intransitive (and any accusative object is thus added), while stören is basically transitive (and any missing object is thus dropped). Correlated with this proposed difference is the fact that covert object addition with stottern has an added result interpretation (2.92 b). However, it remains to be seen whether there is really a substantive difference between these two kinds of verbs (see Sec­tion 5.8.1 for an extensive discussion).

2.7.4.3 Locatives & delocatives

[2.141] A locative diathesis (ø › pbj) is a diatheses that adds an obligatory location phrase to the clause. For example, the transitive befehlen ‘to order’ marks the ordered person as an accusative (2.93 a). With a (directional) locative phrase an die Front ‘to the frontline’ the sentence obtains a caused-motion reading (2.93 b), see Sec­tion 6.8.4. Note that there is no profound association between such a locative diathesis and the widespread phenomenon of a locative case. Both terms simply use the term “locative” to describe the fact that the marking of location is concerned.

[2.142] Even more noteworthy, such a caused-motion diathesis is also possible with many intransitive verbs like schwitzen ‘to sweat’ (2.94 a). With such verbs, a locative diathesis not only adds a location, like in mein Hemd ‘in my shirt’, but also an added-result accusative object, like einen Fleck ‘a stain’ (2.94 b), see Sec­tion 6.8.3.

[2.143] The reversal of a locative diathesis is a delocative diathesis (pbj › adj). In such a diathesis an obligatory location loses its obligatory status and is often completely dropped. An example of such a diathesis is shown in (2.95) with the alternation between stecken ‘to put into’ and verstecken ‘to hide’. The verb stecken needs an obligatory location (2.95 a,b). Such an obligatory location is classified here as a pbj prepositional object (see Sec­tion 2.2.2). The situation is different with the verb verstecken. With this verb the location is an adj optional adjunct and can be left out (see Sec­tion 8.7.11 for an extensive discussion).

2.7.4.4 Symmetrical object diatheses

[2.144] Symmetrical object diatheses are rare in German. A case change (obj › obj) is illustrated in (2.96) by the alternation between folgen ‘to follow’ (with dative) and verfolgen ‘to chase’ (with accusative), see Sec­tion 8.9.2.

[2.145] A governed preposition change (pbj › pbj) does occur in German, but such diatheses have not been explicitly collected in this book. Possible examples are arbeiten an ‘to work on’ (2.97 a) changing into sich durcharbeiten ‘to work through’ (2.97 b) or sorgen für ‘take care of’ (2.97 c) changing into sich sorgen um ‘to worry about’ (2.97 d).

[2.146] An adjunct change (adj › adj) is, according to my definitions, not a diathesis at all, as adjuncts are not lexically specific. However, the change between a possessor dein ‘your’ (2.98 a) and a non-governed prepositional phrase von dir ‘from you’ (2.98 b) can be seen as as a borderline examples, see Sec­tion 6.9.3.

2.7.5 Chained object diatheses

[2.147] Chains of object diatheses (i.e. chains with the object in the middle of the chain) can always be interpreted as a combination of two isolated object diatheses from the previous section. However, not all theoretically possible combinations are attested (see Figure 2.4). The most frequently attested chained object diatheses are the highlighted variants of object exchange (see Sec­tions 2.7.5.1-2.7.5.4). A few incidental examples of chained case change are also attested (see Sec­tion 2.7.5.5).

2.7.5.1 Object exchange

[2.148] The highlighted diatheses in Figure 2.4 are collectively called object exchange because as part of the role-remapping the accusative marking is exchanged from one role to another. These diatheses are used with verbs that involve some kind of part/whole relation between the two roles involved. A typical example (2.99) is the diathesis between schmieren ‘to smear’ and beschmieren ‘to spread’ (discussed in detail in Sec­tion 8.7.13). In this example an auf prepositional phrase turns into a new accusative, while the old accusative is turned into a mit prepositional phrase. So, syntactically the role marked as an accusative object is exchanged from Salbe ‘ointment’ to Wunde ‘wound’. Semantically, the Wunde is the “whole” to which the Salbe is applied.

[2.149] Different variants of such object exchange show an astonishingly strong correlation between syntactic structure and semantic interpretation. Basically, promotions have the effect that the new object role is a part of the old object role (i.e. the new object is a meronym), while demotions have the reverse effect in that the new object role encompasses the old object role (i.e. the new object is a holonym). To appreciate this generalisation it is important to recall how demotions and promotions are defined for chained diatheses. This definition is not trivial because chained diatheses are always a combination of both a promotion and a demotion. So the question is which of the two “wins”.

[2.150] By definition (cf. Sec­tion 2.4.2), a chained diathesis is deemed to be an overall demotion when the demotion-part is stronger than the promotion-part, and vice versa. The strength is measured as the size of the “jump” on the macrorole hierarchy, repeated here in (2.100). Additionally, the overall chain exhibits varying intensity: the larger the difference in jump size between demotion-part and promotion-part, the more extreme the overall chain.

[2.151] Concretely, the promotion of an object exchange can be either an objective (ø › obj) (“three steps up”), an applicative (adj › obj) (“two steps up”) or an oblig­a­tory-loc­ation applicative (pbj › obj) (“one step up”). In reverse, the demotion can be either a deobjective (obj › ø) (“three steps down”), an antipassive (obj › adj) (“two steps down”) or an oblig­a­tory-loc­ation antipassive (obj › pbj) (“one step down”). Combining two of these leads to an overall assessment of the chained diathesis. This whole concept of chained demotions/promotions can be visualised by considering the top-right to bottom-left diagonal in Figure 2.4. The top-right corner (obj › obj › ø) is the most extreme demotion (“net four down”) and the bottom-left corner (ø › obj › obj) is the most extreme promotion (“net four up”). All other possibilities are situated somewhere in between these extremes on this diagonal.

[2.152] For example, consider again the diathesis from schmieren to beschmieren, repeated below in (2.101). It consist of a promotion from a prepositional phrase auf die Wunde to an accusative die Wunde and a demotion from an accusative eine Salbe to a prepositional phrase mit einer Salbe. The promotion starts from an obligatory location (2.101 a,b), i.e. this is an oblig­a­tory-loc­ational applicative (pbj › obj). In contrast, the demotion ends in an optional instrumental phrase (2.101 c,d), i.e. this is an antipassive diathesis (obj › adj). Now, the antipassive demotion (“two steps down”) is a bigger jump on the macrorole hierarchy than the oblig­a­tory-loc­ation promotion (“one step up”), so the whole object exchange (pbj › obj › adj) is classified as a demotion, be it a minor one (“net one step down”).

[2.153] According to the above mentioned generalisation, such an overall demotion coincides with the fact that the new object Wunde ‘wound’ is a holonym to which the old meronymic object Salbe ‘ointment’ is applied. To be precise, the terms meronym and holonym are language-specific classifications as observed in the structure of German. So, not all examples necessarily correspond to any (universal) semantic conceptualisation of the terms holonym/meronym. For example, German verbs that describe an act of covering (e.g. schmieren ‘to smear’) or wrapping (e.g. wickeln ‘to wrap around’) consistently treat the cover/wrap alike to other meronyms and the covered/wrapped object alike to other holonyms. That is no statement about what it semantically means to be a meronym or holonym. It is just a statement about the distribution of syntactic structures among German verbs as they take part in object exchange.

2.7.5.2 Demoted object exchange

[2.154] There are two different kinds of object exchange with demotion, namely the filled holonym and the emptied holonym object exchange. These two kinds of exchange correlate with the intensity of the demotion. A minor demotion manifests a fil­led-holo­nym object exchange, while a more extreme form of demotion expresses an emptied-holonym object exchange.

[2.155] Typically, a filled holonym diathesis is expressed by a minor demotion (“net one step down”), as illustrated with schmieren/beschmieren above in (2.101). Note that in that example, the holonym Wunde is not literally ‘filled with’ the meronym Salbe. However, with many other examples, like pflanzen/bepflanzen ‘to plant’ below (2.102), the holonym Garten ‘garden’ is literally filled with the meronym Tulpen ‘tulips’. In accordance with this being a minor overall demotion (“net one step down”), the new accusative object after the object exchange is a filled holonym.

[2.156] Note that with pflanzen/bepflanzen (2.102) the pre-diathesis location phrase in den Garten is not obligatory (2.102 a,b), so the resulting diathesis is symmetric (adj › obj › adj). This is the kind of minor syntactic variation that is indicated in Figure 2.4 with the unnamed boxes in the centre of the highlighted object-exchange domain. Still, this diathesis is clearly an example of a fil­led-holo­nym object exchange because the change in prepositions from in (with unmarked pflanzen) to mit (with marked bepflanzen) fits in perfectly with other fil­led-holo­nym examples (cf. Sec­tion 8.7.13).

[2.157] The second kind of demoted object exchange is the emptied holonym diathesis, expressed with a more extreme demotion (“net two steps down”). This is illustrated in (2.103) with the diathesis between the verb klopfen ‘to pound’ and ausklopfen ‘to beat out’ (cf. Sec­tion 8.7.12). The unmarked verb klopfen (2.103 a) takes an accusative object role that expresses the result of the pounding (Staub ‘dust’). The pounded object role (Mantel ‘coat’) is expressed as an obligatory location phrase (2.103 b). Crucially, the accusative object role in this construction is a component part (meronym) of the locational object role (holonym). The diathesis from klopfen to ausklopfen (2.103 c) completely drops the meronym Staub from the sentence (obj › ø) and promotes the holonym Mantel to accusative (pbj › obj). The meronymic role Staub cannot be expressed anymore at all after the diathesis (2.103 d). In accordance with this large overall demotion (“net two steps down”), the new holonymic accusative object Mantel is semantically “emptied” from its old meronymic accusative object Staub by the action klopfen. So the new accusative object after this object exchange is an emptied holonym.

2.7.5.3 Promoted object exchange

[2.158] There are also two different kinds of object exchange with promotion, namely the joined meronym and the parted meronym object exchange. These two kinds of object exchange correlate with the intensity of the promotion. A minor promotion manifests a joined-meronym object exchange, while a more extreme form of promotion expresses a parted-meronym object exchange. Basically, these diatheses are reversals of the two demoted object exchanges discussed in the previous section.

[2.159] The joined meronym object exchange occurs with less extreme promotions (“net one step up”). This is illustrated here with the diathesis from nähen ‘to sew’ to festnähen ‘to fixate by sewing’ (2.104). The original object role Bluse ‘blouse’ (2.104 a) turns into an optional an prepositional phrase (9.32 b,c). This part of the chain is an antipassive diathesis (obj › adj), i.e. “two steps down”. At the same time a new object role Knopf ‘button’ is introduced. This role cannot be expressed in the construction before the diathesis (2.104 a). So, this part of the chain is an objective diathesis (ø › obj), i.e. “three steps up”. The promotion is larger than the demotion, so the whole diathesis overall is a promotion, although a minor one, i.e. “net one step up”. As predicted for promotions, the new object Knopf semantically is a meronymic part of the original object Bluse. Further, in accordance with the promotion being minor, the verb nähen describes a situation in which the new object Knopf is physically connected to the holonymic Bluse. In summary, the new object is a joined meronym.

[2.160] In some examples the joined-meronym object exchange allows for optional prepositional phrases at both sides of the diathesis, resulting in a symmetric diathesis (adj › obj › adj). For example, this is attested with the diathesis from massieren ‘to massage’ (2.105 a,b) to einmassieren ‘to massage in’ (2.105 c,d). Still, this diathesis is clearly an example of a joined-meronym object exchange because the change in preposition from mit (with unmarked massieren) to in (with marked einmassieren) is completely parallel to all other joined-meronym examples (cf. Sec­tion 8.9.1). Semantically, the new accusative object role (Balsam, ‘balm’) is a meronym of the old object role (Muskeln ‘muscles’), and this new object role is applied to the old object role to become a part of it by the verb massieren. So, even though the diathesis is symmetric, the new object is syntactically and semantically a joined meronym.

[2.161] The parted meronym object exchange occurs with more extreme promotions (“net two steps up”). For example, the verb waschen ‘to wash’ normally takes an accusative object role that is the washee, here Hose ‘trousers’ (2.106 a). There is a covert diathesis that introduces a new object role that cannot be expressed earlier, here Fleck ‘stain’ (2.106 b). this addition is an objective promotion (ø › obj), i.e. “three steps up”. At the same time, the original accusative is turned into a locational prepositional phrase and this location phrase cannot be left out (2.106 c). This is an oblig­a­tory-loc­ation antipassive (obj › pbj), i.e “one step down”. This diathesis is thus an extreme promotion overall, i.e. “net two steps up”. Accordingly, the new meronymic object Fleck is semantically a component part of the original holonymic object Hose and it is removed from it by the action waschen. In summary, the new object is a parted meronym.

2.7.5.4 Other kinds of object exchange

[2.162] There are a few other examples of object exchange that do not fit in with the general pattern described above, for example with zwingen/erzwingen ‘to force’ (2.107 a,b). This chained diathesis is a remapping of the form (pbj › obj › adj) because the preposition zu is a governed preposition (2.107 c). This diathesis is attested with various verbs of persuasion (cf. Sec­tion 8.7.14). The demotion is more prominent than the promotion, so this chain is overall a demotion. Consequently, because it is the person being persuaded that is demoted I call this a persuadee demotion object exchange.

[2.163] Another example of an object exchange is illustrated here with the verb bewundern ‘to admire’ (2.108), see Sec­tion 6.9.3. This verb (and others like it) show a combination of a possessor-raising applicative promotion (adj › obj) and a governed antipassive demotion (obj › pbj) leading to the object exchange (adj › obj › pbj). The promotion is more prominent than the demotion, so this chain is overall a promotion. I propose to call this a possessor raising object exchange.

2.7.5.5 Chained case changes

[2.164] Finally, there are a few object chains involving a change of case, shown at the top and the left of Figure 2.4. Note that a case change of dative/genitive to accusative can be interpreted as a promotion, and the reverse as a demotion, cf. paragraph 2.43, but that perspective will not be expanded upon here.

[2.165] Example (2.109) shows a combination of a dative-to-accusative case change with an antipassive, resulting in a chain (obj › obj › adj). The verb schenken ‘to gift’ takes a recipient in the dative and a patient in the accusative, while the derived beschenken ‘to gift’ turns the accusative into a prepositional phrase (i.e. antipassive) and changes the dative dir into an accusative dich (see Sec­tion 8.7.8).

[2.166] The reverse situation, i.e. a chain (adj › obj › obj), is attested with the diathesis between drängen ‘to urge’ and the derived aufdrängen ‘to impose’ (2.110), see Sec­tion 8.8.13. In this example a prepositional phrase changes into an accusative (i.e. applicative), while the accusative dich changes to dative dir.

[2.167] Finally, an idiosyncratic diathesis is attested with the verb rauben ‘to rob’ (2.111), see Sec­tion 8.9.4. When this verb is changed to berauben ‘to rob’ then two case changes happen simultaneously: first a dative-to-accusative change (dich becomes dir) and second an accusative-to-genitive change (das Buch becomes des Buches). This is thus an example of a remapping pattern (obj › obj › obj), here called double case change.

3 Summary of major diatheses

3.1 German names for German grammar

[3.1] Among the almost 250 diatheses that are distinguished in this book there are many that are frequently attested and that can be used with very many different verbs. In contrast, there are also many diatheses that only occur in very specific circumstances or that might otherwise be considered to be exceptions or incidental instances. Only the major diatheses, those that are of central importance to the grammatical structure of German, will be summarised in this chapter. Such a summary would normally be presented at the end of a book, but because of the often long-winding data-driven details of the subsequent descriptive chapters, I decided to present this summary here at the end of the introductory deliberations. Take it as a quick appetiser of things to come, with ample links to the actual discussion in later chapters. This chapter also provides a sketch of how diathesis could be approached in practical grammars of the German language.

[3.2] To reiterate the basic premise of this book: in the Chapters 5 to 13 I aim to present a complete list of all coherent, and thus monoclausal, clause structures in German (cf. Sec­tion 1.3.1 on defining monoclausality). All in all, in those chapters there are more than 300 separate sub-subsections that describe (often minor) variations of monoclausal structures. This diversity is condensed into about 120 major clause alternations as summarised here. Of those, about 80 are diatheses (i.e. clause alternations with role-remapping, discussed in this chapter), while only about 40 are epitheses (i.e. clause alternations without any change in role marking, discussed in the next chapter). So, diathesis (“grammatical voice”) is a much more diverse grammatical phenomenon than epithesis (“tense-aspect-mood marking”). All these counts should be taken with some leeway, because a lot depends on individual decisions about splitting or lumping structures into groups (e.g. how many lassen+In­fi­ni­tiv constructions are counted separately, cf. Sec­tion 11.2.5). Although the analysis of German clause alternations might look cleaner when lumping structures into larger groups, that would not reduce the attested diversity, it would only hide the variation at the cost of larger within-group complexity.

[3.3] Besides providing a basic summary, I also propose German names (sometimes based on Latinate terms) for all 120 major derived monoclausal sentence structures. Using suitable names is a central aspect of (scientific) communication. In grammar, names are like instruments that allow us to abstract away from individual details and manipulate classes of utterances that show a specific abstract structure. However, naming is hard and can also lead to miscommunication. When re-using available terminology, the terms are easily recognised and remembered, but they carry the weight of history. Even when detailed definitions are given (as I have tried to do throughout this book), unintended interpretations of previous usage inevitably seep through. In contrast, inventing new names introduces more precision, but the downside is often cumbersome terms that are difficult to remember.

[3.4] In naming diatheses in this book I have tried to strike a balance between precise naming and good readability. For the English names in the detailed discussions in the coming chapters, I have decided in favour of precision. Each phenomenon is newly named with often long descriptive and unique names. In contrast, for the German names in this chapter I try to reuse available terminology as much as possible. When necessary, I propose new names that attempt to evoke a functional description like Reziprokativ or Erlebniskonversiv. However, the semantic characterisation has not been the main focus of this book, so it might become necessary to rename diatheses in the future once more detailed investigations have been performed. In some cases I have not been able to find a suitable semantic characterisation. For those diatheses I have resorted to using formal characteristics in the name, always written as separate words, like Reflexiv Erlebniskonversiv (i.e an Erlebniskonversiv that needs a reflexive pronoun) or Resultativ Delokativ (i.e. a Delokativ that needs a resultative preverbial).

[3.5] In this chapter, the diatheses are organised in sections according to the grammatical macro-role remapping patterns as introduced in Sec­tion 2.7. The different diatheses in each section are thus functionally highly similar, but they are structurally different. Inversely, there are various diatheses that are structurally highly similar, but are nonetheless repeated in separate sections under different names. This is necessary because superficially identical diatheses can have rather different structural repercussions depending on the verbs to which they are applied. This happens for example with different instantiations of the sein+Partizip or the lassen+In­fi­ni­tiv constructions.

3.2 Naming clause types

[3.6] Before diving into the daunting diversity of German diathesis, a short note on German names for different clause types is in order (summarised in Table 3.1). The distinction between sentence (German: satz) and clause (German: teilsatz) is customary made in the German grammatical literature when a precise description is needed. However, the term Satz is often used as a shorthand for both. When subdividing clauses, there is of course a basic distinction between main clause (German: hauptsatz, more precise would be selbständiger teilsatz) and subordinate clause (German: nebensatz or alternatively untergeordneter teilsatz).

[3.7] Yet, a central thesis of this book is that there is a further subdivision for both main and subordinate clauses. First, a “basic clause” is a clause with a single finite verb in the Präsens or Präteritum. For German I propose to use the term basissatz, or, to be more precise, grundlegender teilsatz. Various kinds of derived clauses can be constructed from a basic clause. For German I propose to call such a derived clause a spezialsatz, or, to be more precise, abgeleiteter teilsatz.

[3.8] There are two kinds of derived clauses. First, an epithesis is a clause alternation without role-remapping. For German I propose to use either the neologism übersatz or the Greek-inspired epithese, or, to be more precise, erweiterter teilsatz. Second, a diathesis is a clause alternation with role-remapping. For German I propose to use the neologism wechselsatz or the Greek-inspired diathese, or, to be more precise, umgestellter teilsatz.

3.3 Insubjective diatheses (sbj › o)

[3.9] An insubjective is a diathesis that completely removes the role marked as nominative subject without introducing a new subject. For details on the definition see Sec­tion 2.7.2, specifically starting at paragraph 2.90.

3.3.1 Auslöserentfall

[3.10] The unmarked auslöserentfall (full discussion in Sec­tion 5.5.1 and subsequent sections) is typically found with dispersion verbs like stinken ‘to stink’, klingeln ‘to ring’ or krachen ‘to crunch’ (3.1 a). These verbs allow for a construction without explicit nominative subject when describing a general situation with unknown cause. An obligatory valency-simulating pronoun es is used as a replacement of the nominative subject (3.1 b).

3.3.2 Aktionsbewertung

[3.11] The aktionsbewertung (full discussion in Sec­tion 9.5.1) similarly replaces the nominative subject by a valency-simulating es. Additionally, this diathesis obligatorily needs a reflexive pronoun and an adverbial phrase describing an evaluation, like gut ‘well’ or angenehm ‘pleasantly’. The Aktionsbewertung is typically used with agentive intransitive verbs like leben ‘to live’ or tanzen ‘to dance’ and describes a habitual situation. This diathesis is closely related to the bewertungsantikausativ for transitive verbs (see Sec­tion 3.7.4).

3.3.3 Zustandsbewertung (sein+In­fi­ni­tiv)

[3.12] The zustandsbewertung (full discussion in Sec­tion 11.5.3) is a somewhat formulaic construction dropping the nominative subject of an intransitive verb. This construction is constructed with sein and an infinitive. Additionally an adverbial phrase describing an evaluation is obligatory, like gut ‘well’ or schlecht ‘badly’. A valency-simulating pronoun es instead of the dropped nominative is mostly not present. This construction expresses an evaluation and it typically used with a location, like with sitzen ‘to sit’ (3.3).

3.3.4 Möglichkeitsbewertung (lassen+In­fi­ni­tiv)

[3.13] The möglichkeitsbewertung (full discussion in Sec­tion 11.5.1) consists of the light verb lassen with the infinitive of an intransitive verb. This construction obligatory includes a reflexive pronoun and an evaluating adverbial expression like gut ‘fine’. A valency-simulating pronoun es appears to be optional (3.4). This construction gives an evaluation about a possible situation. It is closely related to the permissivpassiv for transitive verbs (see Sec­tion 3.8.6).

3.3.5 Unpersönliches Vorgangspassiv (werden+Partizip)

[3.14] The unpersönlicher vorgangspassiv (full discussion in Sec­tion 10.5.1) is a construction consisting of the light verb werden with a participle of an intransitive verb. Only agentive (“unergative”) intransitive verbs like tanzen ‘to dance’ (3.5) or schlafen ‘to sleep’ allow for this construction without any nominative subject (not even a valency-simulating es is needed). The name “passive” is rather unfitting for this diathesis, but it is retained here because of widespread usage. This construction is closely related to the vorgangspassiv for transitive verbs (see Sec­tion 3.8.1).

3.3.6 Unpersönliches Modalpassiv (sein+zu‑In­fi­ni­tiv)

[3.15] The unpersönlicher modalpassiv (full discussion in Sec­tion 12.5.1 and subsequent sections) consists of a light verb sein with zu and an infinitive. It is found with incidental intransitive verbs, but more typically with verbs with a dative argument (but no accusative) like helfen ‘to help’ or trauen ‘to trust’ (3.6). In this diathesis the nominative subject is dropped and cannot be retained in any other form. There is also no valency-simulating es present. The name “passive” is actually beside the point for this diathesis, but it is used here because this construction is closely related to the modalpassiv (see Sec­tion 3.8.4).

3.4 Desubjective diatheses (sbj › adj)

[3.16] A desubjective is a diathesis that removes the role marked as nominative subject, though this role can still optionally be expressed as a prepositional phrase. For details on the definition see Sec­tion 2.7.2, specifically starting at paragraph 2.94.

3.4.1 Möglichkeitsdesubjektiv (geben+zu‑In­fi­ni­tiv)

[3.17] The möglichkeitsdesubjektiv (full discussion in Sec­tion 12.5.4) uses a subjectless light verb geben with zu and an infinitive (3.7). The removed nominative subject is replaced by a valency-simulating pronoun es, so the light verbs are always in the third person singular, resulting in a fixed expression es gibt. The removed subject can optionally be retained with a für prepositional phrase, though this is less frequent compared to the gelten+zu‑In­fi­ni­tiv diathesis (see Sec­tion 3.4.2). Any other argument is simply preserved, like the accusative den Koffer ‘the suitcase’ in the example below. The Möglichkeitsdesubjektiv semantically invokes an option that is available to the original subject, i.e. a modal-like ‘can’ meaning. The same construction geben+zu‑In­fi­ni­tiv is also used in a semantically and structurally quite different diathesis, namely the Möglichkeitskausativ (see Sec­tion 3.11.6).

3.4.2 Notwendigkeitsdesubjektiv (gelten+zu‑In­fi­ni­tiv)

[3.18] The notwendigkeitsdesubjektiv (full discussion in Sec­tion 12.5.5) uses a subjectless light verb gelten with zu and an infinitive (3.8). The removed nominative subject is replaced by a valency-simulating pronoun es, so the light verb gelten is always in the third person singular, resulting in fixed expressions es gilt. The removed subject can optionally be retained with a für prepositional phrase. Any other argument is simply preserved, like the accusative den Koffer ‘the suitcase’ in the example below. This construction semantically invokes some kind of (self‑)assignment that should be fulfilled, i.e. a modal-like ‘must’ meaning.

3.4.3 Aufforderungsdesubjektiv (heißen+In­fi­ni­tiv)

[3.19] The aufforderungsdesubjektiv (full discussion in Sec­tion 11.5.4) consists of the verb heißen with an infinitive. The meaning of this constructions is very close to the previous Notwendigkeitsdesubjektiv (see Sec­tion 3.4.2). The removed nominative subject is replaced by a valency-simulating pronoun es, so the light verb heißen is always in the third person singular, resulting in fixed expressions es heißt. The removed subject can optionally be retained with a für prepositional phrase. However, different from gelten+zu‑In­fi­ni­tiv, the construction heißen+In­fi­ni­tiv can only be applied to intransitive verbs. Note that there also exists a completely separate causative usage of heißen+In­fi­ni­tiv, but that Aufforderungskausativ appears to be rather old-fashioned (see Sec­tion 3.11.8).

3.5 Conciliative diatheses (adj › sbj › o)

[3.20] A conciliative is a diathesis that completely removes the role marked as subject and promotes an instrument to be the new subject. For details on the definition see Sec­tion 2.7.3.1, specifically starting at paragraph 2.103.

3.5.1 Instrumentsubjektiv

[3.21] The instrumentsubjektiv (full discussion in Sec­tion 6.5.4 and subsequent sections) promotes an instrument to nominative subject. For example, the instrument Schlüssel ‘key’ of the verb öffnen ‘to open’ is expressed with a mit prepositional phrase in (3.10 a). Alternatively, it can be expressed with a nominative as in (3.10 b). In that construction, the original agent cannot be expressed anymore. This diathesis looks very similar to the Kreationsubjektiv (see Sec­tion 3.6.1), but there are crucial semantic and structural differences (discussed below).

3.6 Fabricative diatheses (pbj › sbj › ø)

[3.22] A fabricative is a diathesis that completely removes the role marked as subject and promotes an fabricated entity to be the new subject. For details on the definition see Sec­tion 2.7.3.1, specifically starting at paragraph 2.102.

3.6.1 Kreationsubjektiv

[3.23] The kreationsubjektiv (full discussion in Sec­tion 6.5.7) superficially looks very similar to the previous Instrumentsubjektiv. In both diatheses a mit prepositional phrase is promoted to nominative subject. However, with a verb like überraschen ‘to surprise’ (3.11) the noun in the prepositional phrase, Aufgabe ‘task’, does not represent an instrument, but a fabrication by the subject of the sentence, Lehrer ‘teacher’. This semantic difference is paralleled by a structural difference, namely that the mit prepositional phrase is a governed preposition (3.11 c). Note that the verbs that allow for a Kreationsubjektiv show a substantial overlap with the verbs that allow for the Reflexiv Erlebniskonversiv (see Sec­tion 3.9.1), though the two groups are not identical.

3.6.2 Auslösersubjektiv (sein+zum-In­fi­ni­tiv)

[3.24] The auslösersubjektiv (full discussion in Sec­tion 13.5.1) is constructed with the light verb sein with a zum‑In­fi­ni­tiv. This diathesis can be applied to verbs of emotion with a governed preposition describing the trigger of the emotion. For example, heulen ‘to whine’ (3.12 a) uses the governed preposition über to describe the trigger, here Schaden ‘damage’ (3.12 b). The result of the diathesis is that the trigger of the emotion is promoted to nominative subject (3.12 c). The original subject, i.e. the perceiver of the emotion, cannot be expressed anymore.

3.7 Anticausative diatheses (obj › sbj › ø)

[3.25] An anticausative is a diathesis that completely removes the role marked as subject and promotes an object to be the new subject. For details on the definition see Sec­tion 2.7.3.1, specifically starting at paragraph 2.99.

3.7.1 Antikausativ

[3.26] The unmarked antikausativ (full discussion in Sec­tion 5.5.5 and subsequent sections) is attested with verbs like öffnen ‘to open’ or kochen ‘to cook’. These verbs occur both as transitive (3.13 a) and intransitive (3.13 b) without any further grammatical marking. Crucially, the object of the transitive is the subject of the intransitive. Because this diathesis is unmarked, there is no formal indication of a direction. So, this diathesis could just as well be interpreted as a causative. However, there is a formal difference between verbs that allow for both a haben and sein in the intransitive (3.13 c,d) and those that only allow for a sein in the intransitive. There seems to be an interesting semantic correlate to this formal difference in that the verbs that allow for both haben and sein seem primarily transitiv. Consequentially this group is called antikausativ (this section), while the second group with only sein is called kausativ (see Sec­tion 3.11.1).

3.7.2 Ortsantikausativ

[3.27] The unmarked ortsantikausativ (full discussion in Sec­tion 6.5.10) is similar to the previous unmarked antikausativ. Verbs like kleben ‘to glue, to stick’ or klappen ‘to fold’ occur both as transitive and intransitive (3.14 a,b) with the object of the transitive being the subject of the intransitive. Likewise, the intransitive is possible with both haben and sein (3.14 c,d). The only difference is the obligatory presence of a location. Note that there is also a parallel ortskausativ (see Sec­tion 3.11.2).

3.7.3 Reflexiv Antikausativ

[3.28] The reflexiv antikausativ (full discussion in Sec­tion 7.5.2 and subsequent sections) is attested with verbs like entscheiden ‘to decide’ or beschränken ‘to limit’. Again, these verbs occur both as transitive and intransitive with the transitive object being the subject of the intransitive (3.15 a,b). However, with these verbs the intransitive needs an obligatory reflexive pronoun (3.15 b). The intransitive with reflexive pronoun typically takes haben in the perfect (3.15 c). The intransitive perfect with sein (3.15 d) can now clearly be identified as a zustandspassiv of the transitive (see Sec­tion 3.8.2).

3.7.4 Bewertungsantikausativ

[3.29] The bewertungsantikausativ (full discussion in Sec­tion 9.5.2 and subsequent sections) is possible with many straightforward transitive verbs, like with ver­kauf­en ‘to sell’ or lesen ‘to read’ (3.16 a). The anticausative intransitive obligatorily needs a reflexive pronoun, and additionally an obligatory manner adverbial is needed (3.16 b). Just like the previous anticausatives, the intransitive occurs both with haben and sein in the perfect. However, haben is clearly used with the reflexive anticausative construction with obligatory adverbial (3.16 c), while sein is used with the zustandspassiv (see Sec­tion 3.8.2) of the original transitive, without reflexive pronoun or obligatory adverbial evaluation (3.16 d).

3.7.5 Inferenzantikausativ (scheinen/erscheinen+Partizip)

[3.30] The inferenzantikausativ (full discussion in Sec­tion 10.5.10) is constructed with one of the light verbs scheinen or erscheinen with a participle of a transitive verb (3.17). This construction expresses an evidential inference by the speaker that something is the case. The retention of the original agent as a prepositional phrase seems to be mostly not possible (3.17 b), so this diathesis is classified as an anticausative here. With intransitive verbs this construction does not show any diathesis and is consequently called perfektinferenz (see Sec­tion 4.6.2).

3.7.6 Sinnesantikausativ (aussehen/wirken+Partizip)

[3.31] The sinnesantikausativ (full discussion in Sec­tion 10.5.11) uses the light verbs aussehen or wirken together with a participle of a transitive verb to form an anticausative diathesis (3.18). This construction expresses that the speaker has sensory evidence about the state of affairs. The retention of the original subject is very rare, though it might to be possible with verbs describing a mental state, like entspannen ‘to relax’ (3.18 b). With intransitive verbs this construction does not show any diathesis and is consequently called sinnesevidenz (see Sec­tion 4.6.3).

3.7.7 Darstellungsantikausativ (geben/zeigen+Partizip)

[3.32] The darstellungsantikausativ (full discussion in Sec­tion 10.5.12) consists of the light verb geben with a participle and an obligatory reflexive pronoun. It expresses a conscious performance to appear in a certain way by the erstwhile accusative. The original nominative cannot be retained. The light verb zeigen can be used alternatively to geben. The difference (if any) between these two light verbs needs more investigation.

3.7.8 Erwartungsantikausativ (stehen+zu‑In­fi­ni­tiv)

[3.33] The erwartungsantikausativ (full discussion in Sec­tion 12.5.6) uses a light verb stehen with zu and an infinitive. The original accusative object is promoted to subject and the erstwhile nominative subject cannot be retained, so this clearly is an anticausative diathesis. However, examples with an explicitly accusative noun phrase as in (3.20 a) are actually rare. Typically, this diatheses is found with cognitive predicates expressing an expectation, like befürchten ‘to fear’, with a dass complement clause (3.20 b). Functionally, this complement clause has the same status as an accusative object. Note that complement clauses typically come towards the end of the sentence in German, and then the first position of the sentence often has to be filled with a position-simulating pronoun es (which is removed when the first position is filled otherwise).

3.7.9 Unmöglichkeitsantikausativ (gehen+zu‑In­fi­ni­tiv)

[3.34] The unmöglichkeitsantikausativ (full discussion in Sec­tion 12.5.7) uses a light verb gehen with zu and an infinitive. This anticausative diathesis is typical for an informal register, but written examples can be found going back to the 19th century. The construction is typically used with an additional negation (3.21 b), though in contemporary online writing it is also attested without negation (3.21 c). Because of the negation, the typical usage of the Unmöglichkeitsantikausativ is to express the impossibility to change something.

3.8 Passive diatheses (obj › sbj › adj)

[3.35] A passive is a diathesis that removes the role marked as subject and promotes an object to be the new subject. The erstwhile subject can optionally be expressed as a prepositional phrase. For details on the definition see Sec­tion 2.7.3.1, specifically starting at paragraph 2.100.

3.8.1 Vorgangspassiv (werden+Partizip)

[3.36] The vorgangspassiv (full discussion in Sec­tion 10.5.15) is the infamous diathesis consisting of a light verb werden with a participle. Passives are very similar to anticausatives in that the transitive object is turned into the intransitive subject (3.22). The special characteristic of a passive is that the transitive subject can be optionally retained, typically as a prepositional von or durch phrase. However, note that this prepositional phrase is normally not used. The same werden+Partizip construction leads to a different diathesis with intransitive verbs, namely the unpersönlicher Passiv (see Sec­tion 3.3.5).

3.8.2 Zustandspassiv (sein+Partizip)

[3.37] The zustandspassiv (full discussion in Sec­tion 10.5.16) consists of a light verb sein with a participle (3.23). Although this diathesis is traditionally called “passive” in German grammar, the status of the retained agent is problematic and appears to be strongly dependent on the verb (3.23 b,c). It might thus be better to consider this diathesis to be an anticausative. However, because of the long tradition I hold on to the term Zustandspassiv and the analysis of it being a passive. The closely related sein+Partizip erlebniskonversiv (see Sec­tion 3.9.2) retains the subject with a governed preposition. Also the sein+Partizip perfekt as attested with some intransitive verbs is arguably a similar construction, though applied to different verbs (see Sec­tion 4.3.1).

3.8.3 Fortsetzungspassiv (bleiben+Partizip)

[3.38] The fortsetzungspassiv (full discussion in Sec­tion 10.5.17) is closely related to the sein-Zustandspassiv (see Sec­tion 3.8.2), but now the light verb bleiben is used with a participle (3.24). This construction expresses that a reached state is maintained. Like with sein, the retention of the original agent with bleiben is possible, but often difficult (3.24 b,c). However, not all verbs can be equally used with sein and bleiben. For example, verbs like drucken ‘to print’ or schreiben ‘to write’ are fine with the sein-Zustandspassiv but not with the bleiben-Fortsetzungspassiv. This construction is only attested with transitive verbs. The same bleiben+Parti­zip construction can be used with intransitive verbs, but then it does not induce a diathesis and is called perfektkontinuativ (see Sec­tion 4.3.10).

3.8.4 Modalpassiv (sein+zu‑In­fi­ni­tiv)

[3.39] The modalpassiv (full discussion in Sec­tion 12.5.8) is constructed using the light verb sein with zu and an infinitive. When applied to transitive verbs like führen ‘to lead’ (3.25 a) or lösen ‘to solve’ (3.25 b) this diathesis promotes the accusative to nominative subject. The erstwhile nominative subject can be retained as a prepositional phrase. This diathesis has two different interpretations. It can indicate either an deontic modality (‘must’) as in (3.25 a) or an ability (‘can’) as in (3.25 b). Note that the subject retention with the preposition für is only possible in the ability-interpretation. The closely related unpersönlicher Modalpassiv is used with intransitives and only allows for the deontic interpretation (see Sec­tion 3.3.6).

3.8.5 Normpassiv (gehören+Partizip)

[3.40] The normpassiv (full discussion in Sec­tion 10.5.18) consists of the light verb gehören with a participle. It is only attested with verbs with accusative objects, like bestrafen ‘to punish’ (3.26). The diathesis expresses that the main verb ought to be applied to the object. The original subject can optionally be retained as a prepositional phrase.

3.8.6 Permissivpassiv (lassen+In­fi­ni­tiv)

[3.41] The permissivpassiv (full discussion in Sec­tion 11.5.5) consists of the light verb lassen with infinitive and an obligatory reflexive pronoun (3.27). The agent can be retained with an optional von prepositional phrase, so this diathesis is a passive. This diathesis expresses that something is permitted (3.27 a) or that something is possible (3.27 b). A similar construction with lassen+sich+In­fi­ni­tiv can be applied to intransitive verbs, which leads to a different diathesis, namely the Möglichkeitsbewertung (see Sec­tion 3.3.4). Also the Permissivkonversiv (see Sec­tion 3.9.3) and the Permissivinversiv (see Sec­tion 3.10.4) use the same construction with lassen, but they also show different role-remappings.

3.8.7 Rezipientenpassiv (bekommen/kriegen/erhalten+Partizip)

[3.42] The rezipientenpassiv (full discussion in Sec­tion 10.5.21) has become a mainstay in the German grammatical literature. It consists of the light verb bekommen with a participle (alternatively, the light verbs kriegen or erhalten can be used). With this diathesis, a dative recipient is turned into the nominative subject. Again, the erstwhile nominative can be retained as a prepositional phrase, though it mostly is not used (as with all passives). Note that the same construction can also be used in a different “achievement” interpretation without diathesis, called Effektiv here (see Sec­tion 4.5.6).

3.8.8 Pertinenzpassiv (haben+Partizip)

[3.43] The pertinenzpassiv (full discussion in Sec­tion 10.5.22) is a special construction because it looks identical to the Perfekt (see Sec­tion 4.3.1), often even being ambiguous among the two interpretations. However, the Pertinenzpassiv is functionally much closer to the Rezipientenpassiv. The new nominative subject der Minister ‘the minister’ (3.29 b) is the (dative) experiencer/beneficiary of the cutting (3.29 a). The original agent of the cutting Friseur ‘barber’ can only be retained with difficulty, so this diathesis looks closer to an anticausative. However, there is a well-known effect that this Pertinenzpassiv becomes much more common when stacked with a modal auxiliary like wollen ‘to want’ (3.29 c). In such a stack, the original agent can clearly be retained.

[3.44] The designation pertinenz refers to the fact that the new subject is necessarily the possessor of the accusative object Haare ‘hair’. Such inherent possessors turn up in various diatheses, and all instances will be designated with the qualifier pertinenz. The most famous one is the pertinenzdativ (see Sec­tion 3.13.4), but there are various others, like the pertinenzinversiv (see Sec­tion 3.10.2) and the ortspertinenzinversiv (see Sec­tion 3.10.3).

3.9 Conversive diatheses (obj › sbj › pbj)

[3.45] A conversive is a diathesis that removes the role marked as subject and promotes an object to be the new subject. The erstwhile subject can optionally be expressed as a governed prepositional phrase. For details on the definition see Sec­tion 2.7.3.1, specifically starting at paragraph 2.101.

3.9.1 Reflexiv Erlebniskonversiv

[3.46] The reflexiv erlebniskonversiv (full discussion in Sec­tion 7.5.7) is a diathesis in which a verb, like empören ‘to appall’ (3.30 a), can be used both with and without a reflexive pronoun. The effect of adding the reflexive pronoun is a remapping of the accusative to nominative and demoting the erstwhile nominative to a prepositional phrase (3.30 b). The prepositional phrase is a governed preposition (3.30 c). The verbs that allow this diathesis are typically verbs that express an experience. This diathesis is functionally similar to the sein-Erlebniskonversiv (see Sec­tion 3.9.2) and the lassen-Permissivkonversiv (see Sec­tion 3.9.3). There are even many verbs that allow for all three diatheses, like empören (3.30 d,e). However, not all verbs allow for all three diatheses, like verärgern (3.31 d,e).

3.9.2 Erlebniskonversiv (sein+Partizip)

[3.47] The erlebniskonversiv (full discussion in Sec­tion 10.5.23) is constructed with the light verb sein and a participle. The form of this diathesis is identical to the Zustandspassiv (see Sec­tion 3.8.2), but there is a crucial difference in the remapping of the original nominative. Verbs that take a Zustandspassiv, like öffnen ‘to open’ only allow for the retention of the nominative with a von prepositional phrase, and only in special circumstances. In contrast, the verbs that take the Erlebniskonversiv can regularly retain the agent with a governed preposition. For example, with the verb verärgern ‘to displease’ the original nominative can be expressed with an über prepositional phrase (3.31 b), which is a governed preposition (3.31 c). Verbs that take the Erlebniskonversiv are typically verbs the express an experience, similar to the next other two conversive diatheses, the Reflexiv Erlebniskonversiv (3.31 d), see Sec­tion 3.9.1 and the Permissivkonversiv (3.31 e), see Sec­tion 3.9.3.

3.9.3 Permissivkonversiv (lassen+In­fi­ni­tiv)

[3.48] The permissivkonversiv (full discussion in Sec­tion 11.5.7) uses a light verb lassen with an infinitive and an obligatory reflexive pronoun. In this diathesis, the original dative is promoted to nominative subject, while the original nominative is demoted to a prepositional phrase, like with the verb begeistern ‘to be enthusiastic’ (3.32 a,b). The preposition is a governed preposition (3.32 c). The verbs that allow for this diathesis are highly similar, but not identical, to the verbs that take the Erlebniskonversiv (3.32 d), see Sec­tion 3.9.2, and the Reflexiv Erlebniskonversiv (3.32 e), see Sec­tion 3.9.1. However, note the different prepositions in these constructions, as illustrated below.

3.10 Inversive diatheses (obj › sbj › obj)

[3.49] An inversive is a diathesis that switches subject and object. For details on the definition see Sec­tion 2.7.3.3, specifically starting at paragraph 2.114.

3.10.1 Restinversiv (bleiben+zu‑In­fi­ni­tiv)

[3.50] The restinversiv (full discussion in Sec­tion 12.9.1) uses the light verb bleiben with zu and an infinitive. This diathesis reverses the expression of the subject and object roles, in that the accusative is promoted to a nominative, while the original nominative is demoted to an (optional) dative. Because the demotion is “larger” than the promotion this diathesis can be interpreted as a demoted inversive. Semantically, this diatheses expresses that (some part of) the patient is still left over to be acted on.

3.10.2 Pertinenzinversiv (haben+am‑In­fi­ni­tiv)

[3.51] The pertinenzinversiv (full discussion in Sec­tion 13.9.1) is constructed with the light verb haben with an am‑In­fi­ni­tiv. This diathesis also reverses the expression of the subject and object roles, though in the different direction from the previous Restinversiv (see Sec­tion 3.10.1). In this diathesis the dative is promoted to nominative, while the nominative is demoted to accusative (3.34). Because the promotion is “larger” than the demotion this can be called a promoted inversive. Further, the dative dem Mieter ‘tenant’ is necessarily the possessor (pertinenz) of the nominative die Wohnung ‘apartment’, so it is a Pertinenzdativ (see Sec­tion 3.13.4). Both in form and meaning this diathesis is strongly connected to the ensuing ortspertinenzinversiv (see Sec­tion 3.10.3), in which the dative is the possessor of the obligatory location.

3.10.3 Ortspertinenzinversiv (haben+In­fi­ni­tiv)

[3.52] The ortspertinenzinversiv (full discussion in Sec­tion 11.9.2) is closely related to the previous Pertinenzinversiv (see Sec­tion 3.10.2). Again, a dative is promoted to nominative, while the nominative is demoted to accusative. Also in both diatheses, the participant expressed by the dative is necessarily the possessor (pertinenz) of another participant. The difference is that with the current Ortspertinenzinversiv this other participant is an obligatory location, e.g. an der Nase ‘on the nose’ in (3.35). The dative in this diathesis is thus an Ortspertinenzdativ (see Sec­tion 3.13.5). An further curious difference to the otherwise highly similar Pertinenzinversiv in (3.34 b) is that the infinitive hängen does not allow for the preposition am in this construction (3.35 c).