Editor

JLLT edited by Thomas Tinnefeld
Journal of Linguistics and Language Teaching
Volume 6 (2015) Issue 1



Forms and Functions of Reflexive Structures in German

Katrin Ziegler (Macerata / Italy)

Abstract (English)
Morpho-syntactic descriptions of reflexivisation processes in German are usually based on a grammatical approach that considers specific elements in terms of categories (for example, as individual words: nouns, verbs or adjectives) without considering the variable paradigmatic and syntagmatic relations of interdependence which link the clause elements. This article, in contrast, adopts a functional approach to the analysis of reflexive structures, in order to provide an in-depth study of clauses as nexuses of grammatical relations. In doing so, it provides a comprehensive description of the various constructions containing German reflexive pronoun sich, and attempts to provide systematic summaries of the complex sets of forms and functions.
Key words: reflexivisation processes, functional approach, clauses as nexuses of grammatical relations


Abstract (Italian)
Dietro le descrizioni morfosintattiche dei processi di riflessivizzazione in tedesco sta solitamente una prospettiva grammaticale che osserva i singoli elementi dal punto di vista categoriale (per es., come singole parole: sostantivi, verbi, aggettivi), non considerando invece i mutevoli rapporti di interdipendenza paradigmatica e sintagmatica che correlano gli elementi proposizionali.
Nel lavoro presente, invece, si procede ad un’analisi delle strutture riflessive adottando un punto di vista funzionale sotto il quale si esaminano le proposizioni in un modo complessivo e cioè come nessi di relazioni grammaticali. In questo modo è possibile fornire una descrizione unitaria di tutte le costruzioni marcate dalla presenza di
sich e tentare sistemazioni organiche degli insiemi complessi di forme e funzioni.
Parole chiave: processo di riflessivizzazione, interdipendenza paradigmatica e sintagmatica, relazioni grammaticali

Introduction

Drawing on a functional syntactic theory, this article examines reflexive structures in the German language with a view to offering a classification of such structures and a reflection on some of the more challenging aspects inherent to the classification task. Despite the existence of a vast bibliography on the topic, no experimental tools have yet been developed to perform a comprehensive survey of the complex set of forms and functions featuring the reflexive pronoun sich, for the purposes of properly considering their dissimilarities and any potential similarities in their features.
As in the case of other languages, traditional studies of German grammar tend to approach constructions involving the reflexive sich by first considering the verbs with which it occurs (cf. Section 2). As such, reference, for example, is made to ‘properly reflexive’ verbs and ‘inherently reflexive’ verbs. The limitations of this approach are clear, however, if we consider that a single verb can occur in a variety of structures and that, as such, these structures are not necessarily linked to the lexical properties of their individual elements. The verb class approach also fails to elucidate the structural relationships among the different constructions in which sich occurs.

This article aims to perform a comprehensive overview of clause structures. To this end, it draws on Relational Grammar (RG)1, a framework which also provides the formal tools used in this study, favoured because they are both easy to apply and effective in highlighting the relevant issues.2 In line with this approach, this study does not focus exclusively on individual words (nouns, verbs, adjectives) but rather includes descriptions of interdependencies and syntactic correlations, from both syntagmatic and paradigmatic perspectives. As such, clauses are considered as nexuses of grammatical relations.

2 Status Quaestionis

In German linguistic studies, ‘reflexive’ verbs are traditionally classified according to the following categories.3


2.1 Categories of Reflexive Verbs
2.1.1 Verbs Used Reflexively

The first category refers to ‘properly reflexive verbs’, the direct or indirect object of which may either be reflexive or any other noun or nominal group. The reflexive pronoun is the direct object (in the case of the accusative) in example (1a), and the indirect object (in the case of the dative) in example (1b):

(1a) Die Frau wäscht sich morgens sorgfältig.
       [The woman washes thoroughly in the morning.]

(1b) Die Studenten erzählen sich die letzten Neuigkeiten.
       [The students tell each other the latest news.]

While the reflexive pronoun in construction (1b) is the argument of the verb and, as such, necessarily part of the phrasal structure, the reflexive in sentence (1c) is a ‘free dative’ (an optional constituent):

(1c) Meine Freundin kauft sich ein neues Auto.
       [My friend buys herself a new car.]


2.1.2 Inherently Reflexive Verbs

The so-called lexicalization of the reflexive occurs both in clauses with verbs that are entirely intransitive (sich) ängstigen (2a) and in (anticausative) clauses that have a transitive counterpart (sich) öffnen (2b):

(2a) Meine Katze ängstigt sich sehr bei Gewitter.
        [My cat is very frightened of thunder storms.]

(2b) Die Tür öffnet sich leicht mit einem Schlüssel.
        [The door opens easily with a key.]


2.1.3 Reflexive Forms with a Passive Meaning

Reflexive forms with a passive meaning feature the reflexive pronoun sich in combination with intransitive verbs (3a) or verbs that are considered detransitivized (3b):

(3a) Es arbeitet sich gut in der Bibliothek. 
       [One works well in the library.]

(3b) Das Buch liest sich gut.
        [The book reads well.]

The impersonal construction (3a) necessarily requires the inclusion of the pronoun es which functions as the formal subject. An adverbial modifier - represented in the above examples by the adverb gut - is required in both the personal and impersonal constructions; without a modifier, the sentence is incorrect.

2.1.4 Passive Reflexive

Constructions of this kind (containing both passive and reflexive morphology) involve verbs that occur in ‘properly’ reflexive structures (4a) and those that are ‘inherently’ reflexive (4b). It should noted that the passive reflexive is considered a rare variant of the impersonal passive (Plank 1993), which, to date, has received little attention. This aspect will be discussed further in Section 4.

(4a) Es wird sich morgens gewaschen; (Morgens wird sich gewaschen.) 
        [One washes in the morning.]

(4b) Es wird sich geängstigt bei Gewitter; (Bei Gewitter wird sich geängstigt.) 
        [People are frightened by the thunderstorm.]

Both structures may contain the indefinite pronoun es; its inclusion is, however, not obligatory under certain syntactic conditions4.

2.2 Four Hypotheses Regarding the Properties of the Reflexive sich

Unlike Italian, for example, which has two series of reflexive pronouns referred to as ‘free’ and ‘clitic’, German has only one form of reflexive pronouns. Reflexive pronouns in the first and second persons are identical with personal pronouns; only in the third person (singular and plural) is there a distinctive reflexive form, sich, which has no number distinction and which syncretises the accusative and dative cases.

In the context of such a lack of morphological clarity, and using the traditional classification method as a starting point, it is difficult to discern whether any commonalities exist between the various instances in which sich is used, and if so, what these might be. The literature on this subject offers four hypotheses, two of which (the first and last) contradict each other:
  • The reflexive is considered a ‘full’ pronoun in all clauses in which it occurs. According to this perspective (Fagan 1992, Steinbach 2002, Bierwisch 2006), sich not only possesses (pro)nominal properties, it also retains argumental status - even in phrasal constructions in which it is not in paradigmatic commutation with the reflexive object. In such cases, it is assumed to have a ‘semantically empty’ theta-role and in the final analysis, the transitivity of all reflexive structures is affirmed. (cf. (1a) - (3b)).
  • In clauses in which the reflexive is not in paradigmatic commutation with a non-reflexive direct object, it is compared with a formal object and therefore, while it retains precise syntactic functions, it does not necessarily have a semantic role (Pittner & Berman 2008). (cf. (2a) - (4b)).
  • The reflexive is considered a syntactic marker of detransitivization and, as such, has pronominal properties which, in some structures, are relegated to the status of residuals (Plank 1986, Hundt 2001). (cf. (2a) – (4b)).
  • The reflexive is considered an integral part of the verbal morphology (Grewendorf 1984, Everaert 1986, Cardinaletti 1999). Assuming that ‘inherent’ reflexive verbs have reflexive particles linked to them through lexical properties, it is not problematic to assign a ‘semantically empty’ status to the reflexives. The value of the commutation with non-reflexive direct objects is, however, lost. (cf. (2a) – (2b)).
Each of the four hypotheses outlined above provides a different approach to a series of related structural problems including, for example, the issue of the transitivity or intransitivity of individual clauses containing a reflexive pronoun (de Alencar & Kelling 2005). Clearly, an evaluation of the transitive or intransitive nature of constructions using sich will vary according to one’s definition of reflexives as arguments or formal objects, syntactic markers or integral parts of the verbal morphology.


2.3 A description of the Pronoun es in Different Syntactic Contexts

The challenge of defining the properties of the reflexive pronoun and, therefore, determining the transitive or intransitive nature of clauses using sich, also entails the problem of establishing a coherent and consistent description of the indefinite pronoun es in impersonal structures containing the reflexive pronoun. As highlighted in the formal representation of constructions (3a), (4a) and (4b), the es form in German grammar is traditionally considered to fulfil the function of both formal subject (3) and so-called place-holder (4a), (4b). This divergent description of the use of the pronoun es in different syntactic contexts, however, gives rise to an incoherent definition of the antecedent of the reflexive pronoun in the structures in question, making it difficult to engage in a uniform formalization of constructions with reflexive morphology. In fact, traditional grammar definitions shed little light on the relationship between the reflexive and passive or between the reflexive and impersonal. As such, the few studies that consider the passive reflexive in which reflexive and passive morphologies converge present contradictory conclusions.

3 A Relational Perspective

The descriptive problems referred to above relate to the perspective from which the matter is approached, and change when that perspective changes. The issue of whether or not to regard sich as a real object, for example - a question obviously linked to whether or not the related structures are considered to be transitive -, is cast in a different light when considered from the perspective of RG. RG starts from the assumption that

(…) one cannot speak of sentences or even clauses as being transitive or intransitive. This is a property of syntactic levels, reconstructed as strata in relational networks. For example, it makes no sense to speak of clauses as being transitive or intransitive because many clauses are transitive in some strata and intransitive in others. (Perlmutter 1983a: 151)

Principles which apply to all structures clearly also apply to reflexive structures. According to RG, clauses may be structured across several syntactic levels, and functional commutations may occur between one level and another. Thus, it follows that the same grammatical relation (e.g. predicate, subject, direct object) may be borne by different elements on separate levels (but not by two nominals in a single stratum) (‘Stratal Uniqueness Law’ (Perlmutter 1983b: 92)).

3.1 Notions of RG Formalization

Fundamental to the formal structure of RG is the concept that a nominal element can relinquish its grammatical relation, either by acquiring a new one (revaluation or removal) or by getting syntactically 'lost'. These nominals, no longer able to engage in morphosyntactic operations, are referred to as chômeur. To clarify this point, the stratal diagram below5 presents the passive structure6 Das Bonbon wurde gegessen von dem Kind (The sweet has been eaten by the child):
Das Bonbon
wurde gegessen
von dem Kind
1
P
chô
2
P
1
Table 1: RG formalization of a structure containing a plain personal passive

The construction Das Bonbon wurde gegessen von dem Kind (The sweet has been eaten by the child) shows the 2→1 advancement from a transitive initial stratum in which the verb essen ‘initializes’ a subject (the person who eats) and a direct object (that which is eaten). The nominal Das Bonbon acquires relation 1, relinquishing its previous relation. The initial subject, das Kind, has been demoted to the chômeur relation, which means that it fails to control agreement with the verb, and occupies an optional circumstantial role in the clause, introduced by the preposition von.

3.2 The Unaccusative Hypothesis

It will be clear by now that concepts of transitive and intransitive do not apply to clausal structure as a whole, but rather to individual syntactic strata7. The concept of the intransitive form, meanwhile, can be further broken down in RG, as demonstrated by Perlmutter (1978) in his Unaccusative Hypothesis. This hypothesis proposes a distinction between unaccusative and unergative syntactic levels. According to this perspective, a structure, the final subject of which also has the properties of a direct object - for example the Italian sentence Lea è partita (Lea has left) - constitutes an initial object (the initial level is regarded as unaccusative). A structure such as Lea ha parlato (Lea has spoken), on the other hand, in which the argument only bears characteristics of a subject, must be considered monostratal and unergative8.

3.3 Medium constructions

According to definitions put forth by La Fauci (1988), constructions in which the relations between the final subject and the direct object are in commutative dependence - that is, paradigmatic (La Fauci 2007: 72) - are considered ‘medium’. The characteristic trait of medium constructions is that the final subject bears the direct object relation at non-final levels of the clause. Medium constructions are in opposition to so-called 'active’ constructions, in which the only possible dependence between the relations of the subject and direct object is syntagmatic, and never paradigmatic, and in which subjects do not therefore bear the object relation at any syntactic level (La Fauci 2007: 72). The oppositions between the active and medium constructions come under the term diathesis. As well as passive structures, the medium type also includes reflexive structures, which will be discussed further in following paragraphs (cf. section 4.2).

3.4 The concept of Multi-Attachment

A sub-set of structural types - formally characterized by the presence of a syntactic level with ‘multi-attachment’ (henceforth MA) - can be identified within the class of the medium constructions.
The MA Hypothesis is a descriptive hypothesis that accounts for “certain clauses with so-called ‘reflexive morphology’” (Perlmutter 1983a: 153). According to this theory, these clauses contain a (non-final) nominal at some syntactic level that simultaneously bears two or more grammatical relations - specifically subject and object relations (La Fauci 1992: 45). MA does not occur at the final stratum of the clause. As such, its resolution in a stratum subsequent to that in which it is produced is indicated superficially by the presence of the reflexive pronoun, causing the lower grammatical relation to be cancelled (La Fauci 1992: 45).

3.4.1 Multi-attachment at Initial-P Level

Multi-attachment can occur at the initial stratum of a clause, at the first level of occurence of a predicate (initial-P). In the initial MA structure, a single nominal element not only bears two grammatical relations, it also serves two separate thematic roles (La Fauci 1992: 47)9. As demonstrated in the RG formalization below, according to this hypothesis, a classic reflexive structure such as La donna si pettina10 (The woman combs her hair) involves the presence of a syntactic level with MA at the first level of occurrence of the predicate11 (Table 2):
La donna
si pettina
1
 P
1, 2
 P
Table 2: RG formalization of a reflexive structure with Multi-attachment at the first level

Rosen points out that MA

introduces a specialized pronoun which takes over from the doubly attached nominal one of its two relations - the lower one, normally and perhaps universally. The specialized pronouns born under this condition are reflexive pronouns. (Rosen 1988: 44)

As shown in the above stratal diagram, the predicate pettinare initializes both the subject (she who combs) and the object (she who is combed). In the final syntactic stratum, however, the nominal abandons the object relation, and only bears the subject relation.

3.4.2 Multi-attachment at Non-initial-P Level

In cases in which MA appears in non-initial strata of the clause - at levels subsequent to the first appearance of the predicate - “il MA non è correlato col possesso da parte del nominale di ruoli tematici distinti, poiché è l’effetto di un processo squisitamente sintattico” [it is not linked with distinct thematic roles of the nominal because the effect is that of a purely syntactic process] (La Fauci 1992: 47).

In structures involving MA in non-initial-P strata, the nominal acquires a new grammatical relation at the advancement level while also retaining the former relation unaltered (retroherent revaluation) (La Fauci 1992: 48).12 In a structure such as Ugo si pente (Hugo repents), an unaccusative or retroherent advancement occurs, as shown in the following stratal diagram:
Ugo
si pente
1
P
1, 2
P
2
P
Table 3: RG formalization of a reflexive structure with MA in non-initial-P strata
The unaccusative verb structure does not have an initial subject, and the 2→1 advancement is referred to as retroherent because the MA occurs at a syntactic level subsequent to the first.

Rosen (cf. La Fauci 1992: 49) observes that structures with an unaccusative valence can be further subdivided according to their (optional) ability to initialize a subject, and to the retroherent and plain modalities by which the advancement of the initial direct object to the subject relation occurs.13

As will be demonstrated later in this article (Section 4.2), it is our belief that the classification of structures with MA, proposed by Rosen for Italian, is also applicable to German. Caution must be exercised in this regard, however, as it is difficult to accurately establish the existence of a class of unaccusative verbs in this language.

3.5 The notion of Dummy

Another basic concept in RG is that the final level of a clause always contains the grammatical subject relation (Final 1 Law (Perlmutter 1983b: 100)) which may be present as a dummy.The dummy can occur syntactically as, for example, in the English clause It’s raining cats and dogs and the French Il pleut des cordes, or can be silent, as in the Italian Piove a catinelle14. In German, the dummy is usually expressed syntactically as, for example, in the clause Es regnet Bindfäden. As demonstrated above, however, this element may be silent in some structures and under certain syntactic conditions, as evident from a comparison of the two constructions (4a) Es wird sich morgens gewaschen and (4b) Morgens wird sich gewaschen.15

It should be emphasised here (though it may already be clear), that RG refutes the possibility of taking a generic approach to the notion of subject: this notion must be considered specifically in relation to each language, and broken down in accordance with the significance of multiple syntactic levels. An analysis of syntagmatic and paradigmatic relations performed by grammatical functions in the various types of structures indicates that the empirical properties attributed to the notion of initial subject differ from those of the concept of final subject.

4 Syntactic Analysis of Structures with Reflexive Morphology in German

In light of the premises outlined above and drawing on the aforementioned tools, this section of the article will seek to present a consistent formalization of reflexive morphological structures in German. This formal description aims in particular to highlight the syntactic conditions under which the passive and reflexive morphologies converge in the passive reflexive structure (4a) – (4b). As mentioned previously, the passive reflexive is considered a (rare) variant of the impersonal passive and has been subject to little analysis to date.

Researchers’ apparent lack of interest in this topic can be attributed primarily to the fact that the structure is considered grammatically erroneous; passive and reflexive morphologies are usually found in complementary distribution and, as such, it is not strictly possible to transform a clause with a reflexive verb into a passive construction16.

This analysis also aims to provide a uniform, formal account of the pronoun es in its various syntactic contexts. So, while in the construction (3a) Es arbeitet sich gut in der Bibliothek, the indefinite pronoun es is allotted the role of antecedent of the reflexive as it is the formal subject, the presumed lack of a syntactic subject in structures with the passive reflexive makes it difficult to solve the problem of the antecedent in this kind of clause. As mentioned above, the pronoun es does not have the role of subject in this case.

What syntactic element serves as the antecedent of sich, then, if the reflexive does not have a subject within the same clause, in this case a passive reflexive? One potential solution is to consider the reflexive, in this case, as a simple detransitivization marker that does not require a co-referring element (Plank 1993: 141-143).

Indeed, in many grammatical frameworks17, reflexivization is regarded as a mechanism of co-reference between nouns and pronouns and, as such, the reflexive is considered on the basis of its relation with other co-indexed elements within the same clause or the same domain.

An analytical approach of this kind, however, does not highlight the interdependence between the reflexive and the passive, or between the reflexive and the impersonal. The MA hypothesis, on the other hand, enables the identification of shared features - for example the fact that reflexive and passive constructions both have a final subject which is (also) a direct object at some level of the structure (La Fauci 1988: 15-27).

What we wish to suggest here is that German reflexive structures can be described in terms of the MA hypothesis; specifically, all relevant reflexive constructions are characterized by a (non-final) syntactic stratum with MA. Moreover, the constructions in question can be further classified as structures with MA at initial-P level and structures with MA at non- initial-P level.

4.1 Structures with MA at Initial-P Level

Let us consider the hypothesis of an MA at the first level of appearance of the predicate for properly reflexive structures, as in example (1a):

Die Frau
wӓscht sich
1
P
1, 2
P
Table 5: RG formalization of a structure with MA at initial-P level

The nominal Die Frau, initialized due to the valence of waschen as the subject and direct object, relinquishes the relation of object at the final stratum and only bears the relation of subject. It follows that the final stratum of the clause is intransitive, as it only contains argument 1. As stated previously (La Fauci 1992: 47), a structure with an initial MA implies both that a single nominal element bears two distinct grammatical relations, and that it has two thematic roles - as agent and patient. The resolution of the MA at the final level of the structure causes the reflexive pronoun sich to ‘inherit’ the lower grammatical relation (Rosen 1988: 44) from the doubly initialized nominal. As such, its direct-object properties are clearly accounted for, highlighted by the accusative case marking.

In example (1b), the grammatical relations also involve an indirect object (Table 5), represented by number 3 in the stratal diagram, which is typically marked in German by the dative case and which, like the subject, participates in the syntactic process of production and resolution of the MA:
Die Studenten
erzӓhlen sich
die letzten Neuigkeiten
1
P
chô
1,2
P
chô
1,3
P
2
Table 5: RG formalization of a structure with MA at initial-P level involving an indirect object

In the case of properly reflexive structures involving an indirect object relation18, we propose that MA occurs at the first level of appearance of the predicate. The verb erzählen initializes the nominal die Studenten as subject and indirect object, and as agent and beneficiary; the 3→2 advancement causes the initial 2 die Neuigkeiten to become a chômeur, thus preventing it from advancing further; and the MA 1, 2 is resolved in the final stratum with the cancellation of the direct object relation. As such, the structure is intransitive, as it only contains the subject relation. The reflexive pronoun in this case inherits the lower relation, and in fact has the properties of an indirect object, something which is apparent, among other things, from the dative case marking.

Let us now consider the hypothesis of an MA at initial-P level in the case of the so-called passive reflexive construction (4a):
Es
Pro
wird sich gewaschen
1

P
2
1
P

1
P

1, 2
P
Table 6: RG formalization of a structure with MA at initial-P level and dummy 2→1 advancement

The structure represented in the table includes a syntactic level with initial MA due to the valence of the verb waschen, which initialises a subject (he who washes) and a direct object (that which is washed) confluent on the same element, ‘Pro’ (“unspecified empathy focus” (Rosen 1988: 1986)). The resolution of the MA, then, occurs in the following stratum of the configuration and manifests in the appearance of the reflexive pronoun sich at the final level of the construction.

The dummy es which occurs at a non-initial level of the syntactic scheme19 is initially attributed to relation 2, as RG specifies (‘Nuclear Dummy Law’ (Perlmutter 1983b: 101)) that es can only bear subject or direct object grammatical relations. Similarly, in constructions involving the impersonal passive, a transitive level is created in the stratum on which the dummy appears, thus enabling the 2→1 advancement of the dummy itself. This advancement is responsible for the passive morphology in the passive reflexive.

Nominal direct objects in conjunction with MA 1, 3 in the initial stratum, as in the clause Es werden sich die Gesichter gewaschen, (They / people wash their faces] (Table 7), can also be found in the passive reflexive:
Es
Pro
die Gesichter
werden sich gewaschen
1

chô
P
2
1
chô
P

1
chô
P

1, 2
chô
P

1, 3
2
P
                     Table 7: RG formalization of a structure with MA 1,3 at initial-P level
                                   and dummy advancement 2→1

In the initial stratum of the stratal diagram, Pro bears the 1, 3 relations simultaneously. The nominal die Gesichter is allotted the relation of direct object, but loses this function with the advancement of Pro to 1, 2 relation. The MA relation, in the stratum subsequent to that in which it first occurs, is marked superficially by the presence of sich.

The dummy es, which bears the direct object relation, occurs at a non-initial level of the syntactic representation. Similarly, in constructions with an impersonal passive, a transitive level is created in the stratum on which the dummy occurs, enabling the 2→1 advancement. This type of advancement is responsible for the passive morphology in the passive reflexive.

4.2 Structures with MA at Non-Initial-P Level

We will now analyse structures with MA at non-initial levels of the clause, specifically at levels subsequent to the first appearance of the predicate, with reference to example (1c), a so-called ‘anti-passive’ structure (Postal 1977):
Meine Freundin
kauft sich
ein neues Auto
1
P
chô
1, 2
P
chô
1
P
2
Table 8: RG formalization of a structure with MA at non-initial-P level

Just as in a passive construction, the initial level of an anti-passive structure is transitive due to the valence of the verb kaufen, which initializes the subject meine Freundin and the object ein neues Auto. In anti-passive demotion, however, there is no 2→1 advancement in the final stratum, as occurs in passive promotion. Instead, the direct object is subject to a removal process - as a result of a syntactic level in which one nominal bears two grammatical relations - and becomes a chômeur. In both anti-passive constructions - such as Meine Freundin kauft sich ein neues Auto - and properly reflexive constructions (La Fauci 1992: 47) - such as Die Studenten erzählen sich die letzten Neuigkeiten - German direct objects that have become chômeurs are distinguished by the accusative case. This does, however, not mean that they continue to fulfil the role of argument (in fact, these structures cannot be passivized).

The constructions (2a) - (2b), meanwhile, are both initially unaccusative; they both contain final subjects that have the properties of a direct object at some level of the syntactic stratal diagram:
Meine Katze
ängstigt sich
1
P
1, 2
P
2
P
                    Table 9: RG formalization of a structure with MA at non-initial-P level
                                  and subsequent retroherent advancement

Let us consider the hypothesis that structures of this kind contain unaccusative verbs without initial subjects20; in this case, the nominal Meine Katze bears relation 2 at the initial level. The subsequent advancement is retroherent, and therefore nominal 2 acquires relation 1 without relinquishing its previous relation. The MA is then resolved according to the rules, and the reflexive sich appears in the final stratum of the clause.

Construction (2b), the so-called anticausative construction, can be similarly represented:
Die Tür
öffnet sich
1
P
1, 2
P
2
P
                      Table 10: RG formalization of a structure with MA at non-initial-P level
                                      and subsequent retroherent advancement


We propose that the German anticausative structure in example (2b) contains a verb - (sich) öffnen - that can initialise either the subject of a transitive clause Der Hausmeister öffnet die Tür leicht mit einem Schlüssel (The caretaker opens the door easily with a key) or a direct object in the transitive structure Die Tür öffnet sich leicht mit einem Schlüssel (The door opens easily with a key).

As such, German anticausative structures can be represented stratally in a manner similar to constructions containing so-called inherently reflexive verbs, as the verb (sich) öffnen initializes the nominal Die Tür in relation 2. The mode of advancement of the initial object to subject relation is retroherent; an MA occurs in a subsequent stratum of the configuration and is then resolved in the final level of the clause.

There are also causative verbs whose corresponding anticausatives do not take sich in the intransitive clause, for example the verb schmelzen (to melt). In the transitive construction Die Sonne schmilzt das Eis (The sun melts the ice), the causative verb initializes a subject; in the unaccusative structure Das Eis schmilzt (The ice melts), meanwhile, the same anticausative verb initializes a direct object with subsequent plain advancement. As such, there is no reflexive marker.

The MA Hypothesis in Relational Grammar facilitates a unitary account both of German structures with atonic reflexive pronouns that have a transitive counterpart, and of those that, at least in the modern language, do not. Using RG, the various reflexive constructions can be compared without reference to diachronic data, based on the valence of their verbs and considering the modalities of syntactic advancement - retroherent or plain. This also offers an explanation as to why, in identical transitive constructions such as Die Sonne schmilzt das Eis and Der Hausmeister öffnet die Tür, the conditions for the presence of the reflexive morphology only partially occur in the anticausative version.
Structures (3a) – (3b) both feature a 2→1 advancement, either from an initial transitive stratum or from a non-initial stratum of the configuration (3a). Both constructions also feature retroherent advancement:
Das Buch
Pro
liest sich
gut
1
Chô
P
1, 2
Chô
P
2
1
P
Table 11: RG formalization of a structure containing a reflexive personal passive

The structure features 2→1 advancement from the initial transitive stratum, in which the grammatical relation 1 is borne by Pro because the verb lesen initializes a subject (the person who reads) and a direct object (that which is read). The advancement is retroherent. The nominal Das Buch acquires relation 1 without relinquishing its previous relation. The MA produced in this way is resolved by cancelling the lower grammatical relation (Blake 1990: 89). Pro becomes a chômeur, but serves the semantic function of agent in the clause.

That Pro becomes a chômeur is further confirmed by the fact that this element cannot be added through an agent introduced by the preposition von (by).21

Furthermore, grammar rules relating to middle constructions in German require the presence of an adverbial modifier at the final level of the clause.

Syntactic descriptions proposed from an RG perspective in relation to structures such as (3a) - the impersonal middle constructions - are not entirely consistent (Rosen 1988: 98-122). As such, Perlmutter & Postal start from the assumption that a construction containing an Unspecified Human Subject (UHS), as in Es arbeitet sich gut in der Bibliothek “involves a 2→1 advancement from a transitive stratum” (Perlmutter & Postal 1984b: 137) and should, thus, be considered, for all intents and purposes, a passive structure. Likewise, “the advancement from 2→1 in reflexive impersonal passives involves a 2-copy of the advancing nominal” (Perlmutter & Postal 1984b: 137), meaning it contains an MA in a non- initial-P syntactic stratum. The stratal diagram below represents a structure containing a reflexive impersonal passive, as proposed by Perlmuttter and Postal (1984b: 137):  
Es
Pro
arbeitet sich gut
in der Bibliothek
1
chô
P
oblique
1, 2
chô
P
oblique
2
1
P
oblique

1
P
oblique
Table 12: RG formalization of a structure containing a reflexive impersonal passive

Perlmutter and Postal’s approach of assigning both a 2→1 advancement and a retroherent MA to a UHS structure, such as the German impersonal middle construction, also offers an explanation as to why constructions of this kind are formed exclusively with unergative verbs. Indeed, clauses such as *Es sinkt sich schnell im Meer (One sinks quickly in the sea) or *Es kommt sich gut an mit dem Zug (It’s easy to get there by train) do not exist in German, as the verbs sinken and ankommen are unaccusative verbs that have the phenomenal properties of an object. Such structures would not only feature the 2→1 advancement of the dummy from a transitive stratum, but also the unaccusative advancement of the initial object to subject relation as in the table below (Table 13):
*Es
Pro
sinkt sich gut
im Meer
1
chô
P
oblique
1, 2
chô
P
oblique
2
1
P
oblique

1
P
oblique

2
P
oblique
                         Table 13: RG formalization of a reflexive impersonal passive
                                         containing double advancement

Double advancement is impossible, however, as it breaks the One-Advancement Exclusiveness Law which does not allow more than one advancement to subject relation in a clause (Rosen 1988: 104, Perlmutter & Postal 1984a).

The final example (4b) of a construction with a reflexive pronoun features a variant of the so-called passive reflexive, as it contains an unaccusative verb:22
Es
Pro
wird sich geängstigt
1

P
2
1
P

1
P

2, 1
P

2
P
                      Table 14: RG formalization of a structure with MA at non-initial-P level
                                      and 2→1 dummy advancement

In the sentence Es wird sich geängstigt, the unaccusative verb ängstigen initializes an object, resulting in the occurrence of the MA in a non-initial-P stratum. Nominal 2 is subject to retroherent advancement, and acquires relation 1 without relinquishing its previous relation. The MA is thus resolved according to the rules, and sich appears in the final stratum of the clause.

Relation 2 is first allotted to dummy es, which appears in a non-initial stratum of the syntactic representation. As with constructions using the impersonal passive, a transitive level is created in the entrance stratum of the dummy. This transitive level enables the 2→1 advancement, responsible for the passive morphology in the Reflexivpassiv, to occur.

5 The Functional Domain [+MA] in German

The above taxonomy of constructions with reflexive pronouns in German enables the grouping of constructions of different structural types in the functional domain [+MA], thus providing a description of all constructions in which sich appears and offering consistent and uniform answers to the questions raised in Section 2 of this article:
  • With regard to the question of the transitivity or intransitivity of reflexive structures, according to RG, such properties do not relate to the clause structure as a whole; this opposition (bearing in mind the potential distinction between two types of intransitivity) relates to the syntactic configuration and, if this is multi-stratal, to each of its strata. As such, although reflexive constructions are all intransitive in the final stratum, they are nevertheless transitive at some levels of the syntactic configurations;
  • With regard to the convergence of passive and reflexive morphology in the passive reflexive, this article argues that German reflexive structures should all be analysed as examples of MA. The resolution of the MA in a syntactic stratum subsequent to that in which it is produced is superficially marked by the reflexive pronoun. The reflexive pronoun sich is therefore a systematic signal of the resolution of the MA at the final level of the clause.
The MA hypothesis offers an explanation regarding the co-existence of passive and reflexive morphology in the passive reflexive: the MA does not conflict with the 2→1 advancement responsible for passivization, and the presence of a 2, meaning a direct object, is a necessary condition for reflexivization. An area of intersection exists between the passive and multi-attachment domains, centred around transitive verbs. Unergatives that permit passive morphology in impersonal structures are excluded from the reflexive morphosyntax (that requires the presence of a 2, meaning a direct object), and while unaccusatives are not entirely unrepresented, they are not common because, despite being perfectly compatible with reflexive morphosyntax, they conflict, in principle, with passive morphosyntax. It therefore seems reasonable to define the passive reflexive as an extension of the impersonal passive beyond intransitive structures, on account of the element sich, which syntactically marks the presence of an object.


References

Abraham, W. (2005). Deutsche Syntax im Sprachenvergleich. Tübingen: Stauffenburg.

Agel, V. (1997). Reflexiv-Passiv, das (im Deutschen) keines ist. In: Dürscheidt, Ch. (Hrsg.) Sprache im Fokus. Tübingen: Niemeyer, 147-187.

Bierwisch, M. (2006). German Reflexives as Proper and Improper Arguments. In: Brandt, P. / Fuβ, E. (Eds.). Form, Structure and Grammar: A Festschrift Presented to Günther Grewendorf on Occasion of his 60th Birthday. Berlin: Akademie-Verlag, 15-35.

Blake, B. (1990). Relational Grammar. London: Routledge.

Brinker, K. (1969). Zum Problem der angeblich passivnahen Reflexivkonstruktionen in der deutschen Sprache. In: Muttersprache 79, 1-11.

Cardinaletti, A. (1999). Pronouns in Germanic and Romance Languages: An Overview. In: van Riemsdijk, H. (ed.). Clitics in the Languages of Europe. Berlin: Mouton de Gruyter, 33 81.

Davis, W. / Rosen, C. (1988). Unions as Multi-Predicate Clauses. In: Language 64, 52-88.
De Alencar, F. / Kelling, C. (2005). Are Reflexive Constructions Transitive or Intransitive?  Evidence from German and Romance. In: Butt, M. / Holloway King, T. (Eds.). Proceedings of the LF605 Conference.

Dubinsky, S. / Rosen, C. (1987). A Bibliography on Relational Grammar through May 1987 with Selected Titles on Lexical Functional Grammar. Bloomington: Indiana University Linguistics Club.

Duden (Ed.) (2005). Die Grammatik. Bd. 4. Mannheim: Bibliografisches Institut.

Everaert, M. (1986). The Syntax of Reflexivation. Dordrecht: Foris.

Fagan, S. (1992). The Syntax and Semantics of Middle Constructions. New York: Cambridge University Press.

Grewendorf, G. (1984). Reflexivierungsregeln im Deutschen. In: Deutsche Sprache 12, 14 30.

Grewendorf, G. (1989). Ergativity in German. Dordrecht: Foris.

Haider, H. (1985). Von sein und nicht sein: Zur Grammatik des Pronomens sich. In: Abraham, W. (Ed.). Erklärende Syntax des Deutschen. Tübingen: Stauffenberg-Verlag, 223 254.

Helbig, G. (2004). Kleinere Schriften zur Grammatik. Sitta, H. et al. (Eds.). München: Iudicium.

Helbig, G. / Buscha, J. (1998). Deutsche Grammatik. Leipzig: Langenscheidt.

Hundt, M. (2001). Formen und Funktionen des Reflexivpassivs im Deutschen. In: Zeitschrift für Deutsche Sprache 2/02, 124-166.

IDS (Ed.): Das Institut für deutsche Sprache Mannheim. www.ids-mannheim.de. (1.2.2010)
La Fauci, N. (1988). Oggetti e Soggetti nella Formazione della Morfosintassi romanza Nuova Collana di Linguistica. Pisa: Giardini Editori.

La Fauci, N. (1992). Capitoli di Morfosintassi siciliana antica. Tassonomia dei costrutti medi e ausiliari perfettivi. In: A.A.V.V. Studi linguistici e filologici offerti a Girolamo Caracausi. Palermo, 43-73.

La Fauci, N. / Mirto, I. (2003). Fare, Elementi di sintassi. Pisa: Edizioni ETS.

La Fauci, N. / Pieroni, S. (2007). Morfosintassi latina-Punti di vista. Pisa: Edizioni ETS.

Perlmutter, D. (1978). Impersonal Passives and the Unaccusative Hypothesis. In: Proceedings oft the 4th meeting oft the Berkeley Linguistics Society, Berkeley, 157-189.

Perlmutter, D. (1983a). Personal vs. Impersonal Constructions. In: Natural Language and Linguistic Theory 1, 141-200.

Perlmutter, D. (Ed.) (1983b). Studies in Relational Grammar 1. Chicago: University Press.

Perlmutter, D. / Postal, P. (1984a). The 1-Advancement Exclusiveness Law. In: Perlmutter, D. / Rosen, C. (Eds.). Studies in Relational Grammar 2. Chicago: University Press, 81-126.

Perlmutter, D. / Postal, P. (1984b). Impersonal Passives and some Relational Laws. In: Perlmutter, D. / Rosen, C. (Eds.). Studies in Relational Grammar 2. Chicago: University Press, 126-170.

Pittner, K. / Berman, J. (2008). Deutsche Syntax. Tübingen: Niemeyer.

Plank, F. (1993). Peculiarities of Passives of Reflexives in German. In: Studies in Language 17-1, 137-167.

Postal, P. (1977). Antipassive in French. In: Lingvisticae Investigationes 1, 333-375.
Primus, B. (1999). Grammatische Hierarchien. Studien zur Theoretischen Linguistik. München: Wilhelm Fink.

Rosen, C. (1988). The Relational Structure of Reflexive Clauses. Evidence from Italian. New York: Garland Publishing.

Rosen, C. (1997). Auxiliation and Serialization: on Discerning the Difference. In: Alsina, A. / Bresnan, J. / Sells, P. (Eds.). Complex Predicates. Stanford: CLSI, 175-202.

Rosen, C. (2012). Dal giardino della sintassi. Pisa: Edizioni ETS.

Schӓfer, F. (2012). The passive of reflexive verbs and its implications for theories of binding and case. In: Journal of Comparative German Linguistics. Vol. 15, issue 3. Dordrecht: Springer, 213-226.

Steinbach, M. (2002). Middles in German. Amsterdam: Anton J. Benjamins.

Steinbach, M. (2004). Unaccusatives and Anticausatives in German. In: A. Alexiadou, / Everaert, M. (Eds.). The Unaccusativity Puzzle, Oxford University Press, New York, 182-206.

Vater, H. (1995). Zum Reflexiv-Passiv im Deutschen. In: Popp, H. (Ed.). Deutsch als Fremdsprache. An den Quellen eines Faches. München: Iudicium, 185-192.


Author:
Katrin Ziegler
Università degli studi di Macerata
Corso Cavour, 2
62100 Macerata / Italy
Email: k.ziegler62@gmail.com




1 Cf. Perlmutter (1983b), Perlmutter & Rosen (1984), Dubinsky & Rosen (1987); for later developments Rosen (1988, 1997, 2012), Davis & Rosen (1988), La Fauci (1988), La Fauci & Mirto (2003).
2 It should be noted that the theoretical framework of RG has not changed over the course of the last 30 years. The grammatical model, however, while only used by Italianists in relation to Romance languages, has evolved in several ways and, as such, provides solutions to many linguistic problems.
3 Terminology for reflexive forms in German is rather unclear and ambiguous. This article draws mainly on the definitions provided in Duden (2005) and in the Deutsche Grammatik by Helbig & Buscha (1998).
4 This term, translated by Fagan (1992: 45) as “place-holder [which] is restricted to sentence-initial position in main declarative clauses”, describes the specific behaviour of the indefinite pronoun whose specific role in this case is to occupy the first position in the syntactic structure. Es can be omitted, enabling the formulation of sentences such as Morgens wird sich gewaschen instead of Es wird sich morgens gewaschen and Bei Gewitter wird sich geängstigt instead of Es wird sich geängstigt bei Gewitter. The syntactic condition for this omission is that the first position in the clause must be held, if not by the subject, then by another grammatical element such as an adverb or complement.
5 Stratal diagrams are read from the bottom to the top; the initial stratum is the lowest, the final stratum is the highest. Symbols The numbers 1, 2, 3 indicate the grammatical relations of subject and object (direct, indirect), respectively, and the abbreviation P stands for Predicate. (Further relations e.g. instrumentalis, locative or temporal are referred to as oblique). It should be noted that the diagrams in this article only include the relevant facts and do not consider, for example, the distinction between the auxiliary and predicative sectors.
6 RG defines four passive forms that are universally valid, irrespective of the great variety of ways the four types are used in different languages (Perlmutter & Postal 1984b: 126-170):
- The plain personal passive as in Das Buch wurde oft gelesen (The book has often been read).
- The plain impersonal passive as in Es wird hier gut getanzt (One can dance well here).
- The reflexive impersonal passive as in Hier tanzt es sich gut (There is good dancing here).
- The reflexive personal passive as in Das Buch liest sich gut (The book reads well).
7 RG defines the individual syntactic strata as follows (Perlmuttter & Postal 1984a: 95), indicating the grammatical relations of subject and object with the the numbers 1 and 2, respectively:
a. a stratum is transitive if and only if it contains a 1 and a 2
b. a stratum is intransitive if and only if it is not transitive
c. a stratum is subjective if and only if it contains a 1
d. a stratum is objective if and only if it contains a 2
e. a stratum is unergative if and only if it is subjective and intransitive
f. a stratum is unaccusative if and only if it is objective and intransitive
8The properties and behaviour that typify unergative and unaccusative verbs must be specified on a language-by-language basis. Such properties range from the choice of the auxiliary in compound tenses to the admissibility of certain constructions. In order to identify the two unergative or unaccusative classes in Italian, for example, a series of syntactic tests have been devised (Perlmutter 1978, Rosen 1988) which are highly accurate and which have made it possible to “grammaticalize” unaccusativity in Italian, which is clearer than in other languages. The definition of tests appropriate to German has proven more difficult, and this has resulted in a semantic definition of unaccusativity being favoured (Steinbach 2004). This situation will be discussed further in Section 4.2.
9 It should be noted that this study exclusively relates to structures with a single semantically active predicate sector and that, as such, causative and other such structures are not taken into account.
10 All relevant studies to date have looked at Italian reflexives (Rosen 1988); as such, the examples used in this article relate to Italian.
11 Rosen (ibid) deals with the Italian clitic reflexive ‘si’ and examines the morphosyntactic differences between the various structures and different syntactic levels in which it occurs. It should be noted that from an RG perspective ‘si’ is considered as a ‘reflexive marker’, “taken to reflect multiattachment” (Blake 1990: 72), whereas the pronoun ‘se (stesso)’ “is treated like other nominals and allotted appropriate initial stratum relations” (Blake: ibid).
12 Rosen states: “
[…] there exist two types of advancement, plain and retroherent. In the [plain] advancement the advancee acquires a new relation and relinquishes its former one. In a retroherent advancement, the advancee acquires a new relation and retains its former relation besides” (Rosen 1988: 22).
13 An unaccusative verb such as pentire (to repent) uses the retroherent modality for the advancement of the initial object, but cannot take an initial subject: Ugo si pente (Hugo repents), *Lea pente Ugo * (Lea repents Hugo(, *Ugo pente (Hugo repents). An unaccusative verb such as riempire (to fill), however, while using the retroherent modality in the same way to advance the initial object in the structure, takes an initial subject in some constructions: Ugo riempie la bottiglia di vino (Hugo fills the bottle with wine), la bottiglia si riempie di vino (The bottle is filled with wine), *la bottiglia riempie di vino *(The bottle fills with wine). By contrast, the verb arrivare (to arrive) never takes an initial subject and the modality to move on the initial object is plain advancement: Ugo arriva (Ugo arrives), *Pia arriva Ugo *(Pia arrives Ugo), *Ugo si arriva *(Hugo arrives himself).
14 See Perlmutter (1983b) and Rosen (1988) on the silent dummy in Italian.
15 As already stated, in certain syntactic contexts, German traditional grammar clearly identifies that the indefinite pronoun es has two functions. It states that es can either be the formal subject or the place-holder. RG however, in the grammatical contexts in question, only attributes one function to this element, that of the dummy which syntactically maintains the subject relation in the final level of the configuration.
16 It should be emphasized that the passive reflexive is a construction used diastratically and diaphasically with growing frequency at the same rate as it is integrated into standard language.
17 As Rosen states:
According to one classic transformational view (e. g. Lees and Klima 1963), Reflexivization is a rule sensitive to coreference between two nominals, among other conditions. But if the Multiattachment Hypothesis is adopted together with the idea of pronoun birth, the result is that the concept of coreference would not figure at all in the conditioning on Reflexivization (Rosen 1988: 44).
18 This article adopts La Fauci’s proposal regarding analogous romance constructions (La Fauci 1988: 83-84), which assumes the presence of an intermediate level with MA [1, 2] for the resolution of MA in reflexive constructions with a direct object. One advantage of this approach is that it explains why it is impossible to make these structures passive. The assumption of an intermediate stratum, and therefore of a 3→2 advancement, “omogeneizza, per effetto della risoluzione del seguente MA 1,2 sotto questo rispetto proposizioni riflessive con MA 1,3 e proposizioni riflessive con MA 1,2” [ harmonises reflexive clauses with MA 1,3 and reflexive clauses with MA 1,2 leading to the resolution of the MA 1,2 that follows] (La Fauci ibid: 83).
19 Es can only occur in post-initial strata, otherwise it would defy the ‚Stratal Uniqueness Law‘.
20 The diagnostics of German unaccusative structures is anything but straightforward. Steinbach states that
“the diagnostics for unaccusativity in German yield no clear evidence for the claim that unaccusativity in German must be encoder (sic!) in syntax […] unaccusative verbs differ from unergative verbs at least in their semantics and selectional properties. Typical unaccusative verbs select a semantic argument which has proto-patient properties as opposed to typical unergative verbs, which select a semantic argument with proto-agent properties”. (Steinbach (2004: 182-183)
Semantically, the verb ängstigen with a reflexive marker describes an event that is not controlled by its protagonist, expressing a state of being or feeling that does not depend on the will of the person involved in the action. A verb of this type can be considered unaccusative and can therefore syntactically initialize an object, in this case the nominal meine Katze. It should be noted that in German, unlike in Italian, it is somewhat difficult to base the distinction between classes of unaccusative or unergative verbs on syntax. A series of diagnostic tests exist to define the different properties of the two sub-groups. For example, unaccusative verbs cannot be found in impersonal passive structures, nor can they be nominalized (Grewendorf 1989, Haider 1985, Primus 1999, Abraham 2005).
21Rosen observes that:
the two varieties of Passive - plain and retroherent - differ not only in their morphological concomitants, but also in the conditions that govern their appropriateness. One clear fact is that the chômeur of a Plain Passive can be either overt or unspecified (…) whereas the chômeur of a Retro Passive can never be overt […]. (Rosen 1988:86)

22 Passive reflexive structures basically involve transitive verbs (as in Es wird sich gewaschen) or unaccusative verbs (as in Es wird sich geängstigt). The passive reflexive never occurs with unergative verbs (as in *Es wird sich gut gearbeitet). An examination of the corpora (Institut für deutsche Sprache, Hundt 2001, Vater 1995) suggests that the passive reflexive is rarely constructed with unaccusative verbs. The few exceptions are not entirely convincing, and the verbs involved seem to have previously been transitive. (Fagan (1992: 237) for further information regarding the transitive variants of inherently reflexive verbs in German.)