On the Syntax of Instrumental Clauses: The Case of Indem-Clauses in German

Abstract

In this article, I examine the external and internal syntax of instrumental indem-clauses in German. As a subordidating conjunction, indem takes a finite TP as its complement and triggers verb final position. I provide evidence showing that instrumental indem-clauses can only operate on the content level and that they cannot be interpreted epistemically, nor can they modify a speech act. Furthermore, I argue that although indem-clauses are restricted to a particular interpretation, they can attach at two distinct heights in the matrix clause. If they are analyzed as central adverbial clauses, they attach as T[ense]P[hrase] adjuncts. If, on the other hand, instrumental indem-clauses are treated as peripheral adverbial clauses, they are taken to be J[udge]P[hrase] adjuncts. Main evidence for the analysis comes from: i) variable binding and Principle C effects, ii) movement to the left periphery of the matrix clause, and iii) licensing conditions of weak and strong root phenomena.

1. Setting the Scence

The main aim of this article is to examine the syntax of subordinate clauses headed by the instrumental complementizer indem ‘by’ in modern German, cf. (1).

(1)

Syntactically, indem heads a subordinate clause, connects it with the matrix clause, and triggers the final position of the finite verb. Semantically, indem introduces, as Sæbø (2015, p. 1435) puts it, an action as an instrument of another action. In (1), the action of quick heading for the hotel Pequeno Paraíso (and questioning its staff) is understood as an instrument of the action of quick arranging of Nadine Weimer’s identification. The task of the instrumental clause is to elaborate on the content of the matrix clause by providing more specific content.

The main argument supporting the view that indem-clauses are instrumental clauses comes from the observation that they are asked for by means of the wh-phrase how that can address methods, means, manners, and instruments, as exemplified in (2) (for more details on the diversity of how-questions the interested reader is referred to Jaworski (2009) and Sæbø (2015).

(2)	A:	How did Judith handle the problem?
	B:	i) With great patience.
		ii) With the help of an advanced software tool.
		iii) By fixing the costs, taking a loan and hiring an expert.
		(Jędrzejowski & Umbach, 2023, p. 3, ex. 6)

The subordinate indem-clause can, as well, be used as an answer to a wie-question, as the dialogue presented in (3) shows:

(3)

I therefore follow Sæbø’s (2015) definition in labeling indem-clauses as instrumental clauses.1

Different studies have addressed the use of indem in modern German, cf. Behaghel (1905), Stojanova-Jovčeva (1976), Pusch (1980), Homberger (1996), Zifonun et al. (1997, pp. 1148, 2277–2278), Fabricius-Hansen (2000, pp. 337–338; 2011), Engel (2004, p. 281), Breindl et al. (2014, pp. 582–588), Bücking (2014, pp. 31–35), Sæbø (2015, pp. 1435–1436), Pittner (2016, p. 517). However, not much is known about the syntax of indem-clauses. The main aim of this article is therefore to examine their external and internal syntax, and to contribute to a deeper understanding not only of manner expressions but also of adverbial clauses.

This article is organized as follows. In Section 2, I set out the syntactic taxonomy of adverbial clauses in terms of their dependency relationship to the matrix clause. Section 3 is devoted to indem-clauses at the syntax-semantics interface. Whereas in Section 3.1, I outline semantic properties and restrictions of indem-clauses, in Section 3.2 external syntax is examined. In Section 4, I investigate internal syntax of instrumental indem-clauses and elaborate on their variation. Section 5 offers some cross-linguistic observations. Finally, Section 6 summarizes the most important findings and addresses open questions for future work.

2. Syntactic Typology of Adverbial Clauses

In what follows, I briefly delineate the syntactic typology of adverbial clauses advocated by Haegeman (2006, 2010, 2012, 2021), Badan and Haegeman (2022), Schönenberger and Haegeman (2023), and Frey (2011, 2012, 2016a, 2023a, 2023b, 2023c, 2023d). The typology is the theoretical basis for the analysis of indem-clauses that is given in the sections to follow. Since this article is about an adverbial clause type occurring in German, I present the typology using German data as well.

Adverbial clauses have been divided into three main groups, depending on the extent to which they are syntactically embedded into the matrix clause and at what structural height they attach to it, cf. (4).

(4)	Three types of adverbial clauses:
	a.	central adverbial clauses (abbreviated as CACs)
	b.	peripheral adverbial clauses (abbreviated as PACs)
	c.	non-integrated adverbial clauses (abbreviated as NACs)

CACs are syntactically embedded clauses and attach at the TP level of the matrix clause, while PACs are considered less dependent and are often analyzed as JP2 adjuncts. NACs are taken to be syntactically independent clauses having their own illocutionary force. Different tests are used to keep them apart, cf.

Table 1. Selected tests for differentiating between the three types of adverbial clauses.

	CACs	PACs	NACs
variable binding	+	–	–
under the scope of matrix negation	+	–	–
movement to Spec-CP of the matrix clause	+	+	–
embedding together with its host clause	+	+	–
possible occurrence of weak root phenomena	–	+	+
possible occurrence of strong root phenomena	–	–	+

.

CACs and PACs differ in two main respects. Whereas the former allow variable binding, (5a), and can fall within the scope of matrix negation, (6a), the latter disallow variable binding, (5b),3 and are not sensitive to the elements occurring in the matrix clause, e.g., to negative operators, (6b).

(5)	a.
		(Frey, 2016a, p. 156, ex. 5a)
	b.
		Intended: ‘No colleague seems recovered, although he has been on a long
		vacation.’                    (Frey, 2011, p. 53, ex. 24b)

(6)
	(Eberhardt, 2017, e71, ex. 17)
	(Frey, 2020, p. 4, ex. 7b)

Based on the contrasts in (5) and in (6), we observe that causal weil-clauses should be analyzed as CACs, whereas concessive obwohl-clauses are either PACs or NACs. Since they can occur in the prefield position of the matrix clause, (9b), they clearly qualify as PACs.

A further difference between CACs and PACs concerns the occurrence of root phenomena. They are divided into weak and strong root phenomena. Weak root phenomena involve a subjective judgement that is often expressed by modal particles and adverbs associated with the C-domain. Strong root phenomena, in turn, are primarily phenomena of spoken language that express a modification or a qualification of a speech act and contain question tags and interjections. An overview of selected root phenomena in German is presented in (7).

(7)	Root phenomena in German
	a.	weak root phenomena:
		i.	modal particles: ja, denn, wohl, eben, doch, halt, einfach, eigentlich
		ii.	adverbs: wahrscheinlich ‘probably’, angeblich ‘allegedly’, leider ‘unfortunately’, glücklicherweise ‘fortunately’, offensichtlich ‘obviously’, offenbar ‘apparently’, anscheinend ‘seemingly’
	b.	strong root phenomena:
		i.	question tags: oder etwa nicht? ‘isn’t that so’, hab ich recht? ‘am I right?’, nicht wahr? ‘isn’t that so?’
		ii.	interjections: Mann ≈‘Dude’, verdammt ‘damn’

By using the modal particle ja, as Thurmair (1989) argues, the speaker asserts p and indicates that p is known to both speaker and hearer. Grosz (2021, p. 9) points out that there are two components of ja: “(i) some type of factuality component – that is, a speaker of a ja-containing utterance conveys at a non-truth-conditional level (e.g., as a presupposition) that the modified proposition is true; and (ii) some type of uncontroversiality component – that is, the notion that the modified proposition is uncontested”, (8).

(8)	$〚ja〛$(p) ⇝$\underset{\underset{\mathrm{FACTUALITY}}{︸}}{p\mathrm{is}\mathrm{true}}$ and $\underset{\underset{\mathrm{UNCONTROVERSIALITY}}{︸}}{\mathrm{speaker}\mathrm{believes}p\mathrm{uncontroversial}}$ (Grosz, 2021, p. 9, ex. 13)

(9)
	(Frey, 2016a, p. 167, ex. 42b)
	‘Although Max was frequently interrupted, he remained calm.’
	(Frey, 2020, p. 8, ex. 23b)

As the contrast between (9a) and (9b) shows, causal weil-clauses used as CACs do not allow weak root phenomena, concessive obwohl-clauses, on the other hand, do not exhibit such a restriction, leading to the conclusion that they cannot be treated as CACs.4

NACs behave differently and contrast to both CACs and PACs in several respects, too. As they are syntactically not embedded and possess their own illocutionary force, they can host expressions modifying a speech act. Questions tags like hab ich recht? (‘am I right?’) add an inquisitive illocutionary act to an assertion and invite the hearer to give a yes-no answer. The interjection Mann ‘Man’, in turn, leads to, as Frey (2020) observes, a (rather rude) strengthening of the illocutionary force, and expresses a strong emotional involvement of the speaker concerning p. Their licensing in adverbial clauses depends on the clause type. Let us stick to concessive clauses to illustrate some differences. Concessive clauses can be headed by the dedicated complementizer although ‘although’ or be introduced by moving the finite verb to the C⁰ position, by leaving the prefield position empty, and by using the modal particle auch (lit. ‘also’).5 While, in the former case, strong root phenomena are ruled out, (10a) and (10c), verb first concessives do not exhibit such restrictions, as (10b) and (10d) persuasively show (taken from Frey, 2020).

(10)

We can therefore conclude that verb first concessives should be analyzed as NACs. By virtue of being NACs, they are also expected to host weak root phenomena. This prediction is borne out, cf. (11a) for the modal particle ja, and (11b) for the epistemic adverb wahrscheinlich ‘probably’.

(11)

Furthermore, verb first concessive clauses disallow variable binding, (12a), and are not sensitive to material occurring in the matrix clause, (12b), similarly to PACs. Unlike PACs, however, they cannot appear in the Spec-CP position of the matrix clause, (12c), and they are not embeddable together with the host clause, (12d). The following examples are taken from Frey (2020) and demonstrate these restrictions.

(12)

In his recent work, Frey (2023c, 2023d) refines the typology of adverbial clauses and proposes to distinguish two distinct types of clauses appearing outside the matrix clause structure, (13).

(13)	External adverbial clauses:
	a.	world-related external adverbial clauses (abbreviated as W-ExtAdvCls)
	b.	discourse-related external adverbial clauses (abbreviated as D-ExtAdvCls)

W-ExtAdvCls refer to the world or to a conceptualization of the world and contribute to the propositional meaning of an utterance, whereas by using D-ExtAdvCls the speaker links his or her speech act to a discourse context and to discourse participants. Furthermore, D-ExtAdvCls do not contribute to the propositional meaning of an utterance and have a conversation controlling function. Verb first concessive clauses, (10b), and irrelevance conditional clauses, (14a), for example, are considered W-ExtAdvCls. In turn, relevance conditionals, (14b), and speech act related adverbials, (14c), are analyzed as D-ExtAdvCls.

(14)
	(d’Avis, 2004, p. 141, ex. 2a)
	(Thim-Mabrey, 1988, p. 56)
	(Pittner, 1999, p. 357, ex. 71)

What W-ExtAdvCls and D-ExtAdvCls have in common are two properties: i) They are NACs. ii) They often trigger the so-called verb third position in the matrix clause, whereby the adverbial clause occurs to the left of the matrix clause, Spec-CP is occupied by a constituent, and the finite verb follows. However, W-ExtAdvCls and D-ExtAdvCls also differ. For example, W-ExtAdvCls must follow D-ExtAdvCls if they co-occur to the left of the matrix clause, cf. (15a) vs. (15b). If, on the other hand, they co-occur to the right of the matrix clause, D-ExtAdvCls must follow W-ExtAdvCls, cf. (15c) vs. (15d) (examples adopted from Frey, 2023c).

(15)

Frey (2023d) also observes further striking differences. i) W-ExtAdvCls can be used parenthetically without strong prosodic signs of separation, whereas D-ExtAdvCls behave more like an interruption. ii) Only W-ExtAdvCls can follow a coordinating conjunction. iii) D-ExtAdvCls can stand alone more easily than W-ExtAdvCls without a uniquely given host. iv) The semantic scope of D-ExtAdvCls, but not that of W-ExtAdvCls, may encompass the content of several subsequent sentences. I will not discuss these differences in detail here because as we will see below indem-clauses cannot appear outside the matrix clause, meaning that they can be used neither as W-ExtAdvCls nor as D-ExtAdvCls. It should be kept in mind, though, that W-ExtAdvCls and D-ExtAdvCls constitute two distinct types of clauses that operate on two distinct levels. While W-ExtAdvCls are handled by sentence grammar, D-ExtAdvCls are handled by discourse grammar.

Bringing together all the observations made in this section leads to five distinct attachment heights of adverbial clauses in general:

(16)	Adverbial clauses can attach at the following heights:
	a.	CACs attach as TP adjuncts,
	b.	PACs attach as JP adjuncts,
	c.	NACs attach as ActP adjuncts,
	d.	W-ExtAdvCls (in the left outerfield (= Außenfeld) position) are ActP adjuncts, and
	e.	D-ExtAdvCls attach as adjuncts via cooptation from the domain of sentence grammar to the separated domain of discourse grammar (cf. Kaltenböck et al., 2011).

In Section 3.2, I examine the syntax of indem-clauses to figure out whether they can be analyzed as CACs, PACs, NACs, W-ExtAdvCls, or D-ExtAdvCls. But first I briefly describe their meaning.

3. Instrumental Indem-Clauses in Modern German

3.1. Semantics

The meaning of indem-clauses has been examined in Behrens and Fabricius-Hansen (2002) and Bücking (2014). In this section, I summarize the most important findings from both studies and elaborate on the issues that are crucial for a better understanding of the syntax of indem-clauses.

Indem-clauses are often considered classical cases of elaboration, whereby, as Halliday (1994, p. 220) puts it, “one clause expands another by elaborating on it or some portion of it by restating in other words, specifying in greater detail, commenting, or exemplifying”. Bücking (2014, p. 25) argues that the embedded event must be conceivable as one of the matrix type. In (1), for example, the author elaborates on the way that the matrix subject acted to identify Nadine Weimer, namely by immediate investigation of the hotel Pequeno Paraíso.

The elaboration definition of instrumental indem-clauses given above means that we take the embedded proposition to describe the same eventuality as the proposition expressed in the matrix clause in a more specific or less abstract manner. Such a dependency relationship between the matrix clause and the subordinate clause may often impose semantic restrictions on the latter. One of the restrictions, as pointed out by Behaghel (1905) and specified by Behrens and Fabricius-Hansen (2002) for instrumental indem-clauses, is subject identity:

“In order for the elaborating clause to contribute with an event which ‘matches’ the underspecified subevent in the matrix, Agent Identity is required.”

(Behrens & Fabricius-Hansen, 2002, p. 47)

“[…], the indem-clause […] specifies the causing subevent of the matrix clause event (complex); thus, indem relates the two clauses by way of Elaboration […], introduces an event referent $e2$ to be identified with the causing event referent $e^{\prime }$ that is presupposed by the main clause (S1). This means, among other things, that there must be an Agent of $e2$ to be identified with x—the Agent of the matrix clause event (complex) $e1$. Likewise, the time and location of $e2$ must be identical to or part of the time and location of $e1$.”

(Behrens & Fabricius-Hansen, 2002, p. 48)

The restriction of the subject’s identity was also noticed by Stojanova-Jovčeva (1976, p. 116), who argues that indem-clauses can be analyzed as adversative, temporal or modal adverbial clauses and who shows by means of corpus examples that it is too strong because subjects may differ. Since I focus solely on instrumental indem-clauses in the present paper, the question of how non-instrumental indem-clauses behave will not be discussed here. However, I think that instrumental indem-clauses do not always require subject identity, as the following example illustrates:

(17)

It clearly follows from (17) that self-regulating forces are expected to make the complaints disappear. They cannot be co-referential, nor can they be activated one by another. No subject identity is therefore involved in (17). The content of the indem-clause itself, in turn, is conceived of as the instrument by which we can make complaints disappear. In addition, (17) highlights two related issues. First, it indicates that agentivity is not the right notion to invoke, as verschwinden ‘disappear’ has no agent argument at all, and although the passive in the indem-clause has an implicit agent, it is not in any obvious sense identified with an argument of the matrix clause. Second, we might want to salvage some sort of identity condition on indem-clauses by arguing that in (17), and similar such examples, the identity requirement plays itself out at the level of implicit arguments of the predicates involved (the implicit agent of anregen ‘stimulate’, and perhaps a source argument in the case of verschwinden ‘disappear’). It seems highly unlikely, however, that in the particular case of (17) such an approach would get much mileage.6

The second lexical restriction observed in the literature refers to the type of eventualities the indem-clause can host. Behrens and Fabricius-Hansen (2002) claim that indem is restricted to, or preferably occurs with, clauses that describe activities or accomplishments. Bücking (2014, pp. 31–32) states this restriction more precisely. First, he shows that statives such as gefallen ‘please’, wiegen ‘weigh’, ähneln ‘resemble’, auffallen ‘stand out’, and constructions with the copula be are deviant with indem, cf. e.g. (18). Second, as the contrast between (19a) and (19b) clearly indicates, the stative restriction applies only to the proposition expressed in the subordinate clause, and not to the proposition encoded in the matrix clause.

(18)
	(Bücking, 2014, p. 31, ex. 45a)

(19)
	(Bücking, 2014, p. 31, ex. 46a)
	(Bücking, 2014, p. 32, ex. 46b)

Correspondingly, indem requires an event controlled by an agent, whereas the predicate of the matrix clause allows for all aspectual classes.

Finally, I would like to elaborate on one issue that has not previously been examined in connection with indem-clauses.

Adverbial clauses can usually operate on three cognitive levels: i) the content level, ii) epistemic level, and iii) speech act level, giving rise to three distinct interpretations. Frey (2016a) persuasively demonstrates that causal weil-clauses in modern German can operate on all three levels:

(20)
	‘Since you are always interested in Maria, she is ill.’ (Frey, 2016a, p. 154, ex. 1a–c)

In the content domain, (20a), the proposition embedded in the causal clause is interpreted as a fact causing another fact. Concretely, the fact that Maria is ill is an explanation for why she is pale. Following Davidson (1967), R. Larson (2004, pp. 31–34), Frey (2016a, pp. 168–169), and R. K. Larson and Sawada (2012), I argue that weil in (20a) relates two eventualities (instead of two propositions), and refer to such clauses as eventuality related clauses (= erc).7 A different interpretation arises on the epistemic level. In (20b), the speaker specifies the reason for why she or he thinks the content of the matrix clause is true. Based on this line of reasoning, the speaker takes the event of Maria’s being pale to be a reasonable argument by which to assume that she is ill. There is some (in)direct evidence perceived by the speaker that allows her or him to make such an assumption. Here I follow Morreall (1979) by labelling such clauses as evidential clauses (= ec). Finally, the weil-clause in (20c) provides the motivation for why the speaker is performing a speech act. A natural paraphrase of (20c) could then be I’m telling you that Maria is ill, and the reason why I’m making this statement is that you are always interested in her. I refer to such clauses as speech act related clauses (= sar). What is interesting about (20a)–(20c) is that they receive a different semantic interpretation, although they are all headed by a single subordinating conjunction.8

The observation that causal weil-clauses can operate on these three cognitive levels is not surprising for two main reasons. First, causal clauses in other languages can express all three relations too, cf. Morreall (1979), Sweetser (1990) for because-clauses in English and Ángantýsson and Jędrzejowski (2023) for af-því-að-clauses in Icelandic. Second, some other types of adverbial clauses can also operate on these three cognitive levels, cf. van Dijk (1977, pp. 68–76), Takami (1988), and Ros (2005) for conditional clauses, Breindl (2015a) for irrelevance conditional clauses, Breindl (2015b), Antomo and Steinbach (2013), and Frey (2020) for concessive clauses, and Jędrzejowski et al. (2022) and Angantýsson et al. (2023) for exceptive clauses.

How do instrumental indem-clauses behave? It has been observed in the literature (see

Table 2. Interpretations of causal weil- and instrumental indem-clauses in modern German.

	eventuality related	evidential	speech act related
weil-clause	+	+	+
indem-clause	+	–	–

) that they are restricted to eventuality related interpretations, (21a):

[…]; vollspezifiziert schließlich sei indem, das ausschließlich propositionale Verknüpfungen zulässt.

(Breindl et al., 2014, p. 189)

‘[…]; finally, the use of indem is fully specified, as it can operate on the content/propositional level only.’

Both evidential and speech act related interpretations are ruled out, as (21b) and (21c) show, respectively.

(21)

It is not clear to me why instrumental indem-clauses cannot operate on the other two levels. As far as the evidential domain is concerned, it is hard to imagine a context in which an event is taken to specify another event and in which, at the same time, the latter event is taken by the speaker to be evidence for the assumption that the matrix clause might be true. As for the speech act level, it is rather difficult to imagine that the speaker uses an instrumental clause as motivation for uttering a speech act. There might be other ontological reasons for these gaps and more reasoning is needed in this respect.

Nevertheless, we can already safely conclude at this point that indem-clauses cannot be D-ExtAdvCls because they cannot operate on the discourse level. If indem-clauses then operate on the sentence level only, and are not available in the discourse domain, we expect them to be a target of linguistic material, i.e. of anaphoric expressions. This prediction is borne out:

(22)

In (22), in the response of speaker B, the anaphoric expression das ‘that’ takes up the proposition that the embedded subject ignored someone tonight. This is supported by the continuation phrase er hat dir zugezwinkert ‘he gave you a wink’ when speaker B disagrees with speaker A.

I now move on to examine the syntactic properties of indem-clauses to better determine the degree of their dependency.

3.2. Syntax

To the best of my knowledge, the syntax of indem-clauses has not been investigated in the literature. The aim of this section is to fill this gap.

3.2.1. Verb Position Within the Indem-Clause

German is an asymmetric language with respect to the position of the finite verb. In independent declarative clauses, the finite verb occupies the second position in the clause structure by moving from the V⁰ to the C⁰ position, (23a), whereas in subordinate clauses it remains in situ, (23b). The V-to-C movement is blocked by a subordinating element that indicates a dependency relationship between the matrix clause and the subordinating clause; cf. den Besten (1983) and Haider (1986; 2010, pp. 45–85), among many others, for more details.

(23)

In (23b), the declarative complementizer dass ‘that’ blocks movement of the finite verb liest ‘reads’ to the C⁰ position. It introduces a subordinate clause and establishes a dependency relationship between two propositions (for the meaning of that see Kratzer (2006) and Moulton (2009)). Likewise, indem blocks movement of the finite verb griff ‘grasped’ to the C-domain in (24). The finite verb must remain within the VP domain.

(24)

Syntactically dass and indem have three properties in common: a) they are base-generated as C-heads, b) they make the finite verb stay in situ, and c) they take finite clauses as their complements.9 The verb final position in the indem-clause points to the presence of a subordinate clause.

One of the anonymous reviewers of this article raised a question about the morphological make-up of indem, and wonders whether it is a complementizer or a preposition taking a case-inflected demonstrative as its complement. I argue that indem is a complementizer, i.e., a C-head, in modern German. In what follows, I outline two arguments supporting this view, one from the diachrony and one from the synchrony of indem-clauses.

First, in older German indem-clauses are often combined with the declarative complementizer dass ‘that’, (25), as pointed out by Paul (1897, pp. 232–233), Behaghel (1928, pp. 189–192), Ebert et al. (1993, p. 477), Breindl et al. (2014, pp. 582–583), and Greisinger (2016, pp. 120–123), to mention just a few.

(25)	so lauffen  die Blättlein, indem daß sie  trocknen, also zusammen
	so run.3pl the leaves  indem  that they dry.3pl  so together
	‘so fall the leaves by drying, that is, together’
	(Peter Pomet, 1717, Der aufrichtige Materialist und Specerey-Händler, p. 176; cit. in Greisinger, 2016, p. 122, ex. 39)

(26)	[PP inP0 [DP demD0[CP dassC0 […]]]] →
	[CP [indem dass]C0 […]] →
	[CP [indem]C0 […]]                  (Greisinger, 2016, p. 122)

I agree that indem used to be a PP, but it changed its status diachronically by developing into a C-head. According to Greisinger (2016), this process happened in the 18th cent., cf. (26). In (25), indem co-occurs with daß ‘that’, whereby the latter’s function is to introduce a subordinate clause and to make the finite verb remain in situ. It does not contribute any meaning. Over time, indem takes over the syntactic function of the declarative complementizer, making its presence redundant. Although this is one of the last steps of this language change process, it clearly shows, at least from a diachronic point of view, that indem is a C-head in modern German.

Second, S. Müller (1999, pp. 100, 202–203) observes that a PP can be extracted from another PP:

(27)
	(S. Müller, 1999, p. 100, ex. 9.30a)

In (27), the PP mit Norwegen ‘with Norway’ is base-generated within the PP in einem langfristigen Stellungskrieg ‘in a long-term positional war’ and moved to the Spec-CP position, crossing the finite verb in the C-position. If instrumental indem-clauses were PPs, we should be able to extract other PPs from them. This is, however, not the case.

(28)

If in were a P-head in indem, we should be able to extract a PP out of the indem-clause. But as (28b) and (28c) convincingly show, PP extraction leads to ungrammatical results. This problem disappears once we assume that indem is a C-head. In this case, it constitutes a barrier in the sense claimed by Chomsky (1986) and blocks movement.

Another synchronic argument underpinning the claim that indem is a C-head is discussed in Section 3.2.4 below.

In what follows, I examine the extent to which indem-clauses depend on the matrix clause.

3.2.2. Constitutent Status

When subordinate clauses are constituents of their matrix clause, they can be usually asked about using a wh-question and used as an answer to this question. (3), repeated below as (29), shows that indem-clauses are constitutents of the matrix clause.

(29)	A:	Wie hatte      er ihr    geholfen?
		how have.3sg.pst he her.dat help.ptcp
		‘How did he help her?’
	B:	Indem er auf Anhieb das Pequeno Paraíso angesteuert hatte.
		indem   he right away the Pequeno Paraíso head:for.ptcp have.3sg.pst
		‘By heading for the hotel Pequeno Paraíso right away.’

Asking a question with the wh-phrase how ‘how’, speaker A asks for the way in which someone was helped. Speaker B replies by using an indem-clause. In doing so, she or he specifies the action of helping.

3.2.3. Position of the Indem-Clause

Behrens and Fabricius-Hansen (2002, p. 48, fn 2) mention in passing that “indem-clauses may follow or precede their matrix clause, or occur in the so-called middle field, in the medial position of a free adverbial”. In fact, instrumental indem-clauses are not restricted to any particular position in the matrix clause. They can occur on its right edge following the matrix VP (= Nachfeld), (30a), in the middle field, (30b)10, and in the prefield (Vorfeld) position of the matrix clause, i.e. in Spec-CP, (30c):

(30)
	‘Presumably, the seaman will get emotional effects by apprehending the victims’
	sorrow, however […].’        (Stojanova-Jovčeva, 1976, p. 113, ex. 2)

Behrens and Fabricius-Hansen (2002, p. 48, fn. 2) point out that indem-clauses occur most frequently in the Nachfeld position. Similar observations are made in Stojanova-Jovčeva (1976, p. 113), who examined 303 occurrences of indem-clauses in different genres. In her corpus, indem-clauses predominanly occur on the right edge of the matrix clause

Table 3. Position of the indem-clause in Stojanova-Jovčeva’s (1976) corpus.

Vorfeld	Mittelfeld	Nachfeld	‘standalone’
11.9%	6.6%	79.2%	2.3%

:11

Based on the order of subordinate clauses proposed by Haider (1994, 1995, 1997) and Pittner (1999), (31), indem-clauses used as adverbial clauses are expected to occur on the right edge of the Nachfeld, i.e., they are expected to follow both relative and complement clauses.

(31)	[Nachfeld relative clause > complement clause > adverbial clause]

This expectation is borne out, as (32) and (33) convincingly show (whereby the symbol ‘>’ means ‘precedes’ in terms of the linear order of subordinate clauses).12^,13

(32)	a.
	b.	indem-clause > relative clause
		??Der Anwalt kümmert sich um die Angelegenheit, indem er mit dem
		Finanzamt verhandelt, die seinen Mandanten nicht schlafen lässt.

(33)

If indem-clauses co-occur with other types of adverbial clauses, they usually precede them. If they occur after them, the less natural word order seems to be driven by information-structural reasons.

(34)
	‘She offended him because she felt hostility towards him by not taking notice of
	him.’                    (Pittner, 1999, p. 314, ex. 222a,b)

While in (34a) the indem-clause precedes the causal weil-clause, the reverse order is given in (34b). The former order is strongly preferred over the latter one, and it sounds more natural. This is to some extent expected if we recall the possibility of occurring indem-clauses in place of canonical complement clauses headed by the declarative complementizer dass ‘that’, cf. Section 3.2.4 below for more details.

Furthermore, indem-clauses cannot appear outside the matrix clause structure, in the so called Außenfeld position, (36), and trigger verb third position in the matrix clause, as the examples in (35), slightly modified versions of (30), show:

(35)	a.	*[CP Indem er Renato unter die Schultern griff], [CP [Spec-CP er] [C0 befreite] sie von der Last.
	b.	*[CP Indem er das Leid der Betroffenen erfasst], [CP [Spec-CP Seemann] [C0 gewinnt] wohl emotionale Wirkungen, aber […].
	c.	*[CP Indem er anderen half], [CP [Spec-CP er] [C0 würde] indirekt auch das Unrecht bekämpfen], das ihm selbst widerfahren war.

(36)	Instrumental indem-clauses as potential W-ExtAdvCls:

The ungrammaticality of (35a)–(35c) provides strong evidence for one important observation, viz. indem-clauses cannot be W-ExtAdvCls, i.e. adverbial clauses that occur in the left outerfield (= Außenfeld) position, that attach as ActP adjuncts, and that are still handled by sentence grammar, contributing to the propositional meaning of the utterance.

So far, we have observed that indem-clauses can be neither D-ExtAdvCls nor W-ExtAdvCls. Now, we need to figure out whether they are CACs, PACs or NACs.

3.2.4. The Proforms Damit and Dadurch

If we take indem-clauses to encode an instrumental relation, we expect them to be able to refer to a corresponding proform. Stojanova-Jovčeva (1976, p. 116) could not identify any examples in her corpus that would contain a proform in the matrix clause and that refers to the content of the indem-clause. Likewise, Pittner (1999, p. 291) and Eisenberg (2020, p. 362) argue that adverbial indem-clauses cannot refer to any correlates in the matrix clause. However, in modern German the proform damit contains the preposition mit ‘with’, which takes DPs as its arguments and often gives rise to an instrumental interpretation of the PP, cf. e.g., [_PP mit [_DP einem Schraubenzieher]] ‘with a screwdriver’. We therefore expect indem-clauses to refer to the proform damit. As the corpus examples registered in (37) demonstrate, this prediction is borne out.

(37)

Another preposition that is often taken to encode instrumental relations is durch ‘through’. It is therefore not surprising that indem-clauses can be co-referential with the proform dadurch:14

(38)

In all the cases listed in (37) and (38), indem can be replaced by the declarative complementizer dass ‘that’:

(39)	a.	Sie behelfen sich [damit]i, [dass sie jede Folge viermal pro Woche wiederholen]i.
	b.	Munsi tröstet sich [damit]i, [dass sie sich auf das Positive ihrer Amtszeit fokussiert]i.
	c.	Leader wird man [dadurch]i, [dass man Leistung erbringt]i.
	d.	Man helfe den Opfern oft schon [dadurch]i, [dass man ihnen einfach zuhöre]i.
	e.	Menschen definieren sich [dadurch]i, [dass sie sich mit anderen vergleichen]i.

I argue that instrumental indem-clauses are base-generated within a PP in the matrix clause when they refer to one of the da-proforms.

(40)

In (40a), the indem-clause is base-generated within the PP and then extraposed to the right edge of the matrix clause, (40b). Evidence for this account comes from the following observations. First, both the proform and the indem-clause occur adjacent to each other in the middle field, (40c), and move together to Spec-CP, (40d). Second, it is not possible to prepose the matrix VP along with the indem-clause and to simultaneously leave the proform in the middle field, (40e). Finally, it is not possible to move the indem-clause to Spec-CP without the proform, (40f).

A detailed syntactic analysis of the cases involving da-proforms is beyond the scope of the present article, and interested readers are referred to G. Müller (1995) and Frey (2016b) and references cited therein for more details. For the purpose of this article it suffices to observe that instrumental indem-clauses can refer to proforms. We should, in addition, stress that the availability of a profom and its co-referentiality with the instrumental indem-clause always seems to be possible when the embedded proposition can be instrumentalized and when the instrumental theta-role is assigned to it.

One of the anonymous reviewers of this article wonders whether instrumental indem-clauses can also refer to the proform darin, which contains the preposition in ‘in’. If so, then the in- in indem cannot be considered a preposition from a synchronic point of view. According to the account developed in the present paper, we expect darin and the indem-clause to co-occur as soon as the criteria mentioned above are met. One case in point is given (41).

(41)

In (41), the indem-clause elaborates on the strength of the oil and specifies its function. They are linked through the clause-embedding predicate liegen in ‘lie in’. It takes the proform darin, which is referentially linked to the proposition in the subordinate clause which, in turn, is headed by the complementizer indem. Its main function is to instrumentalize the embedded proposition. Cases like (41) strongly suggests that instrumental indem is a C-head, and not a PP consisting of the preposition in ‘in’ and its nominal dative-marked complement dem ‘that’, contrary to what Axel (2002, pp. 35–37) claims for adverbial clauses in German in general; see also Section 3.2.1 for more arguments against the PP status of indem.

Another anonymous reviewer raises the question of whether instrumental indem-clauses can refer to the proform so ‘so’. Based on my understanding, this is barely possible and found only occasionally:15

(42)

Since the pattern illustrated in (42) instantiates a peripheral phenomenon of German grammar, I will not examine it here. However, the question of why so appears to be hardly compatible with the instrumental indem-clause still remains an important one and should be addressed in future work;16 for more details on so, see König and Umbach (2018).

3.2.5. Negation Scope

We have seen in Section 2 that adverbial clauses can fall under the scope of a negation operator or attach higher and be beyond its scope. Based on this diagnostic test, we can determine how deeply an adverbial clause is syntactically embedded in the matrix clause at the LF level. Behrens and Fabricius-Hansen (2002, p. 59) point out that “[indem]-clauses apparently fall outside the scope of the negator in the matrix; they do not—at least not easily—attach to the matrix clause below propositional negation”, cf. (43).

(43)
	(Behrens & Fabricius-Hansen, 2002, p. 47, ex. 1a)

However, they also observe that indem-clauses can occur under (contrastive) ‘phrasal’ negation, by referring to examples such as those given in (44).

(44)
	(Behrens & Fabricius-Hansen, 2002, p. 47, ex. 1a)

Similar examples are also present in the DeReKo corpus. Some of them are given in (45).

(45)

It is crucial to note here that cases given in (44)–(45) differ from what we have observed in Section 2 about causal weil-clauses (cf. (6a)). While in cases where the propositional negation scopes over the adverbial clause and the matrix event is also affected, this is not the case in the examples containing the (contrastive) phrasal negation nicht (… sondern). In such cases, only one of the subordinate propositions is negated, not the matrix event itself. This made Behrens and Fabricius-Hansen (2002, p. 59) conclude that “the indem-clause has scope over the negated matrix proposition”. I disagree with this claim.17

First, it does not play a role that the negation does not affect the matrix event itself. What is more important is the observation that the negation can target the content of the instrumental subordinate indem-clause. Keep in mind that neither evidential nor speech act related causal weil-clauses can fall under the scope of negation operators, regardless of whether the matrix clause is negated or not, or whether a propositional or a (contrastive) ‘phrasal’ negation is used. If a negation operator can target the content of the indem-clause, we should conclude that it also has the possibility to attach deeper in the clause structure, as in particular (45a) and (45b) clearly show. Pittner (1999) also mentions an example in which nicht ‘not’ has a scope over the indem-clause, (46), and argues that indem-clauses can fall under the scope of negation.

(46)
	(Pittner, 1999, p. 291, ex. 168b)

Second, there might be two explanations for why cases like (43) are considered unambiguous by some native speakers of German. One explanation is synchronic, the other is diachronic.

Synchronically, instrumental indem-clauses can be also employed as PAC, i.e., as JP adjuncts that are not in the c-command domain of negation. Section 4.2 provides more details on indem-PACs. In this case, we expect them not to fall under the scope of negation.

Diachronically, indem as a C-head developed out of a PP. As Greisinger (2016, p. 122) observes, this process happened in the 18th cent. The PP harks back to the preposition in ‘in’ and the dative-marked demonstrative pronoun dem ‘that’. It seems that indem might still be interpreted as a correlative PP that refers to the content of the following explicative sentence, which, in turn, modifies the correlate. The negation is then restricted only to the matrix clause and does not target the relative clause. Although this is not semantically transparent from the perspective of modern German, the internal structure of the complemenizer indem may still have some morphosyntactic reflexes. This observation is, of course, tentative and requires a deeper investigation.

3.2.6. Variable Binding and Principle C Effects

It is well-known that a quantifier can bind an agreeing pronoun occurring in the subordinate clause if and only if the quantifier c-commands the pronoun. Otherwise variable binding is not possible, cf. Chomsky (1981, pp. 183–230), Enç (1989, pp. 62–64), Büring (2005, pp. 83–93), von Wietersheim (2016, 2022), and von Wietersheim and Featherston (2019) for more details. We have seen in (5a) that variable binding into the causal weil-clause is possible. Instrumental indem-clauses pattern with causal weil-clauses in this regard.18

(47)
	(Homberger, 1996, p. 23, ex. 1; slightly modified by author)
	(Behrens and Fabricius-Hansen, 2002, p. 58, fn. 9; slightly modified by author)

Furthermore, independently referring terms such as Maria in (48) are not allowed to be c-commanded by co-referential expressions according to binding Principle C, cf. (48a) vs. (48b).19

(48)	a.
	b.

3.2.7. Embeddability with the Host Clause

Finally, the instrumental indem-clause, as (49) shows, can be embedded along with its host clause.

(49)

3.3. Interim Summary

An overview of the properties instrumental indem-clauses exhibit is given in

Table 4. Selected properties of indem-clauses in modern German.

property	indem-clause
eventuality related interpretation	+
evidence related interpretation	–
speech act related interpretation	–
target of anaphoric expressions	+
verb final position	+
constitutent status	+
Nachfeld position	+
Vorfeld position	+
middle field position	+
Außenfeld position	–
proform coreference (damit and dadurch)	+
(phrasal) negation scope	+
variable binding	+
embeddability with the host clause	+

.

The data considered so far clearly show that instrumental indem-clauses should be analyzed as CACs, i.e., as adverbial clauses that are in the c-command domain of the constituents of their licensing clauses. They cannot attach outside the matrix clause structure, nor can they operate on discourse level and express non-at-issue content. They can be targeted by anaphoric expressions and co-referential with instrumental proforms. As syntactically integrated subordinate clauses, instrumental indem-clauses can occupy the Spec-CP position of the matrix clause and be embedded along with its host clause. But most importantly, they allow variable binding and obey Principle C of the binding theory. Taken all together, I conclude that instrumental indem-clauses attach as TP adjuncts.

In the next section, I examine their internal syntax and discuss first thoughts on whether indem-clauses operating on the content level exhibit variation and can be analyzed as PACs or NACs, too.

4. Internal Syntax

This section is divided into two parts. In Section 4.1, I outline Krifka’s approach to decomposition of assertive speech acts, whereas in Section 4.2, I apply this approach to instrumental indem-clauses to determine their variation.

4.1. The Decomposition of Assertive Speech Acts

Krifka (2019, 2023, 2024) decomposes an assertion into four layers: i) proposition, ii) judgement, iii) commitment, and iv) speech acts. All of them are formally represented through syntax. A proposition $\phi$ is represented by a Tense Phrase, TP. Private judgements are attributed to JP, Judge Phrase, which is equipped with a syntactic head that turns a proposition $\phi$ into the propositional function that a judge x judges the proposition $\phi$ to be true. Syntactically, this is rendered as $xJ-\phi$, whereby $J-$ stands for the head of the JP. Public assertions are taken to be encoded in a Commitment Phrase, ComP, possessing the head ⊢ that turns a proposition $\phi$ into the propositional function that the speaker x is publicly committed in world i to $\phi$: $x{⊢}_i\phi$. Finally, to keep different speech acts apart, Krifka (2024) makes use of the functional projection ActPhrase, ActP, and takes • to represent an assertion operator, whereas ‘?’ stands for a question operator. Either operator is the syntactic head of ActP, which is the highest structural category in the clause structure, (50).

(50)

Evidence for the existence of JP, ComP and ActP comes from dedicated adverbial modifiers. A sample of selected modifiers associated with the dedicated layer is presented in (51). It is, of course, by no means exhaustive, and for their contribution the interested reader is referred in particular to Krifka (2023).

(51)	Adverbial modifiers in modern German:
	a.	JP modifiers: angeblich ‘allegedly’, wohl ‘presumably’, anscheinend ‘apparently’, wahrscheinlich ‘probably’, sicherlich ‘certainly’, scheinbar ‘seemingly’.
	b.	ComP modifiers: wirklich ‘really’, tatsächlich ‘in fact’, bei Gott ‘by God’, Gott sei mein Zeuge ‘let God be my witness’, bei meiner Seele/Mutter ‘at my soul/mother’, ungelogen ‘truly’, im Ernst, allen Ernstes, ernsthaft ‘in earnest’.
	c.	ActP modifiers: offen gesagt ‘frankly speaking’, mit Verlaub gesagt ‘if I may say so’, übrigens ‘by the way’, mit anderen Worten ‘in other words’, am Rande bemerkt ‘as a marginal note’, jedoch ‘however’.

An example containing adverbial modifiers that represent all functional layers is given in (52a). Its partial derivation follows in (52b).

(52)	a.
	b.

Offen gesagt ‘honestly speaking’ is a standard ActP modifier; tatsächlich ‘in fact’ is a commitment phrase by means of which the speaker increases the strength of their commitment; and sicherlich ‘certainly’ is an epistemic adverb expressing the speaker’s attitude towards what is embedded. The most natural word order is when the ActP modifier offen gesagt occurs on the left edge of the clause, where it c-commands the commitment phrase tatsächlich and the epistemic adverb sicherlich. Correspondingly, the adverbial modifiers are base-generated in the dedicated functional projections ActP, ComP, and JP. Derivationally, the finite verb, ist ‘is’ in (52), moves as a V-head to the Act⁰ position, where the speech act itself is determined. What role do adverbial clauses play here? They behave like modifiers in (51) by attaching to a dedicated functional projection. According to Frey (2023a), CACs are TP adjuncts, PACs attach as JP adjuncts, and NACs are ActP modifiers. Adverbial clauses attaching as ComP modifiers have not been attested yet.

I take instrumental indem-clauses to be subordinate assertions. The question then arises whether they can project up to ActP. This is addressed in the next section.

4.2. Variation of Instrumental ‘Indem’-Clauses

So far, we have seen that instrumental indem-clauses are CACs attaching at the TP layer of the matrix clause (see Section 3.2). However, this is not the only position they can attach to. The following corpus examples convincingly show that instrumental indem-clauses can host expressions encoding private judgements, i.e. JP modifiers, cf. wohl ‘presumably’ in (53), anscheinend ‘apparently’ in (54), wahrscheinlich ‘probably’ in (55), and angeblich ‘allegedly’ in (56).

(53)

(54)

(55)

(56)

Hentschel (1986), in addition, mentions one case containing the discourse particle eben (lit. ‘flat’, ‘smooth’):20

(57)
	(Hentschel, 1986, p. 203)

By using eben in (57), the speaker indicates that the way in which the embedded subject killed someone could or even should have been obvious to his or her interlocutor(s); see Abraham (1991), Karagjosova (2003), and Autenrieth (2005) for more details on the meaning of eben.

The occurrence of weak root phenomena in indem-clauses clearly indicates that they can be PACs too. They can host speaker-oriented adverbs, and the presence of discourse particles strengthens this claim, cf. (53) for wohl and (57) for eben. The use of other discourse particles is also attested; cf. (58) for ja, (59) for eigentlich, and (60) for halt.

(58)

(59)

(60)

The examples (53)–(60) provide evidence for the claim that instrumental indem-clauses can be considered PACs, too, and that they attach as JP adjuncts. This is corroborated by the other diagnostic tests registered in Table 1. Instrumental indem-clauses treated as PACs can occur in the Spec-CP position, (61a), disallow variable binding, (61b), can be embedded together with their host clause, (61c), can host weak root phenomena, (53)–(55), and, finally, cannot host strong root phenomena, (61d).

(61)

Now, the question should be addressed whether instrumental indem-clauses can project up to ActP. Since they seem not to be able to host ActP modifiers, this option is ruled out, as seen in (62).

(62)

We can safely conclude that instrumental indem-clauses can be employed either as CACs or as PACs, but not as NACs.

5. Cross-Linguistic Observations

We have seen in the present paper that indem selects for a finite clause and makes the finite verb stay in situ, as seen in (1). Cross-linguistically, this is rather an uncommon strategy because manner elaboration on the clause level is usually expressed by non-finite strategies, as the examples collected in (63) persuasively show.

(63)

Based on this language sample, we can identify five different syntactic strategies. To the first group belong languages that use prepositions. One case in point is Indonesian, (63b), which uses the preposition dengan ‘with’/‘by’ to elaborate on the matrix proposition. Languages like Italian and Polish constitute the second group. They do not make use of prepositions; instead they convert the verb occurring in the subordinate clause into a gerundive form by attaching a suffix to it, -ando/-endo in Italian, (63c), and -ąc in Polish, (63d). English combines both strategies, making it the third group. As shown in (63a), the gerund ing-clause is headed by the preposition by. A similar strategy is to be observed in Icelandic, Norwegian, Faroese and Dutch, where a (grammaticalized) preposition takes an infinitive clause as its argument, (63e)–(63h). The preposition varies from language to language, cf. með ‘with’ in Icelandic, við ‘by’ in Faroese, or door ‘through’ in Dutch, but the infinitive itself is always accompanied by an infinitive marker, að in Icelandic, å in Norwegian, at in Faroese, and te in Dutch. These Germanic languages form the fourth group. Another strategy is used in Ukrainian, (63i), and in Czech, (63j), where the subordinate clause headed by the declarative complementizer ščo/že ‘that’ is referentially linked to the DP tym/tím ‘this’ marked for the instrumental case.

We do not know much about the instrumental clause types and their syntax in the languages registered in (63). Some studies examine the meaning of by in English, cf. Anscombe (1957), Bennett (1994), Fabricius-Hansen (2006), Sæbø (2008), Schneider (2009), and Bücking (2014), but less attention has been paid to the degree of dependency instrumental clauses exhibit.

From a comparative point of view, Sæbø (2011, pp. 1435–1436) briefly discusses the meaning of by-phrases in English and mentions, in passing, their French and German equivalents. Fabricius-Hansen and Behrens (2001), Behrens and Fabricius-Hansen (2002), and Fabricius-Hansen (2006, pp. 54–56) observe, mainly based on translation studies, that indem-clauses have a wider field of use than their counterparts in other languages, i.e. by in English, en in French, or ved å in Norwegian. A deeper semantic look into by-clauses in English and indem-clauses in German is offered in Bücking (2014). A comparative study between Italian and German is also presented in Pusch (1980). However, the comparative findings still need to be embedded into the syntactic typology of adverbial clauses presented in Section 2.

6. Conclusions and Outlook

The main aim of this article was to examine the external and internal syntax of instrumental indem-clauses in modern German and to embed the findings into the syntactic typology of adverbial clauses. I discussed evidence showing that instrumental indem-clauses should be analyzed either as central or as peripheral adverbial clauses, attaching at two distinct heights as either TP or JP adjuncts. I could, in addition, present evidence that instrumental indem-clauses never occur as non-integrated adverbial clauses and adjoin outside the clause structure of the matrix clause.

Furthermore, the question should be addressed how instrumental indem-clauses came into being and how they develop(ed). Not much is known about their diachrony. Behaghel (1905) points out that they developed out of temporal indem-clauses. In (64), an example from New High German (1650–1900), indem can be paraphrased as ‘when’, ‘while’ or ‘after’:

(64)

Diachronic corpus data is expected to provide answers to at least three questions. First, if instrumental indem-clauses are taken to have developed out of temporal indem-clauses, the cirumstances that paved the way for this development should be examined. Second, temporal clauses are usually considered CACs. It would be interesting to learn whether temporal indem-clauses were CACs too and, more generally, whether the development of CACs into another semantic type of CACs led to any consequences for the syntax of adverbial clauses. Finally, to have a complete diachronic picture of instrumental indem-clauses, it is crucial to figure out when they began to be used as PACs.

I leave all these questions open for future research.

7. Data Sources

DeReKo	–	Das Deutsche Referenzkorpus, version 2.4.5.1, http://www.ids-mannheim.de/cosmas2/ (accessed on 15 December 2024).
Karlsbad	–	Franz Hieronymus Br$\stackrel{\mathrm{e}}{\mathrm{u}}$dmann, 1785, Bemerkungen auf einer Reise nach Karlsbad, Braunschweig: In der F$\stackrel{\mathrm{e}}{\mathrm{u}}$rstl. Waysenhausbuchhadlung.
PE	–	Luis Sellano, 2016, Portugiesisches Erbe. München: Wilhelm Heyne.
PS	–	Luis Sellano, 2023, Portugiesische Sünde. München: Wilhelm Heyne.