1.1 The problem addressed

The problem addressed in this study is the sentence type restrictions of MPs. MPs are restricted to certain sentence types in an item-based distribution. For instance, bloß and nur (both meaning ‘only’) only function as MPs in sentence types with an exclamative or imperative sentence mood, and the MP denn (‘then’) only occurs in WH- and yes/no questions (in other grammatical environments bloß and nur function as adverbs or focus particles, see König 1991; and denn as an adverb or a conjunction). Table 1 illustrates the possible uses of MPs in different syntactically independent sentence type formats according to the analyses of Altmann (1993), Thurmair (1989; 2013) and Kwon (2005).¹ The list is however not exhaustive; there may be more sentence types than is accounted for here, and there is no final agreement on which words count as MPs; furthermore, what is generally referred to as restrictions often refers only to the prototypical usage patterns of MPs; thus, under specific circumstances, often exceptions are possible and have been outlined in corpus-linguistic work (cf. Métrich et al. 2002). Nevertheless, it is uncontroversial that some combinations are much less common than others; we thus take sentence type restrictions to refer to the fact that MPs are unequally distributed across different grammatical contexts.

To approach the usage restrictions of MPs, researchers have discussed whether these restrictions should be attributed to the form or the meaning of the host utterance – and both formally and functionally based restrictions have been suggested. Nevertheless, most researchers acknowledge that there are arguments for both perspectives even if they decide to argue for one side more than for the other themselves.⁴

Before we address different suggestions on how to account for the patterning observable, we need to clarify some terminology.

1.2 Definition of sentence type, sentence mood and illocution

It is well-known that the use of individual MPs is restricted to certain “sentence types” (e. g. Thurmair 1989; 2013). However, sometimes the term “sentence mood” is considered to be the more correct term (e.g. Diewald 2008a; b). Yet again, other researchers prefer to relate MPs directly to illocution or illocutionary type, a concept that originates in speech act theory (Searle 1969). In the literature on MPs we are thus confronted with three different terms, which potentially stand for three different concepts. We therefore start by disentangling these different concepts.

A sentence type is a linguistic structure that is associated with a certain basic communicative function. The four basic sentence types in English and German (cf. Sadock & Zwicky 1985 and Altmann 1987, respectively), representing three different basic communicative functions, are the V2 declarative, the V1 yes/no question, the WH-V2-question and the morphological imperative sentence type:

The basic sentence types are considered to be universal by Sadock & Zwicky (1985), but there are also other, more language specific sentence types, such as, in German, the exclamative and the optative sentence types. The term sentence type is used in the literature either to refer only to the form side of these linguistic structures (e.g. Altmann 1987) or to refer to the combination of a certain linguistic format and its associated meaning/function. It seems to us that the latter is the most common usage of the term sentence type, especially in the English literature (e.g. Sadock & Zwicky 1985). We use sentence types in this latter sense, see examples (2)–(4) above.

The term sentence mood is also used in different ways in the literature. To Altmann (1987; 1993) sentence mood is a complex sign consisting of a form side and a meaning side, pretty much corresponding to our use of the term sentence types. However, sentence mood is also often used to refer exclusively to the meaning side of the sentence types (cf. Meibauer, Steinbach & Altmann 2013: 5), and this is the meaning we adopt for this term. We take sentence mood to stand for the “semantic meaning” of sentence types (cf. Zaefferer 2006: 459; Meibauer, Steinbach & Altmann 2013: 5), such as the traditional declarative, interrogative and imperative meaning. In this view, sentence mood is a broadly defined, basic communicative meaning. There are usually several different linguistic formats that express the same kind of sentence mood. This means that very different linguistic formats will still be considered as tightly interrelated since they are representatives of the same sentence mood. Thus, the two different question type formats in (3a–b) above are usually grouped together despite their striking syntactic differences because they are both considered to express an interrogative sentence mood, i.e. their basic meaning is suited for the function of asking questions.

Finally, we have to position the concept of sentence mood in relationship to the concept of illocution/illocution type, where uses also differ between researchers. For example, Jacobs (1991) uses the term illocution type to refer to the general illocutionary potential of sentence types, much like the notion of sentence mood outlined above. In contrast, Brandt et al. (1992: 1ff.; see also Gutzmann 2015: 167) define sentence mood as the semantically encoded counterpart or “intermediary” between the sentence type and its illocutionary potential. They consider sentence mood to be semantically encoded while the illocutionary force is considered as a pragmatic, more specific instantiation of the sentence mood in a certain context of use. Thus, the illocution of an utterance is a more specific function than that of sentence mood. This kind of thinking is also reflected in the work of many other researchers who consider the general meaning of sentence types to be broader than specific, individual illocutions; for example, the imperative sentence types, which first and foremost express the general imperative sentence mood, can be differentiated further to express more specific imperative subcategories, for example commands, requests, suggestions, or advice (cf. Kwon’s 2005: 25 discussion). We adopt this narrow meaning of illocutionary function here for all practical purposes and use sentence mood for the broader meaning of sentence types.

In the next sections, we examine the usage restrictions of MPs with respect to sentence types and other constructions in which they can occur, striving to identify the relevant factors that govern the usage of MPs.

2.1 The role of meaning in the usage restrictions of modal particles

In this section, we summarize the arguments from the MP literature on the role of meaning and functional features for the use of an MP in a given host structure.

In Table 1, we have chosen to represent the sentence types grouped together according to their sentence moods. Sentence mood is a linguistically established concept and has been suggested to be an important factor for describing the usage restrictions of MPs. For example, Gutzmann (2015: Ch. 3) argues that the semantics of the MPs interacts with the grammatically encoded sentence mood of sentence types. He presents examples demonstrating how semantics, i.e. sentence mood, takes prevalence over “pragmatic effects”, i.e. in this case over the specific illocutionary force of an utterance: If the distributional restrictions of MPs were based on the pragmatic illocution rather than on the semantic sentence mood, the MP ja should be incompatible with the V2 declarative format when this is used to ask a check-question because ja is incompatible with the interrogative meaning, but this is not the case (see example (5a) below). The reverse also holds: the MP ja is not allowed in rhetorical questions although this kind of questions functionally serve to make a statement and not to ask a question (example (5b)):

Abraham (1995) also advocates the meaning-orientation of MPs, but uses a different line of argumentation, as is illustrated in the exclamation using the format in example (6a) below: From a strictly formal perspective, the structure in (6a) looks like a WH-V2-question, and it can be used to ask a WH-question (6c). However, this format can also be used to make an exclamation (6b). In terms of MP restrictions, Abraham argues that the MP aber (literally indicating an opposition, similar to ‘but’) can only be used in a format like (6a) if this is used in the function of an exclamation (6b). If the question-format is actually used in the function of a question, Abraham (1995) does not consider aber to be an MP (6c):⁵

Examples (5) and (6) show that one and the same linguistic format can be associated with quite different illocutions. These illocutions sometimes seem to have the status of being mainly special “pragmatic effects”, like the formats in (5), and sometimes the same linguistic format seems to have been routinely encoded for two different sentence moods like the format in (6a), where the exclamation in (6b) does not feel like a question that has been converted for a special purpose. However, the picture is getting still more complicated: In contrast to the sentences in (5), some MPs seem to be usable in sentence types encoding the “wrong” sentence mood, provided that these sentence types are used to express a so-called indirect speech act of the right kind for the MPs under consideration (e.g. Thurmair 1989; Oppenrieder 2013). The exact meaning of the concept indirect speech act is usually taken for granted in the MP literature, but for clarity’s sake we define it here: We consider each sentence type to grammatically encode a certain sentence mood. Each sentence mood is associated with a certain illocutionary potential (see section 1.2 above): for instance, the declarative sentence mood is associated with an informative illocutionary potential whereas the imperative sentence mood is associated with a directive illocutionary potential. With an indirect speech act, we mean that a sentence type is used to express an illocution which is not part of the illocutionary potential of the grammatically encoded sentence mood. Instead, the expressed illocution is normally associated with the illocutionary spectrum typical of another sentence mood. For example, given the right circumstances a V2 declarative sentence can be used to perform a directive speech act (see below), although its encoded declarative sentence mood is associated with an informing speech act potential and not with giving directions.

The use of MPs in such indirect speech acts, as defined by speech act theory (Searle 1969), speak against a purely semantic orientation of the MP restrictions, and it is also an argument against binding MPs semantically to a/one specific sentence mood, as is often done in formalistic analyses such as Coniglio (2011) and Struckmeier (2014).⁶ The MPs ruhig, bloß and nur are a case in point. While typically occurring in sentence types with an imperative sentence mood (see example (7a) below), they can be used in the verb-second declarative sentence type if this is modified in such a way that it functions as an indirect directive (Thurmair 1989: 202). This can be achieved by using a modal verb expressing permission together with a 2^nd person address term, as is shown in example (7b) below. In contrast, verb-second declarative sentences that lack the modal verb expressing permission are obviously not modified enough to support an MP interpretation of ruhig. As a result, the particle can only receive its literal interpretation, as is demonstrated in (7c), which is a manipulated version of (7b). Although it is imaginable to use sentence (7c) in an advisory or instructive context, this pragmatic possibility alone is obviously not enough to allow for the MP interpretation of ruhig.⁷

The sentence in (7a) is a direct imperative sentence type expressing a piece of advice, i.e. ruhig is an MP. The example comes from a summary of arguments in favor of starting to play tennis as a grown-up. The sentence in (7b) is an indirect directive, here also expressing a piece of advice, i.e. ruhig is an MP. The example comes from a discussion on when to start giving babies real food. The sentence in (7c) is a declarative sentence without a modal verb and thus presenting a factual situation rather than an appeal, i.e. ruhig is an adverbial of manner. This example is our manipulation of the sentence in (7b):

However, there is one group of declarative sentences without modal verbs that routinely take MP ruhig and which at first glance seem to make an exception, as is demonstrated in the following example from Métrich et al. (2002):

In German, sentences like the one in example (8) have a clear established directive/permissive function despite their declarative syntax, which makes it possible for the MP ruhig to occur in the declarative sentence without the presence of a modal verb.⁸ Actually, it is quite possible that the sentence in example (8) is no ordinary declarative at all but an instance of the long established polite imperative sentence-type format used for a polite/formal form of requests; the polite imperative construction with explicit subject Sie was the first environment beyond directive sentence type formats with imperative verb morphology to which the MP ruhig extended in its grammaticalization process (see Diewald 2008a).⁹ In this case, it is no surprise at all that a sentence like the one in (8) can take the MP ruhig.

To conclude, the sentence moods and the illocutionary spectrum associated with each of them play important roles in defining the usage restrictions of MPs, but also indirect speech acts seem to play a role. The behavior of MPs in indirect speech acts is one of the reasons why we choose a construction grammar approach: This framework offers the flexibility to state the same construction several times on different levels of specificity, see section 4.

2.2 The role of formal factors in the usage restrictions of modal particles

As for formal features, many researchers suggest that the usage restrictions of MPs orient more to the form-side of sentence types than to the meaning-side (where meaning is taken to be the sentence mood, cf. Weydt 1969; Hentschel 1986; Altmann 1987; Thurmair 1989; but see also Zifonun et al. 1997: 1212f.). Thurmair (1989; 2013) provides a summary of reasons for favoring form-based explanations of the usage restrictions of MPs.

(i) The first argument for sentence type restrictions emphasizing the influence of the form-side (see Altmann 1993: 1013; Thurmair 1989: 202; 2013: 647; Kwon 2005: 182ff.) refers to the fact that one sentence mood can be expressed by different sentence type formats (as Table 1 in section 1.1 above illustrates); since MPs are often restricted to only some of these different formats, it seems to be the form, and not the meaning in the sense of grammatical mood that determines the usage of an MP. Especially striking is the restricted use of MPs in sentence type formats of the verb-final type in comparison to the MP usage in the verb-initial or verb-second sentence type formats expressing the same sentence mood (Thurmair 1989: 202; 2013: 647). For instance, the MP denn (‘then’) is a typical MP in questions, but Thurmair (1989: 62 + 202) and Kwon (2005: 185) claim that it cannot be used in question formats of the WH-verb-final type that are not regarded as instances of indirect speech. Verb-final question formats representing indirect speech, however, behave more like their verb-second or verb-initial counterparts and are regarded as elliptical structures. The sentences in (9a–c) from Kwon (2005) illustrate denn in the regular WH-V2-question format (9a), in the controversial verb-final format (9b), which however, according to Kwon, is fine if you imagine a left-out main clause clarifying its function as indirect speech (9c):¹⁰

(ii) A second argument from the approaches emphasizing the form-side (see, for instance, Thurmair 2013: 49) is that there are clear differences in the usage of MPs in yes/no questions and in WH-questions, even though these are often treated together as representatives of the same interrogative sentence mood (see section 1.2 above). For example, the MP etwa (literally: ‘approximately’) can be used in the yes/no question type, but according to Thurmair not in the WH-question type; and the MP schon (literally: ‘already’) can be used in the WH-question type, but not in the yes/no question type.¹¹

(iii) Conversely, the literature points to an overlap in the use of MPs in WH-questions and WH-exclamative sentences, even though these sentence types do not belong to the same sentence mood (Thurmair 2013: 632; see also Table 1 in section 1.1 as well as example (5)): For instance, Thurmair points out that the MPs bloß (‘only’), doch (‘however’) and nur (‘only’) can occur in WH-questions but not in yes/no questions; they can occur in WH-exclamatives (both the verb-second and the verb-final kind), but bloß and nur cannot be used with exclamative formats without a WH-element (see Table 1). Thus, it seems that bloß and nur are sensitive to the existence of a WH-element for their use in both questions and exclamatives.

(iv) A fourth argument for the primacy of formal factors determining sentence type restrictions is the “obligatory” use of an MP in some of the “marginal” German sentence moods, like the optative sentence mood (but see the discussion in section 2.3 below) (see Altmann 1993: 1012; Kwon 2005: 226ff.; Jacobs 2008: 31; Thurmair 2013: 637). Altmann (1989; 1993), Jacobs (2008) and Thurmair (2013) consider MPs as a constitutive feature of certain sentence types. As constitutive features they would belong to the form-side of the description of those sentence types since in these cases it seems to be more important that a suitable MP is used than which one of the suitable MPs is actually chosen.

Given the arguments for both perspectives, it seems that MPs interact both with the formal and meaning features of each sentence type.

2.3 Modal particles as illocutionary operators

A third attempt at explaining the sentence type restrictions of MPs in the literature is to analyze them as illocutionary operators. According to such analyses, the main function of MPs is to modify the specific illocution of an utterance (e.g. Altmann 1987; Helbig 1988; Thurmair 1989; 2013; Jacobs 1991; Engel 1996; Helbig & Buscha 1998; Coniglio 2011).

MPs are claimed to interact with the illocutionary potential in three different ways, which we call specification, alternation and constitutive feature; we will discuss them in turn in this section (see Table 2).

First, in the specification function the use of an MP would be a means of selecting one specific sub-illocution within the broad illocutionary spectrum which is associated with the grammatically encoded sentence mood expressed by the host utterance (cf. section 1.2 above). This is illustrated by Zifonun et al. (1997: 1218) in example (10a) by the use of nur (literally: ‘only’) in the verb-first imperative sentence type. Both with and without the MP, the sentence format in (10a–b) expresses a directive function, but the version containing the MP in (10a) expresses a more specific subtype of the directive function: Depending on the context, it can express an encouragement to indeed do what the speaker suggests; a permission to do what the speaker suggests; or even a threat of sanctions if the addressee keeps on doing what is described in the utterance. The same utterance format without an MP in (10b) simply expresses a directive function in general and the addressee will have to do the whole interpretation all on his or her own:

Second, in a few cases the presence of MPs is supposed to alter or convert the sentence mood affiliation of their host utterance completely (cf. Altmann 1993; Coniglio 2011). We take alter to mean that the host utterance, as a result of the MP usage, is changed so that it can express an illocution that does not belong to the illocutionary spectrum of the grammatically encoded sentence mood. For instance, in the following example from Coniglio (2011: 21), the modal-particle cluster doch wohl nicht etwa (literally: ‘but surely not approximately’) is suggested to make the verb-second declarative sentence type of its host utterance express the illocution of a rhetorical question, which Coniglio regards as an instance of the interrogative sentence mood – so basically, the presence of the MP seems to override the grammatically encoded sentence mood of the host structure:

Third, in rare cases the presence of an MP is even said to be a constitutive feature of the sentence type of the host utterance (cf. Altmann 1993: 1012; Kwon 2005: 226ff.; Jacobs 2008: 31; Thurmair 2013: 637). This is illustrated in example (12) from Grosz (2014): The linguistic structure in example (12a) is most often used for forming hypothetical conditional subordinate clauses. This is a dependent clause format that does not usually express a sentence mood or an illocutionary force of its own. Now, according to Grosz (2014: 90; see also among others Thurmair 1989: 117, 182; 2013: 637; Jacobs 2008: 29), the addition of an MP to the kind of linguistic format shown in (12a) turns it into a syntactically independent utterance with an optative illocution, i.e. it becomes an optative sentence type format. This is illustrated in (12b). The use of an MP in (12b) is claimed to be an obligatory part of this sentence type definition, because without it, you only have the hypothetical conditional clause. Furthermore, the choice of one MP over another in (12b) does not seem to be accompanied by any meaning contrast, as it is normally the case when there is a choice between different MPs; instead, any compatible MP functions equally to create an optative interpretation of the sentence:

In this last case, the MPs do not only influence the illocutionary interpretation function of their host utterance, but even seem to be obligatory formal features for constituting a certain sentence type in the first place (cf. Jacobs 2008; Thurmair 2013). This role of MPs applies mostly to the “marginal” optative and exclamative sentence types, but Altmann (1993) and Thurmair (2013) also discuss obligatory MPs for some sentence types with a subordinate clause format (i.e. verb-final formats). They notice that the “individual effect” of one MP over another in these marginal sentence types is small. Instead, the important aspect is that there has to be an MP at all (e.g. Thurmair 2013: 637).

The “obligatory” use of MPs in marginal sentence types is treated in the literature as an important reason to describe the function of MPs in terms of illocutionary effects. However, in his analysis of the optative wenn-clause format (“if”-clause format) illustrated in example (12) above, Grosz (2014) argues that MPs such as bloß and nur (both literally meaning ‘only’) are not obligatory for achieving an optative reading of the utterances – they can just as well be substituted by other expressive particles, like for example interjections, such as ach (Engl. “alas”/“oh”) (Grosz 2014: 91); and if the situational context supports the use of optative utterances, the particle words can even be left out completely (Grosz 2014: 99ff.).¹² Grosz argues that generally, the hypothetical conditional meaning of the wenn-clause is more common than the optative meaning and therefore the hypothetical conditional use is more salient and unmarked, whereas the optative use is unexpected and marked. He suggests a general principle that recommends the use of cues whenever necessary:

“If a marked use of an ambiguous utterance can be made more salient by adding certain elements (e.g. particles, interjections, intonational tunes) to this utterance, the addition of one (or more than one) such element is obligatory. Such elements qualify as cues for the respective utterance use.” (Grosz 2014: 93)

Grosz goes on to state that the principle to use cues is obviated if the context independently supports the intended interpretation of the utterance – and such dispensable cues cannot be constitutive in an Aristotelian understanding of definition.¹³

We conclude that MPs may indicate certain functional interpretations even though they are not obligatory constituents of certain sentence types (see Alm & Behr in preparation for further discussion of this perspective on MPs); nevertheless, none of the perspectives presented in section 2.1–2.3 alone solves the problem of explaining the sentence type restrictions of MPs.

2.4 The interaction of formal and meaning/functional features in the usage restrictions of modal particles

While all three perspectives considered in section 2.1–2.3 have advantages and account for some aspects of the sentence type restrictions of MPs, none provides the whole picture. The evidence in section 2.4.1 illustrates how sentence type restrictions can be related to formal, functional and illocutionary factors at the same time. This complex synchronic situation may be the result of the variety of factors that influence the historical development of MPs (see below); from this perspective, each MP grammaticalizes in an idiosyncratic way, such that there is no single definition of restrictions that works for all MPs. This is exemplified in section 2.4.2 by Diewald’s (2008a; b) diachronic findings on the MP ruhig (original meaning: ‘calm, without disturbances’; MP-meaning: ‘there is no obstacle to (doing) X’).

4.1 Sentence type constructions

In the course of this investigation, we have seen that the different sentence types belonging to the same sentence mood are not freely interchangeable such that one MP could occur in all of them. At the same time, these different grammatical configurations differ from each other by additional meaning- or functional specifications. For example, the verb-initial declarative sentence type in German (example (18)) can only be used in certain contexts, namely for building up the background for a joke or an emotionally involved story (cf. Oppenrieder 2013):

Since we are working from a construction grammatical perspective, we argue that the non-interchangeability between the individual sentence type formats is reason enough to postulate different, very specific sentence type constructions (cf. Wulff 2006). Sentence types have been analyzed as constructions in the construction grammar literature before (cf. Michaelis & Lambrecht 1996; Välimaa-Blum 2005; Jacobs 2008; Hoffmann 2013; compatible with a construction grammar perspective is also Sadock & Zwicky 1985, for instance), corresponding to the fact that sentence types are traditionally characterized as form-meaning pairings of linguistic form and sentence mood.

The relevant form-features for describing the sentence type formats are usually divided into four classes. Following Altmann (1987; 1993) these are: (i) verb morphology, above all the difference between finite verbs and infinite verbs, and among the finite verb forms between the imperative mood and the non-imperative mood;¹⁹ (ii) the position of the finite verb and of the WH-marked question elements inside the clause; (iii) the “categorical filling”, i.e. the presence of WH-elements, of certain performative or modal verbs, or of MPs etc.; (iv) and finally, there is the intonational realization. The representation of intonation as a defining feature for sentence types is somewhat controversial: First, it is not agreed yet whether intonation should be included in a sentence type definition at all. Second, if it is included, the question arises on what level of detail intonational features should be analyzed. Most work on sentence types includes an intonational characterization, usually in the form of holistic nuclear intonation contours (“end intonation”). There is, for example, a large body of literature on the use of falling vs. rising intonation in declarative sentences and questions (for an overview, cf. Alm 2016; Gunlogson 2003; Välimaa-Blum 2005).

On the meaning side of the sentence type constructions, we have the traditional sentence mood, but also the pragmatic meanings associated with the respective sentence type construction, for instance discourse functional meanings or the usage conditions of a given construction (Fillmore, Kay & O‘Connor 1988: 501f.; Goldberg 1995: 6f.; Kay & Michaelis 2012). This means, for instance, that the verb-initial declarative sentence type in example (18) above (Comes the Pope to heaven…) can be specified with information about the special kinds of contexts in which it can be used (cf. Michaelis & Feng 2015). Furthermore, also information-structural differences between different sentence type constructions (cf. Michaelis & Lambrecht 1996) and other non-propositional meanings may be relevant.

In our analysis, sentence mood can be regarded as a common meaning component that accounts for the affiliation of all the sentence type constructions belonging to the same sentence mood – but additionally each sentence type construction will have its own differentiating meaning components. We illustrate this in Figure 1, where the arrows correspond to post hoc classification links (Croft 2001), which can be seen as the generalization from specific examples to schematic structures on the one hand but also the instantiation of general structures as specific instances on the other; hence, the relationship is bidirectional.

Figure 1

Individual declarative sentence type constructions²⁰ grouped together according to their shared sentence-mood meaning.

Given this network of different but related sentence type constructions, we can specify now the possible matches between individual MPs and the specific sentence type constructions. For example, the (unstressed) MP ja (literally: ‘yes’) can only be used in the prototypical verb-second (V2) declarative sentence type construction (cf. Thurmair 1989: 49); it cannot be used with the verb-initial (V1) declarative sentence type construction used to introduce jokes (Oppenrieder 2013: 40ff.), which is demonstrated in example (18’):

We therefore need to specify the interaction between ja and the V2 declarative sentence type construction at a sentence type specific level in the hierarchy.

At the same time, we assume that both the form-side and the meaning-side of the definition of the sentence type constructions may contain prototypical features that can be overridden by contextual factors in the actual communicative situation (cf. Välimaa-Blum 2005; Panther & Köpcke 2008) or by more specific constructions. In prototype theory, categorization is not about absolute category memberships; in contrast, some members may be better representatives of their category than others, and they form the core of the category. The members at the periphery of the category do not exhibit as many relevant features of that category as the core members do and are therefore more difficult to recognize as category members (for a discussion of different notions of prototypicality cf. Geeraerts 1988).

A prototypical sentence type is maximally independent of situational context (Panther & Köpcke 2008: 91): On the form-side, a prototypical format should be recognizable as a sentence type construction with as little contextual support as necessary. On the meaning side, the prototypical meaning of a sentence type is the meaning that we intuitively associate with that format with as little contextual support as possible. This means, for instance, that the prototypical meaning of imperative sentence type constructions is the function of commanding rather than the function of giving advice (Panther & Köpcke 2008: 100). In contrast, non-prototypical sentence type meanings/functions, such as the invitation reading of the imperative, as in “please sit”, need contextual support if the hearer is to arrive at the correct interpretation intended by the speaker (in this case, the support comes in the form of the politeness particle please).

This position requires a definition of what counts as “contextual” support and in what way the meaning/function of sentence type constructions can be modified in actual communicative use. We will not be able to answer this question exhaustively here, but we can make some suggestions. The “prototypical” form and meaning of sentence type constructions seem to correspond to what Altmann (1987; 1993) calls the structural meaning of a sentence type in his “three-level model of the meaning constitution of sentence types”.

The first level consists of the structural meaning of the sentence type, namely the grammatically specified meaning. In our analysis, this is the meaning associated with the schematic sentence type construction formats. On the next two levels, this grammatically specified meaning can be functionally (though not grammatically) modified or even “altered”: On the second level, there is the lexical filling of a schematic sentence type construction. According to Altmann, the lexical filling can influence the interpretation of the function of the sentence type, e.g. through the use of performative verbs like ask or promise or by certain modal verbs. For example, in the declarative sentence type, the combination of an address term like you and a modal verb expressing obligation usually signals a request: You have to leave now. On the third level, we find all kinds of (yet to be specified) contextual factors embedded in the concrete communicative situation.

Returning to the sentence types and their meanings, in construction grammatical terms, these form-meaning pairings are assumed to constitute an inventory of sentence type constructions. As Diewald (2008a; b) and our analyses in section 2 have shown, the licensing of MPs may be strongly tied to the form-side of their host utterance; for example, the original use of the MP ruhig (‘there is no objection to (doing) X’) was constrained to the verb-initial imperative sentence type construction with imperative verb morphology and without an explicit subject (see example (15a) in section 2.4.2). Yet MPs can also be licensed by the meaning-side of the sentence type construction; this was the case when the use of the MP ruhig extended to other sentence type constructions due to their lexical filling (see example (15b) in section 2.4.2). Thus, MPs interact with different aspects of the individual sentence type constructions.

The relevant aspects for describing the usage restrictions of MPs consequently vary both on the form-side and on the meaning-side of a sentence type construction. The relevant form-side features of sentence type constructions range from completely schematic to partially lexically filled structures. The meaning-side of the sentence type constructions comprises both the prototypically encoded sentence mood of that sentence type construction and much more concrete pragmatic functions that can be used to perform indirect speech acts.

Depending on the individual MP under consideration, it may thus be necessary to choose different levels of specificity of the host utterance in order to explain when the particular MP can or cannot be used. Some MPs are compatible with a certain sentence type construction at a high level of schematicity; for instance, the MP aber (literally: ‘but’) can be used in all exclamative sentence type constructions (cf. Table 1 in section 1.1). For other MPs, individual sentence type constructions within the same sentence mood have to be specified; for instance, for ja (‘as you and I both know’), we will want to specify that it only occurs with V2 declarative sentences to account for instances such as example (18’) above. And for some MPs it may even be necessary to specify some of the lexical filling of the individual sentence type construction, as was illustrated for the MP ruhig (‘there is no objection to (doing) X’) in the indirect use of declarative sentence type constructions involving verbs of permission, such as the modal verb können (‘be able to’) in example (7) above.

To sum up, we need to postulate a fine-grained network of sentence type constructions with detailed form- and meaning descriptions as part of the constructional inventory. The individual MPs will then orient to different levels of form and meaning in an idiosyncratic way.

4.2 Particle constructions

In a construction grammatical model, also lexical items are considered to be constructions since the knowledge of a language is taken to consist entirely in the knowledge of the constructional inventory of a language (cf. Kay & Fillmore 1999; Goldberg 2003). We thus need to assume individual particle constructions. Each MP is considered to have an invariant meaning across all of its uses that distinguishes it from other items that can be used as MPs (cf. Fischer & Diewald 1998; Fischer 2000; 2006); for instance, ja has an invariant affirmative meaning and aber an invariant contrastive meaning (cf. Tables 3–4 in section 3 above). Further, the items functioning as MPs often have “doubles” (i.e. heterosemes, cf. Diewald 2015) in other word classes, e.g. as adverbs, conjunctions, discourse particles etc. – and the core meaning associated with the respective phonological form needs to be sufficiently abstract to account for all of these uses (Fischer 2000; Diewald 2013).²¹ For instance, for the particle ja, Fischer (2000) has shown that an abstract meaning of accordance can be assumed, which is then instantiated concretely in the various word class constructions (answer particle, discourse marker, modal particle). Similarly, Diewald and Fischer (1998) identify abstract meanings for the particles aber (literally indicating opposition), auch (literally indicating augmentation), doch (literally indicating opposition) and ja (literally indicating accordance), which are taken to be specified when these particles are used as discourse marker or modal particle. Diewald (2008a) furthermore makes the same argument for ruhig. In order to account for the function of a particle lexeme as an MP, we therefore also have to assume a general MP construction, which describes the work a particle does when functioning as an MP. This is the topic of the next section.

4.3 The modal particle construction

In order to explain how the phonological forms ja, also, aber and other particles that have “doubles” (heterosemes, cf. Diewald 2015) in other word classes come to be used as MPs, we need to assume a construction that accounts for the grammatical word class (see Fischer 2006; see also Diewald 2006 concerning MPs as a grammatical word class). Diewald and Fischer (Diewald & Fischer 1998; Diewald 2006; Fischer 2006) suggest that the word class “modal particle” constitutes a grammatical construction by itself. Like every grammatical construction, it has a form-side and a meaning-side. The form-side consists of the description of a particle slot in the easily recognizable, characteristic mid-sentence position. According to the particle model by Diewald and Fischer (e.g. Diewald & Fischer 1998; Diewald 2006; Fischer 2007), the meaning-side consists in anchoring the host utterance to a proposition of the interlocutors’ common ground, presenting the utterance as based on already established, mutually accepted assumptions (see Tables 3–4 in section 3).

An argument in favor of the assumption of such a construction is the fact that many German particles fulfill systematically different functions depending on their grammatical position, in particular whether they occur turn-initially or utterance-medially in the syntactic middle field.²² This is demonstrated in the constructed sentences in (19a–b):

The turn-initial use of particles concerns the management of conversation in some way or other in a situated exchange (Fischer 2000; Clark & Fox Tree 2002), the middle field use concerns a previously ratified proposition which is assumed to be common ground and which is thus presupposed in the official business. This shift between situated exchange and proposition assumed to be common ground is systematic and applies to many different German particles (e.g. Diewald & Fischer 1998; Diewald 2006), as is illustrated in example (20), using the same sentence as in (19) with the particles exchanged:

This systematic contribution of the grammatical position suggests that certain features are contributed by the grammatical construction, such that the utterance-initial or sentence-medial positions impose certain interpretations on the lexical items involved (cf. Fischer 2000; 2006).

However, there are also problems with the assumption of an MP construction. First, if we propose these grammatical positions to be constructions, then all items in these constructions should be affected by the constructional meaning. This has indeed been established for utterance-initial discourse particles (Fischer 2006) but not for the syntactic middle field. Thus, in the same syntactic slot in which MPs occur, also adverbs and downtoners can be used without invoking a proposition from the common ground. As illustrated in example (21) on the sentence I can do it, only the MP ja in (21b) evoke common ground; the middle-field items in (21c–e) do not:²³

Second, prosody may play a role in distinguishing between the MP-function and a non-MP-function for particles in the middle field, but only for some items. In example (22), the placement of prosodic prominence distinguishes between the interpretation of the particles as a temporal adverb (22a) and a focus particle (22c) (emphasis on the particles themselves) and the function of the same particles as MPs (22b, d) (emphasis on the verb):

While prosody may thus play a systematic, i.e. grammatical, role in the definition of some MPs, stress placement does not change the grammatical role of other particles, even though it influences their semantic interpretation; for example, doch is an MP in both the unstressed and the stressed middle-field use because in both cases it evokes a communicatively given proposition to which it anchors the current utterance:

Thus, the role prosody plays may be rather item-dependent.

To account for these findings, we propose to assume not only an MP construction, but also a middle-field particle construction, of which the former is an instance. The MP construction accounts for the stable form-meaning pairing, specifying their syntactic position and the peculiar pragmatic function of MPs, which is to relate the current utterance to a particular aspect of the common ground. This however leaves open how the more general middle-field particle construction, which also comprises adverbs, downtoners, focus particles etc., can be described and how the relations between individual items and the constructions in which they can occur can be represented, i.e. why some items in the middle field position are MPs while others are not. In the middle field, four groups of particle items emerge: 1) those with “doubles” in the utterance-initial position (discourse particles, conjunctions), which all function only as MPs when used in the middle-field particle construction; 2) those that are potentially ambiguous in the middle-field position and rely on prosody for disambiguation; 3) downtoners; and 4) those with proposition-modifying functions, such as adverbs. Membership in these different classes seems to be item-based, and it also changes historically for individual items.

There are thus no specific formal or functional properties that determine class membership in the middle-field particle construction; this means that for particles that can be used in the same position with other meanings than the MP function, for instance as focus particles or adverbs, the form-side alone is not sufficiently distinctive, but constitutes only one of several possible interpretations. The general middle-field particle construction is thus rather schematic, basically describing only the particular grammatical slot for the kinds of items that can occur. In contrast, the MP construction, one level down in the hierarchy, is informative particularly for those particles whose function as MPs is indeed defined by their grammatical position. That is, for ja, also or denn, for instance, which also occur independently or utterance-initially and thus in different constructions, the MP construction unambiguously contributes meaning components and disambiguates their uses. Consequently, like for the relationship between verbs and argument-structure constructions (Stefanowitsch 2011), we need to assume two processes at work: On the one hand, constructions express generalizations over particular items; on the other hand, category membership may have to be dynamically defined in an item-based manner.

4.4. Modal particles in sentence type constructions

In construction grammar, individual constructs, i.e. instantiations of constructions, are created by an interaction between different constructions, including lexical constructions (e.g. Goldberg 2003). Such interactions comprise the unification with maximally general, schematic constructions, but also sub-generalizations may be captured by postulating partially filled constructions, and even idiosyncrasies can be accounted for in individual constructions. Accordingly, the compatibility of specific particles with specific sentence type constructions can be represented in an item-specific construction to specify the conventional aspects of the interactions between sentence type constructions and a given MP. For instance, there is no difference encoded in the consecutive meanings of denn and also (both mean approximately ‘then’) that would suggest that denn refers to the illocutionary act of asking questions, hence tying it to interrogative sentence types, whereas also refers to the propositional content of an utterance or the topic discussed and therefore is not bound to any specific speech act (see Fischer 2007; Fischer & Alm 2013). Thus, this information cannot be specified as part of their lexical meaning. Likewise, it does not make sense to specify for interrogative constructions that they may take denn and also, yet each in different uses (see the discussion of denn and also in section 3). Consequently, we have to assume that item-specific MP constructions exist that specify for each of the two particles what work they do as MPs. We therefore postulate (in addition to the general MP construction, the denn- and also-particle constructions and the sentence type constructions) constructions describing the particular relationships between specific MPs and certain sentence type constructions, for instance, a denn-in-WH-question construction, which specifies that denn in this construction refers to the speech-act meaning of asking the question, presenting the act of asking the question as a consequence of a shared background. In contrast, in the also-in-WH-question construction, we would specify that it presents the contents of the question as a potential consequence of what was discussed before.

Much work in cognitive linguistics supports the idea that language users retain such item-specific information and even specific exemplars in addition to generalizations and sub-generalizations (cf. Langacker 2008; Taylor 2012; from a language acquisition perspective see Abbot-Smith & Tomasello 2006). Such a model is also plausible from a diachronic perspective since we have seen that historically, MPs are restricted to certain contexts of use in an item-based manner (see section 2.4.2 above). Hence, we need to assume conventional relationships between MPs and sentence type constructions (i.e. the existence of item-specific constructions that specify attested combinations between particles and sentence types such as the also-in-WH-question construction discussed above). This is unproblematic for construction grammar since, following the usage-based model, this framework allows the redundant encoding of schematic and fully instantiated structures if they are only sufficiently entrenched (see Goldberg 2006). Nevertheless, the constructions involved do express the generalizations that exist across different uses.

An alternative view would be to encode all sentence type restrictions in the MPs themselves; this would also account for the item-based relationships that we have seen. However, such an approach does then not account for the fact that there are generalizations identifiable across items; a construction grammatical approach allows the flexible encoding of (sub-)generalizations and idiosyncrasies at various levels.

To sum up the discussion in section 4, we hold that all MP instances inherit from the MP construction the general function to relate the current utterance to common ground, i.e. to some communicatively given background which is taken to be “at hand”. Furthermore, the model comprises the different sentence type constructions, namely pairings of form (verb morphology, word order, grammatical category, and possibly intonation, see section 4.1 above) and meaning (grammatical mood and further functional contributions). The third component of the model are the lexical constructions, i.e. the individual particle lexemes.

Then, in order to specify the interaction between these three resources, we describe constructions of different degrees of specificity and lexical filling. Since construction grammar does not necessarily target a characterization of those structures licensed by the grammar of a language and only those (cf. Goldberg 2006), but instead assumes entrenchment and preemption to guide users to common, accepted structures, we can state generalizations on a general level and constrain them by representatives of attested uses. For instance, for verb-second (V2) declaratives, we can assume a sentence type construction on a high level of generality which includes an optional MP slot; at the same time, there are more specific constructions of specific MPs in this construction, for instance, ja as an MP in the V2 declaratives, aber as an MP in V2 declaratives or doch as an MP in V2 declaratives, which guide the speaker into accepted, entrenched structures. In addition, we assume a lexically filled V2 declarative construction with können/dürfen (‘can’/’may’) which can take the MP ruhig. Thus, to prevent the licensing of constructs that are not generally acceptable, it is sufficient to represent that (i) instances of ruhig as an MP in V2 declaratives are very infrequent unless associated with a directive interpretation (see the examples in Métrich et al. 2002: 28–34; and our discussion in footnote 9) and (ii) that there is the construction with the MP ruhig in V2 declaratives with können/dürfen and other modal verbs, which preempts other kinds of uses of ruhig in V2 declaratives. That is, the scarcity of ruhig in general V2 declaratives serves as evidence for speakers that the use of the construction is restricted (Stefanowitsch 2006; Taylor 2012), given the relative frequency of both the lexical item and the construction.²⁴ In this way, the item-specific constructions account for the fact that MP lexemes occur only in some and not in other sentence-type constructions.

Based on the sentence type restrictions of MPs, we see several advantages with the construction grammatical framework over other approaches to sentence types and MPs.

First, in contrast to most previous work, a construction grammatical representation accounts naturally for the interplay between form and meaning components relevant for MP use. Similarly, in most formal approaches semantic meanings and pragmatic meanings are usually strictly separated, whereas the example of MPs in indirect speech acts shows that they interact with each other in complex ways. It is one of the defining features of construction grammar that the meaning-side of a construction can comprise both semantic and pragmatic meaning on equal terms.

Second, in construction grammar it is unproblematic to state the same construction several times on different levels of abstraction, which is needed in the case of partially filled constructions with specialized uses (see the discussion of the MP ruhig in section 4.1 above).

Third, in our model, sentence type restrictions are handled on a constructional level, not as a property of each MP. By distinguishing between the lexical MP constructions, the modal particle word class construction, the sentence type constructions and the lexically specified sentence type constructions, we can state generalizations across MPs and across sentence types (see discussion in section 4.4).

In sum, the complex interaction between sentence types and MPs reveals the need for a theoretical framework that is flexible enough both to handle several different abstraction levels in parallel and to account for the simultaneous interaction of semantic, formal and functional meaning with each other as well as with lexical items, and construction grammar can provide such a framework.