Bare Root Infinitives in German: Facets of Meaning in Poetry and Non-Literary Discourse

The aim of this article is to show how linguistic and literary studies can benef it f rom the joint analysis of linguistic structures in poetry. Firstly, the analysis of poetry has an important impact on linguistic theory as it leads our attention to specif ic structures and meanings that so far have not been considered. Secondly, a close linguist ic analysis can reveal hitherto overlooked facets of meaning which have a great signif icance for the overall interpretation of a poem. We focus on Bare Root Inf initives (BRIs) in German. As they lack the features for tense, mood, person and number, they are more f lexible in meaning than f inite forms. When looking at poetry, besides the well known deontic and bouletic meanings (cf . Reis 1995, 2003; Gärtner 2014) a third meaning that we call reactive meaning stands out. Remarkably, this reactive meaning can also be found in everyday language. Its specif ic semantic properties show that a semantic analysis of BRIs in the style of Kaufmann (2012) is adequate: modality, but not non-referentiality, is a grammatically given semantic property of BRIs. The specif ic case study of the poem ‘muster f ixieren’ (‘f ixing patterns’) by Nico Bleutge reveals how the restricted context of the poem interacts with the dif ferent interpretations of BRIs, resulting in a complex interpretation of the text.


Introduction
In this article, we intend to show that both linguistic theory and literary scholarship can benefit from a joint analysis of poetic texts. 1 We will argue that, firstly, poetry is an important data source for theoretical linguistics (cf. Bade and Beck 2017). Quite often, the close examination of poetic texts reveals structures and interpretations that have not been looked at by linguists in detail. Moreover, literary analysis benefits from a detailed linguistic analysis: pointing out all the possible interpretations for a linguistic structure helps to come to a more complete understanding of the text and its overall meaning.
One of the aims of this paper is to investigate if these two readings are the only possible interpretations for BRIs, or if there are other readings that are both possible and salient. To answer this question, we will look at BRIs occurring in poetic texts. Our preliminary assumption is that poetic language is part of natural language and thus adheres to the same rules of grammar that apply in everyday language (cf. Kiparsky 1973;Bade and Beck 2017). Nevertheless, poetic language is special in some respects; in particular with regard to syntactic and semantic flexibility, poetry exceeds the limits of everyday language, especially where non-conventional meaning-making is desired (as is often the case in poetry). Hence, poetic language can provide us with data that might have been neglected in linguistic research so far. One of the peculiarities of poetry is that we are often left without a discourse context that would indicate more specifically how to interpret the BRI. What we mean by this is that poems are not necessarily accompanied by additional information or a certain framing within which they are made to 'make sense'; rather, poetry often is read and analysed detached from its context of creation, but still provides enough of its 'own' context to be interpreted coherently and cohesively with what the text itself provides linguistically. We here understand poems as self-sufficient, holistic units of text, as opposed to e.g. individual passages taken from larger bodies of text, or occurrences from everyday discourse (such as text messages sent by an unknown number) for which we similarly lack conversational context. Thus on the one hand, finding an interpretation for BRIs is more challenging than in the examples given in (1) since we assume that the poem's discourse context is contained within itself. On the other hand, this 'freedom of interpretation' can help us figure out rare and exceptional readings that might be the most plausible ones in the context given by the poem. Our central example in this paper is the contemporary German poem ʻmuster fixieren' (2013) by Nico Bleutge (*1971). We will consider in particular the single BRI occurring at the very beginning (muster fixieren l.1): (2) muster f ixieren, zweige der abend wuchs schon ins abteil als der zweite zug herankam. sitze schwanden, wurden durchsichtig f üllten sich mit jenen dünnen anderen reihen, nur kurz, ein aufsehen, ansehen, das auf einer höhe stillzustehen schien am bahndamm klaf f ten kabel gußpf eiler, of f en der schnelle glanz auf einem f ührerhaus 1 2 3 4 5 6 7 8 9 10 11 f ixing patterns, twigs the evening already grew into the compartment as the second train came in. seats ran out, became translucent f illed up with those other thin rows, just brief ly, a look up, a glance that seemed to stand still at one level at the railroad embankment cables were gaping cast iron pillars, open the quick sparkle on a driver's cab muster fixieren can have a bouletic or a deontic reading (along the lines of our examples in (1)), butsurprisingly!it can also have a descriptive function for the overall text: the rail traveller, presumably the speaker, is staring at the patterns that can be seen while looking out the window. Note that in the context of the poem this reading is not at all far-fetched: it aligns very well with the observant attitude the speaker adopts throughout the poem.
Immediately, this poem makes us aware that there are more readings for BRIs than are usually considered in the linguistic literature, i.e. we have to add a 'descriptive' reading to the deontic-directive and the bouletic-optative readings. The descriptive reading is characterised by the fact thatin contrast to the two other readingsthe BRI-event actually takes place: in this case, 'to fix patterns' does not remain an order or wish but is an action or event in the present.
In section 2 we will present a linguistic analysis of BRIs. Special emphasis will be placed on the question of whether such descriptive readings also exist in everyday language, and if so, what their impact is on the theoretical approaches to BRIs. In sections 3 and 4 we will turn once more to poetry: after giving a brief overview of the different functions of BRIs in poetry (section 3), we will analyse Nico Bleutge's poem in more detail, and subsequently review to what extent our theory can account for the different interpretations of the BRI muster fixieren, paying special attention to their meaning for the overall interpretation of the poem (section 4).

Modality, (non-)referentiality and illocutionary type
Beyond the modal meaning there are two other meaning components that are often attributed to BRIs. One of them is non-referentiality, the other one the specific illocutionary type of the BRI. In this section we will comment on these three meaning components and the way they interact. a) Modality: By the concept of modal meaning we understand a specific attitude towards the propositions that are expressed, i.e. beliefs, wishes, requests, etc. We have already seen that for BRIs, bouletic and deontic modality are crucial. Following Palmer (2001), we assume that the distinction between bouletic and deontic modality reflects whether the conditioning factors are actor-internal or actorexternal. Bouletic modality expresses that someone hopes or intends to perform an action or to participate in an event, hence it has an action-internal source. Deontic modality expresses that an action or event is possible, necessary, permissible, or obligatory, given an external source; this external source can either be given by law, by moral principles or by an action-external person. 6 Both bouletic (3) and deontic (4) modality can be expressed by a variety of linguistic means; in (3a) and (4a), the modality is due to a modal verb. In (3b) and (4b), it is due to a BRI: (3) a. Peter will schlafen. Peter wants.to sleep b. Endlich schlafen! finally sleep (4) a. Peter muss arbeiten. Peter has.to work b. Arbeiten! work b) (Non-)referentiality: (Non-)referentiality refers to the relation between a proposition and the world. Referential propositions describe situations occurring either in the actual world or in a possible world. Non-referential propositions are not descriptive; they are 'pure meanings' without a grammatically given relation to the world: a nonreferential proposition is neither part of the actual world nor of any possible world. Reis (1995Reis ( , 2003 argues that referentiality is closely connected to finite inflection; thus, in her approach, BRIsas non-finite verb formsare nonreferential and have no descriptive power. c) Illocutionary type: The illocutionary type belongs to the communicative domain of language. Illocutionary types establish specific interpersonal relations (between speaker 6 von Fintel (2006: 2) defines deontic modality as being obligatorily based on ' a body of law or a set of moral principles or the like'. As a consequence, he classifies an order of a parent to his child as bouletic: You have to go to bed in ten minutes [stern father; bouletic] (von Fintel 2006: 2 (9)). In contrast, we classify all actor-external obligations as deontic, whether they result from law, principles or personal orders. and addressee) with regard to the proposition expressed. Utterances with a deontic BRI (see examples (1a), (4b)) are directive speech acts, whereas utterances with a bouletic BRI (examples (1b), (3b)) are optative speech acts. If a BRI is used as a directive speech act, the speaker imposes an action to the addressee. If a BRI is used as an optative, he commits himself to a specific wish. In other words, directive speech acts are hearer-oriented whereas optative speech acts are speaker-oriented. 7 Modality, (non-)referentiality and the illocutionary type are related features of language; however, the illocutionary type is a feature of pragmatics, whereas modality and non-referentiality are semantic properties. It remains to be shown in which way they depend on each other in the case of BRIs. We assume that there is only an indirect relation between grammatical structures and illocutionary type, whereas semantic properties can be grammatically anchored, i.e. directly determined by grammatical structures. In the remainder of this section we will discuss the question of whether modality or non-referentiality are determined by the morphosyntactic structure of BRIs.
According to Reis (1995Reis ( , 2003 it is the non-referentiality of BRIs that is grammatically encoded. In a nutshell, she assumes that the morphosyntactic structure of BRIs, including a non-finite verb form, obligatorily leads to nonreferentiality, since only finite verb forms can refer to referentially anchored events. Furthermore, Reis argues that in non-referentiality modal meaning arises only through implicatures. The specific modal flavour (deontic/bouletic) and the interpretation of the implicit subject is determined by the context. The illocutionary type of the BRI, in turn, is dependent on the interpretation of the subject. With deontic BRIs, the implicit subject corresponds to the addressee(s); as a consequence, deontic BRIs are always directive. For bouletic BRIs, the implicit subject corresponds to the speaker; hence, bouletic BRIs are always optative.
Alternatively, one could assume that modality is a grammatically encoded property of BRIs. Such an analysis would be in line with Kaufmann (2012) who proposes that imperatives carry a modal layer of meaning in their logical form (LF). BRIs share a considerable number of properties with imperatives: like imperatives, they have a modal interpretation, they have an implicit subject with the feature [+human] (or at least [+intentionality]) and they show a certain flexibility in their semantic and pragmatic analysis (cf. Fries 1983;Reis 1995;Gärtner 2013Gärtner , 2014. A difference between imperatives and BRIs concerns the modal flavour: whereas imperatives are always deontic BRIs can be either deontic or bouletic. 8 Let us reconsider example (1): (  (2012) and Kratzer (1991) would be to assume an LF for BRIs that contains a covert modal of universal force. For (1a) the modality must be described as deontic, especially if there is a contextually defined addressee, for example a pupil in a given classroom context: (5) λw. ∀w' [w' adheres to the speaker's requests in w → add (c) sits down in w'] 'In all worlds that adhere to the speaker's requests in the evaluation world, the addressee sits down.' 'You must sit down!' In (1b), instead of deontic modality, we are rather confronted with the wishes of the speaker with regard to her or his own actions. Accordingly, the sentence should receive a bouletic interpretation: (6) λw. ∀w' [w' adheres to the wishes of the speaker that are relevant in w → sp (c) sits down in w'] 'In all worlds that adhere to the speaker's wishes in the evaluation world w, the speaker sits down.' 'If I only could sit down!' Hence following Kaufmann (2012), the covert modality of BRIs would be interpreted as being inherent to the semantic structure of BRIs. With regard to the other meaning components, there is no difference to Reis (1995Reis ( , 2003: the different modal flavours, the interpretation of the subject and the illocutionary type would arise from the context in the exact same manner.
Overall, the outcomes of analyses in line with Reis (1995Reis ( , 2003 and Kaufmann (2012) arrive at the same meaning of the BRI; yet going along with Kaufmann (2012), modality is determined on the level of LF, whereas Reis (1995Reis ( , 2003 assumes that the modal reading is derived in two steps: the syntactic structure obligatorily leads to non-referentiality, and non-referentiality causes modal meaning via an implicature. In short, Kaufmann (2012) proposes that modality, but not non-referentiality, is a structurally given component, whereas Reis (1995Reis ( , 2003 takes the opposite view, deriving non-referentiality from structure and considering modality as an implicature. In a way, Reis (1995Reis ( , 2003 is more restrictive than Kaufmann (2012). Following Reis, we expect all BRIs to be non-referential, and consequently also modal. Kaufmann's semantic approach only predicts that BRIs have a modal meaningit does not exclude referential uses of BRIs. Assuming that only modality is a grammatically encoded property of BRIs, there could be more variety of meaning, and we would not be surprised to encounter referential readings of BRIs.

The reactive reading: A 'touchstone' for the theory
In the preceding section, we have shown that the deontic and the bouletic BRIs are interpreted as both modal and non-referential. It is hard to decide whether both properties are grammatically given, or whether one of them is only a pragmatic effect of the other one. The descriptive reading mentioned in section 1 can be useful to discriminate between these possibilities. Interestingly, the availability of this third reading has not received any attention from a linguistic point of view. Hence, after discovering this descriptive potential in Nico Bleutge's poem 'muster fixieren' we tried to find out whether descriptive BRIs occur exclusively in poetry. Internet research quickly revealed that descriptive BRIs occur in everyday German. Consider the following excerpt from a post to an internet forum where young parents write about their experiences: It is clear from the context that the infinitive is used for the description of a referentially anchored situation. Furthermore, since (7) is a kind of everyday life diary, the missing subject is automatically understood as referring to the writer. In this specific context there is no need to indicate the agent of the actions describednote that the BRIs alternate with impersonal and personal passives . At first sight, we might assume that in such unambiguous cases the BRI is simply used for matters of convenienceit is shorter than a finite clause (which would include an inflected verb and an obligatory subject), and obviously it can refer to a referentially anchored situation if the context ensures this referential anchoring and specifies the implicit subject. However, such a use of BRIs would be a major problem for both of our approaches: Reis (1995) claims that non-referentiality is syntactically determined and following Kaufmann (2012), we would have to assume that a modal component is semantically fixed. Neither approach can account for using BRIs like finite clauses, i.e. in a referential and non-modal way.
If we take a closer look at (7), however, we notice an interesting difference between the use of BRIs and finite clauses. The BRIs are only used for the description of actions that are considered to be in a kind of unchangeable, recurring course of events, i.e. they have to be done every day and there is no choice for the agent (here: the speaker) about whether or not to perform them. As soon as there is a choice, a finite clause is used (Entweder mach ich noch Haushalt, oder hab auch mal Pause / Entweder bleiben wir dann noch auf dem Spielplatz, oder aber gehen hier in den Garten). This leads to the impression that the BRIs in this diary post are indeed referential but nevertheless include a kind of modality: the actions they describe are 'no choice'-actions. We therefore call them reactive.
Note that this 'modal flavour' of referentially anchored BRIs is not necessarily connected to the fact that the BRIs in (7), stemming from an 'everyday diary', are used to describe recurring situations. In (8) and (9), the BRIs are used to describe actions within a series of unique events. Nevertheless, they also express reactive events, i.e. they present 'no choice'-actions that have to be performed given a specific situation: Significantly, most of the BRIs in (8) and (9) are introduced by also (ʻthus'), indicating that the events follow causally or (quasi-)automatically from the situation described before.
Understanding that BRIs always include modality in their semantic representation even if they refer to referentially anchored events, we henceforth explore an approach in the spirit of Kaufmann (2012). We do not yet completely discard an analysis according to Reis (1995Reis ( , 2003, as one could assume that the non-referentiality of BRIs is overridden in specific contexts; note that Reis (1995: 125) considers the non-referentiality of the infinitive marker to be a default interpretation. However, it remains to be shown how the modal reading of referential reactive BRIs arises if modality is an implicature of nonreferentiality. Hence, we prefer the analysis inspired by Kaufmann (2012), as its grammatically given modality can more clearly derive the readings we observe. In (10), we give a formal representation of the reactive modality observed with also aufstehen 'so get up' in (7i): (10) λw. ∀w' [the speaker considers w' as a very likely consequence of the circumstances in w → p (c) gets up in w'] 'In all worlds w' that the speaker considers as a very likely consequence of the circumstances in w, a contextually prominent person gets up' Reactive modality is defined as an immediate and necessary consequence of the circumstances given in the actual world. Like deontic and bouletic modality, reactive modality has universal force. Thus, BRIs always involve universal force, independent of the specific modal flavour at work. The representation in (10) also makes clear that reactive modality is a sub-kind of circumstantial modality: 'Circumstantial modality […] concerns what is possible or necessary, given a particular set of circumstances' (von Fintel 2006: 2). We will, however, stick to the term reactive modality, since it captures the properties of BRIs more precisely than circumstantial modality: BRIs are restricted to quasi-automatic reactions of human beings. whereas circumstantial modality includes all sorts of events resulting from the circumstances (i.e. all kinds of natural facts besides human reactions).
The representation in (10) does not exclude a referential anchoring of the BRI. In fact, it even makes it plausibleif an action is considered a possible and very likely reaction to a specific situation, it is also expected to take place. 9 Hence, the reactive modality is completely compatible with referential anchoring. 10 Note, however, there are also examples with the reactive reading that are nonreferential. Consider the following excerpt from an online travel diary (Matzmobil 2007): We conclude that the reactive reading is not necessarily referentialit is just compatible with referentiality. 11 Next, we have to ask where the obligatory nonreferentiality of deontic and bouletic BRIs stems fromif it is not determined by syntactic structure. We assume that (non-)referentiality is due to the specific illocutionary type: deontic BRIs are directive, bouletic BRIs are optative; and it is a felicity condition for directive and optative speech act types that their propositional content is not yet fulfilled. It follows that for deontic-directive and bouletic-optative BRIs, the proposition cannot be referentially anchored, whereas it is very likely to be referentially anchored in the case of reactiveassertive BRIs.
In sum, the difference between the bouletic/deontic BRIs on the one hand and the descriptive BRIs on the other hand is not that the reactive BRIs lack modalityall BRIs have a grammatically given modal component. The difference is that the illocutionary force of the bouletic/deontic BRIs is not 9 A similar phenomenon can be observed with modal verbs expressin g deontic or bouletic modality. Very often, the modal meaning is perfectly compatible with a factual interpretation of the proposition: (i) Gestern musste Kerstin arbeiten. yesterday had.to Kerstin work 'Kerstin had the obligation to work yesterday (and in fact she did).' (ii) Ich fuhr nach Paris, weil ich das unbedingt wollte. I went to Paris because I that absolutely wanted 'I went to Paris because I absolutely wanted to go there.' compatible with a referential anchoring of the proposition, whereas the illocutionary force of the reactive BRIs is compatible with it.
This difference also accounts for the way the infinitive's implicit subject is interpreted. With deontic-directive BRIs, the subject obligatorily corresponds to the addressee, while with bouletic-optative BRIs, it corresponds to the speaker (cf. Reis 1995Reis , 2003. For reactive-assertive BRIs there is more choice. In some cases, the implicit subject corresponds to the speaker (see example (7)) or to a group including the speaker (see example (8)). However, this is not a necessary condition; in particular, if the BRI is part of a longer narrative structure, it can also refer to another contextually salient person, as it does in example (9).
Let us summarise the three readings for BRIs by placing them within a framework along the lines of Kaufmann (2012): − We assume one LF with a modal operator and an implicit subject.
− The modal operator always carries universal force. This is the part of modality that is grammatically given for BRIs.
− The modal flavour and the variable assignment for the implicit subject are not fixed grammatically but dependent on the context. The illocutionary type varies accordingly:  (7), we expect a reactive or bouletic-optative, but not a deontic-directive reading. If we further consider the lexical meanings of the BRI-verbs in (7) and consider their enumeration in the post, we clearly see that only a reactive-descriptive reading is plausible here. Other factors for the pragmatic interpretation of BRIs could be clusters of prosodic cues in the case of spoken language or, in the case of written texts, punctuation. These factors are however not absolutely comparable in influence. In spoken language, an intended interpretation might be singled out by prosodic factors. In written language, punctuation can be taken to distinguish between at least the deonticdirective and the bouletic-optative reading on the one hand and the reactiveassertive reading on the other: due to their illocutionary type (directive, optative), the first two readings often occur with an exclamation mark, whereas the reactive readingas an assertionusually occurs with a full-stop (or a comma, if the sentence continues). We only consider written poetry in this study and therefore neglect the role of prosody in potentially disambiguating an utterance. However, in our study of poetry we do not consider punctuation a reliable means to discriminate between the different readings either, since it is often used more flexibly and unconventionally in poetry than it is in everyday language. Accordingly, when analysing the poem 'muster fixieren' in detail in section 4, we will show that the interpretation of the BRI exclusively relies on the context and the overall interpretation of the poem.

Evidence for the interpretive potential of BRIs from poetry
Before proceeding with an in-depth analysis of Nico Bleutge's 'muster fixieren', we will look at some other German poems that reveal the versatility of BRIs. BRIs in poetry are widespread and not limited to a specific use. We have collected a few examples that we will discuss only briefly to illustrate this claim. Consider first this excerpt from Adelbert von Chamisso's poem ʻKüssen will ich, ich will küssen' (' Scheiden, meiden in the given context have to be analysed as BRIs and not as nominalisations (cf. the occurrence of muss in the beginning of the poem, which selects a bare infinitive). The reading of these BRIs is definitely not bouleticoptative as the BRIs do not correspond to a wish of the implicit subject. A deontic-directive reading seems to be excluded as well, given that the implicit subject corresponds to the speaker. We argue that the BRIs must be attributed a reactive meaning, since the events described seem to be an inevitable consequence of a specific situation. In the case of the BRI scheiden, the departing is already taking place, whereas the BRI meiden cannot assume this descriptive meaning. Both cases are consistent with the semantic description given for the reactive BRI in section 2.2.
To give another example, consider the following lines from Kurt Tucholsky's ʻDer Sucher' ('The Searcher'), in which the deontic reading prevails:  (Tucholsky 2011: 170) Verdienen could be a mere description (reactive reading), a wish (bouletic reading) or a kind of order (deontic reading). The notion of motivating the addressee to carry out different actions is prevalent throughout the poem; see, for instance, the first lines: Suchsuch / suche immer nach dem Geld ('search -search / keep searching for money'). The speaker continuously appeals in this manner, and finally turns on the addressee, who is to adopt the same mode as wellwe see this in the line preceding the BRIs denk immer nur: ('always keep in mind:'). The line can thus be read as a form of self-address by someone who is telling him-or herself to 'earn' money. In the overall context of the poem, there seems to be a transition from a deontic-directive to a bouletic-optative reading of the BRI as the addressee begins to desire what the speaker commands.
Finally, to show the variety of poetic expression, we also want to present an example from poetic language that eludes syntactic regulations, and thus conventional interpretation. Many of August Stramm's poems are representative of expressionist poetry. They are not coherent on a syntactic or semantic level, but rather offer rapidly switching images and impressions that can be likened to the art technique of pointillism, but relating to linguistic meaning; the notion is emphasised through frequent neologisms and transformation of verbs. 12 Consider the following excerpt from 'Zwist' ('Quarrel'):  (Stramm 1990: 41) The verbsall of which appear to be in their infinitival formare not connected to any other sentence elements but are merely strung together in a chiefly associative manner. Given the lexical semantics of the verbs, the deontic and the bouletic reading, however, seem to be excluded. A reactive reading is not implausible; it is possible to imagine a situation in which the involved events take place almost automatically without being controlled by an agent. The strongly restricted context of the poem does not really clarify what function the BRIs serve: the text might simply imitate the visual medium in evoking associations in the reader. In Stramm's poem, we are thus confronted with the limits of interpretabilitythe reactive reading of the BRI is not excluded, but language use here moves away from structure and coherency, and is transformed into a form of communication characterised by its recurrent rhythmic and phonemic patterns and its resulting self-referential quality rather than coherence and the distribution of information.
To summarise, if we assume that poetry typically offers the reader different interpretive possibilities, then the semantic underspecification of BRIs seems to fit poetry very well. In the next section, we will show that the ambiguity of BRIs plays an important role in the interpretation of Nico Bleutge's poem 'muster fixieren'.

'muster fixieren': An illustrative interpretation and its consequences for linguistic theory
The poem 'muster fixieren' is part of a cycle under the heading of ʻfarnkraut' ('bracken') and is set off from surrounding poems by whitespace and page breaks, so that a distinction between each individual poem can be made even though they are not designated by titles. The cycle was published in 2013 in the volume verdecktes gelände.
We will first comment on functions of non-finite form and their interpretation (4.1) in this poem, and then consider in particular the single BRI occurring here (muster fixieren) (4.2).

Non-finite (and finite) verb forms in the context of the poem: Function and interpretation
The poem 'muster fixieren' focuses on visual and acoustic impressions collected during a train ride. The human subject experiencing these sensations remains implicit. This covert speaker functions as an observer position that relates elements of and events taking place in the environment. In lines 2 and 3, for instance, both time frame and location are specified: it is in the evening (der abend), and abteil ('compartment') as well as der zweite zug ('the second train') indicate that the deictic centre is located within a train compartment. The subject is thus an observer, speaking from the position of a rail traveller.
Given that the poem focuses exclusively on the observations of the traveller, it is not surprising that there is only one BRI: a description of inanimate objects or nature could not happen through BRIs, since they are restricted to subjects with the property [+human]. There are, however, many other non-finite forms in the poem, in particular nominalised infinitives and present participles. In contrast to the BRI, these non-finite forms are suitable for a description of the environment: they allow a syntactic realisation of their base verb's subject argument, and this subject argument is not restricted to the feature [+human]. Accordingly, it is perfectly accurate that the subject argument of the present participles kreisend ('circular') l. 20 and rufend ('calling') l. 22, and of the nominalised infinitives rufen ('calling') l. 19 and wachsen, anwachsen ('growth') l. 20/21 refers to a flock of geese.
It is interesting to note that at the beginning of the poem the description is based on finite verbs, whereas in the second part present participles and nominalised infinitives are used abundantly. This correlates with the content. The first visual impressions are connected to an oncoming train. This encounter is composed of a complex sequence of short and bounded events. Accordingly, telic predicates with finite inflection are used here (herankam 'came in' l. 3, schwanden 'ran out' l. 4, wurden durchsichtig 'became translucent' l. 4, füllten sich 'filled up' l. 5). The impressions following this encounter, however, seem to lack a fixed temporal order. Their simultaneity is first expressed by the use of stative verbs (stillzustehen schien 'seemed to / stand still' l. 8, klafften 'were gaping' l. 9); in the further course of the poem deverbal nouns (glanz 'sparkle' l. 11; durchdringung 'pervasion' l. 14) and the aforementioned present participles and nominalised infinitives are increasingly frequent. All these deverbal forms are suitable to describe simultaneous and intermingling impressions that cannot be clearly separated and put into a chronological order: they are eventive but lacking tense, i.e. they do not fix a specific topic time. With regard to nominalised infinitives, Lübbe and Trott (2017) have shown that they should be analysed as imperfective, hence as unbounded mass nouns, i.e. they describe an event from the inside and do not present it as completed, even if the verb lexically refers to a change of state (ein auf-sehen 'a look up' l. 6/7). Note, however, that towards the end of the poem the base verbs of the nominalised infinitives are themselves atelic degree achievement verbs (wachsen, anwachsen 'growth' l. 21), i.e. unboundedness is indicated here by both lexical aspect (Aktionsart) and grammatical aspect (via the nominalised infinitive). For present participles the effect of simultaneity is even stronger as they explicitly express an overlap of their event time with a contextually given topic time (cf. Lübbe and Rapp 2011;Rapp 2015). To summarise, nominalised infinitives and present participles are particularly adapted for the description of non-discrete and overlapping impressions. What is, however, the role of the BRI?

Different interpretations for the BRI 'muster fixieren'
Consider the opening lines: (15) muster fixieren, zweige der abend wuchs schon ins abteil als der zweite zug herankam. Sitze schwanden, wurden durchsichtig […] (Bleutge 2013: 20) fixing patterns, twigs the evening already grew into the compartment as the second train came in. seats ran out, became translucent (16) muster fixieren patterns.N fix.INF 'fixing patterns' BRIs 13 cannot be used to refer to natural phenomena or inanimate objects. It therefore has to be a human referent who is fixing the patterns. The denotation for the German verb fixieren ('to fix') offers a range of different meanings. For simplicity's sake, we translate fixieren as 'to fix' throughout the article. It should be noted, however, that while the German verb fixieren encompasses a wide range of ambiguity, these different denotations would be disambiguated in English, for instance by differentiating between 'to fix (on/to) sth.' (in a material sense, e.g. 'to fix the roof' or 'to fix the weathervane to the roof') and 'to fixate on sth.' (in a more psychological sense, to focus on something or even obsess over it). The potential for ambiguity in the German fixieren leads to the following lexical ambiguity: (17) Possible readings a. 'to stare at someone or something; to focus one's vision on one thing' b. 'to fix something, e.g. by attaching it to sth. else or by writing it down' Note that in (17a), fixieren refers to an atelic event whereas in (17b) it is clearly a telic event. 14 Both variants involve a human agent; however, the reading in (17a) seems to be less intentional than the reading in (17b). Our analysis will explore how these two lexical meanings interact with the possible meanings of the BRI: the bouletic and the deontic reading preferably combine with the telic variant of fixieren, whereas the descriptive reading combines with the atelic variant. In the following sections, we will present these different options and their consequences for the interpretation of the text as a whole, taking the properties and functionality of BRIs into account. Beginning the poem with the words muster fixieren the author immediately alerts the reader to a variety of semantic fields or concerns to possibly arise in the following. In combination with the absence of capitalization and punctuation, the text primes its reader to expect more unconventional meaning-making. From the very first line, in other words, the reader is encouraged to remain open to alternative readings, and to revise his or her interpretation as s/he goes along in the reading process.

Interpretation I: The BRI 'muster fixieren' as expressing bouletic modality
As the phenomena observed are incoherent, non-discrete and transient, they are hard to remember in a structured way. It could therefore be a wish of the speaker to fix them, i.e. to put them in a specific mental place. This wish would give rise to a bouletic reading of the BRI, e.g. 'If only I could fix patterns and create a mental order out of this incoherent reality (but I can't).' This bouletic reading combines with the telic lexical variant of fixieren, i.e. the verb then refers to a mental action whose result is that visual and acoustic sensations are mentally well ordered and therefore easy to memorisehence fixed. Given the incoherence of appearances, such a wish for an ordering of things is not implausible.
Recall that at the beginning of the poem the speaker describes the changing state of the 'seats' (sitze, l. 3), which 'ran out, became translucent' (schwanden, wurden durchsichtig, l. 4), and consequently 'were filled up with those other / thin rows' (füllten sich mit jenen dünnen / anderen reihen, ll. 5-6). In lines 13-16 we are pointed to patterns more explicitly than before, when the speaker notes: eine ungleichmäßige durchdringung von dächern und balken fachwerkkanten an irregular pervasion of roofs and beams framework edges However, the vision of partitioned items, all relating to the construction of houses, is apparent in the patchiness or irregularity of the description. A pattern, but more distinctly the lack thereof, is described here. As the poem proceeds, this idea is amplified not merely by repetition but at one point through its extension by other sensual experiences. In line 19 following, the speaker begins to talk about the movement of geese:  ' (rufen, l. 19) and 'screams' (schreie, l. 20). Yet this calling is itself 'scattered' (vereinzelt, l. 19) and subject to permanent change (ihr wachsen, anwachsen 'their growth' l. 20/21). Shortness and irregularity are also indicated by phase particles (schon 'already' l. 23, noch 'still' l. 25) and adverbials (nur kurz 'just briefly' l. 6) throughout the whole poem. Given the incoherence and transience of all visual and acoustic impressions, patterns must be difficult to detect. This makes a bouletic interpretation plausible.
This reading can be seen in parallel to (5) ('Einmal Rom besuchen!'), where the speaker wishes to visit Rome. In fact, the same bouletic accessibility relation can be applied to 'muster fixieren' (see an interpretation of the infinitive in (20) below).
(20) λw. ∀w' [w' adheres to the wishes of the speaker that are relevant in w → sp (c) fixes patterns in w'] 'In all worlds that adhere to the speaker's wishes in the evaluation world w, the speaker fixes patterns, i.e. she creates a mental order out of her incoherent reality.' The implicit subject of 'muster fixieren' is of course the fictional speaker of the poem, and the illocutionary type is optative.

Interpretation 2: The BRI 'muster fixieren' as expressing deontic modality
A second possibility for interpretation is to read the BRI as deontic-prospective.
In this case muster fixieren can either be read as an imperative by the speaker to him-/herself or as an imperative to the addressee. Both interpretations go along with the telic, intentional variant of fixieren. However, there are conceptual differences between the two interpretations with regard to the kind of fixing. If the speaker addresses the imperative to him-/herself, fixieren could assume the meaning 'to fix something by writing it down.' This fits the overall context very well: an adequate way of fixing the transient impressions would be to write them down, keeping a tangible record.
But what happens if the deontic interpretation is directed towards the reader? Clearly, the reader cannot fix the visual and acoustic impressions directly as s/he is not an observer of the scenebut s/he can fix the linguistic patterns of and within the text. This option regards the meta-textual level in more detail, as linguistic structure and semantic fields become the focus of the addressee's investigations. The beginning imperative muster fixieren retroactively turns into a para-textual element, an instruction to the reader of how to deal with the lines that follow.
Formally, the text is marked by phonetic patterns. Consonance, for instance, pervades the poem at various points; the soundscape is especially dominated by sibilants: Similarly, assonance (see, for instance, above: am bahndamm klafften kabel, l. 9) emerges as a significant element in structuring the poem phonetically. Throughout the poem, various sounds emerge as tiny clusters. Furthermore, expressions constituting semantic fields can be considered as patterns; see, for instance, expressions of transience and motion, of shortness and irregularity, and of lack of vision. Hence, there is an interesting contrast between the content concerning the lack of visual and auditive patternsor at least the difficulty in fixing themand the obvious linguistic patterns that can be observed and fixed by the addressee.
The deontic reading can be captured as in (22): (22) λw. ∀w' [w' adheres to the speaker's requests in w → sp (c) or add (c) fixes patterns in w'] 'In all worlds that adhere to the speaker's requests in the evaluation world w, the speaker or the addressee fixes patterns' The implicit subject corresponds either to the addressee orin a self-addressto the speaker. Correspondingly, the illocutionary type is directive.

Interpretation 3: The BRI 'muster fixieren' as expressing reactive modality
As we have seen in the Introduction, it is also possible to understand muster fixieren as a description. We may thus be dealing with a description that tells us what is happening in a fictional world w. Ostensibly, the poem as a whole is about a train ride and the many things that can be seen along the way, whereas the speaker remains hidden. In this case, the aim of using a BRI could be to not draw attention to the observer, the rail traveller; in other words, the speaker could wish to remain as covert as possible in this poem. The effect is a depersonalisation of the text and an overall sense of generalisation. We are then less likely to read the poem as one particular experience of an individual, and more likely to understand it as a generalised snapshot of the described world, unrelated to any one person. In this descriptive reading, the covert speaker does not have any active role, s/he does not try to fix her or his impressions or to put things in their places/he is simply staring out of the window, and the perception of the ever-changing world passing by happens in a quite passive and reactive manner. This purely descriptive scenario goes with the atelic variant of fixieren. And it would imply that, despite the apparent incoherence of the visual and acoustic impressions, there are underlying patterns that can be observed.
The three possibilities (bouletic, deontic, reactive) remain valid throughout the poem. Strikingly, the order of events that is described works with both meanings of fixieren in (18)and with the three possible interpretations of the BRI: either the events are just transient impressions of an inactive observer, or there is abouletic or deonticperspective to actively fix the ever-changing world around, and to figuratively keep it in place. To summarise, we can read the poem as either a reactive description of what is observable, or as an attempt to fix singular elements in order to create coherence out of chaos. We argue, however, that a reactive-referential interpretation for this BRI seems to be the most plausible one. Deontic and bouletic readings are usually marked by clusters of prosodic cues. As these linguistic means are absent in the written text, it requires a larger amount of processing to make the deontic and the bouletic interpretations available to the reader. Furthermore, the reactivedescriptive reading fits best the local and the global context of the BRI. If we read the rest of the line as an ellipsis, i.e. 'muster fixieren, zweige [fixieren]', both the deontic and the bouletic readings are very unlikely. Against the background of the overall text, the descriptive-reactive reading is the most plausible one: the majority of the poem describes the environmental impressions of a rail traveller; as we imagine him or her staring at them as they pass by, we recognise that to fix patterns would be a suitable reaction to these ever-changing impressions. Interestingly, it is only in the lines following the BRI that the reader understands that this 'fixing' can be seen as a consequence of another event, namely the movement of the train. Thus, the reactive reading which might at the beginning of the text compete with alternative readings is supported once the further course of the text is considered. Furthermore, we argue that the deontic-directive reading that involves a meta-linguistic level will only emerge on a second reading against the background of the overall text. Hence, the interesting thing about the BRI muster fixieren is that this line opens the field for several different interpretations, of which the reactive-descriptive reading is most plausible, while the other readings are possible but less likely.
Most importantly, this openness of the initial BRI allows for an interaction between the different interpretations. A descriptive reading (as given in section I3) may leave the speaker covert and focus on a world projection, but the world described can also be considered the basis for I1, the bouletic reading of the text, in which the speaker wishes to be able to fix patterns him-or herself. The deontic reading as given in I2 is curious in that it elicits a transformation of and through language: the speaker receives visual and auditive impulses, which s/he in turn can describe only by means of language. If the initial BRI is then read as deontic and as an appeal to the reader to fix patterns, the (inaccessible) original impulses are being translated into textual and linguistic features that the reader can recognise and fix as patterns. Language creates its own, equivalent patterns in what the speaker is fixing, so to speak; moreover, patterns in language and textuality can be perceived as both visual (i.e. looking at the words on the material page) and auditive (i.e. sound patterns such as alliteration, consonance etc. which become especially apparent when reciting the poem). This mirroring of what the speaker does and what the reader is called to do goes hand in hand with the world designed in which the fix ing of patternsthrough a variety of meansis made paramount.

Conclusion
The analysis of Nico Bleutge's poem 'muster fixieren' has produced important insights for the intersection of semantic and pragmatic research, drawing our attention to the so far neglected reactive reading of the BRI. Once aware of this reading, one discovers more and more examples, not only in poetry but also in our everyday language. In terms of the linguistic analysis of the BRI, we see that an approach extending Kaufmann's (2012) analysis of imperatives can easily result in the three readings of BRIs, i.e. the bouletic-optative, the deonticdirective and the reactive-assertive readings, whereas the (facultative) referentiality of the reactive reading is a severe problem for Reis (1995Reis ( , 2003.
For literary studies, the in-depth discussion of 'muster fixieren' by Nico Bleutge can reveal how the limited availability of discourse context in poetry interacts with the analysis of BRIs to arrive at the complex text interpretation given above. The underspecification of the BRI gives rise to several different interpretations, some of which emerge immediately, while others come into play only at the global text level, leading to a complex meaning of the overall poem.