How to threaten in Russian: a constructionist approach

The article analyzes various linguistic means to carry out threats in Russian with special focus on 27 constructions tagged as “Threat” in the Russian Constructicon, a linguistic repository of more than 2200 constructions in the Russian language. The major purpose of the current study is to investigate what constitutes a threat in Russian and how threats are related to other constructions. Unlike talking about threats, performing them in Russian does not involve the verbs ugrožatʹ and grozitʹ ‘threaten’. Instead, speakers prefer to use various indirect strategies, such as the construction Pogovori mne eščë!‘Don’t you dare talk like that!’. Although the constructions involve considerable variation in form and content, they share a common structure. The proposed taxonomy suggests that threats comprise three components that can be referred to as “Cause” (the undesired action of the threatenee), “Condition” (the action that the threatenee should take to avoid the Content of the threat), and “Content” (the harmful action that the threatener promises to carry out). In most cases one or two components are left out and the remaining components are often referred to through metonymy. The article furthermore contributes to Construction Grammar: it is proposed that lateral relationships between constructions can be of two types, referred to as “Overlap” (sharing a common semantic schema) and “Disambiguation in context” (sharing a common constructional schema).


Introduction
In Russian, the way people describe threats and perform threats are quite different. For talking about threatening, it is common to use the special verbs ugrožatʹ and grozitʹ 'threaten' or verbs morphologically related to them. It turns out, however, that these verbs are hardly ever used in performing a threat in Russian. In other words, in a speech act directed from one person to another one generally does not find an utterance like Ja tebe ugrožaju 'I am threatening you'. Indeed, only two utterances are found in the Russian National Corpus (RNC) for search queries for the verb ugrožatʹ accompanied with pronouns in Nominative and Dative cases, and these two examples do not involve performing a threat. 1 No utterances are found in similar searches for the verb grozitʹ accompanied by two pronouns, one in the Nominative and one in the Dative case. Padučeva (1982Padučeva ( , 1985, who refers to Austin's (1962) classification of speech acts, suggests that the verb ugrožatʹ 'threaten' in Russian is not performative, unlike obeščat' 'promise' or sovetovat' 'give advice, suggest' that have an illocutionary force, as in Ja obeščaju prijti v sredu 'I promise to come on Wednedsay'. Similarly, Apresjan (1995, p. 212) mentioned the verb ugrožatʹ among non-performative Russian verbs. 2 To perform a threat, speakers of Russian instead resort to various indirect strategies. The primary focus of the present study is to analyze the linguistic constructions that are used to perform a threat in Russian. The analysis I propose will be couched in Construction Grammar, and more broadly, in cognitive linguistics (Fillmore et al., 1988;Croft, 2001;Goldberg 1995Goldberg , 2006. Construction Grammarians regard constructions to be the basic unit of language and define them as "conventional, learned form-function pairings at varying levels of abstraction and complexity" (Goldberg, 2006, p. 17). Since such form-function pairings vary in size and complexity, they occur on different levels of language "including morphemes or words, idioms, partially lexically filled and fully general phrasal patterns" (Goldberg, 2006, p. 5 Vendler (1976, pp. 142-143), an affirmative sentence 'I threaten you' is an "illocutionary suicide" and, therefore, is never a performative.
My argument can be summarized as follows. A number of constructions are used to perform threats. Syntactically, these constructions are diverse and the threat that they convey is often indirect. Semantically, threat constructions normally include two participants (which I will refer to as a THREATENER and a THREATENEE) and up to three components: Cause (the undesired action of the THREATENEE that the THREATENER does not approve and wants to stop), Condition (the action of the THREATENEE suggested by the THREATENER), and Content (what the THREATENER promises to do to the THREATENEE). Threat constructions are closely related to a diverse and extensive list of constructions with various semantics. The analysis of this constructional neighborhood offers new insights into the notion of lateral and vertical relations between constructions.
This article has the following structure. After a brief exposition of previous research and a presentation of my data (Sect. 2), in Sect. 3 I propose a simple model of threats in Russian, which I elaborate in the subsequent sections. Section 4 investigates the participants of this situation and the ways they can be encoded in various Russian constructions. Section 5 describes three possible components of threat and their interaction in particular constructions. In Sect. 6, I discuss the semantic neighborhood of threat constructions in Russian, which, in the terminology of Construction Grammar (Cappelle, 2006;Grafmiller et al., 2018;Diessel, 2019), is structured by the means of lateral and vertical relations. Section 7 contains concluding remarks.

What is a threat? Earlier approaches and data
The notion of threat has received considerable attention in linguistics and related disciplines. The Russian word ugroza 'threat' is defined in dictionaries as a "promise to do harm to another person" (Ušakov, 1935(Ušakov, -1940Evgen'eva, 1999;Ožegov & Švedova, 2006). Although this is helpful as far as it goes, we will see that a dictionary definition along these lines is too narrow to do justice to the meaning of the constructions under scrutiny in the present study. As we will see in Sect. 5, constructions for performing threats often involve additional semantic components. Moreover, in many cases the relevant constructions are polysemous, and frequently also serve the function of prohibitive or apprehensive constructions, as we will see in Sect. 6. Letučij (2007), who refers to Mel′čuk (1995), explores one construction: ja PronPers-Dat VP-Fut! as in (1), for which he coins the term ugrozativ (from ugroza 'threat').
(1) Ja According to Letučij, this construction combines two meanings: 1) the Cause of threat (a participant is carrying out an unwanted activity, in this case shouting) 2) the Content of threat (a speaker is promising to do something bad to the participant) Therefore, Ja tebe pooru indicates that the speaker 1) disapproves of shouting and 2) makes it clear that the interlocutor will face negative consequences if (s)he continues shouting, and the consequences will originate from the speaker. Although threatening is a distinct type of speech act in Russian, little is found in scholarly literature on what other linguistic strategies are used in Russian to perform threats. The present study aims to fill this gap.
For the purposes of the present study, I will analyze data from the Russian Constructicon, which is a free open-access searchable electronic database of multiword grammatical constructions of Russian. 3 The Russian Constructicon contains 2250 constructions (as of August 2022). These data show that the semantics of threat may be more complex than Letučij's (2007) ugrozativ would suggest. The constructions in the database are tagged in accordance with their semantics and syntax, and each construction is accompanied with 5 corpus-based examples, as well as a number of other tags. The semantic classification of constructions is created as a system of tags, where more than one tag can be assigned to each construction in order to capture the semantic nuances of the construction. For instance, constructions of the semantic type "Threat" commonly overlap with 14 other semantic types, such as Prohibition, Condition, Apprehension, etc. This indicates that the meaning of the relevant constructions can be rather complex. Although the Russian Constructicon offers an elaborate semantic classification and a description of each construction in the database, the detailed analysis of each semantic type and the nature of the relations between constructions both within one type and across neighboring types is a task for additional research. Similar studies for selected semantic types in the Russian Constructicon have been undertaken in recent years, cf. Endresen and Janda (2020)  The present study provides a thorough analysis of 27 constructions from the Russian Constructicon that bear the semantic tag "Threat" and are used to perform threats. 4 Although the number of threat constructions is not very large compared to some other semantic classes, the systematic expansion of the Russian Constructicon on previous stages created a rather representative list of constructions . Therefore, the number of threat constructions in the Russian Constructicon is sufficient to draw some conclusions about threats in Russian. Importantly, the Russian Constructicon makes it possible to compare threats to constructions with related meanings by exploring the overlapping semantic tags. 5 In the next section, I will argue that the semantics of threat constructions involves the interaction of three components. However, before we turn to the meaning of constructions, it is worth mentioning that there are no formal features that are shared by all threat constructions. Consider the phrase "Ja znaju, gde ty živešʹ" 'I know where you live'. Pronounced with a certain intonation and in a certain context this seemingly innocent sentence may be perceived as a threat. In other words, in appropriate contexts many sentences can be used to perform a threat. The use of certain gestures (such as drawing a finger across one's neck or mimicking a gunshot) may also convey the meaning of threat. However, since this article analyzes examples from written texts, the role of intonation, gestures and the broader situational context is beyond the scope of the present study. Instead, my main focus is on conventionalized, entrenched linguistic patterns, i.e., constructions that serve the function of performing threats. 3 The Russian Constructicon is available at https://constructicon.github.io/russian/. 4 Overall, the Russian constructicon contains 31 constructions tagged as "Threat", but 4 constructions are used to describe threats, rather than to perform them, and therefore they are outside the scope of the present study. A detailed analysis of constructions with the verbs ugrožatʹ and grozitʹ 'threaten', as well as with the nouns ugroza and groza 'threat', has been conducted by Nesset and Makarova (2022). These constructions are not directly relevant for the present study, since they are about describing threats, not performing them. 5 The full list of constructions under scrutiny is available via the Advanced Search function of the Russian Constructicon, and the specific subset of Threat constructions with additional annotation for this article is available at the separate TROLLing post (https://doi.org/10.18710/ZFYCOG). The notion of "threat" has attracted the attention of researchers in various research fields, such as psychology, forensic linguistics, pragmatics, and cognitive linguistics. Research in the discipline of psychology is mostly concerned with the effect that the speaker wants to inflict on the addressee. For instance, the goals of the speaker may vary from expressing the speaker's anger or scaring the addressee to punishing them for some action or coercing the addressee into some action that is desirable for the speaker (cf. Tedeschi et al., 1971;Bonoma & Tedeschi, 1973;Sinaceur & Neale, 2005). Psychologists also investigate the power imbalance between participants in situations of threat (cf. Harris, 1984;Culpeper, 1996). Forensic linguists are interested in identifying a threat in an utterance, since the perception of threat is a matter of interpretation (cf. Shuy, 1993;Fraser, 1998;Tiersma & Solan, 2012;Baranov, 2013;Gales, 2015). Research in pragmatics deals with the speech acts of threatening (introduced and revised by Austin, 1962 andSearle, 1969) and their interpretations (Nicoloff, 1989;Blanco Salgueiro, 2010;Muschalik, 2018), as well as the politeness strategies involved in such speech acts (cf. Limberg, 2009). Within Cognitive linguistics relative research has focused on the verbs for threatening in English, Dutch, and Spanish (cf. Langacker, 2000;Verhagen, 1995 andCornillie, 2004). The grammaticalization patterns of verbs for threatening have received considerable attention (cf. Miyashita 2007 and2008;Narrog & Heine, 2021). Although these approaches shed important light on the notion of "threat", they are not directly relevant for my research, which focuses on the linguistic constructions that are used to perform threats in Russian, in particular their semantics and their relations to other constructions.

A linguistic model of threat in Russian
The situation of threatening another person includes two participants who are in an asymmetric relationship: the THREATENER and the THREATENEE. 6 The THREATENER imposes a threat on the THREATENEE in a speech act, so the THREATENER usually represents the speaker and the THREATENEE -the interlocutor. Like psychologists who have studied threats (cf. Watts, 1991), I consider the relationship between the THREATENER and the THREATENEE to be asymmetric because one participant in this situation has power over another, thus being able to turn a potential threat into a real action. Even the notion of pustaja ugroza 'empty threat' arguably testifies to this power asymmetry: it implies that ugroza 'threat' without any specifications is usually loaded with Content, and therefore, that the THREATENER is normally able to carry out the action that (s)he promised in the threat. 7 In special cases, we may consider threatening a sign of weakness and an ultimate act of self-defense by the speaker, which makes it questionable if the THREATENER necessarily has more power than the THREATENEE or if it can be the other way round. However, even in special situations where threats may not be backed up by the power to turn them into action, the relation between the THREAT-ENER and the THREATENEE is asymmetric, although in such cases it is the THREATENEE that is 6 Alternatively, these two participants are called in the literature the SENDER/SPEAKER and the ADDRESSEE (cf. Fraser, 1998), the SOURCE and the TARGET (cf. Bonoma & Tedeschi, 1973;Muschalik, 2018). For the purposes of the present article, I will use THREATENER and THREATENEE, since these terms are self-explanatory and do not coincide with widely used semantic roles, which otherwise might have led to ambiguity. 7 The notion of pustye obeščanija 'empty promises' employs the same metaphor. In many respects, promises and threats are similar speech acts that differ in the polarity value of the effect that they have on the addressee: positive for promises and negative for threats. Notice that dictionary definitions of ugroza 'threat' typically use words meaning 'promise' in the explanation, as mentioned in Sect. 2. However, while the verb obeščat' 'promise' serves as a performative in Russian, ugrožatʹ 'threaten' does not normally have this function. Fig. 1 The structure of the Content of threat with two sub-components (Promise and Action). The left portion illustrates a situation where both Promise and Action are expressed, while the right portion illustrates an underspecified situation where only Promise is expressed more powerful. It seems reasonable to conclude that a threat always involves an asymmetric relationship between the participants. A simple and straightforward example of a threat is (2) In example (2), the THREATENER makes a promise to kill the THREATENEE. However, in most cases constructions have more complex semantics, such as the construction analyzed by Letučij (Sect. 2). In a thorough analysis of such constructions, we need to consider the two participants (THREATENER and THREATENEE), but in addition I propose to distinguish between three semantic components of the threat: 1) The Cause of the threat (the undesired action of the THREATENEE that the THREAT-ENER does not approve and wants to stop or punish for) 9 2) The Condition to avoid the threat (the desired action of the THREATENEE that the THREATENER suggests doing instead or the undesired potential action that the THREATENEE should avoid doing in the future) 3) The Content of the threat (what the THREATENER promises to do to the THREATENEE. This action is supposed to affect the THREATENEE in a negative way) a. Action (i.e. what exactly the THREATENER will do to the THREATENEE if (s)he does not obey) b. Promise (i.e. the promise to affect the THREATENEE in a negative way) Content is a complex component that can be analyzed as consisting of two sub-components, which I suggest calling "Action" and "Promise". My data suggests that a construction can include both sub-components as in the example (2), where the THREATENER promises to carry out the action of killing, or only include Promise as in the examples (3)-(4) below. Having only one sub-component, therefore, makes a threat underspecified. I have not found any clear examples where the Action is present but not the Promise. Both attested scenarios are illustrated in Fig. 1.
The distinctions between the two sub-components of Content are not always clear-cut. Prioritizing Promise over Action can be strategic for the THREATENER in order to leave the details of the threat to the imagination of the THREATENEE. Discussing the pragmatics of Promise is, however, beyond the scope of the present study. It is also an open question of how many components of threat are present in the minds of speakers, therefore, I can only limit myself to what components are expressed linguistically. In this paper, I discuss which components can be omitted and which are overtly expressed. While analyzing which components are present, I only look at the sentence that contains the construction and mostly do not analyze the larger context. The context can be surrounding sentences or a situational context, such as gestures, etc.
In order to clarify let us consider two examples. In example (3) the construction of threat includes the THREATENER (Ja), the THREATENEE (tebe), the Content (dam) and the Cause (tak sebja vesti). Admittedly, the Content here is rather abstract: the THREATENER simply indicates that (s)he will punish the THREATENEE but does not specify how. Arguably, therefore, only one sub-component (Promise) of Content is present here. The Cause identifies a previous or ongoing behavior of the THREATENEE that the THREATENER does not approve of and wants to stop or prevent from happening again. (3) PronPers-Nom PronPers-Dat datʹ VP-Inf! 10 Ja Another construction where a concrete Content is not specified is (4): I.GEN 'You must come back before ten, or it will be bad for you!' In example (4), the Content (smotri) simply indicates that the negative consequences will reach the THREATENEE, but the THREATENER does not verbalize what exactly will happen. Here again the Promise is present, but the Action is not. However, unlike (3), this construction does not involve a Cause but rather includes the Condition, which is encoded by the first clause. In (4), coming back before ten is what the THREATENEE must do in order to avoid the Content of the threat.
The examples discussed above show that we need to consider three components of the relationship between the two participants. To capture this, I propose the model in Fig. 2. 11 The two participants of the threat are on opposite sides. The typical power imbalance between Fig. 2 The model of the threat situation the participants is reflected by the relative size of the two circles in the model. The large rectangle between the participants represents the threat proper, i.e., the relationship between the participants. The smaller rectangles inside denote the three components of the threat. The arrows between the rectangles symbolize the relationships between the components. The Content usually appears as a reaction to the Cause and Condition, so the Content is located to the right and the arrow is directed from the Cause through the Condition to the Content. The Cause and the Condition are opposed as real vs. hypothetical actions that are related to the Content.
As I will show in Sect. 5 below, it is possible to express all three components in one construction. However, this is more an exception than a rule. Therefore, the next step is to find out what the attested combinations of components are. Furthermore, we must ask how the various attested combinations affect the meaning of the threat, and how each component is expressed. The next sections are devoted to a detailed analysis of the model. However, before we turn to the central part of the model (the three components of threats) in Sect. 5, it is necessary to discuss the participants. This is the topic of Sect. 4.

How are the participants of a threat expressed?
The participants of threats can be expressed in various ways in Russian. In this section I will explore the lexical and morphological patterns that encode the THREATENER and the THREAT-ENEE. The first distinction we must make is whether a participant is encoded as a separate lexeme or not. The lexemes in question are typically personal pronouns of the 1st person for the THREATENER and of the 2nd person for the THREATENEE.
ubʹju! 12 kill-FUT.1SG '[I] will kill you, I will kill you' In (5), the THREATENER is expressed with the personal pronoun ja 'I' in the Nominative case and the THREATENEE is expressed with the personal pronoun tebja 'you' in the Accusative case.
Four logical combinations for expression of the participants are possible. Table 1 gives an overview of the possibilities for lexical encoding of the THREATENER and the THREATENEE by means of pronouns. 13 "+ pronoun" indicates that the participant is encoded with a personal pronoun, while "− pronoun" shows the lack of a pronoun for the participant. Check marks show what combinations are attested for the constructions of threat in Russian. As shown, all 4 combinations are attested: both the THREATENER and the THREATENEE are expressed by a pronoun in (6), one or the other of them in (7) and (8), while in (9)  The participants can also be encoded morphologically by means of agreement affixes on verbs.
(10) znatʹ-Fut, kak VP-Ipfv.Inf! Budešʹ znatʹ, kak ubegatʹ ot roditelej! be-FUT.2SG know-INF how run-INF from parent-GEN.PL 'You'll know it's bad to run away from your parents!' In (10), the THREATENEE is encoded in the affix of the verb, which is in the 2nd person singular. The THREATENER is not expressed with a pronoun either, but one can interpret the speaker of the utterance. I consider this encoding of the speaker to be a "no marking" strategy since there is no lexeme or morpheme that expresses the THREATENER in the construction.  (6), (7), and (5), respectively. If the THREATENEE is not expressed by a pronoun, it is encoded in the verb (i.e., as a suffix) that can be in the imperative form or in the future tense form, as exemplified in (9) and (10). If the THREATENER is not expressed by a pronoun, it is assumed that the speaker is the THREATENER.
The marking strategies are summarized in Fig. 3. The circles symbolize each strategy, while the line with end nodes shows that they tend to form a scale from a more explicit way of encoding to a less explicit one.
Double marking is a combination of lexical marking by means of a pronoun and morphological marking by means of an affix, as in examples (5)-(6). In (5), the THREATENER is encoded both with a personal pronoun, i.e. lexically, and in the affix of the verb, i.e. morphologically. In (6), it is the THREATENEE that is marked both lexically and morphologically. While in (6) the pronoun is an obligatory part of the construction, in (5) one can easily drop the pronoun with no changes in the meaning of the whole construction. This is, however, a regular syntactic feature of first-person pronouns and Russian verbs in the indicative. The double marking is more explicit than either lexical or morphological marking since it marks the meaning twice: both with an affix and a lexeme.
An interesting peculiarity of threat constructions is a possible mismatch between the lexical and morphological encoding within one construction. In (11), the THREATENEE is lexically marked with a personal pronoun of the second person, while morphologically it is marked with the affix of the first person. The syntactic mechanism behind this construction is thoroughly analyzed by Letučij (2007).
The discussion so far concerns variation across constructions, but there is also variation within constructions when it comes to the encoding of the participants. For instance, in (8) the THREATENER can be absent since mne/u menja is an optional element of the relevant construction, which is marked in the Russian Constructicon with parentheses (). At the same time, in (9) the THREATENEE can be present as in Pustʹ on tolʹko poprobuet ne prijti vovremja! 'Just let him try being late!', since this construction involves variation (marked with a slash "/"). The latter also illustrates that the THREATENEE is not always the interlocutor but can be a different person in the situation. ' "And you must either get yourself a ticket to some destination, or go out!" told the conductor to Minin and Pozharski.' Examples (12a) and (12b) contain the verbs in the form of future tense and second person, while (12c) has two imperative verb forms. The THREATENEE is encoded with a personal pronoun in all three sentences, whereas the THREATENER can be missing, as in (12c) where the THREATENER is named in a subsequent clause. Constructions like (12), where the morphosyntactic formulae of the constructions in the Russian Constructicon do not indicate how the participants are encoded, are not included in Table 1. Examples like (12) testify to the flexibility of the Russian language when it comes to the expression of the participants in constructions of threat. To summarize, we have seen that there is considerable variation in expressing the participants since one or both participants can be omitted. We now turn to the relationship between participants. As suggested in Sect. 3, the situation of threatening in Russian can include three components -the Cause, the Condition, and the Content. These components can interact within one construction to create a certain meaning or yield pragmatic implications. In the following, I explore the patterns that emerge from my data. Figure 4 summarizes the combinations of components of threat that are attested in constructions in the Russian Constructicon. Circles with minus signs indicate the absence of a component. The rounded rectangle serves as a schema for component combinations that all overtly express the Content. The rectangles to the right from the general schema represent an extension where the Content is missing, and either the Cause or the Condition is present. The extension relationship is represented as a dotted arrow. Numbers in parentheses give the distribution of threat constructions in the Russian Constructicon in raw numbers. 15 Although these numbers are not very high, they suggest that the combinations are of similar frequency, except for the maximal combination "Cause + Condition + Content", that is only attested in two constructions.

Content
The first option is to express only the Content of the threat. The action that the speaker promises to do to an interlocutor can be very concrete, as in (13) In (13), the THREATENER specifies the action that will happen to the THREATENEE, which is to kill them. This is expressed by the verb ubʹju in the future tense. The construction can be used for a real threat to someone's life or as a hyperbolization meaning to cause a THREATENEE some trouble. In this construction, both Promise and Action sub-components are present.
However, the Content is frequently expressed in rather general terms, as in the construction (14) NP-Dat Cop kryška/konec/kranty/kirdyk/kaput/kajuk 17 Nu vsë, vam kryška! well all you.DAT lid-NOM.SG 'Well, that's it, you're done for!' Here the Content is encoded by one of the nouns that have the meaning 'the end of someone', implying 'death', and are to a larger or lesser degree colloquial. Interestingly, kryška literally means 'lid' (as on a coffin), but metonymically, similarly to the other words that are separated with the slash sign, means 'someone's death'. The Content of the threat in (14) is thus the same as in (13), insofar as the THREATENER promises to kill the THREATENEE. But in (14) this Content is only expressed indirectly through metonymy. This construction is also typically used to hyperbolize a threat.
A threat may be so general and abstract that it almost becomes subtle and implicit, as in the construction (15a) and its variants (15b-c).   (2000)] 'We won't let them do this. We will alert everyone. They will remember us one day, they will know how to offend those who originated in St.-Petersburg…' As noted by Kustova (2012, p. 360), nothing in the form of the verb vspomnišʹ 'remember-2.Fut' indicates the negative assessment of an action of remembering. Kustova highlights that neither the meaning of threat nor the negative assessment is attested in dictionaries. And in fact, the verb vspomnišʹ 'remember' is quite neutral, which is demonstrated by the fact that the other forms of this verb do not have any negative assessment. This indicates that the meaning of threat is created by the construction as a whole. The semantic mechanism, according to Kustova, is based on the following implicature: "I [THREATENER] will do something bad to you [THREATENEE]. That is why you will remember me". 19 Even though the verb vspomnišʹ 'remember' refers to the THREATENEE, it metonymically means 'the things that the THREATENER will do and the THREATENEE will remember', and that is why it can be considered to be the Content of a threat.
Construction (15), therefore, only contains Promise sub-component of the Content, while (14) occupies an intermediate position in the system. Constructions of this kind show that it is possible to convey a threat by just mentioning the Content without elaborating on it. Very often the Content is mentioned indirectly through metonymy.
A threat can be elaborated in two ways: 1) by adding the Condition to avoid the threat or 2) by adding the Cause for the threat. I will consider these options in turn.

Condition + Content
The Condition is typically presented in a biclausal construction where the second part expresses the Content of threat, while the first part expresses the alternative action that the speaker suggests doing, i.e. the Condition. It is implied that if the interlocutor performs the alternative action instead, the threat will not turn into a real action. 20 Therefore, in the Russian Constructicon the semantic type Threat here overlaps with the semantic type Condition, as it combines both meanings in one construction. Consider construction (16)  Here the THREATENER orders the THREATENEE to do what they want. The THREATENER also states that the THREATENEE will be punished for not fulfilling their order but does not specify exactly how (tebe ne pozdorovitsja literally means 'you will not feel physically well'). The order is usually given in a strict commanding tone. A more elaborate analysis of the neighborhood of the semantic type Threat is given in Sect. 6, where I will explore the relationship between threats and related meanings.
The presence of a Condition involved in the meaning of the conjunction a to / a ne to 'otherwise', as in construction (17) was pointed out by Podlesskaja (2000), Inkova-Manzotti (2005, p. 495), and Uryson (2010, pp. 67-69). The Condition here is referred to as a "negative condition" (Podlesskaja, 2000)

Cause + Content
The Cause of threat is an action performed by the interlocutor that the speaker does not approve. Expressing the Cause may be optional, as in the construction ( The combination of these two components (Cause and Content of threat) often results in a prohibitive construction. Thus, the threat can be used to stop a person from performing some undesired activity.

Cause + Condition + Content
So far, we have seen that either one or two components of the threat can be expressed simultaneously. But can all three components combine in one construction? The answer is "yes": the intersection of the three components (Cause, Condition, and Content) is found in the two following constructions (20) and (21)  These constructions encode that there has been some action made in the past that the speaker evaluates negatively and does not approve, i.e. that is the Cause. The speaker warns the interlocutor by providing them with a Condition: if this happens again, 24 the Content of threat will take place.
In other words, the example (20)  The Cause is not overtly expressed in this construction and comes as an implicature of the adverbial phrase eščë raz 'one more time' as follows: 'if something can happen again, it must have already happened at least once in the past', otherwise the speaker would not have included eščë raz in the sentence. Only two constructions in the Russian Constructicon simultaneously express all three components, while the others omit one of the components. In other words, to express all three components is possible, but seems to be rather unusual.

Relation between Cause and Condition
In fact, the Cause and the Condition are very similar as they refer to an action that is linked to the Content. I argue that these two components should nevertheless be treated separately for the following reasons. First, while the Cause is a real event that has taken place, the Condition is a potential event in the future and therefore belongs to the domain of irrealia. Second, at least in some cases the Cause and the Condition can be clearly separated. Such cases include situations where the action that the THREATENER does not approve of is irreversible and cannot be stopped or undone. For example, consider the construction znatʹ-Fut, kak VP-Ipfv. Inf! in (22). In this example, the THREATENER'S intention is to punish the THREATENEE for causing trouble to the co-workers, i.e. for an action that has already happened. Here we are clearly dealing with the Cause.  (2015)] 'It's so that sometimes I want to break into the driver's cab with a pistol and shout to him: -Drive to Yuzhnaya, you bastard, otherwise I'll make a sieve out of you!' This example captures Condition as a positive action that the THREATENEE must carry out. The THREATENEE here must drive the vehicle to Južnaja station. This action will help the THREATENEE not to face the Content of the threat. This type of conditional threats can be formulated as following: 'Do X, otherwise I will do Y'. Example (24) captures Condition as a negative action that the THREATENEE is potentially capable of doing but must avoid in order not to face the Content. As a model, it can be formulated as 'Do not do X, otherwise I will do Y'. This second option with negative evaluation resembles the Cause more than the first option where the Condition is positively evaluated. Although the three issues I have discussed suggest that it is helpful to keep Condition and Cause distinct, there are borderline cases where it is not straightforward to distinguish between the two components. One such case is when the undesired action is ongoing. The THREATENER'S intention then is twofold: to stop an action that (s)he evaluates as negative and to prevent it from happening again. Example (25) can refer to a situation where a THREATE-NEE has already tried to interfere in the kolxoz business in which case we are dealing with a Cause. However, another possible interpretation is that the THREATENER suspects that the THREATENEE can potentially interfere in the future. Under this interpretation we are dealing with a Condition rather than a Cause. For the purposes of classification in the present study, I apply the following rules. If we are dealing with a potential event in the future, I classify this as a Condition. If there is a clear indication that the relevant action is ongoing, I treat this as a Cause. If there is a clear indication that the action happened before and should not be repeated in the future, I classify these examples as both a Cause (referring to what happened before) and a Condition (what should not be repeated in the future), as in constructions (20) and (21) above.

Cause or Condition alone
Finally, 7 constructions from the semantic type Threat overtly only express either the Cause of the threat or the Condition, whereas the actual threat appears as an implicature. Consider example (26): (26) po-Verb-Imp eščë (mne/u menja)! 25 Pogovori mne eščë! DELIM-talk-IMP.SG I.DAT also 'Don't you dare talk like that (in my sphere of influence)!' The verb denotes an undesired activity that is currently performed by the interlocutor. The personal pronoun mne/u menja encodes the speaker, while the presence of the interlocutor is implied from the imperative verb form. The adverb eščë has an emphatic function in this construction. It might also be interpreted in its quantitative meaning as 'talking more/again', similar to (23). While there is no Content of threat expressed overtly, the construction still functions as a construction of threat, as well as a prohibitive construction. The situation is similar in example (27) However, unlike (26), the action here, i.e. not coming in time, can only be interpreted as hypothetical. The speaker suspects that the interlocutor can be late and warns him or her about it. The verb in the imperative form thus indicates the Condition, while the Content is neither present as a promise nor as an action. The construction has nevertheless a flavor of threat that comes as an implicature from the speaker's warning.
In this section, I have shown the 6 combinations of meanings that are attested in Russian constructions for performing threats. It turns out that the speakers' main strategy for threats is to be less than 100% explicit: threats in Russian are typically represented indirectly in constructions. I have identified several mechanisms for the speaker to be indirect. In most cases one or two components of a threat are left out. The most prominent component is the Content, which is present in most threat constructions. In some constructions, it can be enough to overtly express only a Cause or a Condition. The Cause and the Condition are closely related components, and it can be hard to distinguish between them in some contexts. The components that are not overtly expressed in the construction are often involved through metonymy. All this indicates that the THREATENER prefers to be vague and hint about the outcome that the THREATENEE will face. An alternative strategy to being vague is an exaggerated threat: the THREATENER can hyperbolize the outcome, so that the threat becomes more solid and intense. 6 The neighborhood of Threat: how are constructions related?
In the Russian Constructicon, a construction may carry several semantic tags. Alongside the semantic type Threat, the constructions under scrutiny are tagged with 14 other semantic tags to represent additional semantic properties. Exploration of the connections between the types allows us to investigate the relationships between constructions of Threat and other semantic types, such as Prohibitive, Condition, Apprehension, etc. This "neighborhood" of Threat can serve as a model for the relationships between constructions in a language. In the following, I propose that two relationships are important, which I refer to as "Overlap" and "Disambiguation in context".
Since a constructicon in Construction Grammar is perceived as a large network of constructions in a language, the "constructional relationships" (Diessel, 2019) in a language have received considerable attention in recent years. Researchers distinguish between vertical, or hierarchical, and horizontal, or lateral, relationships.
Vertical relationships are often described via a family metaphor, i.e., the parent-child relationship between two or more constructions (cf. Lyngfelt, 2018, p. 7). In a family of constructions, one may consider the "macro-construction", "meso-construction" and "microconstruction" levels (Traugott 2008a and2008b) that are connected by means of vertical relationships. The macro-constructions are perceived as rather abstract and general syntactic or morphosyntactic patterns, while meso-and micro-constructions are more specific. In Langacker's (2008, pp. 17-18) terminology, meso-constructions are "instantiations" of macro-constructions, since the meso-constructions are fully compatible, but more specific than macro-constructions. Micro-constructions are in turn more specific instantiations of meso-constructions.
While vertical relationships focus on structural or syntactic similarities between constructions, horizontal (also referred to as "lateral", "paradigmatic" or "alternation/allostructional", cf. Goldberg, 2002;Cappelle, 2006) relationships arguably focus on semantic similarities "at approximately the same level of abstraction" (Endresen et al., forthcoming). This means that constructions share a common, nearly identical meaning but may vary structurally. Alternation between constructions has recently become a matter of particular interest in Construction Grammar, suggesting that horizontal relationships are as important as vertical ones (cf. Grafmiller et al., 2018;Nesset & Janda, 2023). Endresen et al. (forthcoming) propose a nu clear VS. extended family opposition for the vertical VS. horizontal relationships. While the nuclear family consists of parent(s) and child(ren), the extended family includes more distant relatives, such as cousins (reflecting constructions that have some syntactic resemblance) and in-laws (reflecting constructions that are "syntactically unrelated but semantically similar"). Figure 5 illustrates two types of relationships that I have identified for constructions of the semantic type Threat. The first type can be perceived as a horizontal relationship, while the second one is based on structural similarity on a higher level of abstraction. However, I argue that constructions of this type also share a common semantic schema that I will elaborate on in this section.
In Fig. 5, the gray ovals represent constructions with open slots on the meso-and macrolevel of abstraction, while white ovals represent more specific "child" constructions on the micro-level; these are items where the slots are filled with lexemes. The numbers in the ovals correspond to examples in the text. Solid lines symbolize inheritance links from a constructional or semantic schema to a construction on a more specific level. The semantic schema, i.e., the generalized meaning of a construction, is represented with a rounded rectangle. Dotted lines between constructions show the semantic similarity between constructions of the same level of abstraction.
I will refer to the two types as "Overlap" and "Disambiguation in context".

Overlap
Constructions of the first type are arguably more widespread in the Russian Constructicon. They possess two or more semantic tags since one realization of the construction can belong to two semantic types at the same time. For example, the speaker can combine the meanings of Threat and Causation (with its subtype Direct) in one construction: (28) a nu bystro VP-Imp! A nu bystro delaj domašnee zadanie! and well quickly do.IMP.SG home-ACC.N task-ACC.SG 'Well, do your homework quickly!' In (28), the threat is used as a tool to coerce the THEATENEE into a desired action for the THREATENER. The THREATENER encodes the desired action with the verb in the imperative form. The desired action thus sounds like a command that the speaker gives the interlocutor or a group of people in a strict, peremptory tone. No matter how one fills the slots of this construction, the resulting utterance will still reflect the semantics of both tags -Causation (stating encouragement to perform an action) and Threat (implying that the THREATENER can punish the THREATENEE for disobedience). The combination of the two (Cause and Content) components of threat often results in a prohibitive construction, such as (29) This sentence focuses on an undesired action (tak sebja vesti 'behave like that'). On the one hand, the sentence is interpreted as a threat since the speaker tries to scare the interlocutor into changing their behavior. On the other hand, this example is also understood as a prohibition, whereby the speaker exerts their power to eliminate the unwanted behavior. The prevention of an undesired action is the primary objective of threatening constructions of this kind. Thus, (29) can express both a threat and a prohibition, and therefore it bears two semantic tags: Threat and Prohibition with its subtype Prohibitive. A slightly different type of overlap can be observed in (30).
(30) ja PronPers-Dat VP-Fut! 28 Ja tebe pooru! I.NOM you.DAT DELIM-shout-FUT.1SG 'Don't you dare shout!' In total this construction has three semantic tags to reflect its complex semantics. Similar to (29), this construction bears the semantic tags Threat and Prohibition (subtype Prohibitive). However, it also has the tag Non-standard subject with its subtype Non-standard subject marking. This tag can be attributed to the structural properties of the construction: while typically the syntactic subject of a sentence is also the agent of an action, ja 'I' is actually not the person who is performing an action, i.e. shouting. This mismatch between the lexical and morphological encoding within one construction was discussed in Sect. 4.
Structurally, constructions (28)-(30) do not have a shared parent construction on a higher level of abstraction. 29 What they share is the meaning of Threat that is involved in all three constructions and is accompanied with additional meanings: Prohibition, Causation, and Non-standard subject marking.
Overall, it is possible to postulate the similarity of constructions (28)-(30) if one compares them directly (cf. Langacker, 1987). In Langacker's terms, one item is a "prototype", and another is an "extension". The prototype and the extension possess some common features that can be observed on the same level. These features, however, can be extracted to create a more abstract "schema". For instance, in the category of fruit, an apple is a prototype and a tomato is an extension. Both belong to the "fruit" schema. Similarly, for our constructions one has to extract common semantic features that contribute to a semantic schema, such as "Threat" or "Prohibition". These schemas are represented as semantic tags in the Russian Constructicon (cf. semantic classification in Endresen et al., forthcoming).

Disambiguation in context
The second type of relationship shown in Fig. 5 is less common and is associated with a set of constructions that are united by vertical relations. This type contains constructions with a broad and rather schematic syntactic structure. Their structure is so general and abstract that when filled with specific lexemes it greatly varies in meaning and function. Interestingly, the meanings can even be antonymous as in Threat and Apprehension. Compare constructions (31) (2000)] 'And don't forget to put a scarf on, or you'll catch cold.' [Apprehension] Example (31) can only be interpreted as a Threat, since a (ne) to polučišʹ/polučite is a fixed ("anchor") part of the construction, which can only be used if the speaker is dissatisfied with the actions of the interlocutor and aims to punish them for these actions, i.e the typical threat situation. However, (32) has two open slots that can be filled with various Content depending on the intention of the speaker. While example (32a) clearly expresses a threat since the speaker intends to harm the interlocutor, example (32b) shows quite the opposite: instead of aiming to harm a person, the speaker shows their care towards another person and warns them not to get sick. The resulting construction rather belongs to the semantic type Apprehension. Therefore, constructions (32a) and (32b) share a common syntactic structure: two clauses connected with the conjunction a (ne) to, but they also share a common semantic meaning, which can be formulated as "a negative consequence for the interlocutor". This meaning is further specified in exact contexts depending on the fillers of the slots. The shared semantic meaning and syntactic structure can be considered a "constructional" schema, as opposed to "semantic" schema for the constructions of the first type in Fig. 5 that lack a shared structure.
The examples above show that the notion of threat is closely connected to the notion of apprehension and warning. While the threat focuses on the Content of threat, i.e. a negatively evaluated action imposed by the speaker on the addressee, the apprehension emphasizes the consequence of the addressee's action. The consequence is also a negatively evaluated action imposed on the addressee, but this action happens independently of the speaker's will or control.
To summarize, both Overlap and Disambiguation in context types of relationship are related in terms of the same cognitive mechanisms that they employ. In Langacker's terms, those mechanisms are "extension" and "instantiation" relationships. The speaker makes a direct comparison between two or more entities and extracts a shared part, which is usually more abstract.
The extracted meanings in both types serve as a "semantic schema" that is shared by two or more constructions. Combined with a syntactic schema in Disambiguation, it becomes a "constructional schema", while in Overlap there is no shared syntactic structure.
Overlap in two or more meanings usually involves constructions within the same level of abstraction, in our case, the meso-level. Disambiguation is an example of constructional polysemy that is only resolved in a specific context. Therefore, disambiguation exists across two different levels, namely, when the speaker makes an abstract construction more specific.

Concluding remarks
In this paper I have analyzed 27 constructions in Russian that are used to express threats. On the basis of the Russian Constructicon, a database of more than 2200 constructions, I have investigated what constitutes a threat in Russian and how threats are related to other constructions.
My contribution can be summarized as follows. First, we have seen that speakers of Russian use completely different linguistic means for describing threats as opposed to those used to perform them. The verbs ugrožatʹ and grozitʹ 'threaten', which are commonly used for describing threats, normally do not function as performatives. Second, it has been shown that there is no specific syntactic or lexical pattern for expressing threats: constructions of threat come in many shapes and are usually indirect or underspecified. Third, I have argued that a threat is a complex relationship between two participants, which I refer to as the THREATENER and the THREATENEE. The participants are in a power imbalance, and are usually represented as personal pronouns, which are, however, frequently omitted. Fourth, I have suggested that semantically this relationship consists of three components: Cause (the undesired action of the THREATENEE that the THREATENER does not approve and wants to stop), Condition (the desired action of the THREATENEE that the THREATENER suggests doing instead), and Content (what the THREATENER promises to do to the THREATENEE). Sixth, it has been demonstrated that although all three components can be overtly expressed, usually at least one of them is omitted. In most cases, the Content is present, but in some cases the Cause or the Condition alone suffices to convey a threat. Seventh, the neighborhood of Threat has been shown to be extensive and diverse: constructions of Threat are related to 14 other semantic types that belong to a number of different semantic classes of constructions.
In addition to shedding new light on constructions in Russian my article has theoretical implications for Construction Grammar. I have argued that lateral relationships between constructions can be of two types, which I refer to as "Overlap" and "Disambiguation in context". The former involves a semantic schema, while the latter is about a constructional schema. Although these two types are different, both can be described in terms of Langacker's extension and instantiation relationships.
Last but not least, my study testifies to the value of the Russian Constructicon as a basis for linguistic analysis. The rich classification in the Constructicon facilitates detailed analysis of semantic types of constructions, such as threats. At the same time, the thorough analysis I propose provides new insights that can be incorporated into the Russian Constructicon. Similar studies of other semantic types of constructions will no doubt be beneficial for the Russian Constructicon and provide new insights about the Russian language. It would also be interesting to compare the situation in Russian with threatening strategies in other languages, ultimately targeting a typology of threat constructions. However, such studies are beyond the scope of the present article.
Funding Open access funding provided by UiT The Arctic University of Norway (incl University Hospital of North Norway). This work was supported by the Research Council of Norway (https://www.forskningsradet. no/, grant number 300002).

Competing Interests
The authors have no relevant financial or non-financial interests to disclose.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.