1. Introduction: emotions, conceptual metaphors and frames

When we talk about emotions, we often draw on conceptual metaphors (Kövecses, Reference Kövecses2000): The abstract domain of emotion is understood in terms of another (typically concrete) domain of experience (Lakoff & Johnson, Reference Lakoff and Johnson1980). For instance, in He’s filled with anger, the emotion is understood as a liquid. Although there has been extensive research with regard to emotion metaphors over the last decades (e.g., Crawford, Reference Crawford2009; Fainsilber & Ortony, Reference Fainsilber and Ortony1987; Kövecses, Reference Kövecses2000; Mon et al., Reference Mon, Nencheva, Citron, Lew-Williams and Goldberg2021), as of today, it is not clear how they can be described systematically as part of our linguistic and conceptual knowledge.

Even though in psychology emotions are usually defined as physical and mental states evoked by various stimuli, including thoughts, sensory inputs or other (physical) experiences (Ekman & Davidson, Reference Ekman and Davidson1994, Reference Ekman and Davidsonp. 291), there is no generally accepted definition that applies across disciplines, such as psychology, neuroscience, anthropology, philosophy and sociology. Although – or precisely because – emotions are highly individual, they can be shared intersubjectively, not least through language. How emotions are conceptualized, linguistically coded and communicated has long been the subject of linguistic research (Lindquist, Reference Lindquist2021).

In this paper, we follow up on this tradition by adopting a frame-semantic approach to metaphor. Sullivan (Reference Sullivan2013) shows that both the source and the target domain of metaphors can be conceived as structured by frames. We extend this approach by combining conceptual metaphor theory, frame semantics and, in methodological terms, ‘constructicography’ (Lyngfelt et al., Reference Lyngfelt, Borin, Ohara and Torrent2018). Specifically, we are drawing on well-documented frames in the German FrameNet-ConstructiconFootnote ¹ (Ziem et al., Reference Ziem, Flick and Sandkühler2019), a resource that documents form-meaning structures of contemporary German on the lexicon-grammar continuum. In doing so, our approach differs from classical analyses of conceptual metaphors that are mostly working with domains that have been postulated in an ad hoc fashion (e.g., Gibbs, Reference Gibbs2008; Kövecses, Reference Kövecses1986, Reference Kövecses2005; Musolff, Reference Musolff2021). Thus, on the one hand, the aim of the present paper is to show potential avenues for a more coherent and empirically motivated approach to metaphor analysis. On the other hand, we present a case study to showcase how such an approach can be applied to the specific domain of emotion metaphors.

While the present paper is more ‘corpus-illustrated’ (Stefanowitsch, Reference Stefanowitsch2020) or ‘corpus-informed’ rather than corpus-based or corpus-driven, the lexicographic description of conceptual metaphor frames serves as a starting point for linguistic annotations of concrete instances, that is, metaphors-in-use, which in turn often feeds back into lexicographic analyses of conceptual metaphors yielding compilations for conceptual metaphor entries in FrameNet (or similar lexicographic resources).

In frame semantics as established by Fillmore (Reference Fillmore, Herbst, Schmid and Faulhaber2014), frames are conceived of as conceptual structures that motivate the understanding and use of language. Each frame is structured by a set of frame-specific semantic ‘slots’, so-called frame elements (FEs). FEs are classified ‘in terms of how central they are to a particular frame, distinguishing three levels: core, peripheral, and extra-thematic’ (Ruppenhofer et al., Reference Ruppenhofer, Ellsworth, Petruck, Johnson, Baker and Scheffczyk2016, Reference Ruppenhofer, Ellsworth, Petruck, Johnson, Baker and Scheffczykp. 23). Core FEs instantiate ‘a conceptually necessary component of a frame, while making the frame unique and different from other frames’ (Ruppenhofer et al., Reference Ruppenhofer, Ellsworth, Petruck, Johnson, Baker and Scheffczyk2016, Reference Ruppenhofer, Ellsworth, Petruck, Johnson, Baker and Scheffczykp. 23).Footnote ² To capture the meaning of metaphorical expressions, we argue that FEs of the source domain map onto FEs of the target domain and that this process gives rise to a frame in its own right: the conceptual metaphor frame. However, as shown in Figure 1 the conceptual metaphor frame does not only comprise mapped FEs but also FEs that are projected directly from the source domain and the target domain, respectively, onto the metaphor domain. Against the background of the ongoing debate in conceptual metaphor research on the presumable difference between frames and domains (e.g., Kövecses, Reference Kövecses2020, pp. 50–92), frames as conceived by Fillmore and in the FrameNet literature do not only include lexical frames, that is, frames that are evoked by lexical units. While FrameNet started out as a lexicographic resource that puts lexical units, and in particular verbs, center stage, the very definition of frames entails that they are not necessarily tied to (specific) lexical units, which is why current frame-semantic approaches assume and operationalize more abstract frame types. These include so-called non-lexical frames, e.g., scenario frames, as well as maximally abstract image-schematic frames (for an overview, cf. Ziem, Reference Ziem2014, pp. 5–48). Thus, we use the notion of frame in a broader sense than studies on conceptual metaphors usually do. As for the notion of domain, we follow Langacker’s (Reference Langacker1987, p. 488) definition of domain as a ‘coherent area of conceptualization relative to which semantic units may be characterized’ (see Dancygier & Sweetser, Reference Dancygier and Sweetser2014, Reference Dancygier and Sweetserp. 17, for a similar definition; also see Rosca, Reference Rosca2013). We refrain from equating domains with frames, as has sometimes been suggested in the literature (see Sullivan, Reference Sullivan and Dancygier2017, p. 402, for an overview), and instead argue that source and target domains are structured, to a large extent but not exclusively, by frames (in line with Sullivan, Reference Sullivan2013, p. 36).

Figure 1. A conceptual metaphor frame and its mapping processes.

In the case of conventional metaphors (i.e., extremely common metaphors in everyday language), which will be the main focus of this paper, the mapping process of FEs is not to be understood as taking place every time metaphorical language is processed. Rather, it is assumed that the conceptual metaphor frame is evoked directly. Following the standard assumptions of conceptual metaphor theory, we conceive of metaphor as a product that is a part of the conceptual system in long-term memory. By contrast, as Kövecses (Reference Kövecses2020) shows, novel (or ad hoc) metaphors can be analyzed more aptly drawing on mental spaces and blending theory (Fauconnier & Turner, Reference Fauconnier and Turner1998, Reference Fauconnier and Turner2002), which focus on online processes of meaning construction. Conceptual metaphors are not ad hoc mappings from a source to a target domain but are cognitively entrenched. But to understand the metaphorical expression, recipients still need an understanding of source and target frame. As the mapping between source and target is incomplete and selective (cf. Figure 1), they also have to know which FEs are mapped and which are not.

Frames can be particularly useful in accounting for emotion metaphors. From a psychological point of view, emotions are internal states that are (to a certain extent) assumed to be intersubjectively shared and thus similar across instances. As they are mental states, their linguistic conceptualization strongly depends on metaphors. The metaphors used must in turn be anchored in the shared world of experience in order to make emotions accessible. Thus, for three reasons we assume that frames are particularly useful for capturing the relationship between source and target domains as well as the metaphor meaning itself (resulting from multiple mappings between source and target domain): First, frames provide a structure identified by FEs and relations to other frames, which enables fine-grained descriptions; second, frames reflect the entire context of experience – a system of concepts – to which both the source and target domains relate (frames thus go far beyond word meanings); and third, meanings of conceptual metaphors can be identified as conceptual structures in their own right: as fully-fledged frames that differ from other (especially lexical) frames.

In the present study, we address the following questions: How can emotion metaphors be described in frame-semantic terms to account for both the source and target domain in terms of frames? Which aspects of source domains play a constitutive role in the conceptualization of emotions, and how can they be described using the analytic toolkit of frame semantics? More specifically, which FEs occur in emotion frames, and how do they interact with each other? And finally, to what extent is a frame-semantic approach able to cover the complexity of emotion concepts, as well as of metaphoric concepts in general, e.g., in terms of their potential hierarchically motivated structures? At the same time, our paper aims at providing a first proof-of-concept for developing a frame-based repository of conceptual metaphors (CoMetNet) within the German FrameNet-Constructicon project. This endeavor follows the lead of the MetaNet project (Dodge et al., Reference Dodge, Hong and Stickles2015; Petruck, Reference Petruck2016). But while MetaNet’s work was largely independent of the Berkeley FrameNet (Ruppenhofer et al., Reference Ruppenhofer, Ellsworth, Petruck, Johnson, Baker and Scheffczyk2016), and source and target frames were mostly created ad hoc, CoMetNet is both conceptually and technically fully integrated into the German FrameNet-Constructicon. We consider this an important advantage and will discuss some challenges that this approach entails in more detail below.

The remainder of this paper is structured as follows: In Section 2, we introduce the small corpus our case study draws on. Section 3 discusses the analytic approach taken to link source and target FEs both on an abstract and conceptual level and with regard to its application to concrete corpus examples. Section 4 zooms in on one particularly salient emotion metaphor, namely, anger is hot fluid in a container,Footnote ³ providing a detailed discussion of how it can be analyzed in frame-semantic terms. Section 5 focuses on a potentially problematic case with (seemingly?) conflicting target domains and looks at ways to resolve this problem. In Section 6, the most important implications of our analysis are summarized.

2. Corpus compilation and method

As our goal is not a large-scale analysis of metaphoric mappings in emotion concepts but rather a proof-of-concept that a frame-based approach to the analysis of emotion metaphors is both feasible and insightful, we draw on a small dataset compiled specifically for this study. More precisely, we compiled a custom corpus (7,147 words) drawing on articles from German popular science magazines (three from Spektrum, two from Psychologie Heute and two from Emotion) dealing with emotions. Articles were obtained by searching for keywords that refer to some of the most basic emotionsFootnote ⁴ (e.g., Hass (‘hate’), Wut (‘anger’), Liebe (‘love’), Stolz (‘pride’), Angst (‘fear’), and Freude (‘joy’)) on each magazine’s website. In fact, the most frequent emotion words that actually occurred denote negative emotion concepts, including Angst, Ärger (‘anger’), Wut and Zorn (‘wrath’). But also generic terms including Emotionen (‘emotions’) and Gefühle (‘feelings’) are used to a considerable extent. Note, however, that these overt expressions constitute only a small part of instances of emotion concepts, namely, these instances in which the target domain of the metaphor is realized linguistically. As a matter of fact, there are many cases of metaphors where only the source domain is expressed, while the target domain has to be inferred from the context.

Using such a small dataset has the advantage that the full texts can be taken into account in the annotation process, which allows for taking the broader context into account in which metaphors occur. The data were coded by two of the authors. First, the articles were checked for instances of conceptual metaphors in general, broadly following the Metaphor Identification Procedure (Pragglejaz Group, 2007). Second, if such instances were detected, they were examined more closely as to whether they refer to emotion concepts, that is, whether their target domain can be regarded as being structured by an emotion frame. Third, metaphors that fulfill these requirements were taken into account for the subsequent analysis. Their source domains were then checked for fitting frames from the German FrameNet-Constructicon, that is, the coders checked if an existing frame plausibly contributes to the meaning of the expression at hand. In the event that none of the existing frames fit, there are (at least) two possibilities: either – on the basis of the frame-semantic tenet that linguistic units of varying degrees of schematicity and abstractness always evoke a frame – the frame is not evoked lexically (but, e.g., grammatically) or a new (fitting) lexical frame has to be compiled. In the latter case, the common procedure in CoMetNet is to draw on additional annotations of authentic data in order to examine carefully if it is useful and necessary to compile a new lexical frame to the German FrameNet-Constructicon repository.

Like other FrameNets that are under active development, the German FrameNet is a continuously evolving repository. As a result, if a frame required for a source or target domain is missing, this does not mean that the metaphor cannot be accounted for in frame-semantic terms. The fourth and final step relates to analyzing conceptual metaphor frames in terms of the mapping processes; in our dataset, more than 50 instances of conceptual metaphors in the domain of emotion were detected, some of which will be discussed in more detail in the following sections.

3. Conceptualizing emotions: mapping of FEs between source and target frame

This section introduces the approach taken here, offering examples of how specific emotion metaphors can be accounted for in frame-semantic terms. In particular, we address two groups of emotion metaphors: the first one comprising emotion metaphors with vision as source domain and the second one including emotion metaphors with the human body as source domain. We use – as is common in frame-semantic analyses – simple assertive sentences such as lexical expression X evokes frame Y, which consists of the frame elements Z. Importantly, these descriptions reflect the lexicographic process of creating and thus modeling frames in CoMetNet and FrameNet on the basis of annotated sample sentences, aiming at a coherent framework for constructicographic purposes. We do not claim that these descriptions necessarily describe cognitively real processes, although we aim for them to be cognitively plausible. Generally, compiling frames can be seen as a way to model linguistic knowledge on the basis of annotated data along the lines of the Berkeley FrameNet annotation policy (Ruppenhofer et al., Reference Ruppenhofer, Ellsworth, Petruck, Johnson, Baker and Scheffczyk2016). Additional data and additional annotations may, at least to some extent, affect the lexicographic process of frame compilation.

3.1. Group 1: Emotions and their visual (im-)perceptibility

Consider the following corpus examplesFootnote ⁵:

In (1) and (2), the expression verdeckte Angst (‘covert anxiety’) evokes the conceptual metaphor frame emotions are invisible objects.Footnote ⁶ It is constituted by the source frame Eclipse,Footnote ⁷ which in turn is evoked by the adjective verdeckt (‘covert’), and the target frame Feeling, evoked by the noun Angst (‘fear/anxiety’). Here we can already see what distinguishes frame-semantic from ‘classical’ analyses of conceptual metaphors. Usually, the latter consider (1) and (2) (and also the examples discussed in the remainder of this section and in Section 3.2) to be instances of the emotions are objects metaphor. While we assume that the more specific conceptual metaphor frame emotions are invisible objects indeed entails the generic emotions are objects metaphor (on metaphorical entailment, see Kövecses, Reference Kövecses2020, Reference Kövecsesp. 4), the focus in our example seems to be more on the occlusion of the emotion (here: anxiety), which is why the more specific frame Eclipse seems to be a better candidate to account for the source domain of this particular metaphor.

The source frame Eclipse integrates the core FEs eclipsed and obstruction, of which only the first is linguistically salient here: The core of the metaphorical conceptualization is a mapping from the source FE eclipsed to the target FE emotion. Note that eclipsed has to be read as a nominalization here, that is, it refers to the entity that is eclipsed from view. Thus, emotions – in our example, specifically anxiety – are understood in terms of eclipsed entities. As a consequence of this metaphorical shift, we could assume that emotions can also be unveiled. As (2) shows, this actually seems to be the case (see below for a more exhaustive discussion of this phenomenon).

However, the frame Eclipse and its lexical instantiations in (1) and (2) describe a static type of situation and a result of an action or a process, respectively, rather than the action or process itself. Consider the following data:

The metaphorical expressions Emotionen verstecken (‘hide emotions’), Frust verstecken (‘hide frustration’) and Gefühle verbergen (‘conceal feelings‘) are instantiations of the conceptual metaphor frame making emotions imperceptible is making physical objects invisible. Note that its source frame is not structured by the frame Eclipse but by Hiding_objects. In this frame, an agent causes a hidden_object to become perceptually inaccessible to potential perceivers. In contrast to Eclipse, there is an agent involved in Hiding_Objects. Also, Hiding_objects is related to Eclipse by a causal relation. We assume the following mapping between FEs from source to target frame: agent maps onto experiencer, and hidden_object onto emotion. In this metaphor, not only emotions are understood as physical, hidden objects, but also experiencers of emotions are understood as acting persons who intentionally hide their emotions construed as physical objects.

If emotions can be covert or hidden, an obvious conclusion would be that they can also be made perceptually accessible to potential viewers. Consider the following corpus examples:

In these examples, the verb zeigen (‘show’) is used to construe emotions like anxiety as physical entities that can be made visually perceivable. The source domain of the conceptual metaphor evoked in these sentences, making emotions perceptible is making physical objects visible, is structured by the frame Cause_to_perceive, in which an agent, actor,Footnote ⁸ entity or medium causes a phenomenon to be perceived by a perceiver. The following mappings between FEs are assumed to be constitutive of the conceptual metaphor frame: agent maps onto experiencer, and phenomenon maps onto emotion. However, there are many more verbs to construe such an understanding of emotions, like preisgeben (‘disclose’) and offenlegen (also ‘disclose’; lit. ‘lay open’):

At this point, our data relate to two types of linguistic expressions representing three source frames, that is, three conceptual metaphors. The first type either refers to a static situation (e.g., verdeckt ‘covert’) that is captured by the frame Eclipse or refers to an intentional act (e.g., verstecken ‘hide’ and verbergen ‘conceal’) which is captured by the frame Hiding_objects. Both subtypes are connected to a setting in which the emotion is construed as a non-visible entity. Corresponding metaphors are emotions are invisible objects and making emotions imperceptible is making physical objects invisible. The second type refers to the intentional act of making an object visible again (e.g., zeigen ‘show’, offenlegen ‘disclose’ and preisgeben ‘disclose’), which is captured by the frame Cause_to_perceive. The corresponding metaphor is making emotions perceptible is making physical objects visible. Thus, the metaphoric understanding and conception of emotions as physical entities draws on the cognitive modality of visual perception. However, this modality is not the only one serving as a source domain, as we show in the following section.

3.2. Group 2: Emotions and their physical (un-)controllability

To illustrate group 2, consider the following examples.

The metaphorical expressions Wut in den Griff bekommen ‘get a grip on anger’ and Wut in Schach halten ‘hold anger in check’ evoke the conceptual metaphor frame handling emotions is controlling physical objects, while the frame Being_in_control is used as a source domain: The experiencer of anger is construed as a controlling entity while anger is construed as a somewhat dependent, that is, controlled, entity. Emotions expressed by this conceptual metaphor frame mostly have a negative connotation, like anger above. This seems to be characteristic of this source domain: Even when used with positive emotions such as Freude ‘joy’, expressions like in den Griff bekommen and in Schach halten construe the emotion that is to be controlled, or at least its (public) expression, as undesirable.

Another frame serving as a source domain to construe emotions as physical, controllable objects is Dominate_competitor. This frame inherits from the frame Being_in_control, that is, all of its FEs correspond to equally specific or more specific FEs of the frame Being_in_control. (15) shows an instance of a conceptual metaphor that is structured by the source frame Dominate_competitor and the target frame Feeling:

In (15), anger is construed as a physical object (which can be human, of course) that is dominated by the ‘cool head’, which is construed as an agent. In this metaphor, the emotional state relates to the patient, and the entity that experiences the emotion to the agent that exercises control over the patient.Footnote ⁹ In this sentence, however, in contrast to the examples discussed so far, the realization of the FE experiencer_is_agent (i.e., the idiomatic expression kühler Kopf ‘cool head’) is itself metaphorical, unlike, e.g., people and patients in (13) and (14), respectively. Its meaning is motivated by the conceptual metaphor reason is cool head, whose source domain exploits the Body_parts frame and whose target domain exploits the Mental_property frame. The construal of the relation between emotion and reason in (15) lies in treating them as two separable and competing entities. Moreover, the well-attested conceptual metaphor emotion is heat suggests to understand reason and emotion as a binary opposition. Corpus examples evoking the emotion is heat metaphor are Hitzkopf (‘hothead’) and hitzköpfig (‘hot-headed‘), both relating conceptually to the exact opposite of der kühle Kopf: If you are emotional, you are hot; if you are emotionless, you are cool. Being or acting emotionless is equivalent to being reasonable (or acting reasonably).

(16) shows that the type of the entity that is controlled by the agent can be further specified:

The metaphorical expression ungebremster Zorn (‘unbraked wrath‘) draws on the frame Operate_vehicle as its source domain, evoking the conceptual metaphor frame handling emotions is controlling a vehicle: Wrath is understood as a vehicle. An important feature of this conceptual metaphor is that the experiencer of an emotion is construed as the driver of a vehicle. This core FE of the conceptual metaphor frame is not realized in (16), but likely to be conceptually salient.Footnote ¹⁰ Arguably, driving a vehicle is a form of exercising control. Hence, in this context, the frames Operate_vehicle and Being_in_control probably interact with each other (although there is no semantic relation between the two frames on the lexical level).

Another way of controlling an entity is affecting its movement abilities up to preventing it from moving in the way it would move if the controlling entity was not there. For illustration, consider (17) and (18).

The conceptual metaphor frames evoked by the expressions Ärger zügeln (‘rein in anger’) and unbändige Wutausbrüche (‘unbridled outbursts’) are – besides the target frame Feeling – structured by the source frame Immobilization. In this frame, an agent immobilizes a patient. This means that the emotion (anger) is construed as a patient while the emotional experiencer is construed as an agent. Thus, the conceptual metaphor frame is handling emotions is immobilizing physical objects. If we look more closely at the lexical units related to the source domain, however, another source frame comes into play, that is, Animals: In its literal sense, zügeln relates to the act of tightening the reins of a riding or draught animal in order to hold it back or halt it; bändigen, in contrast, relates to the act of fitting an animal with a lead or a bridle. Hence, the FE patient of the frame Immobilization maps to the frame Animals by means of the conceptual metaphor frame emotions are animals, because only if emotions are construed as animals, they can be conceived of as animals that can (or ought to) be immobilized. As a result, the FE animals maps to the FE patient and, more importantly, the conceptual metaphor frame handling emotions is immobilizing physical objects is specified to handling emotions is immobilizing animals. Figure 2 illustrates the mapping processes in this metaphor:

Figure 2. The structure of the conceptual metaphor frame HANDLING EMOTIONS IS IMMOBILIZING ANIMALS.

Overall, the analyses presented in this section suggest the following two conclusions. (1) At least four source frames may be used to conceptualize emotions in terms of their physical (un-)controllability: Being_in_control (in den Griff bekommen, in Schach halten), Dominate_competitor (die Oberhand gewinnen), Operate_vehicle (ungebremst) and Immobilization (zügeln, unbändig). Thus, conceptualizing emotions does not only draw on visual perception as its source domain (cf. Section 3.1) but also on motor control in the sense of directed movement: Most of the activities related to the source frames mentioned above are embodied in that they include the use of one’s hands and fingers. Thus, conceptualizing emotions as physical objects has broader implications: Before being able to control an entity, we have to (be able to) see it first. Hence, metaphors of group 2 (Emotions and their physical (un-)controllability) conceptually imply metaphors of group 1 (Emotions and their visual (im-)perceptibility). (2) Certain aspects (FEs) of some metaphors are themselves structured by metaphors (cf. reason is cool head and emotions are animals). To investigate such hierarchies of metaphors (or metaphor networks) in detail, the following section looks more closely at the conceptual metaphor anger is a hot fluid in a container and its underlying metaphor network.

4. The complexity of emotion metaphors: anger is a hot liquid and its network of metaphors

In the previous sections, we have already hinted at some challenges arising from modeling conceptual metaphors with frames. In this section, we provide a detailed discussion of some major issues of analyzing (families of) metaphors as frames, drawing on the example of the conceptual metaphor anger is a hot liquid (see, e.g., Sullivan, Reference Sullivan and Dancygier2017, Reference Sullivan and Dancygierp. 388). The first challenge relates to varying degrees of schematicity specific to the conceptual structures involved in a given metaphoric mapping. Metaphors may relate to both barely schematic and highly schematic structures, such as those provided by image-schematic frames. Consider (19).

In (19), Gemüter hochkochen (‘boil up moods/tempers’) instantiates the very productive conceptual metaphor anger is hot fluid in a container (Kövecses, Reference Kövecses2000). Now, what is the structure of this metaphor? In terms of the hierarchical dimension, the highest (and therefore most abstract) metaphor is emotions are objects, structured by the source frame Physical_Entity and the target frame Feeling (see Section 3). Moreover, in (19), the verb hochkochen relates to a specific kind of physical entities, namely, to liquids. In this example, it is thus the more specific conceptual metaphor emotions are liquids – a daughter of emotions are objects – which features a conceptually adequate level of schematicity.

Note that an important attribute of liquids relates to temperature. For example, boiling liquids yields a gradual rise in their temperature. This is why the metaphor emotion is heat is not only structured by the frames Temperature and Feeling but also by the image-schematic frame Scale. As has been pointed out elsewhere (e.g., Stefanowitsch, Reference Stefanowitsch, Stefanowitsch and Gries2006, Reference Stefanowitsch, Stefanowitsch and Griesp. 75), both metaphors emotions are liquids (or, more specifically, anger is a liquid) and anger is heat combined give rise to the conceptual metaphor anger is a hot liquid. To account for such relations between conceptual metaphor frames, we use hierarchical relationships, first and foremost inheritance, just like in Berkeley FrameNet.

In the CMT literature, we often find more specific descriptions of this metaphor, and the metaphor is usually called anger is a hot liquid in a container. It is said to interact with another fundamental (e.g., Kövecses, Reference Kövecses2000, Reference Kövecses2020) and well-attestedFootnote ¹¹ metaphor, namely, body is a container for emotions. In our frame-semantic approach, we address this relation in such a way that the FE Contents relates to the conceptual metaphor frame emotions are substances. In many corpus examples, liquids instantiate substances: Aside from hochkochen, other examples from our corpus data include eruptive Reaktion (‘eruptive reaction‘), which draws on the source domain lava, and überbordender Zorn (lit. ‘overboarding wrath‘) relating to the source domains floods or river.

See Table 1 for an overview of the frames structuring the conceptual metaphor anger is a hot liquid in a container:

Table 1. The complex frame-semantic architecture of Anger is a hot fluid in a container

This description is hierarchical in such a way that it moves from the generic metaphor emotions are objects to the specific metaphor anger is a hot fluid in a container line by line. It works for every instance of the metaphor anger is a hot fluid in a container. Some instances of this metaphor, for example, Gemüter hochkochen, may even be more complex. This metaphor interacts with the metaphor becoming angry is moving upwards, which is structured by several frames: First, Motion_directional, which is in turn structured by the frames Motion, Direction, and the image-schematic frame Source_path_goal.Footnote ¹² Second, the image-schematic frame Verticality because direction can be on a vertical (up versus down) or a horizontal (left versus right) scale. In our example, feelings understood as liquids move upward. Hence, the vertical scale matters. Third, the metaphor relates to Change_position_on_a_scale. This metaphor inherits from the image-schematic frame Scale that – as mentioned above – turned out to be relevant for the metaphor emotion is heat. Specifically, given our example, it is a salient feature of the underlying metaphor that the liquids get hot (i.e., they change their position on a scale of temperature). Thus, the focus is rather on a process than on a state; if liquids are boiled too long and get too hot, respectively, during this process, the container will flow over. In addition, an apt understanding and use of the dynamic metaphor becoming angry is moving upwards presupposes knowledge of the static metaphor angry is up: Only if feeling anger is conceptualized as being located at a high point on a vertical scale, becoming angry can be understood as moving toward this point.

5. angry is up and happy is up: the polyvalence of scalar source domains

As (20) indicates, some examples attested in the corpus suggest that metaphors that express an undesirable emotional state are located in an upper region of an imaginary vertical space axis. This seems to be in conflict with the happy is up metaphor (i.e., one of the most basic emotion metaphors) that also uses up as a source domain while at the same time its target domain happiness refers to a desirable emotional state.

In the context of (20), the expression an die Decke gehen (‘hit the roof’; lit. ‘go towards the ceiling’) evokes the conceptual metaphor frame becoming angry is moving upwards that is structured by the source frame Motion_directional and the target frame Anger (which is not overtly realized here, cf. Section 2). Other instantiations of this metaphor attested in our corpus include sich aufregen (‘be upset’) and its counterpart sich abregen (‘come down’; lit. ‘be downset’). Here, the particles of these particle verbs refer to a movement in a direction: up and down, respectively.

In much of the CMT literature, the expression hit the roof is considered an instance of the metaphor anger is heat (e.g., Lakoff & Kövecses, Reference Lakoff and Kövecses1983) and anger is a hot fluid in a container, respectively, as discussed in Section 4. However, unlike in the case of boil, the source domain heat or hot liquid is not represented at all in hit the roof. Of course, as far as the interpretation of this metaphoric expression is concerned, it is quite possible to take into account conceptual knowledge such as ‘When the substance in a container becomes too hot, the container explodes’ and ‘When a container explodes, parts of it go up in the air’. Nonetheless, we argue for an alternative analysis that emphasizes the motion of the experiencer as a source domain; following our analysis, this domain is evoked directly by the expression hit the roof.

Reference to directional motion applies to (20) in such a way that the ceiling (which is the literal English equivalent to German Decke) in a room is located above the experiencer, that is, to move toward the ceiling means to move upward. Since being or becoming angry (or being or becoming upset) is usually considered a socially undesirable emotional behavior associated with, e.g., loss of control and irrationality, while being happy is arguably an emotional state evaluated as desirable, angry is up at first glance seems to be in conflict with happy is up. This issue has recently been addressed in a cognitive-linguistic and anthropological study by Wnuk and Ito (Reference Wnuk and Ito2021). In Mlabri, an Austroasiatic language, there are spatial metaphors mapping down onto desirable, and up onto undesirable emotional experiences. Wnuk and Ito (Reference Wnuk and Ito2021) are pointing out that this mapping can be explained in terms of arousal: In Mlabri, low-arousal states reflect an ideal affect ‘centered on contentment and tranquility’ (Wnuk & Ito, Reference Wnuk and Ito2021, Reference Wnuk and Itop. 213). This seems to apply to German (and English) as well: Reconsidering the expression an die Decke gehen and situations in which people get angry, verbs like runterkommen (‘come down’) relate to changing an agitated emotional state to a neutral state. In terms of an die Decke gehen, this means that if someone has to ‘come down’ they have to move away from the ceiling and back to the ground. Thus, the ground represents the neutral state that is, in the context of anger, a desirable state, while down maps on a state evaluated as positive.

Note that both metaphors, angry is up and happy is up, build on the primary metaphor more is up, which suggests that these metaphors are not in a general conflict. Primary metaphors are described by Grady (Reference Grady1997, Reference Gradyp. 47) as ‘metaphors which have a direct experiential basis, and which motivate highly predictable sets of data’. For our examples, this means that conceptually it makes no difference if up is mapped to a positive (happiness) or negative (anger) emotion concept as long as there is an emotion and the experiencer, in metaphorical terms, is located above sea level (sea level meaning current lack of emotions).

Importantly, both happy is up and sad is down constitute an opposition pair; in contrast, however, angry is up does not have a counterpart. As a result, on the imaginary scale evoked by happy is up, the distance between ‘positive’ and ‘negative’ differs from the distance between ‘positive’ and ‘negative’ on the scale evoked by angry is up (see Figure 3).

Figure 3. The difference between HAPPY IS UP and ANGRY IS UP in terms of their scalar source domain.

Following this analysis, the metaphor being down relating to sadness means being underground. Here, it is the ground that serves as a reference point. In contrast, the metaphor down relating to ‘having moved away from being angry’ means being on ground and thus takes a spot above the ground as its reference point. Note that in the first case a state rather than a process is described, while in the latter case a dynamic process is described rather than a state.