There can be numerous reasons for change in scientific theories. For example, the elaboration or introduction of new methods may provide new data, but it also makes possible the re-evaluation of previous research results, which may have theoretical consequences and, in the end, the emergence of a new model in the field of research. Yet it is also possible that certain problems lead to a theoretical transformation of certain parts of the dominant theoretical paradigm. The book under review fits into the latter line, when with the past forty years of cognitivist metaphor research in the backpack, the author, Zoltán Kövecses, proposes a partial revision of the previous theoretical model, the standard Conceptual Metaphor Theory.

Kövecses’ work gives not only a broad and in-depth overview of the developments of contemporary research in the cognitive linguistic metaphor research paradigm in a comprehensible manner. It also develops on the basis of the standard version of the Conceptual Metaphor Theory originated from Lakoff and Johnson's seminal work Metaphors We Live by (Lakoff & Johnson 1980 – henceforth CMT), an innovative and challenging novel framework, the Extended Conceptual Metaphor Theory (henceforth ECMT). The book consists of 8 chapters in total, and it is framed by an introductory (Chapter 1) and a summary chapter (Chapter 8). The presentation of CMT in Chapter 1 serves as a stepping stone to the main topic of the book: to the developing the Extended Conceptual Metaphor Theory in the next chapters (Chapter 2–6). On the basis of the ideas elaborated in these chapters, in Chapter 7, Kövecses provides a brief summary of the general framework of ECMT. The concluding chapter (Chapter 8) compares ECMT with Ray Gibbs' theory of the dynamic systems view of metaphors (Gibbs 2017).

In the introductory chapter, Kövecses presents not only a concise summary of the standard version of CMT, but the basic questions that motivate the need to extend the classical model are briefly summarized as well. For those who are familiar with the work of Zoltán Kövecses, it is also clear that the answers to the questions raised in the Chapter 2 to 6 are the results of a longer research period of about ten years. These issues and related major publications include:

• The question of literal and figurative meaning, i.e., the problem of the relationship between figurative and non-figurative language in CMT (Kövecses 2017a).

• The question of priority, i.e., the problem of the metonymical foundations of primary metaphors (Kövecses 2013).

• The hierarchy problem of metaphors, i.e., the topic of the levels of hierarchical schematicity (Kövecses 2017b).

• The role of contexts in the production and understanding of metaphors (Kövecses 2015).

• The ontological status of metaphorical meaning, i.e., the question of the stability of offline metaphorical meaning and the relation between offline and online metaphorical meaning (Kövecses 2016).

The first chapter in medias res starts drilling deep and goes right to the foundations in order to identify the basic problems. For this, Kövecses gives a very clear and focused overview of the classic version of the conceptual metaphor theory. He defines the notion of conceptual metaphor, elaborates on conceptual structures that play a role in metaphorical mapping such as the domains and provides a broad range of examples to illustrate the pervasiveness and the thematic variability of metaphors as well as the role of metaphorical inference in the construction of realities. However, the aim of the first chapter is not merely to introduce notions and concepts or a certain enumeration or appraisal of the findings of the cognitive research of metaphors. The genuine objective of the first chapter is to expose the questions above for further investigations, which will be the basis of the developing of ECMT in the later chapters of the volume.

Chapter 2 examines the relationship between figurative and non-figurative language from the perspective of figurativeness. Following the research tradition of CMT, Kövecses does not grasp the gist of figurative language from the definition of literal meaning but the other way round. He starts from the theorem of Lakoff (1993, 205) that proposes “those concepts that are not comprehended via conceptual metaphor might be called literal” (p. 24), but he also extends this to the (language) historical level. By analyzing a number of examples, he points out, on the one hand, the fact that several current literal meanings today have figurative roots in the linguistic historical sense: “a large number of expressions (that belong to abstract domains) are taken to be literal, but in synchrony their corresponding concepts are metaphorically comprehended and from a historical perspective they have metaphorical or metonymic origins” (p. 25). On the other hand though, Kövecses moves much further beyond this well-known statement and he formulates the “radical view of the extent of the literal” (p. 23). Thus, he not only claims that abstract notions from a synchronic perspective have metaphorical or metonymical roots but also that a great many expressions with a concrete meaning today are also metaphorical historically, (e.g., book), or developed orignally as a result of metonymical projection (e.g., storm). By now, however, since nothing refers morphologically to projection, or to the source domain, in these expressions, we cannot sense the figurative bases of the literal language.

At the same time, Kövecses does not only argue in this chapter for extending the arguments of CMT, but he also emphasizes the continuity between CMT and ECMT by defending the principle of unidirectionality, according to which metaphorical projection moves from the concrete source domain to the abstract target domain. In order to save unidirectionality, he also formulates an assumption concerning the general structure of concepts, which proposes that “much of what we take to be literal is only ontologically literal (i.e., constitutes basic concrete experience), but as regards its cognitive status it is also figuratively construed/constructed experience” (p. 28 – emphasis in the original). This assumption, as Benczes (2021, 267) also points out in her review of the book, “accounts for one of the perplexing aspects of CMT: why source domains can sometimes function as a target domain.”

With these steps, Kövecses extends the principle of “via conceptual metaphor” concerning the figurative meaning creation of conceptual metaphor theory to the level of synchronic literal meanings, too, “which greatly reduces or even eliminates the literal – at least in a certain sense” (p. 23). The example analysed by Kövecses is smell, an “ideal source domain” (p. 29), and the analysis is conducted on the basis of three dimensional concepts, existence, intensity and lack of control, through which he points out that these generic concepts, which can be applied to a multitude of other concepts, too, elevate the domain of smell into the abstract dimension at the same time. With the help of the dimensional concepts, smell can already appear as an abstract domain and, as such, it can become the objective of metaphorical mapping. In Kövecses’ interpretation, smell is viewed as a substance, which is located in the air as a container. Concepts of air and smell share the same dimension of existence; therefore, they can be connected by the interpretation of air as a location, i.e., container and smell as an object, a substance, which can be located in a container. Kövecses argues for the assumption of the following conceptual metaphors in the case of smell in the construction of filled with + smell:

SMELL IS A SUBSTANCE

THE OBJECT/LOCATION THAT HAS THE SMELL IS A CONTAINER

THE EXISTENCE OF SMELL IS FOR THE SUBSTANCE TO BE IN THE CONTAINER

His analysis, which is based on a former study (Kövecses 2019), provides a striking answer to the classical problem of how it is possible that, in light of the unidirectionality thesis, even concrete experiential domains such as smell can be not only a source domain but also a target domain in the metaphorical mapping.

Nevertheless, one point in the analysis at least allows for an alternative interpretation in the sentence used for the analysis of the dimension of existence, The air was filled with a pervasive smell of chemicals (p. 30). According to Kövecses's analysis, the conceptual metaphor smell is substance structures the domain of smell here. Yet, smell in this sentence can also be interpreted metonymically, too, the same way as Radden and Kövecses understand metonymy in their classic study (Radden & Kövecses 2007, 336): “Metonymy does […] not simply substitute one entity for another entity, but interrelates them to form a new, complex meaning.” Therefore, we can understand smell in The air was filled with a pervasive smell of chemicals not simply as the presence of a substance in the air (i.e., an object in a container), but as the effect of inhaling air (that is, experiencing the substance). Smell is only there if there is somebody who is experiencing the presence of the substance. In other words, smell is an effect of the sensation of a substance. That is to say, the interpretation of smell can be motivated by a general metonymy: the effect for the source of the effect, or generally, effect for cause. In other words, the metonymic entity, here filled with smell, is the reference point that affords mental access to an experience, which is a sensation of a special kind. And this special effect of experiencing the substance is realized linguistically. This sketchy analysis is also in line with the “domain matrix” of smell, when sensation also lacks control and can have weak or strong intensity. The metonymic interpretation of smell in The air was filled with a pervasive smell of chemicals is also compatible with the idea of Kövecses and many other cognitive linguists, which claims that at least some kinds of metaphors have metonymic roots (cf. Barcelona 2011; Tóth 2021).

The relation of metonymy and metaphor is the topic of Chapter 3 titled Direct or Indirect Emergence? Kövecses argues that “correlation-based metaphors emerge from frame-like mental representations” (p. 35) in two steps. The first, metonymical step concerns a specific element of a frame that affords access to the whole frame and, in the second, metaphorical step, “the element inside the frame involves or is characterized by a particular frame-specific concept, which is, then, generalized, or schematized, into a concept that exists outside the frame […]” (p. 39). This element outside the frame then serves as a metaphorical target or source. Kövecses demonstrates this two-stage process with several example analyses (control is seeing, purposes are destinations).

In Chapter 4, Kövecses’ starting point is the diversity of theoretical concepts which authors in cognitive linguistics works use in describing and analysing the conceptual structures that constitute conceptual metaphors. Domains, schemas, frames or spaces, this “terminological confusion” (p. 50) is a sign in Kövecses’ interpretation of a “serious, deep-seated theoretical-conceptual dilemma” (p. 50) to identify the most appropriate building block in the structure of conceptual metaphors. In line with his main goals in the book, Kövecses allocates these terms to different levels of schematization in order to show the capability of an extended CMT in grasping the complexity of conceptual metaphors. He distinguishes four levels of schematicity: “the level of image schemas, the level of domains, the level of frames, and the level of mental spaces” (p. 51). For demonstrating the effectivity of his multilevel view, he analyses the well-known source domain of building, but also deliberate metaphors (in the mountain stage), metaphorical idioms (cash cow), and visual metaphors (a painting by Mark Tansey, entitled Picasso and Braque). It is intuitively a striking idea that using metaphors activates complex structures of knowledge. Therefore, the analyses should also be able to tackle different aspects of this complexity. Kövecses’ multilevel approach shows convincingly how to grasp the different levels of metaphors as a mental and as a linguistic phenomenon, respectively.

The idea of handling the different aspects of conceptual structures on different levels has its precursors in the cognitive linguistic literature (e.g., Benczes 2011; Musolff 2016; Semino et al. 2018) but Kövecses’ approach is most probably the first one that also takes into account the methodological consequences of his multilevel approach. After having described the weaknesses and strengths of the methods, he comes to the conclusion that “most of the approaches can be used to study several different levels” (p. 87), which means that there is no single method which can claim a superior status in the research of conceptual metaphors. Instead, there are preferential methods in researching the different levels of schematicity: psycholinguistic methods and neuroscientific experimentation on the level of image schemas, intuitive and lexical approach, and the Framenet-type approach at the levels of frames and domains, corpus linguistic methods and discourse linguistic analysis, but also computational modelling as well as psycholinguistic and neuroscientific experimentation at the level of mental spaces.

In Chapter 5, Kövecses takes a further important step in elaborating the extended CMT by completing it with a contextual component. He argues for a “contextualist view of metaphor” (p. 116), and differentiates four relevant types of context that influence the use of metaphors in discourse: (i) the situational context, which includes the physical environment, the social situation, and the cultural situation as relevant factors; (ii) the discourse context, which includes the surrounding discourse, and the knowledge of the conceptualizers about the speaker, the hearer, and the topic; (iii) the conceptual-cognitive context which encompasses the metaphorical conceptual system, the ideology, knowledge about past events, and interests and concerns; and last (iv) the bodily context, which is the influence of bodily conditions. Additionally, Kövecses also discusses the main cognitive processes of metaphor producing in a given context. Despite his eloquent argumentation about the connection between the influential contextual factors and the schematicity hierarchy of conceptual metaphors, the interrelatedness of the four types of conceptual factors is not really clear. Can they be in conflict which each other? Is any of them prior in sense of production or understanding of metaphors? Nevertheless, it would be deeply unjustified to criticise Kövecses’ argumentation in this chapter because a comprehensive investigation of the conceptual factors that influence the production and use of metaphors would unquestionably exceed even the limits of a single book, and it should be a project all on its own.

A different aspect of contextuality forms the starting point in Chapter 6. How can conceptual metaphor theory be capable of accounting for metaphorical language in real discourse contexts, i.e., online? Kövecses provides an explanation for the online production of metaphorical meaning on the basis of his multilevel approach. He argues that online metaphorical meaning is constructed in mental spaces, the highest level in the hierarchy of schemata. He exemplifies the capability of the multilevel view by discussing two phenomena, mixing metaphors and conceptual integration. Considering the first, mixing metaphors, Kövecses emphasizes that these metaphors, which occur in a metaphorically not homogeneous discourse, are very common and causing “incompatibility in imagery” (p. 124) but without causing serious problems in understanding. He suggests that mixing metaphors can be handled effectively in the framework of a multilevel approach in assuming that in mixing metaphors “the various (near)-adjacent source domains are activated to a low degree” (p. 131) and this low-level of activation does not cause an “interference” between the source domains. Moreover, he assumes that, “the various degrees of conventionalization for the source domains and source expressions are correlated with differing degrees of neural activation. Lower levels of conventionality and, hence, activation do not produce consciously recognized incompatibility between source domains and their metaphorical expressions, while higher levels do” (p. 131). However, how to test the assumed correlation between the degree of conventionality and the degree of neural activity remains an open question. As for conceptual integration, he argues that CMT and the Conceptual Integration Theory (Fauconnier & Turner 2002, 2008) are compatible and, moreover, they are complementary within the framework of ECMT. Kövecses shows that mental spaces as the highest level of the production of metaphorical meaning are the level of the online conceptual integration but the integration is strongly anchored in offline conceptual structures and inferences, like domains and mappings.

ECMT takes on a definitive shape in Chapter 7. Kövecses extends the conceptual pathways for creating and comprehending conceptual metaphors in CMT from two to three (schematicity hierarchy pathways, ad hoc pathways, and shared image schema pathways). Next, in order to show the differences and also the advantages of his multilevel model in comparison to models (cf. Cameron 2008; Steen 2017; Zinken et al. 2008; for an overview see Gibbs 2017) that “downplay the role or even eject the existence of conceptual metaphors of the kind mostly discussed in the CMT literature […] and, at the same time, they emphasize the role of context in the creation and use of metaphors” (pp. 154–155), Kövecses makes a distinction between three types of metaphorical meaning: (i) bodily motivated image schema based metaphorical meaning at the subindividual level – e.g., primary metaphors, (ii) decontextualized meaning on the level frames and domains on the supraindividudal level – e.g., conventionalized metaphorical models, and (iii) contextual meaning constructed in mental spaces on the individual level – contextualized metaphorical meanings. In consequence, Kövecses claims that his multilevel approach is able to capture the experiential basis and the contextual factors priming the use of a metaphor in a discourse, respectively. Relying on aforementioned distinctions, Kövecses also presents a new classification of conceptual metaphors: “systematic metaphors (based on schematicity hierarchy pathways), non-systematic (isolated) metaphors (based on ad hoc pathways), and systematic resemblance (analogy) metaphors (based on shared image schema pathways)” (p. 168).

In the closing sections of Chapter 7, Kövecses gives a concise summary of the components of his multilevel and contextual approach and, in the last chapter of the book, Chapter 8, he revisits the main topics, i.e., his initial questions, and formulates the answers once again within the framework of ECMT. He also briefly compares his multilevel and contextual approach with other current models in metaphor research, like Deliberate Metaphor Theory (Steen 2008), Structure-Mapping Theory (Gentner 1983), or Relevance Theory (Sperber & Wilson 2008).

However, in my view, more attention should have been devoted to the critical analysis of relevance theoretical considerations and also to the notion of metaphorical and literal meaning in Relevance Theory. The latter should have been dealt with in depth especially because Sperber & Wilson (2008) not only reject, as Kövecses remarks, the “existence of conceptual metaphors” (p. 180), but also, as we can add, they do not even accept the notion of metaphor in the description of meaning construction processes at all. And moreover, Sperber & Wilson (2012) do not only call into question the explanatory force of the notion metaphor, but also that of literal meaning: “In our account, we give a theoretical status to the notions of explicature and implicature […], but not to the notions of literal meaning and what is said” (Sperber & Wilson 2012, 77). In the light of the relevance theoretical literature, relevance theoretical accounts could be treated as rival models of the construction of meaning in general and figurative meaning in particular. Therefore, a critical comparison of the multilevel approach of Kövecses and the relevant theoretical accounts of meaning construction might produce interesting insights for both side. Interestingly, so-called hybrid theories have also been provided, which set themselves the objective of connecting relevance theory and conceptual metaphor theory (Tendahl 2009; Stöver 2010; Csatár 2014).

In contrast to the relevance theoretical accounts, the multilevel approach of Kövecses shares more assumptions with the Dynamic System Model of Raymond Gibbs (Gibbs 2013, 2017). In the last section of the book, Final Words, Kövecses considers that these two theoretical models are not only compatible but can also be complementary of “a single unified theory” (p. 185).

The building of the Conceptual Metaphor Theory now appears in an entirely new light. The extended CMT assumes much of the standard version but it extends it in various ways. The multilevel approach provides a systematic order of schematicity hierarchy which, on the one hand, integrates image schemas, domains, frames, and mental spaces in one model and, on the other hand, can also take the contextual factors of metaphor producing and understanding into account. Kövecses’ new approach, ECMT, is a complex model that covers the relevant conceptual components and communicative dimensions of understanding, producing and storing metaphorical meaning, and it will open new ways to look at old problems. It offers a detailed description of the analytical tool that can be employed in the analysis of different types of metaphors, and also presents further topics for his framework, which can be treated as new challenges for the cognitive metaphor research.

However, some old challenges still remain. The empirical, experimental foundation of the model is not well developed. Testability would be especially important when correlation is assumed between neural activity and the degree of conventionality of metaphors: “the various degrees of conventionalization for the source domains and source expressions are correlated with differing degrees of neural activation” (p. 131). Also, the proposal of the schematicity hierarchies, e.g., the assumption of four distinct levels of conceptual structures (image schema, domain, frame, and mental space) should be supported by experimental and neural evidence. Nonetheless, both of these proposals are highly intriguing theoretical innovations, and Kövecses demonstrates with a series of example analyses the capability of his model as a linguistic tool.

During the course of the past ten years, several authors have come up with similar innovative proposals. We have seen successful examples for the enrichment of methodology and the justification of the efficiency of corpus linguistics in metaphor research (Deignan 2005; Stefanowitsch 2006). Experimental procedures have always been present in Conceptual Metaphor Theory and exerted significant theoretical influence on it (Gibbs 1994, 2017). We have also come across new directions of extension, too (Steen 2011). However, within the framework of CMT, it has definitely been much less frequent to encounter theoretical innovations in this respect in the past couple of years (Grady 1997; Fauconnier & Turner 2002). From this perspective, Kövecses’ book may indeed be a significant milestone in the almost fifty-year history of conceptual metaphor theory. This is why Extended Conceptual Metaphor Theory is considered an important reference point for all those who regard the possibility of building the theory further through the issues highlighted by Zoltán Kövecses.