Reviewed by Florent Perek, Freiburg Institute for Advanced Studies.
Introduction
New Directions in Cognitive Linguistics is a collective volume edited by Vyvyan Evans and Stéphanie Pourcel. The book contains 19 peer-reviewed papers based, for the most part, on communications at the inaugural conference of the UK's Cognitive Linguistics Association, held in Brighton in October 2005. The topics dealt with are very diverse and exemplify a wide range of current trends in Cognitive Linguistics.
The book is organized into five sections. The first section is concerned with cognitive approaches to semantics and contains two contributions on theories of semantic representation (Harder and Evans) and two more methodogically-oriented papers on quantitative approaches to lexical semantics (Gries & Divjak and Glynn).
The second section addresses various questions related to conceptual metaphors and blending, such as the interpretation of conventional and novel metaphors (Huang), the status of metaphors in cognition (Casasanto), the generalizability of blending patterns (Fauconnier), and the role of blending in grammatical composition (Dancygier).
The third section deals with cognitive approaches to grammar. Zeschel discusses the definition of constructions from a usage-based perspective. Dabrowska presents an account of the acquisition of non-basic vocabulary based on the notion of collocations in a constructional view of lexical semantics. Langacker draws on a usage-based approach to grammar to clarify the distinction between verbal and constructional meaning. Moravcsik presents an epistemological discussion of the use of partonomical structures in syntactic description.
The papers in the fourth section are concerned with embodiment in language and cognition. Sinha discusses the relation between language and culture in an evolutionary perspective. Altman sees traditional systems of medicine as a window into the embodied mind. Chilton provides an embodiment-based description of the polysemy of get through a notational variant of image-schemas. Pourcel addresses the question of linguistic relativity (the relation between language and thought) as it pertains to the conceptualization of motion events.
The fifth section contains a collection of various papers dealing with either new developments of Cognitive Linguistics or new applications of its concepts to particular kinds of data. Croft argues that more attention should be given to the social and interactive dimensions of language in Cognitive Linguistics, and redefines the basic tenets of the field in a social-interactional perspective. Berman and Nir study the development of text construction with age, following a new heuristic method for the measure of text quality. Van Fliet extends van Hoek's model of sentential anaphora to account for the alternation between proper noun and pronoun when referring to characters in narrative discourse, highlighting the similarities and the differences induced by the discursive dimension. Rubba analyzes the dream that makes up most of the scenario of David Lynch's movie Mulholland Drive in the framework of blending theory. Pascual reports that the depiction of the same murder in the prosecutor's closing argument and in a literary text interestingly contain similar blends relying on fictivity and imagination.
Given the high diversity of topics represented in the book, it is not possible to present a concise and exhaustive summary of the scientific contribution of this volume as a whole. In the next section, I will therefore summarize each paper independently by highlighting its main goals and findings. I hope such an individual description will prove useful for readers to appreciate which contributions are most relevant to their own research (since the book does not contain abstracts). In the concluding section of this review, I will give a more general evaluation of the book as a whole.
Summaries of the contributions
Part I. Approaches to semantics: Theory and method
Peter Harder argues for a view of linguistic meaning intended to avoid the pitfalls of two extreme positions, namely the objectivist view of meanings as abstract objects detached from language use, and the opposite view that the only meanings are those of actualized utterances ('usage fundamentalism'). Harder's approach is based on a finer-grained partition of the usage event, holding for comprehension as well as production, into input (retrieval of pre-existing structures), processing (the combination and adaptation thereof) and output (the utterance meaning that was intended to be delivered, which can never be known in advance). Harder argues that while it is important for linguists working in a usage-based approach to look at the meaning of expressions in full utterances, they should not overlook the task of defining the meaning of these expressions. In his approach, input meanings are defined as instructions about how the corresponding expressions are allowed to be used in a given language community and are precisely derived through exposure to language use.
Vyvyan Evans presents the approach to semantic representation adopted in the Theory of Lexical Concepts and Cognitive Models (LCCM). In LCCM, an explicit distinction is made between cognitive models, which consist of rich perceptual and conceptual information and constitute the conceptual system, and lexical concepts, which are units of semantic structure paired with a phonological vehicle. Lexical concepts encode specifically linguistic content, which is highly schematic and tailored to compositional integration. In addition, lexical concepts of open-class lexemes provide an indirect access to one or more conceptual model(s); together, the lexical concept and the conceptual structure(s) linked to it constitute a semantic representation. Evans argues that the linguistic system of lexical concepts emerged over the evolutionary earlier conceptual system to facilitate the manipulation of conceptual representations for purposes of linguistically-mediated communication.
Stefan Gries and Dagmar Divjak introduce a usage-based method for the study of lexical semantics, based on the concept of behavioral profiles. This method is based on the widespread assumption that differences in meaning are associated to differences in usage, which translate into divergences in the distributional patterns of features in corpus data. The behavioral profile of a word or word sense consists of a set of frequency distributions of grammatical and semantic features of various sorts (word morphology, embedding syntactic structure, semantic features of dependent words, collocations, etc.), obtained by annotating instances of this item extracted from corpora. The behavioral profiles thus obtained can then be submitted to various kinds of data analysis (similarity metrics, multifactorial statistics, cluster analysis, etc.), which, for example, allow to determine on empirical grounds the structure of polysemy networks (i.e., how different senses of a word are related, which meanings are closer to each other, and which one should be considered as prototypical), and to characterize fine-grained differences between near-synonyms.
Dylan Glynn presents an empirical method to derive descriptions of the multiple functions of lexical items from corpus data, in terms of patterns of use. The method starts with annotations of linguistic (formal and semantic) as well as extra-linguistic (e.g., dialect and text topic) features on naturally-occurring instances of a given lexical item. Then, Correspondence Analysis is used to analyze (and plot) the correlation between features. Drawing on the example of hassle in British and American English, the author shows how this method can be applied as an exploratory technique to analyze the co-occurrence of features, conceived of as form-meaning pairs. It is also shown that this approach helps to inform traditional analyses of polysemy by identifying the central senses of a lexeme, which correspond to clusters of features.
Part II. Approaches to metaphor and blending: Theory and method
Mimi Ziwei Huang's paper is concerned with the interpretation of metaphors. The author adopts Giora's Graded Salience Hypothesis, according to which the literal meaning of an expression is always activated in comprehension, along with any conventional metaphorical reading it might have; novel metaphors, however, require a biasing context for the metaphorical reading to be activated. Huang argues for an alternative notion of salience, understood as the accessibility of a meaning at a given point in a given discourse. The author illustrates how the proposed notion of salience can contribute to the study of metaphors with the analysis of a short story, where the accessibility of the metaphor is shown to vary as the story unfolds, as supported by reader interviews.
Daniel Casasanto discusses the status of conceptual metaphors in cognition, by first pointing out that most of the evidence that people conceptualize abstract domains through metaphors is linguistic in nature, and thus fails to establish conceptual metaphors as a theory of mental representation in general. The question of whether metaphors found in language also structure conceptual knowledge in the context of non-linguistic tasks remains open, as the evidence collected in other studies and cited by Casasanto turns out to be in many respects inconclusive. In the article, the author reports a series of experiments investigating the metaphor similarity is proximity, which predicts that our notion of similarity depends of physical distance. These experiments evaluate whether two stimuli are rated as more similar when they are presented closer on a computer screen. Casasanto finds that such a correlation only holds for conceptual similarity (between abstract nouns or between the function of physical objects represented by pictures) but is inverted in the case of perceptual similarity (between faces or pictures of objects), where stimuli are rated as less similar when presented closer to each other than when presented further away. The author concludes that metaphors relate conceptual domains in different ways than those manifested in language, and that they should merely serve as hypotheses about the structure of abstract concepts.
Gilles Fauconnier refutes three fallacies surrounding the theory of conceptual blending: (i) blending is not a new operation posited because a new type of data was discovered, it might well be at work in more familiar data too, (ii) blending is not cognitively more costly than other cognitive operations; blends are everywhere and we apprehend them without difficulty, and (iii) conceptual integration does not purport to explain "everything" about language and cognition, even though it is a central human faculty. Fauconnier attempts to dispel these fallacies through an analysis of a newspaper column in terms of a conceptual integration network (the "smoking ears" network). It is shown that parts of this blend involve familiar, and to some extent, conventional conceptual mappings (such as metaphors and counterfactuals), but other aspects of it require novel mappings that emerge opportunistically in the context of utterance. This example highlights the role of integration as a precise meaning construction operation whose products (integration networks) can be novel and creative, but can also be generalized as templates, which leads to the familiar distinctions between the different phenomena that conceptual integration purports to subsume (metaphor, analogy, counterfactual, etc.).
Barbara Dancygier argues for treating grammatical composition as cases of conceptual integration, drawing on a study of particular copular constructions, exemplified by such sentences as Iraq is George Bush's Vietnam or The adjective is the banana peel of the parts of speech. She analyzes such sentences as cases of blending which involve the construction of a new blend based on the activated frames of the subject's referent (Iraq in the example mentioned above), enriched with properties of the predicate noun (Vietnam). Dancygier notes that the s-genitive and of-variants of the construction are not always equally possible, and argues that this fact follows from the meaning contribution of the s-genitive, which is claimed to profile an experiencing entity. This assumption is consistent with data from the "one person's X is another's Y" construction (e.g., One's person trash is another's treasure) which involves two different viewpoints on a same entity and in which the experiencers of these two viewpoints are expressed by s-genitives. Dancygier concludes that these copular constructions are better seen as instances of constructional blends, in which frames evoked by lexical items and the semantic contribution of grammatical markers are meshed into a complex integration network, giving rise to emergent aspects of meaning.
Part III. Approaches to grammar: Theory and method
Arne Zeschel addresses a fundamental question in constructional approaches to grammar: which criteria determine whether a given schema is stored in the mental grammar of speakers? The author illustrates this discussion with a corpus-based study of the semi-fixed phrase good for NP (instantiated by the idiom good for a laugh). Zeschel shows that this construction is associated with particular semantic preferences which are largely unaccounted for by a constructional description taking non-predictability as the decisive criterion for unit status. Instead, he suggests that knowledge of this construction is primarily centered on a few highly frequent exemplars, from which various analogical extensions can be made. The more extensions of this type are encountered, the more an abstract schema is likely to be stored as a construction. Entrenchment as a function of frequency is thus the decisive criterion for unit status in the usage-based approach that Zeschel advocates.
Ewa Dabrowska is concerned with an important and yet unresolved question in language development, namely the acquisition of non-basic vocabulary. While the meaning of some non-basic words may plausibly be learned through explicit definitions or inferred from the context of use, the author shows that such strategies are not very helpful for learners to figure out exactly in which contexts many of such words may be used, as it is the case with the many verbs of manner-of-motion in English (e.g., prance, stagger, trudge). Dabrowska suggests that speakers more successfully learn how to use non-basic words by attending to their frequent collocations (e.g., the fact that trudge often co-occurs with a reference to some difficult terrain), and puts forward a view of lexical entries as partially specified syntactic constructions. In these "words as constructions", information about regular collocations is stored directly in the syntactic dependents of the verb. Dabrowska presents evidence for this account from two empirical studies with English non-basic motion verbs. The first study shows that adult speakers are well-aware of the typical collocational patterns of particular words, as shown by the clear differences between verbs manifested in elicited sentences. The second study confirms this finding by showing that speakers are more successful at choosing the right verb if they are given sentence contexts containing typical collocations than when they are merely provided with a dictionary definition or referential information (in the form of video clips).
Ronald Langacker presents a unified account of various grammatical phenomena in terms of categorization either by elaboration of a schema (full recognition) or by extension from a prototype (partial recognition), with a particular focus on grammatical composition, which he characterizes as a case of complex recognition (i.e., categorization of a composite target by multiple standards). In particular, Langacker addresses the problem of the division of labor between lexical items and grammatical constructions, i.e., the question of whether some aspect of the compositional meaning of a composite assembly should be attributed to a special sense of a lexical item, or to the construction itself. Langacker shows that this debate makes no sense in Cognitive Grammar, since the apprehension of a lexical item as belonging to a certain category (grammatical category or lexical sense) cannot not teased apart from the occurrence of that item in certain constructions. As a lexical item becomes entrenched in one or more constructions, the composite structure gradually loses its analyzability, eventually leading to a new "extended" sense of the lexical item accessible in isolation. Langacker thus adopts an intermediate position between maximal polysemy (one sense per construction in which the lexeme occurs) and minimal polysemy (at best a single sense combined with additional meaning provided by the constructions): whether a lexeme is associated with different senses depends on the entrenchment of that item in the corresponding constructions, and is therefore a matter of degree.
Edith Moravcsik discusses the use of partonomic structures (i.e., part-whole relations) in the analysis of the syntax of natural languages. Evidently, the traditional notion of constituency is based on partonomic structures: different parts may form similar wholes (e.g., a noun phrase may consist of different numbers and kinds of words). However, partonomic structures in syntax often deviate from ideal ones in that they often turn out to be complex (e.g., constituent can overlap), and that evidence for phrasehood may be contradictory. Moravcsik shows that constituent overlap and discontinuous constituents is accounted for by layering, i.e., the positing of two simultaneous analyses of the same parts into different wholes. Syntactic theories differ as to how they implement layering. Generative transformational grammar captures layering by syntactic derivations, while cognitive grammar achieves the same goal by positing a discrepancy between the organization of phonological structure and semantic structure at each pole of the symbolic assembly.
Part IV. Language, embodiment and cognition: Theory and application
Chris Sinha discusses the relation of language to human nature and culture. The author argues for a milder position between the two extreme views that language is either a unique manifestation of the human nature (or in modern terms, a phenotypic expression of a genetic endowment), or a byproduct of human sociocultural behavior. Drawing on recent discussions of evolutionary theories, Sinha proposes that language may be analogized to animal artefacts (like bowerbirds' nests involved in male mating display), in the sense that it is a manifestation of the human behavioral repertoire and at the same time supports other behavioral strategies. The language faculty hinges not on a human-specific language acquisition device, but on a set of cognitive abilities which, while shared with other species, attain higher levels of organization in humans. Under this account, human beings developed language not genetically (i.e., not in the genotype of the entire species) but epigenetically (i.e., in the phenotype of generations of individuals). The sociocultural environment that characterizes human behavior was dramatically shaped by the human semiosphere (which language is only a part of), and at the same time allowed language to arise as a social institution.
Magda Altman investigates embodiment under the light of traditional medicine. The author argues that since representations of the body in traditional systems of medicine are derived from an intensive study of the subjective experience of "being in a body", they offer a source of information on the body schema, i.e., how the body is naively experienced and conceptualized. Altman notes that many conceptions of the internal dynamics of the body in traditional Chinese medicine bear striking similarities to embodiment-based analyses of the semantics of linguistic expressions, such as Talmy's Force Dynamics and Lakoff's image-schemas. These findings suggest that basic image-schemas may be derived from proprioceptive information related to the body schema, rather than perceptual information gathered in the external world.
Paul Chilton presents an account of the polysemy of the English verb get. This account is couched in a new theoretical framework based on spatial concepts formalized in geometrical terms, called Discourse Space Theory. In this approach, the meaning of lexical items such as get consists of a combination of basic image-schemas, represented by vectors in a multi-dimensional space. According to Chilton, the basic meaning of get is the so-called prehension schema. The others meanings of get (possession, benefaction, caused motion, self-motion, becoming, causality, modality, obligation) are derived from the basic meaning through operations on the vectors of the basic schema. This new account sheds a new light on the polysemy of get by showing how the many meanings of the verb may be motivated.
Stéphanie Pourcel addresses the long-debated issue of the relation between language and cognition as it pertains to the domain of motion events, i.e., whether speakers of typologically different languages (according to Talmy's distinction between satellite-framed and verb-framed languages) conceptualize motion events differently. Experimental studies in this domain report conflicting evidence, which Pourcel attributes to the stimuli involved, either because of their lack of naturalness with regards to human experience, their decontextualized character or because they simply differ between studies in ways that were not controlled for. She presents evidence that subjects speaking typologically different languages are nonetheless equally biased towards expressing manner or path according to properties of the motion event (such as figure type, path telicity, manner type and causality). Pourcel argues that this crucial dependence on the event being expressed largely explains the apparent divergences of previous findings. To tackle this methodological issue, a new study is presented in which English and French subjects were shown the same short movie (taken from Charlie Chaplin's City Lights) and were asked to complete three tasks: a free recall task right after viewing consisting in the production of a narrative description of the movie, a prompted recall task 24 hours after viewing consisting in questions about manners and paths of motion events. Unsurprisingly, English speakers show a greater attendance to manners of motion in their narratives. In addition, their recollection of manners is more accurate than that of paths (compared to the actual events shown in the movie), and this effect persists after 24 hours. The pattern of results is inverted for French speakers, who show a greater attendance to and a better recollection of paths over manners of motion. Both groups follow the same trends in the subjective inferences they produce in their narratives, in that English speakers make more inferences related to manner than path, and vice versa for French speakers. These findings are indicative of different patterns of thought about motion events between French and English speakers, which gives credence to the linguistic relativity hypothesis.
Part V. Extensions and applications of cognitive linguistics
William Croft makes a plea for a closer and more explicit integration of the social dimension in cognitive linguistics, thus defining a 'social cognitive linguistics'. He argues that CL has so far focused too exclusively on the psychological aspects of language, neglecting the fact that language is not just a psychological phenomenon, but also a central feature of human social interaction. In an attempt to bring together the social and cognitive aspects of language, Croft proposes to reformulate the four basic tenets of cognitive linguistics (language instantiates general cognitive abilities, grammar is symbolic, meaning is encyclopedic, meaning involves construal) under a social-interactional perspective, drawing on previous research in this field. The first tenet is re-defined as 'language instantiates general social cognitive ability', thus adding joint-action, coordination and convention to the well-known triad of categorization, perception, and memory. Second, the symbolic nature of language must incorporate a third component in addition to form and meaning: the community in which the symbolic pairing is conventional, which can be defined at any level of granurality, thus covering (inter alia) dialectal, register and genre variation. Third, the encyclopedic nature of meaning stems from the fact that it is shared common ground, in the sense that it is socially constructed from the cultural tradition of a community of practice (which relates to the general social cognitive ability of joint-action). Fourth, construal is taken to fulfill a communicative function, in that it serves for the achievement of the speech act at more discursive levels than that traditionally dealt with in mainstream cognitive linguistics. Finally, Croft illustrates with a case study how this thus-defined 'social cognitive linguistics' approach offers new perspectives for addressing traditional linguistic issues, by analyzing how it sheds light on the variation in the verbalization of experience by different speakers, taken from Chafe's Pear Stories corpus.
Ruth Berman and Bracha Nir study the development of text construction, with a focus on expository discourse (i.e., essays, as opposed to narratives). They point out that previous approaches to the issue of evaluating text quality, mostly based on subjective ratings by independent judges, are methodologically flawed since the criteria used by each judge are unclear, which often leads to disagreements. Besides, more objective criteria proposed in the literature unfortunately do not lend themselves well to the analysis of expository discourse, as they are mostly based on narrative structure. To mend this methodological gap, Berman and Nir propose an analysis of texts at two levels: the 'local' level of linguistic usage (syntax and lexis) in individual clauses, captured by quantitative measures, and the 'global' level of textual structure and discourse organization, as measured by several heuristic (but precisely defined) criteria. This method is applied to the analysis of texts (both narrative and expository) elicited from English and Hebrew speakers of different age groups. They find that while text quality generally increases with age (as expected), the local and the global level follow different patterns of progression. In addition, the progression at the global level is slower for expository discourse than for narratives. These results show that text construction abilities develop largely independently of the purely linguistic competence, and that they involve higher cognitive demands, especially in the case of expository discourse.
Sarah van Fliet proposes a cognitive linguistic account of anaphora in connected discourse by extending Van Hoeck's 'reference point model' of sentential anaphora. She suggests that discourse anaphora is governed by less entrenched and more flexible patterns than sentential anaphora, but is not qualitatively different from sentential anaphora, and thus may be accounted by with similar factors as those proposed by Van Hoeck. On the basis of examples drawn from English novels, van Vliet shows that the choice between proper noun and pronoun when referring to a character in narrative discourse is governed by the interaction between five factors: (i) the topicality of the character in the local discourse context, (ii) referential distance, (iii) the presence of possible competing referents, (iv) narrative structure (defined in terms of the succession of narrative episodes, which mark conceptual breaks in referent accessibility), and (v) the perspective adopted in a given stretch of discourse (either an objective perspective or the subjective perspective of a character). Van Vliet validates this account through an analysis of elicited Dutch narratives.
Johanna Rubba presents an analysis of the plot of David Lynch's movie Mulholland Drive in terms of conceptual blending. As in many of Lynch's movies, dreaming is an important aspect of the apparently incoherent scenario of Mulholland Drive: the first two-third of the movie must actually be interpreted as a dream by one of the main protagonists (Diane), who then awakes and confronts the reality of a life that has gone terribly wrong (and finally commits suicide). Rubba analyzes Diane's dream as a blend between a reality space (events as they actually happened) and a wish space (events as they should have happened). In that blend, many characters change identities, and relations between them are changed as well to match the wish space. The blend also includes unique elements that help fulfilling the dreamer's psychological goals: relief of guilt, achieving success in life, punishing a betraying lover and at the same reclaiming her love. Rubba's study, in addition to revealing the applicability of conceptual blending theory to the analysis of movies, offers perspectives into more convergence between the analysis of dreams as blends and their traditional treatment in psychoanalysis.
Esther Pascual compares two descriptions of the same murder, one drawn from a non-fiction book, and the other from the prosecutor's closing argument at the murder's trial. Both descriptions are based on very few factual evidence (since there was no witness other than the defendant and no aspect related to the murder had been irrefutably proven), hence they both present a depiction of events which is halfway between reality and fiction. Drawing on the concepts of conceptual integration theory, Pascual finds that, despite their differences in genre, both descriptions are rich with many conceptual blends (sometimes very similar ones) depicting imagined events, between 'content' (i.e., facts related to the murder) and 'context' (i.e., elements of the 'Here-and-Now' where the discourse itself unfolds). Pascual argues that these discourse strategies are employed to achieve a goal shared by both writer and prosecutor: sympathy for the victim. This shows that despite the accepted assumption that legal discourse is organized around the presentation of facts and truth, it draws on fictivity and imagination to a similar extent as literary work. In addition, the study shows the importance of taking into account the broad, situated discourse context in the analysis of linguistic communication.
Conclusion
New Directions in Cognitive Linguistics is a high-quality volume that deserves the attention of any scholar with an interest in the cognitive aspects of language; it is a must-read for cognitive linguists proper. All contributions in this volume are of a very high scientific quality, in content as well as in form. The papers address a wide range of topics, and exemplify an equally wide range of methodologies: theoretical discussions, corpus-based studies, discourse analysis and experiments. The huge amount of work and time implied by the collection and careful selection of these many papers, compiled in a sizable 519-page volume (this is to this day the biggest collective volume in the 'Human Cognitive Processing' collection), should be duly acknowledged and the quality of the result highly praised.
As usual with collective volumes, the contributions might seem quite heterogeneous at first blush, but many of them complement each other nicely, either because they present further empirical evidence for some theoretical claim (for example, Fauconnier's article and the other contributions on conceptual blending, especially van Fliet's and Rubba's), or discuss different aspects of the same question (for example, Harder's and Evans' contributions on semantic representations, and Zeschel's and Langacker's papers on usage-based grammar). As far as the structure of the book is concerned, it is overall quite coherent, in particular parts I, II and III. Part IV and V unfortunately appear more like 'miscellaneous' categories, since they contain contributions which are only loosely related in terms of content (if at all). Conversely, some thematic connections are not explicitly reflected in the book's structure; for example, one might wonder why van Fliet's and Rubba' papers appear at the end of the book since they present some continuity with Fauconnier's and Dancygier's contributions, as mentioned earlier; perhaps they would have found a better place in part II (at the risk of unbalancing the otherwise neat distribution of papers into the sections of the book), or conceptual blending would have deserved a section of its own, given the high representation of this topic (4 articles). That being said, it must be acknowledged that the problem of structuring content in collective volumes is often a Gordian knot, given the inherent heterogeneity of this type of publication, and that the structure proposed by Evans and Pourcel is more than satisfactory.
In conclusion, New Directions in Cognitive Linguistics undoubtedly achieves the objective spelled out in its title. The editors managed to compile a rich synthesis of state-of-the-art research in the field, which reports recent advancements in traditional issues of central concern since the advent of Cognitive Linguistics, and also opens new perspectives on the theoretical extensions and the scope of application of many fundamental concepts. In its pages, many readers will surely find more than one relevant piece of research, and without a doubt many sources of inspiration for their own investigations into .
Commissioned Jan 05, 2011
Submitted Aug 15, 2011
Final version submitted Sept 12, 2011