This page intentionally left blank chapter 17 COGNITIVE GRAMMAR ronald w. langacker 1. Background Research leading to the formulation of Cognitive Grammar began in the spring of 1976. On the American theoretical scene, it was the era of the ‘‘linguistics wars’’ between Generative Semantics and Interpretive Semantics. The research was stim- ulated by the realization that this dispute was vacuous and sterile, that making sense of language required a wholly different way of thinking about it. Within three years, the overall architecture and basic descriptive constructs of the new frame- work were established. The first published descriptions, under the rubric ‘‘Space Grammar,’’ were Langacker (1981) and (1982). Its rechristening as ‘‘Cognitive Gram- mar’’ in the first full-length presentation (Langacker 1987a) was not the result of any modification. To this very day, in fact, changes have been matters of elabo- ration and refinement—the basic notions remain intact. Cognitive Grammar was not derived from any other theory, nor is it particularly close to any. While it does bear certain resemblances to numerous other frameworks, these are limited in scope and apparent only when stated in general terms. With Generative Semantics, for instance, Cognitive Grammar shares only the general vi- sion of treating semantics, lexicon, and grammar in a unified way. The most ex- tensive similarities are with Construction Grammar (Fillmore 1988; Goldberg 1995; Michaelis and Lambrecht 1996;Croft2001; this volume, chapter 18). Though developed independently, the two frameworks share a number of basic ideas: that constructions (not ‘‘rules’’) are the primary objects of description; that lexicon and grammar are not distinct, but a continuum of constructions (form-meaning pairings); and that constructions are linked in networks of inheritance (or catego- rization). Yet their extensive differences are also quite apparent. A glance at their respective diagrams reveals radically different formats symptomatic of substantially different theoretical devices and descriptive constructs. In representing meanings, Construction Grammar largely ignores the construal factors (e.g., profiling) taken as fundamental in Cognitive Grammar. Moreover, it does not embrace the pivotal claim of Cognitive Grammar that all valid grammatical constructs have a conceptual characterization; notions like noun, verb, subject, and object are essentially treated as unanalyzable syntactic primitives. 1 Still, Cognitive Grammar is part of the wider movement that has come to be known as Cognitive Linguistics, which, in turn, belongs to the broad and diverse functionalist tradition. The categorization of linguistic approaches as ‘‘function- alist’’ versus ‘‘formalist’’ is of course simplistic and increasingly irrelevant. While their prototypes are sharply distinct, a schematic characterization valid for the categories overall is the rather tenuous one of whether functional considerations are taken as foundational to an account of language structure, or merely subsidiary (Langacker 1995c, 1999a). By this criterion, Cognitive Grammar is strongly func- tional, granted that the two basic functions of language are symbolic (allowing conceptualizations to be symbolized by sounds and gestures) and communicative/ interactive. The symbolic function is directly manifested in the very architecture of Cognitive Grammar, which posits only symbolic structures for the description of lexicon, morphology, and syntax. A manifestation of the communicative/inter- active function is the fundamental claim that all linguistic units are abstracted from usage events. What, then, is ‘‘cognitive’’ about Cognitive Linguistics and Cognitive Gram- mar? Within the functionalist tradition, they are distinguished by the notion that properly describing language from the communicative/interactive perspective re- quires an explicit description of the conceptual structures involved. These struc- tures include the interlocutors’ apprehension of each other, of their interaction, of the context, and of the ongoing discourse itself. With respect to generative doc- trine, Cognitive Linguistics merits the label by virtue of treating language as an integral facet of cognition rather than a separate ‘‘module.’’ Insofar as possible, language is seen as recruiting more general cognitive phenomena (e.g., attention, perception, categorization, memory) from which it cannot be dissociated. Cognitive Grammar is envisaged as fitting into a more comprehensive theory of language structure comprising three interdependent levels. At the first level is a descriptive framework allowing the explicit characterization of the full range of linguistic structures empirically encountered. Work in Cognitive Grammar has aimed primarily at articulating such a framework. Because it needs to accommodate even the most unusual structures, the framework must be quite flexible and will thus define a very large space of structural possibilities. The appropriate restric- tiveness comes from level two, an enumeration of what kinds of structures are universal or prototypical in the world’s languages and to what degree. On the basis of cross-linguistic surveys, this enumeration will specify just how the space of 422 ronald w. langacker structural possibilities is ‘‘warped,’’ such that certain options are exploited far more readily and frequently than others. Level three, then, consists of functional ex- planations for the findings at levels one and two. A primary commitment of Cognitive Grammar is thus to provide an opti- mal set of constructs for explicitly describing linguistic structure. Its formulation has been guided throughout by a number of principles thought to be helpful in achieving such optimality. The first principle, already alluded to, is that functional considerations should inform the process from the outset and be reflected in the framework’s architecture and descriptive apparatus. Because the functions of lan- guage involve the manipulation and symbolization of conceptual structures, a second principle is the need to characterize such structures at a reasonable level of explicit detail and technical precision. To be revealing, however, descriptions must be natural and appropriate. Thus, a third principle is that language and languages have to be described in their own terms, without the imposition of artificial bound- aries or Procrustean modes of analysis based on conventional wisdom. As a cor- ollary, formalization is not to be considered an end in itself, but must rather be assessed for its utility at a given stage of investigation. That no attempt has yet been made to formalize Cognitive Grammar reflects the judgment that the cost of the requisite simplifications and distortions would greatly outweigh any putative ben- efits. Finally, a fourth principle is that claims about language should be broadly compatible with secure findings of related disciplines (e.g., cognitive psychology, neuroscience, and evolutionary biology). Nevertheless, the claims and descriptions of Cognitive Grammar are all supported by specifically linguistic considerations. Radically different when initially formulated, Cognitive Grammar seems much less so today simply because ‘‘mainstream’’ linguistic theory has steadily evolved in its direction. Widely contemplated today, if not generally accepted, are Cognitive Grammar notions such as the following: prototype categorization; conceptual se- mantics; the semantic basis of most grammaticality judgments; the inseparability of grammatical and semantic analysis; lexicon and grammar forming a continuum; constructions as the primary objects of description; inheritance networks; ‘‘rules’’ as schemas (or templates)—see also Tuggy this volume, chapter 4; a nonderivational (‘‘monostratal’’) view; well-formedness as simultaneous constraint satisfaction; composition as ‘‘unification’’; a ‘‘usage based’’ model. Despite these points of con- vergence, Cognitive Grammar remains unique (and in some eyes notorious) by virtue of certain strong and controversial claims, notably the conceptual charac- terization of basic grammatical notions (e.g., noun, verb, subject, object) and the full reduction of lexicon and grammar to assemblies of symbolic structures. It is further distinctive in its overall vision and at the level of specific descriptive detail. At the same time, Cognitive Grammar is arguably quite conservative and down- to-earth. Care is taken not to invoke any cognitive phenomena that are not well known or easily demonstrable. In adopting descriptive constructs, a strategy is systematically employed of seeking converging evidence from three independent sources: in addition to being cognitively plausible, a construct must prove nec- essary for describing and distinguishing meanings and must further be shown to cognitive grammar 423 play a role in grammar (Langacker 1993b, 1999a). Moreover, the content require- ment imposes a kind of restrictiveness and theoretical austerity unmatched by most other theories. First, it limits the linguistic units one can posit to semantic struc- tures, phonological structures, and symbolic structures (which pair the other two). Second, the units posited must either be part of the primary data (occurring ex- pressions) or else be derivable from it via the basic psychological processes of schematization and categorization. These conservative properties are quite desirable provided that the framework nonetheless permits an adequate characterization of language structure. Cognitive Grammar has, though, been criticized (e.g., by Huffman 1997: 331–32) as being overly conservative on another score, namely in adopting a number of traditional gram- matical notions considered problematic for a universally applicable descriptive in- ventory. Indeed, an array of traditional terms are in fact employed with something approximating their familiar values: terms like noun, verb, subject, morpheme, constituency, and subordinate clause. Attentive readers will notice, however, that in each case the notion in question has been thoroughly reconceptualized and refor- mulated in a way that avoids classic problems and makes it potentially appropriate for universal application. 2 The idea is not to stick with concepts known to be inadequate in their standard form, but to rescue what is useful with suitable mod- ification in a new overall theoretical context. In any case, Cognitive Grammar is better known for the proliferation of new concepts and terminology. 2. Architecture Viewed as a mental phenomenon, a language resides in organized processing ac- tivity (patterns of neural activation). The convenient use of terms like ‘‘linguistic knowledge,’’ ‘‘linguistic system,’’ and ‘‘internalized grammar’’ should not be allowed to obscure its essential dynamicity or to suggest a discretely bounded module. Knowing a language is having mastered a set of skills: a vast number of perceptual, motor, and cognitive operations that can be recruited and executed along with many others in speaking and understanding. A language is defined in Cognitive Grammar as a structured inventory of conventional linguistic units. A unit is a pattern of processing activity that is thoroughly mastered and can thus be carried out more or less automatically (a ‘‘cognitive routine’’). Reference to an inventory of units is meant to indicate the framework’s nongenerative and nonconstructive nature: linguistic units do not constitute an autonomous derivational system itself responsible for constructing well-formed expressions, but are merely resources that speakers can exploit in doing so. This inventory is structured in the sense that, instead of being separate and discrete, units relate to one another in various ways (overlap, inclusion, sym- 424 ronald w. langacker bolization, categorization, integration into higher-level units). Of course, the units recognized as part of a language must also be linguistic in nature and conventional within a speech community. So defined, a language cannot be precisely delimited. A particular structure achieves the status of a unit through progressive psychological entrenchment, which is clearly a matter of degree. Also a matter of degree is conventionality: how widely a structure is shared among speakers (and accepted as such). Nor are there definite boundaries between ‘‘linguistic’’ and ‘‘extralinguistic’’ structures (Langacker 1987a: section 2.1.2). The delimitation of ‘‘linguistic meaning’’ is notoriously problematic. Hardly less so is the demarcation of ‘‘linguistic symbolization.’’ While segmental phonology may be central, the range of conventional symbolizing structures further includes both intonational and gestural phenomena. In these domains, we observe an apparent gradation leading from established patterns to spontaneous expres- sivity, such that any specific boundary has to be arbitrarily imposed. A usage event is defined as an actual instance of language use. It resides in the pairing of a comprehensive conceptualization, representing a full contextual un- derstanding, with an elaborate expression, in all its phonetic and gestural detail. All linguistic units are abstracted from usage events. The abstractive process is just a matter of reinforcing whatever commonalities recur across a number of usage events, being inherent in these events at any level of granularity. Features which do not recur fail to be reinforced and are therefore filtered out. Thus, all linguistic units are selective and schematic vis-a ` -vis the usage events from which they arise. As en- trenched processing patterns, these units are available for subsequent processing. Their activation in the context of subsequent usage events effects the latter’s lin- guistic categorization (i.e., their interpretation with respect to the currently es- tablished system). In principle, any facets of a usage event, or a sequence of events in a discourse, are susceptible to being abstracted and conventionalized as a unit. For analytic purposes—and with the caution that they must not be reified as separate, discretely bounded boxes—it is useful to posit a number of sectors, as sketched in figure 17.1 (Langacker 2001a). One sector is the ground, comprising the speaker (S), the hearer (H), their interaction ( 3 " ), and their immediate circumstances. Central to their linguistic interaction is the directing and focusing of attention ( " ). This involves the conceptual analogue of the visual field, a subjective ‘‘space’’ within which a conceptualization is manifested. Called the viewing frame, this space delimits the general locus of viewing attention (metaphorically, it can be referred to as the ‘‘onstage region’’). 3 Any facet of the interlocutors’ conceptual universe can appear in this frame, within which they direct their attention to a specific focus of attention. All of this is embedded in a larger context, which in turn is embedded in the body of knowledge presumed to be shared by the speaker and hearer. To the extent that such factors are apprehended and mentally accessed, they figure in a usage event as part of the conceptualization constituting an expression’s full contextual understanding. Of course, a crucial dimension of this understanding is appre- hension of the ongoing discourse itself, comprising both previous and anticipated cognitive grammar 425 usage events. This is called the current discourse space, that is, the mental space comprising whatever is shared by the speaker and hearer as a basis for commu- nication at a given moment in the flow of discourse. Hence, the overall conceptualization evoked in a usage event includes far more than what appears onstage within the viewing frame. Supporting the onstage con- ception is an elaborate conceptual substrate , which shapes it and renders it coher- ent. What appears onstage, moreover, is also elaborate and multifaceted. As shown in figure 17.2, the conceptualization and the expression can each be resolved into a number of channels (which nonetheless interact in complex ways). Two particular channels are usually most central and have the most substantial content. For con- ceptualization, the focused channel is the conception of the situation being dis- cussed; as the object of discussion, this is dubbed the ‘‘objective situation.’’ Other conceptualization channels (more peripheral to speaker awareness despite their im- portance) are those pertaining to information structure and speech management. On the expression side, the focused channel (in the case of spoken language) is that of ‘‘segmental’’ phonological content. Other major channels of expression are in- tonation and gesture. 4 Since all the sectors and channels depicted in figures 17.1 and 17.2 factor into usage events, they potentially factor into the conventional linguistic units ab- stracted from such events. A particular unit reflects a recurring usage configuration by making specifications in certain sectors, but remains unspecified (or maximally schematic) in regard to others. At the very least, however, each unit incorporates an indication of its own conventional status: a schematized representation of inter- locutors using it in speaking the language in question. 5 A unit’s conventional im- port with respect to various factors often excluded from the scope of linguistic description (e.g., register, affect, discourse function, relative social status of the in- terlocutors) is also specified in sectors not focused in the viewing frame. The two global facets of a usage event are conceptualization and expression. Corresponding to these are the two global facets of abstracted linguistic units, re- ferred to as the semantic pole and the phonological pole. Interpreted broadly, a unit’s Figure 17.1. Structures relevant to discourse 426 ronald w. langacker phonological pole comprises all the channels of expression. Interpreted more nar- rowly, the phonological pole is limited to expressive channels in which a signifi- cant specification is made. 6 A unit’s semantic pole can likewise be defined either narrowly or broadly. Narrowly defined, its semantic pole consists of central and significantly specified channels of conceptualization. Conceived more broadly, how- ever, the semantic pole includes all the sectors in figure 17.1, regardless of specificity. It is even taken as subsuming the channels of expression, on the grounds that these are also apprehended and for various purposes are advantageously treated as facets of conceptualization (Langacker 1987a: section 2.2.1). Under the broad definitions, every linguistic unit has both a semantic and a phonological pole. Under the narrow definitions, three basic types of units can be distinguished depending on which sectors have salient and significant specifica- tions. Semantic units are those that only have a semantic pole (in the narrow sense), while phonological units (e.g., a phoneme or a phonotactic pattern) have only a phonological pole. A symbolic unit has both a semantic and a phonological pole, consisting in the symbolic linkage between the two. 7 These three types of units are the minimum needed for language to fulfill its symbolic function. A central claim— embodied in the content requirement—is that only these are necessary. Cognitive Grammar maintains that a language is fully describable in terms of semantic struc- tures, phonological structures, and symbolic links between them. Linguistic units are further limited to those arising from occurring expressions via schematization and categorization. On this view, lexicon and grammar form a continuum consisting solely of symbolic structures. Lexicon is defined as the set of ‘‘fixed’’ expressions in a lan- guage, that is, conventional expressions with the status of units. This set is not sharply bounded, for both psychological entrenchment and conventionality in a speech community are matters of degree. Fixed expressions vary along two basic parameters: specificity and symbolic complexity. At the phonological pole, expres- sions are quite specific, since to be expressions they have to be overtly manifested. 8 Semantically they run the gamut from highly specific to highly schematic (e.g., tack hammer > hammer > tool > implement > object > thing). An expression’s symbolic complexity is the number of constitutive symbolic elements it contains: sharp < sharpen < sharpener < pencil sharpener < electric pencil sharpener. The expressions traditionally recognized as lexical items are generally fairly specific and of limited Figure 17.2. Channels cognitive grammar 427 symbolic complexity. However, imposing any particular line of demarcation would be arbitrary. Thus, the highly schematic meanings of ‘‘grammatical’’ elements— such as the infinitival to, the preposition of, or the auxiliary verb do—do not prevent them from also counting as lexical items. Nor is lexicon limited to words, compounds, and short phrases. Provided that they are learned as conventional units, expressions of any size qualify as lexical items. 9 The two parameters of specificity and symbolic complexity define a continuous field of possibilities in which particular symbolic structures can be situated. While there are no definite boundaries, various subfields roughly correspond to traditionally recognized domains. Lexical items as classically conceived occupy the subfield char- acterized by limited symbolic complexity and a high degree of semantic and pho- nological specificity. To the extent that symbolic structures are schematic rather than specific, they tend to be regarded as grammatical rather than lexical. Gram- matical markers are phonologically specific, semantically schematic, and symbolically noncomplex. Basic grammatical categories (notably noun and verb) are defined ab- stractly by symbolically minimal structures that are highly schematic both seman- tically and phonologically. Corresponding to grammatical rules (combinatory pat- terns) are symbolic structures which are both schematic and symbolically complex. A linguistic system is merely a vast inventory of conventional units (not a self- contained device wholly responsible for constructing or enumerating expressions). It provides an array of resources which, along with many others, can be drawn upon in speaking and understanding. Among the further resources employed are general andcontextualknowledge, basic cognitive abilities(e.g., memory, attention,planning, aesthetic judgment), as well as such ‘‘imaginative’’ capacities as metaphor, blending, mental space construction, and the evocation of ‘‘fictive’’ entities (Talmy 1996; Langacker 1999d). Linguistic units themselves reflect such factors internally. These same factors figure as well in the formation of novel expressions, which thus in- corporate many features not solely derivable from the linguistic units invoked. Hence, linguistic knowledge is inextricably bound up with numerous other resources exploited in a dynamic processing system. It resides in routinized ‘‘packets’’ of processing activity, some of which are activated as part of the overall process- ing done by the system as a whole in producing or understanding a new expression. If a unit is strongly activated as part of an expression’s apprehension, 10 their rela- tionship amounts to categorization: an instantiation of the unit is immanent in the processing constituting the expression. It is through such categorization—referred Figure 17.3. Coding 428 ronald w. langacker to as coding in Langacker (1987a)—that an expression is seen as manifesting a par- ticular language and attributed a particular structure with respect to it. An instance of coding is depicted in figure 17.3, where L represents the lin- guistic system, and U a usage event. It consists of unit A (from L) being activated and thus effecting the categorization of structure B (a facet of U). There are two basic possibilities, indicated by different kinds of arrows. On the one hand, A can be fully manifested in B, without distortion, though B is generally more specific. In this case the categorizing relationship is one of elaboration, drawn with a solid arrow. On the other hand, A might be manifested in B only partially or imperfectly. In this case the categorizing relationship is one of extension, drawn with a dashed arrow. But in either case, A is exploited as a resource in the processing activity con- stituting facet B of the usage event. How are units selected for the categorization of usage events? At the process- ing phase when linguistic units are still being recruited for exploitation, a usage event is only incipient. Before the units employed are selected and fully activated, neither the conceptualization nor the vocalization has yet been fully developed and struc- tured in accordance with their specifications. It is precisely the activation of a par- ticular set of units that results in a full-blown usage event interpreted as manifesting a particular linguistic expression. Let T then represent a potential target of catego- rization, that is, some facet of an incipient usage event. 11 On the basis of overlapping features, T tends to activate a set of units each of which has the potential to categorize it; this activation set is shown as an ellipse in figure 17.4a. The set’s members are all initially activated to some degree. Mutually inhibitory, they compete for the privilege of being fully activated, thereby effecting the categorization of T. A number of factors contribute to a unit’s selection: degree of entrenchment (inherent ease of activation), contextual priming, and extent of overlap with the target. As shown in figure 17.4b, the competition eventually results in one unit suppressing all the others and be- coming fully active as the categorizing structure. Figures 17.3 and 17.4 show just a single categorization. However, a usage event is simultaneously categorized by many conventional units, each pertaining to a particular facet of its structure. Collectively, these categorizations constitute the event’s structural description, that is, its interpretation with respect to the linguistic system. It is by virtue of being interpreted in a certain way that a usage event can be Figure 17.4. Activation of categorizing units cognitive grammar 429 . involved. These struc- tures include the interlocutors’ apprehension of each other, of their interaction, of the context, and of the ongoing discourse itself. With respect to generative doc- trine, Cognitive. 17 COGNITIVE GRAMMAR ronald w. langacker 1. Background Research leading to the formulation of Cognitive Grammar began in the spring of 1976. On the American theoretical scene, it was the. of the communicative/inter- active function is the fundamental claim that all linguistic units are abstracted from usage events. What, then, is ‘ cognitive ’ about Cognitive Linguistics and Cognitive