said to manifest a particular linguistic expression. Moreover, the set of categori- zations provide an assessment of the expression’s degree of well-formedness (or conventionality). The two kinds of categorizing relationships depicted in figure 17.3, elaboration and extension, respectively correspond to whether the target conforms to the conventional unit invoked or distorts it in some manner. If all the catego- rizations effected on a given occasion are elaborative, the expression is fully well- formed. Each categorization involving extension represents a measure of noncon- ventionality (ill-formedness). Since linguistic resources are always being stretched to accommodate new circumstances, a certain amount of nonconventionality is typical of normal language use. It is only when the distortions are drastic enough (individually or collectively) that an expression is judged as being deviant. Figure 17.3 depicts a particular unit, A, being employed for the categorization of B in the context of a single usage event. Whether it constitutes an elaboration or an extension, B represents the contextual value assumed by A on that occasion. Suppose, now, that A is used with comparable value on a number of occasions (e.g., a lexical item might be used repeatedly with the same extended meaning). If both B and B’s categorization by A occur across a series of usage events, they—like any other facet of such events—are subject to progressive entrenchment and con- ventionalization. The result, as shown for the case of extension in figure 17.5, is that both achieve the status of conventional linguistic units and are thus incorporated in the linguistic system (as a matter of definition). Starting from a single unit, A, successive developments of this sort can eventually yield a network of related units linked by categorizing relationships (which can themselves be recognized as units). This network is a complex category, with A as its prototype (Lakoff 1987; Langacker 1987a: chapter 10; Taylor 1995). It is through the constant impact of usage that linguistic units maintain themselves and evolve. The activation of a unit reinforces and further entrenches it. Conversely, a unit that is not exploited tends to ‘‘decay’’ and may eventually be lost. Through elaboration and extension, units spawn variants which can then take on lives of their own. Schemas emerge by the reinforcing of abstract com- monalities that consistently recur, and complex units arise from simpler struc- tures that consistently co-occur. Thus, every instance of language use has some impact, however slight, on the linguistic system as currently constituted. In this usage-based perspective (Barlow and Kemmer 2000), synchrony and diachrony are inseparable. Figure 17.5. Extension 430 ronald w. langacker 3. Semantics A central claim of Cognitive Grammar is that only symbolic structures (form- meaning pairings) need be posited for the characterization of lexicon and gram- mar, which form a continuum. This symbolic view of grammar implies that the elements, structures, and constructs employed in grammatical description must all be meaningful (just as lexical items are). Their meaningfulness can be recog- nized and accommodated only given an appropriate view of linguistic meaning, one that is strongly justified in its own terms. Required is a conceptualist seman- tics. Meaning is identified with conceptualization, broadly defined as encompassing any kind of mental experience: (i) both established and novel conceptions; (ii) not only abstract or intellectual ‘‘concepts’’ but also immediate sensory, motor, kinesthetic, and emotive experience; (iii) conceptions that are not instantaneous but change or unfold through processing time; and (iv) full apprehension of the physical, linguistic, social, and cultural context. A conceptualist semantics has none of the dire consequences sometimes as- cribed to it. There is first the doctrine, still expressed in semantics textbooks, that concepts are intrinsically mysterious, and their investigation unscientific. 12 On the contrary, cognitive semantic research (e.g., Fauconnier 1985, 1997; Lakoff and Johnson 1980, 1999; Fauconnier and Turner 1998, 2002; Talmy 2000a, 2000 b) has yielded detailed, explicit, empirically grounded, and increasingly principled de- scriptions of conceptual structure that have greatly altered received wisdom about cognition. Within Cognitive Grammar, validity of the descriptive constructs em- ployed is assured by requiring converging evidence from three sources (Langacker 1993b, 1999a). The basis for adopting a particular construct is to show that it is (i) necessary just for purposes of semantic description (e.g., to distinguish the mean- ings of semantically similar expressions), (ii) the manifestation of an indepen- dently attested cognitive ability, and (iii) critical for the explicit characterization of varied grammatical phenomena. Nor does a conceptualist semantics imply solipsism or an inability to deal with social interaction and discourse (cf. Harder 1996). Though conceptualization takes place internally (residing in processing activity of the brain), it is not autonomous or encapsulated, and it is not itself the object of contemplation. 13 Conceptuali- zation is always the conceptualization of something, a facet of either the real world we inhabit or a constructed world ultimately grounded in real world experience. Conceptualization is precisely the act of engaging the world, the experiential aspect of our interaction with it. Broadly understood, conceptualization includes percep- tual experience, as well as the central control of motor activity and the kinesthetic sensations it induces. It further includes the interlocutors’ apprehension of the dis- course and the interactive context supporting it. Also erroneous is the supposition that equating meaning with conceptualiza- tion implies the absence of any distinction between semantics and pragmatics or implies that the conceptualizations invoked for linguistic purposes are unaffected cognitive grammar 431 by being so employed, hence exactly equivalent to nonlinguistic conceptions. 14 Rather, semantic structures are specifically seen as representing the adaptation of conceptualization for expressive purposes, thus conforming to both the strictures of linguistic convention and the exigencies of language function. Cognitive Gram- mar’s claim that the semantics/pragmatics distinction is ‘‘largely artifactual,’’ a false dichotomy (Langacker 1987a: 154), pertains to the arbitrariness of any spe- cific line of demarcation. It does not deny either the existence of pragmatics or the possibility of distinguishing it from semantics. It merely posits a gradation, such that notions which are indisputably semantic or pragmatic lie toward op- posite extremes of a scale, the status of those in the middle being mixed or indeterminate. This gradation is observed even in lexical semantics. At issue is whether a lexi- cal item’s meaning comprises everything speakers know about the type of entity designated (the encyclopedic view of linguistic semantics) or whether—as tradi- tionally assumed—its meaning is limited to a small, strictly delimited portion of this knowledge (the dictionary view). 15 In adopting the encyclopedic view, Cognitive Grammar is not claiming that there are no limits or that all knowledge is equally significant. For one thing, knowledge counts as being linguistic only to the extent that it is psychologically entrenched and conventional in a speech community, both of which are matters of degree. Moreover, specifications which are familiar and widely shared vary greatly in their degree of centrality, that is, the likelihood of their being activated when the lexical item is used. Some specifications are so central that they can hardly be suppressed, while the activation of others is variable depending on the context, and some are so peripheral that they are activated only in very special circumstances. On the encyclopedic view, a lexical item provides a particular way of accessing associated domains of knowledge. The access it affords is flexible and subject to contextual influence, but not at all random or unconstrained. Distinct (though not unrelated) is the issue of how many conventionally es- tablished meanings a lexical item has. As is usual in Cognitive Linguistics, Cog- nitive Grammar maintains that an expression used with any frequency is generally polysemous, having a number of different but related ‘‘senses.’’ 16 These senses arise through usage and are linked by categorizing relationships (elaboration and ex- tension) to form a network, usually centered on a prototype. They represent the conventional range of established usage, as well as whatever schemas are abstracted on the basis of commonalities inherent in sets of more specific senses. Since the same prototypical meaning can be extended in different directions, often forming chains of extensions, there may not be any schematic meaning that all the other senses instantiate. This contrasts with a monosemous account (e.g., Ruhl 1989), which claims that only a single abstract meaning need be posited, from which all specific uses are derived by interpretive principles. While a schematic meaning should always be sought, three basic considerations argue for its insufficiency. First, the mono- semous view is seemingly at odds with the privileged status of a category prototype, including its role in spawning extended and more schematic senses (both develop- 432 ronald w. langacker mentally and diachronically). Second, an all-subsuming schema is often too ab- stract to be clearly distinct from those of other categories or to capture what seems distinctive about a particular category. Third, a single sense—either a schema or a prototype—cannot represent everything speakers know about an expression’s conventional usage. Out of all the specific senses that could in principle be derived from a schema or a prototype by plausible interpretive mechanisms, only a small proportion are actually exploited and conventionally established. Knowing these is part of knowing how to speak a language. Hence they are part of the linguistic system, as defined in Cognitive Grammar. Valid questions have been raised about the ‘‘psychological reality’’ of a network as the representation of a complex category (e.g., Sandra and Rice 1995; Rice 1996). Though it undeniably has a certain utility, the network model is only a metaphor, whose appropriateness cannot be taken for granted. The discreteness inherent in the metaphor—suggesting a determinate number of nodes and a particular set of links—cannot necessarily be ascribed per se to the cognitive phenomena it models. An alternative metaphor, arguably less distorting, likens an expression’s range of meanings to a mountain range, which is continuous but very uneven due to rises, depressions, peaks, and valleys. Counting the senses of a lexical item is analogous to counting the peaks in a mountain range: how many there are depends on how salient they have to be before we count them; they appear discrete only if we ignore how they grade into one another at lower altitudes. The uncertainty often expe- rienced in determining which particular sense an expression instantiates on a given occasion is thus to be expected. In terms of the metaphor, such uses correspond to points in a valley lying between two peaks. Whether we assign such points to one peak, to the other, to both, or to neither is essentially arbitrary. How, in a conceptualist semantics, can a particular meaning (or sense) be characterized? What should beadopted as conceptual ‘‘primitives,’’ the basic elements from which more elaborate conceptions are constructed? An essential point, too often ignored, is that something can be ‘‘basic’’ in many different ways, some of them mutually contradictory. 17 From the Cognitive Grammar perspective, two kinds of basicness are especially important linguistically. Basic in one respect are conceptual archetypes, Gestalt conceptions of some complexity representing salient aspects of our everyday experience that are highly frequent and seemingly fundamental. Here are a few examples: a physical object, an object moving through space, the human face, the human body, a physical container and its contents, a whole and its parts, seeing something, holding something, handing something to someone, exerting force to effect a desired change, speaking, a face-to-face social encounter. Basic in another respect are certain minimal and maximally schematic notions that can be manifested in any domain of experience: point versus extension, contrast, boundary, change, continuity, contact, inclusion, separation, proximity, multiplicity, group, and so on. In view of their abstractness and minimality, they can be thought of as either sche- matic concepts or basic cognitive abilities. Instead of extension, contrast, and group, for instance, one can just as well speak of mental scanning, the ability to detect a contrast, and the capacity for grouping constitutive entities. 18 cognitive grammar 433 Conceptual archetypes and basic cognitive abilities have different roles in gram- mar. Being more specific and cognitively salient, the former tend to be adopted as category prototypes. Being highly abstract, the latter are possible candidates for the schematic description of a category (one valid for all instances). 19 A basic pro- posal of Cognitive Grammar is that certain fundamental and universal grammatical notions—including noun, verb, subject, object, and possessive—are semantically characterized at both levels. A noun, for example, is characterized prototypically in terms of the physical object archetype and schematically in terms of the cognitive abilities of grouping and reification (section 4). Developmentally, the abilities in question are initially manifested in prototypical instances, giving rise to the arche- type, and subsequently extended to other kinds of conceptions. Basic in yet another way are certain realms of experience not reducible to anything more fundamental. Among these basic domains are space, time, and do- mains associated with the various senses, such as color space (the range of color sensations we are capable of experiencing). 20 A basic domain is not itself a concept, but rather provides the experiential potential for conceptualization to occur. Min- imal concepts exploiting this potential include such notions as line, angle, cur- vature, focal colors, and temporal precedence. These, too, are basic in the sense of being incorporated in countless other conceptions, both simple and complex. Starting from these various sorts of basic elements, successively more elaborate conceptions can be constructed, with no upper bound on their ultimate com- plexity. A concept or conceptual complex of any size, at any level of conceptual organization, is called a nonbasic domain. 21 One can then make the general state- ment that a linguistic expression evokes a set of domains (basic and nonbasic) as the basis for its meaning. Collectively these domains are referred to as the expres- sion’s conceptual matrix. A domain representing any level of organization or de- gree of complexity can be part of an expression’s matrix and crucial to its semantic characterization. Thus, red evokes the basic domain of color space, and arm the archetypal conception of the human body (a nonbasic domain). For a term like castle, the pivotal domain consists of an elaborate body of knowledge pertaining to the rules and strategies of chess. The domains of a matrix are often multitudinous, representing facets of speakers’ encyclopedic knowledge of the entity designated. This entity (the ex- pression’s profile) has some manifestation in all the domains, which are not dis- jointed but overlapping. Among the domains for the count noun glass, for exam- ple, are the following: a specification of its typical shape (presupposing the basic domain of space); the conception of its typical orientation (incorporating the shape specification); its function as a container for liquid (involving shape and orientation, as well the notions liquid, spatial inclusion, potential motion, force, and constancy through time); its role in the process of drinking (including the container function, the conception of the human body, of grasping, of motion with the arm, and ingestion); a specification of its material (usually the substance glass); its typical size (easily held in one hand); and numerous other, more peripheral 434 ronald w. langacker conceptions (e.g., cost, washing, storage, possibility of breaking, position on a table during a meal, matching sets, method of manufacture). Obviously, this encyclopedic knowledge cannot all be accessed on every oc- casion. The likelihood of particular domains being activated (their degree of cen- trality) is part of an expression’s conventional semantic value. These default ex- pectations can, however, be adjusted and overridden by any number of contextual factors. The specific array of domains activated, and their degree of activation, may never be exactly the same in any two usage events. In this sense, even lexical meanings are anything but fixed and determinate. The domains activated provide an expression’s conceptual content. Its mean- ing, however, is not just the content evoked—equally important is how that content is construed. Construal is our multifaceted capacity to conceive and portray the same situation in alternate ways. The construal imposed on its content is intrinsic and essential to the meaning of every expression and every symbolic unit. The many aspects of construal fall in a number of general categories: specificity, prominence, perspective,anddynamicity (see also Verhagen, this volume, chapter 3). 22 In describing a situation, we can present it with any degree of precision and detail, depending on communicative needs and speaker objectives. Lexical items form hierarchies ranging from schematic to successively more specific character- izations, e.g., do ? act ? move ? run ? lope. Novel expressions of any size can likewise be arranged in such hierarchies. Naturally, an expression can be highly schematic in regard to certain facets of the situation while specifying others in fine- grained detail. Grammatical elements tend to be quite schematic in their content, their primary semantic contribution residing in the construal they impose. Something can be prominent (or salient) in many different ways, which need to be distinguished for linguistic purposes. 23 Two particular kinds of prominence prove especially important in grammar: profiling and trajector/landmark alignment. Within the array of content it evokes—its conceptual base—an expression des- ignates (i.e., refers to) a particular substructure. This is called its profile.Anexpres- sion’s profile is thus its conceptual referent and, as such, is prominent in the sense that the expression serves to single it out and focus attention on it. Some examples are sketched in figure 17.6 (the profile is drawn in bold). The base for hypotenuse is the conception of a right triangle (a nonbasic domain), and its profile is the side opposite the right angle. The overall configuration of an eye functions as the base for terms like iris and pupil, which profile different portions of it. In diagram (c), the dashed arrow stands for an experiential relationship, wherein a sentient creature entertains a positive mental attitude toward some other entity. The verb admire profiles this relationship, while the noun admirer designates just the sentient individual. 24 Ex- amples like these demonstrate that expressions with the same content can nonetheless differ in meaning by virtue of the profile they impose on it. This constitutes semantic evidence for adopting profiling as a descriptive construct. Either a thing or a relationship can be profiled, both terms being defined quite abstractly (section 4). When a relationship is profiled, we need to recognize a cognitive grammar 435 second kind of prominence pertaining to its participants. It is usual for one par- ticipant to be singled out as the entity conceived as being located, described, or evaluated. Called the trajector (tr), it can be characterized impressionistically as the primary focal participant (the primary ‘‘figure’’ within the profiled relation- ship). Often some other participant is made salient as a secondary focal participant, called a landmark (lm). The need for these descriptive constructs is shown by the existence of pairs of expressions that contrast semantically despite having the same conceptual base and profiling the same relationship. One example is before and after, diagrammed in figure 17.7. Each designates a relationship of temporal precedence between two events. Indeed, they profile precisely the same relation— referentially, a before relationship is also an after relationship. Their semantic contrast resides in whether the later event is invoked as a landmark for purposes of situating the earlier one, or conversely. Turning now to perspective, the most obvious aspect of construal is vantage point. A simple illustration is Come up into the attic versus Go up into the attic, which presuppose different speaker locations: in the attic and down below, respectively. Of course, the vantage point adopted for a particular purpose need not be the speaker’s actual one. Thus, in Joe said to come up into the attic, it is Joe’s vantage point that is used by the speaker as the basis for choosing come . Nor is vantage point limited to space and vision. Consider next year, used as a noun phrase. As shown in figure 17.8, it profiles the year immediately following the year containing a pre- supposed vantage point in time. Once again, this need not be the speaker’s actual temporal location (the time of speaking). In Joe believed that next year would be full of surprises, the vantage point adopted is the time when Joe entertained his belief. Closely related to vantage point is the extent to which a particular entity is subjectively or objectively construed. As used in Cognitive Grammar, these terms pertain to whether the entity functions as a subject or object of conception. A tacit conceptualizing presence, a locus of consciousness that is not itself conceived, is construed with maximal subjectivity. Conversely, something explicitly singled out Figure 17 . 6. Profiling 436 ronald w. langacker as the focus of attention is construed with maximal objectivity. In apprehending the meanings of linguistic expressions, the speaker and addressee function as sub- jects of conception; in that role they are subjectively construed and always remain implicit. An expression’s profile—defined as the entity it directs attention to—is construed quite objectively. This basic configuration was depicted in figure 17.1, where the profile is the ‘‘onstage’’ focus of attention, which the speaker and hearer view from their offstage vantage point, the ground. Of course, the viewing frame can be directed at the ground itself, so that some facet of it is put onstage and profiled. When the speaker and hearer are explicitly mentioned in this fashion, by pronouns like I and you , they function as objects of conception in addition to their tacit role as conceptualizing subjects. This ability to direct the viewing frame wherever we like, at any facet of our conceptual universe, is another aspect of perspective, referred to in Cognitive Gram- mar as scope. An expression’s scope is the extent of the conceptual content it evokes as the basis for its meaning, its ‘‘coverage’’ in active domains. This coverage is often less than exhaustive. For example, a central domain for next year is the conception of one year following another, in an endless sequence. To apprehend the expression’s meaning, however, the portion shown in figure 17.8 is all that needs to be consid- ered, and is thus its scope in this domain. Within the portion evoked, moreover, a particular region often stands out from the rest as being directly relevant for the purpose at hand. In such cases, a distinction is made between the maximal scope and the more restricted immediate scope, described metaphorically as the ‘‘onstage re- gion.’’ 25 As the general locus of attention, the immediate scope represents the portion of the overall content evoked that is potentially available for focused viewing. An expression’s profile is the specific focus of attention within its immediate scope. The conception of an eye, for instance, is the immediate scope on which iris and pu- pil impose alternate profiles (figure 17.6b), but since an eye is itself characterized as part of a face, their maximal scope includes the conception of a face (and more pe- ripherally, of a head and even the body as a whole). A final aspect of construal, dynamicity, pertains to how a conceptualization unfolds and develops through processing time. In this respect word order exerts a constant influence. Because the symbolizing elements occur in a certain sequence, the conceptual components they evoke must also be accessed in that sequence as one facet of the overall processing activity involved. Strictly speaking, semantic non- equivalence must even be recognized for pairs like She argued about religion with her dentist and She argued with her dentist about religion, reflecting the different orders in which the components symbolized by the prepositional phrases are incorporated in Figure 17.7. Trajector/landmark alignment cognitive grammar 437 the overall event conception. The semantic contrast is more apparent when alternate word orders embody different global strategies for presenting a scene, as in Adeadrat lay on the counter versus On the counter lay a dead rat. However, processing is not limited to a single word-by-word pass through an utterance, nor is word order always responsible for sequenced mental access. In The roof slopes steeply upward versus The roof slopes steeply downward, the contrasting directions of mental scanning are lexi- cally induced. Such scanning occurs even in highly abstract domains, as in this example: Forget about calculus—elementary algebra is already too difficult for him.The domain invoked is the conception of different subjects in mathematics, arranged in order of difficulty. The mental scanning proceeds along the scale from easier subjects to harder ones, the word already indicating that an excessively difficult subject is encountered sooner than would be expected (see Michaelis 1991). This last example illustrates the elaborate mental constructions we routinely create and invoke in apprehending linguistic expressions. Though traditionally ne- glected, ‘‘imaginative’’ phenomena—metaphor, metonymy, mental spaces, blending, fictivity—are not at all negligible but actually foundational for a viable linguistic semantics (Lakoff and Johnson 1980; Talmy 1996; Fauconnier 1997; Fauconnier and Turner 1998, 2002; Sweetser 1999; Langacker 1999d). 26 Researchonthesetopicshas clearly shown that the conceptualizations invoked as the meanings of expressions are not mechanically derivable from the meanings of their constitutive elements. Elab- orate layers of ‘‘meaning construction’’ commonly intervene between lexical mean- ings and the complex mental constructions required for a coherent understanding of the whole expression. While rules of semantic composition are certainly part of the story (in Cognitive Grammar, they constitute the semantic pole of constructional schemas), by themselves they are not in general sufficient to compute anything recognizable as linguistic meanings. For this reason linguistic semantics is viewed in Cognitive Grammar as exhibiting only partial compositionality. 4. Grammar Lexicon and grammar form a continuum consisting solely of assemblies of symbolic structures. An assembly can exhibit any degree of symbolic complexity and any degree of specificity (or conversely, schematicity) at the semantic and phonological poles. The assemblies usually recognized as lexical items can be characterized as fixed expressions, ‘‘fixed’’ referring to their status as conventional units, and ‘‘ex- Figure 17 . 8. Vantage point 438 ronald w. langacker pressions’’ to a substantial degree of semantic and especially phonological speci- ficity. 27 More schematic symbolic assemblies are traditionally viewed as belonging to grammar rather than lexicon. Grammatical markers are phonologically specific but schematic at the semantic pole. Schematic at both poles are the symbolic as- semblies representing grammatical classes and constructions. The schemas for constructions are symbolically complex, being merely the recurring commonalities abstracted from symbolically complex expressions. These constructional schemas function as templates exploited in the formation of novel expressions. The Cognitive Grammar claim that basic grammatical classes can be charac- terized semantically has to be properly understood. First, it applies to a limited set of categories that are useful in describing many languages (if not all) and numer- ous phenomena in a single language. The classes in question, starting from the positive end of a scale defined by universality and susceptibility to uniform con- ceptual characterization, include noun and verb, their major subclasses (e.g., count and mass), adjective, adverb, and adposition. At the other end of the scale are idio- syncratic classes reflecting a single language-specific phenomenon (e.g., the class of verbs instantiating a particular minor pattern of past-tense formation). Semantically the members of such a class may be totally arbitrary. 28 Second, reference in Cognitive Grammar to traditional parts of speech is selective and qualified (Langacker 1987a: section 6.3.3). The traditional scheme is highly problematic, and of the standard classes only noun and verb correspond to fundamental Cognitive Grammar cate- gories. To some extent the others do, however, have a semantic rationale, which Cog- nitive Grammar notions allow one to explicate. But in each case a new conceptual description is offered which defines the class in its own, nonstandard way. 29 The most fundamental categories are noun and verb. Their semantic charac- terizations are polar opposites, at both the prototype and the schematic levels. It is widely recognized that the respective prototypes for the noun and verb categories are two conceptual archetypes: a physical object and a force-dynamic event (Talmy 1988), specifically an Agent-Patient interaction (Hopper and Thompson 1980; Rice 1987a). More controversial is the claim that each category has a schematic de- scription, that is, one valid for all instances. Standard arguments against this pos- sibility ignore some crucial factors. For one thing, an expression’s grammatical class is not determined by its overall conceptual content, but rather by the nature of its profile. Thus, despite their identical content, admire is a verb and admirer a noun due to the alternate profiles they impose (figure 17.6c). More generally, stan- dard arguments presuppose an objectivist view of meaning that ignores the im- portance of construal. Not even considered, for instance, is the possibility that nominalizing a verb (e.g., arrive " arrival) might involve a process of conceptual reification so that the verb and noun are semantically distinct. Finally, the only definitions usually contemplated are conceptual archetypes (e.g., person, object, event, property, location), whereas a schematic characterization would have to be considerably more abstract. As schematic definitions, Cognitive Grammar proposes that a noun profiles a thing, while a verb profiles a process. 30 These notions have only tenuous intrinsic cognitive grammar 439 . of the brain), it is not autonomous or encapsulated, and it is not itself the object of contemplation. 13 Conceptuali- zation is always the conceptualization of something, a facet of either the. established. Knowing these is part of knowing how to speak a language. Hence they are part of the linguistic system, as defined in Cognitive Grammar. Valid questions have been raised about the ‘‘psychological. like castle, the pivotal domain consists of an elaborate body of knowledge pertaining to the rules and strategies of chess. The domains of a matrix are often multitudinous, representing facets of speakers’