Jump to ContentJump to Main Navigation
Cognitive GrammarA Basic Introduction$

Ronald Langacker

Print publication date: 2008

Print ISBN-13: 9780195331967

Published to Oxford Scholarship Online: May 2008

DOI: 10.1093/acprof:oso/9780195331967.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2020. All Rights Reserved. An individual user may print out a PDF of a single chapter of a monograph in OSO for personal use.  Subscriber: null; date: 29 January 2020

Engaging the World

Engaging the World

(p.500) 14 Engaging the World
Cognitive Grammar

Ronald W. Langacker

Oxford University Press

Abstract and Keywords

Because it unfolds through time, conceptualization (and hence linguistic meaning) is inherently dynamic. There are numerous natural paths that it tends to follow, and which tend to coalign in linguistic structure. In one kind of path, a salient reference point provides mental access to a target. Certain basic grammatical phenomena are analyzed in terms of reference point relationships, including possessives, pronominal anaphora, topic constructions, and trajector/ landmark organization (subject and object). A subject differs from a discourse topic by being structurally internal to a clause and conceptually intrinsic to the clausal process. Trajector and landmark are characterized dynamically as the first and second reference points evoked in building up to the full conception of a profiled relationship. This explains their general grammatical accessibility as well as their role in certain specific constructions. The mental world we construct is grounded in our experience as creatures with bodies who engage in motor and sensory interactions (embodiment). In constructing it, we transcend direct experience through abstraction, conceptual integration, and subjectification: the application of mental operations immanent in certain conceptions to situations for which their occurrence is extrinsic. Examples include fictive motion, fictive change, and the covert invocation of imagined scenarios. Mental simulation is a fundamental aspect of conception and linguistic meaning. Subjectification is an important factor in grammaticization (the evolution of grammatical elements from lexical sources). Many grammatical notions are subjective counterparts of basic aspects of everyday experience. Grammar reflects the means of disengagement through which we transcend immediate experience and construct our mental world. It is thus a key to conceptual analysis.

Keywords:   abstraction, accessibility, anaphora, conceptual integration, dynamicity, embodiment, fictive change, fictive motion, grammaticization, landmark, mental simulation, natural path, object, possessive, pronominal anaphora, reference point, subject, subjectification, topic, trajectory

Some have questioned whether a “cognitive” approach to language can accommodate either its social function (as a means of interaction) or its referential function (as a means of describing the world). These concerns are unfounded. They stem from the erroneous idea that what goes on inside the skull is isolated from everything outside it, including other minds. But this is simply not so. In the sense that cognition resides in activity of the brain, it does indeed take place inside the skull. The brain, however, is the nexus of a nervous system that runs all through the body, connecting with the sensory and motor organs through which we perceive and act on the world. Neither is the brain's activity isolated from other minds. An essential aspect of cognition is our awareness of other people and our recognition that they, too, are cognitive agents. We are quite adept at reading their intentions, as well as imagining the nature of their mental experience. Thus cognition, far from being insulated from the world and the other people in it, is our primary means of engaging them.

In this final chapter, I deal with two related properties of cognition, both essential to understanding grammar. The first is temporal sequencing. As neurological activity, cognition necessarily takes place through time. Precisely how it does so—the time course of conception—is often critical. A second key property is that cognition consists of far more than sensory and motor interactions. What happens in the social, cultural, and imaginative spheres is as real and important to us as physical occurrences. Moreover, we are not just concerned with immediate reality. We further engage the world through memory, anticipation, prediction, generalization, and contemplation of alternatives. Involving many kinds and levels of mental construction, these phenomena transcend immediate bodily experience. However, they also prove to be grounded in it.

14.1 Dynamicity

Because it occurs through time, conceptualization—even the activation of established concepts—is inherently dynamic. In cases of any complexity, different facets of the (p.501) total conception are activated at each successive instant. This may or may not result in their all being active simultaneously, but either way the sequence of activation is part of the overall mental experience. It is consequential even when the time scale is small enough that it stays below the threshold of conscious awareness.1

14.1.1 Paths of Mental Access

Some conceptions lead to others. If you think of the letter A, you are likely to think of B. This facilitates the activation of C, which in turn leads to D, and so on. One conception can lead to another due to an established connection between them, as with the alphabet, or just by virtue of creating the conditions for its emergence. For example, imagining a hypothetical circumstance makes it possible to imagine what would happen in it: if you won the lottery, you could quit your job. Connections of this sort tend to be asymmetrical, conception moving more readily in one direction than the other. Learning the alphabet, so that you can quickly run through the progression from A to Z, does not automatically give you the ability to smoothly recite it backward. It does, though, provide the means to figure out the opposite sequence. For example, to determine what “follows” T in the reverse ordering, I can recite a string containing it (… Q > R > S > T …) and thereby observe that the adjacent letter is S.

A series of conceptions where each leads readily to the next is called a natural path. Quite a number of natural paths have a significant role in language structure. An obvious one is the order of presentation, where words are produced or encountered in a certain sequence. Another is a chain of elaborative relationships—that is, a series of elements, each of which contains a schematic elaboration site specified by the next. One such case is a chain of complement clauses, as in (1), where the clausal landmarks function as e-sites (fig. 12.4).

(1) [Alice said] [that Bill believes] [that Cindy claims] [that Doris swallowed a spider].

The paths just mentioned involve the linguistic expression itself. Others inhere in the conception expressed. At the conceptual level, two very basic natural paths—closely correlated in our experience—are the order of event occurrence (where X precedes Y) and the sequence of causation (where X induces Y). Also very basic are paths consisting in a series of whole-part relations, such as body > arm > hand > finger > knuckle. At each step in such a chain, the conception of the whole provides the context (immediate scope) for conceiving of the part (fig. 3.3). Additionally, any kind of scale (like cost, weight, or temperature) is a natural path defined by its successive values.

The origin of a natural path is a starting point. It should not be surprising that the starting point is often either the conceptualizer or something to which C has immediate access. The default conceptualizer (C0) is the current speaker, who anchors a (p.502) number of linguistically relevant paths. One such path is a chain of conceptualizers (C0 > C1 > C2 > C3 …), each of whom apprehends the next and to some extent simulates their mental experience. We have such a chain in (1): S > Alice > Bill > Cindy > Doris. A path of this sort is also a path of access to successively embedded mental spaces. The starting space is the speaker's conception of reality. Successively embedded within it are the spaces representing the content of Alice's statement, Bill's belief, and Cindy's claim (fig. 12.4). Other paths correspond to “distance” from the speaker in various dimensions: speaker > hearer > other; human > animate > inanimate; concrete > abstract; actual > virtual; given > new. They reflect the speaker's special status as initial conceptualizer and as an actual person always accessible in the discourse.

There is a tendency for starting points to coincide and natural paths to correlate with one another. In (1), at least four natural paths coalign: order of presentation; sequence of elaboration; chain of conceptualizers; and successive embedding of mental spaces. Moreover, all of them start with Alice (metonymically speaking): she is mentioned first, her clause provides the first elaboration site, she is the first onstage conceptualizer, and hers is the first mental space introduced.2 Being overt and ever-present, word order tends to correlate with numerous paths of mental access. The coalignment of paths facilitates processing. Thus (2)(a), where the order of locatives corresponds to a natural path of search, is easier to process than (2)(b), where the order is random. And in (3) to (5), the (a) examples flow more smoothly than the (b) examples because the order of presentation follows a natural sequence.

(2) (a) The article is in today's paper, in the sports section, on the last page, near the bottom.

(b) ??The article is in the sports section, near the bottom, in today's paper, on the last page.

(3) (a) The distance from San Diego to Los Angeles is about 120 miles.

(b) ?The distance to Los Angeles from San Diego is about 120 miles.

(4) (a) The rainy season begins in January and ends in March.

(b) ??The rainy season ends in March and begins in January.

(5) (a) In the evening stores are open between 7 and 10.

(b) ?*In the evening stores are open between 10 and 7.

We can plausibly assume that any conception of ordering or directionality requires seriality at some level of cognitive processing. Additional processing effort is thus implied when a conception incorporates natural paths requiring that the same elements be accessed in opposite sequences. Consider (3)(b), where from and to impose a directional construal. We apprehend its directionality by tracing a mental path from San Diego to Los Angeles—at some level of processing, we mentally (p.503) access the cities in that order. But the order of presentation leads us to invoke them in the opposite sequence: to Los Angeles from San Diego. To grasp the import of this expression, we must therefore reconceptualize the path (§3.4.2), construing it in the manner specified by the prepositions.

The conception of a scale has an inherent directionality, normally depicted by an arrow. A scalar adjective, such as long, heavy, angry, or intelligent, indicates the extent to which the trajector manifests a defining property. The origin of a scale—its starting point—corresponds to the lack of anything noteworthy: to either a default value or the property's total absence.3 Successive values on the scale, representing degrees of departure from this baseline, thus constitute a natural path of mental access. Although we have the ability to move along a scale in either direction, there is some inclination for this scanning to observe its inherent directionality. Suppose we want to describe the relative intelligence of two individuals. If they are equal in discourse salience (neither being the current topic), we would tend to say X is more intelligent than Y, rather than the logically equivalent Y is less intelligent than X. The reason is that more induces scanning that conforms to the scale's directionality, while the scanning induced by less runs counter to it.

The preference for coalignment emerges quite clearly when comparison is conceived metaphorically in terms of motion. We can say that X surpasses Y in intelligence, or X is beyond Y in intelligence, where X moves along the scale in a positive direction. But we do not find expressions implying motion in the opposite direction (e.g. *Y subpasses X in intelligence).4 There is an alternative where Y is the mover—namely, Y falls short of X in intelligence. Here, though, Y moves in the positive direction, being construed metaphorically as a projectile that fails to reach its target. In another alternative, X and Y are both conceived as moving, so either can function as subject: X is ahead of Y in intelligence; Y is behind X in intelligence. In this case, the motion of both X and Y conforms to the scale's inherent directionality.

The processing efficiency achieved by coaligning natural paths is reflected linguistically in different ways. A minimal result is for one expression to be slightly favored over another (e.g. X is more intelligent than Y > Y is less intelligent than X). In more egregious cases, nonalignment results in diminished acceptability, exemplified by the (b) examples in (3) to (5). Processing efficiency also tends to correlate with coding efficiency: that is, simpler forms. It is no accident, for example, that more alternates with the suffix -er (longer, brighter, easier, etc.), while less has no reduced alternative. Likewise, an active (X broke Y) is formally simpler than a passive (Y was broken by X), where the order of presentation (Y > X) runs counter to the action chain (X ↠ Y). With the coalignment of natural paths, coding efficiency can also be achieved through greater reliance on iconicity. For instance, (6)(a) strongly (p.504) encourages the inference that events occurred in the order stated. If the opposite order is intended, it has to be made explicit, so (6)(b) is more complex.

(6) (a) She criticized him. He went to his room and kicked his dog.

(b) She criticized him. He had gone to his room after kicking his dog.

Paths of mental access are many, varied, and linguistically important. When the elements of a path are discrete and have a certain amount of salience, it can be described as a chain of reference point relationships. Depicted in figure 14.1(a), a reference point relation consists in the mental progression from a reference point (R) to a target (T) accessed through it. The set of entities accessible via R (each a potential target) constitute its dominion (D). Figure 14.1(b) shows a chain of such relations, where each successive target (Ti) functions in turn as the next reference point (Ri+1).

Being a matter of sequential mental access, reference point relationships are intrinsically dynamic but have no intrinsic content. They represent a general aspect of conceptual organization that proves essential to the semantic characterization of some basic linguistic phenomena. One of these is metonymy, in which an expression's usual referent provides mental access to the entity it is actually construed as designating. In (7), for example, Vietnam does not refer to the country per se, but rather to a war that was fought there. Coherence demands that the subject designate an event, and knowledge of recent history leaves little doubt as to which event is intended. Owing to their strong association, naming the country readily calls the war to mind.5

(7) Vietnam marked a turning point in American history.

Reference point relationships also figure in nominal compounds, where two nouns combine to form a composite expression that is also a noun: jar lid, basketball net, sheep dog, baseball glove, bicycle seat, axe handle, window shade, fishing pole, book

Engaging the World

Figure 14.1

(p.505) cover, fingernail, trout stream, pencil sharpener, tomato worm, tree root, wine bottle, etc. The basic pattern is for the first noun to evoke a range of knowledge with respect to which the second is interpreted. There are many kinds of nets, for example, but only one accessible via basketball. By virtue of this association, basketball net is understood as referring to a net of this type (as opposed to a fishing net, butterfly net, mosquito net, or hair net). The interpretation is usually based on familiar scenarios. Sheep dog means what it does due to the cultural model of dogs herding sheep—it would mean something different if the standard model involved dogs being docile like sheep or eating them. Naturally, a compound can also derive its meaning from the context or any situation we might imagine, no matter how outlandish. A given compound can therefore be interpreted in different ways. While an airplane diaper would probably be understood as a diaper carried on an airplane for infant emergencies, it would also be appropriate for a massive piece of cloth wrapped around a plane to soak up leaking fuel.

14.1.2 Possession

Reference point relations are the key to understanding the linguistic phenomenon known as possession. What (you might ask) is the meaning of a possessive marker like English 's, or the import of a possessive construction like Zelda's quilt? The term notwithstanding, possession encompasses far more than relationships of owning or possessing. These are at best prototypical, along with kinship and whole-part relations (my sister, the rat's tail). Indeed, possessive expressions are used for an extremely diverse array of relationships: Zelda's drink, the kitten's fleas, our bus, her trial, the store's location, my headache, Sean's attitude, their average intelligence, the diamond's value, his predicament, the year's top story, the photo's glossy finish, our existence, the bullet's trajectory, Lincoln's assassination, and so on. Thus a general characterization can hardly be based on specific conceptual content. What, then, do possessives all have in common? Various considerations point to a reference point relationship being the shared feature (Langacker 1995; Taylor 1996; GC: ch. 6). A schematic description, valid for all instances, is simply that the possessor functions as a reference point providing mental access to the entity possessed, its target.

As noted previously (§3.4.2), a reference point characterization is sufficiently abstract to accommodate the full range of possessive expressions. Since it does not specify any particular content, it is compatible with any kind of conceived relation between possessor and possessed. But there are also limitations, and these too are accommodated. For the most part, possessive relationships are irreversible; we would not, for example, say *the quilt's Zelda, *the tail's rat, *the value's diamond, or *the trajectory's bullet. This irreversibility reflects the inherent asymmetry of reference point relationships—the path of mental access leads from R to T rather than in the opposite direction. The direction of access correlates with certain natural paths, including whole > part, concrete > abstract, and human > animate > inanimate. If two entities are equivalent in this respect, it may be possible for either to function as possessor, as in the doctor's lawyer and the lawyer's doctor.6

(p.506) In a usage-based approach, the schematic description of possessives does not stand alone. The schema coexists with any number of conventional instantiations characterized at different levels of specificity (including many expressions with unit status). Within this array of established uses, some have good claim to being prototypical, among them ownership, kinship, and whole-part relations.7 These are all clear examples of reference point organization. By their very nature, kin terms invoke a reference individual (fig. 3.6)—one is not a sister, a grandson, or an aunt in any absolute sense but only in relation to a certain person. Likewise, a part is only apprehended as such in relation to a larger whole (fig. 3.2). Though we might well recognize a tail in isolation, we can only identify it as a tail by invoking its position within the overall configuration of a body. In the case of ownership, reference point organization reflects the cognitive salience of people vis-à-vis nonhuman entities, as well as the cultural model whereby every person owns, controls, or has privileged access to a certain set of entities (which we call “possessions”). There are many fewer people than possessions, and we are much more likely to know them as individuals. To identify possessions with reference to their owners (Zelda's quilt) is thus a more efficient strategy than the opposite.

Possessives illustrate a general proposal of CG (§2.1.2): namely, that fundamental and universal grammatical notions have semantic characterizations at both the prototype and schema levels. Functioning as their prototypical values are conceptual archetypes that reflect basic aspects of everyday experience (e.g. physical object, in the case of nouns). The schematic meanings—valid for all instances—consist in basic cognitive abilities not tied to any specific conceptual content (e.g. grouping and reification). The abilities are initially manifested in the corresponding archetypes, providing the basis for their apprehension, and are later extended to other, less central cases.

In possession, the archetypes of ownership, kinship, and whole-part relations function as prototypical values. The reference point ability provides the schematic meaning. It is due to this ability that we are able to conceptualize ownership, kinship, and whole-part relations in the first place—that is, a reference point relationship is inherent in their conception. These archetypes have a basic directionality, involving a mental progression from R to T. We apprehend a kinship relation, such as cousin, by tracing a mental path from reference individual, through linking relatives, to profiled target: R —> parent —> sibling —> child (T). Likewise, the conception of a part implies sequential access to the levels of a whole-part hierarchy, as body (R) —> arm —> hand —> finger —> knuckle (T). With ownership, a key notion is that of the possessor controlling the possessed: R ↠ T. The asymmetry of this relationship, the directionality in the flow of influence, is apprehended through seriality in the evocation of participants, whereby we first conceptualize R and then T. Their content being (p.507) very different, what the possessive archetypes all share is precisely the invocation of R to mentally access T. Less typical uses of possessives result from applying this same ability to other kinds of circumstances.

Reference point relationships are a special case of sequential mental access, where R and T are discrete and salient enough to be individually recognized. As an aspect of conceptual organization, they are independent of any specific linguistic phenomena. They can thus be exploited linguistically in different ways, each of which imposes its own construal on the content so organized. The intrinsic salience of R and T is therefore not to be identified with the kinds of prominence imposed by grammatical constructions. The same reference point relationship can be reflected in multiple expressions involving alternate choices of profiling and trajector/landmark alignment.

There are, for example, both nominal and clausal possessive constructions, which differ in these respects. We have thus far focused on nominal possessives, such as Zelda's quilt. The same relationship can often be expressed in clausal form: Zelda has a quilt.8 The basic difference between them is that the former profiles the entity possessed, whereas the latter designates the possessive relation itself. The verb have is highly polysemous and appears in a wide array of constructions (Brugman 1988). In simple clauses like Zelda has a quilt, it profiles an imperfective process with two focal participants. The range of senses it exhibits in this construction includes the possessive archetypes, established extensions from the prototype, and schemas at different levels of abstraction. Its most schematic value, devoid of any specific conceptual content, is simply that of a reference point relationship, with R focused as trajector and T as landmark. In forming a clause, these are specified by full nominals in accordance with the general subject and object constructions.

Nominal and clausal possession have different functions. Zelda has a quilt profiles a possessive relationship, which it introduces in the discourse. By contrast, Zelda's quilt presupposes this relationship and invokes it to identify a nominal referent. This identifying function qualifies Zelda's as a nominal grounding element. The noun quilt specifies a thing type of which there are many instances. The possessive marker 's evokes a reference point relationship and, more generally, the notion of there being many possible reference points, each with its own dominion. This provides the basis for distinguishing instances of a type from one another: one way to single out a quilt, distinguishing it from all other things of this sort, is to identify it as the instance occurring in the dominion of a particular individual. Of course, that individual must itself be identified. This is the function of the possessor nominal, in this case Zelda, which bears the possessive marking. Zelda's quilt is thus identified, for discourse purposes, as the one accessible via Zelda (whether through ownership or some other relation evident in the context).

This construction is sketched in figure 14.2. The conceptual base for 's is a reference point relationship. As a schematized grounding element, 's profiles the grounded entity, namely T, rather than the grounding relationship (§9.3.1). The schematic reference point functions as an e-site at the first level of composition. Its elaboration by Zelda produces the specific grounding element Zelda's, in which the (p.508)

Engaging the World

Figure 14.2

profiled target remains schematic.9 At the higher level of organization, quilt elaborates T to specify the type of the grounded instance. The integration of Zelda's and quilt exhibits a characteristic property of grounding constructions (fig. 9.4): since their profiles correspond, neither stands out as profile determinant.

14.1.3 Pronominal Anaphora

Possession is one example of reference point relationships playing a role in the identification of nominal referents. They also help determine use of the definite article, which implies that only one instance of the specified type is accessible in the current discourse space. They do so by providing a scope of interpretation where a profiled instance has the requisite contextual uniqueness. In (8), the first sentence introduces a quilt into the discourse and establishes it as a topic. The following sentence is thus interpreted with respect to it. In particular, the quilt serves as reference point for interpreting the pattern and the colors, its dominion constituting the immediate scope for this purpose. And since a typical quilt has just one pattern and one set of colors, within this scope the nominal referents are the only instances of their types.

(8) I really like Zelda's quilt. The pattern is neat and the colors are striking.

Pronominal anaphora is another reference point phenomenon, with R being the antecedent nominal and T the pronoun. T is mentally accessed via R in the sense that the antecedent determines the pronoun's reference. In (9), for example, it is interpreted via Zelda's quilt. Like the pattern and the colors in (8), it is taken as referring (p.509) to an element of the quilt's dominion: as something associated with the quilt (hence mentally accessible through it). The special property of a pronoun is that this element is identified as the reference point itself—of all the elements in its dominion, R itself is the easiest to access. So with Zelda's quilt as the antecedent nominal, it is taken as referring to the quilt. The two are said to be coreferential.

(9) Zelda's quilt is beautiful. Everyone likes it.

A reference point tends to function as such in multiple respects, pertaining to different aspects of linguistic organization. In (8) and (9), R gives access to T in terms of both content and expression. With respect to content, R's dominion comprises the entities mentally associated with R through either established knowledge or a conceived situation in which they figure. Our standard knowledge of quilts is such that Zelda's quilt primes us to think about its pattern and its colors, not to mention the quilt itself. At the level of expression, R's dominion consists of that portion of the ongoing discourse within which R is sufficiently prominent to impose itself as the basis for interpretation. We see in (8) and (9) that a nominal in one sentence is able to exert its influence at least through the following sentence.

In regard to pronominal anaphora, the extent of a reference point's dominion is a classic problem of great complexity (Langacker 1969; Reinhart 1983). Its solution in CG (van Hoek 1995, 1997) can be sketched here only in the briefest terms. Stated most generally, the likelihood of a nominal being invoked as reference point depends on its prominence, and the likelihood of an element being included in its dominion depends on the closeness of their conceptual connection. Among the factors contributing to a nominal's prominence are profiling, trajector status, discourse salience, and role as conceptualizer. In (9), Zelda's quilt is highly prominent due to being trajector of the process profiled by the sentence containing it, which further establishes it as a discourse topic. Thus it is readily invoked as antecedent for a pronoun in the sentence that directly follows. The topic relationship is one conceptual connection between Zelda's quilt and the elements of the latter sentence. They are further connected through emergence of a coherent overall conception, in which the beauty of the quilt is taken as being responsible for everyone liking it. Hence the pronoun it falls within the dominion of Zelda's quilt and will almost certainly be interpreted as coreferential with it.

The closest conceptual connections hold among the elements of a single sentence, as reflected in its grammatical organization. Perhaps the strongest is the connection between the trajector and the landmark of a profiled relationship. This is the prototypical configuration for using a reflexive pronoun, as in (10), where coreference of the president and himself is the only option. But since every grammatical link represents a conceptual connection, coreference is generally possible even between elements whose grammatical association is only indirect. Thus in (10) the possessor pronoun his can also be taken as referring to the president.10 Their connection is effected through a chain of grammatical relations: private office specifies the target of his; his private (p.510) office is the landmark of in; in his private office modifies the clausal process; and the president elaborates the trajector of that process. Despite this indirectness, the connection is clear, and due to its prominence the subject functions as antecedent. Here, though, it is not the only option—in the proper discourse context, his could be taken as referring to another individual previously established as the topic of discussion.11

(10) In his private office the president was admiring himself.

Usually the positions of a pronoun and its antecedent cannot be happily reversed. With coreference intended, for example, (11)(a) is at best rather awkward (we tend to interpret he as referring to someone other than the president) and (11)(b) is completely unacceptable:

(11) (a) ??In the president's private office he was admiring himself.

(b) *In his private office himself was admiring the president.

These asymmetries reflect the inherent directionality of anaphoric relationships. Because it provides the basis for interpreting the pronoun, the antecedent nominal has conceptual priority with respect to it. Consequently, the felicity of anaphoric relationships depends on the antecedent preceding the pronoun along natural paths of mental access. The most obvious path is order of presentation—normally the antecedent precedes the pronoun in the flow of speech. The opposite order is often problematic. In contrast to (9), the following is very marginal (assuming that the quilt has not yet been established in the CDS):

(12) ??It's beautiful. Everyone likes Zelda's quilt.

But order of presentation is not the only relevant path, or even the most important. Thus a pronoun sometimes does precede its antecedent, as in (10), where his precedes the president. In such cases, the antecedent has conceptual priority with respect to other natural paths.

One relevant path is a chain of elaborative relationships, notably with complement clauses (as in (1) ). Another is the natural path of access from a conceptualizer to the conception entertained. In (13)(a), the antecedent precedes the pronoun with respect to both these paths, as well as order of presentation: the president is in the matrix clause, he in the complement; the president functions as conceptualizer in regard to the complement; and the president comes before he in the temporal sequence. This represents the optimal configuration for an anaphoric relationship, since all three paths coalign with its inherent directionality. Conversely, in (13)(b) we observe the worst configuration: the pronoun precedes its supposed antecedent along all three natural paths.12 (p.511)

(13) (a) It never bothers the president [that he lies].

(b) *It never bothers him [that the president lies].

Other examples show that all three paths are relevant for anaphora. Note first that (14)(a) is fully acceptable even though the pronoun precedes the antecedent in the temporal sequence, while (14)(b) is marginal (assuming coreference) even though the pronoun follows. Evidently this factor is outweighed by the other two: despite the effect of word order, the more acceptable sentence is the one in which the antecedent is in the matrix clause and functions as conceptualizer with respect to the complement.

(14) (a) The fact [that he lies] never bothers the president.

(b) ??The fact [that the president lies] never bothers him.

While these latter factors often work together, each makes its own contribution to sanctioning anaphoric relationships. How might one demonstrate that path of conceptualization has an effect? We can see this by comparing the sentences in (14) to those in (15), which are parallel except that the antecedent (the book) is not a conceptualizer. In this case the sentences are equally acceptable. The relative infelicity of (14)(b) must therefore be ascribed to the pronoun's role as conceptualizer for the clause containing its antecedent.

(15) (a) The fact [that it's full of lies] doesn't diminish the book.

(b) The fact [that the book is full of lies] doesn't diminish it.

Likewise, we can assess the role of a complement chain by comparing sentences with complements to analogous sentences with modifiers. In (16)(a)–(b) the clause introduced by that is a preposed object complement (the landmark of realize), while in (16)(c)–(d) the when-clause is adverbial. The crucial examples are (b) and (d). In the former, the president resists interpretation as the antecedent of he even though it precedes it in the linear sequence. This is due in part to he being a conceptualizer with respect to the complement. But only in part, as shown by comparison with (d). Here too he functions as conceptualizer in regard to the preposed clause, yet coreference is unproblematic. The difference is that the complement clause in (b) bears an elaborative relationship to the matrix predicate, whereas the adverbial clause in (d) does not. It is only in the former that realize provides an elaborative path connecting the subject pronoun to the clause containing its antecedent.

(16) (a) [That he is a liar] the president certainly realizes.

(b) ??[That the president is a liar] he certainly realizes.

(c) [When he is lying] the president realizes it.

(d) [When the president is lying] he realizes it.

(p.512) Pronominal anaphora has often been approached from a purely syntactic standpoint. Attempts to describe the position of a pronoun vis-à-vis its antecedent have tended to focus on formal properties, especially linear order and grammatical constituency. But while these are not irrelevant, a comprehensive description has to be based on dynamic conceptualization. In anaphoric relationships, the antecedent nominal provides mental access to the pronoun in the sense of determining its reference. Whether it can do so—whether the pronoun falls in its dominion—depends on its prominence and the closeness of their conceptual connection. These in turn reflect the complex interaction of numerous factors, each involving precedence along a natural path. Two such factors deserve to be looked at in greater depth: the status of a nominal as a topic or as a subject.

14.1.4 Topic vs. Subject

Topic and subject have an important role in pronominal anaphora. A nominal is readily invoked as antecedent for a pronoun in the stretch of discourse over which it has topic status. For example, this computer antecedes it in (17)(a), an explicit topic construction. An explicit topic has a strong tendency to be invoked as reference point. We observe in sentence (b) that the positions of the pronoun and antecedent cannot be reversed: the topic nominal can only function as R, not as T, in the anaphoric relationship. Indeed, we note in (c) that a nominal in the pronoun's position cannot escape the topic's influence. If coreferential with the topic, it has to be expressed as a pronoun—repeating the full nominal (this computer) wrongly implies that its referent would not otherwise be apparent.

(17) (a) This computer, I just can't get it to work.

(b) *It, I just can't get this computer to work.

(c) *This computer, I just can't get this computer to work.

In similar fashion, a subject tends strongly to be invoked as reference point with respect to all other clausal elements.13 The effect of subject status is shown by the data in (18). In sentence (a) the anaphoric relationship is optimal because the antecedent Harvey not only precedes the pronoun but is also the clausal subject. Conversely, sentence (b)—where a subject pronoun precedes its antecedent—is flatly ungrammatical. The difference is not solely due to linear order, which proves to be a secondary factor. We observe this in (c) and (d), where the initial element in the anaphoric relationship is not the clausal subject but rather its possessor. In this case both sentences are acceptable, though (d) is somewhat marginal because the pronoun comes first. Still, its relative acceptability indicates that the pronoun's status as subject is primarily responsible for the ungrammaticality of (b). (p.513)

(18) (a) Harvey resembles his dog.

(b) *He resembles Harvey's dog.

(c) Harvey's dog resembles him.

(d) ?His dog resembles Harvey.

Their role in pronominal anaphora suggests that topic and subject are themselves reference point phenomena. As a reference point, a topic or a subject provides mental access to the elements of its dominion. Should one of these elements be a pronoun, the topic or subject will tend to be invoked as the basis for its interpretation. This reflects the inclination—a matter of cognitive efficiency—for natural paths to coalign and for reference points to serve as such in multiple respects. If a nominal and a pronoun function as R and T in regard to a topic or subject relation, they will also tend to do so for anaphoric purposes, with the same alignment.

How, then, can a topic or a subject be characterized? The two notions are closely associated grammatically and sometimes hard to distinguish. It happens, moreover, that topic constructions develop historically into subject constructions.14 Given their evident affinity, and their common description as reference point phenomena, what precisely is the difference?

Let us start with topics. As with many fundamental notions, it is hard to find a verbal definition that does not seem hopelessly vague. Almost invariably, linguists merely say that the topic of a sentence is “what the sentence is about”. This is quite compatible with its CG characterization, which does, however, have the advantage of relating this notion of “aboutness” to a general and basic feature of conceptual structure. Topics represent a particular kind of reference point organization, distinguished from others by the nature of the target.15 In contrast to possessives and pronominal anaphora, where the target is a thing, for topics the target is a proposition (P), i.e. a grounded process. The basic configuration is thus as shown in figure 14.3.

A topic's dominion is the range of associated knowledge—a set of propositions in which it has some role and which can thus be accessed through it. The import of

Engaging the World

Figure 14.3

(p.514) a topic relationship is that the target proposition belongs to this body of knowledge. P is an integral part of D, whether it is already established there or has to be introduced. For both the speaker and the hearer, D provides the context for apprehending P and interpreting it. A proposition can only be interpreted in R's dominion if R somehow figures in its content. The small circle in figure 14.3 represents R's manifestation in P: that is, the element of P it corresponds to. This is called the pivot.16 While the pivot may have any role in P, it tends to be a focal participant, as in (19)(a)–(b). And while it is often expressed overtly by a pronoun, it may also remain implicit. The pivots in (19)(c)–(d) are quite peripheral (at least in terms of grammar), and in (d) the pivot is left unexpressed.

(19) (a) That mural, it's really starting to depress me.

(b) That mural, I like it less and less every day.

(c) That mural, we would be better off if someone sprayed graffiti all over it.

(d) That mural, we never should have given permission.

As in the case of anaphora, a topic has a dominion in regard to both content and expression. At the level of expression, R's dominion is the stretch of discourse during which it enjoys topic status. Minimally this is just a clause, but it can also be a complex sentence or a discourse of any length. In (20), for example, the first sentence introduces the topic of writing a dissertation, and the succeeding sentences are all interpreted with respect to it. There is clearly the potential for the disquisition to continue indefinitely, until someone “changes the topic”.

(20) It's really hard to write a dissertation. You have to find a subject. Then you have to come up with some ideas and do lots of preliminary analyses. When you do the background reading, you find that most of those ideas have already been proposed and rejected. So you have to work for a number of years before anything viable starts to take shape. You have to worry about continued financial support. Then you have to satisfy five committee members with mutually incompatible notions about what you should be doing. You have to go through about seven drafts. Then

The topic in (20) bears no special marking but is simply a verbal object introduced via the regular object construction. Nor does anything explicitly indicate that the following clauses are interpreted with respect to it; in fact, they fail to even mention it. Topic relationships are first and foremost a discourse phenomenon. As a discourse unfolds, the organization they impose is inherent in the conceptual structure being built, even when they have no distinct grammatical realization. There are, however, topic markers and explicit topic constructions. The sentences in (19) represent a basic topic construction in which the initial nominal is established as a topic at least (p.515)

Engaging the World

Figure 14.4

for the proposition that follows. In unhurried speech, the nominal and target clause are separated by the slight pause written as a comma, suggesting that each occupies its own attentional frame.

An instance of this construction is sketched in figure 14.4. The two component structures are a nominal, that mural, and a finite clause, I really hate it. Semantically, their integration hinges on a correspondence between the nominal profile and the clausal element functioning as pivot—in this case, the referent of the object pronoun. Phonologically, their integration consists in temporal adjacency and the suspended intonation between them. There being no topic marker, the topic relationship emerges only at the composite-structure level; it is not a property of either component structure, considered individually, but rather of the construction as a whole.17 The composite structure is shown as having two distinct profiles: the reference point and the target process. This is because the expression comprises two attentional frames, each a window of attention with its own attentional focus. An analysis positing a succession of profiles (instead of a single overall profile) captures the evident dynamicity of this construction.

In this construction, the topic nominal is clearly external to the target clause, as shown by intonation as well as by the fact that the clause is structurally complete. The same considerations lead to the opposite conclusion in sentences like the following: That mural I really hate. In this expression, there is no intonational disjuncture between the topic and the remainder, nor is the latter a complete clause in and of itself. Here the topic nominal is an integral part of the target clause, functioning not only as topic but also as clausal landmark. It is thus internal to the clause, both semantically and syntactically.

Shown in figure 14.5, this construction amounts to a blend of the topic and object constructions. It represents a variant of the object construction because that mural elaborates the schematic landmark of I really hate. It resembles the topic construction (p.516)

Engaging the World

Figure 14.5

in that the topic comes first, but it differs by virtue of there being just a single attentional frame with a single overall profile. Thus the mural's invocation as reference point is simultaneous with its invocation as clausal landmark. Another way to put it is that there is no distinction between topic and pivot, the topic nominal being internal to the clause expressing the target proposition.

Topics thus occur at different levels of organization, holding sway over dominions of different sizes.18 The dominion can be a passage of any length, a complex sentence, or a single clause. The smaller the dominion, the closer the connection tends to be between the reference point and the target proposition. Structurally, the topic can be in a separate sentence, in the same sentence but external to the target clause, or an integral part of that clause. But even in the latter case, a topic functions as reference point for discourse reasons, not by way of coding the clausal proposition. Being a matter of information structure, a nominal's role as topic is superimposed on its role as clausal participant. Conceptually, its status as topic is extrinsic to the proposition, determined by discourse factors rather than by objective content. Hence the same proposition can usually be expressed with alternate topics (That mural, I really hate it vs. As for me, I really hate that mural) or none at all (I really hate that mural).

Like a topic, a subject functions as reference point for a target proposition. This is the basis for their similarity. The difference is that a subject is both structurally internal to a clause and conceptually intrinsic. A topic is a reference point in discourse, directing the hearer to the proper realm of knowledge for interpreting the target proposition. Its role as reference point is thus extrinsic to the clause's objective content. By contrast, the subject's role as reference point is inherent in the apprehension of the clausal content. A subject specifies the trajector of a profiled process, and the trajector—it will be argued—functions as reference point for the very purpose of conceptualizing that process.

An entity invoked as reference point for one purpose is likely to serve in that capacity for other purposes as well. So it stands to reason that a subject, being a (p.517) clause-internal reference point, would also tend to be a discourse topic. Still, their coincidence is only a tendency. Topic and subject are distinct notions, pertaining to different channels. When the same nominal serves in both capacities, it is a reference point with respect to both information structure and objective content.19

14.1.5 Subject and Object

The subject of a clause is a nominal that elaborates the trajector of the process it designates. I am now suggesting that the trajector of a profiled relationship, previously described as its primary focal participant, can also—and more fundamentally—be characterized as a reference point with respect to it. Similarly, an object elaborates the landmark of a profiled relationship, previously described as its secondary focal participant. By analogy, then, a relation's landmark can be characterized as a secondary reference point with respect to it. These notions are hardly self-explanatory, but when properly understood they prove both psychologically plausible and linguistically revealing.

Let us start with the idea that the focal participants of a profiled relationship are intrinsic reference points with respect to it. In contrast to a topic, which locates a proposition in a larger conceptual context, the trajector and landmark of a process are inherent in its conceptualization—they function as reference points for the very purpose of building up to its full conception. This follows from the earlier observation (§4.2.1) that relations are conceptually dependent on their participants: we cannot apprehend a relationship without invoking the participants that support its manifestation. Because they make its conception possible, participants are reasonably described as providing mental access to a relation. This, of course, is tantamount to saying that they function as reference points in regard to it.

Figure 14.6 represents the contrast between extrinsic and intrinsic reference point relationships with a profiled process as target. Diagram (a) corresponds to a clause-external topic construction, like the one in figure 14.4. The reference point relation is above and beyond the target's characterization and is thus extrinsic to it, even though R somehow figures (as pivot) in the overall content of this proposition. R's dominion is the range of associated knowledge, within which T is interpreted. Diagram (b) depicts the intrinsic reference point relation involving a process and its

Engaging the World

Figure 14.6

(p.518) trajector. In this case R and T are not distinct. “Reaching” the target is simply a matter of conceptualizing it, so the dashed arrow representing the path of mental access is coextensive with the solid arrow repesenting the temporal unfolding of the profiled relationship. And since T is conceptually dependent, the path of mental access runs through R. Here R's dominion (the set of potential targets) consists of all those processes it might anchor in this fashion.

Granted that trajector and landmark are intrinsic to the conception of a profiled process, how does their reference point characterization relate to their description in terms of primary and secondary focal prominence? Describing them as focal participants is equivalent to describing them as reference points. Focal participants are commonly referred to by linguists as central (or core) participants. This notion of centrality cannot, of course, be understood in a literal, spatial sense. Rather, it is invoked metaphorically to indicate that the participants in question are essential to the relationship: without them, it would either be incoherent or constitute a different relationship. This amounts to saying that a relation's central participants provide mental access to it. And since a profiled relationship is by definition a focus of attention, its essential participants are focal participants.

To the notion of centrality (obligatory access) the reference point characterization adds the notion of dynamicity (sequential access). Here we find the basis for distinguishing trajector and landmark: they are primary and secondary focal participants by virtue of being the first and second reference points evoked in building up to the full conception of a profiled relationship. According to this analysis, trajector/landmark alignment consists in a natural path of mental access, with the trajector as its starting point: trajector > landmark > relationship. It is thus suggested that the two degrees of focal prominence are at least partially attributable to sequentiality.

A relationship with both a trajector and a landmark is sketched in figure 14.7. The labels R1 and R2 indicate their status as first and second reference points. The overall target process, T, has been factored into two subprocesses, T1 and T2, representing those facets of T that directly involve R1 and R2. With a verb like smash, for instance, T1 consists in the trajector's exertion of force, and T2 consists in the landmark's resultant change of state.20 The proposal, then, is that the conception of a two-participant relation involves a mental progression from R1 to R2, and thus from

Engaging the World

Figure 14.7

Engaging the World

Figure 14.8

T1 to T2, the subrelations they respectively anchor. In view of this progression, it is also apparent that R1 heads a path of access subsuming both T1 and T2. Figure 14.7 can thus be seen as a special case of figure 14.6(b), where R = R1 and T = T1 + T2.

For ease of representation, I will adopt the simplified diagrams of figure 14.8. The labels R, T, D, and C are suppressed, and neither the conceptualizer nor the dominions are shown explicitly. It suffices to indicate the path of mental access.

Importantly, a path of mental access is usually not exclusive. With a situation of any complexity, there are alternate ways to access constitutive elements in building up to a full conception of it. Comprising one range of alternatives are different options in regard to profiling—which facets of the overall situation will be singled out as the focused relationship. Depending on the choice of profile, different participants are central enough to its conception (if not indispensable) to be considered reference points with respect to it.21 A relationship can also be accessed by evoking the same participants in alternate sequences. These differences are matters of construal and are possible even if the total conceptual content should be identical. Since meaning includes both content and construal, expressions that contrast in these ways are semantically nonequivalent.

As brief illustration, consider the three-way contrast among expressions of the form X give Y Z, X give Z to Y, and Y receive Z. Applied to a simple act of physical transfer, they share the conceptual content sketched in figure 14.9(a). The effect of the transfer is that Z moves from X's dominion (sphere of control) into Y's. The dominions are shown as ellipses, and Z's movement as a single arrow. The double-shafted arrow indicates that X initiates this movement. The double-headed arrow represents the event's interactive nature: X intends for Y to have Z, Y is cognizant of this intent, the exchange has social import, and so on. Finally, multiple arrows stand for Y's role in assuming possession of Z: observing its approach, physically accepting it, and then controlling it.

Shown in figure 14.9(b)–(d) are the alternate construals the expressions impose on this content.22 The verbs are distinguished in terms of profiling. Whereas give designates the overall event, comprising all the relationships indicated, receive's profile is limited to Y's interaction with Z. This difference correlates with the choice of trajector, the starting point for mentally accessing the profiled process. With the agent X as starting point, the profile extends to all the occurrences X initiates: Z moving, Y interacting with Z as a consequence, and Y being affected at the mental and social levels (e.g. in apprehending the change and being recognized as the new controller). (p.520)

Engaging the World

Figure 14.9

On the other hand, with the recipient Y as starting point, the profile is confined to the occurrences Y initiates: the physical and experiential aspects of accepting and controlling Z.

In the case of give, a further difference stems from the choice of landmark, the second reference point evoked in mentally accessing the profiled process. The landmark can be either the recipient, in what is known as the ditransitive construction, or the mover, as an instance of the caused-motion construction (§11.3.3). With either option, the profile encompasses all the relationships shown. The reference points do, however, represent alternate ways of building up to the full conception. Naturally, choosing one or the other serves to highlight those facets of the overall relationship it figures in most directly. This is shown by the heavier lines in figures 14-9(b) and (c). When the access from X proceeds via Y (note the dashed arrows), there is greater emphasis on the action affecting Y, mainly by effecting its status as controller. By contrast, access via Z serves to highlight the causation of Z's movement.23

In addition to there being alternate paths of access, processing occurs at multiple levels of organization. This is exemplified by the relation between a transitive verb, such as throw, and its corresponding passive, be thrown. At the lower level of organization, the verb throw profiles the causation of motion, with the agent as trajector and the mover as landmark. Shown at the left in figure 14.10, this represents an optimal alignment, in that the path of mental access follows the natural paths of causation and temporal sequencing. The passive provides an alternative for cases where, for discourse purposes, the same content is more conveniently accessed with the mover as starting point. The effect of the passive construction (and more specifically, the past (p.521)

Engaging the World

Figure 14.10

participial inflection) is to superimpose on the basic conception a higher-level construal in which the mover functions as trajector. In the derived expression be thrown, the same content is accessed through (hence viewed in relation to) the mover. It is not that the basic alignment disappears—it is simply overshadowed at the composite-structure level. To some extent we have to backtrack, and follow the natural path anchored by the agent, in order to apprehend what happens to the mover.24

The proposal that trajector/landmark alignment is based on sequence of mental access, and is thus essentially temporal in nature, may not be intuitively obvious. This is only to be expected in view of the small time scale involved. At the most basic level, the sequentiality is intrinsic to the meaning of a single lexeme (e.g. throw), whose conception is measured in milliseconds. It is only at higher levels of organization, where the time scale is larger and elements are separately expressed, that we can reasonably anticipate their sequential access being subject to introspection. The proposal is a special case of the broader notion (§14.1.1) that conceptions of ordering and directionality reside in serial processing, which is a basic aspect of conceptual experience even when it remains below the threshold of conscious awareness. In this respect trajector/landmark alignment is seen as comparable to the intrinsic directionality of a scale.

Though not self-evident, neither is the proposal merely speculative. There is quite a bit of evidence that trajector and landmark are properly characterized as sequentially accessed reference points. Some evidence is only circumstantial, like the well-known fact that in most languages a subject precedes an object in the basic (most neutral) word order. This is only circumstantial because the subject and object nominals, while they express the trajector and landmark, cannot (strictly speaking) be identified with them. Trajector and landmark are conceptual entities, inherent in the meaning of a verb or a larger predicate. Their sequential access is thus internal to the predicate, a matter of apprehending the profiled relationship, so per se it has nothing to do with the order of presentation. On the other hand, subject and object nominals are symbolic structures. As such, they have phonological expression and occur in a certain order. Their status as subject and object does not depend on word order, however, but on correspondences between their profiles and the clausal trajector and landmark. Even when used to identify subject and object, clause-level (p.522) word order is influenced by information structure and other considerations. There is, nonetheless, a definite tendency for a subject to precede an object in the absence of overriding factors, which is only to be expected if trajector/landmark alignment also has a temporal basis. The default ordering reflects the processing efficiency achieved through the coalignment of natural paths.

Also suggesting the sequentiality of trajector/landmark alignment is the well-known grammatical accessibility of the subject and object roles. Compared with other clausal elements, the two are highly active grammatically: they are most likely to trigger verb agreement, to function as pivot in relative clauses, and so on. In autonomous theories of syntax, grammatical relations are often ranked for “syntactic prominence”, with subject and object the first two items on the list. In such approaches, however, terms like “prominence” and “accessibility” have no independent content—they simply label the fact that some grammatical roles figure in more phenomena than do others. By contrast, the CG account derives the ranking from something more fundamental. The syntactic prominence of subject and object reflects the conceptual prominence of trajector and landmark, established independently as essential constructs for semantic description (§3.5). We have now taken the further step of explicating trajector/landmark alignment in terms of sequence of mental access, providing a direct basis for their grammatical accessibility.

One manifestation of their accessibility is the key role of subject and object in pronominal anaphora. Though many factors are involved, one can make the basic generalization that a subject can serve as antecedent for any other clausal element, and an object for any element except the subject (van Hoek 1995, 1997). We see in (21)(a) and (b), for instance, that a subject can serve as antecedent for the possessor of an object, but not conversely. Likewise, (21)(c) and (d) show the unidirectionality of an anaphoric relationship between a clausal object and an element of a prepositional phrase.

(21) (a) The kitten was chasing its tail.

(b) *It was chasing the kitten's tail.

(c) We observed the baboons in their native habitat.

(d) *We observed them in the baboons' native habitat.

Accepting that anaphora is itself a reference point phenomenon, the generalization dovetails with the characterization of trajector and landmark as first and second reference points for apprehending the clausal process. Reference points invoked for one purpose tend to be used as such for other purposes as well.

Supporting the analysis even more directly is the use of possessives to specify the participants of a nominalized verb, as in Booth's assassination [of Lincoln] and Lincoln's assassination [by Booth]. When a verb is nominalized, the process it designates is construed as an abstract thing; the noun assassination profiles a thing comprising one instance of the process assassinate. The event's construal as a thing is represented by the heavy-line ellipse in figure 14.11. If the trajector and landmark of a process are reference points with respect to it, and the process is construed as (p.523)

Engaging the World

Figure 14.11

a thing, then each participant's relation to it constitutes a possessive relationship (defined schematically as a reference point relation between two things). Hence the analysis correctly predicts that possessives should commonly be used to specify the participants of a reified process.

The expressions are quite analogous to basic possessives like Zelda's quilt (fig. 14.2), the only difference being that the target is a reified process. So in both expressions assassination elaborates the schematic target of the possessor phrase (Booth's or Lincoln's). The special feature of this periphrastic construction is that the target noun itself incorporates reference point relations—those inherent in the verb's trajector/landmark alignment—one of which is identified with the possessive relation. The possessor thus corresponds to either the trajector or the landmark of the reified process, and its reference point relation to the target is the same one it has intrinsically as part of the verbal meaning. In this way, the possessive construction serves to specify a processual participant.

A final piece of evidence comes from equative expressions, as in (22). These are of two basic sorts. In (22)(a)–(b), the subject and predicate nominals refer to specific, actual individuals. What these sentences profile is the relationship of referential identity: that is, the nominals refer to the same individual.25 In such examples the nominals are often reversible, identity being a symmetrical relation. But in other examples, like (22)(c)–(d), the nominals do not refer to specific individuals. And in this case they are typically not reversible, as witnessed by the infelicity of (22)(d). How can we describe such expressions? And why are they irreversible?

(22) (a) My cousin Harvey is the guy who got drunk at our wedding.

(b) The guy who got drunk at our wedding is my cousin Harvey.

(c) A tiger is a feline.

(d) *A feline is a tiger.

(p.524) Expressions of this latter sort also profile a relation of referential identity. The basic difference is that the nominal referents are virtual rather than actual: they are fictive instances of their types, conjured up in order to make a general statement. The key factor is a kind of directionality that tends to be obscured with actual referents but emerges more clearly with fictive referents evoked as representatives of their types. The import of (22)(c), very roughly, is as follows: if you start with an instance of tiger, you will always find that it coincides with an instance of feline. This accords with our standard taxonomic model, in which tigers are a subclass in the class of felines. But (22)(d) runs afoul of this taxonomy: if you start with an instance of feline, it is not always true that it coincides with an instance of tiger (e.g. it might be a leopard).

So even though it predicates identity, the equative construction implies a direction of assessment, invoking the first nominal's referent as a starting point—the one whose identification is at issue. Where does this directionality come from? The only evident source is trajector/landmark alignment. As clausal subject, the first nominal specifies the trajector of the profiled identity relationship. The trajector is the starting point by virtue of being the first reference point evoked in conceptualizing this relation.

14.2 Fictivity

The tiger and the feline referred to in (22)(c) are fictive (or virtual) instances of their types, as opposed to actual individuals. They join the growing menagerie of fictive entities that we have seen to be linguistically significant. Among these (to mention just a few) are the products of metaphor and blending (§2.2.3), an imagined vantage point (§3.4.1), virtual bounding (fig. 5.4), the conceptualizer (C) invoked by a grounding element (§12.3.2), and the fictive invocation of a speech-act scenario (§13.2.3). Its prevalence is such that fictivity has to be recognized as a basic feature of cognition with a fundamental role in language structure.

14.2.1 Disengaged Cognition

We live in a real world.26 Since our view of this world is mentally constructed on the basis of experience, each of us apprehends it somewhat differently, and quite differently from creatures with other mental capacities. Despite this variability, the constructive process is shaped and constrained by the world's actual nature—otherwise, there would be little chance of survival. The process is further constrained by our position in the world. We always apprehend it at the present moment, from our current location, through our own senses, and with our own mental faculties. Omniscience is not an available option.

Ultimately, the world we construct is grounded in our experience as creatures with bodies who interact with their surroundings through physical processes involving sensory and motor activity. This is known in cognitive linguistics as embodiment. (p.525) But obviously, our mental life transcends the limits of immediate bodily experience. Various cognitive processes give rise to mental structures, at successive levels of organization, whose connection with such experience is progressively more remote. Not only do these structures allow us to cope with the real world more efficiently, but also they define—and vastly expand—what constitutes it. From our standpoint, the world we inhabit and engage has not just physical but also social, cultural, and intellectual dimensions. Once they are cognitively established, we can operate in these realms in largely autonomous fashion. Discussions of investment strategies, or ruminations concerning linguistic theory, are basically independent of immediate physical reality; nonetheless, they are means of engaging certain aspects of our mentally and socially constructed world.

Thus a great deal of our cognitive activity is disengaged (to varying degrees) from immediate bodily experience. How do we manage to transcend it? One way is through memory, consisting in the partial revival of a previous experience. Another is anticipation, where the observation of a present situation affords a basis for projecting its future development. This can be as elemental as the default expectation that a physical object will continue to exist from moment to moment. But it can also be based on patterns that are learned through previous occurrences. Patterns are learned by abstraction, a fundamental means of transcending immediate experience.

Abstraction comes about through the reinforcement of what is common to multiple experiences. Since features that fail to recur are not reinforced, an abstracted structure is always impoverished relative to the experiences it derives from. And since commonalities are often apparent only in a coarse-grained view, involving lesser precision, abstracted structures are usually schematic relative to these experiences. Though immanent in all of them, an abstracted structure is independent of any particular instantiation. It represents a generalization with the potential to be invoked in subsequent processing. Without the capacity for abstraction, every experience would be unique and unrelated to every other. A structured view of the world could not emerge.

The conventional units of a language are abstracted from usage events. Once learned, a unit transcends the events giving rise to it, with the potential to be employed in further events involving new expressions (ch. 8). It is disengaged from immediate experience in the sense that it is part of a speaker's linguistic repertoire, available for implementation, even when not currently being used. As a consequence of this disengagement, an entity crucial to a unit's meaning may be virtual in nature when the unit is considered independently of its use. Obvious cases are the pronouns I and you, which—viewed as conventional units—refer to the speaker and hearer in abstracted, generalized fashion. It is only in the context of a particular usage event that they refer to specific, actual individuals. The same holds for a speech-act scenario, as well as for the conceptualizer invoked by grounding elements and predicates of propositional attitude (§12.3.2). More generally, the entity profiled by a lexical noun or verb is merely a type of thing or process, and a type per se is always virtual. It is only when grounded to form a nominal or a finite clause that the profiled entity is conceived as an instance of its type. And it is only in the context of a particular usage event that this instance can be identified as an actual individual in the world.

(p.526) It is not the case, however, that abstracted entities are always identified with actual ones in the context of usage events. In some linguistic structures, their abstracted nature is precisely the reason for invoking them. This is so when, instead of being grounded, a noun or verb is used as the first element of a compound or as the base for morphological derivation. A mosquito net is intended to offer protection from mosquitoes in general, rather than any one in particular. Similarly, describing someone as a complainer would normally be taken as indicating a general propensity to complain, not a specific instance. But while the absence of grounding results in virtuality, its presence does not itself ensure actuality. For example, A tiger is a feline is a full, finite clause containing two grounded nominals. Yet all the referents are virtual: the tiger, the feline, and thus the profiled identity relation. Instead of being actual, an instance of a type may simply be conjured up for a special purpose, one of them being to make a general statement.27 In this case, the instance referred to is construed as being representative of its type (§9.3.5), hence an abstraction vis-à-vis particular instances.

Types and representative instances arise from actual entities through different kinds of abstraction, indicated in figure 14.12. Recall that an instance is thought of as having a particular, distinguishing location in the domain of instantiation (DI). We obtain a type conception by abstracting away from this property. While the overall description of a type includes its connection with instances (this is, after all, what it means to be a type), instances and their locations remain in the background (fig. 9.3). In the conception of a representative instance, on the other hand, the notion of instances and distinguishing locations is itself an aspect of the abstracted commonality. The virtual instance is abstracted from actual ones precisely to represent their shared property of being instantiations of the type (fig. 9.13).

Representative instances of a type figure in numerous linguistic phenomena. They are pivotal to the meaning of the grounding quantifiers every, each, and any (§9.3.5). They are further used for making both local and global generalizations. For example, the following sentence describes what is common to three distinct events involving different customers and different snakes:

(23) Three times this morning a customer bought a python.

Engaging the World

Figure 14.12


Engaging the World

Figure 14.13

Despite this multiplicity, the subject and object nominals occur in the singular. The profiled customer and python are not any actual instances, but virtual instances of their types, construed as being representative of the actual ones. They participate in an instance of buying that is also virtual, being representative of three actual instances. As shown in figure 14.13, this fictive occurrence is the one the sentence puts onstage as the profiled process. It is abstracted as a generalization over three actual occurrences, the connection between them being specified by the adverbial expression three times this morning. The sentence's meaning includes this entire configuration—both levels and the nature of their relationship.28

Analyzed in similar fashion are generic statements employing indefinite nominals, e.g. A tiger has stripes. With generics, the generalization is global rather than local: instead of applying to a limited number of occurrences that happen to be analogous, it is offered as a characterization of the world's essential nature (Goldsmith and Woisetschlaeger 1982; Langacker 1997). It thus applies to an open-ended set of occurrences, and is expected to be valid for every instance of the qualified type (in this case, tiger). Generics make no explicit reference to the notion of the world having an essential structure, or to the type's maximal extension (Et). These mental constructions are part of the supporting conceptual substrate.

Virtual instances also play a role in quantifier scope, where one quantifier figures in the conception to which the other applies. On the relevant interpretation, for example, Two boys ate seven apples ascribes the feat of eating seven apples to each of two boys. It is said that two has seven in its scope (or that two has wide scope, and seven narrow scope). Interpreted in this manner, the sentence implies that a total of fourteen apples were consumed. Why, then, does it mention only seven? The reason is that the profiled occurrence eat seven apples is a fictive event abstracted to represent the commonality of two actual events, one on the part of each boy. The two boys are actual, but the seven apples are virtual.

By nature, an abstraction conforms to the structures it is based on but is less detailed. As a representation of what they share, it is immanent in these structures but not exhaustive of them. Another basic means of transcending direct experience, one that exhibits the opposite properties, is conceptual integration (Fauconnier and Turner 2002). Conceptions are integrated through correspondences between their elements. The result is often a new conception substantially different from any previously entertained. An obvious (p.528) case is metaphor, in which a source domain is used to apprehend a target domain, resulting in a blended structure (fig. 2.9). Another is semantic composition, where component structures are integrated to form a composite conception. Patterns of composition—the semantic poles of constructional schemas—allow the formation of conceptions that are familiar (e.g. lazy cat), novel but possible (purple bread), purely imaginary (invisible elephant), or even conceptually incoherent (square circle).

Conceptual integration lets us deal with the ever-changing circumstances of real life. At the other extreme, it is used in producing works of fiction where the characters, the story, and even the world itself are imaginary. In between is the practice of invoking fictive entities as an indirect means of dealing with actuality. Large-scale examples include mathematics, scientific theories, and systems of philosophy (Lakoff and Núñez 2000; Lakoff and Johnson 1999). On a more modest scale, we find linguistic devices that specifically indicate the nonactual status of occurrences. In (24)(a)–(c), for instance, the fictive nature of the dog speaking French is indicated by negation, the verb imagine, and the conditional construction with if, respectively. Invoking this situation does however serve a purpose in regard to actuality. Knowing that something is not the case, or will only be the case under certain conditions, may very well have consequences for what we actually do. And if Jane is deluded about her dog speaking French, the fact that she imagines it is nonetheless a real situation we may have to cope with.

(24) (a) Jane's dog does not speak French.

(b) Jane merely imagines that her dog speaks French.

(c) If her dog speaks French Jane can make a fortune.

In a final means of transcending direct experience, mental operations inherent in a certain kind of experience are applied to situations with respect to which their occurrence is extrinsic. This is called subjectification, indicating that the operations come to be independent of the objective circumstances where they initially occur and whose apprehension they partially constitute. A previous example is the reference point characterization of possessives (§14.1.2). The mental operation of invoking R to access T (C —> R —> T) is immanent in the conception of the possessor controlling the possessed (R ↠ T). In many possessive uses, the objective notion of control is attenuated to the point that sequential mental access is all that remains. Another example is nominalization, e.g. assassinate > assassination, whereby an event is conceived as an abstract thing (fig. 14.11). Here the mental operations are grouping and reification (§4.2.2), which apply transparently in collective nouns such as stack, team, and archipelago, and below the level of conscious awareness with prototypical nouns such as dog, rock, and pencil. With a noun like assassinate, the entities grouped and reified are the component states of the verbal process (the relationships profiled at each successive instant).

One product of subjectification is the phenomenon known as fictive motion.29 For instance, the expressions in (25)(a)–(b) incorporate elements used primarily for (p.529) spatial movement: the motion verb run, as well as the path prepositions from and to. Moreover, they appear to indicate movement in opposite directions. Yet both describe the same situation, which is static and has no inherent directionality. In such expressions, cognitive operations inherent in the conception of spatial motion are applied to static scenes as a way of mentally accessing them.

(25) (a) An ugly scar runs from his elbow to his wrist.

(b) An ugly scar runs from his wrist to his elbow.

(c) The pitcher ran from the bullpen to the mound.

We conceptualize an actual motion event, such as (25)(c), by tracking the mover's progress along a spatial path. This is shown in figure 14.14(a): through processing time (T), we successively conceptualize the mover as occupying—through conceived time (t)—a series of locations that collectively constitute the path. An inherent aspect of this conception is that the conceptualizer scans mentally along the same path which the mover traverses physically; to properly apprehend this event, C must access the successive locations in the same order that the mover reaches them. The dynamic conception of a path is therefore immanent in the conception of actual motion. In fictive motion expressions, the same mental operations are applied to a static scene, as shown in figure 14.14(b): through processing time, C scans along the path by successively invoking the constitutive locations. Here, though, the analog of the mover is a spatially extended object (like a scar) that occupies all these locations simultaneously. Instead of tracking an object's movement, C scans along the path by way of building up to a full conception of the object's spatial configuration. And at least for this purpose, conceived time has no significant role in the expression's objective content (OC).30

Engaging the World

Figure 14.14

(p.530) Through subjectification, the dynamicity inherent in the apprehension of events is transferred to the conception of static scenes. A verb like run, which profiles objectively construed motion by its trajector, comes instead to designate a configuration apprehended through subjectively construed motion (i.e. sequential mental access) by the conceptualizer. We observe a similar transfer of dynamicity in cases of fictive change (Matsumoto 1996c; Sweetser 1997). Let us briefly examine two basic kinds.

One kind involves past participles used as adjectives, as in broken vase, detached retina, and scattered marbles. Derived from change-of-state verbs, these participles designate the state resulting from the verbal process (fig. 4.15). A vase is described as broken, for example, only when it has undergone the process break. However, not every use of a stative-adjectival participle implies an actual change. A broken line has never undergone the process of breaking. Likewise, a detached garage has never been attached, nor have scattered villages ever been clustered together. In such uses, the change designated by the verb stem is only virtual, serving to specify how the actual situation deviates from one considered neutral or more typical. A broken vase and a broken line are comparable in terms of the profiled state—the vase and the line are both in pieces—which in each case differs from the state of being whole. But broken vase invokes an actual change through time, a physical progression manifested in the vase itself, hence objectively construed. By contrast, the change invoked by broken line is subjectively construed. It does not inhere in the line itself, but rather in the conceptualizer, as a mental progression in which the profiled state is viewed as departing from the canonical one. Being only virtual, the change is not conceived as unfolding through time. The mental progression (residing in sequential access through processing time) is, nonetheless, a vestige of break's dynamicity.

The following sentence exemplifies a second kind of fictive change:

(26) Our Christmas tree gets smaller every year.

This may describe an actual change, of course: we cannot afford a new tree every Christmas, so we use the same one over and over, and each year it loses additional needles and branches. More likely, though, the sentence means that the tree we buy each year is always smaller than the (different) one we bought the previous year. On this interpretation the change is only fictive—no tree actually gets any smaller. Nor does the subject refer to any actual tree. The sentence invokes the abstracted conception of a family celebrating Christmas, a cultural scenario in which a tree has a prominent role. Our Christmas tree is a role description: the tree it refers to is the virtual one that occurs in this scenario (as it applies to the speaker's family). This role is filled by a series of actual trees, each presumably constant in size. It is through the fictive identification of these instantiations, imagining them to be a single entity, that we obtain the notion of a tree changing size.

Even this small sample should indicate how often we resort to fictive entities and other mental constructions. What explains their prevalence? Why is cognitive activity so often disengaged from immediate experience? For the most part we are not attempting to escape reality by constructing an imaginary world. More typically, it is in fact the real world that concerns us. While it does not refer to them directly, (p.531) for example, (26) does pertain to actual trees and an aspect of real-world experience. The main purpose of disengaged cognition is not to escape but to cope. The mental capacities we have discussed here are crucial for constructing and negotiating the world we live in. Though disengaged from immediate bodily experience, they allow us to engage the world at other levels.

14.2.2 Covert Scenarios

We understand (26) by invoking the cultural practice of decorating a pine tree during the Christmas season. We have an explicit basis for doing so: the compound Christmas tree provides direct access to this familiar scenario. Quite commonly, however, we rely on scenarios that are left implicit or suggested only indirectly.31 Faced with seeming incoherence, we may simply infer them as a way of making everything make sense. Many invoked scenarios are fictive in nature. Like fictive motion and fictive change, they are often a source of dynamicity in the conception of stable situations.

An example that nicely illustrates these points is the following (cited in Talmy 1988b):

(27) There's a house every now and then through the valley.

This sentence comprises an existential expression, there's a house, together with two adverbs, now and then and through the valley. The sentence is natural and readily understood, but when we try to analyze it, questions arise about its semantic and grammatical coherence. The adverb now and then specifies that an event occurs intermittently. Usually, however, the existence of a house is a stable situation. And even though through the valley describes a path of motion, (27) does not contain a motion verb and does not refer explicitly to anything that moves. What, then, do the adverbs modify?

We make sense of (27) by invoking the scenario of a traveler (I imagine someone riding on a train) who observes the scenery along the way. The adverb through the valley describes this imagined path of travel. A traveler has a limited field of view, so as he moves along the path, only a portion of the valley is visible at a given moment. This provides the basis for the adverb now and then. It describes the intermittency of a certain kind of viewing event: occasions when, at the moment in question, a house appears in the field of view. It is, of course, a different house on each occasion. In the context of this scenario, there's a house construes the house and its existence as virtual entities representing the abstracted commonality of multiple viewing experiences. The travel and the traveler are also fictive. While it is not precluded that the speaker might be recalling the observations made during an actual journey, the sentence itself does not imply this. The import is rather that anyone traveling through the valley would have this experience.

Travel is only one activity allowing the sequential observation of multiple entities. If they can all be observed from one place, like the fielders on a baseball team, we can access them sequentially just by shifting our gaze. If they are moving one by (p.532) one across our field of view, like the cars of a passing train, we can simply watch them. We also perform various actions to bring them successively into view: turning the pages of a book, removing plates from a stack, pressing a button to scroll through a text, and so forth. From these varied kinds of experience, we abstract a generalized conception of sequential examination. Let us call it the scanning scenario.32 Like the travel scenario in (27), the schematic scanning scenario is tacitly invoked in many sorts of expressions. The scanning is usually fictive—we do not actually conceptualize the entities one by one but merely imagine doing so. This simulated scanning lends dynamicity to the apprehension of static situations. It also provides the connection between a generalization and the range of instances supporting it.

One element invoking the scanning scenario is the quantifier each (§9.3.5).33 It is used quite naturally in expressions describing actual sequential observation, like (28)(a). The scanning, though, is only simulated—understanding this sentence does not require that we actually observe all the graduates one by one. And since the scanning is only simulated, each is also used in expressions like (28)(b), where, in actuality, there is no sequential examination. Each profiles a virtual instance of a type taken as being representative of a set of actual instances. The notion of serial access provides a link between the representative instance and those it represents: the property ascribed to the former (e.g. having its own recipe for tiramisu) holds for all the instances reached in this manner. If we examined them one by one, checking them for the property, in every case we would find it.

(28) (a) As they filed across the stage, she called out the name of each graduate.

(b) Each restaurant has its own recipe for tiramisu.

Even when in the background of awareness, covert scenarios are not only part of the meanings of expressions but are also reflected in their forms. We see this in (29), where frequency adverbs appear to function as nominal quantifiers. Normally these adverbs specify the frequency of events (e.g. She {always/usually/often/seldom/never} pays cash). Here, though, the more likely interpretation concerns the proportion of instances of a type who exhibit a certain property. On this reading, the sentences in (29) are equivalent, respectively, to those in (30), with the quantifiers all, most, many, few, and no. How can this be? How can adverbs quantify nouns?

(29) (a) Linguistic theorists are always arrogant.

(b) Professional basketball players are usually tall.

(c) Moral crusaders are often closet perverts.


(d) University professors are seldom rich.

(e) Movie stars are never good role models.

(30) (a) All linguistic theorists are arrogant.

(b) Most professional basketball players are tall.

(c) Many moral crusaders are closet perverts.

(d) Few university professors are rich.

(e) No movie stars are good role models.

This apparent anomaly is due to a covert scenario. The adverbs in (29) do in fact specify the frequency of events, but these cannot be identified with the profiled clausal process. Each sentence ascribes a property (e.g. being arrogant) to its subject (linguistic theorists). We understand the sentences by invoking a version of the scanning scenario: the notion of progressing through life (or moving through the world), in the course of which we encounter enough instances of a given type to constitute a representative sample. What the adverbs describe is the frequency of events in which the instance encountered exhibits the property in question. And since the frequency of events correlates directly with the proportion of instances with the property, the effect is the same as with nominal quantification. The events, of course, are only fictive—(29)(a) does not imply that the speaker has ever been fortunate enough to actually meet a linguistic theorist (let alone all of them). They serve as a means of apprehending the static distribution of properties in terms of the dynamic process of exploration.

Numerous adverbial expressions prompt the simulation of a scanning experience. Compare the uses of still in (31). Canonically, as in sentence (a), it indicates that the profiled situation continues longer than expected. In this case, both the situation (being undecided) and its continuation through time are aspects of the expression's objective content—that is, the scene being viewed and described. To a basic apprehension of the scene, still adds the instruction to scan along the temporal axis by way of assessing the duration of the profiled relationship, and it specifies that the requisite scanning goes beyond an expected cut-off point. While this scanning per se is a mental operation, hence subjectively construed, it does have an onstage counterpart: the progression through time of the profiled situation.

(31) (a) She is still undecided about buying a new car.

(b) She can't stand sports like football or hockey, and golf is still too violent for her.

By contrast, the scanning prompted by still in (31)(b) lacks a counterpart within the objective scene. It proceeds along a scale comprising a series of sports ranked in decreasing order of violence: football > hockey > soccer > …> volleyball > tennis > golf.34 Here, also, still indicates that the profiled situation (a sport being too (p.534) violent for her) continues longer than expected. But how do we interpret “longer than expected”? It does not pertain to the duration of the situation (presumably this is permanent). Rather, it reflects the process of scanning along the scale by way of assessing the situation's extension on it, i.e. the set of values for which it exhibits the property. Hence the import of still in sentence (b) is not that the situation continues longer than expected through conceived time but rather that the scanning continues longer than expected through processing time.

The preposed adverbials in (32) illustrate various ways of inducing fictive scanning. In sentence (a), we understand through the ages by simulating the experience of tracing a mental path through the time span of human history. In (b), from the brightest to the dumbest invokes the scenario of examining all the students one by one in the order of their intelligence. Sentence (c) recalls the experience of reading a graph, its two axes representing body size and gestation period. As body size increases is a case of fictive change, obtained by viewing the sizes of different species as if they were a single entity. The matrix clause is likewise a case of fictive change, obtained by identifying different gestation periods. We further conceptualize these changes in terms of movement along the respective axes. Finally, we understand the changes as affecting the same virtual creature, representing the different species. Based on this mental construction, the sentence tells us that the changes occur in tandem: as the virtual creature's body extends along the scale of size, at the same time its gestation period stretches out along the scale of length.

(32) (a) Through the ages, some great intellects have changed our view of the world.

(b) From the brightest to the dumbest, the students all work very hard.

(c) As body size increases, the average gestation period gets longer.

Covert mental constructions are also the basis for nonpresent uses of the English present tense. Three such uses are exemplified in (33): the “scheduled future”, the “historical present”, and generics (sometimes described as “timeless”). Prototypically, the English present specifies that the grounded process coincides with the time of speaking (§5.2.3). Schematically, it indicates that this process is immediate to the conceptualizer (C) invoked by clausal grounding (§9.4.2). Nonpresent occurrences seem problematic with respect to either characterization. Future, past, and timeless events can hardly coincide with the time of speaking. And what does it mean to describe them as immediate to C?

(33) (a) The party starts at midnight. [scheduled future]

(b) I get home last night and see a note on my door. [historical present]

(c) A kitten chases a piece of string. [generic]

The key is to recognize that the profiled occurrences are only virtual. They figure in tacit mental constructions, and while they correspond to actual occurrences, they cannot be identified with them. Sentences like (33)(a) invoke the scenario of consulting a plan or schedule (hence the term “scheduled future”). A schedule comprises (p.535) representations of events, together with their anticipated times, and once in place it is always available for consultation. In this sense a schedule and its entries are directly accessible—hence immediate—to anyone who knows it. Using (33)(a) amounts to consulting a mental schedule and “reading off” an entry. Apprehending the event consists in activating its representation—that is, simulating its occurrence at the specified time. Although it pertains to the future, this simulation coincides with the time of speaking.

Other uses of the present are comparable, apart from being based on different mental constructions. The historical present reflects our capacity for reliving past events by replaying them in our minds. We do not confuse this mental replay with the original events themselves—we know full well that they are simulations or re-creations of those events, available through memory for “viewing” at the present moment. Producing a sentence like (33)(b) is not too dissimilar from watching a videotape and describing each event as we see it. Generics are quite distinct. They arise through generalization instead of memory, and rather than particular events they represent the abstracted commonality of many occurrences. Generics invoke a cultural model that views the world as having an essential structure we can discover and describe. Producing a statement like (33)(c) amounts to “reading off” an item in such a description. Because they inhabit a representation of the world, not the world per se, the event and its participants are only virtual (representative instances of their types).

14.3 Simulation and Subjectification

Cognition is embodied. It resides in processing activity of the brain, which is part of the body, which is part of the world. At the most basic level, we interact with the world through our senses and physical actions. There are other levels, of course: much of the world we live in is mentally and socially constructed. But either directly or indirectly, the world we construct and apprehend is grounded in sensory and motor experience.

In the last section we explored various ways of disengaging cognition from immediate physical experience. They all share the property shown abstractly in figure 14.15. Diagram (a) represents an act of engaged cognition, where a person

Engaging the World

Figure 14.15

(p.536) interacts directly, at the physical level, with something in the world (W). This interaction (double arrow) is effected through the body, primarily via sensory and motor organs. The box labeled A indicates the role of the brain in this engagement: A is the processing activity, minimally including sensory input and motor commands, that constitutes the interactive experience. Diagram (b) shows comparable processing taking place without engagement. Certain facets of A—labeled A′—come to occur autonomously, in the absence of any current interaction with W. Though not necessarily either well delimited or easily segregated, A′ is immanent in A, so it occurs whenever A does. Thus its independent occurrence amounts to a shadow version of the experience constituted by A.

A′ is said to be a simulation of A. In various guises and under different labels, simulation is widely recognized as having a fundamental role in conceptualization and cognitive semantics (e.g. Johnson 1987; Barsalou 1999; Matlock 2004; Hampe 2005; Bergen 2005). One of its guises is sensory and motor imagery, well established as a psychological phenomenon (Shepard 1978; Kosslyn 1980). Without the usual perceptual stimulation, we can conjure up the visual image of a cat, the auditory image of a baby crying, or the tactile image of sandpaper. Without actually moving, we can imagine what it feels like to walk, swim, or throw a rock. These kinds of images have a significant role in lexical semantics. Included in the meaning of apricot, for example, are images of what one looks like and how it tastes. Included in the meaning of throw are visual and motor images of throwing. Activating appropriate images—simulating the experiences they represent—is a nontrivial aspect of apprehending such expressions.

Simulation is not confined to lexical semantics. As a general feature of cognition, it has many linguistic manifestations. It is a significant factor, for example, in expressions that invoke a fictive vantage point or viewing circumstances. In grasping the import of (34)(a), one thing we do is simulate the experience of seeing Catalina under the conditions indicated. Simulation is also essential for recognizing other conceptualizers and the nature of their mental experience. In understanding (34)(b), we have to imagine being in the senator's place in order to figure out where the wife and lover are in relation to him. And to some extent we simulate his mental state by way of apprehending the finite clause. Other obvious cases include the fictive travel invoked in (27) [There's a house every now and then through the valley], as well as the mental replay narrated in the historical present, as in (33)(b) [I get home last night and see a note on my door]. The basic point, however, is that simulation—in the broad sense of figure 14.15—occurs to some degree in virtually all expressions.

(34) (a) If it were clear, we could see Catalina from the top of that mountain.

(b) With his wife seated on his left and his lover on his right, the senator was getting nervous.

Simulation is always attenuated relative to engaged experience. Because it is not driven by immediate perceptual input, or harnessed to actual motor activity, it lacks the intensity or “vividness” of such experience. (Given the choice between burning my hand and merely imagining the pain this involves, I would probably (p.537) choose the latter.) Simulations are also less elaborate, A′ being just a portion of A. The visual image of a cat is bound to omit certain features that are evident when actually seeing it. The image is more schematic, lacking fine-grained detail.

Attenuation is a matter of degree. Naturally, lesser intensity and more rarified content translate into diminished awareness of the simulations carried out. Consider some previous examples. We understand (27) by imagining the experience of traveling through the valley. This being a fairly specific notion, we are easily made aware of it. Less so in the case of (29)(b), Professional basketball players are usually tall. This too invokes a scenario involving travel and successive encounters. But since no specific path is indicated, the travel conception is rather nebulous—the generalized notion of moving through the world in the course of life. Indeed, its spatial component is tenuous enough that it might fade away altogether. If so, we are left with something approximating the scanning scenario: the abstract conception of sequential examination. The scanning scenario is central to the meaning of each (distinguishing it from every and any). But owing to its rarified content, speakers are not explicitly aware of it.35

One dimension of attenuation is the extent to which elements are objectively or subjectively construed (§9.1). Elaborate conceptual content lends itself to being construed objectively. Thus (27) may well engender explicit awareness of a person traveling through the valley and watching the scenery. But we can also construe this content more subjectively by imagining how things look through the eyes of the traveler. In this case we have lesser awareness of the mover, as well as the circumstance of traveling and observing: instead of being onstage as objects of conception, they remain implicit as features of the (imagined) viewing situation. With more rarified content there is less to construe objectively. In this respect, the travel scenario of (27) and the scanning scenario invoked by each lie toward opposite ends of the spectrum. On the one hand, travel makes possible the sequential observation of scenes along an extended spatial path. On the other hand, the sequentiality of each is fully generalized: it is not limited to travel, to spatial extension, to visual observation, or even to the physical realm. This abstracted notion has little by way of tangible content. Instead of presenting a situation to be viewed, it is better described as a manner of viewing, potentially applicable to any sort of content. It is then subjectively construed, inhering in the subject rather than the object of conception.

Sequential examination functions as an object of conception when we conceive of someone engaging in it (e.g. in watching a general inspect a line of troops). In relation to this onstage role, its status in each exemplifies subjectification (§14.2.1): mental operations inherent in experiences of a certain kind are used in abstraction from their content and applied to other circumstances. The individuation effected by each, reflecting the discreteness of examining objects one by one, can thus be imposed on any sort of conception. Through subjectification, many abstract meanings can be related to everyday experiences. Among the other grounding quantifiers, every and any are based respectively on simultaneous viewing and random selection. The proportional quantifiers all, most, and some reflect the basic experience of (p.538) superimposing two objects and assessing their relative sizes. The scanning involved in fictive motion (A scar runs from his wrist to his elbow) mirrors the continuous observation of movement along a spatial path.

Subjectification is often manifested diachronically. The use of verbs like run for fictive motion develops historically from their original import of describing actual movement, objectively construed. Possessive verbs like have derive historically from verbs of physical control, with meanings like ‘seize’, ‘catch’, ‘hold’, ‘carry’, ‘get’, and so on (Heine 1997). Immanent in the conception of R controlling T is a reference point relation, where R is invoked as a basis for apprehending T. This mental progression from R to T is all that remains when a verb is extended to general possessive use, becoming independent of any specific conceptual content. Owing to its highly schematic nature, a verb of this sort is usually regarded as “grammatical” rather than “lexical”. Subjectification is thus a factor in the diachronic process of grammaticization: the evolution of grammatical elements from lexical sources.36

Recall, for example, that the English modals (may, will, must, etc.) derive historically from verbs with meanings like ‘want to’, ‘know how to’, and ‘have the power to’: they describe a potential force or potency tending toward the execution of an action (§9.4.3). Even in their epistemic uses, the grammaticized modals retain a vestige of their force-dynamic origin. For instance, must indicates compulsion (a force that cannot be resisted), and may the absence of a barrier (Sweetser 1982; Talmy 1988a). This force is subjectively construed, experienced as part of a mental simulation. It is the force we experience in extrapolating our current conception of reality so that it “reaches” the grounded process.

To take just one more example, it is common for a verb meaning ‘go’ to grammaticize into a marker of future time. English be going to is well along this evolutionary path. Tom is going to mail a letter can still describe Tom's spatial motion toward a goal with the intent of mailing a letter upon reaching it. More likely, though, it simply means that Tom will mail a letter (perhaps just by clicking a mouse). In the former case, the conceptualizer scans through time by way of tracking the subject's movement through space. On the future interpretation, this subjective temporal scanning occurs independently of any conception of spatial motion. It is merely a way of mentally accessing an event's location in time.

I do not claim that all grammatical markers arise in this manner. But it is striking how many grammatical notions are plausibly described as subjective counterparts of basic aspects of everyday experience, i.e. conceptual archetypes. This brings us back to a general CG proposal (§2.1.2) that has framed a sizeable portion of our discussion. It is suggested that certain fundamental and universal grammatical notions can be characterized semantically in terms of both a prototype and a schema. Providing the prototypical meaning is an objectively construed conceptual archetype. (p.539) The schematic meaning resides in a domain-independent cognitive ability, initially manifested in the archetype and later extended to other domains of experience. Clearly, this relation between the prototype and the schema is nothing other than subjectification: mental operations immanent in the archetypal conception come to be used in abstraction from its content and applied to other circumstances.

At a minimum, the proposal was made for the notions noun, verb, subject, object, and possessive. Let us briefly review how it applies to them.

  1. 1. The schematic basis for possessives is the conceptual operation of invoking a reference point to mentally access a target. This mental progression is immanent in the conception of ownership, kinship, and whole-part relations, the possessive archetypes.

  2. 2. Subject and object are also defined schematically in terms of reference points. They correspond to the trajector and landmark of a profiled relationship, i.e. its primary and secondary focal participants. Their focal prominence consists in the trajector and landmark being the first and second reference points accessed by way of building up to a full conception of the profiled relation. This mental progression from trajector to landmark is immanent in the conception of an agent acting on a patient. It is in clauses describing such interactions that subject and object assume their prototypical values.

  3. 3. Agent-patient interactions are prototypical for verbs as well. In actually observing such events, we scan them sequentially: at a given point in time, we can only observe the situation manifest at that instant. The schematic characterization reflects this aspect of the archetypal experience while abstracting away from all specific content. A verb profiles a process, a relationship scanned sequentially in its evolution through time.

  4. 4. Finally, a noun profiles a thing, defined abstractly as the product of grouping and reification. These mental operations are immanent in the conception of physical objects, the prototype for nouns.

14.4 Mind, Meaning, and Grammar

As their names suggest, cognitive linguistics and Cognitive Grammar view language as an integral part of cognition. Conceptualization is seen (without inconsistency) as being both physically grounded and pervasively imaginative, both individual and fundamentally social. Being conceptual in nature, linguistic meaning shares these properties. And being symbolic in nature—hence intrinsically meaningful—grammar does as well.

Grammatical meanings are schematic. At the extreme, they are nothing more than cognitive abilities applicable to any content. The more schematic these meanings are, the harder it is to study them, but also the more rewarding. Grammatical analysis proves, in fact, to be an essential tool for conceptual analysis. In grammar, which abstracts away from the details of particular expressions, we see more clearly (p.540) the mental operations immanent in their conceptual content. These often amount to simulations of basic aspects of everyday experience: processing activity inherent in conceptual archetypes is disengaged from them and extended to a broad range of other circumstances. In this respect, grammar reflects an essential feature of human cognition.

What sets us apart from other creatures is the extent to which conceptualization transcends immediate experience. Though grounded in such experience, the world we live in and talk about is mentally constructed through processes involving abstraction, conceptual integration, and subjectification. These means of disengagement are clearly reflected in grammar. At the semantic pole, grammar consists in abstracted patterns of conceptual integration. Grammatical meanings are schematic and often represent the subjectification of basic experience, consisting in the autonomous occurrence of mental operations inherent in such experience. In this way, grammar itself is a means of transcendence. Through the conceptions it allows us to construct and symbolize, we can engage the world in all its richness and complexity.


(1) For various reasons, however, conceptualization is not strictly linear. Processing occurs simultaneously in various dimensions and at multiple levels of organization. There is not invariably any natural sequence of access for the elements of a complex conception, nor is one fully adhered to in actual practice. And given the pressures of online processing, any actual rendition is likely to be discontinuous and complicated by factors like backtracking and reconceptualization.

(2) As default starting points, the speaker and the speaker's conception of reality have an ambivalent status with respect to natural paths. They are indeed starting points, but since defaults are taken for granted, they can be ignored for certain purposes.

(3) In expressions like This cord is only a foot long, the starting point for assessing length is zero. By contrast, This cord is long bases the assessment on the usual length of cords. For properties based on a norm, there is often a complementary property representing departure from it in the opposite direction—for example, This cord is short (meaning that its length is less than the norm, not that it is less than zero). See also Croft and Cruse 2004: ch. 7.

(4) Of course, we can achieve the same effect by using the complementary scale, with the opposite inherent directionality: Y surpasses X in {stupidity/lack of intelligence}.

(5) This metonymic interpretation is a well-established sense of Vietnam. It instantiates a general pattern productively applied to new occurrences. If everything goes wrong on your vacation in Rhode Island, you can subsequently describe the experience by saying Rhode Island was a disaster.

(6) Additional factors bear on the choice of possessor. One is information structure: the doctor's lawyer takes the doctor as given and introduces the lawyer as a discourse referent. Another is contextual accessibility. If you find a disembodied tail on the kitchen counter, you might wonder where the tail's rat has gone. Expressions like the car's owner, the hotel's occupants, and the country's ruler reverse the usual alignment (human > nonhuman) because the possessed noun derives from a verb (own, occupy, rule) and profiles its trajector. The identification of its landmark is thus a natural basis for singling out an instance of the type it specifies.

(7) These uses are frequent and readily come to mind as examples of possessive expressions. In some languages, the relevant nouns have to be possessed.

(8) While these are not atypical, there are numerous other kinds of nominal and clausal possessive constructions, in English and other languages (Langacker 2004b).

(9) Evidence for this profiling is the fact that Zelda's can stand alone as a nominal used anaphorically: This quilt is nicer than Zelda's.

(10) Stated more precisely, his refers to the entity possessed (the private office), with the president identified as reference point. Although his is only partially analyzable morphologically, at the semantic pole it still incorporates a pronominal element that participates in anaphoric relationships.

(11) For example, (10) might appear in a story about the director of the CIA, with prior indication that the president has come to CIA headquarters.

(12) The sentence is perfectly grammatical if him and the president are taken as referring to different people, but only judgments based on coreference are relevant here. Well-formedness is always assessed relative to specific interpretations.

(13) We saw this previously in the contrast between (10), In his private office the president was admiring himself, and (11)(a) ??In the president's private office he was admiring himself. The former is more acceptable despite the order of presentation.

(14) This process is happening in French, being most apparent in questions. It comes about when a sentence like Sophie, est-elle heureuse? ‘Sophie, is she happy?’ (where Sophie is a clause-external topic) is reanalyzed as Sophie est-elle heureuse? ‘Is Sophie happy?’ (where it functions as a clause-internal subject). For a general discussion of topic vs. subject, see Li and Thompson 1976.

(15) In the cases to be considered, the reference point is always a thing. A question worth pursuing is whether preposed adverbials (like the one in the previous sentence) should be analyzed as topic expressions that profile relationships.

(16) The term is also used with relative clauses (§12.2.2). The difference between topic and relative clause constructions is that a relative clause is part of the nominal expressing R and helps identify its referent, whereas a topic is a separate nominal whose reference is established independently.

(17) In some languages, the topic nominal does take an overt marker, Japanese wa being a well-known example. English as for is roughly comparable: As for that mural, I really hate it. The topic construction is then analogous to the one shown for possessive 's in figure 14.2 (apart from profiling and the nature of the target).

(18) This illustrates the fractal organization of language structure (§13.3.2).

(19) This is not inconsistent with the common observation that certain constructions, like the passive, allow a discourse topic to be expressed as clausal subject. The nature of a subject should not be confused with the factors that influence its choice.

(20) This is not to say that T1 and T2 are always nonoverlapping or clearly distinguishable. In the case of smash, the notion of causation (T1) is conceptually dependent on the change induced (T2) and thus invokes it schematically. Likewise, the change of state is hard to conceptualize independently of the force inducing it.

(21) Does the profiled relationship determine the focal participants, or conversely? Let me simply say that these factors are interdependent. It may well be a chicken-and-egg situation.

(22) For sake of clarity, the dominions are not represented.

(23) This contrast results in different patterns of usage. For instance, the caused-motion construction allows omission of the recipient, which is not on the main path of access: She could only give $5. In cases where there is no actual movement on the part of Z, the only option may be to use the ditransitive, which highlights the end result: The noise gave him a headache; *The noise gave a headache to him.

(24) A similar kind of backtracking occurs with clause-internal topics, as in That mural I really hate (fig. 14.5). Here, though, it is not a matter of deriving a higher-level predicate with alternate trajector/landmark alignment. Instead, backtracking occurs with a finite clause containing subject and object nominals, as a function of the special word order. The reference point relation it introduces is extrinsic to the conception of the profiled clausal process (fig. 14.6).

(25) I noted previously that identity is the limiting case of a reference point relation, where the path from R to T has a length of zero. Since it has so little content, identity is commonly expressed simply by juxtaposing two nominals, emerging as an aspect of constructional meaning (fig. 11.11). As in other uses, be gives temporal extension to this relationship (in case you were wondering what the meaning of is is).

(26) Or at least we think we do. If not, the illusion is quite compelling.

(27) Examples of other reasons for invoking a virtual instance are negation (I don't have a dog) and the description of desires (I would like to have a dog).

(28) These levels can also be described as mental spaces: “actuality” and a “generalization space”. Fictive entities occupy special mental spaces by their very nature.

(29) See, for example, CIS: ch. 5; Langacker 2005a; Matsumoto 1996a, 1996b; Talmy 1996. There is experimental evidence that the processing of fictive motion expressions is indeed linked to the conception of actual motion (Matlock 2004; Matlock, Ramscar, and Boroditsky 2004).

(30) Expressions like (25)(a)–(b) involve summary scanning along this path (fig. 4.7). At a higher level of conceptual organization, the entire configuration built up in this fashion is portrayed as stable through time, so run designates an imperfective process. Fictive motion expressions can also be perfective, e.g. The trail rose quickly near the summit. These involve sequential scanning, reflecting the experience of a moving viewer. Though actually different, the successively encountered portions of the object traversed are fictively construed as the same entity, which is thus perceived as changing position through time, just as in fig. 14.14(a).

(31) This is usually the case with speech-act scenarios (§13.2.3).

(32) This is not to be confused with sequential scanning, defined quite narrowly as the processing mode characteristic of verbs (§4.2.3): as the profiled relationship is tracked through time, its component states are accessed serially but without summation—that is, only one is focused at each moment of processing time.

(33) Each contrasts in this respect with every and any, which invoke abstracted scenarios that are based, respectively, on simultaneous viewing and random selection (fig. 9.13).

(34) Once more, the scanning is only fictive: we do not actually have to run through every sport in sequence in order to understand the sentence. It is enough to imagine doing so by means of a small-scale simulation of that experience.

(35) They do find the characterization reasonable when it is presented to them, however.

(36) See Heine, Claudi, and Hünnemeyer 1991; Hopper and Traugott 2003; CIS: ch. 12; GC: ch. 10. The term “subjectification” is commonly used in a related but slightly different sense. As defined by Traugott (1982, 1989), subjectification indicates a shift in meaning from something objectively discernible to something in the mental and textual realms—for example, the semantic extension of while from temporal to “concessive” import (‘at the same time as’ > ‘although’). For different perspectives on subjectification, see Athanasiadou, Canakis, and Cornillie 2006.