Being About table of contents   Chapter 7 bibliography   Chapter 8 Representing and the IPL   Ellie Epp home page


BEING ABOUT  Chapter 7. Representational effects

Abstraction and structural segregation

Metaphor

Act metaphor

Structural metaphor

Thinking


This chapter outlines a structural and dynamical way of understanding two sorts of specialized cognitive effect: abstraction and metaphor. Studies of both have in fact been pressing for structural/dynamical description; interaction or blending theories of metaphor, for instance, have seemed implicitly dynamical. As will be seen below, both abstraction and metaphor have needed an account of sublexical semantic evocation, which is handily understood as subnet activation.

Abstraction is usually studied in relation to the visual arts, or, in what has seemed an unrelated sense of the term, it has been a topic in cognitive psychology. Metaphor is most often considered as part of rhetoric or literary theory. In this chapter abstraction and metaphor are taken as instances of related cortical effects. Abstraction is described here as segregated use of structure that would normally be part of a wide net responding to a whole circumstance. The discussion of metaphor is divided into two sections: transient or inflectional metaphor is thought of as dynamical alteration of cortical nets as an effect of representational forms not normally used in the contexts in question. Structural metaphor is thought of as use of structure relevant to some particular base level domain in order to think about something else entirely -- usually some theoretical domain for which no base level capabilities exist.

Both abstraction and metaphor build off basic perception-action abilities, but they are also supported and enabled by structure formed through high-cultural practices. Especially when they are simulational, abstraction and metaphor tend to require continuing management by means of representational forms, which may themselves be perceived and produced, or only imagined.

One of the general claims of this book is that the humanly prestigious abilities we call thinking are culturally developed variants of base level aboutness. Abstraction and metaphor can be thought of as parts of thinking that exceed base level perception/action capabilities, and even exceed simpler abilities with language and other representational events, but as this chapter will begin to show, base level aboutness can be understood as foundational even to culturally dependent forms of cognition quite remote from base level presence.

Abstraction and structural segregation

The so-called concepts of extension, of far and near, gravity, rigidity, horizontal, and so on, are nothing but partial abstractions from a rich but unitary perception Gibson 1979, 261

Earlier chapters have described base level perception/action as facilitated by a wide net of cortical response to whole circumstances including objects, locations and events. Nodes or subnets of this wide net, integrated by recurrent connections, may at the same time be thought of as responding to aspects of whole circumstances.

Response to 'aspects' may be quite high order response to combinations of lower order facts. In Chapter 3 we saw neuronal groups in auditory matrices accomplishing differentiative response of multimodal covariant kinds. I described these matrices as complex or multidimensional filters. In Chapter 4, we saw neuronal groups in anterior temporal areas acting as object recognition, or object-constant, filters. We also saw neuronal groups in the hypothalamus responding as location-constants, and neuronal groups in parietal and frontal cortex responding in an act-constant way. An example of the latter would be the high order grasp constant neural groups in parietal-premotor gradients -- neurons that respond in the course of grasping, whether with the hand or the mouth -- or what have been called command neurons, that are active before a behavior such as reaching is initiated.

Unimodal sensory discriminations among color, form, object motion, and other sub-object features occur as part of object recognition and response, but these discriminations are also cross-categorical: an apple and an airplane may both be red, an elephant and an airplane may both be loud. There are also cross-modal feature discriminations: we can know something to be nubby either visually or by touch. Other sorts of discriminations are so high-order they might be part of any sort of sensory response, including perception of action; a judgment of beauty is an example. The existence of cross-categorical, cross-modal, and very high-order feature response suggests that subnets accomplishing these discriminations must be active in many kinds of wide net.

Act-, object-, and location-constant response, and sensory feature response of different kinds, would be included in the wide net of particular function as kinds of generalization accomplished automatically in the process of particular response. Different levels of general response must be active simultaneously in perception-action through-streams: we will deal with any particular object, presumably, by setting up a net that includes many of them. Dealing with other kinds of objects will set up nets with some of the same and some different subnet activity.

Very comprehensive aboutness may sometimes be called up through what would usually be a small part of a circumstance. We can respond globally on the basis of something we detect with one sense rather than all, or on the basis of a sensory sub-ability like timbre, color, or visual motion perception. We can begin to be about dinner by one whiff of scent.

At the same time, even when we are in a position to be in contact with many parts or aspects of something, we may respond by being about only one particular part or aspect. We are able to narrow attention, to focus and ignore selectively. Instead of a wide net of contextualized base level contact, what we have in these instances would be the segregation or isolation of a responsive subfunction. The possibility that this functional segregation may also be a structural segregation, at least to some degree, is supported by evidence of selective loss with local lesions.

In Chapter 5 I described the role of prefrontal cortex in 'directing attention' to parts, aspects, and features, as well as to particular objects and events. Prefrontal counterflow influences will be acting to inhibit activity in some subnets while increasing it in others. Recall Hanna and Antonio Damasio's hypothesis that particular aspects of simulational aboutness may also be evoked through counterflow activation of sensory or motor subnets, and that simulational attention to aspects, including simulation evoked by language, would be organized by means of convergent/divergent zones in association areas. Prefrontal cortex is likely to be involved when simulation of these kinds is deliberate.

Under this hypothesis we could guess that representational practices are often concerned with this sort of segregated activation of parts of wide nets. The use of names would be an example: the Damasios hypothesize that names activate convergence zones that in turn evoke distinct though distributed subnetworks. Calling the name of a person or thing has the effect of making everyone attend, for a moment, to that person or thing -- and not much else. Similarly, the name of something that isn't present can have the momentary effect of bringing us to be focally though simulationally about that person or thing.

Base level ability to segregate can occur at feature and superordinate categorical levels as well as at object and act level, where segregation at these levels is important to survival. Feature names or superordinate category names (vehicle or animal) are ways of evoking segregated response relevant to something about things or acts, rather than the thing or act as a whole. We can talk about red, in either the presence or the absence of something red, because we can selectively evoke neuronal groups responsive to the redness of things rather than to the things that are red. Response evoked by red is in this instance segregated or abstract.

Names effecting segregable feature and superordinate response form a reduced vocabulary in comparison to base level names of objects and events (Rosch 1978). Our response to these names can be felt as very general meanings, which are called abstract when they are evoked without their usual wide networks of concomitant response. Our more complicated sorts of play with sensory or high-order isolations of this kind have been elaborated through great cultural effort over millennia. Apart from a small cross-cultural core of segregational priorities, different cultures will have different sets of, for instance, color names, because they have taken different kinds of trouble with segregational evocation.

Situational generics: schemas

In English, schema, like its Greek original, includes at least three different but related senses. They include: 1) form or order, in the sense of systematic spatial or temporal arrangement; 2) a diagram or plan representing the spatial or temporal order of something; and 3) the planning (plan in the psychological sense, like the English scheme) thought of as responsible for this order. All of these senses can be felt in the various technical usages of schema.

Kant in the 1700s used schema as a name for mediating structures thought to be responsible for the experienced orderedness of cognitive function. Schemas of the imagination were for instance said to mediate between raw nature and perception. In its twentieth century uses by neuroscientists and psychologists, schema has continued to be a way of talking about causes of mental or behavioral order in the absence of detailed knowledge of physical structures responsible. Schema is thus related to other placeholder terms of this sort: concept, idea, category, conceptual structure. A difference is that, with schema, there is an emphasis on structure responsible for sequences of behavior related to things in the actor's surrounding world.

Schemas are postulated in terms of discrete acts, that is, of acts in relation to discrete objects or aspects of a circumstance. A chair schema would include memory for behaviors having to do with chairs. Because locating and sitting down in a chair require complex sequences of coordinated movement, a chair schema is thought of as including smaller schemas that must be combined. Arbib (1995, 831) uses reach and grasp as instances of schemas composable out of combinations of simpler schemas.

In the representationalism typical of functionalist contexts, a schema is thought of as a representation type that may be tokened or instanced or implemented in various kinds of material structure whose organization need not be studied. With the move to more fully imagined physical structure that attends distributed computation, schema has become a way of thinking about the relation between function and structure in terms of discrete but combinable structure units responsible for function units as analyzed into tasks and subtasks. Arbib thus describes schema as "a functional analysis which is linkable to a structural analysis, as this becomes appropriate" (1995, 831).

Understood in terms of competitive/cooperative distributed networks, schema instances can be envisaged as neural networks, which may be combined by recurrent linkages. Any given schema may be widely distributed, and a particular subnet may participate in many schemas. In this context, schema enables talk about perception and action as inherently related and temporally extended: schemas can be perception-action units rather than perception or action units. Arbib invokes as an example McCulloch and Pitts' (1943) distributed model of visual response linked to motor response in one continuous process.

Where schemas can be imagined as the wide nets integrating perception, action, and motivation -- whole behaviors in relation to objects or events in whole contexts -- schema, like frame, comes to have an added connotation of global situational organization. We speak of spatial schemas, event schemas, social schemas, schema complexes.

Hebb (1949) and Piaget (Beth and Piaget 1966) have thought of skilled act sequences as learned through the strengthening of associative connections among neuronal groups facilitating successful actions. New schemas might also be assemblages, or modified versions, of existing schemas. They might arise to modulate existing schemas: "Thus, for example, schemas for control of dextrous hand movements serve to modulate less specific schemas for reaching with an undifferentiated grasp unadapted to the shape or planned use of an object" (Arbib 1995, 832). Once learned, a schema or perception-action process could be evoked in response to sensory contact or in response to other sorts of cortical activity.

In a non-functionalist, non-representationalist framework that tries to imagine structure from the outset, the notion of a schema is closely related to the notion of structural aboutness. If we think of aboutness as a state of a whole body, not just of a brain, schema would be embodied in all the spatiotemporal structures that are the means of being able to do something or be with something. A chair schema would involve chair-related structure of muscles and eyes as well as brain.

In psychological or linguistic contexts, schemas are often called image schemata; this usage is unfortunate in the sense that, like other sorts of aboutness, global perception-action structure is also evoked when we are perceiving and acting rather than imagining. Schemas are simulational when global structure relevant to objects or situations is evoked in the absence of those objects and situations. In these instances, structure may or may not include simulational visual structure. What is neglected when we think of schemas in terms of images is the fact that they are overall perception-action complexes whose sensory involvement may for instance be primarily kinesthetic.

Abstract use of schemas

The second of the senses of the schema named above, schema as a diagram or schematic, brings with it a sense of abbreviation, simplification. Neural shortcuts wired-in with practice could be considered schemas, in the schematic sense, for a behavior. Hebb and Piaget describe behaviours and the neural structure facilitating them as simplifying with practice: a motion that requires much neural tissue, many connections, when it is being learned, would form more compact assemblies with time, and these more compact assemblies would be the basis for skilled perception and motion. A decrease in active tissue that accompanies increasing skill has been observed in PET studies including Ungerleider (1995, 771-2) There might similarly be schematized or abbreviated forms of comprehensive perception-action or situational organization. Chapter 8 will describe the formation of summaries of these sorts in the language-specialized left hemisphere.

I have introduced this discussion of schemas in a section on abstraction because there is evidence is that schematic aspects of perception-action wholes, like other sorts of sensory and motor generic, can also be evoked abstractly. Generic response elements (object constants, feature constants, location constants, and so on) normally included as parts of wide net response to whole circumstances, but sometimes activated in isolation, might also be evoked as parts of a sort of skeleton network. Generic elements connecting generic elements might also be needed. Edelman calls schemas "high order functional complexes", "mappings of mappings that have to do with positions and states of the body as it relates to objects/events" (1987, 195).

Active in isolation, generic substructures might organize a sort of action precis, simplified and temporally abbreviated, and yet including whole sequences with motor preparation and release, simulational sensation and eye motion.

Evoked abstractly -- for simulational purposes, in representation contexts -- the high order generic would organize a sketchy simulation. Suzanne Langer notes that when children play that they are eating, their gesture is a sketch gesture, a back and forth wave from table to mouth with fingers held as if closed over a spoon (1942, 156). In play, in daydreaming, in reading, in thinking, we do not always need to imagine in detail; but the sketch-evocation can nonetheless include viewpoint, oriented location, agent, act, objects and products, motion and force. There can be imagined motor participation too, imagined gesture or eye motion or orienting shifts of the whole body.

About abstraction

Like schema the notion of abstraction has several senses. It can suggest superordinate or generic response to something like chairness or eatability 'in the abstract'; this sort of abstraction can be thought of as part of concrete dealings with chairs or eatables, an 'abstract concept' being embodied during particular encounters. It can also suggest segregated function, where aspects of aboutness are evoked apart from their usually integrated functions in base level presence: the grin without the cat, color without colored things. And it can suggest perceptual thinness. These senses of abstraction are not equivalent but thought of in cortical terms they can be seen as related. Superordinate or feature response evoked in isolation is likely to be the means of relatively meager perceptual or simulational experience. Where a wide net is working to create a comprehensive specific aboutness -- where it is being used to perceive or imagine something by means of many matrices at many levels of many through streams, we could talk about semantic depth of response. Where a relatively sparse aboutness is being invoked by means of only a few matrices, we could talk about semantic shallowness or abstraction.

In this way of thinking about it, abstraction can happen at different levels in a net: there can be abstraction by means of primary sensory cortex, or abstraction by means of high-level act constants in the premotor, or abstraction by means of high-level object-constants in the ventral stream, or many other versions of unimodal or multimodal segregated effect.

If action-perception networks evoked in isolation are very high order, or very abbreviated -- if high level generic response is evoked without much primary or unimodal associative activity in the net -- there might be little or no sentient participation. This meagerness of experience -- which is more likely in simulation, because presence will set up sensory activity in primary areas -- seems to be a characteristic of schemas generating linguistic structure.

When it is simulational, linguistic function seems always to be abstract in at least some and often all of the senses of the word given above. We can speak, read, calculate, and think as fast as we do because we do not set up deep simulational nets; we are using circuits trained to a very abstract calculus of simulational effect.

The various sorts of abstraction occur with non-linguistic media too. Cultures play with segregated, high level, and semantically thin evocation. The arts are often elaborate specialized dissociations from full environmental perception and action. A simple outline drawing of an object may evoke object constant and some minimal form response but little other sensory simulation -- we're minimally about that object when we see the drawing.

Cultural play and traumatic disorder can have limitations in common: art practices may build networks that mimic the effects of cortical damage when practitioners, with great effort, rebuild base level connectivity to be able to isolate and hyperdevelop local regions. Idiot savants of music and mathematics are evidence that isolated representation-supported function can be very highly developed.

Black and white movies and color field painting set up kinds of seeing that abstract from the semantic depth of environmental vision, but are not abstract in the sense of being sensorily thin. Music abstracts in similar ways: we usually listen to voices, not to persons singing, for instance, but many acoustic subfields are active; musicians arrange for more timbral and rhythmic response than is usual with environmental hearing, and so we don't experience music as thin.

Smalley describes kinds of musical evocation that are abstract also in the sense that they engage high level generic networks, and these are some of the rare wonders of electronic (and other) music composition. A piece can evoke extraordinary exact abstract visualities. We can seem to see/hear textures, contours, transparent planes intersecting each other at angles, colored lines moving (left to right) in relation to each other. As hearing they are part of a rich mix, but as vision, they are thin.

It should be remembered that although cultural practices are abstractive in these ways, the extent of abstraction depends partly on the user of a representation. The bare name of something may, in the right circumstance, evoke a wave of memory with great semantic depth. The same is true even of rich media like film: when Meryl Streep reaches a hand to touch a shoulder, one viewer, but not another, will seem to feel muscle under a shirt.

Metaphor

What I am calling transient metaphor is sometimes, in linguistic contexts, referred to as lexical metaphor, but since transient metaphoric effect can occur also in nonlinguistic media, I am giving it a name primarily designed to distinguish it from structural metaphor, which will be discussed later. An example of transient use of metaphor is Wordsworth's "brooks / Muttering along the stones" (1799/1979).

A transient/structural distinction should not be taken as implying any sharp division between metaphoric kinds.

The long history of the theory of lexical or transient metaphor has pointed up problems with taking the representational object as the effective locus of representational effect. Puzzlement about metaphor derives from thinking of metaphor as something about the representational object itself: this phrase, this photo, this gesture, is 'a metaphor'. It is better to say that a representing object or event has metaphoric effect in certain contexts -- that is, in contexts that set us up to take it metaphorically.

If we start with thinking of the wide net active when we use representational forms, we can understand metaphor as like perceiving, like comparing, like similes, like illustrations, like basic predication -- one among other ways of organizing cognitive structure so that we are about particular things in particular ways by means of it.

I.A.Richards (1936) describes metaphor as perceiving or thinking or feeling about one thing in terms of another. What might this mean in terms of cortical dynamics? Metaphoric effect might be thought of as a kind of inflection, a partial bending or warping of established simulational aboutness, like the effect of the subjunctive case of a verb, which also evokes non-assertionally. We could say a metaphoric term in a paragraph drops a pebble into a pool where a larger stone has very recently been dropped; wave structure on the surface of the water will be the integrated result of an interaction of the effects of both stone and pebble.

These preliminary metaphors should be taken as introductory prestructurings of effects to follow.

Referring from and polysemy

Here, however, we are looking at it from the other end, asking not "How is this word used?" but "How is this meaning expressed?" A meaning may be realized by a selection of words that is different from that which is in some sense typical or unmarked. From this end, metaphor is a variation in the expression of meanings. Halliday 1994, 341

Traditional questions about metaphor have been put in this way: if a word 'has a meaning' which is its relation to an object, how can a metaphoric use of a word succeed, since in such use the name by definition cannot be thought of as having any prior relation to the object it is used to name? Traditional questions about polysemy have a similar form. How does a word get a new meaning -- how does it develop a relation to a different class of objects?

When we initiate one action rather than another -- when we turn our eyes to the left -- some particular configuration of some particular subnet activates effectors. Pronouncing one word rather than another is an action like others, gated from a wide net.

If we think of meaning as the structural aboutness from which wordings are organized, we can think of homonyms (bold and bowled) as instances where different states of a wide net can nonetheless sort toward the same, or almost the same, articulatory motion, the way a saccade can be organized from different areas of premotor cortex.

Polysemy can be understood as a variant of homonymy. Many words are polysemic: they can be used in conditions that vary largely but may have a range of partial commonalties, the kinds of family resemblance Wittgenstein noticed (1956), that allows grain to be used for seeds, cereal plants, minimal quantities, the direction of wood fibers, the size and texture of the particles of any surface, the direction of cleavage of a mineral, one's natural disposition or temperament. When we read that list we feel not complete differences but as if a series of partial shifts. It makes sense that a net configured to remember the surface texture of leather would release the same word as a net configured to remember the surface texture of a cut round of wood.

Words are used in new ways as new activities need naming:

A plot first meant a physical site and building plan, then the stage director's plot or blocking plan, then the action or story blocked out. In the sixteenth century a frontispiece was a decorated front of a building, then the title page of a book ornamented with a figure, usually the allegorical representation of a building, and finally the illustrated page that precedes the title page. Sweetser 1990, 219

Different cognitive conditions gate the same words necessarily. When we're stuck for a word we can feel a meaning but the net isn't setting up a wording. Then we'll gate off some part of our meaning. Jesse's dad asks him if he knows the difference between a breaking plow and an ordinary plow. Jesse is very bright but he's only two. He says "The breakin' plow has bigger ... knives." He had to sit with it for a minute. He knows the fact but he doesn't know the word. When he's sitting with the fact he knows, imagining the breaking plow, dwelling on it, he's energizing the structure by means of which he's thinking it. Some part of it, the cutting-edge part, the dangerous part, is energized enough to gate a word. He has worded from the meaning he was.

Eva Sweetser (1990) describes the orderedness of shifts in word use as metaphoric. How verbal metaphors are made is not so different from other kinds of talk. There's structure and it gates a word. When Jesse says the breaking plow has bigger knives he is not being metaphoric but he is doing what people do when they use metaphors -- he is referring from a complex meaning -- gating a word that runs off the part of a net that's hyperenergized for some reason.

Hyperenergizing a subfield

The question why predicates apply as they do metaphorically is much the same as the question why they apply as they do literally. Goodman 1968, 78

Polysemy is like metaphor in the sense that it has to do with unusual wordings. An understanding of the systematic extensions of word use registered in polysemy shows that words can be used in new contexts because the new context shares structure with the old. Metaphor can also be understood as an effect of shared structure which regulates emphasis.

Anomalies of emphasis can happen when we are seeing different things in the same context. Think of a blue-eyed person wearing a blue sweater. It's a basic perceptual effect, organized by co-presence.

A less obvious form of the effect is if I'm looking at the beech tree and my friend, looking at a runner, says "muscles" (or even "no muscles"). I see long muscles on the beech tree's trunk.

Taking a further step from perception into simulation, we have a phrase that reads: "The birds rose swiftly like a handful of grain flung" (Woolf 1977). In what respect, she doesn't say; we seem to see the separate points moving together, but not exactly together, accelerating and decelerating in dynamically related arcs.

Now think of syntax as a way of routing activation to subnets of simulational structures. Predication in a sentence can have the same effect as putting on a blue sweater: one part of a wide net is organizing the way another part gets hyped.

One more step and we're at metaphor proper. My friend and I are looking at the beech tree. "It's so muscular" she says. Or the whole set-up can happen by means of a sentence; "Cloaked in red velvet, the sumac's seed heads stand pensive in autumn light". It doesn't always work. The blue sweater doesn't work either if it isn't the right shade.

Synaesthesia and similarity

Synaesthesia is a systematic but perceptually irrelevant coactivation of different subnets of a wide net, so that we feel a sensory quality as belonging to an object that doesn't, or can't, have that quality. Word-color synaesthesics who experience vivid color when hearing non-color words show unusual activity in the area of left ventral temporal cortex that is active when normals remember colors of common objects (Paulesu et al 1995 cited in Martin, Ungerleider and Haxby 2000, 1025-6).

When I'm tired a sudden noise can make me seem to see a dim round patch spread evenly with little dots. Presumably some midbrain bimodal audiovisual matrix is sending a split signal into auditory and visual cortex (Barth et al 1995, 177). I'm not tempted to ascribe the dim dotted spot to the sound, or the sound to the spot, so I don't call this synaesthesia. It's perceptually-entrained simulation of a rudimentary sort. (The Welsh woman who says Mary is pale mauve and Charles dull red may not really be ascribing either -- just noticing that it usually happens that way.)

When I seem to see a pain as a three-dimensional yellow shape, I half-ascribe. I'm experiencing it as yellow while knowing it's not. I'm experiencing it as changing shape, which it is -- it is a changing shape of muscle tension -- and experiencing that shape as visible, which it is not. That is synaesthesia.

When I look at the beech tree and see it as muscular I'm half-ascribing in the same way. I'm seeing the beech and I'm imagining it as muscular. I'm not imagining muscles separately from seeing the tree. Like feeling the shape of a pain and imagining it yellow, I'm seeing the long fiber-bundles of the beech trunk and imagining them muscles.

In both instances the imagining is felt as intrinsic to the sensing. Seeming to see what I'm feeling helps me feel it clearly. Seeing the tree trunk as muscular helps me see the shape of the fiber-bundles clearly. Muscle-tension-feeling structure is in some sort of helpful interaction with shape-seeing-structure; tree-trunk-seeing structure is in facilitative interaction with muscle-seeing or maybe muscle-feeling structure.

We tend to think of the means by which we see or imagine different things as if they were themselves different things -- which can resemble each other, correspond to each other, be mapped onto each other, or be superimposed like screens whose interaction makes 'similar features' 'stand out'. It is possible, though, that we can be thinking different things by partly shared means. If we are looking at a room and imagining a field, the nets by which we do so will have basic orientation in common -- we are seeing something below, above, to the right, to the left. And maybe much more than that; the means by which we understand field and room may be deeply interwoven in virtue of common cognitive origin.

If we can think of ourselves seeing the beech tree and imagining muscle by means of a net that is already one net, we do not have to think of transport, transfer, or projection. There will be interaction, but there is always interaction.

We are not comparing our perception of the tree and an image of muscle and finding 'a similarity'. We are structured in a way that enables us to see different things as similar. That may mean we are using common structure to see and/or seem to see different things.

Metaphor dynamics

selects, emphasizes, suppresses and organizes features of the primary subject by applying to it statements isomorphic with the members of the secondary subject's implicative complex Ortony 1993, 28

I suggested above that a term used metaphorically can be articulated in the first place because the structural aboutness being spoken from already includes subnets whose connections tend to gate that term. When it is heard or read, conversely, the term will evoke substructure within a wide net that has already been established, and the result may be an inclusion or attachment, like a synaesthesia.

Metaphor is from meta - pherein: to carry over or beyond. Where cognitive effects are thought in terms of inner representations, there is often talk of mapping, projection or transference from one domain of representation to another. Imagining cortical activity in terms of standing reentrantly-connected wide nets makes it possible to suggest that metaphor, like synaesthesia, like predicative emphasis, can more easily be thought as an effect of a patched-in subnet which is transient and unusual but effective under the circumstances.

Since activation in a wide net is reentrant, changes to the network as a whole could result from the activity of these unusual subnets -- a strengthening of some connections and an inhibiting of others, allowing some parts of the working net to fall below the level of activation that makes them sentient and hyperactivating others so we are suddenly noticing something we didn't notice before. Metaphoric effect can then be understood as related to abstraction, as a kind of representationally managed dynamics of subnets within wider networks.

There have been hints of this relation between abstraction and metaphor in earlier writers: Suzanne Langer suggests metaphor is possible because perception is "abstractive" (1942, 145-49), and Arnheim says the "pairing of two images throws into relief a common quality and thereby accomplishes a perceptual abstraction without relinquishing the context" (1969, 62).

Another way to say it is that metaphor and abstraction both press for a theory of sublexical semantic effect, as do other phenomena like synaesthesia, comparison, and even adjectival and other predication. The Groupe de Liege describes metaphor as an operation that has to do with "suppression or more exactly ... bracketing of one portion of the constitutive semes of the lexeme used" (Leguern quoted in Ricoeur 1975, 345).

Does a semantics that speaks of sub-lexical units imply word-response constituents that should be thought of as composable inner representations? Not necessarily. What is implicit in sublexical effect is a notion of meaning as a constellated aboutness which is not necessarily a constellation of representation parts that are productive and compositional. Semantic analyzability falls easily out of a wide net description of representational effect. We can think of nodes or subnets in distributed semantic effect as responding to parts or aspects of something, as found in PET studies by Martin, Ungerleider and Haxby (2000, 1032), without thinking of them as 'representing' properties, features, attributes and all the rest. If we talk in terms of internal representation-parts or symbols combined in different ways to produce the array of meanings we find for a single word, we cannot imagine dynamical effects of the combinations. We can't, for instance, imagine phase effects or other forms of dynamical interaction.

Metaphor is like punning, in the sense that both are out-of-context effects. With metaphor, one context is used transiently, to inflect another. With puns, one word has two or more simultaneous, or possibly not quite simultaneous, effects, each tending to evoke a different contextual whole. How would a punning term work? We are normally about one context at a time through phase integration over the whole of a wide net, Edelman suggests (see Chapter 2). When we're about another context, the wide net would be in phase at a different periodicity. Could a punning term set up a transient phase shift -- just a blip?

Like a punning term, a metaphoric term might set up a transient out-of-phase network. The effect might be a limited inflection of the dominant phase. How much of an effect there is might depend on how much time the alternate phase is given.

Thought of in terms of network dynamics, rhetorical figures like metonymy and synechdoche probably are not very different from metaphor in their manner of operation: they are all ways of evoking substructure effects in wide nets constructed in response to representing forms.

Act metaphor

functional complexes ... akin to gestures, postures, and their transitions Edelman 1987, 220

The MacIntosh desktop lets me 'drag' files to a trash icon at the bottom right of the screen by dragging the mouse physically. Many of us have a waste paper basket just there, under the desk next to our right foot. An actual arm movement in its direction is used to get something out of sight -- a remarkably intuitive interface metaphor.

'Moving a cursor' by moving a mouse on a mousepad is itself a version of the reach and drag metaphor (since, on screen, pixels are being activated but nothing is being moved). A variant of this metaphor has paraplegics, wired to computers via electrodes embedded in a frontal reach region, moving a cursor by imagining that they are moving it (Begley 1999, Headlam 2000).

With training and with electrodes placed differently, they would be able to move the cursor by imagining apples, but the actual arrangement has the advantage of analogy. In this case it uses simulated rather than actual motor control. In both my examples, reaching or seeming to reach are used metaphorically, to accomplish something that isn't a reach.

Action used metaphorically may not be not purely motor. It may, often in a sketchy way, include the whole of an act, that is, it might include emotion, somatosensory as well as other sorts of sensory response, and also prefrontal and premotor aspects of selection and intention. It could include the temporality of action in schematic form -- readiness, starting, continuing, finishing, and so on.

Somatic-motor act metaphor is part of punctuation and spacing practices too. We write lists so they look like stacks or rows of objects we examine one after the other. When we want to present part of a text separately from its context we enclose it in parentheses or quotation marks that curve like palms facing each other around an object.

Another sort of act metaphor, called syntactic metaphor in Lakoff and Turner (1989), and iconic syntax by Freeman (1991, 161), is the literary effect by which a linguistic form "does what it says". It is difficult to think up such a device intentionally; poets know it just 'occurs to them'. Such an occurrence, inexplicable in classical linguistics, can be more easily understood if we imagine linguistic form as being generated from the very same central structure that is the meaning expressed.

The fact that a single overall network may be organizing all aspects of a discourse also suggests why examples, analogies, and graphic illustrations often are keys to understanding how someone is thinking a topic. An implicit metaphor structuring someone's discourse can be discovered in the examples given, because these figures are also being generated from the structure being used to think. The figure may do its work without our noticing it: when we understand the example, the structure that is our understanding of the example is the structure we need to understand the sentence or text as its creator is understanding it.

Structural metaphor

figures of thought Langer 1942, 193

This chapter has been describing significant representational effects -- representation-supported ways of using wide nets active when we are about something, either in its presence or simulationally. Abstraction has to do with segregated evocation of subnets; polysemy, synaesthesia and transient metaphor, with effects of shared substructure evoked in segregated ways. With structural metaphor we come to the most thorough-going of the representational effects, which is so comprehensive that thinking as we know it is impossible without it.

Structural or conceptual metaphor is the systematic structuring of one domain in terms of another (Sweetser 1990). Structural metaphor relates the notion of schema to the notion of metaphor: we use schematic structure metaphorically, to think about one context in terms learned in relation to another.

Babies younger than two will sort objects by touching them sequentially, and later, by aggregating them in a pile (Mandler 1992, 601). Categorizing operations can in this instance be seen to be the same thing as sequential touching and aggregating. In Piaget's sense of it (1971), later means of more abstract categorizing would continue to include structure having to do with sequential touching and aggregating, as the child assimilates different kinds of objects to the schema -- sorting new kinds of objects, first, and then using a covert schematic remnant -- simulationally, in act metaphor -- to think about non-things, like numbers or sets, that are felt to be objects as a consequence of the schemas used to think them.

Many of our topics are so deeply a tissue of this sort of metaphor that there is no non-metaphoric remnant. Others are partially metaphoric, or are thought by means of complexes of schemas. Lakoff gives as an example the spatial vocabulary we use in daily talk about nonspatial things: time as motion past a point, state as location, category as containment, communication as object transfer, and so on (Lakoff and Johnson 1999).

Act and spatial metaphor are found on a concrete-to-abstract continuum. Act aspects evoked as part of a metaphoric schema might be high-order act constants segregated from participation of sensory tissue. Piaget describes a progression in development from sensory-motor thinking, to pre-operational, to operational thinking, the latter characterized by a lack of evidence of sensory-motor involvement. Examples would be abstract notions of support, balance, causation, or force, which Lakoff suggests have been built on schemas originally formed from kinesthetic experience of "compulsion, blockage, collision, diversion, enablement, attraction and release," which continue as a barely experienced gestalt.

Much of the evidence that has been gathered for structuring participation of high order perception-action schemas has come from linguistics, where schema theory is used to investigate linguistic constructions like object-action duals and the caused motion construction (NP-V-NP-prep-NP). George Lakoff calls the structures generating these forms generic level schemata -- very general or superordinate schemas of practical action. Other generic level schemas found in linguistic constructions are containment, contact, paths, center-periphery, source-goal-trajectory, object/location pairs, object/aspect pairs, and the subject-verb-object construction. The subject-verb-object construction, ubiquitous in pidgins, is for instance the form of first sentences in every language.

Evidence for the existence of base level schematic structure of this kind has also been drawn from studies of lexical change and innovation, inference patterns, language acquisition, posture and gesture during speech, ideogramic writing, and from ASL studies (Lakoff and Johnson 1999).

Thinking

What we mean by thinking can range from the very abstract to the very concrete. Thinking can be experienced as part of located perception-action, of presence: we can experience ourselves as thinking when we are deeply engaged in building a rock wall, although little of our involvement is verbal or 'abstract'. An eagle fledgling lifting from a snag for the first time can be seen to be thinking about its next move.

Even when thinking happens as part of present action, though, thinking tends to be ontologically mixed; in humans, even thinking on site, thinking for purposes of presence, most often includes simulation and is representation-dependent. Thinking processes may use actual representational objects/events -- written or spoken language, numbers, sketches, gestures -- but often the thinker imagines using them, and uses or imagines using them to support imagining something else. To talk about thinking, then, we need to have in place a prior account of perception and action, simulation, and representing practices.

Recap

The progression described in this work begins with base level perception and action. For some of the primates, base level function comes to include gestures and glances used to coordinate joint attention or to manage aboutness in another creature. Among humans, rudimentary language may be used in a similar way, to coordinate social action in the here and now of situated presence.

In a next step, representing objects or events can be used to evoke aboutness relevant to non-present objects or events: something that is not, but was or will be, present: something out of sight that can be reached by walking over there. Next, representation can be used to evoke simulational aboutness that is largely unrelated to the here and now, as when a cave painter or a story teller 'transports' an audience, organizes the aboutness of other people so they are seeming to perceive, and to feel and understand, a circumstance that may actually exist nowhere.

By a further shift, this ability to organize simulational aboutness is used, as structural metaphor, to organize a much more abstract play of simulational aboutness -- that is, the sort of aboutness we call abstract thought.

To accommodate the whole range, from the first stage through to the last, there have had to be many changes in the physical structure of brains and bodies. Representing practices need hand and cortical hemisphere specialization (to be described in chapter 8). The full play of representation-evoked simulation in its metaphoric employments has also needed the structural alterations each human child undergoes in the process of enculturation.

Given this sort of progression -- in which each stage must build on the last -- every kind of thinking, no matter how remote from concrete concerns, must make use of structure evolved, developed, or learned in contact with the physical world. What is common to all stages and levels of cognitive function is the embodied nature of structural aboutness and the wide net character of cortical activity.

Thinking and structural metaphor

Aspects of base level aboutness found to be vividly present in the most non-basic cognition include many sorts of structural metaphor.

Francis Yates (1964) describes mnemonic systems that use imagined motion through memory palaces; nurses may be taught medical diagnosis procedures that involve imagined navigation through a series of branching paths. In these sorts of thinking, act metaphor is used strategically and consciously.

In many other sorts of thinking we use base level structural metaphor unconsciously and necessarily.

Most of our stock of abstract and technical terms is derived from names used in practical contexts at earlier times, that is, in farming, food preparation, family, community, war and trade. (Filter, for example, derives from the Latin feltrum, a cloth, which would have been used for straining fluids.) Evidence is accumulating that our stock of theories similarly makes use of schemas developed in practical contexts. Schemas of practical aboutness in these instances are the structures of theoretical discourse. The implication is that thinking is simulational not only in its evocation of remembered things or events, but also in its very structure.

Theoretical metaphor can be emotional or attitudinal, as when we relate to the cosmos as if it were a parent, deal with illness as if we were engaged in a war, or try to 'rise above' a topic and see 'the whole field' at once. There are scale metaphors, as when we think long periods of historical time as if they were acts or events with beginnings, middles and ends. Spatial and act metaphor are ubiquitous.

Our ways of thinking armchair domains have been notoriously unstable, Lakoff remarks, because their metaphoric foundations are unconstrained:

Our most important abstract philosophical concepts ... are all conceptualized by multiple metaphors, sometimes as many as two dozen. What each philosophical theory typically does is to choose one of these metaphors as 'right', as the true literal meaning of the concept. One reason there is so much argumentation across philosophical theories is that different philosophers have chosen different metaphors as the 'right' one, ignoring or taking as misleading all other commonplace metaphorical structuring of the concept. Philosophers have done this because they assume that a concept must have one and only one logic. Lakoff and Johnson 1999, 71

Giovanni Battista Vico, who could be said to have founded cultural studies in the 1600s, commented that even the sciences are formed upon a basis of "imaginative universals" or root metaphors. Vico assumed that the sciences work toward a less metaphoric structure as they mature, but Lakoff and other contemporary metaphor theorists suggest that for many sorts of high cultural discourse, there is and can be no non-metaphoric base, so that progress is a question of discovering what can and cannot be done with the implicational structure of any given metaphor: "Not to dis-use but to re-use metaphors, in the endless search for other metaphors, namely a metaphor that would be the best one possible" (Ricoeur 1975).

In this spirit, a complexity theorist at the Santa Fe Institute describes changes in the contemporary physical sciences in this way:

Instead of relying on the Newtonian metaphor of clockwork predictability, complexity seems to be based on metaphors more closely akin to the growth of a plant from a tiny seed, or the unfolding of a computer program from a few lines of code, or perhaps even the organic, self-organized flocking of simpleminded birds. Mitchell Waldrop 1992, 329

This view of the metaphoric organization of scientific thinking has implications for the nature of innovation in science, Waldrop observes:

... from this point of view, the purpose of having a Santa Fe Institute is that it, and places like it, are where the metaphors and a vocabulary are being created in complex systems ... So what the SFI will do, if it studies enough complex systems, is to show us the kinds of patterns we might observe, and the kinds of metaphor that might be appropriate for systems that are moving and in process and complicated, rather than the metaphor of clockwork. Waldrop quoting William Brian Arthur (1992, 334)

If we understand that our theoretical discourses are grounded in structural metaphors, which in turn are elaborated uses of base level schemas built into the body through the whole of a childhood, we could consider what it might cost, in terms of body-restructuring, to change the metaphors we think with, and what that cost may have to do with resistance to paradigm shift.

Thinking as act metaphor

Thyncan, > OE to seem

A dictionary defines thinking as mental action. We speak of acts of thought. There is a sort of rarefied muscularity that may be felt in thinking, and sometimes a sense of immaterial motion. The sensations of decision, often part of thinking, are like muscular holding and releasing. More generally, thinking seems often to involve structural metaphor which is also act metaphor. We may 'lay out' 'the parts' of a problem by arranging papers in piles. We may say that in pondering a topic we 'turn it' this way and that, or that we 'look at it' first one way and then the other.

Seeming to speak is one among many kinds of seeming to act that are concatenated as thinking. Theorists describing an imagined speech component of thinking have included Piaget, Dennett, and Watson. The Russian psychologists Luria and Vygotsky describe a process they call internalization, by which children think first in overt, social speech, and later think by speaking when they are alone. Covert or simulational speech, developed later, gradually takes over cognitive operations that were earlier supported by speaking and hearing speech. Children will comfort themselves with imagined reassurances, or resort to seeming to hear themselves saying what they do not dare say aloud. Vygotsky reports that in this internalizing process, nouns may be replaced by visual simulation, but verbs continue to be as if spoken and heard (1986, 173).

Ong (1982) notes that sustained thought is often described as a dialogue, and that thinking may become more speech-like, as well as more complex, in the literate, who have long-sustained practice in the sorts of seeming to speak and to hear speech that are given with reading and writing. Many writers report as-if taking dictation by seeming to hear a voice. At least one writer says she seems to read her text off a sort of teleprompt (personal communication).

Speech is not of course the only action with representational media that may be imagined as part of thinking. Thinking of London can include seeming to look at a map of the Underground. Chapter 9 will describe examples of other kinds.

Frontal cortex and act metaphor

certain aspects of the so-called 'inner life' ... which have formal properties similar to those of music -- patterns of motion and rest, of tension and release, of agreement and disagreement, preparation, fulfillment, excitation, sudden change, etc. Langer 1942, 228

There is also a tradition describing thought without sensory simulation. The Wurzburg group (which included Wundt, circa 1908) investigated forms of 'imageless thought' they considered "unconscious, holistic, non-sensory and at the core of meaning" (Honeck 1980, 26).

The Wurzburg description is plausible in several ways. Scrutiny and manipulation of inner images is a poor metaphor for any sort of cognition, as I have said. Neither is thinking limited to visual or even to any sort of sensory simulation, since it can include activity of motor and other frontal areas. The Wurzburg Group may also have had in mind some of the more subtle aspects of attending, planning, and will or decision -- aspects that are act-simulational but not sensory.

In Chapter 5 I described prefrontally initiated back-connections active with deliberate attention, in the course of perception and action or in the course of deliberate simulation. LaBar et al, who found an area of prefrontal cortex active both during perception and action and during simulation, propose that the neuroanatomic overlap observed

relates to the shifting of an attentional focus, irrespective of whether the shifts occur over space, time, or cognitive domains ... Covert shifts of attention activate this common set of brain regions regardless of whether the shifts are endogenously or exogenously driven. LaBar et al 1999, 702

Deliberate attention can include deliberate considering of options, that is, deliberate remembering or imagining of alternative courses of action with their outcomes. Mesulam guesses that the prefrontal's widespread corticocortical connections could allow it to organize possible scenarios while suspending motor response (Mesulam 1990, 608). Considering alternatives would need prefrontal participation in hyperenergizing, hyperconnecting, and sometimes segregating, structure within simulational perception-action nets.

Sensations of decision included in thinking could derive from covert activity in eye and other motor areas. Motor involvement in deliberate imagining usually includes the frontal eye fields, as described in Chapter 5. There seems in general to be a particular relation of intentional action and prefrontally initiated eye movement. We could even guess the experience of willing, the muscular feel of will, is built around the ability of the prefrontal to direct the eyes: as if intention -- whether in act or in thought -- uses eye shifts as its first move.

Spatial schemas in thinking

The locus of reason (conceptual inference) would be the same as the locus of perception and motor control. The question from the viewpoint of the brain is whether conceptual inference makes use of the same brain structures as perceptual motor inference. Lakoff 1987, 231

abstract reasoning is a metaphorical version of spatial reasoning Lakoff 1995, 122

We stand looking at a scene or rush into it; we keep track of the many things there are on a desk or in the hidden shelves of a cupboard: spatial function is a way of being about many things at once, and so it is a paradigm of complex comprehension -- "that faculty of keeping a large number of relationships present in one's mind that distinguishes all mental achievement" (Gombrich 1977, 261).

Orientation, or task axis organization, is central to many kinds of thinking, as it is to the coordination of perception and action in located function. The act-metaphoric operation of attention maintenance and attention shift networks is required for deliberate simulation, which is often part of thinking; as described in Chapter 5, deliberate imagining generally seems to include or be driven by motor areas directing eye movements. Further, since we cannot use representational objects and events without orienting toward them, the part of thinking that involves use of representations -- real or imagined language, numbers, diagrams -- would necessarily include aspects of actual or simulational sensory orientation. These simulation- and representation-related forms of task-axis organization seem also to be present in act-metaphoric use to organize what could be called a basic posture in thinking, an i-it relation by which we 'face' or 'confront' our topic.

Evidence accumulated by Lakoff and others is that structure important to sensorimotor spatial function is used act-metaphorically in a very large class of thinking operations. Lakoff describes base level spatial schemas as sources of logics used abstractly. Various sorts of inference or reasoning employ "neural structure that is actually part of, or makes use of, the sensorimotor system of our brains" and can therefore be called sensorimotor inference (Lakoff and Johnson 1999, 20). Phase spaces in physics, state spaces in engineering, measure spaces in mathematics, the year thought as a row of days or as a cycle, Venn diagrams in logic, sentence diagrams in linguistics, and the many kinds of graphic metaphor used in scientific visualization, are all representationally managed uses of inferential capability given with spatial intelligence. "So-called propositional inferences" also arise from the "inherent topological structure" of schemas of spatial perception and action, Lakoff suggests (1993, 229). Basic spatial schemas used to structure verbal logics could include center-periphery, link, cycle, adjacency, straight-curved, near-far, verticality, horizontality, front-back.

Looking back and forth from one thing to another is a basic form of relational sensing that often seems to be used act-metaphorically in thinking about serial relation; when we think about time, for instance, we often seem to be fixating points, scanning from one to another across intervening space. Musical time is thought in terms of horizontal position, while pitch relations are understood as vertical relation to a baseline. Greater-than/smaller-than relations can be thought either horizontally or vertically.

Relational sensing operations that could be used to think non-spatial topics also include the tracking of changes of position of an object relative to the self or other objects and the tracking of changes of one's own orientation toward stable or partly stable wholes. This more motor-involved form of i-it relation used in whole-body behaviors such as object-targeting and transport of materials can be used when we organize thinking as a trajectory in a scene: we can aim for a solution and submerge ourselves in the specifics of a question.

Task axis organization also involves foreground-background segregation and shifts, which support object perception and perception of part-whole relations both at the level of object parts and at the level of objects in environments. Basic logics available on the basis of visual object-gestalt and figure/ground abilities would include logics of spatial inclusion, exclusion, separateness, contact and overlap. Categorical relation is often understood in terms of these spatial logics.

Another scale of spatial function available for act metaphor is the scale of navigation, where place is understood as a simultaneous whole with internal relations, some perceived and some remembered. This scale might be used to think large abstract domains, theoretical wholes or systems such as 'a language' or 'a culture'. These theorized systems do not exist as physical entities and cannot even be thought simultaneously as wholes, but they are nonetheless understood as 'existing' with parts and relations of parts that may be surveyed, with change of location and perspective possible.

More recently evolved very fine-scale ability with small near objects requires detailed hand-eye coordination through the sequence of a complex task. In Chapter 9 I will show how this scale of spatial ability is part of act-metaphoric mathematical action with count objects.

The presence of spatial schemas in thinking seems often to be tacit or nonconscious. In Chapter 4 I discussed a multisensory relational or structure subsystem that seems to be centered in parietal tissue. It seems that activity in the parietal often is not included in sentience subnets. If the subnets embodying spatial schemas employed in act-metaphoric logics are largely parietal, their participation in thinking could be generally unnoticed.

How thinking is thought

...a metaphoric mapping may be apt in some respects, but not in others. The point here is that one cannot ignore conceptual metaphors. They must be studied carefully. One must learn where metaphor is useful to thought, where it is crucial to thought, and where it is misleading. Conceptual metaphor can be all three. Lakoff and Johnson 1999, 73

Our thinking about knowing, speaking, thinking, and metaphor is itself metaphorically structured: that is, our abstract talk about knowing and speaking is based on ways of talking about base level social and physical doing.

An example already described is talk about imagining in terms of activities and experiences with pictures: we speak of assembling images, storing images, examining images, manipulating images, comparing images with things. Various terms, questions and distinctions are brought into the discussion along with the picturing metaphor. We talk about qualities of the mental image in ways we talk about visual qualities of pictures: sharpness, color, likeness and unlikeness. Similar observations apply to other sorts of 'inner' or 'mental' representation talk.

Metaphor itself is often described as use of image-schemas (Lakoff and Nunez 1997), but this description is itself an effect of a schema by which schema is being thought.

Mental space theory (Fauconnier 1985) is a branch of cognitive linguistics that describes linguistic effects in terms of a spatial metaphor. It is based on the assumption that understanding a text involves the creation of domains or spaces. That is, it is based on the assumption that linguistic comprehension takes its structure from situational construals that may include orientation in a location, with self, objects, actions, and implications proper to these aspects. In my terms, a 'mental space' would be the usually simulational comprehensive aboutness achieved in the process of linguistic comprehension.

Mental space theory is used by metaphor theorists to account for cross-contextual effects, or blends, across separate domains. In mental space accounts, metaphoric and other sorts of blends are said to result when structure in one space or domain maps onto structure in another. Fauconnier says, for instance, that he thinks of mental spaces as structured, incrementable sets; the relations of mental spaces then are thought of the way the relations of sets are thought, as projections from the elements of one set to those of the other.

When we talk about 'projecting' or 'mapping' schemas onto domains, or one domain onto another, it is as if we imagine lines drawn between two subspaces, the way we draw diagrams of set relations with lines joining counterpart elements. Thinking of mental spaces this way, we set up an unthought assumption that the wide net organizing one simulational space cannot share substructure with the net organizing another: "I assume for the moment that ... they have no elements in common", Fauconnier says (1985, 16). But shared substructure may be just the explanation we need for cross-contextual effects like metaphor. If the very same structure can participate in sensing or imagining two different things, nothing need be transferred or projected. Even conceptual or structural metaphor may be understood as a dynamic effect of shared substructure. Sejnowski's (1989) discussion of skeleton filters as internal organizers of wide net activity is a helpful corrective.

 

 


Chapter 8. Representing and the IPL