5
CREATIVE BLENDS

… nor did Alice think it so very much out of the way to hear the Rabbit say to itself, “Oh dear! Oh dear! I shall be too late!” (when she thought it over afterwards, it occurred to her that she ought to have wondered at this, but at the time it all seemed quite natural).

Lewis Carroll, Alices Adventures in Wonderland

WE TYPICALLY CONCEIVE of concepts as packets of meaning. We give them labels: marriage, birth, death, force, electricity, time, tomorrow. Meanings seem localized and stable.

But parable gives us a different view of meaning as arising from connections across more than one mental space. Meaning is not a deposit in a concept-container. It is alive and active, dynamic and distributed, constructed for local purposes of knowing and acting. Meanings are not mental objects bounded in conceptual places but rather complex operations of projection, binding, linking, blending, and integration over multiple spaces. Meaning is parabolic and literary.

We have seen that parable carries narrative meaning across at least two mental spaces. In fact, other spaces are involved, and their involvement is not a modest addition to parable, but instead its most important aspect. We can detect a hint of this new aspect of parable by looking at a curious event in the tale of the ox and the donkey: The ox and the donkey, like Alice’s hurried rabbit, talk.

Talking animals are a conceptual blend. The talking ox and the talking donkey do not reside in the space that treats everyday farm labor and mute beasts of burden, nor do they reside in the hypothetical space that treats the vizier’s apprehension of Shahrazad’s future disaster. Where, conceptually, do they reside? It may seem perverse to ask this question. Talking animals are as natural as the nursery. They are obvious in every national literature. But in the study of the mind, whatever looks natural is most suspect. Talking animals, seemingly so trivial, are created through a general and central parabolic activity of the everyday mind—blending. Blending has been studied in detail by Gilles Fauconnier and me jointly and separately, and by Seana Coulson, Nili Mandelblit, Todd Oakley, and Douglas Sun.

The blending involved in the tale of the ox and the donkey is extensive. Let us approach it by looking at the central inference of the tale. The central inference is that the donkey has outsmarted himself; he should have foreseen that, on a farm, the inevitable background of ploughing and milling requires work animals, so that if one work animal is excused, another becomes liable for the job. The donkey’s failure to foresee the obvious likelihood that he will suffer justifies our judgment that he is too smart for his own good. His conceit makes him blind to the obvious. We blame him for being blind.

Can this central inference be constructed in the target space independently of projection? Can it be constructed in the source space and projected from source to target? The answers to these questions are no and no. This central inference does not arise in the target space independently of projection from somewhere, which is why the vizier exerts himself so ingeniously to project it. This central inference requires an agent who outsmarts himself by failing to take account of inevitable and unavoidable background. This inevitable and unavoidable background does exist in the source space: Farm labor is performed by work animals. But it does not exist in the target space. There, the counterpart of labor by work animals is suffering by wives, but Shahrazad explicitly disagrees that the suffering by the wives is inevitable in the way ploughing is inevitable on a farm. In this target space, she views King Shahriyar’s practice as contrary to courtly decorum, to his own disposition, and to traditions of order. So does he. His subjects believe that only exceptional and remarkable events have led Shahriyar to this bizarre practice. They believe they are right to rebel against it. Secretly, he may think so, too. The enormity of his exceptional practice is unquestioned, which accounts in part for the plausibility of his discontinuing it under appropriately contrived conditions.

Shahrazad’s plan to change Shahriyar’s mind depends on bringing him to invoke traditional deeply held normative expectations—his own—against his recent abnormal behavior. In the source space, labor by work animals is inevitable background: standard, expected, and unavoidable. But in the target space, suffering by wives has the opposite status: abnormal, surprising, and possibly avoidable. According to the logic of this target space, therefore, Shahrazad cannot be judged as blind for overlooking the inevitability of the background practice, because the inevitability does not exist; it is explicitly denied. Her denial is reinforced by the narrator of The Thousand and One Nights, who, before the vizier tells the tale of the ox and the donkey, classifies Shahriyar’s behavior as quirky and unstable.

Because the inevitable background necessary for the central inference does not exist in the target, the central inference cannot arise in the target. We cannot infer in the target that the inevitable background requires someone to suffer; that if Shahrazad manages to get the virgins off the hook, she must suffer in their place; or that Shahrazad is foolish for working against inevitable background. The central inferences of the tale of the ox and the donkey cannot arise in the target independently of projection. The vizier must construct them according to the logic of a different frame—farm labor—and then project the inference to the target without projecting the details of the frame that made it possible for him to establish it.

There are additional reasons that this central inference cannot arise in the target. The agent who outsmarts himself does so because he is blinded by pride. But this condition does not apply to Shahrazad in the target. Quite unlike her counterpart, the donkey, she does indeed see the risk and explicitly insists upon taking it. The donkey is foolish for blinding himself to the risk but Shahrazad looks at it without blinking. Again, the central inference of the tale of the ox and the donkey is simply unavailable exclusively from the target.

But the central inference is not available from the source, either. In the source space of farm animals, it is predictable that if one beast of burden is excused from ploughing and milling, another will be used. A human being on the farm who did not see this likelihood would be thought to have failed in not seeing it. But a donkey cannot see it. Much less can a donkey scheme or foresee. In the source, the inferences that a donkey is responsible for the ox’s reprieve, that he should have foreseen how the ox’s reprieve would result in his own grievous employment, and that he is to blame for having outwitted himself are simply unavailable. Farm animals do not have these capacities.

Where are the central inferences constructed? They are constructed in the blended space of animals with human characteristics. The blend includes abstract information that is taken as applying to both source and target, such as schematic event shape and force-dynamic structure. Additionally, specific information from both source and target is projected into the blended space. The scenario of ploughing and milling as inevitable background, the relation of work animals to this background, and the classification of the ox and the donkey as work animals come from the source space. From this information, we can deduce that when the ox is excused, the donkey is the likely replacement. We can deduce this only because the donkey is a donkey and a donkey can pull a plough, which is information from the source space. But to obtain the central inferences also requires that the donkey be intelligent, cunning, and able to articulate his complicated plan to the ox so that it can be acted upon. This information comes from the target space. The blended space incorporates intentionality, scheming, talking, foreseeing, and surprise from the target space. The result is an impossible blend with animals that are simultaneously beasts of burden and intentional agents with sophisticated mental capacities.

The central structure of this blended space does not come from the source space alone. The blended space has, for example, causal structure that cannot come from the source of the farm: The donkey’s thought leads to a plan, which leads to the execution of the plan, which leads to the reprieve of the ox, which leads to the donkey’s suffering; but this causal structure cannot come from the space that contains farm animals who do not plan. Only the causal relation between the ox’s languor and the donkey’s use as a plough animal can come from a space that contains regular farm animals. The blended space also has “modal” structure (of possibility, enablement, and so on) that cannot come from the source space of the farm: The donkey’s cunning enables him to devise a plan, and his talking enables him to communicate; but farm animals are not cunning in this way and do not communicate in this way.

The vizier intends to project the causal and modal structure developed in the blend to the hypothetical story of Shahrazad and Shahriyar. He also wants to project the blend’s framing of agents because it yields the judgment he prefers. The central inference in the blend is that the schemer blindly causes his own suffering through his pride and his schemes. It is this inference that the vizier wishes to see projected to the target, in the hope of dissuading his daughter from her plan. It cannot be projected directly from the source to the target because it does not exist in the source. It is instead constructed in a blended space, and then projected to the target.

In the previous four chapters, we used a model of projection from one space to another where the projection was direct, one-way, and positive. This model needs refinement. The refinement is blended spaces.

A blended space has input spaces. There is partial projection from the input spaces to the blend. In the tale of the ox and the donkey, the input spaces include the story of real farm animals and the story of Shahrazad. Sometimes, those input spaces will be related as source and target, in just the way we have seen so often. Crucially, blended spaces can develop emergent structure of their own and can project structure back to their input spaces. Input spaces can be not only providers of projections to the blend, but also receivers of projections back from the developed blend.

BLENDED SPACES

In the tale of the ox and the donkey, the specific agent in the blended space who develops the cunning plan has four legs and enormous ears, eats barley, sleeps on well-sifted straw, and talks. We are not to project the specific attributes four legs, enormous ears, eats well-winnowed barley, and sleeps on sifted straw onto Shahrazad in the target space. Talks is projected onto Shahrazad in the target space, or rather, talks is returned to the target whence it came.

One of the great cognitive advantages of a blended space is its freedom to deal in all the vivid specifics—ploughing, straw, barns, planning, talking, deceiving—of both its input spaces. Although the blended space will conform to its own logic, it is free of various constraints of possibility that restrict the input spaces. By means of these specifics from both input spaces, the blended space can powerfully activate both spaces and keep them easily active while we do cognitive work over them to construct meaning. Upon that circus of lively information, the mind can dwell and work to develop a projection.

Let us consider a literary example, Dante’s celebrated portrayal of Bertran de Born in the Inferno. While living, Bertran had instigated strife between the king of England and the king’s son and heir, tearing father and son apart. When seen in hell, Bertran consists, spectacularly, of two parts: a headless body and its separate head. The body carries the head in its hand, lifting the head manually to talk to Dante on his journey through hell. Bertran cites his punishment as the appropriate analogue of his sin:

Because I parted people so joined,
   I carry my brain, alas, separated
   from its root, which is in this trunk
Thus is to be seen in me the retribution.

Perch’io parti’ così giunte persone,
   partito porto il meo cerebro, lasso!
   dal suo principio ch’è in questo toncone.
Cosi s’osserva in me lo contrapasso.

This is an impossible blending, in which a talking human being has an unnaturally divided body. The blend has many parts. First, there is a conventional metaphoric understanding: Dividing people socially is understood metaphorically as dividing a joined physical object. This metaphoric projection is not at all novel. We can say conventionally that a home wrecker has “come between” a married couple by creating “distance” between them. “Till death do us part” is not a vow to hold hands; “what God has joined together, let no man put asunder” does not mean that husband and wife are surgically sutured. We can speak of breaking a bond of business, a bond of belief, a bond of loyalty, a bond of trust. None of this inherently involves the specific information of dividing a head from a body.

In this conventional metaphor, spatial proximity, junction, and separation are projected to create an abstract generic space that applies to many different targets, including stories of social and psychological actions. We have already seen such a generic projection in the interpretation of “Look before you leap” as it appears on a slip in a fortune cookie: It suggests an abstract story that applies to many target stories—a business deal, a romantic involvement, even standing up to leave the restaurant. In general, we understand proverbs out of context by projecting a generic interpretation. These generic spaces are a new kind of mental space in addition to source and target input spaces. Impossible blending does not occur in them. The information they contain applies to both source and target input spaces. In the case of “separating” father and son, the generic space contains only a unit that is separated and an agent who causes the separation.

In Dante’s portrayal of Bertran de Born, this generic space provides the beginnings for a much fuller space, a “blended” space. The blended space contains the abstract information of the generic space; it additionally contains specific information projected from its input spaces. Dante’s blended space takes, from the target, the specific sin and sinner, and, from the source, the source counterpart of the sin: the separation of a joined physical object. In the blended space, the source counterpart of the sin is visited upon the target sinner as punishment. We can see the justice of this punishment: The sinner has his own sin visited upon him not literally but figurally; the projection to the sin is traced backward to its source, and this source analogue of the sin is visited upon the sinner. The specific information from the source input space—physical separation of a joined physical object—is applied impossibly to the target human being in a blended space. The blended space contains something impossible for both source and target: a talking and reasoning human being who carries his detached but articulate head in his hand like a lantern.

In the case of the portrayal of Bertran de Born, just as in the tale of the ox and the donkey, the power and even the existence of central inferences of the projection come not from the source input space and not from the target input space but only from the blended space. The portrayal of Bertran de Born is often quoted out of context as an example of the kind of horrible punishment found in the Inferno. It is not merely bad, but bad in a special way: unnatural, ghastly, violent, destructive of integrity. The bodily division is taken as a sign of this special kind of badness. To be sure, a sophisticated reader of this passage in its context may have already concluded that Bertran has sinned, given that he is in hell, and that his sin is of a particular sort, given his location in hell. But even such a reader may derive the central inferences from the portrayal itself rather than from an abstract definition of the sin.

Where does the inference arise that this scene signals this kind of badness? Let us consider the background metaphoric projection. In the source space, there may be nothing at all wrong with separating a joined physical object, like shelling a pistachio nut. In the target space, there may be nothing wrong with setting two people against each other, or, more specifically, in setting son against father (perhaps the father is an evil infidel warrior, for example). The background metaphoric projection does not necessarily carry the implication that division is wrong—social “parting” can be good. Many readers, informed of the relevant history, would not even agree that Bertran de Born’s actions were sinful, much less treacherous. But we all know there is something ghastly and horribly wrong about a decapitated human body that operates as if it were alive. We see the amazing spectacle of Bertran carrying his detached head and interpret this division as symbolizing something unnatural, ghasdy, violent, and inappropriately destructive. The inference is established in the blended space before Bertran de Born begins to tell his story to Dante in hell—which is to say, before we are told the history of the target space.

As we will see often, blending is not restricted to combining counterparts. In the set of metaphoric correspondences, the divided object in the source is the counterpart of the “divided” father and son in the target, not of Bertran de Born in the target. But in the blend, the divided object and Bertran de Born are combined.

Blends arise often in Dante. “Mahomet,” regarded as the great schismatic who “divided” Christianity, has split himself in hell. The adulterers Paolo and Francesca, who yielded to the “forces” of passion, are blown hither and yon in hell by a forceful wind over which they have no control. The “uncommitted,” whose sin was that they never took a “stand” or a “position,” must scurry ceaselessly over a place that is—in Dante’s scheme of cosmic geography—nowhere, lacking all status.

In all these cases, a source story (division, being swept away by wind, moving without stopping) has been projected onto a target story of a particular sin as a way of conceiving that sin. The punishment of that sin resides in a blended space that is fed by source and target input spaces. The punishment comes from the specifics of the source. The damned human being comes from the target. The specific information from the source is applied to the human being from the target in a blended space of retribution.

Dante’s Inferno is an encyclopedic display of local blended spaces, but additionally, at a higher level, it is a single monumental synoptic blended space. Its target is the story of Dante’s instruction in theology and philosophy. Its source is the story of a journey. The blended space combines all the aspects of the story of the journey with all the aspects of the story of instruction in theology and philosophy.

Dante’s blended spaces are explicitly marked as exotic and literary. Blends of this sort are—like talking animals—vivid. Paradoxically, because they present striking spectacles, they may mislead us into thinking that blends are incongruous visual cartoons, unimportant to everyday thought. To be sure, cartoons that use blends are ubiquitous and familiar. Consider the cartoon of the angry character: His skin grows red from his toes to his head in the manner of a boiled thermometer; he flips his lid while steam shoots from his ears. This is certainly a blend, based on the conventional metaphoric projection that connects heated objects to angry human beings. But it is only one kind of blend, and not the most interesting kind. Most blending is covert and undetectable except on analysis. Most blends are unrelated to visualization, to exotic incompatibilities, to structural clash, or to emotion. The essential cognitive work done by blended spaces is often invisible.

Let us consider a slighdy less exotic example of a blended space, from Shakespeare’s King John. It occurs in a scene that involves no cartoonish impossibilities and no visualization of clashes. On the contrary, a member of the audience, watching the scene, would see only a messenger, looking fearful, who comes before the king. King John, recognizing the disturbance in the messenger’s face, says:

So foul a sky clears not without a storm.
Pour down thy weather.

We could read this passage simply as asking us to project directly from source to target: The appearance of the sky projects to the appearance of the face; the bad weather projects to the bad message; the event of precipitation projects to the act of delivering the message; and so on. This projection is familiar to us as an instance of the conduit metaphor: a message is an object in a container; communication is transferring an object spatially from the speaker to the hearer.

So far, so good. But there is a more sophisticated customary reading. King John speaks these lines at a moment in the play when everything is falling apart, when there is a confrontation of the first magnitude over who or what is in command, and it is a question whether King John can survive the forces gathering against him. He has apparendy just succeeded in having the rightful heir to the throne, Arthur, killed, which causes powerful nobles to defect, no doubt taking with them a worrisome following of lesser nobles. In certain subtle ways, King John is becoming something other than king, increasingly represented as having never been king. He is a king who is not a king. The activities and economies King John normally commands are rapidly escaping him. He tries to command what he no longer commands. He may appear to be in command, events may happen that seem to conform to his command, but his command is losing status. He is not naive—he realizes that his command is a conundrum. This is a subtle reading. To arrive at this reading requires a blended space, in which the messenger, the prime example of something absolutely under the king’s command, is also nature, the prime example of something that is absolutely above the king’s command. This is a combination of contraries impossible outside the blend. King John is commanding what he can command, but what he can command turns out to be simultaneously what he cannot command. The tension and instability in King John’s command is presented symbolically. It is a powerful and significant paradox.

The blend necessary for this reading combines, from the source, the inevitable release of weather that is subject to no one’s control, and, from the target, King John’s intentional control over the intentional messenger. The manifest tension between the lack of control and the exercise of control provides the central inference of the subtle reading. But the tension is not in the source: in the source space that includes nature, human command plays no role at all, and King John is not mad like Lear and would not dream of trying to command the weather. The tension is not present in the target, either: King John can indeed command messengers absolutely to deliver their bad news. The tension cannot be imported to the target directly from the source because the target would defeat it: the messenger is absolutely and rigidly under the command of King John. Indeed, the messenger’s complete subordination to the king in the target is reinforced by this very scene in which John commands and the messenger performs accordingly. We do not construe this passage as saying of the target that just as King John cannot command nature, so he cannot command the messenger. That is clearly false: He can and does command the messenger.

It is only in the blended space that King John is revealed in a situation of conflict: He both commands and lacks command at the same moment and in the identical respect. In the blended space, he is giving commands that are simultaneously appropriate and inappropriate, simultaneously routine and absurd. This is a fundamentally unstable position. It is the basis of the sophisticated reading that King John’s command is profoundly troubled and conflicted.

The tension of the blended space is reinforced by a corollary blending of impossibilities. King John is above the messenger metaphorically, in the sense of having power over him. He is probably above him spatially: The messenger may be kneeling. By contrast, any human being on earth is below a raining sky both metaphorically and spatially: The human being is subject to the power of the sky and is also literally below it. In the blended space, King John is literally and figuratively above the messenger but literally and figuratively below the sky. He is simultaneously above and below the messenger-sky. This paradox is compacted into a single expression: “Pour down!” This would be a very odd thing to say to a messenger who is kneeling. In the source space of people and skies, John is simply below, in the target space of kings and their dominions, he is simply above. Only in the blended space is he both, a clash of the most significant violence.

This presentation of John as inhabiting an impossibly unstable position has been read as constituting the power and memorability of this passage, and as distinguishing it from insipid expressions such as “he has a stormy countenance.” But the central inferences of this subtle reading are unavailable except from the blended space. The reading of these two lines as conveying on their own that something is seriously wrong in the structure of John’s kingship comes neither from the source of weather and people nor the target of kings and messengers. It comes only from the blended space. The projection is not direct from source to target.

The work of this blended space involves disanalogy between the source and the target. We do not simply suppress lack of correspondence between source and target. On the contrary, information from the source that does not correspond to the target and cannot be projected onto the target is brought into the blended space exactly so we can understand the difference between the source and the target, and thereby recognize the clash. The projection is therefore significandy negative. We have seen that the projection is not direct; now we see that it is not essentially positive. The model of projection as direct and positive from source to target will not capture the central inference of these two lines.

The scene is profoundly ironic, exactly because in the blended space there is an ironic tension between the image schemas. Without the blend, there would be no tension, and without the tension, there would be no irony. To explain the cognitive result of irony in this case requires a model of conceptual projection that acknowledges the role of blended spaces.

We often encounter linguistic marks of blended spaces. The passage from King John contains a lovely example: “thy weather.” Possessive pronouns like “thy” or “your” can be used with a noun to indicate possession and control, as in “thy riches” or “your coat,” since riches and coats can be under human control. They can be used with a noun like “weather” to indicate association, as in “Your weather is better than mine.” But in the blended space, they can exceptionally be used to indicate possession and control of “weather”: The messenger-nature possesses and controls the message-weather. “Thy” with this meaning of control comes from the target; “weather” with its meaning comes from the source; “thy weather” combines vocabulary from the source and the target. But “Pour down thy weather” combines more than vocabulary: Semantically, the verb “pour” is connected to the weather of the source space, but its use in the imperative mode—to indicate command—comes from the target space, where one intentional agent can command a second agent to perform the action encoded in the verb. The single word “pour” in this sentence blends linguistic structure associated with both the source and the target.

Because I have introduced blended spaces with examples from the high literary canon—The Thousand and One Nights, Dante, and Shakespeare—it would be easy to make the mistake of thinking that blending is a special device of literature. The everyday mind is essentially literary. Everyday logic and language depend upon blending. Fauconnier and I originally demonstrated this point with the following example, taken from the world’s premier sailing rag, Latitude 38:

As we went to press, Rich Wilson and Bill Biewenga were barely maintaining a 4.5 day lead over the ghost of the clipper Northern Light, whose record run from San Francisco to Boston they’re trying to beat. In 1853, the clipper made the passage in 76 days, 8 hours.

In the space of 1853, Northern Light makes a passage. In the space of 1993, Great America II makes the same passage, although, given the inevitable differences in weather and sea conditions in the two spaces, and the considerable differences in performance between a high-tech racing catamaran and a huge wooden clipper ship laden with cargo, they surely did not follow the identical course in detail. In neither space is there a race. But in the blended space, there is a race between Great America II and the ghost of Northern Light. Readers easily distinguish and manipulate these three spaces, and they know what each space is good for. No one imagines that the writer believes in ghost ships. No one imagines that, should Great America II capsize (as did its predecessor, Great America), Northern Light will come along from behind to rescue the rival crew.

Here we come to a fundamental widening of our investigation of parable. So far in this book, for the sake of pedagogical sequence, we have considered only those cases of parable where one of the stories is a source and the other is its target—the target is conceived, at least in part, by projection from the source. After looking at the principles of source-target parables, we detected the role of blended spaces: We saw that the source and the target are inputs to the blend and that the blend can project back to the source and target input spaces. Often, the essential projections go from the blended space to the target input space. This is the case for the tale of the ox and the donkey, for Dante, and for King John.

But now we see something quite new. Input spaces to the blend do not have to be related as source and target. Consider the boat race. Clearly this is conceptual projection from input stories, which I will call 1853 and 1993. In each story, there is only one boat, no race, and no chance of a “lead.” Projecting structure from 1853 and 1993 to a blend helps us create a blended story of a boat race, where there are two boats, one with a “lead” over the other. That blended story helps us understand the relationship between the inputs. It gives us a way to integrate the entire situation into one blended story, without erasing what we know of its independent inputs. But 1853 is not a source for 1993; 1993 is not a source for 1853. Neither is a target for the other; neither is conceived by projection from the other. For example, we do not understand the catamaran by projection from the clipper ship the way we understand Death-in-general by projection from an agent who enforces departure or the way we understand a cause by metaphoric projection from a mother. Hereafter, I will speak of input spaces and blended spaces. When the input spaces are related as source and target, I will call them the source input space and the target input space, or, where there is no possibility of confusion, just source and target.

In parabolic blending, the input spaces are often rhetorically unequal. For example, in the boat race, it is 1993 that the reporter cares about and talks about. It is 1993 that he wants to understand and report fully. I will say that 1993 is the topic space of the parabolic projection. A topic space is not necessarily a target space. Ninteen ninety-three is a topic space but not a target space with 1853 as its source. It is possible, as we will see later, for there to be more than one topic space. It is also possible for the topic space to shift: If we are descendants of the captain of Northern Light, it maybe 1853 that we care about understanding. It is even possible, as Seana Coulson has shown, for the source input space to be the topic space.

The blended space includes an imaginary race; inferences can be made in that blended space and projected to the topic space of 1993: Great America II doing well, is moving fast enough, is accomplishing its goal. The blended space does not merely compare positions of two ships; it calls up the conceptual frame of a race. The blend therefore contains structure not contained in either of the two input spaces. This frame provides emotions and intentions of the crew and of fans, which can be transferred to the space of 1993. A victory party, with all the standard rituals and conventional photographs, was held for the crew of Great America II and reported in the customary fashion in Latitude 38. The blended space made a profound mark on the feelings and actions of everyone involved. The blended space left its trace on reality.

In the cases of Bertran de Born and the messenger in King John, impossible conjunctions were exploited inferentially—for example, we were able to draw an inference of irony from the discordant image schemas in the King John blend. But in the sailboat race, the impossibility of the two ships’ racing is merely pragmatic—there is nothing in the structure of the race that is impossible, merely the extraneous difference in the years during which the two boats existed—and this merely pragmatic impossibility is irrelevant to the central inferences. Blends are in general not constructed merely to present some spectacular clash or exotic impossibility. Suppose someone says, “Four and a half days may seem like an insurmountable lead, but maybe it isn’t: Great America II looks as if she will be becalmed when she enters the Caribbean, but Northern Light will have a gale behind her and could pick up enough time to sail right through Great America II.” In this case, the blend incorporates fabulous elements impossible outside the blend, such as one ship’s sailing through another unharmed, but that impossibility is merely pragmatic and we draw no inferences of irony from that impossibility.

The blended space of the boat race is not constructed exclusively by fusing counterparts from the input spaces. There are many counterpart connections between the space of 1853 and 1993: The catamaran and the clipper are counterparts, as are their courses, their starting points, and their destinations. Many of these counterparts are fused in the blend: The two courses are fused into one course, for example, the way the messenger in King John is fused with nature. But the catamaran and the clipper are not fused into each other in the blend. The catamaran and the clipper are distinct in the blend.

The passage in Latitude 38 sets up an independent conceptual domain—of ghost ships and imaginary races. It is specific to the blend; it does not belong to either input space. This extra and fantastic conceptual domain helps us to notice the existence of the blend. But blended spaces do not have to set up their own independent conceptual domains and do not have to make us realize consciously that we are doing any blending. For example, the passage in Latitude 38 could easily have read, “At last report, Great America II was 4.5 days ahead of Northern Light.” Here, with the same blend but without the mention of phantom ships, we might not realize consciously that we have constructed the blend. Another way to set up the “impossible” blended space so as to provide the right inferences would be to exploit the standard counterfactual construction: “If the two ships were racing, Great America II would be 4.5 days ahead of Northern Light at this point.”

Projecting inferences from the blend to an input space is not a simple matter of copying all of the inferences in the blend to the input space. We know how the blend connects to its input spaces, and we know how inferences in the blend correspond to inferences possible in the input spaces. Projecting inferences from the blend to an input space often involves selecting or translating them to fit the input space. For example, the inference in the blend that Great America II is a long way ahead of Northern Light is not appropriate for the space of 1993; but the corollary inference that Great America His doing well is appropriate for the space of 1993, and can be projected from the blend to the space of 1993.

There are several different ways to make the appropriate inferences in the blend and project them appropriately to the space of 1993. For example, how do we construct a scenario in the blend in which Great America His “4.5 days ahead” of Northern Light? There are several possibilities. In one reading, Great America II reached a particular geographical position after sailing a certain length of time, and Northern Light did not reach that position until it had sailed a length of time longer by 4.5 days. In another reading, Northern Light and Great America II sailed the same length of time, at the end of which Northern Light had reached a geographical position that Great America II had reached after sailing a length of time shorter by 4.5 days. It is tempting to imagine that these are equivalent statements of the same situation, but they are in general not, and, given the realities of sailing, they are in practice almost certainly not. In the first reading, the difference of 4.5 days comes from the space of 1853, and it is Northern Light that does the 4.5 days of sailing. In the second reading, the difference of 4.5 days comes from the space of 1993, and it is Great America II that does the 4.5 days of sailing.

In yet another reading, experts predict that Great America II, given its performance so, far, should complete its voyage to Boston in a time that is 4.5 days less than the time of the voyage for Northern Light. In yet another reading, the distance between the positions of Great America II and Northern Light after each has sailed a certain length of time is what would be covered in 4.5 days at Northern Light’s average speed made good—its overall speed actually accomplished relative to the ground along the course line. In yet another, experts compute that Northern Light, under the weather and sea conditions in the relevant area faced by Great America II during the past several days, would have covered the relevant distance in 4.5 days. In another, experts compute that Northern Light, under the weather and sea conditions coming up for the next several days, would cover the relevant distance in 4.5 days. And so on. Any of these constructions can be established in the blended space, all of them leading to the inference, projectable to the space of 1993, that Great America II succeeding in its ambition.

There are evident linguistic traces of this conceptual blending. In the blended space of the boat race, there are two boats simultaneously racing, and we can refer to them accordingly, but there are vestiges of Northern Light’s having come from a space that provides the landmark frame, so that in the blend, Northern Light provides the landmark and Great America II is conceived as a trajector relative to that landmark: We prefer to say,” Great America II is ahead of Northern Light,” rather than “Northern Light is behind Great America II.” For a real race, we might say either because in general either boat can be the landmark The phrase “barely maintaining” also shows a trace of blending, since it presupposes intentionality: The crew of Great America II is trying to maintain a lead. By contrast, if Northern Light were ahead by a day in the blend, it would be strange to say, “The captain of Northern Light is barely maintaining his one-day lead,” since the presupposed intentionality could not be projected back to the space of 1853 without some work to make it possible there, like the following: “Historically, the captain of Northern Light took the view that he had the fastest ship imaginable and that his record would stand forever. But now, he is barely maintaining a tiny lead over Great America 7/and maybe proved wrong.” It is important to see that the construction of the blend is not meant to erase or dispense with the input spaces. Blending provides a way to integrate efficiently and effectively over many spaces while maintaining the network of connections across all those spaces.

The same constraints of landmark and intentionality can be shown to operate in a case that involves metaphor. To understand, “President Franklin Delano Roosevelt accomplished a great deal in his first one hundred days, but President Clinton has accomplished by comparison little,” we must build two mental spaces and an intricate comparison between them. Both of these mental spaces can themselves be understood through a conceptual metaphor according to which accomplishment is travel along a path. We might then say, “FDR covered a lot of ground during his first one hundred days. President Clinton by comparison has only just started to move.” In one blended space, FDR is moving along a path whose locations are goals; reaching the location is accomplishing the goal. In the other blended space, President Clinton is beginning to move along a similar path. These two blends are conventional and share the identical generic space. These two blends can themselves be the input spaces to a new blended space, as when we say, two months after President Clinton has taken office, “Clinton was supposed to hit the ground running. He implied that he was going to accomplish as much in his first one hundred days as FDR accomplished in his. So far, Clinton has failed completely to keep pace with FDR.”

“To keep pace with” requires the construction of a conventional blend that has both agents competing simultaneously along the same track. For most readers, this conventional blend will have been constructed and used entirely unconsciously. The construction of that conventional blend includes the connection of structures in the blend to corresponding structures in all of the spaces that led to it, so that we know the implications for Clinton in the space of 1992 of the fact that Clinton in the FDR-versus-Clinton blend does not “keep pace with” FDR in that blend.

We can force the blend into consciousness by drawing attention to it: “Clinton is in a race with the ghost of FDR”; alternatively, “At this rate, Clinton’s term will be over before he gets anywhere near the finish line.” We know implicitly that “the finish line” in the blend corresponds to FDR’S degree of accomplishment on his hundredth day in office in the relevant input space of “FDR’S first year in office” (beginning March 4, 1993). But protest will arise if we say, “So far, FDR has succeeded completely at keeping well ahead of Clinton.” We could say this of a real race, but not of this blended race, because in the blend, some aspects of intentionality from the input spaces are indispensable for the reasoning in the blend, but only certain aspects of that intentionality are allowed to enter the blend. Yet this restriction is not simple. If we knew of a passage in President Roosevelt’s diary in which he represents himself as competing with all presidents, past and future, and then claims that he has set a record for accomplishment in his first one hundred days that no other president will ever be able to equal, then we might be able to say without provoking protest, “So far, FDR has succeeded in staying well ahead of Clinton.”

Structure that is developed in a blended space can change our view of the input spaces, as in the following riddle, which Arthur Koestler attributed to psychologist Carl Dunker:

A Buddhist monk begins at dawn to walk up a mountain. He stops and starts and varies his pace as he pleases, and reaches the mountaintop at sunset. There he meditates overnight. At dawn, he begins to walk back down, again moving as he pleases. He reaches the foot of the mountain at sunset. Prove that there is a place on the path that he occupies at the same hour of the day on the two separate journeys.

The reader might pause here to try to solve the riddle before reading further. It can be solved ingeniously by imagining the Buddhist monk walking up as his double walks down on the same day. In that blended space, it is clear that there is a place on the path that the two Buddhist monks occupy at the same hour of the day: The place is where he meets himself. He must of course meet himself, since someone who traverses a path in the space of a day must meet someone who traverses the path in the opposite direction in the space of the same day, however much they both walk as they please. This inference, that there must be a place that the two travelers inhabit at the same time of day, is projected from the blend back to the input spaces to create a point of connection between the input spaces of the two journeys, although no encounter occurs in either of them. Interestingly, people who count this blend as supplying a proof often cannot supply an alternative proof that does not make use of the blend.

It is worth pausing to observe that this is another case in which the input stories to the blend are not related as source and target. Indeed, this is a projection over stories that does not seem to involve metaphor or analogy in any central respect. There is conceptual projection from the input stories to help us create quite a different blended story—one involving two monks and an encounter. This blended space helps us to integrate the entire situation without erasing the independent input stories. It helps us organize the input stories and see relationships between them. It is also worth observing that both of the input stories count equally as topic spaces of the parable: we are interested in establishing for both of them that there is a place on the path that satisfies the requirement of the riddle. Notice also that this blend does not always fuse counterparts—the two monks in the two inputs stay distinct in the blend, although the two days are fused into one.

The blends in the boat race and the riddle of the Buddhist monk do not come from literary works, and neither can be mistaken as merely spectacular, since their purpose is to perform inferential work. They both allow the reader to reason successfully to a goal. Consider a similar case of everyday problem-solving: Suppose we hand a volleyball to someone who has played tennis but not volleyball and we say, “Serve it overhand the way you would a tennis ball.” The volleyball becomes, for the moment, a tennis ball in part, but only in part. Partial blending of the volleyball space with the tennis ball space produces a suitable action.

Blending, used to reason about everyday problems, often succeeds but also often fails. If we sit down late at night to a dimmed computer screen at a desk illuminated by an architect’s lamp and attempt to brighten the screen by turning the knob on the lamp instead of on the computer, we say it is a “mistake,” but it is not an arbitrary or unmotivated mistake. We are very familiar with “absentminded” mistakes like putting on our eyeglasses to hear better or punching a button on the television remote control device to turn the air conditioning off. Strange slips of action like these reveal the covert existence of momentary cognitive blends. The connections they manifest do not come from abstract analogies—they are not the transfer of an elaborate schema from one domain to another. Seeing and hearing are not “like” each other when we put on our eyeglasses to hear better; in the unsuccessful blend, seeing and hearing are fused, and they respond to the same corrective lenses. In the same way, the volleyball and the tennis ball are not merely similar to one another; in the blend, they are fused and respond to the same manipulation. Of course, it goes without saying that we know the difference: Seeing and hearing are different in the two input spaces; the volleyball served with the hand and the tennis ball served with a racket are different in the two input spaces. We know what is appropriate to each of these spaces. We often try to solve a problem by dealing with it as if it resides in the blend, but we know where it actually resides and we know how to move between the useful blend and the real problem.

These blends are not cases of misframing or reframing. If the actor is asked immediately before performing the action whether his hand and a racket are identical, or seeing and hearing are identical, or a television and an air conditioner are identical, he would of course say that they are not. He would give the same response after performing the action. The two spaces and their frames are kept quite distinct, but a blended space has momentarily been put together for the purpose of local problem-solving. When the blend produces a successful action, as in the case of the volleyball serve, the action may be permanently projected to the target, to be developed further there.

These connections are not metaphoric, either. In the metaphoric projection of seeing onto hearing, eyeglasses would correspond to hearing aid, not identically to eyeglasses. In the blend, however, eyeglasses come from one space but apply identically to the entity in the blended space that results from the fusing of seeing and hearing. These actions are essentially nonverbal and so clearly are independent of literary “deviation,” metaphoric language, or strange idioms.

Blended spaces do cognitive work in the strongest sense. They provide inferences, emotions, and novel actions, and consequently leave their mark upon the real world. In the example of the boat race, the sailors of the catamaran were able, through the blend, to live their action as a race, and Latitude 38 was able to cover the adventure as a race. In a series of articles in successive issues, Latitude 38 reported on the preparation, the initial race, the initial failure, the rescue and salvage attempts, the second preparation, the second race, the details of the race and the victory, and, finally, the victory celebration. Readers of Latitude 38 were able to follow the “race” and to “assist,” in the French sense. Everyone involved was able to plot and compare strategies that arose from considerations of the relative positions of the two ships. They could wonder whether Great America II was winning or losing, and they could, through projection from other spaces, actually find out: the blended space of the race comes with very precise, quantifiable truth conditions. By living as a race what is not a race, everyone involved—from the designer of the boat to the photojournalists—could feel, react, and reason differently, and perhaps more efficiently, and with more pleasure.

Wayne Booth, in The Rhetoric of Fiction, gives evidence of blends that lie between two specific kinds of spaces—the space of a story narrated and the space of narration. Booth does not recognize these blends as blends because his instruments of analysis include only two spaces, not a third blended space between them. Consequendy, he describes the blends as “intrusions” by the space of narration into the space of what is narrated. But it is important to recognize that in general we keep the space of what is narrated and the space of narration separate, while blending them in a distinct blended space.

In the anonymous novel Charlotte Summers, for example, the narrator and readers are not part of the space of the story narrated; the characters of the story, their towns, and their residences are not part of the space of narration. But in the blended space, the narrator and readers can journey to the locales of the characters, to have a look at them.

Before I introduce my Readers into the Company of Miss Charlotte Summers, I must make them acquainted with some of her Friends … for which Purpose, I must beg their Company as far as Carmarthenshire, in Wales. Tho’ the Journey is pretty long, and, in the ordinary Way of travelling, may take up some Days, yet we Authors are always provided with an easy flying Carriage, which can waft our Readers in an Instant, much longer Journeys than this we are now setting out on.

In the space of the story narrated, the narrator does not exist and necessarily has no special powers there. In the space of narration, the narrator does exist. He has no special powers with respect to most of that space, but he does have special powers with respect to the story he is telling: He can shift the focus of time and place in the narration. In the blend, the narrator, the readers, and the characters can inhabit one world. In the blend, shifting spatial focus is identical to actually moving bodily from one place to another; shifting temporal focus is identical to moving bodily from one time to another. The power to shift spatial and temporal focus is identical to the power to move in superhuman ways from one place to another, or one time to another, and to move readers as well. We keep distinct the space of the story narrated, the space of narration, and their blend.

This flying “Carriage” is a specific blend derived from the general blend. Some of the features of the carriage come from the story narrated: It is a carriage like the carriages in the story narrated—it is a container, makes spatial journeys, and so on. But some of its features come from the space of narration, specifically, from the way the author’s mind operates in that space—the “event shape” of the event performed by this blended carriage and the various “modal” relationships of what it can do and how it can do it are taken from the way the author’s mind moves in the space of narration. In the blend, this carriage has features from both spaces.

Booth’s Rhetoric of Fiction can be read (between the lines) as a taxonomy of such blends. Often these blends are easily noticed. The more interesting blends, however, are covert, and they are nearly ubiquitous in literature: Whenever a narrator moves in and out of a character’s mind, shifts point of view from character to character, or provides an inside view of any sort, she is doing something impossible for the space of the story narrated, and she is doing it to a world that is not real inside the space of narration. Similarly, whenever anyone in the story, including the author, makes a statement that, given the logic of the narration, we must take to be absolutely reliable, or whenever the author confers a badge of reliability on a character, or alters “durational realism” or sequence, or tells us exactly how intensely a character experienced certain emotions, the author is thereby doing something impossible in the space of the story narrated and doing it to a world that is not real in the space of narration. The space with respect to which these actions are both possible and real is the blend.

Mental images to accompany blended spaces are easy to imagine. An advertisement for a tax consultant might talk about “tax bite” in voice-over while a visual image of an anthropomorphized Internal Revenue Service Form 1040 sinks its teeth into a terrified paycheck. Allegories and political cartoons make routine use of these impossible blendings. Gluttony is an anthropomorphized banana cream pie, Study a walking and talking dictionary.

One of the most common uses of blending is in counterfactual expressions such as “If I were you, I would have done it,” said by a man to a woman who declined earlier to become pregnant: the woman did not do it, the man cannot do it, but the blend combines the man’s judgment with the woman’s conditions, enabling the man-woman to become pregnant in the counterfactual blend.

Personification is perhaps the most thoroughly analyzed consequence of blended spaces. If we revisit the various personifications we have considered—of the wind as a torturer, of Death as Thanatos, of the rain as a violent and spiteful destroyer, of situations we are trying to master as intentional adversaries, and so on—we will detect instantly the impossible blendings of specific information from source and target.

Let us consider a case of personification that shows these complexities of parable as well as some new complexities: Death the Grim Reaper. This personification involves robust impossible blending: The biological cause of death is simultaneously an animate, intentional skeleton who walks, carries a scythe, wears a robe with a cowl, and comes with the goal of killing you in particular. We do not confuse this interesting blend with the input space of death: no one ever actually confuses the Grim Reaper with the biological event of death. But the Grim Reaper blend is usefully connected to its input spaces and helps us think about them.

What are those input spaces? Here we make the crucial observation that nothing limits blending to only two input spaces. A blend can have as many input spaces as can be mentally juggled. Blending can also be recursive, happening in steps, so that a blend can be an input space to another blend. In the case of the Grim Reaper, there are at least four input spaces. Let us begin with two of them, which have a partial source-target relationship. They are (1) a space with an individual dying human being and (2) a space with reapers in the scenario of harvest. We can recruit a conventional metaphor to connect these two spaces, thereby putting them into a source-target relationship. That conventional metaphor is PEOPLE ARE PLANTS WITH RESPECT THE LIFE CYCLE; it underlies everyday expressions like “She’s withering away,” “He’s a late bloomer,” and “He’s a young sprout.” It guides us in connecting the person in the space of person-death with the plant in the space of reaping, EVENTS ARE ACTIONS guides us in doing more: The action-story of reaping can be projected onto the event-story of death. Dying (an event) can be understood by projection from reaping (an action).

But there is an additional input space to the Grim Reaper blend that does not stand as source or target to any other input space. This space is the very abstract story of causal tautology. We say Death causes dying, Sleep causes sleeping, Hunger causes hungering, Desire causes desiring, Lust causes lusting, Starvation causes starving, Fire causes things to be on fire, and so on. In the abstract story of causal tautology, events that belong to the same category have the same abstract cause. We saw this causal tautology earlier, in our discussion of Death-in-general as an abstract cause. But we had not yet begun to discuss blending at that point, so we skipped a crucial step there. Let us do it here.

In the abstract story of causal tautology, all the events in a given category share the same abstract cause. Of course, the abstract story of causal tautology does not stand up scientifically. There are many different individual event-stories of being ill, brought about by many different specific causes. When we blend any one of those specific stories with the abstract story of causal tautology, then, in the resulting blend, the specific event of being ill acquires an abstract cause—Illness-in-general—which is shared by every other event of being ill. This is such a standard and conventional blend that it is difficult even to recognize that any blending lies behind it. But it is necessary: When we are thinking in medical terms, we can manipulate a space in which a specific illness causes specific biological realities and a specific event of death without rooting that causality in a hypothetical Illness-in-general or a hypothetical Death-in-general; to get the abstract causes, we must use the conventional blends that draw upon the abstract story of causal tautology.

There are many individual events of human death, brought about by many different causes. Any particular event-story of death may have a particular cause—bone cancer, malaria, starvation, bleeding, stroke, suffocation, injury, and so on. When we blend any one of them with the abstract story of causal tautology, then, in the resulting blend, the specific event of death acquires an abstract cause, Death, shared by every other event of death. When we blend the event-story of Alcestis’s death with the abstract story of causal tautology, then Alcestis’s individual death has an abstract cause: Death-in-general.

That blend can then be an input to a second blend in which Death-in-General becomes an actor, not just an abstract cause. We have already seen a partial analysis of how this is done: EVENTS ARE ACTIONS guides us to map an action story of departure onto the event-story of death as caused by Death-in-General, personifying Death-in-General as an agent who enforces that departure. Here we add the observation that the personification of Death as an agent who enforces the departure resides not in the departure story nor in the death story but only in the second blend. We can now see that the story of Alcestis’s death is not just a projection of an action-story of departure onto an event-story of death. It is more complicated: it is a two-step blend. The first two inputs are the story of Alcestis’s individual death and the abstract story of causal tautology. The resulting first blend is blended with the action-story of departure to produce a final blend in which Death-in-general is personified as an agent who enforces the departure. The inputs to the second blend are related as source and target.

The Grim Reaper blend also has a fourth input space, a space containing a prototypical story of a human killer killing a victim. This space stands in partial metaphoric relation to the space of the human death caused by Death-in-General. The Grim Reaper is not just a reaper and not just Death-in-general. He also has features of an intentional human killer who comes specifically to kill a specific victim.

So let us walk through the blending in the case of the Grim Reaper. Blending the individual event-story of a human being dying with the abstract story of causal tautology gives us a blended space in which the individual event of human dying is caused by Death (Death-in-general). This gives us an event-story of an individual human death with an abstract cause. We can blend that story with two different action stories, guided in each case by EVENTS ARE ACTIONS. The first of these input action stories is the prototypical story of a human killer who comes to kill a human victim. The second of these input action stories is the story of reaping grain. The resulting blend then has an entity that is simultaneously Death-in-general, a killer, and a reaper. The person who dies is simultaneously a human being dying from a specific cause (say, heart disease), a human being dying from an abstract cause (Death), the victim of a killing, and a plant that is harvested.

The Grim Reaper resides in a blend but cannot reside in any of the input spaces to that blend. The Grim Reaper is not in the input space with the individual event of human dying, since there are no plants or reapers there. Obviously, he is too specific to reside in the input space of the abstract story of causal tautology. He is also absent from the input story of prototypical killing by a killer, which does not have reapers. He is not in the input space with the story of reaping and harvest, either: the stereotype of reapers in the reaping story is incompatible with features of The Grim Reaper.

Let us consider some of these incompatibilities. Stereotypical reapers are subject to persuasion and argument. But the blending of the abstract story of causal tautology and the event-story of individual human dying gives a Death that is beyond persuasion. The Grim Reaper is inhuman.

The individual authority of any actual reaper is unknown: perhaps he takes his orders from others; perhaps he is a slave. But Death has authority that is blended with the reaper to create an absolutely transcendent authority completely incompatible with that of normal reapers.

Actual reapers are numerous and essentially interchangeable. But Death is conceived of as a single abstract cause, which is projected to the blend, making Death-the-Reaper single and definite. This explains the appropriateness of the definite article: the Grim Reaper.

Actual reapers are mortal and are replaced by other reapers. But Death is neither. Projecting these features of Death to the blend creates a Death-the Reaper who is immortal: the same Grim Reaper who cut our ancestors down will cut us down.

Stereotypical reapers are strong, productive, healthy, and attractive. But the killer is destructive, unhealthy, and works on us, so the Grim Reaper must be unattractive, or “grim.”

Stereotypical reapers perform heavy labor for long intervals and wear clothing suitable to these conditions of labor. But the killer acts once only, on the person who dies, so Death-the-Reaper can wear clothing suited to repose; and this clothing can further carry connotations of the killer as grim and isolated.

Stereotypical reapers use their scythes, but the Grim Reaper is often thought of as doing his work merely by appearing. In that case, there is yet another, minor input space to the blend, in which the herald of death brings death merely by appearing. In the blend, the Grim Reaper has aspects of the reaper and the killer, but his effective action comes from the input story of the herald rather than the input story of the reaper or the killer.

Stereotypical reapers work in daylight, reaping the entire field indiscriminately, ignorant of the individual existence of plants of wheat, and they harvest rather than kill. But Death and the killer have an entirely different set of meanings, so the Grim Reaper comes for a specific person at a specific time, and he kills. He can stalk you like a killer.

Lakoff and I originally noticed a constraint on personification: We must feel about the personification the way we feel about the event, and the appearance and character of the personification must correspond to the way we feel about the event. As long as we think grimly about the event of death and its cause, we must take a grim view of Death-the-Reaper. We can now see a reason for this constraint: We project to the blend our view of the event, including its cause. In the case of the Grim Reaper, we also project to the blend an action story of killing consistent with our feelings about the event of death. The reaper in the blend is simultaneously a cause we feel grimly about and a killer we feel grimly about. The Reaper must therefore actually have these features.

If we look at the linguistic elements involved in referring to Death-the-Grim-Reaper, we see that the terms reflect the conceptual blending. The definite article “the” comes from the causal tautology, since it picks out a single general cause. The name “Death” comes from the blending of the causal tautology with the individual event of dying. The adjective “Grim” comes from both the space with the prototypical killer and the space with the individual event of human dying. The noun “Reaper” comes from the input space of harvest.

The Grim Reaper shows us again that combining in the blend is not restricted to counterparts. Reapers and skeletons are not counterparts in the input spaces: PEOPLE ARE PLANTS does not connect them, nor does EVENTS ARE ACTIONS. But Death as a cause is metonymically associated with skeleton as an effect. In the blend, the reaper is combined with both Death and the skeleton. Similarly, priests, monks, mourners, and members of lay brotherhoods that are associated with dying, funerals, burial, and afterlife are metonymically associated with Death. They are not counterparts of Death, but in the blend, the attire we associate with them—robe and cowl—can be the attire of the Grim Reaper. The cowl, pulled over the head of the Grim Reaper, at once evokes both connotations of death and the impression of Death as mysterious, unknown, and set apart from human society. This cognitive construction of meaning is independent of historical and scientific accuracy: maybe priests, monks, or lay brethren in fact never wore robes with cowls. What matters is only that we know the conceptual association, from any source, including cartoons. Someone who knows that association can use it to make sense of the attire of the Grim Reaper.

The possibility of combining metonymic elements—like Death and a skeleton—gives blending a great power: The blend can combine elements that contribute to the desired effect even though those elements are not counterparts. The combined elements “go together” in evoking the same effect even if they do not “go together” according to the counterpart connections between the input stories. Consider for example the personification of Heroism. Blending a story of heroic behavior with the abstract story of causal tautology yields a blended story in which heroic behavior is caused by an abstract cause, Heroism-in-general. Just as we use EVENTS ARE ACTIONS to personify Death-in-general as a reaper, so we can use it to personify Heroism-in-general as a human actor who causes heroic behavior. Nothing so far requires that this human actor be a hero. Moreover, the actor who causes the heroism corresponds to Heroism-in-general, not to any particular hero or kind of hero. However, Heroism-in-general and that causal human actor are indeed both associated with actual heroes. We can therefore pick such an actual hero, or a kind of hero, and use it in the blend to give Heroism-in-general a particular human form: Heroism-in-general can be Galahad or a generic veteran knight of the Round Table or a battle-seasoned samurai or Queen of the Amazons or Ajax or any other heroic person, whether specific or generic. This possibility of combining noncounterparts allows the blend to combine elements for effect even if they are not counterparts.

The Grim Reaper also shows us that when a blend involves a conventional metaphor (like PEOPLE ARE PLANTS), the recruitment of that conventional metaphor to the blend is partial, selective, and transforming. The conventional blend of the Grim Reaper does not simply recruit the conventional metaphor in which stages in the human cycle of life correspond metaphorically to stages in the plant cycle of life. Instead, the blend takes only a part of that metaphor and alters it to suit its purposes. In the conventional metaphor, the stages of a person’s life correspond to the stages of plant life: Youth is a sprouting or burgeoning; full maturity is full flowering; old age is withering; and the person-death is the final decline and disintegration of the plant or of the relevant part of the plant, such as the flower. But the scenario of harvest interferes with that set of mappings.

The original analysis George Lakoff and I offered of the Grim Reaper included the observation that “the plants at the end of their life cycle are harvested.” This inexact characterization came from not yet having discovered blending and its partial and flexible use of conventional metaphors. Reaping occurs not at the end of the life cycle of a plant, but at the end of its cultivation cycle, which is in the middle of its life cycle. Lakoff and I had recognized this in remarking that the “Gazing Grain” in Emily Dickinson’s poem “Because I could not stop for Death” stands for maturity—it follows “Children” at school (youth) and precedes the “Setting Sun” (old age). There is a fundamental mismatch between the harvesting of the grain and the typical disintegration and death of a plant: The harvesting happens in the middle of the life cycle but the death happens at the end of the life cycle. There is also a fundamental mismatch between the harvesting of the grain and the dying of the person: Harvesting takes place at a fixed stage in the life cycle of the plant—when it has matured but has not begun to decline—yet death happens at any stage of the human life cycle.

Because of these mismatches, the conventional metaphoric connections between stages of plant life and stages of human life are ignored in the blend. In the personification of Death the Grim Reaper, it is the event of human dying we are concerned with, and so we require the Grim Reaper to be able to reap at any stage of human life. He can reap his victim in young adulthood, full flower, or old age. The conventional metaphoric connection in PEOPLE ARE PLANTS between the final stage of plant-death (in the declining plant’s life cycle) and the final stage of person-death (in the declining person’s life cycle) is kept out of the blend. Similarly, the fact that reaping takes place only at the middle stage in the plant’s life cycle is kept out of the blend. In the blend, the stages of human life and the stages of plant life are not fused. Instead in the blend, the moment of harvest is fused with the moment of death, but that moment can occur at any point in the cycle: the Grim Reaper is a bad farmer.

Interestingly, although in principle blending could make more use of the person-plant correspondence—a painting might portray the Grim Reaper as mowing down a field of people-as-wheat (there are historical examples of this sort)—in fact the modern conventional blend does not emphasize this connection. The Grim Reaper simply arrives, and his arrival causes the dying. In this way, the Grim Reaper acts not at all like a reaper, but like any herald of death. The person who dies of course does not act like a plant.

Not just any specific information from input spaces is likely to be projected to the blended space. Prototypical information that cannot be avoided unexceptionably, given the medium of representation, is always likely to appear: In a picture of the blend, the talking donkey would probably have four legs and donkey ears, regardless of the conceptual use of this information in the parable. Much more important, however, the blended space typically includes specific details that serve as cues for projecting to the topic input space. The plough and the millstone in the tale of the ox and the donkey are specific details that do not project as specific details onto the story of Shahrazad. But they are metonymic for generic conditions of servitude and suffering, which do appear in the blend and do ultimately project onto the virgins who suffer in the story of Shahrazad. The sifted straw and the well-winnowed barley do not project as specific details onto the story of Shahrazad. But they are metonymic for generic conditions of luxury and comfort, which do appear in the blend and do ultimately project onto Shahrazad’s current situation.

The varied examples we have considered make it clear that blended spaces are useful for inferential work. When we widen our scope yet further, we find blended spaces not only in literature and everyday linguistic utterances, but also in dreams and in our attempts to make sense of anything new by blending its specific details with structure from something we already know, so as to categorize it provisionally and act accordingly. The various figural projections we encountered in earlier chapters all reveal blended spaces, once we look for them. Proust’s ouverture to À la recherche du temps perdu, for example, which is a projection of the spatial onto the mental, works through a fantastic blended space. In that blended space, spatial events of change, including body action, are blended with mental events of memory in strikingly impossible ways. As memories come and go, or shift into one another, the actual room that contains the sleeping narrator changes to match the memory. As a memory of one room becomes the memory of another, the wall before his eyes shifts direction and grows longer. The physical space around him takes on the aspectual nature and powers of his memory. This is an impossible blending, of the kind commonly recognized in dreams, memory, cartoons, and literature.

Blending is a dynamic activity. It connects input spaces; it projects partial structure from input spaces to the blend, creating an imaginative blended space that, however odd or even impossible, is nonetheless connected to its inputs and can illuminate those inputs. A blend can produce knowledge. It is not constructed by union or intersection of the inputs. It is not a skeletal or static mock-up of a few elements from the inputs but has a life of its own, in the sense that it contains structure that is not calculable from the inputs and that can be developed, once constructed, on its own. The blend counts as a unit that can be manipulated efficiendy as a unit, providing full access to the input structures without requiring continual recourse to them.

Blending in parable has the following general principles:

• The blend exploits and develops counterpart connections between input spaces.

• Counterparts may or may not both be brought into the blend, and may or may not be fused in the blend.

• The projection from the input spaces is selective.

• Blends recruit a great range of conceptual structure and knowledge without our recognizing it.

• What has been recruited to the blend can be difficult to discover.

• A blend may have many input spaces.

• Blending is a process that can be applied repeatedly, and blends themselves can be inputs to other blends.

• Blends develop structure not provided by the inputs.

• Blends can combine elements on the basis of metonymic relation.

• The recruitment of a conventional metaphor to the blend is in general partial, selective, and transforming.

• Inferences, arguments, ideas, and emotions developed in the blend can lead us to modify the initial input spaces and change our views of the knowledge used to build those input spaces.

Now let us consider two final related questions:

• How does structure develop in the blend?

• How does structure in the blend lead us to reconsider input spaces?

Blends develop by three mechanisms: composition, completion, and elaboration.

In blending, we project partial structure from input stories and compose structure in a blended story. We are guided in doing so by counterpart connections between the input spaces. For example, the riddle of the mountain-climbing Buddhist monk has elaborate counterpart connections between the two input spaces. Some of those counterparts, like the paths and the dates, are brought into the blend and fused. Others, like the two monks, are brought into the blend as separate entities.

In other blends, only one counterpart is brought into the blend, as in the blend for “If I were you, I would have done it,” which brings into the blend the judgment of the man but not the counterpart judgment of the woman. Partial composition provides a working space for further composition.

Completion provides additional structure not provided by composition. Given a minimal composition of two boats on a course in the example from Latitude 38, we can complete the blend with a large amount of structure from our conceptual frame of a race. Given a minimal composition of two monks traversing a path during the same day starting from opposite ends, we can complete that structure by recognizing it as an instance of a familiar frame that contains an encounter. Completion provides the structure in the blend that gives us the solution to the riddle.

Elaboration develops the blend through imaginative mental simulation according to the principles and internal logic of the blend. Some of these principles will have been brought to the blend by completion. Continued dynamic completion can recruit new principles and logic during elaboration. But new principles and logic may also arise through elaboration itself. Blended spaces can become extremely elaborated, as in literary fantasies.

Composition and completion often draw together conceptual structures previously kept apart. As a consequence, the blend can reveal latent contradictions and coherences between previously separated elements. It can show us problems and lacunae in what we had previously taken for granted. It can equally show us unrecognized strengths and complementarity. In this way, blends yield insight into the conceptual structures from which they arise.