1 Drawn Into the Future or Driven by the Past Martin Seligman University of Pennsylvania Peter Railton University of Michigan Roy Baumeister Florida State University and Chandra Sripada University of Michigan 2 Abstract Prospection (Buckner and Carroll, 2007), the internal representation of possible futures is a ubiquitous feature of the human mind, with roots going back into our animal ancestry. Much psychological theory and practice, however, has proceeded in a framework that understands human action as determined by the past, while viewing any sort of teleology as mysterious and metaphysical. Understood as prospection, teleology is neither mysterious nor metaphysical--it simply is guidance by one’s internal, evaluative representations of possible future states. We call this phenomenon being drawn into the future. These unmysterious representations can in turn be understood as evaluative “If X, then Y” conditionals, and the process of prospection can be understood as the imaginative simulation of the possibilities encoded in these conditionals. We review the failures of behaviorism and psychoanalysis to account adequately for action as driven by the past, and we suggest that prospection is the crucial missing piece. Intriguingly, a priori considerations, too, point to the centrality of “If X, then Y” expectations in systems that learn from experience and regulate their behavior accordingly. We speculate that prospection may illuminate the function of consciousness and the role of subjectivity in consciousness. Speculating further, we suggest that prospecting provides a way of thinking about what is “free” and “willing” in “free will.” Finally, we consider how clinical practice might benefit by considering pathologies of prospection and from a shift in attention from past history toward being drawn into the future. 3 William James famously wrote, “My thinking is first and last and always for my doing” (James, 1890, I: 333). He might have added: “… and all my doing is in the future”. Even the simplest act, a slight movement of the eye tracking a moving object, say, extends forward, never backward, in time. And this movement will be more successful and more efficient if it accurately anticipates in the here and now what will happen in the then and there. Efficiency considerations alone, not to mention competitive pressures, mean that natural selection will have relentlessly favored organisms that do better at anticipation. Thinking designed for doing thus will be thinking designed for reliable prospection into the future. And ‘thinking’ here must incorporate evaluation as well as description. A good prospector must know more than the physical landscape—what is to be found where, with what probability—but also at what cost in effort and risk, and with what possible gain. The prospecting organism similarly must construct an evaluative landscape of possible acts and outcomes. The organism then acts through this evaluative representation, guided by the attractive or aversive prospects it holds. And the success or failure of an act in living up to its prospect will lead not simply to satisfaction or frustration, but to maintaining or revising the evaluative representation that will guide the next act. At any given moment, improving our chances for survival and reproduction lies entirely in the future, so learning and memory, though grounded in the past, had better contribute to successful prospection. Memory, for example, should be useable memory, 4 encoding information in ways that facilitate the simulation of new possibilities And learning should not be simple acquisition of past facts awaiting some use, or the entrenchment of previously successful habits—at its heart it must be about forward expectation. To see intelligent animals as drawn into the future is not to deny that the past influences the present and future, but it is to reject a pervasive and seemingly natural picture of how the influence of the past is transmitted. In this picture, past history, present circumstance, and inner states drive behavior, much as, in a classical dynamical system, the vector sum of forces operating upon and within a particle uniquely determine its subsequent trajectory. Intelligent animals, in contrast, use the past, it does not use them. The influence of the past is thus not the impulsion of cueball-like forces. Rather the influence is routed via the animal’s representation of possible futures, i.e., present representations of absent acts and outcomes. Intelligent billiard balls would have “a mind of their own”, and simple cue-strokes would not suffice to determine their trajectories—we would have to know how the world, and especially possible future acts and outcomes, look to them. Animals and humans are at least this intelligent. The Elimination of Telos As intuitive as our “teleological” account of thinking and acting might seem, it is decidedly not the account that has prevailed throughout the history of modern psychology. Instead, teleology has largely been anathema, to be replaced wherever possible with mechanical determination by the past. A history of the hold that the rooting 5 out of teleology has had over science begins with Galileo, Bacon, Hobbes, and Laplace, continues through Locke and Hume, and then wends its way through Darwin, Marx, and Freud to reach an apotheosis in Behaviorism. A recent survey argues that it persists today: … the [contemporary] social-cognitive approach to higher mental processes, like cognitive science in general, shares with behaviorism a basic deterministic stance toward psychological phenomena. By determinism we mean, quite simply, the position that for every psychological effect (e.g., behavior, emotion, judgment, memory, perception) there exists a set of causes, or antecedent conditions, that uniquely lead to that effect. … although [the] distinction between the two schools is certainly substantial and consequential, behaviorists and cognitive scientists do share certain assumptions about the nature of human volition and educe them from the same general philosophical foundations. [Bargh and Ferguson, 2000, p. 925] We, however, are less than sanguine about the prospects for a deterministic psychology. Reliant as they must be upon an underlying stochastic physics, neurological—and thus psychological—processes should be expected to involve an element of indeterminism, and understanding this could be an indispensable part of understanding them (see Grabenhorst and Rolls, 2011), a point to which we will return. But the metaphysics of causation and randomness is not our chief focus here. Our concern is with the heuristic force of psychology’s “canonical conviction” to determinism, which we believe has led to a distorted picture of human motivation and action. 6 The idea that people are drawn into the future might seem a truism or something that psychologists already “know.” But it is neither. An unabashedly over-simplified walk through the history of modern psychology will help throw the contrast we wish to make into relief. More importantly, this history will enable us to forge links in a long chain of evidence leading to our conclusion. The origins of experimental psychology lie in the early modern period in philosophy and physics, a time when the project of banishing “teleology” from the study of nature was indeed indispensable for scientific progress. Aristotelian or pseudoAristotelian explanations invoked “final causes”—causation in terms of ends: “Why does a heavy body fall to the earth? Because that is its natural resting place, where it strives to be.” “Why then does a cannon ball fall faster than a feather? Because it is heavier, and thus has a stronger affinity with the earth.” “Why then does the massive sun circle the earth rather than fall rapidly toward it? Because it belongs to the perfect and eternal celestial realm, and the circle is the perfect geometric shape.” Such explanations had powerful appeal because they seemed both to fit appearances and to explain the natural order in terms of intelligible purpose—a paradigmatic and humanly satisfying way of saying why something happens. However, once the early modern experimentalists learned how to measure nature with quantitative accuracy and use mathematics to formulate law-like regularities, these teleological accounts proved to be both at odds with observable facts and explanatorily otiose—in short, anathema to the growth of scientific knowledge. Sir Francis Bacon (1561-1626) wrote, “Inquiry into final causes is sterile, and, like a virgin consecrated to God, produces nothing” (1627, III:5). Galileo Galilei (1564-1642), rather 7 than posit invisible entelechies with a natural end or telos, studied the “science of motion,” establishing experimentally that the rate of free fall near the surface of the earth is the same (setting air resistance aside) for bodies independent of weight. Using his telescope, he observed changing sunspots in the “unchanging” heavens, while Johannes Kepler’s (1571-1630) detailed calculations of planetary motion led to heliocentric elliptical orbits prohibited by the Aristotelian natural order of things. The English philosopher Thomas Hobbes (1588-1679), rightly impressed by these astonishing discoveries, sought to apply a Galilean method to create a human science: break behavior down into its parts, observe these assiduously, extract law-like regularities, and then explain the whole in terms of the recombination of these lawgoverned parts. Hobbes (1651) decomposed human motivation into “appetites and aversions”, developing a law-based “knowledge of consequences” of how these forces operate in individuals to result in the impossibility of an Aristotelian “natural social order”—only by creation of an artificial political society do humans live together in peace. Hobbes’s method had the advantage of seeming obvious. After all, once all the motions of the parts are accounted for, what else could enter into the determination of behavior of the whole? One cannot imagine that humans, any more than billiard balls, can escape the “laws of motion” governing their nature. The concern to replace all appeals to teleology with law-based mechanical causation reverberated throughout the sciences. In physics, the French mathematician Pierre Simon Laplace (1749-1827), applied the new “mathematics of motion” on the largest scale, promising to bring the entire universe into one system. He wrote: 8 We may regard the present state of the universe as the effect of its past and the cause of its future. An intellect which at a certain moment would know all forces that set nature in motion, and all positions of all items of which nature is composed, if this intellect were also vast enough to submit these data to analysis, it would embrace in a single formula the movements of the greatest bodies of the universe and those of the tiniest atom; for such an intellect nothing would be uncertain and the future just like the past would be present before its eyes. [1814/1902] The universe needed no purpose to explain its order—Laplace’s demon did not need to divine any intelligent plan or posit any Aristotelian “striving” of matter toward final ends in order to see into the future. Knowledge of the distribution of matter and of the laws governing the forces that impart motion to its parts will suffice. The great 19th century intellectual triumvirate, Darwin, Marx, and Freud, each made a contribution to ridding their domains of teleology. Charles Darwin (1809-1882) argued that the seeming planfulness of living nature—or the apparent telos of organs, organisms, or species to behave so as to promote future survival and reproduction— could be understood instead as the result of a wholly non-teleological process of planless variation and selection. Karl Marx (1818-1883) replaced Hegelian teleology with historical materialism—the progression in ideas and social practices that Hegel had seen as the self-realization of Reason was seen instead as a sequence of unintended results of struggle for control of the means of production, spurred along by a mechanism of technological change and periodic crises. 9 And Sigmund Freud (1856-1939) removed from the individual psyche any vestige of the Aristotelian sovereignty of reason with its telos of the Good and the True. Freud saw a divided, not unified self, and sought a theory encompassing the nonrational and irrational as well as the rational. Conscious human thought and action could be understood as expressing an unconscious psychodynamic of drives, inherently conflictual, and wholly rooted in our individual pasts and human nature. Seemingly inexplicable behavior could be explained by digging back into a person’s childhood to unearth the sources of the unresolved inner conflicts that continue to drive adult behavior and determine its shape and meaning. These theories were major achievements, and their elimination of objectionable teleology from physics, biology, history, and psychology was of the first importance. But none went so far in eliminating teleology as 20th century Behaviorism. Behaviorism: Precursors Hobbes sought a mechanical understanding of human behavior, but he relied heavily on instinctive dispositions to act, and seemed to stumble when it came to explaining choice or intentionality. If all is motion and the consequences thereof, where is agency? For Hobbes, one’s “decision” or “will” was simply the “last appetite, or aversion” that happened to be in one’s mind before acting. What then gives behavior an intelligible semblance of rational purpose? For this a more powerful psychology was needed, and John Locke (1632-1704) and David Hume (1711-1776) laid its foundations: Associationism. Indeed, in the abstract Hume anonymously published of the Treatise of Human Nature, hoping to win 10 it some readers, he advertised that “if anything can intitle the author to so glorious a name as that of an inventor, ’tis the use he makes of the principle of the association of ideas, which enters into most of his philosophy” (Hume,1740b). For it permitted him to formulate an empirically respectable account of reasoning and intentional action—he could even account for the purposive self without the “rational homunculus” that so bedeviled Rationalist accounts. The first step toward understanding apparently rational self-guidance is to see how association leads to causal learning—sensations and ideas occurring in spatial contiguity or temporal succession give rise to cause-effect dispositions of thought. Impressions of sugar, an opaque white powder, being poured into bitter tea and mixed, are regularly followed with impressions of a still-transparent amber liquid with a distinctively changed, sweeter, and better-liked taste. With repeated exposure, the first images in this sequence come to entrain the later ones, and so we speak of the adding of sugar as ‘causing’ the tea to be sweet. When next we face an overly bitter cup of tea, this same train of images is awakened in us, calling to mind adding some sugar. And so we do. Asked why, we need only cite what is, in fact, on our mind—“Sugar sweetens tea, and I like sweet tea.” Similarly for William James, more than a century later, learning and motivation are not orchestrated in terms of “final ends.” Not one man in a billion, when taking his dinner, ever thinks of utility. He eats because the food tastes good and makes him want more. If you ask him why he should want to eat more of what tastes like that, instead of revering you 11 as a philosopher he will probably laugh at you for a fool. The connection between the savory sensation and the act it awakens is for him absolute and selbsverständlich. (1890, II: 386) An impulse to eat what is savory—like sipping agreeably sweet tea—calls for no further explanation. But not all impulses to act are so intelligible, many are instinctual and ‘blind’. In this respect, man is simply a scaled up animal: Man has a far greater variety of impulses than any lower animal; and any one of these impulses, taken in itself, is as ‘blind’ as the lowest instinct can be. (1890, II: 390) Why, then, isn’t human behavior more nearly blind and impulsive? It is obvious that every instinctive act, in an animal with a memory, must cease to be ‘blind’ after being once repeated, and must be accompanied with foresight of its ‘end’ just so far as that end may have fallen under the animal’s cognizance. [Ibid.] Resurrecting a Humean solution, James imagines a hen that has already hatched a brood, and that therefore associates the birth of chicks with nesting. Since the idea of chicks interests her, it strengthens the association and helps supply added motivation to the ‘blind’ nesting instinct. An ‘end’ will have been introduced, but ‘end’ is in quotes because no more than a past associative sequence is involved. He then imagines a boy whose first instinct, upon seeing a large toad, is to smash it with a rock. Which he proceeds to do. But the pitiful sight of the final moments of the squashed toad gives him new associations with smashing a toad, giving rise to a future aversion. That 12 we are beings capable of conscience and rational agency, and the mother hen not, can be put down to “man’s memory, powers of reflection, and power of inference” (James, 1890 II: 390), each, itself, an associative process. So a mechanical, driven-by-the-past picture survives translation all the way from atoms and billiard balls to the consciencestricken human. Behaviorism’s Triumph James’s Principles was, however, complex and mentalistic, and hardly free of all teleological elements. Later, he repented, seeking a much more radical empiricism under the banner of Pragmatism (James, 1904 and 1909). Complete victory in the elimination of teleology from the science of human beings, however, came a decade later with the rise of Behaviorism in the 1920’s. Like all of the preceding views, its chief motivation was methodological—and admirable: to permit the construction of a more scientific, more strictly evidence-based theory, without reliance upon teleology. To demonstrate and quantify mechanical causation was the order of the day, with observable stimuli causing observable responses. Indeed this is the very meaning of the terms “stimulus” and “response.” The animal—any animal—became an input-output device, whose behavior was directly determined by its past history. As J.B. Watson famously wrote in Behaviorism: Give me a dozen healthy infants, well-formed, and my own specified world to bring them up in and I'll guarantee to take any one at random and train him to become any type of specialist I might select – doctor, lawyer, artist, merchant- 13 chief and, yes, even beggar-man and thief, regardless of his talents, penchants, tendencies, abilities, vocations, and race of his ancestors. [1930, 82] This was, in many ways, a liberating vision at the time—especially in a society where imputing racial or ancestral “disabilities” was commonplace, even within the scientific community. And Watson did have the good sense to admit that he was speaking broadly, beyond his evidence, having never done the experiment (Watson, 1930, 83). The point, however, is not that all behaviorists held this picture, but rather that something akin to this picture might have held them: the behavior of animate nature, from flatworm to human, came under the comprehensive governance of Thorndike’s Law of Effect (1911) or, in B. F. Skinner’s (1904-1990) hands, under the “contingencies of reinforcement.” Again, this view had on its side a kind of obviousness—behavior is reinforced by the force of reward and decreased by the force of punishment. A rat comes to a T-junction in a maze it has run before. Which way will it turn, and why? Behaviorist learning theories (we will often say “learning theory” for short) said, in essence, “Just as, in Laplacian physics, initial conditions plus impinging forces determine the subsequent trajectory of a particle, so do the rat’s initial conditions (food deprivation plus a history of reinforcement) plus impinging stimuli (reaching the junction) determine her subsequent trajectory.” The distinguished experimental psychologist K. S. Lashley wrote in 1923: The problem which confronts the behaviorist is to find in the physical world deterministic relations between non-qualitative [i.e., non-mentalistic] discrete 14 entities in time and space which fulfill certain conditions of relationship laid down by subjective experience. [1923, 329] Returning to a Jamesian image of a diner, Lashley showed how both mentalism and teleology could be removed at a stroke: Our current psychological language is a weird composite of teleological and mechanistic terms. … I may say that I am hungry and purpose to have steak and onions for dinner. The subjectivist and the man-in-the-street gets the meaning clearly. Yet my words have only been accepted names for the facts that stomach contractions, salivary secretion, changes in visceral tonus, specific laryngeal and tongue movements, contractions of trunk musculature, and the like are occurring within my body. [1923, 349] He adds, “An introspective description of my purpose would not reveal an influence of the future on the present, nor does the behaviorist account” (1923, 349) Cause necessarily precedes effect, excluding notions of purpose from any serious explanatory role. Thinking otherwise is a metaphysical as well as scientific mistake. The two systems, mechanistic explanation and finalistic valuation, stand out as incompatible points of view, scientific versus humanistic. To the writer, the most serious defect of current psychology is the confusion of these points of view in the attempt to develop a science. [1923, 346] “Other sciences,” he continues, set the model for psychology, having “escaped from this thralldom” to “finalistic valuation”, and the associated “attempt to inject metaphysics into the science” (1923, 346-7). 15 Which gives rise to a question: Why aren’t determinism and “mechanical explanation” themselves metaphysical assumptions? The best answer is that they have functioned in the natural and human sciences less as metaphysical axioms than as methodological heuristics—prodding investigators to be unsatisfied with any account that cannot accurately predict behavior, or whose mechanisms remain obscure. These methodological virtues, along with an antipathy to older, introspective psychology, perhaps explain why the discoveries of quantum mechanics, or the muchearlier recognition in physics of the intractability of calculating three-dimensional trajectories in a classical system once it involves three or more interacting particles (see Wolfram, 2002), did not deter behaviorists from sticking to observables and following determinism as a “canonical conviction”. While the rest of science moved toward the view that prediction at best can provide only a probability distribution over possible outcomes, for the behaviorist, the scientific route was to push the Laplacian billiard ball model as far as possible. Of course, there are many variations in the details of different learning theories, but perhaps our somewhat cartoonish characterization captures the essentials for present purposes. In principle, it seemed, something like this picture must be right. No objectionable reference to “finalistic evaluation” should be needed to explain the rat’s turning left in the maze, despite the apparent forward-looking intelligence of its behavior. Humans are no exceptions. Our drives and reinforcers might be more varied and, mostly, more expensive to produce. The behavior we “emit” might exhibit wider “stimulus generalization”, longer stimulus-response chains, and more individual 16 variation. The role of “nature” (innate dispositions) vs. “nurture” (learning) might be disputed. But the mechanical model proved irresistible. This is the school that Martin Seligman cut his teeth on. As a student he was taught to eschew mentalistic explanations, like those in old-fashioned Associationism or Freudianism. But learning theory’s greatest bugaboo was “teleological” explanations. Such explanations were mentalistic—that was bad. And they were untestable—that was worse. But worst of all they were unintelligible—for they had the future causing the present, a stark violation of all natural law. Notions such as “expectations” of as yet nonexistent future events were “folk” explanations, examples of old wives’ tales, bad science. The overarching program of learning theory was to dispense with such notions and cure psychology of all impulse to impute unobservable purpose rather than establish observable behavioral regularities of classical and instrumental conditioning. Many are now tempted to say that the failure of Behaviorism lay in its over-reach, trying to use a theory that worked for “rats and pigeons” in the experimental setting to explain human psychology in unconstrained situations (Bargh and Ferguson, 2000; Schwartz, Schuldenfrei, and Lacey, 1978). But we believe that the crucial failure was in eschewing teleological explanation, and that this is perhaps clearest on its home ground: Behaviorist learning theory did not even work for white rats in the laboratory. Rats. How did this seemingly self-evident approach fail? From early behavioral research on, white rats, Behaviorism’s subjects of convenience, looked suspiciously future-oriented in ways that did not seem easily reducible to past habits—even in the venerable T-maze. Rats who had been reinforced for making the response of turning 17 left to get food had the right motor cortex of their brains ablated, making it physically impossible to turn left. If reinforcement were about motor habits, they should have walked up to the choice point and stalled—unable to “emit” the instrumental response. Instead, they walked up to the choice point and promptly made 270-degree right turns (Lashley, 1929), even if this meant rolling, somersaulting, or dragging paralyzed limbs. These experiments, and incidental reports of chance observations, e.g., of maze-trained rats escaping the starting box and walking diagonally across the top of the maze, directly to the food (Lashley, 1929), pointed clearly away from the idea that behavior was under the control of past motor “habits”, suggesting instead that an acquired “cognitive map” governed navigation flexibly, permitting goal-directed behaviors of unprecedented kinds. This idea gained further support from doing what Behaviorists most strenuously urged: making close observations of actual behavior. Psychologists who looked up from their automatic recording systems noticed that, when not “overtrained”, rats at the choice point would occasionally turn their heads to the left, to the right, and back again. Such “vicarious trial and error”, along with a suite of experimental evidence, suggested that “intervening brain processes are more complicated, more patterned, and often, pragmatically speaking, more autonomous than [is accepted by] the stimulus-response psychologists.” (Tolman, 1948) Even more trouble for banishing what had been seen as “teleology” came from careful experiments on “information processing” in Pavlovian conditioning. Pavlovian conditioning was the non-mentalistic operationalization of pure Associationism, and it was supposed to occur based on the mere temporal contiguity of conditioned stimulus 18 (CS) and unconditioned stimulus (UCS). Unfortunately, Pavlov’s original Russian was mistranslated—his terms ‘conditional’ and “unconditional” were rendered ‘conditioned’ and “unconditioned” (McGuigan and Ban, 1987, 138). ‘Conditioned,’ the past participle, has the connotation of a fixed relation, whereas ‘conditional,’ Pavlov’s term, implies the “if-then” representation of possibilities. Translation might seem beside the point—all that counts are the experiments. But in this case, words matter—as experiments later showed. It turned out that informationally redundant, but nevertheless contiguous, conditioned stimuli, did not “condition” (it had become a verb)—contrary to what Associationist principles would lead one to expect. Degree of conditioning was found to be an exquisitely sensitive mirror of how much information the CS gave about the likelihood of the UCS (Rescorla, 1988). This suggests not the passive ingraining of a blunt associative “connection”, but the dynamic learning of a fine-grained conditional probability–exactly what one would expect if animal learning operated via an “if CS, then UCS” expectation and revision. The final fight over “teleology” in learning theory, underappreciated at the time, came down to avoidance learning and “two-process theory.” Avoidance learning, to the uninitiated, looks exactly like learning an if-then conditional: “if I jump, I will avoid future shock.” Animals are first given escape training in which a tone is followed five seconds later by a shock. The animals can escape the shock by jumping over a barrier to the safe side. Jumping over the barrier also turned off the tone. Animals soon learned to jump as soon as the tone went on and before the shock would have gone on. This was called “avoidance” learning because the animals avoided getting any more shock by jumping when the tone went on—which both prevented shock and terminated the tone. 19 If any animal behavior “looks” future-oriented, it is avoidance, so reducing it to the present and the past was considered a tour de force for learning theory (Rescorla and Solomon, 1967). Here’s how the reduction worked. The behaviorists denied that the animals expected anything at all, claiming that a nonoccurrent event—the avoided shock—could have no role in conditioning. Instead an occurrent event, the tone coming on, had become fear-evoking by the usual process of Pavlovian conditioning of pairing with shock. So jumping was actually reinforced by getting the fearful tone to stop, and was not at all motivated by “avoiding” a shock that never came on. By contrast, the cognitive theorists claimed that the animals learned that the tone predicted shock and thus acquired an expectation that if they jumped they would get no shock (Seligman and Johnston, 1973). That expectation—an evaluatively-coded if-then conditional about a merely possible outcome—was in itself sufficient to guide the jumping behavior. The stage was set. This was a rare instance of a head-on collision between two theories permitting a crucial test, in this case by “extinction.” Once the animals became steady jumpers, they never got shocked again, and so tone was no longer paired with shock. This is called “Pavlovian extinction” which should cause the tone to lose its fearful properties. If the cause of jumping were to turn off the fearful tone, as the behaviorists contended, the animals should stop jumping once they had numerous trials in which fear of the tone was extinguished. If on the other hand they were jumping to prevent the anticipated shock, as the cognitive theorists contended, the animals should keep jumping. Their if-then expectation that jumping would prevent shock was born out 20 every time they jumped—extinction ends association but is not disconfirmation of the conditional. So if the conditional guides their behavior, the jumping behavior should persist. Which is exactly what happened. Hundreds of trials later, the animals were still jumping. The simplest explanation—to which no counter-response by learning theorists was ever offered—is that the animals had acquired a well-confirmed evaluative representation of the future. Even so, behaviorist learning theory limped on for another forty years, actually starting to use “expectations” to explain conditioned behavior, and gradually giving up wholesale anti-mentalism and the grand project of demonstrating that apparently future-directed behavior was wholly reducible to past-driven behavior. What was this but an admission that understanding how these animals make their way in the world requires us to see that they act with an eye to the future, drawn toward good outcomes, and drawn toward avoiding bad ones? Rats, it seems, are more predictable when we postulate that they have expectations portraying positive or negative future actions and outcomes, than if we posit drives and “conditioned responses” (Berridge, 2004). Perhaps humans should be given as much credit? All these lines of research suggested that complex learning could not be explained without positing “forbidden” internal representations of the future. But why were they “forbidden”? A conceptual error seems to have animated Behaviorism, in which something genuinely suspect—a metaphysical teleology of causation backward in time, of the present by the future—was conflated with something not at all mysterious, namely, guidance by a system bearing causal and evaluative information about possible futures. 21 Freudianism and its Discontents As Behaviorism evolved into cognitive psychology and behavioral neuroscience, so did classical Freudian psychology evolve into contemporary psychiatric practice. In this case, too, there was movement beyond the orthodox framework of hydraulic drives, developmental stages, and personality complexes. Analytic theorists felt free to introduce novel models and techniques. However, what often remained in theory and practice was the attempt to explain present thought and action as an interplay of unresolved and largely unconscious conflicts, grounded in the past. The psychoanalytic assumption that such unresolved conflicts drive adult personality is much less amenable to testing than the simpler mechanisms posited by the behaviorists. And analytic psychotherapy itself is a lengthy and multi-faceted process, impossible to standardize. Even so, and even though recent experimental research has provided increasingly strong and detailed information about the importance of unconscious processes (Hassin, et al. 2003), fifty years of psychoanalytic practice aimed at uncovering repressed childhood conflicts has failed to provide convincing evidence of efficacy. More telling, perhaps, is the accumulating weight of carefully-done longitudinal studies that have found disappointingly small effects of childhood events on a range of adult behaviors. For example, in an 8000 twin-pair study of the onset of adult depression, childhood events had almost no predictive value, whereas genetic factors had great predictive power and recent events had moderate predictive power (Kendler, Walters, and Kessler, 1997). What the careful reporter of the psychoanalytic literature 22 would be entitled to conclude is that remote past experiences, even when they seemingly are important, greatly underdetermine present feeling or action, and may not point to a particular underlying psychodynamic that must be unearthed if dysfunctional behavior is to be changed. The failure of Behaviorism’s favored mechanisms to account for animal behavior could be established—insofar as the failure of any global scientific paradigm can be “established”—in a wide variety of experiments and in exquisite detail because of the theory’s allegiance, in the end, to testing and evidence. In the case of psychoanalysis, it remains possible that improvement in measuring techniques, more accuracy in the finegrained reconstruction of the past, a more complete reconstruction of recent events, and a more complete set of psychodynamic laws, might greatly improve psychoanalytic prediction. We cannot gainsay this, except to say that no such improvements have yet emerged. The Logic of Expectation The reluctant acceptance of expectations by Behaviorism was no small concession. Indeed, it is a fundamental reorientation in thinking about how past experience influences future behavior. Rather than seeing the past as contributing drives and habits, we should see it as primarily contributing information about the prospects of possible acts or outcomes. This information updates the if-then evaluative representations that draw the individual forward by making some actions or outcomes more, or less, attractive than others. 23 The first part of our argument has been that prospection into the future is an indispensable element of psychology, and we advanced this idea on historical and empirical grounds. But it is possible to provide a more philosophical, a priori argument that points to the same conclusion. Anyone setting out to design an organism to respond appropriately to its needs and environment would put the formation and revision of if-then conditionals at its heart. Such conditionals are the simplest form of prospecting, but achieve its essential function: to give future, i.e., at the moment absent, possibilities a role in guiding the individual comparable to present experience. Present, rewarding experience can directly move an organism. But saturation by the present can make an organism blind to what comes next, exposing it to risks or lost opportunities it could have predicted—and responded appropriately to—if only it had represented likely futures as well as it represents the present. Suppose a diner contentedly chewing his roast beef notices on a chalkboard an announcement of today’s dessert—caramel banana bread pudding. This he would enjoy greatly, but only if he has some appetite remaining when the main course is cleared away. For that to happen, a conditional value, a mere idea (“If still hungry, then rich dessert will be delicious”), must compete favorably with actual value, the savory beef. He knows what bread pudding is like, loves the idea of having some (“I so seldom see it on menus these days. And this place will do it well!”), and in a moment we see him pushing aside the roast beef aside to leave room for the wonderfully rich and heavy pudding. What changed for the diner was not his hunger or the beef’s savory taste, but his expectations. Change expectations, and you change behavior—quickly and automatically in most cases, more consciously and deliberatively in others. And 24 expectations can change behavior without needing additional reinforcement. New information alone—a man noticing a menu or a rat discovering a way to escape the maze’s starting box—can do it. Expectation is a matter of cottoning on to patterns in the world, and intelligent action makes use of them. Acting on expectations is a form of pattern completion, of contributing one’s part in a synchronization with the world, satisfying an available if in order to obtain an attractive then—whether one is a rat jumping a barrier to be free of shock or a diner pushing aside a tasty dinner to save room for dessert. The crux of our argument might be put in this way: the chief way that behavior becomes more intelligent through experience is via the acquisition and modification of expectations, which turns passive experience into active experimentation. This deep connection between expectation and learning was perhaps first explored in the work of the philosopher Rudolf Carnap (1891-1970). He was studying confirmation functions, hoping to understand their basic logic (Carnap, 1950). And he arrived at a startling conclusion: intelligent learning from experience requires bias, One might think the opposite—an ideal learner would begin without bias. Put in formal terms, this learner would assign to all possible states of the world the same initial probability. For simplicity, imagine that time is finite and that there are only a finite number of properties and entities in the world, so that imagining a possible state of the world is imagining, for each time, entity, and property, whether the entity does or does not have that property at that time. Call such a possible way things might be a “statedescription” of the world. To learn, one simply records at each moment which entities have which properties, crossing off the list of possible state-descriptions any 25 incompatible with this. If the sun comes up this morning, cross of all state-descriptions in which it does not. Now imagine that one has been doing this diligently for several years, the sun coming up each and every day, and someone asks, “Will the sun come up tomorrow?” The learner would then survey his list, count the number of remaining state-descriptions in which it does come up, and report the ratio with the number remaining in which it does not. Doing this, the learner answers, “50:50”, indistinguishable from chance, just as it was before any observations had been made. What went wrong? The initial, bias-free assumption that all possible statedescriptions are equally probable in effect made each event probabilistically irrelevant to any other event. Thus, despite the obvious pattern of sunrises, no number of past observations of sunrises could change what one expects about the next sunrise. The same point can be put in logical terms, without reference to probability. No one particular atomic fact—the possession of a given property by a given entity at a given time—logically implies any other particular atomic fact. So, a perfectly logical, unbiased system of inference would leave the learner nothing to say about tomorrow’s sunrise. Carnap (1950) went on to show that if, instead, our learner had begun with certain notions about which combinations of entities and properties are more likely than others, then observations of individual facts could quickly displace one’s view about tomorrow’s sunrise away from chance. But what of it? Bias could be a self-fulfilling prophecy, a form of epistemic pre-destination in which one’s starting point distorts forever the future conduct of one’s inquiry. Any connection with overcoming error or gaining accuracy would be merest happenstance. A dilemma: either be free of bias, and so unable to learn, or be biased, and so unable to unlearn. 26 It was subsequent development of a truly dynamical theory of learning that provided the way out of this seeming dilemma and laid the groundwork for contemporary formal learning theory. For Carnap, probabilistic relations were fixed a priori, but in more recent models initial probabilities are simply “priors”, and revisable in the face of new experience (Jeffrey, 1953). We can think of these theories somewhat generically as having a feedforward/feedback structure: an initial prior (feedforward) generates an expectation which then is compared with new evidence as it comes in (feedback); the prior is then adjusted in a manner to reduce discrepancy with the observed outcome. (*) expectation (feedforward) observation discrepancy detection (feedback) error-reducing expectation revision (new feedforward) observation discrepancy detection … Systems of kind (*) have been shown in formal learning theory and formal epistemology to permit learning very generally, and to have the important features that, for a wide range of cases, as evidence increases (i) the influence of initial expectations tends to diminish (“bias washes out”), and (ii) the resultant expectation tends to approximate the actual relative frequencies in the sample (“convergence”). Thus, whether the learner begins with a low (0.1) or high (0.999) initial expectation of a sunrise on the morrow, sufficiently many observations will lead his expectation of a sunrise tomorrow to converge asymptotically to 1.0 (see also Miller, Galanter, and Pribam, 1960). 27 (*)-like processes turn observation into experimentation. Each experience is more than a fact to be recorded—it is also an opportunity for success or failure that can generate an information-bearing “teaching signal”. That is how expectation works, and for this reason engineers can design (*)-like circuits which function to attune themselves to any of a variety of environmental variables. “Smart” shopping programs succeed in encouraging you to buy more by using if-then conditionals learned from others’ choices to anticipate what might appeal to you, and these “expectations” create the possibility of feedback and refinement based upon which recommendations you accept or reject. Like smart animals, these programs do not simply accumulate facts—they seek out advantageous patterns in the world by continually forming and acting upon, and thereby testing and revising, expectations. We have now presented two prongs of the argument that prospection is required for psychology: one historical and empirical—that animal learning could not be successfully explained without teleology, and the second logical—that on a priori grounds adaptive learning must involve expectations. This mention of adaptation brings us to the third prong: the evolutionary underpinnings of prospection. We will then present the fourth prong—the prospecting brain. Evolution and Realism The premier example of (*)-like learning is Bayesian updating, but isn’t it well known that this is impossibly unrealistic—requiring the learner to keep track of a large number of conditional probabilities, and continually revise them with the arrival of new 28 information? And didn’t Kahneman and Tversksy (1979) show that even beings as intelligent as humans, much less rats, are terrible at probabilistic reasoning? Probabilistic learning, however, is largely implicit rather than reasoned. The saccade of an eye following the trajectory of a moving object is seldom the object of reasoning. Yet the circuitry governing the visual systems of humans and animals learns from experience to follow trajectories more reliably and efficiently (see Abegg et al., 2011). Like any act, simple eye movements are extended in time, and do not end exactly where and when they began. The brain structures that govern action would therefore be selected to be anticipative in nature, “looking ahead”. Expectation of where a moving or ambiguous object might next appear can, if accurate, save the brain energy and time, improve one’s chance of success, and perhaps even avert disaster. Efficiency and efficacy will be gained in large and small ways thousands of times a day if the near future can be anticipated more reliably. Indeed, the acquisition of the presaccadic expectations encoded in the circuitry that controls the movement of the eye appears to approximate Bayesian updating rather closely (Abegg et al., 2010). Similar gains in efficiency can be made in any of the subsystems of the mind and body—regulating everything from reaching and grasping movements to hormone release to metabolic rate. “Animals run on batteries,” the evolutionary ecologists tell us, and there is no recharging once they’ve run out. Energy spent exploring must be constantly offset by energy gained exploiting, so effectiveness in both, and efficiency in balancing them, is at a premium. As the so-called “Good Regulator Theorem” from systems theory suggests, a good regulator for a system will build a model of that system and its 29 environment (Conant and Ashby, 1970). Effective regulators do not simply detect errors, they develop expectations to anticipate and avoid errors, metabolizing experience with the same efficiency that they metabolize food, to attune the individual to future opportunities and perils. Prospecting thus makes evolutionary sense. We hypothesize that animals become more effective and more populous, more capable of intelligent responses to the world around them, if they live not simply in the present, but rather anticipate what might lie ahead and initiate action proactively. So we think evolution rewarded adaptations that shifted current behavior to be ever more oriented by the projected future, and see this as a vital force in shaping animal cognitive, affective, and motivational systems. Even the pattern completion found in elementary associative learning could have been seen from the outset, not as mechanical repetition of past responses, but as a form of “moving ahead” from the present into the future—as the term ‘completion’ suggests. Organisms that could form and act on reliable forecasts of future benefits and costs would have gained advantage by natural selection over organisms whose reliability was concentrated on the past and present. Indeed, millennia of natural selection have so rewarded such gains in efficiency and effectiveness that foraging mammals placed in environments with varied food sources, costs, and risks will tend over time to allocate their energy and time in nearoptimal ways (Krebs et al., 1978, Dugatkin, 2004). They are equipped via experiential learning processes to solve what would otherwise be a complex linear programming problem. 30 Moreover, they appear to do so not through the entrenchment of habit, but through the acquisition and representation of complex factual and evaluative information that permits continuous, fine-tuned behavioral variation as circumstances, or their needs, change. Foraging mammals have systems of neurons whose functional organization and firing rates separately correlate with: the identity of stimuli, their intensity, the magnitude of specific positive vs. negative hedonic rewards or food values, the internal state of the animal, the relative value of a stimulus (i.e., its value in the current context, e.g., of satiation), the absolute value of a stimulus (i.e., its long-term value, e.g., physiological need), the probability or expectation of a given positive or negative outcome, the occurrence of a better-than-expected reward prediction error, the occurrence of a worse-than-expected reward prediction error, and the absolute risk and expected value of given actions (Grabenhorst and Rolls, 2011; Schultz, 2002 ; Tobler et al., 2006 ; Singer et al., 2009; Craig, 2009 ; Quartz, 2007; Preuschoff et al., 2006; Kringelbach and Berridge, 2009; Rolls et al., 2008). A recent series of experiments with rats in T-mazes makes the action-guiding role of these evaluative representations vivid, and enables us to revisit the question of “vicarious trial and error” with the advantage of improved neurological recording. Hippocampal “grid cells” and “place cells” been identified providing direct evidence of the construction of cognitive spatial maps (Langston et al., 2010), and activations in these maps when an animal is at the choice point spread alternately down the two arms ahead of the rat’s current location, as mental attention seems to shift forward to explore what lies in store down the two arms (Johnson, van der Meer, and Redish, 2007; Johnson and Redish, 2007). Without leaving the choice point, the rat has “looked down” 31 both of the paths before her; having previously encoded greater expected value for one than the other, it turns in that direction. Guidance by such “value laden” cognitive mapping is possible because of the flexible learning of (*)-like systems, and because the “ancient”—or, better, highlyevolved—affective system can both encode expectation and entrain motivation, so that it can shape behavior implicitly or directly. Perhaps we should not be surprised, then, that today’s foragers, the products of eons of selection for efficiency, have evolved a highly efficient brain that learns surprisingly like a Bayesian and, within the temporal and computational limits of their brains, allocates its time and energy surprisingly like a rational investor (Quartz, 2007). Prospecting and Empathy Effective prospection must extend across space and time and in a social species such as ours across other individuals. It appears that a key capacity for accomplishing this is to mentally simulate social outcomes empathetically, and individuals deficient in empathy are likely do poorly at planning and following plans (Demurie, et al., 2011). The costs or benefits of a given action here and now depend heavily upon what is happening elsewhere, including especially what is present in the hearts and minds of fellows or foes. Here the human capacity for empathetic simulation appears to be critical to an efficient solution to the “other minds” problem (Decety and Chaminade, 2003). To survive, flourish, and reproduce effectively, social animals must forge and maintain ties with kith and kin, most notably mating partners, offspring, and near relatives. Natural selection would tend to favor any gain in the efficiency or accuracy with which 32 individuals can prospect the likely internal states, needs, and actions of those around them. The wrong move, or a missed opportunity, can quickly result in a failed mating, a dead infant, a lost alliance, or a fight one has no hope of winning. For highly social animals such as humans, a well-attuned map of the social landscape is therefore no less important than an accurate map of the physical environment. And for all their obtuseness and seeming short-sightedness, humans are extraordinary among animals in their capacity for the social prospection necessary for long-term, shared enterprises such as government, law, schools, commerce, collective bargaining, and retirement planning. Even the “self” as a persisting entity essentially involves such prospection. Commitments, relationships, values, and convictions, for example, are not just a matter of how one has acted or is acting, but of how one thinks about the future, and how one will or would act in various future contingencies. Mental simulation enables one to envisage these possibilities concretely, and to think through or develop one’s responses to them before they arise. In so doing, one is partly determining who one is. Learned information in general can do the organism some good only in the future— the same asymmetry of time makes it impossible to gain past benefits or avoid past harms. We should thus expect that the fundamental orientation of perception and cognition will be prospective, not retrospective. Vision involves calculating expectations of motion, and beliefs are distinguished from stray thoughts in the mind not by their representational content—since stray thoughts are also representations—but by the fact the belief generates expectations and shapes subsequent behavior and interpretation of new evidence accordingly. 33 Desire versus Drive We have argued that perception and cognition should be prospective. So, too, we claim, should motivation. Here we come to the critical difference between drives and goals or desires. Theorists as diverse as Freud, Konrad Lorenz (1903-1989), and many behaviorists (Hull, 1943; see also Dollard et al., 1947), placed the concepts of drive and drive reduction at the core of motivational psychology. A drive was seen as a motive force that could point the individual in a given direction and propel him toward it. This behavioral trajectory would then continue until the drive was reduced, e.g., by the satiety produced by obtaining food. Such drive-reduction would both instigate and reward the behavior, making learning possible. Thus behavior that seems intelligently goal-directed will be produced, but with no need for any representation of a goal to guide it. Mechanical causation replaces telos. The idea is wonderfully simple, even obvious. Kent Berridge writes: If motivation is due to drive, then, the reduction of deficit signals should satisfy this drive and essentially could be the goal of the entire motivation. … The drive reduction concept of reward is so intuitive that it was thought to be self-evident for decades. The power of this idea is so great that some behavioral neuroscientists today still talk and write as though they believe it. All the more pity, perhaps, that the idea turns out not to be true. … Even for food and hunger, reducing drive via intravenous feeding proves to be relatively ineffective at stopping eating (Berridge, 2004, p. 191). 34 If action is not propelled forward by drives and reinforced by drive reduction, how are we to conceive of it? Our argument thus far has been: it should be seen instead as drawn into the future by the prospects of possible events, acts, and outcomes, and modulated by changes in these prospects. To borrow Berridge’s terms, positive affective interest (“liking”) in an act or outcome elicits motivational incentive (“wanting”) to bring that act or outcome about. And this is precisely how desires or goals differ from drives. Desires, for example, are not inarticulate urges, but, as philosophers have long noted, involve attraction by an appealing idea to pursue what that appealing idea represents. Aristotle spoke of an “apparent good”, and Elizabeth Anscombe of a “desirability characteristic” (Aristotle, 1999; Anscombe, 1957). This is known all too well by advertisers, who sell their product by first selling a seductive image of it. If this image is taken on board, desire then lays an alluring path ahead, with all of the elements of narrative—from initial attraction and promise, to the tension of longing, to the challenges of overcoming obstacles, to the climax in union with the object of desire, and the denouement in its consumption or use. Small wonder that novels and plays as well as advertisements are written around the drama of desire, not the relief of drive reduction. Of course, not all desiring is as explicit as a Harlequin Romance. Most likely, the greatest part of the evaluative representations encoded in the affective system, as described above, are implicit rather than conscious and deliberative. This is the truth in James’ remark that “not one man in a billion, when taking his dinner, ever thinks of utility”. 35 So in speaking of being drawn into the future we are not claiming that the thought of future utility is always before one’s conscious mind when acting. James imagines that the best explanation of the man’s eating his dinner lies in impulse triggered and sustained by present sensory stimulation. We believe that the man was drawn to the dinner table well before any savory flavor met his tongue or nose. Most likely, he was attracted to the table because he simply liked the idea of having dinner at this restaurant. Such affective representations then motivated him to think of how he might bring such a meal about, and the glow of these “desirability” representations placed in a positive light such unappetizing activities as putting on a coat and walking out into the cold night—enough to move him to do it. Reasoning and motivation worked “backward” from the attractive future goal—not forward from a hunger drive. (For evidence of working backward from goal to means in monkeys as well, see Saito et al., 2005.) Of course, as James noted, l’apétit vient en mangeant. Perhaps once he is eating his roast beef—a food he heartily enjoys—motivation has become a matter of a drive, and he’ll eat until he has his fill and the drive is gone. But recall the bread pudding. Though the beef is as savory as ever, the diner stops eating. Not because hunger has been sated—indeed, part of the point of his stopping is to maintain some hunger. Rather, a tempting idea has emerged, which competes with a tempting taste— and wins. Prospection makes this sort of response to the greater, though distant, good possible. In desire, unlike the “blind” drives of which James wrote, our attention and motivation are directed forward, toward possibilities that “feel” attractive when imagined, and away from possibilities that “feel” bad or wrong. No declarative concept of utility or 36 expected value need play any role in this, even though the subjective “feeling” one experiences when contemplating alternatives is the product of an affective system that is highly sensitive to utility and expected value (Railton, 2002). Are we denying that it is ever useful to analyze behavior in terms of drives or habits? No. But we are claiming that drives and habits are a poor general model for the behavior of intelligent animals of most sweeping scope, their prospecting possibilities. The Scope of Prospection Prospection is at the core of four kinds of mental simulations. The first is navigational: Imagine your residence and the nearest supermarket. Now mentally walk block by block from your front door to the nearest supermarket. If you turned right out of your front door, now do the exercise again starting with a left turn and taking therefore a different route to the supermarket. (If you turned left, start by turning right.) The second is social, about other minds: Imagine that you have been invited to have a chat with President Obama about whom he should choose as the “White House Person of the Year.” Whom would you nominate? How would the President react to that nomination? Now imagine telling the President three reasons why he should choose your nominee. Imagine the objection that you think he would most likely raise. Now imagine your response to that objection, and how effective it would be for the President. 37 The third is intellectual: As you are reading this article, with its grand claim that action is drawn by the future rather than driven by the past, what mental activity are you engaged in? If you are like most active readers, you are mentally trying out various reactions to the material. You are making arguments against the idea, or finding holes in our reasoning or weaknesses in our evidence, or thinking how you might improve, qualify, or defend the paper’s conclusions. You are imagining trying to explain the argument to someone else, perhaps a class or a colleague, or how you might use it to advance your own positions. You are asking whether the idea is really so radical after all, or building toward a decision not to waste any more time on this article. The fourth exercise is memorial. Recall a happening in your life that turned out badly because of something you said or did. What could you have said or done that might have made it turn out better? Run through that scenario. How would things have been better? Really? What would have been the negative consequences as well? What the first three exercises demonstrate is the process of prospection, the mental running of hypothetical simulations of the future. What the fourth demonstrates is the mental running of counterfactual simulations of the past. Each of these exercises demonstrates your enormous facility for generating, exploring, and evaluating alternatives to the present you now confront or the past you can’t help but remember (Buckner and Carroll, 2007). In each case you free yourself from your actual conditions to take advantage of your powerful mind to “do,” explore, assess, and perhaps learn from, mere possibilities. 38 Ordinary mental life is chock-full of this activity. Mind wandering, day-dreaming and fantasy are the least of it. In an ordinary conversation, for example, the mind is continually “running ahead” of present stimuli, exploring possibilities for the meaning of the other’s remarks or for one’s own next response. mentally completing sentences that have only just begun (see Havas, 2007). Navigation, inference, interpretation, appreciation of other minds, reliving and rehashing past episodes, are all commonplace examples. Dreaming itself appears to involve the exploration of possibilities, thereby contributing to problem-solving in rats and humans alike (Lee and Wilson, 2002; Wamsley, et al., 2010). What proportion of mental life is consumed by prospection? Our guess would be that it is a very substantial proportion, and especially of those many times a day when we are implicitly or explicitly “making up our minds” about what to do, think, or feel (Smallwood, Nind, and O’Connor, 2009; Morsella, Ben-Zeev, Lanska, and Bargh, 2010). The Prospecting Brain What parts of our minds are being “made up”? Full blown human prospection encompasses a number of kinds of imaginative simulation: navigation, other minds, intellectual and emotional evaluation, memory, and others. What evidence is there that these share something in common? Intriguing new evidence from neuroscience, especially neuroimaging, suggests there is a single core brain network that plays a role in these diverse forms of imaginal simulation. The story of how this network was discovered is fascinating in its own right (see Raichle and Snyder, 2007, Buckner et al., 2008 for reviews), and, as we will see, reveals much about the ubiquity of prospection in our mental lives. 39 Functional brain imaging has been used for more than two decades to identify patterns of neural activation associated with cognitive tasks. Studies are typically structured with alternating time blocks – blocks of task-related activity alternating with resting blocks, which, among other purposes, serve as a control condition. During resting blocks, subjects are typically told to look at a blank screen. Contrasting levels of brain activation during a task block against a resting block reveals neural activation specifically associated with the task. But what is going on when human participants are at “rest?”Is it really “rest?”How should one interpret the reverse contrast in which the resting block is compared with the task block? During the early years of neuroimaging (the 1990’s), researchers knew that this contrast shows a highly reliable pattern of neural activation in the brain’s midline and the lateral parietal lobes (Shulman et al., 1997). One might have thought that when subjects are given no explicit instructions (other than to look at a blank screen), then the resulting patterns of brain activation would be highly unstructured and thus would vary widely. Yet these studies show that the resting state is associated with a highly reliable and uniform pattern of activation. What might be the significance of this pattern? Marcus Raichle, Debra Gusnard, and colleagues proposed that the regions exhibiting enhanced neural activations at rest relative to task constitute a functionally integrated neurobiological system, not mere noise (Raichle et al., 2001, Gusnard et al., 2001). They called this the ‘default mode’ system, and proposed that it supports internally-directed mentation during intervals when there are no externally cued cognitive demands. Their key insight was that during these resting blocks, the brain was not actually at rest, but rather actively engaged a highly organized, but default 40 functioning. Important additional support for the default mode hypothesis came from parallel research into the phenomenon of spontaneous fluctuations in neuronal activity. It has long been known that neural activity exhibits a pattern of slow oscillation (<0.1 hertz) in individual neurons and larger collections of neurons (Biswal, 1995). Researchers discovered that the spontaneous fluctuations measured in the regions of the default mode system are highly correlated with each other (Greicius et al., 2003). This suggests that these regions are not simply oscillating at random, but rather seem to constitute a highly coherent network. But what activity are they organized around? Dozens of neuroimaging studies have assessed the neural basis of tasks involving the various forms of imaginal simulation that we have emphasized such as mentally navigating an environment, thinking about the future, remembering the past, and thinking about others’ mental states. Quantitative meta-analyses of these studies (e.g., Spreng et al., 2008) demonstrate there is substantial overlap in the regions implicated in these tasks, suggesting that a common underlying network subserves these functions. Moreover, the regions implicated in these tasks correspond to the regions identified by Raichle, Gusnard and colleagues as part of the default mode system. This supports the intriguing idea that when people are at rest and not engaged in some externally-directed task, their mental lives are largely occupied by imaginative simulations of one sort or other, projecting oneself into other situations and circumstances (Mason et al., 2007, Buckner and Carroll, 2007). This hypothesis gains additional support from two recent studies. The first found that people with a more pronounced tendency to engage in internally-directed mentation (such as day-dreaming and other unstructured prospective activities) generated more robust activity in the 41 default network during rest intervals of fMRI scans (Mason et al., 2007). The second study used ecological momentary sampling during fMRI scanning of a rote cognitive task, and found enhanced activity in the default mode network during intervals when subjects’ minds wandered off task (Christoff et al., 2009). Remembering the Past and Imagining the Future The default mode hypothesis set the stage for research on how these various forms of imaginative simulation are interconnected (Buckner and Carroll, 2007; Goldman, 2006). One particularly fruitful line of investigation has characterized the rich connections between episodic memory of the past and prospective representations for the future. Episodic memory refers to autobiographical memory for specific prior events, including information about who was present, what occurred, and what was felt (Tulving, 1985). Interestingly, episodic memory appears to be a fundamentally constructive process. Each time an event is remembered, past episodes are reconstructed anew, and in most cases a little bit differently than the last time. At first it was thought that this reflected an initial process of memory consolidation, by which memories had to be replayed several times until they stabilized, after which each remembering would be more or less identical. However, more recent work has concluded that memories continue to change and evolve as long as the person lives. Memory never “nails things down” once and for all. It continues to be a flexible and fallible re-creation of what probably happened. 42 This has puzzled researchers for decades (Bartlett, 1932; Neisser, 1966). Why is memory designed in such a fallible way? An emerging view suggests the answer may have to do with intimate ties between episodic memory and the constructive processes important for prospection of the future. According to the constructive episodic simulation hypothesis put forward by Daniel Schacter, Donna Addis, and their colleagues (Schacter and Addis, 2007a, 2007b), episodic memory provides details needed to construct prospective simulations of future events, and thus the two rely on some common component processes and neural substrates. Both episodic memory and prospective simulation rely on such common processes as the storage and recall of individual details, mental imagery, and self-referential processing. In addition, both involve constructive operations that bring together these elements into a coherent episode. Prospection, however, engages these constructive processes more vigorously, as the task of constructing multiple possible futures is inherently more complex and requires extrapolation beyond the given evidence to project what would be the case under various hypothetical and even counterfactual conditions. Support for Schacter and Addis’s hypothesis comes from recent neuroimaging studies that identify a core network (which encompasses the regions of the default network discussed earlier) that is engaged during both retrospection and prospection (Schacter and Addis 2007; Spreng 2009). Prospection preferentially engages certain regions that are required for flexible recombination of information, including anterior hippocampus and frontopolar cortex (Addis 2009; Addis 2010). 43 So episodic memory is fallible because it is in the service of prospection. While flexible construction might be seen as a flaw in the reconstruction of past events, it is very much a virtue in attempting to forecast future events—and all the benefit memory might confer upon the individual lies in the present and future, not the past. If episodic memory functioned only to file away singular snapshots of individual events in the past, it would be more veridical, but it would have much more limited value for prospective simulation, which must redeploy stored information to represent a multiplicity of possible futures--many of which will not be mere repetitions of past events—and also their dependency upon which candidate action is undertaken. Flexibility in combining and connecting past information to represent possible future actions and outcomes contributes to the identification and selection of behaviors that will maximize survival and reproduction (Buckner and Carroll, 2007; Schacter and Addis, 2007; Suddendorf and Corballis, 1997, 2007). Evaluating Alternative Futures Merely imagining the future bloodlessly is not enough. To survive and thrive, a creature must be able to rapidly evaluate the acts and outcomes imagined. How does the brain encode such evaluative prospections? We have already reviewed evidence that animal brains are equipped with sophisticated systems, especially affective systems, that encode the future expected value of possible states. One natural hypothesis about how brain might compute the expected value of future states, has its roots in the thinking of early economic theorists (Rae 1834; Boehm-Bawerk, 1889): to evaluate the value of future states of affairs, project one’s self into a simulation of these states, then allow one’s affective systems to register the evaluative valence of the 44 simulated states as if they were actually happening to you. Drawing on Buckner and Carroll (2009), we call this the “self-projection” hypothesis, and we now review two lines of evidence that provide initial support for it. A number of neuroimaging studies have investigated how the brain represents stimuli that are related to one’s self (Northoff et al., 2006). In some studies, subjects viewed valenced pictures and were asked to rate the pictures in terms of how selfrelevant they are (Phan et al., 2004; de Greck et a.l, 2008). In other studies, subjects are presented with trait adjectives or personality descriptors, and evaluate to what extent the adjectives describe one’s self (Kircher et al., 2000; Kelley et al., 2002; Schmitz et al., 2004). In still other studies, subjects imagine actions (Ruby et al., 2001), thoughts and beliefs (Ruby et al., 2003), and feelings (Ruby et al., 2004) from a firstversus third-person perspective. These studies reliably find that more self-relevant material (or are viewed from a first-person perspective) activates regions of the prospective cognition network, including medial prefrontal cortex, posterior cingulate/retrosplenial cortex, and inferior parietal cortex, an observation that has been quantitatively corroborated in a recent meta-analysis (Northoff et al., 2006). The selfprojection hypothesis requires that to evaluatively engage with simulated alternative futures, processes underlying self-related experience and processes underlying simulations of alternative futures must somehow be joined. Such extensive overlap between regions engaged by self-related stimuli and regions that subserve thinking about future states of affairs thus provide some support for the self-projection hypothesis. Evaluating Prospective Rewards 45 Roughly a dozen neuroimaging studies have examined the neural basis of intertemporal decision-making (see Peters and Buchel, in press, for a review). In this task, subjects choose between smaller rewards available earlier versus larger rewards available later. These studies reliably find activity in regions overlapping with the brain’s prospection network during intertemporal choice (Kable and Glimcher, 2008), leading several authors to speculate that intertemporal choice involves prospective simulations of the available rewards (Weber et al., 2008; Luhman et al., 2009; Sripada et al., 2010). Two studies tested these speculations more directly by manipulating imaginative engagement with future rewards. In one study (Benoit et al., 2011), subjects imagined themselves spending an amount of money in a specific scenario, for example, $35 dollars at a pub, versus merely estimating abstractly what that dollar amount could purchase in that scenario. In the second study (Peters et al., 2010), subjects wrote down future plans (e.g., taking a vacation) during pre-scan interviews, and verbal tags referring to these plans were presented during fMRI scanning prior to a subset of choices between smaller earlier versus larger later rewards. In both studies, the high imagination condition preferentially activated regions of the brain’s prospection network, including medial prefrontal cortex, posterior cingulate cortex, and hippocampus. Moreover, greater activity in these regions immediately prior to intertemporal choices predicted more future-minded choices. By drawing links between among prospective simulation, the valuation of future rewards, and future-mindedness, these studies provide additional evidence for the hypothesis that imaginative simulation of future outcomes plays a role in evaluation and choice for these outcomes, which is a key element of the self-projection hypothesis. 46 Taken together, we find it compelling that a common neural network seems to subserve such key forms of prospecting as mental simulation, episodic memory, other minds, and the evaluation of future possibilities, as diverse as their content is. Prospecting beings, it seems, would have held an evolutionary edge for long enough to shape neural architecture itself, and the existence of a prospecting brain constitutes the fourth and final prong of our argument. Prospecting Prospection So far we have argued that prospection is central to action-guidance in intelligent animals. It ranges from simple animal expectation and the implicit evaluative representation of if-then conditionals to full-blown conscious simulation of alternatives from multiple points of view in the human case. We intend the idea that we are thus drawn into the future rather than driven by the past as a heuristic. Our hope is that this new heuristic will help promote a science of brain and behavior that will be more fruitful at predicting and understanding action than the heuristic that we are driven by the past. One way of judging the fruitfulness of a new heuristic is whether it illuminates major issues that are opaque to the heuristic it seeks to replace. Restricting ourselves for now to the human case, we suggest that taking prospection seriously might re-orient thinking and illuminate (a) consciousness and subjectivity, (b) free will and self-control, (c) and psychopathology and therapy. Prospecting: What is Consciousness for? We now pursue a bald speculation, that a major, perhaps the major, function of human consciousness is to permit better prospection of the future. In prospection, 47 states of affairs not present to the senses or portrayed in memory nonetheless play a guiding role for action akin to the triggering role that the stimulus plays for the driven-bythe-past stimulus-response heuristic. We speculate that one of the most important and adaptive functions of consciousness is to construct mental simulations of future possible actions and outcomes. One function of consciousness, acknowledged by nearly all writers about the functions of consciousness, is integration of information. Morsella (2005) called this the “integration consensus” — but he pointed out that not all inconsistencies and conflicts reach consciousness. He proposed instead that only conflicts resolving competing impulses for actions that involve skeletomotor muscles become conscious. Thus, only when bodily action is required to resolve competing impulses is consciousness needed. His implication is that animal consciousness resolves inner conflicts so that the animal can produce overt behavior in a coherent manner. When an animal has impulses to go left and to go right, its conscious mind integrates the relevant information and decides which way to go. We do not intend to take a position on animal consciousness in this paper, but it is clear to us that prospection goes beyond this more rudimentary ability that we share with many animals. Prospection mentally simulates possible sequences of actions and outcomes, evaluates them, and selects among them. Some of this can be done unconsciously. But there are two crucial things that prospection does better than unconscious thought, as delineated by Baumeister and Masicampo (2010): social communication and complex sequential thought. Baars (2002) reviewed evidence that consciousness of the sort that we call prospection is crucial for sequences of thought. Priming studies that rely on 48 unconscious processes have generally failed to work with even two-word combinations (e.g., Draine, 1997; Greenwald & Liu, 1985). In dichotic listening tasks, people can process and respond to single words in the unattended channel, but meanings of sentences do not get through (Mackay, 1973). Logical reasoning, which requires building a particular sequence of thoughts, is heavily dependent on conscious processing (De Neys, 2006; DeWall et al., 2009; Lieberman, Gaunt, Gilbert, & Trope, 2003). Baumeister and Masicampo (2010) proposed that conscious thought can be understood as a place where the unconscious mind constructs meaningful sequences of thought. This is highly relevant for prospection, because prospection is always about sequences: If I do A, X will happen, whereupon I would have to do B, which would lead to Y, after which I could do C, and so forth. Consciousness enables prospection because it enables the mind to simulate multiple rounds of actions and consequences The social functions of consciousness are also highly relevant. Although many acts can be executed without conscious supervision, talking coherently is not one of them: People have to be conscious of what they are saying. The mental processes that produce conscious thought are closely linked to those that produce speech, and in fact Fox et al., (2011) showed that having people voice their thoughts usually has little or no effect on performance, which suggests that talking does not constitute much of an extra mental load over and above thinking. Furthermore, the use of logic, morality, and other forms of reasoning may have evolved to resolve social disputes. Davidson (1982) proposed on conceptual grounds that rational thought is inherently social, such that only beings who communicate can 49 enjoy full rational thought. Haidt (2007) likewise has argued that moral reasoning serves essentially interpersonal purposes, such as justifying one’s actions to others. Human evolution diverged from other animals with its emphasis on communication. Humankind became superior hunters not by virtue of speed, strength, or fangs, but by group coordination and this depended on planning. The formation of a group plan for hunting required all the three functions of consciousness we have discussed thus far: resolving competing possibilities, understanding a sequence of acts and consequences, and communicating that with others. Tomasello (1999) noted that although chimps may hunt as a group, each individual is essentially acting on his own and looking for individual opportunities. In contrast, human groups form plans with interlocking, complementary roles. In order to form a group plan, however, it is essential that all members of the group can share the same understanding of what is about to happen (and can debate alternative versions of the plan). Consciousness facilitates communication among social beings, who teach each other how to think about the future and can debate different possible courses of joint action (Smallwood et al., 2009; Smallwood, Schooler, Turk, Cunningham, Burns, & Macrae, in press).It is hard to imagine a better aid to this than prospection. We do not know how the snowball of language and culture first began to roll. But perhaps it grew so dramatically and quickly because consciousness and communication extruded the processes and products of prospection into the light. Language and culture are multipliers of the effectiveness of prospection since many minds are so often better 50 than one—creating a wider pool of evidence, shared examination, functional specialization, and coordinated responses. So our speculation is that consciousness makes for better prospection—and better sharing of prospection. Prospecting: Subjectivity This thought suggests in turn a different way of thinking about the question, “Why do we have subjectivity at all?” What does subjectivity—the phenomenology or “felt qualitative character” of experience itself—add to whatever can be accomplished without it? Put in our terms, “Why does conscious prospecting also have a subjective face?” Our only excuse for speculating about this aspect of the “hard problem of consciousness” (Chalmers, 1996) is that there are not many other plausible answers to “Why subjectivity?” If we adopt Block’s distinction between access consciousness and phenomenal consciousness (Block, 2004), our discussion of consciousness and prospection thus far has focused upon access. Consider, for example, the problem of developing the kinds of communication and coordination essential for complex cultural life. Signaling systems relying upon implicit processes possess a number of advantages. They can make deception more difficult, for example. But an animal that cannot access or describe its own mental states is at a distinct disadvantage where non-deceptive communication and cooperation is concerned. Devoted parents trying their best nonetheless experience a great deal of frustration trying to figure out what their pre-linguistic baby is crying about, or what it is trying to do, or what it wants. Co-planning with such an individual is impossible, despite 51 the best will in the world. Five-year-olds, by contrast, are fairly adept at telling parents how they feel, what they want, and what they intend to do, and so can engage in a good deal of co-planning. In turn, adults can communicate their thoughts, feelings, and intentions to the five-year-old, an essential part of the transfer of culture and the child’s social development. It might seem as if the main capacity that accounts for these differences is language, but if we imagine a linguistically endowed individual who somehow lost the ability to gain subjective access to her own mental states—from simple feelings to complex intentions—very similar barriers to effective communication and coordination would exist. Similarly, how would one plan one’s life or deliberate about what to do if one’s own possible future thoughts, desires, feelings, or aims were a closed book to consciousness. Present sensory stimulation does not in itself afford this information about future possibilities, so an alternative “sensory” route must be provided by imagination or simulation, just as much as, when dealing with a young child, one needs an alternative—verbal—route to his thoughts and feelings. Sensory experience and prospective experience would not, however, be on the same playing field as long as actual experience lay entirely within one modality and possible experience lay entirely within another modality, such as off-line simulation. Inherently different, they might nonetheless be brought into comparison if they could appear with equal presence—a common-coin “felt” character—in some mental workspace the content of which was “subjective” and accessible. This is the same condition for the social common-coin we mentioned above: if others have access to 52 their thoughts, feelings, and intentions, too, then such content can in principle be shareable, and interpersonal comparison, planning, and principles possible. We have been suggesting, perhaps somewhat paradoxically, that consciousness might be relatively unimportant simply to record concurrent experience or synthesize multi-modal sensory information. Certainly many non-human animals have multiple, extremely acute sensory capacities that work together to guide behavior with great speed and deftness, without giving any behavioral or brain evidence of conscious experience. Instead, the subjective face of consciousness might be important for providing an effective competitor to one’s own concurrent experience—so prospective conditions and conditionals can be commensurately experienced. Prospection will be an effective competitor to actual experience only if it can be equally affecting. Consider the problem of making an affecting film or staging an affecting play. Here one must provide a convincing artificial experience, and this is done by manipulating actors, props, and scenery in ways that reproduce the impact of the real. Appearance is what reality, on the one hand, and films or plays, on the other, have in common. Prospection is also the making a film or staging a play, providing a impactful, artificial experience. But what are its actors, props, or scenery? They must be pure appearances, so to speak, wholly mental mock-ups—in short, subjective states. And note that this point is independent of conscious access. Subliminal experience and implicit prospection could share appearances, and be equally affecting, despite lack of conscious awareness. Prospecting: Consciousness and Commensuration 53 Consider a concrete example of self-conscious prospection. (We ask the scholarly reader to forgive our venture into phenomenology here, but the point is difficult to make without an example.) Martin Seligman is sitting at his computer in the process of deciding what to do in the next hour. Here is what he was conscious of in the course of about 30 seconds: 1) I could go on to the internet and play bridge. Who might be available? Mark Lair, but he’s usually at lunch. Peter Friedland, he’s in Taiwan, probably going to sleep, but there are a few lesser lights, likely to be available. They make errors. Anyway, I’ve wasted a lot of time playing bridge lately. 2) I might help Mandy teach Carly and Jenny about the Silk Road. I don’t know much about Asia. But the kids would love it, and I haven’t spent any time with them today. Mandy might find it intrusive, having prepared the lesson. But she thinks I have not done my share of teaching lately. 3) I might make myself some lunch. There’s some Moroccan chicken left over in the Fridge. It’s pretty high calorie though. And I’m meeting Phil Voss for dinner at Le Bec Fin in only five hours. But I could order only their three course meal. Maybe Mandy was saving the chicken for the kids’ dinner. 4) I could keep working on this damn paper. But I’m having trouble thinking through examples of compelling counterfactual simulations. Maybe a bridge break will help. But this is itself a pretty good example, 54 so maybe I should keep plugging. Why bother? I don’t have a deadline, since this paper is for my own amusement. Peter will be disappointed if I don’t follow up soon. 5) All those tulip bulbs need planting. I could use the exercise, particularly with Le Bec Fin coming up. The temperature is good, but the ground is soggy. Tulips can be planted even if it gets really cold, no rush. They might rot. I did lift weights for twenty minutes already today. But I could use some fresh air. It would calm me down. I need it particularly after the argument with my Dean. Notice how multidimensional these simulations of the future are, notice that each is evaluative, and notice also how incommensurable the evaluations seem. By what metric does the pleasure of playing bridge with Mark Lair stack up against annoyance of letting tulip bulbs rot or the satisfaction of seeing brightly colored tulips in six months or the anticipation of Moroccan chicken or the guilt of not working out? In the market, disparate human actions achieve commensurability through the unifying metric of money: so fixing a hole in the roof and a loaf of bread become commensurable by attaching a dollar value to each and computing how many loaves of bread you can buy in exchange for fixing your neighbor’s roof, or how many loaves of bread you would have to forgo to have your own roof fixed. Affect is the brain’s common currency for value, and conscious, subjective affect would permit the possible futures to be brought into the open for explicit comparison with each other. We have argued that conscious subjective affect attached to prospections would enable them to compete effectively with ongoing experience. Here 55 we are asking how such prospections can compete with one another. Each one of Seligman’s competing prospections, has a (perhaps nebulous) global affective valence, but within each there are also conflicting valences (he doesn’t know much about Asia, but the kids would love it, and Mandy thinks he’s not done his share). Properly responding to many situations, and especially social situations, is not a matter of feeling one particular shade of one particular emotion. On the contrary, taking the full measure of the situation will involve keeping before the mind a mixture of “unblended” feelings of different shades and degrees, sometimes, even conflicting feelings representing diverse perspectives that should not be combined or resolved. Non-conscious summing methods may be adequate when the values at stake and appropriate responses are straightforward, even if complex—as in optimal foraging or operating from a fixed set of preferences over multiple dimensions. But what if certain foods are taboo and yet one’s need is desperate and one’s elders disapprove yet one’s children are at risk and more powerful neighbors look down upon anyone who respects this taboo? One will need to act, but doing so successfully is not something easily passed along to an algorithmic learning system that produces a net action tendency. And in such cases there is as much a question about how to feel or what to show yourself as feeling as how to act. Often every available act has costs, so that even choosing the best act may also require making amends to those disadvantaged by it, or steeling oneself for their disapproval. Given multiple incommensurable dimensions and conflicting values and perspectives, none of which will go away even a sum could be struck, we don’t quite see how implicit algorithms could do the job alone if humans are to be capable of all the cultural and social innovation and adaption lying within human reach. If they could 56 instead feed into a final common but non-summative path of explicit mental prospecting and affective assessment, then perhaps humans would be able to use their intelligence and imagination to best effect. Human life might be much the same if humans saw in black and white rather than color, or saw the colors differently. But it would be incalculably different if humans could not keep before their minds the often conflicting thoughts and feelings and memories afforded by experience, or use voluntary thought processes to consciously prospect alternatives in light of these “unblended” or “unsummed” facts and values. To be sure, in daily life and in most of the regulatory decisions made by the executive brain, there are genuine constraints of cognitive resources and of time. Consciously keeping track of all the component elements would be out of the question. But even in such cases, it can be best to act in awareness of complex feelings and conflicting thoughts—indeed, this seems to be the kind of thing that humans specialize in. This is to say that mirrored subjectivity is a real time way of solving “other minds” problems—and, we speculate, crucial to understanding why humans have subjectivity. Prospecting: Freedom of the Will The concept of free will is entirely opaque under the driven-by-the-past heuristic. We believe that the drawn-by-the-future heuristic allows a tractable recasting of what ordinary people understand by “free will.” A recent investigation by Stillman et al. (in press) sought to discover how ordinary people understood free will. They asked participants to narrate an event from their lives in which they acted of their own free will or, in another condition, not of their 57 own free will. Actions reflecting free will were more likely than the non-free-will actions to emphasize pursuit of long-term future goals. There was no difference with regard to short-term goals. Free actions were more likely than the non-free ones to be about conscious deliberation and reflection. The free actions were also more likely to be consistent with the person’s moral values, and the free actions were also more likely than unfree ones to bring about positive outcomes. Taken together, these findings show that everyday understandings of free will are about the of long-term, beneficial outcomes, aided by conscious reflection and principled commitments. These features point to the centrality to the experience of freedom of guidance via prospection. Acting freely or willingly is a matter of generating and evaluating multiple possible future courses of action and so enhancing freedom is thus a matter of enhancing powers of generating and evaluating options. Three apparently distinctive advantages of human prospection are most enhancing: complexity, time horizon, and accuracy about the future. Complexity. Plans consist of sequences of actions linked in a coordinated way to achieve a goal. Plans are usually decomposable into parts, each of which achieves some proximal subgoal (Miller, Galanter, and Pribam, 1960). In order to get to Boston, I need to fill the tire with air. Once that is done, I can drive to the gas station. With the tires filled and the gas tank full, I can drive all the way to Boston. Often parts of plans are themselves decomposable into further parts. Given their decomposable structure, efficient construction of plans requires a distinctive kind of prospective ability. It not enough to simply prospect the outcomes of 58 single actions. Rather, one must be able to perform sequentially linked prospections and as we suggested above, consciousness is needed for sequential thought. The general form of such prospections is: If I do A1, O1 will occur. Given O1 the case, if I were to do A2, O2 would result. And so on. Thus sequential prospection, and the attendant ability to rearrange and build more flexible plans, not only expands the size of option sets, it makes available a vast array of new options that are dramatically more likely to achieve one’s goals. Time horizon. Some actions unfold over a short-duration: the rat in a T-maze selects going either right or left and receives a reward in a matter of moments. An extensive literature review by Roberts (2002) concluded that most animals can project at best 20 minutes into the future. In contrast, humans can project years ahead and adjust current behavior accordingly. This contrast between humans and rats is set out not to deride the achievements of animal minds. When at a fork in a maze, expecting that danger lurks down one path and not another is an impressive feat. How much more impressive is it then that when at a fork in life, let’s say choosing a major in college, we have a the ability to prospect not just the immediate consequences of each option, but we can mentally ‘see’ in rich and vivid detail the various ways one’s life might unfold. Accuracy about the future. Improvements in accuracy about the future includes being able to imagine oneself in different subjective states that one will experience in the future. Unlike animal expectancies, human prospection includes representations whose content is mental representations of states that one is not in now—metarepresentations. We can also readily prospect among motivational states that we are not in now. According to the Bischof-Kohler hypothesis (Bischof, 1980; Suddendorf and 59 Corballis, 2007), non-human animals are unable to represent future events in which they have motivational states that differ from their current motivations, and this impairs their capacity for planning for their future needs. Thus, a sated animal cannot represent the fact that it will be hungry in the near future and thus plan for this eventuality. If the anticipated motivation of its future self is one that the individual does not want to have, he can instead formulate options which prevent these anticipated future desires from arising. In one of the most famous works on free will in the modern philosophical literature, Harry Frankfurt (1988) argued that freedom of the will consists in having the will that one wants to have. According to Frankfurt, both humans and animals have the ability to have ‘first-order desires’, i.e., desires directed at doing this or that action. But only humans have the ability to step back and form second-order desires, desires about which among their various first-orders desires should or should not be acted on. For example, a person genuinely trying his best to shake his addiction to narcotics both has a desire for the drug, but also has a higher-order desire that the desire for the drug should not win out. If this addict uses the drug, he is not free, according to Frankfurt, because his first-order desires conflict with his desires of higher order – he does not have the will he wants to have. So prospection expands options in at least three ways. Sequential prospection enables complex, flexible planning, expanding the number and quality of options. Prospection with a long time horizon enables options that unfold not just over days, but rather over years and decades. And prospection with meta-representation enables better accuracy about the future. 60 Prospection gives sense to what it means to be free: being free means having options, and so beings with these prospective powers have vastly more (and vastly better) options, and they are commensurately more free. Prospecting: Willing We now turn from “freedom” to “willing” in order to explore the mechanisms by which a particular option is selected. There is one sense of “will’ that comes into play when we engage in the spontaneous or deliberate prospection of future possibilities. This feels “free” because the mind freely explores possibilities, and it feels like “freely willing” because we act simply by making up our mind one way rather than another in the course of this free exploration. The experience of “freely willing” is running through these prospections until one feels that one’s mind is made up, and then letting the settled outcome of this process entrain our behavior, and nothing more. The “idea of what to do” upon which we settle is in an obvious sense our idea, since it came about through our own uncoerced mental activity. And it is this “idea”, an evaluative representation of a possible act, that then entrains our subsequent action. No “intervention” by a transcendental will is needed for the idea to translate into action. Nor is a rational homunculus needed to “freely endorse” this act–settling the mind after freely exploring options by following what “seems best” or what one cares about most is already a free endorsement. Indeed, were we to describe the activity of such a homunculus, it would be no different from this. We will be “doing what we want to do because we want to do it”, but with a sense of freedom, not causal necessity. That is because each of these simulations has an emotional valence and the one that emerges 61 as most positive (or least negative) will be the future toward which we are drawn. This fits the phenomenology of willing, but not of causal impulsion. Indeed, at no point need one “ask permission” of motive force in order to generate the intended act—the affective character of our evaluative representations naturally induces an orchestrated response of directing attention, thought, feeling, and motivation “under the idea” upon which we settle. Had we been more drawn to another possibility, that one would have entrained subsequent action instead. This is the sense in which we will feel, appropriately, that “I could have done otherwise if I’d wanted to.” Our formulation is far from the last word about the mechanism of willing, and we are acutely aware of two inadequacies in our account. The first is “how is one’s mind made up” in leading to a “settled” evaluation among the options. In the section (above) on consciousness and commensuration, we postulated that a complex evaluative process among alternative options must exist, but we made no claims about the details of this process. We simply do not know what the rules of evaluative summation among the options are, but once prospection is put at the core of free will, this now becomes a tractable, empirical issue amenable to the kinds of analysis traditionally used in the field of judgment and decision-making. The second inadequacy is the role that controlled versus automatic processes play in prospecting. Under some conditions, external circumstances frame the question which is prospected. “If you had ten million dollars to spend over the next year, what would you do?” sets off a panoply of prospections, many of these being automatic associations (“a clinic in North Philadelphia, no, in South Sudan.”) Under some circumstances, a controlled process frames the question which is prospected. “I want to 62 play internet bridge over the next hour, so how should I proceed?” Under some circumstances, an automatic process such as being thirsty frames the question. It is clear that a mixture of automatic and controlled processes are involved in the initiation, maintenance, and conclusion of the prospecting process, and the best we can say is that this becomes a tractable, empirical issue. The central point of our analysis of willing is, however, that there is no such thing as a “will.” It is worth remembering that the modern notion of a “will” as a thing is a relatively recent reification, with an origin based in religion rather than observation or science. Aristotle’s term boulesis, used for the desire which combines with an idea of an act to yield rational action, comes from the notion of “taking counsel” or “thinking over”, a form of prospection, and Aristotle’s rational man chooses by via “deliberative appetition”, not an inner act of willing (1999). In ancient Greek thought and law, the equivalent of “willing”, ἑκών, meant being favorably disposed to seeing one’s idea of an act brought into being, and murder, φόνος, was a matter of foreseeing the death of another, wishing for it, and doing it as a result. In Old English and in many contemporary Germanic languages, “I will it” is synonymous with “I like it” or “I want it”, and in many Latinate languages, “willfulness”, (e.g., volonté in French) has the same root as “desire”, (vouloir). “Free will” for the French is libre arbitre, roughly, “free weighing and judging”, with no reference to a special volitional faculty. Prospecting: Temptation and Self-Control But isn’t there another, heftier notion of “free will,” distinct from this “freely willing” kind of prospection, that requires the “will?” What about those cases in which we face 63 temptation, and in which, having already come to a settled representation of, say, avoiding dessert, we do face a problem when the alluring dessert tray actually arrives? Here we need the notion of effortful “self-control.” Isn’t a powerful will—above and beyond focused prospection—necessary to resist temptation? We believe that the facility of human self-control, like the facility of spontaneous willing, relies on prospections and that no additional entity called “will” is needed even here . As for prospection, laboratory economic studies of self-control often require it, namely the choice of future outcomes relative to present ones, such as whether a person would rather receive $100 today or $120 one month from now. (e.g., Ainslie, 2006; Loewenstein, Reed, & Baumeister, 2004). As for an entity, “will,” let us look more closely at temptation. The difficulty in resisting temptation is that the here-and-now temptation activates impulses to enjoy it, while at the same time the environment offers little to activate contrary impulses that can resist longer term harm (Baumeister & Scher, 1988). However, if the person can mentally prospect the future harm, and the prospection of future harm is subjectively vivid, he can activate those more enlightened impulses. For example, coming home after being paid, the young man is attracted by the saloon. To resist, he imagines the dismal scene some hours hence, drunk, no money for rent, sleeping on couch, wife angry all weekend. Vividly prospecting future outcomes is thus a valuable, perhaps indispensible, support for self-control. Does effortful self-control require that we postulate a transcendental will? We think not; instead, we believe that competing motivational systems are sufficient to explain resisting temptation. Temptation usually takes the form of a highly salient stimulus associated with immediate, strong wanting, even if one at the same time does 64 not like the idea of breaking with one’s long-term goals. One may like the idea of achieving those more distant goals, and have a good deal of positive affect invested in them, and actively dislike the idea of giving in to temptation, so that one’s settled prospection is to resist. Even so, the abstract evaluative representations may pale in comparison to the attention-grabbing, proximate immediately gratifying stimulus. So how does one resist? Mischel and others found that resisting such a locally-powerful force appears not to arise from an iron will, but from a set of skills that weaken the salience of the tempting stimulus or strengthen the salience of the competing, longer-term goals (Moore, et al., 1976). This sort of “affective regulation” is indeed effortful, in the sense that it requires the expenditure of cognitive energy and the mobilization of affect to shift attention, refocus cognition, and thereby arouse different feelings. It “feels like” work because it is work—just as sustained focus on a difficult task is work (Baumeister & Tierney, 2011). Understanding the “willpower” of Mischel’s subjects is a matter of understanding the control of prospective cognition, not understanding the nature of a mysterious, contra-causal force within. The self-control needed to resist a strong temptation does not need an overridingly powerful mental “brake” or “will”, it requires only enough energy to keep attention focused away from tempting stimuli and toward longer-term prospects or principles. Debating metaphysics is beyond the scope of this paper and we are neither affirming nor denying that human action is “free” in any metaphysical sense. We firmly believe, however, that the prospection point of view frames what people experience as 65 “free will” into a tractable, psychologically credible form: For ordinary choices, free will amounts to evaluative prospection among possible futures with the most valued prospect entraining action. For effortful self-control, just the same, plus the activation of skills that strengthen the vividness of prospection or that weaken the salience of the immediate temptation. One long theme of evolutionary progress was to enable organisms to move from the past to the future in shaping their actions. To the extent that there is a reality behind the experience of free will, it likely reflects this theme and what goes by the name of free will in humans is best understood as an enhanced ability to base current behavior on the future. Thus, the evolutionary progress from being driven by the past and bound by current stimuli, to being oriented to future possibilities is, in our view, an increase in freedom. Prospecting: Therapy and Psychopathology The question of whether we are drawn by the future rather than driven by the past is not just academic. The major psychotherapies invented in the twentieth century were founded on the premise that humans are driven by the past, and so the therapies naturally focused on undoing the dirty work of the past. Psychoanalytic theory holds that present symptoms are caused by unresolved sexual and aggressive conflicts from the past, usually the distant past. So therapy focuses on reliving past events and gaining insight into them. It is not an overstatement to say that the results of one hundred years of this kind of therapy are disappointing. 66 Behavior therapies, similarly, derive from the premise that psychopathology consists of maladaptive habits learned in the past. So therapy focuses on extinguishing those habits and reinforcing more adaptive habits. Unlike psychodynamic therapy, this modality has been subject to serious outcome study, and overall it is mildly to moderately effective (Seligman, 2007). Even in its most cutting edge forms, however, undoing the past is the central task of all the behavior therapies. Exposure therapy (Foa et al., 1999), for example, used to treat posttraumatic stress disorder following rape, consists in rehearsing the traumatic experience in the safety of the therapist’s consulting room in order to extinguish anxiety and habituate. Cognitive therapy is somewhat less concerned with the past, or at least the distant past, than are psychodynamic therapy or behavior therapy. Its concerns center on the present and the recent past, with therapy focused on realistically disputing “automatic thoughts.” The thoughts can be about the past (“everything I have ever touched turned to shit,”), the present (“my boss hates me,”), and one’s character (“I am unlovable.”) But it also has elements of dealing more realistically with the future (“I lack the ability to achieve my goals”). Of these three modalities, which together make up the dominant schools of modern psychotherapy, cognitive therapy is the one that is most easily supplemented (or reformulated) to illustrate our thesis and so this is the modality we use to illustrate our speculations about prospection and psychopathology. Consider as examples the following prospective reformulations of several of disorders: 67 Agoraphobia: the fear that if I go out in public, I will become sick or go crazy,. Panic Disorder: the fear that if my heart starts pounding, I will have a heart attack. Obsessive-compulsive Disorder: the fear that if I don’t flush the toilet in multiples of three, disaster will strike. Generalized Anxiety Disorder. Chronic expectation that something unspecified, but awful, will soon occur. Major depressive disorder: Chronic expectation that the future will be miserable and that if I try to improve things, I will be helpless to make it any better. In such formulations, what has gone wrong is a maladaptive (and often mistaken) if-then prospection. We do not doubt that the past plays a large role in bringing such beliefs about. We speculate, however, that working directly on the mistaken belief about the future will be at least as effective as revisiting the source of the belief, which is often inaccessibly buried under the detritus of the past. There are at least five ways a therapist can assist the patient by dealing explicitly with maladaptive prospection. 1) Enhancing the prospection of alternatives. 2) Developing more effective prospection 3) Disconfirming unrealistic prospections. 4) Incentivizing the future. 68 5) Building meaning and purpose. Enhancing the prospection of alternatives. In the cognitive therapy of panic disorder, for example, a patient might believe that when she feels her heart pounding, she will go on to have a heart attack, and this belief leads to a spiral of increasing panic. The therapist poses the possibility that that a pounding heart will not lead to a heart attack. She suggests an alternative future, previously unconsidered, that a pounding heart is just a normal symptom of mounting anxiety, and that knowing this will itself stop the upward cascade of anxiety (Clark, 1986). In depression, for example, the only prospection the patient might entertain is “no matter what I will try, I will still not get into graduate school.” The therapist might suggest broadening of prospection by working through an entire range of other possible routes to graduate school (“volunteering for work in the professor’s lab,” or “taking a summer course in advanced statistics”) or alternatives to going to graduate school (“Teach for America”). Increasing a patient’s ability to generate appropriate alterative prospections on her own will involve increasing her affective as well as imaginative skills. If we are right that perspective-taking and empathetic simulation play a vital role in effective prospection, then the importance of both kinds of skills becomes yet clearer. Forwardlooking therapies therefore would devote some of the time now dedicated to the exploration of past events instead to helping the patient explore possible future situations. Training in empathy is now commonly part of medical school curricula, and a recent study found that such training could have a noticeable effect on clinical practice and physician attitudes if it involved behavioral and affective, as well as cognitive, 69 elements (Jenkins and Fallowfield, 2002). Something similar seems worth trying in a therapeutic setting. We confess that we ourselves know very little about how to broaden and nuance prospections. Indeed one of the purposes of this section is to call attention to the need for creative interventions in this modality, as well as the development of appropriate measures of prospecting skills and styles. Developing more effective prospection. The translation from settling upon a favored action or outcome and action is often prompt and seamless, but failures can occur, especially when the goal—however highly valued—is distant and indistinct while temptation and distraction are proximate and concrete. As the sad history of dieting shows, simply increasing emotional investment in the goal may not solve the problem, nor will ever-deeper probing into individual. Prospective-oriented therapy can, however, teach individuals strategies that increase their chance of success. For example, Peter Gollwitzer and colleagues have shown that working with the if-then structure of the mind through the formation of situationally-cued “subplan implementation intentions” can help individuals achieve better self-regulation and greater success in achieving long-term ends (Gollwitzer and Sheeran, 2006). This prospective strategy can be made yet more effective if combined with another, “mental contrasting,” in which people vividly imagine a positive future in which they have succeeded at their goal and contrast this to the negative reality that they now face and that stands in their way (Adriaanse, Gollwitzer, Oettingen, et al., 2010). A related class of strategies for more effective prospection was discussed above (Mischel, et al., 1992); techniques for shifting or sustaining prospective 70 focus help long-term goals compete more effectively with temptations, as do techniques of making the longer-term aims more salient. Disconfirming unrealistic prospections. In agoraphobia, for example, the prospection that I will inevitably get sick, throw up and make a fool of myself if I go out in public is a false and an unrealistic distortion. The cognitive (or behavior) therapist can disconfirm this by going with the patient on gradually increasingly difficult trips, culminating in a visit to a shopping mall, showing the patient that it will not lead to panic, vomiting, and making a fool of herself (Marks ,1987). In parallel for depression the prospection that no matter what I do, my boss will despise me is often an unrealistic distortion. Proposing new ways of doing things (handing in reports one week early, being in the office before the boss arrives and staying until after she leaves) will disconfirm the distortion of helplessness. The heart of this tack is evaluating whether the prospections that the patient has are unrealistic distortions, and then making plans for achieving better futures while disconfirming the distortions. We are acutely aware that the examples above are moves that good therapists already might make, at least implicitly and in passing, and that we have merely reformulated them to display more accurately the future orientation that is at their core. Our overall speculation is that explicitly spending more time on maladaptive prospections directly will yield better therapeutic results. Incentivizing the future. Future-oriented treatments have often been surprisingly effective. Volpp et al. (2009) approached all smokers employed by a large corporation (including many who had no particular intention of quitting) to take part in a study. All were given information about the benefits associated with quitting and about locally 71 available resources to facilitate quitting. By random assignment, some were also offered a program of financial incentives for quitting, which would pay up to $750 for successful abstinence over a year and a half (to be verified by biochemical test). In principle, longterm financial incentives should make no difference to someone in the grip of a compulsive addiction. Yet the incentive tripled the quit rate. Remarkably, over a dozen participants quit for a year and a half so as to earn the full reward and then resumed smoking. . Using incentives to enhance the value of difficult-to-achieve futures, especially if done in ways that are easy to call vividly to mind, is an important complement to providing better information about such futures. Even apart from therapy, incentives are often influential. The theory of addiction as involuntary or compulsive behavior implies that addicts should be essentially indifferent to changes in price of their preferred substance, but the facts contradict that theory. Tobacco smokers cut back when the price of cigarettes is raised, a fact that has remained true over time (Burns & Warner, 2003). In fact, heavy smokers seem more responsive than light smokers to changes in price, contrary to the view that heavy smokers have no voluntary control over smoking. Even heroin users respond to changes in price, and if they are asked to choose between heroin and money, their decisions depend on how much money is involved (Hart & Krauss, 2006). Many influential theories about addiction have been based on being driven by the past. One holds that addicts have a “habit” and are trapped by withdrawal: each attempt to quit smoking or taking drugs causes so much physical distress that the person must continue using to avoid those highly aversive symptoms, even though he would quit if only he could get past withdrawal. That theory reigned in the 1970’s but was gradually 72 discredited, as observations showed that many addicts quit, go through withdrawal, and then start using again, thus seemingly willfully re-entering the ostensible trap from which they escaped. This forced addiction theorists to realize that addicts may smoke and use drugs in pursuit of (what are to them) positive goals (see Ainslie, 2001). Another approach held that the longer a person had the habit of using drugs (especially smoking), the harder it would be to quit. This theory was especially important to the “target-hardening” hypothesis, which was proposed to explain the gradual historic decline in effectiveness of medical treatment programs for smokers: Over the decades, quit rates for those programs declined, leading many to suggest that everyone who could quit easily had quit, leaving only the most addicted smokers, such as those who had been heavy smokers for decades, behind. Yet Coambs, Li, and Kozlowski (1992) found that older, heavy smokers had among the best quit rates. Rather than the strength of the past habit, stopping smoking was apparently due to the emergence of smoking-related health problems, which gave smokers a much clearer anticipation of escalating future health problems. Building meaning and purpose. There is growing evidence that a strong sense of meaning and purpose—which we regard as a paradigm instance of robust futureorientation—is highly protective against psychopathology (Damon, 2010) In one dramatic example, 84 soldiers who committed suicide had all taken the same test of strengths and weaknesses months before; Those soldiers in the very lowest percent of meaning (strongly disagreeing with “my life has meaning”) were at extreme risk for suicide (Cornum et al., 2011). This suggests that building a foundation of meaning and purpose in life should be a major focus of therapy. The lifetime of work by Brian Little 73 (1996) also exemplifies this prospective approach: Little has his subjects identify their core sustainable personal projects, e.g., becoming a physician, making oneself more lovable, helping to solve world hunger, and he argues that identity is centered on such prospection. Much of our own thinking about therapy emerges from work on hope and optimism (Seligman,1990; Snyder, 2002) which then finds its way into positive psychology and prevention (Seligman, 2011). In this endeavor, school-children (Positive Education) and soldiers (Comprehensive Soldier Fitness) are taught a set of skills in a preventive mode. Three sets of skills are taught: “mental toughness,” “identifying and using signature strengths,” and building new “social skills,” such as active-constructive responding to good events. As preventive skills, all instruction focuses on how to use these skills in the future. The use of the past in these preventive modalities is largely to illustrate previous failures which make the need for new skills poignant. From a therapeutic standpoint, perhaps the most problematic aspect of the view that we are driven by the past is that it transformed the discipline of psychology into the modern equivalent of predestination. In our view, this heuristic warped therapeutic theory and practice, by focusing scrutiny on the individual’s past, which now lies beyond her control, and turning attention away from the future and the ways in which it will depend upon the choices she can make. Conclusion The view that behavior is driven by the past, a kind of psychological Laplacianism, has held sway for many years. Originally, this heuristic helped the human sciences escape the idle teleology that the physical sciences cast off with the Galilean 74 Revolution. The result was a long evolution in thought that began with Associationism and culminated in Behaviorism. Behaviorism is now largely behind us, but a “canonical commitment” to a Laplacian style of explanation appears to live on in psychology (Bargh and Ferguson, 2000). Productive as this commitment has been, we believe it has outlived its usefulness, and that research in a wide range of domains within psychology is now pointing to a new heuristic: behavior is not driven by the past, but drawn into the future. After all, a distinction is needed. If an ice cube melts, we can explain this in terms of statistical mechanics. To posit a telos of matter seeking an equilibrium temperature would be mere mystification—no information about any possible future state plays a role in the melting process. But if a friend helps you to move out of your apartment, very likely a positive representation of a yet-to-be-realized state did play an essential role. Explanations involving the active prospection of such states are indeed teleological, but that is because nature contains purposeful or goal-oriented organisms, not because Nature herself has purposes or goals. So there need be nothing illicit or contrary to the natural order about invoking representations of the future to explain the behavior of such organisms in the here and now. On the contrary, as increasing knowledge of the brain reveals, explanations that leave out this teleological element in the guidance of action are empirically as well as theoretically inadequate. Being driven by the past is as unsuitable as a heuristic for living as it is for theorizing in psychology. Daily life is lived on the assumption that how we choose will make a difference to what actually comes about. Hoping, planning, saving for a rainy day, worrying, striving, risking or minimizing risk, even undertaking therapy, all have in 75 common the presupposition that which future will come about is contingent upon our deliberation and action. We have argued that this is no illusion. Prospection—guidance by running and evaluating simulations of possible futures—is not mysterious, and it is at the very core of human action. A heuristic is just that—we are not trying to have the last word on the subject, or adjudicate ultimate metaphysics. But after over one hundred and fifty years of failing to establish that the past drives human action, we suggest that the old, backward-looking heuristic is no longer productive, and that the new, forward-looking heuristic has much brighter prospects. 76 References Abegg, M., Manoach, and D., Barton, J. (2011). Knowing the future: Partial foreknowledge effects on the programming of prosaccades and antisaccades. Vision Research. 51, 215-221. Addis, D. R., Cheng, T., R, P. R. and Schacter, D. L. (in press). Hippocampal contributions to the episodic simulation of specific and general future events. Hippocampus. Addis, D. R., Pan, L., Vu, M. A., Laiser, N. and Schacter, D. L. (2009). Constructive episodic simulation of the future and the past: distinct subsystems of a core brain network mediate imagining and remembering. Neuropsychologia, 47, 2222-38. Adriaanse, M., Oettingen, G., Gollwitzer, P., et al. (2010). When planning is not enough: Fighting unhealthy snacking habits by mental constrasting with implementation intentions. European Journal of Social Psychology, 40, 12771293. Ainslie, G. (2001). Breakdown of will. New York: Cambridge University Press. Anderson, S. W., Bechara, A., Damasio, H., Tranel, D. and Damasio, A. R. (1999): Impairment of social and moral behavior related to early damage in human prefrontal cortex. Nature Neuroscience, 2, 1032-7. Anscombe, G. E. M. (1957). Intention. Oxford: Blackwell. 77 Aristotle (1999). Nicomachean Ethics. Translated. T. Irwin. Indianapolis: Hackett. Baars, B. J. (2002). The conscious access hypothesis: Origins and recent evidence. Trends in Cognitive Science, 6, 47-52. Bargh, J.A. and Ferguson, M.J. (2000). Beyond behaviorism: On the automaticity of higher mental processes. Psychological Bulletin, 126, 925-945. Bartlett, R. (1932). Remembering. Cambridge: Cambridge University Press. Baumeister, R. and Masicampo E. (2010) Conscious thought is for facilitating social and cultural interactions: How mental simulations serve the animal-culture interface. Psychological Review, 117, 945-971. Baumeister, R. F., & Scher, S. J. (1988). Self-defeating behavior patterns among normal individuals: Review and analysis of common self-destructive tendencies. Psychological Bulletin, 104, 3-22. Baumeister, R. F. & Tierney, J. (2011). Willpower. N.Y.: Free Press. Bechara, A., Damasio, H., & Damasio, A. (2000). Emotion, decision making and the orbitofrontal cortex. Cerebral Cortex, 10, 295-307. Bechara, A., Damasio, H., Tranel, D. and Damasio, A. (1997). Deciding advantageously before knowing the advantageous strategy. Science, 275,12931295. Berridge, K. (2004). Motivation concepts in behavioral neuroscience, Physiology & Behavior , 81,179– 209. 78 Bischof, N. (1980). On the philosophy of human morality. In Stent, G. S. (Ed.), Morality as a Biological Phenomenon. Berkeley: University of California Press, 48-66. Boehm-Bawerk, E. V. (1889/ 1970). Capital and Interest. South Holland: Libertarian Press. Buckner, R. L., Andrews-Hanna, J. R. and Schacter, D. L. (2008): The brain's default network: anatomy, function, and relevance to disease. Annals of the New York Academy of Sciences, 1124, 1-38. Buckner, R. and Carroll, D. (2007). Self-projection and the brain. Trends in Cognitive Science, 11, 49-57. Burns, D. and Warner, K. (2003). Smokers who have not quit: Is cessation more difficult and should we change our strategies. In Burns D., (ed.) Those Who Continue to Smoke: Is Abstinence Harder and Do We Need to Change Our Interventions? National Cancer Institute, Smoking and Tobacco Control Monograph No. 15. Bethesda , MD : USDHHS, PHS, NIH, NCI. NIH Publication No. 03-5370 pp. 1131. Busby, J., & Suddendorf, T. (2005). Recalling yesterday and predicting tomorrow. Cognitive Development, 20, 362-372. Carnap, R. (1950). The logical foundations of probability. Chicago: University of Chicago Press. Chalmers, D. (1996). The conscious mind: In search of a fundamental theory. New York: Oxford University Press. 79 Chomsky, N. (1967). A review of B.F. Skinner’s Verbal Behavior. In Leon A. Jakobovits and Murray S. Miron (eds.), Readings in the Psychology of Language, PrenticeHall, 142-143. Christoff, K., A. M. Gordon, et al. (2009). Experience sampling during fMRI reveals default network and executive system contributions to mind wandering. Proceedings of the National Academy of Sciences, 106, 8719-8724. Clark, D. (1986). A cognitive approach to panic. Behaviour Research and Therapy, 24, 461-470. Coambs, R., Li, S., and Kozlowski, L. (1992). Age interacts with heaviness of smoking in predicting success in cessation of smoking. American Journal of Epidemiology, 135, 240-248. Cornum, R., Lester, P., Harms, P. et al. (2011). What we know about completed suicides and the army global assessment tool. Technical report to the US Army. Craig, A. D. (2009). How do you feel—now? The anterior insula and human awareness. Nature Reviews Neuroscience 10, 59-70. Damasio, A. R. (1996). The somatic marker hypothesis and the possible functions of the prefrontal cortex. Philosophical Transactions of the Royal Society London, B: Biological Science, 351, 1413-20. Damon, W. (2008). The path to purpose. N.Y.: Free Press. Davidson, D. (1982). Rational animals. Dialectica, 36, 317-327. 80 Deary, I. J., Weiss, A., and Batty, G. D. (2010). Intelligence, personality and health outcomes. Psychological Science in the Public Interest, 11, 53-79. Decety, J. and Chaminade, T. (2003). When the self represents the other: A new cognitive neuroscience view on psychological identification. Consciousness and Cognition, 12, 577-596. de Greck, M., M. Rotte, et al. (2008). Is our self based on reward? Self-relatedness recruits neural activity in the reward system. Neuroimage, 39, 2066-2075. De Neys, W. (2006). Dual processing in reasoning: Two systems but one reasoner. Psychological Science, 17, 428-433. DeWall, C. N., Baumeister, R. F., & Masicampo, E. J. (2008). Evidence that logical reasoning depends on conscious processing. Consciousness and Cognition, 17, 628-645. Dollard, J., Doob, L., Miller, N., Mower, O., and Sears, R. (1947). Frustration and aggression. New Haven: Yale University Press. Draine, S. C. (1997). Analytic limitations of unconscious language processing. Unpublished doctoral dissertation, University of Washington. Dugatkin, L.A. (2004). Principles of Animal Behavior. New York: W.W. Norton. Elga, A. (2007). Reflection and disagreement. Noûs, 41, 478-502. 81 Fellows, L. K. 2007: The role of orbitofrontal cortex in decision making: a component process account. Annals of the New York Academy of Science, 1121, 421-30. Foa, E., Dancu, C.,Hembree, E., Jaycox, L., Meadows, E., Street, G. (1999). Comparison of exposure therapy, stress inoculation training, and their combination for reducing posttraumatic stress disorder in female assault victims. Journal of Consulting and Clinical Psychology, 67, 194-200. Fox, M. C., Ericsson, K. A., & Best, R. (2011). Do procedures for verbal reporting of thinking have to be reactive? A meta-analysis and recommendations for best reporting methods. Psychological Bulletin, 137, 316-344. Frankfurt, H. (1988). The Importance of What We Care About. Cambridge: Cambridge University Press. Frederick, S., Loewenstein, G., O’Donoghue, T. (2002). Time discounting and time preference: A critical review. Journal of Economic Literature, 40, 351–401. Gollwitzer, P. & Sheeran, P. (2006). Implementation intentions and goal achievement: A meta-analysis of effects and processes. Advances in Experimental Social Psychology, 38, 69-119. Grabenhorst, F. and Rolls, E .T. (2011) Value, pleasure, and choice in the ventral prefrontal cortex. Trends in Cognitive Sciences, 15, 56-67. Grafman, J., Vance, S. C., Weingartner, H., Salazar, A. M. and Amin, D. (1986). The effects of lateralized frontal lesions on mood regulation. Brain, 109, 1127-48. Greenwald, A. G., & Liu, T. J. (1985). Limited unconscious processing of meaning. 82 Bulletin of the Psychonomic Society, 23, 292-313. Greicius, M. D., Krasnow, B., Reiss, A. L. and Menon, V. (2003). Functional connectivity in the resting brain: a network analysis of the default mode hypothesis. Proceedings of the National Academy of Sciences, 100, 253-8. Gusnard, D. A., Akbudak, E., Shulman, G. L. and Raichle, M. E. (2001). Medial prefrontal cortex and self-referential mental activity: relation to a default mode of brain function. Proceedings of the National Academy of Sciences, 98, 4259-64. Haidt, J. (2007). The new synthesis in moral psychology. Science, 316, 998-1002. Hart, C. and Krauss, R. (2008). Human drug addiction is more than faulty decisionmaking. Behavioral and Brain Sciences, 31,448-449. Hassabis, D., Kumaran, D., Vann, S., & Maguire, E. (2007). Patients with hippocampal amnesia cannot imagine new experiences. Proceedings of the National Academy of Sciences, 104, 1726-31. Hassin, et al. (2003). The New Unconscious. Cambridge: MIT Press. Hume, D. (1740). Treatise of human nature. ed. L.A. Selby-Bigge, revised P.H. Nidditch, Oxford: Clarendon Press, 1978. Hume, D. (1740b) An Abstract of a book lately published entitled A Treatise of Human Nature, reprinted in An Inquiry concerning Human Understanding, 1748, ed. C. Hendel, Indianapolis, IL: Library of Liberal Arts, 1955. (Hume's own anonymous review of the Treatise.) 83 Hull, C. L. (1943). Principles of behavior. New York: Appleton-Century. James, W. (1890). The principles of psychology. Cambridge: Harvard University Press. James, W. (1904). A world of pure experience, I and II. Journal of Philosophy, Psychology, and Scientific Methods 1, 533-543, 561-570. James, W. (1907). Pragmatism. New York: Longmans, Green, and Company. Jenkins , V and Fallowfield, L. (2002). Can communication skills training alter physicians? beliefs and behavior in clinics. Journal of Clinical Oncology, 20, 765– 769. Johnson, A., van der Meer, M., and Redish, A. (2007). Integrating hippocampus and striatum in decision-making, Current Opinion in Neurobiology ,17, 692–697. Johnson, A. and Redish, A. (2007). Neural ensembles at CA3 transiently encode paths forward of the animal at a decision point, The Journal of Neuroscience, 27, 12176–12189. Kahneman, D. and Tversky, A. (1979). Prospect Theory: An Analysis of Decision under Risk, Econometrica, 47, 263-291. Kelley, W. M., C. N. Macrae, et al. (2002). Finding the self? An event-related fMRI study. Journal of Cognitive Neuroscience, 14, 785-794. Kendler, K. S., Walters E. E., and Kessler, R. C. (1997). The prediction of length of major depressive episodes: Results from an epidemiological sample of female twins. Psychological Medicine, 27, 107-117. 84 Kircher, T. T., C. Senior, et al. (2000). Towards a functional neuroanatomy of self processing: effects of faces and words. Cognition and Brain Research, 10, 133144. Kringelbach, M. and Berridge, K. (2009). Towards a functional neuroanatomy of pleasure and happiness. Trends in Cognitive Science, 13, 479-87. Langston, R. F., Ainge, J. A., Couey, J.J., Canto, C. B., Bjerknes, T. L., Witter, M. P., Moser, E.I., and Moser, M-B. (2002). Development of the spatial representation system in the rat. Science, 328, 1576-1580. Laplace, Pierre-Simon (1814/1902). A philosophical essay on probabilities. N.Y.: Wiley. Lashley, K. (1929). Brain mechanisms and intelligence. University of Chicago Press, Chicago. Lieberman, M. D., Gaunt, R., Gilbert, D. T., & Trope, Y. (2002). Reflexion and reflection: A social cognitive neuroscience approach to attributional inference. In M. P. Zanna (Ed.), Advances in experimental social psychology (pp. 199-249). San Diego, CA: Academic Press. Little, B. R. (1996). Free traits, personal projects and idio-tapes: Three tiers for personality research. Psychological Inquiry, 8, 340-344. Loewenstein, G., Read, D., & Baumeister, R. F. (Eds.) (2003). Time and decision: Economic and psychological perspectives on intertemporal choice. New York: Russell Sage. 85 Mcclure, S. M., Berns, G. S. and Montague, P. R. 2003: Temporal prediction errors in a passive learning task activate human striatum. Neuron, 38, 339-46. McGuigan, F. J. & Ban, T (1987). Critical issues in psychology, psychiatry, and physiology: a memorial to W. Horsley Gantt. Gordon and Breach Science Publishers: OPA Amsterdam. Mackay, D. G. (1973). Aspects of the theory of comprehension, memory and attention. The Quarterly Journal of Experimental Psychology, 25, 22-40. Marks, I. M. (1987). Fears, Phobias, and Rituals. N.Y.: Oxford. Mason, M. F., Norton, M. I., Van Horn, J. D., Wegner, D. M., Grafton, S. T. and Macrae, C. N. 2007: Wandering minds: the default network and stimulus-independent thought. Science, 315, 393-5. Miller, G., Galanter, E., & Pribam, K. (1960). Plans and the structure of behavior. N.Y.: Holt, Rhinehart, and Winston. Mischel, W. (1974). Processes in delay of gratification. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 7, pp. 249-292). San Diego, CA: Academic Press. Mischel, W., Shoda, Y. & Peake, P. (1988). The nature of adolescent competencies predicted by preschool delay of gratification. Journal of Personality and Social Psychology, 54, 687-696. Mischel, W., Shoda, Y., and Rodriguez, M. L. (1992). Delay of gratification in children”, in G. Lowenstein and J. Elster (eds.), Choice Over Time. New York: Russell Sage Foundation. 86 Montague, P. R., King-Casas, B. and Cohen, J. D. (2006). Imaging valuation models in human choice. Annual Review of Neuroscience, 29, 417-48. Moore, B., Mischel, W., and Zeiss, A. (1976). Comparative effects of the reward stimulus and its cognitive representation in voluntary delay. Journal of Personality and Social Psychology, 34, 419-424. Morsella, E. (2005). The function of phenomenal states: Supramodular interaction theory. Psychological Review, 112, 1000-1021. Neisser, U. (1966). Cognitive psychology. Appleton, Wisconsin.: Appleton-CenturyCrofts. Northoff, G., A. Heinzel, et al. (2006). Self-referential processing in our brain—A metaanalysis of imaging studies on the self. Neuroimage, 31, 450-457. O'Doherty, J., Kringelbach, M. L., Rolls, E. T., Hornak, J. and Andrews, C. (2001). Abstract reward and punishment representations in the human orbitofrontal cortex. Nature Neuroscience, 4, 95-102. Phan, K. L., S. F. Taylor, et al. (2004). Neural correlates of individual ratings of emotional salience: a trial-related fMRI study. Neuroimage, 21, 768-780. Pierce, C. S. (1878/1955). How to make our ideas clear. In J. Buchler, (Ed.) Philosophical writings of Peirce. New York: Dover. Preuschoff, K., Bossaerts, P., & Quartz, S. (2006). Neural Differentiation of Expected Reward and Risk in Human Subcortical Structures. Neuron, 51, 381-90. Quartz, S.R. (2009). Reason, Emotion, and Decision-Making: Risk and Reward Computation with Feeling. Trends in Cognitive Sciences, 13, 209-215. 87 Rae, J. (1834). The Sociological Theory of Capital. London: Macmillan. Raichle, M. E., Macleod, A. M., Snyder, A. Z., Powers, W. J., Gusnard, D. A. and Shulman, G. L. (2001). A default mode of brain function. Proceedings of the National Academy of Sciences, 98, 676-82. Raichle, M. E. and Snyder, A. Z. (2007). A default mode of brain function: a brief history of an evolving idea. Neuroimage, 37, 1083-90; discussion 1097-9. Railton, P. (2002). Kant meets Aristotle where reason meets appetite. C. U. Moulines and Karl-Georg Niebergall (eds.), Argument und Analyse. Paderborn: Mentis. Rescorla, R.A. (1988) Pavlovian conditioning: It's not what you think it is. American Psychologist, 43, 151-160. Rescorla, R. & Solomon, R. (1967) Two-process learning theory: relationships between Pavlovian conditioning and instrumental learning. Psychological Review, 74, 151182. Roberts, W. A. (2002). Are animals stuck in time? Psychological Bulletin, 128, 473-489. Rolls, E. T., Tromans, J. and Stringer, S. M. (2008). Spatial scene representations formed by self-organizing learning in a hippocampal extension of the ventral visual system. European Journal of Neuroscience, 28, 2116-2127. Ruby, P. and Decety, J. (2001). Effect of subjective perspective-taking during simulation of action: a PET investigation of agency. Nature Neuroscience, 4, 546-550. 88 Ruby, P. and J. Decety (2003). What you believe versus what you think they believe: a neuroimaging study of conceptual perspective-taking. European Journal of Neuroscience,17, 2475-2480. Ruby, P. and J. Decety (2004). How would you feel versus how do you think she would feel? A neuroimaging study of perspective-taking with social emotions. Journal of Cognitive Neuroscience, 16, 988-999. Saito, N., Mushiake, H., Sakamoto, K., Itoyama, Y., and Tanji, J. (2005). Representation of immediate and final behavioral goals in the monkey prefrontal cortext during an instructed delay period. Cerebral Cortex, 15, 1535-1546. Schacter, D. L., & Addis, D. R. (2007). The ghosts of past and future: A memory that works by piecing together bits of the past may be better suited to simulating future events than one that is a store of perfect records. Nature, 445, 27. Schacter, D. L., Addis, D. R. and Buckner, R. L. (2008). Episodic simulation of future events: concepts, data, and applications. Annals of the New York Academy of Sciences, 1124, 39-60. Schmitz, T. W., T. N. Kawahara-Baccus, et al. (2004). Metacognitive evaluation, selfrelevance, and the right prefrontal cortex. Neuroimage, 22, 941-947. Schultz, W. (2000). Multiple reward signals in the brain. Nature Reviews Neuroscience, 1, 199-207. Schultz, W. (2002). Getting formal with dopamine and reward. Neuron, 36, 241-263. 89 Schultz, W., Dayan, P. and Montague, P. R. (1997). A neural substrate of prediction and reward. Science, 275, 1593-9. Seligman, M. E. P. (1990). Learned optimism. New York: Knopf. Seligman, M.E.P. (2007). What you can change and what you can't. (2nd ed.) New York: Vintage. Seligman, M. E. P. (2011). Flourish. N.Y.: Free Press. Seligman, M. & Kahana, M. (2009). Unpacking intuition. Perspectives on Psychological Science, 4, 399-402. Seligman, M.E.P. and Johnston, J.C. (1973). A cognitive theory of avoidance learning. In F. J. McGuigan and D. B. Lumsden (Eds.), Contemporary Approaches to Conditioning and Learning. Washington: Winston and Sons, 69-110. Shoda, Y., Mischel, W., & Peake, P. K. (1990). Predicting adolescent cognitive and self regulatory competencies from preschool delay of gratification: Identifying diagnostic conditions. Developmental Psychology, 26, 978-986. Shulman, G. L., Fiez, J. A., Corbetta, M., Buckner, R. L., Miezin, F. M., Raichle, M. E. and Petersen, S. E. 1997: Common Blood Flow Changes across Visual Tasks: II. Decreases in Cerebral Cortex. Journal of Cognitive Neuroscience, 9, 648-663. Singer, T., Critchley, H. D., and Preuschoff, K. (2009). A common role of insula in feelings, empathy and uncertainty. Trends in Cognitive Science 13, 334-340. 90 Smallwood, J., Nind, L., & O’Connor, R. C. (2009). When is your head at? An exploration of the factors associated with the temporal focus of the wandering mind. Consciousness and Cognition, 18, 118-125. Smallwood, J., Schooler, J.W., Turk, D. J., Cunningham, S. J., Burns, P., & Macrae, C. N. (in press). Self-reflection and the temporal focus of the wandering mind. Consciousness and Cognition. Snyder, C. R. (2002). Hope theory: Rainbows in the mind. Psychological Inquiry, 13, 249-275. Spreng, R. N., Mar, R. A. and Kim, A. S. (2009). The Common Neural Basis of Autobiographical Memory, Prospection, Navigation, Theory of Mind and the Default Mode: A Quantitative Meta-analysis. Journal of Cognitive Neuroscience, 21, 489-510. Sripada, C. (2011). Free will and prospective guidance. Stillman, T. F., Baumeister, R. F., & Mele. A. R. (in press). Free will in everyday life: Autobiographical accounts of free and unfree actions. Philosophical Psychology. Suddendorf, T. and Corballis, M. C. (1997). Mental time travel and the evolution of the human mind. Genetic Social and General Psychology Monographs, 123, 133-67. Suddendorf, T. and Corballis, M. C. (2007). The evolution of foresight: What is mental time travel, and is it unique to humans? Behavioral and Brain Sciences, 30, 299313; discussion 313-51. Tomasello, M. (1999). The cultural origins of human cognition. Cambridge, MA: Harvard 91 University Press. Thorndike, E. L. (1911). Animal Intelligence. New York: Macmillan. Tobler, D. N., Dougherty, J. P., Dolan, R. J., Schultz, W. (2006). Reward value coding distinct from risk attitude-related uncertainty coding in human reward systems. Journal of Neurophysiology, 97, 1621-1632. Tolman, E. C. (1948). Cognitive maps in rats and men. Psychological Review, 55, 189208. Tulving, E. 2002: Episodic memory: from mind to brain. Annual Review of Psychology, 53, 1-25. Vohs, K. D., & Baumeister, R. F. (Eds.) (2011). Handbook of self-regulation: Research, Theory, and Applications (2nd Edn). New York: Guilford. Volpp, K., Troxel, A., Pauly, M., et al. (2009). A randomized, controlled trial of financial incentives for smoking cessation. New England Journal of Medicine, 360, 699709. Watson, J.B. (1930). Behaviorism. Chicago: University of Chicago Press.