Inference and Imagination Authors: Ruth M. J. Byrne Published: 2026 ABSTRACT AbstractOur ability to imagine alternatives to reality is central to everyday cognition. Our competence in reasoning depends not only on our capacity to envisage the relevant facts, but also on our proficiency in simulating alternatives to them. I discuss how our skill in making deductive inferences depends on imagining possibilities. To illustrate, I consider evidence on the suppression of inferences, the cognitive processes based on mental models that explain how we make inferences, and corroboration from studies of counterfactual reasoning. I also discuss how our ability to create alternatives to reality depends on identifying mutable aspects of mental simulations. To illustrate, I consider evidence on imagining alternatives to aspects of situations within our control, the cognitive processes based on simulation availability that explain how we create alternatives to reality, and corroboration from studies of moral decision‐making. I conclude with an exploration of the imagination as the engine of the mind. Consider the chilling statement by the President of Ukraine, Volodymyr Zelenskyy, in a CBS 60 Minutes interview in April 2025, “The security of the world is at stake. If we do not stand firm, he [Putin] will advance further… I believe it could escalate into a world war.” 1 The crux of your understanding of the truth of Zelenskyy's assertion lies not just in your thoughts about whatever relevant facts you know, but also in your attempts to imagine various alternative possibilities. Our capacity to understand and reason about the world depends on our ability to grasp the facts of whatever is happening, but also on our skills to model alternatives to current reality, to mentally simulate how things could turn out in different ways in the future, and how they could have turned out in different ways in the past. In this article, one in a series designed to introduce new Fellows of the Cognitive Science Society, I sketch some of the research questions about the relationship between reasoning and imagination that have guided my experiments over the past 40 years. I start with the first clues I had that reasoning depends on imagination. They emerged from observations about how knowledge can suppress inferences, and their explanation led to new predictions on differences in inferences about factual and counterfactual matters. I then turn to the first clues I had that the imagination, in turn, depends on rational principles. They emerged from observations about commonalities in how we create alternatives to reality, and their explanation led to new predictions about how our imagination affects our decisions, including our moral judgments. Throughout, I argue that the imagination is the powerhouse of the mind, and I consider the mechanics of simulation as the engine that drives our thinking, reasoning, and decision‐making. We reason by imagining possibilities I start with a simple illustration of the proposal that our ability to make deductive inferences depends on our skills in imagining alternative possibilities, from experiments on the effects of knowledge and context on reasoning. Next, I outline the cognitive processes underlying its explanation based on the construction of mental models. Then, I sketch an example of its corroboration from experiments on how we reason about counterfactual matters. A clue about how we reason: The suppression of inferences An enduring discovery about human reasoning is that there are some inferences that we all make very readily. From the earliest treatises on reasoning by Aristotle over 2000 years ago, to the first experimental data on reasoning collected in the early 1900s, the idea that we all agree that some inferences are inviolably valid has been considered a gold standard of rationality. For example, suppose you are told that if Mark watered the plants, they grew. And suppose you find out that Mark did indeed water the plants. You will find it trivially easy to infer that the plants grew. This simple modus ponens inference has been considered the bedrock of hypothetical thought, and many experiments have shown that close to 100% of participants make it (e.g., Nickerson, 2015). It is an inference so obvious that it seems inherent to the meaning of “if,” and so compelling that we would find it hard to understand anyone who routinely made the opposite modus schmonens inference, that the plants did not grow (Rips, 1994). Yet, my earliest experiments contradicted this cherished logical and psychological doctrine. My first interest was in how knowledge and beliefs influence reasoning, and my PhD studies, carried out in Dublin in the early 1980s, showed that it is, in fact, very easy to create circumstances in which we will not make a modus ponens inference. For example, suppose you are told that if Mark watered the plants, they grew. And you are also told that if the sun shone, the plants grew. Now, when you find out that Mark watered the plants, will you infer the plants grew? About three‐quarters of participants in experiments reject the modus ponens inference in this context (Byrne, 1989). The inference is suppressed because participants do not know whether the additional background condition holds, whether the sun shone. The demonstration that valid inferences can be suppressed undermined a core tenet of the then‐dominant explanation of the cognitive processes underlying human reasoning. At that time, the prominent view for many decades had been that reasoning depends on a mental logic of abstract inference rules (e.g., Braine & O'Brien, 1998; Rips, 1994). On this account, when you reason, you strip away the content of premises to uncover their logical skeleton, such as “If A then B, A,” and match it to a mental inference rule, such as the one corresponding to modus ponens, “if A then B, A therefore B.” But suppose you are told that if Mark watered the plants, they grew, and this time you hear that the plants grew. You might be tempted to infer that he watered the plants. About 70% of participants in experiments make this affirmation of the consequent inference (e.g., Nickerson, 2015). But the inference is considered a mistake if human reasoning is compared to the normative standard of propositional logic (e.g., Jeffrey 1981). Mental logic theorists resisted the idea that we might have a rule corresponding to it because such rules would undermine rational competence. Instead, they proposed that participants often misunderstood such premises to mean “If A then B and If B then A,” and applied their modus ponens rule to the misinterpreted premise. Errors of reasoning were explained away as errors of interpretation. A key test of this explanation was to give participants information to block such misinterpretations (Rumain et al., 1983). When participants were given premises such as, “if Mark watered the plants, they grew. If it rained the plants grew,” and they heard that the plants grew, they no longer made the affirmation of the consequent inference that Mark watered the plants. Mental logic theorists concluded that since information about alternatives blocked the affirmation of the consequent inference, there could be no rule corresponding to it in the mental logic. It is a persuasive argument, but it applies equally well to modus ponens. By parity of argument, my demonstration that additional background conditions blocked the modus ponens inference indicates there must be no rule corresponding to it in a mental logic either. I proposed instead that the suppression of inferences depends on our interpretation of what we are reasoning about, our knowledge and understanding of the context (Byrne et al. 1999; Espino & Byrne, 2020). The theoretical challenges that inference suppression presents, not only for formal rule theories but for theories based on Bayesian probabilities, have been extensively debated (e.g., Oaksford & Chater 2017; Cariani & Rips 2017; see Knauff & Gazzo Castenada 2023). It poses difficulties too for artificial intelligence algorithms to model the defeasibility or nonmonotonicity of inference (e.g., Dietz et al. 2014; Ragni et al., 2016). Its proper explanation is crucial because its influence is so widespread: it emerges at an early age in children, and it occurs when participants are encouraged to think of enabling conditions themselves (e.g., Cummins et al. 1991; de Chantal & Markovits, 2017); it underlies reasoning about multiply determined causal outcomes, and it arises even for noncausal content (e.g., Kominsky et al. 2015; Juhos et al., 2015). It requires working memory resources, and it is associated with activation of specific brain areas (e.g., De Neys et al. 2005; Bonnefond et al. 2014). Its practical applications range from health decisions to consumer choices (e.g., Chandon & Janiszewski, 2008; Gazzo Castaneda & Knauff, 2016), and its clinical implications range from delusions in schizophrenia to enhanced reasoning in autism (e.g., McKenzie et al., 2011; Pijnacker et al., 2010). Most importantly, the suppression of inferences provides a striking clue to help us understand reasoning: we reason by using our imagination. An explanation for how we reason: Mental models What sorts of reasoning processes does the mind rely on that could explain the suppression of inferences? We understand a conditional such as “If Mark watered the plants, they grew” by constructing mental models, to simulate how the world would be if the assertion were true (Johnson‐Laird & Byrne, 1991). When you understand the conditional, you simulate that it is possible that Mark watered the plants and they grew, but your understanding is not merely a conjunction of these events. Instead, you understand that it is also possible that he did not water the plants and they did not grow, and you may even appreciate that it is also possible that he did not water the plants and they grew anyway, perhaps because it rained (Byrne & Johnson‐Laird, 2009; Khemlani et al., 2018). Initially, you tend to think about just one of these possibilities, that Mark watered the plants and they grew (Johnson‐Laird & Byrne, 2002). A principle of parsimony constrains the construction of mental models so that we tend to envisage as few models as possible, because of the limitations of working memory. But you can “flesh out” your initial models to be more explicit, and your knowledge can modulate the models you construct. You might think only about the possibility that Mark watered the plants and they grew, and the possibility that Mark did not water the plants and they did not grow. Or you might think of these two possibilities, and also the possibility that Mark did not water the plants and they grew anyway, perhaps because you know of an alternative, it rained. Or instead, you might think about the first two possibilities and a third possibility that Mark watered the plants and they did not grow, perhaps because you know of a disabler, the sun did not shine. In fact, we can construct many distinct interpretations of “if” assertions, influenced by meaning and knowledge (Johnson‐Laird & Byrne, 2002). A principle of truth constrains the construction of models so that we tend to think only about the way the world would be if the assertion were true, not about the possibilities ruled out by the assertion. The models you construct affect the inferences you are willing to make (Johnson‐Laird et al., 2023). Say you have constructed models corresponding to the possibility that Mark watered the plants and they grew, the possibility that he did not water them and they did not grow, and the possibility that he did not water them but they grew anyway, perhaps because it rained. You will readily make the modus ponens inference that the plants grew when you hear that Mark watered them, because the information is consistent with the first model, in which the plants grew, and your other models can be eliminated. You will not make the affirmation of the consequent inference that he watered the plants when you hear that the plants grew, because the information is consistent with two of your models, in one Mark watered the plants, in the other, he did not, it rained. Suppose instead you have constructed models corresponding to the possibility that Mark watered the plants and they grew, the possibility that he did not water them and they did not grow, and this time, the possibility that he watered them but they did not grow, perhaps because it was not sunny enough. You will make the opposite pattern of inferences to the previous case. Now, when you hear that Mark watered the plants, you will not make the modus ponens inference, that the plants grew, because the information is consistent with two of your models, in one, the plants grew, but in the other, they did not, because it was not sunny. And when you hear that the plants grew, this time you will endorse the affirmation of the consequent inference that Mark watered them, because the information is consistent with the first model only, in which Mark watered them, and your other models can be eliminated. During my postdoctoral years in Cambridge in the late 1980s, I had the opportunity to develop these ideas in collaboration with Phil Johnson‐Laird. For each of the main domains of deduction, not only conditional inferences but also Aristotelian syllogisms, spatial inferences, and meta‐deductions about truth, we showed that people reason by relying on cognitive processes that construct, combine, and describe models (Johnson‐Laird & Byrne, 1991). We tested the theory in many experiments and wrote a computer program in LISP to simulate it. The model theory situates the role of imagination at the heart of the process of reasoning: we simulate possibilities to understand premises and, in so doing, we discover counterexamples to putative conclusions. I turn now to an example that is particularly illuminating with regard to the role of imagination in reasoning. An example of corroborating evidence: How we make counterfactual inferences We reason to understand the world and to communicate our beliefs about it to others. Most experiments on human reasoning examine how we can interrogate a set of assumed facts about the world, and extract a further, hitherto nonobvious fact from them. It is a crucially important skill that provides the basis on which we advance, in individual learning, in personal engagement with others, in our collective social and political organization, and in science and mathematics. Establishing facts based on other facts is the solid ground of progress. Yet, our studies on counterfactual inference showed that inferences can be made very readily about imagined alternatives to facts, even more so than about facts. Consider a counterfactual conditional about how things could have been different, “if Alicia had been in Dublin, Laura would have been in Cork” (Byrne & Tasso, 1999). It can seem to mean something very different from a factual conditional, "If Alex was in Galway, Katie was in Donegal" (Stalnaker, 1968; Lewis, 1973; Fillenbaum, 1974). When you understand the counterfactual, you mentally simulate the conjecture, you imagine that Alicia is in Dublin and Laura is in Cork. But you also spontaneously try to recover what you think must be the corresponding facts. You think about the situation in which, in fact, Alicia is not in Dublin, and Laura is not in Cork. Your understanding of the imagined conjecture is anchored in your presumptions about the corresponding reality. Its interpretation requires you to envisage two different possibilities, Alicia is in Dublin and Laura is in Cork, and Alicia is not in Dublin and Laura is not in Cork, and to keep track of their epistemic status, that one corresponds to the facts of the matter, whereas the other corresponds to an imagined situation that once was possible but is so no longer (Espino & Byrne, 2021). When we imagine how things could have been different, we seem to be able to think readily about multiple possibilities. Such effortless entertainment of multiple possibilities does not occur when we reason about facts. Suppose you are told instead, “if Alex was in Galway then Katie was in Donegal.” You think about the facts, Alex is in Galway and Katie is in Donegal. You are aware there are other possibilities, but at the outset, most of us tend to think explicitly about just this single possibility, because of our working memory limitations. And this partial initial representation can have drawbacks. Suppose you are told Katie was not in Donegal, and you are asked to infer, what follows? About half of the participants in experiments say Alex was not in Galway, but many say that nothing can be inferred (e.g., Nickerson, 2015). The modus tollens inference requires you to think about possibilities that are consistent with the conditional other than that Alex is in Galway and Katie is in Donegal, such as that Alex is not in Galway and Katie is not in Donegal. But you may not have immediately thought through such possibilities at the outset when you understood the conditional. As a result, the modus tollens inference from a factual conditional presents us with challenges. In contrast, the inference is very easy for the counterfactual conditional about an imagined alternative to reality. Consider again that you are told if Alicia had been in Dublin, Laura would have been in Cork. Now, suppose you are also told that Laura was not in Cork, and you are asked to infer, what follows? Most participants in experiments readily say that Alicia was not in Dublin (Byrne & Tasso, 1999). The modus tollens inference requires you to have thought about other possibilities that are consistent with the conditional, such as that Alicia was not in Dublin and Laura was not in Cork. And in fact, you will have immediately thought about such possibilities at the outset when you understood the counterfactual, because you will have grounded your understanding of the imagined conjecture in your recovery of the presupposed reality. The inference seems easy because you have already thought about the possibility that Alicia is not in Dublin and Laura is not in Cork in your mental representation of the presupposed facts (Byrne & Tasso, 1999). When I returned to Dublin in the early 1990s, I worked extensively with many collaborators to find out more about how people make inferences from counterfactual conditionals. We discovered that people make more inferences from counterfactual conditionals for many different sorts of content, including causal, definitional, and deontic counterfactuals (Quelhas & Byrne, 2003; Thompson & Byrne, 2002). They even make immediate inferences for counterfactual inducements, such as “if you had put that box on the shelf I would have grounded you,” from which they infer directly that the hearer did not put the box on the shelf (Egan & Byrne, 2012). And the frequency of some inferences is also elevated for “semi‐factual” conditionals, such as “even if the button had been pressed the machine would have started,” which describe an antecedent event that is contrary‐to‐fact: in reality, the button was not pressed, but with an outcome that corresponds to the facts: in reality, the machine started (Moreno‐Rios et al., 2008; see also Egan et al., 2009). Overall, the evidence shows that we overcome our difficulties in entertaining multiple models when we understand assertions that invite us to imagine how things could have been different. Strikingly, our understanding of multiple possibilities for counterfactuals occurs very rapidly (Frosch & Byrne, 2012; Santamaria et al., 2005). Suppose you hear that Michael goes to a flower shop with his sister. She says to him that if there had been roses in the shop, there would have been lilies. In some of our experiments, participants listened to such descriptions as they observed images on a computer screen, such as an image of roses and lilies, an image of carnations and poppies, an image of roses and lilies with a cross through it, and an image of carnations and poppies with a cross through it. Cameras tracked where their eyes looked as they listened to the brief descriptions. At the outset, most participants looked at the image of the roses and lilies. But after they heard the first part of the counterfactual, “if there had been roses…” within just a few milliseconds, their eyes moved to the image of the roses and lilies with a cross through it. As they heard the rest of the counterfactual, “…there would have been lilies,” they continued to look as often at this image corresponding to no roses and no lilies, as they did at the image of roses and lilies, alternating their gaze between the two images (Orenes et al., 2019; 2021). In contrast, when the description was just about the facts, such as “if there are roses there are lilies,” they tended to look only at the image of roses and lilies. The results of such experiments show that swiftly, within milliseconds, we think not only about the imagined conjecture for a counterfactual, but also about the presumed facts. A counterfactual primes us to recover the facts very rapidly indeed, to anchor our imagination of its conjecture (de Vega & Urrutia, 2011; Ferguson & Jayes, 2018; Kulakova & Nieuwland, 2016). Even when we imagine impossibilities, such as, “if people were made of steel, they would not bruise easily,” we try to imagine them just as we do real possibilities, by constructing a consistent simulation of the impossible conjecture, informed by our knowledge. As a result, we consider that some impossibilities could be true, but others could not be (Byrne, 2024). These few examples illustrate how reasoning depends on imagining possibilities. From early clues about the effects of our knowledge on suppressing inferences, to later discoveries on how we reason about counterfactual matters, the evidence indicates that we simulate models of the world to enable us to reason. My exploration of how we reason about alternatives to reality led me to become curious about a related question, how do we create such alternatives to reality in the first place? We create alternatives to reality by simulating available possibilities I start with a simple illustration of the proposal that our ability to create alternatives to reality depends on identifying mutable aspects of our mental simulations. I describe our experiments that show we often imagine how things could have been different by modifying things within our control. Next, I outline the cognitive processes underlying its explanation based on the availability of alternatives. Then, I sketch an example of its corroboration from experiments on how we make moral decisions. A clue about how we create alternatives to reality: The case of controllability We create alternatives to reality often in our daily lives, when we think “if only” about an episode in the past and imagine how it could have turned out better or worse, or when we think “what if” about an episode in the future and imagine how it could turn out in different ways (Kahneman & Tversky, 1982; see Byrne, 2016 for a review). Suppose you get caught in the rain, you may wish you were not getting wet and imagine, “if only I'd worn a raincoat!” Such thoughts may lead us to feel regret, or disappointment, or a range of other emotions (e.g., Morrison & Roese, 2011; Sweeny & Vohs, 2012). They help us to identify causal relations, for example, you got wet because you did not wear a raincoat (e.g., Frosch & Byrne, 2012; Gerstenberg, 2022; Lagnado et al., 2013). And they may even help us to prepare for the future, by learning from our mistakes, “next time I'll wear a raincoat!” (e.g., Markman et al., 2008a; Roese & Epstude, 2017). With many collaborators during the early 2000s, I carried out experiments to test not only how we reason about counterfactuals, but also how we create them. Remarkably, we all tend to create the same sorts of alternatives to reality. We tend to imagine how things could have turned out differently by imagining a change to an unusual event, for example, if you are in an accident when driving home by an unusual route, you may tend to imagine “if only I'd driven home by my usual route!” (e.g., Dixon & Byrne, 2011; Kahneman & Tversky, 1982). We tend to imagine changes to our actions, something we did more than something we failed to do, for example, when you lose money because you switched shares to a different company, you may tend to imagine “if only I hadn't switched!” (e.g., Byrne & McEleney, 2000; Henne et al., 2019; Kahneman & Tversky, 1982; see also Gilovich & Medvec, 1995). We tend to imagine changes to the first cause in a causal sequence, for example, a traffic jam that led to subsequent delays on our way home (e.g., Segura et al. 2002; Wells et al., 1987), but we tend to imagine changes to the more recent event in a temporal noncausal sequence, for example, the last penalty shoot‐out in a football match (e.g., Byrne et al., 2000; Miller & Gunasegaram, 1990; Walsh & Byrne, 2004). These commonalities reflect our detection of a turning point in an episode, the identification of a moment when events could have taken an alternative course. Our imagination can even appear constrained by “fault‐lines” in our representation of reality, in that the changes we imagine often seem restricted to a limited set of pivotal junctures (Kahneman & Tversky, 1982). Yet, our studies showed that each of these fault‐lines—the tendencies to imagine changes to exceptions, or to actions, or to first causes, or to more recent events—can be eliminated or even reversed (Byrne, 2005). Consider, as an illustration, one of these fault‐line tendencies and its reversal. We often imagine how things could have turned out differently by envisaging a change to an event within our own control. For example, participants wrote diary entries after they read a fictional account in which they had moved to a new town and had not made any friends. They spontaneously imagined how the outcome could have been different by envisaging changes to decisions within their control, “if I had chosen to go to other functions, I could have made more friends,” and they did so twice as often compared to their causal thoughts (McEleney & Byrne, 2006; see also Davis et al., 1995). We tend to imagine how things could have turned out differently by envisaging a change to an event within our own control, even when there are many other sorts of events we could change instead. For example, suppose you are delayed on your way home, and your partner becomes ill while you are not there. You were delayed by a series of events, a traffic jam, a fallen tree in the road, and your decision to call into a bar for a beer. You may tend to imagine you could have made it home on time to help your loved one if only you had not called into the bar for a beer. You imagine a change to the event within your control, even though there are many other events you could change (Girotto et al., 1991; Markman et al. 1995). But our imagination is flexible. The tendency to imagine changes to events within our control can be eliminated by our knowledge. Suppose the series of events that delayed you included a traffic jam, a fallen tree in the road, and this time, your decision to call in to check on your elderly parents. This time, you will not tend to imagine you could have made it home in time if you had not called into your parents. The tendency to change the event within your control is eliminated, and you will tend to imagine changes to events outside your control instead. In this case, the controllable event is not mentally undone because it is constrained by our knowledge of the obligation to care for our parents (McCloy & Byrne, 2000; see also Walsh & Byrne, 2007). When we imagine how things could have been different by changing an aspect of a situation within our own control, we are able to identify causes of an outcome that we can change in the future (Roese & Epstude, 2017; Ferrante et al., 2013; Mercier et al., 2017). In contrast, when we imagine how things could have been different by changing an aspect of a situation outside our own control, we can identify ways to excuse our past failures, justify our decisions, or even abrogate responsibility (e.g., Brockbank et al., 2024; Catellani & Covelli, 2013; Markman et al., 2008b; McCrea, 2008). Another way that the tendency to change an event within our control is eliminated is when we adopt a different perspective on the situation. Suppose you read about Anna who could win some chocolates if she does a multiplication sum in her head in 30 s. The sum will either be an easy one with two one‐digit numbers (e.g., 7×3) or a difficult one with two two‐digit numbers (e.g., 67×83). The sums are in two sealed envelopes, marked A and B. Anna decides to try it and chooses envelope A, and unfortunately, it contains the two‐digit sum. She tries to do the sum but fails and does not win the chocolates. When participants read about the situation and imagine how things could have turned out differently for Anna, they tend to say, “if only she had chosen the other envelope,” that is, they exhibit the common tendency to change the decision within her control (Girotto et al., 2007). But now consider instead that you yourself are placed in this situation. You are given the opportunity to win some chocolates if you do the multiplication sum. You choose one of the envelopes, and it turns out to contain the difficult sum. You try to solve it and you fail. When participants experience the situation themselves and then imagine how things could have turned out differently for them, they tend to say, “if only I had had more time,” “if only I had had pen and paper” or “if only I had been able to concentrate better” that is, they change an aspect of the situation outside their control (Girotto et al., 2007). People do so even when they observe someone else experiencing the situation (Pighin et al., 2011; 2021). The different perspectives of reading about an event, or taking part in the event yourself, or observing someone taking part, lead to differences in how you imagine that things could have turned out differently. Our studies demonstrated the extraordinary flexibility of the counterfactual imagination. This flexibility occurs not only in the tendency to imagine changes to events within our control, which can be switched to a tendency to imagine changes to events outside our control (McCloy & Byrne, 2002; Pighin et al., 2011). It occurs for the other commonalities in our imagination too. For example, the tendency to imagine changes to exceptions can be switched to a tendency to imagine changes to normal events (Dixon & Byrne, 2011); the tendency to imagine changes to actions can be switched to inactions (McEleney & Byrne, 2006; Gilovich & Medvec, 1995), the tendency to imagine changes to recent events can be switched to earlier events (Walsh & Byrne, 2004; see also Segura et al., 2002). It is a striking discovery that when we imagine how things could have been different, we all tend to imagine changes to the same aspects of a situation (Kahneman & Tversky, 1982). But it is equally informative that such tendencies can be so readily eliminated. It provides an important clue about how we create alternatives to reality. An explanation for how we create alternatives to reality The evidence from our experiments on reasoning, and on the creation of alternatives to reality, suggested to me that just as our ability to reason depends on imagination, so too our imagination depends on rational processes (Byrne, 2005). Our ability to simulate an alternative to reality depends on a core set of processes to identify what aspects to change in the mental representation of reality. To imagine an alternative possibility requires you to construct a new model that contains some changes to your existing model of a situation. Our counterfactual imagination is guided by a set of rational principles that underlie the simulation of multiple possibilities. Suppose you hear that Alicia cycled to the beach on a cloudy day and went for a walk in the dunes. You can imagine how things could have turned out differently by modifying one or more aspects of your model of Alicia cycling to the beach and going for a walk. You may mentally delete an aspect, say, you might imagine Alicia cycling to the beach and cycling immediately back home again. Your imagined possibility has changed one aspect of the original representation, deleting Alicia going for a walk, but it has left other aspects the same, cycling to the beach, the cloudy day, and so on. Or you may mentally add a new aspect, say, you might imagine Alicia cycling to the beach and going for a swim instead of a walk. Your imagined possibility has changed an aspect of the original representation, deleting Alicia going for a walk, and it has also added something new to replace it instead, Alicia going for a swim; and other aspects remain the same, cycling to the beach, the cloudy day, and so on (e.g., Markman et al., 2007; Roese & Epstude, 2017). What you change and how you change it will affect how similar your imagined alternative seems to reality, how plausible it is, and how likely it is (e.g., De Brigard et al., 2013; Petrocelli et al., 2011; Stanley et al., 2017; see also Lewis, 1973). When children begin to create alternatives to reality, they often just delete an aspect of their mental representation of a situation, but as they get older, they tend to replace the deleted aspect, substituting something new (e.g., Begeer et al., 2009; Rafetseder et al., 2013; see also Feeney et al., 2017). We also tend to imagine changes to outcomes that are bad more often than ones that are good, so if you know that Alicia did not enjoy the walk or something unpleasant happened during the walk, you will be inclined to imagine alternatives to that aspect. In contrast, if Alicia had a wonderful walk, you might change some other aspect, such as the cycle to the beach (Roese & Epstude, 2017). And we tend to change bad outcomes to be better rather than worse, so you will be more inclined to imagine that Alicia went for a swim if you know Alicia enjoys swimming, rather than imagining Alicia had a picnic if you know Alicia dislikes picnics (De Brigard et al., 2013; Rim & Summerville, 2014). Our knowledge modulates the imagined possibilities we construct, guiding us to identify aspects of a situation to delete, and providing potential replacements by our retrieval of information from memory. A salient aspect may be identified as a candidate for modification by its exceptionality, perhaps you know Alicia never usually walks in the dunes. We apply our knowledge about norms, such as intrapersonal norms about what Alicia generally likes to do, or moral norms about what we are all obliged to do, say calling to check on elderly parents. An exceptional aspect recruits from memory its normal counterpart, and so it is readily identified as mutable (Kahneman & Miller, 1986). The “fault‐lines” in the representation of reality that we tend to identify as candidates for mutability may thus reflect our knowledge of alternatives and their ready availability in memory, as moderated by norms or default possibilities (e.g., Feldman et al., 2020; Kahneman & Miller, 1986; Phillips et al., 2019; Tepe & Byrne, 2022). It is unsurprising, then, that the “fault‐lines” are flexible and can be eliminated or reversed. For example, in the story about Anna trying to solve a sum, we simulate a set of possibilities that includes the factual events, the choice between envelopes, the rules about the time allowed, and about not using pen‐and‐paper, which all lead to the outcome, she did not solve the sum. Various counterfactual alternatives may come to mind in which the outcome is different, the sum is solved, including the possibility in which the other envelope was chosen with a simpler sum, the possibility in which there was more time allowed, the possibility in which pen and paper was allowed, and so on. Our knowledge may be constrained by defaults, for example, the aspect chosen from a potential set of modifiable possibilities may be an enabling background condition that is necessary for the outcome to have occurred (Deighan & Byrne, 2025; Frosch & Byrne, 2012). To construct an alternative to reality requires enough working memory resources not only to envisage multiple possibilities, but also to keep track of their epistemic status, one model corresponding to an imagined alternative, and the other corresponding to the facts (Espino & Byrne, 2021). It also requires sufficient representational flexibility to switch attention from one representation to another, so that you can compare multiple possibilities. Children develop such executive function skills throughout middle childhood, and their imagined alternatives to reality, and ascriptions of emotions such as regret, begin to more closely resemble those of adults (e.g., McCormack et al., 2020; Meehan & Byrne, 2005; Rasga et al., 2016; 2017). You need to be able to make changes to the model of the imagined possibility while at the same time maintaining the integrity of the original model corresponding to the facts, as you update each one (e.g., Beck et al., 2006; 2009). Yet, our choice of which aspect to mentally modify when we create alternative circumstances is remarkably rapid (Tepe & Byrne, 2022). I turn now to an example that further illuminates the role of rational principles in imagining alternatives to reality. An example of corroborating evidence: How we make moral judgments The counterfactual imagination has widespread effects on many of our judgments and decisions in daily life, ranging from causal judgments to moral decisions. For example, the idea that we understand that A caused B, by thinking that if A had not happened then B would not have happened, has a long history (e.g., Hume, 1739/1978; Mill, 1843/1967; see also Lewis, 1973; Mackie, 1974). There are close links between the imagined alternatives we create and our causal ascriptions (e.g., Einhorn & Hogarth, 1986; Hart & Honore, 1959; McGill & Klein, 1993; see also Pearl, 2013). For example, when you can imagine how an outcome could have turned out differently if only some event had not happened, you will tend to judge that the event caused the outcome. Consider a runner who sustains a minor ankle injury in training and then loses her race by milliseconds when she experiences side effects of fatigue after taking a legal painkiller. Suppose there was another painkiller available that did not cause side effects. Do you think the chosen painkiller's side effects caused the runner to lose the race? Participants tended to judge that the painkiller's side effects caused the runner to lose her race because they can readily imagine a counterfactual alternative, “if only she had taken the other painkiller, she would have won the race” (McCloy & Byrne, 2002; see also N'gbala & Branscombe, 1995). Suppose instead there was another painkiller available and it had the same side effects. Now, do you think the chosen painkiller's side effects caused the runner to lose the race? Participants tended to judge that the painkiller's side effects did not cause the runner to lose the race as much in this situation, because they can readily imagine a semi‐factual alternative, “even if she had taken the other painkiller she still would not have won the race!” that is, they can imagine the outcome turning out just as it did even though some alternative action was taken beforehand (McCloy & Byrne, 2002). When we try to work out the causes of events, we are influenced by the various alternative possibilities that come to mind (e.g., Gerstenberg et al., 2017; Icard et al., 2017; Kominsky et al., 2015; Quillien & Lucas, 2023). Relatedly, our imagination plays a powerful role in our moral decisions. With several collaborators during the 2010s, we explored the moral imagination. Our studies demonstrated how moral judgments of fault and accountability, punishment and blame, are influenced by factors such as our intuitions and emotions (Gubbins & Byrne, 2014; see Greene et al., 2001; Haidt, 2001), our inferences about the intentions of an actor (Parkinson & Byrne, 2017; see Gray & Schein, 2012; Young & Saxe, 2011), and even by how tired we are (Timmons & Byrne, 2019; see Danzigera et al., 2011). Importantly, our moral judgments are affected by whether we can imagine how things could have turned out differently. Suppose you hear that the outcome of an attack would not have been as bad if the victim had acted differently. You may be inclined to blame the victim, more so than when you hear that the outcome would have been as bad even if the victim had acted differently (e.g., Branscombe et al., 1996; see also Alicke et al., 2008). Just as our judgments of causal responsibility are affected by whether we can imagine how things could have turned out differently, so too our judgments of moral responsibility are affected by whether we can imagine how things could have, and should have, turned out differently (Malle et al., 2014). The power of the moral imagination is illustrated by the effects of imagined possibilities on how we assess outcomes. Outcomes matter to us. Consider two boys throwing bricks from an overpass bridge, trying to hit cars passing underneath. We tend to blame the boy whose brick landed on a car and injured someone far more than the boy whose brick landed on the pavement without injuring anyone. We do so even though they both had the same intentions and carried out the same actions. And we do so even though the second boy's failure to cause a bad outcome occurred because of circumstances outside his control rather than because of any decision on his part (Parkinson & Byrne, 2017; see also Lench et al., 2015). Likewise, a severe outcome such as murder receives a much harsher legal sentence compared to an attempt to murder (e.g., Hart & Honore, 1959). Bad outcomes may have diagnostic value (Uttich & Lombrozo, 2010). The influence of outcomes on our thoughts is so widespread that we even tend to believe with hindsight, once we know an outcome, that we would have predicted it all along (Fischhoff, 1975). So too, our judgments of blame reflect a moral hindsight bias (Byrne & Timmons, 2018; see also Fleischhut et al., 2017; Oeberst & Goeckenjan, 2016). We infer, with hindsight, that the boy throwing bricks who injured a driver could have known that his bad action would harm someone (e.g., Baron & Hershey, 1988; Martin & Cushman, 2016). We construct a model that explicitly links the action to its intended and expected outcome (Byrne & Timmons, 2018; Cushman, 2013). And we infer, with hindsight, that the lucky boy whose brick did not harm anyone would have known that his bad action would fail. Our moral hindsight leads us to allocate more blame to the boy whose brick injures a driver than to the one whose brick fails to injure a driver. But our studies show that the imagination is an effective antidote to this moral hindsight effect. When you imagine an alternative to an outcome, it affects your judgments about the morally bad action that led to it. A counterfactual, such as, “if the brick had hit a car, the driver would have been injured,” emphasizes that the brick almost hit the car and a bad outcome nearly occurred. When you imagine how the outcome could have been worse, you appreciate that it was a close call that the outcome turned out as it did, and that it very nearly turned out badly, the brick almost hit a car. As a result, condemnation of the “lucky” boy whose brick failed to harm anyone increases when participants imagine how things would have turned out worse, such as if he had thrown the brick from another section of the bridge (Parkinson & Byrne, 2017). The counterfactual leads participants to judge the boy who did not harm anyone just as harshly as the one who injured a driver, increasing their judgments of blame, punishment, and moral responsibility. And this antidote to the moral hindsight bias occurs even for judgments about morally good outcomes. We generally tend to appreciate moral goodness and praise self‐sacrificial acts (e.g., Algoe & Haidt, 2009; see also Martin & Cushman, 2016). Suppose you hear about Ann, who sees a young child, Jill, trip and fall near a busy intersection as a large truck approaches. Ann can see that the truck driver will not be able to stop in time, and she decides to run into the road to hold Jill down as the truck passes over them. Participants showed a moral hindsight effect in that they judged that Ann should do so, before they knew the outcome; and when they heard the good outcome that Jill sustained only minor bruises, they judged even more so that Ann should have acted as she did. Morally good actions tend to be considered worth doing even if they do not succeed, and so when they heard the bad outcome that Jill sustained serious injuries, they still judged that Ann should have acted as she did (Byrne & Timmons, 2018). But suppose you heard that the police examined videos of the incident and observed that if Ann had not rushed to help, Jill's injuries would have been worse, do you think Ann should have acted as she did? Participants’ judgments that Ann should have acted as she did, increased even more. Now, suppose you hear the police observed that even if Ann had not rushed to help, Jill's injuries would have been the same, do you think Ann should have acted as she did? Participants’ judgments that Ann should have acted as she did, decreased in this case (Byrne & Timmons, 2018; Timmons et al., 2019). When we imagine how things could have been different, our moral judgments change, not only about attempts to harm, but also about attempts to help. The moral imagination exerts tremendous influence over our judgments of good and bad. The consequences of the influence of the moral imagination have timely relevance for many current challenges, from how morally unacceptable actions may become licensed, to how misinformation may be confronted (Bertolotti & Catellani, 2023; Effron, 2018; Tepe & Byrne, 2022). These few examples illustrate how we create alternatives to reality by identifying mutable aspects in our mental simulations. From early clues about how we imagine that things could have turned out differently by imagining changes to aspects of a situation within our control in some circumstances, and outside our control in others; to later discoveries about how our imagined alternatives affect our moral judgments, the evidence indicates that we simulate alternatives to reality guided by our knowledge of possibilities. Concluding remarks My research on reasoning and imagination has led me to the conclusion that the imagination, far from being an optional extra in our set of cognitive abilities, is in fact the powerhouse of the mind, the engine that drives our thinking, reasoning, and decision‐making. There are many reasons why it is important to find out more about how we imagine how things could have been different, and one final example from our recent research illustrates its timely practical relevance. Decisions made about us by artificial intelligence (AI) decision support systems are becoming more common in our everyday lives, from decisions about our health treatments and legal entitlements, to decisions about our finances, education, and commerce. Many of us may have some concerns about the fairness and transparency of decisions made by AI systems. And our willingness to accept AI decisions can be influenced by different factors, including the provision of an explanation (e.g., Dodge et al., 2019; Hoffman et al., 2018; van der Waa et al., 2021). Increasingly, AI decision support systems are programmed to provide us with explanations that invite us to imagine how things could have been different. Suppose your bank manager, assisted by an automated system, refuses your loan application; the AI system may provide a counterfactual explanation, such as “if you had asked for $2k less, you would have received the loan.” Counterfactual explanations describe how an AI system's output decision would have been different, if some of the input features had been different. Several hundred distinct algorithms in eXplainable AI now generate automated counterfactual explanations for AI decisions (Karimi et al., 2022; Keane et al., 2021). This remarkable proliferation drives home the need to find out more about how we understand explanations for AI decisions based on imagining how things could have turned out differently (Byrne, 2019; 2022; Miller, 2019). In experiments over the past few years carried out with Mark Keane and our collaborators, we have found that participants often self‐report that they find such counterfactual explanations to be very helpful. And counterfactual explanations can help people to predict an AI system's next decisions. Yet, our studies show that other sorts of explanations can help people just as much (Celar & Byrne, 2023; see also Lim et al., 2009; Lucic et al., 2020; Lage et al., 2019). The potential dissociation between our subjective appreciation of counterfactual explanations of AI decisions and their objective efficacy in helping us understand AI decisions may raise some ethical questions. We may be somewhat led astray if we feel subjectively that we are being helped by an AI explanation based on imagining how things could have been better, yet objectively the explanation is not particularly better than others in helping us to understand the AI decision (Warren et al., 2024; see also Dai et al., 2022). The example illustrates that how we imagine alternatives to reality has considerable relevance for our daily lives. Our skill in making inferences depends on imagining possibilities, and our ability to create such alternatives to reality depends on cognitive processes that identify mutable aspects of our mental simulations. Our capacity to imagine how things could have been different in the past or could be different in the future is central to everyday cognition and has widespread implications in our daily lives.