Byrne's (Reference Byrne2005) book, despite its clarity of analysis and clarity of exposition, advances a claim that is far from clear: Imagination is rational. This claim is unclear because the word rational typically denotes conformity to some normative standard, yet, as Byrne herself notes, “There is no normative standard against which to judge whether an imaginative thought is best” (p. 209). Even if one adopts Byrne's view of rationality as the capacity to draw normatively valid conclusions (as opposed to the disposition to do so), one is still left with the problem of deciding what constitutes such a capacity in the domain of counterfactuals.
Byrne acknowledges this problem, but rather than confront it directly by explaining why the content of counterfactual inferences should be considered rational, she circumvents it by explaining why the process of counterfactual reasoning should be considered rational. Her argument proceeds as follows: (1) the process by which most individuals make deductive inferences (i.e., searching for counterexamples to an argument's conclusion among a set of possibilities consistent with the argument's premises) is capable of yielding normatively valid conclusions; (2) counterfactual reasoning shares many similarities with deductive reasoning; (3) therefore, if deductive reasoning is considered rational, then counterfactual reasoning should be considered rational as well.
The problem with this argument is that, without any independent measure of what constitutes a valid counterfactual inference, we have no reason to believe that the analogy between deductive reasoning and counterfactual reasoning is, itself, valid. After all, the literature on strategy development (e.g., Siegler Reference Siegler1996) has documented many instances in which failures of reasoning are attributable to the misapplication of domain-specific strategies. For instance, when children are first introduced to decimal notation, they often compare decimals on the basis of digit length rather than digit location, judging a decimal like .125 to be larger than a decimal like .25 because the former contains more digits than the latter (Moss & Case Reference Moss and Case1999; Smith et al. Reference Smith, Solomon and Carey2005). Although this strategy is reliably correct when applied to integers, it is not reliably correct when applied to decimals.
Is the application of deductive-reasoning strategies to counterfactual-reasoning problems as inappropriate as the application of integer-comparison strategies to decimal-comparison problems? Perhaps not, but Byrne provides no reason for us to believe otherwise. By focusing on processing similarities between deductive reasoning and counterfactual reasoning, Byrne overlooks potential dissimilarities in their application. One such dissimilarity is the nature of the space of possibilities over which each type of inference is drawn. That is, when reasoning about factual conditionals of the form “if A, then B,” individuals are limited to a small, well-defined space of possibilities (i.e., A and B, A and not-B, not-A and B, not-A and not-B), but when reasoning about counterfactual conditionals, they are confronted with the space of all possible worlds (Lewis Reference Lewis1973; Stalnaker Reference Stalnaker2003). Thus, the absence of a counterexample specifies a normatively valid conclusion in the former space of possibilities but not the latter. Indeed, to conclude that reality is immutable because no changes to reality are conceivable is, in Dennett's words (Reference Dennett1993), to “mistake a failure of imagination for an insight into necessity” (p. 48).
Consistent with this idea, most adults recognize, at least implicitly, that failures of imagination do not count as evidence of necessity (Shtulman & Carey Reference Shtulman and Carey2007). That is, when asked to judge the possibility of events that violate physical laws, like walking through a wall or walking on water, most adults not only deny the possibility of such events, but also justify their judgments with positive evidence of the events' impossibility (e.g., “both walls and people are solid,” “water doesn't have enough surface tension”). In other words, rather than appeal to the perceived absence of a counterexample (e.g., “there's no way a person could walk on water”), adults tend to provide principled reasons for why no such counterexamples exist.
In contrast to adults, preschool-aged children do not tend to provide principled reasons for their judgments. Instead, they appeal to their own failures of imagination, either explicitly (e.g., “it just doesn't seem possible”) or implicitly, via the comparison of a seemingly impossible event to a possible one (e.g., “you can't walk across water but you could swim across”). Such appeals suggest that preschoolers reason about physical possibility similarly to how adults are purported to reason about counterfactuals: by searching for counterexamples to the status quo. If they can identify such a counterexample, they judge the event possible; if they cannot, they judge the event impossible. Although this strategy does, in fact, lead children to deny the possibility of events that violate physical laws, it also leads them to deny the possibility of events that, although difficult to imagine occurring, do not violate any physical laws, like making pickle-flavored ice cream or finding an alligator under the bed.
The point of this illustration is not to suggest that the process of searching for a counterexample is irrational but to suggest that this process is rational in some contexts (i.e., small, well-defined domains) and not in others (i.e., large, ill-defined domains), and that the appreciation of this fact is a normal developmental achievement. Moreover, by considering whether the application of an inference strategy is rational – as opposed to the strategy itself – one can better appreciate what constitutes a valid conclusion in the domain at hand and what does not. Admittedly, the aforementioned findings come from studies of hypothetical reasoning, not counterfactual reasoning; yet they pertain to Byrne's claims in so far as reasoning about the mutability of particular events in the past is structurally similar to reasoning about the mutability of generic events, past or present. At the very least, this comparison points to the need for additional research on how individuals justify their counterfactual inferences, for such data are likely to shed light on how those inferences were made.
In sum, imagination can be put to rational purposes but it should not be considered inherently rational. Although Byrne's careful analysis of the similarity between counterfactual reasoning and deductive reasoning provides evidence of imagination's systematicity, it does not provide evidence of its rationality.
