The FPORA

Davidson’s argument brackets any methodological issues, but he does make a distinction between a rationalization that provides the actual reason why the action was performed and one that is reasonable/plausible, but does not appeal to the reason that was the agent’s actual reason. This distinction is crucial for any methodological discussion of motives. From this distinction arises an issue that is common to all social sciences, which I call the fundamental problem of reason attribution:

This is also true in daily life. Why my partner, colleagues, parents, or a stranger acts the way they do can only be attributed on the basis of observation of their behavior, including speech acts. I think I know why I am doing something. By using introspection, or self-examination of our mental states, we have a mode of access to our own mental states that is unavailable when we are trying to figure out the motives of other people. I can introspect my reason for leaving a conversation with a colleague early, whether it be that I do not want to be late for a meeting or that I am uninterested in their conversation. When that same colleague stops our conversation short, I am left to deal with the FPORA.

There is a history in IR of a widely perceived problem stemming from the roots of the discipline in a debate between idealism and realism, encapsulated in the well-known difference between the logic of consequences and the logic of appropriateness (March and Olsen Reference March and Olsen1998). Numerous middle-range debates, such as the one about greed vs. grievance in civil war studies, or whether the power of the UN Security Council is about information revelation or normative legitimacy, are fundamentally about reason attribution. This is the understandable doubt that a political actor ‘really believes’ what he/she is saying when they appeal to altruistic or moral values. For example, Lebovic and Voeten (Reference Lebovic and Voeten2006) claim that ‘the empirical implications’ of an argument that states shame human rights violators because it provides information on a state’s reputation for good and bad behavior and an argument that such shaming is done because non-violators morally disapprove of violators are ‘difficult to disentangle’. This doubt elides the distinction between different rationalist reasons and reifies the distinction between two classes of reason. The more fundamental issue is the FPORA.

Introspection is not foolproof. Humans can suffer from self-deception, or ‘denial’ of their true motivations. This can be owing to the desire to conform our thoughts to socially acceptable reasons so as to preserve our sense of self-worth. Some research suggests that we lack the vocabulary to describe some of our thoughts, feelings, or attitudes and so we borrow labels and models from society to help us understand our own mental processes (Wittgenstein Reference Wittgenstein1953; D’Andrade and Strauss Reference D’Andrade and Strauss1992; Jansz Reference Jansz1996). A line of psychological research investigating self-perception theory has found repeatedly that a behavior comes before a related emotion, despite people’s self-reports based on introspection. If people are induced to smile they report being happier, if they are told to stare into a stranger’s eyes they report more romantic attraction to the stranger, if they are told to stand up straight they report higher self-esteem than if told to stand slumped, and so on (Laird Reference Laird2007). Another serious issue is confabulation, such as when people manufacture reasons for their behavior, even when they did not perform the behavior in the first place (Hirstein Reference Hirstein2005). The conclusion reached is that the individuals themselves make mistakes when introspecting the causes of their actions. These lines of research suggest that the FPORA can apply even to first-person reason attribution.Footnote ¹⁵

How can we deal with the FPORA? In this paper, I distinguish between three types of approach to the FPORA: assumption, avoidance, and using empirical evidence. I lay out precisely the form of explanation underlying each type of approach and identify the attractions and limitations of each. Both the assumption and avoidance approaches have costs and rest upon over-stated objections to the use of evidence in reason attribution.

The plausibility of reasons for action

In one example of an explicit appeal to plausibility, Katznelson and Weingast (Reference Katznelson and Weingast2005), commenting on the tendency of rational choice institutionalists studying the US Congress to assume the motive of maximizing the probability of reelection, defend the assumption as justified by its general plausibility. They reference Mayhew (Reference Mayhew1974), who concludes that the individualistic, open US electoral system, and a minimally restrictive party affiliation procedure provide an incentive to focus on reelection at the expense of, say, party loyalty or bureaucratic wrangling, as it might be in the United Kingdom or continental Europe.Footnote ¹⁶ However, Mayhew makes clear that this is contingent on a desire to make a career in Washington, something that Mayhew asserts was unusual in the pre-Civil War era. Even in Mayhew’s study, there is no empirical evidence provided that any action by any individual was actually motivated by the desire for reelection, except a single generalized out-of-context quotation. Katznelson and Weingast’s appeal to Mayhew’s work thus does not mitigate the fact that their response to the FPORA is to assume the motivations of actors.

Under the assumption response, there are no grounds for resolving a challenge based on the superior plausibility of an alternative reason, except in terms of plausibility. If any empirical evidence is used to decide between reasons, then the knowledge claim is subject to all of the issues inherent in such an enterprise (see below). This is a crucial point. Assessing plausibility, however it is done, is a separate task from weighing evidence in a reason attribution.

What does plausibility mean? Is it an unjustifiable intuition? Are there general properties of plausibility with regard to reason attribution, such as logical consistency or phenomenological familiarity? The key feature of rational choice explanations is that acceptable reasons must conform to some minimal standard of rationality, such as completeness and transitivity of preferences. Maybe there are some features, apart from any evidence about the particular case, of the current state of knowledge that mean that there is some sort of distance from our prior beliefs to the postulated reason. This is a question for future research.

The second consequence of the assumption response to the FPORA is that this form of warrant does not allow for any claims that the action is evidence supporting the truth of the reason s or the set of reasons S. Instead, the assumed reason is explaining the action to you. That is, the assumed reason makes the action comprehensible.

Imagine that I see the United States limiting its power by the construction of international organizations (IOs) that hamper its freedom of action (Ikenberry Reference Ikenberry2001). This is puzzling because it seems as if freedom and flexibility are things that states should pursue. Ikenberry dissolves this puzzlement and explains the action by saying that the United States was trying to transmit credible information of benign intentions to other states in the system. If I am resorting to the assumption response to the FPORA, then this is as far as I can go. It would be inconsistent to use statements of intention to communicate benign intentions as supporting evidence that was the motivation because then I would be appealing to a different warrant for my knowledge claim, that is, using evidence to adjudicate between possible reason attributions. However, under the assumption response, it is consistent to claim that statements of other reasons are irrelevant to your knowledge claim. That is, if you are relying on the assumption response, then statements by participants that they were constructing IOs in order to, say, exert control in particular areas of international activity, or to placate a domestic constituency, do not affect your knowledge claim. Rather than using evidence to decide between different possibilities, explanations based on the assumption response explain actions in terms of the assumed reasons. All of the analytic work happens at the level of the logical consistency or the plausibility of the attributed reason.

The costs and benefits of the assumption response

Why would you want to merely assume reasons? A possible objection is that to only assume or postulate reasons is not a valuable task. One major limitation of simply assuming a reason is that there is no claim that the ‘explanatory’ reason is the actual reason driving behavior. It thus seems to be merely a preliminary step before the valuable part of research, the empirical investigation. Clarke and Primo (Reference Clarke and Primo2012) argue against this line of thinking. They point out that elucidating a model, such as a coherent set of reasons for action, constitutes a valuable use of time and effort. Rubinstein (Reference Rubinstein1991) agrees, pointing out that the value of game theory is that it aids in the exploration of reasoning in strategic settings. Schelling reports a reader who had ‘simply not comprehended that an inherently non-zero-sum conflict could exist’ before reading theoretical elaboration of such a scenario in Schelling’s (Reference Schelling1980, vi) work. Laying out or elaborating a reason or set of reasons and its implications clearly, explicitly, and in detail, is hard and uses different skills from those employed in empirical work. Elaboration of reasons for action may also suggest sources of variation in action that would not have been apparent otherwise.

In addition, the evidence available for or against a particular reason may be non-existent or so sparse and underdetermining that simply explaining an action may be the best we can do. However, there can be serious costs to not engaging with the available evidence. Wendt points out that knowing how actors ‘were actually thinking and motivated’ is central to choosing between equally plausible alternative explanations for international institutional design choices. That is, in order to get to ‘the real explanation … we need to get inside the heads and discourse of decision makers and see what is motivating their behavior’ (Wendt Reference Wendt2001, 1028). An excellent example of the drawbacks of merely assuming reasons is the audience costs literature. Since Fearon’s (Reference Fearon1994) elaboration, there has been a profusion of work utilizing the idea that a crucial way state leaders can make credible threats is to engage domestic political costs, ‘audience costs’, for backing down if the threat is not followed through on. This idea solved several theoretical and empirical puzzles, including providing a mechanism for the democratic peace finding (Schultz Reference Schultz2012). However, recent work has empirically investigated whether the audience costs mechanism actually motivates crisis behavior. Trachtenberg (Reference Trachtenberg2012) studies whether state leaders intentionally try to use audience costs to lock in a bargaining position and whether the adversary understands that the threatening state leader ‘would find it hard to give way for fear of incurring audience costs’. His question is do governments ‘actually make this kind of calculation?’ (Trachtenberg Reference Trachtenberg2012, 7). Despite the plausibility of the reasons involved, Trachtenberg finds that audience costs do not motivate behavior in the cases he studies. Similarly, Snyder and Borghard (Reference Snyder and Borghard2011, 437) find ‘hardly any’ evidence that audience costs motivate behavior in post-1945 crises. Mercer encapsulates the primary drawback of the assumption solution: ‘Although one might use audience costs to solve a variety of puzzles, if audience cost mechanisms are imaginary, then so are the solutions that rely on them’ (Reference Mercer2012, 399).

Defining the avoidance approach

The perceived need for intentional explanation is strong; however, there are other types of causes that are fruitfully hunted in political science. Many causal inferences have been made without relying on an answer to the question of reasons for action. One example is Fearon and Laitin’s (Reference Fearon and Laitin2003) work on the duration of civil wars. Among other claims, they argue that mountainous terrain, a proxy for the availability of hard-to-eliminate rebel safe havens, causes an increase in the duration of civil wars. We can use this to explain why some civil wars last so long, or longer when compared with others. This argument is compatible with numerous reasons for pursuing rebellion or for government counter-insurgency actions and thus does not rely on any one particular reason. Similarly, some explanations rely upon network analysis to explain patterns of behavior. Nexon explains the seemingly radical shift in allegiances from the Taliban to the United States in Afghanistan after 2001 with reference to dependency in patron–client relationships; because warlords were bribed to switch, most of their followers switched with them. Clients may have acted out of loyalty, honor, or calculation of the benefits of not breaking ties with the patron. This claim about social ties ‘provides significant explanatory power independent of specific microfoundations’ (Nexon Reference Nexon2009, 41). The key underlying characteristic of the avoidance response is that the knowledge claim must be fully compatible with more than one (themselves incompatible) reason attributions:

An innovative recent line of work that relies upon the avoidance response to the FPORA appeals to the reasons and justifications for action that individuals express, but avoids the attribution of a specific reason to an individual at a particular time. Instead, the constellation of explicit justifications for an action is charted, with the analyst coming to a decision on which justifications were socially sustainable (see also Mills Reference Mills1940). Jackson (Reference Jackson2006, 42) argues that the collective shape (or topography) of the legitimation debate over a policy or set of policies is causally sufficient for an outcome. Not only are policy makers restricted to enacting ‘those policies that they can justify in a manner acceptable to their audience’, but the ways in which a policy is justified ‘makes the policy proposal possible and helps it to win out over alternative courses of action’ (Jackson Reference Jackson2006, 25, 29). This position is premised on agnosticism as to the reasons inside individual actor’s heads, and instead operationalizes the acceptability of policies in terms of how other actors speak and write about, or act toward the policy. The argument is formalized and extended in the study by Krebs and Jackson (Reference Krebs and Jackson2007). In an explicit rejection of reasons as a response to the FPORA, Krebs and Jackson (Reference Krebs and Jackson2007, 41) allow that actors possess motives that shape behavior relevant to political outcomes; however, they propose a type of explanation to which ‘purposive accounts’ (i.e. reason attributions) are irrelevant.

Reasons here are still explanatory, but not in the sense that actor X did action a at time t for reason s. Instead of this sort of judgment, a judgment is made as to what the articulated justifications for action were and which of these were accepted by the relevant actors in the sense of not resisted. Regardless of what the individual motivations actually were, or whether reasons are constituted by internal mental states, this form of the avoidance response allows the analyst to use evidence to determine what the actors involved thought a convincing justification would be. Given that the justification was provided, it must have been thought convincing to some actor, even if that is the same actor who is expressing the justification. An actor makes a justification for an action or provides a reason why an action should be done in order to legitimate that action in the eyes of others. The aggregation of those reasons at a historical point in time constitutes the boundaries of legitimate discourse at that time for that action. Identifying socially sustainable justifications is especially useful when a group of actors are trying to come to a decision on a joint action, like in a government or a group of governmental actors. Individuals are being honest about their reasoning, or they are trying to be convincing to others, or they are trying to avoid censure for violating shared rhetorical standards. In all three of these situations, speech acts are potential evidence of the reasons that are thought convincing.

The form of the knowledge claim for a socially sustainable justification is:

Notice that individual reasons for action do not appear in this knowledge claim. The actors could be motivated by j, they could simply be willing to accept j as a reason for a, or they could be motivated by j*, but know that others would not be willing to accept j* as a reason for action.

Here the distinction between private and public justifications is important. As public pronouncements are the only ones that are being used to legitimate action, they are more important than private statements as evidence to a socially sustainable justification explanation.

The costs of avoidance

Is anything lost by recourse to an avoidance explanation? That is, can investigating reasons in an avoidance explanation add anything to our understanding of the causes of action? Let us return to Fearon and Laitin’s civil war example, in which we know that mountainous terrain increases length of civil wars on average. It would be a distinct improvement in our causal understanding to also know that the rebels retreating to mountain hideouts do so because they intend to regather their strength for future fighting, rather than that they want to surrender, but they have to hide from authorities intent on eradicating them. Similarly, Krebs and Jackson’s claim that conscription helped the Druze get awarded Israeli citizenship would be supplemented by knowing what the Israeli authorities’ motives were. Did they hope that including Druze in the Israeli Defense Force would bolster their claim to liberalism and undercut international criticism (Krebs and Jackson Reference Krebs and Jackson2007, 50)? Or something else? Advocates of analyzing international practices have appealed to the ‘great advantage of ridding [the analyst] of the need to make problematic claims about the state of mind amongst the people who perform the practice’ (Andersen and Neumann Reference Andersen and Neumann2012, 458). As Ringmar explains, however, merely describing the practices gives us ‘no clue’ as to the intentions behind the practices. Implicitly conceding this point, some practices analysts deal with this problem by adding in a reconstruction of the intentions and aims behind practices, for example, of deterrence in the Cold War (Ringmar Reference Ringmar2014, 13). The point here is not that explanation is impossible without reasons. It is that reasons add to our understanding of the causes of action.

There are at least three ways in which avoiding reason attribution lessens our understanding of the causes of action. First, obviously, even though we may have a coherent and empirically supported explanation of an outcome, we do not know why the individuals involved performed the actions making up the outcome. However, there are other consequences important even if we are locating explanatory leverage elsewhere. The second cost of avoidance is the loss of some sources of variation in the actions of interest. For example, rebels regathering their strength might be less amenable to a negotiated settlement than rebels looking to surrender. An Israeli government on the lookout for legitimation in terms of international liberalism may do different things from a government looking to recruit the fiercest fighters. Third, avoidance also means losing a way to generate new theoretical arguments. Empirical investigation of a particular type of action might unearth a hitherto unimagined reason or perhaps one that is well known, but that no one had previously thought was driving this type of action. For example, the actions of US President Richard Nixon and his adviser Henry Kissinger during the 1971 war between India and Pakistan can only be explained with reference to their secret diplomacy with the People’s Republic of China, ultimately motivated by grand strategy with regard to the USSR. To others, this motivation was unthinkable during the crisis surrounding the war (Hollen Reference Hollen1980).

Statements of motivation

There is an intuitive plausibility to the following procedure. I want to know why actor X did action a. Actor X says, ‘I did action a because I wanted to achieve outcome b, or for reason s’. Assuming that the given reason s is comprehensible to me, s explains a. This initially seems to be a best case scenario for empirically determining reasons for action. However, there is a prominent criticism of this procedure; the potential for dishonesty or misrepresentation in proclaiming one’s reasons. The misrepresentation principle is:

There are two positions on dishonesty or misrepresentation regarding reasons. The first is that political actors’ incentives for misrepresentation are so pervasive that it should be the default starting point of any analysis. The other is that misrepresentation should not be assumed but should instead be based on evidence, for example, some inconsistency between various pieces of evidence. Scholars often make a case for pervasive, systematic strategic misrepresentation that has prima facie considerations for reason attribution. For example, Mackie (Reference Mackie1998) notes the tendency to assume that ‘all men are liars’ in political economy or public choice models of voting. Elster places this position front and center of a discussion of reason attribution. He points out that ‘there are many reasons why people might want to misrepresent their motivations and those of their opponents’ (Elster Reference Elster2007, 59). For example, societies have a normative hierarchy of motivations and people gain an advantage by appearing to be motivated by a ‘better’ motivation (this is often a significant part of legitimation contests). However, one inherent problem with this argument is that it denies that we can know people’s real motives by appealing to people’s real motives, thus flirting with circularity (Bruce and Wallis Reference Bruce and Wallis1983, 63).

Excessive weight is often placed in IR on the possibility of misrepresentation. This can have pernicious effects on research practice. For example, excessive skepticism of statements leads Simmons and Danner to ignore relevant evidence of reasons and instead resort to making claims based on wildly insufficient data. They ask why the International Criminal Court (ICC) was set up and ‘more importantly, why do states agree to join this institution?’ (Simmons and Danner Reference Simmons and Danner2010, 225). Noting that ‘Evidence on governments’ motive for joining the ICC is hard to come by’, they assert that private statements are inaccessible and public statements are too coarse-grained, that is, they are consistent with numerous analytically distinct reasons. That said, they provide a couple of statements that the ICC will reduce conflict, restore confidence in the country acceding, and add to the accountability of the country’s leaders. They even say of one statement, ‘One can infer from this statement that the speaker acknowledges that domestic processes are often less than effective, and that the ICC can in these cases provide a more effective—because more credible—substitute’ (Simmons and Danner Reference Simmons and Danner2010, 237). However, they then assert, in a statement of excessive skepticism, that these statements ‘reveal very little about the way supporters expect the Court actually to operate’. They then say that they will look at actions; however, they do not. Instead, they look at coarse-grained properties of states, like Freedom House democracy scores or whether a country is involved in a civil war, and infer directly from those properties the reasons for ratification. This evidence is insufficient to distinguish between alternative reason attributions, and so does not add to our understanding of the causes of ICC ratification. This is one example where excessive skepticism of statements has led to unconvincing reason attribution.

Historians attempt to mitigate the problem of misrepresentation with reference to expressions of reason from sources they deem less likely to be used for a political purpose, primarily private documents including classified internal memoranda and personal letters. As Elster notes, the underlying principle behind the credibility of these private documents is that they are ‘less likely to be motivated by a desire for misrepresentation’ (Reference Elster2007, 61). However, positing that some types of source [such as private letters and diaries (George and Bennett Reference George and Bennett2005, 100, fn 17)] are systematically less likely to be subject to misrepresentation is no more plausible than the reverse. That is, there is no a priori systematic difference between types of sources in terms of whether they more or less truly indicate reasons. As Broockman advocates, ‘both public and private statements should be viewed through a strategic lens’ (Reference Broockman2012, 105).

Private settings qua private settings do not systematically favor the true revelation of reasons. If there is a strategic incentive to conceal or misrepresent reasons in the most public setting, there is also a strategic incentive to misrepresent reasons in other less public settings (Krebs and Jackson Reference Krebs and Jackson2007, 40). Jacobs (Reference Jacobs2014) argues that, in general, political actors ‘have incentives to exaggerate the importance of “good policy” motives and broad social benefits’ of their policy positions. However, in general, political actors also have incentives to frame their behavior positively to any audience. If they are speaking privately to a small stakeholder group, they have an incentive to highlight the narrow individual benefits accruing to that group. If they are speaking to racists they are more likely to appeal to racist reasons than they are to non-racists.Footnote ¹⁹ It is easy to construct simple scenarios in which decision makers would not lay out their actual motivations in a private letter to a friend or family member. For example, maybe the actual reasoning is thought to be too complicated, morally questionable, or too reliant on special knowledge for the letter recipient to understand. If they are writing a letter to their spouse or family member, they have an incentive to make themselves seem as favorable as possible to that person, maybe by emphasizing the private benefits that they, and hence their family, will accrue from an action or policy. Even letters or memos to colleagues could fall prey to common distortions, such as a desire not to appear biased, idealistic, or naïve. For example, if they are writing an intragovernmental memo, they have an incentive to conceal their idealistic motives in favor of a cynical national interest-oriented analysis, so as to advance their standing and reputation for calm competence in that organization.

Such speculations could be multiplied. However, what they demonstrate is that in the absence of evidence, there is no good prima facie reason to accord one type of expression of reason primacy over another. Instead, misrepresentation should be a judgment arrived at on the basis of the evaluation of evidence. There is the principle of nemo gratis mendax, that is, no one lies freely. Under this principle, expressions of reason should only be doubted if there is a concrete reason to do so in a particular circumstance. That is, the construction of a reason for dishonesty or misrepresentation must use some sort of evidence. Otherwise, expressions of reason must be taken at face value. This principle focusses attention on specific reasons for doubt, rather than a generalized assumption that anything said in public by a political actor is necessarily dishonest or disingenuous. There is even some systematic empirical evidence that political leaders’ public speeches accurately convey their actual beliefs. Utilizing the fact that US President John F. Kennedy recorded private discussions with key advisers in the summer of 1962, Renshon (Reference Renshon2009, 656) compares the operational code (a set of political beliefs) of Kennedy’s public and private speech, finding a ‘striking similarity’ between the two.

However, when we do have evidence of misrepresentation, then we should incorporate that into our reason attribution. One notable example of evidence of dishonesty is that with which Nexon opens his book on the political impact of the Protestant Reformation. Holy Roman Emperor Charles V, while publicly justifying his war against the Protestant states Hesse and Saxony as that they were ‘transgressors of the peace against the Duke of Brunswick and his territory’, wrote to his sister that this ‘pretext will not long disguise the fact that this is a matter of religion, yet it serves for the present to divide the renegades’ (quoted in Nexon Reference Nexon2009, 1). Here the fact that we have this evidence makes preferring the attribution of an ulterior motive, in the face of public statements otherwise, warranted.

The general principle of consilience is again a useful guide here. The attribution of reason misrepresentation only makes sense when there is an inconsistency between different classes of evidence, like different classes of statements. A good example of the utility of the consilience principle is Broockman’s study of the negotiating positions behind the passage of a 1964 healthcare law in the United States. Broockman, addressing the ‘problem of preferences’, that is, that actors might strategically misrepresent their preferences during policy contestation and bargaining, advocates ‘considering how actors’ expressed preferences vary across strategic contexts’ (Reference Broockman2012, 84). Specifically, Broockman found that a previous study had misattributed the preferences of business interest groups on the basis of a letter from the National Association of Manufacturers (NAM) and Congressman John Byrnes, ranking Republican on the House Ways and Means committee. The letter includes reasons why industry would support a different version of the healthcare law than was currently being proposed, which the previous work had said meant that the NAM and business interests supported passing a version of Medicare. Broockman argues that evidence of statements from different strategic contexts, including from before the Democratic landslide victory in the 1964 election, shows that in fact business interests ‘were totally and unmistakably opposed to Medicare’ (Reference Broockman2012, 93) and were proposing an alternative in order to dilute some provisions in the bill. Further, Byrnes’ writings to other Republicans and other statements indicate that Byrnes proposed an alternative to Medicare, whereas expecting it to be rejected so that Republicans could convincingly argue that they were not opposed to healthcare for the aged. Here, in accordance with the consilience principle, Broockman’s attribution of reasons to Byrnes and industry groups is superior because it is consistent with different classes of evidence, that is, statements from different strategic contexts.

Consilience can also accommodate the use of different classes of evidence of the absence of a particular set of beliefs. For example, when discussing the failure of German officials to consider the future misuse of funds collected for the first public pension scheme in 1889, Jacobs (Reference Jacobs2014) makes the case that the attribution of this state of mind is supported by the fact that there is no mention of it across several deliberative venues.

As intuitively attractive as statements of the form ‘I did action a for reason s’ are, much historical evidence is not as starkly related to the reason for an action. This means that, in practice, using statements as evidence of reasons requires an interpretation of those statements. Further, the conceptual vocabulary used by social scientists is often not the same as that used by historical actors in their self-ascriptions of reasons. As Taylor notes, in a discussion of how our representations, for example, justifications for action, may not map perfectly onto the distinct categories of action created by social scientists, ‘… the person concerned may not even possess the appropriate descriptive term. For instance, when I stand respectfully and defer to you, I may not have the word “deference” in my vocabulary … This understanding is not, or is only imperfectly, captured in our representations’ (Taylor Reference Taylor1993, 51). Using a statement or utterance as evidence in support of a reason attribution, requires making an interpretive judgment about the fit of that utterance into a particular theoretical framework. This can be a serious challenge, especially when dealing with actors and contexts with which we may be phenomenologically unfamiliar (see Bevir Reference Bevir1999 for a thorough evaluation of this and related issues). Skinner argues that one tactic for improving the quality of interpretive judgments is to ‘trace the relations between the given utterance and this wider linguistic context as a means of decoding the intentions of the given writer’. Analysis of the social context is ‘the ultimate framework for helping to decide what conventionally recognizable meanings it might in principle have been possible for someone to have intended to communicate’ (Skinner Reference Skinner2002, 87). Hopf summarizes this position by saying that, ‘Evidence does not consist of the actor’s words alone’ and that ‘there must always be an accompanying account of the relevant sociohistorical context’ (Reference Hopf2007, 61).

Another interpretation issue is that statements sometimes cannot be accurately understood out of the context of the other statements made by an actor.Footnote ²⁰ A statement may appear to indicate a particular reason or intention or belief but when viewed amongst other statements made by the same person, the statement may indicate something else. One example involves Stanley K. Hornbeck, the Chief of the Far Eastern Affairs Division at the US State Department in 1931. Hornbeck was considering potential reactions to the Japanese invasion of Manchuria earlier that year (the Manchurian Crisis). In particular, Hornbeck was evaluating the proposed policy of declaring that the United States would not recognize any result of the invasion. This policy later came to be called the Stimson Doctrine, after the US Secretary of State. Hornbeck wrote a memorandum, in which he writes that the proposed policy of non-recognition would, ‘show the powers “mean business”. It would give the Pact of Paris “teeth”. It would answer the charge that the League and the various governments are impotent’ (Doenecke Reference Doenecke1981, 85). Several secondary sources use this quote as evidence that Hornbeck believed that the non-recognition policy would be effective. However, Hornbeck places ‘mean business’ and ‘teeth’ in quotation marks. This, in addition to numerous other statements made by Hornbeck that non-recognition would be relatively useless, suggests that he is responding to criticism made in the press and elsewhere, and that the policy of non-recognition might deflate that criticism.

This issue of the fallibility of interpretation places a heavy burden on the analyst, as establishing the context of a statement can be time consuming and difficult, and there may not be much available evidence. However, the evidentiary value of statements is dependent upon such interpretation.

Actions as evidence of reasons

Fear that actors misrepresent their true reasons for action is widespread. A common reaction to this problem is to rely instead on action as an indicator of reasons, dismissing expressions as ‘cheap talk’. Only actions that incur enough costs to discriminate between actors really motivated by reason s and thus willing to bear those costs in the pursuit of s can be used as evidence of reason. This is the underlying logic of the resort to ‘costly signaling’ models from game theory (e.g. Fearon Reference Fearon1997). Elster, asking ‘do they put their money where their mouth is?’, uses the example of the Bush administration and its professed reasons for invading Iraq. The key stated reason was to seize or destroy Saddam Hussein’s weapons of mass destruction (WMD). Elster (Reference Elster2007, 63) argues that a useful piece of evidence relevant to whether this was the true motivation would be whether the Bush administration took steps to protect the invasion forces from WMD. Here, the key principle is not the costliness of the action per se. Rather, if an action would only make sense if performed by someone with a particular reason, then it is good evidence in favor of that reason being the true reason. However, this strategy seems prey to the same sort of problem as afflicts the use of statements. Any action seems liable to a redescription that makes more than one reason equally supported. For example, poor planning, arguably in evidence in other respects during the Iraq War, could have meant that even a sincere Bush administration did not make adequate preparation for US troops to face WMD.

Another way of using actions as evidence of reason involves assessing if actions, other than the action to be explained, by the same actor can plausibly be described as motivated by the same motivation.

In the abstract this can sound confusing and counterintuitive. However, this format for warranting a knowledge claim is widespread in lay discussions of foreign policy and is sometimes used by scholars. For example, Girard (Reference Girard2004) asks why President Bill Clinton decided to invade Haiti in 1994. Girard uses the actions of the Clinton administration as evidence that the pursuit of democracy and human rights in Haiti were not primary reasons for the 1994 invasion, despite Clinton’s explicit appeal to the need to restore democracy as a justification for the invasion in an address to the nation in September 1994. Girard points to the absence of US action to restore democracy in other countries, US willingness to accept the fraudulent 1994 electoral victory in the Dominican Republic in exchange for Dominican participation in the embargo on Haiti, and Clinton’s refusal to ask for congressional approval of the invasion as evidence that the desire to restore democracy was not actually the motivation behind Operation Restore Democracy.

Underlying the idea that actions can be used as evidence of motivation for other actions is the concept of reason consistency. That is, people who are motivated to perform an action in pursuit of some goal are also motivated to perform other actions in pursuit of that goal. I have two reactions to this consistency premise. The first is that it is reasonable to suppose that an actor could be driven at two different times by two different reasons (this is another place where equifinality is relevant). I might want to eat spaghetti with marinara sauce now because it tastes nice, but later not eat spaghetti because I worry about embarrassing myself in a sensitive social situation. The latter action is not evidence against my eating spaghetti because I like it. My second reaction is that it is intuitive to hold some consistency requirement for reason attribution. Politicians tout their record on voting the same way for the same issues as evidence that they truly believe in that position. They are implicitly relying on the idea that someone motivated to act in a particular way for a particular reason is more likely to do so in the future. This reasoning is common in scholarship. For example, Saunders takes a strong position on the interpretation of statements made in times of crisis:

Here, Saunders is implicitly appealing to a form of the reason consistency idea. She is holding that beliefs (a central part of reasons for action) remain constant over time and can drive and explain behavior even in the face of contradictory contemporary statements. If the reason consistency idea is rejected, then we would have no reason to think that presidents’ beliefs before office are relevant to decisions made in office.

The consilience principle again proves a useful guide. If different actions are all consistent with a particular reason, then that reason attribution is to be preferred to one that is not consistent with those actions. However, as in the spaghetti example, changes in context should also be taken into account.

Context and reason attribution

Another type of evidence that might be used in a reason attribution is some property of the strategic context of the action performed. A contextual condition is some feature of an actor’s situation, or environment, that plausibly provides him with an incentive to perform an action. This kind of evidence has the benefit of often being more easily accessible than statements made in a situation where there is no incentive for misrepresentation. Elster (Reference Elster2007, 61) raises the idea of using the ‘objective interests’ of an actor as a proxy for the actor’s subjective motivation. However, as he also notes, this can suggest useful hypotheses, but does not dissolve the FPORA.

A recent example demonstrates both the importance of the strategic context in informing and motivating policy choices and the limitations of context as evidence of motive. Harvey (Reference Harvey2011) uses extensive evidence, including private and public speeches and writings, to argue that the contextual conditions facing US president George Bush after 2001 would counterfactually have led ‘president’ Al Gore to make the same choices for the same reasons, culminating in going to war on Iraq in 2003. Harvey does not simply state the context and infer the reason from that. Instead, he carefully eliminates alternative motives for Bush’s decisions, using a variety of types of evidence. What confidence we have in Harvey’s conclusion derives from his use of different classes of evidence, leading to an inference to the most consilient reason.

Merely establishing that the context could provide an actor with a reason is only the first step in a reason attribution. Yet, sometimes work in IR relies too heavily on strategic context as evidence in reason attribution. The mere presence of a contextual condition cannot distinguish between reasons that are similarly supported by the presence of other contextual conditions. In addition, the actors in the strategic context, as described by scholars, may not be aware of that context. Some work does not adequately appreciate these points. One example is Simmons and Danner’s analysis of the ICC (see above). They use the fact that a state is involved in a civil war as their only evidence that state’s reason for joining the ICC was to credibly commit to abiding by a peace agreement in that civil war. They argue that states who have no other way to credibly commit to peace in a civil war join the ICC in order to increase the costs of returning to war, and hence more credibly committing to a peace agreement. However, states in a civil war might also be involved in negotiations over trade deals (perhaps aimed at the post-war period) and want to use accession to the ICC as a bargaining chip in those negotiations.

Moreover, there can be more than one reason to perform the same action. That is, two actions that are of the same type can be motivated by two different reasons. One state in a civil war might want to credibly commit to peace using the ICC, but another might want to join the ICC because they think that this will enhance their chances of receiving external support in the civil war. The mere fact of being in a civil war cannot distinguish between these two possibilities.

Strategic context is thus useful for suggesting hypotheses of reasons, but by itself is only weak evidence. However, if a reason attribution is consistent with statements of reason and the strategic context, then it is to be preferred to a reason that is not. Here again, the principle of consilience subsumes our intuitions.

General comments on using evidence to determine motivation

Some scholars take a strong stand against the position that evidence of reasons can be used. For example, Jackson points out that ‘there is no way to tell whether these reconstructed motives are the real motives possessed by the historical actors in question’ (Reference Jackson2002, 748) and Krebs and Jackson go further to deny that there is ‘evidence [that] could even in principle clinch the case’ (Reference Krebs and Jackson2007, 40). Similarly, Frieden argues that ‘preferences are unobservable independent of outcomes’ (Reference Frieden1999, 48). Such a radical skepticism about evidence of reasons means resorting to an alternative response to the FPORA. Jackson also claims that ‘given time, one can usually find evidence to support virtually any position in the documentary record’ (Reference Jackson2002, 748). This latter position is too strong. If you are accepting the possibility that you can use evidence to attribute reasons and you have a delimited time period and a clear group of actors you think were instrumental in the key decisions, the positions that receive support must be relatively limited and the weight of evidence may even be in only one direction, making inference to a historical reason more clear-cut.