Somaliland is a democratically governed, autonomous region that maintains an independent police force, defends its borders, and issues currency in its own name.¹ Despite claims of statehood, it has not been officially recognized as a country by any state-level actors. Instead, it is considered an “autonomous region of Somalia.” Scotland is a semi-autonomous region that neither controls nor defends its borders² and does not govern its own affairs to the degree that countries typically do. Despite being atypical in these respects, it is officially recognized as a country.

Atypical cases like Somaliland, Scotland, and others immediately raise the question as to the ontology of institutional objects like countries, presidents, money, borders, and traffic lights. The issue is particularly pressing among social scientists who study such phenomena. Suppose one is doing a cross-country comparative analysis of some social or economic trend. The trend does hold in Somalia (or the United Kingdom) but does not apply in Somaliland (or Scotland). In such a case, does Somaliland (or Scotland) serve as counter-examples, thus weakening any potential claim to generality? Or does Somaliland (or Scotland) not “count,” hence not affecting the generality of any claim as to how widespread the trend actually is?³

The default answer one typically encounters when asking about what makes it the case that \(X\) is a country is that \(X\) is a country if, and only if, regulative bodies consider it a country. Of course, such regulative bodies have declared that ketchup is a vegetable,⁴ that Microsoft is a person,⁵ and that an arm bent 15 degrees is straight.⁶ Botany, psychology, and mathematics have ignored these uses of “vegetable,” “person,” and “straight” to no ill effect. So why should we care what regulative bodies have to say when individuating the institutional world for the purposes of social science?

In this paper, we argue that social scientists should not feel compelled to individuate the social world in the same way that institutions do. The institutional use of language differs from the descriptive use proper to social science in at least two ways, and both serve to make the regulatory schemes of individuation used by institutions unsuited for descriptive work. The first bad consequence of regulatory individuation is due to the familiar fact that institutional judgments have constrained revisability criteria. This implies that the facts picked out by institutional judgments will almost inevitably be non-identical to the facts picked out by our best epistemic practices. The second bad consequence of regulatory individuation is due to the fact that institutions amend their declarative judgments based on the syntactic properties of the judgments and in response to regulatory pressure, not based on semantic properties and in response to matters of descriptive adequacy. This implies that “regulatory kinds” (countries, borders, kings) will almost inevitably be disjunctive kinds that are ill-suited for scientific theorizing. In making this argument, we reject the account of Searle (2005), whose position implies that social scientists should respect the individuation schemes of institutions.

1 Searle on Institutional Facts

John Searle, in a number of publications (1995, 2010), has defended an elegant view of institutional facts. The object of study is institutional objects, i.e., objects that serve some social purpose in virtue of having certain deontic powers (rights, duties, obligations). These deontic powers cannot be sufficiently explained by the intrinsic or natural properties of the object itself, but are the result of some institutional structure collectively endowing the object with such properties by recognizing it as having such properties.

A typical example is that of a president. A president is not a president in virtue of his or her physical or intrinsic properties, but in virtue of being recognized as a president by the governing institution of the country that he or she is president of. Paradigm cases of institutional objects also include countries, borders, driver’s licenses, the playing field of a football game, and so on. Our social reality is filled with such objects, and we interact with them all the time.

Two aspects of Searle’s view are of particular interest. First, he claims that institutional facts have the logical structure “\(X\) counts as \(Y\) in \(C\)” (1995, 28).⁷ The \(X\)-term denotes the natural object, the \(Y\)-term is the institutional specification of the object, and the \(C\)-term denotes the context in which the institutional object has its function. In this way, Joe Biden counts as the president of the United States at present, a specific line counts as the goal line during a game of football, and so on. Second, Searle claims that the recognition that is constitutive of the existence of an institutional fact is essentially collective recognition. Institutions are collectives, and the collective recognition by which they endow an object with deontic powers is irreducible⁸ to individual recognition (1995, 24–25).⁹

Searle (2005) considers his view to have particular relevance for the social sciences. He introduces an article on the relevance of his view for economics (and social science in general) as follows:

When I was an undergraduate at Oxford, we were taught economics almost as though it were a natural science. The subject matter of economics may be different from physics, but only in the way that the subject matter of chemistry or biology is different from physics. The actual results were presented to us as if they were scientific theories. So, when we learned that savings equals investment, it was taught in the same tone of voice as one teaches that force equals mass times acceleration. And we learned that rational entrepreneurs sell where marginal cost equals marginal revenue in the way that we once learned that bodies attract in a way that is directly proportional to their mass and inversely proportional to the square of the distance between them. At no point was it ever suggested that the reality described by economic theory was dependent on human beliefs and other attitudes in a way that was totally unlike the reality described by physics and chemistry. (Searle 2005, 1)

Searle sets up a basic distinction between the objects of the physical sciences and the objects of social sciences and advises social scientists to heed the fact that their objects are fundamentally unlike those of the physical sciences. The objects of social science are frequently institutional objects and, as such, should be understood as explained above, i.e., in terms of the collective recognition of objects as having certain deontic powers.

Of particular importance to the current discussion is his claim that such objects can only exist for as long as they are represented as existing (1995, 13) and his claim that such objects should be understood as having a logical structure (1995, 22). This implies that if one is a social scientist and wishes to study borders, countries, or presidents, then one must understand one’s area of study as pertaining to those things recognized to be borders, countries, and presidents. In other words, those things that exist in virtue of the collective acceptance of a declaration of the form “\(X\) counts as \(Y\) in \(C\).” This implies, although it is not explicitly stated by Searle, that the social scientist must individuate the institutional world as the institutions that create it do. For if it is constitutive of borders, countries, and presidents that they must be recognized to be these objects, then studying objects not so recognized is to not study borders, countries, and presidents at all.

2 Two Peculiarities of the Institutional Use of Language

2.1 A Toy Example

Below, we will argue that social scientists would be ill-served if they employed the individuation schemes used by institutions themselves. The argument is based on the fact that institutions use language in peculiar ways, and these ways make institutional standards of individuation ill-suited for the purposes of scientific description. This is not to say that there is any specific problem with describing institutions; rather, the claim is that the social scientist should not feel compelled to use the regulatory schemes of individuation adopted by institutions when describing institutional facts.

For purposes of exposition and illustration, it will be useful to have a toy example at our disposal. Suppose there is a village in the Scottish Highlands that has a cultural ritual called “Firecasting” that takes place annually on the first day of spring. They celebrate the end of winter and the reduced need for heating by letting each member of the village attempt to light a torch on fire, run to the firecasting line, and hurl it into a lake, extinguishing the flame within twenty seconds. Those who succeed get a medal (and receiving such a medal has significant prestige in the village).

In Firecasting, there is an umpire who keeps time, adjudicates whether a flame has been extinguished, etc. Every time a torch has been extinguished, the umpire proclaims, “Player \(x\) is a firecaster,” i.e., a flame has been extinguished.

2.2 First Peculiarity Constrained Revisability

The umpire in Firecasting has to judge whether a specific state of affairs obtains, namely, whether the flame has been extinguished. This is a judgment that any spectator can also make. The umpire’s judgment, however, counts in a way that the judgments of spectators do not. Consider the following judgment:

If a spectator makes a judgment by using (1), then this is a speech act of description that asserts that John extinguished the flaming torch. As it is a standard instance of description, it can be straightforwardly true or false. Call the use of the institutional term “firecaster” in a speech act of description the descriptive use of the term.

If (1) is used by the umpire, however, the situation is different. The umpire’s use of (1) is based on his assessment of whether the flame has been extinguished, yet his speech act is that of declaration. His speech act has the function of creating a certain institutional fact, namely, the institutional fact that John is a firecaster. In Searle’s terminology, such a judgment has certain deontic consequences, namely, that the player is entitled to be awarded a medal by the village. Call such a use of the institutional term “firecasted” the regulative use.¹⁰

Note that if the umpire makes a mistake in adjudging whether John has extinguished the flame and erroneously declares that he is a firecaster, then the descriptive content of (1) is false, yet the regulative content of (1) can still be affirmed. This is so that, even if the umpire makes a mistake, the deontic consequences of his judgment will still obtain, i.e., John will still be entitled to the medal. What the umpire commits himself (and the village) to through the speech act of declaration is, above all else, that John is entitled to receive the medal. The umpire’s judgment might be based on whether the descriptive content of (1) obtains, yet what is affirmed by the umpire in making his judgment is something else.¹¹

It is a staple of the literature on the philosophy of law (see, for instance, Hart 1961) that institutional declarations cannot be revised in light of future evidence in the same way that descriptive judgments can. Even if the umpire and the village see conclusive evidence that John did not extinguish the torch, they may choose to “let the judgment stand,” i.e., to remain committed to enforcing the deontic consequences of the original regulative judgment. While the village may choose to explicitly adopt regulative rules that do allow for the revision of prior judgments, there is nothing inherently irrational about not doing so, as affirming the regulative content of (1) does not logically commit them to any specific position as to the truth value of the descriptive content of (1). In this way, the regulative judgment contrasts sharply with the descriptive judgment, as it is a sine qua non of descriptive practice that such judgments are always revisable in light of future evidence.

In fact, most real-world sports do not, except in extreme cases, allow such later reviews of umpiring decisions. Mistakes inevitably happen, and sports fans everywhere use the institutional term itself in a descriptive way in order to register their disagreement with the referee. Consider judgments like “That was never a strike!” “He was miles offside!” “It pitched outside leg stump!” and so on. In such cases, the utterer uses a non-institutional, descriptive standard for applying the terms “strike,” “offside,” and “outside.” Here, the institutional term is used in order to voice disagreement with the factual basis of an umpiring decision (and also to draw attention to the unfairness of the deontic consequences of such a decision).

The phenomenon of constrained revisability is found in all institutional settings. While institutional judgments can sometimes be overruled—i.e., appealed in various ways—such revisability is constrained in a way that that open-ended, epistemic inquiry is not. For example, legal systems in a wide variety of countries recognize a principle of “double jeopardy,” whereby an accused cannot be retried for an offense that they have already been acquitted of. This remains so even if definitive evidence of prior guilt is produced and no one believes that the descriptive judgment underlying the institutional declaration was accurate.

Some legal systems do allow various, tightly restricted exceptions to this principle. In general, though, the revisability of the legal declaration is constrained in a way that commitment to the underlying descriptive claim is not. The distinction between the descriptive use and the regulative use of legal terms is again well-recognized in our ordinary discourse. Consider judgments like “Andy Dufresne was innocent,”¹² “OJ Simpson was guilty,” “Jimmy Saville was a criminal,” and so on.¹³

The fact that (1) can express distinct speech acts with distinct criteria of revisability means that the denotation of the descriptive use and the regulative use of a term can diverge. In the case of the game Firecasting, the denotation of the descriptive use of the term “firecasters” will include those who succeeded in extinguishing a flame. The denotation of the regulative use will include all those who were adjudged to have extinguished a flame. If a scientist were to study firecasting, then the nature of the study might force her to take the distinction seriously and use one or the other criterion. In this way, if the scientist were tasked with determining what physical characteristics allow one to firecast, then she would be ill-served by the regulative use. This is because, if a number of serious umpiring errors have occurred in the history of Firecasting, any law-like generalization that the scientist seeks to uncover will be much more likely to apply to those who actually achieved the feat of extinguishing a flame and not merely those adjudged to have done so. The denotation of the regulative use of “firecasting” will almost inevitably be non-identical to the denotation of the descriptive use of “firecasting”; the denotation of the former will be more heterogeneous with regards to physical characteristics (as it includes both those who succeeded and those who did not) and, as such, less likely to be the object of useful law-like generalizations of the required type.

The opposite is likely to be true for the historian of the game. The historian who writes about the stars of the game is implicitly, and correctly, writing about the regulative use when writing about those recognized to have firecasted. Here, the main interest lies in those falling under the denotation of the regulative use, and as such, reports of prior regulative use are appropriate to the study. As this is the main topic of interest, any law-like regularities that the typical sports historian seeks are supposed to concern those who were adjudged to have firecasted.

The same distinction applies to the academic study of less frivolous matters. Consider a criminologist who aims to make discoveries about the causes of crime in order to determine how law-breaking can be prevented. Here, the interest is likely to lie in determining what causes people to break the law. Breaking the law, of course, is not the same thing as being adjudged to have broken the law. In this manner, the criminologist would feel vindicated if their theory applied to someone who did commit a crime but was never caught or convicted. In the same way, they would be untroubled if their theory did not apply to someone who was wrongfully convicted. When we express an interest in preventing crime, we are typically not expressing an interest in having fewer of the people who break the law caught but in having the law broken less. This makes the descriptive use of the term “criminal” the one appropriate to such a study.

The criminologist could, of course, decide to try and find out what distinguishes those convicted of committing a crime from those not so convicted. Here, the regulative use of “criminal” would be appropriate to the study. Note, however, that we have some reason to believe that in typical cases, law-like generalizations are more likely to apply to the descriptive use. The denotation of the descriptive use of criminal would include all those who broke the law. The denotation of the regulative use would include those who broke the law and were convicted and those who did not break the law and were convicted, while excluding those who did break the law and were not convicted. As the latter category is individuated in terms of a more heterogeneous mix of properties, one would suspect, ceteris paribus, that the descriptive use of the term “criminal” would be more suitable for obtaining law-like generalization. Simply put, it will typically be easier to obtain general truths among a group whose members were rightfully classified as belonging to the group than among a group that includes a mixture of those correctly and incorrectly classified as members of the relevant group.

The above reasoning implies that the constrained revisability of regulative judgments sometimes gives the social scientist a good reason to, despite using the terms used by some specific institution itself, reject the individuation scheme of the institution. This is so as, firstly, the ultimate goal of the inquiry (i.e., crime prevention) may demand it. Secondly, the descriptive use of the institutional term will be more suitable for law-like generalization and, so, more useful to social science.

2.3 Second Peculiarity Institutional Judgments Are Amended Based on Their Inferential (Syntactic) Properties, Not Descriptive (Semantic) Properties

Suppose that the village that practices Firecasting notices that players are sometimes prevented from hurling the torch by other players kicking them just as they are about to hurl it and, in so doing, making it less likely that the player throws across the line. They wish to make such behavior pointless and, so, announce that players who are kicked as they are about to throw will receive a medal anyway, even if their throw did not cross the line.

The required rule change can be made in two distinct ways. Prior to the rule change, the relevant rules of Firecasting are as follows:

The first way to change the rule so as to award those who were kicked just prior to throwing would be to amend the definition of firecasting so that those who were kicked also “count” as firecasters. This option is analogous to a “penalty try” in rugby. If a rugby player is illegitimately prevented from scoring a regular try, the referee may award a so-called “penalty try” to the team prevented from scoring. A rugby team awarded a penalty try is awarded five points in the same way that a team that scores a regular try is awarded five points. In this way, the penalty try “still counts,” despite the fact that the attacking team was prevented from scoring a regular try. In the same spirit, the village can amend (1) as follows:

The village, however, need not amend the definition of “firecasting” in order to secure the result that those who are kicked in order to prevent them from firecasting still receive a medal. They can leave (1) intact and simply amend (2) so that it states that those who are kicked also receive a medal from the village. In this way, (2) can be amended as follows:

(2\(^{*}\)) directly regulates the result of the person attempting to firecast being kicked, whereas (1\(^{*}\)) does indirectly by changing the concept of “firecasting.” Yet these two ways of amending the rules are equivalent; both of the above rule changes would have the effect that those who were kicked receive a medal. The change can be made in distinct ways, as, in the context of enforcement of such rules, the rules of Firecasting constitute a set of premises that can be amended in distinct ways so as to, in conjunction with judgments about an instance of the game, imply the statement that some player who has been kicked should receive a medal. In other words, the aim of the village, when amending the rules of Firecasting, is to appropriately link the following two statements concerning some specific instance of the game.

(3) is a specific judgment concerning some specific instance of the game of Firecasting, and (4) is the regulatory response to what was adjudged to have happened in some such specific instance of the game. The village aims to formulate rules that, in conjunction with (3), imply (4). The combination of (1\(^{*}\)), (2), and (3) implies (4), and the combination of (2\(^{*}\)) and (3) also implies (4). In this way, the fact that the rules can be amended in distinct ways reflects no more than the fact that the same conclusion can follow from distinct sets of premises.

In the above case, the first way of changing the rules amounted to changing the definition of “firecasting,” whereas the second amounted to changing the statement of rewards given out by the village. The regulatory equivalence of these changes in our toy example is a phenomenon that applies to law in general. When we wish to amend the law in order to secure a specific consequence, there will always be distinct ways of doing so, and the only criteria for choosing whether to amend a definition or amend some statement of penalties or awards is, where rational, pragmatic.

Legal language turns out to be holistic in an almost Quinean way (Quine 1951). The law is holistic in two distinct ways. First, there is no one correct way to change the law so as to secure some regulatory response. Second, the list of claims we call definitions has no special status that prevents them from being changed so as to secure the desired regulatory response.

The fact that the law is holistic serves to explain why the law often uses perfectly ordinary terms in peculiar ways. The claim that a company is a “person” is just a tool to secure a regulatory response concerning the legal liability of the members of a corporation; the claim that ketchup is a “vegetable” is a tool to effectively lower the legally mandated nutritional requirements for school lunches. In the same way, the claim that an arm bent 15 degrees is “straight” is a tool to secure the result that cricketers may bowl with a slightly bent arm.

Cases of atypical use of terms like “ketchup,” “person,” “straight,” and the like serve to demonstrate something important. When the lawmaker changes the law, it has no overriding reason to respect the semantics of the term (as used in non-legal contexts). Rather, specific statements only matter inasmuch as they help to, in conjunction with other statements, secure the desired regulatory response when the law is applied. Such regulatory responses in specific instances can be represented as the conclusions of arguments that have legal statements among their premises. As the overriding factor governing the formulation and emendation of laws is the regulatory response to which it gives rise, this implies that the overriding factor governing the emendation of statements within a system of law is the role of such statements in facilitating inference. This, in turn, implies that we can expect changes to the law to end up radically changing the denotation of terms that were originally used in a perfectly ordinary sense. In the final analysis, this is due to the fact that institutional judgments are amended in virtue of their inferential (syntactic) properties and not their descriptive (semantic) properties.

The first peculiarity of language that was noted was that constrained revisability meant that the denotation of the regulative use of a term would not exactly coincide with the denotation of the descriptive condition based on which the term is applied. The second problem, however, is much more basic and would apply even if judges never made mistakes. Lawmakers will change the content of perfectly ordinary terms in order to secure regulatory consequences. This implies that the legal system will tend towards a scheme of individuation designed to serve regulatory and not descriptive purposes. This much is obvious enough, but it has the less commonly understood consequence that institutional judgments will be amended based on their syntactic properties and not on their semantic properties.

This implies that technical terms introduced for some regulative purpose (like “firecasting”) are not constrained so as to include only relevantly similar elements under their denotation. Furthermore, even when the regulative term is taken from ordinary language (like “straight,” “vegetable,” etc.), the term will tend to start out being used as legal terms in their familiar sense but will often end up including unlike objects in the same category. In this way, “regulatory kinds” will end up, if judged against a standard of descriptive adequacy, becoming disjunctive kinds that are ill-suited to scientific theorizing.

No one would expect mathematicians to do useful work while treating an arm bent 15 degrees as straight, and no one would expect a botanist to employ the term “vegetable” so as to include ketchup. These cases, however, are just the tip of the iceberg that serve to make the general phenomenon visible. The physical sciences pay no attention to regulative bodies when individuating the world, as such bodies are simply involved in another kind of activity altogether. In the same way, there is no reason for the social scientist to consider herself uniquely encumbered by, and beholden to, a schema of individuation that does not serve her purposes.

Note that the point concerning ordinary terms being introduced into law is not merely that such regulatory terms “change their meaning.” The problem, rather, is that such changes occur due to inferential (syntactic) considerations. Legal changes may be phrased as changes in definition (as when the definition of “firecasting” is amended so as to include being kicked) or as changes in regulation (such as when those kicked during firecasting also receive a reward). Whether these changes are phrased as one or the other change has little, if anything, to do with descriptive adequacy, and so we end up with categories that mix unlike things together, i.e., “regulatory kinds” become disjunctive kinds.

To illustrate the above point, consider the difference between the descriptive term “computer” and the regulative term “king.” The term “computer” was originally used to denote people, specifically those employed to engage in tedious tasks of rote calculation.¹⁴ The project to mechanize such tasks was originally described as the project of creating a “mechanical computer,” and this description was no mere tautology. Once the project succeeded, however, and the human computers disappeared, the meaning of the term “computer” underwent a social shift until it denoted only machines designed to perform such calculations. In fact, the change in the use of the term exhibits a nice symmetry; today, if we call someone a “computer,” it is a metaphorical use of the term that suggests extreme proficiency at calculation, a tendency to act without emotion, or some such.

The above change is generally socially recognized as a fundamental change in the meaning of the term “computer.” In principle, we could have treated the shift differently, for instance, by saying that the term merely expanded its denotation so as to include both human and mechanical, and eventually electronic, computers. There would be no point in doing so, however, for then “computer” would be a disjunctive kind that groups two radically distinct kinds of things together. Our descriptive language is guided by descriptive adequacy, and so simply treating the content of the term as having changed completely individuates the world in a much more useful way.

The same is not true for the institutional pair “king”/“queen” when used in a regulative manner. Kings and queens, historically, are paradigmatic persons who rule a state by decree and who obtain their position by right of birth. Today, however, in the vast majority of countries that still recognize a “king” or “queen,” being a king or queen is primarily a symbolic or ceremonial role. While today’s kings and queens do have some influence, this influence is so different in kind from the right to rule by decree that the two kinds of “king” or “queen” are beyond any meaningful similarity or comparison. While we may loosely say that “the meaning of being a king or queen has changed,” we do not generally consider the term to have changed its semantic content in the same way that the term “computer” has. This is despite the fact that the term categorizes together entities with vastly different social roles. If a historian or sociologist were to uncritically accept the institutional use of the term “king” (or “queen”) and try to determine commonalities or differences between kings, the very category of analysis would serve to unnecessarily complicate the inquiry. The term would group together those who ruled by decree as well as those whose social role is effectively a more dignified version of a mascot. If our purpose is descriptive adequacy, then little is to be gained by an individuation scheme that treats those who ruled by decree (old-style kings, the present-day King of Swaziland, etc.) with the current Queen of England or the current Queen of Denmark. Furthermore, it would exclude those whose social role is similar to that of old-style kings and queens, i.e., dictators who de facto rule by decree and have their position in virtue of birth, e.g., the North Korean leader Kim Jong-Un.

The point of the above is not to criticize the existence of present-day royalty or to suggest a change in linguistic habit. The point, rather, is that a social scientist who accepts an individuation scheme in which an all-powerful king and the current Queen of Denmark are the same sort of thing¹⁵ is as absurd as a mathematician who treats all lines that bend less than 15 degrees as “straight.”

The above considerations concern both cross-institutional identification (i.e., whether the current King of Denmark and the current King of Swaziland are the same sort of thing) and inter-temporal identification (i.e., whether the kings and queens of centuries ago are the same sorts of things as the current Queen of England). It is a fundamental constraint upon inquiry that our criteria of individuation remain constant, and here, the regulative use of institutional terms is a poor guide to scientific individuating practices. It is for this reason that the social scientist should feel under no obligation to accept institutional standards of individuation; in fact, she should rearrange the conceptual world as she sees fit.¹⁶

3 Cases Where the Regulative Use of Institutional Terms Is the Correct Use

The point of the above is not that social scientists should never employ the regulative use of institutional terms as basic terms of inquiry. Institutions do manage to affect the world via declarations, and Searle is correct that this phenomenon is important to understanding our social and institutional world. We can distinguish three reasons for adopting the regulative use as a term of inquiry.

First, our interest may lie precisely in the objects grouped together by the regulative use of an institutional term. In this way, as mentioned earlier, we may wish to inquire into the difference between those who are convicted of committing a crime and those who, while having committed a crime, are acquitted. Or, alternatively, we may be interested in the difference in severity of sentence among those convicted of a crime. Such topics are a staple of criminological and sociological studies that try and determine what effect categories of identity (race, gender, etc.) or socioeconomic attainment have on rates of conviction and severity of sentence. In such cases, our interest lies precisely in a category that exists in virtue of regulative declarations; hence, the regulative category is proper to the study.¹⁷

Second, the declarations made by institutions have a causal impact on the world, and our interest may lie precisely in studying the effect of such an impact. In this way, the criminologist may be interested precisely in the impact of a criminal conviction on one’s life prospects. In this case, again, the regulative use is proper to the study in virtue of the causal role of such regulative judgments.

An interesting sub-class of the causal impact that institutional declarations can have is where such declarations have a symbolic impact on the objects of such a declaration; we may well be interested in studying the nature and effects of such a symbolic impact. In this way, a historian or sociologist may be interested in changes in self-conception that occur among those people occupying a territory that is widely recognized as being a “country,” or changes in self-conception among those recognized as “criminals,” and so on.¹⁸

Third, the social scientist may be interested in a category that need not be governed by regulative use, but the institutional use is close enough, for the purposes of the study, to what they are trying to identify that it is a useful proxy for the descriptive use. If a country’s rules remain relatively stable over time, the judiciary does a decent job of applying the laws, and the concepts involved happen to individuate reality in a descriptively useful way, the institutional category should be good enough for useful inquiry. Consider, for instance, a scientist who wishes to study whether the color of a motor vehicle has an impact on people’s propensity to speed. Strictly speaking, some people who speed will not be among those convicted of speeding, whereas some of those convicted will have been innocent. But if all the scientist is looking for is a rough correlation in aggregate and the legal system has been reasonably efficient, then counting all those convicted of speeding as “speeders” should be a good enough sample to do meaningful statistical work.

In endorsing the above regulative uses, we also embrace something close to pluralism about general institutional terms. Good usage will be polysemic; the social scientist will inevitably have to craft the terms of their inquiry to the topic at hand. What we object to, on the grounds discussed, is the idea that regulative bodies should be implicitly granted the power to set the terms of our descriptive agenda.¹⁹

4 Is Somaliland a Country?

Somaliland is a country. More specifically, we think that, except in the very specific types of cases previously explained, i.e., cases where our epistemic interest is precisely in those objects grouped together by institutional declarations, the social scientist should view Somaliland as a country. Here, we do not base this claim on any specific definition of the term “country.” Rather, our judgment reflects the fact that Somaliland, once we ignore regulatory schemes of individuation for the reasons outlined in this paper, seems entirely like paradigm cases of countries, i.e., Kenya, Germany, Chile, Japan, etc.

In this paper, we have explained why we think that the social scientist should, in principle, be very wary of adopting institutional schemes of individuation. This matters, as the currently dominant theory of institutions, i.e., the Searlean theory, effectively adopts and legitimizes institutional schemes of individuation, and, hence, it is worth knowing why the social scientist should feel free to disregard Searle’s view. This is so, especially as Searle explicitly recommends that social science adopt his theory of institutions (Searle 2005). Also note that reflexive definitions, i.e., definitions on which an entity gains its identity from being considered to be the things that it is, long predate Searle.

More important, however, is the question as to the scope of the problem, i.e., the question of how much harm is done by social scientists adopting regulatory schemes of individuation. Our argument is compatible with quietism about regulatory individuation, i.e., the view that Somaliland is an edge case, a mere curiosity whose exclusion from the list of countries does no real harm to social analysis. Our view is also compatible with revisionism about regulatory individuation, i.e., the view that Somaliland and cases like it serve to make visible a deep problem that calls for social scientists to abandon regulatory schemes of individuation in favor of a series of successor concepts more suited to descriptive purposes.

The question of which position on the continuum between quietism and revisionism is most justifiable is beyond the ambition of the present work. We can see the appeal of quietism; it would appear ridiculous to expect an economist writing about the correlation between countries in the measured link between inflation and unemployment to worry too much about whether his fundamental categories of analysis are making his job harder than it needs to be.

We can also, however, see the appeal of revisionism. Consider the definition below, intended to capture the regulatory notion of a “country”:²⁰

The above definition—in addition to being vague—is disjunctive to an extreme degree. It is akin to a definition of “vegetable” that includes not only ketchup but also all bottles of Worcestershire sauce that are older than three months. The problem with such a disjunction is plain: What possible reason could we have to expect that some underlying causal process could produce similar effects across entities that have been grouped together merely as a matter of a series of historical regulatory contingencies?

Current practice seems to imply at least some deviation from quietism. Social scientists are not naive and have not stayed slavishly faithful to institutional categories of individuation. The CIA World Factbook, for example, on its list of countries by gross domestic product, does not list England, Scotland, or Wales among the entries, even though they are generally recognized as countries.²¹ It does, however, list the European Union, despite the fact that it is not recognized as being a country. This makes sense, as the interest of the economist would be in finding a category of individuation that identifies individual units of action, i.e., units with a fair degree of autonomy qua matters of economic production and exchange. When it comes to such practices, the present paper serves to justify how such deviations from institutional schemes of individuation are, contra Searle, perfectly justified.²²

The question, however, is whether current practice occupies the appropriate position on the continuum between revisionism and quietism. Note that the CIA World Factbook does not list Somaliland, despite there being very little reason not to do so once we abandon the purely regulatory use of “country.” In fact, once we take the matter of individuation seriously, we may well have reason to include sub-units of various “countries” under the de facto control of some entity other than the recognized government, i.e., parts of “countries” under the control of rebel groups or drug cartels.²³ This may sound radical, but if our interest lies in discovering the units of political and/or economic action—and hence in groups that have a high degree of autonomy over running their own affairs—then there is little reason to exclude them. We may well learn interesting things by considering such entities qua units of economic and political action, for they are effectively no different from “countries” under military or dictatorial control.

5 Conclusion

In this paper, we have argued that social scientists should not be Searleans when it comes to their own categories of analysis, i.e., be wary of employing the regulative use of institutional terms for purposes of individuation. There are two main problems. First, the revisability of institutional judgments is non-epistemically constrained, i.e., mistakes do not get corrected in the same way that we correct them when dealing with descriptive assertions. This means that the social scientist would frequently be better served by employing the descriptive use, and not the regulative use, of institutional terms as a basis of individuation. The second problem, and by far the most important one, is due to the fact that institutions individuate in order to regulate, not to describe. Such regulation is holistic, and, hence, the usage of terms will change based on their syntactic properties and in response to regulatory pressure, not based on their semantic properties and in response to matters of descriptive adequacy.²⁴ This means that “regulative kinds” will inevitably tend to become disjunctive kinds, and so the law will be prone to the seeming absurdity of classifying ketchup as a vegetable, considering an arm bent 15 degrees to be straight, and so on. This implies that the social scientist will sometimes be better advised to ignore both the descriptive and regulative use of institutional terms and to invent institutional categories that have never been subject to regulative declaration at all.