The discovery of meaning in legal texts is one of the core competencies of law and society scholarship. How legal text becomes socially efficacious, how it admits of multiple symbolic and instrumental meanings for different audiences, how those meanings eventually become settled (or do not)—these questions are asked and answered continuously in contemporary law and society work, in empirical domains ranging from the enforcement of hate crime law (Phillips and Grattet Reference Phillips and Grattet2000; Grattet and Jenness Reference Grattet and Jenness2005, Reference Grattet and Jenness2008) and the historical development of affirmative action policies (Skrentny Reference Skrentny1996) to international nuclear nonproliferation treaties (Mallard Reference Mallard2014). Ambitious attempts to develop something like a general theory of meaning in legal language (e.g., Constable Reference Constable2014) still attract serious attention.

Within this discursive world, however, there is not much systematic attention to appellate court judicial decisions, despite the expressly interpretive character of judicial decisions and their strong hold on the public interest (the strong public interest, at least, in the steady stream of highly politicized cases decided by the US Supreme Court). Phillips and Grattet's (2000) close reading of thirty‐eight appellate court decisions on the constitutionality of hate crime law is a notable exception. There does exist a large body of “behaviorist” studies of judicial decisions, including the books under review by Posner and by Epstein, Landes, and Posner, but because of divergent methodological approaches and divergent explanatory ambitions, the scholarship on judicial behavior occupies an intellectual world that overlaps relatively little with law and society. The goal of this essay is to clarify the nature of these two divergences and to point out what more could be said about the social production of judicial decisionsFootnote ¹ than what the judicial behaviorists are currently saying.

Research into the social production of judicial decisions could benefit from an expanded repertoire of evidentiary sources and corresponding analytical tools as well as a reconsideration of theoretical premises, and law and society scholars are well positioned to make decisive and constructive interventions in this line. I will also argue that my own field of the sociology of knowledge can provide important resources for this research agenda. Although sociologists of knowledge have hitherto paid little or no attention to judicial decisions as data, their analyses of knowledge work in other settings provides them with a highly relevant and flexible set of analytical tools.

Sociologists of knowledge tend to employ nominalist rather than essentialist definitions of “knowledge,” “knowledge production,” and “knowledge work.” “Expert knowledge” is what scientists, artists, professionals, political ideologists, and the like produce when they advance propositions in their respective domains of expertise. The outcomes of this knowledge work are generally judged by sociologists by heuristic criteria—knowledge work is effective if it wins recognition or acceptance, or if it helps its producers and consumers to navigate the world. For sociologists studying knowledge production in physics, for example (e.g., Collins Reference Collins1975; Pickering Reference Pickering1984; Knorr Cetina Reference Knorr Cetina1999), it matters not at all whether the physicists have uncovered the true nature of reality. What matters for sociological analysis is our ability to follow the internal logic, the heuristic utility, and the practical and theoretical implications of their expert knowledge claims.

I adopt a nominalist approach here. I regard appellate judges as knowledge workers, which is to say that I regard them as advancing interpretive claims about the law that are neither purely mechanical nor purely instrumental. In the course of this essay I also argue that there is good evidence that judges understand themselves as knowledge workers, and that this self‐understanding ought to figure into our explanations of the decisions they produce. The normative question of how judges ought to interpret the law, in general or in particular, is beside the point.

The foundation for much of the contemporary research on judicial behavior, including Epstein, Landes, and Posner's Behavior of Federal Judges, is a set of large datasets: the comprehensive Spaeth Supreme Court database (Spaeth et al. Reference Spaeth, Benesh, Epstein, Martin, Segal and Ruger2013), the Songer courts of appeals database (Songer Reference Songer1998; Kuersten and Haire Reference Kuersten and Haire2007), and the Sunstein courts of appeals and district courts database (Sunstein et al. Reference Sunstein, Schkade, Ellman and Sawicki2006). Epstein, Landes, and Posner also rely on four additional datasets that provide partial coverage of district court decisions, focused in particular on sentencing and civil procedure, and they rely on these datasets to the near‐total exclusion of other empirical sources. These datasets encode information about a huge number of US appellate and trial court cases, and they formally operationalize features of those cases (by, e.g., taking the party of the president who nominated a given judge to the bench as a proxy measure for judge ideology) in ways that allow for the construction of formal judicial utility functions that can in turn be fit to the data.

This statistical modeling approach, which is frequently accompanied by a rational choice theoretical framework, is broadly characteristic of judicial behaviorist research in law, political science, and economics. It has revealed many highly suggestive patterns. Among the notable empirical conclusions in Epstein, Landes, and Posner are that “dissent aversion” is stronger in the courts of appeals than in the Supreme Court (255ff., 279);Footnote ² that apparently ideological voting is much less pronounced in the courts of appeals than in the Supreme Court (189); and that dissents have a jurisprudential impact in high impact cases—specifically, “the more frequently the majority opinion is cited, the more frequently the dissent is cited” (289). The authors find statistically significant differences between the voting records of conservative and moderate judges and between the voting records of moderate and liberal judges in civil liberties cases, economic and labor law, criminal law and constitutional law, consistently over time (115, 165).Footnote ³

The behaviorist datasets, however, need not be taken as the exclusive sources of useful evidence when we seek to explain judicial decision making and judicial decisions. Several sources of data are mostly neglected by behaviorist scholarship—I discuss three broad categories of sources here. In the exploration of these auxiliary sources, and in the refinement of methodological and theoretical approaches appropriate to the available data, there is great potential for growth in this area. I argue, accordingly, for a closer engagement from law and society with judicial decisions as data and with the existing behaviorist scholarship. When empirical examples are called for in the discussion below, I draw on my own ongoing research into judicial decision making in the realm of political asylum law.

A first broad evidentiary source for this agenda is data on the structure, functioning, and culture of courts and the legal profession as institutions. Organizational analyses of courts are already plentiful (e.g., McCloskey Reference McCloskey1960; Shapiro Reference Shapiro1981; Ginsburg Reference Ginsburg2003; Rosenberg Reference Rosenberg2008) and could be read for insights into how institutional context informs legal reasoning of judges (or of other actors in the legal system). These legal organizational analyses have not focused on knowledge production as the outcome of primary interest, but we can look to the sociology of knowledge for guidance on how to give substantive meaning to the rough idea that institutional structure, function, and culture can affect knowledge outcomes. To note just two highly successful examples, Fourcade (Reference Fourcade2009, 38–40) reviews the nineteenth‐century emergence of the modern research university to frame her international comparative analysis of the discipline of economics in the twentieth century; Abbott (Reference Abbott1988, 280–314) looks back to the inchoate cultural awareness of “personal problems” in 1850 to ground his account of the eventual emergence of professional psychiatry. Fourcade and Abbott successfully mobilize sociological understandings of the path dependence of organizational developments, the interplay of structure and function, and the exogenous influences of competition and political culture to explain how two bodies of professional knowledge—academic economics and professional psychiatry—came into being.

The judicial behaviorist data allow us to identify some suggestive patterns of institutional variance in the federal courts. Examples include (1) variance in decision‐making tendencies between the several appellate circuit courts that exceeds the within‐circuit variance and (2) greater ideological divisions in the US Supreme Court than in the lower courts. But the standard behaviorist datasets alone cannot sustain very complex arguments about institutional culture or functioning. They provide us no information, for example, about the structure of communication and information flow within the courts, which may vary by judge and by court, or about the role of lawyers in influencing case outcomes, which since Galanter's foundational work (1974) has been of central interest to law and society scholars. When Epstein, Landes, and Posner do discuss institutional structures, they treat them rather narrowly as responses to economic inefficiencies: for example, the system of appointing the longest‐serving judge under sixty‐five as the chief judge is described as “[trading] off competence against the danger of infighting” (Epstein, Landes, and Posner Reference Epstein, Landes and Posner2013, 32).

Second, research into the social production of judicial decisions could benefit from a close attention to rhetorical meaning making in the texts of judicial decisions themselves. I have already mentioned this as a core competency of law and society scholarship. Sociologists of knowledge have likewise paid close attention to language use—particularly the uses of nonliteral language and the social efficacy of so‐called speech actsFootnote ⁴ —to great effect. They have found that successful knowledge work often depends on highly particularistic language and invokes cultural understandings that are only narrowly shared and understood. Karin Knorr Cetina's (Reference Knorr Cetina1999) and Bruno Latour's (Reference Latour1987) ethnographies of scientific laboratories point out the allusive, metaphorical language regularly employed by scientists in the working out of precise quantitative results: a biologist complains of a “bad gut” and discards the specimen (Latour Reference Latour1987, 66); high energy physicists rely on symbolic language to interpret the signs and traces of events that cannot be directly observed (Knorr Cetina Reference Knorr Cetina1999, 49–50) and speak of their technical instruments as if they were conscious agents—as “looking,” “watching,” and sometimes “blind” (114).

These sociological observations on the role of language use in knowledge work are clearly applicable to judicial decisions as data. Judicial decisions are paradigmatic examples of speech acts, and they can be at the same time complex rhetorical documents. Most of the time, judicial decisions are written in response to narrow and clearly defined questions of law, but there is no second‐order rule (or “rule of recognition,” to borrow the language of legal theorist H. L. A. Hart [Reference Hart1961]) that dictates how federal judges will approach the interpretation of legal rules. In some areas of law, the outcome is a bewildering variety of approaches and reliances on different sources of law from case to case. In political asylum decisions, we sometimes find judges making recourse to “practical realities” over and against “semantic niceties” (see the dissent by Justice Powell in INS v. Cardoza‐Fonseca 1987) and sometimes insisting that such practical considerations are legally unsupportable. In other instances, judges focus on the importance of methodological consistency in their own construction of concepts, or on the question of Chevron deference,Footnote ⁵ or on international legal norms or on the adequacy of the factual information available to make a determination. In still other instances, they frame their legal reasoning with reference to standing and procedural questions or to the constraints imposed by common law precedent.

Understanding how legal language “acts,” both in its technical function as speech act and through the rhetorical framing that makes a complex argument either persuasive or unpersuasive, is a problem basic to the understanding of the social production of judicial decisions. Sociologists and nonsociologists alike have written on the social efficacy of legal language (e.g., Matoesian Reference Matoesian2001; Mertz Reference Mertz2007; Ng Reference Ng2009; Constable Reference Constable2014), although usually not with explicit reference to the central sociology of knowledge problem—how does legal language affect legal judgment and other forms of knowledge work within the legal system? It is possible to integrate key insights into the social efficacy of language use into broader institutional or cultural analyses of the social production of ideas, as in Knorr Cetina's comparative ethnography of science laboratories (see also Collins [Reference Collins1975] on the linked challenges of communicating results and facilitating replication in scientific experiments). This work can and should be done with reference to judicial decisions and courts.

Third, research into judicial decision making could benefit from attention to the broad array of methodological approaches—textual interpretation, interviews, ethnography, and laboratory experiments—that can provide clues to the relationship between the self‐understandings of judges and the processual, phenomenological experience of judging, on the one hand, and the production of written judicial decisions that have the force of law, on the other hand. Sociologists of knowledge have tackled this problem of linking the self‐understandings of individual actors to knowledge production in, for example, Randall Collins's sprawling Sociology of Philosophies (1998), and on a much smaller scale, Charles Thorpe's “sociological biography” of Robert Oppenheimer (2006) and Neil Gross's account of philosopher Richard Rorty's early career (2008). Thorpe and Gross both rely primarily on archival and other textual evidence, modeling a method that analyses of judicial decisions could follow.

There is good evidence that something like “intellectual self‐concept” (Gross Reference Gross2008) plays a role in the framing of some judicial decisions. In political asylum cases, we can readily find judges signaling moral commitments that are irrelevant to the law. In the Tenth Circuit case of Nalwamba v. Holder (2010), Judge Henry's separate concurrence stresses the point that “had I been the hearing judge in this matter, I believe I would have calculated differently,” apparently lamenting the deferential standard of review in the federal courts required by the law. In a case before the Federal Court of Canada, a decision against a sympathetic appellant (a stateless Palestinian) was made “with no joy whatever” (Elastal v. Canada (Minister of Citizenship and Immigration) 1999). We find judges expressing irritation in their decisions when political actors speak publicly on a case and so preemptively stain the judiciary with the implication of political motive.Footnote ⁶ We find some judges proclaiming their commitments to certain interpretive standards, sometimes even by impugning the competence or integrity of their fellow judges (see, e.g., Justice Scalia's decision in INS v. Elias‐Zacarias (1992) overturning a Ninth Circuit decision [Elias‐Zacarias v. U.S. INS 1990]). In Elias‐Zacarias, Scalia declared the Ninth Circuit's reasoning in part “irrelevant” and in part “untrue.” In the long term, this kind of adversarial stance undermines the judiciary's claim to be disinterested interpreters of the law, presumably an undesirable outcome from the perspective of any professional judge. Judges who act this way may be privileging their ideological purity, their sense of self‐importance, or their intense desire to model the right way to decide cases for misguided colleagues, above considerations of collegiality or even long‐term strategic interest.Footnote ⁷

I am not proposing that we should appropriate judges’ self‐accounts as a true model of their decision‐making procedures. For one thing, judges are often not the sole authors of their opinions (clerks can play a large role) and, as Posner has written, “cats are not consulted on the principles of feline psychology” (2008, 2). Nonetheless, what judges think they are doing when they author decisions, and what they wish the public to think they are doing, can be interpreted as evidence for what they are actually doing. This is particularly true if we focus narrowly on explicit disputes over questions of law that appear within and between cases. Judges hearing cases are required to issue an opinion, so the issuance of an opinion does not necessarily signal a judgment that might, under different circumstances, have turned out differently. But neither ideology nor professional obligation can fully account for which cases judges will dispose of with simple memorandum dispositions and which ones will prompt them to write long opinions, which can be by turns impassioned, exhaustively analytic, speculative, moralistic, combative, and so forth. The straightforward and convincing explanation is that they are sometimes engaged in genuine intellectual disagreement about the application of legal rules. When we find judges raising explicit disputes with one another over questions of law (as opposed to merely issuing opinions at different points on an ideological scale), the sociology of knowledge may provide the best framework to make sense of their activity.

Cognitive anthropologists and cognitive psychologists have made complementary contributions to our understanding of the phenomenological process of judgment and decision making. The anthropologists have observed judges in action but have not had the ability to test hypothesized cognitive mechanisms under experimental conditions. A foundational example is Hutchins (Reference Hutchins1980), whose ethnographic analysis of Trobriand land dispute settlement made persuasive reference to laboratory experiments (Wason and Johnson‐Laird Reference Wason and Johnson‐Laird1972) to show that logical inferences are made easier when they can be performed with culturally familiar objects arranged in culturally meaningful ways (Hutchins Reference Hutchins1980, 10–11). Cognitive psychologists, by contrast, have been able to conduct controlled laboratory experiments (e.g., Simon and Scurich Reference Simon and Scurich2011) but without professional judges as subjects, and without being able to re‐create fully in the laboratory the remarkable social import of judicial decisions as speech acts. Psychological studies of legal judgment and decision making tend to focus on jury rather than judicial decision making (e.g., Hastie Reference Hastie1994; Connolly, Arkes, and Hammond Reference Connolly, Arkes and Hammond2000, 197–240; Dhami et al. Reference Dhami, Hastie, Koehler and Wiener2007), likely because it is much easier to model adequately the situation of a juror than a judge in an experimental setting.

There are, then, many sources and associated analytical tools that could be brought to bear on the explanation of judicial decisions and judicial decision making but are at present underutilized in the dominant behaviorist scholarship. I hope that many law and society scholars and sociologists of knowledge—particularly those who work on textual interpretation or on how individuals understand their own identities and social roles in relation to the law—will already see promise in these lines of potential inquiry. It remains a challenge to explain why the behaviorist research has not already adopted a broader methodological and empirical base, and in turn to assess what would be gained and lost in a move to greater evidentiary and analytical pluralism in studies of the social production of judicial decisions.

This problem manifests even within the behaviorist literature. Posner begins How Judges Think with an explicit reflection on the question of what a complete explanation of the social conditioning of judicial decision making would have to include, and his answer, which Epstein, Landes, and Posner expressly adopt as their interpretive theoretical framework in Behavior of Federal Judges (25), seemingly aligns with the multipronged research agenda I have outlined. Posner delineates nine dimensions of explanation: the attitudinal, the strategic, the sociological, the psychological, the economic, the organizational, the pragmatic, the phenomenological, and the legalist. He presents this as a list of “nine theories” but makes it clear that they are not mutually exclusive and, indeed, he argues that there are many overlaps and that all the theories will be relevant to some degree (Posner Reference Posner2008, 19). His framework has much to recommend it as a set of organizing propositions to guide empirical research. There is nothing obviously left out,Footnote ⁸ and the framework matches up well with the ways that sociologists of knowledge have collectively described the social conditioning of knowledge production. Posner does not make any reference to the sociology of knowledge, so we have, encouragingly, a kind of independent attestation of the likely social sources of knowledge from Posner, on the one hand, and the field of sociology of knowledge, on the other.

Posner and Epstein, Landes, and Posner give more definite shape to their explanatory framework with the proposal that the nine theories can be “integrate[d] … into a single theory, that of the judge as a participant in a labor market—that is, as a worker” (Epstein, Landes, and Posner Reference Epstein, Landes and Posner2013, 25). My conceptualization of judges as knowledge workers—that is, as workers whose work is neither merely deductive nor entirely instrumental, for whom inferential reasoning and judgment mediate between their initial situation and the output (judicial decisions) that they produce—finds support from Posner and Epstein, Landes, and Posner in some passages, particularly when they stress the conceptual impossibility of judge‐made law being completely rule determined (e.g., Posner Reference Posner2008, 7–9; Epstein, Landes, and Posner Reference Epstein, Landes and Posner2013, 44–45). But their gambit in positing the judge as participant in a labor market narrows the scope of the thing to be explained much more dramatically than does my focus on knowledge work. The authors of both volumes axiomatically assume a rational actor model. Epstein, Landes, and Posner, working with systematic empirical data, present judges’ preferences in the form of utility functions. In How Judges Think, Posner does not systematically analyze any empirical data, but he relies on the language and heuristics of rational choice theory throughout.

If we probe the rationale for Epstein, Landes, and Posner's choices of data and methodology (statistical datasets and utility function modeling), we ultimately come to an epistemological justification. Posner, on the basis of many years of service writing appellate court decisions on the Court of Appeals for the Seventh Circuit, argues consistently that the rhetoric of judicial decisions tends to conceal rather than reveal the true motivation and reasoning behind them. This position is wholly characteristic of judicial behavior scholarship more broadly (for a literature review that emphasizes just this point, see Bybee Reference Bybee2012, 70ff). Furthermore, the version of rational choice theory that Posner maintains takes no account of the self‐descriptions or self‐understandings of actors: “rationality means little more to an economist than a disposition to choose, consciously or unconsciously, an apt means to whatever ends the chooser happens to have selected, consciously or unconsciously … it would not be a solecism to speak of a rational frog” (Posner Reference Posner2007, 15). In light of this epistemological positioning—or to put it in slightly different terms, this theorizing about what would count as an adequate explanation of judicial decision making—Posner's and his coauthors’ decision to eschew any close textual analysis, analysis of institutional culture, or any systematic interest in the self‐understandings of judgesFootnote ⁹ seems to be a positive and categorical choice rather than merely a matter of emphasis.

Here the sharp edges of the epistemological question emerge and provide a key to understanding the communicative breakdown between law and society scholars and the judicial behaviorists. I tend to think that a detailed answer to the epistemological question—what would count as an adequate explanation of judicial decision making?—matters fundamentally to the methodological question—how should we organize empirical research in pursuit of a complete (or, as complete as possible) explanation of the social conditioning of judicial decision making? Others may consider the first concern more easily avoidable for the purposes of practical research.

Posner's and Epstein, Landes, and Posner's integrated theory of judicial behavior places its emphasis on the economic (which subsumes the attitudinal, the sociological, and parts of the psychological) and a version of the pragmatic (which subsumes the legalist in certain instances) dimensions of explanation. In one respect this does look like a genuine integration of the chaotically diverse explanatory dimensions initially posited: political ideology (the focus of the attitudinal theory) really can be modeled in a utility function (the framework of the economic theory), as can “dissent aversion” and other consequences of small‐group dynamics (the focus of the sociological theory). Incomplete information and cognitive limitations (the core of the psychological theory) “can be modeled as costs of processing information” (Posner Reference Posner2008, 37). It is right to say that a judicial decision that treats law as determinate and deductively applied in the courts (the legalist view) is sometimes the best practical strategy for a judge to adopt. Finally, practical designs and strategies for achieving them can be understood in a utility‐maximization framework, so Posner ultimately claims that the nine theories are wholly commensurate when expressed in terms of an individual judge's utility function. There is an admirable elegance and formal clarity to this explanatory gambit, but it is deceptive when presented as a genuine unification of the nine dimensions outlined at the start.

Specifically, the assumption of a rational actor model precludes the possibility of genuine unification of all that the nine dimensions of Posner's framework have to offer, although Posner and Epstein, Landes, and Posner do not consider all the constraints imposed by their theoretical assumptions. The pragmatist and the phenomenological theories are the most severely diminished in Posner's combinatorial exercise, and I discuss them in turn below.

Posner glosses pragmatism as “an extension of the scientific method into all areas of inquiry” (2008, 231) and as a school of thought that bears a “family resemblance to utilitarianism” on account of its instrumental focus on outcomes (Posner Reference Posner2008, 40). Epstein, Landes, and Posner only touch on pragmatism obliquely, but they follow Posner's lead in treating the term as a stand‐in for instrumental practicality: “shortsighted pragmatism … inflects the consequences for the [political] parties with the judge's political preferences or personal sentiments” (Epstein, Landes, and Posner Reference Epstein, Landes and Posner2013, 28). Posner and Epstein, Landes, and Posner do not recognize pragmatism as the basis of a theory of meaning making through social action as developed by Dewey (Reference Dewey1930), Mead ([1934] Reference Mead1967), Joas (Reference Joas1993, Reference Joas1996), and the symbolic interactionists (e.g., Goffman Reference Goffman1974). This is not merely a matter of definitional difference—Posner cites the names of Peirce, James, and Dewey to indicate that he intends to invoke the highest philosophical meaning of pragmatism (Posner Reference Posner2008, 231). Epstein, Landes, and Posner contribute to the same flattening of philosophical differences when they identify Bentham's Introduction to the Principles of Morals and Legislation (1780) as “the first full articulation of the realist conception of judging” (Epstein, Landes, and Posner Reference Epstein, Landes and Posner2013, 27) and trace a line from Bentham to Oliver Wendell Holmes, Jr. without mention of the pragmatist influence on Holmes (cf. Menand Reference Menand2002) and by extension on the legal realists who followed in his footsteps.

The failure to recognize pragmatism as a distinctive theory of action and a profound critique of utilitarianism really matters. The major pragmatist philosophers and social theorists have all rejected the essential utilitarian premise that preferences are well defined in the abstract in such a way that they can straightforwardly guide action. The pragmatist perspective is, rather, that interpretation, continuously developed in concrete situations of lived experience, is crucial to the conduct and therefore to the explanation of social action. Some of the major pragmatists, especially James ([1902] Reference James2002) and Joas (Reference Joas2000), have further challenged the notion that the individual is wholly analytically separable from social forces—norms and values—that the individual internalizes in moments of self‐transcendence. By contrast, the clearly bounded individual whose identity persists over time is the indispensable unit of analysis in utilitarian thought. It is difficult to see how a pragmatist theory of judicial decision making could be reconciled with the basic epistemological assumptions of the behaviorists. The rational choice framework they adopt does not seem capable of adjudicating between the processual, interpretive meaning making that interests pragmatist philosophers and sociologists, on the one hand, from instrumental rationality, on the other.Footnote ¹⁰

The place of phenomenology is likewise problematic in Posner and Epstein, Landes, and Posner, although from omission rather than simplification and distortion. Whereas pragmatism's greatest relevance to social science is as an epistemology and a theory of action, the basic meaning of phenomenology—and the meaning that Posner apparently intends—is the designation of a level of analysis. It is of particular interest to cognitive anthropologists and psychologists who study decision making, as noted above, and there is at least one well‐known example of applied phenomenology in the realm of legal studies: Duncan Kennedy's essay, “Freedom and Constraint in Adjudication” (1986). Kennedy proposes a “phenomenology of adjudication,” which is to proceed by assessing what might happen when a judge is faced with an intuited gap between “how‐I‐want‐to‐come‐out” [sic] and “the law.” The subjective experience of judging is the key phenomenon to be explained. It is important to recognize that this explanandum is something that cannot be read from a utility function, so Kennedy's essay would seem to be an important foil to the behaviorist account of judge as labor market participant. Yet Posner does not return to the phenomenological perspective in How Judges Think after introducing its baseline possibility (although he has recently written a volume in part on his own subjective experience of judging; see Posner Reference Posner2013). Epstein, Landes, and Posner do not mention phenomenology at all. It is, therefore, difficult to know on what grounds Posner claims that the model of judge as rational labor market participant can successfully incorporate phenomenological insights.

Embracing analytical pluralism in the study of the social production of judicial decisions would mean giving up on some of the formal clarity and parsimony that appears in behaviorist scholarship. It is not immediately clear that the tradeoff would amount to substantive progress. Interpretive approaches to analyzing judicial decisions are not new, and often they resort to the well‐worn division between legalism and realism as a basic interpretive framework. This tendency is exemplified in a Annual Review of Law and Social Science by Keith Bybee (Reference Bybee2012), in which Bybee advocates a move in studies of judicial decision making toward “paying attention to what judges say.” Bybee makes common cause with my argument here, although he does not provide many resources to advance the constructive project of coordinating judges’ self‐understandings with cultural and institutional context in a single explanatory account. It does seem right that US judges at least sometimes think of their work and the decision‐making process in terms of a choice or balance between legalist and realist considerations.Footnote ¹¹ Bybee is also persuasive that the tension is productive in certain ways—and that “useful pretenses” are important in supporting the legitimacy of the US federal appellate courts. But to make legalism versus realism the central tension in a broad theory of judicial decision making is to sidestep or distort a great deal of complexity in how judges write and self‐present, especially beyond the borders of the United States.

Those judges who are most explicit and vocal about their interpretive approaches tend to adopt more elaborate models than a balancing of legalism and realism—textual originalism, the economic analysis of law or pragmatism, for example. Other judges have expressed interpretive affinity with legal theorists such as H. L. A. Hart or Ronald Dworkin.Footnote ¹² If the interpretive work of judicial decision making bears any relation to developments at the leading edge of legal theory, then legalism versus realism will be a poor model for understanding the substantive problems facing judges in the twentieth and twenty‐first centuries. Furthermore, the legalism versus realism division assumes a focus on the United States—with its highly politicized and politically powerful courts—and is of limited use elsewhere. Two major issues in contemporary judicial decision making are (1) the status of international law and (2) the appropriateness of borrowing legal reasoning from the court systems of other nations governed by different constitutional orders (see Slaughter Reference Slaughter2004; Black and Epstein Reference Black and Epstein2007). Debates over these issues play out both within and outside the United States. In political asylum law, the UN Handbook and Guidelines on Procedures and Criteria for Determining Refugee Status (UN High Commissioner for Refugees [1979] 2011) and interpretive guides to comparative jurisprudence (e.g., Goodwin‐Gill and McAdam Reference Goodwin‐Gill and McAdam2007; Hathaway and Foster Reference Hathaway and Foster2014) are regularly cited in judicial decisions. One of the few asylum cases to reach the Supreme Court included an extended discussion of the meaning of the French word refoulement and the legislative intent of the Netherlands’ delegate before the UN Convention.Footnote ¹³ These debates often do not manifest as problems of choosing between legalist and realist interpretive approaches.

Much like the various strands of legal realism, the sociology of knowledge has found analytical pluralism to be a stumbling block to collective progress. The subfield is methodologically pluralistic, as many areas of sociology are, and its boundaries are sometimes blurry (e.g., Marion Fourcade's work cited above might easily be reclassified as sociology of organizations). In general, sociologists of knowledge pose empirical instantiations of the general theoretical question: “Why did these particular ideas emerge in these particular times and places?” A persistent and somewhat discouraging feature of sociology of knowledge as a disciplinary endeavor has been that individual scholars faced with that same basic question have pitched their responses at quite different levels of analysis. They have treated knowledge production as the product of competition within an ecology (Abbott Reference Abbott1988; Bourdieu Reference Bourdieu1988), as the product of institutional culture (Camic Reference Camic1992; Knorr Cetina Reference Knorr Cetina1999; Lamont and Huutoniemi Reference Lamont, Huutoniemi, Camic, Gross and Lamont2011), national culture (Fourcade Reference Fourcade2009), local culture and micro‐level interaction (Collins Reference Collins1998), personal biography and self‐understanding (Gross Reference Gross2008), or position within a communication network (Knorr Cetina and Bruegger Reference Knorr Cetina and Bruegger2002; Knorr Cetina and Preda Reference Knorr Cetina and Preda2005; Vilhena et al. Reference Vilhena, Foster, Rosvall, West, Evans and Bergstrom2014).

Judicial decisions as data, however, are especially well suited to the adjudication between social effects on knowledge production at different levels. Appellate court judicial decisions—in the United States as in many other modern constitutional democracies—respond to narrow and clearly defined questions of law, they are strongly bounded in time (i.e., it is clear what is part of the decision and what is not, and it is clear when a decision has been finally determined), and in substantively difficult areas of law they are repeated over and over again under only slightly variant conditions as new appeals raising the same formal questions of law come to the courts. The social identity of the judge is clear, and professional judiciaries in modern states have in general been extremely successful in maintaining jurisdictional control over their key social function—namely, producing interpretations of law that carry the authority of the state. These key features of judges and judicial decisions are a gift to scholars who seek to identify the social factors that condition decision making and to isolate inference and judgment from potential confounding factors.

To see that knowledge work makes up some part of judicial decision making does not require us to ignore the robust behaviorist finding that some part of judicial decision making is the strategic enactment of personal or policy preferences. We need only to refocus attention on different parts of the evidentiary record and reframe our theoretical conception of judges (and their coauthor clerks) as social actors. There is thus good reason for sociologists of knowledge to care about judicial decisions, just as there is good reason for law and society scholars and judicial behaviorists to invest in understanding the toolkit of sociology of knowledge.

Political asylum law is one example of a substantive area of law where a relatively simple statutory law has produced richly suggestive interpretive disputes in case law. The core of political asylum law is the definition of a “refugee” as someone who has a well‐founded fear of persecution for reasons of race, religion, nationality, political opinion, or membership in a particular social group. This is the definition that governs the recognition of refugee status under the UN Convention and Protocol Relating to the Status of Refugees, and it has been adopted in essentially the same language in national legislation in the United States and in many other countries.Footnote ¹⁴ The application of this legal rule in the federal courts has been complex and sometimes contradictory, most especially through the efforts of the courts to clarify the meaning of the residual protected class, “particular social group.” Nuclear families have been recognized as a “particular social group,” but Senegalese wives abused by their husbands have been denied status.Footnote ¹⁵ Gang membership has been recognized, but “tattooed youth” have been denied status, even though the appellant in question claimed a well‐founded fear precisely because his facial tattoos signified gang membership.Footnote ¹⁶ Iranian feminists are out but parents of Burmese student dissidents are in.Footnote ¹⁷ Two mutually exclusive tests have been proposed for the recognition of a “particular social group” in the Ninth Circuit alone: that it should be characterized by a “voluntary association,” on the one hand, or by an “immutable characteristic,” on the other hand.Footnote ¹⁸ No court has yet succeeded in producing a determinate rule to settle the legal question.

Judges deciding these cases must make complex conceptual judgments about what constitutes a “particular social group,” about the material and the intersubjective social and emotional conditions for the existence of a “well‐founded fear,” and so on. They must do so, furthermore, in cases that frequently turn on empirical details of events that played out in distant lands under morally ambiguous, poorly documented circumstances. It is no surprise that they have regularly disagreed. Descriptive statistical and behaviorist work on political asylum law has already pointed to some of the significant patterns of that disagreement. David Law (Reference Law2005) reviews the apparently political calculus behind the publication or nonpublication of asylum decisions in the Court of Appeals for the Ninth Circuit. Ramji‐Nogales, Schoenholtz, and Schrag (Reference Ramji‐Nogales, Schoenholtz and Schrag2007, 363) show the dramatic variation in remand rates in asylum cases by circuit court, ranging from 2.4 percent (Fourth Circuit) to 37.7 percent (Seventh Circuit) in 2005. There is similarly striking variation in remand rates by applicant country of origin. In 2001, appellants from Haiti were admitted at a rate of about 7 percent; appellants from Armenia, Brazil, India, Iran, and Russia were all admitted at rates greater than 50 percent (Ramji‐Nogales, Schoenholtz, and Schrag Reference Ramji‐Nogales, Schoenholtz and Schrag2007, 361).

A two‐level analytical approach to judicial decisions has the potential to enrich the behaviorist reliance on statistical data alone. At one level would be historical and institutional accounts of the courts and the legal systems in which they are embedded. A historically informed institutional analysis may help us explain otherwise mysterious between‐circuit variations. The Ninth Circuit, which processes more political asylum cases than any other federal court, also generates panel disputes at a higher rate than any court save the Supreme Court and the District of Columbia Circuit, is reversed by the Supreme Court at a higher rate than any other court of appeals and is implicated in by far the greatest number of circuit court splits. This disputatiousness cannot be attributed to a small number of especially zealous or outlying judges. Of the sixty‐seven judges who were on the Ninth Circuit bench for any period of time between the introduction of the Refugee Act in 1980 and 2011, forty‐seven of them (70 percent) at least once authored or joined a separate opinion expressing a legal disagreement with the panel majority over a question of political asylum law.Footnote ¹⁹ Ten have authored or joined twelve or more disputes (O'Scannlain, Pregerson, Kleinfeld, Reinhardt, Kozinski, Thomas, Callahan, Rawlinson, Bea, and Paez). By contrast, in the Second Circuit, the only other court that processes a similar number of asylum cases as the Ninth Circuit, thirteen judges have authored or joined a disputing opinion on a panel dealing with a question of political asylum law. That amounts to 28 percent of the forty‐five judges who were on the Second Circuit bench for any period of time between 1980 and 2011. Only three of those authored or joined more than one such opinion. We ought therefore to look for an institutional explanation for the highly disputatious Ninth Circuit decision making in political asylum law. Institutional factors such as the size of the court (more judges sit on the Ninth Circuit bench than on any of the other several circuits), the structure of communication within the court, and the role of clerks and other staff vis‐à‐vis the judges are clear examples of institutional dynamics that might make a decisive difference to the decision‐making process within the Ninth Circuit.

The second level of analysis would be pairwise comparisons of formally similar cases within the large network of judicial decisions. Over 20,000 political asylum cases have been heard in the US federal courts under the modern legal regime of refugee protection, and interpretively suggestive disputes have emerged at every level at which dispute is possible: within panels, within courts over time, across the courts of appeals with equal jurisprudential authority (Circuit Court splits), and in cases overturned in the judicial hierarchy. These disputes raise two core questions for the sociology of knowledge and law and society scholarship: “Why do we find different answers to formally identical questions of law in different social contexts?” and “Why do some legal interpretive questions remain unsettled, while others become settled or even pass uncontested through appellate court cases?” When behaviorist studies have focused specifically on case law disputes, it has usually been with respect to the dynamics of dissents within panels, and disputes are usually taken to be manifestations of fixed underlying preferences (Epstein, Landes, and Posner Reference Epstein, Landes and Posner2013, 255–303). The explanandum for behaviorist scholars is, then, “under what conditions will judges express their disagreements?” rather than “how does social context affect inference and judgment such that judges produce different legal judgments in response to formally identical questions?”

Instances of notable consensus can also be subject to investigation—for instance, we can interrogate the lack of any dispute in the US appellate courts over the interpretation of “race” or “religion” for the purposes of asylum protection, in contrast to “particular social group” and “political opinion.”Footnote ²⁰ Judicious reliance on such surprising nonresults can rescue the case analysis procedure I am proposing from the problem of selection on the dependent variable. With enough such cases to compare, we will be able to assess how some fairly narrow and well‐defined, but still substantively interesting, knowledge problems have been answered differently in response to social contextual variation and path dependencyFootnote ²¹ at the level of individual experience and self‐understanding, at the level of small group dynamics, at the level of institutional culture, and at the level of national culture.

Ultimately, my proposal to bring sociology of knowledge tools to the analysis of judicial decisions will stand or fall on the empirical results that it produces. Within the scope of this essay and with the examples I have mentioned in passing, I mean only to propose a plausible idea. I close with two comments to meet potential objections that may already have arisen for the reader.

First, I recognize that what I outline here is only antagonistic to parts of the behaviorist agenda—most fundamentally, to the notion that modeling utility functions can in principle be the basis for a complete explanation of the social conditioning of judicial decision making. The behaviorist datasets will be an important descriptive base for any work in the near future on judicial decision making. I have embraced and elaborated on Posner's basic framework for what a comprehensive investigation of judicial decision making should look like, although I have disagreed about how his nine dimensions of explanation can be made conceptually commensurate and how they can be most effectively investigated in empirical work. Finally, we should leave open the possibility that for some judicial decisions, a behaviorist account may be the essentially correct and complete explanation. If we were to take Posner himself as the subject of analysis (at some level this seems unavoidable—he is the most prolific and highly cited US courts of appeals judge of his timeFootnote ²² ), even the phenomenological and pragmatist dimensions of explanation, which I have insisted are conceptually speaking not expressible in terms of a utility function, might add little. Posner makes it clear in his writings that his intellectual self‐concept is of someone who searches for efficiency rationales when making legal decisions (Posner Reference Posner2013, 26) and who understands his own action or behavior in terms of utility maximization. In some areas of law, such thinking in terms of utility maximization may turn out to be a generalized practice of judges, and hence “the logic of the law might be economics” (Posner Reference Posner2007, 26). Antitrust law is one likely candidate. In such cases, the best account of the social production of judicial decisions would converge on a behaviorist model, albeit with some presentational differences.

Second, I recognize the limited empirical applicability of the approach to judicial decisions I have outlined in this essay. A great advantage of the behaviorist research agenda is that at least basic models can be constructed with respect to any area of law for which systematic judicial decision data have been collected. To implement a rigorous interpretive analysis of judicial decisions along the lines I have suggested requires that we make empirical reference to areas of law where formally similar questions of law are answered differently in the courts with some frequency. Not every area of law will be amenable to that kind of analysis. Areas of law governed by complex and regularly changing statutes—intellectual property law, for example—may not produce a set of judicial decisions that can easily be analyzed in rigorous comparative context. Consequently, the sociology of knowledge as an interpretive lens cannot claim the same scope as the behaviorists’ rational actor models. Nonetheless, it remains worth pursuing because of the light it can shed on some areas of law and because it would be useful to show by example the limitations of the behaviorist explanatory mandate.