LAW AS AN ETHNOGRAPHIC OBJECT

There are many ways to take law as an ethnographic object—that is, a social phenomenon that both reflects and affects the society that produces it. For example, the law and society tradition often contrasts law “on the books” with law “in action.” Following law as it descends from abstract heights to on-the-ground situations, we show how it works, and changes, through social practices, how people in different social positions experience it differently, and how its effects are explicable but never completely predictable. At the same time, formal law itself is also a human product, made by people in different social positions bound by social strictures, whose experience of lawmaking is not completely predictable. In this chicken-and-egg sense, law on the books is already law in action.

In this article, I take as my ethnographic object the law that guides officials’ evaluation of law itself. I focus on doctrines that tell courts how to review administrative agency decisions about statutes. Law about law: meta-law.Footnote ¹ I also look a bit beyond law to some theories and ideas that are not themselves inscribed into the formal legal system but would probably like to be—ideas that instruct legal actors how to reach legal understandings. In explaining how these meta-laws and would-be laws make sense to their users, I hope also to show how they use ideologies about language to subtly support a very particular notion of political legitimacy—one that leaves a lot of modern governance hanging.

Taking agency judicial review doctrine as my ethnographic object allows me to ground myself in a cultural product of American legal discourse—one that reflects enduring language ideologies at the same time as it influences the development of American law. I try to unearth the underlying conceptions of law, language, and governance that allow the doctrine to make sense—and even seem intuitive—to its practitioners. That is, I take an ethnographic attitude, using anthropological tools to analyze legal artifacts.

In this, I treat the content of the law seriously—as seriously as I would treat a social movement, a governance decision, a linguistic pattern, or any other ethnographic object from my field sites (Bernstein Reference Bernstein, Washington, Goodall and Rosier2006, Reference Bernstein2008, Reference Bernstein2017b). But, again like any ethnographic object, I do not take it credulously; I do not take it at its word. Analyzing an ethnographic object involves recognizing it as a social product: shot through with cultural forces and exerting social effects. It also involves taking multiple perspectives on the object of analysis to explain how the social phenomenon makes sense to the people who engage in it, exploring how that sense-making fits into larger contexts, and illuminating some of what it leaves out or glosses over. Ethnographic inquiry seeks to approach its objects from many directions at once. We view our objects from the inside, but we also take perspectives that diverge from those the objects themselves suggest, and we stay on the lookout for implicit underpinnings and effects. In taking this approach, I assume that formal law is—soup to nuts and chicken to egg—a social product, available to anthropology’s analytical tools.

Below, I first introduce two of the primary doctrines governing judicial review of agency action. One guides the review of agency interpretations of law, the other, of agency policy decisions. Whether or not judges believe that agency action actually takes these two discrete forms, they must choose which doctrinal line to apply in any given case. So, I then suggest, this bifurcation of the doctrine frames agency action through an implicit dichotomy between the interpretation of statutes and their implementation. I trace out the way this dichotomy sets up an implicit contrast between a relatively passive understanding of inherent linguistic meaning, on the one hand, and an active utilization of statutory power, on the other: a contrast between saying and doing. But, I note, this contrast is not strongly anchored in the realities of agency action. Some of the main work that this doctrinal contrast does might thus be precisely to support a dichotomous vision of agency action as coming in these two discrete flavors.

The second half of the article brings to bear scholarship on language use. I show why this saying-doing contrast finds no basis in the qualities of legal language, which presents a form of doing by saying that is recognized in work on speech acts or performative utterances. Drawing on research in linguistic anthropology, I introduce the theory of language ideologies: ways of talking about language that tie ideas about communication to ideas about other aspects of society. Language ideologies help people make sense of their world by presenting normative visions of social ordering as though they were natural or inherent. I explain how the dichotomous judicial review doctrine that splits the saying of law from the doing of it fits the primary characteristics of language ideology: iconicity, erasure, and recursivity.

Interviews with agency employees yield a non-dichotomous image of regulation production as a gradual institutional consolidation around a plan of action regarding a statute. But the language ideology that splits interpretation from implementation supports a normative vision in which law has inherent meaning assayable through value-neutral interpretive approaches, while its implementation depends on the value-laden practices. I suggest that the structure of legal language belies the empirical veracity of this vision, while the practice of agency action undermines its claim to normative desirability in a democratic society. Given this, I suggest that it may be time to replace this language ideology with one that is both more empirically defensible and more normatively attractive. And I take a first stab at imagining what that might look like.

No ethnographic explanation is complete or final. Any ethnographic object worth its salt participates in, and dynamically develops, multiple sense-making processes and contexts (Bernstein Reference Bernstein2017b). This article, like any other, leaves out a lot more than it addresses. I do not address, for instance, the role of Congress in making statutes or, indeed, separation-of-powers issues more broadly. I ignore the other doctrines that surround the doctrines I discuss here, not to mention the very process by which something recognizable as a doctrine emerges—and much more. I present a narrow view of a narrow object. Still, I hope that taking an ethnographic attitude toward formal law in this way unearths some presuppositions running through some legal doctrines and illuminates their less visible political effects. What may seem like an inert text—law on the books—turns out to be a sociocultural microcosm.

Like any ethnographer, moreover, I hope that viewing my analytic object in novel contexts and asking questions that are not expected reveals counterfactual possibilities—ways the law could have gone and ways that it may yet go. Although an ideology naturalizes a particular social and normative order, other ideologies wait in the wings, perhaps eventually taking its place to naturalize other orders (Mouffe Reference Mouffe2018). As Friedrich Nietzsche ([Reference Nietzsche1874] 2018, 29–30) reminds us, any seemingly natural social fact “was, at some point, a second nature”; that point is made invisible by processes like the ones I discuss here, processes by which a “victorious second nature becomes a first.”Footnote ²

JUDGING AGENCY JUDGMENT

Way back in 1803, the US Supreme Court staked a famous position for the American judiciary. “It is emphatically the province and duty of the judicial department to say what the law is,” wrote Justice John Marshall in Marbury v. Madison.Footnote ³ Marbury claimed for the federal courts the power to invalidate statutes for incompatibility with the Constitution and to determine the content of those statutes. Even in 1803, though, there were other contenders for the role. Administrative agencies set up by Congress and the president to realize statutory provisions also had wide latitude to determine what the law was (Parrillo Reference Parrillo2021). Although legal scholarship generally treats the interpretation of law as, indeed, the province and duty of the judicial department, agencies have shared the responsibility for such determinations from the start (Mashaw Reference Mashaw2012; Mortenson and Bagley Reference Mortenson and Bagley2021).

The distribution of that responsibility currently follows a formula set out in the 1980s. Chevron v. Natural Resources Defense Council laid down the doctrine, repeated and elaborated over the following decades, for how courts should evaluate an agency’s interpretation of a statute when it is challenged in litigation.Footnote ⁴ As a binding Supreme Court precedent, Chevron provides, somewhat unusually, a legal rule about the interpretation of legal rules (Gluck Reference Gluck2014). When a party challenges an authoritative agency statutory interpretation, Chevron instructs courts to first determine whether the statutory term is “ambiguous”—that is, multivalent and susceptible of more than one correct interpretation (Bernstein Reference Bernstein2016, 6–7).^{Footnote 5} If the court can discern only one possible correct interpretation, any incompatible agency interpretation is invalid. If, however, the court concludes that the statutory term can in principle bear more than one correct meaning, it goes on to the second step: determining whether the agency’s interpretation is “reasonable.”Footnote ⁶ So, American courts have developed a particular way for courts to proceed in litigation about an agency’s interpretation of a law.^{Footnote 7}

When an agency’s policy decision is challenged in court, courts apply other standards. The most prevalent rests on the Administrative Procedure Act instruction that courts “hold unlawful and set aside” litigated agency actions that are “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.”^{Footnote 8} The currently canonical elaboration of this approach—somewhat hilariously called the “arbitrary and capricious standard”—came in Motor Vehicle Manufacturers’ Association v. State Farm, a case decided the year before Chevron.Footnote ⁹ State Farm held that, to determine whether a policy decision was arbitrary and capricious, a court should consider whether the agency “relied on factors which Congress has not intended it to consider, entirely failed to consider an important aspect of the problem, offered an explanation for its decision that runs counter to the evidence before the agency, or [the decision] is so implausible that it could not be ascribed to a difference in view or the product of agency expertise.”Footnote ¹⁰

INTERPRETATION AND IMPLEMENTATION

There are thus two approaches to reviewing agency action. Agency action labeled interpretation flows into the Chevron doctrine, while that understood as implementation follows State Farm. The doctrines also have considerable overlap. Judicial review of both things shares a concern with rationality: unlike legislative enactments, which need only be constitutional, the fundamental requirement for agency action is reasonableness under existing circumstances (Bernstein Reference Bernstein2016, 46–48). State Farm requires agencies to show that their policy decisions further statutory directives; Chevron requires agencies to show that their interpretations are reasonable. What exactly reasonableness entails in the interpretation context is a matter of some uncertainty, and strong arguments have been made that Chevron’s reasonable interpretation requirement merges at some point with State Farm’s rational policy one (Foote Reference Foote2007; Pierce Reference Pierce2007; see also Bernstein Reference Bernstein2016, 13–15). One could also argue that State Farm implicitly incorporates an ambiguity inquiry that resembles Chevron’s: if Congress’s policy preference is clear from the statute, any contrary agency decision would presumably be arbitrary and capricious. Thus, the two doctrines may not be as distinguishable as they are phrased to be (an issue I discuss further below).

Yet when a court evaluates agency action, it chooses whether to apply the Chevron approach or the State Farm one. Doctrinally, there is no option to simply review agency action for rationality and concordance with unambiguous statutory directives. The court must either follow Chevron’s two steps or work through the State Farm rationality factors. Thus, despite the interweaving similarities of the approaches and their frequently analogous results, the doctrine remains bifurcated: the paths courts follow to evaluate an agency’s reasonableness, and even the rules for deciding whether to evaluate the reasonableness at all, fall along interpretation-versus-implementation lines. Even if judges do not believe that agency action comes in these two discrete flavors, the doctrine channels their evaluations into one or the other. Courts, after all, must review some particular thing; the need to choose between Chevron and State Farm pushes judges to categorize that particular thing in the terms the doctrine sets. This bifurcated approach to judicial review can thus make interpretation and implementation seem like distinct qualities of agency action itself: a quality of the thing being reviewed.

It does not take a huge perspectival shift, though, to see the interpretation-implementation distinction as a quality of the judicial framing of agency action instead. The details of Chevron demonstrate why. Amendments enacted to the Clean Air Act in 1977 held new “stationary sources” of pollution in some states to the lowest achievable emissions rates and required them to be offset by decreasing pollution from other sources.Footnote ¹¹ The Act defined a stationary source as “any stationary facility or source of air pollutants which directly emits, or has the potential to emit, 100 tons per year of any pollutant.”Footnote ¹² The Environmental Protection Agency (EPA) initially treated any object capable of emitting one hundred tons per year as a stationary source. So, in a factory with several large smokestacks, each smokestack might constitute one “source.” Under President Ronald Reagan, the agency switched to treating groups of objects operating as a joint economic unit as a single stationary source. Our multi-smokestack factory would thus count as just one “source.”

The US Supreme Court treated the issue as one of statutory interpretation: the agency had changed its definition of “stationary source,” and the court had to determine whether the new interpretation was permissible.Footnote ¹³ Yet it would not have been difficult to classify this issue as a matter of policy. The “bubble concept” narrowed the universe of objects that the EPA treated as regulatable, making it easier to build new polluting structures—an extra smokestack in a factory of smokestacks.Footnote ¹⁴ The change loosened restrictions on industrial development, despite its predictable damage to the environment, with a view toward spurring economic growth. It is not self-evident that this decision is primarily about determining the meaning of the term. Instead, it seems easy to characterize it as a policy choice—a part of the deregulatory agenda President Reagan had promised.

The Supreme Court could, in other words, have applied the State Farm factors to the Chevron question, asking whether the agency had taken into account the considerations that mattered to Congress and had offered a rational explanation for implementing the bubble concept. Indeed, the Chevron opinion itself says as much: “[A]n agency … may … properly rely on the incumbent administration’s views of wise policy to inform its judgments…. [I]it is entirely appropriate for th[e executive] branch … to make such policy choices.”Footnote ¹⁵ Yet, even while calling the bubble concept a policy, the Chevron opinion proposed a separate approach for linguistic decisions interpreting statutes, as opposed to State Farm’s approach to policy decisions implementing them. Later cases subscribed to and strengthened this dichotomy.

The doctrine, in other words, lays out two separate streams for the judicial evaluation of agency actions. But the choice of which stream to follow can depend less on the nature of the agency action and more on the way the court frames the inquiry. The doctrine thus projects underlying assumptions about law onto the agency actions that bring it to life.

SAYING AND DOING

Anthropological analysis, the old saying has it, strives to make the strange familiar and the familiar strange. That work involves elucidating how ethnographic artifacts make sense to the people who make and use them. It also means showing the contingencies and broader contexts that give rise to such presumed realities. So, what can make sense of the artifact I have presented—the doctrine that makes it seem like interpreting and implementing describe contrasting agency actions rather than enacting a decision about how to frame agency action?

The bifurcated approach makes sense if we assume that understanding the linguistic meaning of a law is a different kind of activity than determining the law’s application. A statute, in this image, has one side that includes just language and communication and another that involves action and application. The language side requires interpretation, which the court can evaluate by considering what the words of the statute are most likely to mean. Courts are supposed to make sure agencies stick to the clear meaning of the statute’s words or, if those words lack a clear meaning, stay within a reasonable range of meanings. The action side is different: it involves making policy decisions about the statute’s practical implementation. Here on the arbitrary-and-capricious side of things, courts should ensure the agency rationally grounds its decision in relevant real-world facts.

Notice how, in this image, interpretation seems more passive and constrained: not a decision one makes oneself so much as the elucidation of a decision that someone else has already made. The implication is that there is something like a right answer to the question of meaning. Unambiguous statutory words either do or do not mean what the agency says. Multivalent statutory words either can or cannot reasonably bear the agency’s meaning. The implementation side appears more agentive, creative, open-ended. The agency gets to choose how to act as long as it can justify its choices as rationally responding to the relevant evidence.

Yet, as I suggested above, a given agency action can often be framed as both interpreting and implementing, depending on the perspective or narrative one chooses. That is because, when it comes to administrative agency action, any decision can enact effects on the world. Just like implementation decisions, interpretations of statutory language might define the reach of agency jurisdiction, articulate a standard for private conduct, set out eligibility criteria, and so on. The interpreting-implementing distinction, then, is not strongly anchored in the characteristics of agency action itself.

This doctrine about law—this meta-law—turns out to be also a doctrine about language. It effectively relegates the linguistic expression of law to a world of understandings and meanings—a relatively passive realm in which there are something like correct answers that preexist a legal decision and that can be evaluated as a matter of pure semantic expression without recourse to practical real-world contexts. Layered on top, or slightly to the side, of those understandings and meanings sit policy decisions—more agentive, creative, open-ended processes that require engagement with practical realities and are less predetermined by the statute itself. In short, the doctrine imagines saying as pitted against doing.

MAKING MEANINGS

The division of judicial review doctrine into one approach for saying and another for doing reflects a broader “drive for reference” in legal theory (Weissbourd and Mertz Reference Weissbourd and Mertz1985, 623). This drive assumes that what language does, primarily, is pick out objects in the world and make assertions about them. That is, this approach focuses on the side of language that allows for “reference and predication,” the “‘semantic’ aspect of speech[,] … which can be analyzed apart from the social context of speaking” (623–24). It obscures the more “creative” aspects of language, which can take many forms and have many effects (Silverstein Reference Silverstein, Keith and Henry1976, 11). Language use “make[s] explicit … the parameters [that] structure … the ongoing events” (33–34). Through the use of shifters—words whose meaning depends on their situation of use—language use can help create social categories, groups, or objects of attention through socially constructed inclusion and exclusion (Silverstein Reference Silverstein, Keith and Henry1976). And, of course, language use can create quite new situations in the world, as with explicit performative utterances (or speech acts), which help constitute the conditions they describe (Austin [1955] Reference Austin, Urmson and Sbisa1962; Searle Reference Searle1974; Rosaldo Reference Rosaldo1982).

In everyday language, in other words, we frequently encounter words that do things. And the way they do things is hard to separate from the way they say things: the word “I” simultaneously communicates a predetermined meaning (a reference to the speaker) and creates a “unique being”—this current speaker as opposed to some other one (Benveniste [1966] Reference Benveniste1971, 218). We may not always realize this aspect of communication—it is easier to notice language in its identifying and describing mode than in its constituting and revealing mode (Silverstein Reference Silverstein and Duranti2001)—but we live it anyway. These effects do not necessarily correspond to individual words or terms. It is more helpful to think of them as aspects of language use rather than as kinds of language. The semantic aspect of language—that part that stays steady across usage—contributes to the production of meaning in its pragmatic context—the circumstances in which language is used.

For example, the word “chair” carries a certain steady semantic meaning across contexts of use—let us say a chair generally is something one sits on (Weissbourd and Mertz Reference Weissbourd and Mertz1985, 626–27). But whether any particular instance of “chair” should mean the easy chair in which you read articles, the rocking chair on my porch, or the miniature dining chair on which a child’s doll sits is something that the people involved in the linguistic interaction have to work out for themselves. They do that by making decisions about which aspects of the surrounding context—co-text, interactional patterns, intertextual clues, and other cultural influences—are most relevant to meaning production in this particular instance (Bernstein Reference Bernstein2017a).

Interactional participants often help each other out in this process—for instance, through the use of deictics, which help “single out objects of reference or address in terms of their relations to the … interactive context in which the utterance occurs” (Hanks Reference Hanks, Duranti and Goodwin1992, 47).Footnote ¹⁶ Deictics “organize[] the field of interaction into a foreground upon a background” by expressing relationships between the referent—this or that object being picked out—and the “context of speech at the moment of utterance” (61). Like other linguistic functions, deictics are easily misrecognized as stable descriptions of enduring objects or essential natures (see, for instance, Green Reference Green2009). But, in fact, “they describe not the referent itself, but the relation between the utterance framework and the referent” (Hanks Reference Hanks, Duranti and Goodwin1992, 50). In doing so, they help interaction participants jointly construct both the backgrounded interactional context and the foregrounded focus of attention—objects that are not stable but must be continuously created through meaning-making practices. Just think of the way a word like “we” allows “speakers in speaking [to] create a social group around them, including some members of the audience and excluding others” (Weissbourd and Mertz Reference Weissbourd and Mertz1985, 626; emphasis in original). The way people create meaning thus relies heavily on linguistic functions that are simultaneously context-dependent and context-altering.

Speech acts, deictics, and related phenomena highlight a larger fact about language: the constant, semantic content of linguistic utterances is not enough to create meaning in any given instance. Semantics is not contrasted with pragmatics but nestled within it. So, just as the say-versus-do configuration of judicial review doctrine finds no strong anchor in the character of agency action, it also does not comfortably map onto the realities of linguistic communication. If anything, legal language tends to be particularly speech-acty, the epitome of doing by saying (Constable Reference Constable2014). Statutes constitute legal categories, issue orders, make promises—classic performative stuff. This performativity is perhaps even a little harder to miss than the everyday kind since the whole point of statutes is usually to have an effect on the world. And the agencies that shepherd statutes into efficacy have to take the communicative with the performative: deciding how to regulate stationary sources of pollution requires deciding what to count as a stationary source of pollution.

LANGUAGE IDEOLOGIES OF LAW

The bifurcated doctrine of judicial review of agency action, then, does not reflect an actual dichotomy either in agency action or in language. That is, this approach to evaluating how agencies handle the laws they are entrusted with “do[es] not simply describe the social world in any direct way; [it is] rather [a] tool[] for arguments about and in that world” (Gal Reference Gal2002, 79). In this sense, it can productively be understood as a “language ideology”—a story that people tell about language that helps tell further stories about other aspects of social life (Schieffelin, Woolard, and Kroskrity Reference Schieffelin, Woolard and Kroskrity1998). Language ideologies tie notions about communication to notions about social ordering in ways that allow participants to present normatively inflected images of society as though they were natural and unavoidable (Woolard Reference Woolard, Bambi, Kathryn and Paul1998, 4). Like any ideology, a language ideology helps people make sense of their world, bringing coherence to dispersed social spheres by subsuming phenomena under some governing logic (Silverstein Reference Silverstein and Saville-Troike1977).

Language ideologies usually center on properties claimed to be inherent to language, allowing people’s understandings of how language works to have profound effects on social dynamics far beyond any particular communicative situation. Prototypically, understandings of how language habits are connected to particular kinds of speakers help spread or cement ideas about people of different ethnic, religious, political, or economic groups, making sociologically attributed characteristics appear inherent and eternal. That in turn can influence how people behave toward others—that is, how they enact social stratifications and divisions. Supposedly empirical qualities of language are thus made available for projection onto society as a normative expectation—say, that a certain kind of speech indicates a certain kind of person—or demand—say, that a cognizable political or sociological group speak one single common language (Irvine and Gal Reference Irvine, Gal and Kroskrity2000).

But language ideology can extend beyond the prototypical insistence that social identity match linguistic form. More broadly, it can structure the semiotic ordering of a wide variety of social categories, giving category schemas a foundation that is both hard to see and easy to regenerate through resonant usages across instances and domains. Crucially, language ideologies are “not … a mere false frame that distorts our vision of ‘reality’” (Mertz Reference Mertz, Bambi, Kathryn and Paul1998). They are, rather, productive of our realities. They influence how people interact with language and with one another through language. In other words, language ideology is one way that societies create and perpetuate important—but not inherent—social categories and distinctions.

Scholars have identified several key ways language ideology does its invisible work. It often presents some aspect of language use as representing, rather than merely being connected to, a particular social object: “[L]inguistic features that index social groups or activities appear to be iconic representations of them, as if a linguistic feature somehow depicted or displayed a group’s inherent nature or essence” (Irvine and Gal Reference Irvine, Gal and Kroskrity2000, 37; emphasis added). So, maybe I take a person’s slow speech to represent their deficient thought processes (rather than, for instance, politeness) or a non-standard grammatical patterning to represent a deficient culture (rather than, for instance, an alternative linguistic form).

Language ideology also usually involves some erasure of things that do not fit comfortably into the ideological image: “Facts that are inconsistent with the ideological scheme either go unnoticed or get explained away” (Irvine and Gal Reference Irvine, Gal and Kroskrity2000, 38). So, if I assume that most people are monolingual in the language I associate with their ethnicity, I might treat nonconforming polyglot realities as aberrational or reinterpret social systems organized by other allegiances as “chaotic” or backward (65).

Finally, language ideology often reproduces central categories or divisions at different levels of generality or in different domains of social action, creating a recursive regeneration of key dichotomies: “[I]ntragroup oppositions might be projected outward onto intergroup relations, or vice versa” (Irvine and Gal Reference Irvine, Gal and Kroskrity2000, 38). The dichotomy marking homes as private and streets as public might be projected into the home—yielding a public parlor and a private bedroom—and onto the street, where public sidewalks mark off the private space of shops (Gal Reference Gal2002). This helps naturalize distinctions between private and public spheres, making privateness and publicness look like characteristics of spaces rather than projections of social categories onto spaces (Gal Reference Gal2002, Reference Gal2005). The ability to recursively reincorporate the same objects—house, parlor, sidewalk, shop—into different relative positions makes it easy to reproduce and reinforce the schema, while allowing participants to continuously reinterpret objects of attention to create the expected dichotomous configuration.

SAYING, DOING, AND INSTITUTIONAL CONSOLIDATION

Once we look at it from the language ideology angle, the doctrine guiding judicial review of agency action fits pretty snugly. As I suggested above, when we examine agency action by itself, it is often difficult to separate out the interpreting from the implementing. Indeed, in a series of interviews on this question, federal administrators generally rejected the idea that interpretation and implementation characterized two different kinds of agency action: “To me, they go hand in hand. Just in my experience, they’re not really separate functions” (Agency Action Study [AAS] 234). “They’re rolled up together because in the process of implementing, we interpret” (AAS 670). “You don’t have the luxury of separating those two questions if … your agency is the one who’s going to be responsible for implementing [the statute]” (AAS 486).Footnote ¹⁷ Such comments typify agency personnel responses.

To the extent that the interviewees did distinguish interpretation from implementation, they usually did so not as a binary contrast between distinct kinds of actions but, rather, as a graduated, relative understanding of different stages of agency process. Interviewees tended to describe settling on the more general, abstract, overarching goals or purposes of a statutory provision as closer to interpretation, while identifying specific, concrete, localized measures to achieve those goals came closer to implementation.Footnote ¹⁸ Moreover, none of this was ever described as a single decision. The entire process, including general understanding and specific measures, involved scores of participants, as statutory questions made their way through levels of administration hierarchy and areas of agency expertise (subject matter, legal, technical, congressional and public relations, interagency negotiation, and so on) and then returned to start the journey again for the next stage in the process. This temporally drawn-out, multi-participant, iterative practice is not easily dissected into contrasting pieces of interpretation and implementation. Rather, it produces a gradual institutional consolidation on a plan about what to do with the statute.

The actual agency practices that go into institutional consolidation, and the internal categories through which it is understood within the agency, remain invisible to the doctrine guiding court review of agency action. When courts choose, as they must, whether to follow the Chevron or the State Farm path, they posit a resemblance—an iconicity—between the bifurcated doctrine and the agency action it reviews. Imagining agency action as coming in these two flavors treats legal saying as clearly separable from legal doing. In other words, it erases the inherently speech-acty quality of law, which changes the world with its words. Moreover, agency review doctrine presents just one example of this dichotomous ideology, which pops up recursively in other areas to separate the doing of law from the meaning of it. One thing to note here is that, in practice, agencies have historically won lawsuits most of the time under both approaches (Eskridge and Baer Reference Eskridge and Baer2007; Barnett and Walker Reference Barnett and Walker2017). Add to this the extensive overlaps between the approaches as they are applied, and it starts to seem like a big part of the work these doctrines do may be maintaining the ideological dichotomy between interpretation and implementation.

RECURSIVE REINSCRIPTIONS

One such recursive move posits a homology between the supposedly distinct acts of interpreting and implementing, on the one hand, and the distinct social groups that have the authority to engage in those acts, on the other. The Chevron doctrine leaves the interpretation of legal text in the first instance to judges: it is judges who decide whether legal text is ambiguous, what it ultimately must mean, and what a reasonable interpretation must look like. The doctrine presents interpretive authority as inhering in the court, with agencies trying to hit the targets that judges set. Arbitrary and capricious review imagines the institutions a bit differently. Agencies have the authority to determine their policies and implement the statute. Courts are there to ensure that agencies have sufficiently considered the issue to come to credible conclusions and that they have sufficiently explained their reasoning to substantiate a connection between what they know and what they decide. There is, however, no claim that courts have the answers or even set the bounds within which answers may be found.

The doctrinal distinction between interpretation and implementation, then, is recursively projected onto governmental institutions. While, from the viewpoint of judicial review, an agency may either interpret or implement a statute, when the view pans out to the larger domain of government branches, it presents the judiciary as essentially in charge of interpretation, in contrast to the agencies, which implement. This is a familiar trope, but note how strange is its implication that a judicial interpretation of law is not, at the same time, also an implementation—as though a court deciding what a statute means had no practical effects on the litigants or the world around them.Footnote ¹⁹

Just as language ideology scholarship would predict, projecting the saying-doing distinction onto separate branches erases agencies’ interpretive work—work recognized when seen from the judicial review vantage. When put in the larger context of the separation of powers, this shifting distinction seems to support a vision of each branch wielding a distinct form of authority that does not overlap with that of the others, rather than a vision of related governmental authorities distributed among branches whose primary differences inhere in factors like personnel, temporal scale, impetus for action, procedures for acting, and so on (Mashaw Reference Mashaw2005; Bernstein Reference Bernstein2020). The dichotomous language ideology of legal saying and doing does not create the brittle, rigid version of separated powers that the American conservative legal movement has supported in recent years (Metzger Reference Metzger2017). But it can make that vision seem more natural or obvious, as though its brittleness inhered in the structure of government rather than being a very particular normative assessment of what government ought to be.

The same saying-versus-doing distinction is reproduced at the level of constitutional theory, where a tradition distinguishing “interpretation” from “construction” has been recently reinvigorated by proponents of originalism claiming the ability to draw clear “distinctions between … semantic content and legal content” (Solum Reference Solum2010, 118). Semantic content, in this telling, is a legal phrase’s “linguistic meaning, associated with the meanings of the constituent words and phrases” (99), while legal content is “what gives a text legal effect (either [b]y translating the linguistic meaning into legal doctrine or by applying or implementing the text)” (96).

Having staked out two discrete kinds of activity with respect to legal text—one that interprets, another that implements—one can then characterize each differently, ascribing empirical qualities with normative implications. In this version of saying versus doing, a legal writer’s analysis of language is removed from politics and contestation: interpretation “cannot be settled by arguments of morality or political theory” (Solum Reference Solum2010, 99–100). It is, rather, a kind of scientific investigation that reveals the true meaning of the text, like one might assay a piece of metal to understand its true composition. Interpretation is thus “‘value neutral,’ or only ‘thinly normative’” (102). Construction, in contrast, is “ultimately normative: … because constructions go beyond linguistic meaning, the justification for a construction must include premises that go beyond linguistic facts” (104). Interpretation, on this view, can be grounded in semantics—the part of word meaning that travels from context to context unchanged—while construction requires recourse to pragmatics—the meaning-making context.

Iconically projected onto the people who work with legal texts, this distinction subtly yields two groups. Those who (claim to) stick to interpretation—to merely saying what the law already inherently means—are less politicized, more constrained, and less able or willing to impose their views or use their discretion. They work in a “value neutral” sort of way. And they need no justification for their conclusions “beyond linguistic facts.” Those in the “construction zone,” in contrast, enact political motivations and impose policy preferences onto a neutral legal baseline—not just saying but also doing things with, and to, the law (Solum Reference Solum2010, 108). This iconicity predictably entails some erasure: the discounting or reinterpreting of realities that do not fit. It imagines that semantics suffice to produce meaning outside of pragmatics and presents these two aspects of meaning in a symmetrical, counterposed relationship, ignoring the way that semantics is just a part to pragmatics’ whole. And it imagines that law has a neutral, policy-free state from which one can choose not to depart, ignoring the way that any rendering of legal text entails a policy position and wiping away that speech-acty power with which legal texts affect their world (Bernstein and Staszewski Reference Bernstein and Staszewski2021).

Because this distinction comes from the framing rather than the object—the legal text—itself, users of the interpretation-construction distinction can recursively re-project the dichotomy at different scales and onto different domains of textual analysis, continuously creating groupings in which some legal actors merely read the law for its inherent meaning, while others impose their normative views on it. This language ideology reverberates through a sociological picture that presents judges as neutral expositors of meanings that inhere in texts and agencies as normative deciders of policies that they come up with. Alternatively, when the image is recursively re-projected onto each sociological group, it is easy to see good judges who are neutral, as opposed to worse ones who seek to impose their own views, and good agencies that decide on the details but leave legal meaning where judges find it. That, at least, is how the conservative legal movement has deployed these language ideologies to propound an image of courts as ideally removed from the consequences of their decisions (Bernstein and Staszewsky Reference Bernstein and Staszewski2021) and of agencies as exceeding legal text in order to constrain individual liberty (Emerson Reference Emerson2021).

CONCLUSION: DOING BY SAYING

Taking an ethnographic attitude toward formal law allows us to explore how laws make sense to those who promulgate and wield them. It makes visible the ways that underlying visions of law (and its language) ramify in and help cohere disparate notional domains—from judicial review, to the separation of powers, to constitutional analysis, to prescriptive theories of judging. By revealing some mechanisms through which governing images and ideologies are entrenched and invigorated, and casting a light into the hidden corners they ignore, this approach also helps us recognize contingencies and imagine alternatives.

Some alternatives suggest themselves already from my discussion above. Recall how categorizing agency action into contrasting flavors of interpreting and implementing neither matches the structure of agency action nor yields consistent results. And recall how administrators described implementing a law as not clearly distinguishable from interpreting it. The interpreting-implementing distinction does not provide a particularly relevant category schema for actual agency action. In litigation, agency personnel must know how to wield the schema skillfully to appeal to judicial preferences. Internally, though, agencies employ processes that involve numerous people over extended time periods iteratively focusing on, considering, studying, and negotiating about statutory topics to gradually achieve institutional consolidation around a plan of action. This practice recognizes a statute as a fundamentally practical kind of speech whose meaning inheres largely in its effects on the world.

Judges like to say what the law is. But it is agencies that bear the primary responsibility for statutes, and it is agencies that statutes primarily address. How might a language ideology look if it viewed statutes from the agency perspective—a perspective in which legal saying counts as doing? It might make the very idea that simply understanding a law’s individual words could suffice to give it meaning seem much stranger. If we assume that interpreting and implementing are interconnected—that understanding a law involves consolidating around a plan for how it will work—then the very possibility of a “value neutral” way of acting based on legal texts starts to look quite unlikely. Legal texts are prescriptions born of normative convictions, so it should not be surprising to find little value neutral ground in them. But that can be difficult to see if we imagine legal language as separable into purely meaning-bearing, as opposed to actionable, parts. Once we dissolve the dichotomies into a more integrated scheme, the inherently normative, value-laden quality of judging—just like of policy making in an agency—comes into clearer view. On this image, moreover, the good judge might no longer be one who claimed the unearthly ability to create meanings from pure semantics. Rather, it might be one who recognized the practical effects of judicial assertions and justified decisions with a view to their consequences.

One could think of Justice Marshall’s assertion that the judiciary gets to “say what the law is” as a simple description. Yet, given that it was this assertion of power that established the authority of judicial review, that seems a bit odd: the situation that Marbury describes did not exist before Marbury described it. Before Marbury, it was at least unclear whether the courts could invalidate congressional action for violating the Constitution—whether, that is, the courts could determine what the law would be or lay down what it should mean. Justice Marshall was surely aware that, with his opinion, he was not just describing but also establishing—or at least trying to establish—a state of affairs.^{Footnote 20} Perhaps Justice Marshall himself recognized that courts “say” what the law is the way I “say” what my child’s bedtime is: not so much describing a situation as setting a policy. In that sense, courts may not be so different from agencies after all. And, taken as an ethnographic object, the dichotomous doctrine of judicial review, which divides legal language into the saying and the doing, may be a familiar image that turns out to be quite strange indeed.