Though sometimes overlooked, the language of court opinions can be as important as the disposition of cases, albeit for different reasons. The outcome of a case has the most direct impact on the parties and issues involved. It announces who wins and who loses, which laws and policies survive and which do not. But the language of opinions is often where the real work of courts is done. Judicial opinions can confine holdings to particular sets of claimants or announce more general principles. They can firmly endorse rules or they can equivocate, inviting relitigation. And opinions can persuade, building up coalitions of judges for majorities and earning the support of interpreting and implementing groups.
This study investigates some of the language choices that U.S. Supreme Court justices make. Specifically, it focuses on what might be termed rhetorical sources: references to prominent authors and texts that are nonbinding on case outcomes. These sources include a wide range of materials, from interpretive authorities such as the Federalist Papers and Blackstone's Commentaries on the Laws of England to the writings of esteemed figures such as Thomas Jefferson and John Locke. Because their inclusion is optional, rhetorical sources can serve as a gateway to understanding how Supreme Court justices assemble their opinions. Why does one justice cite the Federalist Papers when another does not? Is it simply a matter of personal preference, or do legal considerations also matter? Are justices motivated by efficiency concerns, or by more complex interests?
One possibility is that justices use rhetorical sources strategically, employing them most frequently when the legitimacy of their holdings is in doubt, such as when they are overturning precedents or invalidating statutes. When deciding hard cases, justices must know that their opinions are likely to be scrutinized by individuals both on and off the bench. Even if such scrutiny does not affect the outcome a justice chooses (Reference Segal and SpaethSegal & Spaeth 1993), it may still affect the language used to defend it. Making an opinion as persuasive as possible can serve any number of useful functions. It can hold together a majority coalition or encourage a wavering justice to sign on. It can facilitate compliance by winning over interpreting and implementing groups. Or it can enhance the reputation of an opinion and its author in the legal community.
For example, when the Court in Furman v. Georgia (408 U.S. 238 [1972]) declared that states could not impose the death penalty because, as applied, it constituted “cruel and unusual punishment,” the justices did not simply present their views as raw expressions of judicial will. The separate opinions justify their expressed policy preferences by invoking such materials as Blackstone's Commentaries, John Stuart Mill's “On Liberty,” the Magna Carta, and the writings of Thomas Jefferson, Joseph Story, and Oliver Wendell Holmes, among other legal authorities. Buckley v. Valeo (424 U.S. 1 [1976]), concerning the constitutionality of campaign finance legislation, contains references to Charles de Montesquieu, while Justice John Paul Stevens's majority opinion in U.S. Term Limits v. Thornton (514 U.S. 779 [1995]) cites extensively from the Federalist Papers when arguing that states may not impose term limits on members of Congress. Strictly speaking, none of these sources was required to resolve the legal questions presented; yet the justices must have felt that their inclusion would enhance the legitimacy of their decisions.
Corley et al. reach similar conclusions in their study of the Federalist Papers (see also Reference DurchslagDurchslag 2005), finding that “the use of … sources in Supreme Court opinions is tactical, and to a more limited degree attitudinal, rather than … based solely on the need for important legal authority” (2005:338–9). The present study broadens the analysis beyond the Federalist Papers and other so-called originalist sources to examine the use of rhetorical sources in general. To be clear, rhetorical sources are not synonymous with historical or originalist sources: they are any materials that justices include in their opinions that are not binding on case outcomes. Rhetorical sources do not include references to precedents, statutes, administrative regulations, or other materials that are necessary to resolve cases. They are instead the optional, supplementary citations that justices include to make their opinions more persuasive.
The point of this study is not to suggest that references to materials such as the Federalist Papers are decisive in establishing an opinion's legitimacy, or even that justices take them very seriously, but to probe the types of factors that go into the assembly of written opinions in order to make better sense of this important feature of judicial policymaking. While it is hypothesized that strategic considerations play an important role in shaping opinion content, it is acknowledged that these factors are probably not the only ones that justices consider. Of likely importance are also the legal issues presented, the ideologies and preferences of particular justices, and the time and resources at a justice's disposal. By considering these factors together, one can assemble a more complete account of the factors that go into the writing of judicial opinions and thus arrive at a better understanding of judicial behavior.
The Puzzle
The behavior of judges has been the subject of much academic scrutiny, with a convincing body of literature finding that case outcomes, at least on the U.S. Supreme Court, are best explained by attitudes (Reference PritchettPritchett 1948; Reference SchubertSchubert 1958, Reference Schubert1959; Reference SpaethSpaeth 1964; Reference DanelskiDanelski 1966; Reference Segal and CoverSegal & Cover 1989; Reference Segal and SpaethSegal & Spaeth 1993). Because justices sit atop the judicial hierarchy and decide hard cases that have divided the lower courts, they are generally freer than other judges to make decisions based on their policy preferences. Yet a growing literature suggests that if case outcomes are mostly shaped by attitudes, the language of court opinions reflects more complex considerations. Supporters of the strategic model have shown that opinion content does not always reflect a justice's sincere preferences (Reference Spiller and SpitzerSpiller & Spitzer 1992; Reference Epstein and KnightEpstein & Knight 1998; Reference Tiller and SpillerTiller & Spiller 1999; Reference MaltzmanMaltzman et al. 2000; Reference Smith and TillerSmith & Tiller 2002; Reference CorleyCorley et al. 2005). More commonly, language is tailored to the desires or objections of other justices and actors whose support is necessary to bring about the opinion writer's goals.
Following the definition used by Smith and Tiller, strategic behavior occurs when justices consider the reactions of others, both on and off the Court, to their decisions and modify their behavior in response: “Strategic judges anticipate the likely responses of other players and make choices that will ultimately maximize their own policy preferences under given constraints” (2002:63). The strategic model thus differs from one of its chief rivals, the attitudinal model, in that it takes seriously the interactive nature of judicial decisionmaking. Supporters of the attitudinal model maintain that judges “decide disputes in light of the facts of the case vis-à-vis their sincere ideological preferences and values” (Reference SegalSegal 1997:28; see also Reference SchubertSchubert 1959; Reference Segal and SpaethSegal & Spaeth 1993). Committed attitudinalists assume that because judges act on their sincere policy preferences, they do not worry about how others will react to their decisions.
Evidence for the strategic model is mixed (see Reference SegalSegal 1997), but researchers have found support for at least two forms of strategic behavior. The first involves policy-choice strategies. In these circumstances, judges change the substance of their policies in order to maintain minimum winning coalitions. By negotiating and manipulating opinion content, a judge may broaden or narrow a holding, or in some other way alter the substance of a court's judgment. Reference Epstein and KnightEpstein and Knight (1998), for example, show that Supreme Court justices circulate bargaining statements in an effort to modify judicial policies, while Reference MaltzmanMaltzman et alia (2000) consider how justices respond to draft opinions or accommodate their own opinion language in an effort to appease colleagues. This research demonstrates that judges are willing to compromise the wording of their opinions in order to advance their policy preferences, even when such modifications limit the effects of preferred policies, because refusal to compromise can result in the collapse of a majority coalition and the enactment of less desirable alternatives.
The second form of strategic behavior involves instrument-choice strategies (Reference Spiller and SpitzerSpiller & Spitzer 1992; Reference Tiller and SpillerTiller & Spiller 1999; Reference Smith and TillerSmith & Tiller 2002). These strategies concern judicial manipulation of the supporting evidence or legal grounds of a decision. Judges do not change the substance of policies, but alter the language used to defend them. For example, Smith and Tiller show that lower courts select the legal grounds for decisions based on how their choice of instrument will influence higher courts: “There are compelling reasons to believe that the choice of a particular judicial instrument has important consequences for determining how resistant decisions are to reversal bodies with authority to review the court's decisions” (2002:61–2). Smith and Tiller find that lower court judges prefer to base decisions on reasoning process review than on statutory grounds because higher courts are less likely to review them.
Most previous applications of instrument-choice theory consider efforts by subordinates to influence their superiors (Reference Spiller and SpitzerSpiller & Spitzer 1992; Reference Tiller and SpillerTiller & Spiller 1999; Reference Smith and TillerSmith & Tiller 2002). But it is reasonable to suppose that Supreme Court justices also use instrument-choice strategies to persuade others, including their peers on the Court, interpreting and implementing groups, and the broader legal community. Persuasion is a central component of legitimacy theory, which holds that the justices' authority depends on how they present themselves, both in the courtroom and in their opinions (e.g., Reference ShapiroShapiro 1964; Reference EdelmanEdelman 1964; Reference JohnsonR. Johnson 1967; Reference PetrickPetrick 1968; Reference CaseyCasey 1974; Reference WalshWalsh 1997). Scholars who advance this theory believe that justices can win support for their rulings by enhancing their legitimacy, emphasizing such features of their decisionmaking as their expertise, prestige, and experience. Many of these studies focus on how the justices use myth to influence the public, relying on the pomp and ceremony of the courtroom as well as the perception that justices guard the Constitution and uphold the rule of law (Reference PetrickPetrick 1968; Reference CaseyCasey 1974). Yet it is doubtful that elites are influenced by myth in the same manner as the general public. Justices must often convey the legitimacy of their directives to an audience that is savvy about the policymaking role of justices and skeptical of their claims.
One way for justices to meet this difficulty is through the use of persuasive opinion content, including rhetorical sources, associating their rulings with the views of well-respected authorities. There is reason to suppose that such techniques can be effective, even among elites. Research on priming suggests that the behavior and attitudes of individuals toward policies are influenced by the prestige or legitimacy of the sources disseminating them (e.g., Reference Iyengar and KinderIyengar & Kinder 1987; Reference Miller and KrosnickMiller & Krosnick 2000). Miller and Krosnick find that even political elites are influenced by priming effects: “Priming does not occur because politically naïve citizens are ‘victims’ of the architecture of their minds, but instead appears to reflect inferences made from a credible institutional source of information by sophisticated citizens” (2000:301). By this logic, when Supreme Court justices use persuasive authorities, they prime their audiences to base their evaluations of judicial policies on their attitudes toward esteemed figures and texts.
One need not have a cynical view of the Court to suppose that justices use rhetorical sources in this manner. Indeed, it is very likely that Supreme Court justices, steeped in the norms and traditions of the legal community, find the materials they are citing persuasive themselves. The point is not that Supreme Court justices use rhetorical sources disingenuously, seeking to pass off arguments and evidence on others that they do not themselves find persuasive. It is that the shape of opinion content depends upon the context in which a justice is writing, with parameters determined by the preferences and anticipated actions of others. When faced with defending a judgment of questionable legitimacy, justices must employ special measures to ensure that their policy choices prevail. While one might take this behavior to be a natural component of judicial decisionmaking, to acknowledge that justices behave in this manner is to acknowledge that they act strategically.
A study by Reference WalshWalsh (1997) supports the legitimacy hypothesis. His analysis of state supreme courts finds that judges cite precedent with greater frequency and breadth when they believe the legitimacy of their holdings is in doubt, such as when they adopt new or far-reaching legal doctrines: “Recognizing a new doctrine engenders particular concern with acceptability, and the judges in these cases clearly felt obliged to cite more often and widely” (Reference WalshWalsh 1997:351). Walsh's findings apply only to state supreme courts, but without much difficulty one can extend the logic of his analysis to the U.S. Supreme Court. Murphy believed that a justice's greatest resource is “his capacity to engage in thorough research, to organize relevant facts and legal principles, to reason with taut logic, to write with eloquence, and to use persuasive rhetoric” (1964:98). Contemporary strategic theorists agree with Murphy that opinion content can be used by justices to make their rulings more appealing. Reference Epstein and KnightEpstein and Knight (1998) argue, for example, that justices frequently cite precedent to appeal to a norm of stare decisis in the legal community, while Reference CorleyCorley et alia (2005) find strategic motivations behind the Court's use of the Federalist Papers. Justices concerned about the legitimacy of their decisions may also embellish their opinions with references to other sources whose authority their readers recognize and accept.
Rhetorical Sources Defined
The concept rhetorical sources is potentially expansive, encompassing a wide array of sources that might reasonably be classified as rhetorical. A glance through the pages of U.S. Reports yields many possible candidates, including references to the Magna Carta, statesmen such as Thomas Jefferson, philosophers such as Montesquieu, and scholars such as Edward Corwin. Most of these sources should be familiar to court watchers. It is common for establishment clause cases to reference Jefferson's “wall of separation between church and state,” or for federalism cases to cite the Federalist Papers or Story's Commentaries on the Constitution, among others. Recognizing the need to be selective, this study focuses on a small set of rhetorical sources that justices frequently employ.
Four criteria were developed to guide selection. Each of the sources chosen are (1) nonbinding on the outcome; (2) well-respected by the legal community; (3) widely applicable, appearing in a broad assortment of cases; and (4) not contemporaneous with the period under analysis. While these criteria permitted some freedom in the choice of materials, they were helpful in generating a manageable number of rhetorical sources for analysis. The fourth criterion was especially important in distinguishing rhetorical sources from other, less discretionary sources that judges commonly use to resolve cases, such as references to precedent, legislation, or administrative proceedings. For a more extensive discussion of how sources were chosen, the Appendix outlines the procedure in some detail.
Table 1 lists the rhetorical sources included in the study. Because most of the sources in Table 1 are cited infrequently by the justices, they have been grouped into four categories: (1) Interpretive Authorities, which are written commentaries on the Constitution or English common law (e.g., the Federalist Papers, Story's Commentaries, Blackstone's Commentaries, and the writings of Sir Edward Coke); (2) Founding Documents, which express underlying principles of American government but have no controlling authority on the outcome of cases (e.g., the Magna Carta, the Declaration of Independence, and the Preamble to the federal Constitution); (3) Respected Jurists, who are highly respected former Supreme Court justices (e.g., Chief Justice John Marshall, and Justices Oliver Wendell Holmes, Louis Brandeis, and Benjamin Cardozo); and (4) Respected Nonjurists, who are highly respected but not judges, including founding fathers (e.g., Thomas Jefferson, James Wilson, and Brutus), philosophers (e.g., John Locke, John Stuart Mill, and Montesquieu), and scholars (e.g., James Bradley Thayer, President Woodrow Wilson, and Edward Corwin).
Table 1. Breakdown of Dependent Variable by Source Type
These categories have been collapsed into a single dependent variable, which measures the total number of different rhetorical sources used by a majority opinion writer.Footnote 1 The decision to use a single variable is theoretically justified if, as argued here, rhetorical sources are persuasive in nature and thus interchangeable to justices: references to the Federalist Papers should function similarly to references to other sources such as the Declaration of Independence or the writings of Woodrow Wilson. Nevertheless, some comparison of sources is useful to verify that they function alike.Footnote 2 The results therefore provide separate estimates for each of the four categories of sources.Footnote 3 Reliability estimates indicate that the measurement of the dependent variable is robust. Using two coders and a random sample of 200 cases, the interagreement generated a Cohen's kappa score of 0.875, suggesting high reliability.
Strategic Explanations
The first set of explanatory variables tested whether rhetorical sources are strategic instruments that justices use to enhance the legitimacy of case outcomes. Justices who are concerned about legitimacy may have a number of different audiences in mind. They may wish to persuade fellow members of the Court to join their opinions or prevent others from defecting. They may want to influence the lower court judges who are responsible for interpreting their decisions or the administrators and other actors who are in charge of implementing them. Or they may wish to write opinions that are well-regarded by the legal community, both now and in the future, thereby enhancing their reputations as jurists. The point is that justices in hard cases should be more likely to modify their opinion-writing in response to the anticipated behavior of others. The more that the legitimacy of a decision is in doubt, the more likely it is for a majority opinion writer to use rhetorical sources.
Testing for strategic influences thus involved identifying rulings in which majority opinion writers are likely to have heightened concerns about legitimacy. Two such categories of cases are those that formally alter precedent or invalidate state or federal law. Overturning laws and overruling precedents are among the most controversial actions that justices can take. When majority opinion writers ignore principles of stare decisis or overturn the actions of elected officials, they can expect opposition from individuals who are sympathetic to old policies or wish to limit the new policy's effects. In confronting these possibilities, justices should be more inclined to cite rhetorical sources to enhance the legitimacy of their positions. Operationalizing these variables involved adapting measures already present in Reference SpaethSpaeth's (1999) database. For “Formal Alteration of Precedent,” cases were coded 1 if the majority opinion altered or overturned precedent and 0 if it did not. For “Invalidation of State or Federal Law,” cases were coded 1 when the majority opinion overturned state or federal law and 0 when no such action was taken.Footnote 4 Because both variables represent conditions in which concerns about legitimacy are heightened, they should be positively correlated with the dependent variable.
The third strategic variable identifies the number of justices disagreeing with the majority. When the number of dissenting justices is high, majority opinion writers can be expected to make extra efforts to enhance the legitimacy of their opinions. Part of this effort may reflect a desire to hold together a majority coalition. But disagreement among justices also signals to outsiders that the majority's holding has been challenged and may be vulnerable to criticism. Interpreting and implementing groups may think that there is room to defy or evade the Court's ruling, since there is no clear consensus for the policy on the Court. Work by C. Reference JohnsonJohnson (1979) indicates that this scenario is quite plausible. He finds that evasive and discordant responses are more common among lower federal appellate and district court judges when a Supreme Court decision has a large number of dissenting justices or opinions. In such cases, majority opinion writers may guard against noncompliant behavior by citing rhetorical sources in their opinions. Operationalizing this variable involved adapting Spaeth's measure of the number of dissenting justices, which is a continuous variable ranging from 0 to 4.
Legal Explanations
Though strategic concerns are likely to be an important explanation for the Supreme Court's use of rhetorical sources, other factors may also contribute to the use of these sources and should therefore be considered. For example, rhetorical sources may function as tools of legal reasoning, necessary for the resolution of certain cases. To test these factors, the first legal variable controls for whether a case raises a constitutional question. Because most of the rhetorical sources included in this study are concerned with the history and interpretation of the federal Constitution, they are more likely to be correlated with constitutional cases than other issue areas. In order to test this hypothesis, a variable was developed, again using Reference SpaethSpaeth (1999), based on whether the authority for a decision was the federal Constitution. It is expected that citations to rhetorical sources are more likely in these cases.
A second potential legal explanation is whether rhetorical sources appear more often in complex cases, where more evidence is necessary to resolve the legal questions. When faced with complex questions, legally minded justices may feel that they need to consult a greater breadth of materials in order to reach the best result. Measuring complexity involved adapting Spaeth's measure of the number of legal issues under consideration by the Court: complex cases were identified as those raising two or more legal issues.
Attitudinal Explanations
Another explanation for the use of rhetorical sources relates to the preferences of particular justices. Table 2 reports the proportion of each justice's majority, concurring, and dissenting opinions that contained a citation to at least one rhetorical source in the years 1953–1997.Footnote 5 It indicates that justices vary considerably in their use of these sources. While some justices, such as Justice Felix Frankfurter (33.3%), Justice Antonin Scalia (26.4%), and Justice Warren Burger (26.0%) frequently cited rhetorical sources in their majority opinions, other justices, such as Justice Sherman Minton (4.5%), Justice Charles Whittaker (7.1%), and Justice Byron White (13.1%) used them less often. Still others had more complex habits: Justice Stephen Breyer, for example, frequently used rhetorical sources in his majority opinions (21.7%), but not in his dissents (4.0%). To understand more fully when rhetorical sources are likely to be used, the individual proclivities of particular justices should therefore be taken into account.
Table 2. Proportion of Opinions with Rhetorical Sources, by Justice, 1953–1997
To measure proclivity, a composite score was developed measuring the proportion of a justice's written opinions—majority, concurring, and dissenting—that included citations to rhetorical sources. These scores are reported in Table 2. A composite score was preferred to more opinion-specific measures of proclivity (e.g., using a justice's proclivity for citing rhetorical sources in majority opinions to predict citations in majority opinions) to avoid concerns about circularity. Thus, a justice such as Justice Stevens, who used rhetorical sources in 26.0% of his majority opinions, has a composite proclivity score of just 0.111 because he used rhetorical sources in only 4.1% of his concurrences and 6.3% of his dissents. The measure of proclivity is expected to be positively associated with the dependent variable.
In addition to the proclivity variable, two more measures of judicial attitudes were incorporated into the analysis. The first is a measure of judicial ideology. Because many of the sources used in this study are historical in nature, there may be an inherent tendency for conservative justices to cite them, especially since conservatism is often associated with originalist methodologies. It is therefore expected that majority opinions penned by conservative justices will be especially likely to include citations to rhetorical sources. To measure ideology, the Segal/Cover scores were employed (Reference Segal and CoverSegal & Cover 1989; Reference SegalSegal et al. 1995). While scholars have observed that these scores most accurately predict the justices' views on civil liberties issues (Reference Epstein and MershonEpstein & Mershon 1996), they remain the best available measures of judicial attitudes. Furthermore, a justice's liberalism or conservatism in the area of civil liberties is probably compatible with the liberalism or conservatism of that justice's interpretive methodologies. The Segal/Cover scores range on a continuous scale, with −1.00 indicating a strongly conservative justice and +1.00 indicating a strongly liberal one. It is expected to be negatively correlated with the dependent variable.
The final measure of attitudes concerns citation habits that derive from a justice's personal background. Specifically, it is hypothesized that justices who were previously law school professors are more likely to use rhetorical sources than non-academics. Justices such as Scalia and Frankfurter, who taught previously at law schools, may have taken from their prior academic experiences a didactic approach to the law that has informed their opinion writing. To assess this claim, opinion authors were classified based on whether they had taught at a law school before coming to the Court, with the expectation that justices with a law school teaching background are more likely to use rhetorical sources than other justices.Footnote 6
Efficiency Concerns
The final set of explanations involves the Court's workload. It is hypothesized that as the Court's workload decreases, justices are likely to be more at leisure to employ rhetorical sources in their opinions. One of the most direct measures of the Court's workload is the docket size. When justices decide fewer cases in a term, they have more time to devote to their opinions, which may allow them to include a greater number of citations to rhetorical sources. The Workload variable assesses whether citation habits vary with the number of opinion assignments that a justice has in a given year, with the expectation that justices are less likely to cite rhetorical sources as their workload increases. The variable is continuous and is based on the number of cases the majority opinion author was assigned in the year a case was decided.
Another factor that has contributed significantly to the Court's efficiency is an increase in the number of clerks. For the first half of the twentieth century justices had at most one clerk, but in 1947 the number was raised to two. Then in the 1960s Justices Potter Stewart and White petitioned the House Appropriations Committee for a third clerk starting with the 1970 term, which was then followed in 1978 with the addition of a fourth clerk. According to one former law clerk (Reference KesterKester 1995; see also Reference Oakley and ThompsonOakley & Thompson 1980), expanded opinions have become the inevitable consequence: “Opinions grow longer, procedures more complicated, doctrine more convoluted” (Reference KesterKester 1995:20). It therefore seems reasonable to hypothesize that the use of rhetorical sources is related to the number of clerks that a justice has employed. To gauge the impact of law clerks on the use of rhetorical sources, the Clerks variable measures the number of clerks employed by the majority opinion writer in the term a case was decided (see Reference BrownBrown 1996), with the expectation that citations will increase as the number of clerks increases.Footnote 7
Results
Table 3 summarizes the coding criteria for the independent variables. Hypotheses were tested with STATA 7.0, using Reference SpaethSpaeth's (1999) database to examine all full opinions (N=4972) decided from 1953 to 1997.Footnote 8 Because the dependent variable is a count variable and corrections for overdispersion were necessary, negative binomial regression was employed (Reference LongLong 1997). Table 4 provides estimates for each of the categories of the dependent variable, as well as for the combined model. All five models perform well, with the LR chi-square value for the combined model statistically significant at 338.580, with a pseudo-R2 of 0.052. The chibar2 value of 336.350 for the combined model is also significant, corroborating the presence of overdispersion and thus affirming the use of negative binomial regression in place of Poisson regression.
Table 3. Summary of Independent Variables
Table 4. Negative Binomial Regression Model Predicting Citations to Rhetorical Sources in Majority Opinions, 1953–1997
Note:
* p<0.05; standard errors in parentheses.
a chibar-square is the test statistic used in Stata for the likelihood ratio test for overdispersion.
To assist with the interpretation of the results, Table 5 reports first differences using the CLARIFY program (Reference KingKing et al. 2003). CLARIFY provides extra precision in the reporting of first differences because it presents the amount of error associated with the generation of predicted probabilities. Table 5 shows how changes in the values of certain independent variables affect the probability that one or more rhetorical sources will appear in a majority opinion. More precisely, Table 5 reports changes in the probability that the dependent variable will assume a value greater than 0 as the values of certain independent variables move from their minimums to their maximums, while holding the values of other independent variables at their means.Footnote 9 Only those variables that attain statistical significance in the full model are included in the calculation of first differences.
Table 5. First Differences, Majority Opinions (Combined Model)
Note: Change in the probability of at least one rhetorical source appearing in a majority opinion, as certain independent variables move from minimum to maximum values.
As hypothesized, all three strategic variables have a statistically significant impact on the use of rhetorical sources. The effect of altering or overturning precedent is the strongest and is significant across each of the categories of the dependent variable. According to Table 5, when majority opinion writers alter precedent, they are 27.8% more likely to cite one or more rhetorical sources in their opinions, give or take 4.9%.Footnote 10 The effect of dissenting justices is less powerful but still significant across three categories of the dependent variable, as well as in the combined model. The presence of four dissenters raises the probability that majority opinion writers will use rhetorical sources in their opinions by 4.8%, plus or minus 1.2%. Invalidating a state or federal law has no significant effect in the combined model but does significantly impact one of the categories of sources, citations to respected nonjurists.
Altogether, the success of the strategic variables suggests that justices do cite rhetorical sources with greater frequency when the legitimacy of their decisions is lowest. Justices are more likely to use rhetorical sources when they are overturning precedents, invalidating laws, or facing opposition from other justices, all scenarios in which justices are likely to be concerned about how others will perceive the legitimacy of their decisions. These variables do not allow one to make conclusions about precisely whom the justices are hoping to influence when concerns about legitimacy arise; indeed, it is quite possible that they have multiple audiences in mind. But at a minimum, the results are consistent with strategic explanations.
The legal variables also perform as expected, with both variables significant across all four categories of sources. According to Table 5, majority opinion authors are 12.8% more likely to cite one or more rhetorical sources in constitutional cases, give or take 1.4%. They are also 6.5% more likely to cite at least one rhetorical source in complex cases, plus or minus 1.4%. The success of these variables affirms that the use of rhetorical sources reflects the legal context in which the justices are writing. Because sources such as the Federalist Papers and the writings of thinkers such as Brutus and Edward Corwin are highly relevant to constitutional interpretation, one should naturally expect to see them cited more frequently in constitutional cases. The fact that the strategic variables remain robust after controlling for the legal variables also helps further substantiate the independent effect of the strategic variables on the use of rhetorical sources.Footnote 11
The proclivities of particular justices also have a strong impact on the use of rhetorical sources, attaining significance in all but one of the categories of the dependent variable. Justices with strong proclivities are 21.0% more likely to use rhetorical sources, give or take 4.5%.Footnote 12 Neither of the other two measures of judicial preferences has an effect on the use of rhetorical sources, with the exception of ideology, which has a significant influence only on the use of interpretative authorities. Most likely, controlling for the proclivities of particular sources has diminished the impact of these variables. Table 6 reports the bivariate correlations of the Proclivity variable with the other two measures of judicial preferences. It shows that justices with a proclivity for citing rhetorical sources are more likely to be conservative and to have taught at a law school; both correlations are statistically significant at p<0.05. Thus, while these two measures of agency preferences do not have an independent effect on the use of rhetorical sources, they do help explain why certain justices develop proclivities for citing sources.
Table 6. Bivariate Correlations for Proclivity, Ideology, and Law School Variables
Note:
* p<0.05.
The final two independent variables also perform as expected, although not across all categories of the dependent variables. The workload of particular justices is statistically significant in the combined model, as well as in one of the categories of the dependent variable. According to Table 5, as a justice's workload approaches its maximum value, the likelihood of a citation in a majority opinion decreases 9.1 percentage points, plus or minus 3.7%.Footnote 13 The number of clerks in a justice's chambers is not significant in the combined model but does influence the use of interpretive authorities by majority opinion writers. Altogether, these results corroborate the impact of efficiency concerns on the use of rhetorical sources. When justices are more at leisure, either because they are deciding fewer cases in a term or because they have a large number of clerks, they are more likely to use rhetorical sources in their opinions.
Another way to examine the effects of the independent variables is to see if their impact on the use of rhetorical sources has changed over time. Table 7 disaggregates the results of the combined model in Table 4, with a separate coefficient reported for the Warren, Burger, and Rehnquist courts. The results show that since the Warren Court, the effects of the strategic variables have declined, with one of the strategic variables (Formal Alteration of Precedent) diminishing in importance and the effect of another (Number of Dissenting Justices) dropping out completely. By contrast, the importance of efficiency concerns has become more prominent, with the Workload variable attaining significance in the Rehnquist Court. The one curiosity is the Clerks variable, which is statistically significant but contrary to expectations is also strongly negative in the Rehnquist Court. A possible explanation for this observation is that Justices Stevens and Rehnquist, who frequently used rhetorical sources during this period, hired fewer than the maximum number of allotted clerks. Apart from the Clerks variable, the hypotheses generally hold up, with at least one of the variables from each of the categories of independent variables performing as expected across the different time periods.
Table 7. Citations to Rhetorical Sources on the Warren, Burger, and Rehnquist Courts, 1953–1997
Note:
* p<0.05; standard errors in parentheses.
a chibar-square is the test statistic used in Stata for the likelihood ratio test for overdispersing.
Concurrences and Dissents
Table 8 provides information on the citation habits of justices writing concurring and dissenting opinions, with results also reported for nonmajority opinions in general. In order to test the effects of the independent variables on nonmajority opinions, some of the variables had to be recoded because their values were based on attributes of majority opinion writers. New scores were thus needed to reflect the ideologies, proclivities, teaching backgrounds, workloads, and clerks of the concurring and dissenting opinion writers. Also, because the unit of analysis in Spaeth's database is the case, rather than the opinion, citations to rhetorical sources could not be linked with the same precision to concurring and dissenting opinion authors as they could for majority opinion writers. The dependent variable for nonmajority opinions thus measures the number of rhetorical sources appearing in all dissents or all concurrences in a given case, rather than in particular opinions. Appropriate revisions were made to the coding of the independent variables to reflect this change in the measurement of the dependent variable.Footnote 14
Table 8. Citations to Rhetorical Sources in Nonmajority Opinions, 1953–1997
Note:
* p<0.05; standard errors in parentheses.
a chibar-square is the test statistic used in Stata for the likelihood ratio test for overdispersing.
Also included in the model is a new strategic variable measuring whether the use of rhetorical sources by majority opinion writers results in citations by the authors of concurrences and dissents. If justices think strategically, one might expect them to be responsive to the use of rhetorical sources in other opinions. Concurring or dissenting justices might respond to citations in majority opinions by using the same sources; or they might use different sources in an effort to make their opinions appear as persuasive as the majority. The point is that the use of rhetorical sources becomes an interactive process, with concurring and dissenting opinion writers employing them not simply as a reflection of their sincere preferences, but because a majority opinion cited them first. Such interaction is the essence of strategic behavior: its presence would suggest that a dialogue is occurring among the justices, in which they are seeking to persuade either their fellow justices or outside observers of the superiority of their policy choices.
It is theoretically defensible to suppose that majority opinion citations are more likely to influence the use of rhetorical sources by concurring and dissenting authors than the reverse. As Reference MaltzmanMaltzman et alia explain, majority opinion writers generally are not interested in accommodating dissenters, since the chances of winning their votes are not great: “Even if changes are made to the majority in response to the dissent, they are probably unlikely to be significant enough to convince many dissenters to join” (2000:71). Majority opinions are almost always circulated before nonmajority opinions, minimizing opportunities for concurring and dissenting justices to influence the content of the majority's initial draft.Footnote 15 Instances of fluidity, in which a justice on the majority coalition switches to the minority, are relatively infrequent.Footnote 16 Finally, data from this study corroborate the predominantly unidirectional nature of the relationship.Footnote 17 Taken together, the evidence suggests that the model has been appropriately specified.
Table 8 summarizes the results, with Table 9 reporting first differences for the significant variables in the combined model. As with majority opinion citations, all three of the original strategic variables perform as expected, especially in the combined model. Of even greater theoretical importance, the new strategic variable, which measures the number of rhetorical sources used by the majority, also performs as expected. As the number of rhetorical sources used in a majority opinion approaches its maximum value, the likelihood of at least one rhetorical source appearing in a concurring or dissenting opinion increases 57.2 percentage points, give or take 13.0%.Footnote 18 The success of this variable is highly important theoretically because it demonstrates that the use of rhetorical sources is an interactive process. While the other three strategic variables indicate that justices are more likely to use rhetorical sources when the legitimacy of their decisions is low, the performance of this variable more directly establishes that justices use rhetorical sources in response to the behavior of others.
Table 9. First Differences, Non-Majority Opinions (Combined Model)
Note: Change in the probability of at least one rhetorical source appearing in a non-majority opinion, as certain independent variables move from minimum to maximum values.
Comparing the results for concurrences and dissents reveals that the effects of the strategic variables are somewhat stronger for dissenting justices than they are for justices writing concurring opinions. Concurring justices are more likely to cite rhetorical sources when the Court is invalidating a state or federal law, but not when the majority is altering precedent. The effects of dissenting justices on the use of rhetorical sources by nonmajority opinion writers are also weaker for concurrences than for dissents. Finally, the inclusion of references to rhetorical sources by majority opinion writers has no impact on the use of rhetorical sources by concurring justices. By contrast, three of the four strategic variables are statistically significant for dissenting justices, with only the invalidation of a state or federal law exhibiting no effect on citation patterns. Dissenters are responsive to citations by other justices, citing rhetorical sources more often when they appear in majority opinions.
The stronger performance of the strategic variables for dissenting justices makes sense theoretically. Because dissenting justices have lost on the merits, they are likely to have a greater interest in persuading justices from the majority coalition to change their views. They may also wish to discredit the majority's opinions in the eyes of the legal community. Dissenting justices should therefore be expected to respond to the opinions of other justices in an effort to make their own opposing viewpoints appear more persuasive. In contrast, concurring justices, who have already won on the merits, are less likely to be motivated by strategic concerns. Indeed, the fact that a justice's background as a law school professor has an impact on the citation of rhetorical sources in concurrences but not dissents suggests that the motives of concurrence-writers are more didactic than persuasive in origin.Footnote 19
Discussion and Conclusions
Overall, the results indicate that the use of rhetorical sources reflects a combination of motives, with legal considerations, ideological preferences, and efficiency concerns all affecting the shape of judicial opinions. Yet of greatest theoretical significance are the findings of strategic behavior. The data affirm that justices use rhetorical sources more often when the legitimacy of their actions is lowest, such as when they are overturning precedent, invalidating state or federal laws, or issuing directives from a divided bench. It also shows that opinion writers respond to the use of rhetorical sources by other justices, with citations more likely in dissents when majority opinion writers have also used them. While this study has not taken a position on whom justices may be trying to persuade when they use rhetorical sources, it has provided evidence that the assembly of judicial opinions is an interactive process.
The findings of this study have the most important implications for the strategic literature. Strategic theorists have long observed that judicial behavior is influenced by fellow judges and outside political actors (e.g., Reference MurphyMurphy 1964; Reference Spiller and SpitzerSpiller & Spitzer 1992; Reference Epstein and KnightEpstein & Knight 1998; Reference Tiller and SpillerTiller & Spiller 1999; Reference MaltzmanMaltzman et al. 2000; Reference Smith and TillerSmith & Tiller 2002). This article bolsters these claims with evidence that strategic considerations shape the use of opinion content by Supreme Court justices. Rhetorical sources are important because they enhance the legitimacy of an opinion writer's raw preferences. As Epstein and Knight observe, “[b]efore the Court can make authoritative policy that other institutions, the states, and the public will view as binding on them, it must have some level of respect” (1998:11–2).
The findings may also be of interest to legal researchers. Legal commentators have long criticized the poor quality of “law-office history,” or “history lite” (Reference RichardsRichards 1997; Reference TushnetTushnet 1996; Reference FlahertyFlaherty 1995; Reference PowellPowell 1987; Reference MillerMiller 1972; Reference KellyKelly 1965). Scholars in this camp have argued that judges and lawyers who employ history and other types of corroboratory evidence typically use them crudely and only for effect. The present analysis has supported these conclusions with empirical evidence, but it has also suggested more sympathetically that rhetorical sources are among the few resources that justices can use to enhance their legitimacy.
At least two important questions remain. The first is whether instrument-choice strategies work. Do rhetorical sources actually persuade other decision makers? Although studying the impact of rhetorical sources is beyond the scope of this analysis, there is reason to suppose that rhetorical sources do have the effects that justices intend. As discussed above, research on priming indicates that even elite behavior can be influenced by the perceived legitimacy or prestige of policy sources (e.g., Reference Iyengar and KinderIyengar & Kinder 1987; Reference Miller and KrosnickMiller & Krosnick 2000). Subsequent research can examine the conditions under which these citation strategies work, and whether some strategies are more effective than others.
A second question for future research is whether Supreme Court justices employ additional instrument-choice strategies. Justices might, for example, make strategic use of precedent, the legal grounds of decisions, or other forms of opinion content. The evidence presented in this study suggests that such research would be promising.
Appendix: Composition of the Dependent Variable
Studies of this nature inevitably invite controversy about why some sources were chosen in place of others. Why, some might ask, was Brutus included but not Cato? Why were James Madison's non-Federalist writings excluded? Why include Locke but not Rousseau? This Appendix explains the selection process in some detail. To those objecting to the omission of a particular source, it should be noted that substituting one source for another would likely have no meaningful impact on the results. For the purposes of this study, the choice of particular sources was of secondary importance, so long as they met the selection criteria, because all sources were hypothesized to function similarly.
The most important part of the selection process was ensuring that the chosen sources were appropriately classified as rhetorical. It was especially important to distinguish rhetorical sources from other materials that judges commonly use to resolve cases, such as references to precedent, legislation, and administrative proceedings. To that end, four criteria were developed to guide selection. To qualify as a rhetorical source, each of the sources chosen were (1) nonbinding on case outcomes; (2) well-respected by the legal community; (3) widely applicable, appearing in a broad assortment of cases; and (4) not contemporaneous with the period under analysis. It is acknowledged that the selection criteria permitted a good deal of discretion in the choice of sources. However, as noted above, the results of the study should not depend on which sources are chosen, so long as they are correctly identified as rhetorical sources.
Using these criteria ensured that other types of citations were not included in error. For example, references to Chief Justice Earl Warren were inappropriate, even though Justice Warren might often be cited rhetorically, because of a failure to meet the fourth criterion. Since Justice Warren served on the Court during the years covered by the study, it is quite likely that many citations to Justice Warren are not rhetorical, referring to recent precedent. To the extent that citations to precedent or references to contemporaneous or other sources, such as social science evidence, might function differently, these materials were not included in the analysis.
Sources were divided into four broad categories. The first consisted of Interpretive Authorities, written commentaries on the Constitution or English common law that are widely respected by the legal community. Though many commentaries might have been used, four of the most prominent were chosen: the Federalist Papers, Story's Commentaries on the Constitution, Blackstone's Commentaries on the Laws of England, and the writings of Sir Edward Coke. To identify cases citing these sources, Lexis keyword searches were conducted for “federalist,”“story and (commentary or commentaries),”“blackstone,” and “coke.” Cases were then screened for their appropriateness and for whether the reference was made in the majority opinion. Four new dummy variables were then created, coding 1 for any mention of the source in the majority opinion and 0 when there was no reference. A similar procedure was employed for the category of Founding Documents, which are documents that express the underlying principles of American government but do not control case outcomes. The founding documents selected were the Magna Carta, the Declaration of Independence, and the Preamble to the federal Constitution. Variables were coded as before, with Lexis searches conducted for “(magna) and (carta or charta),”“declaration and independence,” and “preamble.” Any reference to the source in the majority opinion counted as a citation.Footnote 20
The third and fourth categories included Respected Jurists and Respected Nonjurists, individuals whose interpretations of the Constitution and the principles of American government are highly regarded. Selecting the four respected jurists involved consulting the Appendix of Reference AbrahamAbraham's Justices, Presidents, and Senators (1999), which reports the results of surveys asking legal experts to provide their impressions of the greatest justices. Only four justices were ranked as “great” by a majority of the surveys—John Marshall, Oliver Wendell Holmes, Louis Brandeis, and Benjamin Cardozo—and each of these individuals was included in the analysis. Justices Hugo Black, Warren, and Frankfurter were also mentioned on the lists, but they were excluded because they served on the Court during the years covered in the analysis and thus do not meet the fourth criterion. To find appropriate cases, Lexis keyword searches were conducted for “chief justice marshall or john marshall or marshall, c,”“cardozo,”“brandeis,” and “holmes.” All references appearing in majority opinions were counted as citations; no distinction was made between references to justices' opinions and their off-the-bench writings.
Respected nonjurists were drawn from three broad categories: founders, philosophers, and scholars. The founders chosen were Thomas Jefferson, James Wilson, and Brutus, who were prominent members of the constitutional or ratifying conventions. James Madison and other contributors to the Federalist Papers were excluded because the Federalist Papers was already included in the analysis. The philosophers, who were influential contributors to American political thought, were John Locke, John Stuart Mill, and Montesquieu. The scholars, prominent writers on American government and constitutional law, were James Bradley Thayer, Woodrow Wilson, and Edward Corwin. To find these sources, Lexis searches were conducted for “jefferson,”“wilson,”“brutus,”“locke,”“mill,”“montesquieu,”“thayer,” and “corwin.” Intercoder reliability statistics indicated that the coding criteria for all four categories of the dependent variable were robust. A random sample of 200 cases was selected and recoded by a second reader. The interagreement yielded a Cohen's kappa of 0.875.
