WAR AND IMAGINED SOVEREIGNTY

How do wars shape the popular imagination of sovereignty? Given the novelty of this research question, there exists little in the way of established theory to guide our reasoning. In this section, we draw upon disparate literatures to identify two potential pathways through which war may affect ideational outcomes. The following sections then apply these ideas to the particulars of the U.S. case in order to derive specific testable hypotheses.

Our approach begins from the observation that wars and their aftermath often constitute critical junctures in a country’s history that relax prior ideational constraints and establish new beliefs, values, and understandings (Capoccia and Kelemen Reference Capoccia and Daniel Kelemen2007; Soifer Reference Soifer2012; Widmaier, Blyth, and Seabrooke Reference Widmaier, Blyth and Seabrooke2007). In particular, wars often entail massive social, economic, and institutional disruptions in everyday life. Conscription, the curbing of political freedoms, the reshaping of the labor force and economy—to say nothing of the death, physical destruction, and displacement caused by fighting—all represent a sudden and violent break from the normal rhythms of life. These disruptions shake loose the structure of established ideas and increase the demand for new explanatory models of traumatic events. It is in this course of “searching for meaning” that new ideas about sovereignty take hold.

Building from this general framework, we identify two potential pathways through which wars may generate ideational change. The first perspective links ideational change to the costs of war and the need to mobilize public support. Specifically, major wars in the era of mass mobilization demand tremendous sacrifice on the part of citizens (Koehler-Derrick and Lee Reference Koehler-Derrick and Lee2023; Scheve and Stasavage Reference Scheve and Stasavage2010; Tilly Reference Tilly1992). These sacrifices include compliance with the state’s extraction of labor to fight wars and the taxes to finance them, as well as the willingness to endure shortages in everyday essentials such as food and fuel.

To persuade citizens to bear the costs of wartime disruptions, leaders must work to justify why such sacrifices are necessary (Chong and Druckman Reference Chong and Druckman2007; Lee and Prather Reference Lee and Prather2020; Maxey Reference Maxey2020; Nelson, Oxley, and Clawson Reference Nelson, Oxley and Clawson1997; Widmaier, Blyth, and Seabrooke Reference Widmaier, Blyth and Seabrooke2007). These justifications are important because ordinary citizens are unlikely to possess the specialized knowledge or political awareness to make sense of war on their own. Political issues can be complex and remote; acquiring political information independently of elites is not only costly but even prohibitive (Downs Reference Downs1957; Gilens and Murakawa Reference Gilens and Murakawa2002; Guisinger and Saunders Reference Guisinger and Saunders2017). Leaders can fill the information gap by framing, interpreting, and explaining the causes of war and issues under dispute.

Themes such as the preservation of national security or the protection of economic interests figure prominently in these appeals to the public. Just as often, however, justifications for war are intimately bound up with ideational issues such as sovereignty. For example, although the Zimmerman Telegram and Germany’s resumption of unrestricted submarine warfare directly precipitated the entry of the United States into World War I, President Woodrow Wilson justified America’s involvement—and the immense sacrifices that this would entail—by invoking the higher principles of national self-determination and the safeguarding of democracy.

Political actors play an important role in advancing such higher principles as justifications for conflict. Political leaders, intellectuals, and activists thus assume the role of ideational entrepreneurs who take advantage of wartime upheavals to rally public support around the issues to which they are committed. The influence of these entrepreneurs lies in their ability to (re)frame issues and ideas in ways that both resonate with the broader public and help to make sense of the traumas of war (Finnemore and Sikkink Reference Finnemore and Sikkink1998, 897).

When such frames are adopted, they can serve to elevate popular commitments to principles such as popular sovereignty or democracy by casting these ideas as “rightful” causes deserving of wartime sacrifice. In this way, justifications based on higher principles can have a psychological effect by introducing normative valence to the conflict. The sense of moral imperative that results not only increases citizens’ willingness to bear the costs of war, but also convinces them that the principle in question is a “good” cause worth fighting and dying for.

In summary, this pathway focuses on processes of framing, persuasion, and justification as critical mechanisms, and political actors as key agents of change. In seeking to explain the costs of war, these actors increase the valence of their preferred ideas. By implication, this perspective holds that we should observe the greatest ideational change among the ideological vanguard and their followers. We may also expect that the vanguard, being comprised of politically sophisticated elites, will more quickly adopt new ideas compared with their non-elite, less politically sophisticated followers.

A second, alternative pathway focuses not so much on the normative effects of war itself, but instead stresses the process of “sense-making” that follows the shock and trauma of defeat. Of course, in many ways, defeat—particularly when it is accompanied by military occupation and the imposition of new institutions and leaders—represents an intensification of the wartime disruptions and uncertainties that increase the demand for new explanatory models. Importantly, however, defeat also generates a qualitatively different set of effects. Unlike the questioning that accompanies wartime sacrifices, the “search for meaning” here is intimately related to a sense of failure, humiliation, and loss of status that defeat lays bare in clear and unequivocal terms (Barnhart Reference Barnhart2021).

These negative emotions play an important role in shaping ideational outcomes. Our argument here builds upon Legro’s two-stage model of change (Legro Reference Legro2005). The first stage underscores the role of defeat in calling into question prior ideational commitments. Here it is important to recognize that the sense of failure accompanied by military defeat is not confined to events on the battlefield alone, but also extends to the implication of those events for the higher-order principles contested in the war itself. For example, the August 2021 fall of Kabul to the Taliban and the chaotic American withdrawal from Afghanistan prompted pundits, scholars, and ordinary Americans to question the proper role of the United States in the world more generally. Similarly, historian John Dower emphasized the delegitimation of the ideational status quo ante among ordinary Japanese following the country’s unconditional surrender in 1945: “When the war ended in disaster and utter defeat, it was obvious that the ‘New Order’ [Greater East Asia Co-Prosperity Sphere] and the ‘New Structure’ had been miserably conceived” (Dower Reference Dower1999, 178–9). As these examples illustrate, defeat can reverberate beyond the battlefield to signify the failure of more basic ideational principles.

Secondly, this sense of failure, as well as the associated feelings of humiliation and loss of status, engenders not only a rejection of the status quo, but also a search for alternative principles and ideas. Legro particularly identifies the quality of success as a key criteria in selecting among available alternatives (Legro Reference Legro2005, 35). Ironically, in the aftermath of a major war, the most salient successful model may be found in the principles and ideals of one’s victorious adversary. The triumph of arms, as well as the resulting difference in status between the winning and losing sides, are power symbols of success. In this context, the attraction of the victor’s model, coupled with the shock and trauma of defeat, may drive the vanquished toward an ideational reversal. Dower describes this process in post-WWII Japan as a “‘the reevaluation of all values,’ even to the point of prompting a thorough-going iconoclastic feeling of ‘damn what you have adored and adore what you have damned”’ (Dower Reference Dower1999, 245).

This reversal pathway thus emphasizes the focal quality of the victor’s ideals in combination with the humiliation of defeat as the key mechanisms driving ideational change. By offering a salient exemplar of success in a context where defeat has discredited prior alternatives, the winning side may establish its ideals among its former foes. Consequently, we expect the vanquished to abandon the old ideational orthodoxy and accept ideas from the opposing side.

To summarize, both pathways of ideational change begin from the observation that war is a highly disruptive phenomenon. They also share an emphasis on interpretation and meaning, and specifically on the ideals and principles over which wars are fought.Footnote ¹ However, they differ in key respects. The valence pathway stresses framing, justification and normative “rightness,” along with the agency of ideational entrepreneurs in providing those frames; the reversal pathway instead emphasizes the search for meaning in the aftermath of defeat, and the focality of the victor’s ideals as replacements for those of the vanquished. To be clear, we do not argue that ideational entrepreneurs play no role in the reversal pathway. Rather, in our conception, such individuals are not critical for the reversal pathway to operate, whereas the valence pathway assigns them a central role as agents of change.

SOVEREIGN AUTHORITY IN THE UNITED STATES

We now apply our theoretical discussion about the ideational effects of warfare to the particular context of nineteenth-century America. The U.S. case is attractive in that it features two clear, competing visions about how sovereign authority was to be organized. Indeed, constitutional debates about the location of final sovereign authority are as old as the United States itself. In designing the Constitution, the founders set up a system whereby sovereignty was “neither wholly national nor wholly federal,” an arrangement that reflected the uneasy union of two visions about how sovereign authority in the United States was to be structured (Holt Reference Holt2019, 24; Madison Reference Madison1788). Importantly, questions revolved around whether the United States was a union of multiple sovereign states, or rather a single nation where final authority rested in the hands of the federal government.

In the years leading up to the Civil War, sectional divisions lent new intensity and urgency to debates over sovereign authority (Morison Reference Morison1968, 48–52). To be sure, the sectional conflict revolved primarily around the future of slavery—a crucial point to which we will return below (Bensel Reference Bensel1990; Bonekemper Reference Bonekemper2015; Loewen Reference Loewen, Loewen and Sebesta2010; McPherson Reference McPherson1988, 241). But the South also couched its position on slavery within a particular view of states’ rights vis-à-vis the federal government and justified the legality of secession as a means of defending those rights. Further, the act of secession and the North’s decision to fight to preserve the Union turned the Civil War into an armed conflict over a constitutional issue. At stake were the very questions that had dogged the country from its birth: the configuration of sovereign authority and the kind of nation the United States was supposed to be (Oakes Reference Oakes2013, 50, 451–2).

A central tenet of the Southern position was the belief that the Union was created as a compact of equal sovereign states, and thus its sovereignty was derivative of its constituent parts. Jefferson Davis aptly summarized this position when he declared, “I am a citizen of the United States, it is true, because I am the citizen of a State.” McPherson further elaborates the Southern view that “State sovereignty…preceded national sovereignty. When they had ratified the Constitution, states delegated some of the functions of sovereignty to a federal government but did not yield its fundamental attributes” (McPherson Reference McPherson1988, 240). And since the Union was a voluntary compact of sovereign states, these states, as sovereign authorities, could resume their original independence when their interests (e.g., in protecting slavery) were threatened.

In contrast, Northern elites believed in a more organic, national conception of sovereignty. The Union was more than a compact, and also more essential than the sum of its parts. Rather, the United States was an “unbreakable, perpetual national community” created by the American people (Maizlish Reference Maizlish2018, 125).Footnote ² Or, as New York Whig Elbridge Spaulding proclaimed, “We are…one people, one country, having mutual interests, reciprocal duties, and a common destiny” (Maizlish Reference Maizlish2018, 126). Sovereign authority in such a national community was naturally located in a “consolidated government of one people” (Maizlish Reference Maizlish2018, 96). The fight to preserve this consolidated government inaugurated one of the critical junctures in American history.

THE CIVIL WAR AND IDEATIONAL CHANGE

The Civil War imposed enormous disruptions on American society, both North and South. Roughly three million men—some 20% of the entire white male population in 1860—served in the conflict (McPherson Reference McPherson1988, 306). Casualty estimates also paint a striking picture, as approximately “1 in 10 white men of military age in 1860 died as a result of the war and two hundred thousand white women were widowed” (Hacker Reference Hacker2011, 311). Losses were particularly acute across the South, where the conflict “killed two-fifths of southern livestock, wrecked half of the farm machinery, ruined thousands of miles of railroad, left scores of thousands of farms and plantations in weeds and disrepair, and destroyed the principal labor system [i.e., slavery] on which southern productivity had been based. Two-thirds of assessed southern wealth vanished in the war” (McPherson Reference McPherson1988, 818).

The valence pathway predicts that such dramatic upheavals should prompt political actors to advance ideas that motivate and mobilize citizens to bear the costs of conflict. By justifying citizens’ sacrifices in service to higher principles, ideational entrepreneurs give “meaning to the war” while popularizing their preferred understanding of American sovereignty.

From the Northern perspective, the key justification held that war was necessary to preserve the existence of a “perpetual Union.” This principle privileged the sovereignty of a national popular majority over the sovereignty of states’ rights. While such views certainly predated the conflict, the South’s secessionist challenge lent renewed vitality to these ideas. Indeed, for Northern reformers of all stripes, “as never before, the war mobilized [their] energies…imbuing their lives with a renewed sense of purpose” (Foner Reference Foner1988, 25). Further, the act of violent secession delegitimatized the opposing ideal of states’ rights, turning “even those who had looked forward with equanimity to separation from the South into violent foes of any compromise” (Trefousse Reference Trefousse1969, 168–9).

The key proponents of perpetual Union—the ideational entrepreneurs—were members of the Radical wing of the Republican Party (Bensel Reference Bensel1990, 11, 18–9). Before the war, Republicans and some of their Whig predecessors had supported economic policies that would have exercised or even increased national power at the expense of the states (Foner Reference Foner1970, 36). For Radicals, moreover, the question of federal authority became deeply bound up with the slavery issue. Indeed, “the driving force of Radical ideology was the utopian vision of a nation whose citizens enjoyed equality of civil and political rights, secured by a powerful and beneficent national state” (Foner Reference Foner1988, 230).

Secession, followed by Emancipation, provided Radicals with a critical opportunity to advance their ideals. Radicals viewed the Civil War as “a ‘golden moment,’ an opportunity for far-reaching change that, if allowed to pass, ‘will have escaped for years, if not forever”’ (Foner Reference Foner1988). Northern victory further propelled optimism at the prospects of far-reaching changes. In the aftermath of the war, Republicans resolved to make the nation a “‘custodian of freedom,’ and some [reformers] questioned whether the states deserved continued existence at all” (Foner Reference Foner1988, 24). Given these considerations, we offer the following valence hypothesis:

We could, at this point, also formulate a parallel hypothesis about the war promoting adherence to the states’ rights ideology in the South. However, we refrain from doing so for several reasons. First and foremost, many historians contend that for Southerners, the war was about the future of slavery, not states’ rights.Footnote ³ Perhaps the most striking illustration of this view can be found in the famous “Cornerstone speech” that Confederate Vice-President Alexander Stephens delivered two weeks prior to the outbreak of the Civil War. Stephens explicitly pronounced that “African slavery…was the immediate cause of the late rupture and present revolution,” and explained that “Our new government is founded upon…the great [moral] truth, that the negro is not equal to the white man; that slavery…is his natural and normal condition” (Stephens Reference Stephens, Loewen and Sebesta2010, 188).

More broadly, Stephens’ speech points to the lack of Southern ideational entrepreneurs pushing the states’ rights issue as a more general principle rather than a narrow and instrumental defense of slavery. Even the most vehement supporters of secession, a group of individuals known as the Fire-Eaters, opportunistically treated states’ rights as a means to an end—that end being the protection of slavery (Heidler and Heidler Reference Heidler and Heidler2014). As such, we would expect that Southern justifications for war would serve to increase popular attachments to the South’s “peculiar institution,” but have little effect on the normative valence of states’ rights.

In contrast to the valence pathway, the reversal pathway stresses the shock of defeat as the catalyst for the reevaluation of ideational attachments in the South. Defeat was traumatic for the Confederacy: as late as May 1864, Southern civilians and soldiers alike remained optimistic about the prospects of prevailing over the Union (Gallagher Reference Gallagher1997, 36–8). However, the capture of Atlanta and Sherman’s subsequent march to the sea, followed closely by the re-election of Abraham Lincoln, dashed Southern hopes for a negotiated peace. With the surrender of the Confederate armies in 1865, “civilians who had maintained faith in their defenders despite material hardship and social disruption…recognized that the end had come” (Gallagher Reference Gallagher1997, 157).

The upheavals wrought by the South’s failed attempt at secession are difficult to understate: “Four years of brutal struggle had ravaged their military-age male population, vastly altered their physical landscape and economic infrastructure, and destroyed their slave-based social system” (Gallagher Reference Gallagher, Gallagher and Nolan2000, 1). Postwar Reconstruction and military occupation further underscored the totality of the South’s defeat and added to its humiliation. In particular, the presence of federal troops was a constant reminder of the Confederacy’s downfall and of Southerners’ status as second-class citizens (Downs Reference Downs2015, 14). Finally, the presence of formerly enslaved Blacks among the ranks of the army and in public office deepened Southern whites’ sense of humiliation (Rable Reference Rable2007, 24, 62).

The reversal pathway predicts that the total shock of Southern defeat should lead to a deep dissatisfaction with ideas and principles upon which the Confederacy stood while precipitating an embrace of opposing ideals. In the South, the individuals who supported the Republican Party and their vision of strong national authority were disparagingly called “scalawags.” The scalawags were a diverse group that counted both wartime Unionists as well as former Confederates among their numbers. Although the former group had never supported secession, the same could obviously not be said for the former Confederates—at least some of whom appeared to have abandoned the old Southern orthodoxy in the wake of defeat.

This position is most evident in the views of some prominent ex-Confederates. General James Longstreet, known as Robert E. Lee’s “Old War Horse,” offers one example. “Since the North had won the war, its principles became law,” Longstreet wrote in a private letter that later appeared in print, “and the South’s duty was ‘to abandon ideas that are obsolete and conform to the requirements of law”’ (Richter Reference Richter1970, 222). Thomas Settle, a former Confederate captain who helped found the North Carolina Republican Party, also illustrates the postwar about-face in attitudes: “Prewar ‘ideas and feelings,’ [he declared,] must be buried ‘a thousand fathoms deep…. Yankees and Yankee notions are just what we want in this country” (Foner Reference Foner1988, 299).

Despite this evidence, other factors suggest that the reversal pathway is unlikely to obtain in the specific case of the American South. Southerners like Longstreet who reversed their ideological positions after the war constituted a minority within the scalawag minority. Although their views attracted considerable attention, they were also vilified in the South as sectional traitors (Richter Reference Richter1970, 215). More broadly, the Southern backlash to Reconstruction offers powerful counter-evidence against the reversal hypothesis. Nonetheless, to give full consideration to a more generalizable theoretical proposition, we test the following hypothesis:

IMAGINING SOVEREIGNTY: FROM PLURAL TO SINGULAR

The ideational nature of sovereignty poses significant measurement challenges. Ideally, we would construct a measure of the popular imagination of sovereignty using public opinion surveys but, unfortunately, our period of study predates the advent of scientific polling. Since we cannot measure what people think about U.S. sovereignty, we instead turn to their writings and speeches about the United States. Our approach assumes that individuals’ language patterns reveal important information about what they think (Liu Reference Liu2021; Pérez and Tavits Reference Pérez and Tavits2019; Tavits and Pérez Reference Tavits and Pérez2019). We therefore surmount the measurement challenge by studying the shared civic language as a window into the popular imagination of sovereignty.

Our measurement strategy leverages a grammatical change in the civic language in which the term “United States” shifted from a plural to a singular noun. Whereas at the beginning of the nineteenth-century Americans said “the United States are,” by the end of the century they were much more likely to say “the United States is.” We treat plural/singular usage as a proxy for the popular imagination of sovereignty. Specifically, we take plural usage as indicating that Americans view the United States as having multiple sovereignties embodied in the several states, and singular usage as indicating that Americans conceive of the United States as possessing a single national sovereignty.

We are not the first scholars to attribute such meaning to the usage of the plural and singular forms of the United States. For example, Civil War historian Shelby Foote linked sovereignty and speech in his observation that “Before the war, it was said ‘the United States are.’ Grammatically, it was spoken that way and thought of as a collection of independent states” (quoted in Ward Reference Ward1990, 273). Similarly, McPherson notes that the Civil War “marked a transition of the United States to a singular noun” and that “the ‘Union’ also became the nation” (McPherson Reference McPherson1988, 859).

Contemporary observers in the nineteenth century also treated language as laden with sovereign meaning. The Washington Post editorial board’s defense of its use of the grammatical singular is illustrative:

Thirty years later, a prominent former Texas legislator, newspaper editor, and columnist at the Rusk County News (Langston Reference Langston1941) linked speech and sovereignty in his criticism blasting a rival newspaper’s grammatical choices:

Perhaps most tellingly, neither Southerners nor Northerners considered the Confederate States of America to be a singular noun, much less a singular entity. For example, in his inaugural address, Jefferson Davis referred to the Confederate States in plural terms when describing “their” militias (Davis Reference Davis, Crist and Dix1861). In our own analysis of rhetoric in newspapers and congressional speeches, we find 95 instances in which the “Confederate States (of America)” appeared as a grammatical subject and was treated as a plural noun, and exactly two instances in which the singular was used.Footnote ⁴

METHODS AND DATA

We follow existing scholarship in the digital humanities by identifying singular usage in our textual sources through subject–verb agreement (Myers Reference Myers2008; Santin, Murphy, and Wilkens Reference Santin, Murphy and Wilkens2016). Specifically, the form of the verb reveals whether its subject is singular or plural. This approach requires the term “United States” to appear as a grammatical subject (i.e., “the United States are unwilling to pay” (Johnson Reference Johnson1864, 2929)), rather than as a grammatical object (i.e., “the commissioners who represented the United States are now dead” (Mann Reference Mann1914, 7715). We further restrict our search to three common verb pairs that are unambiguously singular or plural: is/are, has/have, and was/were. Details on the automated text mining processes, we use are provided in the dataverse appendix. The resulting measure of our dependent variable is an indicator taking 1 if a mention of the “United States” in the subject position treats it as a singular noun and 0 otherwise.

We employ two types of textual sources: American newspapers and congressional speeches. Our newspaper data cover the period of 1800–99 and are drawn from Nineteenth-Century U.S. Newspapers and America’s Historical Newspapers; both are proprietary databases containing OCR text (Gale 2021; Readex 2021). Because we are interested in examining geographic differences in language use, we impose the additional restriction that newspaper content must be generated locally. The practice of reprinting material from other publications, including wire service content after the founding of the Associated Press, violates this restriction. Because there is no consistent way to track wire service and reprinted content across the multiplicity of publications in our dataset, we limit our attention to content that is likely to be locally generated: editorials and letters to the editor.

Congressional speech data come from OCR text conversions of the Congressional Globe, a private journal founded in 1833 to report on the daily proceedings of Congress, and the Congressional Record, the journal that succeeded the Globe in 1873 and carried official status (Phillips Reference Phillips2015; Gentzkow, Shapiro, and Taddy Reference Gentzkow, Shapiro and Taddy2018, respectively). By 1850, improvements in stenographic technology and the professionalization of the reporting corps allowed these publications to provide verbatim accounts of congressional speech (Byrd Reference Byrd1988, 314; McKinney Reference McKinney2002, 17; McPherson Reference McPherson1942, 147). We are therefore able to link speeches to individual members of Congress from the 32nd Congress (1851–53) and all meetings thereafter until 1899.

Each textual source offers distinct advantages for our analysis. The newspaper corpus covers the entirety of the nineteenth century and contains both political and non-political content authored by non-elite (or at least less elite) writers.Footnote ⁵ The newspaper corpus thus serves as the main focus of our analysis. In contrast, the congressional speech corpus provides additional analytical depth. In particular, since we can link statements to specific (elite) individuals and factors such as party identification, we can directly examine the responses of ideational entrepreneurs to the war.

Importantly, the ability to distinguish between elite speakers in the congressional corpus and non-elite authors in the newspaper corpus also allows us to examine an observable implication from the theory. Specifically, the valence pathway’s emphasis on the role of ideological entrepreneurs implies that postwar change in ideational adherence should appear more rapidly among the elite in comparison to their non-elite, less politically sophisticated followers. We therefore expect to observe more discontinuous or immediate ideational shifts in the elite speech data, and more gradual (i.e., “smoothed”) change in non-elite newspaper data.

Because our hypotheses concern the sectional effects of the Civil War, we restrict our attention to newspapers headquartered in U.S. states and territories that (eventually) participated in the conflict. For consistency, our congressional speech sample includes mentions from Congressmembers and delegates from the same set of states and territories. In both cases, we omit mentions from the border states of Delaware, Kentucky, Maryland, and Missouri. All four were slave states that did not (or were prevented from) joining the Confederacy. As such, we do not consider their participation in the war to be comparable to other members of the Union. For similar reasons, we exclude all mentions from West Virginia, which seceded from Virginia after Virginia seceded from the Union. Finally, our main analyses also omit Washington, DC as a “border” area.Footnote ⁶

SECTIONAL EVIDENCE

We begin with a visual inspection of sectional patterns in language usage. We focus on the newspaper corpus given its longer temporal span. Figure 1 depicts singular usage as a proportion of all grammatical subject mentions by year, separately for Northern and Southern newspapers. In general, we observe a consistent movement toward singular usage throughout the course of the nineteenth century. Our data thus replicate the broad temporal patterns that prior studies have documented using different textual corpuses (Myers Reference Myers2008; Santin, Murphy, and Wilkens Reference Santin, Murphy and Wilkens2016).Footnote ⁷ That said, we also uncover evidence of novel sectional patterns. Specifically, whereas Northern and Southern trends appear similar at the beginning of the century, we begin to observe a sectional divergence around time of the Civil War, with singular usage appearing to accelerate in the North but not in the South.Footnote ⁸

FIGURE 1.

Singular Usage in Northern and Southern Newspapers

Note: The top and middle panels show temporal trends in singular usage in editorials and letters to the editor fitted using lowess and a 60% bandwidth. Size of the bubbles indicates the number of publications in each year. The bottom panel shows the first differences of the lowess lines (i.e., estimated singular in year t – estimated singular in year t–1). The vertical bars show the years of the Civil War.

Overall, these patterns provide suggestive evidence in favor of the valence hypothesis predicting a positive effect of the Civil War on singular usage in the North. Further, the continuation of prewar trends in the South indicates that defeat had little effect and speaks against the reversal hypothesis. We next turn to a regression framework to more rigorously evaluate these findings.

Because singular usage was already increasing over the course of the nineteenth century, change in the context of those trends could take different forms. First, the Civil War could have increased the rate of adoption of the singular, which would appear in the data as a change in slope. Second, the Civil War could have immediately increased the level of singular usage, which would appear as a “jump” or intercept shift. Changes in the level of usage imply a rapid shift, and we are more likely to observe that shift in our elite congressional sample. However, because both types of changes are possible, we test for both effects. Our regression models thus take the form:

(1) $$ \begin{array}{rl}\begin{array}{rll}{S}_{ij}=\alpha & +{\beta}_1{Y}_{ij}+{\beta}_2{P}_{ij}+{\beta}_3{N}_{ij}& \\ {}& +{\beta}_4{Y\cdot P}_{ij}+{\beta}_5{Y\cdot N}_{ij}+{\beta}_6{P\cdot N}_{ij}\\ {}& +{\beta}_7{Y\cdot P\cdot N}_{ij}\\ {}& +\delta {W}_{ij}+{\epsilon}_{ij},\end{array}& \end{array} $$

where S is a dummy variable indicating an instance i of singular usage in newspaper j, Y represents the year of publication and captures time trends in the data, P is a period indicator (either post-1860 for tests of the valence hypothesis or post-1865 for tests of the reversal hypothesis), N indicates that the newspaper was published in the North, and W represents a vector of covariates.

We estimate Equation 1 using a linear probability model with standard errors clustered within newspapers. In secondary models, we also include newspaper fixed effects to help address concerns about unobserved heterogeneity related to the entry and exit of newspapers into our dataset. The number of newspapers grew rapidly in the early nineteenth century, in part a consequence of population growth, territorial expansion, and the advent of new technology (American Antiquarian Society 2022). While some publications still exist today, others ceased production or merged with other publications. Newspaper fixed effects help ensure comparability between these publications.

Our covariates W attempt to account for potentially confounding variables that are plausibly correlated with political geography, time, and the adoption of the grammatical singular. In the nineteenth century, the most likely confounders relate to economic and political modernization. We operationalize modernization using historical and geospatial data and code the following variables at the county level: urbanization, whether the county has a steamboat-navigated river or canal, ruggedness, and post office density as a proxy for contact with the government. Covariates are described in detail in Appendix D1.

We begin by examining sectional differences in singular usage before the Civil War. We estimate Equation 1 centering Y at the year 1860 and setting $ P=1 $ if $ Y>0 $ for ease of interpretation. To facilitate presentation of the interaction effects, we display estimated linear combinations of the relevant coefficients in Figure 2. Full results, including estimates for the control variables, appear in Appendix Table D3.

FIGURE 2.

Singular Usage in Northern and Southern Newspapers before 1861

Note: This figure displays coefficients, linear combinations, and 95% confidence intervals from Equation 1, centering Y at the year 1860 and setting $ P=1 $ if $ Y>0 $ . Full results are available in Appendix Table D3.

Our baseline model indicates that singular usage in both sections increased at a rate of about 0.6% per year during the prewar period, such that by 1860, around 35% of statements referred to the United States as a singular noun. Importantly, we detect no statistically significant differences in usage between Northern and Southern newspapers. Our fixed effects models indicate a slower rate of increase of about 0.3% per year, but again no significant sectional differences. Finally, we note that, in general, there are few significant associations between singular usage and our control variables.

We next turn to the valence hypothesis which predicts an increase in singular usage in the North as commitments to the idea of national sovereignty deepen as a result of the war. While we are somewhat agnostic about the exact year where one might begin to observe this effect, the outbreak of war in 1861 strikes us as a reasonable break point in the data. We therefore look for differences in both the level (i.e., an intercept shift) and rate of change (i.e., a slope change) of singular usage around this date. Figure 3 presents the relevant coefficients (and their linear combinations); full results appear in the dataverse appendix.

FIGURE 3.

Singular Usage in Northern Newspapers before and after 1860

Note: This figure displays coefficients, linear combinations, and 95% confidence intervals from Equation 1, centering Y at the year 1860 and setting $ P=1 $ if $ Y>0 $ . Full results are available in Appendix Table D3.

Our results confirm the visual patterns from Figure 1. Specifically, we observe a marked increase in the rate of change in singular usage in Northern newspapers around 1860. In particular, the difference between Northern postwar and prewar time trends is about 0.9% per year. To put this result in perspective, recall that our fixed effects estimate yields a prewar time trend of around 0.3% per year, which now accelerates to 0.3% + 0.9% = 1.2% per year after 1860. Our baseline estimate without fixed effects yields a similar result (from 0.6% to 1.4%). In substantive terms, these estimates imply that, while the transition for plural to singular was only about one-third complete by 1860, the process would reach a rapid conclusion in the North by the turn of the century.

We also re-estimate Equation 1 centering Y iteratively at all years from 1820 to 1880. This exercise is important for identifying the Civil War as a unique period in the nineteenth century that altered the trajectory of singular adoption. However, because the trend in singular usage in the North is convex and upward-sloping (i.e., taking the form of an elongated “J”), we are likely to observe statistically significant slope changes at alternative “placebo” time points, simply as an artifact of the shape of data (see Section S3 of the Supplementary Material for a demonstration using simulated data). Thus, our placebo tests need to account for this underlying issue.

Rather than statistical significance alone, we also focus on the substantive size of slope changes at various points in time. Conceptually, the difference in pre- and post-trends should be greatest at the “true” inflection point. Thus, if the Civil War is indeed the turning point around which the trend in singular adoption accelerates, then the coefficient of slope change should reach a local maximum around this date. Figure 4 shows this is indeed the case: although there are some anomalies relating to time points at the very extremes of our time period, overall we observe the largest slope changes around the early 1860s. We take these patterns as further evidence that the Civil War was a critical juncture in the North.

FIGURE 4.

Slope Changes in Northern Newspapers at Each Year 1820–80

Note: This figure displays slope changes in the time trend of singular usage in Northern newspapers, centering Y iteratively at each year from 1820 to 1880 and setting $ P=1 $ if $ Y>0 $ . Slope changes that are not statistically significant are depicted as hollow circles.

We conclude this section by examining language patterns in Southern newspapers to test the reversal hypothesis. Since this pathway focuses on the effects of Southern defeat, we re-estimate Equation 1 using 1865 as the break point. The relevant coefficients, presented in Figure 5, show that the downfall of the Confederacy does not affect singular usage in Southern newspapers: we observe neither a statistically significant one-time increase in singular usage, nor a significant steepening of the postwar time trend. Further, as we show in Figure S4 in the Supplementary Material, the Southern time trend in singular usage appears roughly constant throughout the entire period 1820–80.

FIGURE 5.

Singular Usage in Southern Newspapers before and after 1865

Note: This figure displays coefficients, linear combinations, and 95% confidence intervals from Equation 1, centering Y at the year 1865 and setting $ P=1 $ if $ Y>0 $ . Full results are available in Appendix Table D4.

Overall, our results thus far suggest that the war did promote greater adoption of national sovereignty in the popular imagination, but only among the Northern population. We take this as evidence in support for the valence hypothesis. Further, consistent with the weight of historical accounts, we find no indications of an ideational reversal in the South. The following section digs deeper into these findings by examining the entrepreneurial influence of the Republican Party in driving the valence mechanism.

EVIDENCE ON MECHANISMS

We begin with the congressional speech data, which allow us to link an individual’s language patterns to his partisan identity. We argue above that Republicans, much more so than Democrats, were most committed to the ideal of a sovereignty embodied in the national government. Our argument implies that we should observe that Northern Republicans adopt the singular at higher rates than Northern Democrats. Further, since Congressmen are political elites, we expect to observe more immediate or sudden changes in their speech patterns.

We create categorical variables for Democrats ( $ DEM $ ) and Whigs+Republicans ( $ REP $ )Footnote ⁹ and estimate the following model (treating $ DEM $ as the reference category and excluding members of third parties):

(2) $$ \begin{array}{rl}\begin{array}{rll}{S}_{isc}& =\alpha +{\beta}_1{Y}_{isc}+{\beta}_2{P}_{isc}+{\beta}_3{REP}_{isc}& \\ {}& +{\beta}_4Y\cdot {P}_{isc}+{\beta}_5Y\cdot RE{P}_{isc}+{\beta}_6P\cdot RE{P}_{isc}\\ {}& +{\beta}_7Y\cdot P\cdot RE{P}_{isc}\\ {}& +\delta {W}_{isc}+{\zeta}_s+{\eta}_c+{\epsilon}_{isc}.\end{array}& \end{array} $$

Equation 2 largely resembles Equation 1 except in some small respects. First, since each statement i is located within a congressional session which spans 2 years, our year variable is now defined as the midpoint of each congressional session.Footnote ¹⁰ Converting congressional meetings to years facilitates comparisons with our previous analyses. As before, we center year around 1860. Second, grammatical subject mentions i are nested within individual speakers s and, to the extent that Congressmen may be influenced by how their counterparts speak, may also be nested within sessions of Congress c. Because of this cross-nested data structure, we estimate a hierarchical linear probability model with crossed speaker ( $ {\zeta}_s $ ) and Congress ( $ {\eta}_c $ ) random effects.

To draw valid inferences from our speech corpus, we must address another issue with our data: the entry and exit of individual speakers from our dataset. This issue arises from the high level of turnover in Congress during this period (Kernell Reference Kernell1977; Swain et al. Reference Swain, Borrelli, Reed and Evans2000). One may worry that Congressmembers elected prior to the Civil War are not comparable to the more professionalized, careerist Congressmembers in office during the late nineteenth century. In order to pool mentions from individuals who appear in only one Congress with those from individuals who contribute mentions in multiple Congresses, we must account for differences that could influence their usage of the singular.

We focus on variables related to a Congressmember’s life experiences and the social milieu of his birthplace. We code the same set of covariates from the newspaper analysis at the county-decade level; these variables reflect birthplace and birth decade characteristics, not contemporaneous characteristics at the time a statement is observed. In addition, we account for biographical experiences that could influence imagined sovereignty by coding the Congressmember’s birth year and whether he was college educated or served in the military. Details appear in the dataverse appendix.

Figure 6 displays results for the relevant coefficients from Equation 2 and their linear combinations. We note that, in contrast to our newspaper corpus, our congressional speech corpus begins in year 1851 and include data at only five time points (1852, 1854, 1856, 1858, and 1860) before the Civil War. Thus our estimates of prewar quantities, as well as prewar versus postwar differences, are admittedly noisy. Nonetheless, we find a discontinuous jump in singular usage of almost 24% among Northern Whigs/Republicans, and a small and non-significant increase for Northern Democrats. Substantively, these estimates imply that whereas roughly one in four statements made by Whigs/Republicans in the prewar Congresses used the grammatical singular, this proportion rises to one in two in the first Congress to meet after the outbreak of war.

FIGURE 6.

Singular Usage among Northern Congressmen, by Party

Note: This figure displays coefficients, linear combinations, and 95% confidence intervals from Equation 2, centering Y at the year 1860 and setting $ P=1 $ if $ Y>0 $ . Full results are available in Appendix Table D5.

To address the potential criticism that these results are an artifact of our noisy prewar estimates, we conduct an additional analysis of partisan differences in singular usage, but restrict our attention to Congresses after 1860 where we have much more data (see Table S1 in the Supplementary Material). We also include Congress fixed effects and compare Northern Republicans versus Democrats within a particular Congress. Results indicate a consistent pro-Republican gap of 6% (se = 3%, p = 0.05), which is roughly the difference we would expect to observe in two congressional speeches made 5 years apart.

Taken together, although our congressional speech data are somewhat noisy, our analyses point to a consistent partisan gap in the adoption of singular usage following the outbreak of war. These partisan patterns are consistent with historians’ arguments that Republicans seized upon the Civil War as an opportunity to advance their own vision of a singular national sovereignty. More broadly, these results are also consistent with our expectations under the valence hypothesis concerning the role of warfare in galvanizing ideational entrepreneurs into action to advance their preferred framing of the conflict.

To complete the empirical picture, we return to the newspaper data and examine whether the North’s accelerated shift toward singular usage was more concentrated in areas of Republican influence. Given the role of the Civil War as a watershed moment in the entrepreneurial narrative, we define areas of Republican influence based on voting results in the presidential election of 1864.

We test whether we observe a more rapid adoption of the grammatical singular in counties voting for Abraham Lincoln as opposed to his Democratic challenger George McClellan. To do so, we restrict our attention to newspapers headquartered in the North, and modify Equation 1 by replacing the Northern newspaper dummy variable (N) with an indicator ( $ LINCOLN $ ) for whether Lincoln won the county in which the newspaper is headquartered.

Figure 7 summarizes the results for the coefficients and the linear combinations of interest. Our results corroborate the findings from the congressional partisanship analysis. Specifically, consistent with the conjecture that ideational entrepreneurs should drive ideational change under the valence hypothesis, we find that the postwar slope shift was twice as large among counties that voted for Lincoln than for counties that voted for McClellan. Read in combination with our congressional speech analysis, these results suggest that the war further strengthened ideational commitments to the vision of a national sovereignty in the places that were most subject to the Republicans’ ideological influence.

FIGURE 7.

Singular Usage in Northern Newspapers, by 1864 Election Results

Note: This figure displays coefficients, linear combinations, and 95% confidence intervals from modified Equation 1, where the Northern newspaper dummy variable (N) is replaced with an indicator ( $ LINCOLN $ ) for whether Lincoln won the county in which the newspaper is headquartered. Y is centered at the year 1860 and $ P=1 $ if $ Y>0 $ . Full results are available in Appendix Table D6.