Abstract
We provide a nuanced quantitative description of the historic evolution of lynching frames in two major White newspapers. We analyze all articles mentioning lynching in the New York Times and Atlanta Constitution from 1880 to 1950. We use keyword-assisted topic modeling, an automated document clustering method, to analyze the frames these papers used when discussing lynching. We document the dominance of a Rough Justice frame, which framed lynching as a legitimate community response to crime through the 1880s and early 1890s and its slow decline thereafter. Due Process and Civilization frames, which framed lynching as illegal or uncivilized respectively, began to rise in the 1890s. Finally, a Civil Rights frame, which gave more voice to Black organizations (e.g., the NAACP), began to rise in the 1920s in the New York Times. Results suggest that lynching frames were more varied and changed more gradually than typically assumed in the lynching and framing literature.
Keywords
Introduction
How does the news media represent racial violence? The news media does not neutrally represent facts as they occur in the world, but instead frames events and issues for their audience. Media frames rarely construct issues out of whole cloth, but instead highlight some aspects of issues over others (Chong and Druckman 2007), organizing complex issues into “interpretive packages” (Gamson and Modigliani 1989:2). Framing is typically a contested process, with multiple coexisting frames, and multiple political actors struggling to advance their preferred frames (Esacove 2004; Snow, Vliegenthart, and Corrigall-Brown 2007). Here we ask descriptive questions about how the ways newspapers represented lynching in the United States evolved over time. These questions tap into theoretical and historical debates about whether media frames change only slowly, or rapidly in a “punctuated equilibria” (Baumgartner, Boydstun, and De Boef 2008; Walgrave and Vliegenthart 2010).
The “Lynching Era” in the United States is typically periodized from the 1880s until the 1930s or 1940s. Conservative estimates suggest that over 4,467 people were murdered by lynch mobs from 1883 to 1941 in the United States. Although lynching occurred nationally, and claimed victims from all racial groups, the vast majority of the victims of this period were Black men and boys murdered by White mobs (Bailey and Tolnay 2015; Beck 2015; Seguin and Rigby 2019:2). Accordingly, most lynchings of this era occurred during the political project to reestablish White political supremacy in the former slave states following the end of Reconstruction (Du Bois 1998; Foner 2002).
We focus not on how newspapers reported these specific acts of violence, but on how the newspapers framed the phenomenon they called “lynching.” Lynching was widely reported and debated in the nation’s newspapers and was the site of a long-running framing contest over the meaning of mob-murder in the United States. Recognizing the power of the press to define, in part, how this wave of mob murder would be understood, celebrated, or condemned by political elites and northern Whites, historians and sociologists have increasingly studied how White newspapers framed lynching for their audiences (e.g., Jean 2005; Seguin and Nardin 2022; Waldrep 2002; Wasserman 2006; Wood 2009). Much work in this vein focuses on a framing contest among White elites and publics over framing lynching as a legitimate form of community-sanctioned “rough justice” or as an embarrassing perversion of the formal justice system (Pfeifer 2004).
Although many scholars and historians argue that a Rough Justice frame was dominant at least through the 1880s in the White newspapers, and was eventually supplanted, there is no agreement on the kind of language that ultimately replaced it and when. Much of what we know about lynching frames comes from studies that are focused on specific actors such as the NAACP (e.g., Bernstein 2005) or Ida B. Wells (e.g., Silkey 2015), and thus their scope is limited by when these actors were active and the frames they employed. Other studies focus on the careers of single lynching frames, such as “civilization” and are hence looking only for specific frames (Bederman 1995; Seguin and Nardin 2022). A few studies follow qualitative changes in the newspapers over time (e.g., Berg 2011; Pfeifer 2011; Waldrep 2002; Wasserman 2006), and we draw heavily from these studies in our analysis. Thus, although scholars tend to agree that there was an important shift in newspaper framing of lynching over time away from a Rough Justice framing, there is no consensus on exactly when this shift occurred, and on the careers of the specific frames that replaced the Rough Justice frame.
Our goals and results are descriptive, rather than causal (Besbris and Khan 2017; Conwell and Loughran 2023). Our goal for this paper is to compare and describe the dynamics of lynching frames in a leading White northern and southern newspaper from 1880 to 1950: the Atlanta Constitution and the New York Times. We gather and digitize all articles mentioning “lynching” or “lynched” in the New York Times and Atlanta Constitution from 1880 to 1950. We use keyword-assisted topic modeling (Eshima, Imai, and Sasaki 2024), an automated document clustering method, to analyze the frames that newspapers used in stories in newspapers. Keyword topic modeling allows us to balance the typically inductive and exploratory nature of topic modeling (Nelson 2017), with the deductive seeding of topics using language drawn from historical work on lynching.
Drawing on the historical lynching literature we seeded four keyword topics, representing a Rough Justice frame which framed the victims of lynching as criminals, a Due Process frame which framed lynching as a violation of due process rights, a Civilization frame which framed lynching as “barbaric” or “uncivilized,” and a Civil Rights frame which gives some voice to Black anti-lynching organizations and anti-lynching debates in Congress. We document the dominance of the Rough Justice frame through the 1880s and early 1890s, which was first supplanted by Due Process and Civilization frames. We later document the rise of a Civil Rights frame within the New York Times. We find that multiple competing frames were present throughout the period, and that frames changed slowly, rather than in a rapid response to specific events as argued by much of the theoretical and historical literature (e.g., Baumgartner et al. 2008; Seguin and Nardin 2022; Walgrave and Vliegenthart 2010).
Lynching Frames and Chronology
The typical academic definition of lynching is: “(1) an extrajudicial killing, (2) motivated or justified by reference to “justice” or tradition, and (3) committed by three or more people” (Seguin and Rigby 2019:2; Tolnay and Beck 1995). Unlike other forms of collective violence, such as violent riots, lynching is not formally organized and is individual in its “liability,” meaning that specific individuals are targeted by mobs for specific alleged offenses (Senechal de la Roche 1996). Anti-lynching groups of the period, particularly those that gathered and recorded data on lynching victims, such as the NAACP and the Tuskegee Institute, vigorously debated details of the definition of lynching, such as how large the mob had to be or whether someone who died before a lynch mob could murder them could be lynched, but largely followed (indeed inspired) the academic definition in general outline (Waldrep 2002:127–50).
By this definition of lynching, thousands of victims were murdered by lynch mobs during this period. Despite this, most people in the United States did not witness lynching directly, but were exposed to lynching primarily through newspaper reporting (Wood 2009). Newspapers were the dominant media technology in the late nineteenth and first half of the twentieth centuries. Between 1892 and 1916 the invention of the web rotary press, spread of the telegraph, and the Post Office beginning its Rural Free Delivery program contributed to a nearly 10 fold increase in newspaper circulation (Perlman and Schuster 2016:776). Newspapers did not have a complete monopoly on the communication of lynching, which was also spread through rumors, lynching postcards (Allen et al. 1999), church pulpits (Downey 2001; Williams 2001), and early cinema (Wood 2009:246–47) but discourse from these other sources also found their way into newspaper content.
Newspapers explicitly debated the nature of lynching in op-eds, but also implicitly in reporting of specific lynchings through how they framed protagonists such as the victim or mob. Frames work by emphasizing some aspects of phenomena over others (Chong and Druckman 2007), and lynching frames worked through emphasizing either the supposed crimes of the lynching victim, the crimes of the lynch mob, the legal implications of “mob law,” or comparisons to the rest of the “civilized world.”
Here we study the dynamics of lynching frames in two newspapers, the New York Times and the Atlanta Constitution from 1880 to 1950. These newspapers were both regionally and nationally prominent, although they cannot be said to be representative. The Times, unlike many other leading northern papers, had shed its Republican loyalties by 1884 when it endorsed Grover Cleveland for president, and under the editorship of Adolph Ochs became a leading voice for the ascendant objectivity norm beginning in the 1890s (Porwancher 2011:188; Schudson 1978). Ochs, who was the Times editor throughout most of this period (1896–1935), was southern by birth and had previously edited both the Chattanooga Dispatch and Chattanooga Times (Porwancher 2011:188). Thus both Ochs’ lack of ties to Republican party politics, whose leadership was at times vociferously anti-lynching (Seguin and Nardin 2022), and his southern origins distinguished the New York Times from other major White northern papers. The Constitution, in turn, may have been relatively less supportive of lynching than typical White southern newspapers owing to its “New South” ideology. The Constitution’s editor throughout the 1880s was Henry Grady, who was dubbed the “New South spokesman” (Gaston 2011:37). The New South ideology, including Grady’s, was fully steeped in White supremacy, but embraced more legalistic methods of racial domination than lynching, such as Jim Crow segregation (Gaston 2011:148–49). Nevertheless, many examples exist of the Constitution’s support of lynching, perhaps most well-known was their reporting during the lead-up to the lynching of Sam Hose in 1899, which Ida B. Wells characterized as: “never a word for law and order, but daily encouragement for burning” (Squires 2015:142).
The historical literature delineates four major lynching frames which we describe in detail below: Rough Justice, Due Process, Civilization, and Civil Rights. We describe the rise and fall of these four frames over this period and compare our descriptive portrait of these dynamics to the claims in the historical literature. Specifically, we look at whether dramatic frame changes follow specific historical events and turning points highlighted in the historical literature. Generally we assess whether frames change rapidly as in the “punctuated equilibria” model (Baumgartner et al. 2008; Walgrave and Vliegenthart 2010), or whether they change more gradually over the course of the period.
Rough Justice
Historians of lynching generally agree that during the 1870s and 1880s lynching was overwhelmingly framed in the national news media as a form of “rough justice” and popular sovereignty against a slow and lenient legal system. This Rough Justice frame was appropriated from justifications of “Wild-West” frontier lynchings from before the Civil War (see especially: Berg 2011; Pfeifer 2011; Waldrep 2002). These frontier Rough Justice themes were coupled with a “folk pornography” theme in the White South in which Black men were constructed as threats to White women (Hall 1993). The Rough Justice frame drew on ideas of popular sovereignty, in which lynching could be justified if enough of the “best citizens” supported it, and White newspaper accounts often stressed the supposed unanimity, rationality, and social status of mob members in lynching articles (Waldrep 2002:93–97). During the 1870s and 1880s, according to the newspapers, “quiet” or “determined” mobs of “citizens” lynched “fiends” or “brutal negroes.”
The Rough Justice frame was dominant until at least the early 1890s. Many argue that this frame began to dramatically decline in the early 1890s. Either beginning with the lynching of Italian nationals in 1891 (Seguin and Nardin 2022) or as a result of Ida B. Wells’ British tours of 1893 and especially 1894 (Bederman 1995; Silkey 2015; Waldrep 2002). Many NAACP-centered analyses argue, however, that the lynching of Jesse Washington in 1916, and the publicity resulting from the NAACP investigation, marked the beginning of the end of the Rough Justice frame (e.g., Bernstein 2005; Ming-Francis 2014). Wood, for instance, argues that the lynching of Washington represented: “A defining moment in the history of lynching, an instance when the spectacle of lynching began to sow the seeds of its own collapse” (Wood 2009:179). Thus, even as studies suggest very specific dates, it is unclear from the exisiting literature when the Rough Justice frame began its decline.
Part of the question of dating the decline of the Rough Justice frame has to do with debates over which frames replaced it, and when. Scholars argue that alternative lynching frames emphasizing Due Process, Civilization, or Civil Rights later replaced the Rough Justice frame. We discuss each frame in turn.
Due Process
The Due Process frame was employed as a critique of the Rough Justice frame. This frame juxtaposed the importance of established legal systems of justice with the “anarchy” of “mob rule” or “mob law.” This frame attacked the idea in the Rough Justice frame that lynch mobs were sovereign and represented the will of the community. Although this frame often treated the lynching victim as presumptively guilty of some offense, in contrast to the Rough Justice frame, it emphasized the criminality of mobs in usurping the legitimate formal justice system. Newspaper articles in the Due Process frame also emphasized the crimes of lynch mobs rather than the supposed crimes of the lynching victim (Waldrep 2002:114–15). Waldrep (2002) argues that the Due Process frame slowly began to take hold in the New York Times in the 1880s, as the paper “began to argue editorially that even guilty persons had inalienable rights to due process” (p. 114), such that by the 1890s the Times “very rarely found community support sufficient to make a lynching explicable” (p. 116).
Pfeifer argues that White southern papers rarely employed a Due Process frame before the twentieth century, because of “a white consensus that favored harsh punishment rooted in white supremacist, communal prerogatives” (Pfeifer 2004:116). Beginning in the twentieth century, however as the South increasingly modernized, and White elites sought to replace lynching with legal social controls such as Jim Crow segregation and racialized capital punishment as “legal lynchings” (Hagen, Makovi, and Bearman 2013; Pfeifer 2004:139–47; Vandiver 2005) it is likely that the Atlanta Constitution would increasingly adopt such frames. Thus, we might expect the Due Process frame to rise more rapidly in the New York Times than in the Atlanta Constitution, although as an early and vociferous proponent of the “New South,” the Constitution might have adopted the frame in the early twentieth century or perhaps even earlier (Davis 1987).
Civilization
Others argue that the Rough Justice frame was replaced with a frame of lynching as “uncivilized”; a practice which shamed the United States before the “Civilized World.” The Civilization frame attacked the Rough Justice frame’s construction of lynch mobs as rational, high status, or otherwise sober citizens. Drawing on colonialist constructions of non-Europeans as “savage” or “primitive” the Civilization frame depicted lynching as a barbaric and atavistic practice that bespoke of a limited level of civilization among the inhabitants in places where lynchings occurred. This frame emphasized international comparisons, often noting that lynching was not found in other developed countries in the “civilized world.”
Many scholars date the rise of the Civilization frame to Ida B Wells’ British tours of 1893 and 1894 (Bederman 1995; Silkey 2015). Wells had toured Great Britain hoping to raise international awareness and condemnation of the lynching of Black Americans. Wells succeeded in mobilizing allies from old abolitionist networks, such as William Lloyd Garrison Jr, who in turn pressured American politicians (Silkey 2015). Seguin and Nardin (2022) date the rise of the Civilization frame slightly earlier as beginning in 1891, following international fallout from the lynching of Eleven Italians in New Orleans, but stress the role of Wells’ British tours in turning British attention from the lynching of Italians to the lynching of Black Americans.
Although these studies identify the rise of the Civilization frame, and mostly agree on dating its origins to the early 1890s, most studies are qualitative and make no claims to the numerical composition of the frame over time. Seguin and Nardin (2022) quantitatively track the rise of the Civilization frame through keyword searches but do not analyze the dynamics of other frames, leaving it unclear how the Civilization frame fared against other frames.
Civil Rights
Early in the Lynching Era, activists such as Ida B Wells, argued that lynching, far from being a natural response to Black criminality, was an expression of White racism. Wells and others cast doubt on the guilt of lynching victims when accused with crimes such as the rape of a White woman (Silkey 2015), and put into question the racist motives of the mob (Waldrep 2002:103–26). Although this frame is dominant in current scholarly understandings of lynching, it’s unclear how much purchase it gained during the Lynching Era itself.
To the extent that a frame emphasizing the racist motives of White lynch mobs would enter White newspapers, it is likely it would have originated with Black activists, beginning with Ida B. Wells. However, Waldrep argues that Wells’ arguments about the racist origins of lynching were aimed more at Black audiences (Waldrep 2002:122–23), and others have found Wells to have a greater impact on Whites when employing a Civilization frame (Bederman 1995; Seguin and Nardin 2022; Silkey 2015). Wood argues that Black activists, most prominently the nascent NAACP, increasingly emphasized a Civilization frame after World War I, rather than explaining lynching as emerging from “deeply rooted racist structures and institutions” as Ida B. Wells had argued in the 1890s (Wood 2009:204). Rather, in a push to appeal to White liberals and moderates, Black activists made “surprisingly little mention of race or racial prejudice” (Wood 2009:205). If Black activists were making little mention of racial prejudice, then it appears unlikely that White newspapers would be. Wasserman argues that beginning around 1919 and into the 1930s and 1940s as Black political power in the North increased following the Great Migration, the major political parties began to sponsor anti-lynching bills and take other anti-lynching stances. In response, White newspapers increasingly framed lynching as factional political struggles, rather than as an expression of racism (Wasserman 2006:183). Thus, it remains unclear if, or how, a Civil Rights frame was used by the White newspapers.
Other Frames
Outside of these core frames, we have little sense of how else lynching was reported in the national newspapers. Although the federal political process failed to deal with lynching throughout this period, it was among the most central political topics of its day and touched most aspects of domestic and international politics in some way. The lynching of Eleven Italians in New Orleans in 1891, for instance, was the most widely reported news story of that year (Dell et al. 2023:9). Given lynching’s prominence as a topic in the news, it is unlikely that a few substantive themes exhaust the meaningful ways it was reported and discussed, making a partially inductive approach to lynching discourse appropriate.
Data and Method
Our data and methodology allow us to track the prevalence of multiple frames across lynching articles in two key newspapers from 1880 to 1950. We first collected all articles mentioning either “lynching” or “lynched” in The New York Times from 1880 to 1950 (n = 8,496 articles) and The Atlanta Journal-Constitution from 1882 to 1945 (n = 10,773 articles). In analyzing articles which mention these keywords we are tracking change in the meaning of the term “lynching,” rather than coverage of lynchings as such. In many cases, these articles do not mention any specific events, but critique lynching as a practice.
In using these keywords, we also miss some references to lynching which do not use those terms (Weaver 2019). For instance, searching the New York Times from 1880 to 1950 for the term “mob violence” yields 1,312 articles. 335 of these articles also contain the terms “lynching” or “lynched” and so are already included in our data, leaving 977 articles containing “mob violence” but not our lynching keywords. Some of these articles discuss lynching under the term “mob violence” and so we necessarily miss some of the lynching discourse, but many discuss riots or pogroms, which scholars and contemporaneous observers understood as distinct from lynchings (Senechal de la Roche 1996) and hence would be “false positives.” Since the term “lynching,” however, was in widespread usage in the United States since 1835 (Waldrep 2002:27), and was used both by opponents and apologists of lynching, what we are capturing, is most of the public debate on the meaning of the wave of mob murder in the United States known as “lynching.”
We cluster our articles using keyword topic modeling (Eshima et al. 2024). Topic modeling is an unsupervised approach to machine classify documents into distinct topics, defined as clusters of word usage (e.g., DiMaggio, Nag, and Blei 2013). Generally, unsupervised text classification is suitable for an inductive approach, when little is known beforehand of the relevant topics (Nelson 2017). A fully inductive approach to lynching discourse is not appropriate, however, as scholars have accumulated historical knowledge to draw from. Keyword topic modeling allows us to incorporate deductive elements, by “seeding” some of the topics beforehand based on keywords drawn from the lynching literature, while continuing to draw on the inductive capabilities of the method (Eshima et al. 2024). In effect, keyword topic modeling allows us to take an exploratory approach, while still benefiting from the accumulated knowledge of prior research.
Optical Character Recognition and Post-Processing
We used the ABBYY FineReader optical character recognition (OCR) software to digitize our PDF image files of both the New York Times and the Atlanta Constitution articles. We trained ABBYY using within-sample training data. We then post-processed the OCRed articles to fix remaining patterned errors. For instance, many “e” characters were misrecognized as “c,” “n” as “h,” and vice versa. To address this issue, we first word-tokenized all the articles and ran a spell-checker on them (Norvig 2009), to identify the fifteen most common OCR error patterns. We then flagged any words that failed the spell-check, and tested letter replacement patterns to find a substitution which would yield a word in the dictionary. The replacement pattern that corrects the misspelled token and has the highest probability score in the spellchecker dictionary was then selected to replace the mispelled word (if none of the candidate substitutions was in the dictionary, we made no changes to the original word). For example, given the word “armcry,” the script attempts to replace “c” with “o” or “e,” checks both potential words (“amery” and “armory”) against the spellchecker, and finally corrects the word to “armory.” In total, we corrected 26,661 unique misspelled words which appeared 190,589 times in our corpus. The remaining “misspelled” words overwhelmingly proper nouns or archaic terms.
We had very poor OCR performance on a small percentage of articles owing to poor-quality scans of the original PDF image. To identify and remove these articles, we calculated the “non-word rate” (Dell et al. 2023:7) for each article and dropped the documents in which over 20 percent of the words were not in the dictionary. This step removed 668 documents: 3.4 percent of our sample.
Topic Modeling and Keyword Assisted Topic Models
Topic models are a class of clustering algorithms within the natural language processing toolbox that allow the soft assignment of documents into multiple topics. Soft clustering means that each document can be assigned to multiple topics with proportional membership scores to all the topics (Grimmer, Roberts, and Stewart 2022). Given the number of topics, the algorithm groups similar documents based on the most frequent tokens and their co-occurrence.
Unlike supervised methods that require labeled data, topic models allow researchers to initially explore and categorize the content of documents without imposing the researcher’s theoretical commitments (Nelson 2017). Such a fully inductive approach is appropriate when expectations are unclear but fails to take advantage of the researcher’s knowledge of the case. Keyword Assisted Topic Modeling (Key ATM) allows the researcher to benefit from both deductive and inductive approaches by “seeding” some of the topics with a set of researcher-provided keywords prior to modeling to guide the algorithm toward some topics of interest, while allowing other topics to emerge fully inductively. Seeding some of the topics with keywords prior to fitting the model improves the performance of topic models and makes them less sensitive to the number of topics (Eshima et al. 2024).
We compiled keywords for seeding four topics: Rough Justice, Due Process, Civilization, and Civil Rights. We seeded the Rough Justice topic with the words: murderer, fiend, brute, slayer, criminal, scoundrel, ruffian, desperado, brutal_negro, culprit, firebug, thief, bad_negro, villain, outrage, outraged, ravished, ravisher, posse; the Due Process topic with: innocent, innocence, alleged, anarchy, anarchic, mob_rule, lawless, lawlessness; the Civilization topic with: civilization, civilized, uncivilized, savage, savagery, barbaric, barbarism, chivalry; and the Civil Rights topic with: bias, segregation, discrimination, race_prejudice, race_feeling, racial.
We then specified an additional 16 inductive topics, for a total of 20 topics overall. We use the KeyATM library in the R programming language to train the topic models (Eshima et al. 2024). After cleaning the digitized documents and correcting the misspelled words, we tokenized the documents and removed stop words, numbers, and punctuation. We did not stem or lemmatize, because in our corpus this would make the language less accurate. For instance, stemming the words “brute,” “brutal,” and “brutality” to “brut” would conflate “brute” which almost always was a way of labeling a Black lynching victim as a depraved, animalistic, criminal with “brutality” which often referred to the brutality of mobs.
Keyword topic modeling requires the analyst to choose the number of topics before the model is run. We experimented with multiple different numbers of topics, and we ultimately chose a 20-topic model (16 unnamed topics and four keyword topics) because we found the unnamed topics of this model most interpretable based on our readings of the topic’s most representative words and the articles that were most representative of each topic. Since our keyword topics were most central to our results, to test whether our keyword topics were robust to different numbers of non-keyword topics, we compared the correlations in the prevalence of keyword topics across articles for our 20-topic model with 10- and 15-topic models. For the Rough Justice topic, correlations were .976 and .979 with the 10- and 15-topic models, respectively, suggesting the Rough Justice topic dynamics were nearly identical across models. For the Due Process model, the correlations were .813 and .422, suggesting modest consistency across models. For the Civilization topic correlations were .142 and .853 suggesting that the Civilization topic was weakly consistent with the 10-topic model but highly consistent with the 15-topic model. The Civil Rights topics were correlated at .631 and .646, suggesting moderate to high consistency across models. Overall, our various models produced qualitatively similar trends over time for the keyword topics.
Results
Here, we focus on our keyword-seeded topics: Rough Justice, Due Process, Civilization, and Civil Rights, and discuss our other more inductive topics in the appendix. Figure 1 shows the topics which were most prevalent in our corpus of lynching articles. This figure shows the dominance of the Rough Justice frame, which was much more common than any of the other topics. The Civilization and Due Process topics are the third and fourth most prevalent respectively. The Civil Rights topic was the least prevalent of our keyword topics, being the ninth most prevalent overall. We discuss the unnamed topics in the appendix.
Most prevalent topics in lynching articles.
Figure 2 shows the evolution of the topics over time. The Rough Justice topic began to decline around 1890, although it remained the most prevalent topic until at least the 1930s. The Due Process and Civilization frames began to rise around 1890, roughly consistent with previous literature (Bederman 1995; Seguin and Nardin 2022; Waldrep 2002). The Civil Rights frame ultimately came to dominate in the New York Times, but does begin to rise until the mid-1910s, consistent with accounts emphasizing the role of the NAACP which was founded in 1908, but whose anti-lynching campaign was argued to have achieved the most success beginning with its response to the lynching of Jesse Washington in 1916 (Bernstein 2005; Wood 2009), and with the introduction of the Dyer anti-lynching bill in 1918 (Ming-Francis 2014).
Keyword topic prevalence over time in lynching articles.
The most representative words for each of our four named topics, which give some sense of the type of language used in articles that are most representative of those topics, appears in Table 1. We discuss these more in the description of the most representative words for our 16 additional unnamed topics, which appear in Table A1 in the appendix. Although the most representative words can at times identify the substantive meaning of topics, in many cases human interpretation is limited and subject to confirmation bias (Chang et al. 2009). We deepen our understanding of these topics through targeted reading of the documents most representative of particular topics (Nelson 2017). We present, in Tables 2–5, the headlines of the articles most representative of each of our keyword topics.
Most Representative Words by Topic.
Headlines of Articles Most Representative of the Rough Justice Topic.
Headlines of Articles Most Representative of the Due Process Topic.
Headlines of Articles Most Representative of the Civilization Topic.
Headlines of Articles Most Representative of the Civil Rights Topic.
Rough Justice
The most representative words of our Rough Justice topic (Table 1) often evoke reporting on specific lynching events, as opposed to more general discussion of lynching, such as the victim’s race (negro), as well as the process through which the mob captured and lynched the victim (jail, mob, night, sheriff, county, shot). Reading the most representative articles on this topic shows that these were not neutral descriptions of lynching but were rather laden with assumptions that the lynching was a natural response to the supposed crime (often an assault of a White woman or girl) of the lynching victim. For instance, a headline from the New York Times in 1910 reads: TWO NEGROES LYNCHED. _______________ Florida Mob Puts Them to Death for Killing a White Girl.
Rough Justice is the most represented of all topics in our analysis for both newspapers, being over twice as common as any of the other topics (see Figure 1). Given that the Rough Justice frame is emphasized in most studies of lynching’s representation in the newspapers (e.g., Jean 2005; Waldrep 2002), and in contemporaneous activist accounts (Wells 1895), its dominance is not surprising. It is surprising, however, that the Rough Justice frame was dominant for as long as it was. The Rough Justice topic, for instance, is the most prevalent of the named topics in both the Times and the Constitution throughout most of the 1920s and roughly the first half of the 1930s (see Figure 2). For instance, a 1933 article in the Times is titled: LOUISIANA NEGRO LYNCHED. _______________ Youth Said to Have Confessed to Killing Girl is Taken From Jail.
The article centers the lynching victim’s supposed crime and reports the authorities’ claims that the victim had confessed to killing a 15-year-old girl. Despite the Rough Justice frame’s continued presence, however, it should be noted that the combined prevalence of the Due Process and Civilization frames challenged the Rough Justice frame’s dominance by the 1910s in the New York Times and by the 1920s in the Atlanta Constitution.
Due Process
The most representative words of the Due Process topic, such as law, courts, crime, states, or justice (see Table 1) evoke the law and its enforcement, which underscores the frame’s focus on the formal legal rights of lynching victims. Although this topic was somewhat more common in the New York Times than in the Atlanta Constitution (Figure 1), many of the most representative articles from this topic come from the Constitution. Fitting with the Constitution’s “New South” focus on development and law and order, the most representative article (see Table 3), for instance, argued that lynching would cease when the law was enforced as rigorously as it is in England, but until then lynching “will be a standing reproach to the people and their machinery of justice.”
The Due Process topic begins to rise in both papers in the early 1890s, closely tracking the Civilization topic (Figure 2). The Due Process and Civilization topics are also correlated across articles at .392, among the higher correlations among our topics, suggesting that these frames are not entirely separable, or that articles using one frame tend to also employ the other frame. A reading of the top articles from both the Due Process and Civilization topics suggests that shared themes of lynching being “shameful” or “disgraceful” for the communities in which it occurred, and that the “lynching spirit” was atavistic. A 1915 New York Times article on the lynching of Leo Frank (see Dinnerstein 2008), for example, in line with the Due Process frame expressed hope that “the authorities will speedily punish” the lynch mob, but also in line with the Civilization frame, argued that Frank’s lynching was “accomplished in a spirit of savage and remorseless cruelty unworthy of our age and time.” Thus, at least some articles linked the Due Process and Civilization frames to the idea that lynch mobs’ uncivilized or “savage” actions should be addressed through enforcement of the law.
Civilization
The most representative words for the Civilization topic (see Table 1), such as great, people, or country gave us little sense of the meaning of the topic. The most representative news articles, however, were more informative. In Table 4, we see the headlines of articles for the Civilization topic, a reading of these articles finds that many are international comparisons where the author finds the “level of civilization” in the United States to be lacking owing to violence such as lynching. An 1899 article in the Atlanta Constitution, entitled “THE VALUE OF THE PHILIPPINES” argues, for example, that “among the 300,000,000 inhabitants of the Indian empire the reign of peace is unbroken” and “outrages such as are practiced by American lynching parties” are unknown. A 1916 article in the Atlanta Constitution, entitled “IN PROPER EDUCATION OF YOUNG LIES CURE FOR LYNCHING” argues that: One of the most serious problems confronting society today, especially in this section of the country, is how to control that element [lynch mobs] who ruled by impulse and passion have committed crimes that the civilized world has shuddered at.
Thus the most representative articles of the Civilization topic display the features emphasized in previous research: that this frame portrayed lynching as shameful before the eyes of the world, particularly when juxtaposed with American imperial ambitions (Bederman 1995; Seguin and Nardin 2022).
Civil Rights
Our Civil Rights topic, although seeded mostly with words relating to interpersonal racism, such as discrimination, bias, or race prejudice, is described by its most representative words (Table 1) such as Party, President, Republican, National, or Democratic, as relating to party politics. The articles which are most representative of this topic confirm that it relates to party politics, but also show that this topic is dominated by reporting on the work of Black organizations, such as the NAACP (Zangrando 1980), Anti-Lynching Congress, or National Negro Congress (Gellman 2012), in pushing for legislation or statements against lynching from political elites. In this sense, the topic modeling supports the idea that Black organizations, particularly the NAACP, increasingly gained an audience with presidents and other power holders, which they used to denounce lynching and push for anti-lynching legislation (Ming-Francis 2014). In an article in the New York Times in July of 1944, for instance, titled: PLATFORM IS CONDEMNED _______________ Democrats Bowed to South’s Bigots Negroes’ Spokesman Says
Walter White, then secretary of the NAACP, was quoted sharply criticizing the Democratic party’s platform as bigoted and without principle.
The Civil Rights frame began to rise in prevalence in the early 1920s (Wasserman 2006:238), perhaps marked by the NAACP support of the 1922 Dyer Anti-lynching bill, which was passed in the U.S. House, only to be filibustered in the Senate (Walter 1934:442). Given White Southern hostility to Black civil rights, it is perhaps unsurprising that this topic is much less common in the Atlanta Constitution than in the New York Times. It bears noting that the Constitution did employ Civilization and Due-Process anti-lynching frames, suggesting that the focus on Black political power, rather than anti-lynching per se, was why the Constitution avoided the frame.
Conclusion
We used automated text analyses to show which frames dominated discussion of lynching in two key newspapers over the period of 1880–1950. We document the dominance of a Rough Justice frame, which portrayed Black lynching victims as depraved criminals who were lynched by determined and democratic White mobs, through the 1880s and early 1890s. This Rough Justice frame was slowly supplanted by Due Process and Civilization frames, beginning in the 1890s. The Due Process frame portrayed lynch mobs as criminals rather than practitioners of “community justice,” while the Civilization frame portrayed lynching as shaming the United States before the “Civilized World.” In the New York Times these frames in turn were later supplanted by a Civil Rights frame, beginning in the 1920s and becoming dominant in the late 1940s, which emphasized political struggles over anti-lynching legislation, but generally fell short of critiquing the racist structures which created and sustained lynching.
Our results find many parallels within the historical literature on lynching discourse in the national newspapers while supporting some specific chronologies and typologies over others. Most generally our results suggest that changes in lynching discourse in White newspapers, although substantial over the seventy years we studied (1880–1950), occurred quite gradually through the replacement of the Rough Justice frame with a heterogenous set of frames that treated lynching more as a social problem than as a form of community justice. Our results contrasted with the historical literature in that many studies characterize the changing of lynching frames as a rapid process, responsive to single events, and characterized by the dynamics of a single critical frame. Seguin and Nardin, for instance, argue that a Civilization frame rapidly gained prominence following the 1891 lynching of Italians in New Orleans. Our results show a small rise in the Civilization frame in 1891, but nothing like the dramatic change shown in Seguin and Nardin (2022:74). Others argue that the lynching of Jesse Washington in 1916 marked a turning point when the Rough Justice frame began to finally decline (Wood 2009:279; see also: Bernstein 2005; Carrigan 2004:185), but our data show a decline in the Rough Justice frame beginning in the late 1880s and early 1890s, and if anything a brief resurgence of the frame after 1915 until the early 1920s (see Figure 2). Future work, both descriptive and causal, on the relationship between lynching and political discourse must take into account the varied, gradual, and uneven dynamics of lynching frames.
Our results contrast also with theories of political and media dynamics that emphasize rapid change or punctuated equilibria in discourse and politics generally (Baumgartner and Jones 2009; Walgrave and Vliegenthart 2010). Media discourse around the death penalty, for instance, rapidly changed in the early 2000s to emphasize the potential innocence of those convicted of capital crimes to the exclusion of other frames (Baumgartner et al. 2008). Such a rapid and encompassing transformation could certainly have happened with lynching frames, but instead, most of our topics show gradual change that takes place over decades, and a continuing coexistence of multiple frames throughout. This is not a mere artifact of our method, as can be seen in some of our unnamed topics discussed in the appendix, which are punctuated and unstable. These topics are driven by external events that are themselves punctuated. Topic 5, for instance, which primarily covered lynching in international contexts, shows dramatic spikes during WWI and WWII (see Figure A1). 1 Thus, the decline in the legitimacy of lynching as reported and understood in White newspapers resembled more a “strong and slow boring of hard boards” (Weber 1946:128), a gradual process in which anti-lynching advocates from activists such as Ida B. Wells or W.E.B. Du Bois and the early NAACP, U.S. presidents such as Harrison, McKinley, Theodore Roosevelt, or Truman, and even White-supremacist advocates of the “New South” and Jim Crow segregation attacked the practice of lynching for their own purposes and with their own framings.
Our results show that by the 1930s and 1940s anti-lynching frames were dominant in newspaper discussions of lynching. However, for pragmatic reasons, anti-lynching frames, even those pushed by Ida B. Wells and the early NAACP, typically fell short of a full critique of racist motivations and structures that enabled lynching (Silkey 2015; Waldrep 2002; Wood 2009). In particular, the civilization frame painted lynch mobs as “crazed” or irrational, in effect painting lynching as a psychological pathology. Such frames are echoed in the discourse around contemporary hate crimes, which are often framed as motivated by mental illness (Waldrep 2002:186–91). Contemporary public figures such as Clarence Thomas, Bill Clinton, or Donald Trump have appropriated the term lynching to draw on public understandings of lynching to paint their political opposition as irrational mobs (Ore 2023; Waldrep 2002:151–53; Wood and Donaldson 2008). Our results suggest that this continuing discursive power of the word “lynching” draws on the unfinished historical framing contest over the meaning lynching, which ultimately left “lynching” and its history misunderstood and continually contested.
Footnotes
Appendix
Author’s Note
Marjan Davoodi is now affiliated with Johns Hopkins School of Medicine, MD, USA.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
