The term “nation” derives from the Latin word nationem for birth, origin, or tribe. The geographical concept of the nation is thus firmly established, and is still evident in the 1840–1914. Ironically, the conceptual stability of nation in public discourse seems more stable than the geopolitical realities of the time. Within the word vector space of TDA, the search term “England,” for example, produced a stable set of names of other nations; “England” shares a semantic space with “Scotland,” “Ireland,” “France,” and “Belgium,” and also with “America.” Perhaps more remarkably, the term “Europe” is also part of that stable presence. “Germany” and “Spain” are mentioned in 1848, as the political revolutions in both countries inject them into public discourse, and ensure they persist after (see Table 2 below).

In the German newspaper corpus (DE-SBB), the conceptual neighbours of “Deutschland” are also its geographical neighbours France, Austria-Hungary, England, Russia, Belgium and Italy, more or less in that order, along with Europe.

The Finnish dataset FI-NLF shows a sense of Nordic insularity, as all the proper names of nation states are from Nordic countries: Finland, Sweden, Norway, Denmark (Suomi, Ruotsi, Norja, Tanska). In Finnish there is a distinction between Suomi (Finland) and suomi (Finnish language), but it is lost here because the word embeddings are not case sensitive. Therefore, many word vectors are also related to the Finnish language (like suomenkieli, opetuskielenä, kieli, suomenkielinen). Historia (history) and Pohjanmaa (Ostrobothnia, a region in Finland) are also included, but references to other European nations or to the concept of Europe are remarkably absent in the Finnish word vector space of Nordic countries.

This semantic stability may be seen as an illustration of the fact that the nation states are defined in relation to each other, as Koselleck observes [Koselleck 1992a, 145–6]. The relative stability in the word vector models stands in sharp contrast to the geopolitical upheavals of this period. The First World War, for instance, is virtually invisible in the long-term representation of the word vector models.

The shifting vocabularies in which the concept of the nation is expressed in public discourse, have changed over time considerably. Although references to the nation appear in all European countries during the nineteenth and early twentieth century, the semantic associations reflect the different contexts in which this concept functions. In TDA, the semantic connection between the term “nation” and “monarchy,” which is initially very strong, disappears after 1884. The nation is consistently associated with “people” over the full period under review, reflecting the broad meaning of the word in which national identity is based on the notion of a people sharing similar language and ethnic origins. Unsurprisingly, nation is also associated with the term “empire” in England, although this connection weakens after 1908 (see Table 5).

A narrower political interpretation of national identity appears only in the 1860s in TDA, when nationalism became associated with conservatism, but also with the competing political affiliations of liberalism, democracy, and Toryism. The growing political awareness and polarization in the 1870s is reflected by the connection between “nationalism” and the term “radicalism” within the same word vector space. That nationalism is also placed in a religious context is suggested by the emergence of religious terms such as “protestantism,” “clericalism,” and “ultramontane” (the latter reflecting the political ambitions of the pope). The role of the state is difficult to assess on the basis of the word vector models because the word “state” possesses several meanings in English. Yet the term “state” seems associated with the more administrative terms “to govern,” “government,” and “administration” from the 1840s on, and the more negative term “disorganisation” from the mid-1850s on, hinting at a tendency perhaps, to describe the state, rather than the nation, in utilitarian terms.

After German unification in 1871, the newspapers reflected the strong cultural origins of the German sense of national identity. The term “nation” is strongly associated with “christianity” (Christenheit) in the 1870s, the period marked by Bismarck’s cultural war (Kulturkampf) against Catholicism. Although the Iron Chancellor faced defeat in the political arena, the emergence of the secular alternative “civilization” (Zivilization) in the 1880s suggests that he won in public discourse. This seems to confirm Koselleck’s conclusion that the German sense of nation is primarily a cultural construct (Kulturbegriff). Nevertheless, the word vector models also show the connection with power. The emergence of geopolitical and military terms such as “sea power,” “great power,” and “world power” (Seemacht, Großmacht, Weltmacht) in the vocabulary around the term “nation” reflects Germany’s global geopolitical ambitions, which were translated into colonial expansion and maritime muscle-flexing at that time. A similar German sense of superiority and the civilizing mission is reflected in the close relation emerging between the term “civilization” (Civilisation) and geopolitical markers such as “sea power” (Seemacht) and “world politics” (Weltpolitik) at the end of the nineteenth century.

The political interpretation of national identity within Germany is expressed by the term “nationalism,” which can be understood as a developing term first associated with foreign influences from the West and East (Bonapartismus, Polenklubs) and the indigenous national spirit (Nationalgeistes) propagated to counter France’s national ambitions. Interestingly, similarly to the UK, the German term “nationalism” becomes contested at the end of the century, as is evinced by its semantic association with pejorative terms with the suffix -ismus such as “clericalism,” “imperialism,” “absolutism,” “radicalism,” “chauvinism,” and “fanaticism” (Klerikalismus, Imperialismus, Absolutismus, Radikalismus, Chauvinismus, Fanatismus). In German discourse, the term “state” (Staat) does not make inroads, as the associated vocabulary suggests competition between governmental intervention and private initiatives, such as “compulsory insurance” (Versicherungszwang), “private firm” (Privatbetrieb), “artisans” (Handwerkerstand), and, at the end of the century, “taxation” (Steuerbezahler, Fiscus). The Hegelian conception of national identity, lastly, is reflected in the frequent use of the term “spirit of the nation” (Nationalgeist) in connection with the German “nation.” In the 1870s, this term is associated with terms such as “idealism,” “spirit,” “patriotism” and “national identity” (Idealismus, Geist, Patriotismus), which may reflect an essentialist, or Hegelian, interpretation of national identity. In the 1880s, terms such as “idealism,” “altruism,” “fatalism,” “naturalism,” and “national character” (Idealismus, Altruismus, Fatalismus, Volkscharacter) dominate, perhaps suggesting a conceptual change towards a more moral and personal interpretation of national identity.

In the German-language newspapers that were published in the United States (DE-CA), the concept of the nation is used in a radically different political context. Across the Atlantic, the nation became immediately connected with “justice,” “politics”, “party,” “government,” and “principles” (Nation, Gerechtigkeit, Politik, Partei, Regierung, Grundsätze), which illustrates how readily German migrants absorbed the constitutional context of their adopted nation. That adaptation may also explain the emergence of the term “republic” (Republik) in the 1850s, and “race” (Rasse) during the last years of the nineteenth century, in the discourse of nation. Similarly, the term “democracy” (Demokratie) became associated with heated party politics in the post-Jackson years of the 1840s, resulting in terms such as “slave owner” and “slavery question” (Sklavenhalter, Sklavenfrage) in the late 1840s, and “Republican” in the 1850s. The occurrence of references to silver and prohibition (Silberfrage, Silberleute, Prohibition and Prohibitionisten) in the vocabulary around democracy during the last decades of the century illustrates that German immigrants absorbed the key concerns of political populism, such as alcohol prohibition and resentment against the silver standard, which divided the rest of the United States [Wells 2015] [Kazin 2007]. For German immigrants in the United States, the term “folk” (Volk) is associated with “germanness” (Deutschthum) from the late 1850s and “fatherland” (Vaterland) from the 1880s. During the 1860s, references to “civil rights” and “citizenship” (Bürgerrecht, Bürgerthum) are associated with people (Volk), and with “workers” (Proletariat) from the 1890s, which may be a reflection of the emerging radicalism in the United States of socialist and anarchist groups, inspired by European movements [Kazin 2012] [Foner 2014].

The case of Finland provides an interesting point of comparison with English and German datasets. Finland was originally a region of the Swedish Kingdom, annexed by the Russian Empire during the Napoleonic Wars. The Grand Duchy of Finland (1809–1917) was an autonomous part of Russia and can be considered as a predecessor of the Republic of Finland, founded in 1917. The fact that Finland is a bilingual country had consequences for Finnish nationalism. In the beginning of the nineteenth century, Finnish intellectuals and civil servants understood only Swedish, whereas the common people tended to use Finnish in their communication. This led to a situation where the early promoters of Finnish nationalism typically published in Swedish, and could not understand Finnish. However, towards the end of the nineteenth century, the Finnish language slowly replaced Swedish as the main language of the press, and during the peak of Finnish nationalism (Fennomania) , many Swedish-speaking families changed their first language to Finnish.[8]

Academic scholarship has emphasized the role of the German example in the early development of Finnish nationalism. In the beginning of the nineteenth century, folk (folk in Swedish or kansa in Finnish) was a term used in the idealized context of national Romanticism. Johann Gottfried von Herder and German Romanticism were important models for early Finnish nationalists like Adolf Ivar Arwidsson (1791–1858), active in the 1810s and 1820s. From the beginning, the press was the most important forum for Arwidsson’s nationalist activity. He founded the radical journal Åbo Morgonblad, which was suppressed by the Russian Tsar in 1821. After the short period of Finnish Romanticism in Turku, the intellectual center moved to Helsinki where the Hegelian philosophy became popular. When writing about the conceptual history of folk (kansa), Liikanen (2003) emphasizes the importance of the Fennomans and their Hegelian understanding of national spirit (kansallishenki, nationalanda from German Volksgeist). Inspired by Hegelian philosophy, the Finnish philosopher Johann Vilhelm Snellman (1806–81) introduced the concept of “spirit” (Geist in German, anda in Swedish, and henki in Finnish) into the Finnish nationalist discourse. Although our word vector models start from the 1840s, the Hegelian tradition is reflected in these German and Swedish word vector models. Both German and Swedish word vector models associated the Hegelian concept of “spirit” with “patriotism.” Fosterlandskärlek (“love for country”) is also strongly linked with the concept of “spirit” in Swedish results. Finally, the concept of “national spirit” (kansallishenki) is also associated with a mixture of nationalist terms and Hegelian technical terminology in the Finnish results (see Table 12).

It has been customary to make a distinction between cultural and ethnic nationalism [Leerssen 2018] [Smith 2008] [Alter 1994]. For example, the late eighteenth-century philosopher Johann Gottfried von Herder is considered an important early theoretician of nationalism, although he in fact objected to the biological race theories of his time. When Herder emphasized the importance of local national cultures, his intention was to defend small states against the imperialism of the multinational empires of the time [Nisbet 1999]Nisbet 1999). However, after the development of positivism, scientism, and Darwinism, late nineteenth-century nationalism became more based on ideas of ethnicity, race, and the shared biological descent of national populations. This is also reflected in our results. In the Fenno-Swedish corpus, “nationality” (nationalitet) is associated with descent and origins. It seems that “nationality” and “ancestry” (härkomst) are also associated with “religious confession” (trosbekännelse). In contrast to these Swedish results, German word vector models for the concept of nationality (Nationalität) indicate a clear association between “race” and “ancestry.” Finally, the Finnish model is very different to the Fenno-Swedish and German results. The concepts in the word vector space linked with “nationality” (kansallisuus) refer to siwistys (cultivation, cf., Bildung), kulttuuri (culture), kirjallisuus (literature), sanomakirjallisuus (press), kieli (language), and aate (idea), which are related more to cultural nationalism than ethnic nationalism (see Table 13).

In German newspapers (DE-SBB) of the same period, the ethnic roots of national identity are reflected by the close association between the term “folk” (Volk), “christianity” ( Christentum) and “army” (Heer). In similar fashion, the term “nationality” (Nationalität) is used in the context of words that strongly suggest heritage, such as “descent” (Abkunft), “heredity” (Abstammung), “religion” (Konfession, Religion, Christian, Mission), and “language,” both “spoken” and “written,” “local” and “national” (Schriftsprache, Landessprache, Muttersprache, Volkssprache). The term “nation” is also closely related to “civilization” (Civilisation) and “mores” (Gesittung).

As discussed, Koselleck emphasizes that the politicization of nation and folk in Germany towards the end of the nineteenth century was connected with the development of the labour movement and concepts related to anonymous multitudes of people, like mass or mob [Koselleck 1992b, 389]. In other words, the politicization of “folk” did not only happen against other nations, but also in relation to the inner power structures of each country. This temporal process, where the connotations of “folk” change substantially, is reflected very clearly in all of the national newspaper corpora analysed here. For example, in FI-NLF the term “the poor” (köyhälistö) becomes associated with “folk” (kansa) only from the 1890s onwards. In SE-NLF, “social class” (samhällsklass) appears in association with “folk” from the 1870s (Figure 1). In DE-SBB, references to “popular representation” (Volksvertretung) appear in the word vector space around “nation” in the 1870s and 1880s, and “workers” (Arbeiter) appears mainly in the 1880s. Similarly, terms such as “mob” (Janhagel) and “rabble” (Pöbel) appear in the vocabulary around “nation” with connotations such as “riot,” “fright,” “alarm,” and “rebellion” (Aufruhr, Schrecken, Allarm, Aufstand). The term “folk” (Volk) becomes semantically linked to “workers” (Proletariat) in the 1890s (Figure 1).

Whereas the labour movement sympathized with the poor and the proletariat, the political conservatives often made a distinction between “folk” – the ideal common people – and the disobedient “rabble.” In SE-NLF it seems that when “common people” are idealized and well-behaving, they are folk, but when they rebel, they become an anonymous rabble. In the DE-CA corpus, for instance, the concept of Volk (folk) is associated with Schicksal (destiny) in the beginning of the nineteenth century, whereas the association with Proletariat rises towards the end of the century (Figure 2). However, the etymological connections between Germanic languages can be misleading in this case. For example, English “people” is relatively neutral, whereas its German and Swedish homonyms Pöbel and pöbeln refer to a pejorative political concept meaning “mob” or “rabble,” i.e. a disorderly crowd of people [Koselleck 1992b, 143]. The division between idealized folk and dangerous rabble is reflected in Fenno-Swedish word vector models as well: the concept of “rabble” (pöbeln) is associated with “rebels” (rebellerna). In Finnish, rahvas (or rahwas) was first used as a neutral term to refer to “common people,” but acquired the pejorative meaning of “rabble” (roskajoukko) towards the end of the nineteenth century.

In TDA, too, the term “nation” becomes associated with the concept of “democracy” from the 1860s on, and the concept similarity score gradually increases until 1900 A contested aspect of national unity within the United Kingdom is suggested by the terms which appear in the context of the “people,” where a consistently strong association with “countrymen” and “peasantry,” as well as with “Englishmen” and “Irishmen,” reflects the growing tensions around industrialization, urbanization, and home rule. References to Ireland typically evoke subaltern terms such as “oppressed,” “colonists,” and “loyalists” in the word vector space. The proximity between “people” and the term “agitators” in the models of the last decades of the century may reflect the rise of trade unions in the UK.

An interesting transatlantic context is presented by the words surrounding the concept of citizenship within the UK, as the term “inalienable” remains dominant until the first decade of the twentieth century. This seems a clear reference to the “inalienable rights” that were promised in the American Declaration of Independence, which may have informed the British debate about citizenship and national identity. The proximity of terms such as “privileges,” “birthright,” and “heritage” also illustrates the British dissension about the concept of nation and the right to be a citizen. If we superimpose the timeseries of “independence” and “citizenship” from the word vector model based on the query of “nationality” on the same plot, we can see that, during the beginning of the nineteenth century, “independence” is more frequently used in association with “nationality,” whereas “citizenship” dominates at the end of the nineteenth century (Figure 3).

The word vector models illustrate that the concept of “nation,” in spite of its common origin, is expressed in divergent semantic contexts in the European countries examined here. The Finnish connotations of “nation” lean towards a cultural interpretation, suggested by terms like spirit and idea, culture, literature and language. German discourse points at a shared heritage that also suggests a cultural interpretation, as is visible in words such as Christianity and civilization. Although the English-language word vector models of TDA were not as clear, the British concept of “nation” seems to emphasize historical continuity as the basis of national identity, apparent in terms such as destiny and fatherland. British and German word vectors demonstrate a conceptual shift from a public to a more personal understanding of “nation,” in which state and government are seen as intruders of the private sphere, suggested by references to disorganization, taxation and private enterprises. Germany and Finland share a tendency to understand the nation in essentialist and spiritual terms, where the term “folk” embodies the body politic. At the same time, all language sets show that the “people” evoked negative connotations by the vicinity of alarming terms such as “rabble,” “mob,” and “revolt”. The transatlantic connections, lastly, were visible in the British corpus (especially in the references to independence, citizenship and inalienable rights), but virtually absent in the German and Finnish newspapers. Although the word vector models of the German newspapers printed in the United States were not as conclusive, they suggested an interesting avenue of transnational comparison for future research, particularly with respect to the way the nation is understood in different geographical and historical contexts.

If, on the one hand, word vectors lack the precision of a thorough conceptual analysis such as the one conducted by GGB, on the other, they offer the advantage of extending the analysis on much larger datasets that, it can be argued, reflect the public discourse of a larger readership. The longue durée has shown, for instance, the relative instability of the connection between nation and monarchy in the UK. Due to the limitations of having to work with small datasets, these are subtle changes that the GGB, which is based on a limited set of key texts, could not have noticed.

More importantly, the parallel presentation of word vector models from newspapers in different languages opens up a multilingual comparison of public debates around common concepts. Although our methodology has had to rely on modern translations of seed words, detailed comparisons of domain knowledge still allow for a thorough interpretation of the research results. This offers many interesting heuristic possibilities to discover changes in newspaper discourse that may result from conceptual change. The current methodology does not allow us to track cross-border migration of concepts, but it does demonstrate commonalities and contrasts that can direct future research.

In the first experiment, we search for the 15 most related terms for the concept term “illness” in the word embeddings of all corpora for the different time spans. For all different corpora, we encounter spelling variations that have resulted from OCR problems. For example, for the German word Krankheit (illness) we also retrieve terms like Kraukheit or Krankhett, and the Swedish word sjukdom (illness) was retrieved as fjukdom and fjuldom. To enable a better view of the results, we merge these spelling variations for all terms and compute their average similarity score.

In addition, we retrieve synonyms of the term illness such as sickness and disease, which validate the reliability of the semantic models. We focus on the most similar terms that appear in at least four of the different models with a continuous time span, confirming their significance in the corpora. This enables us to identify discrepancies across the different languages and corpora.

The modern era has been characterized by pandemics [Caduff 2015]. The word pandemic itself is a Greek loanword that refers to diseases that affect all (pan) people (demos). In the nineteenth century, new forms of transport and mobility enabled the rapid spread of diseases. Cholera and influenza in particular raged on a global scale [Hamlin 2009] [Honigsbaum 2014]. This effect of globalization is visible in our results. Since the corpora cover slightly different timespans, they do not offer any comprehensive picture of how pandemics emerged and developed conceptually. However, the results refer to different regional perceptions of pandemics. In themselves, “epidemic” and “pandemic” are ancient words that have been used in many European languages since at least the seventeenth century [OED 2019]. In SE-NLF, the Swedish word “epidemi” appears eleven times from 1858–68 onwards. At the same time, it also shows up in DE-CA, first as “Seuche,” which was retrieved as similar to “Krankheit,” 16 times. The loanword “Epidemie” appears in the results slightly later, in 1864–74. In DE-CA, the term “cholera” is retrieved in 1858–68, obviously as a result of the so-called third cholera pandemic of 1846–60. In Europe, cholera raged particularly violently during the Crimean War of the 1850s, but the pandemic soon reached the Americas in the late 1850s, which accounts for discussions in the German immigrant press that label cholera the disease. While the closeness of real pandemics to their conceptual reflections is obvious, the word embeddings indicate that pandemics also caused conceptual changes.

In addition to cholera, influenza constitutes a further interesting example that gained weight in the nineteenth century. The so-called “Russian flu” pandemic was particularly lethal in 1889–90. In the Swedish corpus, too, “influenza” appears in 1888–98. In FI-NLF “influenza” appears for the first time as “influentsa” in 1884–94, not as a synonym for “sairaus” (see above) but for “tauti,” another Finnish term for illness. As a concept, “influenza” went almost as viral as the disease itself. There were, of course, references to influenza before the 1880s, but the international loanword (and its variations) was not used, according to our empirical data. Semantically, influenza hides behind different words, such as flu, or breast or lung disease.

At the beginning of the nineteenth century, the understanding of the mechanisms of nervous diseases was mostly pathological. However, in the course of the century, modern psychology developed a more internal view of the human mind [SChultz and Schultz 2012]. In the newspaper corpora, nervous diseases appear increasingly towards the end of the nineteenth century. At the turn of the twentieth century there were particularly vivid discussions about neurasthenia. This term for weakness of the nerves was coined in the 1820s, but became popular from the 1860s onwards, especially after the publication of George Miller Baird’s article “Neurasthenia, or nervous exhaustion,” in The Boston Medical and Surgical Journal in 1869 [Beard 1869, 217–21]. Neurasthenia was often seen as a disease caused by the accelerating rhythms of modern culture [Uimonen 2000] [Salmi 2013]. Nervous diseases start to appear in our models in the 1860s, and remain an essential conceptual aspect of illness going forward.

The same observation applies to mental diseases. Nervenkrankheit (nervous disease) is retrieved in DE-SBB first in 1882–892, and four times thereafter. In the DE-EU, it comes up in 1888–98, in total eight times during the whole timespan, and in the SE-NLF in 1898–1908 three times in total as nervsjukdom. Mental health issues also appeared through a plethora of similar words: in Swedish as sinnessjukdom (mental disease), as many as forty times after 1862–72; in DE-EU as Geisteskrankheit ten times after 1872–82; and in the DE-SBB four times during the same period. This is not to argue that nervous or mental problems had not been diagnosed before, but conceptually they became more eminent as defining principles for illness.

There are clear differences between the corpora, however. The examples from the Swedish and German material refer to different timescales in the discourses and articulations on nervous and mental issues. It also appears that Nervenkrankheit is not being retrieved in DE-CA or in FI-NLF. Although the models do not make the reasons for this absence clear, these results may be explained by the different audiences of the Finnish-language and German immigrant press. The former was addressing a rural population, whereas neurasthenia was mainly discussed in a middle-class context. This is also supported by the fact that the Finnish-language models yielded many words that refer to animal and plant diseases, which were important for the rural readership of the Finnish-language press. These findings support the conclusion that “illness” fuelled conceptual migrations across borders, while at the same time reflecting social boundaries for the conceptualization of diseases. However, the word vector models do not allow for further exploration of the context of these concepts, which means that the semantic interaction between human and natural domains in a period when Darwinian evolution theory was debated worldwide would require more research [Hawkins1997] [Bowler 2003, 224–324]

The etiological standpoint, coined by Robert Koch in the 1870s, refers to the assumption that diseases are best controlled and understood by means of their causes. Since this theory has dominated medical discourse for the past two centuries, much of contemporary medical practice concentrates on identifying specific causes of disease. A disease, however, is regularly defined by specifying its nature. To study the representation of illness at community-level in the nineteenth century, our results show that this identity can be expressed by more than causes. Our word embeddings reflect this by showing vocabularies which conceptualize illness from different perspectives such as symptoms, their consequences, synonymous terms, and treatment. By breaking up illness into categories (e.g. treatment, consequences, and symptoms) we can demonstrate its multimodal nature as a concept without fixed entities. Moreover, the word “illness” is used in different senses or meanings. The aim of this use case is to investigate whether the dominant sense of illness changed over time. The findings support the claim that “there is no single way of defining, interpreting, experiencing or managing disease” [Jackson 2017, 6]. Our results show that, in contrast to medical discourse, the nineteenth-century press did not concentrate on causes and aetiology of diseases, but rather focused on their human consequences.

To investigate the change in word sense, we first translate all similar terms for “illness” into nineteenth-century English. Then, two annotators assign categories to each term. After a close reading of the vocabulary, we define the subcategories listed in Table 16:

We assigned words such as “death,” “crisis,” or “incurable” to the consequences category, “sore throat,” “tiredness,” or “vomiting” to the symptoms category, “ill,” or “sickness” to the synonymy category, “cholera,” “influenza,” or “mental disease” to a category for specific diseases, “surgery,” or “inoculation” to treatment, and words that were not context-related to the other category.

The annotation and comparison of the different subcategories highlights that the majority of generated words are synonyms and consequences of diseases. These categories are dominant in all six corpora. As Figure 5 illustrates, analysis of word embeddings in diverse newspaper corpora suggests the discussion of consequences of diseases rarely respects local, regional, or national borders. This leads to the conclusion that the press constructed illness as a concept with an emphasis on its consequences. This emphasis is supported by mainly negative characteristics such as, for instance, “epidemic,” “poverty,” “died”, and “suffering.” Thus, diseases are represented as agents of suffering, misery, and death, which have a strong negative impact on human lives.

However, shifting temporal and corpus-specific patterns can be observed, as illustrated by Figure 5. Whereas DE-SBB only textualizes consequences from 1870 onwards, no words categorized as consequences seem to occur in DE-EU after 1878. Moreover, words that refer to consequences are widely discussed in FI-NLF and SE-NLF throughout the entire time span.

Analysing the different categories in which the concept illness is discussed in newspapers allows us to trace the fluctuating meanings of disease and illness as a superordinate category. The overview of subcategories also helps us to consider the role of the press in defining diseases. The fact that there was a strong emphasis on the consequences of illness may be explained by the dual nature of newspaper discourse. On the one hand, the press constructed fear of diseases by describing the disastrous consequences of illness, becoming a platform for emotional expression. On the other, the press expressed social observations and shared them with a wider audience. Pandemics such as influenza and cholera were strongly present in the nineteenth-century public sphere, and papers informed their readers of their impact on the life of an individual as well as on society.

The combination of synchronic and diachronic analysis of word vector models allows us to trace the development of an abstract concept such as illness over time in newspaper corpora in different languages. We can discover how specific diseases emerged in public discourse in Europe and the United States at different times in history, such as the epidemics of cholera and influenza that affected these parts of the world. More interestingly, word vectors are an effective instrument to trace the vocabularies with which national publics discussed – and constructed – diseases, and their supposed causes and consequences. Our study shows the influence of public discourse, newspaper publishing in particular, on definitions of illness and the multidimensional network of relations between symptoms, specific diseases, and their consequences to construct the concept of illness. Although this case study was limited to newspaper collections from a small number of nations, the differences and silences in discourse, and the changing vocabularies that emerge from the word vector models, offer interesting heuristic starting points for a close reading of these corpora within their proper historical context.