Reasoning COVID-19: the use of spatial metaphor in times of a crisis

The paper investigates how spatial metaphors structured reasoning and decision-making in early COVID-19 discourse and how such metaphors were used to justify political measures. Positioned within geography’s linguistic turn, the study bridges cognitive linguistics (metaphors and image schemas), discourse analysis (hegemonic framing and power), and pragmatics/argumentation (Toulmin’s model) to understand how metaphors such as containers, borders, hotspots, waves, and personified agents shaped public narratives. The purpose is to provide a theoretically grounded and empirically supported account of spatial metaphor usage in German-language media and to propose argumentation theory as a hinge between cognitive and discourse approaches for analyzing crisis communication.

The background situates geography’s move from spatial essentialism to linguistic constructivism, highlighting how space/place are socio-linguistic constructions (Barnes & Duncan; Jackson; Werlen; Paasi). Cognitive linguistics (Lakoff & Johnson; Lakoff 1987) emphasizes image schemas (CONTAINER, PATH, CENTER-PERIPHERY, etc.) as foundational to spatial thinking; Schlottmann demonstrates persistent container metaphors in socio-spatial imaginaries (e.g., post-unification Germany). Place images remain stable despite material change (Weichhart & Weixlbaumer; Shortridge). Discourse theory/analysis (Kendall & Wickham; Wodak & Chilton; Chilton) shows how hegemonic discourse frames social/political phenomena via spatial/territorial distinctions (Glasze; Mattissek; Rheindorf & Wodak). Cognitive and discourse approaches are not mutually exclusive; CDA integrates cognitive metaphor theory (Semino et al.; Musolff; Hampe). Pragmatics and argumentation (Wittgenstein; Austin; Searle; Brandom) foreground language as social action with inferential commitments. Toulmin’s model (Claim, Data, Warrant, Backing) captures the interplay of explicit statements and implicit assumptions, linking cognition and discourse. Prior research on crisis metaphors documents cross-domain mappings to natural disasters, war, and disease to simplify complexity and mobilize action (Ziem; Wengeler & Ziem; O’Mara‑Shimek et al.; Charteris‑Black; Semino; Craig; Chapman & Miller). In COVID-19, war, wildfire, hotspot, wave, and epicentre metaphors were prevalent; TPSN framework (Jessop et al.; Brinks & Ibert) relates measures to territory, place, scale, and networks.

Design: Two-phase approach combining a qualitative explorative pre-study with a semi-automated distant reading analysis. Data: German-language COVID-19 corpora from DWDS (publicly funded Digital Dictionary of German Language). Explorative analyses used 2020 data via the DWDS web interface; automated analyses used the GitHub coronakorpus snapshot (retrieved Aug 13, 2020). Comparative subcorpora: high-quality journalism, boulevard press, and private websites. Qualitative pre-study: Identified recurrent spatial metaphor patterns in media discourse—toponyms, containerisation/spatialisation (borders, hotspots, epicentres), naturalisation (storm, wildfire, tsunami, waves), and personification (virus as an agent; territories as actors). Applied Toulmin’s argumentation schema to illustrative statements (e.g., containment logic; place-based causation such as Ischgl; personification of the virus to justify measures). Operationalisation for distant reading: Derived search patterns (Table 3):

• Personification: structural collocations where COVID-term directly collocates with a finite verb (++ or ++).
• Toponyms: Named Entity Recognition with tag ‘LOC’.
• Spatialisation: search terms including ‘Grenze’ (Region initially considered but dropped).
• Naturalisation: search terms ‘Epizentrum’, ‘Hotspot’, ‘Sturm’, ‘Welle’ (Tsunami dropped later).
• Analogy: structural collocations with comparison particles () linking COVID terms and nouns. Implementation: POS tagging and NER using established NLP toolkits (e.g., Stanza; NLTK). Filtering by pattern, then context reduction to (proper) nouns. Cooccurrence analysis via association rule mining to visualise semantic context graphs (nodes = nouns; directed edges weighted by confidence of cooccurrence within a sentence/paragraph window). Thresholds: minimum term frequency ≥ 3; association rule confidence ≥ 0.1. Analysis workflow: Iterative filter–refine strategy—apply patterns to subcorpora, generate frequency lists of contextual terms/toponyms, visualise context graphs around selected markers (e.g., ‘breiten’, ‘Hotspot’, ‘Grenze’, ‘Welle’, ‘Sturm’), and conduct close reading of representative statements surfaced by the automated steps.

• Personification: Across media types, the predominant verb collocating with COVID terms was ‘(aus)breiten’ (to spread), reinforcing the virus-as-agent narrative. Context graphs (e.g., for boulevard press filtered on ‘breiten’) consolidated the origin storyline (e.g., Wuhan, Hubei, subsequent lockdowns).
• Toponyms and containerisation: Early discourse was strongly national in focus. Frequent toponyms (Table 4) included, for high-quality journalism: Deutschland (2639), China (1598), Italien (941), USA (835), Europa (734), Berlin (582), Wuhan (544); boulevard: Deutschland (1672), China (1458), USA (817), Italien (712), Berlin (502), Wuhan (592), Europa (495); private websites: Deutschland (66), China (43), Italien (30), Europa (22), Welt (19), USA (17), Wuhan (16). Narratives often equated spatial origin with causation (e.g., tracing Austrian cases back to Ischgl).
• Borders (Grenze) and measures: ‘Grenze’ co-occurred frequently with ‘Maßnahmen’, highlighting borders as levers of control; in German, ‘Obergrenze’ appeared as a lexicalised threshold metaphor for case limits. Context graphs tied ‘Pandemie’ to ‘Kontrolle’ and framed ‘Krise’ and ‘Corona’ as mutually dependent markers.
• Naturalisation metaphors: ‘Hotspot’ was highly productive and associated more with city/municipality scale (e.g., Heinsberg, Mitterteich, New York, Ischgl) than regions/nations; proportional emphasis was greater in boulevard press. In the ‘Hotspot’ context (Table 5), frequent toponyms included USA (14) and Heinsberg (9) for high-quality journalism; and Spanien (9), USA (5), Mitterteich (4), Deutschland (4), New York (4), Ischgl (3) for boulevard. ‘Epizentrum’ skewed toward Wuhan/Hubei (especially boulevard), with Italy/Europe and New York also framed as epicentres. ‘Welle’ was widely used (also metaphorically for solidarity waves). ‘Sturm’ was comparatively unproductive (Table 6), with co-occurrences like ‘Noah’, ‘Arche’, ‘Wunder’ (2 each) in high-quality journalism, indicating alternative political-religious imagery (e.g., ‘Arche Noah’).
• Analogy patterns: Comparison constructions surfaced recurring metaphors such as ‘Brandbeschleuniger’ (fire accelerator), ‘Lauffeuer’ (wildfire), ‘Bedrohung’ (threat), ‘Damoklesschwert’ (sword of Damocles), and ‘Angreifer’ (attacker) (Table 7). Example uses flagged worry about uncontrolled spread among vulnerable populations.
• Scaling and generalisation: Media narratives frequently ‘zoomed out’ from specific places/events to generalised risk regions, enabling quick inferential judgments (e.g., if from Bielefeld/Niedersachsen, then risk), even when premises were factually mistaken.
• Media-type differences: Boulevard press proportionally relied more on hotspot/epicentre framing and place-based origin stories; private websites showed more idiosyncratic personification verbs and unique analogies, though overall counts were lower. Overall, spatial metaphors positioned space as the primary target domain for reasoning: national containers and borders for initial control; hotspots and epicentres for localised containment; natural disaster metaphors framing the pandemic as rapid, intense events despite the multi-year temporal reality.

Findings demonstrate that spatial metaphors provided a powerful cognitive and argumentative scaffold for justifying policy responses in the early pandemic. The national container served as the initial scale of action (e.g., border closures) consistent with legislative competencies. As asymptomatic transmission undermined coarse containment, the discourse adapted within the same spatial domain—identifying hotspots and epicentres to rationalise curfews and local measures, sometimes retroactively attributing widespread infections to single locations (e.g., Ischgl). Personification of the virus enabled agency-oriented reasoning (‘the virus spreads/acts’), while naturalisation metaphors (waves, wildfires, epicentres) heightened urgency but risked suggesting brief, externalized catastrophes, potentially misaligning expectations with the protracted nature of pandemics. Argumentation analysis (Toulmin) revealed how explicit claims (need for restrictions) drew strength from implicit spatial warrants (containers/borders are effective units of control) and backings (virus as an agent, territories as responsible actors). The study argues that argumentation theory effectively bridges cognitive metaphor analysis and discourse analysis by exposing the inferential structure linking metaphors, scale-jumping, and policy legitimation.

The study shows that, in early COVID-19 discourse, spatial metaphors—containerisation, borders, hotspots/epicentres, waves—were central to explaining the crisis and legitimising measures. Guiding narratives cast Wuhan as the epicentre sending shockwaves to global hotspots; borders and national containers were initial levers for containment; as their limits emerged, hotspot framing sustained spatial reasoning at finer scales. Natural disaster metaphors framed the pandemic as intense but temporally limited, contrasting with the long duration of pandemics. Conceptually, the paper proposes argumentation theory as a hinge between cognitive linguistics and discourse analysis to reveal how metaphors function within inferential structures of public reasoning. Methodologically, an iterative filter–refine distant reading approach combining POS/NER patterns, cooccurrence analyses, and association rule mining efficiently surfaced contexts and statements for close reading. Future research should systematically evaluate retrieval performance (precision/recall), extend pattern sets and languages, assess temporal dynamics of metaphor usage, and test generalisability across media ecosystems and crises.

• Corpus scope limited to German-language sources from early pandemic stages (DWDS/coronakorpus snapshot as of Aug 13, 2020), constraining temporal and cross-linguistic generalisation.
• Selective, exploratory design prioritised variability over exhaustive coverage; findings are illustrative and not statistically generalizable.
• No systematic evaluation of the toolset’s information-retrieval performance (e.g., precision/recall); thresholds (frequency ≥3, confidence ≥0.1) were heuristic.
• Pattern lists were incomplete and adjusted during exploration (e.g., ‘Region’ and ‘Tsunami’ dropped), potentially omitting relevant metaphors.
• Media-type categorisation was coarse; differences observed may reflect corpus composition as much as stylistic tendencies.