‘Is climate science taking over the science?’: A corpus-based study of competing stances on bias, dogma and expertise in the blogosphere

The study investigates how climate science blogs function as politicized, multivoiced arenas where scientific evidence, values, and claims to expertise are negotiated. Using a subset of the Genealogies of Knowledge Internet corpus, the Climate Science Blogger Corpus (CSBC), the research asks how bloggers with different forms of expertise (core, contributory, interactional) position themselves vis-à-vis competing perspectives. It focuses on how lexical items—bias, dogma, and peer review—are used to construct intersubjective stances, claim or challenge expertise, and mediate the dialectic between epistemic and non-epistemic values. Grounded in Studies of Expertise and Experience (SEE) and Martin and White’s engagement framework, the study aims to reveal how climate knowledge is contested beyond the binary ‘alarmists vs deniers’ framing and why these dynamics matter for public understanding and policy debates.

Prior work shows media significantly shape public perceptions and politicization of climate change, with increased exposure to competing discourses and audience selective news seeking based on cultural congruence. Linguistic and corpus-based studies have examined frames, metaphors, and lexical patterning in mainstream media (e.g., Brazilian news corpora indicating prevalence of gradualist views) and social media/blogs (including grammar induction approaches). Theoretically, second-wave science studies (SSK) emphasized democratization and the co-production of science and society, but offer limited tools to specify what expertise is and who counts as an expert. SEE (third wave) introduces a taxonomy of expertise—core, contributory, and interactional—separating technical from political inputs while recognizing experience-based expertise. Climate science has also been discussed as value-laden, with values playing direct and indirect roles in scientific reasoning and policy translation, which can fuel polarization. The study draws on Martin and White’s appraisal/engagement framework to analyze heteroglossic positioning and dialogic (contractive vs expansive) strategies in climate blog discourse.

Design: A corpus-based analysis using the Climate Science Blogger Corpus (CSBC), a five-blog subset of the Genealogies of Knowledge Internet corpus capturing diverse positions in the climate debate. Selection criteria included explicit stance on climate change, varied authorship/agendas (scientists, journalists, researchers, lobbyists), and consent for inclusion. Composition: 448,608 tokens from five blogs (US, UK, Australia), with most posts published 2007–2019 (majority 2014–2018). Subcorpora: Contrarian (CSBC-CON: Australian Climate Madness; Science Defies Politics; Climate Depot; total ≈ 91,021 tokens) and Acceptor (CSBC-ACC: DeSmog UK; Union of Concerned Scientists; total ≈ 357,587 tokens). Post counts and token totals by outlet (as reported): Australian Climate Madness: 169 posts, 34,080 tokens; Science Defies Politics: 49 posts, 48,414 tokens; Climate Depot: 9 posts, 8,527 tokens; DeSmog UK: 218 posts, 252,968 tokens; UCSUSA: 89 posts, 104,619 tokens. Analytical framework: Systemic Functional Linguistics—appraisal/engagement (Martin & White, 2005), focusing on dialogically contractive vs expansive positioning and ‘pronouncing’ resources. Targets of analysis: Three lexical items associated with epistemic standards and expertise—bias/biases, dogma, and peer review—selected from frequency lists (≥15 occurrences), prioritizing nouns and pre-modifiers unambiguously tied to epistemological evaluation; items functioning as verbs or ambiguous POS were excluded. Tools and procedures: Concordances generated via GoK interface; visualization with Metafacet (distribution across outlets) and Mosaic (positional collocation statistics) using MI3 (EXP scale), typically in local mode to display broader collocate sets. For bias*, verb uses were removed; analyses emphasize qualitative interpretation of collocational networks and engagement strategies rather than statistical salience, given relatively small subcorpora. The analysis contrasts how CSBC-CON and CSBC-ACC deploy these lexical items to construct intersubjectivity and to legitimize/de-legitimize expertise.

Bias: 102 total occurrences across CSBC, appearing in 4 of 5 blogs; despite CSBC-ACC being ~4× larger, absolute counts are similar across subcorpora, implying proportionally higher use in CSBC-CON. In CSBC-CON (34 noun uses after removing verbs), bias co-occurs with lexis linking mainstream science to ideology/politics (e.g., alarmist, IPCC, media/tech companies), foregrounding non-epistemic, value-driven flaws and institutional/corporate influence. In CSBC-ACC, bias is used to critique methodological shortcomings (e.g., biased modeling, poor protocols, instrument calibration) and to highlight problematic funding sources, with occasional acknowledgment that values can directly and legitimately inform societal priorities. Dogma: 16 total occurrences in CSBC; 12 occur in CSBC-CON, framing climate science as ‘dogma’ or ‘climate (cult) dogma,’ often contrasted with ‘science’ (e.g., ‘calls its dogma science’; ‘settled science’), and associated with religious metaphors (sin/repentance; Galileo vs Church), and with actors like UN bodies, corporations, and ‘alarmists’ reinforcing dogma. The 3 DeSmog UK instances are quoted speech from contrarian figures (dialogically expansive), used to expose prejudice and politicization. Peer review: 19 total occurrences across CSBC; present in both subcorpora but proportionally higher in CSBC-CON (7 lines analyzed vs 12 in CSBC-ACC). CSBC-CON associates peer review with ‘alarmism’, ‘IPCC’, and pejoratives such as ‘schmeer-review’, ‘pal-review’, ‘skewing’, ‘corruption’, challenging legitimacy as politically captured rather than critiquing specific procedural mechanics. CSBC-ACC associates peer review with journal standards, rigorous editorial policies, and gatekeeping; negative collocates (e.g., ‘sloppy’, ‘corruption’) typically appear within quoted contrarian critiques (dialogically expansive) and are counter-argued (e.g., coverage of Energy and Environment’s overhaul and editor change). Overall patterns: Contrarian bloggers rely on dialogically contractive pronouncements to portray mainstream climate science as politicized and dogmatic; acceptor bloggers predominantly use dialogically expansive strategies to attribute and rebut contrarian claims, while grounding their own stance in epistemic values and methodological rigor.

Findings demonstrate how climate bloggers construct intersubjective positioning to claim or contest expertise. Contrarian authors, despite presenting themselves as dispassionate and process-focused, predominantly frame consensus science via non-epistemic values (political, ideological, corporate), elevating perceived biases to ‘dogma’ and casting peer review as partisan legitimation mechanisms. This reveals a mismatch between their avowed neutrality and their politicized, contractive stance. Acceptor bloggers largely treat bias as arising from methodological decisions and instrumentation, maintain peer review as a core epistemic gatekeeping process, and employ dialogically expansive strategies that quote and contextualize opponents’ claims to highlight subjectivity and inconsistency with scientific norms. The contrast maps onto SEE’s taxonomy: contestation of who holds contributory vs interactional expertise and how values are mobilized (direct vs indirect roles). These dynamics clarify how epistemic authority and legitimacy are negotiated in public, influencing perceptions of climate knowledge and its translation into policy.

The study shows that targeted lexical items—bias, dogma, peer review—serve as key pronouncing resources through which climate bloggers position themselves, assert or question expertise, and foreground the role of values in climate knowledge production. Contrarian blogs tend to deploy contractive, politicized framings (bias as ideological, science as ‘dogma’, peer review as corrupted), while acceptor blogs foreground methodological rigor, institutional standards, and use expansive attribution to expose opponents’ subjectivity. Methodologically, the paper demonstrates how corpus-assisted tools (Metafacet, Mosaic) combined with appraisal theory can yield systematic insights into intersubjective positioning in blog discourse. Future research should expand the evaluative lexicon studied, probe additional genres within the Genealogies of Knowledge Internet corpus, and assess whether observed stance patterns generalize across larger and more diverse datasets.

The corpus is relatively small compared to large-scale studies and restricted to five blogs that granted inclusion consent, potentially limiting representativeness. The analysis focuses on three lexical items, prioritizes qualitative interpretation over statistical salience, and excludes certain parts of speech to ensure comparability. Most posts fall within 2014–2018, which may not capture longer-term shifts. The study emphasizes engagement patterns rather than comprehensive procedural audits of peer review or exhaustive mapping of all evaluative lexis.