1. ICMs: The Architecture Beneath Political Language
George Lakoff proposed Idealized Cognitive Models in 1987 as an account of how human cognition organises knowledge into schematic, idealised structures that support rapid inference and categorisation. An ICM is not a description of how things actually are — it is a cognitive schema of how we expect things to be in typical cases. The 'ideal' in ICM is cognitive, not normative.
In political communication, ICMs function as the pre-conscious infrastructure of interpretation. When a politician invokes 'hard-working families,' they activate the FAMILY ICM with all its associated inferences: unity, care, shared resource, children's futures, sacrifice. The factual content of any policy attached to this ICM is processed through the schema's lens — accelerating endorsement when there is ICM alignment, and generating resistance when there is ICM mismatch, regardless of the policy's objective merits.
2. The ICM Activation Toolkit
Our corpus analysis identified 23 dominant ICMs recurrent across the political texts studied. The five most frequently activated were: NATION AS FAMILY, ECONOMY AS PERSON/BODY, POLITICS AS WAR, SOCIETY AS BUILDING, and GOVERNMENT AS PARENT.
Each ICM carries a characteristic set of entailments. ECONOMY AS BODY/PERSON entails that economic policy is analogous to personal health management — implying individual agency, recovery narratives, and the naturalness of boom-bust cycles. This ICM powerfully constrains what economic policy responses feel intuitively correct, systematically favouring interventions that match 'treatment' and 'recovery' schemas over structural or redistributive approaches.
ICM blending — the combination of multiple ICMs in a single communicative act — proved to be the most sophisticated rhetorical strategy in our corpus. The most effective political communicators consistently demonstrated the ability to activate compatible ICMs simultaneously, creating reinforcing inference chains that made their policy positions feel multiply validated.
3. ICM Alignment as a Predictor of Policy Support
To test whether ICM alignment predicted policy support independently of other factors, we combined our corpus analysis with YouGov electorate profiling data for each country. ICM profiles of electorates were constructed from qualitative survey data, deliberative polling transcripts, and focus group materials — capturing the dominant ICMs operative in each political context.
Regression analysis revealed that ICM alignment was the single strongest predictor of policy support across all six countries (β = 0.61, 95% CI [0.54, 0.68]). This effect held controlling for factual accuracy (β = 0.23), speaker credibility (β = 0.31), and policy specificity (β = 0.17).
The implication is stark: what matters most for policy support is not whether a policy is accurate, detailed, or proposed by a trusted speaker, but whether it is framed through ICMs that match the electorate's existing cognitive architecture.
4. The Suppression of Counter-ICMs
One of the most practically significant findings concerns the dynamics of ICM suppression. Standard 'fact-checking' and 'correction' models of political communication assume that providing accurate information will override misinformation or misleading framing. Our data provide large-scale empirical confirmation of Lakoff's 'don't think of an elephant' argument: corrections delivered within the counter-ICM frame showed near-zero persuasion effect (β = 0.04, ns).
More concerning, corrections that explicitly named and engaged the target's ICM framing — even to refute it — showed modest backfire effects in two countries (Australia, d = −0.18; India, d = −0.21), consistent with the psychological reactance literature. The implication is that ICM counter-communication must involve ICM replacement, not ICM refutation — a substantially more demanding communicative task.
5. Implications for Democratic Communication
These findings raise important questions about democratic communication and political epistemology. If ICM alignment is a stronger predictor of policy support than factual accuracy, this implies that democratic deliberation is substantially less fact-driven than normative democratic theory assumes.
For progressive policy communication specifically, our data suggest that policy positions are often not unpopular in themselves — they are ICM-misaligned. The same policy delivered through an aligned ICM consistently outperformed the misaligned version in controlled vignette experiments. This is not a licence for political manipulation; it is an argument for taking the cognitive architecture of democratic communication seriously.
Journalists, policy advocates, and political communicators who invest in ICM analysis and alignment are not subverting democratic discourse — they are meeting citizens where their minds already are.