Paul Thibodeau and Lera Boroditsky gave people a paragraph about crime in a fictional city called Addison. The paragraph included statistics — crime rates, arrest figures — and asked what the city should do about it.
For half the readers, one sentence described crime as a virus infecting the city. For the other half, the identical sentence described crime as a beast preying on the city. Everything else, including all the numbers, was the same.
Readers given the virus framing were substantially more likely to propose social reform: investigate the causes, improve education, address poverty, treat the underlying conditions. Readers given the beast framing were more likely to propose enforcement: more police, harsher sentences, capture and contain.
The difference produced by one word was larger than the difference between self-identified Democrats and Republicans in the same sample.
Then came the finding that makes this a manipulation mechanism rather than a curiosity about writing style. Participants were asked what had influenced their decision. They cited the crime statistics. Almost nobody identified the metaphor, and when it was pointed out, many did not accept that it had mattered. In a follow-up, participants were asked to identify the most influential part of the report and overwhelmingly pointed at the numbers — the part that had been held constant.
This is the confabulation effect Nisbett and Wilson documented in 1977, in which people confidently report the reasons for judgments whose actual causes they have no introspective access to. The reported reason is not a lie; it is a reconstruction, generated after the fact from what seems like it ought to have been the reason.
What a metaphor does, in the account George Lakoff and Mark Johnson developed, is import an entire inferential structure. Beast is not a synonym for bad. It brings a whole schema with it: predators are hunted, caged, killed; they are external to the community; they act on instinct and cannot be reasoned with; the appropriate response is force and containment. Virus brings a different one: infections have vectors and conditions that allow them to spread; they are diagnosed and treated; the appropriate response is to understand transmission and improve the underlying health of the system.
Neither schema is argued for. Both arrive complete inside a single noun, and the noun is not treated by the listener as a claim requiring support, because it presents itself as a description.
Which makes metaphor the most efficient smuggling mechanism in political language, and the catalogue is easy to build once you are looking. Immigration described as a flood, a wave, a surge — natural forces that are contained with barriers, not people who are processed by institutions. A war on drugs, on terror, on cancer, on poverty, importing the schema in which there is an enemy, dissent is disloyalty, and casualties are acceptable. An economy that is sick or healthy, a market that is nervous, a company family, a talent pipeline.
The countermeasure is to read the metaphor as a claim and ask whether it holds. Is crime like a predator, in the specific respects that matter for the policy conclusion? Sometimes the answer is partly yes, and the metaphor is a genuine argument that can now be examined. Very often the answer is no, and the entire conclusion was arriving inside a word nobody had to defend.
Cross-ref: framing effects chapter above