The Metaphor Must Pay Rent

Thesis

Metaphors are indispensable in science and philosophy, but they become misleading when their emotional force is mistaken for evidence. A useful metaphor must say what it illuminates, what it omits, and what work it enables next.

Why we need borrowed language

Supporting/contextual references: [black-1962] [hesse-1966] [lakoff-johnson-1980]

No field begins with a vocabulary perfectly fitted to its subject. Scientists borrow from ordinary life: particles have spin, fields have momentum, genes have code, and black holes have horizons. Philosophers borrow images too: the mind is a theater, society a contract, time a river. These phrases help a reader approach unfamiliar structure. They are not failures of rigor simply because they are figurative.

Trouble begins when a metaphor receives the authority of the theory from which it was borrowed. A “quantum leap” becomes a sudden personal transformation; “information” becomes a cosmic substance; “entanglement” becomes proof that lovers share a mystical channel. The borrowed word sounds technical while the claim has quietly left the conditions that gave the word meaning. The metaphor has stopped carrying a reader toward an idea and started carrying an idea past scrutiny.

The problem is not solved by banning figurative language. A reader needs images before they can manipulate abstractions, and experts use models that are partly visual, analogical, and narrative. The better practice is to make the handoff visible. Say when a term is technical, when it is a heuristic, and when it is being used for existential resonance. A metaphor that announces its status can remain vivid without acquiring counterfeit authority.

Rent as an epistemic test

Supporting/contextual references: [hesse-1966] [keller-1995]

A metaphor pays rent when it performs identifiable work. It may compress a complex relation, suggest a fruitful question, organize observations, or reveal a limit in an older picture. Darwin’s tree of life highlights branching descent and common ancestry, while also requiring revision when horizontal gene transfer complicates the tree. The metaphor helps because its structure can be compared with evidence and corrected where it fails.

The rent test has three parts. First, state the source domain and the target domain: what is being compared with what? Second, identify the preserved relation rather than the shared atmosphere. Third, name the point where the comparison breaks. “The brain is a computer” can point toward information processing and architecture; it does not settle consciousness, embodiment, or the nature of meaning. Explicit limits make a metaphor stronger, not weaker.

The machine in the mind

Supporting/contextual references: [lakoff-johnson-1980] [keller-1995]

The computer metaphor transformed cognitive science by making representation, memory, and computation researchable. It encouraged models in which a system takes inputs, transforms internal states, and produces outputs. Those models generated experiments and useful distinctions. But a metaphor can become a default ontology. If the brain is treated as a computer in every relevant sense, a researcher may overlook metabolism, plasticity, bodily regulation, affect, and the role of an organism embedded in a world.

The point is not to reject computational models. It is to ask what the model explains. A program can specify formal transitions without specifying the meaning of its symbols; a brain’s significance-bearing activity depends on a living system with needs, history, and social learning. Competing frameworks can coexist when they answer different questions. The metaphor pays rent when it predicts or organizes behavior, not when it is used to declare that subjective experience is “nothing but software” without an argument about realization.

Information, code, and purpose

Supporting/contextual references: [shannon-1948] [landauer-1961] [keller-1995]

“The universe is information” is a powerful sentence because information theory has changed physics and engineering. Shannon’s measure quantifies uncertainty in a source; quantum information describes allowed states and operations; Landauer’s principle connects logical erasure with thermodynamic cost. None of these uses requires information to be a free-floating material. Information is realized in physical distinctions and correlations under a specified description.

Biology adds another layer. DNA can function as a sequence in a regulatory and developmental system, where molecular machinery responds to patterns. Calling it a code can illuminate mapping and translation, but the analogy does not imply a programmer outside biology. Evolution explains how systems capable of using such sequences can arise through variation, inheritance, and selection. “Code” names a relationship among molecules and processes; it is not, by itself, a proof of intention.

The universe as a computer

Supporting/contextual references: [gleick-2011] [landauer-1961] [ball-2016]

The computational-universe metaphor has several distinct forms. One says physical systems can be simulated or described computationally. Another says the laws of nature are computational processes. A stronger proposal says computation is ontologically fundamental and matter emerges from it. These are not three versions of one claim. The first can be a practical engineering observation; the second is a theory about laws; the third is metaphysics that needs an account of what runs, what counts as a state, and why this computation has these rules.

Digital physics can generate sharp questions about discreteness, complexity, and physical limits. It can also tempt us to mistake a representational choice for a discovery about reality. Any physical theory must recover continuous phenomena where observed, explain Lorentz symmetry if it claims a preferred computational grid, and make predictions that differ from rivals or clarify why its ontology is superior. “It is all code” earns no rent until it does more than redescribe equations in a new idiom.

The map is not the territory

Supporting/contextual references: [black-1962] [hesse-1966] [lakoff-johnson-1980]

The map metaphor is itself instructive. A map is not the terrain, yet it can be objectively better or worse: it preserves some relations, omits others, and supports navigation. Scientific models work similarly. A model may idealize frictionless planes, point particles, or rational agents without claiming that such things exist literally. Its success depends on the contexts in which the retained structure tracks the world well enough for prediction and intervention.

The warning cuts both ways. Saying that a model is not the territory does not make every model equally fictional. Scientific representations are constrained by calibration, measurement, and failed predictions. Nor does a map become the landscape because it is mathematically elegant. The metaphysics of a model must be argued separately from its utility. This is why a beautiful equation can support a realist interpretation without forcing one, and why an evocative picture can remain a picture even when it is memorable.

A map also has a user and a purpose. A subway diagram preserves connections while discarding geography; a geological map preserves strata while ignoring the commuter's route. Neither is a deficient street map. The same is true of a scientific model. Asking what it leaves out is not an accusation but a way to learn what question it was built to answer. Confusion begins when one map is treated as the whole territory simply because it is excellent for one journey.

There is a further ethical reason to mark the edge of a metaphor. Images influence which possibilities feel natural and which people become visible. Describing a society as a machine may make maintenance seem more important than care; describing a person as a data point may erase a history that matters to the decision. No metaphor is politically neutral simply because it is familiar. Paying rent includes asking what the image foregrounds, what it hides, and who bears the cost of the hidden remainder.

Open research directions

Supporting/contextual references: [black-1962] [hesse-1966] [shannon-1948] [landauer-1961] [ball-2016] [gleick-2011] [keller-1995]

Research continues on how metaphors function inside scientific change. Historians and philosophers study when analogies generate concepts, how models transfer structure, and how a figurative term becomes technical. Cognitive science tests where analogy supports reasoning and where intuitive mappings create systematic errors. These programs examine language and practice; they do not turn poetic association into discovery.

In physics, quantum information, emergence, computation, and spacetime provide a useful stress test. Researchers ask whether information-theoretic principles derive known laws, whether computational descriptions add predictions, and which features of an image survive formalization. The answers remain unsettled. A metaphor may guide a research program without being a result of it, so a productive heuristic should remain distinct from an established ontology.

Failure is especially informative. When a model misses a scale, omits an interaction, or encourages a false intuition, the mismatch shows which features never transferred. Education can use the same discipline: teach the memorable picture alongside the experiment or equation that limits it. Knowing where an image breaks is part of knowing what it means.

Keep the image, mark the edge

Supporting/contextual references: [black-1962] [hesse-1966] [lakoff-johnson-1980]

The alternative to inflated metaphor is not sterile prose. Images make difficult ideas portable; equations and experiments keep them accountable. Carry an image far enough to reveal a relation, then stop before it impersonates a mechanism. A horizon can illuminate the limits of access while remaining a precise structure in spacetime rather than a symbol that explains every personal mystery.

The best image leaves a better question behind: what physical system stores this information, which relation survives the analogy, what observation would distinguish a literal claim, and where does the comparison fail? Those questions keep meaning in contact with the world. Wonder becomes more durable when the picture knows what it cannot carry.

Sources & references

Supporting/contextual references, not claim-level proof.

1. Claude E. Shannon — A Mathematical Theory of CommunicationBell System Technical Journal 27(3), 379–423; 27(4), 623–656, 1948.10.1002/j.1538-7305.1948.tb01338.x
2. Rolf Landauer — Irreversibility and Heat Generation in the Computing ProcessIBM Journal of Research and Development 5(3), 183–191, 1961.10.1147/rd.53.0183
3. Max Black — Models and MetaphorsCornell University Press, 1962.
4. Mary B. Hesse — Models and Analogies in ScienceUniversity of Notre Dame Press, 1966.
5. George Lakoff and Mark Johnson — Metaphors We Live ByUniversity of Chicago Press, 1980.
6. Philip Ball — Patterns in Nature: Why the World Looks the Way It DoesUniversity of Chicago Press, 2016.
7. James Gleick — The Information: A History, a Theory, a FloodPantheon, 2011.
8. Evelyn Fox Keller — Refiguring Life: Metaphors of Twentieth-Century BiologyColumbia University Press, 1995.