English French German Suite & Topic XII — Minds and Thoughts

Core idea

The English French German Suite dialogue presents Lewis Carroll's Jabberwocky poem alongside its French translation by Frank L. Warrin and its German translation by Robert Scott. The three versions share structure — the nonsense words ("brillig," "slithy," "vorpal") have analogues in French ("brilgue," "lubricilleux") and German ("brillig," "schlichten") that preserve the feel of the originals. But what exactly is being preserved? The literal words are different, the rhythm survives, some onomatopoetic effects survive, others do not. Translation reveals that meaning is not a single thing: it has layers, and different layers translate with different fidelity.

Topic XII applies this lesson to minds. When two people communicate, they are translating between their symbol systems. Two brains never instantiate identical symbols — your "kitchen" symbol and mine are differently embedded in different memories, different aesthetic preferences, different associations. Communication works because the structural roles of our symbols are similar enough to enable mutual prediction. But complete understanding between minds is, Hofstadter argues, structurally impossible — and he uses Cantor's diagonal argument to show that even at the level of pure mathematics, no procedure can capture every reasoning pattern of every mind.

Hofstadter's argument: Two systems can be deeply isomorphic without being identical. Communication, translation, and shared understanding work to the extent that the relevant structural roles align — and they fail to the extent that the surface details do not. The dream of complete understanding between minds is the same dream as the dream of a complete formal system: both crash on the rocks of irreducible structural difference.

Why it matters

Layers of translatability

The Suite's three Jabberwocky versions translate at different layers:

• Sound and rhythm: the iambic tetrameter survives. "Twas brillig, and the slithy toves" / "Il brilgue: les toves lubricilleux" — same meter, same rhyme positions.
• Texture of nonsense: in each language the nonsense words feel like coherent words of that language. "Brillig" feels Anglo-Saxon; "brilgue" feels French; "schlichten" feels German. The translators worked hard to produce nonsense that obeys the host language's phonotactic rules.
• Onomatopoeia: "snicker-snack" (the sound of the vorpal blade) has been replaced in French and German by language-appropriate equivalents that approximate the same affective payload.
• Specific etymological jokes: Carroll's "slithy" combines "slimy" and "lithe." French and German require new portmanteaus — they cannot translate the joke per se, only the structural move (a portmanteau).

The deepest layer is the rule by which Carroll constructed the poem — a procedure that the translators infer and reapply in their own languages. This is the layer Hofstadter cares about. Translation succeeds when the receiver can recover the rule; the surface details are the rule's local instantiations.

Communication as structural translation

The topic argues that all communication between minds works this way. When you describe a vacation, the listener does not copy your kitchen-symbol or your beach-symbol into their head. Their symbols, similar in structural role to yours, get triggered by your words and recombined to produce a representation that is structurally analogous to your memory, not identical. Two people who have read the same book have different brain-representations of the book that play similar functional roles.

Two consequences. First: communication is lossy. Not in the trivial sense of bandwidth but in the structural sense that the receiver's representation cannot be exactly the sender's. Second: communication is possible because the structural roles of symbols across human brains overlap deeply — we share a species, a developmental program, an environment, much vocabulary, many memes. The richer the cultural overlap, the higher the translation fidelity.

Cantor's diagonal and the limits of formalization

The topic introduces Cantor's diagonal argument, which Hofstadter uses to make the case for irreducible structural difference at the level of mathematics. Cantor showed that the real numbers are uncountable: any list of real numbers necessarily omits some real (you can construct an omitted real by taking each listed real's n-th digit, flipping it, and assembling the flipped digits into a new real that differs from every listed real in at least one digit).

Hofstadter uses the diagonal to show that no fixed formal system can capture every reasoning pattern of every mind. Suppose someone proposed: "here is the formal system that captures human mathematical reasoning." Whatever the system, Gödel's construction provides a sentence the system cannot prove that humans can see is true (this is Lucas's argument — Hofstadter rehearses it sympathetically while also undermining it). The system would need to be extended; but the extended system is again incomplete; and so on. The mind seems to escape any single formalization the way the real numbers escape any single list.

The catch — Hofstadter is careful to note — is that no one mind can outrun every formal system either. The mind's superiority is over any one fixed system, not over the limit of all systems. The relationship is closer to "creative diagonalization" than "magical insight." The mind has the same mechanism of meta-level reasoning that lets us step outside any given formal system; that mechanism is the strange loop.

What can be shared between minds

The topic ends by asking: if minds cannot be identical, what can they share?

• Active symbols whose structural roles align. Two people can both understand "free will" without their internal representations being the same — they need only play similar roles in similar reasoning patterns.
• Procedures and methods. Mathematics is shareable not because we share the same mental images but because we share derivations.
• External artifacts. Books, recordings, code. The artifact carries enough structure that minds with different internal representations can converge on a shared interpretation.
• The reasoning patterns themselves. Logical inference, analogy, decomposition into cases.

What cannot be shared, Hofstadter argues, is the exact subjective texture of a particular thought. Your experience of red and mine differ in unprovable ways. But the structural role of "red" — what other symbols it primes, what behaviors it elicits, what categories it belongs to — is shareable, and that is enough for the work of communication.

Key takeaways

Mental model

Copy sourceZoom

Practical application

The topic trains a discipline of locating the right layer of translation for any cross-domain transfer.

1. Identify the surface, the structure, and the construction rule. Most failed translations preserve the surface and lose the structure. A literal translation of Jabberwocky would not be the poem. A literal port of architecture from one language to another is rarely the right design.

2. Translate the structure, recreate the surface. Good translators across language preserve the structural role (the texture of nonsense, the meter, the affective shape) while recreating the surface in the new medium. Good idiomatic ports do the same in code: a Pythonic version of a Lisp algorithm rewrites the surface to match Python's idioms while preserving the algorithm's structure.

3. When the structural role does not exist in the target, recreate it. Some surface details have no analogue in the target. Carroll's onomatopoeic "snicker-snack" needs a French equivalent invented for the French translation. In code, monadic patterns in Haskell often need to be reimagined as middleware in Express, decorators in Python, or context managers — same role, different shape.

4. Accept the residual loss. Some details cannot transfer. The exact prosodic feel of "Twas brillig" cannot be replicated in any other language. The exact subjective experience of your insight cannot be transferred to a colleague. Stop trying to do the impossible and concentrate on what can be shared — the structural role and the construction rule.

Example

Consider porting a software pattern between programming paradigms. A skilled object-oriented programmer is asked to write the same logic in Haskell.

A naive port translates surface tokens: classes become records, methods become functions taking the record as the first argument. The result usually feels wrong in Haskell — wordy, ceremonious, missing the language's strengths.

A structural port asks: what role does each class play, and what is the corresponding role in Haskell? An immutable value class becomes a Haskell record. A class hierarchy used for polymorphism becomes a type class. A class used as a namespace becomes a module. A class managing internal state becomes a monad or an IORef. The structural roles map; the surface details are reconstructed for the target.

The deepest layer is the construction rule — the design intent. "We wanted to encapsulate database access behind an interface that can be swapped for tests" maps to "we will use a class with a sum-typed interface and a parameter that selects an implementation," which in Haskell becomes "we will use a type class with several instances and have functions accept the implementation as a parameter." The construction rule survives both translations. The surface is wholly different.

The same lesson applies to teaching. A professor with a deep grasp of a topic does not translate their own internal representation literally to students — they identify which structural roles the students' existing symbols can play, which need to be built up, and which construction rules can be invoked. Bad teaching is literal transcription of expert understanding; good teaching is structural translation tuned to the receiver's existing symbol system.

Related lessons