How do nonsense words create meaning without referents?

Lewis Carroll's "Jabberwocky" makes sense to most readers despite containing words ("brillig," "slithy," "toves") that have no referents. The reader knows what kind of sentence is being uttered, what mood is being established, what kind of action is happening. The meaning is real even though the words denote nothing. Compositional theories of meaning struggle with this — if the words have no meanings, the sentence cannot have meaning by composition.

Deleuze's Logic of Sense (1969) takes Carroll seriously as a philosopher of meaning. His central distinction: sense-of-nonsense vs nonsense-of-sense. Sense-of-nonsense is what Jabberwocky achieves — material that lacks referential meaning but produces sense through frame-activation. The reader hears the rhythm of an English sentence about something fierce being slain, hears the prosody of a heroic-tale frame, hears the morphology of monster-words, and the sense follows from frame-activation alone. Nonsense-of-sense is the opposite — material that has perfect referential meaning but fails to make sense because no frame coheres ("colorless green ideas sleep furiously" in Chomsky's example).

This bears on the broader theoretical claim that How do readers actually build meaning from words?. Jabberwocky is the limit case: the word-meanings are zero, the frame-activation is sufficient, and meaning emerges. The example proves that meaning-making does not require referential content; it requires successful frame-activation. The compositional path is not the only path to meaning.

For AI, Jabberwocky is a test of whether the system can engage in the meaning-making operation human readers perform on it. AI can recognize that Jabberwocky is poetry, can identify it as Carroll, can analyze its meter — these are content operations. The harder question is whether AI experiences the sense-through-nonsense the way human readers do, or whether AI stops at the level of "these words have no referent" and treats the poem as a string failure. The frame-activation operation Does the mind selectively activate frames from only some words? enables the human experience; whether AI does anything analogous is an empirical question with theoretical stakes.

The strongest counterargument: AI trained on Jabberwocky and discussions of it will produce competent commentary on it, which is the only operational test. True for output, but the question is internal: is the sense-of-nonsense detected as sense, or only described as something humans report as sense? The distinction matters because it determines whether AI participates in meaning-making or only describes it from outside.

Source: Making Sense - brief for co-authored essay on language