Language…or Logic?

1500 words, 6 minute read

Few ideas in Intellectual History remain constant across diverse cultures and through successive eras. One exception? The privileged relationship between Language and Logic.

The Old Testament story of the Tower of Babel directly connects our ability to solve physical problems (logic, engineering) with our use of language. In the creation narrative (Genesis 1: 3), “God said, ‘Let there be light’, and there was light.”

Across the Bosporus, the father of Western philosophy, Parmenides (On Nature, 5^th century BCE) also linked speaking with being: “What there is for speaking and knowing, there is for being.” (Fragment 6) . Apparently, physis (being) and logos (language) are two sides of one coin on both sides of the Strait.

500 years later, the Apostle John retold the creation story from the Christian perspective, incorporating material from the Old Testament and from the Pre-Socratics, resulting in an even heavier emphasis on the linguistic component: “In the beginning (arche) was the Word (logos)…through whom all things were made and without whom nothing came to be. In him (logos) was life and that life was the Light of the world.” (John 1: 1 – 4)

Now fast forward a few millennia and see various schools of North Atlantic Analytic Philosophy, from Logical Positivism through Deconstruction, attempting to ‘reduce’ philosophy to linguistics once and for all. Do names like Ayer, Wittgenstein, Austin, Derrida and Chomsky ring any bells?

But it’s all for naught! (Don’t you just hate it when real-physics spoils meta-physics?) A recent study conducted at MIT's McGovern Institute for Brain Research concludes that “language isn't necessary for logical reasoning.” Wow! That’s 2500 years we’ll never get back!

Our ability to induce from experience and deduce from assumptions is neurologically unrelated to our ability to use language. Brain imaging shows that language-processing parts of the brain are not involved in logical reasoning.

In research published in the journal PNAS, and reported by Medical Xpress (July 9, 2026), researchers led by MIT associate professor of brain and cognitive sciences Evelina Fedorenko showed that people can perform well on tasks that require logical reasoning even if their language abilities are severely impaired.

Researchers presented participants with challenges involving geometric shapes, numerical sequences, etc. As participants solved increasingly difficult puzzles, it became clear that people don't need language for this kind of reasoning. Patients with language impairments solved the problems as well as the ‘verbo-typical’ control group.

The MRI scans showed the brain's language system was not engaged for either inductive reasoning (when participants identified hidden rules) or deductive reasoning (when they assessed the validity of syllogistic arguments).

How come? Why don’t the language and logic functions converge, or at least overlap, neurologically? For one thing, language is naturally, and probably necessarily, fuzzy. Wittgenstein (above) spoke of word meanings as familial relationships. We’ve all experienced frustration when talking with someone whose speech is unwaveringly literal.

Logic, on the other hand, is inherently precise. Much blood has been shed, figuratively and literally, because someone applied rigorous logic to an imprecisely defined concept (language). Plus, language is multifunctional. It can ‘compare thee to a summer’s day’, it can talk you through an IKEA assembly maze, and it can inspire soldiers at Agincourt. Logic is, well, more logical.

Finally, most languages are linear in structure, progressing one word at a time, whereas evaluating available information to reach logical conclusions often requires thinking along several tracks at once, connecting those tracks via reciprocity and recursion. The closest thing we have to that in English is the work of James Joyce, especiallyUlysses and Finnegans Wake.

In these ‘novels’ the narrative follows multiple story lines simultaneously: Leopold and Stephen in Dublin, Ulysses and Telemachus in Ithica, King Hamlet’s ghost and the Prince in Denmark, YHWH and his Christ in Roman Catholic Liturgy.

As in polyphonic music, each line maintains its integrity but harmonizes with others. These extraordinary works are to the rest of literature as Beethoven is to Gregorian Chant…or as today’s quantum computers are to their analog ancestors.

In truth, Ulysses and Finnegans Wake are exceptions that prove the rule. They demonstrate both the limitations and the potential of linear language foe modeling non-linear reality. When we approach the whole with language, we necessarily break the entirety into a sequence of stages (so-called causes and effects); we think horizontally.

By contrast, Logic is threadbare; but it is also inherently holistic and recursive. We conceptualize our problem as a whole and then break it down into discrete modules, each modifying and being modified by the whole. We reason vertically. We trust that we can gain information about the whole by understanding its self-similar parts (aka fractals). Because I know that 1 + 1 = 2, I can calculate x + y for any values of x and y.

***

"(This research) really upends a(ny) theory that says that symbolic rule induction is not possible without linguistic capacities." Yes, it does…and it upends more besides. For example, I no longer have any reason to assume that my poor pussy cat, who struggles mightily but unsuccessfully to participate in family conversations, is any less conscious than I am (and perhaps even more so after I’ve had a couple of bone dry martinis).

Fedorenko’s team has provided a major boost to advocates of panpsychism and/or animal rights. On the other hand, the implications for AI are less sanguine. Large language models like ChatGPT and Claude are trained entirely on text and use text as their output—yet they convincingly simulate some kinds of human reasoning (Turing, Searle). My LLM bot, for all its apparent genius, relies exclusively on language to reach logical conclusions. Knowing that, it is hard to assign any logical validity to its output.

Is this the greatest con job in history? Like many a fast talker I’ve known (and one I’ve caught sight of in a mirror), does Claude use language facility to mask logical inadequacy? Does he have a career in politics in mind? Earlier we smirked that physics was overturning metaphysics; could it also be said that rhetoric (LLM) is overturning logic?

Or is it just possible that at the performance extremes achieved by today’s AI, horizonal reasoning (language) genuinely reproduces the results of vertical reasoning (logic). If so, can we treat that confluence as significant or must we consider it mere coincidence?

I am reminded of Aristophanes’ Clouds. In this play, the students in Socrates’ Thinkery are trained to apply logic to nonsense…with predictably hilarious results. Are today’s LLMs extensions of Socrates’ academy? Could it be that Kitty is conscious and Claude is not?

And what of Turing’s Test? And Searle’s Room? These philosophical paradigms draw their power from an assumed relationship between Logic (reasoning) and Language (communication). But what if there is no such relationship?

***

“What if there is no such relationship?” What if? There is none! (assuming we trust the research of course) So now what? What can we say we know? And how do we live?

Imagine a triangle. Label the apex R for Reality. Label the base points L1 and L2 (L1 is Language, L2 Logic). This figure lies, unrecognized of course, at the foundation of all of our conceptions of the World; no more!

R

↙        ↘

L1      ↔      L2

We have concluded (above) that Language and Logic provide irreconcilable accounts of Reality; our triangle is inherently unstable. That leaves us with a choice of three possible models:

(1)    The world of Lewis Carroll (Alice and her Lookingglass): Language but no Logic.

(2)    The world of Jacques Ellul (Propaganda) and George Orwell (1984): Language and Logic but no Reality.

(3)    The world of Samuel Beckett and Eugene Ionesco: Logic and Reality but no Language. (Of course, all three employ language masterfully in their works but it has no agentic force, it does not advance the narrative, its function is subservient to Logic and Reality.)

Hold up! If there are three models of the World that fit the data with equal fidelity, don’t we have to conclude that ALL of them are true? Yes indeed, we do! No single model of the World that includes Logic, Language and Reality can be stable; but no model can be complete without all three.

So, we must accept that we live in a world of wave-particle duality, of quantum complementarity, of Feynman’s Sum over Histories, or Bell’s entanglement. No single model can account for the World as it is but multiple models, applied separately but simultaneously, just might; but you’ll have to stay on your toes to make sure that they never meet. And they say Atlas had a tough job.