Every few months a neuroscience paper appears with headlines announcing that we have finally discovered where language lives in the brain.
This month's contribution (here) reports an impressive technical achievement. By recording the activity of individual neurons in the human frontotemporal cortex during spontaneous conversation, researchers identified neurons whose activity reliably correlates with particular aspects of sentence production. Some neurons preferentially fire before nouns, others at the end of phrases, and still others appear sensitive to broader syntactic structure.
The experimental accomplishment is remarkable. Recording single-neuron activity during natural conversation was unimaginable only a few decades ago. The study represents an extraordinary advance in our ability to observe the neural dynamics accompanying speech.
The interpretation, however, deserves rather more caution.
The accompanying commentary announces that researchers have discovered "specialised linguistic building blocks" within the brain. One investigator is quoted as saying:
"We used to think language was this diffuse, whole-network phenomenon. But it turns out you have specific neurons that only care if a word is a noun."
This is an extraordinarily strong conclusion.
Unfortunately, it is not the conclusion the experiment demonstrates.
The experiment shows that particular neurons participate selectively during linguistic activity. It does not show that nouns themselves exist inside neurons.
That distinction matters.
The study begins with an existing linguistic analysis. Spoken utterances are first segmented into words, phrases, syntactic dependencies, semantic relations, and grammatical categories using contemporary linguistic models. Only then is neuronal activity compared with those pre-established categories.
In other words, the neuroscience depends entirely upon prior linguistic theory.
The neurons are not discovering nouns.
The linguists already have.
The experiment asks whether neural activity systematically covaries with distinctions that linguists have independently identified.
That is a perfectly legitimate scientific question. It is also a rather different question from asking whether nouns are physically located inside the cortex.
The shift from one claim to the other occurs almost invisibly.
Correlation quietly becomes constitution.
This move is familiar throughout contemporary cognitive neuroscience. We are repeatedly told that neurons "encode", "represent", or "store" meanings. These expressions are often useful shorthand. The trouble begins when the shorthand hardens into ontology.
Within a relational ontology, meaning is not an object that can be stored anywhere—not in books, not in computers, and certainly not in neurons.
Meaning is actualised through construal.
The symbolic phenomenon exists only as an instance within a relation between a symbolic potential and a construal that actualises it. Neural activity is undoubtedly one of the biological conditions that makes such actualisation possible. But conditions of possibility should not be confused with the symbolic phenomenon itself.
A neuron no more contains a noun than ink contains a sonnet.
Ink participates in the conditions under which a poem may be actualised. Remove the ink, and the printed poem disappears. But no one concludes that Shakespeare resides chemically inside carbon pigments.
Likewise, the selective firing of neurons before the utterance of a noun demonstrates that those neurons participate in the biological organisation of language production. It does not demonstrate that nouns are biological objects.
The title often given to these discoveries—"how the brain builds a sentence"—contains a similar ambiguity.
Does the brain build a sentence?
Or does the brain participate in the biological actualisation of speech, from which symbolic phenomena subsequently emerge under construal?
The difference is subtle but profound.
From a relational perspective, the sentence is not hidden inside the cortex awaiting expression. What exists is a coordinated biological system capable of actualising symbolic potential. Speech, gesture, sound waves, hearing, and interpretation together constitute the event through which a sentence comes into being as a symbolic phenomenon.
The sentence is not found in the neurons.
Neither is it found in the sound waves.
Nor is it found in the printed marks on a page.
It exists only as an actualised symbolic relation.
Ironically, some of the study's most interesting findings point in precisely this relational direction. The authors describe neurons whose activity depends upon context, whose behaviour changes as sentences unfold, and whose responses reflect evolving grammatical dependencies rather than isolated words.
These are fundamentally relational observations.
The neural organisation appears remarkably sensitive to changing patterns of relation across an unfolding utterance.
Yet the interpretation repeatedly retreats into the comforting language of internal representations and building blocks.
Perhaps that language reflects less what the experiment has discovered than the representational assumptions neuroscience continues to inherit from twentieth-century cognitive theory.
None of this diminishes the scientific achievement.
On the contrary, the experiment reveals something genuinely fascinating: the extraordinary degree to which biological systems become differentiated in support of symbolic activity.
What it does not reveal is where language is stored.
Because language is not the sort of thing that can be stored.
The remarkable neurons identified by the researchers are not miniature grammatical categories hiding inside the cortex. They are biological participants in the actualisation of symbolic potential.
Neurons do not contain nouns.
They participate in the conditions under which nouns become possible.
No comments:
Post a Comment