Hugh Foley
Bagel City
.
There are days when I feel that language was a mistake. Nights, most keenly. My daughter, a few months shy of two, wakes up at 3AM to shout, Some day / I’ll be / living in a bagel city. I know she is saying bagel rather than ‘big old’, the correct lyric to the Taylor Swift song (‘Mean’) which someone must have sung to her, because she then says you want bagel, by which she means I want a bagel. But what she means when she almost sings is more complicated than that, and it keeps me up, even after she settles.
A few years ago, I had a job teaching a university literary theory module; and foolishly I set an essay by the philosopher Donald Davidson, ‘A Nice Derangement of Epitaphs’, which claims that ‘language does not exist’. The normal picture of language, Davidson argues, is one in which we work out what someone is saying by first appealing to a conventional usage of the word. When someone departs from those conventions – by being sarcastic, for example, or saying Hugh speaks clearly in lectures, when asked if I am any good at my job – we work out the ‘speaker meaning’ by understanding how what they are implying modifies the conventional, or ‘sentence’ meaning.
And yet when Mike Tyson says, after losing to Lennox Lewis, that he will now ‘fade into Bolivian’, we can understand without any reference to the conventional meaning of ‘Bolivian’. Here, we have to treat the ‘speaker meaning’, what one wants to say, as primary. Davidson thinks such special cases disprove the general one – we are always adjusting our theories of what someone means, trying to figure each other out. If there is a rule, it is not linguistic, it is interpersonal. It’s something like: we believe that others can understand us. The students didn’t really get it.
I find myself thinking of this problem at three in the morning, as I receive my instructions: Daddy lie down on floor. If my daughter sings bagel city, what does she mean? Is her intention to say the word ‘bagel’, a word I know she ‘knows’, or is it to imitate the song, and therefore to say big old? When she sings the second part of the chorus (all you’re ever gonna be is mean) it just sounds like Ogga-be-me. On the one hand, this seems like an ideal interpersonal act of translation. My job, when talking to my daughter, is to guess what she means, her job is to guess what I mean. We believe things about each other. But how do we have a concept of meaning before we have a whole language? When does an infant have a meaningful sense of meaning?
I am soon singing with her in order to return her to sleep. She shouts the words back at me. Eventually her eyes smooth over in a way that I can only compare to a lake after the breeze has subsided, and then she goes to sleep.
*
Charles Darwin had a theory about the origins of language in song. Our ancestors must have wooed each other with singing; thanks to the forces of sexual selection, the capacity to make beautiful music resulted in a throat and brain that could also handle words. It’s as if peacocks, which evolved their elaborate, impractical feathers purely for mate-luring purposes, then developed flight, leaving all the rest of the birds on the ground.
I like this theory, of what scientists call a ‘costly signal’ becoming a whole sign system – so do a number of cognitive linguists, who call it the ‘musical protolanguage’ thesis. On the brain scans of songbirds and humans, there is a matching pattern, a ‘white-matter fibre tract’ between the vocal motor cortex and the brain stem. Macaques and unmusical birds don’t have this line, but humans have adapted their brains to song.
The brains of babies, however, don’t have this fully developed connection. Subsequent investigations have suggested that this ‘white-matter fibre tract’ is as likely to be a consequence of language as a cause. Genetic testing has found that speech capacity has something to do with a gene called FOXP2 – humans without the standard human form of this gene cannot speak grammatically – but the mutation by itself does not explain language.
Instead of studying how we can speak, many people think we need a story of our wider capacities to understand the process of meaning-making. This has to involve children, and how we usher them into the world of language.
Biologists used to tell themselves a general story: ontogeny recapitulates phylogeny; how a trait develops in an individual reflects how it developed in our species. A physiological example – the human larynx descends in early childhood, allowing us to speak more easily, and the same is true in our species’ descent from earlier hominids. Their larynxes are higher in the throat. To understand language and its arrival, some think, we can study how it arrives in an individual. How my daughter learns to speak might tell us about how humans first meaningfully spoke.
What needs to exist for language, beyond the cognitive or physical capacities, is a spirit of trust. At the heart of accounts favoured by psychologists is the development of a theory of other people. This seems to begin with what they call ‘joint attention’, looking at something together, and realising that we are together. Pointing precedes language in children, and looking at the same thing as your mother precedes that. My daughter’s first word was ‘this’, which, after I started reading all this stuff, seemed almost too on the nose.
I learned about this need for shared attention reading on my phone, while my daughter played. I put down the phone guiltily, to find her looking at me, or possibly the phone. I felt shame for my neglect, prioritising abstract understanding over the person she was, right there with me. Shit, I said. Shit, she said back. Shit shit shit, running around gleefully.
From joint attention, then, comes ‘theory of mind’, the ability to assign intentions to others. In the late 1990s, the psychologist Amanda Woodward began to study this theory’s emergence in babies. By examining how long infants looked at objects that adults moved – longer looking being taken to indicate surprise – she concluded that they attributed predictable intentions to humans from at least six months. If a hand that had been seen to move a set of toys in sequence changed the sequence, the babies looked longer, in surprise. If the hand was gloved, they did not.
Underlying our language-learning is our willingness to believe in other people’s attention. Again, this is sometimes threaded into an evolutionary origin story. Because human infants are helpless for longer, they need to trust more. And perhaps they need to trust a wider kinship network than other animals do, which might lay the groundwork for what philosophers call a form of life. This is a sort of boundary line, within which thinking or acting make sense. We are social animals of a particular kind because we share intentions. This is our species boundary.
To go back to Donald Davidson and my daughter; we grow up in a world where we are tasked with being social. So we work out what we mean. This dyadic exchange is really an expression of a more general being together. Often when my daughter would point things out to me she would form questions. Is it big? Is it a blossom? Is it yummy? This was because, I think, she was imitating how I had spoken to her. I read her as saying that lunch was indeed yummy, but that wasn’t exactly her meaning. She was playing a game with me of attending to things, trying to get it right. In that sense, to speak is to sound out the prehistory of trust.
But that also seems to me to be where the problems begin. Some people worry about their children being poisoned by microplastics or mould spores. Some worry about contamination from 5G or vaccines. Some worry about their little ones picking up needles. I worry about her picking up words.
Words might emerge from love, from trust, from wonderful things, but they have specific designs of their own. As I bring my daughter into the world, the human world of language, I sometimes worry I am training her to produce a kind of air pollution. We go on a train journey together, to visit my mother. See it. Say it. Sorted. says the automated message about calling the police to snitch. See it, say it, sordid, she says. I worry that I am binding her not just into a trust between us, but into a form of life that took advantage of this trust and ruined it.
*
Words were first created to feast on the dead. At least, the linguist Derek Bickerton’s ‘confrontational scavenger’ theory suggests that hominins first made articulate sounds to coordinate the eating of large carcasses. Scavenging, which requires talking about a body that was not present, encouraged communication about absent things. To find and butcher carcasses in time (after a mammoth’s tough skin had softened but before the meat had rotted) required words, sounds or gestures that could be used to refer to things that aren’t there. That’s where human and early hominid infant trust comes in. Chimpanzees, by comparison, only produce specific calls in the presence of the things they refer to. They don’t trust each other enough to co-ordinate actions on the scale needed. Eventually, because these scavenging, word-using hominins thrived on meat, the ability to articulate created a selective pressure that drifted finally into language.
By contrast, the linguist Noam Chomsky thinks language was a miracle. A mutation, he doesn’t speculate how, must have given humans a strange gift. For him, it is ‘something like a snowflake, assuming its particular form by virtue of laws of nature – in this case principles of computational efficiency.’ By this he means that we developed the ability to perform an operation he calls ‘Merge’. This is where we create sets out of concepts, which then can be endlessly combined with other concepts. In language this takes the form of things like noun phrases. ‘The stupid song’ gathers together ‘the’, ‘stupid’ and ‘song’ into a unit, a set governed by ‘song’, which I can then ‘recursively’ add into larger units; the stupid song that I sing to my daughter that she sings back to me that I sing to her that she then sings to me once again is now over.
With this power over what Chomsky calls ‘digital infinity’, the first mutant human began talking not to others, but to himself or herself. Because they could now plan in advance, etc., they thrived reproductively, and passed on the snowflake of symbolic thought, which evolved into words.
Robert Planer and Kim Sterelny, meanwhile, think that the evolution of language had no distinct revolutionary phase. Hierarchy itself is necessary for the mapping of social worlds; you can see it in other apes. Baboons use their barks in specific ways depending on where they are in a dominance hierarchy. But syntax, in the hierarchical sense of set-making that Chomsky means, becomes infinite over the scale of evolutionary time to abet composite stone tool development. The demands of making a particularly complex hand-axe, say, have to be broken down into steps, which are nevertheless part of the general set, ‘make an axe’. Language is weapons-grade thinking.
Chomsky’s, Bickerton’s, and Planer and Sterelny’s theories differ dramatically in their basic sense of how language came about. They are not the only ones, by any means. But they all (rightly) gesture to the kind of evolutionary advantage that language conferred on humans by letting us think things that were not present. In thought, we can make anything into the same kind of thing: concepts, which can be corralled into increasingly complicated forms. In a way, they all owe something to one of the first modern arguments about language evolution, that of Etienne Bonnot, Abbé de Condillac, who said that humans invented words to give ourselves ‘empire over our imagination’.
As a story, I prefer the drift of trust into representation over the freak mutation (the evidence also seems to favour it). There’s a bittersweet character, a fortunate fall, to the narrative: words are dependent on our sense that people want to mean things, but they come to be ways of thinking about the world’s objects in the absence of their own distinctive meaningfulness.
What do I mean by this? I mean that when we begin to manipulate concepts, and develop our ideas, we treat them as things that do not have an independent intentionality, or the interest of our mother, but as members of a set. ‘The gazelles that feed at this pond’ are a set; even though the words each add specificity, the set is drained of what matters to its members. These gazelles are thinner than the real animals, eyeing us suspiciously. To think ourselves into being apex predators, we have to lose some of our love of the world. We narrow our sense of what is meaningful, of what is alive.
To be clear, I have no coherent objection to hunting, carcass harvesting or symbolic thought. But something about this need for absence at the heart of language strikes me as presaging some unfortunate later developments. If we take trust theories of human language and the love of mothers and others for their children as true, they nevertheless seem to evolve into faculties of exploitation and domination, a kind of unloving way of being in the world. The linguist W Tecumseh Fitch refers to the cognitive capacities by which humans develop language as ‘Dendrophilia’ (love of trees) – the creation of hierarchical, tree-like structures in our concepts. That’s how linguistic recursion works. But dendrophilia, we might say, seems also to lead to the kinds of thinking that destroy the Amazon rainforest.
*
Two months after my daughter was born, in November 2022, Open AI released ChatGPT. This Large Language Model is an attempt to generate language in ways that mirror human communication. The company claims to hope, also, to use this as a path to a more general reasoning capacity, artificial intelligence. The probabalistic model they work with is incompatible with the model of symbolic thought in language proposed by Chomsky, but the linguist and the tech companies share a stance towards language as a tool of computation that, I think, is as importantly similar as it is different.
If thought is a way of processing information, if we evolved a capacity that captures some true information substrate of existence – just as we evolved cells that capture light and that became the eye – then, just as we invented the camera, we can invent genuine reasoning machines. As a result of believing in this description of thought, many of the people involved in making these LLMs claim to think it would be desirable to create a disembodied superintelligence that had no ties to humanity beyond a shared computational function. They talk about its coming birth with a strange, religious fervour.
In autumn 2022, I also became unemployed, ending my previous career in literary studies. I took up various forms of entry-level content work just as this technology hit the market. It struck me that I was being replaced, not by an AI deity, but by a shoddy tool. One day, working in a café, I watched several undergraduate students use ChatGPT to write their essays. Even if the bots weren’t very good, they seemed to be creeping in on my world in all directions. I was being replaced by the words themselves.
There, too, was my daughter, also my successor. And as she lay on the floor, in our care, exhausting us and terrifying us and awing us, her presence seemed to me a thing beyond words. But words came soon. First ‘this’, then ‘banana’, and then I began to lose track of the objects she could name. Sometimes I worried that words would come between her and the aliveness of the world, that everything she was learning brought her closer to a purer kind of information processing, into the human world of exploitation and domination, the world that was actually the world not of humans but of machines. And yet, watching her do things – even things that seemed like computation – with words, remained a revelation. Whenever she sits on a bench, she says adventure, because the two words sound similar. When she saw Battersea Bridge, she said it was the Eiffel Tower, when she saw an eagle, she told her mother that it had orange polka dots in its eyes, like her dress. She was constructing a world with concepts – she was doing things with trees – but her dendrophilia was not just for manipulating the objects of the world. It was for us.
Whether or not speech comes from song, I like the idea of some vestigial, phylogenetic element in all words that speaks against the reduction of things to concepts. Like language, music is a ‘digital infinity’, but its meaning is simply in its being experienced. Meaning is found, too, in the attention to the surface shapes of the words, of the shapes they make in the mouth. The shapes we make for others. It reminds us not of how we might use the object we are describing, but how it resonated inside us. Everything has at some point been the medium through which a child and a parent have made meaning. Songs can remind us of this condition. When my daughter sings, though she has no ear for music yet, only words, she brings me closer to the world where meaning and being mean and are the same thing.
.
.
Hugh Foley’s poetry and criticism have appeared in Poetry Review, The White Review, Poetry London, PN Review and The Rialto, among other places. He is the author of an academic work on American poetry (Lyric and Liberalism in the Age of American Empire) and several study guides for children. He writes a Substack newsletter, Useless Concentration.
To read this and more, buy our latest print issue here, or subscribe to receive a copy of The London Magazine to your door every two months, while also enjoying full access to our extensive digital archive of essays, literary journalism, fiction and poetry.