Seven thousand ways to say hello
Dr Paul Ibbotson tells Shaoni Bhattacharya about the linguistic superpowers of the average five-year-old, and what AI can learn from them.
06 September 2023
Why did you write a book on how children acquire language?
I wrote this book to celebrate an extraordinary journey that children take. In a few short years, they go from no language to babbling, to producing their first words, to stringing words together into sentences, to holding conversations and telling stories. Despite 7,000 different possible languages to learn, each with its own complex system of sounds, grammar and meaning, children seem equally able to learn any one of them and have mastered many of their intricacies by their fifth birthday. I wanted to share how they do it and what a remarkable part of our human nature this is.
Why is the human capability for language so special and how does it tie in with our evolution?
These are still open questions. Some people have argued that language requires heritable specialised cognition, located in a discrete faculty of the mind. Others see it as a by-product of a broader adaptation to uniquely human cultural and social life. Either way, language recycles many other parts of our cognition – things like categorisation, executive functions, and intention-reading – that were designed for different purposes, both in child language acquisition and over the course of our evolutionary history. We are currently researching how deep this recycling goes and how it might help children learn language.
How do our linguistic abilities compare with that of animals? Your example of Whisky the border collie who could understand 59 English object words sticks in my mind.
A sideways look at what other animals can and can't do helps us work out what is unique about our language and its place in the natural world. In general, it seems that the communicative systems of other animals are much more context-bound and use a more fixed repertoire of symbols compared with the variety of forms and meanings humans use. Human language also appears to be underwritten by a richer social inferential system that helps us work out the communicative intentions of others.
Your book highlights how at 35 weeks after conception, foetuses can recognise different vowel sounds while in utero, and that they also start learning prosody (the linguistic patterns of stress, intonation and rhythm). Can you tell us more about this – do psychologists know when and how language acquisition starts?
We know babies are becoming sensitised to the world of sound while they are still in the womb, and this includes a preference for frequently heard voices and their native language(s). Following birth, infants continue this process of homing in on the language(s) they are exposed to, at the expense of those that they are not, as part of a broader orientation to their cultural surroundings.
How does language learning relate to the social-economic context that children grow up in? You highlighted the '30 million word gap' debate – what is this and does it hold relevance today?
Research has estimated that children from families with low Social Economic Status (SES) hear about 13 million words by age four whereas their peers from families with higher SES hear about 45 million words, 'a gap' of about 30 million words. The basic premise – that SES influences language development, which influences educational achievement and related outcomes – has been important in educational policy, prompted intervention trials and continues to feature in wider public debate, but the magnitude and even existence of this gap remain controversial. The debate shows how the wider socioeconomic context can create individual pathways in language acquisition.
It's clear from your book how pivotal the first 2,000 days in a child's life are for acquiring language. What hope is there for older people wanting to learn a new language in adulthood?
There is hope! But there are obvious differences between adults and children. It's partly a numbers game, for example, English-learning children are exposed to something like 7,000 utterances a day, embedded in meaningful communicative interactions that involve sharing, requesting and informing. Few adult language-learning contexts can match that intensity or immediacy. It is also a qualitatively different experience for adults; for example, adults' knowledge of their first-acquired language can impede learning multiword chunks of a new one. The biggest difference is arguably that adults often learn language in educational contexts whereas children construct language for themselves with no formal instruction.
You mention that having a 'deep-and-meaningful chat with an android is some way off in the future'. Why is this, despite our advances in voice-recognition software and AI, such as ChatGPT?
This area is moving very fast, and there is no reason why AI won't match human language abilities, perhaps sooner than we think. The challenge for AI is that language – particularly speech, dialogue and conversation – is a messy business. There are different meanings to what someone says; there are different ways for them to say what they mean; different accents and tones of voice in which to say it; and speech is full of incomplete sentences, overlaps, restarts, abrupt topic shifts, and non-sequiturs.
It is still easier for a computer to calculate pi to a billion decimal places than it is for it to understand a joke, and that difference between AI and us tells us something interesting about the way humans work. Also, the best current AI language generators typically use a text interface, which sidesteps a major challenge that children solve for themselves, namely figuring out where the words are in a stream of speech.
There are at least 7,000 spoken human languages, and 2-300 signed languages in the present day: when you were researching this book, was there one that you found particularly interesting and why?
They each have their interesting qualities and in the book, I try to celebrate the diversity of forms languages can take, for example, Rotokas, spoken in Papua New Guinea has as few as six consonants while !Xóõ, with modern-day speakers in Botswana, has around 77. Despite being able to use examples from 73 languages in the book, this still only represents about one per cent of all languages in use today. If we want to understand how learning is the same and how it is different for children around the world, then we need to study a much broader range of languages.
What languages do you speak, and what drew you to this area of psychology?
I speak English, and fragments other languages, which I typically use in my writings to illustrate a particular aspect of how language works. I was attracted to this area because there are interesting (and hard) problems to solve, and it gives me the opportunity to draw on ideas from other areas of psychology, anthropology, and evolution.
Who are your ideal readers for this book, and what do you hope they will take away from it?
The book is aimed at anyone with an interest in language and how we acquire it. I hope readers will take away the same sense of wonder that I have about what children achieve and I hope it nurtures an interest to explore the topic further.
Language Acquisition: The Basics is published by Routledge.
Dr Paul Ibbotson is a senior lecturer in developmental psychology at The Open University.