The science of linguistics

While Professor Cercignani and her colleague Dr Leandro Beltrachini brought technical expertise to the project, researchers at the School of Welsh have been crucial to its success – not only for their knowledge of the language but also their expertise on the science of linguistics.

“Very often when we think of Welsh as a discipline, we think of core humanities elements such as literature, creative writing, history and culture,” says Dr Iwan Wyn Rees, a Senior Lecturer at the School of Welsh.

“But we should also remember there’s a scientific element to Welsh too. I’m not only talking about grammar but also phonology – its sound system.

“There still hasn’t really been enough academic work done on Welsh phonetics and there's never before been a phonetic study on Welsh using MRI technology.

“So, in that sense the project is completely new and a real opportunity to work with colleagues from across the University to focus on the Welsh language.”

Dr Rees and research collaborator Dr Jonathan Morris drew on their networks to recruit participants for the project.

They created a list of words for participants to recite while they lay in the MRI scanner, starting with simple tasks like counting from one to 10, and progressing to more complicated scripts that incorporated regularly mispronounced words and phrases, and the sounds learners find most difficult.

The team included words that are pronounced differently depending on the dialect of the speakers.

A participant from Patagonia, a region of Argentina where there are thought to be as many as five thousand Welsh-language speakers, also took part in the study.

“Kind of like a sci-fi experience”

An Argentinian national himself, and the project’s neuroimaging expert, Dr Beltrachini, said the project posed technical challenges in obtaining both the images and the sounds produced by the speakers.

“MRI is a fairly slow technique, capturing images of the whole brain over a period of a few minutes – far too long to capture the movement associated with pronouncing a word,” he said.

“So, the challenges were initially scanning fast, and making sure the images weren’t distorted.

“We had to test different sequences to check which was most appropriate for capturing the images we needed together with the sounds.”

To overcome this, the team focused their scans on the central section of the head, going down to the bottom of the neck to capture the mouth, the tongue, and the vocal cords.

Dr Beltrachini joined Cardiff University in 2016. He originally trained in engineering, before completing his doctoral training in neuroimaging, which he describes as a convergence of computer science, engineering, physics, and mathematics.

He says the motivation behind his research on the technical aspects of MRI is to get people to think about the same problems in different ways.

“The scanners we use at CUBRIC are very specialised, very technological pieces of equipment, that allow us to get into the function and structure of the brain non-invasively.

“We work with these machines, and all of the complexities and challenges that are inherent to this kind of technology, so that it is optimised for the kinds of MRI research we carry out. This kind of work might be done with the hardware of the machines, data processing, and visualisation too.

“The scanners are more open to us than they are for clinicians in hospital, so we can do this optimisation work. We can therefore come up with new ways of using our scanners, so they can reveal different types of information. We programme them like computers to test our developments, and in some cases, these can find applications in clinics, too.”

The team also had to consider how they would enable participants to read the Welsh-language script and capture the audio without the background sounds generated by the MRI scanners.

Professor Cercignani explains: “When the participants came in, they would lie in the scanner and make themselves comfortable. There is a screen connected to a computer outside, and a system of mirrors which allows them to see the screen behind their head. We projected the Welsh words onto the screen for them to read aloud, and a specially designed microphone captured their voice.

“It’s a bit like when you do an ultrasound of a pregnant woman, and you want to measure the heartbeat of the baby, but the heartbeat of the mum is much stronger. So, you measure the heartbeat of the mum, and subtract it from the global signal, leaving you with the baby’s heartbeat.

“Similarly, we recorded a few seconds of the MRI without the human voice, and then used an algorithm to cancel the noise of the MRI from the recording.

“This enabled us to capture both images and sound from which we are able to make the audio and video resources.”

With so many elements to the experiment, the team worried participants could get overwhelmed.

Dr Beltrachini recalls: “It’s not always easy to find people to participate in our studies. And although people are excited to go into a scanner, when the time comes, it can be a little bit scary. Luckily, all our participants were up to the challenge, and some of them even left saying it was kind of like a sci-fi experience.”

Increasing use and confidence in the language

The team used existing MRI technology, but hopes to look into the possibility of developing a faster acquisition sequence that still provides the same sort of anatomical detail.

They say this might enable them, in the future, to scan participants reading whole sentences and even conversations, rather than just isolated words.

The resources were recently tested and benchmarked by tutors and learners in workshops organised by the National Centre for Learning Welsh, and were delivered on-line and on location in various parts of Wales by Dr Rees and Dr Morris.

“It is really important to take advantage of both research and technology to enhance learners’ experiences,” said Helen Prosser, the Director of Learning and Teaching at the National Centre for Learning Welsh.

“The team held a virtual session for learners, which was very well attended, and where learners felt that seeing the graphics could help them with their pronunciation.

“We will be signposting our learners who have trouble pronouncing the Welsh language in the new academic year.”

The team also hopes their work has a role to play in achieving the Welsh Government’s ambitions to inspire a million people to enjoy speaking and using Welsh by 2050.

“I think the Welsh Government captured the imagination of a lot of people with the target of one million Welsh language speakers by 2050,” says Dr Rees.

“But if you look at the full Cymraeg 2050 strategy, the second target, which is to increase use and confidence in the language, is arguably just as important – if not more so.

“Investing in resources like ours that support this second ambition is therefore absolutely essential.”