Super-voice-recognisers

Sana Qadar: To kick off the show today, a test. Get ready because I'm going to play you a few clips. Have a think about whether you can recognise these voices:

Audio of different voices:

If all of that money that had been put away in super had just been put into the housing market…

The sort of chaos I used to invite into my life…

And I was just so startled, I was, like, what, like it just didn't compute to me for a minute.

And this is an issue that we are raising today is the issue, frankly, of being overweight.

And then one day you are sitting there with your friend doing yoga and you go, oh, I get it.

Sana Qadar: How did you do? It depends on how much news, TV and music you listen to of course, but even then, some people are much better at identifying voices than others. All in the Mind producer James Bullen has been looking into this and he's here with me now. Hey James.

James Bullen: Hey Sana. How'd you do on that one?

Sana Qadar: I think I've got most of them figured out, but there's one or two I'm not totally sure about.

James Bullen: Yes, I think there might have been a politician or two lurking in there, and we are going to reveal the answers to that quiz at the end of the episode.

Sana Qadar: Nice. I'll wait until then to find out if I was right. So, what do we know about this voice recognition ability?

James Bullen: We know it's a spectrum that most people will be decently good at remembering voices, but it's a bell curve, so there's people who are excellent at the top and people who are not so good at the bottom, and it's starting to emerge that this could be linked to another ability that some people have, which is a terrific memory for faces. Keen listeners of All in the Mind might remember that Lynne Malcolm and Genelle Weuledid an episode about this back in 2018, and I want to take you back for a quick visit to the University of New South Wales Face Research Lab because they've made some new findings, and because understanding how people recognise faces so well could be very important in working out whether there are people with exceptional voice recognition too.

James Dunn: A super-recogniser is someone who is naturally gifted at face recognition. They can remember faces more accurately than the average person, and we believe they are in the top percentile of people who do this naturally, yes.

James Bullen: Dr James Dunn is a postdoctoral research associate with the University of New South Wales School of psychology, and his research focuses on super-recognisers.

James Dunn: The research tells us that face recognition is largely based on your genes, so it looks that you are as good as you are at face recognition because you are born that way. So we see quite a lot of high heritability between twins who both have face recognition ability, and we believe this is also true for super-recognisers. And so when we speak to them they do seem to have parents or siblings who are also super-recognisers as well. So it's seen to be largely genetic based.

James Bullen: Do we have an understanding at all of what's going on in the brain and what's happening?

James Dunn: So we know that face recognition is related to a particular area of the brain known as the fusiform face area, and it's located just above the right ear. What we believe we will see in super-recognisers is that this area may be more active or more densely populated with neurons, but we don't know that for sure, it's still early days.

So we also think this area is related to development of visual recognition and expertise. So we see people who, for example, become bird experts, identifying birds. This area also tends to light up as they are recognising and distinguishing between different types of birds. Similarly car enthusiasts when they are looking at images of cars and trying to work out what they are, that area also lights up. So it's seen to be related to discriminating things that we are experts in.

James Bullen: There have been a few different ways to test for super-recognition ability. The original tests were actually designed to pick up prosopagnosia, or face blindness. That's where you can't tell faces apart, sometimes even members of your own family. Someone with prosopagnosia might score 50% on one of these tests, but most people in the typical range of ability still make a bunch of errors. They might score 70%, 80%. Not so super-recognisers.

James Dunn: So, although these tests were designed to discriminate if people have disorders, they also distributed the natural variability in people's ability as well, so people would do the tests and some would score really highly and some would do quite poorly. We would expect a super-recogniser to score maybe 90% to 100% on those types of tests.

James Bullen: Because a whole heap of super-recognisers will get a perfect score, the researchers needed a way to tease them out further.

James Dunn: One of the reasons why we decided to make the UNSW face test was to make a test that really challenged super-recognisers, because it was quite common for them to score really highly, in some cases 100% on those previous tests, and I think we have succeeded in doing that because average performance on our test is just 60%, while super-recognisers score as much as 70%, there's still a lot of room to improve on that. And one of the reasons why our test is so difficult and what super-recognisers are doing really well is recognising the face despite a large variability in performance. We are using naturalistic images taken from people's social media profiles, and they are recognising those people despite years and differences between those photographs. So when we say if these people are good, they are really good at recognising faces. So now we can say even though all super-recognisers are doing really well scoring above 70%, there's now a big difference between someone who scores 90% and someone who scores with 75%. So we are thinking now we have a way to find quite possibly a superduper recogniser.

James Bullen: So before we go on, I have a confession to make.

Sana Qadar: What is it?

James Bullen: I have a sneaking suspicion that I am a super-recogniser.

Sana Qadar: How do you know that? Have you taken one of these tests?

James Bullen: A couple of years back I did one of these tests. It's a bit hazy, I can't quite remember, but I think I got a pretty decent score, and I've always had a good memory for faces. But I wanted to test it out for real, and to find out more about these superduper recognisers that Dr James Dunn just mentioned, so I got him to test me out.

Okay, so the tests are starting, there are three main parts. I'm beginning the first part which is the memory learning phase. I am shown 20 faces, one after another and I have to memorise those faces.

So this is the UNSW face test.

Okay, so I just saw 20 faces, it felt like a lot of faces, I'm not so confident in my ability to remember them all. But now we have to decide whether the faces that I see coming up were some of those that I've just been shown, and they are going to be looking at different to how they looked previously…

There are different parts of the test. In one section I had to memorise and recall faces, in another I had to decide whether faces matched or whether they were different people. The whole thing felt pretty tricky. But then came the final result.

All right, let's submit. Like this?

James Dunn: Yep.

James Bullen: Oh, okay, I got a score of 81%. How's that?

James Dunn: 81% is phenomenal. So as I mentioned before, the top 5% of people score over 70% on the test, so you are well above that. Based on this test, yes, you are definitely in the ballpark of a super-recogniser or one of the top super-recognisers on this test.

James Bullen: Does this mean that I can hang up the headphones and get a job with Customs?

James Dunn: Unfortunately not, I think you should keep your job working for ABC. Unfortunately we don't know anyone hiring super-recognisers at this point, although there is a lot of interest in the research, no one out there is hiring people specifically because they are a super-recogniser right now.

James Bullen: And, full disclosure, I should say I have done this test, a number of years ago now, but some of the faces did feel a little bit familiar to me. Could that have influenced the results?

James Dunn: Yes, we do see when people do the test again, they do get a bit better, especially those people who are already naturally good because people who are bad aren't going to remember any of the faces they've seen. So we've seen people might get 1% to 2% better on this test on a resit, but it's not guaranteed, so it's still a fantastic result.

James Bullen: Feels good, but I'm still no superduper recogniser. Late last year, the UNSW team released their findings after testing more than 25,000 people.

James Dunn: One of the people we've managed to find from that population was a young woman from Melbourne who scored the highest we've ever seen on that test, and when we brought her into the lab she also scored the highest we've ever seen on quite a number of different tests as well. She was quite exceptional at recognising faces, one in a million basically.

James Bullen: The woman scored 97%, and there were 11 more people (that's 11 out of 25,000) who scored somewhere in the 90s.

And with this group of potential superduper recognisers, the cream of the crop, are you doing further research with them in particular?

James Dunn: So one of the reasons why we are interested in super-recognisers is they perform a particular group of natural expertise. In similar ways that other people become experts, as I mentioned, in identifying birds or cars, we know that people become experts in different ways but we don't really understand how someone becomes an expert or what natural abilities support those things.

The great thing about face recognition is it's something we are all naturally able to do, but these super-recognisers just seem to be able to do it better. What we want to do by studying these people is try and understand the broader mechanisms of how people become experts, and one of the best ways to do that is to get the best of the best. So if you want to learn how to play soccer, it makes sense to study the footwork of Lionel Messi. If you want to become a chess player, you look at a Grand Master. If you want to learn how to recognise faces, you look at these best super-recognisers.

If you think about it, you might have an everyday experience of somebody you know really well and you bump into them and you spot them from a mile away, from the back. So one of the things we think is special about super-recognisers is they get to that level of familiarity we have for our friends, our family, the people we work with every day, just much faster. So that quick, automatic recognition, we feel like they can get there after seeing the person for just a short period of time, just an acquaintance kind of person, just really quickly.

We also don't know if it's face specific. So although all the tests that we have given you today and other ones people are doing use faces and showing that you are good at recognising faces, you may actually be clueing onto other signs, ways of identifying that person too; their body shape, their posture, the gait, all of these things too also help us recognise the person. So you mentioned voice before, but we don't actually know how much of those different cues to recognition are supported by the same ability. So if you are a super-recogniser, can you also recognise body shapes really well and identify someone based on their body.

James Bullen: Right, it could be some underlying ability which is across these things and we just don't know it yet.

James Dunn: That's right, yes, it's still early days, there is still so much we don't know about super-recognisers. And one of the great things about having this test online is that we've got so many people reaching out to us to let us know, hey, I think I might be special and be a super-recogniser, and now we are able to bring them to the lab and do more studies to understand this special ability. Before we only had five, ten people at a time, now we are finding hundreds of these people. Although this is a rare gift, when you test thousands, you get hundreds.

Sana Qadar: You're listening to All in the Mind, I'm Sana Qadar, and today we are looking at the range of abilities when it comes to recognising faces and voices. Our producer James Bullen has been looking into this. And, James, I'm curious, has your talent for recognising faces ever freaked anyone out?

James Bullen: I've never really thought about it before, but looking back there were times where I remembered people from primary school and I remembered them from years above or below me in school and I'd see them on the train or in the city and be able to spot them out, and I always felt a bit bad about coming up and saying, 'hey, it's me', because most of the time they wouldn't have a clue who I was or what I was doing, this random stranger coming up to them.

Sana Qadar: Oh my gosh, that's hilarious, but also pretty awkward.

James Bullen: It is quite awkward. But one of the interesting things I found out about this whole area is that with facial recognition you have people who are super-recognisers at one end of the spectrum, and those who have face blindness, prosopagnosia, at the other, but there is also some evidence of a similar spectrum when it comes to recognising the human voice. Phonagnosia is the term used to describe voice-blindness, that's where someone can't distinguish voices.

Kristina: I'm 29 years old, I have cerebral palsy which is…it's a disability. I call it glorified brain damage because that's basically what it is. It has a wide spectrum of symptoms, and it's basically caused by either brain damage that occurs either before, during or after birth, and I've lived with it my entire life. So that has kind of been the lens through which I tend to see most of my life. Right now I'm a doctoral student. I'm working on a dissertation and I'm aiming to be a psychologist.

James Bullen: Kristina was in her teens when she first started realising she couldn't tell voices apart.

Kristina: Where my family used to live we had this pizza parlour that had been sort of like a mainstay of my childhood, we would order from there all the time. The owner knew us by name and it was a constant in our lives. But for some reason when I'd call to order pizza I wouldn't know who was talking, I would never know if it was the owner or one of his staff, but he would know me, and so that's when I first started realising that…I couldn't put a name to it but I realised that this was a problem that I had where if somebody sounded too similar to somebody else, even in the slightest way, if it was, like, two men within the same age range with a similar cadence, I wouldn't be able to tell them apart unless I was directly looking.

James Bullen: It wasn't a major issue for Kristina until a couple of years back when she began her professional career.

Kristina: What would happen is say I had a contact in my phone that had the same name as another contact, and I would call that contact and mistakes that person for the other contact because I didn't put a differentiator in the name, like I didn't put their last name in there, and I would carry on full conversations thinking I was talking to this other person, until I would realise either after the phone call or during the phone call that I had the wrong person. It becomes almost like a phobia. In my line of future work I'm going to be providing a service, and the last thing I would want to do is mix up my clients.

James Bullen: But she has developed ways to compensate.

Kristina: Even if it sounds odd to the person I'm speaking to, I've made it a habit to ask them to verify who I'm speaking with, even if I'm the one that has called them, just so I'm 100% aware I'm talking to the correct person. And I've also sort of tried to pay attention to differences in cadences, which is a lot harder to do but I'm trying to train my brain to recognise what it can't, if that makes sense.

The strange thing is I can tell different accents apart geographically, I can tell…for example, I can tell you are not from New Zealand. I know that some people can't. It has been pointed out to me that I'm very good at recognising different accents geographically, but unless I'm looking at you…like, if you contacted me from a different account tomorrow, I would know that I was hearing a man with an Australian accent but I wouldn't know it was you.

I did not know phonagnosia was a thing, which is strange because as a psychologist one of the first things you learn in your undergraduate studies is that brain damage can cause blindness in a variety of areas, and everybody knows about prosopagnosia, but I've never heard of voice-blindness. So all this time I thought I was weird. It turns out I'm not.

Ryan Jenkins: They are unable to recognise the voices of their own family members. For instance, they may not recognise the voice of their son, their daughter, whoever, if they were to call them.

James Bullen: Ryan Jenkins is a researcher at the University of Greenwich in London.

Ryan Jenkins: However, it has generally been found that other vocal information, such as a speaker's gender, age, emotion, alongside general speech recognition and musical processing remains mostly preserved. So these sorts of things aren't affected, it's just purely voice identification. And it has been generally found that there are two ways in which it can occur. The first one, found in the early 1980s, is acquired phonagnosia, which follows the onset of some brain injury. And so some research has identified damage to the right parietal lobes, which is more associated with voice recognition, auditory perception. So that's acquired phonagnosia. The other one, which is found much more recently, is developmental phonagnosia, which has been found to just be present from birth.

James Bullen: The literature suggests that two in 100 people will have phonagnosia, but the condition isn't the main focus of Ryan's research. He's looking at the other end of the spectrum, the other 2%.

Ryan Jenkins: Yes, so generally it has been found that people's abilities to recognise a voice is a lot poorer than recognising a face. So we focus more…we gather more information from our eyes, from our sight than we do from our hearing. But for voices, again, there is this idea of this spectrum where you have most people with typical or above or below typical abilities, and then at the lower end you have people with phonagnosia which is an inability to recognise familiar voices, whereas on the other end of the spectrum, theoretically would be these super voice-recognisers, who are people that I'm trying to look at in my own research.

James Bullen: Like with face recognition, there are tests that can help determine how well you can recognise voices.

Ryan Jenkins: There are two standardised voice tests that have come out, and these are the Bangor Voice Matching Test which is a test of voice matching or discrimination. So over 80 trials participants are presented with two voices presented side-by-side, and a participant's task is to decide whether or not the two voices belong to the same person or to two different speakers.

James Bullen: That section sounds a little bit like this:

[voice sounds]

You have to decide if those clips are from the same person or two different people. And then:

Ryan Jenkins: The other test is a voice recognition or voice memory test, and this is the Glasgow Voice Memory Test, and this one is split up into two stages. For stage I, the participant will end up learning eight voices and then will do a 16 trial voice recognition task where they have to do say or decide whether or not a voice they heard was from the learning phase and they'd say this is an old voice or whether or not it is a brand-new voice.

James Bullen: Then you hear a whole heap of bells as well and you have to tell them apart. It's quite challenging. Finally there is a section where you hear the voices of well-known people and have to work out who it is. Here's an easy one:

Barack Obama: I don't see him as the cause for our divisions and the problems with our government, I think he is an accelerant.

James Bullen: This battery of tests is designed to examine two distinct skills, one being how well you remember voices, and the other how well you can discriminate between them. Ryan was interested in how well super-recognisers did. He split them up into people who were good at remembering faces and those who are good at telling faces apart, and then he tested their voice recognising abilities compared to control groups.

Ryan Jenkins: And I generally found that those who performed exceptionally well at either face memory or face matching tended to outperform those who were typical range face memory or face matching at their respective voice tests. So those who were really good at face memory did really well on voice memory, those who were really good at face matching did really well on voice matching. This provides some quite good initial results to say that there is a possibility of super voice-recognisers.

James Bullen: I did these tests too, keen to cement my newfound role as the office super-recogniser, and I did okay, above average, but not by much.

Ryan Jenkins: The Bangor Voice Matching Test, the mean was around 69.5 out of 80. The voice section of the Glasgow Voice Memory test is 11.7 out of 16, and the bell section is 12.2 out of 16, and you got 71, 12 and 13, so they are all higher than average, yes, above average.

James Bullen: I did do decently on the famous voices test though.

Ryan Jenkins: So in this instance with a score of six out of 10 this would be a conditionalised naming hit rate of 0.6. But the overall mean I found for my participants was around 0.3 out of one.

James Bullen: But why would there even be an overlap in recognising voices and faces when they use totally different sensors?

Ryan Jenkins: So there have been some models and theories of voice and face recognition that do propose that they can be linked. It has been suggested that although initial coding and recognition processes are physiologically independent, so this being specific face and voice brain regions. Once these face or voice stores have been accessed, it has been proposed that there is some cross-modal familiar identity related information can be linked, such as the name, the face, the voice all comes into the same unit under a person.

So if you were, for example, to remember someone, you may remember the name, face, voice or something, and it's like all this is linked. And some brain imaging research has found that despite these independent physiological areas, some evidence has found activations in a place called the fusiform face area, which is an area commonly associated with face processing, and that was activated in response to a familiar speaker's voice alone without their face being there. So this then further demonstrates cross-modal processing between voices and faces, suggesting some underlying cognitive mechanism that may drive both. This is part of an underlying root for my research.

James Bullen: So the brain activates in similar ways when it's recognising faces and voices, but the voice recognition part is different to the noise recognition part. Put another way, a different part of your brain will light up when you are hearing your mother's voice versus a church bell ringing.

Ryan Jenkins: So there has been a few studies that have looked at various different areas, and some areas tend to interact, but strong research has suggested that through functional magnetic resonance imaging studies have proposed that there is a voice specific area, particularly temporal voice areas which are responsible for identity processing in the temporal lobe areas, and these have also been found to exhibit a stronger response to voice compared to non-vocal sounds. So this is indicating a voice specific area rather than a more auditory general area, which is sort of the important bit between distinguishing sound and a voice as identity, which is the main important part of what I'm looking at, for instance. So because it's suggested that there may be different areas or different parts of that area that relate to more auditory perception in general rather than, say, specific voice recognition, yes, it's probably fair to say that things like musical processing in phonagnosics tend to be intact.

James Bullen: But more work needs to be done before we can say with complete confidence that super voice recognisers exist.

Ryan Jenkins: Super voice-recognition is certainly something brand-new. My recent research has been the first to really explicitly focus on it. It has been mentioned of being, oh, maybe there is a possibility. However, my research has been the first to properly see super face-recognisers, also super voice-recognisers. So it's certainly very early days yet.

The next areas of research would be to basically continue assessing what could potentially be a super voice-recogniser to develop tests that focus more for the upper end of the voice recognition ability spectrum, and then from then, once a good standardised test has been designed, test and retest these people just so that they have consistently high scores, maybe even giving them more real-life tests, for example the things they use in forensic psychology based tests, so like voice line-ups or something. And the overall reason why as well is to ensure that if they just do one test, it's, okay, they scored highly but was this just a lucky guess, like a one-off? They need consistent testing and then scoring consistently highly on each test to really be classified as a super-recogniser.

James Bullen: If the link between these abilities bears out, then it could be a valuable asset for law enforcement.

Ryan Jenkins: With super face-recognisers, they have been deployed in several successful police operations, and across the world, there's a team in the Metropolitan Police Force in London who have done a lot of work in helping with investigative roles. And so it may be possible that similar settings can be used with super voice-recognisers, for example in cases of fraud, terrorism, over the phone voice analysis sort of things.

James Bullen: And understanding the connections between voice and face recognition could be helpful for people all over the spectrum of ability.

Kristina has trouble with faces and voices, and she wonders how they are linked.

Kristina: If I see you alone and I'm familiar with your face, I'll know I'm looking at you, but when it comes to picking people out of a crowd or in an unfamiliar setting, I all of a sudden will be either unable to recognise them or unsure whether I'm looking at the right person.

I've never thought to label myself as phonagnosic at all. Is it just the same issue coming at me from two different distinct ways? I've thought about it before. I think that the link between those two things, given what I've said, is also interesting to look into because I do wonder if they are connected.

Sana Qadar: That's Kristina, ending that story from All in the Mind producer James Bullen. You also heard from Ryan Jenkins, a PhD student at the University of Greenwich, and Dr James Dunn, a postdoctoral research associate with the University of New South Wales School of Psychology.

Now, for that test we had at the start of the show with the various voices. If you were wondering who they were, well, the first one was former Prime Minister Malcolm Turnbull:

Malcolm Turnbull: If all of that money that had been put away in a super had just been put into the housing market…

Sana Qadar: The second was actor Daniel Radcliffe, a.k.a. Harry Potter:

Daniel Radcliffe: The sort of chaos I used to invite into my life…

Sana Qadar: The next was Kim Kardashian:

Kim Kardashian: …you know, what, like it just didn't compute to me for a minute.

Sana Qadar: Followed by British Prime Minister Boris Johnson:

Boris Johnson: …the issue, frankly, of being overweight.

Sana Qadar: And finally, actress Jennifer Aniston:

Jennifer Aniston: And then one day you are sitting there with your friend doing yoga and you go, oh, I get it.

Sana Qadar: That's it for the show this week. Our sound engineer is John Jacobs. I'm Sana Qadar, thanks for listening, catch you next time.