What am I going to say next?

How do our brains decode speech?
03 April 2018

Interview with 

Dr Thomas Cope, University of Cambridge


How do our brains effectively 'cheat' when decoding speech, and what does that have to do with tinnitus? Georgia Mills found out from Cambridge University's Thomas Cope...

Thomas - The important thing to always remember is that your brain is not just representing what's going on in the outside world, it’s representing an interpretation of what’s coming into you through your sensors. So, if I’m talking to you now in a quiet room, you’re predicting as you’re going along what the next word that I’m going to say is. If I said to you “I’m just popping out to the shop to buy some…” You would have a list of things that are most likely for me to say next; top of that list might be milk, and then bread, and then there’d be less common words lower down. When you then get information from your senses, you are comparing that information against your prediction to see was I right or wrong? Do I need to do a lot of work in processing this auditory information to work out what it was, or was the start of it, like milk, and therefore it’s probably milk and can I just move onto the next thing saving energy in my neural architecture?

Georgia - I see. So I think you’re going to say one word, my brain prepares itself for that word, and if it’s the same word job done, move on, and if it’s not then we have to take a step back and think a bit harder?

Thomas - Yeah, that’s exactly right. Your auditory cortex, which is the first bit of the brain that auditory information goes into, sets up a prediction and it only sends forward to other parts of the brain errors from that prediction. So, if there are no errors, it was correct; then it has to send nothing further forward and nothing else has to happen. If what comes in is a different word or you’re not sure what it was; then it has to work harder in sending information higher up the processing stream to other parts of the brain to work out okay, he didn’t say milk he said eggs. What was it, how does that change what I’m going to predict next?

Georgia - How did we find out this is what our brains were doing?

Thomas - The grand idea is called predictive coding. And there’s evidence for predictive coding, not just in speech. This is a process that underlies everything that the brain does. The brain is really a predictive engine, and we know that errors in predictive coding underly a lot of different diseases. There’s evidence that in schizophrenia predictions go wrong and that can lead to abnormal perceptions like hallucinations.

In tinnitus, predictions go wrong, so you have a lack of input from one part of the ear because of damage there, and that sets up a little bit of noise, and the brain starts to predict that maybe the noise is meaningful and maybe I should pay attention to this noise and, over time, that’s perceived as tinnitus. Even if the noise goes away, the prediction that there will be tinnitus remains. And there’s nothing to counteract that because the ear has died in that part and then you perceive tinnitus as an interaction of strong predictions about hearing something, and nothing meaningful going in.

Georgia - It’s interesting you should mention tinnitus because that’s a very common condition as I understand, so we think that this is caused by our predictive system going into overdrive?

Thomas - Yeah; that’s the emerging view. About half of all adults will have some tinnitus if they’re in a quiet room and they’re really concentrating on listening to what’s coming into their ears, they’ll hear a quiet buzzing or ringing and that’s completely normal. When tinnitus become troublesome is when that quiet buzzing or ringing gains an abnormal perceptual salience, so people start to attend to that buzzing or ringing. They start to expect that buzzing or ringing, and any information that does come in is interpreted very precisely, and that prediction of buzzing or ringing is verified and then the tinnitus essentially ramps up in how irritating it is independent of what the volume of the tinnitus actually is. Predictive mechanisms, together with some problem in the ear, are what causes tinnitus.

Georgia - When Thomas says the buzzing or ringing can ramp up, it can get really, really bad. It can stop people from concentrating or from sleeping and potentially ruin lives. This is what it can sound like. That’s from the British Tinnitus Association. So does knowing about this mechanism help us treat tinnitus?

Thomas - Well, yeah. This is one of the reasons that things like cognitive behavioural therapy for tinnitus can be quite effective because it’s not just a problem of what’s going on in the ear causing the perception. It’s a problem of how the brain is working subconsciously to really eek out every bit of information it can get from what’s essentially noise ramping up the intensity of this perception. We know that therapies that reduce the loudness of tinnitus don’t necessarily reduce the distress of tinnitus, and these two thing need both to be tackled.


Add a comment