Latest / The Psychology Undergrad Podcast / Sound Localization & the Cocktail Party Effect: Hearing in Noisy Rooms
Transcript
- 0:00Welcome to the Psychology Undergrad. I'm so excited
- 0:03for today's deep dive because, well, we are tackling
- 0:06Chapter 7. Oh man, Chapter 7. Hearing in complex
- 0:10environments, it's a dense one. It really is.
- 0:14But if you're listening to this right now, just
- 0:15consider yourself our third study buddy. Our
- 0:18whole mission today is to break down this totally
- 0:21overwhelming science of spatial hearing and speech
- 0:25processing. Right, because we want to turn all
- 0:27this textbook jargon into actual memorable nuggets
- 0:29so you can completely ace this imaginary psychology
- 0:32exam with us. Exactly. And I mean, auditory processing
- 0:36is just wild. If you close your eyes, you can
- 0:38perfectly point to where your fridge is humming.
- 0:40Yeah, you're building this flawless 3D audio
- 0:43map in the dark. It feels effortless. But the
- 0:46computational miracle your brain is performing
- 0:47just to locate a single sound or understand a
- 0:50spoken word in a loud room is staggering. So
- 0:53to understand how we do that, we have to start
- 0:55with the actual coordinate system, right? The
- 0:573D map. Right. The brain uses a strict three
- 0:59-part coordinate system for spatial hearing.
- 1:01You've got azimuth, elevation, distance. Okay
- 1:05let's unpack this because I actually came up
- 1:06with a memory trick for you our study buddy using
- 1:09moral geometry. Oh wow moral geometry. Okay let's
- 1:14hear it. Yeah so think of azimuth as the horizontal
- 1:16path of life. It's the left and right. You're
- 1:19constantly choosing between the right path and
- 1:21the left path. I like that. It tracks. Right.
- 1:23And then elevation is the spiritual vertical.
- 1:26So you're either looking up toward the heavens
- 1:28or you're looking down toward the underworld.
- 1:30Okay. Very dramatic. And distance is just simply,
- 1:33you know, how far away you are from reaching
- 1:36your ultimate moral goals. Well, that is definitely
- 1:39going to stick in my head for the exam. So taking
- 1:42that vertical axis. your spiritual elevation.
- 1:46The brain actually uses a monaural cue to figure
- 1:49that out. Monaural meaning just one ear. Exactly.
- 1:52Just one ear is enough. And it relies on the
- 1:54head -related transfer function or HRTF. Which
- 1:57is just a fancy way of talking about the actual
- 1:59physical shape of our ear cartilage. Yeah, those
- 2:05weird fleshy folds on the outside of your ear.
- 2:08Because depending on whether a sound comes from
- 2:10above or below, those physical folds will dampen
- 2:13or boost specific frequencies before the sound
- 2:16even enters your ear canal. Wait, I was reading
- 2:19the Hoffman et al. study from 1998 about this
- 2:22in the text, where they gave subjects fake ear
- 2:25molds. Oh, yeah, that study is incredible. They
- 2:28completely changed the physical shape of the
- 2:30subject's pinna. And it ruined their elevation
- 2:32perception, right? Like they couldn't tell up
- 2:34from down. But the text says they learned a brand
- 2:36new HRTF in just a few weeks. Right, their brains
- 2:39adapted. But here's my question. When they took
- 2:41the fake ears off, they instantly regained their
- 2:44old HRTF. Does our brain just like keep multiple
- 2:47Macs of our ears on file, like hoarding old pairs
- 2:50of prescription glasses? It basically does. The
- 2:52neuroplasticity is amazing. It retains that original
- 2:55blueprint perfectly. which is super important
- 2:58because even with your custom ear shape, your
- 3:00radar has a massive blind spot called the cone
- 3:03of confusion. The cone of confusion? I am just
- 3:06picturing a dog wearing a plastic cone from the
- 3:08vet. I mean, visually it's actually a similar
- 3:11shape. It's this region pointing out from the
- 3:13side of your head where the audio cues are mathematically
- 3:17identical. So a sound in front of you and a sound
- 3:19behind you produce the exact same acoustic data
- 3:21So you just mix up front and back that seems
- 3:24like a terrible evolutionary flaw It would be
- 3:27but it assumes your head is perfectly still the
- 3:30second you move your head even a tiny bit You
- 3:31break the symmetry the confusion vanishes. Okay,
- 3:34so just look around that's elevation But what
- 3:36about the azimuth the left and right paths of
- 3:39life because that requires both ears, right?
- 3:41Yes, binaural cues and the primary one is internal
- 3:45time difference or ITD because sound takes time
- 3:48to travel. But it travels so fast. It does. A
- 3:50sound on your left hits your left ear slightly
- 3:52before your right ear. But the maximum delay
- 3:54is only about 0 .676 milliseconds, so roughly
- 3:59686 microseconds. Wait. Microseconds. As in millionths
- 4:03of a second. Yeah. But your brain stem, specifically
- 4:06the superior olivary complex, is so insanely
- 4:09sensitive they could detect differences as small
- 4:11as 10 to 20 microseconds. That is just incomprehensible
- 4:14to me. It uses Lloyd A. Jefferson's coincidence
- 4:17counter circuit. It's basically a neural track
- 4:19where signals from both ears travel and a neuron
- 4:22only fires when the signals crash into each other
- 4:25at the exact same time. Oh, so it turns time
- 4:28into a physical location on the neural track.
- 4:30They proved this in barn owls in 1988, right?
- 4:33Exactly. Barn owls have perfectly optimized brains
- 4:36for this because they hunt in the dark. But surprisingly,
- 4:39scientists couldn't find the same optimal setup
- 4:41in guinea pigs. Probably because their heads
- 4:44are too small to get a good time delay. But my
- 4:46analogy for this ITV thing is that the brain
- 4:48is like this omniscient moral judge weighing
- 4:51the absolute truth between two witnesses. The
- 4:54two witnesses being the two ears. Right. Even
- 4:56if one witness speaks just a fraction of a millisecond
- 4:59later, the judge knows exactly where the truth
- 5:01originated. That's a great way to think about
- 5:03it. But wait, what if the sound is too high pitched?
- 5:06Because the text says ITD stops being useful
- 5:08at around 1 ,500 hertz. Does the judge just give
- 5:11up? No, the judge just uses a different metric.
- 5:14For high frequencies above a thousand hertz,
- 5:17the brain switches to ILD, the interaural level
- 5:19difference. Level, so volume. Right, because
- 5:22high frequencies have really short wavelengths.
- 5:24So instead of wrapping around your head, they
- 5:26literally crash into your skull. Oh, so your
- 5:29head casts an acoustic shadow. Precisely. The
- 5:32sound gets up to 20 decibels quieter in the far
- 5:35ear, and your brain uses that volume drop to
- 5:38pinpoint the azimuth. OK, so we have the angle.
- 5:41But how does the brain know about the physical
- 5:43room we're in? Or the distance? That's where
- 5:46echoes come in. Reverberation. We measure reverberation
- 5:49time, which is just how long it takes for a sound
- 5:51to decay 60 decibels. So, like, hardwoods and
- 5:53metal would make it echo longer, but carpets
- 5:56and foam would muffle it. Right. And your brain
- 5:58compares the crisp, direct sound to the smeared
- 6:01echoes. It's called a direct -to -reverent ratio.
- 6:04Okay, and this is where I use the spherical expectation
- 6:06analogy for distance. Oh, another one? Go for
- 6:09it. Yeah, so our expectations totally dictate
- 6:11our reality here. If you expect a divine whisper
- 6:14to be very quiet, but you hear it loudly, your
- 6:16brain immediately knows it's right next to your
- 6:19ear. It's all about what you expect the truth
- 6:21to sound like. That is very profound. And it
- 6:24works for non -divine things, too. Like the textbook
- 6:27example of hearing a train. We expect a train
- 6:30to be loud, so if we hear it quietly, we know
- 6:32it's miles away. Or my favorite example from
- 6:34the text, hearing that angry 28 -year -old yelling
- 6:36about smiling shark swim shorts in a retail store.
- 6:39Oh my gosh, yes. Like you hear that loud voice
- 6:42mixed with the massive echoes of a huge store,
- 6:45and you know exactly how far away he is. So you
- 6:47can completely steer clear of the children's
- 6:49section. It is a very practical survival skill.
- 6:52But let's take all of this, the coordinates,
- 6:54the shadows, the echoes, and put it in a really
- 6:57crowded room. It's chaos. How does the brain
- 6:59isolate what it actually wants to hear? Sound
- 7:02segregation. And there are two main strategies
- 7:04for this, right? Right. Primitive strategies,
- 7:06which are bottom -up, and schema -based strategies,
- 7:09which are top -down. So primitive is just grouping
- 7:12sounds by physical similarities, like pitch or
- 7:15timbre. separating the blare of trumpets from
- 7:18the whisper of flutes and an orchestra. Exactly.
- 7:20It's raw acoustic data. But schema based strategies
- 7:23rely on memory. It's matching a signal to something
- 7:26you already know, like picking out a specific
- 7:28bird song at a noisy park. OK, here's where it
- 7:31gets really interesting, though. If I'm at a
- 7:33super loud frat party and I'm trying to eavesdrop
- 7:36on a very specific, juicy conversation across
- 7:39the room, which strategy am I using? You're using
- 7:42both to solve what we call the cocktail party
- 7:45problem. It is an incredibly difficult cognitive
- 7:48task. You have to use intense selective attention
- 7:51to tune out the background noise and hyper -focus
- 7:54on that target sound. Okay, so I finally segregate
- 7:56the sound. I'm focusing on human speech. But
- 7:58before my brain can even understand it, it has
- 8:01to decode how it was physically produced, right?
- 8:03Right, which involves three mechanisms. Respiration,
- 8:06phonation, and articulation. Respiration is just
- 8:09the lungs providing air energy. And phonation
- 8:11is that air passing through your vocal folds
- 8:13in the larynx to create a buzzing sound. And
- 8:16articulation is using your mouth. tongue, lips,
- 8:19and teeth to shape that buzz into words. And
- 8:21we can actually see these sounds on a graph called
- 8:24a spectrogram. Yes, time frequency plots. Vowels
- 8:28appear as these thick bands of constant acoustic
- 8:31power called formants. Right, and consonants
- 8:33appear as transitions, where the power bands
- 8:35swoop up and down because you're rapidly moving
- 8:37your lips or tongue. Exactly. But real world
- 8:40speech... on a spectrogram is incredibly messy.
- 8:44To decode it, your brain has to solve the segmentation
- 8:47problem, knowing where words even start and stop,
- 8:51because we don't pause between words. And the
- 8:53co -articulation problem, where sounds change
- 8:55based on context. Right, plus just speaker variation.
- 8:58People talk fast, they slur, they mumble. So
- 9:01this is where my last Meryl analogy comes in.
- 9:03The cheat code for speech, categorical perception.
- 9:06And phonemes. Yes. distinct units of sound, like
- 9:10voice onset time or VOT, the delay before your
- 9:13vocal cords vibrate. It's what changes the T
- 9:16in tie to the D in die. Right, if you artificially
- 9:19shorten the VOT, it physically blends the two
- 9:22sounds. But our brains hate the gray area. Even
- 9:24if the sound variation is literally continuous
- 9:26on the recording, the brain forces it into a
- 9:29strict category. It's either completely tie or
- 9:31completely die. It rounds the data, yeah. Just
- 9:33like separating good and evil, you know? There
- 9:35is no in -between sound, a loud... in our perception.
- 9:38It's absolute. I mean, that's incredibly accurate.
- 9:42So what does all this mean? When the brain is
- 9:44guessing and categorizing this aggressively,
- 9:47it can make mistakes. Like when it gets confused
- 9:49by our eyes? Yes, the McGurk effect. Visual cues
- 9:53can literally override auditory ones. If you
- 9:56see someone mouthing, gah, but the audio is,
- 9:58bah, your brain will actually change what you
- 10:01hear. to match what you see. That is terrifying.
- 10:04It is. And if the categorization system breaks
- 10:07down entirely in the brain, you get things like
- 10:09Wernicke's aphasia. That's receptive aphasia,
- 10:12right? Yeah. It's tragic. The person can speak
- 10:14with perfect grammar, but they lose the ability
- 10:16to make sense of words. It makes zero logical
- 10:19sense because the brain can't map the sound to
- 10:21a meaning anymore. Wow. Okay, study bad. If you've
- 10:24made it this far, congratulations on getting
- 10:26through chapter seven. Basically, hearing isn't
- 10:28just passively recording the world. It is this
- 10:31highly active computational illusion generated
- 10:34by your brain. Mapping space, weighing time,
- 10:38aggressively categorizing messy data. And it
- 10:40really leaves you with a provocative thought,
- 10:42honestly. Oh, like what? Well, building on the
- 10:45McGurk effect and how our expectations change
- 10:48distance. If our brain literally alters the audio
- 10:52we hear based on what we expect or what our eyes
- 10:54see, can we ever truly trust our own sensory
- 10:57perception of reality? Oh, man. Or are we just
- 11:00living in a highly edited simulation created
- 11:02by our own minds? Okay, I am going to be thinking
- 11:04about that all night. Good luck on your psychology
- 11:07exam, everyone. You're going to crush it. Catch
- 11:09you next time on Psychology Integrate.