Chapter 10: Sentences: Cognitive Processes in Language Structure

Welcome to Last Minute Lecture.

This free chapter overview is designed to help students review and understand key concepts.

These summaries supplement, not replace, the original textbook and may not be redistributed or resold.

For complete coverage, always consult the official text.

Welcome back to the Deep Dive.

It is great to have you with us again.

Good to be here.

Today, we are doing something a little bit meta.

Usually, we look outward.

We're analyzing history or technology or, you know, the business world.

But today, we are turning the spotlight inward.

Or maybe I should say we're turning the microphone inward.

I like that.

We have a stack of fascinating material here, and it's all centered around a single, incredibly dense chapter from the classic cognitive psychology text.

Specifically,

chapter 10,

sentences.

A big one.

It is.

And our mission today is to unpack something so fundamental that we usually just, we just take it for granted.

How the human mind actually processes spoken language.

It is fundamental, isn't it?

But it's also invisible.

That's the strange paradoxical thing about language.

How so?

Well, it is arguably the most complex cognitive feat we perform as a species, and yet we do it effortlessly, continuously, every single day without breaking a sweat.

We don't even realize we're doing it.

Exactly.

And I want to start with a hook that I think really frames why this is so cool.

I want you, the listener, to just pause for a second and really think about what is happening right now.

As you are hearing my voice, a legitimate biological miracle is taking place.

You are understanding sentences that you have likely never, ever heard before in this exact order, with this exact intonation, in this exact context.

Your brain is taking a stream of physical noise,

just vibrations in the air,

and instantly converting it into complex, abstract thought.

That is the core miracle of auditory cognition.

And what we are looking at today involves the intersection of psychology and linguistics, a field we often call a psycholinguist.

Psycholinguistics.

And the material we're diving into, it really challenges that old school idea that language is just a habit.

You know, the idea that we speak the way we do because we've been conditioned like Pavlov's dogs.

Just stimulus response.

Exactly.

It turns out language is much more about creativity and crucially about structure.

Structure is the key word there.

We are going to come back to that a lot.

But let's get right into the first big hurdle the text identifies, which really sets the stage for everything else.

The problem of novelty.

Yes.

Because, and I hadn't really thought about this before reading this chapter,

most of what we say is brand new.

It's irrepressible novelty, as the text describes it, which is a great phrase.

If you stop and think about your daily conversations, you aren't just a parrot repeating phrases you've heard a thousand times.

Right.

We're not just playing back recordings.

No.

I mean, sure, we have some stock phrases like, how are you?

Or, pass the salt.

But the vast majority of our communication involves constructing unique sentences to fit unique situations.

The source quotes Noam Chomsky on this, and he's really the heavy hitter in this field.

He argues that a mature speaker can produce a new sentence on the fly.

On the fly, yeah.

And a listener can understand it immediately, even if it's completely new to both of them.

And that observation is what really broke the back of the older psychological theory, specifically behaviorism.

Okay, so let's dig into that.

You have to remember, for a long time in the early 20th century, the behaviorists wanted to explain language as a series of habits or stimulus response chains.

Like a dog learning to sit when it hears a whistle.

Exactly.

They thought, okay, you see a chair, you say chair.

Simple.

But Chomsky pointed out the obvious flaw.

You cannot have a habit for a sentence you have never heard or spoken before.

That makes total sense if I say that.

The fluorescent hamster piloted a submarine made of cheese.

Right.

A perfect example.

I have definitely never practiced that sentence.

I've never heard that sentence.

But I can say it, and you, and you, the listener, can visualize it instantly.

Exactly.

If language were just habit, you'd be stumped.

You'd stare at me blankly because you have no conditioned response for cheese submarines.

But I don't.

I get it immediately.

You get it.

So Chomsky and the cognitive psychologists argue that to understand this infinite creativity, we have to stop looking at habits and start looking at rules.

Mental rules.

Deep mental rules that determine how words relate to one another.

It's not about which words are used.

It's about how they are organized.

It's all about the structure.

Okay.

So let's unpack this idea of organization because the text draws a really interesting comparison between modern linguistics and old school Gestalt psychology.

Yes.

Now, usually when I think of Gestalt, I think of visual puzzles.

Is it a duck or a rabbit or how we group dots together into a shape?

That's the classic application, for sure.

Gestalt psychology is famous for that maxim.

The whole is more than the sum of its parts.

When you look at a square, you don't see four unrelated lines floating in space.

You see a square.

The shape, the Gestalt, depends on the relationship of the parts, not just the parts themselves.

And the text argues that this applies to listening, too.

We don't hear a list of words.

We hear a shape of a sentence.

Precisely.

Just as a visual shape depends on the whole figure, a sentence's meaning depends on its overall structure.

The text uses a great example from the researcher Lashley to demonstrate this.

Listen to this sentence carefully.

Rapid writing with his uninjured hand saved from loss the contents of the capsized canoe.

Okay, wow.

Let's break that down in slow motion.

When you start that sentence,

rapid writing.

When you hear rapid writing, your brain immediately starts making assumptions.

You hear the word writing, and because of the context of hand, you probably think of writing with a pen W -R -I -T -I -N -G.

Right.

Of course.

Rapid writing with his uninjured hand.

Makes perfect sense so far.

Someone is writing a letter quickly because their other hand is hurt.

But then you get to the end of the sentence, saved from loss the contents of the capsized canoe.

And suddenly the image in my head just crashes.

Writing a letter doesn't save a canoe.

Your brain has to perform a full and surprising reorganization of what you just heard.

You realize writing didn't mean using a pen, it meant R -I -G -H -T -I -N -G, correcting the balance of the boat.

So you have to reinterpret everything that came before.

But here's the kicker.

You can't understand the beginning of the sentence until you have the whole structure.

The meaning of the very first word was determined by the very last word.

That is a perfect segue into one of my favorite concepts from this deep dive.

Ambiguity.

The text uses ambiguity as a tool to prove that we are actively organizing what we hear.

It compares it to the Peter Paul Goblet.

That's the famous visual illusion, yeah, where depending on how you focus, you either see a white vase in the center.

Or two black silhouette vases looking at each other from the sides.

And the key is that you can't see both at once.

Your brain imposes a structure.

It decides this is the figure and this is the ground.

You decide what is the object and what is the background.

And there's an auditory parallel to this.

The text gives the example, they are eating apples.

It sounds so simple, doesn't it?

They are eating apples.

But stop and analyze the structure.

There are two completely different ways to organize those words grammatically.

Okay, let me try to pull this apart.

So interpretation A is the obvious one.

They refers to people.

Our eating is the verb, the action.

So some people are currently consuming fruit.

Correct.

That is the most common reading.

But what if they refers to the apples themselves?

Oh, I see.

Like if someone asked, are these the kind of apples you cook with?

And I said, no, they are eating apples.

Exactly.

In that case, are is the verb and eating becomes an adjective describing the type of apple.

Wow.

The words, the raw sound waves hitting your ear are identical in both cases.

But the structural organization in your mind is completely different.

Which proves that the mind isn't just a passive recorder.

Not at all.

It is actively imposing a structural interpretation on the input.

We are deciding the shape of the sentence.

That is fascinating.

It really highlights that meaning isn't just in the words themselves.

It's in the architecture we build around them.

And this leads us to a bit of a philosophical clash mentioned in the chapter, nativism versus empiricism.

The classic nature versus nurture debate.

Indeed.

The behaviorists were empiricists.

They thought language was learned entirely through experience and conditioning.

You start with a blank slate and you learn everything from the world around you.

But the Gestaltists and later Chomsky argue for nativism.

They suggest that the brain has innate genetic structures for organizing language.

So just like we are hardwired to perceive visual depth, I mean, we don't have to go to school to learn how to see in 3D.

We're hardwired to perceive grammar.

That is the argument.

Children acquire complex grammar so quickly and universally that it seems impossible they are listening solely through trial and error.

There's this concept called the poverty of the stimulus.

Poverty of the stimulus.

It basically means kids don't hear enough examples, especially of complex or rare sentences, to learn all the complex rules they know just by guessing.

There seems to be a grammar template built into the human operating system.

Which brings us to the nuts and bolts of that template.

We need to talk about grammar.

And the text makes a very important point to distinguish this from the grammar we learned in school.

We aren't talking about where to put a comma or avoiding the word ain't.

Yes.

This is crucial for the listener to understand.

We are not talking about prescriptive grammar, the rules that say don't say ain't or don't split infinitives.

Those are just social etiquette, basically.

They really are.

We're talking about generative grammar.

Generative grammar.

That sounds wonderfully industrial.

What does it actually mean?

It refers to the internal mental rules that allow us to generate all possible grammatical sentences and this is the key part, generate zero ungrammatical ones.

It's the code running in the background that defines what English or French or Mandarin actually is.

It's the engine.

It's the engine that produces language.

Okay.

So to explain how this code works, the text first shows us how it doesn't work.

It introduces something called Markov grammar or left to right grammar.

This is an early theory, right?

A very early one, yes.

The idea was that maybe language is just a chain reaction.

Maybe we just predict the next word based on the previous word.

So if I say thee, you expect a noun.

If I say the dog, you expect a verb.

A simple probability chain linking one word to the next like dominoes falling.

It sounds plausible on the surface.

I mean, we do kind of predict what people are going to say.

We do, but the text just tears this apart with the concept of long distance dependency.

Long distance dependency.

This is where it gets really interesting.

A simple left to right system is short sighted.

It can't remember what happened 10 words ago, but humans can.

The text gives this complex example.

The people who called and wanted to rent your house when you go away next year are from California.

That is a mouthful.

Let me look at that again.

The people are from California.

Exactly.

But look at everything happening in between.

The people who called and wanted to rent your house when you go away next year are.

There are 15 words separating people and are.

But they are like chemically bonded.

People is plural, so it needs the plural verb are.

Right.

If you said the people is from California, it would sound completely wrong.

Immediately wrong.

A left to right system would get lost in that middle section.

It sees the word year, a singular noun right before the verb.

It would likely try to match the verb to year and say is.

But our brains don't do that.

No, our brains effortlessly hold the people in suspension waiting for its partner verb are bridging that massive gap.

That proves we aren't processing in a flat line.

We are processing in hierarchical groups.

And to explain this hierarchy, the expert in the text uses a visual description that I found really helpful, but a little tricky to visualize at first.

The rolling wheel.

OK, let's try to visualize figure 42 from the text together.

Imagine a wheel rolling along the ground in the dark.

OK.

Now imagine there is a single glowing light attached to the rim of the wheel.

OK, I'm picturing it.

A glowing dot spinning and moving forward at the same time.

Right.

So if you are standing on the sidewalk watching this, that dot traces a very weird path.

It goes up, it loops over, touches the ground, starts for a split second and then goes up again.

It's a curve called a cycloid.

It's complex and it looks, well, chaotic because it's combining two motions, the spinning and the moving forward.

But now imagine we turn on a light at the hub, the center of the wheel.

Suddenly, everything changes.

If you look at the glowing dot relative to the hub, it's just a perfect circle.

It's just a simple, perfect circle.

Relating it to the hub makes the movement simple and organized.

I see the analogy now.

The text argues that language is the same.

If you listen to words as just a flat string, that's the cycloid, it's a mess.

It's just one word after another with no apparent logic.

But if you see the hub, the phrase, it becomes organized.

We hear words circling a central concept.

This leads us to the famous tree diagram or constituent structure.

This is the map of those hubs.

The tree diagram is the visual representation of this hierarchy.

Imagine an upside down tree.

At the very top, you have the sentence or S.

OK.

That splits into two main branches, the noun phrase NP and the verb phrase VP.

So let's take a simple sentence.

Paul saw Mary.

Paul would be the noun phrase and saw Mary would be the verb phrase.

Correct.

Paul is the subject.

Saw Mary is the predicate.

The brain processes these as chunks or constituents.

We don't hear Paul saw Mary as three equal beads on a string.

We hear Paul and saw Mary.

And then inside saw Mary, we break it down again.

Exactly.

Into the verb saw and the noun Mary.

This grouping, this chunking, is what allows us to handle those long distance dependencies we talked about.

The people who that whole thing is one giant noun phrase and are from California is the verb phrase.

We match the phrases, not just the words next to each other.

So we have this hierarchical structure.

We know it exists.

But here's the practical question.

How do we actually build it in real time?

Right.

We don't get the tree diagram handed to us on a piece of paper.

We just get sound waves.

This brings us to section three, how we decode.

And the theory here is called analysis by synthesis.

Analysis by synthesis.

That sounds like we are actively building the sentence ourselves as we listen.

That is the idea.

The listener isn't a passive bucket catching words.

We are actively reconstructing, synthesizing the sentence structure as we hear it, trying to match the input.

It's like we're running a simulation of what the speaker is saying in order to understand it.

So we're constantly guessing the structure and then confirming it.

Constantly.

But we need clues to do that.

If I'm building this tree in my head, I need to know where the branches are.

The text lists a few specific cues that help us figure that out.

There are three main cues the text identifies.

First, and this is a big one, function words.

The little words.

The little words.

The a, in, of.

They don't carry the heavy meaning like dog or run, but they mark the skeleton of the sentence.

They are the traffic signals.

They tell you, hey, a noun phrase is starting here.

It's like seeing a road sign that says entering noun phrase city.

Exactly that.

When you hear the, your brain opens a new noun phrase folder.

Second, we have effixes.

The endings of words.

The endings.

Things like leg -ly, legs -ovation, legging.

These tell us the grammatical category of the word.

If it ends in lug -ly -ly, I know it's probably an adverb, so I know where to hang it on the tree.

And the third cue.

Prosody and rhythm.

The pauses, the stress, the musicality of speech.

This is huge for resolving ambiguity.

Can you give an example of how rhythm changes meaning?

Sure.

Compare these two.

First,

Sam the mechanic can't come.

And second, Sam the mechanic can't come.

In the first one, Sam paused the mechanic can't come.

I hear a pause after Sam, so I know you are talking to Sam telling him about the mechanic.

But in Sam the mechanic can't come, there is no pause.

The rhythm groups Sam and mechanic together.

So mechanic becomes his title, Sam the mechanic.

Exactly.

The prosody tells you how to group the words.

It is literally the glue that holds the phrase structure together.

To prove how powerful these cues are, especially the function words and effixes, the text discusses the Jabberwocky effect.

It mentions a study by Epstein in 1961.

This is a classic.

It's so cool.

Epstein gave people nonsense sentences to learn.

One was, the Yigs were vomly rixing hum in Jejistmiv.

Sounds like an alien language or bad poetry.

It does.

But notice the structure.

The Yigs sounds like a noun phrase because of the and the S at the end of Yigs.

Where vomly rixing sounds like a verb phrase because the Lee and Ing.

So even though the words are fake, the frame is English.

The frame is perfect English.

And the finding was that people learned this nonsense sentence much faster than they learned a random list of the same nonsense words presented without that grammar frame.

That is incredible.

So even without any meaning, the grammar itself acts like a hook to hang the memory on.

Exactly.

The frame, the Yig spin, it allows the brain to build a phrase structure tree.

It gives the nonsense words slots to sit in.

Without the structure, it's just noise.

There was another experiment mentioned by Miller and Selfridge involving approximations to English.

That one really tripped me up at first.

Can you explain what an approximation is in this context?

It's a really clever way to test how much context matters.

They generated strings of words based on probability.

A low order approximation is just random words picked from a dictionary.

Total chaos.

OK.

But a high order approximation selects words based on the probability of them following the previous few words.

So they ask people to guess the next word in a sentence and then build chains from that.

Basically, yeah.

And the result is something that sounds like English but is total gibberish.

A high order one might be something like,

they saw the play Saturday and sat down beside him.

They saw the play Saturday and sat down beside him.

I mean, that sounds like a real sentence.

It flows.

It does.

It follows the local rules of probability.

Saturday often follows play.

Sat often follows Saturday.

But it doesn't actually mean anything coherent in a broader sense.

Right.

It's just a string of likely connections.

But here's the thing.

Listeners remembered the high order approximations much, much better than the random words.

Why?

Is it just because the words go together more often?

That's part of it.

But the text argues the real reason is that these approximations allow us to build a phrase structure.

Even if the meaning is nonsense, if the grammar holds up, our brains can latch onto it.

It proves that we crave structure.

We need to hang the words on a tree to keep them in memory.

OK.

So far, we've established that we build these tree structures to understand sentences.

But the deep dive goes even deeper.

The text introduces a problem where the tree diagram, the phrase structure, isn't enough.

Sometimes the tree can lie to us.

Here's where it gets really interesting.

Phrase structure explains the surface.

But sometimes the surface is deceptive.

The lion's example.

I stared at this one for a while.

Yes, it's a great one.

Consider two sentences.

Sentence A, growling lions can be dangerous.

Sentence B, subduing lions can be dangerous.

On the surface, those look identical.

You've got an adjective type word, a noun, a verb phrase.

Grammatically, on the surface, they map to the exact same tree.

But think about the meaning.

In growling lions, who is doing the growling?

The lions, of course.

The lions are the actors.

Right.

But in subduing lions can be dangerous.

Are the lions subduing anyone?

No.

Someone else is subduing the lions.

The lions are the ones being acted upon.

They are the object.

Precisely.

The surface grammar is identical.

But the deep relationships are totally different.

In one, the lion is the subject.

In the other, the lion is the object.

A simple tree diagram of the surface words can't capture that difference.

So this leads to Chomsky's big gun,

transformational grammar.

This is the idea that every sentence has two levels.

There is the surface structure, what we actually hear a final output.

And then there is the deep structure, the core meaning.

And how do they connect?

Chomsky argues that we start with a kernel sentence.

A kernel sentence.

That's the simplest active declarative form, like the secretary typed the paper.

Correct.

That's the kernel.

It's the raw semantic data.

But we can apply transformations to that kernel.

Think of them like Instagram filters for grammar.

I like that analogy.

We can apply a passive transformation.

The paper was typed by the secretary.

We can apply a question transformation.

Did the secretary type the paper?

So the kernel is the raw material.

And the transformations are the filters we put it through to change its style or focus.

Exactly.

And this solves the lion's problem.

Growling lions comes from a deep kernel, like lions growl.

Subduing lions comes from a kernel, like someone subdues lions.

The deep structures are completely different, even if the surface transformations make them look the same.

And the idea is we understand the sentence by reverse engineering it back to the kernel.

That's the theory.

Now, this is all very elegant linguistically.

It makes for a nice flow chart.

But the text asks a crucial question.

Is this just a theory for linguists to draw on chalkboards, or is it psychologically real?

Yes.

Does it actually happen in our heads?

Do our brains actually sit there and unwind these sentences back to the kernel in real time?

It sounds exhausting, doesn't it?

Like we're doing algebra every time we speak.

But the evidence suggests we actually do.

The text details an experiment by Miller that supports this.

Miller's experiment.

Right.

He had subjects memorize different sentence types, kernels, passives, negatives.

And he found something fascinating called the shift toward the kernel.

When people memorized a complex sentence like, the ball was hit by the boy, which is passive, and then they started to forget it a little.

They didn't just forget random words.

No.

They didn't say the ball hit something.

They tended to simplify the sentence back to the kernel.

They would say, the boy hit the ball.

Exactly.

They stripped away the passive transformation.

That is wild.

It's like the brain naturally wants the default to the factory settings.

It implies that we store the core, meaning the kernel, plus a mental tag for the transformation.

So your brain stores content, boy hit ball plus tag,

passive.

When memory fades, the tag falls off first, and you're left with the kernel.

There was another experiment that illustrated this mental load concept even more clearly.

The overflow experiment by Savin and Purchinok.

I love the design of this one.

Imagine your short -term memory is a bucket with limited space.

OK, I've got the bucket.

The researchers gave subjects a sentence to remember, plus a list of unrelated words.

The task was to remember the sentence perfectly, and then as many words as possible.

So if the sentence takes up a lot of space in the bucket, fewer words will fit.

They will overflow and be forgotten.

Exactly.

And the results were crystal clear.

Complex sentences, passives, negatives, questions caused more unrelated words to be forgotten.

A passive negative sentence took up significantly more mental storage space than a simple kernel sentence.

So the ball was not hit by the boy, physically or cognitively, takes up more room in your head than the boy hit the ball.

Because you have to store the kernel, plus the instruction tags for negative and passive.

It confirms that transformations cost mental energy.

Every time you complicate the grammar, you are burning calories in the listener's brain.

However,

the text does play devil's advocate here.

We can't let Chomsky have all the fun.

It mentions a counter -argument called Ying's depth hypothesis.

Science is never settled right.

Ying Wzil looked at the tree diagrams and said, maybe it's not about transformations.

Maybe it's just about how lopsided the tree is.

Lopsided, what does that mean?

He called it left branching.

Imagine you are building a sentence.

If you put a ton of description before the main verb, you are building a heavy left branch.

Like our example from earlier.

The people who called and wanted to rent your house when you go way next to your R -E.

You have to hold that whole left chunk in your brain before you get to the point.

Ying Wzil argued that this depth, holding all those uncompleted phrases, is what creates the memory load, not the grammatical transformation itself.

It's a storage problem, not a processing problem.

And Martin and Roberts did a study supporting this, showing that deeper sentences were harder, regardless of whether they were passive or active.

But the text offers a rebuttal to that, right?

The transformation theory fights back.

It does.

Savin and Puchinot came back and showed that, with questions like who hit the ball.

Take up extra memory space too.

Who hit the ball.

That's a very short sentence.

It is short and structurally on the tree.

It isn't deep at all.

It's very simple.

Yet it still carries that cognitive load.

Which suggests the depth hypothesis doesn't explain everything.

It doesn't seem to.

There really is something about changing the form of a sentence,

transforming it from a statement to a question that requires extra processing power, independent of the tree's shape.

So the transformation theory still seems to hold a lot of water.

There's something about twisting the shape of a sentence that really taxes the CPU.

It seems so.

The debate continues, as it always does in psychology, but the consensus in this chapter really points toward the psychological reality of these transformations.

So what does this all mean?

Let's try to recap the journey we've been on because we've covered a lot of ground here.

Well, we started with the idea that speech is novel.

We don't memorize sentences.

We generate them.

And to do that, we don't just chain words together like beads on a string.

We build hierarchies.

We build trees.

Correct.

And we use cues like rhythm, affixes, and those little function words to synthesize these structures on the fly.

We build the hub so you can understand the wheel.

We found that memory relies on this structure.

We remember Jabberwocky sentences better than random lists because of that grammatical frame.

The structure gives the nonsense a place to live.

And finally, we saw that there is likely a deep structure, a kernel of meaning that we manipulate with transformations.

And every time we twist that kernel into a passive or a negative, we pay a small tax and mental energy.

It really makes you appreciate the incredible complexity of what we're doing right now,

just chatting.

It is the most sophisticated software in the known universe running on biological hardware.

Before we sign off, I want to leave the listener with one final provocative thought from the text.

We've talked so much about syntax and grammar and structure.

It feels very mathematical.

Very mechanical.

But the text ends by pivoting to the final frontier, meaning.

Yes, this is crucial.

We can map the trees and count the transformations.

But at the end of the day, the mind cares about making sense of the world.

The text compares two sentences to prove this point.

Sentence A, accidents kill motorists.

Sentence B, hunters simplify motorists.

Hunters simplify motorists.

Now, grammatically, hunters simplify motorists is a perfect sentence, noun, verb, noun.

It has the exact same tree structure as accidents kill motorists.

But it's incredibly hard to remember.

My brain doesn't want to hold onto it.

Why?

Because it is semantically anomalous.

It doesn't mean anything in our world.

Hunters don't simplify people.

Accidents kill motorists is sticky because it aligns with our knowledge of reality.

So syntax is powerful, but it ultimately serves the master,

which is meaning.

Precisely.

The mind uses both grammatical rules and semantic knowledge to construct reality.

We can't separate the grammar from the world it describes.

We use the structure to find the meaning.

But if the meaning isn't there, the structure just sort of collapses.

A huge thank you to the learner for joining us on this deep dive into your own mind.

Hopefully the next time you hear a sentence, you'll picture that little tree growing in your head,

organizing the chaos.

And maybe cut yourself some slack if you forget a passive negative sentence.

You know, it's taking up a lot of bucket space.

Exactly.

This deep dive was brought to you by the last minute lecture team.

Thanks for listening and we'll catch you on the next one.