Original subtitles

Viewers like you make this program possible.

Support your local PBS station.

Have you ever wanted to know

what your dog is thinking?

Water.

Now some dog owners say

they do know. Walk.

Thanks to buttons that seem

to give dogs the ability to talk.

Snuffle mat.

It's, like, how is this even possible?

Come outside.

Some dogs are saying so much

that scientists have decided to study them.

Button.

Treat.

Yeah, so good-- go get it.

We need to look into this.

They've set up the largest

animal communication study in history...

Play, play, play.

...with thousands of dogs from around the world.

Later.

Come. Come.

Play. Play. Play.

What we want to see is,

where can we take this and how far can dogs go?

Dog on site!

We know we can train dogs to do incredible things.

But could we really teach them

to speak our language?

Snug, snuggle.

You want a snuggle? Come here.

She knows what she's saying.

She's communicating with me.

Treat. Fridge.

Treat. Treat. Treat.

I know what dogs are capable of

in terms of their linguistic ability.

But I think that people

read a little bit too far into things.

Hungry.

So, what's really going on here?

Outside.

Are they just pushing our buttons?

I love you.

Or can dogs talk?

Food.

Right now, on "NOVA."

This is Stella,

an Australian cattle dog/ Catahoula mix.

It's late, but Stella isn't ready for bed.

Yes.

Park.

Want.

Stella appears to be talking.

You want to go to the park?

It's a little late to go to the park.

This video has been sped up a little,

but nothing has been edited in or out.

Do you want to go for a walk instead?

Yes.

Want.

Outside.

Okay, we'll go for a walk.

Let's get your leash on.

Leash.

On.

Stella has been taught to use word buttons.

Walk.

Her owner, Christina Hunger,

can record different words on each of the buttons.

Stella.

Stella is the world's first talking dog.

It's a big claim.

So what makes Christina think it's true?

Christina.

Christina worked as a speech therapist

and used communication devices with children

who struggled to speak.

What?

It was my job as a speech therapist

to find other ways for kids to talk

if they couldn't use their verbal speech.

When I first brought Stella home,

I had so many observations

about how well she was communicating

with her gestures and her body language.

She would whine to get our attention.

She would paw at her water dish when it was empty.

She would look at me, then look to an object that she wanted.

This is all so similar to how kids start learning language.

And it got me thinking, if she's already communicating

so similarly to kids right before they start talking,

what would happen if she had another way to talk?

Christina got some buttons for Stella

to see what would happen.

I was pushing the button as I was saying the word.

Outside?

And for the first couple of weeks,

she didn't pay attention to the buttons at all.

Even when Stella became interested in the buttons,

it took a while before she managed to press one.

You're so close, Stella.

But finally, she cracked it.

And the next morning, Christina caught it on camera.

Outside!

Let's go outside!

I was ecstatic.

And then when she went outside,

she went to the bathroom right away.

So, to me, the most important part

was seeing the communication intent behind it.

Outside.

She was using a word appropriately and functionally.

Stella is now six years old.

What do you want, girl?

And uses far more buttons than the three she started with.

On Stella's main board here,

we have some places that she likes to go.

Park.

And... Beach.

The beach.

Where.

If she wants to know where we're heading out.

Happy.

As Stella started to use her buttons more,

Christina began sharing her videos online.

We left our apartment one day and we set up a camera

just to see, you know, what she would do while we were gone.

Where?

I had never heard her made a sound like that.

It was heartbreaking.

Stella's videos went viral

and have turned her into a social media phenomenon.

We're joined now by Christina Hunger,

a speech pathologist in San Diego,

owner of the very clever Stella.

Outside.

And so, Christina, what does Stella say the most?

She definitely says "outside" the most.

She absolutely loves being outside.

Outside.

As news of Stella and her word buttons has spread...

No, stand up.

...other so-called talking dogs have flooded social media

with their seemingly amazing feats of communication.

Mommy.

These dogs seem to be using our words.

Box.

So is that enough to say they're talking?

Paw.

Do they really understand what they're saying?

Walk, walk, walk.

Or is their button-pushing just random?

Love you.

I love you, too, Cop.

I love you, too.

Federico Rossano is director of the Comparative Cognition Lab

at the University of California, San Diego.

As dogs using word buttons

gained a following on social media,

Federico was approached to study them.

But he was skeptical.

I most certainly did not want to get involved in it.

I have said no so many times.

One of the thing about social media

is that we try to portray ourselves

always happy and excited

and at the best of our life.

Let's go!

And we usually don't post

all the things that go wrong

and all the accidents.

If dog owners only post clips of their dogs

pressing word buttons that seem to make sense

and ignore all the random presses,

it would be completely misleading.

The risk of social media, of course,

is that because there is, you know, money coming with it,

there is attention coming with it,

there might be a way of exploiting social media,

so creating clips that you think would attract more attention.

Treat.

But despite Federico's concerns about social media,

he saw some clips that stunned him.

Mad.

Why are you mad?

Why mad?

Ouch.

Where is your ouch?

Where ouch?

Stranger.

In your ear?

Where stranger?

Paw.

In your paw?

Let me see your paw.

Okay, I'm gonna put this down and check.

This was the stranger in her paw.

She's got a mat between her-- ow!-- toes.

It looked as if they were having a conversation.

And I was, like, wow, if that is possible,

where, like, a human can talk

and the animal can just respond with the buttons,

that would be amazing.

What we want is to look at these anecdotes

and think, is this even possible?

And if it has not been trained extensively,

if it seems to be spontaneous,

then the next question is,

can we elicit it in a more systematic way?

Okay.

Pour it, pour it there?

The trouble is, past efforts to teach animals

to use our language have been incredibly controversial.

Which color?

That one? The blue one.

The way it was done was basically,

a researcher would get one animal,

and then train this animal to learn,

for example, sign language,

or to learn how to use other devices.

Sometimes it looked like something extraordinary,

but if you only have one animal, it's very hard to tell

how that came about

or what that really means

from the perspective of cognition

and what that means about how they're thinking.

After several studies, focused on one or two animals

taken out of their natural habitats,

and with concern about researchers

over-interpreting results,

the entire field of research was called into question.

Eat. Eat.

So when the button dogs emerged on social media,

many of Federico's colleagues

thought he'd be unwise to get involved.

I had scientists calling me

at the beginning of this study,

saying, "Do you know what you're getting yourself into?

You might actually lose your job because of what you're doing."

But this time,

Federico realized that instead of all the research

being based on one or two highly trained subjects...

Ball.

...there was a huge pool of button-pressing dogs.

So he decided to study them.

What happens if you start tracking thousands of animals

at the same time?

He asked Amalia Bastos,

a researcher in animal cognition,

to join him.

Where can we take this

and how far can dogs go?

We need to actually test things out empirically

to find out what's going on.

Federico and Amalia have been

recruiting participants from around the world.

WOMAN

While early contributors in the study

used simple spreadsheets to log their dogs' button presses,

many owners are now using buttons

that automatically record which words are pressed and when.

Working.

Every time a button is pressed...

Snuggle.

...that goes straight into the app,

and then we can download through the app

the kind of data that has been collected.

And what we ask you to do

is to basically tell us, was it you

or was the dog?

As of now, I think we have approximately

6,500 people using the app.

Play.

We know that some dogs seem to make sense some of the time

with their button presses.

But the data that Federico is collecting

will help his team... Outside.

...understand the bigger picture:

do the thousands of dogs taking part

know what they're doing with buttons?

Or is most of it just random?

Later.

First, Federico needs to check

that the dogs actually understand

the words on the buttons when they hear them.

Good boy.

His colleague Amalia has visited several dogs

with a team of researchers to conduct an experiment.

This is a clip from our first

experimental study.

This is Mokai.

You're so calm, hm?

Federico and Amalia need to test whether a word

generated by a button

means the same thing to a dog as when that same word

is spoken by a person.

Who's having some water?

Because when we say words,

we add all sorts of extra details:

pitch, tone, body language.

You are the messiest drinker.

But a word from a button is the same every time.

The first study had to be:

do the dogs that are trained with these buttons

understand the meaning of the words

that are associated with it?

And can we show that they would respond

not just to their owners,

but they will respond to others pushing those buttons?

All right.

This is Kai, trial number four.

The experimenter is wearing headphones,

so that she could not hear the word that is produced

once you push that button.

And the instruction that she has in this case

will be something like,

push the red button.

Play.

The research team wore headphones

and owners were blindfolded,

to make sure neither could inadvertently give clues

about which button had been pressed.

And then you see the dog goes straight for the toy basket,

retrieves the ball,

like, "Let's play."

And so it's contextually appropriate

with respect to what button had just been pressed.

It's the simplest possible test.

58 other dogs were also tested.

Chrissy!

Chris!

Outside.

And the results showed that the dogs responded appropriately

to the "play" and "outside" buttons...

Outside.

...with no other cues,

just as they'd respond

to those words spoken directly by a human.

Play.

That was critical for us,

because if they did not understand

the meaning of words...

Play. ...if they did not respond in a way

that would suggest they understand it,

then there's no point in moving forward with any other research.

The dogs didn't react appropriately

to hearing a "food" button.

Scout!

But it's clear that they can respond to at least

some of their buttons in a useful, communicative way.

Play.

But that doesn't tell us whether they could use them

correctly themselves.

We know that dogs can respond

appropriately to words,

but that could just be conditioning,

that could just be training.

So whether or not they understand

what the word itself means

is much harder to test.

Arik Kershenbaum studies animal communication

at the University of Cambridge in England.

So this ability to hear the word "sit,"

and to know that if they sit, they will get a reward,

is a very, very powerful mechanism.

But just because an animal responds correctly

to a particular command that is given in language

doesn't mean that they understand the language of it.

Whether they're responding to commands

or pushing buttons themselves,

the point is,

it's incredibly hard to know

how much dogs actually understand

the words they're hearing or using.

Good boy-- you gonna try not to bark at stuff today? Yeah?

Can you go to bed?

This is Rohan.

Yeah, good boy.

Amalia's 23-month-old Australian shepherd.

Ready?

Break.

Button.

Treat.

Yeah, so good, go get it.

Thanks.

Love you.

Good boy.

Go to bed.

So what we saw looked like a conversation,

where the dog understood what it was saying,

and we had a back-and-forth.

In actual fact, these were

trained interactions, where the dog

was actually just trained to press the buttons as a trick.

Button.

Treat. Treat. That's the one.

He doesn't actually know what that button means.

Button.

It could have had a bird chirp or it could have had,

you know, siren sound.

It could have been literally anything,

and he would have performed the behavior exactly the same way.

Love you, thanks, love you, love you.

You know, the fact that they're

human-language words is completely arbitrary.

Love you. Thanks.

It's just associative learning.

Rohan has learned

to associate an action, like pressing a button,

with an outcome, like getting a treat.

Treat. Treat. Treat.

That's so good, I know.

He likes treats, so he keeps pressing the buttons.

It's known as "associative learning"

or "conditioning." Button.

Treat. Yes.

But it doesn't mean he knows what the words

"treat" or "thanks" or "love you"

actually mean.

Everything in life starts through associative learning,

and actually, humans learn through associative learning.

Good boy.

Associative learning is common to pretty much all animals.

Break.

And it doesn't require

any sort of complex cognition that we should be impressed by.

Treat.

You could easily teach a mouse to press buttons.

Okay, you got that one-- button.

And that's the big problem.

Just because they can press buttons that say words

doesn't necessarily mean

the dogs have any real understanding

of what those words mean.

Perhaps they just enjoy the reward they get,

whether that's a treat or just some attention.

Good boy.

Down.

So could dogs learn to use different buttons

to get different things?

Could they choose which ones to press

because of the specific words attached to them?

It's something Federico is finally starting to answer.

He's gathered data on the word buttons being used

by an astonishing 10,000 dogs in almost 50 countries.

It's the largest animal communication study in history.

Just to give you an example,

in the last three months,

we have received

more than one million button presses

from dogs in our pool.

All done.

All that data can help to show

whether dog button use is random or intentional.

I love you.

Look.

In principle, the dog could just be spamming

the, the soundboard, just coming there,

touching buttons randomly.

Love you. Treat.

And if it's random,

we should stop doing this research, right?

Why are you do, looking into something

that's completely random?

And what if the dogs are just pressing buttons

that are easy to reach?

What can happen, of course,

is that the dog might just end up pushing, for example,

these two buttons

that are close to each other

just because they are adjacent to each other

and it's just easy for them.

And so, it is very important for us to know

not just which buttons they have, participants have,

but also how they're laid out, so that we can, for example,

run analysis to see

the dogs seem to be actually doing something intentionally.

Federico's analysis of over ten million button presses

by thousands of dogs over almost two years

shows that these are the most popular words.

Of course, all the words on soundboards

are initially put there by humans.

But these are the ones that the dogs are pressing most.

And among these words,

Federico has found that some are much more popular.

The dogs are choosing to press them more than others.

So here, the size of the word is representative of the number

of presses across the population.

And "play," "outside," "food" are the buttons that the dogs

tend to press the most,

which is very much in line with being a dog.

You can also see "water," "potty," "treat," "all done."

They don't use a lot, other words that might not be

quite in line with what we expect,

like "training" or "mad or "friend."

The fact that some words are used more than others...

Bed.

...shows that across the population,

the button use isn't random.

Help.

If it was random, you would just have all the buttons

that you have on the soundboard

have an equal probability of being pressed.

So across the data set,

dogs definitely show preference for certain words.

Outside.

Seemingly, things they want:

to go outside, to get food, to play.

But should we be impressed or surprised by that?

We know a lot of animal communication is asking others,

can be humans or non-humans, to do things for them.

In principle, there is no difference

between pushing the buttons and saying, "Outside,"

and trying to tell you by scratching on the door

that they want to go outside.

It's another way of communicating.

So could dogs actually use buttons in a way

that goes beyond what they can already communicate?

Outside.

Could they start to use language?

All animals communicate.

All animals communicate

and can understand communication,

but not all communication is language.

Humans communicate using language.

Obviously, this is our big difference

from all other animals.

My definition of language,

just a simple rule of thumb that seems to work really well,

is, language is the ability or at least the potential

to convey an unlimited number of concepts.

Doesn't matter how you do it,

but what makes humans special

is that we can communicate to each other

anything, any concept.

There's no end to the books that will ever be written

or the ideas that we can convey,

and no animals that we know of can do that.

No one believes dogs could communicate

absolutely any concept to us using words.

But how far could they go on the spectrum

between communication and language?

I think really, communication becomes language

when you cross that boundary from a signal

that manipulates other individuals

to a system by which information is exchanged between individuals

who understand that their conversation partner

is also participating in that conversation.

Could dogs use words to do more than what they already do

with body language and vocalizations?

Could they communicate about more than just what they want?

The thing is, dogs could be uniquely positioned

to learn our language because of how they evolved.

All dogs evolved from wolves, their wild ancestors.

More than 10,000 years ago,

some wolves were taken in by humans and domesticated.

Over generations spent living with and around us,

they evolved in different ways, according to how we bred them.

We selected traits that we valued,

leading to breeds that were smaller and cuter,

or faster and slimmer,

or bigger and stronger.

And of course, over time, we favored dogs

that could communicate well with us.

Dogs that were good at understanding

what we're trying to say.

There does seem to be something very special

about what happened during that domestication process

that led dogs to be excellent understanders of humans,

and humans to be excellent understanders of dogs, as well.

That communication between humans and dogs

can be so clear...

...that we can form incredibly constructive

working relationships with them.

Yeah!

Amritha Mallikarjun works at

the University of Pennsylvania's Working Dog Center.

I think the human-dog relationship is so special.

The way that we

work together as human-dog teams

is just unlike any other

relationship that you have.

Amritha and her colleague Emma Gaalaas Mullaney

help the dogs here to go on

to careers in search and rescue.

It's all about training,

and training is all about communication.

Dog on site!

Come on, go find.

Today, they're working with Andy,

an 11-month-old black Labrador,

who has to find someone hidden in the rubble.

Andy, turn! Good boy!

Go find!

Because there are always people

standing around on the rubble pile,

he knows not to find the people that he can see,

but to find hidden people, specifically.

When Andy finds his target,

he starts barking,

and he won't stop until his team knows

what he's done.

Yes!

What a good boy!

That was so good, yeah, yeah! Whoo-hoo!

There is a remarkable array of skills on display.

Good job, buddy!

But they are only helpful to the team

because Andy has learned what's expected of him.

These dogs could not do the work they do

if we didn't have the strength of our communication.

The communication between dogs and humans is

how we save lives.

That was so good! That was so good!

In addition to word forms,

dogs are also paying attention to the way our body moves.

They're paying attention to where we're looking.

They're paying attention if we're pointing at things

or if we're gesturing at things.

They're paying attention to the tone of our voice.

So they're reading things off of us all the time,

you know, even as subtle as an arm gesture.

With this extraordinary ability and desire of dogs

to communicate with us,

could they be better-suited to learning our language

than the animals that we've tried to teach in the past?

Because there are some dogs

that seem to have an incredible ability

to recognize large numbers of words.

This is Gaia, a six-year-old border collie.

And this is Isabella Ruiz, Gaia's owner.

While most dogs are good at learning action words

like "sit" and "stay,"

very few are able to easily learn the names of objects.

But that's not the case with Gaia.

Gaia has 215 toys now.

And she knows all of them by name.

It's a lot of toys now for us to remember, so,

we did a catalogue of the toys to remember their names,

but Gaia always knows the toys.

So yeah, I think her memory is better than ours.

Isabella enrolled Gaia in a study

called the Genius Dog Challenge.

It's run by Claudia Fugazza,

who studies canine cognition.

She's traveled here all the way from Hungary,

where she works

with one of the world's leading canine research groups.

Gaia is one of the, what we call the gifted word learner dogs,

and these are dogs that have a very special talent

for learning the name of toys, and they learn very fast.

Only a few dogs around the world seem to have this capacity.

Go, go!

One of the most capable dogs

at learning object names

was Chaser, a border collie

who knew the names of over 1,000 toys.

So far, Claudia has found fewer than 50 dogs

around the world

who are good enough at learning object names

to take part in her study.

Hello! Hi, Gaia, hello, girl!

She's traveled here to put Gaia to the test.

Thank you.

So I got something for Gaia.

Gaia!

Look at here-- maybe there's something for you.

That's for you.

Yeah.

Coffee?

Coffee.

A little test shows just how quickly

Gaia can learn the name of a new toy.

RUIZ

Hey?

RUIZ

Good girl, she got it.

So she learns really quick.

That was a matter of

couple of minutes playing,

and she already got this new name.

But Claudia already knows

that Gaia is exceptionally talented

at learning the names of toys.

Coffee!

She wants to find out whether Gaia's cognitive skills

go a step further.

Gaia knows that "tug" and "fetch" are games to play.

Last week, she was given two new toys,

one to tug and one to fetch.

New toys.

Can she figure out which toy goes with which game,

based on how she plays with it?

This one, I will play like this.

Isabella hasn't given the toys any names

and hasn't said the words "tug" or "fetch"

when playing with them.

This one, we've been playing like this,

but I, I don't say the name of the category, so, yeah.

Is Gaia smart enough

to figure out which is a "fetch" toy

and which is a "tug" toy,

even though these words haven't been used with these toys?

So are they able to sort those items

into the tug or the fetch category

just based on the way you play?

RUIZ

...Fetch?

Come on, come on, come on.

Good girl! Good girl!

Good girl, you got the fetch. Good girl!

Good girl!

Gaia.

...Tug?

Good doggy! Good girl!

Good doggy. Yeah.

Yeah. That's a good girl.

Yeah! Go, go, go!

Yeah!

Well, I have to say, I had no idea how this would go.

But that was really, really exciting.

Totally, I believe this is a big step

in our understanding of dogs' minds.

So, she was able to assign this toy to the named category

without ever having heard that name paired with that toy.

What Gaia has done seems different

from just simple conditioning or associative learning.

For sure,

Gaia could not have associated the name "fetch" or "tug"

to the specific item,

because she has never heard that name

while seeing that toy or while playing with that toy.

This is not possible.

She could not have learned that this way.

It seems some dogs can understand

that one word can refer to a whole category of objects--

not just one individual thing.

But that's just words that they're hearing.

Could some dogs also understand

that they could use a button word

in different situations?

Meet Parker.

She's a beagle cross and a prolific user of word buttons.

Water.

So Parker is a bit atypical.

When we first got them, it took her

about 30 seconds, maybe, not even, to press her first button.

And by the, you know, end of that kind of evening,

she had pressed, at least experimentally, all six of them.

Snug, snuggle.

You want to snuggle? Come here.

Her owner, Sascha Crasnow,

enrolled Parker in Federico's study

as soon as the buttons arrived.

She definitely says things with the buttons

that she would not be able to communicate to me otherwise.

Snuffle mat.

So a snuffle mat has,

you know, sort of these layers or these little pockets,

little places you can kind of hide food,

and then a puzzle...

...is sort of what it sounds like.

It's another way to kind of do mental enrichment

while they're eating.

But if I give her food in a snuffle mat,

and she wants a puzzle that day,

she will refuse to eat that food.

And if I take the exact same food and pour it into a puzzle,

she will then eat that food.

And if I didn't have the buttons,

I would think she was just off her food, right?

That maybe there was something wrong with it.

But no, she's just very particular

about the experience she wants while she's eating.

Of course, there may always be a temptation

for owners to over-interpret their pets' use of word buttons.

Food.

Ball.

But Parker certainly seems more capable than most.

One time, when I was

kind of just out for maybe an hour or so,

she pressed "potty."

Potty, potty.

And then, on the video, you can see that she throws up.

And then she goes and presses "help."

Help.

Amalia, Federico's colleague,

felt that use of the word "help"

suggested greater cognitive abilities

than most of the button words used by dogs.

So with most of the buttons that dogs have,

like "food" or "play" or "walk,"

these are simple associations

that are related to one particular context.

Help.

"Help" is interesting,

because it's a little bit more general

than a lot of the simpler buttons

that these dogs might have.

She decided to set up an experiment

called The Impossible Task.

Come on!

Sascha and Parker took part.

Sit-- wait.

So the way that the study is set up is,

we have this mat on the floor here,

and it's got a Tupperware in the middle.

First, the owners give the dogs opportunities

over four different trials

to obtain a food reward from inside the Tupperware,

and at that point, the lid is open.

And at all these stages, the dog can knock off the lid

and eat the food inside the Tupperware

without much difficulty.

Until, in the very final trial,

the fifth trial, the Tupperware is completely locked,

and that's the point at which the dog might need help

to access the food.

The owners are not allowed to speak to the dog,

point or gesture at the dog,

look at their dog-- nothing at all.

I didn't know what Parker would do

in that circumstance,

whether she would use the buttons or not.

We can see she's immediately gone to Sascha,

and she's climbed onto Sascha's back,

trying to get her attention.

But what we're interested in is,

can these dogs then make that leap and use a word?

Help. Look.

She presses "help," "look,"

and then she looks to Sascha.

Parker had never before been presented with this scenario,

and so had never used the words "look" or "help"

in this context.

Help. Help. Look. Look.

So because it's the very first time that they've seen this,

it tells us two things.

It tells us the behavior wasn't trained

and it tells us that the dog is having to think about what this

requires from them,

in terms of communication to their owner,

and that they need to generalize a button

that they've not used in this context before.

Help. Help. Help.

It seems from this experiment... Look. Look.

...that Parker understands that the word "help"

can be used in a broad range of scenarios.

Oh, hi! Do you need help?

There you go.

As well as using words in new, untrained ways,

a huge step would be if dogs were using words

to do more than just ask for things.

What we're actually interested in

is whether dogs might be able to use

the soundboards not just as a vending machine,

where they're requesting things from humans,

but actually understand it as a communicative device.

So humans do not just request things,

humans can inform.

It's a thing that humans do all the time.

Um, toddlers start doing it-- they will point at the moon

and say "moon," and, you know, they're not requesting the moon,

they're just letting you know the moon is there.

Right or wrong,

it's something that some dog owners believe

their dogs are capable of.

Though of course, they may have a tendency

to read too much into their dogs' button presses.

Parker uses the buttons to narrate a decent amount.

Um, she has narrated me watering the plants quite a bit.

She'll narrate also things she does sometimes.

Settle.

She's pressed "settle" before,

and then just gone into her bed and lie down.

She didn't get anything from me for it.

She just pressed "settle"

and then she settled.

It's impossible to know for sure whether Parker

really was commenting on what she was doing.

But other dogs, like Copper,

a three-and-a-half-year-old Labrador,

also seem to be using buttons to do more than just make requests.

Outside. Pool.

Outside pool.

Outside pool maybe tomorrow.

Soon.

Outside pool later.

Outside. Pool.

Later. Later.

Sad.

Oh, my goodness-- Copper.

It looks like Copper was using her buttons

to express emotion.

But of course, we don't know for sure.

An animal could learn that pressing the "sad" button

gets attention.

And, and once they've learnt that,

they can manipulate these, these different concepts

in, in quite complex ways,

without understanding there's a semantic meaning behind it.

Outside pool later.

From the data being submitted by dog owners,

Federico has found that many believe their dogs

are using buttons to do more than just ask for things.

While 50% of button presses

were categorized as requests,

more than five percent of presses were believed by owners

to be sharing thoughts and feelings.

Six percent were informing,

as well as 6.4 percent described by owners as "narrating."

Crucially, these aren't

scientifically validated reports.

They're just what owners believe their dogs are doing.

Knowing for certain what a dog means

when it uses words, or thinks when it hears words,

is much harder.

But in Budapest, Hungary,

Marianna Boros wants to find out what's happening

inside a dog's brain

when it hears words it knows.

For most people, we have

a mental representation

of things from the external world,

so we have a mental representation of an object,

some kind of memory of that thing.

We can imagine it in a way that, we hear the word "ball,"

we will see in our mental imagery a ball.

Marianna wants to test

whether dogs' brains might work similarly.

When a dog hears a word,

does it conjure up an image of that thing in its head?

Hello.

To find out,

she's going to study the brain activity of Demi,

an 11-year-old Swiss shepherd dog.

So right now, I'm putting on

the electrodes on Demi's head.

These will measure the brain signal.

Demi's owner, Timi,

has brought in some of Demi's toys.

Demi hears a message telling her what toy

she's about to see.

RECORDED VOICE

Sometimes, Timi shows her that toy.

RECORDED VOICE

But sometimes,

Timi shows a different toy

than what Demi is expecting.

The idea is that if the dog sees a different object,

it should be somehow surprised,

and this surprise effect

should be visible in the E.E.G. signal.

RECORDED VOICE

What we can see here is that if a dog saw

an object that matched the word that he just heard,

this is the response that we get.

But when we violate their expectations,

so we show a different object,

then we get a more positive response

in the E.E.G. signal.

Marianna believes

that the difference in electrical activity

indicates that when dogs hear a word,

it can evoke a mental representation of an object.

We can claim, based on this study,

that when they hear the name of an object,

they activate something, the so-called memory of this object.

So we think that the dog is actually surprised

when it sees a different thing,

but it is surprised because it has an expectation

and because it understands the meaning of that word.

That's a qualitatively different way

of using language,

as opposed to associative learning.

So yes, I would say

there's a possibility that they understand words

similarly to us.

There are limitations in interpreting

E.E.G. signals like these

and figuring out exactly what they tell us

about how much an animal really understands.

But if dogs do understand words

in similar ways to us,

could they start putting them together

and using them in more complex ways?

To understand that,

we need to look at the dogs that use a lot of buttons.

Come. Outside.

Okay, we can go outside.

Come on, girl.

Of all the dogs taking part in Federico's study,

most use just a handful of buttons.

But there are at least 30 dogs

who routinely use at least 50 different buttons.

The holy grail would be to find dogs that are capable

of using a limited number of words

to express new ideas

that they don't have specific buttons for.

We humans have

a limited vocabulary.

But the reality is, there's no constraint

on how many different things we can express.

The only way that we can get that unlimited ability

from a limited vocabulary

is by combining those different terms,

those different words, that vocabulary,

in different ways to create new concepts.

This ability to say anything with limited words

is known as productivity.

I can give you an example.

My two-year-old kept telling me

that he wanted to be dressed as a green lion,

and I was, like, "What are you talking about?

There is no green lion."

And at some point, he sees somebody dressed as a dinosaur,

points to it, and says, "Green lion."

And I'm, like, "There you go.

"You don't have a word for dinosaur.

"But by saying 'green lion,'

"you're getting as close as you can with your vocabulary

to telling me, 'This is what I want.'"

Linguistic productivity is absolutely central

to the way that humans do language.

So that cognitive ability is very, very powerful

and very, very rare in the animal world.

It would be a huge step up in terms of cognitive ability.

So could some dogs be showing signs of it?

In order to look at productivity,

you need to have dogs that can combine at least two buttons.

As of yesterday, we have 790 dogs in our pool

that do multi-button combinations.

It is possible that out of this 700,

almost 800 individuals,

you can now look at some,

and maybe there's gonna be one, two,

three, five, ten, 20 that actually have

the ability to kind of refer to things

that you don't have a word for

or by combining other signals.

We have anecdotal reports from some of them

that they seem to be doing it.

One time, when Stella's "beach" button broke,

she went right over,

looked exactly where her "beach" button was,

she kind of sniffed it, looked at the empty space,

and then immediately after,

said, "Help," "Water," "Outside."

Help. Water.

Outside.

That was something that had never been taught.

I had never referred to the beach as "water outside."

But she put those together

in order to communicate that concept to me

when her "beach" button wasn't available anymore.

It's impossible to know

whether Christina's interpretation is correct.

But clips like these are helpful for Federico,

to highlight possible areas for future research.

Water. Outside.

This is the reason I am doing it,

even though people told me I might lose my job.

It's because I've seen things like that.

I've seen animals putting together two or three buttons

in ways that look like they were producing sentences.

But of course, you should not stop there,

because that thing could have been trained.

We'll go to the beach.

It might look like productivity,

but we don't know for sure that it is.

I do find interesting

that she pressed buttons that are near "beach."

Help. Water.

I would have loved

to know if she would have done the same

if, say, "beach" was on the other side of the soundboard.

Although Stella had never before used those buttons together,

testing what she meant by them after the event is impossible.

After that set of buttons is hit,

if Stella then goes to the beach,

now that set of buttons

definitely means "beach."

So then it's almost impossible for us to know

what, the first button press occurrence,

what Stella had intended, because now

it always means "beach."

So, can dogs talk?

I think Parker can communicate using

English-language words.

Do you want to do training?

No.

Okay!

We won't do training, then.

If you want to call that talking, great.

I say that Stella talks

because she's using words to communicate.

Outside.

The fact that it's not a trained behavior,

and that it's independent,

and that she can use words

to mean a lot of different things

indicates language, not just a conditioned trick.

It's not surprising that owners,

with their close relationships and strong communicative bonds,

might feel their dogs are talking.

Love you.

Stella.

Love you, Stella.

But scientific understanding

of their linguistic capabilities

is still in its early days.

Bedtime.

We're very much at the beginning of what we're doing

and we need to be ready

to collect data for years and years.

Out, outside.

And to me, the exciting part is

that we now get a better sense

of what might be happening out there.

Outside.

Maybe you can see something that nobody else could see before,

because we just didn't have access to all this data.

Chewy.

We're really at the start of this journey

of understanding animal cognition,

animal communication, and where that lies

on that spectrum between language and non-language.

Where. Bye.

Where...

Bye.

Where bye?

Are you asking where you're going?

You're going to go see Natalie!

It's clear that dogs will never be capable

of using our language to generate the rich,

grammatically correct, infinite range of sentences

that humans can create.

Does it matter

if the sentence you're saying is grammatical or not,

as long as you're getting your point across,

and you're actually benefiting from that two-way communication?

I don't think so.

Whether dogs are using language or not,

with a new tool in their communication kit

and access to our words,

perhaps the close bond between humans and dogs

could become even closer.