NOVA

(dogs barking, panting) NARRATOR: Have you ever wanted to know what your dog is thinking?

AUTOMATED VOICE: Water.

NARRATOR: Now some dog owners say they do know.

AUTOMATED VOICE: Walk.

NARRATOR: Thanks to buttons that seem to give dogs the ability to talk.

AUTOMATED VOICE: Snuffle mat.

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

AUTOMATED VOICE: Come outside.

NARRATOR: Some dogs are saying so much that scientists have decided to study them.

Button.

AUTOMATED VOICE: Treat.

Yeah, so good-- go get it.

We need to look into this.

NARRATOR: They've set up the largest animal communication study in history... AUTOMATED VOICE: Play, play, play.

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

AUTOMATED VOICE: Later.

Come.

Come.

Play.

Play.

Play.

AMALIA BASTOS: What we want to see is, where can we take this and how far can dogs go?

WOMAN: Dog on site!

NARRATOR: We know we can train dogs to do incredible things.

(barks) NARRATOR: But could we really teach them to speak our language?

AUTOMATED VOICE: Snug, snuggle.

You want a snuggle?

Come here.

SASCHA CRASNOW: She knows what she's saying.

She's communicating with me.

AUTOMATED VOICE: Treat.

Fridge.

Treat.

Treat.

Treat.

AMRITHA MALLIKARJUN: I know what dogs are capable of in terms of their linguistic ability.

(barking) MALLIKARJUN: But I think that people read a little bit too far into things.

AUTOMATED VOICE: Hungry.

NARRATOR: So, what's really going on here?

AUTOMATED VOICE: Outside.

NARRATOR: Are they just pushing our buttons?

AUTOMATED VOICE: I love you.

NARRATOR: Or can dogs talk?

AUTOMATED VOICE: Food.

NARRATOR: Right now, on "NOVA."

NARRATOR: This is Stella, an Australian cattle dog/ Catahoula mix.

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

AUTOMATED VOICE: Yes.

Park.

Want.

NARRATOR: Stella appears to be talking.

CHRISTINA HUNGER: You want to go to the park?

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

NARRATOR: This video has been sped up a little, but nothing has been edited in or out.

HUNGER: Do you want to go for a walk instead?

AUTOMATED VOICE: Yes.

Want.

Outside.

HUNGER: Okay, we'll go for a walk.

Let's get your leash on.

AUTOMATED VOICE: Leash.

On.

NARRATOR: Stella has been taught to use word buttons.

AUTOMATED VOICE: Walk.

NARRATOR: Her owner, Christina Hunger, can record different words on each of the buttons.

AUTOMATED VOICE: Stella.

Stella is the world's first talking dog.

NARRATOR: It's a big claim.

So what makes Christina think it's true?

AUTOMATED VOICE: Christina.

NARRATOR: Christina worked as a speech therapist and used communication devices with children who struggled to speak.

AUTOMATED VOICE: 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.

(whining) HUNGER: 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.

(growling) HUNGER: 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?

NARRATOR: Christina got some buttons for Stella to see what would happen.

(barks) I was pushing the button as I was saying the word.

Outside?

(barks) HUNGER: And for the first couple of weeks, she didn't pay attention to the buttons at all.

NARRATOR: Even when Stella became interested in the buttons, it took a while before she managed to press one.

HUNGER: You're so close, Stella.

(barks) NARRATOR: But finally, she cracked it.

And the next morning, Christina caught it on camera.

AUTOMATED VOICE: Outside!

HUNGER: Let's go outside!

(barks) 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.

AUTOMATED VOICE: Outside.

She was using a word appropriately and functionally.

NARRATOR: Stella is now six years old.

What do you want, girl?

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

HUNGER: On Stella's main board here, we have some places that she likes to go.

AUTOMATED VOICE: Park.

And... AUTOMATED VOICE: Beach.

The beach.

AUTOMATED VOICE: Where.

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

AUTOMATED VOICE: Happy.

NARRATOR: As Stella started to use her buttons more, Christina began sharing her videos online.

HUNGER: 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.

AUTOMATED VOICE: Where?

(Stella whimpering, howling) I had never heard her made a sound like that.

(barking, howling) It was heartbreaking.

(howling) NARRATOR: Stella's videos went viral and have turned her into a social media phenomenon.

CYRIL VANIER: We're joined now by Christina Hunger, a speech pathologist in San Diego, owner of the very clever Stella.

AUTOMATED VOICE: Outside.

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

HUNGER: She definitely says "outside" the most.

She absolutely loves being outside.

AUTOMATED VOICE: Outside.

NARRATOR: As news of Stella and her word buttons has spread... WOMAN: No, stand up.

NARRATOR: ...other so-called talking dogs have flooded social media with their seemingly amazing feats of communication.

AUTOMATED VOICE: Mommy.

NARRATOR: These dogs seem to be using our words.

AUTOMATED VOICE: Box.

NARRATOR: So is that enough to say they're talking?

AUTOMATED VOICE: Paw.

NARRATOR: Do they really understand what they're saying?

AUTOMATED VOICE: Walk, walk, walk.

NARRATOR: Or is their button-pushing just random?

AUTOMATED VOICE: Love you.

WOMAN (chuckling): I love you, too, Cop.

I love you, too.

NARRATOR: 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.

(chuckling) Let's go!

(all cheering, laughing) ROSSANO: And we usually don't post all the things that go wrong and all the accidents.

NARRATOR: 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.

AUTOMATED VOICE: Treat.

NARRATOR: But despite Federico's concerns about social media, he saw some clips that stunned him.

AUTOMATED VOICE: Mad.

WOMAN: Why are you mad?

Why mad?

AUTOMATED VOICE: Ouch.

WOMAN: Where is your ouch?

Where ouch?

AUTOMATED VOICE: Stranger.

WOMAN: In your ear?

Where stranger?

AUTOMATED VOICE: 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.

ROSSANO: 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?

NARRATOR: The trouble is, past efforts to teach animals to use our language have been incredibly controversial.

Which color?

That one?

The blue one.

ROSSANO: 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.

BASTOS: 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.

NARRATOR: 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.

AUTOMATED VOICE: Eat.

Eat.

NARRATOR: So when the button dogs emerged on social media, many of Federico's colleagues thought he'd be unwise to get involved.

ROSSANO: 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."

NARRATOR: But this time, Federico realized that instead of all the research being based on one or two highly trained subjects... AUTOMATED VOICE: Ball.

NARRATOR: ...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?

NARRATOR: He asked Amalia Bastos, a researcher in animal cognition, to join him.

BASTOS: 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.

NARRATOR: Federico and Amalia have been recruiting participants from around the world.

WOMAN (speaking French): NARRATOR: 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.

AUTOMATED VOICE: Working.

ROSSANO: Every time a button is pressed... AUTOMATED VOICE: Snuggle.

ROSSANO: ...that goes straight into the app, and then we can download through the app the kind of data that has been collected.

(phone chirping) 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.

AUTOMATED VOICE: Play.

NARRATOR: 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... AUTOMATED VOICE: Outside.

NARRATOR: ...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?

AUTOMATED VOICE: Later.

NARRATOR: First, Federico needs to check that the dogs actually understand the words on the buttons when they hear them.

BASTOS: Good boy.

NARRATOR: His colleague Amalia has visited several dogs with a team of researchers to conduct an experiment.

ROSSANO: This is a clip from our first experimental study.

NARRATOR: This is Mokai.

BASTOS: You're so calm, hm?

NARRATOR: 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?

NARRATOR: Because when we say words, we add all sorts of extra details: pitch, tone, body language.

You are the messiest drinker.

(all chuckling) NARRATOR: But a word from a button is the same every time.

ROSSANO: 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?

MAN: All right.

This is Kai, trial number four.

ROSSANO: 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.

AUTOMATED VOICE: Play.

NARRATOR: The research team wore headphones and owners were blindfolded, to make sure neither could inadvertently give clues about which button had been pressed.

ROSSANO: 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.

NARRATOR: It's the simplest possible test.

58 other dogs were also tested.

Chrissy!

Chris!

AUTOMATED VOICE: Outside.

NARRATOR: And the results showed that the dogs responded appropriately to the "play" and "outside" buttons... AUTOMATED VOICE: Outside.

NARRATOR: ...with no other cues, just as they'd respond to those words spoken directly by a human.

(calls dog) AUTOMATED VOICE: Play.

ROSSANO: That was critical for us, because if they did not understand the meaning of words... AUTOMATED VOICE: Play.

ROSSANO: ...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.

(people laughing) NARRATOR: The dogs didn't react appropriately to hearing a "food" button.

Scout!

NARRATOR: But it's clear that they can respond to at least some of their buttons in a useful, communicative way.

AUTOMATED VOICE: Play.

NARRATOR: But that doesn't tell us whether they could use them correctly themselves.

ARIK KERSHENBAUM: 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.

NARRATOR: Arik Kershenbaum studies animal communication at the University of Cambridge in England.

KERSHENBAUM: 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.

NARRATOR: 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?

NARRATOR: This is Rohan.

Yeah, good boy.

NARRATOR: Amalia's 23-month-old Australian shepherd.

Ready?

Break.

Button.

AUTOMATED VOICE: Treat.

Yeah, so good, go get it.

AUTOMATED VOICE: Thanks.

(Rohan whimpers, barks) AUTOMATED VOICE: Love you.

(chuckling): Good boy.

Go to bed.

(laughs) BASTOS: 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.

(growls) AUTOMATED VOICE: Treat.

Treat.

That's the one.

BASTOS: He doesn't actually know what that button means.

Button.

(growls) BASTOS: 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.

AUTOMATED VOICE: Love you, thanks, love you, love you.

BASTOS: You know, the fact that they're human-language words is completely arbitrary.

AUTOMATED VOICE: Love you.

Thanks.

It's just associative learning.

NARRATOR: Rohan has learned to associate an action, like pressing a button, with an outcome, like getting a treat.

AUTOMATED VOICE: Treat.

Treat.

Treat.

That's so good, I know.

NARRATOR: He likes treats, so he keeps pressing the buttons.

It's known as "associative learning" or "conditioning."

Button.

AUTOMATED VOICE: Treat.

Yes.

NARRATOR: 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.

BASTOS: And it doesn't require any sort of complex cognition that we should be impressed by.

AUTOMATED VOICE: Treat.

You could easily teach a mouse to press buttons.

Okay, you got that one-- button.

NARRATOR: 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.

NARRATOR: 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.

ROSSANO: 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.

AUTOMATED VOICE: All done.

NARRATOR: All that data can help to show whether dog button use is random or intentional.

AUTOMATED VOICE: I love you.

Look.

ROSSANO: In principle, the dog could just be spamming the, the soundboard, just coming there, touching buttons randomly.

AUTOMATED VOICE: Love you.

Treat.

ROSSANO: And if it's random, we should stop doing this research, right?

Why are you do, looking into something that's completely random?

(button clicking repeatedly) NARRATOR: And what if the dogs are just pressing buttons that are easy to reach?

(automated voice speaking indistinctly) ROSSANO: 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.

NARRATOR: 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.

ROSSANO: 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."

NARRATOR: The fact that some words are used more than others... AUTOMATED VOICE: Bed.

NARRATOR: ...shows that across the population, the button use isn't random.

AUTOMATED VOICE: Help.

ROSSANO: If it was random, you would just have all the buttons that you have on the soundboard have an equal probability of being pressed.

NARRATOR: So across the data set, dogs definitely show preference for certain words.

AUTOMATED VOICE: Outside.

NARRATOR: 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.

NARRATOR: So could dogs actually use buttons in a way that goes beyond what they can already communicate?

AUTOMATED VOICE: Outside.

NARRATOR: 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.

NARRATOR: 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.

NARRATOR: 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.

(howling) NARRATOR: 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.

KERSHENBAUM: 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.

NARRATOR: That communication between humans and dogs can be so clear... (barks) NARRATOR: ...that we can form incredibly constructive working relationships with them.

(Mallikarjun calling out) (Andy growling) MALLIKARJUN: Yeah!

NARRATOR: Amritha Mallikarjun works at the University of Pennsylvania's Working Dog Center.

MALLIKARJUN: 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.

NARRATOR: 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.

WOMAN: Dog on site!

WOMAN: Dog on site!

Come on, go find.

NARRATOR: Today, they're working with Andy, an 11-month-old black Labrador, who has to find someone hidden in the rubble.

WOMAN: Andy, turn!

Good boy!

Go find!

MALLIKARJUN: 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.

(Andy barking) (barking) NARRATOR: When Andy finds his target, he starts barking, and he won't stop until his team knows what he's done.

(barking) WOMAN: Yes!

(cheering) What a good boy!

That was so good, yeah, yeah!

Whoo-hoo!

NARRATOR: There is a remarkable array of skills on display.

WOMAN: Good job, buddy!

NARRATOR: 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.

(Andy panting) MULLANEY: The communication between dogs and humans is how we save lives.

WOMAN: That was so good!

That was so good!

MALLIKARJUN: 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.

NARRATOR: 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.

(sirens blaring, car horns honking) This is Gaia, a six-year-old border collie.

And this is Isabella Ruiz, Gaia's owner.

(speaking Portuguese): NARRATOR: 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.

NARRATOR: 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.

MAN: Go, go!

NARRATOR: 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.

(knock at door) Hello!

Hi, Gaia, hello, girl!

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

Thank you.

So I got something for Gaia.

Gaia!

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

(chuckles, speaks softly) (gasps) That's for you.

Yeah.

Coffee?

Coffee.

NARRATOR: A little test shows just how quickly Gaia can learn the name of a new toy.

RUIZ (speaking Portuguese): Hey?

RUIZ (speaking Portuguese): FUGAZZA: Good girl, she got it.

FUGAZZA: So she learns really quick.

That was a matter of couple of minutes playing, and she already got this new name.

(cooing) NARRATOR: But Claudia already knows that Gaia is exceptionally talented at learning the names of toys.

Coffee!

NARRATOR: 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.

RUIZ: New toys.

NARRATOR: Can she figure out which toy goes with which game, based on how she plays with it?

This one, I will play like this.

NARRATOR: 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.

NARRATOR: 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?

FUGAZZA: So are they able to sort those items into the tug or the fetch category just based on the way you play?

RUIZ (speaking Portuguese): ...Fetch?

Come on, come on, come on.

Good girl!

Good girl!

Good girl, you got the fetch.

Good girl!

Good girl!

Gaia.

(speaking Portuguese): ...Tug?

(repeats with Portuguese) 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.

NARRATOR: What Gaia has done seems different from just simple conditioning or associative learning.

FUGAZZA: 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.

NARRATOR: It seems some dogs can understand that one word can refer to a whole category of objects-- not just one individual thing.

(Ruiz speaks indistinctly) NARRATOR: 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.

AUTOMATED VOICE: Water.

CRASNOW: 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.

AUTOMATED VOICE: Snug, snuggle.

You want to snuggle?

Come here.

NARRATOR: 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.

AUTOMATED VOICE: Snuffle mat.

CRASNOW: 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.

NARRATOR: Of course, there may always be a temptation for owners to over-interpret their pets' use of word buttons.

AUTOMATED VOICE: Food.

Ball.

NARRATOR: But Parker certainly seems more capable than most.

CRASNOW: One time, when I was kind of just out for maybe an hour or so, she pressed "potty."

AUTOMATED VOICE: Potty, potty.

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

(Parker retching) CRASNOW: And then she goes and presses "help."

AUTOMATED VOICE: Help.

NARRATOR: Amalia, Federico's colleague, felt that use of the word "help" suggested greater cognitive abilities than most of the button words used by dogs.

BASTOS: 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.

AUTOMATED VOICE: Help.

BASTOS: "Help" is interesting, because it's a little bit more general than a lot of the simpler buttons that these dogs might have.

NARRATOR: She decided to set up an experiment called The Impossible Task.

Come on!

NARRATOR: 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.

CRASNOW: I didn't know what Parker would do in that circumstance, whether she would use the buttons or not.

BASTOS: 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?

AUTOMATED VOICE: Help.

Look.

BASTOS: She presses "help," "look," and then she looks to Sascha.

NARRATOR: Parker had never before been presented with this scenario, and so had never used the words "look" or "help" in this context.

AUTOMATED VOICE: Help.

Help.

Look.

Look.

BASTOS: 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.

AUTOMATED VOICE: Help.

Help.

Help.

NARRATOR: It seems from this experiment... AUTOMATED VOICE: Look.

Look.

NARRATOR: ...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.

NARRATOR: 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.

BASTOS: 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.

ROSSANO: 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.

NARRATOR: 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.

AUTOMATED VOICE: Settle.

CRASNOW: 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.

NARRATOR: 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.

AUTOMATED VOICE: Outside.

Pool.

WOMAN: Outside pool.

Outside pool maybe tomorrow.

AUTOMATED VOICE: Soon.

Outside pool later.

AUTOMATED VOICE: Outside.

Pool.

Later.

AUTOMATED VOICE: Later.

AUTOMATED VOICE: Sad.

Oh, my goodness-- Copper.

NARRATOR: 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.

(grumbles) WOMAN: Outside pool later.

(grumbles) NARRATOR: 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.

BOROS: 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.

NARRATOR: 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.

NARRATOR: To find out, she's going to study the brain activity of Demi, an 11-year-old Swiss shepherd dog.

BOROS: So right now, I'm putting on the electrodes on Demi's head.

These will measure the brain signal.

NARRATOR: 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 (in Hungarian): NARRATOR: Sometimes, Timi shows her that toy.

RECORDED VOICE (in Hungarian): NARRATOR: But sometimes, Timi shows a different toy than what Demi is expecting.

BOROS: 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 (in Hungarian): 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.

NARRATOR: 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.

(yawns) BOROS: That's a qualitatively different way of using language, as opposed to associative learning.

(whimpers softly) BOROS: So yes, I would say there's a possibility that they understand words similarly to us.

NARRATOR: 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?

(whimpers softly) NARRATOR: To understand that, we need to look at the dogs that use a lot of buttons.

AUTOMATED VOICE: Come.

Outside.

HUNGER: Okay, we can go outside.

Come on, girl.

NARRATOR: 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.

KERSHENBAUM: 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.

NARRATOR: This ability to say anything with limited words is known as productivity.

ROSSANO: 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.

NARRATOR: 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."

AUTOMATED VOICE: Help.

Water.

Outside.

HUNGER: 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.

NARRATOR: 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.

AUTOMATED VOICE: 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.

HUNGER: We'll go to the beach.

BASTOS: 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."

AUTOMATED VOICE: Help.

Water.

BASTOS: I would have loved to know if she would have done the same if, say, "beach" was on the other side of the soundboard.

NARRATOR: Although Stella had never before used those buttons together, testing what she meant by them after the event is impossible.

MALLIKARJUN: 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."

(Parker barking, Crasnow laughing) NARRATOR: So, can dogs talk?

CRASNOW: I think Parker can communicate using English-language words.

Do you want to do training?

AUTOMATED VOICE: No.

(laughs): Okay!

We won't do training, then.

If you want to call that talking, great.

HUNGER: I say that Stella talks because she's using words to communicate.

AUTOMATED VOICE: Outside.

HUNGER: 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.

NARRATOR: It's not surprising that owners, with their close relationships and strong communicative bonds, might feel their dogs are talking.

AUTOMATED VOICE: Love you.

Stella.

Love you, Stella.

NARRATOR: But scientific understanding of their linguistic capabilities is still in its early days.

AUTOMATED VOICE: Bedtime.

ROSSANO: 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.

AUTOMATED VOICE: Out, outside.

And to me, the exciting part is that we now get a better sense of what might be happening out there.

AUTOMATED VOICE: Outside.

Maybe you can see something that nobody else could see before, because we just didn't have access to all this data.

AUTOMATED VOICE: Chewy.

KERSHENBAUM: 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.

AUTOMATED VOICE: Where.

Bye.

WOMAN: Where... AUTOMATED VOICE: Bye.

WOMAN: Where bye?

Are you asking where you're going?

You're going to go see Natalie!

NARRATOR: 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.

BASTOS: 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.

(growling) NARRATOR: 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.