Imagine Max, a well-trained border collie, manages to ignore a squirrel in the park when his owner tells him to sit. His owner says, "Max, stop chasing that squirrel and sit down," and Max obeys. Can dogs learn and understand words the way humans do?
Author
- Juliane Kaminski
Associate Professor of Comparative Psychology, University of Portsmouth
A new study found dogs like Max may have learnt the names of objects (like a squirrel) from overhearing their owners talking. The study is the latest to try and understand whether intelligent dogs and humans can have real conversations.
A widely reported case in 2004 brought this question into the spotlight. Rico, an eight-year-old border collie was the first dog who demonstrated under experimental conditions that he knew the names of over 200 different toys.
Dogs like Rico seem different to other ones. Scientists have a name for them: label-learner dogs. They seem so exceptional, it's easy to wonder if they're learning words in a similar way to humans. Research is starting to give us some answers. But first, it's important to understand how these dogs have been studied.
In 2004, researchers, including myself, wanted to make sure Rico wasn't simply reacting to subtle, unconscious signals from people. So Rico was tested in a room where he couldn't see anyone. He still fetched the correct toys upon hearing the command "Fetch, xy". That meant he was not using visual cues from his owner.
The next big question was whether Rico could learn new name-object combinations the way young children do. Children often learn new words through a process called fast mapping. They hear a new word, look at the options and figure out what it must refer to. For example, if a child knows what "blue" means but not "olive," and you show them a blue object and an olive-green one, they'll probably choose the olive-green one when you ask for "olive".
Rico showed something similar in his behaviour. When researchers placed a brand-new toy among familiar ones and asked for a name he had never heard before, he picked the new toy. He even remembered some of these new name-object pairs weeks later. That means Rico could pick up new names for things without seeing people point at them or look at them or give any other obvious hints.
He just heard a new name and figured out what it referred to.
It seems that there is a group of gifted dogs that have realised that objects have names . These dogs appear to have an exceptional ability to learn the names of many objects. Like Rico's ability to learn names through a process of elimination, these dogs can also learn independently, without needing additional cues to identify the object being named.
But what is it, that makes these dogs gifted in this way? To explore this question, my colleagues and I recently studied a group of these unusually talented dogs, of various breeds (border collies, mixed breeds, a Spanish water dog and a pug). Many label-learner dogs are border collies but lots of other breeds seem to have this ability too.
My colleagues and I gave them a set of cognitive puzzles to solve. Each dog completed eight tasks designed to measure curiosity, problem solving, memory, learning ability and their ability to follow human communicative cues like pointing or gazing. A second group of dogs - matched by age, sex and breed - (and without any special name-learning skills) took the same tests so we could compare the two groups.
The label-learner dogs consistently showed three key traits. They were obsessed with new objects. They showed strong, selective interest in particular items. And they were better at controlling their impulses when interacting with objects. However, more research will need to investigate whether these traits appear naturally in some puppies or whether they can be shaped through training as a dog grows.
The findings may eventually lead to something like a puppy "IQ test" that identifies young dogs with the potential to learn many object names. This could help trainers select dogs well suited for important roles such as assisting people with sight or hearing impairments or supporting police work.
But does this all now mean dogs learn words like children do? After all the new paper about overhearing used a approach designed to study understanding in human toddlers.
The answer is: not quite. Children learn thousands of words, and they do it rapidly and flexibly . Even at 18 months, children don't just match a word to whatever they see at the moment.
They can understand what an adult intends to talk about by realising when a person is referring to something that isn't there. For example, if a parent says, "Where's the teddy we played with this morning?" even though the teddy is not in the room, the child may still understand what the parent means and go look for it. Children use shared context to understand others.
Even the highly skilled label-learner dogs seem to struggle to understand object-name links this way.
Although there is ample evidence that dogs seem specifically adapted to human use human given gestural communicative cues, like pointing and gazing , when it comes to "word-learning" the evidence we have is just that dogs can form object-name associations. We also know that some dogs can acquire hundreds of these associations or might have understood a rule that objects have names.
This is not comparable to word learning in children. By around age two, typical English-speaking children learn approximately ten new words each day, reaching an average vocabulary of about 60,000 words by the age of 17.
When they learn words, children apply rules and principles. Their language acquisition is based on the understanding of others as "intentional beings", that other people have goals and intentions . They recognise that when someone talks, points or gestures, they are trying to share an idea, ask for something, or draw attention to something. For example, when a parent says "Look at the dog!" the child typically understands that the parent wants them to notice the dog, not that the words are just random sounds.
However, there is currently no conclusive evidence to suggest that this core principle, underpins dogs' interactions with humans.
Dogs are amazing learners, but their abilities are not the same as human language learning. They learn names for objects, not language.
Juliane Kaminski does not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.
