EMOTAG
Science18 August 2026

Can Dogs Learn Toy Names Just by Listening to Us Talk?

Some dogs can. In a 2026 Science study, ten Gifted Word Learner dogs learned new toy names by overhearing two people talk. Typical family dogs did not.

By the Emotag team · 7 min read

A golden retriever lying down with its head on the floor, looking up and to the side

Some dogs can—but this ability appears to be extremely rare. In a 2026 study, ten Gifted Word Learner dogs learned the names of new toys by watching their owners speak to another family member. They performed about as well as when the owners spoke directly to them. Ten typical family dogs, however, showed no convincing evidence that they had learned which name belonged to which toy.

You are showing a new toy to someone in your home. You hold it up, say its name and pass it back and forth. Your dog is nearby, watching, but nobody speaks to the dog.

Could your dog still learn the toy’s name?

For most dogs, the answer is probably no. But a small group of dogs with an exceptional ability to learn object names may be listening more closely than anyone realised.

A study published in Science found that these unusually talented dogs could connect a new spoken label with the correct toy simply by observing an interaction between two people. They could even form that connection when the toy was no longer visible as its name was spoken.

What are Gifted Word Learner dogs?

Most dogs can learn action cues such as “sit” or “down.” Learning the names of individual objects is much less common.

Gifted Word Learner dogs are a rare group that already know the spoken labels of many objects, usually toys. They can spontaneously learn new toy names during natural play with their owners, without the kind of deliberate training normally used to teach commands.

In this research, the word word has a narrow meaning. It refers to an auditory label connected to a particular object. The study did not test whether dogs understand grammar, sentences or human language in the broader sense.

The researchers wanted to know whether the dogs’ unusual learning ability would still work when the human speech was not directed at them.

How did researchers test whether dogs learn by overhearing?

The study included several related experiments. Ten Gifted Word Learner dogs took part in both a direct-learning condition and an overhearing condition. Ten typical family Border collies were also tested in the overhearing condition. A further experiment with eight of the gifted dogs tested whether they could connect a name with a toy that was no longer in view.

Study detailWhat the researchers did
Gifted dogs10 dogs with an existing vocabulary of object labels
Typical dogs10 family Border collies without a known object-label vocabulary
New objectsTwo unfamiliar toys in each experiment
ExposureEight minutes of labeling per toy, spread across four nonconsecutive days
Learning testThe owner asked the dog to retrieve each new toy by name
Memory testEight gifted dogs were tested again two weeks later

Every dog was introduced to two unfamiliar toys with two new names. Each exposure began with one minute of labeling. This was followed by three minutes of play in which the owner did not say the toy’s name, and then by a period when the dog could explore the toy freely. In total, each dog heard each new label for eight minutes over four separate days.

In the direct condition, the owner played with the gifted dog and introduced the toy in the usual way. In the overhearing condition, the owner and another family member sat opposite each other, passed the toy back and forth and used simple phrases such as “This is [toy name]” and “Do you want [toy name]?” They looked at each other and the toy and spoke with enthusiasm, but they did not look at or communicate with the dog during the labeling period.

Two people sitting across a table from each other, talking
The overhearing condition, from the dog’s side of the room: two people, a toy, and a conversation that does not include her.

After the exposure days, the two new toys were placed in another room with nine familiar toys. The owner remained out of sight and asked the dog to retrieve a toy by name. Because both new toys were always available, choosing the correct one showed whether the dog had connected each label with its intended object rather than simply choosing something unfamiliar.

Can gifted dogs learn toy names by overhearing people?

Yes. As a group, the ten Gifted Word Learner dogs selected the correct new toys significantly more often than expected by chance in both conditions.

Their median accuracy was 92% when their owners addressed them directly and 83% when they merely overheard the owners interacting with another family member. The difference between the two conditions was not statistically significant. Seven of the ten dogs performed above chance individually in each condition.

The researchers also examined only the first request for each new toy. The gifted dogs chose correctly in 16 of 20 first attempts after direct learning and in all 20 first attempts after overhearing. This helped rule out the possibility that the dogs only worked out the correct names through repeated testing.

The result suggests that some exceptionally gifted dogs can learn an object’s name even when they are observers rather than active participants in the interaction.

Did typical family dogs learn the names too?

No convincing evidence showed that they did.

The researchers tested ten typical family Border collies using the same overhearing procedure. At first glance, six dogs appeared to choose the correct toy more often than expected if they were selecting randomly from all 11 toys in the room.

However, when these dogs made mistakes, they usually selected the other new toy rather than one of the nine familiar toys. This pattern suggests that they recognised which toys were new but had not necessarily learned which name belonged to which one.

A dog lying on a wooden floor surrounded by scattered toys and a red frisbee
Nine familiar toys and two new ones. Having a second new toy in the room is what separates learning a name from preferring whatever is unfamiliar.

When the researchers used the more conservative comparison of choosing between the two new toys, none of the typical dogs performed above chance individually. Their behaviour was therefore more consistent with a preference for novelty than with learning two new object-name connections.

This distinction matters. A dog may hear an unfamiliar name and infer that it probably refers to an unfamiliar toy. That does not mean the dog has stored a lasting connection between that exact name and that exact object.

Can gifted dogs learn a name when the toy is out of sight?

The researchers next tested eight Gifted Word Learner dogs under a more difficult condition.

The owner first showed the dog a new toy and then placed it inside a specially designed bucket. Once the toy was hidden, the owner spoke its name while looking between the bucket and the dog. False-bottom buckets were used to control for possible scent cues.

The dogs later retrieved the correct new toys with a median accuracy of 79%, which was above chance at the group level. Five of the eight dogs also succeeded above chance individually.

This result suggests that the gifted dogs did not always need to see the object at the exact moment its name was spoken. They could connect information separated in time: the toy they had just seen and the label they heard after it disappeared.

Two weeks later, the eight dogs were tested again. Their performance had not significantly declined, suggesting that the new object-name connections had been retained rather than disappearing immediately after the first test.

Are these dogs learning words like 18-month-old children?

Only in a specific, limited sense.

Children as young as 18 months can learn a new object name by observing two other people interact, even when the conversation is not directed at them. The gifted dogs in this study completed a functionally similar task: they followed a human interaction and connected an overheard label with an object.

That does not mean the dogs have the language abilities of an 18-month-old child. The study tested one narrow skill—learning labels for objects. It did not test grammar, conversation, symbolic thought or the full range of abilities involved in human language.

The researchers also do not claim that dogs and children use identical mental processes. The dogs may have relied on a combination of gaze, enthusiasm, the people’s attention, the novelty of the toy and their previous experience of learning toy names. Future studies will be needed to separate these possibilities.

The comparison with children is useful because it shows that some of the social and cognitive building blocks used in early word learning are not necessarily unique to humans. It is not an intelligence-age comparison between dogs and toddlers.

Can you teach your own dog toy names by talking nearby?

This study does not show that overhearing is an effective teaching method for most dogs. The successful dogs had been selected because they already possessed an unusually large vocabulary of object labels. The typical Border collies did not show clear evidence of learning the new object-name mappings under the same conditions.

The overhearing interaction was also carefully structured. The two people repeatedly used the new label in short sentences, passed the toy between them, looked at the object and each other, and expressed enthusiasm. This was different from a dog casually hearing an ordinary household conversation in the background.

If your dog does not learn the names of toys, that says very little about the dog’s overall intelligence. Object-label learning is a specialised ability, while dogs can be skilled in many other areas, including reading gestures, learning routines, following action cues and responding to human emotions.

A husky asleep on its side on a hard floor indoors
A dog lying quietly across the room may not be part of the conversation.

For dog owners, the most interesting lesson is not that every dog is secretly building a vocabulary. It is that a few dogs may learn simply by paying attention to interactions that do not include them. A dog lying quietly across the room may not be part of the conversation—but sometimes, it may still be learning from it.

The study

Dror, S., Miklósi, Á., Morvai, B., Năstase, A.-S. & Fugazza, C. Dogs with a large vocabulary of object labels learn new labels by overhearing like 1.5-year-old infants. Science 391, 160–163 (2026). Read the paper →

Read next · Science

Do Anxious Owners Have Anxious Dogs?

← All articles