Â鶹ÒùÔº

A study in Current Biology by the Department of Ethology at Eötvös Loránd University reveals that dogs with a vocabulary of toy names—known as Gifted Word Learners—can extend learned labels to entirely new objects, not because the objects look similar, but because they are used in the same way.

In humans, "label extension" is a cornerstone of early language development. In non-humans, until now, it had only been documented in a few so-called language-trained individual animals, after years of intensive training in captivity.

But learning to extend labels to objects that share the same function, rather than visual similarities, is considered an even more complex skill. A toddler learns that the word "cup" can apply to mugs, tumblers, and sippy cups, or that both a spoon and a ladle are "for scooping." While individuals of many animal species can group items by appearance, extending a learned label to a functionally similar but visually different object has long been considered an advanced skill.

The time and efforts needed to train animals in captivity to learn verbal labels, as well as the very limited number of subjects that successfully acquired such vocabulary, have until now limited the feasibility of this type of research.

But here comes the twist: "Gifted Word Learner dogs offer a unique possibility to study this phenomenon because they rapidly learn verbal labels—the names of toys—during natural interactions in their human families," said Dr. Claudia Fugazza, lead author of the study.

"Our results show that these dogs do not just memorize object names," continues Dr. Fugazza. "They understand the meaning behind those labels well enough to apply them to new, very different-looking toys— by recognizing what the toys were for."

A play-based experiment

Researchers of the Department of Ethology at Eötvös Loránd University tested seven Gifted Word Learner dogs—(six border collies and a blue heeler)—known for their unusual ability to learn the names of dozens of toys naturally, through everyday play.

The experiment had four stages, all of them conducted in a natural setup, at the house of each dog owner, during playful interactions:

1. First, in the learning phase, dogs learned two new labels, such as "pull" and "fetch," each referring not to a single item, but to a group of toys that looked completely different but were used in the same way during play (tug or retrieve).
2. Second, during a formal assessment, the dogs showed that they had successfully learned those labels and could appropriately choose the "pulls" and "fetches" when asked.
3. The crucial part of the experiment was carried out after this assessment: in the generalization phase, the dogs were introduced to new toys, also with diverse physical features, and the owner played in the same two ways as before, but this time saying no labels.
4. As a test, when asked for a "pull" or "fetch," the dogs selected the correct unlabeled toy significantly above chance, indicating they had generalized the labels to a functional category.

The study provides the first evidence that dogs can generalize verbal labels to functional categories during natural-like playful interactions in their human families—mirroring, in functional terms, the natural context of human language development.

"This ability shows that classification linked to verbal labels can emerge in non-human, non-linguistic species living in natural settings," said Dr. Adam Miklosi, co-author of the study. "It opens exciting new avenues for studying how language-related skills may evolve and function beyond our own species."