This month in canine science – September 2026

Every month I go through what’s been published in canine science, looking for studies that might shift how we understand our dogs.

September had a themer: what is going on in the dog before we see the behaviour. In the brain, in the mood, and even in the body a few seconds before a mistake.

Here are the studies that caught my attention.

Does your dog’s brain listen for the consonants?

When we pick words out of continuous speech, we mostly follow the consonants. Vowels are louder, but consonants carry more of the information about where one word ends and the next begins. Babies start leaning on them at around one year old. The open question is whether that needs a brain built for language, or whether hearing a lot of speech is enough.

A team at ELTE in Budapest tested it on dogs. They recorded EEG from 20 people and 20 companion dogs while they listened to streams of made-up speech, built so that either the consonants or the vowels formed repeating three-syllable “words”.

In both species, brain activity followed the rhythm of the words more strongly when the pattern was in the consonants. Both also showed a change in brain activity about 400 milliseconds after each word, again only for consonant-based words. And dogs that had spent their first three months on the street or in a shelter, rather than in a home, showed the same bias.

There were differences too. People could pick up the vowel-based patterns as well, dogs could not. And human brains tracked individual speech sounds, while dogs seemed to work at the level of syllables.

This does not mean dogs understand what we say. It tells us how the brain handles the sound of speech, in 20 dogs. What it suggests is that regular exposure to speech may be enough to change how a brain processes it, without a brain built for language.

Your dog’s brain is doing a lot with your voice. Whether it acts on it at the park is a separate question.

Source:

Morvai, B., Tóth, K.G., Boros, M., Rácz, D.S., Iotchev, I., Szabó, K., Andics, A. Neural evidence that dogs segment the speech they hear with a humanlike consonant bias. Science (2026). https://doi.org/10.1126/science.adw7709

Speech is one channel. Another is what we look like when we are upset.

Do dogs react to crying the same way wherever they live?

We like to say “dogs know when we’re upset”. This study asked what a dog’s own life history does to that.

Researchers compared 112 dogs from three backgrounds: companion dogs, shelter dogs and free-ranging dogs. A familiar person or an experimenter simulated either crying or singing near the dog, and the team scored looking, physical contact and behaviour directed away from the person.

Crying produced more person-directed behaviour than singing in every group. Whether the person was familiar made little difference, which suggests it was the crying itself that mattered. Companion and shelter dogs paid more attention to the crying person than free-ranging dogs did, and the free-ranging dogs spent more time on their own business and the surroundings.

A few things to keep in mind. Paying more attention to crying does not show that a dog felt distress or wanted to comfort anyone. And the three groups were tested in different countries and settings, so life experience can’t be neatly separated from everything else that differed.

My favourite detail is habituation, which simply means losing interest when the same thing keeps happening. As the exposures were repeated, the dogs engaged less with the task and turned to other things. Which is either reassuring, or slightly insulting.

Source:

Generoso, C., Nandi, S., Ayrosa, F., Resende, B., Bhadra, A., Savalli, C. Life experiences affect dogs’ responses to human crying. Sci Rep (2026). https://doi.org/10.1038/s41598-026-71958-5

Noticing distress is one thing. Working out what a dog is feeling in the first place is another.

Fear and frustration can look the same. Are they?

Researchers at the University of Lincoln looked at how clinical animal behaviour uses the words fear and frustration, and found they are used loosely, often without any definition. They screened the literature, ended up with 86 papers on fear and 35 on frustration, and proposed a framework.

In their definition, fear is a response to a predicted threat of physical harm to the individual. Frustration is what happens when a motivated behaviour cannot be carried out: being restrained, someone invading personal space or a guarded resource, a reward that doesn’t arrive, two competing motivations at once.

One consequence surprised me. Resource guarding is often described as fear. The authors argue it fits frustration better, because the threat is to the dog’s control over the resource, not to the dog’s safety. They also point out that “anxiety” can mean either, and suggest saying which one you mean.

Why it matters: in their view a fearful dog and a frustrated dog may need different things. Less hypervigilance for one, coping skills when control is taken away for the other. That is a hypothesis, not a tested result.

This is a proposed framework, not an experiment, and the authors label several of their categories as plausible extrapolations or hypotheses for future testing. Still, it made me think about how quickly I file a dog’s behaviour under one word and move on.

Source:

Wilson, B.M., Soulsbury, C.D., Mills, D.S. Development of a psychobiological nosology for fear and frustration in clinical animal behavior using dogs as a model species. Front. Psychol. (2026). https://doi.org/10.3389/fpsyg.2026.1888439

If frustration is its own thing, does it help to have someone there while it happens?

Your presence may not help him solve it, but may help him cope when it stops working.

Dogs in this study learned to remove plastic bones from an apparatus to get the food underneath. Then the food stopped coming, and later came back. Thirty dogs took part, 15 with the owner in the room and 15 with the owner absent.

When the food stopped, dogs with their owner present showed fewer frustration and stress behaviours. But the owner’s presence made no difference to any of the problem-solving measures. The authors suggest that is because the dogs could solve the task on their own.

So the dogs weren’t better at the puzzle. They were less bothered when it stopped working.

The groups were small, it was a single task, and it compares different dogs rather than the same dog with and without the owner. It also builds on earlier work from the same lab, where the owner’s presence reduced frustration in a social task.

My reading, not the authors’: every training plan has moments where an expected reward doesn’t come. Being there may matter for how the dog gets through those moments, even before you do anything clever.

Source:

Munafó, F., Rial, L.A., Carballo, F., Bentosela, M. Social buffering in domestic dogs: Are the owners able to decrease dogs’ frustration during a non-social task? Behav Processes 242:105439 (2026). https://doi.org/10.1016/j.beproc.2026.105439

Frustration, fear, plain over-excitement, and a kennel with all of them at once. That brings me to shelter dogs who can’t settle.

What actually helps a shelter dog who can’t settle?

A team from Texas Tech ran a randomised trial with 50 shelter dogs that showed excessive arousal, meaning a lot of jumping, circling, pacing and vocalising. After a baseline week, dogs spent a week in one of five groups: usual care (two short walks a day), calm enrichment (longer walks plus one-to-one time in the kennel), active enrichment (play sessions plus one-to-one time), medication (gabapentin and trazodone), or medication with calm enrichment.

Dogs were scored when they came back to the kennel after their daily walk or play session. Compared with the dogs that only got the usual two short walks, arousal was lower in three groups: play sessions, medication, and medication with calm enrichment. The dogs that got longer walks and one-to-one time but no medication did not differ from the usual-care group. The combined group had the most consistent changes, with less lip licking, panting, vocalising and jumping.

Before anyone reads this as “medicate the excited dog”: it lasted one week, in one shelter. There was no placebo, and the pills were given in peanut butter or cheese, which is enrichment of its own. All dogs also moved into better-equipped kennels, and staff chose which dog went where. Most importantly, dogs were enrolled because of how they behaved, not why. Frustration, fear and plain high energy can all produce a jumping dog, which is exactly the problem from the previous study.

Still, I like the question it asks. Not “why is this dog badly behaved”, but “what is his arousal doing, and what actually changes it”.

Source:

Cisneros, A., O’Hanley, K., Moody, C.M., Stellato, A.C. Combined enrichment and pharmaceutical intervention reduces excessive arousal in shelter dogs in a randomized controlled trial. Front. Vet. Sci. 13:1931203 (2026). https://doi.org/10.3389/fvets.2026.1931203

One last study, about a state that shows up before the behaviour does.

Can you see a detection dog about to miss?

Even well-trained detection dogs sometimes fail to indicate a target odour. The usual assumption is that a miss just happens. This study asked whether there are warning signs.

Dogs worked on a treadmill with computer-controlled odour delivery, so the researchers knew the exact moment each odour arrived. They looked at the 4.5 seconds before it. Body posture, tracked at 29 points on the dog with markerless pose estimation, was fed into a machine-learning model. It predicted a coming miss with an AUC of 0.81. Adding heart rate, heartbeat timing and core body temperature increased the share of misses it caught from 77% to 85%. What did the model actually pick up on? Mostly how the dog carried its head and neck, and what the tail was doing. Of the physiological measures, the tiny variation in the gap between one heartbeat and the next mattered most, more than heart rate itself.

The dogs kept trotting at the same steady pace throughout, with no visible disruption to their gait. The differences were subtle, and they appeared before the odour was even presented, so the dog had not yet had the chance to smell it and miss it.

Now the caveats. Four dogs took part. It was a treadmill with a fixed odour port and a nose-poke as the response, which is far from a real search. In the test set, misses made up two thirds of the trials, which is why the researchers rely on AUC rather than plain accuracy. And the idea of a running estimate of a dog’s miss risk was shown on a single session, not tested properly. The authors say clearly that this is not ready for the field.

But I can’t hide that as someone who builds a tool around tracking dog training data, I’m drawn to signals that show up before the behaviour does.

Source:

Schultz, J., Rothkoff, L., Aviles-Rosa, E.O., Hall, N.J., Maughan, M.N. Predicting missed alerts in detection dogs. Sci Rep 16, 28712 (2026). https://doi.org/10.1038/s41598-026-71367-8

Which of these surprised you most?

For me it’s the detection dogs. The idea that a miss has a run-up, and that the run-up can be measured.