Can dogs tell time?


It seems pretty obvious that dogs can tell time. No one who has a dog really doubts that dogs can tell when it’s time for dinner, or a walk. Freddie (pictured) stamps his feet and cries if dinner does not appear at the time he is expecting.



But until very recently, there had been no studies on whether dogs are able to measure time. It could be that dogs are unable to tell how much time has passed, but that they use other cues to determine when it’s time for a walk, or for bed, or for their human to come home. If dinner, or a walk, or a returning human, tend to happen after certain cues, then dogs’ apparent capacity to tell that it is time for those events to happen might really only be a capacity to detect that the relevant cue has just occurred.


So the question is, can dogs tell time, and, if they can, how do they do it?

There are already two hypotheses about how dogs tell time, assuming that they do.

Alexandra Horowitz (2016) has suggested that dogs might smell how much time has elapsed by using decreases in the amount of scent in the air from some particular event, until the current time. So for instance, dogs might work out when their human is likely to return by monitoring the decrease in the amount of scent from the human throughout the day. When the scent reaches a certain level of dissipation, the dog knows that a certain amount of time has elapsed and that the human is likely to be returning.


Another theory is that a dog is able to detect how much time has passed using much the same mechanism as it hypothesised to exist in other species, including humans. On that view, we each, dogs included, have inside us a little internal timing system made up of a pacemaker, an accumulator, and switch. We track time by tracking the number of pulses emitted by the pacemaker, which are then collected by the accumulator. If we want to measure the time between two events, the system empties the accumulator and then starts the accumulator collecting pulses at the beginning of the interval to be timed. The switch then stops the accumulator collecting any more pulses at the end of the interval to be time. The system then counts the number of pulses collected, and stores this information in working memory, where this can be compared to pulses collected during some other measured duration. We can then determine that one temporal duration is longer, or shorter, than another.

Two recent studies, (Domeniconi and Machado 2017 and McPherson and Roberts 2017) aimed to see whether dogs measure temporal durations.


Both studies used what are known as temporal bisection tasks. That sounds grisly, but it isn’t. Temporal bisection tasks begin with a training session in which the aim is for participants to pair some stimulus (say a yellow box) with experiencing an interval of a certain length (say one second) and pair a different stimulus (say a blue box) with experiencing an interval of a different length (say 4 seconds). The aim is to train participants to choose the yellow box when they experience a 2 second interval, and to choose a blue box when they experience a 4 second interval.


In the Domeniconi and Machado study, 5 dogs were initially trained to choose a yellow stimulus and not a blue one, when they heard a 1 second tone, and were trained to choose a blue stimulus and not a yellow one, when they heard a 4 second tone. It took, on average, 34 trials to train the dogs to do this task, and they had an above 80% success rate.

Once the dogs had learned to choose the right colour when they were presented with the tone of the relevant length, the experimenters then offered them tones that were neither 1 second not 4 seconds in length, but instead, somewhere in between. The aim is to see which stimulus, yellow, or blue, the dogs choose, in response to each of the tones they are given, and to graph the results.


After the training phase the dogs were randomly given 12 tones, 6 long and 6 short. They were then positively reinforced if they chose the yellow stimulus for the short tones, and the blue stimulus for the long tones.


Finally, the dogs were given 18 tones in random order, and their choice of stimulus was not reinforced. If dogs can measure temporal intervals, then we would expect to find that they choose the yellow stimulus the closer the tone is to 1 second, and begin to choose the blue stimulus as the tone gets closer to 4 seconds. In fact, the study found that dogs switch from choosing the yellow, to choosing the blue stimulus at about the geometric mean between 1 and 4 seconds. (The geometric mean is calculated by taking the square root of the product of the numbers in question. Hence the geometric mean of 2 and 10 is the square root of 20 (~4.47) while the arithmetic mean of 2 and 10 is the sum of the numbers divided by 2 (6).)

The dogs’ responses suggest that they are about as sensitive to intervals of time as pigeons and possums, more sensitive than fish and turtles, but less sensitive than cats, rats, monkeys and humans, at least when compared using a temporal bisection task. It is unclear whether this is the result of some loss of capacity to track time due to domestication, or an artefact of small differences in study design.


At any rate, even if dogs are not quite as sensitive to temporal intervals as cats, rats and monkeys, they still do pretty well on this task.


The second study, by McPherson and Roberts also used a temporal bisection task. But their task was a little more complicated. They had 6 participant dogs, and they divided the dogs into two groups, one of four and one of two. The group of four dogs were trained on the bisection task in much the same way as in the experiment just described, except that the temporal intervals they were trained on used both a tone and a light to mark the extent of the interval. Also, in this experiment the dogs were trained on an interval of 2 seconds, and one of 8 seconds. So these dogs either heard a tone and saw a light go on for 2 seconds, and were taught to approach feeder 1, or they heard a tone and saw a light go on for 8 seconds, and were taught to approach feeder 2. The second group of dogs was trained on the same intervals, but they were trained using only a light to mark the extent of the interval. They were taught to approach feeder 1 if they saw the light go on for 2 seconds, and to approach feeder 2 if they saw the light go on for 8 seconds.


Once the dogs were 75% accurate they were gives trials of signals of 2, 3, 4, 5, 6 and 7 and 8 seconds. The dogs trained on both the tone and light sometimes saw both signals, and sometimes they heard only the tone, or saw only the light. The dogs in the other group saw only a light.


Interestingly, the dogs trained on both the tone and light were accurate when they were presented with both signals, and when they were presented with only the tone but no light. In both these conditions the dogs responded by switching which feeder they went to at around the geometric mean between 2 and 8 seconds. By contrast, they were relatively inaccurate when they saw just the light. However, the dogs in the condition that were trained to respond just to the light showed that they were sensitive to the intervals elapsed, though they were not as sensitive as the dogs trained with the tone and light.


This suggests that the dogs trained using both the tone and the light were using the tone, and not the light, to determine the duration of time that had passed, while the dogs that were trained using only the light were using the light to determine the elapsed duration, but were less effective at determining the duration than were the dogs who were trained using the tone. The experimenters concluded that when the light and tone were presented, the tone overshadowed the light, which was then not salient as a cue they could later use in the task. It remains unclear, though, why the tone so overshadowed the light, given that both were of moderate intensity.


All up, the studies jointly show that dogs can measure durations of time, and that they do it fairly well.


But the studies don't tell us how dogs do this: by smell or through an internal timing mechanism. Given that the intervals they tested were very short (in the range of seconds) the results suggest that the dogs were probably using something like an internal mechanism rather than tracking differences in smell. But of that doesn’t mean that dogs don't use smell to track longer intervals of time. Since these are the first studies of this kind on dogs, though, we really want to see more evidence about dogs’ capacity to measure time, and the mechanisms by which they do this.


In the meantime, when your dog complains that they’ve been waiting for 10 minutes for their dinner, you should take them seriously.

Horowitz A. (2016). Being a dog. New York, NY: Scribner.

Domeniconi, C. and A Machado (2017). “Temporal bisection tasks with dogs: an exploratory study.” Psychology and Neuroscience10(1): 101108.

McPherson, K and W. A. Roberts (2017). “On the clock: interval timing and overshadowing in domestic dogs (Canis familiaris). Journal of Comparative Psychology 131(4): 348-361.