Is it possible to measure the working memory of a non-human animal? And if so, have there been any studies that have quantified the working memory of animals, and that have compared that to the working memory of humans? Or are the two not comparable?
The problem with comparing them is that the actual stimuli used for each species are different (e.g. odors for bees, shapes for monkeys) and this is known to affect the results.
In humans, if you run similar tasks, just with small verses large sets of possible stimuli, for example digits verses general words, you can get very different results. Even if you did try to use the same stimuli set, the result will probably not be "subjectively" equal to that of a different species. Think of yourself trying to do the odor task of the bees; it's not a fair game.
Other seemingly small changes in the paradigm can also affect the results, for example a delay between the trials, etc. So the only way to seriously compare them would be to conduct the same experiment, by the same lab, of different species. I personally don't know of any research that was directly aimed at comparing them. Maybe someone using (lesioned) animals as a model for dementia would know more.
Sure, visual working memory capacity has been assessed in humans, chimps and mac using a change-detection tasks and trail-making tasks.
For example, the chimp Ayumu at Kyoto University easily outperforms all humans on a simple number sequencing task. In the task, the numbers 1-n appear briefly on the screen in random positions and then the numbers must be touched by the participant in numerical order. Ayumu can sequence more than nine digits in less than 200 ms, according to the 2007 paper. In contrast, human performance was at chance levels at that exposure duration.
In another visual working memory task, the change-detection task, colored shapes appear on the sample screen, disappear for some brief amount of time (typically about a second) and then reappear. 50% of the time all the squares are the same, 50% of the time one square has changed. The subject indicates if a the test screen matches the sample screen. This a delayed-match-to-sample task. To measure memory capacity, the sample and test array sizes are increased, 2,4,8, etc. Humans and rhesus perform equivalently on this task, in that both species are able to perform at or above 80% correct for array sizes of about 3~4 items. earl miller study
io9 published a pop-sci discussion of the chimp study, "This chimp will kick your ass at memory games — but how the hell does he do it?", and links to a flash game by Lumosity Games that replicates the original study.
When classes come visit the Duke Canine Cognition Center, we ask them how they would test the memory of their pet. The kids come up with a solution pretty similar to how its done. One kid put down 5 different cups, showed the dog a treat, and placed it under one of the cups. Then the dog was either distracted, or simply restricted from approaching the cups, for a fixed amount of time. The dogs are very eager to retrieve the treat, and will often just stare at the cup until the time has elapsed, so in another version we shield the cups with a giant occluding board so the dog must keep an internal representation. This task is no problem for many dogs. We do 20 second, 40 second, and then 60 second trials with only 3 cups, and there is a lot of variability in success.