Friday, June 25, 2010

Cats, Dogs, Strings, and Causality

During a recent cat-related Web search I came across a whole slew of articles that all had similar titles and content, and seemed to be referencing the same study, i..e., the following:

- Cats outsmarted in psychologist's test (Guardian)

- Dogs are smarter than cats, research shows (Telegraph)

- Study claims cats have limited intelligence (San Francisco Examiner)

- Study Finds Dogs Are Smarter Than Cats (KTLA Los Angeles)

The above articles all refer to the study: Domestic cats (Felis catus) do not show causal understanding in a string-pulling task, published in the journal Animal Cognition, September 20091.

In those articles above wherein a comparison is drawn between cats and dogs, an earlier study is also relevant: Dogs ( Canis lupus familiaris) fail to show understanding of means-end connections in a string-pulling task, also published in Animal Cognition, January, 20052.

Now, as someone with an admitted fondness for cats, and who currently resides in a household including no less than four felines, I must of course acknowledge the possibility of a bias in my reading of the articles.

Nonetheless, the main point of my analysis here is not to "defend" cats. I have no fear of data and I am certainly prepared to accept whatever a well-designed experiment might demonstrate. I also welcome any feedback indicating whether I've somehow managed to misinterpret or misconstrue the results.

Moreover, as I don't have the university connections or what-have-you to access scientific journals for free, I decided to only pay for and read in full the cat study. I have only read the abstract of the dog study. Given these disclosures, take this writing for whatever it is worth.

But anyway. On to the analysis itself.

In each case, the target animals (either cats or dogs) were presented with three tasks in which a food treat was attached to a length of string.

The baited string was then placed in a low-sided box under a clear plastic screen, either alone or in conjunction with a "dummy" (unbaited) string. Three scenarios were presented: one in which a single string with a treat at the end was employed, one in which two strings (one baited and one not) were placed parallel to one another, and a third in which the two strings were angled and/or crossed.

In all scenarios, the ends of the string jutted out far enough for the animals to reach, but the animals could not (due to the dimensions of the setup) directly access the treat itself; in short, they needed to pull on the string in order to get the food.

The results? Per the dog study abstract:

...the dogs were successful if the treat was in a perpendicular line to the barrier, i.e. straight ahead, but not when the string was at an angle: in the latter condition, the typical response was a proximity error in that the dogs pawed or mouthed at a location closest in line to the treat. When two strings that crossed were present, the dogs tended to pull on the wrong string.2

...and per the cat study abstract:

All cats succeeded at pulling a single string to obtain a treat, but none consistently chose the correct string when two strings were parallel. When tested with two crossed strings one cat chose the wrong string consistently and all others performed at chance level.1

Next, let us examine the conclusion suggested by the researchers as a valid interpretation of each study:

Regarding dogs, it was stated that "The combined results from the experiments show that, although dogs can learn to pull on a string to obtain food, they do not spontaneously understand means-end connections involving strings."2

Regarding cats, it was stated that "There was no evidence that cats understand the function of the strings or their physical causality."1

When I read the conclusion of the cat study (in the abstract), that was basically the point at which I determined I needed to read the full text of the study. This certainly revealed quite a bit more information about the experiment, its premise, and its (per the authors) implications than the abstract, and gave quite a different picture of the situation than the popular articles.

If I hadn't been convinced previously that it is both useful to read actual papers and question popular media interpretations of said papers, I certainly am now.

For the paper does not actually say anywhere in it that "dogs are smarter than cats" or that "cats are not actually all that clever". And it would not be correct to interpret the study's conclusion as being that cats do not comprehend the behavior of physical objects at all, considering that the tasks this study entailed were all highly specific string-pulling tasks.

Rather, what the study (or rather studies) point out is that per particular models of developmental cognition, cats' performance on the string tasks indicates one level of causality understanding, whereas dogs' performance indicates a slightly different level. The study isn't perfectly written and it seemed like there were some inferential gaps between data and conclusion, however, the media articles referencing this study seem to me to be vastly over-generalizing this to imply something about dog and cat cognition overall.

Moreover, in the cat paper at least, it is acknowledged that performance differences could be due to the canine and feline species' different types of optimization, i.e., cats as solitary hunters of small prey that tends to dart in and out of sight evolved to exhibit higher-level object permanence abilities. Whereas dogs as pack hunters, due to the need to both track large prey and coordinate efforts with other dogs, may have developed a better grasp of specific types of physical-object relationships than cats.

All that said, I am finding myself rather perplexed by the paper's conclusion that the cats do not understand the function of the string. I have seen my cats yank on string, and on the tails of toy mice, etc., on many occasions when these items are stuck under another object (like the couch). So unless I am misconstruing what it means to "understand" something, I have difficulty seeing how a statement that cats do not understand what string does could possibly be valid (again, I welcome corrections if in fact I am wrong on my interpretations here).

I was also unable to find (in reference to the dog paper) a ready definition of "means-ends connection". But at any rate, the studies seemed to be suggesting that if an animal truly comprehends the physical properties of the string attached to the treat, s/he will take advantage of those properties in order to obtain the treat.

Hence if the animal in a given trial either (a) fails entirely to obtain the treat even after considerable effort, or (b) obtains the treat inconsistently and/or inefficiently, it is often concluded that the animal simply does not understand that grasping, pulling, or otherwise manipulating the string in a particular way will guarantee or hasten access to the treat.

But all that said, I am not convinced that the experimental setup (in the cat study at least, as that is the one I actually read the paper describing) would have been adequate to test cats' understanding of the function of the string. For one thing, it was not clear to me in either reading the experimental equipment description or viewing a photograph of said equipment how the cats were expected to detect the presence of a treat (attached to a given piece of string) in the first place.

The paper noted several animals that had been successful in many string tasks that supposedly demonstrated causal understanding in excess of cats (or dogs, for that matter). And while I do not doubt of course that different species can and do indeed exhibit different sorts of cognitive optimization, it does strike me as interesting that all the string-test-passing animals mentioned (primates and corvids, for instance) have relatively high visual acuity compared to cats.

Felines, being crepuscular hunters of small prey, have evolved visual systems optimized for detecting tiny, subtle movements in low light conditions.4 Cats hence see vastly better than humans and somewhat better than dogs in relative darkness, and are highly adept at detecting even the smallest hint of motion in their peripheral vision.

Nevertheless, felines cannot distinguish as many colors as, say, primates or birds (most mammals, including both cats and dogs are actually dichromatic). Moreover, their visual acuity for fine details is relatively poor, especially at close range. Dogs also have fairly poor visual acuity as compared to humans, however, theirs is still estimated to be about twice that of cats. 3

This raises the question of whether the cats in the string-pulling study failed to pull the "correct" string in part simply because they could not see where the treat was attached. From the photos I found of the setup it looked as if the treat was fairly physically small (not much larger than the end of the piece of string) and that the cats were expected to perform the task when positioned quite close to the setup.

Additionally, the presence of the plastic screen, while certainly vital to the setup in terms of blocking direct access to the treats, would not have allowed the cats to identify the location of the treat by smell.** Nor could the cats touch the treat or string under the screen with their paws, or brush against it with their whiskers. I have watched my cats chase after small treats I toss across the room, and what I have noticed is that their eyes "lock on" to the treat while it is in motion flying through the air, and then when it lands, they will sniff around on the floor near the end of its trajectory until they find it. In other words, they do not seem to be using their eyes to find the treat at close range, but their noses.

Add to that the fact that the treat was, from all appearances, just sitting there motionless at the end of the string, and you've essentially removed all the major perceptual modalities that could actually assure the cat of the treat's location. Unless there was some provision made not described in the paper to account for this, I would be inclined to figure that a "random guess" strategy would in fact be the most logical one available to the cats tested.

After all, as opportunistic predators, cats are not so much inclined toward "efficiency" but rather toward trying something that seems like it might have the potential to lead to something tasty or otherwise interesting. Which could certainly contribute toward a decision to pull on both strings, or on a random one, etc.

** CORRECTION (7/16/2010): Following a closer reading of the study's apparatus description, I realized that the referenced cat study had not employed a solid transparent plastic lid with gridlines drawn on it (as I had presumed from the setup photo) but rather a wire mesh screen. Hence, my original statement that the study apparatus would not have allowed the cats to smell the treats was in error. This, however, does not negate my other observations, and moreover, given the apparent height of the box used in the study (see photo again) it is unclear to me whether the cats would have been able to precisely pinpoint the location of the treats by smell.


Primary Reference:

1-Whitt, E., Douglas, M., Osthaus, B., & Hocking, I. (2009). Domestic cats (Felis catus) do not show causal understanding in a string-pulling task Animal Cognition, 12 (5), 739-743 DOI: 10.1007/s10071-009-0228-x

Secondary References:
2- Britta Osthaus, Stephen E. G. Lea1 and Alan M. Slater (2005).
Dogs ( Canis lupus familiaris) fail to show understanding of means-end connections in a string-pulling task
Animal Cognition, Volume 9, Number 1 / January, 2005

3- From

Visual acuity is the ability to see the details of an object separately and unblurred. Acuity is measured in "cycles per degree", which means how many lines you can distinguish as being separate in a degree of the visual field. Humans see 30 cycles per degree, horses 18, dogs 12 and cats 6. Acuity in dogs is 0.4 times that of people, 0.67 times that of horses, and twice that of cats. Acuity in cats is 0.2 times that of people, 0.33 times that of horses, and 0.5 times that of dogs. If normal human vision is 20/20, then that of the dog between 20/50 to 20/100, the horse 20/33, and that of the cat is 20/100.

4- Fundamentals of veterinary opthamology, Douglas H. Slatter, p8

Thursday, June 03, 2010

Nifty 1953 Popular Mechanics Issue (via Google Books)

OK so this might just be one of those things I am always one of the last to know about but since it is just so nifty, I figured I would point it out here. Anyway, the "it" I am referring to is the fact that you can find whole issues of various magazines on Google Books, some of them quite old!

I discovered this by accident this evening because I was trying to find out who the builder and/or architect was of the 1954 ranch-style house I reside in (architect is still at large, but the builder appears to have been David Bohannon).

My search terms somehow led me to a 1953 issue of Popular Mechanics that contains a large section on house-related issues (building, buying, renovating, etc.)

Anyway, I don't know how many readers are into this sort of thing, but this issue in particular has a lot of VERY cool pictures of mid-century furniture and decor (eeee!) and some impressively detailed plans for such projects as a chair with folding arm-rests, a bed that folds up into a cupboard, and an ice fishing shelter (I kid you not).

So, yeah, I am definitely enjoying this and thought I might as well share! That is all for now.