Is it Possible to rank animals by intelligence?

Many find the above question one of the scariest because it allows absolute answers. Take the idea that ‘killing X species of whales is murder’. If there is no such way to rank species then, save for extreme cases, we are forever free to hold either affirmative or negative opinions on that whaling X is murder. We can entertain a personal degree of certainty beholden only to our culture. Alternatively, if we can so rank animals, then one of these groups will likely be devastated by a certainty borne of science. Species X may prove no more intelligent than a pig (our most intelligent farm animal) at the 99% confidence level, or more intelligent than a human four year old (the lowest age at which, I think, we would all agree it has our level intelligence). Given the gravity of the consequences, we must examine the evidence that would allow ranking with particular care.

How Hard is it to Rank Humans in Intelligence?

The bulk of all psychometric analysis has been performed on a single species: ours. Although the methods by which one would attempt to measure intelligence variation within a species are not identical to the ways you would compare differences between species, this is the place we must start.

We are now a hundred years on from when Spearman discovered that humans exhibited a ‘g factor’. Before that time it was quite reasonable to argue that each person might be so talented in the skill of their vocation, that comparisons between people were useless. After Spearman we knew there existed a unique intelligence factor that predicted all other such factors best. Don’t get confused, individual and specialist factors of intelligence really do exist, and a person with low g might well have a special ability that might help save the lives of others. What Spearman proved was not merely that one could create a hypothetical construct of ‘general intelligence’ above all special factors, but that this new single factor could explain most of the intercorrelations of those special factors. There could be no reason for this unless it represented something real or tangible. We call quantitative efforts to render g with a mean of 100 and standard deviation of 15 as ‘IQ’. Perfecting IQ tests became an analytical task, and statistical methods show exactly how close we are getting. IQ isn’t the be-all and end-all of intelligence, but it is the ultimate measure if we must compare.

For humans, the best IQ tests quickly neared a g factor correlation of .7. Some modern tests, such as Raven Progressive Matrices, can achieve a correlation of .8, and cross referencing with other tests we can get closer still. Even at .7 this made the IQ test a powerful tool to assist at ranking people in a (then) brand new concept of general intelligence.

 

How Hard is it to Rank Primate Species in Order of Intelligence?

The next most studied group in cognitive ability is the nonhuman primates.

If you google search for current data on which are Earth’s smartest well-studied species, one of the first ideas you come across is that ‘there are so many ways to measure intelligence that the question is meaningless’. The premise of that statement is correct, but they reach for the very same false conclusion as we had with humans pre-Spearman. Of cause, this conjecture can only be true if there doesn’t exist a general intelligence factor that manifests for all primates.

Let’s look at that seminal meta-analysis of primate intelligence by Deaner et al more closely. It took six separate measures of intelligence, and cross referenced them against each other. Rather like Spearman did for humans, this paper did for the whole primate order: it showed that there existed a general intelligence factor common to all primates – one that could be referenced across species. It also showed that this wider g-factor was reasonably easy to extract. Once more I must remark that every species will have its own areas of special skill, it is just that we now know that these are seldom strong enough to confound the unique ranking of a cohort of primate species in order of intelligence if several different tests are available to us. That order would be unique, and its measure is without parameters that can be adjusted. Below is a list of the ten most intelligent primate genera, as we would pick them today. Future data could only produce minor adjustments to it.

 Rank  Primate
 1  Human
 2  Orangutan
 3  Chimpanzee
 4  Gorilla
 5  Spider Monkey
 6  Grey Langur
 7  Surill
 8  Macaque
 9  Mandrill
 10  Guenon

To this website, an even more significant result from that paper was that this primate general intelligence correlated highly with raw brain size, but little with anything else. That finding was the first great strike against the Encephalisation Quotient Hypothesis.

In humans, meta-analyses places the correlation of individual intelligence and brain size at only around 1/3, but this one meta-analysis across all primates places it much higher, at .8. It may seem paradoxical that elaborate tests within that best studied species of primate shows a much lower correlation of brain size with extracted g, than tests across primate species. This becomes less of a mystery when you consider that the brain size differences within a species would typically be a few percent. Differences across species are typically by factors of several fold, making other factors less important by comparison.

 

Can We Rank all Mammals and Birds by Intelligence?

This next question places us in fresher fields. Psychometry’s frontier, if you like.

Much work has also been done in measuring the intelligence of non-primate animals, with some surprising results that confound prior expectation. Who would have thought the short-lived and solitary octopus would turn out to be so intelligent! When informally describing such findings it is common to compare the animal being studied with an animal whose intelligence is better known among the public. There have been efforts to give Earth’s cleverest animals a rough ranking to further help this process. These have typically been based on tasks each animal can and can’t be trained for. Here is a typical recent example of such an attempt.  

Animal-Intelligence-Hierarchy

Note the use of crude categorisations, implying that this was never meant to be taken too seriously. But could any of these possibly be meaningful or accurate? Such animals are so different from each other. We humans tend to rate tool usage highly, and this might exaggerate the placing of dexterous animals and animals closer to our genetic heritage such as primates. Could a general intelligence factor rescue us once more, and compensate for our own human biases? Before I answer, let me take you on a diversion to see what general intelligence might look like.

Imagine three hypothetical animals; A, B and C. A can account for the exact number of nuts up to six, but not for anything else no matter how much training it receives. B can account the same for apples up to six but, once more, cannot transfer that skill to any other object. C can account for any class of objects up to five. The question is: which animal shows the most numeric intelligence? Some scientists today would still claim ‘it depends on what is being counted’, but a better answer might be that it is most likely to be C. Its skill has been shown to be transferable – even if its ability is lower than A or lower than B in highly specific cases. Though we may suspect this is the right answer, the question can’t be answered beyond all doubt unless a g-like factor can be extracted across the entire group, and that would require cross referencing this numerical ability with several other ways to measure intelligence. This finally brings us to a remarkable paper published in 2014.

This was the very first paper to attempt to create the framework for a quantitative gauge of intelligence across all mammals and birds. It only used one criteria. Even though it was a well picked gauge that was known to correlate highly with intelligence within several species (including humans), it was still just a single measure. As such, we cannot attempt to extract general intelligence, so its findings might eventually be overturned by much later studies that are more comprehensive. Nevertheless it is groundbreaking, fresh, and the best we have to date. The idea was to find what evolutionary processes led to higher intelligence, and it delivered a clearer picture than anticipated.

They looked at many things, but only two factors resolved as important: the diversity of food choices present in the test species’ natural habitat and, most of all, brain size. There is debate as to whether dietary breadth is a ‘special factor’ only apparent in this study because every test was done with food, which was one of the reasons I gave my ABC case above. These two factors together explained a staggering 82% of all the variance in its primate subset. For an alternative interpretation of this study and the significance of its findings read here.

The study looked to test many theories, including the current favourite: that social interactions play the larger part in creating higher intelligence. Animals that live in larger social groups tend to be more highly encephalised than closely related species living in small groups. There seems no doubt that animal species that form larger groups in the wild have higher social intelligence, but this type of intelligence didn’t seem to transfer to tasks of individual food choice.

The study suggests that brain size is a very good predictor of general intelligence across warm blooded tetrapods as a whole. They could also show how the brain size effect almost disappeared if adjustments were made for body size, and that no other factor they could think of had any significant predictive value for intelligence. Once more let me clarify what this does and doesn’t mean…

Animals in larger social groups definitely do display more social intelligence, and evolution almost certainly does add more neural circuits for that purpose to their brain. This might add to the general intelligence of more social animals over time, but it is now apparent that that trend is not particularly significant.

So that is the state of the art in animal psychometry as it stands in 2015 – brain size is a far more significant predictor of intelligence than anything else our scientists can yet think of and test for…. and the animal with the largest of them all is the sperm whale.

For those who are itching to know what the smartest genera on Earth might turn out to be if a universal g is established, here is my current educated guess.