Finding a Relationship between Whale Sociability and Intelligence

This post should be thought of as a followup to a previous post where I examined the data in a recently published meta-analysis of cetacean studies, to see if species/genus specific cognitive data could be extracted from it. They showed a strong correlation between brain size and habitat range. Brain size and habitat range, in turn, correlated highly with directly tested intelligence in primates. This opened the possibility of creating a similar scale for cetaceans, but what we really needed was a third metric, independent of the first two, also known to correlate to intelligence. Below we will attempt to construct that within the dimensions of social data provided by Kieran et al.

As I mentioned in that previous post, we have reason to believe that sociability should have a mild negative correlation with intelligence, which is the relation we find between their corrected social repertoire and brain size. However, by analogy with primates, we should also expect that some social behaviours should have a strong positive correlation with intelligence. You should note that below, I have made no attempt to adjust for research effort and, since, more intelligent species tend to be more interesting to science, a neutral trait is likely to have been found among the smartest, than would be expected if research had no such skew. I will now go through those given behavioural dimensions given by Kieran et al., one by one.

Group Hunting

Group hunting seems to show no relationship at all with our proxies for intelligence (raw brain size, r=-.006, latitude range, r=+.074).

Interspecific Cooperative Behaviour

This is a tough one. A problem with using this as a marker, is that high cognitive demands might fall only on one side of the partnership. Another is that different sets of cooperative behaviour could have wildly variant causes. Even the very same trait, might require markedly different cognitive abilities. An example of this might be plant domestication and the advent of agriculture. In humans this was associated with sophisticated techniques, and seemed to occur only after we acquired language ability. It was, arguably, one of our cleverest achievements, yet ants and termites have also done it. Every now and then a top scientist postulates that hive minds can be capable of great intelligence, and this could be true of the insect species that made the original transitions to agriculture that we are now witnessing in their descendant lines. A more conventional  answer is that there may be less cognitively demanding paths to agriculture.  This example was only given to illustrate how complex the relationship of this trait with intelligence may be. For interspecific cooperation the correlations are raw brain size, r=+.198, latitude range r=+.160.

Caregiving or Alloparenting

These traits require strong reciprocal social contracts that require years to unfold. Unsurprisingly, they correlate positively with our proxies for intelligence, though not as highly as you might expect (brain, r=+.202), range, r=+.052).

Vocalisation Complexity

There is much anecdotal evidence that vocalisation complexity and variety is higher in more cognitively gifted species. This seems to apply to cetaceans, and is one of three social metrics from Kieran et al. with a strong correlation to our proxies for intelligence. (brain, r=+.419, latitude range, r=+.351)

Social Play

Here we come across a metric with the most significant negative correlations to our proxies for intelligence. This will prove very useful as it allows us to separate species that are simply sociable from those that are truly intelligent (brain, r=-.220, range r=-.127)

Social Learning

By contrast with social play, social learning has a high positive correlation with intelligence (brain, r=+.438, range, r=+.392).

Social Defence

The third social metric that seemed heavily correlated with intelligence was social defence. Again, both brain size (r=+.487) and latitude range (r=+.279). Show high correlations, suggesting a deep cognitive divide between cetaceans who resist predation as individuals, and those who team up to do so. Even those high r values may understate the sharpness of the divide, because killer whales are seldom observed under attack. From accounts of Bering Strait fishermen, when they are attacked (always by male sperm whales), they go into a marguerite formation. Once that behaviour is videoed, or documented by experts, that correlation will become brain (+.556) and range (+.394), making this, the best of all the social indicators of intelligence.

Alliance Forming

There were too few cases of alliance forming for any meaningful analysis, only two having met the criteria for inclusion. In this context ‘alliance’ is similar to the human phenomenon of ‘war’. As such, many may like to think it would turn out to be negatively correlated to intelligence, but I suspect otherwise.

Making an Index

From the above correlations, the recorded social metrics fall into four classes: those that correlate highly and positively with our proxies for intelligence; those that correlate moderately positively; one that shows no significant correlation; and one that correlates negatively with those metrics. As such a reasonable index as to social intelligence, SI, as opposed to sociability, should be:

SI(1) = i +c + 2v + 2l + 2d – h – 2p (equation 1)

where i = interspecific cooperative behaviour, c = caregiving and alloparenting, v = vocalisation complexity, l = social learning, d = social defence, p = social play, h = group hunting,

The index of equation 1 shows a low positive, correlation with the csr, which was the species specific sociability index calculated Kieran et al*. It is desirable for the index of social intelligence to be normal to the index of sociability. This is particularly so both because of the strength of objective evidence that sociability correlates negatively with intelligence in other groups, and because their csr also correlates negatively with our proxies for intelligence in cetaceans. To this ends, I created equation 2, and solved its matrix for x: the SI and csr had a correlation of zero.

SI(x) = i +c + 2v + 2l + 2d – h – 2p – x.csr (equation 2)

The value returned was x=1.2. The usefulness of this methodology seemed confirmed by this value for x being very close to that which would have maximised the correlation of SI with both brain size and latitude range. This generates a final equation in which we can have moderate confidence.

SI(x) = i + c + 2v + 2l + 2d – h – 2p – 1.2csr

Then, as is standard, I converted these values to z scores. Now I can give you a list of the cetaceans with the highest Social Intelligence.

 z-score  top cetacean species by genus
 +2.49  sperm whale
 +2.25  bottlenose dolphin
 +1.93  humpback whale
 +1.92  beluga whale
 +1.48  short-finned pilot whale
 +0.71  fin whale
 +0.56  bowhead whale
 +0.51  killer whale
 +0.51  false killer whale
 +0.26  southern right whale
 +0.21  Chinese white dolphin
 +0.20  grey whale
 +0.20  striped dolphin
 +0.16  harbour porpoise
    0  common dolphin

This social intelligence index correlates very highly with brain size (r=.655) and latitude range (r=+.543).

I should note here that for primates (= the only group for which we have comprehensive data), a few (such as gorillas and gracile capucins) score far more highly in metrics of social intelligence than by more direct testing, but, for the majority of primate species, social intelligence follows general intelligence quite closely. I suspect that the bottlenose dolphin may also have an anomalously high social intelligence. Even if not, and even if the data collected on Kieran et al.’s dimensions were perfect, we would not expect the above figures to be accurate to within half a z-score. Combining the SI with raw brain size and latitude data would give us better estimates of cetacean general intelligence by species, though our confidence in their cognitive scores would still be significantly less than we can hold for most primate genera/species… to that purpose I have more work to do before year’s end.

*Their index is corrected for research effort, but the other factors are not. This introduces a small distortion into my calculations that I must fix in the coming year.