Forming a Scale for Cetacean Intelligence

Nonhuman primates were the only group for which we had comprehensive cognitive data. From their study came two surprising findings that feature heavily on this website. The first of these is that, across multiple cognitive fields, a single factor can be extracted that explains most of the interspecific variation in ability. This means, quite simply, that we can describe the intelligence of any primate species by a single number. Furthermore, I have noted that mammals, but not birds, on which multiple types of cognitive test have been performed, tend to slot to an unambiguous position on this single dimension primate scale. That is why I feel it a reasonable assumption that a general intelligence factor exists across all mammals.

The second key finding among those primates, was that raw brain size explains most of the variance in this interspecific g factor, and that adjustments of body size for brain size tend to weaken the correlation. The most traditional adjustment is know as the encephalisation quotient, and turns out to be particularly good at destroying the signal of a connection between brain size and cognition. Certain observations, such as the comparison of elephants with humans, or cows with chimps, indicate that this strong link to raw brain size doesn’t generalise across the all mammal groups. That said, Suzana Herculano-Houzel has made a strong case that if the metric of brain size is replaced by forebrain neuron count, the correlation with intelligence will retain its strength across all (warm blooded) tetrapods.

If neuron count does turn out to be the best single physiological indicator of intelligence, and we gained quality count data on far more species than available today, this still wouldn’t be a particularly comprehensive metric, as the expected error between actual intelligence and neuron-calculated intelligence should have a standard deviation that is about the magnitude of the intelligence difference between a chimp and a gorilla, or that between a seven year old human and an adult. For whales, we have very little neuron count data, but the analysis of Kieran et al. does give the most comprehensive data yet on cetacean brain size by species. Creating a rough scale by this alone, so that we may flag the most intelligent cetaceans, is exacerbated by whale brain weights not being log-normally distributed. Unfortunately, their weights cluster near the upper physiological limit (this is due to rotational and shearing wave problems). By way of comparison, the sixth largest cetacean brain by genus is less than the largest by a factor of two, whereas the fifth largest primate brain is smaller than ours by a factor of seven.

Thankfully, Kieran et al. give several other types of data, also known to be associated with a species cognitive level. Let us now go through these.

Latitude Range

For primates, the extent of their historic range correlates highly with intelligence. It has been argued that, the flexibility needed to inhabit many different types of environment requires high levels of general intelligence. As a proxy for this Kieran et al. used a whale species recorded latitude range. Their metric correlates highly with raw brain size (r = +.595). This relationship is sufficiently strong that I will, from now on, use a whale’s recorded latitude range as a second indicator of its intelligence.


In primates, dietary range correlates highly with intelligence. In fact, when combined with brain size, it explains about 80% of all variance. With cetaceans, it shows a low correlation to brain size (+.085) or latitude range (+.243). There could be several reasons for this poor correlation. It might be that their very broad classification of prey types hides the true range of variation, or it might be that whaling has marginalised the largest cetaceans, with the biggest brains, to a fraction of their natural food range. In sperm whales this can be seen in early whaling records indicating that most of their stomachs were fish-filled, to more scientific surveys of a century or two later, recording that their diet consisted mostly of deep-water squid. A third option is that it might be due to the limitations of baleen system used by the largest whales. If due to post-whaling, or baleen. diet distortions, then this metric should be improved by restricting the dataset to cetaceans up to the size of orca. If we do, the correlation becomes +.285 with brain size and +.399 with range. This verges on useful. I suspect that dietary range will turn out to be a better predictor of cetacean intelligence than this suggests, however it looks like the particular metric they used in this meta-analysis can only be of limited use to us here.

Corrected Social Repertoire

Here Kieran et al. find the principal component of sociability, then correct it for the research effort. Large datasets on other mammals tend to go against the Machiavellian Intelligence Hypothesis, and find the more sociable the species, the less intelligent, with solitary species displaying the highest intelligence of all. This almost shows up here, with Kieran et al.s sociability index have a negative sign to its correlation coefficient for both raw brain size (-.003) and latitude range (-.064). Note that, as in the other surveys, the correlation is very weak, but in large enough datasets it has, so far, tended towards significance. That said, some sociable tasks require more intelligence than others, and that subset should correlate positively with intelligence. Kieran et al. give data on eight  different behaviours. This allows the possibility of creating a new index made by subtracting the least cognitively demanding tasks from the most demanding. This is a task in itself, and will be the topic for a separate post.

Group Size

As you might have expected by now, the most significant negative correlation with intelligence came from group size. Some social tasks may indicate very high cognitive skills, but this is a metric of sociability in its rawest form. Log group size inversely correlates with log brain size (-.162) though it shows a slight positive correlation to latitude range (+.067)


The measures of raw brain size from the latest meta-analysis, and its strong correlation to their latitude range data, give us a basis on which to draw a fair to moderate scale for cetacean intelligence. If we can find a third independent metric that correlates highly with these first two, then we might be able to construct a scale of more practical use. By my calculations, we have just that within the dimensions of their social repertoire, as I will explain in a later post..