Notes on Progress is a new diary-style format from Works in Progress. If you only want to hear when we have a new issue, you can opt out of Notes in Progress here.
In this issue, Rosalind Arden writes about how cognitive ageing in dogs can help us to understand dementia and Alzheimer’s in humans. If you enjoy it, please share it on social media or by forwarding it to anyone you think might find it interesting.
My esteemed friend and colleague, Mark Adams, and I published a couple of papers about intelligence in dogs. This has led, occasionally, to the mistaken belief that I have dogspertise. I don’t. I feel much the same about dogs as I do about children: I like mine but have zero interest in the general child phenom. I am extremely fond of my ‘Significant Other’ but am not interested in the class of Significant Others in general. I agree with Ricky Gervais that dogs can be better than people; yet I assert that some dogs are the devil’s spawn. As are some owners. Like people’s, dogs’ characters vary. And morphologically, even more so – they are the most bodily variable species on earth. Adult dogs can vary in weight between 2 lbs (Teacup Chihuahua) and 175 lbs (Great Dane). That’s a factor of 87. Humans are the second most morphologically variable species. Our species’ average adult weight varies between around 130 (in Asia) to 180 lbs (North America); that’s a factor of about 1.3. So, the dogs have it. But I never wanted to own one.
Very much against my wishes, the family Last Lamb ordered a dog so she’d have something that would “love me when you’re beastly cross”. Spot the precocious manipulator. I shouldn’t have fallen for it. Especially as she was nearing her 30’s or was it earlier? Motherhood is so tiring; one gets muddled about timelines. Anyway, there we were with a puppy looking like a damp guinea pig. You get used to sleeping, it turns out, with one arm dangling to the floor so doglet has the comforting touch of the warm, circulation-free hand. All night. Naturally one’s days are a little taken up with the 20-minute timers set for an outside run so the item at issue can pee. That fitted in fine with doing a PhD; one doesn’t have to concentrate all day, surely? Last Child was Too Busy, apparently. Taken up with either elementary school or running a bank, I forget which.
Sorry, I’m getting distracted by bitterness… So, the months passed and the piglet-like puppy grew into a magnificent border collie with a distinctive temperament: immense loyalty, quick to catch on, and a strong sense of what was unsuitable. She would tolerate neither ugly headwear nor partial nudity in the wrong setting. Having been left holding the dog-baby, and knowing nothing about the species, it was highly salient when my expectations about dog behaviour were violated. I had expected that Jenny would learn to connect the sound ‘sit’ with the required action, that she would learn to come when I called her and flung my arms out with a sound approximating ‘think how fun it will be when chase-the-squirrel is over, and you can have your non-extendable leash clipped back on’. What Violated the Expectations was that she learned to correctly assign various sounds to their referents without their having been taught or reinforced. For example, I’d say aloud to a human, ‘hey, I bet David will be here soon’, and Jenny would go to the only window from which the drive could be seen. Or, I’d say to David, ‘hey look at that bird’ and Jenny would immediately look up to the sky. That kinda made me think.
Dogs are, it must be said, well-known for their ability to form associations. The whole Pavlov thing. But the bird thing surprised me because it was free of any reward or motivation. It showed me that dogs could a) learn stuff and that b) they could likely learn quite a bit of stuff and that c) some dogs would be much better at learning stuff than others. You can see what a star student I was; how quick to catch-on.
Robert Plomin generously gave me the opportunity, as a post doc at King’s College London, to ask a question we were both interested in. The question was this: do dogs’ cognitive abilities overlap, as ours do? Let me spool back a bit to explain why this was worth asking. You’ll likely know, gentle reader, that a potted history of research on human intelligence, goes something like this. Francis Galton, the half-cousin of Chas Darwin, notices that intellectual ability tends to run in families.
Charles Spearman, a retired Major from the British army, while teaching late schoolboys and sour prentices, noticed that boys who are good at English are not too shabby at sums. He developed Galton’s correlation coefficient which measures the extent to which different test scores are in agreement. He then examined grade scores from a range of subjects and found a general tendency for like to go with like. A student who is above average in one subject tends to do fine in others. Spearman re-worked the correlation coefficient (technically tetrad differences) and brought forth what became factor analysis. This statistical technique is a major advance. Instead of asking ‘is A associated with B?’, he can explore all the pairwise associations among A,B,C,D, E and F simultaneously.
Now, Spearman can capture any overlap among grades across all the subjects he teaches. The statistical overlap that he found is the most useful and important property ever to have emerged from psychology. It’s an index of how dreadful academia can be that students can gain a degree in psychology (or the social sciences) and have neither heard of it nor understand its significance. Spearman called his discovery the g factor where g is lower case, italicised, and stands for general intelligence. Big nomenclature mistake there Charles my lad, but we’ll park that for another time.
Spearman’s small god of the overlaps, g, emerges as a statistical property when, for example, you administer to a decent number of people (100 or 10,000 say), a set of at least four cognitive ability tests (verbal reasoning, find the next shape in a series, even vocabulary or straight mathematics tests will do. But g is not an ability; g taps into something about our brains, but it’s important to note that it isn’t a ‘thing,’ like a structure, or even a place in the brain. It’s useful because it carries more signal than any single test. You likely know someone whom you consider bright but they’re a horror at long division or spelling. The value of g is partly that noise is ineluctably present in a single test; since g captures a general tendency across several tests it is a more useful signal.
The measurement of intelligence improved after Spearman because of the statistical noise reduction techniques that he and others developed. By the time large-scale studies had been established (such as the massive Twins Early Development Study, founded by Robert Plomin at King’s College London), one could test the reliability of g empirically, by using four different cognitive tests on the same people at various ages and then assessing whether the extracted g roughly preserves the rank order of scores as the cohort ages and takes different tests. Because we can assess intelligence more reliably by using g to index it, we are now in a much better position to find out what life outcomes are associated with being brighter or less bright. The answers are somewhat surprising.
It turns out that intelligence or g, assessed in adolescence, is among the most potent single predictors of later health (brighter people have better health). Early life intelligence gives even smoking – the giant among epidemiological risk factors – a run for its money, coming in first or second in many diseases. And low intelligence is the earliest single predictor of dementia risk. Please pay attention to the word single. A person who is very bright can develop any illness including dementia. Lower intelligence does not necessitate later ill-health or dementia. Intelligence is a ‘risk factor’ among other risk factors (smoking, deprivation, binge-drinking and so on). Yet it is surprising that it packs so much punch as a single predictor. Cancers seem to be excluded from this relationship – intelligence is uncorrelated with non-smoking related cancers. But an association or correlation between X and Y, is uninformative about causes. Intelligence may play no causal role in the epidemiology of disease. But we were talking about dogs, and I’ve drifted. Dogs may be useful to us for an oblique reason. The age-pyramid of our species is changing for the first time since people evolved. I’ll explain the relevance.
Since we’ve been capering about on the solid part of our Earth, there have been more young ‘uns than old ‘uns at any one time. Imagine a pyramid with the young at the base and the oldest old at the top. But this pattern is changing. As countries go through the ‘demographic transition’ in which family size drops – sometimes below replacement – the pyramid is slowly melting and transforming. Some predictions suggest that it will even invert; then the few Zoomers at the bottom must hold up a huge weight of Boomers and Crusties. This is a global phenomenon. What happens if the Boomers get sick? Alzheimer’s and dementia are diseases of the old. They are deeply not fun. So far dementia is untreatable, and expensive. Zoomers and their grandchildren might be overwhelmed by this aspect of the living old with its burden of care and expense. Making progress with understanding and treating dementia is highly consequential. It’s proving a gnarly problem. Animal models have contributed much to understanding other diseases. But it turns out that not many species suffer from dementia. But dogs do.
Polar bears, I’ve read, can get dementia, but they are challenging laboratory animals preferring seal-sandwiches at luncheon, and there is that problem of their large claws. Mice are more tractable on the claws/luncheon front, but they don’t become demented. Transgenic mice can be built to reproduce some of the features of dementia, but it costs around $400 per mouse, and you don’t see natural disease progression in the mouse. You just engineer a snapshot of the disease state. Yet around 14-20% of dogs naturally acquire dementia as they age. Their form, canine cognitive dysfunction, shares many features with ours. A dog that knew her way around the home, will come into a room, and forget how to get out, standing at the wrong side of an open door for ages. She may lose interest in food and her responses to people can change, becoming fearful. Problems with continence arise. Alzheimer’s in people is associated with brain changes including a burden of amyloid protein, tau tangles, and neuronal loss. All of the above clinical signs have been found in dogs post mortem. As with humans, no single sign (such as the presence of amyloid) is a perfect predictor of dementia, but the pattern of pathology in dogs and people looks relevant and similar.
One thing that we could do with dogs, I thought, would be to develop a full natural animal model of this dreadful disease, so that we could learn more about its pathophysiology, find out if there are early life signs indicating who might be at risk, and test treatments. It would be a win for both species. Dogs have many advantages over mice, as animal models of disease. They don’t need to live in the laboratory, they are widely available for study recruitment, they are variable, they are natural models (they get the disease without intervention), they don’t drink alcohol, smoke weed or cigarettes, they share lots of our environment and don’t have the range of socio-economic advantage or disadvantages that people do. The life of a corgi raised by a janitor does not differ significantly from the life of a corgi raised by the Monarch of the Realm. Yes, there are environmental differences, but nothing like the scale of the economic differences between the lives of children raised by janitors versus Monarchs.
These things matter because they otherwise confound how we understand relationships between measured traits and life outcomes. For example, reverting to humans and intelligence, brighter people often have cushier jobs, fewer environmental hazards, more cash for healthy food and so on. All these facts can confound a link between intelligence and health. So, dogs could play a role in understanding ageing and dementia. Perhaps. Worth a try anyway? Maybe? I asked John Hardy (now Sir John) the world-class dementia expert at University College London. John reckoned it would be worth a shot. Note for the record: John is incredibly generous, extremely unassuming, and everybody I met who knows him has a favourite ‘Hardy story’ showcasing his complete disregard for pomp coupled with a commitment to truth.
There’s a domino train of dependencies in the line of dog-coggery to dementia that I’ve outlined here. On the first tile there is a simple question – do dogs’ cognitive abilities overlap? If they do, overlapping cognitive abilities might be the animal-typical ‘build’; we could use that insight to learn more about how brains work. If we can assess intelligence in dogs reliably, we would have another way of probing what causes the association between being brighter and being healthier in a mammal that rarely falls down drunk on Fridays. In applied terms, if we found a g factor in dogs, we could use that g as a more reliable measurement of cognitive ability than a single test, a clinical outcome in dementia against which to test a treatment. Going to the dogs could be a smart move.