Monday, July 28, 2014

What percent nature? What percent nurture?

The Nature versus Nurture debate is over: we no longer ask if genetics governs personality or if environment does. They work together, and it’s hard to pick their effects apart. But surely we can pick their effects apart a little? For example, if a dog trainer is trying to impress upon their students the importance of getting a puppy from a good breeder who takes behavior into account — or conversely, the importance of bringing a new puppy to a puppy class: what should she tell them? 50/50? 60/40? Surely there are some numbers we can cite?

It’s a tough question, but one that researchers have tackled. The concept is called heritability: the measurement of how much of a trait is due to genetic influences, and how much is due to environment.

Human researchers have it easier than dog researchers, because humans sometimes produce identical twins, and twin and adoption studies form the basis of human heritability studies. Some twins are identical (100% identical genetics), some are fraternal (around 50% similar genetics); some are raised in the same home, and some are adopted out and raised separately. You can do some complex math to all of these situations and come out with conclusions about particular traits. Identical twins more similar than fraternal twins for a particular trait? Strong genetic component. Raised together twins more similar than raised apart twins? Strong environmental component.

These studies have given us some numbers: IQ (how someone scores on a particular standardized test) is about 40-50% heritable. Environment does the rest.

Dog studies are harder. Dogs don’t have identical twins. Theoretically, the best way to study the heritability of personality traits in dogs would be to breed parents who do or do not show the trait in question and assess the puppies, then rinse, wash, and repeat for several generations. But this is expensive and somewhat ethically fraught to do in a laboratory, so we fall back on finding populations of dogs whose personality traits have been well measured and whose pedigrees are well known.

How often does that happen? Not very. But there is a test, the Swedish Dog Mentality Assessment (DMA), which is given to a large percentage of dogs in Sweden and some other European countries. Those crazy, overly-responsible Europeans measure their dogs’ personalities before breeding them, to make sure they're breeding stable dogs. Researchers have mined this resource repeatedly to learn more about the heritability of a variety of personality traits.

As lucky as we are to have this resource, it’s not an ideal one. The DMA is a suite of behavioral assessments which are given to a dog on a particular day in a strange environment by a judge who doesn’t know the dog well. Ideally, personality is best measured over time, by someone who knows the animal very well — its owner. And, in fact, every study I read that evaluated heritability of personality using the DMA noted that one of the most important factors was not genetics but the identity of the judge who gave the test. Did some judges tend to judge more severely than others? Did dogs respond differently (more or less fearfully, perhaps) to different judges? Hard to say, but we know that the reliability of the test suffered as a result.

Perhaps more alarmingly, we’re not really sure about the validity of the test, either. What are these assessments actually measuring? They’re measuring the response of a dog to a particular stimulus in a particular situation. Can this response be generalized to a personality trait? If the dog reacts fearfully to a person wearing a sheet over his head so he looks like a ghost, does that mean the dog is fearful or just that this was a particularly surprising experience? The DMA asserts that it measures playfulness, chase-proneness, curiosity/fearlessness, and most interestingly, aggressiveness. But does it? Studies of the validity of behavioral assessments in shelter dogs — a similar situation in which a series of small tests are given to a dog by a stranger in a strange situation — have repeatedly shown that the subtleties of personality are really hard to measure in this way.

Ideally, a personality heritability study would be designed using the canine behavioral assessment and research questionnaire (C-BARQ), a questionnaire which relies on the dog's owner to assess the dog’s personality through 101 questions. This test has been found to be valid and reliable. And the University of Pennsylvania has a database of the results of this test when given to thousands of different dogs. Except... they don’t have the pedigree information for many (or perhaps not for any) of these dogs. So this isn’t a practical solution, either.

So it’s hard, and I don’t really trust the studies that are out there as a result. What do these studies find? Most studies out there use the DMA or tests like it, and find roughly 20%-50% heritability for most personality traits studied. These numbers might be artificially low, though, because the tests may not be testing real traits — behavior that is stable over time.

I was able to find one study using the C-BARQ, which had much higher heritabilities, around 70%-100%. It's a dramatic difference, but I would hesitate to assign the responsibility for that difference entirely to the C-BARQ. This study used a non-random set of samples, selecting aggressive golden retrievers and dogs related to them. With no control set of non-aggressive goldens and unrelated animals, it’s hard to know how to interpret the study’s results.

So what are the real numbers? I still want to wriggle away from an answer. I don’t think we really know. I’d love to see a C-BARQ study using a random sample — maybe by finding pedigrees for dogs already in their database, if that’s possible. Until then, I’ll guess that the real answer falls in the 30%-60% range for most traits. But, in the end, does it really matter? Genetics are important and environment is important. The best genetics can fail in the face of a poor environment, and the best environment can fail in the face of poor genetics. We can’t predict everything about our next dog; we can just do our best to make a good decision, and then provide the best possible environment for whoever comes home with us.

I owe the inspiration for this post to my students in APDT's Canine Behavioral Genetics course, who asked about the balance of nature versus nurture and would not be satisfied with vague answers.

  • Strandberg E. & Peter Saetre (2005). Direct genetic, maternal and litter effects on behaviour in German shepherd dogs in Sweden, Livestock Production Science, 93 (1) 33-42. DOI:
  • Liinamo A.E., Peter A.J. Leegwater, Matthijs B.H. Schilder, Johan A.M. van Arendonk & Bernard A. van Oost (2007). Genetic variation in aggression-related traits in Golden Retriever dogs, Applied Animal Behaviour Science, 104 (1-2) 95-106. DOI:


  1. Svartberg did an interesting study where he compared DMA results to C-BARQ results... Sort of. And van der Waaij et al. suggested heritability changed dramatically between breeds and traits. I am not a geneticist, so don't really understand why heritability is kind of frowned upon. How reliable is it?

  2. K. Svartberg, B. Forkman
    Personality traits in the domestic dog (Canis familiaris)
    Appl. Anim. Behav. Sci., 79 (2) (2002), pp. 133–155
    apparently? I didn't know that one, I'll check it out. Thanks!

    I think the real problem with heritability is it's hard with behavior to tell what you're assessing. If you want to estimate heritability of height, it's easy to quantify height. If you want to measure heritability of aggression -- what's aggression? How do you quantify it? If a dog hasn't displayed any, does that mean they never will? Are there different kinds of aggression? (Obviously yes, but what are they?)

    So for me, that's why I don't put so much stock in measuring behavior traits. I think it's interesting to get heritability estimates, but important to remember that they're estimates -- take them as ranges or guesses rather than facts.

  3. Comparison to C-BARQ: Svartberg, K. 2005. A comparison of behaviour in test and in everyday life: evidence of three consistent boldness-related personality traits in dogs. Applied Animal Behaviour Science, 91, 103-128.

    Test-retest: Svartberg, K., Tapper, I., Temrin, H., Rades├Ąter, T. & Thorman, S. 2005. Consistency of personality traits in dogs. Animal Behaviour, 69, 283-291.

    Interesting blog post, as usual! :)

    Ann-Sofie Sundman (PhD student working with behavioural genetics in dogs)

    1. Thanks so much, Ann-Sofie.

      I'm curious what you work on specifically!

  4. Working on labradors and the two different lineages of show and gundogs. More or less seperated during the last couple of decades and thus a recent split. First of all I hope to be able to find some quantifiable and distinct behavioural difference between the two and then with for example a GWAS be able to find interesting genes responsible. Started out only half a year ago so only starting up everything. I do have access to DMA data for almost 3000 labs (this tradition of testing dogs in Sweden is a real nice thing :D ) and I'll be doing a new but quite similar test during the fall with invited dogs. We'll see where everything will end up :)

  5. Very cool. I'll be really interested to see how that works out! You and I should definitely stay in touch.