Thursday, July 26, 2012

Wrong, wrong, wrong

This week I read Simon Baron-Cohen’s book Zero Degrees of Empathy. I didn’t like it. He may well be best known among laypeople like me for his work on differences in the distributions of psychological traits among people of different sexes. That comes up in this book, but mostly incidentally, and in any case my problems with it aren’t anything to do with that.

Zero Degrees of Empathy is a book about empathy. Baron-Cohen (yes, he is Borat’s cousin) thinks that the common characteristic of three personality disorders and two types of autism is that people with the conditions have little or no empathy. He also thinks that we should stop explaining human cruelty in terms of evil, which is a bit of a non-explanation, and start explaining it in terms of a temporary or permanent deficit of empathy on the part of the perpetrators. I’m not sure how common it is these days to give a heavyweight role to evil in the explanation of cruelty, but if studying empathy helps us understand cruelty better, then empathy’s worth studying. No problems there.

My problem with the book is that it seems to be riddled with errors. Here are some of them:
  • On page 14 there is a graph showing that he’s found people’s empathy levels to be normally distributed. He says it’s a spectrum, but divides it into seven evenly spaced levels, from level 0 (no empathy) to level 6 (Desmond Tutu). The graph shows the mean/mode/median to be at level 3. Then on pages 17-20 he describes what each level is like. In the descriptions, people at level 3 have to ‘pretend to be normal’, and may realize that they ‘just don’t understand jokes that everyone else does’. Level 4 is ‘low-average’, and level 5 is ‘marginally above average’. So the descriptions suggest the average is between 4 and 5, and contrary to the graph empathy isn’t normally distributed. He repeats the claim that empathy levels are normally distributed many times in the book.
  • The same thing happens with his ‘systemizing’ spectrum: the graph on page 80 shows a normal distribution from 0 to 6 averaging 3, while the descriptions (page 80-81) suggest that the average is between levels 3 and 4, which isn’t what the graph says. It also means the distributions of empathy and systemizing are different, contrary to the impression given by the graphs looking exactly the same.
  • Here’s a quote from page 48: “If you have empathy you will be capable of feeling guilt, while if you lack empathy, you won’t. This might make you think that guilt and empathy is one and the same thing: clearly this cannot be true, since a person can feel guilt (e.g. that they went through a red traffic light) without necessarily feeling empathy. So empathy can give rise to guilt but guilt is not proof of empathy.” (Italics added.) The two italicized phrases contradict each other, don’t they? He also uses the expression ‘clearly’ a lot when making claims which are not clear and may not even be true, which is well known to be an annoying thing to do.
  • There is a diagram on page 31 illustrating the three forms of what he calls ‘zero-negative’ personalities. It looks just like a Venn diagram showing that zero-negative is the intersection of the borderline, psychopathic and narcissistic personality disorders. However, he wants to convey the quite different information that the three forms of zero-negative are those three personality disorders, and the diagram is pointlessly misleading.
  • He defines empathy on page 12 like this: “Empathy is our ability to identify what someone else is thinking or feeling, and to respond to their thoughts and feelings with an appropriate emotion.” That’s fine: it’s a reasonable definition. But in chapter 6 (‘Reflections on human cruelty’) he says that it follows from the definition that people commit various acts harming people (his examples involve murders and terrorist bombings) must have their empathy switched off at the time of action. This isn’t true, because there is no inconsistency in feeling an emotion and acting in spite of it. The line of thought that empathy is incompatible with some actions, even some praiseworthy actions (see note iv), appears frequently, especially in chapter 6, and it isn’t right.
  • Here is a quote from page 26: “…we react in a very sensory way when we identify with someone else’s distress. This clear brain response is telling us that even without any conscious decision to do so we must be putting ourselves into the other person’s shoes, not just to imagine how we would feel in their situation, but actually feeling it as if it had been our own sensation. No wonder we wince involuntarily when we see someone else get hurt. Of course, not everyone will have this strong empathic response to such emotionally charged situations. If our somatosensory cortex is damaged or temporarily disrupted, our ability to recognize other people’s emotions is significantly diminished. Surgeons may, for example, be well suited to their job precisely because they don’t have this emotional reaction, a prediction that was confirmed by Yawei Cheng who found that physicians who practise acupuncture show less somatosensory cortex activity while watching pictures of body parts being pricked by needles.” (Italics in original.) This suggests that a relaxed response to needles makes people more likely to go into acupuncture. He doesn’t explain why he doesn’t think the causation might be the other way round.
  • His definition of truth (pages 77-78) appears to be formulated to annoy philosophers. “Philosophers and theologians have long debated what we mean by truth. My definition of truth is neither mystical, nor divine, nor is it obscured by unnecessary philosophical complexity. Truth is (purely and simply) repeatable, verifiable patterns. Sometimes we call such patterns ‘laws’ or ‘rules’, but essentially they are just patterns.” I am a philosopher, and I am annoyed.
  • His view is that the variation in incidence of zero-negative personality types is explained in significant parts both by variation in genes and by variation in upbringing. This may well be true. He presents evidence for the genetic component (from twin studies and gene testing), but his evidence for the upbringing component seems to be that the parents of zero-negative people disproportionately often mistreated them. Such a correlation would (of course) result from a genetic connection even if there was no causal contribution from upbringing. Perhaps he has evidence that upbringing makes a difference, but it isn’t in the book as far as I can tell. This makes it irritating when on page 89 he says “we have seen bucket-loads of evidence for the importance of early experience”.
  • While he does present actual evidence for the genetic contribution, he also says this (pages 88-9): “there are parents who used the empathic, non-authoritarian style of parenting, discussing things reasonably with their child, yet their child still turns out to be a psychopath. Equally, we all know individuals who have thrived despite growing up in difficult environments… [Professor Dante Cicchetti] is proof that growing up in what James Blair calls a ‘dangerous and criminogenic’ environment does not totally determine your outcome. In his studies he found that as many as 80% of children who suffered abuse or neglect went on that [sic] to develop ‘disorganized attachment’. But clearly it takes more than a harsh environment to make a psychopath. There must be a genetic element.” Well, must there? It seems to me that this argument is easily parodied: disproportionately many people without access to clean water die in infancy, although not all do, and not all people who die in infancy lack clean water. Would this mean there was a genetic component? There could be a genetic component, but the rest of the variation might also be down to non-genetic chance factors. Drinking the wrong bit of water, perhaps. Or in the original case, having an encounter with the wrong person on the wrong day. Witnessing the wrong murder. Having the wrong friends. The possibilities are endless. These chance factors – what I understand the pros refer to as the unique environment – could explain the rest of the variation, without a genetic contribution. If I know this, Professor Baron-Cohen should know this.
  • Here’s his guess at why empathy is distributed as it is (page 128): “Presumably the reason that empathy is a bell curve (with the majority of people showing moderate rather than high levels of empathy) is because moderate empathy levels are most adaptive.” He goes on to tell a story about why this might be. But does he really think that when evolved traits are normally distributed we can presume that moderate levels are most adaptive? Does he think this applies to intelligence, beauty, physical fitness, height, fertility, resistance to disease and so on? Perhaps these things aren’t normally distributed, but I’m fairly sure some of them are. Normal distributions appear all over the place, and they don’t need explaining in terms of the mode being the most adaptive. Again, if I know this, so should he.

I could go on, but that’s enough. Perhaps some of my criticisms could be given satisfactory responses, but I'd be astonished if there wouldn't still be a lot for him to think about. Simon Baron-Cohen is a professor at Cambridge and has been in the business for thirty years. He shouldn’t be making mistakes like these. It’d be nice if you could just come across a book like this, write the author off as a hack, and ignore them in future. You can’t do that, though. People trust experts. Laypeople trust them, of course, but academics in related fields trust them too, because you can’t be an expert in everything. We trust mathematicians that the proof of Fermat’s Last Theorem doesn’t have any errors in it (anymore), and Steven Pinker (in chapter 18 of The Blank Slate) trusts Simon Baron-Cohen [see correction]. There’s a lot of division of labour in our collective search for truth, and misinformation spreads. Of course people make mistakes, but if experts took a reasonable amount of care in their work then we could trust that their simple mistakes were reasonably rare. But it seems they don’t, and it seems we can’t.


I said Steven Pinker trusted SBC in the chapter of The Blank Slate on gender. Sorry about this: I misremembered. He doesn't mention him there at all. In that chapter and others he relies on the work of a lot of experts in fields he isn't an expert in, so that part of the point still stands. SBC isn't one of those experts though, and is only briefly cited in the book, for his work on autism. Pinker and SBC do both hold the view that the average man isn't psychologically interchangeable with the average woman and that this is probably partly biologically caused, based on some of the same evidence. SBC's views on this (see especially his The Essential Difference) are however much more involved than anything Pinker defends in The Blank Slate. I might talk about TED in another post.


  1. Brilliant post - Pinker would normally be all over that antepenultimate bullet point as well so odd that you say he trusts Baron-Cohen (don't have the Blank Slate to hand so can't check exactly the context of what you're referring to.
    Am not sure what the cause of diagramm abuse is. Either he doesn't understand diagramms or he thinks he can get away with them not matching with what he's written because he thinks his readership don't understand them. Maybe this is one of the perils of people operating cross-disciplinarily.

  2. Pinker's book came out before this one of SBC's, so he can't be blamed for not mentioning it. I was referring to the stuff in The Blank Slate about psychological differences between men and women, which draws on SBC's work. SBC's popular book about that is called The Essential Difference, and I haven't read that.