Yesterday I read a disturbing news report
Feeding children a diet
rich in fish could prevent violent and anti-social behaviour in their
teens, according to research to be announced this week which suggests
the root causes of crime may be biological rather than social.
Why is this disturbing? Primarily because this is an extraordinary
claim and so it needs extraordinary evidence - evidence that does not
appear to exist. There are two major problems that I'm going to
discuss. First, whether the newspaper report is an accurate
representation of the scientific paper and, second, whether the claims
in the scientific paper are justified. I'm going to talk about both as I go because they are interwoven.
Here are the resources you might find interesting if you want to see the original reports:
Guardian newspaper article
Abstract of research paper
From the abstract:
children were assigned to an experimental enrichment program from ages
3 to 5 years and matched on temperament, nutritional, cognitive,
autonomic, and demographic variables with 355 children who experienced
usual community conditions (control group).
The children studied in this group all live in Mauritius. This brings
us to the first interpretational problem. Criminality is extremely
culture-dependent. That means that the study here has little
generalizability. (i.e. we can't extrapolate from these results and
think that they would be similar in other cultures, such as the US or
the UK, both very different from Mauritius.) Lack of
generalizability is not a problem as such, as long as it is always kept
in mind. Seeing as the study is done in Mauritius, there is a chance it
could be of great benefit to Mauritiuan communities. However, the
newspaper article does not even mention Mauritius until the 5th
paragraph and implies that the results are generalizable to other
countries. The abstract of the research paper doesn't mention Mauritius
Something else in the abstract set off warning bells in my mind - the
idea of matching on temperament. I have spoken with a number of
criminologists and psychologists who say that the idea of being able to
quantify or classify three-year old children on the basis of
"temperament" is ludicrous and impossible. But we'll let that slide for
the moment. We'll also let slide the strange use of 83 experimental
subjects and 355 control subjects. Well-known theorems in statistics
suggest this is a silly way of doing things - most of the control
subjects are wasted and the results would be nearly as precise with a
far smaller number. It would have been better use of resources to do
The main conclusion of the paper is that children who went through the
enrichment program between ages 3 and 5 ended up being less likely to
show schizotypal personalities and criminal behaviors at ages 17 and
23. The abstract states:
The beneficial effects of the intervention were greater for children who showed signs of malnutrition at age 3 years, particularly with respect to outcomes for schizotypy at ages 17 and 23 and for antisocial behavior at age 17.
Notice here that the connection between malnutrition and criminality is
not the most significant of the findings - which seems at odds with the
implications of the newspaper story.
But really, these points are relatively minor compared with
something else that is apparent from the newspaper article and
abstract. The experimental group went through an "intensive programme of
enriched diet, exercise and cognitive stimulation, which included being
read to and involved in conversation," according to the newspaper article.
We already know that there is a strong link between good socialization
and lowered criminality. We also know there is a strong link between
higher education levels and lower criminality. So why would you claim
that the main reason for lowered criminality in the study was the fish?
The situation is a little like hearing "listening to Bach makes you a
better runner" and then finding out that the experimental group were
provided with "Bach listening sessions and running coaching".
Now it is possible that the fish played a role. That is because there
is a small amount of evidence suggesting that eating fish helps
learning, etc. So perhaps eating fish made the children better able to
learn and therefore less likely to be criminal. But we expect this only
to be really significant where malnutrition is a problem, and so we
confront the generalizability problem again.
From reading the abstract, the authors do not make strong claims that
it is the fish that is key, however the following quote appears in the
newspaper article from one of the authors of the research paper:
'Could it be the exposure
to increased omega 3 fatty acids, which we know are the building blocks
of cell membranes, leads to better brain function which we did discern
at age 11 - and better outcomes at 23?' said Raine.
The way the newspaper article is written implies that the author is
suggesting the fish diet is responsible for the reduced criminality. I
am not convinced that the researchers would support such an
implication. What this does tell us is that Raine is probably aware of
the intermediate step in this process from fish diet to better learning
to lower criminality. That being the case, why design an experiment
that appears to cut out exploration of the middle step and confound the matter so heavily?
One reponse to this whole issue is to just say that the study doesn't
really say much that is interesting that we didn't already know.
However, the way it is reported has very serious ramifications.
Notice that the newspaper article is written by a political reporter
rather than the science reporter. This already suggests something about
the way this research is seen in the wider community. Then look at the
subtitle of the article: "New study reveals that the root cause of crime may be biological, not social."
Does this study really support the idea that the cause of crime is
biological? Absolutely not, except in the sense that amount of
learning, etc. is biological. It certainly doesn't support the fairly
widely discredited theories that criminality is dependant on intrinsic
biological properties such as genetics, but those are the types of
ideas that people usually associate with "biological causes of crime".
The topic is important because if biological causes of crime become
strongly accepted, we suddenly have a situation where defendants claim
that they have some biological imperative to be criminals and so the
crime is, in some sense, not their fault. This argument is already
doing the rounds in US courts and is generally dismissed, although not
sufficiently often (remember the "Twinkie defense"?). It also gives
excuses to avoid dealing with crime and criminality in other ways (ways
that tend to be politically difficult or expensive). If we are going to
start claiming there is a biological imperative for crime and making
policy based on those claims, we had better have some good evidence.
This paper is not that evidence.
All in all, the newspaper report does not seem to fairly represent what
was involved in the research (but when did a political correspondent
ever get a science story right?) and the science itself does not seem
to make any supportable extraordinary claims that are new.
To conclude, perhaps it worth repeating what one criminologist
commented to me about this paper and news report: "It is this sort of
thing that gives psychology a bad name."