Jonathan Gordall quotes me in his article. Allow me to clarify.
I was asked by the Swedish Research Council to review some of the
critique against Gillberg, specifically the harsh critique against
Gillberg's work formulated by Eva Kärfve in her book"Brain Ghosts"
(available only in Swedish). My judgment (see below), as well as that of
professor Ottoson, concerned what was written by Eva Kärfve and whether
her critique of the Gillberg group was unfair and could be said to
constitute scientific misconduct. We were not asked, and found no
particular reason, to look into whether she was collaborating with the
scientology church or not - in fact this is irrelevant to the primary
conflict between Kärfve and Gillberg. Neither were we asked to, or did we,
look into the events leading up to the destruction of Gillberg’s large
data material. Both of these aspects are central in Gordall’s article and
in the 22 (so far) rapid responses.
The reviews commissioned by the Swedish Research Council in 2006 have
only been available in Swedish. I think they would help the reader of BMJ
to understand the roots of the controversy and to move beyond the bitter
accusations and counter-accusations. My review and that of professor Jan
Otto Ottoson came to similar conclusions. Finally the Swedish Research
Council followed our advice that Kärfve’s criticism should not be
dismissed as scientific misconduct.
Below is my statement for the Swedish Research Council.
To the Swedish Research Council
I, the undersigned, have been requested by the Swedish Research
Council to present my views on the Lund University communication of 26
March 2005 concerning the claim that Eva Kärfve had been guilty of
scientific misconduct. A preliminary approach was made to me in the late
autumn of 2005, and this was followed by a definite request in February
2006. The Lund University communication raises a number of issues. The
Swedish Research Council, however, primarily wanted me to take a position
on the question of whether Eva Kärfve’s research could be described as
scientifically dishonest on the basis of three passages in her book
Hjärnspöken (‘Brain Ghosts’) specified in the university communication.
My opinion on these three points is as follows:
1. Kärfve (page 15) writes that Gillberg and Landgren ignored or
dismissed findings that indicated the relevance of social factors, such as
social class or housing conditions, for minimal brain dysfunction (MBD).
Is this claim misleading or untrue?
Gillberg discussed psychosocial conditions and social class in a
number of different places in his doctoral thesis. He finds, for instance
(page 103, table III), that social class, poor housing areas and rented
flat accommodation are highly correlated to the MBD diagnosis. Similarly,
the mother’s stress load is significantly correlated to the MBD diagnosis
in the child. He nevertheless concludes (page 112) that “Social
disadvantage is in itself not an etiological factor”.
This certainly looks like a dismissal.
The keywords in seeking to understand Gillberg’s conclusion are in
itself. On page 112, he explains: “Social class, although in itself highly
correlated to the MBD diagnosis, was not in any way a factor directly
affecting the background variables studied.” Instead, the etiologically
operative background factors to which weight is attached are for instance
“prenatal non-optimal factors” and “hereditary non-optimality” etc.
Gillberg appears to be arguing that since social class and housing area
are not linked to these background factors, they cannot be of etiological
interest. But if social class and housing area are highly correlated to
MBD, despite not being correlated to variables such as “prenatal non-
optimal factors”, a reasonable conclusion would instead seem to be that
social class is an (‘upstream’) etiological factor that operates via some
other mechanism than the ones discussed above. Thus an important discovery
is left hanging in the air, without any interpretation.
Alternatively, social class may nevertheless have affected the
factors grouped under the heading “prenatal non-optimal factors”
(including for instance low weight at birth and premature birth) without
such a link being detected in this particular study, targeting as it does
a relatively limited number of persons (= low statistical power). In
Sweden, low birth weight and premature birth were more common among
working-class mothers and mothers with little education during this
period. Gillberg himself notes in his thesis the relevance of low birth
weight and “small for gestational age” as etiological factors for MBD
(pages 110–111). Thus it would have been reasonable to expect that
“prenatal non-optimal factors” would mediate the observed correlation from
social class to MBD in Gillberg’s study. Here, too, an unanswered question
is left hanging in the air.
The factors grouped under the heading of “hereditary non-optimality”
also include some with a social content. Late puberty among older
relatives is taken to be suboptimal heredity. But the age of entry into
puberty has been shown to be highly differentiated by social class in all
countries where the matter has been examined. This background factor,
therefore, might equally well be interpreted as a social factor as an
hereditary one.
Kärfve may be wrong to argue that Gillberg ignored the impact of
social class and housing area on the development of neuropsychiatric
diagnosis. She is right, however, to argue that he dismissed them as
significant causal factors. Gillberg’s reasons for dismissing them are
hardly convincing, at least not in light of our current knowledge in this
area. Even if Gillberg had devoted greater attention to this issue, it
goes almost without saying that a sociologically trained person would want
to analyse this point in greater depth and to partly dispute it. Kärfve’s
criticism in this respect cannot therefore be described as illegitimate.
2. Kärfve (pages 49–55) discusses what is termed the Mariestad study
by Magnus Landgren, Christopher Gillberg et al. The study is included in
Landgren’s thesis. Lund University’s communication asks for comments on
what Kärfve says about this study on page 52 in her book. In describing
the authors’ work, she talks about them “rummaging through old patient
records”, accuses them of cynicism and urges that their work be rejected.
Kärfve’s tone is bantering. Is she misleading, scientifically dishonest or
propagating an untruth?
Kärfve comments on the fact that five children who were screened as
positive and whose parents subsequently declined to take part in the
clinicial study were nevertheless included in it. Landgren and his
colleagues give the children neuropsychiatric diagnoses with the aid of
patient records (“a thorough evaluation of all previous records and of the
screening results” [page I:5 in Landgren’s thesis]). The diagnoses are
given without the team having met the children. Examination of the records
led to five children being given the following diagnoses: motor perception
dysfunction, mental retardation, DAMP, ADHD, and in one case a combination
of ADHD/DAMP.
This procedure contrasts sharply with the account of how other
children in the study were given their diagnoses, namely through “…in-
depth neurodevelopmental/neuropsychiatric assessment. This comprised a
detailed history, psychiatric and neurodevelopmental examination,
neuropsychological assessment and evaluation of speech and language
performed by the author (ML), psychologists and speech therapists…. a
medical, developmental and behavioural history was taken at interview with
the parents, using a standardised interview schedule…etc.” (Page I:3 in
Landgren’s thesis.)
Diagnosis setting and diagnosis criteria are one of the most
controversial aspects of the Gillberg group’s research. There is good
reason to critically discuss the way these five diagnoses were made. The
study is a limited one, at least in terms of statistical ‘power’, and it
is not clear to what extent the addition of these five diagnosed children
to the other 58 diagnosed children has affected various conclusions in the
study. The methodological problem is left unsolved. Landgren’s discussion
fails to tackle the problem; instead, the mothers’ reluctance to take part
in the study is seen as possible confirmation of the neuropsychiatric
diagnosis given to the children. Nor is there any discussion of the
ethical problem of including the five children in the study against their
parents’ wishes.
I am of the opinion that Kärfve’s criticism on this point – despite
the severe language it is couched in – is neither dishonest, untrue nor
unreasonable.
3. Point 3 in the Lund University communication principally concerns
pages 45–55 in Kärfve’s book. These sections mainly deal with how
Gillberg’s and Landgren’s theses estimate the prevalence of MBD (Gillberg)
and DAMP and other diagnoses (Landgren). Gillberg makes specific estimates
of the prevalence in Sweden based on their studies. Kärfve is highly
critical of how Gillberg’s prevalence estimate is strongly influenced by
two cases that were transferred from the control group to the group with
MBD. As a result, MBD prevalence among boys is estimated at almost 10 per
cent. This illustrates how small changes in the material can have a major
impact on estimates. Kärfve has similar objections to Landgren’s
estimates.
Probably a more important problem concerning the estimation of
prevalence of such neuropsychiatric diagnoses among children in Sweden is
the extent to which the authors’ material is selected. Even if the
intention is to base the study on the population as a whole, a step by
step process occurs until those who are to take part in the study are
finally selected. The selection covers such aspects as the researchers’
choice of study venue, whether the children attend preschool, parental
decisions whether or not to take part in a survey, decisions by preschool
staff whether or not to distribute the survey questionnaires, parental
decisions as to whether their children should be clinically examined or
not, and the researchers’ decisions whether to expand groups or move
people between groups. It is by no means certain (especially in the case
of the Göteborg study) that prevalence estimates are actually based on a
sample of children that is representative of the child population in each
venue, not to mention Sweden as a whole. A full discussion of possible
bias in the estimation is needed. Nor have Landgren and Gillberg included
any statistically calculated confidence intervals with their estimates.
This is otherwise common practice, especially if working with
representative samples. Thus it is difficult to express any opinion at all
on the value of Gillberg’s estimation from 1981 that 7.1 per cent of
Swedish children have MDB.
Viewed objectively, therefore, Kärfve’s criticism of what she calls
Gillberg and Landgren’s neuropsychiatric mathematics is not particularly
startling.
Conclusions
Lund University has asked for an assessment of certain passages in
Eva Kärfve’s book, ‘Brain Ghosts’. Are they examples of scientific
misconduct? Kärfve does not pursue any neuropsychiatric research of her
own – in her book she makes no reference to publications of her own in
this field – and can therefore hardly be accused of scientific misconduct
in the sense of having invented her observations, falsified her findings
or showing negligence in the presentation of her data. Original research –
the base on which scientific knowledge is built – must of course be the
area of activity subjected to the closest scrutiny, whether cheating or
dishonesty is suspected or not.
In the present case, the question is whether Kärfve’s critique of
research undertaken by others is dishonest. All three points raised in the
Lund University communication concern Kärfve’s discussion of the Gillberg
group’s research, not her own studies (in the same book) concerning the
ideological roots of some of the ideas that both laypersons and
professionally trained experts possess/have possessed concerning mental
ill-health and its causes. Scientific critique should also be subject to
scrutiny, of course, but can the same criteria be applied?
Scientific critique, whether strongly polemical or not, should in my
opinion be considered a legitimate activity even when the person levelling
the criticism does not primarily belong to the research community being
criticised. Advancing one’s criticism outside academia, as part of the
public discourse, is also legitimate. Normally, scientific critique helps
improve the research in question. Research controversies, even when marred
by irrelevancies, often generate new perspectives on old truths or
unresolved scientific issues. In that sense, scientific critique is one of
the conditions of research and a prerequisite for knowledge growth.
Kärfve’s sociological expertise means that she is competent to assess
various aspects (but not all) of the Gillberg group’s research. Research
methodology and analyses of causal links are (or should be) essentially
the same in all disciplines primarily concerned with studying human beings
and human society. If they nevertheless differ, there is every reason to
express oneself with care and to carefully encourage interdisciplinary
understanding.
Kärfve’s book is largely a polemical publication of a general nature
– primarily intended, perhaps, to influence the community at large and
policymakers, and only as a secondary consideration addressing the
specialised circle of people working scientifically with these matters.
One might take the view that Kärfve is unnecessarily disputatious in tone,
or sometimes goes a bit far, but writing a polemical publication on
scientific issues that is partly or largely aimed at a general readership
can hardly be equated with scientific misconduct. On the contrary, it is a
time-honoured tradition in many scientific fields.
I believe it would be of benefit to the scientific discourse if Eva
Kärfve were also to express her views precisely and scientifically in the
sociological or medical science press. An unfortunate aspect of the
conflict currently surrounding neuropsychiatry is that it risks drawing up
unproductive battle lines between social scientists and the medical
profession. Ranging the ‘biological’ against the ‘sociological’, or
‘biologism’ against ‘sociologism’, may be popular nowadays, but it is
totally fruitless. Most biological processes are affected by people’s
relations to one another, i.e. by society. Equally, social processes are
affected by biology. If we are to understand how, we need a dialogue
between disciplines. If the Swedish Research Council has a part to play in
connection with the Kärfve-Gillberg conflict, it should be to promote such
a dialogue and to prevent bloc-building and disciplinary trench warfare.
Some unique research material has been destroyed. The development of
children’s mental health in modern Swedish society is in many ways a cause
for concern. We need to bury the hatchets.
Stockholm, 20 March 2006
Denny Vågerö
Professor, Member of the Royal Swedish Academy of Sciences
(Translation by Stephen Croall)
Competing interests:
None declared