refrigerator mothers

You could be forgiven for assuming that the ‘refrigerator mothers’ theory for the cause of autism has been consigned to the wastebasket of history. That might be true for children with a formal diagnosis of autistic disorder, but parents are still often under suspicion if their children have autistic characteristics but no diagnosis, or indeed any unusual behavioural characteristics but no diagnosis. Bruno Bettelheim is often credited with inventing the term ‘refrigerator mothers’, but Leo Kanner appears to have come up with the refrigerator analogy first.

Leo Kanner

Leo Kanner

In the comment section at the end of his 1943 paper, Kanner weighs up the evidence for the possible causes of autistic behaviour. The children have schizophrenic characteristics, but their condition differs from schizophrenia because it’s been present from birth – suggesting a biological origin. On the other hand “in the whole group, there are very few really warmhearted fathers and mothers” (p. 250), suggesting that development could have been disturbed by parental behaviour. In the end, Kanner concludes “The children’s aloneness from the beginning of life makes it difficult to attribute the whole picture exclusively to the type of the early parental relations with our patients. We must, then assume that these children have come into the world with innate inability to form the usual, biologically provided affective contact with people, just as other children come into the world with innate physical or intellectual handcaps [sic]” (p. 250).

But Kanner later changes his mind. In 1949 he describes his patients as in refrigerators which did not defrost.* In his 1956 paper with Leon Eisenberg he contrasts the low incidence of psychosis and neurosis in the children’s relatives (a sample of around 1000) with the much higher incidence in families of children with schizophrenia and concludes “Thus, if one limits his search for genetic factors to overt psychotic and neurotic episodes in family members, the results would appear to be negative” (p.8). After some discussion of parental characteristics, he decides; “The emotional frigidity in the typical autistic family suggests a dynamic experiential factor in the genesis of the disorder in the child” (p.8) and “These children were, in general, conceived less out of a positive desire than out an acceptance of childbearing as part of the marital contract” (p.10). But less than a decade later, when Bernard Rimland, in his book Infantile Autism: The Syndrome and Its Implications for a Neural Theory of Behaviour suggested that autism might have its origins in the brainstem, Kanner wrote the foreword. And in 1969, Kanner told the first annual meeting of the then National Society for Autistic children “I herewith especially acquit you people as parents” (Feinstein, 2010).

The discussion in the 1956 paper sheds light on what initially appears to be wavering on Kanner’s part about the cause of autism – now he’s blaming biology, now he’s blaming the parents. Kanner wasn’t so much undecided as aware that both factors could be involved. He goes into some detail about the interaction of biological and environmental factors in producing autistic characteristics and describes early infantile autism as ‘a total psychobiological disorder’ – in other words, it isn’t a case of its cause being either genetic or parental. As Michael Rutter observed; “What we have to differentiate is evidence of a broader phenotype. Kanner switched back and forward, which is a mark of his integrity” (Feinstein 2010).

Bruno Bettelheim

Bruno Bettelheim

For Bruno Bettelheim there was no uncertainty about the cause of autism. At first glance, Kanner and Bettelheim appear to have a good deal in common. They were both born into Jewish families in central Europe around the beginning of the 20th century. Both had their studies interrupted; Kanner by military service, Bettelheim by the death of his father. Both fled to the USA as ethnic refugees, both married and raised families there, and both became successful, respected figures in the field of child development. There the resemblance ends.

Bettelheim was born into a well-to-do Viennese family in 1903. As a teenager, he was fascinated by psychoanalysis and read all he could about it. He enrolled as a student of philosophy and history of art at the University of Vienna, but postponed his studies when his father died from syphilis and he had to take over the family lumber business. In 1930 Bruno married his first wife, Gina, who worked at a Montessori nursery. The couple took in a young American child, Patsy, whose mother had sent her to Vienna for therapy with Editha Sterba, a member of the Vienna Psychoanalytic Institute. Bruno was in therapy with Editha’s husband Richard for a while, and his connections with Patsy, the nursery and the Sterbas were to prove a turning point in his career.

Eventually, he was able to resume his studies and was awarded a doctorate in February 1938. Within a month, German troops had entered Austria and Gina had left for the USA with Patsy. Bruno remained in Vienna with his mother and sister. In June, he was arrested, jailed and then taken to Dachau. In September he was moved to Buchenwald and released the following April in an amnesty to mark Hitler’s 50th birthday. Patsy’s mother, Agnes, had managed to arrange a visa for him, and Bruno was reunited with Gina in the USA in May 1939. By then it was obvious that their marriage was over – both had had affairs – and Bruno settled down with Trude, a former girlfriend. His experience with psychoanalysis and child development also got him a job in the education department at the University of Chicago, which led to his appointment as director of the university’s Orthogenic School in 1944, where he was to remain until the early 1970s.

Most of these biographical details are from Richard Pollak’s fascinating biography of Bettelheim The Creation of Dr B. Pollak is aware that his account might be seen as biased. His younger brother, Stephen, had attended the Orthogenic School until his death in an accident whilst on holiday. Bettelheim’s disdain for the boys’ parents and his claim that Stephen had committed suicide despite Richard witnessing Stephen’s fall from a hayloft, and Bettelheim’s suicide in 1990, were what prompted Pollak to research Bettelheim’s life. Although Bettelheim probably had the children’s best interests at heart and certainly changed the Orthogenic School for the better, reactions to him were mixed to say the least. Comments from school staff, parents, children and his students are peppered with reports of admiration, intimidation and humiliation. There’s little doubt that he beat the children and there are some stories of sexual contact, although these accounts, if true, need to be set in context; corporal punishment was common at the time and the psychoanalytic theory embraced by Bettelheim saw sexual expression as natural. Some people hated him, but others felt that Bettelheim’s methods, even if frowned upon, were well-intentioned.

Refrigerator mothers

Probably more widespread harm was caused by Bettelheim’s view of parents, especially mothers. Bettelheim disapproved of children at his school going home for visits and viewed mothers as cold, uncaring and responsible for their children’s behavioural problems. Bettelheim set out his ideas in a series of books including The Empty Fortress: Infantile Autism and the Birth of the Self. Despite the popularity and influence of this book, it’s not clear how much experience of autistic children Bettelheim actually had. He claimed that two autistic children had lived at his home, although Patsy appears to have been the only one, and she wasn’t diagnosed as autistic. Visitors to the Orthogenic School commented on the fact that the children appeared to be normal, if troubled, kids. Bettelheim admitted only children with no physical or intellectual impairment, thus ruling out more severely autistic children. And the success of his techniques was called into question too. In the 1980s a study showed that during Bettelheim’s period as director Orthogenic School, of 220 children entering the school, only 13 were admitted with a diagnosis of autism (Bettelheim diagnosed many children himself) and not all the children had made the progress he claimed.

Aside from contrasts in their life experience, with regard to their contribution to autism research I want to highlight two key differences between Kanner and Bettelheim; the way they used evidence and their level of relevant expertise.

Use of evidence

Kanner had a medical background, derived testable hypotheses from the best theory available at the time, and wasn’t afraid to change his conclusions if the evidence dictated. Bettelheim had trained in philosophy and appears to have made up his mind in advance about the cause of autism and then selected evidence to support his theory. Pollak refers to an essay Bettelheim wrote about the philosophers who shaped his thinking as a student. Significantly, the philosophers – Lessing, Lange and Vaihinger – all saw historical truth as a construct of the mind. Vaihinger argued in his book The Philosophy of “As If” that even though fictions should not be mistaken for true propositions, they can work As If true (Pollak, 1997; p.15). Bettelheim seems to have put Vaihinger’s ideas into practice; in a cv compiled in 1942 he exaggerated his credentials and frequently reported events in a way which conflicted with the recollection of other witnesses.

Level of relevant expertise

Kanner was aware that his theories about the cause of autistic characteristics were limited by the biological knowledge available at the time. Nonetheless, he clearly understood the complexity of child development and was careful to rule out a number of possible causes for autism before arriving at his conclusions. Without doubt, Bettelheim was also knowledgeable, about psychoanalysis and theories of child development that is; the reference section in The Empty Fortress is extensive. But he appears to have had little knowledge about biology and his explanations for the children’s behaviour are in terms of psychoanalytic concepts only. Indeed, he was actively opposed to biological theories for the causes of autism, attacking Kanner and Rimland in The Empty Fortress and telling Thomas Kemper that his brain studies must indicate a different kind of autism (Feinstein, 2010).

Bettelheim’s influence

What puzzles me is how Bettelheim’s book could become so influential amongst professionals with medical training, long after research into genetics and brain function had shown that psychoanalytic and psychodynamic theories of child development were lacking. Bettelheim was preoccupied by psychoanalytic symbolism. He saw the children’s interest in balls, balloons, light fittings and automobile headlamps as symbolic of their relationship with the breast. Their words had deep symbolic meaning – ‘breakfast’ meant ‘break breast’, ‘Connecticut’ meant ‘connect-I-cut’ and even an interest in the weather on the part of a non-verbal child symbolized her fear that she might be devoured (‘weather’ meant ‘we/eat/her’). This should have been enough to suggest that his theory might not have a solid grounding. Bettelheim’s use of symbolism extended to his lectures, resulting in the (in)famous knitting/masturbation story. Bettelheim is reported to have said to a female student, knitting during one of his lectures; “Don’t you realize your knitting is nothing but a sublimated form of masturbation? You’re sitting in front of the entire class masturbating.” The student is alleged to have replied; “Dr Bettelheim, when I knit, I knit! And when I masturbate, I masturbate!”

Despite the lack of evidence for his theories they were very influential. The Empty Fortress was the first book on autism translated into Spanish, Bettelheim’s lectures were shown on French national tv, in the 1980s most of the books on autism in Danish libraries were by Bettelheim, and it wasn’t until 1998 that the International Association for Child and Adolescent Psychiatry and Allied Professions decided that “parents have absolutely no responsibility for their children’s autism” (Feinstein 2010). As Judy Barron, the parent of an autistic child, observed when she read The Empty Fortress, “I wasn’t a scientific reader and I certainly wasn’t a researcher; I was a twenty-four-year-old mother; but I just didn’t see any evidence to support his pronouncements” (Pollak, 1997, p. 275).

The use of evidence and levels of expertise; two factors that I plan to explore further in the next post.


Bettelheim, B. (1967). The Empty Fortress: Infantile Autism and the Birth of the Self. The Free Press.
Eisenberg, L. & Kanner, L. (1958). Early infantile autism 1943-1955. In C. F. Reed, I. E. Alexander and S. S. Tomkins (eds.) Psychopathology: A Source Book, Harvard University Press.
Feinstein, A (2010). A History of Autism. Wiley Blackwell.
Kanner, L. (1943). Autistic disturbances of affective contact. Nervous Child, 2, 217-250.
*Kanner L (1949). Problems of nosology and psychodynamics in early childhood autism. American Journal of Orthopsychiatry, 19, 416–26.
Pollak, Richard (1997). The Creation of Dr B, Simon & Schuster

*Paper behind paywall – the citation is from various sources.

image of Bettelheim: Ottofroehlich under Creative Commons licence


autism and its co-morbidities

In the previous post I looked at the different levels of complexity involved in classifying things, including behaviour. I want to apply this idea to the behaviours involved in autism.

Classifying autistic behaviours

Autism is defined in terms of impairments in three types of behaviour; social interaction, communication and flexibility (people diagnosed with autism show ‘restricted, repetitive and stereotyped patterns of behaviour’). The definition is at a high level of complexity – each type of behaviour involved is the equivalent of a class in the Linnaean system. Many behaviours come under the heading ‘communication’, an even larger number under ‘social interaction’ and ‘restricted, repetitive and stereotyped patterns of behaviour’ could refer to almost anything. It’s true that the DSM narrows things down a bit by listing certain aspects of impairments in these behavioural classes, but there’s still a lot of scope for variation between individuals. Even the specific behaviours mentioned, such as impaired eye-to-eye gaze, delayed speech or persistent preoccupation with parts of objects could have different causes in different people. Jon Brock has calculated the number of possible combinations of the 12 types of behaviour listed in the DSM criteria here; the figure would be much higher if individual variations in these were taken into account.

I’ve illustrated the complexity of the behavioural classes by mapping out the relationship between a low-level behaviour that would qualify as an impairment in the class of behaviours that we call ‘communication’. I’ve chosen one of the behaviours noted by Kanner – ‘pronominal reversal’ (reversal of pronouns). Pronoun errors are common in young children. Very young children often don’t use them at all – ‘Harry do it’ – or sometimes use only one as a default – everyone is referred to as ‘he’. ‘Pronoun error’ is ambiguous because it could mean several different things, but it’s very clear what’s meant by pronoun reversal. Ironically, despite listing ‘pronomial reversal’ as one of the essential common characteristics of his syndrome, in five of the six cases where Kanner records pronoun errors, the children don’t consistently reverse pronouns. The reversals appear to be due to the use of echolalia e.g. ‘you want candy’ instead of ‘I want candy’ or ‘want me to draw a spider’ for ‘I want you to draw a spider’. Jon Brock expands on this point in his blog.)

classification of pronoun reversal

My classification of this behaviour is not exhaustive or authoritative. Nor does it show all the linkages between levels; it simply illustrates the levels of complexity between communication in general and pronoun reversal in particular. Note also that the diagram illustrates the way pronoun reversal could be classified – it doesn’t reflect the way children actually develop language. As I pointed out in the previous post, each level of complexity is equally valid; pronoun reversal could accurately be described as an impairment of pronoun use, or of pragmatics, syntax, language, verbal communication or communication in general. But even if a description at each level of complexity is equally valid, it doesn’t mean each level of complexity is equally useful for all purposes.

Here’s the same diagram again, but this time showing the impairments in communication a doctor diagnosing autism would be looking for. One feature would need to be observed.

diagnostic criteria for impairments in communication

Levels of complexity and autism

Although the diagnostic criteria listed in the DSM appear at first glance to be precise, when mapped onto a diagrammatic representation of the set of behaviours involved in communication, it’s clear that the criteria are at quite a high level of complexity, so the impairments might manifest themselves in different ways and there might be more than one possible cause for each impairment.

A key problem with the definition of autism, as I see it, is that it is framed at too high a level of complexity to be useful in determining what’s causing an individual’s developmental impairments. There is no doubt that a child diagnosed with autistic disorder has impairments in social interaction, communication and flexible behaviour, because that’s how autistic disorder is defined. But because those classes of behaviour encompass so many low-level behaviours, people with totally different impairments to each other can all end up with a diagnosis of autistic disorder. And the atypicalities of people with autism aren’t confined to behaviours; some somatic (bodily) symptoms frequently occur alongside autism. They are referred to as co-morbidities.

Autism and its co-morbidities

When diagnosing a somatic disorder, doctors tend to look at a patient’s low-level symptoms – the type of breathing difficulty, cough, pain or rash. Many diagnostic tests are at an even lower level – the shape of cells or presence of specific molecules. The reason for this is that many somatic disorders have very similar symptoms, so it’s important to identify exactly what symptoms each patient has. In the case of autism the diagnostic process is different. A doctor might first rule out any medical conditions known to cause autistic characteristics (e.g. Fragile X) and might note all the patient’s symptoms, but what’s seen as important is not all the patient’s symptoms, but whether or not the patient’s behaviour fits the criteria for autistic disorder.

Because autistic disorder is defined in terms of behaviour, and because autistic behaviours are often the most obvious of a child’s symptoms, autistic disorder frequently becomes the primary diagnosis and any minor somatic symptoms associated with autism are relegated to the category of ‘co-morbidities’. (Also see Paul Whiteley’s blog on this topic). So sensory atypicalities, epilepsy, sleep disorders, hypermobile joints, growth abnormalities, dietary intolerances, digestive problems and allergies, unless serious, are seen as features that might or might not be caused by whatever causes the autistic characteristics. But because these somatic features aren’t typical of everyone diagnosed with autism, they don’t qualify as part of the autistic syndrome.

That’s fair enough if what you want to do is refine the definition of autism. But if you want to find out what might be causing an individual child’s developmental abnormalities, you need to look at all the child’s symptoms, autism being only one of them.

Harry and Sam

To illustrate this point, I want to look at the symptoms shown by Harry and Sam. Harry’s intelligent and has an excellent grasp of language and abstract ideas. He loves maths and science and has a photographic memory. But he can’t discriminate between some speech sounds or between some letters. He can’t keep track of what people are saying in noisy environments or unless they speak slowly and clearly, and his own speech is hesitant and mumbled. High-pitched sounds, such as children screaming, hurt his ears. He loses track of things that move quickly, like balls or other children, and his motor co-ordination is poor. If he puts his head back he feels sick and dizzy and sometimes blacks out. He finds light touch unbearably tickly but has a high pain threshold. He has a persistent dry cough, can fold his legs behind his head and often gets pain in his legs and chest when walking. He’s a fussy eater, is extremely skinny and his chest is sunken. Certain foods give him diarrhoea, and a casein-free diet has improved his energy levels considerably. Not surprisingly, Harry struggles with conversation, avoids social interaction if he can, and, because he finds almost everything he tries to do difficult, sticks to things he can do best and with least discomfort, like reading, lego, computer games or watching tv. But Harry’s diagnosis is ‘autistic disorder’; his specific problems with receptive and expressive speech, visual discrimination and tracking, auditory hyperacusis, motor and vestibular function, pain, joint mobility, growth and dietary issues just don’t come into the picture. There’s a group of children like Harry with almost identical symptoms at the ‘high-functioning’ end of the autistic spectrum. Their symptoms aren’t just behavioural, but involve sensory processing, motor function, skeletal development and digestion as well. Because the children have symptoms that are so similar (unlike the wide variation seen in ‘autistic disorder’) it’s quite possible that this group of children all have the same cause for their symptoms.

There are other groups with distinctive symptoms on the autistic spectrum too, such as ‘low-functioning’ children like Sam. Sam has little speech, intellectual impairment and poor motor control. He’s big for his age, has epilepsy, pica, is doubly incontinent and smears faeces. He often has tantrums, is aggressive to others, beats his head and throws himself against the walls and furniture. Sam also has a diagnosis of autistic disorder. Do Harry and Sam have the same medical condition? Unlikely, I’d say.

Note that I’m not saying the autistic spectrum can be divided into two sub-groups – ‘high-functioning’ and ‘low-functioning’. There are children across the autistic spectrum who have different symptoms to both Harry and Sam. I’ve just described patterns of symptoms that I’m most familiar with. I also want to emphasise that neither group, Harry’s or Sam’s, would form a sub-group of ‘autism’ because autism is only one of their symptoms. They have symptoms other than behavioural ones, but these are currently classified as ‘co-morbidities’ of autism, rather than being seen as symptoms in their own right.

Biological markers for genetic variations

There’s been a great deal of interest in recent years in the genetics of autism. The Autism Genome Project identified a number of genetic ‘rare variants’ associated with autistic disorders. What geneticists are keen to identify are biological markers that indicate the endophenotype that results from a specific genetic variation. A major problem with this approach is that because the diagnostic criteria for autistic disorder lump together everyone who has autism as one of their symptoms, scientists often aren’t looking for the biological markers for genetic variations in a specific individual, but for the biological markers for genetic variations in everybody diagnosed with autism. And only for the markers associated with autism, not with the other symptoms that individuals might have. It’s quite possible that these much sought-after biological markers are actually in plain sight in the form of co-morbidities, but are being ignored because autistic disorder is seen as the primary diagnosis and is framed only in terms of behaviours.

Next, I want to look at some of the possible causes of autism and its co-morbidities.

autism: syndrome under suspicion

This is the last post about syndromes. Promise. The reason I keep referring to syndromes is this: In recent years ‘autism’ as a label for a set of symptoms has become confused with ‘autism’ as a label for an underlying medical condition assumed to cause those symptoms. This confusion, in my view, has placed a major obstacle in the path of autism research. In this post, I round up what we know about syndromes, imagine being a doctor in the heyday of syndromes a century ago, and then expand on the two different uses of the word ‘autism’.

What we know about syndromes

Emil Kraepelin classified mental disorders in terms of syndromes. This approach makes perfect sense because:

1. A syndrome consists of signs and symptoms that tend to co-occur. (Technically, a sign is something someone else can detect, and a symptom is what the patient reports, but for the sake of brevity, I’ll refer to both as symptoms).

2. Symptoms that tend to co-occur are more likely to be causally linked than symptoms that don’t, so identifying a syndrome can provide an important clue as to what’s causing the symptoms.


3. The fact that symptoms tend to co-occur doesn’t guarantee that they are causally linked. Correlation doesn’t indicate causality.

4. The symptoms could all share the same cause; some symptoms could be causally linked, but not others; the causal links could be complex; or the symptoms could co-occur by chance.

Dr Smith’s syndrome

To illustrate the advantages and disadvantages of classifying symptoms in terms of syndromes, I’m going to imagine a doctor – Dr Smith – who has a successful medical practice in a large country town and tries to keep up to date with advances in medicine. It’s 1912. During the last month, Dr Smith has seen over a hundred patients with the same symptoms; breathing difficulties, a persistent cough and chest pain. His experience tells him that the sudden onset of such symptoms in a large number of patients living in the same geographical area indicates that the cause is likely to be a contagious disease. The course of the illness in the first few cases allows Dr Smith to reassure his patients that if they rest, keep warm and take plenty of nourishing broth they will probably be as right as rain within a week or two.

Like any diligent doctor, Dr Smith keeps notes on his patients’ progress. After the first half-dozen cases, he stops listing all the symptoms and writes ‘respiratory disorder’ as shorthand for breathing difficulties, cough and chest pain. If patients have other symptoms besides the three main ones, he makes a note of that. If patients don’t have all the core symptoms, he notes that too. The core symptoms Dr Smith has observed are so common that he wouldn’t dream of claiming that he’d found a new syndrome, but his ‘respiratory disorder’ is a syndrome nonetheless, and he would refer to it as such. It consists of core symptoms that tend to co-occur, but the severity of the symptoms varies between patients and some patients have additional symptoms.

Dr Smith is aware that although most patients with the syndrome ‘respiratory disorder’ will recover quickly because their symptoms are caused by a mild infection, the core symptoms of ‘respiratory disorder’ can also be caused by more serious illnesses such as pneumonia, tuberculosis or an allergic reaction to dust, so he keeps a careful eye on how his patients’ symptoms change. His patients have every confidence in him – but what does Dr Smith’s syndrome have to do with autism?

Autism as a descriptive term

The syndrome labelled ‘autistic disorder’ in DSM-IV has three core symptoms – impairments in social interaction and communication, together with restricted, repetitive behaviours. As far as the process of diagnosis is concerned, ‘autistic disorder’ is an exact parallel to Dr Smith’s ‘respiratory disorder’ – except of course that the causes, core symptoms and outcomes are different. Dr Smith used the term ‘respiratory disorder’ to describe the core symptoms of a syndrome, but he knew that there could be several causes for those symptoms. For many years, ‘autism’ and ‘autistic’ were also used descriptively. Bleuler, Kanner, Asperger, Eisenberg, Wing, Gould and a host of other medical practitioners between 1911 and 1979 used ‘autism’ to describe symptoms. DSM-IV also uses ‘autistic’ descriptively; autistic disorder is a disorder characterised by autistic features, ‘autistic’ being an adjective that describes behaviour, just as ‘excitable’ or ‘lethargic’ do.

Autism as a medical condition

Since 1979, a different use of the word ‘autism’ has crept into general use, and even into specialized use. It’s now used to refer to an underlying medical condition that is assumed to cause autistic behaviour. Why does that matter? It matters because what has also crept in is the assumption that if people meet the diagnostic criteria for autistic disorder, that means they have the underlying medical condition that causes autistic behaviiour – that everybody’s autistic characteristics must have the same cause.

That assumption is of crucial importance, because even if you find a genuine syndrome – a group of symptoms that do reliably co-occur – it doesn’t follow that all those symptoms are causally related, something that our fictitious Dr Smith was well aware of. Kanner was aware of it too; he refers to his syndrome as ‘unique’. This wasn’t any old group of symptoms that happened to crop up in eleven children but overlapped here and there in other children; his syndrome consisted of twenty (count ’em) essential common characteristics – how likely was that to happen by chance? Furthermore, Kanner also thought he’d found the cause of all those symptoms; what tied them together was a disturbance of affective contact. That explained everything, including the children’s abnormalities in feeding, speech and motor movement. But Kanner’s causal link between symptoms – a disruption of the social instinct – was based on an assumption made by psychodynamic theory. Not only does later research suggest that there’s little evidence for the existence of a social instinct, it also shows that Kanner was wrong about the symptoms he observed making up a syndrome.

As the number of children diagnosed with his syndrome increased, Kanner found he had to omit some of the symptoms because many children didn’t show them. He also had to group the remaining symptoms under five headings, because not all the children showed all the ones that were left. What Wing and Gould found in their epidemiological study was that although some children did meet the criteria for Kanner’s syndrome, there were no clear-cut differences between them and the rest of the socially impaired group. Wing and Gould pointed out that although social impairment was reliably associated with two other symptoms – i.e. it was part of a syndrome – that syndrome wasn’t clearly differentiable from social impairments in general. Indeed its symptoms formed a spectrum because they varied considerably in severity and blurred into each other.

Despite what Wing and Gould’s conclusions, a common current perception of autism is not that it’s a descriptive term for a rather vague group of symptoms that might have lots of different causes, but that it is a single medical condition that manifests itself in different ways in different individuals. The DSM and ICD, although technically using the word ‘autistic’ or ‘autism’ descriptively (the ICD equivalent of ‘autistic disorder’ is ‘childhood autism’), reinforce that idea in respect of autism and many other so-called mental disorders. That’s because they are both arranged in the form of labels for syndromes followed by a list of their symptoms, giving the impression that we know that these disorders exist, and we know what their symptoms are. All that remains is to find those elusive and complex causes.

Diagnostic criteria: the effect on research

This way of thinking about autistic characteristics in particular and mental disorders in general, poses a major problem for researchers. Many researchers are aware that we don’t actually know that the symptoms labeled ‘autistic disorder’ in the DSM (‘childhood autism’ in the ICD) are causally linked, and there might be different causes for the same symptoms in different cases, but the only way researchers can locate autistic research participants is by using the diagnostic criteria set out in the DSM or ICD. In many cases, for research purposes autistic participants are required to be diagnosed using specific standardized assessments. This process might look as if it ensures that all the autistic participants form a homogeneous group. They meet the same diagnostic criteria, certainly, but if those diagnostic criteria by definition put people with similar symptoms but different causes for those symptoms into the same basket, the researchers are in effect trying to find differences between a group of apples-and-oranges and a group of pears. Or more accurately, a group of apples-and-oranges and a group of mixed fruit. There will be some differences between the groups, but they are unlikely to shed any light on what causes apples to develop as apples, or what causes oranges to develop as oranges.

One by one, theories about the causes of autistic characteristics have been sidelined because none of them has explained the characteristics of everyone who meets the diagnostic criteria for autism. Studies investigating ‘refrigerator mothers’, sensory processing, theory of mind, executive function, central cognitive coherence, the extreme male brain, vaccines and dietary intolerances have all produced interesting and useful findings, but have all fallen by the wayside because those findings have also tended to be inconclusive and contradictory – not for the specific individuals involved but for people with autism in general. If everyone diagnosed with autistic disorder showed identical symptoms and we knew that all those people had the same cause for their symptoms, inconclusive and contradictory findings would indeed be frustrating and puzzling. But since autistic characteristics vary widely and we don’t know that all autistic people have the same causes for their symptoms, the most likely explanation for inconclusive and contradictory research findings is that there are several different causes for autistic characteristics (different causes in different people), but the diagnostic criteria lump them together regardless.

In the next post, I want to explore why that happens.

Illustration: this work by Phillip Martin is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License.

turning Kanner’s model of autism upside-down

Kanner’s model of autism should be turned on its head.
The idea that Kanner’s syndrome was caused by a fundamental impairment in social interaction has prevented us discovering the causes of autism.
Those are quite bold claims. In the next couple of posts I’ll explain why I made them.

Previously, I suggested that Kanner was using two theoretical frameworks to analyse the behaviour of the 11 children with his unique syndrome; Kraepelin’s classification of mental disorders and psychodynamic theory. Kanner’s comments imply that he interpreted abnormalities in feeding, speech and movement in social and sexual terms because of the psychodynamic framework. I want to look more closely at psychodynamic theory and explain my claim that viewing Kanner’s syndrome as caused by a fundamental impairment in social interaction has proved an obstacle to research into the causes of autism.

How the brain works: the psychodynamic model

As I pointed out earlier, three of the founders of psychodynamic theory, Freud, Jung and Adler, had neurological experience and would have known a fair amount about brain anatomy and about the specific functions of different parts of the brain. What they also would have known about was the patterns of electrical activity that had been observed in the brain. What was little understood at the time was how the brain worked, so the psychodynamic theorists had to make an educated guess based on their observations of people’s behaviour.

What Freud came up with was the idea that human behaviour is driven by life instincts such as the will to survive, to eat, to seek pleasure and sexual gratification. He eventually grouped these instincts together into a primary life instinct or ‘drive’ (Eros) that created flows of energy through the brain. The life drive was the origin of all behaviours that increased the chance of survival, such as an awareness of one’s surroundings and the ability to interact and communicate with others. (Freud later added a death drive (Thanatos) to explain aggression and destructive behaviour). In the light of what we now know about how the brain works, it’s clear that some aspects of Freud’s model were very insightful but others weren’t.

How the brain works: the information-processing model

Fifteen years after Kanner first described his syndrome, David Hubel and Torsten Wiesel began publishing their work on the visual cortex of the cat. Hubel and Wiesel’s research was significant, not just because it told us a lot about how cat vision develops, but because it provided some important insights into how brains in general process information – the mechanism that the psychodynamic theorists were trying to figure out.

Hubel and Wiesel’s research, and the work that followed it, revealed some key principles about how the human brain processes information.

1. All information about the outside world enters the brain via the senses.

2. Sensory information is processed in a hierarchical way, from simple to complex – broadly speaking, from the back to the front of the brain.

3. Different parts of the sense organs respond to different aspects of sensory information and that information is then integrated increasingly as it’s transmitted through the brain.

Say, for example, that I am looking at a table. Different cells in my retina respond to different properties of the pattern of light reflected off the table and entering my eye; some cells respond to the boundaries between light and dark areas, others to light of different wavelengths (colour). This very basic information about the pattern of light from the table is then integrated as it passes through my brain; first it’s chunked up to form a pattern that represents the table I’m looking at. This representation is then linked with stored representations of other images, such as other tables and chairs, and is then integrated with information from other sensory modalities such as what the word ‘table’ sounds like, what the written word ‘table’ looks like, what tables feel like and so on. This information ends up in the frontal area of the brain, which has been described as having an ‘executive function’ – it integrates information from all over the brain and makes decisions on the basis of that information. Except that information doesn’t actually ‘end up’ anywhere, because the brain has a complex series of feedback loops that send information from higher-level areas back to lower-level ones.

What does the way sensory information is processed have to do with Kanner’s syndrome?

Kanner concluded that the ‘…fundamental disorder is the children’s inability to relate themselves in the ordinary way to people and situations’ because he was working within the psychodynamic framework. He saw the children’s abnormal behaviours essentially as caused by a disruption in ‘the usual biologically provided affective contact’ – an instinct that emerged from the life drive. Kanner doesn’t use the terms ‘instinct’ or ‘life drive’ – he doesn’t need to because all his readers would be familiar with that framework – but he makes it clear that he sees affective contact as an innate, biologically-based ability that the children didn’t have.

I can understand where the psychodynamic idea of instincts came from. Most children develop skills like eating, walking, responding to the world around them, interacting with and communicating with other people, without any apparent effort on their part or any significant intervention from adults. Indeed, psychodynamic theorists felt that adult intervention often disrupted normal development. It made sense to assume that natural selection had ensured the ability to relate to people and situations developed instinctively – as automatically as growth or sexual development, or the way the heart, lungs and digestive system function.

I think Kanner’s conceptual model of his syndrome could be represented like this:

Kanner's conceptual model

But from what we now know about brain function, as far as the brain is concerned the ability to relate in the ordinary way to people and situations is exactly the opposite of a basic instinctive drive. There are areas of the brain that specialise in relating to people and situations; they are in the frontal lobes where information from many other brain areas is integrated. Relating to people and situations requires monitoring a constantly changing flow of complex information from a wide range of sources and constant feedback to other parts of the brain. The frontal lobes and their functions develop slowly and mature late – often not until early adulthood. Even after maturity, because of the plasticity of the brain, the frontal lobes continue to change in response to the environment.

Highly over-simplified schematic showing flow of sensory information to frontal lobes

We’re not usually aware of all this complex integration, monitoring and feedback of information; what we usually experience in ourselves, and observe in others, is that responses to the environment and to one another happen instinctively and automatically – that is, until something goes wrong. When we find we have too much or too little information, or that information is ambiguous, or we feel tired, hungry or anxious, then behaviours that most of the time feel and look instinctive and automatic, feel and look a bit less instinctive and automatic.

An impairment in processing any of the streams of information about people and situations would, to some extent, disrupt normal responses to people and situations. In the light of what we now know about the way the brain works, I propose that Kanner’s causal model should have looked more like this;

alternative model for Kanner's syndrome

What we now know about how the brain works suggests that Kanner’s conceptual model of his syndrome should actually be reversed; that relating to people and situations is the outcome of some very complex information-processing requiring input from many parts of the brain, rather than a basic, automatic instinct that drives other behaviours. The implication is that rather than a disturbance in affective contact causing problems with feeding, speech, body movements, social interaction and communication, what’s more likely is that problems with motor function, sensory processing, speech and language resulted in the children’s problems with relating to people and situations in the ordinary way; that Kanner’s model should be turned upside-down.

More on this later, but next I want to look again at Kraepelin’s classification system and find out what happened to Kanner’s syndrome after 1943.

a critical look at Kanner’s autism

In this post I want to take a close look at Leo Kanner’s ground-breaking paper ‘Autistic Disturbances of Affective Contact’ and explain why I consider Kanner’s analysis of the children featured in the paper to be critically flawed. I can understand why Kanner came to his conclusions; he was working in the light of knowledge that was available at the time. What concerns me is that his reasoning has been perpetuated in the diagnostic criteria for autism and in autism research. I suggest it’s because of this that 70 years later we are still scratching our heads about what causes autism.

In 1938, a five year-old boy named Don (later identified as Donald Triplett) was referred to Kanner’s clinic. Don’s was the first and most detailed of 11 case studies (eight boys and three girls) featured in Kanner’s paper published in the journal Nervous Child in 1943. Kanner claimed that he had identified a “unique ‘syndrome’, not heretofore reported” (p.242) consisting of “inborn autistic disturbances of affective contact” (p.250). He reviewed the children’s previous diagnoses and then outlined a list of “essential common characteristics” (p.242) that supported his hypothesis.

Previous diagnoses

Previous diagnoses included idiocy, imbecility, deafness or being hard of hearing, and schizophrenia. (At the time, ‘idiot’ and imbecile’ were technical terms; an imbecile was someone with an IQ score between 20 and 49, and an idiot had a score below 20. Anyone with an IQ between 50 and 69 was described as a ‘moron’). Kanner ruled out idiocy and imbecility because all the children had what he described as ‘good cognitive potentialities’ (p.247). He appears to have ruled out hearing difficulties on the grounds that another doctor had observed that one of the children (Virginia) “does not seem to be deaf from gross tests” (p.230). Kanner rejected the diagnosis of schizophrenia because the children had shown their unusual characteristics from birth.

The term ‘autism’ wasn’t invented by Kanner; it had been coined thirty years earlier by Eugen Bleuler, born near Zürich, and a contemporary of Kraepelin and Freud. Bleuler used ‘autism’ to describe the self-absorbed, withdrawn characteristics seen in schizophrenia, another label he came up with to replace Kraepelin’s ‘dementia praecox’. Bleuler’s book Dementia Praecox or the Group of Schizophrenias was published in 1911, ten years before Kanner graduated in Berlin and twenty years before Hans Asperger did so in Vienna, so both of them would have been familiar with the word they used to describe the children in their case studies.

Kanner’s syndrome

Kanner based his conclusion that he had identified a previously undiscovered syndrome on 20 “essential common characteristics” shared by the 11 children. Statistically, it’s highly unlikely that 11 children could show 20 common characteristics without having the same syndrome – until you take a closer look at the characteristics. Here’s the list (Kanner gives a short explanation for each of them);

•inability to relate themselves in the ordinary way to people and to situations from the beginning of life
•extreme autistic aloneness
•failure to assume at any time an anticipatory posture preparatory to being picked up
•ability to speak
•excellent rote memory
•personal pronouns are repeated just as heard
•loud noises and moving objects
•monotonously repetitious
•anxiously obsessive desire for sameness
•limitation in the variety of spontaneous activity
•good relation to objects
•masturbatory orgiastic gratification
•relation to people different to that of objects
•good cognitive potentialities
•anxious tenseness
•physically normal
•from highly intelligent families.

What’s interesting about the ‘common’ characteristics is that they are not common to all the children, nor do they all reflect what Kanner describes in his case studies. The second point applies to seven items on the list, which I’ll address in turn.

Ability to speak, excellent rote memory, literalness and personal pronouns are repeated just as heard. Seven of the children had previously been considered deaf or hard of hearing because of their abnormal receptive and/or expressive speech. Three children (Richard, Herbert and Virginia) presented as mute. Richard had once been heard to whisper “good night” and pupils at Virginia’s school claimed she had said ‘chocolate’, ‘marshmallow’, ‘mama’ and ‘baby’ (p.231). There’s no record of Herbert saying anything. Despite speech or language impairments, other children in the group had clear enunciation, sophisticated vocabularies or good sentence structure, and some had specific problems with pronouns. Yet despite this range of differences, Kanner concludes “there is no fundamental difference between the eight speaking children and the three mute children” in terms of “the communicative functions of speech” (p.243).

Food. Six of the children had feeding difficulties as infants (one vomited repeatedly and another was tube-fed). But Kanner interprets these characteristics not as feeding difficulties per se, but in terms of “our patients…anxious to keep the outside world away, indicated this by the refusal of food” and contrasts their behaviour with those of “affect-hungry” children demanding excessive quantities of food when placed in foster-care (p.244).

Failure to assume at any time an anticipatory posture preparatory to being picked up. Kanner notes this in only two specific cases, but says that ‘almost [my emphasis] all the mothers …recalled their astonishment’ at the children’s failure to respond (p.242).

Masturbatory orgiastic gratification. Kanner’s evidence for this is the children’s preoccupation with spinning objects, jumping up and down with glee and rolling and rhythmic movements, despite his records indicating masturbation in only two cases.

Two questions

So, two questions: First, why did Kanner list as “essential common characteristics” characteristics that weren’t common to all the children? Second, why did he interpret some very somatic (bodily) characteristics (feeding, speech, jumping and rolling) in social and sexual terms rather than in terms of feeding, speech and motor movements? The answers, I suggest, lie in the theoretical frameworks that Kanner was using at the time – Kraepelin’s classification of mental disorders, and psychodynamic theory. (Kraepelin’s classification system and the psychodynamic model had both been in use for half a century, so were well-established. Adolf Meyer, who appointed Kanner to his post at Johns Hopkins, had been a keen advocate of both frameworks.)

Why weren’t Kanner’s ‘essential common characteristics’ common to all 11 children?

I mentioned earlier that Kraeplin’s classification system was based on syndromes – patterns of symptoms that tended to co-occur. Two observations about syndromes: First, syndromes are about correlation, not causality. It’s true that symptoms that tend to co-occur are more likely to be causally linked than symptoms that don’t, but all you can deduce from co-occurring symptoms is that the symptoms sometimes co-occur. You can only make deductions about how the symptoms might be causally linked once you know something about what causes them.

Second, although syndromes have consistent core symptoms (if they didn’t they wouldn’t be syndromes), the symptoms shown by individuals tend to vary. So everyone diagnosed with Syndrome Alpha, say, would show the core symptoms A, B and C, but together those people might also show symptoms D to T (making a total of 20 symptoms associated with Syndrome Alpha). Kanner couldn’t tell from a sample of 11 children which symptoms were core ones and which might have appeared by chance. He would no doubt have been interested to hear from clinicians who’d seen cases of any of the symptoms, so he wouldn’t have wanted to miss anything out. Hence he cast his net quite wide. In short, his “essential common characteristics” were characteristics of his proposed syndrome rather than of the individual children.

Why did Kanner interpret somatic abnormalities (feeding, speech and body movements) in social and sexual terms?

I noted in the previous post that psychodynamic theory recognised the link between behaviour and the brain. (Two of the founders of the psychodynamic school, Freud and Jung, had worked with renowned neurologists and Adler had begun his career as an ophthalmologist, so they would have been well aware of the brain-behaviour connection.) Because the brain was involved and the brain is an organ of the body, behavioural drives, like other bodily characteristics, could be inherited. Characteristics were inherited via sexual reproduction, so any characteristics that reduced the chances of sexual reproduction taking place, such as impairments in affect or communication, or atypical sexual behaviour, would be less likely to be passed on to the next generation, would not be ‘normal’ for the species and so could be considered pathological. Psychodynamic theory saw social and sexual behaviour as fundamentally important in normal child development, and impairments in social and sexual drives as being capable of causing problems with feeding, speech and motor control (Kanner cites David Levy and Hilde Bruch, both of whom used a psychodynamic approach, to support his argument). As Kanner says “the outstanding, “pathognomonic”, fundamental disorder is the children’s inability to related themselves in the ordinary way to people and situations…” (p. 242).

Why I see Kanner’s analysis as flawed

I see Kanner’s analysis as flawed because of three assumptions he made. First, that because the individual children’s characteristics overlapped, despite not all the children having all the symptoms, all the children had the same syndrome. Second, that the existence of a syndrome meant that the characteristics of the syndrome must be causally related. Third, the assumption that social and sexual drives are so fundamental that they must be the cause of problems with feeding, speech and motor movements. Although children certainly can refuse to eat or speak because of an underlying issue with affect, knowing what we now know about brain development, it’s highly unlikely that a babe-in-arms would not suckle or that an older child wouldn’t speak for years on end for that reason. Since motor function is implicated in all three behaviours it should at least be considered as a possible cause.

At the time, Kanner’s assumptions weren’t unreasonable, and at the end of his paper, he makes it clear that he’s making assumptions and that there’s uncertainty about how affect could result in the syndrome he proposes. At the time a good deal was known about brain structure and what brain areas controlled what types of behaviour, but brain function was still something of a mystery. In 1943 Karl Lashley was still attempting to find the location of memory, it would be a decade before Crick and Watson unveiled their model of DNA, and two decades before Hubel and Wiesel published their work on cat visual cortex.

In the light of current knowledge about brain function, Kanner’s conclusion that all the children’s unusual behaviors were attributable to an “inability to relate themselves in the ordinary way to people and situations” doesn’t hold water. Despite this, his assumptions have persisted in the diagnostic criteria for autism, and thus in autism research. I suggest the assumption that autism is caused by a fundamental impairment in social interaction has been the main reason why, half a century later, we are still trying to find the causes of autism. In my next post I propose that turning Kanner’s model on its head could break this log-jam.


Kanner, L. (1943). Autistic disturbances of affective contact. Nervous Child, 2, 217-250.
(pdf available on line)