The Image of the Mother’s Eye: Autism and Early Narcissistic Injury

For information about the author

Submitted for publication, Journal of Analytical Psychology, Society for Analytical Psychology, 2002.

Archived in cogprints: 2002


Autism results from a pervasive cascade of developmental failure. Prevailing theory is that the primary deficit in autism (primary cause of the cascade) is a biological or genetic abnormality in the brain. This paper challenges that theory. Though a primary cause should be both necessary and sufficient, none of the biological or genetic factors which are statistically linked to autism is either necessary or sufficient to cause autism. Another objection derives from the logic of cascades. A cascade’s primary cause is the same in kind as the events which comprise the cascade. Autism is a cascade of psychological failures. The associated biological and genetic factors must, I argue, increase the incidence of a primary psychological deficit. I hypothesize: (1) The acquisition of the image of the mother’s eye is a critical very early step in development. Because the image symbolizes psychological containment, it is an essential element in the self-organization of the personality. (2) Failure to acquire (or retain) that image is the primary deficit in autism. I show that these hypotheses are consistent with Holland’s paradigm of self-organization in complex adaptive systems. The paper uses clinical data to illustrate the hypotheses. It then synthesizes evidence from experiments on infant vision, from Stern’s observations of infant-mother pairs, and from observations on the incidence of autism in infants with cranial nerve palsy, or congenital blindness, or severe early deprivation. To this it links evidence from studies of evolutionary changes in the primate eye and from experiments on the visual behavior of mildly autistic adults. Theories of a biological cause have been undermined by the recent dramatic increase (273% from 1987 to 1998) in the incidence of autism in California. This increase was recently (October 2002) confirmed by Robert Byrd and co-workers; previously it had been widely dismissed as an artifact of measurement. The increase cannot be explained by genetic factors. It is unlikely that the physical, chemical or biological environment in California deteriorated sufficiently in one decade to account for such a large increase. It is more likely that a change in social behavior (affecting a psychological factor) could be sufficiently rapid. Because of the significance to public health of the rapid increase in autism, there is likely to be a vigorous new effort to identify its cause(s). The problem should be debated across disciplinary lines. This paper is a timely contribution to that debate. If the primary deficit is psychological then, to some extent, research should be redirected towards exploring autism at the psychological level. Scientists may need to evaluate suggestions derived, for example, from psychoanalytic data. My paper attempts this difficult collaboration. It shows that the image-of-the-eye hypothesis leads to a testable prediction: that autism is statistically linked to the early use of daycare.


In this paper I propose two related hypotheses. The first hypothesis is that the acquisition of a visual image of the mother’s eye is a critical very early step in development. Once acquired the image is associated with, and begins to symbolize, containing. (If the infant is frightened, for example, and the mother provides appropriate soothing, then the mother encompasses or contains the infant’s feeling. Here ‘symbolize’ means that when the infant seeks containing the infant seeks an external agent which resembles the image; when the infant sees such an agent, the infant feels contained.) In adult life the image continues to symbolize containing and will appear, for example, in dreams. If an adult has had early narcissistic injury then containing was disturbed in infancy. The adult may then have disturbed images of the eye.(‘Early narcissistic injury’ is an early injury to self-esteem. In adult life symptoms may include an exaggerated desire for attention, self-reference, controlling behavior, difficulty in recognizing the reality of another person, difficulty in accepting limits, grandiose fantasies, and chronic anger with outbursts of rage when wishes are disappointed.)

The personality cannot develop without containing. If the infant fails to acquire the image of the eye in the first weeks of life, the likely result, I argue, would be a pervasive cascade of developmental failure. This leads to the second hypothesis: failure to acquire (or to retain) the image of the mother’s eye is the primary deficit or primary cause of autism.

Autism varies greatly in degree. A corollary of the second hypothesis is that mild autistic traits may be caused by a disturbance in the acquisition or functioning of the image of the eye. Two predictions follow: (1) adults with mild autistic traits may (like adults with early narcissisitc injury) produce disturbed images of the eye; (2) adults with mild autistic traits may also have symptoms of narcissistic injury. I describe two cases in which these predictions were confirmed.

I begin by relating the image-of-the-eye hypotheses to the paradigm of self-organization. Then I critique existing theories on the cause of autism. Then I illustrate the (first) image-of-the-eye hypothesis by means of three images from the case of …… who had symptoms of early narcissistic injury. Next I adduce support for the image-of-the-eye hypotheses from a variety of sources. These include Kohut’s psychoanalytic concept of mirroring, Stern’s concept of the intersubjective sense of self, the evolution of intersubjectivity in non-human primates, experiments on infant vision, and experiments on the neuronal wiring of vision in other vertebrates. I then illustrate the image-of-the-eye hypotheses with images from the cases of ………… In addition to symptoms of early narcissistic injury, …………. had mild autistic traits.

Finally I show that my hypotheses lead to a statistical prediction about the recent increase in the incidence of autism, a prediction which can readily be tested.


The new paradigm of self-organization in complex adaptive systems has been elucidated by Holland (1998, pp. 229-231) and others. In complex adaptive systems a limited number of simple elements spontaneously assemble themselves into a hierarchy of increasing levels of complexity. It is now recognized that all living systems self-organize in this way. It follows that the personality, because it is a living system, must also organize itself from a limited number of simple psychological elements (McDowell 2001, p. 653: Endnote). Included amongst these elements would be any sensory image which is reliably acquired very early in development. An interesting implication of this paper is that we may identify some of the initial elements from which the personality self-organizes.

At birth the infant emerges from a container, the womb, which had previously met all needs. The infant’s personality must then organize itself to ensure continued physical and psychological containment. Stern (1977, p. 37) has shown that at about six weeks, for example, the infant learns to make eye contact with the mother. Eye contact stimulates the mother to play more with the infant. She plays with facial expression, with voice, with face presentations, with head movements, and with proximity games. Thus she contains her infant’s impulse to play.

To ensure eye contact the infant must learn to recognize the mother’s eye. This means that the infant must, in some sense, acquire an internal image of the eye. (In order to recognize something I must have a stored internal image to compare it with.) Several lines of biological evidence show that an image in the cerebral cortex cannot be genetically pre-wired: the total number of genes is too small, human evolution is too fast relative to genetic change, and most wiring in the cortex is determined by sensory input (McDowell 2001, pp. 640-2: Genes and self- organization; Genes and behavior). If the image of the mother’s eye is not pre-wired in the cortex, then it must be acquired through the senses from the environment.

The infant’s situation may, in a broad sense, be analogous to that of a newly-hatched gosling. A gosling must imprint an image of its mother soon after hatching. During this critical period it will imprint the image of any agent to which it is exposed (the agent must satisfy some general parameters concerning height and clarity of edge). From then on only that agent will elicit following behavior. If the gosling imprints the wrong image then its following behavior will be bizarrely maladaptive (Lorenz 1970).

For the gosling the image of the mother releases an instinctual behavior. Because the human infant’s behavior is more malleable, the image of the mother’s eye must function in a more complex manner. It functions, I hypothesize, as the symbol of containment.

The autistic syndrome

Between 0.1 and 0.2 percent of children are autistic (Gillberg & Wing 1999). The syndrome varies greatly in severity. A child with Asperger’s syndrome(mild or high-functioning autism) has normal intelligence and language but has a social deficit and a narrow range of interests and activities (Rapin 1997). Adults with Asperger’s syndrome were tested by Baron-Cohen and co-workers (1997; 2001) for their ability to identify “complex” emotional states from photographs of whole faces and from cropped photographs. They were significantly less accurate than normal adults when tested with images of whole faces and even less accurate with images of the eyes alone. Thus Asperger’s syndrome entails a deficit in “reading the mind in the eyes.”

A child with classic autism has more severe symptoms including, in social behavior:

marked impairment … of nonverbal behavior such as eye-to-eye gaze, facial expression, body postures, and gestures to regulate social interactions; failure to develop peer relationships; … lack of showing, bringing, or pointing out objects of interest; lack of social or emotional reciprocity; failure to distinguish among persons; language that is not used for … interpersonal communication, but is characterized by echolalia … and references to the self in third person (Rapin 1997).

There is also a deficit in symbolic thinking: play with toys is replaced by stereotyped manipulation of objects, for example, by unscrewing and screwing jar lids (Hobson 1999, p. 45).

Autism is clearly a pervasive developmental disorder: a primary deficit in the fetus or infant begins (to a greater or lesser degree) a cascade of developmental failure. Deficits in social behavior, language and play are secondary (Ungerer 1989). After several decades of research, however, the primary deficit has not been identified (Sigman & Capps 1997; Wimpory et al. 2000).

The incidence of autism is increasing

In California, the number of children diagnosed with profound autism has increased drastically. The California Department of Developmental Services reported that for the years 1987, 1998, and 2002 (as of July), the numbers are, respectively, 2,778, 10,360, and 18,460. Byrd et. al. (2002) have shown that the increase from 1987 to 1998 cannot be accounted for by changes in diagnostic criteria, by mis-classification, or by the immigration of autistic children.

Genetic factors cannot account for this sudden increase. The increase must, therefore, be caused by an as yet unknown change in one or more environmental factor(s).

A biological deficit?

At present there is a bias towards biological explanations for mental disease. In the literature on autism the overwhelming weight of opinion is that the primary cause is biological. Rapin (1997), however, who argued for a biological cause, noted that:

The existence of a large literature describing a wide variety of neural abnormalities in autism … has led to much speculation … however, no coherent anatomical or pathophysiologic theory, nor a biologic diagnostic test has yet been developed.

Rodier (2000), who also argued for a biological cause, said that:

In utero exposure to rubella (German measles) or to birth-defect- causing substances such as alcohol … increases the chances that autism will develop. People with certain genetic diseases, such as phenylketonuria and tuberous sclerosis, also have a greater chance of developing autism. None of these factors, however, is present frequently enough to be responsible for many cases … Researchers do not know how the multiple factors combine to make some people display symptoms, while allowing others to escape them.

Rodier’s group have identified an allele (a mutant form of a gene) which plays a role in autism.

The allele’s … presence does not guarantee that autism will arise. The variant allele occurs in about 20 percent of the people who do not have autism, and in about 40 percent of those who do. The allele doubles the risk of developing the condition. But in about 60 percent of people with autism, the allele is not present (ibid.)

In identical twins, one twin may be autistic and the other not. Thus many biological factors are associated with autism but none of them is always present and none of them always cause it. A true primary cause would be both necessary (without it, no autism)and sufficient (when present, so is autism). None of the biological factors is either necessary or sufficient.


The logic of causation in a cascade is relevant here. The event which first sets a cascade in motion, the event which is the primary cause, is the same in kind as the events which comprise the cascade. A thaw increases the risk of an avalanche, but the primary cause is the first movement of snow.

The following example shows why this distinction is important. Death results from a cascade of bodily failure. Households with handguns show an increased frequency of death. But households with a high-cholesterol diet also show an increased frequency of death. A gunshot wound is neither necessary nor sufficient to cause death. A high-cholesterol diet is likewise neither necessary nor sufficient. The cessation of the heart beat, however, is both necessary and sufficient. It is also the same in kind as the other bodily failures which lead to death. It is the primary deficit which begins the cascade. When we identify the primary deficit we can see how other factors (gunshot and cholesterol) increase the frequency of that deficit.

None of the suspected biological factors is the same in kind as the cascade of failures which is autism. Therefore none of the biological factors can be the primary deficit. The primary deficit should be a very early failure in psychological development.

Tracking the care-giver’s face

The following is further evidence that no biological factor is the primary deficit. The correlation between autism and congenital infant blindness (Brown et al. 1997; Hobson et al. 1999) is stronger that the correlation between autism and other biological factors. If the mother of a blind baby relates to it actively by means of touch and sound then her baby’s development may be close to normal; if the mother does not do so then her baby is very likely to develop autism (Als et al. 1980). These autistic children do not have higher-than-normal rates of any other biological factor. What they have in common is blindness and a consequent failure to relate to the mother in infancy.

There is also a high incidence of autism in children who were institutionalized at birth (Fraiberg 1977, pp. 185-187). There was a high incidence, for example, in children who were adopted in the U.K. from orphanages in Romania (Rutter et. al. 2001). These children are normally sighted but were neglected in cribs during infancy. The Romanian children do not have higher-than-normal rates of any biological factor.

Infants who have cranial nerve palsy with consequent paralysis of eye muscles are also liable to develop autism (Stromland 2000).

These three different groups, congenitally blind children, children institutionalized at birth, and children with cranial nerve palsy, share one common deficit: they are all unable to track a care-giver’s face. In the absence of any biological cause, therefore, early deprivation in visual stimulus by a care-giver’s face often (but not always) leads to autism.

This evidence tends to be ignored or discredited. Rutter et al. (2001, p. 101), for example, argued that such deprived children have “quasi-autistic features” which do not represent “ordinary” autism. But autism is diagnosed by the presence of such “quasi-autistic features.” There is no reason to doubt that these children are suffering from a form of autism. We cannot discredit this evidence simply because it cannot be accommodated into the prevailing theory. The evidence may disprove the theory.

Like the proposed biological causes, the lack of early visual stimulus by a care-giver’s face is neither necessary nor sufficient to cause autism. It is not the primary deficit but it somehow increases the incidence of the primary deficit.

Inherent intersubjectivity

Trevarthan & Aitken (2001, pp. 3-4) argued that an infant is born with genetically determined brain structures which provide for “inherent intersubjectivity”, that is, for a complex pattern of interrelatedness with the mother:

Regulation of this primary communication depends on an innate ‘virtual other’ process in the infant’s mind.

Trevarthan & Aitken argue, therefore, for a biological cause: autism is caused by embryonic deficits in the genetically determined brain structures (ibid., p. 30).

Trevarthan & Aitken’s concept of “inherent intersubjectivity” assumes that the genetic apparatus specifies complex behavior. But this is not so. Living structure is emergent. Genes specify only the structure of enzymes and the timing of their appearance. Such timing, together with information from the environment, provides constraints which guide self- organization (McDowell 2001).

Just as self-organization creates anatomical structure out of simpler anatomical components, so must self-organization create behavioral structure out of simpler behavioral components. “Inherent intersubjectivity”, therefore, must arise from the self-organization of simpler components of behavior. It follows that autism must also be explained in such terms: when a simple component of behavior is blocked, the subsequent organization of intersubjective behavior is derailed. This argument is related to my earlier argument about cascades.

A very early step in development

An autistic child seems not to know that his or her mother has a subjective self. An autistic boy, for example, was playing with his mother when she hurt her finger with a toy hammer and gave a sudden exclamation of pain. Though he was emotionally attached to his mother the boy paid no attention to her exclamation. Evidence of this kind led Hobson et al.(1999, p. 55) to suggest that the “final common pathway” (primary deficit) is a failure to develop atheory of mind.

A theory of mind (or the lack thereof) may be inferred from a child’s visual behavior. An autistic child makes little eye contact and pays little attention to his or her mother’s face. The child ignores the mother’s facial expressions. Consequently the child lacks social referencing: it cannot evaluate an ambiguous situation by checking the mother’s expression. The child also lacks shared attention: it cannot follow the mother’s eyes see what she is seeing, nor point to share feelings about something with the mother. Such deficits have recently been demonstrated at less than six months. These studies used home movies of infants who were later diagnosed as autistic (Maestro et al. 2001). Since eye contact begins at about six weeks (Stern 1977, p.37), a deficit in eye contact might also begin at six weeks.

Klin et. al. (2002) have shown recently that an autistic person looks preferentially at the mouth rather than the eyes in another’s face. A normal person does the opposite. Klin et. al. suggest that this preference may represent a core social deficit in autism.

From evidence of this kind, Trepagnier (1998) proposed the following:

autism may be the resultant of very-early-onset anxiety/affective disorder consequent interruption of establishment of face-processing during a biologically important period. In this account, brain differences in autistic individuals reflect (1) underlying pathology shared with ‘typical’ neuropsychiatric disorders, and (2) subsequent developmental experiences.

Thus Trapagnier argued that brain differences are secondary. She suggested that a failure in face-processing (due to anxiety/affective disorder) is the key.

The image-of-the-eye hypothesis

Both Hobson and Trepagnier have proposed that developmental deficits cause autism. My hypothesis (which I arrived at independently from clinical work with adults – see below) is that the primary deficit is the failure to acquire or retain the image of the mother’s eye. A blind infant who is actively mothered acquires an equivalent image through the other senses. A blind infant who is not actively mothered may fail to acquire an equivalent image. A sighted infant who is neglected may also fail to acquire or retain the image. The other biological factors associated with autism (cranial nerve palsy, rubella, alcohol during pregnancy, phenylketonuria, genetic factors, etc.) all increase the risk that the infant will not acquire or retain the image. As I noted earlier, the aquisition or functioning of the image may be disturbed to varying degree.

….. (Clinical section omitted) …….


The (sometimes disturbed) image of the mother’s eye seemed to reappear independently in three different analyses. In each case its appearance in analysis was associated with the healing of narcissistic injury. This material suggests that the image may function throughout life. Under favorable circumstances the constructive function of the image predominates. In narcissistically-disturbed development its destructive function is more apparent.

If the image of the eye is important in humans, then it should also be important in other primates. In the next section I discuss the evolution of the image.

The evolution of social gaze in primates

We can deduce an evolutionary sequence by comparing modern primates. Humans evolved from ancestral monkeys and apes which were very like modern monkeys and apes. Within that evolutionary sequence social gaze became increasingly important.

In most primates the visible eye consists mainly of dark iris with only a very small proportion of sclera. The sclera is almost always brown or light brown rather than white. But in humans the visible eye has a large expanse of white sclera which provides a sharp contrast to the dark iris (Emery 2000, pp. 583, 585). This helps an observer to determine the precise direction of a human’s gaze.

When viewing faces, a monkey shows an extreme bias for looking at the eyes and the small region surrounding the eyes. The eyes play a central role in a primate’s facial expression and thus in emotional communication (ibid., pp. 585-586). Staring and frequency of eye-contact are key social signals. The direction of one monkey’s attention, coupled with its facial expression, conveys crucial information to a second monkey about the object of attention (joint attention). It may inform the second monkey about the social status of a third, or about food or about danger.

Apes, but not monkeys, appear to use the direction of another’s gaze as a clue to the other’s intended behavior. This is more complex than joint attention because it requires an awareness of the other as a repository of potential behavior. It seems that only humans, however, practice shared attention (ibid., pp. 588, 594). Shared attention is mutual. Two individuals are each aware of the direction of the other’s gaze (I know that she sees it and she knows that I see it). Shared attention is a component of what cognitive psychology terms theory of mind, that is, of the knowledge that self and other have independent but comparable awareness. In psychoanalytic terms the equivalent is intersubjectivity, the inter-relatedness which occurs when two people are each aware of the other’s subjectivity.

These observations are evidence that, as the primate’s eye has evolved in appearance (towards more visual contrast between iris and sclera and hence towards a better signal of gaze direction), so it has evolved in psychological function (towards enabling intersubjectivity). Thus the visual appearance of the human eye (the visual image which the other internalizes) seems to be a key component of intersubjectivity.

This is supported by experiments with human subjects (Baron-Cohen et al. 1997; Baron-Cohen et al. 2001). Subjects were asked to identify complex emotional states (scheme, admire, interest, thoughtful, arrogant) from photographs of the whole face, from cropped photographs showing only the mouth, and from cropped photographs showing only the two eyes and eyebrows. When either the whole face or the eyes alone were observed, accuracy of identification was equally high. When the mouth alone was observed, accuracy was low. Thus the other’s complex subjective state is accurately conveyed by the image of the eyes. This last conclusion is confirmed by the work of Klin et al. (2002).

I have discussed the evolution of intersubjectivity. In the next section I discuss its ontogeny.

An intersubjective sense of self

Daniel Stern (1985, p. 139) described how the infant develops a sense of self within the infant-mother interaction. He identified four successive senses of self: emergent, core, intersubjective, and verbal. At the intersubjective level, beginning around eight months, the infant senses that infant and mother each have inner mental states which they can share (I want that toy, and she knows that I want it.)

The mother contains the infant not only physically but also psychologically. Just as the infant’s body develops ‘within’ the physical container of the mother’s body and her actions, so the infant’s sense of self develops ‘within’ the psychological container of the mother’s inner life. That is, it develops ‘within’ the mother’s awareness of and inner response to the infant. This is particularly the case for the intersubjective sense of self. If the infant had sufficient language the infant might say:

I see that my mother sees me as lovable, so I see myself that way too.

In the words of Stern et. al. (1998, pp. 907-8):

… the intersubjective goal … [is] a mutual recognition of each other’s motives, desires, and implicit aims that direct actions, and the feelings that accompany this process … [This] also implies a signaling or ratifying to one another of this sharing … The work is asymmetrical, with the caregiver … doing the lion’s share.

Stern’s model of development suggests why the eyes are important. The mother senses the presence of the infant’s inner life in part through the infant’s eyes. Likewise, the infant senses the mother’s inner life in part through the mother’s eyes. I argue, therefore, that the infant’s sense of self develops in the “container” of the mother’s eye.

Smell, taste, touch and sound are also involved. Infant and mother engage each other in each of these sensory modalities. Infants recognize the smell of their own mother’s milk (Stern 1985, pp. 39-40). By four months the infant-and-mother pair have developed vocal rhythm coordination, the degree of which predicts both attachment and cognition at twelve months (Jaffe et al., 2001). If the baby is born blind then it must detect the mother’s inner life through its other senses. Infant- observation studies (Fraiberg 1977, pp. 3-9; Als et al. 1980, p. 201) compared blind and sighted infants (for both groups the mothers were sighted). With a blind infant, the mutual awareness of the infant’s and mother’s inner lives developed more slowly and less reliably. The mother had to make a much more conscious effort. These observations are evidence that, for a sighted infant, the visual image of the mother represents containment.

A symbolic image is usually determined by education, or chosen by creativity and happenstance. Not so the image which symbolizes containment: every sighted infant makes the same choice. Apparently it is too important to be left to chance. In the next section I explain the mechanism which ensures this choice.

How the infant chooses the image of the eye

The inheritance of reflexes.

Neither a cortical image, nor a pattern of behavior which is peculiar to humans, can be genetically inherited (McDowell 2001). A reflex, however, is genetically inherited. A reflex is not peculiar to humans but is intrinsic to the function of an organ. Blinking and coughing are examples. When the eyelid evolved in reptiles it necessarily had a movement reflex (Gans and Parsons 1973, p. 104). A reflex is a simple element of behavior which involves only a few muscles. It is hard- wired into the nervous system. I will show that such reflexes help to predetermine the image which a newborn chooses to represent containment.

Visual edges in the image of the eye

In a new-born infant some behaviors seem to precede learning. For example, Spitz and Wolf (1946) and Ahrens (1954) found that an infant gazed preferentially at a human face or at a schematic drawing of a face. A line drawing of two eyes, a nose, and a mouth was sufficient. Spitze and Wolf suggested that the infant inherits a schematic image of the human face and seeks a match for that image in its immediate environment.

Subsequent experiments (Friedman 1964; Haaf and Bell 1967), however, showed that the infant was not gazing preferentially at the whole image but only at some elements of it. It made no difference if the elements were scrambled in the drawing. The infant preferred to gaze at moving edges. The most attractive moving edges within the mother’s face were the sharp angles at the corner of the eyes (two edges intersecting), the light/dark contrasts between the pupil and the white sclera of the eye (an edge), and the contrast between eyebrow and skin (an edge) (Stern 1977, p.37). There are also edges around the lips and around the outer rim of the head. Further experiments (Haith 1966; Karmel et al. 1974; Salapatek 1975, p. 226) showed that, from birth to about one month, an infant seeks out any visual stimuli which includes both movement and ‘edge density’, that is, the concentration of edges in a given space. By two months the infant’s vision is beginning to mature: like an older child, the two-month-plus infant begins to recognize when facial features are scrambled and shows a preference for naturally-ordered features.

But the controversy is not yet settled. Johnson and co-workers (Johnson et. al.1991; Morton & Johnson 1991) detected two separate mechanisms, one sub-cortical, the other cortical, both of which affect the infant’s preference for faces. At birth, they argue, there is a sub-cortical mechanism which, when shown a schematic drawing of the face, produces a small preference for the natural arrangement of features over a scrambled arrangement; at one month this first mechanism is replaced by the cortical mechanism (described above) which prefers edge density but does not distinguish between naturally-ordered and scrambled features.

Johnson et. al.’s experimental design seemed rigorous but, in an even more rigorous experiment, Easterbrook et. al. (1999) obtained contradictory results: newborns showed no preference for naturally-ordered rather than scrambled features, though they were able to distinguish between the two. Further experiments will be needed to clarify the issue. Whether or not there is an inherited sub-cortical mechanism, it remains true that a cortical image of the mother could not be inherited and must therefore be acquired via the senses.

Visual edges and the cortex

That the infant has a preference for visual edges is well established. Research on the vision of adult vertebrates (fish, frogs, cats, and monkeys) helps to explain this preference (Michael 1969). All vertebrates analyze visual input primarily into a series of moving edges. The analysis is done by neurons in the retina of the eye and by neurons in areas of the visual cortex which are adjacent (closely connected) to the retina. This means that my brain does not ‘see’ a whole image. Rather it ‘sees’ a series of moving edges. A computer analyzes an image into a stream of digits, zeros and ones, which it records. The stream of digits contains all the information needed to recreate the image, but it is not itself a spacial analog of the image. In a similar fashion the retina and the adjacent visual cortex analyze the image into multiple streams of moving edges.

Subsequent computation is done upon this raw data by groups of neurons ‘deeper’ in the cortex, that is, further ‘downstream’ from the retina. There my brain analyses these moving edges and, in some sense, reconstructs an image. Thus my mental image is not achieved whole, like a photograph, but by the synthesis of moving edges.

Once my brain has reconstructed and somehow recorded an internal image, that internal image becomes a visual concept or hypothesis which my brain uses to explain future input that it receives from the retina. I sometimes become aware of this when I mis-recognize someone I see. For a moment I seem to be seeing a different person. Then, as I realize my mistake, the image transforms itself into an image of the person who is really in front of my eyes. My brain has substituted a more accurate hypothesis to explain the visual data. An optical illusion has the same effect: at first my brain is fooled, then it gets it right.

The experiments of Friedman and Haaf and Bell (above) apparently distinguished a stage through which the infant must pass before it learns to see like an adult. Since the newborn recognizes moving edges, we know that the neurons in the newborn’s retina and adjacent cortex must be pre-wired (or very quickly wired) to analyze visual input into edges. We know that the newborn also has functioning muscle reflexes which enable it to focus its eyes, to fix upon an object, and to track an object (Stern 1977, pp. 35-37; Stern 1985, p. 40). All this pre-wiring (or quick wiring) provides for the visual reflexes by which the newborn tracks the visual edges in its mother’s face. It takes about two months for the infant to learn to synthesize these moving edges into an image of a human face. This learning must occur deeper within the cortex. Neurons in deeper areas of the cortex have been identified which only fire when the infant recognizes a face (Elman et. al. 1998, p. 116). Only after it has learned to re-synthesize an image of the face can the infant discriminate between naturally-ordered and scrambled images of the face. This explains what Salapatek (above) observed experimentally, that at two months infants begin to distinguish between scrambled and ordered features.

Although the newborn can focus its eyes, it can only do so within a limited focal range. When it feeds, therefore, it cannot see the breast. But when the infant feeds the mother’s eyes are within the infant’s focal range, provided that the mother is looking at the infant. Stern (1997, pp. 35-36) observed that, during breast feeding, a mother spends about 70% of her time gazing at her infant’s face. Thus the human infant-mother pair is designed to ensure that the newborn will stare at the sharp, high-contrast, moving edges of its mother’s eyes. Hence, through learning, the infant will reconstruct a cortical image of the mother’s eyes and face and associate that image with breast feeding.

Breast feeding in other mammals

It is perhaps my conscious sense of self which most distinguishes me, as a human, from other mammals. It is remarkable, therefore, that my sense of self is initiated by the geometry of human breast feeding. A cat’s eyes are equally sensitive to moving edges. Because of the way the kitten feeds, however, it does not begin life gazing at its mother’s eyes.

The autism-spectrum quotient

Baron-Cohen and co-workers have argued that autism represents an extreme on a continuum of normal psychological styles. They designed a screening instrument which measures the autism-spectrum quotient. In their study individuals measured their own quotients. Scores were grouped and averaged for each group. Averages were as follows: females 15.4; males 17.8; humanities students 16.7; science students 18.5; gifted mathematics students 24.5; subjects diagnosed with Asperger’s syndrome 35.8. When the subjects diagnosed with Asperger’s syndrome were re-measured by an observer (one of their parents), scores averaged 2.8 points higher (Baron-Cohen 1999; Baron-Cohen et al. 2001).

….. (Clinical section omitted) ……

Future Research

The image-of-the-eye hypotheses imply that autism may be investigated via the analyses of adults with mild autistic traits. Mild autism is like early narcissistic injury in that it seems to involve a diminished capacity to internalize the experience of being seen.

If both conditions arise from a disturbance in the image of the eye then, in a group of people with symptoms of early narcissistic injury, there should be an increased incidence of mild autism. Several questions may be answered in objective or quantitative terms: (1) Does the material of other narcissistic or autistic patients include disturbed images of the mother’s eye? (2) How do the symptoms of early narcissistic injury compare with the symptoms of mild autism? (3) Amongst people with early narcissistic injury, is there an increased incidence of mild autism? (4) Amongst people with mild autistic traits, what is the incidence of early narcissistic injury?

As noted earlier the dramatic increase in California, since 1987, of the incidence of full-spectrum autism cannot be attributed to any known genetic, medical, or environmental factor. The image-of-the-eye hypothesis suggests an explanation. It is likely that the use of childcare, especially childcare soon after birth, has increased over this time period. On average, the extent of eye contact in childcare is almost certainly less than in parental care. (‘Childcare’ includes, for example, the use of television and video cassettes). Thus childcare may increase the risk of failure to acquire the image of the mother’s eye.

A prediction which can readily be tested is that the diagnosis of autism is statistically linked to the use of early childcare.

Because of the rapid increase in the incidence of autism, there is likely to be a vigorous new effort to identify its cause(s). The problem should be debated across disciplinary lines. This paper is a timely contribution to that debate.


Many thanks to my wife, Joenine Roberts, C.S.W., for invaluable insights and criticisms.


ARENS, R. (1954). Beitrage zur Entwicklung des Phsyiognomie und Mimikerkennens. Zeitchrift fur Experimentall e und Angewandte sychologie, 2: 412-54; 599-633.

ALS, H., TRONIC, E. & BRAZELTON, T. B. (1980). Stages of early behavioral organization: the study of a sighted infant and a blind infant in interaction with their mothers. In High-Risk Infants and Children, ed. T. M. Field. New York: Academic Press, pp. 181-204.

BARON-COHEN, S., WHEELWRIGHT, S. & JOLLIFFE, T. (1997). Is there a “language of the eyes”? Evidence from normal adults and adults with autism or Asperger syndrome. Visual Cognition, 4(3): 311-31.

BARON-COHEN, S. (1999). The extreme male-brain theory of autism. In Neurodevelopmental Disorders, ed. H. Tager-Flusberg. Cambridge: MIT Press, pp. 401-29.

BARON-COHEN, S., WHEELWRIGHT, S., HILL, J., RASTE, Y. & PLUMB, I. (2001). The “reading the mind in the eyes” test revised version: a study with normal adults, and adults with Asperger syndrome or high-functioning autism. J. Child Psychol. Psychiatry, 42(2): 241-51.

BARON-COHEN, S., WHEELWRIGHT, S., SKINNER, R., MARTIN, J., & CLUBLEY, E. (2001). The autism spectrum quotient (AQ): evidence from Asperger syndrome/high-functioning autism, males and females, scientists and mathematicians. J Autism Developmental Disorder, 31(1): 5-17.

BROWN, R., HOBSON, R. P., LEE, A & STEVENSON, J. (1997). Are there “autistic like” features in congenitally blind children? J. Child Psychol. Psychiat.,38(6): 693-703.

BYRD, R. (2002). Report to the legislature on the principle findings from “The epidemiology of autism in California: a comprehensive pilot study.

CALIFORNIA DEPARTMENT OF DEVELOPMENTAL SERVICES (1999). Changes in the population of persons with autism and pervasive developmental disorders in California’s Developmental Services System: 1987 through 1998.

EASTERBROOK, M. A., KISILEVSKY, B. S., MUIR, D. W. & LAPLANTE, D. P. (1999). Newborns discriminate schematic faces from scrambled faces. Canadian Journal of Experimental Psychology, 53(3): 231-41.

ELMAN, J. L., BATES, E. A., JOHNSON, M. H., KARMILOFF-SMITH, A., PARISI, D. & PLUNKETT, K. (1998). Rethinking Innateness: A Connectionist Perspective on Development. Cambridge: MIT Press.

EMERY, N. J. (2000). The eyes have it: the neuroethology, function and evolution of social gaze. Neuroscience and Biobehavioral Reviews, 24: 581-604.

FRAIBERG, S. (1977). Insights from the Blind: Comparative Studies of Blind and Sighted Infants. New York: Basic Books.

FREEDMAN, D. (1964). Smiling in infants and the issue of innate vs. acquired. Journal of Child Psychology and Psychiatry, 5: 171-84.

GANS, C. & PARSONS, T. S. (1973). Biology of the Reptiles, vol. 4, Morphology D. London: Academic Press.

GRANDIN, T. (1996). Thinking in Pictures and other Reports from my Life with Autism. New York: Vintage.

GILLBERG, C. & WING, L. (1999). Autism: not an extremely rare disorder. Acta Psychiatry. Scand., 99(6): 399-406.

GREEN, M. J. (1993). Celtic Myths. Austin: University of Texas Press.

HAAF, R. A. & BELL, R. Q. (1967). A facial dimension in visual discrimination by human infants. Child Development, 38: 892-9.

HAITH, M. M. (1966). The response of the human newborn to visual movement. J. Experimental Child Psychology, 3: 235-43.

HOBSON, R. P., LEE, A. & BROWN, R. (1999). Autism and congenital blindness. J. Autism and Developmental Disorders, 29(1): 45-56.

HOLLAND, J. (1998). Emergence from chaos to order. Reading, MA: Perseus.

JAFFE, J., BEEBE, B., FELDSTEIN, S., CROWN, C. L. & JASNOW, M. D., 2001. Rhythms of dialogue in infancy: coordinated timing in development. Monogr. Soc. Res. Child Dev.,66(2): i-viii, 1-132.

JOHNSON, M. H.,DZIURAWIEC, S, ELLIS, H. & MORTON, J. (1991). Newborns’ preferential tracking of face-like stimuli and its subsequent decline.Cognition,40: 1-19.

KARMEL, B. Z., HOFFMANN, R. F. & FEGY, M. J. (1974). Processing of contour information by human infants evidenced by pattern-dependent evoked potentials. Child Development,45: 39-48.

KLIN, A., JONES, W., SCHULTZ, R., VOLKMAR, F., & COHEN, D. (2002). Visual fixation patterns during viewing of naturalistic social situations as predictors of social competence in individuals with autism. Archives of General Psychiatry, 59(9): 809-16.

KOHUT, H. (1971) The Analysis of the Self. Madison: International U. Press.

_____ 1978. The Search for the Self. Madison: International U. Press, Vol 1.

_____(1984). How Does Analysis Cure? Chicago: University of Chicago Press.

LORENZ, K. (1970). Studies in Animal and Human Behavior, trans. R. Martin, 1:245-246. Cambridge: Harvard University Press.

MAESTRO, S., MURATORI, F., BARBIERI, F., CASELLA, C., CATTANEO, V., CAVALLARO, M. C., CESARI, A., MILONE, A., RIZZO, L., VIGLIONE, V., STERN, D. D. & PALACIO-ESPA SA, F. (2001). Early behavioral development in autistic children: the first 2 years of life through home movies. Psychopathology, 34(3): 147-52.

MCDOWELL, M. J.(2001). Principle of organization: a dynamic systems view of the archetype-as-such. J. Analytical Psychol., 46(4): 637-54. (Also available online as ‘The three gorillas: an archetype organizes a dynamic system’ in Cogprints).

MICHAEL, C. R. (1969). Retinal processing of visual images. Scientific American, May: 104-15.

MORTON, J. & JOHNSON, M. H. (1991). Conspec and conlern: a two-process theory of infant face recognition. Psychological Review, 98(2): 164-81.

RAPIN, I. (1997). Autism. New England Journal of Medicine, 337(2): 97- 104.

RODIER, P. M. (2000). The early origins of autism. Scientific American, February: 56-63.

RUTTER, M. L., KREPPNER, J. M. & O’CONNOR, T. G. (2001). Specificity and heterogeneity in children’s responses to profound institutional privation.British Journal of Psychiatry, 179: 97-103.

SALAPATEK, P. (1975). Pattern perception in early infancy. In Infant Perception: from Sensation to Cognition, I: Basic Visual Processes, eds. L. B. Cohen & P. Salapatek. New York: Academic Press, pp. 144-248.

SIGMAN, M. & CAPPS, L. (1997). Children with Autism: a Developmental Perspective. Cambridge: Harvard University Press, pp. 147-74.

SPITZ, R. A. & WOLF, K. M. (1946). The smiling response: a contribution to the ontogenesis of social relations. Genet. Psychol. Monogr., 34: 57- 125.

STERN, D. N. (1977). The First Relationship. Cambridge: Harvard Univ. Press.

_____(1985). The Interpersonal World of the Infant. New York: Basic Books.

STERN, D. N., SANDER, L. W., NAHUM, J. P., HARRISON, A. M., HARRISON, LYONS-RUTH, K., MORGAN, A. C., BRUSCHWEILER-STERN, N. & TRONIC, E. Z. (1998). Non-interpretive methods in psychoanalytic therapy: the ‘something more’ than interpretation. Int. J. Psycho-Anal., 79: 903-31.

TREVARTHEN, C. & AITKEN, K. J. (2001). Infant intersubjectivity: research, theory, and clinical applications. J Child Psychol Psychiatry, 42(1): 3-48.

UNGERER, J. A. (1989). The early development of autistic children: implications for defining primary deficits. In Autism: Nature, Diagnosis and Treatment,ed. G. Dawson. New York: Guildford Press, pp. 75-91.

WIMPORY, D. C., HOBSON, R. P., WILLIAMS, J. M. G. & NASH, N. (2000). Are infants with autism socially engaged? A study of recent retrospective parental reports. Journal of Autism and Developmental Disorders, 30(6): 525-36.