462 four-month-old infants. A battery of unfamiliar sights, sounds and smells. Then follow-ups at two years, four and a half, and seven and a half.
20%
high-reactive
40%
low-reactive
40%
neither
Somebody has described you in one word and you have carried it around ever since. Quiet. Difficult. Easy. Sensitive. The word arrived early, from an adult who meant nothing much by it, and then it became a thing you knew about yourself.
Mine was shy. I have spent a long time treating it as a personality flaw with a training plan attached.
Then a course reading put a mechanism underneath it, and — more usefully — put a boundary around what the mechanism is allowed to claim.
What the assignment asked
The weekly response papers in this course came with a short, unfussy specification:
These papers are reactions to the Questions of the Week that are posed at the end of each Monday’s lecture… roughly 1½ – 2 pages in length.
It’s okay to have your opinion feature prominently in the response papers. However, to receive full credit (a check), your own thoughts must be skillfully integrated with material from the lectures, sections, and/or readings.
In other words, we are looking for evidence that you’re thinking critically about the material you’ve encountered in the course.
“Skillfully integrated” is doing a lot of work in that sentence, and it is the whole difficulty of this particular reading. When a study describes a child who sounds like you were, the temptation is to write two pages of recognition and call it integration. Recognition is not evidence. It is the reader agreeing with themselves.
The study
Jerome Kagan’s team ran a battery of unfamiliar events past 462 healthy four-month-old infants — the mother looking down without speaking, brief sentences through a speaker, novel smells and sights. Nothing painful. Just new.
About a fifth of the babies responded with vigorous motor activity, arching of the back and crying. Those were classified high-reactive. Forty per cent showed little movement and minimal distress: low-reactive. The rest fell into neither group.
The proposed mechanism is a threshold. An amygdala that fires readily at the unfamiliar produces a baby that thrashes and cries; a higher threshold produces one that watches. Kagan is careful about how far this goes — he notes there are at least 150 candidate brain molecules involved, that no gene or set of genes has been found to be consistently associated with a personality independent of the child’s gender, ethnicity and life history, and that the biases discovered so far are a small fraction of those that could exist.
Then they followed the children.
At four and a half years — classified inhibited
Were high-reactive
46%
Were low-reactive
10%
At seven and a half — classified “anxious”
Were high-reactive
45%
Were low-reactive
15%
One thing is worth knowing before carrying that 45% anywhere. “Anxious” at seven and a half was not a diagnosis. Mothers filled out a questionnaire about their child’s fears; the ones who described a highly anxious child were phoned to check that the description held up; then teachers ranked those children against classmates for shyness, fearfulness, aggression and distractibility. Only where mother and teacher agreed did a child go into the group Kagan labels — in his own quotation marks — “anxious.” Forty-three children, about a quarter of the sample. It is a careful instrument, and it is still two adults agreeing about a word.
Read those numbers the way a headline would and you get your baby’s future, decided at four months. Read them properly and they say something much stranger and much more interesting.
The finding is a constraint, not a prediction
Here is the sentence that reorganised the reading for me. Kagan reports that not one high-reactive four-year-old showed the fully complementary profile: low fear in the second year and uninhibited behaviour at four and a half. At seven and a half, 29% of low-reactives had run the whole uninhibited, non-anxious course — against 2% of high-reactives.
And he states the implication directly: it is easy to sculpt high- or low-reactive infants into children who are neither extreme, and hard to produce a consistently uninhibited profile in a high-reactive child.
That is not a prediction. A prediction says where you will end up. This says where you will not.
The closed corner is consistently, extremely uninhibited — not "confident," not "capable in public," not "fine." One end of a distribution, not a life.
I deal in constraints for a living, and I still misread this on the first pass. A constraint is the more modest claim and the more durable one. “This system will handle 10,000 requests per second” is a forecast and it is usually wrong. “This system will never return a result faster than the disk can read” is a bound, and it holds. Temperament, on this evidence, is the second kind of statement.
Not one — high-reactive four-year-old showed low fear at two and uninhibited behaviour at four and a half.
Of low-reactives — ran the full uninhibited, non-anxious course to seven and a half.
Of high-reactives — did the same.
The moment in the study I keep thinking about
At four and a half, an examiner ran the children through a battery of tests. In one episode she opened a photo album, took out a large colour photograph, handed it to the child and said: “This is my favourite picture; tear up my favourite picture.”
More low-reactives than high-reactives either asked her why, or — five of them — simply refused. Kagan is precise about what this was and wasn’t: the refusal was not accompanied by anxiety. These children were less afraid of disobeying an authority figure when the request required violating a norm they had accepted.
Almost all the high-reactives were reluctant to disobey and, after a short delay, tore a small corner from the photograph.
I have read that paragraph several times. It is a small laboratory episode and it is also, obviously, a description of every meeting I have ever sat in where something was clearly wrong and the objection stayed in my throat. The children who pushed back weren’t braver in the heroic sense. They just found the unfamiliar act of refusing less expensive. And the detail that does the work is the size of the compliance: not tearing it up, but pausing, and then removing the smallest piece that would still count as having obeyed.
Where I have to stop myself
This is the part of the response paper that mattered more than the recognition, and it is the reason the guidelines say critically.
The sample was 462 healthy, Caucasian, middle-class infants in the United States. I did not grow up in that population. What “inhibited with an unfamiliar adult” means, and what an adult does about it, is not constant across households or countries — a quiet child gets read as respectful in one setting and as a problem in another, and the reading changes the experience that does the sculpting. Kagan’s own framing insists that outcome depends on life history. The mechanism may generalise; the base rates and the developmental path should not be assumed to.
Recognising yourself in a temperament category is the weakest evidence in the building. Forty-five per cent of high-reactives were classified “anxious” by that procedure at seven and a half, which means a clear majority were not. The category is compatible with almost every adult outcome. If I read the profile and feel seen, that tells me about the readability of the profile, not about my amygdala. Personality descriptions are engineered to be recognisable — I have written about how an early self-description survives contact with a real instrument, and it mostly doesn’t.
And the direction of the inference is backwards. Kagan measured babies and followed them forward. I am an adult reconstructing an infancy I cannot remember, from a story my family tells, to match a category I have just read about. That is not the study’s design. It is the exact retrospective move the study was built to avoid.
So: I don’t know that I was a high-reactive infant. I know the category exists, that it is measurable at four months, that it constrains one end of a range, and that the constraint is far narrower than the word shy ever implied.
“Shy” — global, edgeless, and quietly asserts a ceiling on the whole person.
One end of a distribution — specific, measurable at four months, with most of the range open and life history doing the work inside it.
What I actually took from it
The useful thing was not being explained. It was getting a smaller and more accurate claim to carry around.
“Shy” is a global label. It attaches to the whole person, it has no edges, and it quietly asserts a ceiling — this is what you are, so this is what you can do. What the research supports is nothing like that. It supports a specific, bounded statement about one end of a distribution of responses to the unfamiliar, with most of the range still open and life history doing most of the work inside it.
Swapping the global label for the bounded one changed what I do with it. A ceiling is something you resent. A bound is something you plan around — you accept that the first thirty seconds of an unfamiliar room will cost you more than they cost the person beside you, and then you build in the thirty seconds. That is not a personality change. It is a scheduling decision.
Thanks to whoever chose to put this reading in the anxiety week rather than the personality week, which is the difference between here is a mechanism and here is a type of person.
One thing I’d ask, if you’ve read this far with your own early label in mind: what would it take for you to find out whether it was ever true?
Comments