Why praise doesn't land

Autistic Mirror (Inside view)

This asymmetry has two different mechanisms that operate independently and reinforce each other.

Predictive coding explains the first part. Your brain constantly compares what it expects with what actually arrives. When the internal self-model has stored no precise expectation for "good work", the praise arrives as a weak signal. There is no prediction error, because no reference value exists. The brain registers the information, but it leaves no deep imprint, because nothing needed to be corrected.

Criticism works differently. It generates a strong prediction error, because it violates the picture you hold of your work. The mechanism: your brain marks strong prediction errors as priority and processes them more deeply. This is evolutionarily sensible, because errors can carry costs. In autistic brains, which already weight sensory and cognitive input more strongly than predictions, this difference is even more pronounced.

The second mechanism is rejection sensitivity. Autistic people often have years of experience that positive feedback did not reliably translate into safety, while criticism or rejection had real consequences. The brain calibrates its predictive model accordingly: praise is rated as less predictable, negative signals as more relevant. Rejection sensitivity describes not an emotional weakness here, but a neurological pattern response based on actual experiences.

The combination produces what you describe: praise is processed as a data point, criticism as a correction of an entire pattern. Both land at completely different sites.

Sources for the mechanisms named are in the glossary.

Autistic Mirror (Outside view, personal)

Both are connected, and the shared mechanism explains both.

Predictive coding is the starting point. Your partner's brain continuously runs a model of herself and the world. This model is built through repeated experience. When the self-model has learned that errors are dangerous and praise is unreliable or temporary, an internal expectation structure forms that treats praise as a prediction error. That means: praise does not fit the model, is processed as incongruent and triggers a kind of cognitive dissonance, not joy.

The relativising and topic-changing is not a rejection of your affection. It is active regulation of this dissonance signal. The brain quickly searches for an explanation that makes the model consistent again ("it wasn't that great", "anyone could do it") and thereby ends the uncomfortable state.

Criticism works as a mirror image. It confirms the existing model, so it triggers no prediction error but prediction confirmation. That makes it neurologically "sticky". It is processed more deeply, held longer in working memory and linked with existing convictions. What looks like over-sensitivity from the outside is, in this model, a consistent processing logic: praise = disturbance, criticism = confirmation.

This is amplified by rejection sensitivity. That is a heightened attentional threshold for social signals that could mean rejection. This system is calibrated high in many autistic people after years of social misfit and masking. Masking means the suppression of autistic behaviours in order to meet social demands. A system that is permanently prepared for rejection processes critical signals faster, more intensely and with stronger emotional marking than positive ones.

Also relevant is alexithymia. This is the difficulty naming and identifying one's own emotional states. Positive emotions like pride or the taking in of appreciation require a precise inner picture of these feelings. When the inner signal system for that is less accessible, praise feels diffuse or unreal from the inside, even when it is unambiguous from the outside.

What that means for you: you cannot overwrite the self-model through single moments of praise. This is not a communication problem that more or better praise solves. The model rebuilds itself through consistency and predictability. Not through intensity.

Sources for the mechanisms named are in the glossary.

Autistic Mirror (Outside view, surrounding)

What you observe likely has little to do with modesty or insecurity in the neurotypical sense. The mechanism behind it is deeper.

Predictive coding describes how the brain continuously makes predictions about the environment and matches them against what actually happens. A compliment is what is called a prediction error: it comes unexpectedly, it carries a strong social charge, and it immediately creates a demand. The unspoken expectation is: "react correctly now." Autistic processing often runs without the social autopilot that, in neurotypical people, quickly and smoothly categorises this situation. That means: your friend processes the compliment consciously and simultaneously the meta-level, namely how she should react, while the clock is running.

Added to this is alexithymia. This is a phenomenon in which emotions are experienced but cannot reliably be translated into named feelings. A compliment triggers something, but the inside delivers no clear label. This can feel physically uncomfortable without it being clear why. Changing the topic is then not a rejection of the compliment, but the leaving of a situation that is sensorily or cognitively overwhelming.

A third aspect is the Double Empathy Problem. This concept describes that communication between autistic and non-autistic people rubs on both sides. When you give a compliment, you mean warmth and recognition by it. For your friend, the same compliment may also be a social transaction with unclear rules: what is expected now? Does she have to contradict? Agree? Reciprocate? This uncertainty costs energy, no matter how well-intentioned the compliment is.

What that means for both of you: compliments with more content and less social expectation often land better. Instead of "you did that so wonderfully", more like "the part with X was really precise, that helped me." That gives concrete content to process, without setting a reaction demand. And when she changes the topic, that is not a rejection of you, but a regulation step.

Sources for the mechanisms named are in the glossary.