It’s common for researchers to compare an “autism group” with a “control group,” on some dimension. Perhaps some part of the brain differs in size or in activation, or the neural network graph has a different shape, or kids with autism have more allergies, or they are less prone to certain cognitive biases, or whatever, the list is a mile long. Then we get a “such-and-such correlates with autism” discovery.

In writing up their research, scientists get out their mental thesaurus and try to avoid saying “more-connected neural network” and “less-connected neural network” over and over. They might start to say things like “connectivity deficit” or “excess connectivity” or “abnormal neural network” or “pathological neural network” or “defective gene.” If the autism group has more then it’s “excess” and if it has less it’s “deficit.”

This is bad science.

Think about what you’d have to prove to justify the words “excess” and “deficiency.” You would need to determine the “best” genotype or phenotype. Then you would need to know thresholds (how far someone can be from “best” before it’s a practical problem). When was the last time an autism study showed either of these?

Just because we demonstrate that something is correlated with autism, we haven’t demonstrated that it is “bad.” We haven’t shown that it results in the negative aspects of autism. We haven’t demonstrated that it’s unequivocally bad; perhaps it’s a mixed blessing. Or perhaps it’s something innocuous and irrelevant (it could share a cause with some forms of autism, but have no functional effect). Or perhaps it’s an effect of autism, rather than a cause.

In most correlation studies, there’s no evidence whatsoever about how or why the thing in question relates to autism, there’s just an unexplained correlation.

Many biology textbooks discuss sickle-cell disease, a genetic trait which can be an advantage: in heterozygous form, it confers resistance to malaria. In tropical regions, evolution tends to select for this gene, even though it has major downsides.

The various conditions which involve autistic traits are undoubtedly more genetically complex. If we are humble and scientific, we might assume that many traits which correlate with autism exist in our genetic heritage for a reason. Since we don’t know what the conditions are, what the endophenotypes are, or how it all relates to observed autistic traits, we need to reserve judgment on whether anything we observe constitutes a “deficit” or “excess.” Stick to saying it’s a difference.

Even when a trait is plainly a deficit in one context, it may be a strength in another. Is sickle-cell disease a net negative? — well, do you live in a malaria-infested region or not?

In my speculative theory about how autism works in our family, it’s easy to see how causes which contribute to “intellect” and causes which contribute to “instinct” would need to be in balance. A person too lopsided in either direction will have trouble. None of the genes or other causes which push the balance one way or the other would be “abnormal” or “deficits” in themselves; only an imbalance or preponderance — potentially in either direction — would cause a problem.

I’m not saying this is how our particular autism condition works, or how any condition works. It’s pure speculation. But the point isn’t that I’m right about this, it’s simply that in many possible scenarios, we don’t know whether the phenomena we’re observing are net positive, net negative, context-dependent, irrelevant, or what.

Outside of autism, think about intelligence. Why are human IQs distributed as they are? Well, some people say that IQ at the very upper limits of the distribution can be a liability. Anecdotally, people with IQs of 150 or 160 may be barely functional. So perhaps intelligence, also, requires a balance. Perhaps we need some genes to make us smart enough, and some other genes that keep us from being too smart. In this balance, there may be many genes which are neither categorically “good” nor categorically “bad.”

Scientific papers should strive to use neutral language, reserving judgment on the merits of the observed phenomena, unless there’s actual evidence about those merits. This goes double for poorly-understood genetic and neurological features.

No “deficits” and “abnormalities” and “dysfunctions,” just the facts.

This should be done to promote sharper, clearer, more precise, more accurate research.

There are ethical implications as well (we should be concerned about stigma and eugenics). But even if one doesn’t care about ethical considerations, factual considerations ought to rule out language based on unproven assumptions about the “best” genotype or phenotype.

(To be clear, I’m not talking about an observable problem described as such; “difficulty conducting a conversation” can be described as a “pragmatic deficit” and that’s fine. The problem comes when we have some proposed underlying genetic or neurological feature, but no proof of that feature’s specific effect on observable behaviors or quality of life.)