I have spent the past week trying to write a single paragraph about Spaceflight-Associated Neuro-Ocular Syndrome, and I have failed seven times. The first draft said SANS is “a condition in which astronauts develop swelling of the optic nerve.” A colleague who reads retinal scans for a living circled the word “swelling” and wrote in the margin: edema, distension, or disc elevation? They are not the same thing. The second draft distinguished optic disc edema from optic nerve sheath distension. A neuro-ophthalmologist then asked whether I was describing a mechanical phenomenon, a vascular one, or a cerebrospinal fluid compartment syndrome. By the fifth draft, I had a paragraph that was technically defensible and completely unreadable. By the seventh, I understood that the writing problem was not a failure of clarity. It was a mirror of the science itself.
This is not a confession of incompetence. It is an argument that the craft of writing about frontier medicine—the kind practiced in orbit, where sample sizes are small, mechanisms are contested, and every conclusion carries an asterisk—is a form of systems thinking. Every qualifier, every cited limitation, every honest admission of uncertainty is a load-bearing component of the final structure. Remove one, and the whole thing collapses into false confidence. Add too many, and the reader cannot move through the prose. The writer’s job is to engineer a text that holds its shape under the weight of what we do not yet know.
I want to use SANS as the case study because it is the most interesting acronym in space medicine and because the debate over whether it is one syndrome or several is exactly the kind of problem that breaks bad science writing and reveals good science writing for what it is: a discipline of judgment, not simplification.
The Acronym and the Argument
SANS stands for Spaceflight-Associated Neuro-Ocular Syndrome. It was named in 2011 after researchers recognized a pattern of ocular changes in astronauts returning from long-duration missions on the International Space Station. The findings included optic disc edema, globe flattening, choroidal folds, hyperopic shifts in refraction, and, in a minority of cases, elevated intracranial pressure measured by lumbar puncture after landing. Not every astronaut developed every finding. Some developed none. The pattern was inconsistent enough that the term “syndrome” was chosen deliberately—it signaled a cluster of signs that appeared together more often than chance would predict, without committing to a single underlying mechanism.
That commitment, or the refusal to make it, is where the writing problem begins. A syndrome is a descriptive label, not an etiological one. It says: these things co-occur; we are not yet sure why. But the gravitational pull of narrative is toward causation. Readers want to know what causes SANS. Editors want a clean lede. The writer feels the pressure to deliver a sentence like “SANS is caused by the cephalad fluid shift that occurs in microgravity.” That sentence is not false, but it is incomplete in ways that matter. The cephalad fluid shift—the movement of blood and interstitial fluid toward the head when gravity no longer pulls it toward the feet—is almost certainly involved. But the same fluid shift happens in every astronaut, and not every astronaut develops SANS. Some develop optic disc edema without globe flattening. Some develop choroidal folds without a refractive shift. The relationship between the fluid shift and the ocular findings is mediated by variables we are still trying to name: individual anatomy of the optic nerve sheath, compliance of the lamina cribrosa, one-carbon metabolism polymorphisms, exercise countermeasure adherence, cabin CO₂ levels, and possibly factors we have not yet measured.
Writing about SANS therefore requires holding multiple causal strands in suspension while still producing a sentence a colleague could explain over coffee. That tension—between narrative clarity and scientific fidelity—is not a bug in the system. It is the system.
The One-Syndrome-or-Several Problem
In 2023, a group of researchers published a paper arguing that SANS should be split into two distinct entities: one driven primarily by elevated intracranial pressure, the other by local ocular changes independent of pressure. The argument rested on the observation that some astronauts with optic disc edema had normal intracranial pressure on lumbar puncture, while others with elevated pressure did not have disc edema. The imaging data suggested that globe flattening and choroidal folds could occur without optic nerve sheath distension, and vice versa. The paper was careful, well-referenced, and immediately controversial.
The controversy is instructive because it is not about data quality. The measurements are as good as orbital constraints allow. The disagreement is about taxonomy: how we decide when a cluster of signs is one thing or several. That decision is partly empirical and partly philosophical. It depends on what you think a syndrome is for. If a syndrome is a clinical tool for predicting outcomes, then lumping or splitting should be judged by prognostic utility. If a syndrome is a research framework for investigating mechanisms, then splitting may be premature until we have evidence that the proposed subtypes have distinct etiologies. The 2023 paper made a mechanistic argument—that pressure-driven and non-pressure-driven SANS are different processes—but the evidence for distinct etiologies was correlational, not causal. The authors acknowledged this in the limitations section, which I read three times before I wrote anything about the paper.
This is where the craft of writing and the craft of science converge. The limitations section of a paper is not an afterthought; it is the part of the structure that tells you how much weight the conclusions can bear. When I write about SANS, I cite the limitations with the same respect I give the findings. I do this not because I am cautious by temperament but because the reader deserves to know whether the claim they are absorbing is a confirmed finding, a promising correlation, or an open question. Distinguishing those categories is the central ethical act of science writing.
The Authors Guild, in its AI Best Practices for Authors, articulates a related principle: that preserving human thinking and voice in writing is not a stylistic preference but a professional standard. The Guild’s guidelines emphasize that AI outputs are “generic mashups of pre-existing works” and that the writer’s unique judgment—what to include, what to qualify, what to leave unresolved—is what makes a piece of writing worth reading. When I am deciding whether to describe SANS as one syndrome or several, I am not performing a mechanical task. I am making a judgment that no language model can make because the judgment depends on weighing evidence that is itself unsettled. The tool can help me structure the paragraph, but it cannot tell me where to place the qualifier.
The Qualifier as a Load-Bearing Component
Consider the following sentence, which I wrote in draft four and then dismantled:
“SANS is caused by the cephalad fluid shift that occurs in microgravity, which increases intracranial pressure and compresses the optic nerve.”
This sentence has the virtue of being clear. It has the vice of being wrong in ways that matter. The cephalad fluid shift does not uniformly increase intracranial pressure; lumbar puncture measurements in astronauts show a range from normal to mildly elevated, and the relationship between pressure and optic disc edema is inconsistent. The optic nerve is not compressed in a simple mechanical sense; the current model involves a complex interaction between the optic nerve sheath, the lamina cribrosa, and the translaminar pressure gradient. The word “caused” implies a level of certainty the literature does not support.
Now consider the version I settled on after draft seven:
“The leading hypothesis is that the cephalad fluid shift of microgravity alters the pressure gradient across the lamina cribrosa, contributing to optic disc edema in some astronauts; however, individual susceptibility factors—including optic nerve sheath anatomy, one-carbon metabolism variants, and cabin CO₂ exposure—appear to modify the response, and the relationship between intracranial pressure and ocular findings remains inconsistent enough that some researchers argue SANS may represent more than one underlying process.”
This sentence is harder to read. It contains five qualifiers. It refuses to resolve the one-syndrome-or-several question. But it is honest, and it is load-bearing. Each qualifier carries a piece of the evidence: the word “leading” signals that other hypotheses exist; “contributing to” replaces “causes”; “in some astronauts” acknowledges the inter-individual variability that is the central puzzle of SANS; “appear to modify” distinguishes correlation from demonstrated causation; “inconsistent enough” tells the reader that the data do not yet converge. The sentence is not elegant, but it is structurally sound. It will not collapse under scrutiny.
This is what I mean by systems thinking in writing. A piece of science prose is a system of claims, qualifiers, and evidence. Each component interacts with the others. If you strengthen one claim without adjusting the surrounding qualifiers, you introduce a stress point. If you remove a qualifier to improve flow, you may cause a failure downstream when a reader who knows the literature encounters an unsupported assertion. The writer’s task is to distribute the weight of uncertainty across the structure so that no single sentence bears more than the evidence can hold.
What Tools Can Do and What They Cannot
I want to be precise about the role of writing tools in this process because the conversation about AI and writing tends to oscillate between utopianism and panic. A well-designed writing environment can help a writer manage complexity. It can surface related papers, suggest structural reorganizations, flag sentences that exceed a readability threshold, and remind the writer that a claim made on page two needs a qualifier on page seven. These are genuinely useful functions. The Purdue OWL’s Creative Writing resources describe the discipline of creative nonfiction as one that requires the writer to balance clarity with complexity, to make choices about structure and voice that no algorithm can prescribe. The same principle applies to science writing. A tool can show me that my paragraph has too many dependent clauses. It cannot tell me which qualifier to keep and which to cut, because that decision depends on my assessment of the evidence—an assessment that is itself provisional and subject to revision when new data appear.
This is why I am interested in tools like the Unsloppy AI Writing App, which approaches the problem from the direction of structure rather than generation. The premise—that writing about complex, uncertain material benefits from an environment that helps the writer organize claims, track qualifiers, and maintain fidelity to source material—is sound. It does not promise to write the article for you. It promises to help you keep the load-bearing components in place while you do the writing yourself. That distinction matters because the central act of science writing is not the production of text; it is the exercise of judgment about what the text should say and what it should withhold. No tool can perform that act for you, but a good tool can reduce the cognitive overhead of managing the structure so that you can focus on the judgment.
I should be clear: I am not describing a tool I have used to write this article. I am describing a category of tool that the craft of science writing needs—one that treats qualifiers as first-class objects, not as clutter to be edited out. The current generation of AI writing assistants tends to optimize for fluency, which often means removing the very qualifiers that carry the scientific weight. A tool that optimizes for fidelity instead would be a genuine contribution to the practice.
The Mirror Between Writing and Science
The reason this matters beyond the craft of prose is that the process of writing about uncertain science recapitulates the process of doing uncertain science. Both require you to hold multiple hypotheses in mind without prematurely collapsing them into one. Both require you to distinguish what you have measured from what you have inferred. Both require you to communicate your findings to colleagues who will scrutinize every claim. And both require you to revise your framework when new data arrive.
When I write about SANS, I am forced to confront the same taxonomic question the researchers are confronting: is this one syndrome or several? I cannot answer it definitively because the researchers cannot answer it definitively. What I can do is represent the state of the question honestly, with the qualifiers that signal where the evidence is strong and where it is thin. That representation is itself a form of knowledge. It tells the reader not just what we think we know but how we know it and how confident we are. In a media environment that rewards certainty, that kind of writing is countercultural. It is also, I would argue, the only kind of science writing worth doing.
The next time I write about SANS—and there will be a next time, because new data are coming from the current cohort of long-duration astronauts and from the first commercial missions carrying researchers—I will probably fail another seven drafts. The failure will not be a sign that I am bad at my job. It will be a sign that the science is still unsettled and that I am trying to write about it without lying. The qualifiers will shift. Some will become stronger; others will be removed because the evidence has firmed up. The structure will need to be rebuilt. That is not a problem. That is the craft.
What This Means for Readers (and Writers)
If you are a reader who encounters a piece of science writing that feels too smooth—every claim declarative, every mechanism explained, every uncertainty resolved—I would encourage you to be suspicious. The frontier of human knowledge is not smooth. It is jagged with contradictory data, small sample sizes, and measurements taken under constraints that would make a terrestrial researcher wince. Good science writing does not sand down the jagged edges. It shows them to you and explains why they are there.
If you are a writer who struggles with the kind of paragraph I have described, the struggle is not a sign that you are failing at clarity. It is a sign that you are taking the material seriously. The qualifier is not your enemy. It is the unit of intellectual honesty. Learn to place it precisely, to weigh it against the flow of the sentence, and to recognize when removing it would make the prose cleaner but the meaning false. That skill is not taught in most writing courses, but it is the core competency of anyone who writes about science at the edge of what we know.
The body becomes a question in extreme environments. Writing about that question with precision and warmth—with every qualifier in its load-bearing place—is how we invite readers into the inquiry without pretending we have already arrived at the answer.