Strong for What?
Thinking of data, citations, claims, and human stories by the jobs they do
When a piece of writing feels weak, one instinct is to add something stronger: more data, a better study, a concrete example, a human story, or a clearer claim.
This makes intuitive sense. If the argument is not strong enough, perhaps it simply needs stronger material.
But I am not sure that is always the right way to think about it.
The same piece of data can stabilize an argument, challenge it, or send it somewhere new. A single human story can provide evidence of existence, make an abstract structure visible, create emotional weight, or reveal a variable the writer had not noticed before.
The material has not changed. What has changed is the job we are asking it to do.
So perhaps there is another question worth asking before adding anything to a piece of writing:
What can this element make the argument do?
Let the data do something other than prove
Data is easy to think of as evidence. You have a claim, you add numbers, and the numbers make the claim stronger.
Sometimes that is exactly what happens.
A dataset can create stable ground. It can tell us that, within certain limits, something has actually been observed. It can establish the point from which the next part of an argument begins.
But data can also do almost the opposite.
Suppose an existing explanation seems to account for what is happening, and then one number appears that does not fit. The useful question may no longer be whether the data supports the explanation, but whether the explanation can account for that number.
The data has become pressure rather than support.
Or perhaps the interesting part is not the average at all. A distribution contains a strange cluster. Two variables that seemed likely to move together suddenly separate. One subgroup behaves differently from the rest.
Now the data is not answering the question. It is generating the next one.
The same dataset can therefore act as ground, pressure, or sensor. Calling all three uses “evidence” is not wrong, but it can hide what the data is actually doing to the argument.
One person can strengthen different things
Human stories make this even less tidy.
Suppose I am writing about a system that seems to create a particular kind of problem. I can describe the system abstractly, and then add one person: what happened to them, what they were told, what they lost, what they tried next.
The argument changes immediately.
The structure becomes easier to see. The distance between the reader and the issue becomes smaller. The piece may become more memorable or emotionally heavier.
And sometimes one person is also very strong evidence.
If someone says, “This never happens,” one well-established case may be enough to defeat that claim.
But that same case does not necessarily establish, “This is what usually happens.”
The person has not changed. The job has.
One case can be extremely strong for existence and weak for prevalence. It can be weak as a basis for generalization and still be excellent at making an invisible structure visible. It can be statistically uninformative and still reveal a variable that aggregate data had hidden.
This matters because a human story often makes a piece feel stronger. But what became stronger may have been the evidential force, the concreteness, the emotional weight, the salience, the memorability, or the visibility of the mechanism.
Those are not interchangeable.
A human story can strengthen a piece without strengthening every claim inside it.
That distinction becomes especially important when the story is vivid. The more memorable the person becomes, the easier it is to let the weight of the story travel farther than the evidence actually does.
“Anecdotal” does not tell us what the case is for
There is a familiar way to dismiss an individual case: it is only anecdotal.
Sometimes that criticism is exactly right. If one person is being used to establish what happens across a population, the limits are obvious.
But sometimes the criticism misses the job the person was given.
Perhaps the writer is not trying to estimate prevalence. Perhaps the person is there to show what an abstract mechanism looks like when it reaches an actual life, to reveal a sequence that aggregate numbers cannot show, to expose a variable missing from the original model, or to provide a counterexample to a universal claim.
In those cases, “n=1” remains true, but it is no longer a complete evaluation.
We tend to ask whether a piece of evidence is strong.
It may be more useful to ask whether it is strong for the job being assigned to it.
Keep the pressure on the idea, not the person
There is another reason human cases are complicated.
A person can carry an idea into a piece of writing. Someone makes an argument, and their words provide a concrete version of an assumption that would otherwise remain abstract.
That can be useful. You can take the assumption seriously, push it farther, and ask what follows from it.
But there is an important shift between saying, “This assumption creates a problem,” and saying, “This person is the problem.”
Get 102Protocol’s stories in your inbox
Join Medium for free to get updates from this writer.
The first keeps pressure on the idea. The second creates a new claim about the person, and that new claim now has to be defended.
The other side also gains an easier route out of the original argument. Instead of explaining the assumption, they can debate whether the characterization of the person is fair.
Attacking an idea can increase the explanatory burden on the other side. Attacking the person often increases your own.
This is one reason arguments can become weaker when they become more personal. The writing may feel sharper, while the logic becomes less efficient.
Calling people “tools” can sound uncomfortable, so the distinction matters here too. I do not mean that people should be treated instrumentally. I mean that once a writer places a person inside an argument, that representation is being asked to perform some function.
It is worth knowing what that function is.
Let a citation disturb the argument
Research citations are often used in a familiar way: “Research shows that…”
The study arrives as reinforcement. The writer has made a claim, and the citation helps protect it.
Again, there is nothing wrong with that. Sometimes a piece needs firmer ground.
But a citation can do other things.
A study can introduce a competing explanation, expose a contradiction, bring in a variable from another field, show that an apparently obvious assumption is not actually settled, or even change the writer’s question.
In that sense, a citation can be a shield. But it can also be a lever, a map, or a doorway.
A paper does not have to tell me that I am right.
It may be more useful when it makes the thing I was trying to prove less interesting than the question I had not yet noticed.
A claim can be a destination — or a probe
Claims themselves can also be treated as tools.
This is easy to miss because we usually imagine a claim as the place the argument is trying to reach. Sometimes it is.
But consider the difference between “Maybe we need less of this” and “What if we removed it entirely?”
The second statement is stronger in one sense, but it does not have to express greater certainty. It may simply have a larger range.
By pushing the assumption farther, we can see where something breaks. If nothing breaks, perhaps the original boundary was more arbitrary than we thought. If something does break, we have found a condition that matters.
The strong claim has functioned as a probe.
This is different from saying, “I have strong evidence that we should remove it entirely.”
One increases certainty. The other increases the amplitude of the thought experiment.
They can look similar on the page.
They are doing very different jobs.
Strength depends on the job
This begins to complicate some familiar hierarchies.
Data is strong. A peer-reviewed paper is strong. An anecdote is weak. An opinion is weaker.
Those rankings can be useful when the task is narrowly defined, for example when we are trying to establish whether a general empirical claim is true.
But writing does more than establish general empirical claims.
Sometimes we need to show that something exists. Sometimes we need to expose a boundary, make a structure visible, destabilize an explanation, or open a question rather than close one.
A hammer is not a stronger tool than a screwdriver in the abstract.
The useful question is what the job requires.
The same may be true in writing.
The question is not simply how strong the material is. The question is what job we are asking it to do.
This does not make evidence standards irrelevant. Quite the opposite.
Once the job is clear, we can be more precise about what kind of strength is needed.
If I want to generalize, one vivid story is probably not enough. If I want to show that something can happen, one well-established case may be decisive. If I want to challenge an explanation, the most useful datapoint may be the one that does not fit. If I want to explore, the best paper may not be the one that supports my conclusion, but the one that introduces a variable I had not considered.
Before adding more, ask what needs to happen
When a draft feels weak, “add more evidence” is an understandable response.
But perhaps there is a step before that.
It may help to ask what the argument needs at that point: firmer ground, pressure on an assumption, a more visible structure, evidence of prevalence, evidence of existence, a new question, a clearer boundary, or a conclusion that can actually close the argument.
Once that is clearer, the choice of material changes.
A number, a person, a study, or a strong claim does not have one fixed function.
Perhaps writing is less about assembling increasingly powerful ingredients than about deciding, moment by moment, what each piece is here to do.
