A decision-maker may think they are assessing facts, candour or insight while also assessing how easily a person’s speech is processed. When fluency becomes a proxy for credibility, a process can be formally open and still misread the evidence.
A lot of disputes arrive already named.
They are called performance problems, attitude problems, credibility problems, failures of insight or failures to take responsibility. Once one of those labels settles over a matter, communication itself begins to look like evidence for the label. A short answer can look evasive. A pause can look uncertain. A careful answer can look rehearsed. A direct answer can look rude. An answer that changes after reflection can look inconsistent.
Sometimes those inferences are justified. Sometimes they are not.
The problem is not that language should be ignored. It is that linguistic performance can become part of the case without anyone deciding that it should.
The original version of this essay began with a concern about meaning being lost between people operating with different linguistic habits, assumptions and degrees of power. The deeper problem is narrower than that. It is what happens when the ease of receiving an answer begins to stand in for the quality of the answer itself.
FLUENCY IS NOT CREDIBILITY
Credibility is an evidential judgment. Fluency is a feature of performance. They can interact, but they are not the same thing.
The research here is less tidy than the popular version. Some studies have found that statements delivered in a foreign accent are judged less credible. Later studies have not always reproduced that effect. Other work suggests that accent, familiarity and processing difficulty can shape how listeners experience a speaker, but there is no sound basis for a rule that foreign-accented speech is simply believed less.
That uncertainty is important. It moves the argument away from a claim that every listener is biased in the same way and towards a better question: what are we actually using as evidence?
If an explanation is difficult to follow, that matters. But difficulty for the listener can have several causes. The account may be incoherent. The question may be badly framed. The terminology may be unfamiliar. The speaker may be searching for a precise word. The listener may be unfamiliar with the accent or be working in an additional language too. The room may be tense. Several things may be true at once.
The mistake is to collapse those possibilities into a character judgment.
A person who sounds hesitant is not therefore dishonest. A person who speaks slowly is not therefore confused about the facts. A person who uses simple language is not therefore lacking insight. A person who speaks polished English is not therefore more reliable.
It is rarely about intelligence. It is about interpretation. And interpretation is where fairness can quietly collapse.
PRESSURE CHANGES THE PERFORMANCE
Everyday fluency and high-stakes fluency are not the same task.
A disciplinary interview, regulatory complaint, performance meeting or investigation asks a person to do several things at once. They have to understand the question, recall events accurately, recognise the implications of formal language, judge how much context is necessary, avoid creating new ambiguity and respond while being observed and evaluated. Doing that in an additional language can add another layer of cognitive work.
Research on second-language speaking gives some support to what practitioners see in real rooms: anxiety and cognitive load can affect visible fluency for some speakers. That does not mean pressure always makes someone less fluent. It means pauses, repairs, slower delivery or a change in verbal confidence can have explanations that are not about honesty or knowledge.
This matters because high-stakes processes often treat the first answer as though it were a clean measurement of the person’s understanding. It may not be. The answer is being produced under conditions created by the process itself.
A question can also contain its own conclusion. “Why did you do this?” is different from “Did you do this?” Formal words such as misconduct, dishonesty, integrity, insubordination and failure to follow instruction carry consequences that are not always obvious from ordinary-language understanding. A person may think they are agreeing to the factual description while missing the institutional meaning attached to it.
When that happens, later clarification can be misread as retreat. The person has not necessarily changed the story. They may finally have understood the question.
WHEN PERFORMANCE BECOMES PROOF
The dangerous turn is quiet. Communication stops being the medium through which evidence is given and starts being treated as evidence in itself.
Once a person is seen as credible, their explanation is given room. Once they are seen as not credible, the same explanation can be read as self-serving. The same pause that looked uncertain now looks evasive. The interpretation begins to generate the evidence that confirms it.
This is not unique to speakers of an additional language. Native speakers are also judged through class, region, education, disability, neurotype, confidence and familiarity of expression. But speaking in an additional language makes the distinction especially visible because linguistic ease is so tempting to mistake for evidential weight.
The same caution applies to culture. “Culture” is often invoked too quickly, as though a person’s nationality can explain eye contact, silence, deference, directness or willingness to disagree. That simply replaces one shortcut with another. People vary within cultures, workplaces have their own norms, power changes behaviour, and individuals adapt differently across settings.
The useful question is not “What does this culture do?” It is “What inference are we drawing from this behaviour, and what evidence supports that inference?”
That is a much harder question. It is also a fairer one.
FAIR PROCESS HAS TO UNDERSTAND THE ANSWER
Fair process does not mean lowering standards. It means assessing accurately.
A process can formally offer a chance to respond and still be badly designed for understanding. A person may be given a meeting but not the material being relied on. They may be given time but not clarity. They may be asked an open question using terms they understand differently from the decision-maker. They may answer orally, then discover that the record has frozen an imprecise phrase into a definitive position.
Good process creates ways to test meaning before turning it into judgment. Putting allegations in writing and explaining loaded terms can establish a shared understanding of what is being asked; a chance to clarify afterwards can prevent a moment of verbal difficulty from being frozen as the person’s position. Depending on the setting, that understanding may also require an interpreter, a support person or a question about what the person understood.
These are not special privileges. They are ways of reducing measurement error.
The key discipline is to separate substance from style: which conclusion depends on the content of the answer, and which depends on how comfortable the answer sounded?
A process that cannot make those distinctions may still look procedurally complete. It may have the letter, the meeting, the notes and the response. But if it mistakes linguistic performance for evidence, its completeness is cosmetic.
THE RECORD CAN CORRECT THE ROOM
High-pressure conversation is fleeting. The record is not.
That makes the written record especially important where language, accent or communication style has become entangled with credibility. A good record can slow the inference down. It can distinguish what the person meant from how the first answer sounded. It can show where a term needed clarification. It can separate an actual contradiction from a later, more precise explanation.
This is not an argument that written English is inherently more truthful. A written response can be edited, translated, supported or professionally drafted. Assistance is not itself a defect; the account must still reflect the person’s position and remain open to testing, with any necessary communication support. The purpose of a record is not to reproduce the speaker’s spontaneous linguistic performance. It is to make the substance of the position testable.
That distinction becomes more important as communication is increasingly mediated by translation tools, writing assistance and AI. A polished sentence may tell us very little about how easily the person would have produced the same sentence aloud. An accented sentence may tell us very little about whether the underlying account is accurate. Presentation and substance are becoming easier to separate technologically at the same time as institutions remain tempted to fuse them psychologically.
The answer is not to distrust polished language or to romanticise imperfect language. It is to return to evidence.
What happened? What supports the account? What remains inconsistent after genuine clarification? Which inference depends on the facts, and which depends on the listener’s reaction to the way those facts were delivered?
Whether a process can ask those questions depends on more than the decision-maker’s goodwill. Its design must leave room for clarification to change the record and for the record to change the conclusion.
A fair process does not need to pretend language is irrelevant. It needs to know what language can and cannot prove.
Fluency can tell us how easily a message arrived. It cannot, by itself, tell us whether the message is true.
Provenance
Originally published on Jan 13, 2026 as “Fair Process When English is Not Your First Language”. Substantially revised for By Ava in 2026.