Choosing Active or Passive Voice by Purpose and Context

Use active voice when the reader needs to know who did what. Use passive when the actor is unknown, irrelevant, or you have a reason to background them. That is the whole rule. An editor once circled every verb in a manuscript and found sixteen of twenty-three sentences with no visible actor. The writer had not chosen a style. The writer had defaulted to a structure that erased agency, and the page read like events that simply occurred.

Voice choice is a structural lever. Active voice puts the agent in the subject slot, where the reader instinctively scans for "who did this." Passive voice shifts the recipient or the action into that slot. The agent moves to a prepositional phrase or vanishes. Both constructions are grammatically sound. Both have distinct jobs. The failure is using one when the other is required by the evidence and the reader's need for source traceability.

Use passive voice when the recipient of the action matters more than the actor, and naming the actor would distract from the claim. This is standard in scientific writing conventions, where a methods section reports that "the solution was heated to 70°C" because the heating device is standard lab equipment and the technician's identity is irrelevant to replication. The register of formal research has trained readers to expect this construction. Switching to active voice in every line would violate the given-new contract by foregrounding information the reader does not need.

Use passive voice when the actor is genuinely unknown. A journalist reporting that "the documents were leaked late Thursday" cannot name a leaker they have not identified. An active construction like "an unknown person leaked the documents" adds no precision and pads the line. The passive here preserves attribution specificity by not inventing a source.

The Reuters Handbook of Journalism permits passive voice when the doer is "unknown or irrelevant." That is the test. If you can name the actor in three words without speculation, the reader probably needs that information. If you cannot, passive voice is not evasion. It is accuracy.

When Naming the Actor Is the Point

Use active voice when naming the actor is the point of the line and when the reader's ability to evaluate the claim depends on knowing who performed the action. In evidence-based writing, this is most of the time. "The committee rejected the proposal" carries more foregrounding weight than "the proposal was rejected" because the committee is the entity with decision-making authority. Omitting it changes what the reader can verify.

Closing the Claim-Defensibility Gap

Active voice in nonfiction is not a style preference. It is a tool for closing the claim-defensibility gap. When a writer reports that "researchers found a 14% increase in recovery rates," the active construction forces the writer to specify which researchers, which study, and whether the reader can access it. If the writer cannot name the researchers because the claim comes from a secondary source that did not name them, the active line exposes that gap. The passive version, "a 14% increase in recovery rates was found," leaves the gap invisible.

The discipline of evidence-based editing treats active voice as a diagnostic. When you revise a passive line to active and the subject slot goes blank, you have found a claim not yet sourced to a traceable actor. That is not a grammar problem. That is an evidentiary problem.

Foregrounding the Rule Over the Rule-Maker

The purpose of passive voice is to control what the reader holds responsible for an action. This is not inherently deceptive. A public health advisory that states "masks are required inside the terminal" uses passive voice to foreground the rule rather than the rule-maker, because the rule-maker is a bureaucratic entity whose name would add friction without adding clarity. The reader needs to know what to do, not which subcommittee voted on it.

When Backgrounding Becomes Evasion

But passive voice purpose becomes evasion when a text systematically backgrounds actors who should be accountable. The nominalization rate of a passage measures how many verbs have been converted into nouns, and this process compounds the passive's agency-hiding effect. A line like "the decision was made after a review of the data" contains two nominalizations, "decision" from "decide" and "review" from "review," and a passive verb "was made." The result is a line in which no human being appears and no human being can be questioned. The presupposition load is high: the line presupposes a decision occurred, a review occurred, and a causal link connects them, without offering any access to who decided or what the data showed.

In stylistics, this construction is called agent deletion. It is not a grammatical error. It is a move, and an editor trained in reading for agency can spot it in seconds. When a paragraph contains more than two agentless passives, stop and ask what the writer is not naming.

Construction	Agent in Subject Position	Traceability	Typical Use
<strong>Active voice</strong>	Yes, the actor is the subject.	High; the reader can verify who did what.	News leads, evidence claims, accountability statements.
<strong>Passive voice (agent retained)</strong>	No, but the agent appears in a prepositional phrase.	Moderate; the agent is named but not foregrounded.	Scientific methods, diplomatic reports, legal documents.
<strong>Passive voice (agent deleted) plus nominalization</strong>	No, the agent is removed entirely.	Low; the reader cannot identify the actor.	Bureaucratic evasion, unsourced journalism, corporate statements.

The Null Hypothesis Discipline

Voice choice in editing is a decision made line by line, not a global setting. The null hypothesis discipline for an editor is to assume every passive line should be active until the context proves otherwise. This is not a ban. It is a workflow rule that prevents passive from becoming the default. When you encounter a passive construction, test it by converting it to active and asking three questions.

Three Tests for Every Passive Line

First, does the active version name an actor who is traceable to a source? If the actor is "a study" or "experts," the line has a pseudo-attribution problem. The reader needs a named researcher, a named institution, or a named study with access. If the active rewrite cannot supply one, the claim is not ready for publication. Second, does the active version violate the given-new contract by putting unfamiliar information in the subject position? If the paragraph has been discussing a policy change, a line that begins "the commissioner announced the policy" might satisfy the contract better than "the policy was announced," even though the policy is the topic. The commissioner is given information if the commissioner was introduced earlier.

Third, does the active version distort the abstraction ladder position the paragraph requires? A methods section reporting standard lab procedures operates at a mid-level of abstraction where the actor is normally irrelevant. Forcing active voice here would pull the paragraph down the ladder into unnecessary specificity. The right choice keeps the reader at the level the argument demands.

The interaction between passive voice and nominalization is the most powerful agency-hiding mechanism in English nonfiction. A verb describes an action performed by an agent. When you nominalize that verb, you turn the action into a thing that exists independently. When you then place that thing in the subject position of a passive line, you create a structure in which things happen to things, and no human being is anywhere in the chain.

Consider a line from an institutional report: "The implementation of the restructuring plan was delayed pending further analysis." The verb "implement" became the noun "implementation." The verb "restructure" became the noun "restructuring." The verb "analyze" became the noun "analysis." The only verb with tense is "was delayed," and it is passive. Who delayed it? Who made the plan? Who is analyzing what? The line answers none of these questions. The nominalization rate is high, the attribution specificity is zero, and the reader has been given a report that reports nothing actionable.

This is not a stylistic failure. It is a structural evasion of responsibility. In a forensic linguistics analysis, a pattern of agentless passives and high nominalization across a document is evidence that the writer is systematically avoiding naming who made decisions, who acted, and who is accountable. The Taking Down Words method flags this pattern not as bad writing but as a feature of the text's argumentative posture. The text is not failing to name actors. It is succeeding at not naming them.

A context-based voice choice follows from the purpose of the line and the reader's need for source traceability. A style preference is a blanket rule like "never use passive voice" or "passive voice is more objective." The first is a professional editing standard. The second is a superstition. The section contract for this page distinguishes the two by asking what the reader can verify after reading the line.

A context-based choice answers this: the reader can verify who acted, what the action was, and what the source is, or the reader can understand why that information is unavailable and why the line is still worth publishing. A style preference answers nothing. It applies the same rule to a news lead naming a corrupt official and a lab report describing a titration, and it fails both when it fails the wrong one.

The Taking Down Words approach to voice choice treats every passive line as a decision the writer made, whether consciously or not, and evaluates that decision against the evidence the writer had. If the writer had the actor's name and did not use it, the passive is a rhetorical move. If the writer did not have the name, the passive is a fact of the reporting. The editor's job is to know which is which.

The claim-defensibility gap is the distance between what a line asserts and what the cited evidence supports. Voice choice widens or narrows this gap directly. An active line that names an actor and cites a specific study closes the gap by giving the reader a path to verification. A passive line that names no actor and cites no study leaves the gap permanently open.

Institutional writing often widens the gap deliberately. A press release stating "mistakes were made" contains no agent, no specific error, and no mechanism for follow-up. The gap is the entire point. The line is not informative. It is defensive. An editor trained in evidence-based editing treats this construction as a red flag. The passive voice is not the problem. The absence of source traceability is the problem, and the passive voice is how the writer achieved it.

When you encounter a line with a wide claim-defensibility gap, do not simply mark it for revision. Identify whether the gap can be closed with the evidence the writer already has. If the writer cannot name the actor, the claim may be unverifiable and should be cut. If the writer can name the actor but chose not to, the revision is a disclosure decision, not a grammar fix.

Attribution specificity is an ordinal scale from no source at all to a named study with access. At the low end, a line like "it is believed that the policy will increase costs" attributes the belief to no one. The passive voice here is agentless, and the attribution is empty. At the high end, a line like "the FDA's 2025 guidance document states that the policy will increase costs by 3 percent" names the actor, the document, and the year. The voice choice is active, and the attribution is specific enough to verify.

Most nonfiction falls between these poles. "Studies show" is a pseudo-attribution that uses active voice with a vague plural agent. "It has been reported that" is a passive construction with no agent at all. Both fail the source traceability test. The editor's standard is not that every line must cite a primary source. It is that every claim that carries a factual assertion must be traceable to a named, accessible source, and the voice choice should not obscure whether that source exists.